d/freebsd-nq
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
f32b457b6e
freebsd-nq/sys/kern/vfs_subr.c

4255 lines
106 KiB
C
Raw Normal View History

/* -> /*- for copyright notices, minor format tweaks as necessary
2005-01-06 23:35:40 +00:00
/*-
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
* Copyright (c) 1989, 1993
* The Regents of the University of California. All rights reserved.
* (c) UNIX System Laboratories, Inc.
* All or some portions of this file are derived from material licensed
* to the University of California by American Telephone and Telegraph
* Co. or Unix System Laboratories, Inc. and are reproduced herein with
* the permission of UNIX System Laboratories, Inc.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 4. Neither the name of the University nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
This is the kernel Lite/2 commit. There are some requisite userland changes, so don't expect to be able to run the kernel as-is (very well) without the appropriate Lite/2 userland changes. The system boots and can mount UFS filesystems. Untested: ext2fs, msdosfs, NFS Known problems: Incorrect Berkeley ID strings in some files. Mount_std mounts will not work until the getfsent library routine is changed. Reviewed by: various people Submitted by: Jeffery Hsu <hsu@freebsd.org>
1997-02-10 02:22:35 +00:00
* @(#)vfs_subr.c 8.31 (Berkeley) 5/26/95
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
*/
/*
* External virtual filesystem routines
*/
Use __FBSDID().
2003-06-11 00:56:59 +00:00
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
Convert DDB to new-style option.
1996-01-04 21:13:23 +00:00
#include "opt_ddb.h"
Introduce support for Mandatory Access Control and extensible kernel access control. Invoke the necessary MAC entry points to maintain labels on vnodes. In particular, initialize the label when the vnode is allocated or reused, and destroy the label when the vnode is going to be released, or reused. Wow, an object where there really is exactly one place where it's allocated, and one other where it's freed. Amazing. Obtained from: TrustedBSD Project Sponsored by: DARPA, NAI Labs
2002-07-31 02:03:46 +00:00
#include "opt_mac.h"
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
#include <sys/param.h>
#include <sys/systm.h>
Separate the struct bio related stuff out of <sys/buf.h> into <sys/bio.h>. <sys/bio.h> is now a prerequisite for <sys/buf.h> but it shall not be made a nested include according to bdes teachings on the subject of nested includes. Diskdrivers and similar stuff below specfs::strategy() should no longer need to include <sys/buf.> unless they need caching of data. Still a few bogus uses of struct buf to track down. Repocopy by: peter
2000-05-05 09:59:14 +00:00
#include <sys/bio.h>
Introduce a mechanism to suspend/resume system processes. Suspend syncer and bufdaemon prior to disk sync during system shutdown.
2000-01-07 08:36:44 +00:00
#include <sys/buf.h>
Remove the use of lbolt from the VFS syncer. It seems we only use `lbolt' inside the VFS syncer and the TTY layer now. Because I'm planning to replace the TTY layer next month, there's no reason to keep `lbolt' if it's only used in a single thread inside the kernel. Because the syncer code wanted to wake up the syncer thread before the timeout, it called sleepq_remove(). Because we now just use a condvar(9) with a timeout value of `hz', we can wake it up using cv_broadcast() without waking up any unrelated threads. Reviewed by: phk
2008-07-30 12:39:18 +00:00
#include <sys/condvar.h>
Introduce a mechanism to suspend/resume system processes. Suspend syncer and bufdaemon prior to disk sync during system shutdown.
2000-01-07 08:36:44 +00:00
#include <sys/conf.h>
Introduce vfs_read_dirent() which can help VOP_READDIR() implementations by handling all the cookie stuff.
2005-09-12 08:46:07 +00:00
#include <sys/dirent.h>
Introduce a new kevent filter. EVFILT_FS that will be used to signal generic filesystem events to userspace. Currently only mount and unmount of filesystems are signalled. Soon to be added, up/down status of NFS. Introduce a sysctl node used to route requests to/from filesystems based on filesystem ids. Introduce a new vfsop, vfs_sysctl(mp, req) that is used as the callback/ entrypoint by the sysctl code to change individual filesystems.
2004-07-04 10:52:54 +00:00
#include <sys/event.h>
Introduce a mechanism to suspend/resume system processes. Suspend syncer and bufdaemon prior to disk sync during system shutdown.
2000-01-07 08:36:44 +00:00
#include <sys/eventhandler.h>
Introduce new extattr_check_cred() function which implements the canonical crential washing for extended attributes. Sponsored by: DARPA & NAI Labs.
2002-09-05 20:38:57 +00:00
#include <sys/extattr.h>
Allow EVFILT_VNODE events to work on every filesystem type, not just UFS by: - Making the pre and post hooks for the VOP functions work even when DEBUG_VFS_LOCKS is not defined. - Moving the KNOTE activations into the corresponding VOP hooks. - Creating a MNTK_NOKNOTE flag for the mnt_kern_flag field of struct mount that permits filesystems to disable the new behavior. - Creating a default VOP_KQFILTER function: vfs_kqfilter() My benchmarks have not revealed any performance degradation. Reviewed by: jeff, bde Approved by: rwatson, jmg (kqueue changes), grehan (mentor)
2005-06-09 20:20:31 +00:00
#include <sys/file.h>
Restored rev.1.31 which was clobbered by rev.1.69 (the big Lite2 merge). This fixes at least hanging in revoke(2) when a somewhat active slave pty is revoked. The hang made the window for the null pointer bug in ufsspec_{read,write} much larger. There are many other bugs in this area (revoke of an active fifo at best leaks memory...).
1998-12-24 12:07:16 +00:00
#include <sys/fcntl.h>
Fix jails and jail-friendly file systems handling: - We need to allow for PRIV_VFS_MOUNT_OWNER inside a jail. - Move security checks to vfs_suser() and deny unmounting and updating for jailed root from different jails, etc. OK'ed by: rwatson
2007-04-13 23:54:22 +00:00
#include <sys/jail.h>
Update for the KDB framework: o Call kdb_enter() instead of Debugger().
2004-07-10 21:47:53 +00:00
#include <sys/kdb.h>
Fixed support for DIAGNOSTIC option. SYSCTL_INT() depends on kernel.h.
1995-11-16 09:45:23 +00:00
#include <sys/kernel.h>
Slight reorganization of kernel thread/process creation. Instead of using SYSINIT_KT() etc (which is a static, compile-time procedure), use a NetBSD-style kthread_create() interface. kproc_start is still available as a SYSINIT() hook. This allowed simplification of chunks of the sysinit code in the process. This kthread_create() is our old kproc_start internals, with the SYSINIT_KT fork hooks grafted in and tweaked to work the same as the NetBSD one. One thing I'd like to do shortly is get rid of nfsiod as a user initiated process. It makes sense for the nfs client code to create them on the fly as needed up to a user settable limit. This means that nfsiod doesn't need to be in /sbin and is always "available". This is a fair bit easier to do outside of the SYSINIT_KT() framework.
1999-07-01 13:21:46 +00:00
#include <sys/kthread.h>
Move the head of byte-level advisory lock list from the filesystem-specific vnode data to the struct vnode. Provide the default implementation for the vop_advlock and vop_advlockasync. Purge the locks on the vnode reclaim by using the lf_purgelocks(). The default implementation is augmented for the nfs and smbfs. In the nfs_advlock, push the Giant inside the nfs_dolock. Before the change, the vop_advlock and vop_advlockasync have taken the unlocked vnode and dereferenced the fs-private inode data, racing with with the vnode reclamation due to forced unmount. Now, the vop_getattr under the shared vnode lock is used to obtain the inode size, and later, in the lf_advlockasync, after locking the vnode interlock, the VI_DOOMED flag is checked to prevent an operation on the doomed vnode. The implementation of the lf_purgelocks() is submitted by dfr. Reported by: kris Tested by: kris, pho Discussed with: jeff, dfr MFC after: 2 weeks
2008-04-16 11:33:32 +00:00
#include <sys/lockf.h>
Include file cleanup; mac.h and malloc.h at one point had ordering relationship requirements, and no longer do. Reminded by: bde
2002-08-01 17:47:56 +00:00
#include <sys/malloc.h>
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
#include <sys/mount.h>
Change NDFREE() from a macro to a function for the time being; the macro version caused intolerable bloat (30k). I'm likely to revisit this with an attempt at a smarter macro. Bloat noticed by: bde
2000-01-08 16:20:06 +00:00
#include <sys/namei.h>
Sweep kernel replacing suser(9) calls with priv(9) calls, assigning specific privilege names to a broad range of privileges. These may require some future tweaking. Sponsored by: nCircle Network Security, Inc. Obtained from: TrustedBSD Project Discussed on: arch@ Reviewed (at least in part) by: mlaier, jmg, pjd, bde, ceri, Alex Lyashkov <umka at sevcity dot net>, Skip Ford <skip dot ford at verizon dot net>, Antoine Brodin <antoine dot brodin at laposte dot net>
2006-11-06 13:42:10 +00:00
#include <sys/priv.h>
Don't attempt to trigger the syncer thread final sync code in the shutdown_pre_sync state if the RB_NOSYNC flag is set. This is the likely cause of hangs after a system panic that are keeping crash dumps from being done. This is a MFC candidate for RELENG_5. MFC after: 3 days
2004-08-20 19:21:47 +00:00
#include <sys/reboot.h>
Switch the sleep/wakeup and condition variable implementations to use the sleep queue interface: - Sleep queues attempt to merge some of the benefits of both sleep queues and condition variables. Having sleep qeueus in a hash table avoids having to allocate a queue head for each wait channel. Thus, struct cv has shrunk down to just a single char * pointer now. However, the hash table does not hold threads directly, but queue heads. This means that once you have located a queue in the hash bucket, you no longer have to walk the rest of the hash chain looking for threads. Instead, you have a list of all the threads sleeping on that wait channel. - Outside of the sleepq code and the sleep/cv code the kernel no longer differentiates between cv's and sleep/wakeup. For example, calls to abortsleep() and cv_abort() are replaced with a call to sleepq_abort(). Thus, the TDF_CVWAITQ flag is removed. Also, calls to unsleep() and cv_waitq_remove() have been replaced with calls to sleepq_remove(). - The sched_sleep() function no longer accepts a priority argument as sleep's no longer inherently bump the priority. Instead, this is soley a propery of msleep() which explicitly calls sched_prio() before blocking. - The TDF_ONSLEEPQ flag has been dropped as it was never used. The associated TDF_SET_ONSLEEPQ and TDF_CLR_ON_SLEEPQ macros have also been dropped and replaced with a single explicit clearing of td_wchan. TD_SET_ONSLEEPQ() would really have only made sense if it had taken the wait channel and message as arguments anyway. Now that that only happens in one place, a macro would be overkill.
2004-02-27 18:52:44 +00:00
#include <sys/sleepqueue.h>
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
#include <sys/stat.h>
Introduce a mechanism to suspend/resume system processes. Suspend syncer and bufdaemon prior to disk sync during system shutdown.
2000-01-07 08:36:44 +00:00
#include <sys/sysctl.h>
This is a forward port of Peter's vlrureclaim() fix, with some minor mods by me to make it more efficient. The original code had serious balancing problems and could also deadlock easily. This code relegates the vnode reclamation to its own kproc and relaxes the vnode reclamation requirements to better maintain kern.maxvnodes. This code still doesn't balance as well as it could, but it does a much better job then the original code. Approved by: re@freebsd.org Obtained from: ps, peter, dillon MFS Assuming: Assuming no problems crop up in Yahoo testing MFC after: 7 days
2001-12-18 20:48:54 +00:00
#include <sys/syslog.h>
Lots of improvements, including restructring the caching and management of vnodes and objects. There are some metadata performance improvements that come along with this. There are also a few prototypes added when the need is noticed. Changes include: 1) Cleaning up vref, vget. 2) Removal of the object cache. 3) Nuke vnode_pager_uncache and friends, because they aren't needed anymore. 4) Correct some missing LK_RETRY's in vn_lock. 5) Correct the page range in the code for msync. Be gentle, and please give me feedback asap.
1997-12-29 00:25:11 +00:00
#include <sys/vmmeter.h>
Introduce a mechanism to suspend/resume system processes. Suspend syncer and bufdaemon prior to disk sync during system shutdown.
2000-01-07 08:36:44 +00:00
#include <sys/vnode.h>
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
Change vprint() to vn_printf() which takes varargs. Add #define for vprint() to call vn_printf().
2005-01-24 13:58:08 +00:00
#include <machine/stdarg.h>
Complete break-out of sys/sys/mac.h into sys/security/mac/mac_framework.h begun with a repo-copy of mac.h to mac_framework.h. sys/mac.h now contains the userspace and user<->kernel API and definitions, with all in-kernel interfaces moved to mac_framework.h, which is now included across most of the kernel instead. This change is the first step in a larger cleanup and sweep of MAC Framework interfaces in the kernel, and will not be MFC'd. Obtained from: TrustedBSD Project Sponsored by: SPARTA
2006-10-22 11:52:19 +00:00
#include <security/mac/mac_framework.h>
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
#include <vm/vm.h>
Untangled the vm.h include file spaghetti.
1995-12-07 12:48:31 +00:00
#include <vm/vm_object.h>
#include <vm/vm_extern.h>
Some performance improvements, and code cleanups (including changing our expensive OFF_TO_IDX to btoc whenever possible.)
1997-12-19 09:03:37 +00:00
#include <vm/pmap.h>
#include <vm/vm_map.h>
This is a rather large commit that encompasses the new swapper, changes to the VM system to support the new swapper, VM bug fixes, several VM optimizations, and some additional revamping of the VM code. The specific bug fixes will be documented with additional forced commits. This commit is somewhat rough in regards to code cleanup issues. Reviewed by: "John S. Dyson" <root@dyson.iquest.net>, "David Greenman" <dg@root.com>
1999-01-21 08:29:12 +00:00
#include <vm/vm_page.h>
Make the maximum number of vnodes a function of both the physical memory size and the kernel's heap size, specifically, vm_kmem_size. This function allows a maximum of 40% of the vm_kmem_size to be used for vnodes and vm objects. This is a conservative bound based upon recent problem reports. (In other words, a slight increase in this percentage may be safe.) Finally, machines with less than ~3GB of RAM should be unaffected by this change, i.e., the maximum number of vnodes should remain the same. If necessary, machines with 3GB or more of RAM can increase the maximum number of vnodes by increasing vm_kmem_size. Desired by: scottl Tested by: jake Approved by: re (rwatson,scottl)
2003-05-23 19:54:02 +00:00
#include <vm/vm_kern.h>
Remove references to vm_zone.h and switch over to the new uma API. Also, remove maxsockets. If you look carefully you'll notice that the old zone allocator never honored this anyway.
2002-03-20 04:09:59 +00:00
#include <vm/uma.h>
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
Add 'show vnode <addr>' DDB command.
2006-09-04 22:15:44 +00:00
#ifdef DDB
#include <ddb/ddb.h>
#endif
sync_vnode() has some messy code about locking in order to deal with mount fs needing Giant to be held when processing bufobjs. Use a different subqueue for pending workitems on filesystems requiring Giant. This simplifies the code notably and also reduces the number of Giant acquisitions (and the whole processing cost). Suggested by: jeff Reviewed by: kib Tested by: pho
2008-05-04 13:54:55 +00:00
#define WI_MPSAFEQ 0
#define WI_GIANTQ 1
Normalize a significant number of kernel malloc type names: - Prefer '_' to ' ', as it results in more easily parsed results in memory monitoring tools such as vmstat. - Remove punctuation that is incompatible with using memory type names as file names, such as '/' characters. - Disambiguate some collisions by adding subsystem prefixes to some memory types. - Generally prefer lower case to upper case. - If the same type is defined in multiple architecture directories, attempt to use the same name in additional cases. Not all instances were caught in this change, so more work is required to finish this conversion. Similar changes are required for UMA zone names.
2005-10-31 15:41:29 +00:00
static MALLOC_DEFINE(M_NETADDR, "subr_export_host", "Export host address structure");
Distribute and statizice a lot of the malloc M_* types. Substantial input from: bde
1997-10-11 18:31:40 +00:00
When we traverse the vnodes on a mountpoint we need to look out for our cached 'next vnode' being removed from this mountpoint. If we find that it was recycled, we restart our traversal from the start of the list. Code to do that is in all local disk filesystems (and a few other places) and looks roughly like this: MNT_ILOCK(mp); loop: for (vp = TAILQ_FIRST(&mp...); (vp = nvp) != NULL; nvp = TAILQ_NEXT(vp,...)) { if (vp->v_mount != mp) goto loop; MNT_IUNLOCK(mp); ... MNT_ILOCK(mp); } MNT_IUNLOCK(mp); The code which takes vnodes off a mountpoint looks like this: MNT_ILOCK(vp->v_mount); ... TAILQ_REMOVE(&vp->v_mount->mnt_nvnodelist, vp, v_nmntvnodes); ... MNT_IUNLOCK(vp->v_mount); ... vp->v_mount = something; (Take a moment and try to spot the locking error before you read on.) On a SMP system, one CPU could have removed nvp from our mountlist but not yet gotten to assign a new value to vp->v_mount while another CPU simultaneously get to the top of the traversal loop where it finds that (vp->v_mount != mp) is not true despite the fact that the vnode has indeed been removed from our mountpoint. Fix: Introduce the macro MNT_VNODE_FOREACH() to traverse the list of vnodes on a mountpoint while taking into account that vnodes may be removed from the list as we go. This saves approx 65 lines of duplicated code. Split the insmntque() which potentially moves a vnode from one mount point to another into delmntque() and insmntque() which does just what the names say. Fix delmntque() to set vp->v_mount to NULL while holding the mountpoint lock.
2004-07-04 08:52:35 +00:00
static void delmntque(struct vnode *vp);
Try to unbreak the vnode locking around vop_reclaim() (based mostly on patch from kan@). Pull bufobj_invalbuf() out of vinvalbuf() and make g_vfs call it on close. This is not yet a generally safe function, but for this very specific use it is safe. This solves the problem with buffers not being flushed by unmount or after failed mount attempts.
2005-02-19 11:44:57 +00:00
static int flushbuflist(struct bufv *bufv, int flags, struct bufobj *bo,
Give flushbuflist() a struct bufv as first argument and avoid home-rolling TAILQ_FOREACH_SAFE(). Loose the error pointer argument and return any errors the normal way. Return EAGAIN for the case where more work needs to be done.
2005-01-11 10:01:54 +00:00
int slpflag, int slptimeo);
When shutting down the syncer kernel thread, first tell it to run faster and iterate to over its work list a few times in an attempt to empty the work list before the syncer terminates. This leaves fewer dirty blocks to be written at the "syncing disks" stage and keeps the the "giving up on N buffers" problem from being triggered by the presence of a large soft updates work list at system shutdown time. The downside is that the syncer takes noticeably longer to terminate. Tested by: "Arjan van Leeuwen" <avleeuwen AT piwebs DOT com> Approved by: mckusick
2004-07-01 23:59:19 +00:00
static void syncer_shutdown(void *arg, int howto);
- Rename vcanrecycle() to vtryrecycle() to reflect its new role. - In vtryrecycle() try to vgonel the vnode if all of the previous checks passed. We won't vgonel if someone has either acquired a hold or usecount or started the vgone process elsewhere. This is because we may have been removed from the free list while we were inspecting the vnode for recycling. - The VI_TRYLOCK stops two threads from entering getnewvnode() and recycling the same vnode. To further reduce the likelyhood of this event, requeue the vnode on the tail of the list prior to calling vtryrecycle(). We can not actually remove the vnode from the list until we know that it's going to be recycled because other interlock holders may see the VI_FREE flag and try to remove it from the free list. - Kill a bogus XXX comment. If XLOCK is set we shouldn't wait for it regardless of MNT_WAIT because the vnode does not actually belong to this filesystem.
2003-10-05 05:35:41 +00:00
static int vtryrecycle(struct vnode *vp);
Make various vnode related functions static
2005-02-10 12:28:58 +00:00
static void vbusy(struct vnode *vp);
- Remove vx_lock, vx_unlock, vx_wait, etc. - Add a vn_start_write/vn_finished_write around vlrureclaim so we don't do writing ops without suspending. This could suspend the vlruproc which should not be a problem under normal circumstances. - Manually implement VMIGHTFREE in vlrureclaim as this was the only instance where it was used. - Acquire a lock before calling vgone() as it now requires it. - Move the acquisition of the vnode interlock from vtryrecycle() to getnewvnode() so that if it fails we don't drop and reacquire the vnode_free_list_mtx. - Check for a usecount or holdcount at the end of vtryrecycle() in case someone grabbed a ref while we were recycling. Abort the recycle, and on the final ref drop this vnode will be placed on the head of the free list. - Move the redundant VOP_INACTIVE protection code into the local vinactive() routine to avoid code bloat. - Keep the vnode lock held across calls to vgone() in several places. - vgonel() no longer uses XLOCK, instead callers must hold an exclusive vnode lock. The VI_DOOMED flag is set to allow other threads to detect a vnode which is no longer valid. This flag is set until the last reference is gone, and there are no chances for a new ref. vgonel() holds this lock across the entire function, which greatly simplifies logic. _ Only vfree() in one place in vgone() not three. - Adjust vget() to check the VI_DOOMED flag prior to waiting on the lock in the LK_NOWAIT case. In other cases, check after we have slept and acquired an exlusive lock. This will simulate the old vx_wait() behavior. Sponsored by: Isilon Systems, Inc.
2005-03-13 11:54:28 +00:00
static void vinactive(struct vnode *, struct thread *);
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
static void v_incr_usecount(struct vnode *);
static void v_decr_usecount(struct vnode *);
static void v_decr_useonly(struct vnode *);
- Solve a race where we could lose a call to VOP_INACTIVE. If vget() waiting on a lock held the last usecount ref on a vnode and the lock failed we would not call INACTIVE. Solve this by only holding a holdcnt to prevent the vnode from disappearing while we wait on vn_lock. Other callers may now VOP_INACTIVE while we are waiting on the lock, however this race is acceptable, while losing INACTIVE is not. Discussed with: kan, pjd Tested by: kkenn Sponsored by: Isilon Systems, Inc. MFC After: 1 week
2006-02-01 00:30:05 +00:00
static void v_upgrade_usecount(struct vnode *);
- Now that there are no external users of vfree() make it static. - Move VSHOULDBUSY, VSHOULDFREE, and VTRYRECYCLE into vfs_subr.c so no one else attempts to grow a dependency on them. - Now that objects with pages hold the vnode we don't have to do unlocked checks for the page count in the vm object in VSHOULDFREE. These three macros could simply check for holdcnt state transitions to determine whether the vnode is on the free list already, but the extra safety the flag affords us is probably worth the minimal cost. - The leafonly sysctl and code have been dead for several years now, remove the sysctl and the code that employed it from vtryrecycle(). - vtryrecycle() also no longer has to check the object's page count as the object holds the vnode until it reaches 0. Sponsored by: Isilon Systems, Inc.
2005-03-15 14:38:16 +00:00
static void vfree(struct vnode *);
- Don't recycle vnodes anymore. Free them once they are dead. getnewvnode now always allocates a new vnode. - Define a new function, vnlru_free, which frees vnodes from the free list. It takes as a parameter the number of vnodes to free, which is wantfreevnodes - freevnodes when called from vnlru_proc or 1 when called from getnewvnode(). For now, getnewvnode() still tries to reclaim a free vnode before creating a new one when we are near the limit. - Define a function, vdestroy, which handles the actual release of memory and teardown of locks, etc. This could become a uma_dtor() routine. - Get rid of minvnodes. Now wantfreevnodes is 1/4th the max vnodes. This keeps more unreferenced vnodes around so that files which have only been stat'd are less likely to be kicked out of the system before we have a chance to read them, etc. These vnodes may still be freed via the normal vnlru_proc() routines which may some day become a real lru.
2005-03-25 05:34:39 +00:00
static void vnlru_free(int);
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
static void vgonel(struct vnode *);
Fix the recent panics/LORs/hangs created by my kqueue commit by: - Introducing the possibility of using locks different than mutexes for the knlist locking. In order to do this, we add three arguments to knlist_init() to specify the functions to use to lock, unlock and check if the lock is owned. If these arguments are NULL, we assume mtx_lock, mtx_unlock and mtx_owned, respectively. - Using the vnode lock for the knlist locking, when doing kqueue operations on a vnode. This way, we don't have to lock the vnode while holding a mutex, in filt_vfsread. Reviewed by: jmg Approved by: re (scottl), scottl (mentor override) Pointyhat to: ssouhlal Will be happy: everyone
2005-07-01 16:28:32 +00:00
static void vfs_knllock(void *arg);
static void vfs_knlunlock(void *arg);
static int vfs_knllocked(void *arg);
Protect check for v_pollinfo == NULL and assignment of the newly allocated vpollinfo with vnode interlock. Fully initialize vpollinfo before putting pointer to it into vp->v_pollinfo. Discussed with: dwhite Tested by: pho MFC after: 1 week
2008-10-28 12:08:36 +00:00
static void destroy_vpollinfo(struct vpollinfo *vi);
Style fixes: * Add lots of comments * Convert a couple of assertions to KASSERT() * Minimal whitespace & misapplied {} fixes * Convert #if 0 to #if COMPILING_LINT for code we presently do not support, but want to keep available. Reviewed by: adrian, markm
2000-09-22 12:22:36 +00:00
- Add the tunable and sysctl for the mpsafevfs. It currently defaults to off. - Protect access to mnt_kern_flag with the mointpoint mutex. - Remove some KASSERTs which are not legal checks without the appropriate locks held. - Use VCANRECYCLE() rather than rolling several slightly different checks together. - Return from vtryrecycle() with a recycled vnode rather than a locked vnode. This simplifies some locking. - Remove several GIANT_REQUIRED lines. - Add a few KASSERTs to help with INACT debugging. Sponsored By: Isilon Systems, Inc.
2005-01-24 10:41:01 +00:00
/*
* Enable Giant pushdown based on whether or not the vm is mpsafe in this
* build. Without mpsafevm the buffer cache can not run Giant free.
*/
- Enable SMP VFS by default on current. More users are needed to turn up any remaining bugs. Anyone inconvenienced by this can still disable it in the loader. Sponsored by: Isilon Systems, Inc.
2005-02-23 10:05:43 +00:00
int mpsafe_vfs = 1;
- Add the tunable and sysctl for the mpsafevfs. It currently defaults to off. - Protect access to mnt_kern_flag with the mointpoint mutex. - Remove some KASSERTs which are not legal checks without the appropriate locks held. - Use VCANRECYCLE() rather than rolling several slightly different checks together. - Return from vtryrecycle() with a recycled vnode rather than a locked vnode. This simplifies some locking. - Remove several GIANT_REQUIRED lines. - Add a few KASSERTs to help with INACT debugging. Sponsored By: Isilon Systems, Inc.
2005-01-24 10:41:01 +00:00
TUNABLE_INT("debug.mpsafevfs", &mpsafe_vfs);
SYSCTL_INT(_debug, OID_AUTO, mpsafevfs, CTLFLAG_RD, &mpsafe_vfs, 0,
"MPSAFE VFS");
Style fixes: * Add lots of comments * Convert a couple of assertions to KASSERT() * Minimal whitespace & misapplied {} fixes * Convert #if 0 to #if COMPILING_LINT for code we presently do not support, but want to keep available. Reviewed by: adrian, markm
2000-09-22 12:22:36 +00:00
/*
* Number of vnodes in existence. Increased whenever getnewvnode()
Correct the comment: numvnodes is decreased on vdestroying the vnode. OKed by: tegge Approved by: pjd (mentor) MFC after: 1 week
2006-10-02 07:25:58 +00:00
* allocates a new vnode, decreased on vdestroy() called on VI_DOOMed
* vnode.
Style fixes: * Add lots of comments * Convert a couple of assertions to KASSERT() * Minimal whitespace & misapplied {} fixes * Convert #if 0 to #if COMPILING_LINT for code we presently do not support, but want to keep available. Reviewed by: adrian, markm
2000-09-22 12:22:36 +00:00
*/
Staticized.
1997-11-22 08:35:46 +00:00
static unsigned long numvnodes;
Add vlruvp() routine - implements LRU operation for vnode recycling. We calculate a trigger point that both guarentees we will find a sufficient number of vnodes to recycle and prevents us from recycling vnodes with lots of resident pages. This particular section of code is designed to recycle vnodes, not do unnecessary frees of cached VM pages.
2002-01-10 18:31:53 +00:00
Move some sysctls from the debug tree to the vfs tree.
2002-06-06 15:50:22 +00:00
SYSCTL_LONG(_vfs, OID_AUTO, numvnodes, CTLFLAG_RD, &numvnodes, 0, "");
Completed function declarations and/or added prototypes.
1995-12-02 18:58:56 +00:00
Style fixes: * Add lots of comments * Convert a couple of assertions to KASSERT() * Minimal whitespace & misapplied {} fixes * Convert #if 0 to #if COMPILING_LINT for code we presently do not support, but want to keep available. Reviewed by: adrian, markm
2000-09-22 12:22:36 +00:00
/*
* Conversion tables for conversion from vnode types to inode formats
* and back.
*/
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
enum vtype iftovt_tab[16] = {
VNON, VFIFO, VCHR, VNON, VDIR, VNON, VBLK, VNON,
VREG, VNON, VLNK, VNON, VSOCK, VNON, VNON, VBAD,
};
Add marker vnodes to ensure that all vnodes associated with the mount point are iterated over when using MNT_VNODE_FOREACH. Reviewed by: truckman
2006-01-09 20:42:19 +00:00
int vttoif_tab[10] = {
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
0, S_IFREG, S_IFDIR, S_IFBLK, S_IFCHR, S_IFLNK,
Add marker vnodes to ensure that all vnodes associated with the mount point are iterated over when using MNT_VNODE_FOREACH. Reviewed by: truckman
2006-01-09 20:42:19 +00:00
S_IFSOCK, S_IFIFO, S_IFMT, S_IFMT
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
};
Style fixes based on comments by bde
2000-10-05 18:22:46 +00:00
/*
* List of vnodes that are ready for recycling.
*/
Style fixes: * Add lots of comments * Convert a couple of assertions to KASSERT() * Minimal whitespace & misapplied {} fixes * Convert #if 0 to #if COMPILING_LINT for code we presently do not support, but want to keep available. Reviewed by: adrian, markm
2000-09-22 12:22:36 +00:00
static TAILQ_HEAD(freelst, vnode) vnode_free_list;
Fix some vnode management problems, and better mgmt of vnode free list. Fix the UIO optimization code. Fix an assumption in vm_map_insert regarding allocation of swap pagers. Fix an spl problem in the collapse handling in vm_object_deallocate. When pages are freed from vnode objects, and the criteria for putting the associated vnode onto the free list is reached, either put the vnode onto the list, or put it onto an interrupt safe version of the list, for further transfer onto the actual free list. Some minor syntax changes changing pre-decs, pre-incs to post versions. Remove a bogus timeout (that I added for debugging) from vn_lock. PHK will likely still have problems with the vnode list management, and so do I, but it is better than it was.
1998-01-12 01:46:33 +00:00
Style fixes: * Add lots of comments * Convert a couple of assertions to KASSERT() * Minimal whitespace & misapplied {} fixes * Convert #if 0 to #if COMPILING_LINT for code we presently do not support, but want to keep available. Reviewed by: adrian, markm
2000-09-22 12:22:36 +00:00
/*
- Don't recycle vnodes anymore. Free them once they are dead. getnewvnode now always allocates a new vnode. - Define a new function, vnlru_free, which frees vnodes from the free list. It takes as a parameter the number of vnodes to free, which is wantfreevnodes - freevnodes when called from vnlru_proc or 1 when called from getnewvnode(). For now, getnewvnode() still tries to reclaim a free vnode before creating a new one when we are near the limit. - Define a function, vdestroy, which handles the actual release of memory and teardown of locks, etc. This could become a uma_dtor() routine. - Get rid of minvnodes. Now wantfreevnodes is 1/4th the max vnodes. This keeps more unreferenced vnodes around so that files which have only been stat'd are less likely to be kicked out of the system before we have a chance to read them, etc. These vnodes may still be freed via the normal vnlru_proc() routines which may some day become a real lru.
2005-03-25 05:34:39 +00:00
* Free vnode target. Free vnodes may simply be files which have been stat'd
* but not read. This is somewhat common, and a small cache of such files
* should be kept to avoid recreation costs.
Style fixes: * Add lots of comments * Convert a couple of assertions to KASSERT() * Minimal whitespace & misapplied {} fixes * Convert #if 0 to #if COMPILING_LINT for code we presently do not support, but want to keep available. Reviewed by: adrian, markm
2000-09-22 12:22:36 +00:00
*/
- Don't recycle vnodes anymore. Free them once they are dead. getnewvnode now always allocates a new vnode. - Define a new function, vnlru_free, which frees vnodes from the free list. It takes as a parameter the number of vnodes to free, which is wantfreevnodes - freevnodes when called from vnlru_proc or 1 when called from getnewvnode(). For now, getnewvnode() still tries to reclaim a free vnode before creating a new one when we are near the limit. - Define a function, vdestroy, which handles the actual release of memory and teardown of locks, etc. This could become a uma_dtor() routine. - Get rid of minvnodes. Now wantfreevnodes is 1/4th the max vnodes. This keeps more unreferenced vnodes around so that files which have only been stat'd are less likely to be kicked out of the system before we have a chance to read them, etc. These vnodes may still be freed via the normal vnlru_proc() routines which may some day become a real lru.
2005-03-25 05:34:39 +00:00
static u_long wantfreevnodes;
Move some sysctls from the debug tree to the vfs tree.
2002-06-06 15:50:22 +00:00
SYSCTL_LONG(_vfs, OID_AUTO, wantfreevnodes, CTLFLAG_RW, &wantfreevnodes, 0, "");
Style fixes: * Add lots of comments * Convert a couple of assertions to KASSERT() * Minimal whitespace & misapplied {} fixes * Convert #if 0 to #if COMPILING_LINT for code we presently do not support, but want to keep available. Reviewed by: adrian, markm
2000-09-22 12:22:36 +00:00
/* Number of vnodes in the free list. */
Do not initialize static/global variables to 0. Use bss instead of taking up space in the data section.
2001-12-19 01:35:18 +00:00
static u_long freevnodes;
Move some sysctls from the debug tree to the vfs tree.
2002-06-06 15:50:22 +00:00
SYSCTL_LONG(_vfs, OID_AUTO, freevnodes, CTLFLAG_RD, &freevnodes, 0, "");
After extensive testing it has been determined that adding complexity to avoid removing higher level directory vnodes from the namecache has no perceivable effect and will be removed. This is especially true when vmiodirenable is turned on, which it is by default now. ( vmiodirenable makes a huge difference in directory caching ). The vfs.vmiodirenable and vfs.nameileafonly sysctls have been left in to allow further testing, but I expect to rip out vfs.nameileafonly soon too. I have also determined through testing that the real problem with numvnodes getting too large is due to the VM Page cache preventing the vnode from being reclaimed. The directory stuff made only a tiny dent relative to Poul's original code, enough so that some tests succeeded. But tests with several million small files show that the bigger problem is the VM Page cache. This will have to be addressed by a future commit. MFC after: 3 days
2001-10-01 04:33:35 +00:00
Style fixes: * Add lots of comments * Convert a couple of assertions to KASSERT() * Minimal whitespace & misapplied {} fixes * Convert #if 0 to #if COMPILING_LINT for code we presently do not support, but want to keep available. Reviewed by: adrian, markm
2000-09-22 12:22:36 +00:00
/*
Style fixes based on comments by bde
2000-10-05 18:22:46 +00:00
* Various variables used for debugging the new implementation of
* reassignbuf().
* XXX these are probably of (very) limited utility now.
Style fixes: * Add lots of comments * Convert a couple of assertions to KASSERT() * Minimal whitespace & misapplied {} fixes * Convert #if 0 to #if COMPILING_LINT for code we presently do not support, but want to keep available. Reviewed by: adrian, markm
2000-09-22 12:22:36 +00:00
*/
The buffer queue mechanism has been reformulated. Instead of having QUEUE_AGE, QUEUE_LRU, and QUEUE_EMPTY we instead have QUEUE_CLEAN, QUEUE_DIRTY, QUEUE_EMPTY, and QUEUE_EMPTYKVA. With this patch clean and dirty buffers have been separated. Empty buffers with KVM assignments have been separated from truely empty buffers. getnewbuf() has been rewritten and now operates in a 100% optimal fashion. That is, it is able to find precisely the right kind of buffer it needs to allocate a new buffer, defragment KVM, or to free-up an existing buffer when the buffer cache is full (which is a steady-state situation for the buffer cache). Buffer flushing has been reorganized. Previously buffers were flushed in the context of whatever process hit the conditions forcing buffer flushing to occur. This resulted in processes blocking on conditions unrelated to what they were doing. This also resulted in inappropriate VFS stacking chains due to multiple processes getting stuck trying to flush dirty buffers or due to a single process getting into a situation where it might attempt to flush buffers recursively - a situation that was only partially fixed in prior commits. We have added a new daemon called the buf_daemon which is responsible for flushing dirty buffers when the number of dirty buffers exceeds the vfs.hidirtybuffers limit. This daemon attempts to dynamically adjust the rate at which dirty buffers are flushed such that getnewbuf() calls (almost) never block. The number of nbufs and amount of buffer space is now scaled past the 8MB limit that was previously imposed for systems with over 64MB of memory, and the vfs.{lo,hi}dirtybuffers limits have been relaxed somewhat. The number of physical buffers has been increased with the intention that we will manage physical I/O differently in the future. reassignbuf previously attempted to keep the dirtyblkhd list sorted which could result in non-deterministic operation under certain conditions, such as when a large number of dirty buffers are being managed. This algorithm has been changed. reassignbuf now keeps buffers locally sorted if it can do so cheaply, and otherwise gives up and adds buffers to the head of the dirtyblkhd list. The new algorithm is deterministic but not perfect. The new algorithm greatly reduces problems that previously occured when write_behind was turned off in the system. The P_FLSINPROG proc->p_flag bit has been replaced by the more descriptive P_BUFEXHAUST bit. This bit allows processes working with filesystem buffers to use available emergency reserves. Normal processes do not set this bit and are not allowed to dig into emergency reserves. The purpose of this bit is to avoid low-memory deadlocks. A small race condition was fixed in getpbuf() in vm/vm_pager.c. Submitted by: Matthew Dillon <dillon@apollo.backplane.com> Reviewed by: Kirk McKusick <mckusick@mckusick.com>
1999-07-04 00:25:38 +00:00
static int reassignbufcalls;
SYSCTL_INT(_vfs, OID_AUTO, reassignbufcalls, CTLFLAG_RW, &reassignbufcalls, 0, "");
Style fixes: * Add lots of comments * Convert a couple of assertions to KASSERT() * Minimal whitespace & misapplied {} fixes * Convert #if 0 to #if COMPILING_LINT for code we presently do not support, but want to keep available. Reviewed by: adrian, markm
2000-09-22 12:22:36 +00:00
/*
* Cache for the mount type id assigned to NFS. This is used for
* special checks in nfs/nfs_nqlease.c and vm/vnode_pager.c.
*/
Instantiate `nfs_mount_type' in a standard file so that it is present when nfs is an LKM. Declare it in a header file. Don't forget to use it in non-Lite2 code. Initialize it to -1 instead of to 0, since 0 will soon be the mount type number for the first vfs loaded. NetBSD uses strcmp() to avoid this ugly global.
1998-09-05 15:17:34 +00:00
int nfs_mount_type = -1;
Style fixes: * Add lots of comments * Convert a couple of assertions to KASSERT() * Minimal whitespace & misapplied {} fixes * Convert #if 0 to #if COMPILING_LINT for code we presently do not support, but want to keep available. Reviewed by: adrian, markm
2000-09-22 12:22:36 +00:00
/* To keep more than one thread at a time from running vfs_getnewfsid */
Convert all simplelocks to mutexes and remove the simplelock implementations.
2001-01-24 12:35:55 +00:00
static struct mtx mntid_mtx;
Style fixes based on comments by bde
2000-10-05 18:22:46 +00:00
- Extend the vnode_free_list_mtx to cover numvnodes and freevnodes. This was done only some of the time before, and now it is uniformly applied.
2002-08-13 05:29:48 +00:00
/*
* Lock for any access to the following:
* vnode_free_list
* numvnodes
* freevnodes
*/
Convert all simplelocks to mutexes and remove the simplelock implementations.
2001-01-24 12:35:55 +00:00
static struct mtx vnode_free_list_mtx;
Style fixes: * Add lots of comments * Convert a couple of assertions to KASSERT() * Minimal whitespace & misapplied {} fixes * Convert #if 0 to #if COMPILING_LINT for code we presently do not support, but want to keep available. Reviewed by: adrian, markm
2000-09-22 12:22:36 +00:00
/* Publicly exported FS */
struct nfs_public nfs_pub;
Style fixes based on comments by bde
2000-10-05 18:22:46 +00:00
Style fixes: * Add lots of comments * Convert a couple of assertions to KASSERT() * Minimal whitespace & misapplied {} fixes * Convert #if 0 to #if COMPILING_LINT for code we presently do not support, but want to keep available. Reviewed by: adrian, markm
2000-09-22 12:22:36 +00:00
/* Zone for allocation of new vnodes - used exclusively by getnewvnode() */
Remove references to vm_zone.h and switch over to the new uma API. Also, remove maxsockets. If you look carefully you'll notice that the old zone allocator never honored this anyway.
2002-03-20 04:09:59 +00:00
static uma_zone_t vnode_zone;
static uma_zone_t vnodepoll_zone;
Style fixes based on comments by bde
2000-10-05 18:22:46 +00:00
Style fixes: * Add lots of comments * Convert a couple of assertions to KASSERT() * Minimal whitespace & misapplied {} fixes * Convert #if 0 to #if COMPILING_LINT for code we presently do not support, but want to keep available. Reviewed by: adrian, markm
2000-09-22 12:22:36 +00:00
/* Set to 1 to print out reclaim of active vnodes */
Do not initialize static/global variables to 0. Use bss instead of taking up space in the data section.
2001-12-19 01:35:18 +00:00
int prtactive;
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
/*
* The workitem queue.
Gratuitous whitespace cleanup.
2002-06-06 15:46:38 +00:00
*
Style fixes: * Add lots of comments * Convert a couple of assertions to KASSERT() * Minimal whitespace & misapplied {} fixes * Convert #if 0 to #if COMPILING_LINT for code we presently do not support, but want to keep available. Reviewed by: adrian, markm
2000-09-22 12:22:36 +00:00
* It is useful to delay writes of file data and filesystem metadata
* for tens of seconds so that quickly created and deleted files need
* not waste disk bandwidth being created and removed. To realize this,
* we append vnodes to a "workitem" queue. When running with a soft
* updates implementation, most pending metadata dependencies should
* not wait for more than a few seconds. Thus, mounted on block devices
* are delayed only about a half the time that file data is delayed.
* Similarly, directory updates are more critical, so are only delayed
* about a third the time that file data is delayed. Thus, there are
* SYNCER_MAXDELAY queues that are processed round-robin at a rate of
* one each second (driven off the filesystem syncer process). The
* syncer_delayno variable indicates the next queue that is to be processed.
* Items that need to be processed soon are placed in this queue:
*
* syncer_workitem_pending[syncer_delayno]
*
* A delay of fifteen seconds is done by placing the request fifteen
* entries later in the queue:
*
* syncer_workitem_pending[(syncer_delayno + 15) & syncer_mask]
*
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
*/
Do not initialize static/global variables to 0. Use bss instead of taking up space in the data section.
2001-12-19 01:35:18 +00:00
static int syncer_delayno;
Gratuitous whitespace cleanup.
2002-06-06 15:46:38 +00:00
static long syncer_mask;
Move the syncer linkage from vnode to bufobj. This is not quite a perfect separation: the syncer still think it knows that everything is a vnode.
2004-10-27 08:05:02 +00:00
LIST_HEAD(synclist, bufobj);
sync_vnode() has some messy code about locking in order to deal with mount fs needing Giant to be held when processing bufobjs. Use a different subqueue for pending workitems on filesystems requiring Giant. This simplifies the code notably and also reduces the number of Giant acquisitions (and the whole processing cost). Suggested by: jeff Reviewed by: kib Tested by: pho
2008-05-04 13:54:55 +00:00
static struct synclist *syncer_workitem_pending[2];
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
/*
* The sync_mtx protects:
Move the syncer linkage from vnode to bufobj. This is not quite a perfect separation: the syncer still think it knows that everything is a vnode.
2004-10-27 08:05:02 +00:00
* bo->bo_synclist
When shutting down the syncer kernel thread, first tell it to run faster and iterate to over its work list a few times in an attempt to empty the work list before the syncer terminates. This leaves fewer dirty blocks to be written at the "syncing disks" stage and keeps the the "giving up on N buffers" problem from being triggered by the presence of a large soft updates work list at system shutdown time. The downside is that the syncer takes noticeably longer to terminate. Tested by: "Arjan van Leeuwen" <avleeuwen AT piwebs DOT com> Approved by: mckusick
2004-07-01 23:59:19 +00:00
* sync_vnode_count
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
* syncer_delayno
Rework syncer termination code: Speed up the syncer when shutting down by sleeping for a shorter period of time instead of cranking up rushjob and using the normal one second sleep. Skip empty worklist slots when shutting down to avoid lengthy intervals of inactivity. Give I/O more time to complete between steps by not speeding the syncer quite as much. Terminate the syncer after one full pass through the worklist plus one second with the worklist containing nothing but syncer vnodes. Print an indication of shutdown progress to the console. Add a sysctl, vfs.worklist_len, to allow the size of the syncer worklist to be monitored.
2004-07-05 01:07:33 +00:00
* syncer_state
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
* syncer_workitem_pending
When shutting down the syncer kernel thread, first tell it to run faster and iterate to over its work list a few times in an attempt to empty the work list before the syncer terminates. This leaves fewer dirty blocks to be written at the "syncing disks" stage and keeps the the "giving up on N buffers" problem from being triggered by the presence of a large soft updates work list at system shutdown time. The downside is that the syncer takes noticeably longer to terminate. Tested by: "Arjan van Leeuwen" <avleeuwen AT piwebs DOT com> Approved by: mckusick
2004-07-01 23:59:19 +00:00
* syncer_worklist_len
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
* rushjob
*/
static struct mtx sync_mtx;
Remove the use of lbolt from the VFS syncer. It seems we only use `lbolt' inside the VFS syncer and the TTY layer now. Because I'm planning to replace the TTY layer next month, there's no reason to keep `lbolt' if it's only used in a single thread inside the kernel. Because the syncer code wanted to wake up the syncer thread before the timeout, it called sleepq_remove(). Because we now just use a condvar(9) with a timeout value of `hz', we can wake it up using cv_broadcast() without waking up any unrelated threads. Reviewed by: phk
2008-07-30 12:39:18 +00:00
static struct cv sync_wakeup;
Style fixes: * Add lots of comments * Convert a couple of assertions to KASSERT() * Minimal whitespace & misapplied {} fixes * Convert #if 0 to #if COMPILING_LINT for code we presently do not support, but want to keep available. Reviewed by: adrian, markm
2000-09-22 12:22:36 +00:00
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
#define SYNCER_MAXDELAY 32
Staticize.
1998-12-21 23:38:33 +00:00
static int syncer_maxdelay = SYNCER_MAXDELAY; /* maximum delay time */
syncdelay, filedelay, dirdelay, metadelay are ints, not time_t's, and can also be made static.
2001-10-27 19:58:56 +00:00
static int syncdelay = 30; /* max time to delay syncing data */
static int filedelay = 30; /* time to delay syncing files */
Get rid of the global variable rushjob and replace it with a function in kern/vfs_subr.c named speedup_syncer() which handles the speedup request. Change the various clients of rushjob to use the new function.
1999-06-15 23:37:29 +00:00
SYSCTL_INT(_kern, OID_AUTO, filedelay, CTLFLAG_RW, &filedelay, 0, "");
syncdelay, filedelay, dirdelay, metadelay are ints, not time_t's, and can also be made static.
2001-10-27 19:58:56 +00:00
static int dirdelay = 29; /* time to delay syncing directories */
Get rid of the global variable rushjob and replace it with a function in kern/vfs_subr.c named speedup_syncer() which handles the speedup request. Change the various clients of rushjob to use the new function.
1999-06-15 23:37:29 +00:00
SYSCTL_INT(_kern, OID_AUTO, dirdelay, CTLFLAG_RW, &dirdelay, 0, "");
syncdelay, filedelay, dirdelay, metadelay are ints, not time_t's, and can also be made static.
2001-10-27 19:58:56 +00:00
static int metadelay = 28; /* time to delay syncing metadata */
Get rid of the global variable rushjob and replace it with a function in kern/vfs_subr.c named speedup_syncer() which handles the speedup request. Change the various clients of rushjob to use the new function.
1999-06-15 23:37:29 +00:00
SYSCTL_INT(_kern, OID_AUTO, metadelay, CTLFLAG_RW, &metadelay, 0, "");
Style fixes: * Add lots of comments * Convert a couple of assertions to KASSERT() * Minimal whitespace & misapplied {} fixes * Convert #if 0 to #if COMPILING_LINT for code we presently do not support, but want to keep available. Reviewed by: adrian, markm
2000-09-22 12:22:36 +00:00
static int rushjob; /* number of slots to run ASAP */
Get rid of the global variable rushjob and replace it with a function in kern/vfs_subr.c named speedup_syncer() which handles the speedup request. Change the various clients of rushjob to use the new function.
1999-06-15 23:37:29 +00:00
static int stat_rush_requests; /* number of times I/O speeded up */
SYSCTL_INT(_debug, OID_AUTO, rush_requests, CTLFLAG_RW, &stat_rush_requests, 0, "");
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
When shutting down the syncer kernel thread, first tell it to run faster and iterate to over its work list a few times in an attempt to empty the work list before the syncer terminates. This leaves fewer dirty blocks to be written at the "syncing disks" stage and keeps the the "giving up on N buffers" problem from being triggered by the presence of a large soft updates work list at system shutdown time. The downside is that the syncer takes noticeably longer to terminate. Tested by: "Arjan van Leeuwen" <avleeuwen AT piwebs DOT com> Approved by: mckusick
2004-07-01 23:59:19 +00:00
/*
Rework syncer termination code: Speed up the syncer when shutting down by sleeping for a shorter period of time instead of cranking up rushjob and using the normal one second sleep. Skip empty worklist slots when shutting down to avoid lengthy intervals of inactivity. Give I/O more time to complete between steps by not speeding the syncer quite as much. Terminate the syncer after one full pass through the worklist plus one second with the worklist containing nothing but syncer vnodes. Print an indication of shutdown progress to the console. Add a sysctl, vfs.worklist_len, to allow the size of the syncer worklist to be monitored.
2004-07-05 01:07:33 +00:00
* When shutting down the syncer, run it at four times normal speed.
When shutting down the syncer kernel thread, first tell it to run faster and iterate to over its work list a few times in an attempt to empty the work list before the syncer terminates. This leaves fewer dirty blocks to be written at the "syncing disks" stage and keeps the the "giving up on N buffers" problem from being triggered by the presence of a large soft updates work list at system shutdown time. The downside is that the syncer takes noticeably longer to terminate. Tested by: "Arjan van Leeuwen" <avleeuwen AT piwebs DOT com> Approved by: mckusick
2004-07-01 23:59:19 +00:00
*/
Rework syncer termination code: Speed up the syncer when shutting down by sleeping for a shorter period of time instead of cranking up rushjob and using the normal one second sleep. Skip empty worklist slots when shutting down to avoid lengthy intervals of inactivity. Give I/O more time to complete between steps by not speeding the syncer quite as much. Terminate the syncer after one full pass through the worklist plus one second with the worklist containing nothing but syncer vnodes. Print an indication of shutdown progress to the console. Add a sysctl, vfs.worklist_len, to allow the size of the syncer worklist to be monitored.
2004-07-05 01:07:33 +00:00
#define SYNCER_SHUTDOWN_SPEEDUP 4
When shutting down the syncer kernel thread, first tell it to run faster and iterate to over its work list a few times in an attempt to empty the work list before the syncer terminates. This leaves fewer dirty blocks to be written at the "syncing disks" stage and keeps the the "giving up on N buffers" problem from being triggered by the presence of a large soft updates work list at system shutdown time. The downside is that the syncer takes noticeably longer to terminate. Tested by: "Arjan van Leeuwen" <avleeuwen AT piwebs DOT com> Approved by: mckusick
2004-07-01 23:59:19 +00:00
static int sync_vnode_count;
static int syncer_worklist_len;
Rework syncer termination code: Speed up the syncer when shutting down by sleeping for a shorter period of time instead of cranking up rushjob and using the normal one second sleep. Skip empty worklist slots when shutting down to avoid lengthy intervals of inactivity. Give I/O more time to complete between steps by not speeding the syncer quite as much. Terminate the syncer after one full pass through the worklist plus one second with the worklist containing nothing but syncer vnodes. Print an indication of shutdown progress to the console. Add a sysctl, vfs.worklist_len, to allow the size of the syncer worklist to be monitored.
2004-07-05 01:07:33 +00:00
static enum { SYNCER_RUNNING, SYNCER_SHUTTING_DOWN, SYNCER_FINAL_DELAY }
syncer_state;
When shutting down the syncer kernel thread, first tell it to run faster and iterate to over its work list a few times in an attempt to empty the work list before the syncer terminates. This leaves fewer dirty blocks to be written at the "syncing disks" stage and keeps the the "giving up on N buffers" problem from being triggered by the presence of a large soft updates work list at system shutdown time. The downside is that the syncer takes noticeably longer to terminate. Tested by: "Arjan van Leeuwen" <avleeuwen AT piwebs DOT com> Approved by: mckusick
2004-07-01 23:59:19 +00:00
Style fixes: * Add lots of comments * Convert a couple of assertions to KASSERT() * Minimal whitespace & misapplied {} fixes * Convert #if 0 to #if COMPILING_LINT for code we presently do not support, but want to keep available. Reviewed by: adrian, markm
2000-09-22 12:22:36 +00:00
/*
Style fixes based on comments by bde
2000-10-05 18:22:46 +00:00
* Number of vnodes we want to exist at any one time. This is mostly used
* to size hash tables in vnode-related code. It is normally not used in
* getnewvnode(), as wantfreevnodes is normally nonzero.)
*
* XXX desiredvnodes is historical cruft and should not exist.
Style fixes: * Add lots of comments * Convert a couple of assertions to KASSERT() * Minimal whitespace & misapplied {} fixes * Convert #if 0 to #if COMPILING_LINT for code we presently do not support, but want to keep available. Reviewed by: adrian, markm
2000-09-22 12:22:36 +00:00
*/
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
int desiredvnodes;
Gratuitous whitespace cleanup.
2002-06-06 15:46:38 +00:00
SYSCTL_INT(_kern, KERN_MAXVNODES, maxvnodes, CTLFLAG_RW,
Add sysctl descriptions to many SYSCTL_XXXs PR: kern/11197 Submitted by: Adrian Chadd <adrian@FreeBSD.org> Reviewed by: billf(spelling/style/minor nits) Looked at by: bde(style)
1999-05-03 23:57:32 +00:00
&desiredvnodes, 0, "Maximum number of vnodes");
fix minor bug in kern.minvnodes sysctl. Use OID_AUTO.
2001-10-16 23:08:09 +00:00
SYSCTL_INT(_kern, OID_AUTO, minvnodes, CTLFLAG_RW,
- Don't recycle vnodes anymore. Free them once they are dead. getnewvnode now always allocates a new vnode. - Define a new function, vnlru_free, which frees vnodes from the free list. It takes as a parameter the number of vnodes to free, which is wantfreevnodes - freevnodes when called from vnlru_proc or 1 when called from getnewvnode(). For now, getnewvnode() still tries to reclaim a free vnode before creating a new one when we are near the limit. - Define a function, vdestroy, which handles the actual release of memory and teardown of locks, etc. This could become a uma_dtor() routine. - Get rid of minvnodes. Now wantfreevnodes is 1/4th the max vnodes. This keeps more unreferenced vnodes around so that files which have only been stat'd are less likely to be kicked out of the system before we have a chance to read them, etc. These vnodes may still be freed via the normal vnlru_proc() routines which may some day become a real lru.
2005-03-25 05:34:39 +00:00
&wantfreevnodes, 0, "Minimum number of vnodes (legacy)");
Do not initialize static/global variables to 0. Use bss instead of taking up space in the data section.
2001-12-19 01:35:18 +00:00
static int vnlru_nowhere;
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
SYSCTL_INT(_debug, OID_AUTO, vnlru_nowhere, CTLFLAG_RW,
&vnlru_nowhere, 0, "Number of times the vnlru process ran without success");
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
- Now that there are no external users of vfree() make it static. - Move VSHOULDBUSY, VSHOULDFREE, and VTRYRECYCLE into vfs_subr.c so no one else attempts to grow a dependency on them. - Now that objects with pages hold the vnode we don't have to do unlocked checks for the page count in the vm object in VSHOULDFREE. These three macros could simply check for holdcnt state transitions to determine whether the vnode is on the free list already, but the extra safety the flag affords us is probably worth the minimal cost. - The leafonly sysctl and code have been dead for several years now, remove the sysctl and the code that employed it from vtryrecycle(). - vtryrecycle() also no longer has to check the object's page count as the object holds the vnode until it reaches 0. Sponsored by: Isilon Systems, Inc.
2005-03-15 14:38:16 +00:00
/*
* Macros to control when a vnode is freed and recycled. All require
* the vnode interlock.
*/
#define VCANRECYCLE(vp) (((vp)->v_iflag & VI_FREE) && !(vp)->v_holdcnt)
#define VSHOULDFREE(vp) (!((vp)->v_iflag & VI_FREE) && !(vp)->v_holdcnt)
#define VSHOULDBUSY(vp) (((vp)->v_iflag & VI_FREE) && (vp)->v_holdcnt)
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
/*
* Initialize the vnode management data structures.
*/
s/MAX_SAFE_MAXVNODES/MAXVNODES_MAX/g
2004-08-16 08:33:37 +00:00
#ifndef MAXVNODES_MAX
#define MAXVNODES_MAX 100000
#endif
Untangle vfsinit() a bit. Use seperate sysinit functions rather than having a super-function calling bits all over the place.
2000-12-06 07:09:08 +00:00
static void
vntblinit(void *dummy __unused)
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
{
Make the maximum number of vnodes a function of both the physical memory size and the kernel's heap size, specifically, vm_kmem_size. This function allows a maximum of 40% of the vm_kmem_size to be used for vnodes and vm objects. This is a conservative bound based upon recent problem reports. (In other words, a slight increase in this percentage may be safe.) Finally, machines with less than ~3GB of RAM should be unaffected by this change, i.e., the maximum number of vnodes should remain the same. If necessary, machines with 3GB or more of RAM can increase the maximum number of vnodes by increasing vm_kmem_size. Desired by: scottl Tested by: jake Approved by: re (rwatson,scottl)
2003-05-23 19:54:02 +00:00
/*
* Desiredvnodes is a function of the physical memory size and
* the kernel's heap size. Specifically, desiredvnodes scales
* in proportion to the physical memory size until two fifths
* of the kernel's heap size is consumed by vnodes and vm
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
* objects.
Make the maximum number of vnodes a function of both the physical memory size and the kernel's heap size, specifically, vm_kmem_size. This function allows a maximum of 40% of the vm_kmem_size to be used for vnodes and vm objects. This is a conservative bound based upon recent problem reports. (In other words, a slight increase in this percentage may be safe.) Finally, machines with less than ~3GB of RAM should be unaffected by this change, i.e., the maximum number of vnodes should remain the same. If necessary, machines with 3GB or more of RAM can increase the maximum number of vnodes by increasing vm_kmem_size. Desired by: scottl Tested by: jake Approved by: re (rwatson,scottl)
2003-05-23 19:54:02 +00:00
*/
Revert VMCNT_* operations introduction. Probabilly, a general approach is not the better solution here, so we should solve the sched_lock protection problems separately. Requested by: alc Approved by: jeff (mentor)
2007-05-31 22:52:15 +00:00
desiredvnodes = min(maxproc + cnt.v_page_count / 4, 2 * vm_kmem_size /
(5 * (sizeof(struct vm_object) + sizeof(struct vnode))));
s/MAX_SAFE_MAXVNODES/MAXVNODES_MAX/g
2004-08-16 08:33:37 +00:00
if (desiredvnodes > MAXVNODES_MAX) {
Put a cap on the auto-tuning of kern.maxvnodes. Cap value chosen by: scottl
2004-08-02 21:52:43 +00:00
if (bootverbose)
printf("Reducing kern.maxvnodes %d -> %d\n",
s/MAX_SAFE_MAXVNODES/MAXVNODES_MAX/g
2004-08-16 08:33:37 +00:00
desiredvnodes, MAXVNODES_MAX);
desiredvnodes = MAXVNODES_MAX;
Put a cap on the auto-tuning of kern.maxvnodes. Cap value chosen by: scottl
2004-08-02 21:52:43 +00:00
}
Minor style cleanups (mostly removal of trailing whitespaces).
2007-04-10 15:29:37 +00:00
wantfreevnodes = desiredvnodes / 4;
Change callers of mtx_init() to pass in an appropriate lock type name. In most cases NULL is passed, but in some cases such as network driver locks (which use the MTX_NETWORK_LOCK macro) and UMA zone locks, a name is used. Tested on: i386, alpha, sparc64
2002-04-04 21:03:38 +00:00
mtx_init(&mntid_mtx, "mntid", NULL, MTX_DEF);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
TAILQ_INIT(&vnode_free_list);
Change callers of mtx_init() to pass in an appropriate lock type name. In most cases NULL is passed, but in some cases such as network driver locks (which use the MTX_NETWORK_LOCK macro) and UMA zone locks, a name is used. Tested on: i386, alpha, sparc64
2002-04-04 21:03:38 +00:00
mtx_init(&vnode_free_list_mtx, "vnode_free_list", NULL, MTX_DEF);
Remove references to vm_zone.h and switch over to the new uma API. Also, remove maxsockets. If you look carefully you'll notice that the old zone allocator never honored this anyway.
2002-03-20 04:09:59 +00:00
vnode_zone = uma_zcreate("VNODE", sizeof (struct vnode), NULL, NULL,
Allow the vnode zone to return the unused memory. The vnode reference count is/shall be properly maintained for the long time, and VFS shall be safe against the vnode memory reclamation. Proposed by: jeff Tested by: pho
2008-04-24 09:58:33 +00:00
NULL, NULL, UMA_ALIGN_PTR, 0);
Remove references to vm_zone.h and switch over to the new uma API. Also, remove maxsockets. If you look carefully you'll notice that the old zone allocator never honored this anyway.
2002-03-20 04:09:59 +00:00
vnodepoll_zone = uma_zcreate("VNODEPOLL", sizeof (struct vpollinfo),
Allow the vnode zone to return the unused memory. The vnode reference count is/shall be properly maintained for the long time, and VFS shall be safe against the vnode memory reclamation. Proposed by: jeff Tested by: pho
2008-04-24 09:58:33 +00:00
NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
/*
* Initialize the filesystem syncer.
Gratuitous whitespace cleanup.
2002-06-06 15:46:38 +00:00
*/
sync_vnode() has some messy code about locking in order to deal with mount fs needing Giant to be held when processing bufobjs. Use a different subqueue for pending workitems on filesystems requiring Giant. This simplifies the code notably and also reduces the number of Giant acquisitions (and the whole processing cost). Suggested by: jeff Reviewed by: kib Tested by: pho
2008-05-04 13:54:55 +00:00
syncer_workitem_pending[WI_MPSAFEQ] = hashinit(syncer_maxdelay, M_VNODE,
&syncer_mask);
syncer_workitem_pending[WI_GIANTQ] = hashinit(syncer_maxdelay, M_VNODE,
&syncer_mask);
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
syncer_maxdelay = syncer_mask + 1;
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
mtx_init(&sync_mtx, "Syncer mtx", NULL, MTX_DEF);
Remove the use of lbolt from the VFS syncer. It seems we only use `lbolt' inside the VFS syncer and the TTY layer now. Because I'm planning to replace the TTY layer next month, there's no reason to keep `lbolt' if it's only used in a single thread inside the kernel. Because the syncer code wanted to wake up the syncer thread before the timeout, it called sleepq_remove(). Because we now just use a condvar(9) with a timeout value of `hz', we can wake it up using cv_broadcast() without waking up any unrelated threads. Reviewed by: phk
2008-07-30 12:39:18 +00:00
cv_init(&sync_wakeup, "syncer");
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
}
In keeping with style(9)'s recommendations on macros, use a ';' after each SYSINIT() macro invocation. This makes a number of lightweight C parsers much happier with the FreeBSD kernel source, including cflow's prcc and lxr. MFC after: 1 month Discussed with: imp, rink
2008-03-16 10:58:09 +00:00
SYSINIT(vfs, SI_SUB_VFS, SI_ORDER_FIRST, vntblinit, NULL);
Untangle vfsinit() a bit. Use seperate sysinit functions rather than having a super-function calling bits all over the place.
2000-12-06 07:09:08 +00:00
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
/*
This is the kernel Lite/2 commit. There are some requisite userland changes, so don't expect to be able to run the kernel as-is (very well) without the appropriate Lite/2 userland changes. The system boots and can mount UFS filesystems. Untested: ext2fs, msdosfs, NFS Known problems: Incorrect Berkeley ID strings in some files. Mount_std mounts will not work until the getfsent library routine is changed. Reviewed by: various people Submitted by: Jeffery Hsu <hsu@freebsd.org>
1997-02-10 02:22:35 +00:00
* Mark a mount point as busy. Used to synchronize access and to delay
Improve VFS locking: - Implement real draining for vfs consumers by not relying on the mnt_lock and using instead a refcount in order to keep track of lock requesters. - Due to the change above, remove the mnt_lock lockmgr because it is now useless. - Due to the change above, vfs_busy() is no more linked to a lockmgr. Change so its KPI by removing the interlock argument and defining 2 new flags for it: MBF_NOWAIT which basically replaces the LK_NOWAIT of the old version (which was unlinked from the lockmgr alredy) and MBF_MNTLSTLOCK which provides the ability to drop the mountlist_mtx once the mnt interlock is held (ability still desired by most consumers). - The stub used into vfs_mount_destroy(), that allows to override the mnt_ref if running for more than 3 seconds, make it totally useless. Remove it as it was thought to work into older versions. If a problem of "refcount held never going away" should appear, we will need to fix properly instead than trust on such hackish solution. - Fix a bug where returning (with an error) from dounmount() was still leaving the MNTK_MWAIT flag on even if it the waiters were actually woken up. Just a place in vfs_mount_destroy() is left because it is going to recycle the structure in any case, so it doesn't matter. - Remove the markercnt refcount as it is useless. This patch modifies VFS ABI and breaks KPI for vfs_busy() so manpages and __FreeBSD_version will be modified accordingly. Discussed with: kib Tested by: pho
2008-11-02 10:15:42 +00:00
* unmounting. Eventually, mountlist_mtx is not released on failure.
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
*/
The big 4.4BSD Lite to FreeBSD 2.0.0 (Development) patch. Reviewed by: Rodney W. Grimes Submitted by: John Dyson and David Greenman
1994-05-25 09:21:21 +00:00
int
Improve VFS locking: - Implement real draining for vfs consumers by not relying on the mnt_lock and using instead a refcount in order to keep track of lock requesters. - Due to the change above, remove the mnt_lock lockmgr because it is now useless. - Due to the change above, vfs_busy() is no more linked to a lockmgr. Change so its KPI by removing the interlock argument and defining 2 new flags for it: MBF_NOWAIT which basically replaces the LK_NOWAIT of the old version (which was unlinked from the lockmgr alredy) and MBF_MNTLSTLOCK which provides the ability to drop the mountlist_mtx once the mnt interlock is held (ability still desired by most consumers). - The stub used into vfs_mount_destroy(), that allows to override the mnt_ref if running for more than 3 seconds, make it totally useless. Remove it as it was thought to work into older versions. If a problem of "refcount held never going away" should appear, we will need to fix properly instead than trust on such hackish solution. - Fix a bug where returning (with an error) from dounmount() was still leaving the MNTK_MWAIT flag on even if it the waiters were actually woken up. Just a place in vfs_mount_destroy() is left because it is going to recycle the structure in any case, so it doesn't matter. - Remove the markercnt refcount as it is useless. This patch modifies VFS ABI and breaks KPI for vfs_busy() so manpages and __FreeBSD_version will be modified accordingly. Discussed with: kib Tested by: pho
2008-11-02 10:15:42 +00:00
vfs_busy(struct mount *mp, int flags)
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
{
Improve VFS locking: - Implement real draining for vfs consumers by not relying on the mnt_lock and using instead a refcount in order to keep track of lock requesters. - Due to the change above, remove the mnt_lock lockmgr because it is now useless. - Due to the change above, vfs_busy() is no more linked to a lockmgr. Change so its KPI by removing the interlock argument and defining 2 new flags for it: MBF_NOWAIT which basically replaces the LK_NOWAIT of the old version (which was unlinked from the lockmgr alredy) and MBF_MNTLSTLOCK which provides the ability to drop the mountlist_mtx once the mnt interlock is held (ability still desired by most consumers). - The stub used into vfs_mount_destroy(), that allows to override the mnt_ref if running for more than 3 seconds, make it totally useless. Remove it as it was thought to work into older versions. If a problem of "refcount held never going away" should appear, we will need to fix properly instead than trust on such hackish solution. - Fix a bug where returning (with an error) from dounmount() was still leaving the MNTK_MWAIT flag on even if it the waiters were actually woken up. Just a place in vfs_mount_destroy() is left because it is going to recycle the structure in any case, so it doesn't matter. - Remove the markercnt refcount as it is useless. This patch modifies VFS ABI and breaks KPI for vfs_busy() so manpages and __FreeBSD_version will be modified accordingly. Discussed with: kib Tested by: pho
2008-11-02 10:15:42 +00:00
MPASS((flags & ~MBF_MASK) == 0);
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
CTR3(KTR_VFS, "%s: mp %p with flags %d", __func__, mp, flags);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
- Add the tunable and sysctl for the mpsafevfs. It currently defaults to off. - Protect access to mnt_kern_flag with the mointpoint mutex. - Remove some KASSERTs which are not legal checks without the appropriate locks held. - Use VCANRECYCLE() rather than rolling several slightly different checks together. - Return from vtryrecycle() with a recycled vnode rather than a locked vnode. This simplifies some locking. - Remove several GIANT_REQUIRED lines. - Add a few KASSERTs to help with INACT debugging. Sponsored By: Isilon Systems, Inc.
2005-01-24 10:41:01 +00:00
MNT_ILOCK(mp);
- Grab a mnt ref in vfs_busy() before dropping the interlock. This will prevent the mount point from going away while we're waiting on the lock. The ref does not need to persist once we have the lock because the lock prevents the mount point from being unmounted. MFC After: 1 week
2006-02-22 06:20:12 +00:00
MNT_REF(mp);
Reviewed by: various. Ever since I first say the way the mount flags were used I've hated the fact that modes, and events, internal and exported, and short-term and long term flags are all thrown together. Finally it's annoyed me enough.. This patch to the entire FreeBSD tree adds a second mount flag word to the mount struct. it is not exported to userspace. I have moved some of the non exported flags over to this word. this means that we now have 8 free bits in the mount flags. There are another two that might well move over, but which I'm not sure about. The only user visible change would have been in pstat -v, except that davidg has disabled it anyhow. I'd still like to move the state flags and the 'command' flags apart from each other.. e.g. MNT_FORCE really doesn't have the same semantics as MNT_RDONLY, but that's left for another day.
1997-11-12 05:42:33 +00:00
if (mp->mnt_kern_flag & MNTK_UNMOUNT) {
1) Fix a deadlock in the VFS: - threadA runs vfs_rel(mp1) - threadB does unmount the mp1 fs, sets MNTK_UNMOUNT and drop MNT_ILOCK() - threadA runs vfs_busy(mp1) and, as long as, MNTK_UNMOUNT is set, sleeps waiting for threadB to complete the unmount - threadB, in vfs_mount_destroy(), finds mnt_lock > 0 and sleeps waiting for the refcount to expire. Fix the deadlock by adding a flag called MNTK_REFEXPIRE which signals the unmounter is waiting for mnt_ref to expire. The vfs_busy contenders got awake, fails, and if they retry the MNTK_REFEXPIRE won't allow them to sleep again. 2) Simplify significantly the code of vfs_mount_destroy() trimming unnecessary codes: - as long as any reference exited, it is no-more possible to have write-op (primarty and secondary) in progress. - it is no needed to drop and reacquire the mount lock. - filling the structures with dummy values is unuseful as long as it is going to be freed. Tested by: pho, Andrea Barberio <insomniac at slackware dot it> Discussed with: kib
2008-12-16 23:16:10 +00:00
if (flags & MBF_NOWAIT || mp->mnt_kern_flag & MNTK_REFEXPIRE) {
- Grab a mnt ref in vfs_busy() before dropping the interlock. This will prevent the mount point from going away while we're waiting on the lock. The ref does not need to persist once we have the lock because the lock prevents the mount point from being unmounted. MFC After: 1 week
2006-02-22 06:20:12 +00:00
MNT_REL(mp);
- Add the tunable and sysctl for the mpsafevfs. It currently defaults to off. - Protect access to mnt_kern_flag with the mointpoint mutex. - Remove some KASSERTs which are not legal checks without the appropriate locks held. - Use VCANRECYCLE() rather than rolling several slightly different checks together. - Return from vtryrecycle() with a recycled vnode rather than a locked vnode. This simplifies some locking. - Remove several GIANT_REQUIRED lines. - Add a few KASSERTs to help with INACT debugging. Sponsored By: Isilon Systems, Inc.
2005-01-24 10:41:01 +00:00
MNT_IUNLOCK(mp);
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
CTR1(KTR_VFS, "%s: failed busying before sleeping",
__func__);
This is the kernel Lite/2 commit. There are some requisite userland changes, so don't expect to be able to run the kernel as-is (very well) without the appropriate Lite/2 userland changes. The system boots and can mount UFS filesystems. Untested: ext2fs, msdosfs, NFS Known problems: Incorrect Berkeley ID strings in some files. Mount_std mounts will not work until the getfsent library routine is changed. Reviewed by: various people Submitted by: Jeffery Hsu <hsu@freebsd.org>
1997-02-10 02:22:35 +00:00
return (ENOENT);
- Add the tunable and sysctl for the mpsafevfs. It currently defaults to off. - Protect access to mnt_kern_flag with the mointpoint mutex. - Remove some KASSERTs which are not legal checks without the appropriate locks held. - Use VCANRECYCLE() rather than rolling several slightly different checks together. - Return from vtryrecycle() with a recycled vnode rather than a locked vnode. This simplifies some locking. - Remove several GIANT_REQUIRED lines. - Add a few KASSERTs to help with INACT debugging. Sponsored By: Isilon Systems, Inc.
2005-01-24 10:41:01 +00:00
}
Improve VFS locking: - Implement real draining for vfs consumers by not relying on the mnt_lock and using instead a refcount in order to keep track of lock requesters. - Due to the change above, remove the mnt_lock lockmgr because it is now useless. - Due to the change above, vfs_busy() is no more linked to a lockmgr. Change so its KPI by removing the interlock argument and defining 2 new flags for it: MBF_NOWAIT which basically replaces the LK_NOWAIT of the old version (which was unlinked from the lockmgr alredy) and MBF_MNTLSTLOCK which provides the ability to drop the mountlist_mtx once the mnt interlock is held (ability still desired by most consumers). - The stub used into vfs_mount_destroy(), that allows to override the mnt_ref if running for more than 3 seconds, make it totally useless. Remove it as it was thought to work into older versions. If a problem of "refcount held never going away" should appear, we will need to fix properly instead than trust on such hackish solution. - Fix a bug where returning (with an error) from dounmount() was still leaving the MNTK_MWAIT flag on even if it the waiters were actually woken up. Just a place in vfs_mount_destroy() is left because it is going to recycle the structure in any case, so it doesn't matter. - Remove the markercnt refcount as it is useless. This patch modifies VFS ABI and breaks KPI for vfs_busy() so manpages and __FreeBSD_version will be modified accordingly. Discussed with: kib Tested by: pho
2008-11-02 10:15:42 +00:00
if (flags & MBF_MNTLSTLOCK)
mtx_unlock(&mountlist_mtx);
Reviewed by: various. Ever since I first say the way the mount flags were used I've hated the fact that modes, and events, internal and exported, and short-term and long term flags are all thrown together. Finally it's annoyed me enough.. This patch to the entire FreeBSD tree adds a second mount flag word to the mount struct. it is not exported to userspace. I have moved some of the non exported flags over to this word. this means that we now have 8 free bits in the mount flags. There are another two that might well move over, but which I'm not sure about. The only user visible change would have been in pstat -v, except that davidg has disabled it anyhow. I'd still like to move the state flags and the 'command' flags apart from each other.. e.g. MNT_FORCE really doesn't have the same semantics as MNT_RDONLY, but that's left for another day.
1997-11-12 05:42:33 +00:00
mp->mnt_kern_flag |= MNTK_MWAIT;
- Grab a mnt ref in vfs_busy() before dropping the interlock. This will prevent the mount point from going away while we're waiting on the lock. The ref does not need to persist once we have the lock because the lock prevents the mount point from being unmounted. MFC After: 1 week
2006-02-22 06:20:12 +00:00
msleep(mp, MNT_MTX(mp), PVFS, "vfs_busy", 0);
MNT_REL(mp);
MNT_IUNLOCK(mp);
Improve VFS locking: - Implement real draining for vfs consumers by not relying on the mnt_lock and using instead a refcount in order to keep track of lock requesters. - Due to the change above, remove the mnt_lock lockmgr because it is now useless. - Due to the change above, vfs_busy() is no more linked to a lockmgr. Change so its KPI by removing the interlock argument and defining 2 new flags for it: MBF_NOWAIT which basically replaces the LK_NOWAIT of the old version (which was unlinked from the lockmgr alredy) and MBF_MNTLSTLOCK which provides the ability to drop the mountlist_mtx once the mnt interlock is held (ability still desired by most consumers). - The stub used into vfs_mount_destroy(), that allows to override the mnt_ref if running for more than 3 seconds, make it totally useless. Remove it as it was thought to work into older versions. If a problem of "refcount held never going away" should appear, we will need to fix properly instead than trust on such hackish solution. - Fix a bug where returning (with an error) from dounmount() was still leaving the MNTK_MWAIT flag on even if it the waiters were actually woken up. Just a place in vfs_mount_destroy() is left because it is going to recycle the structure in any case, so it doesn't matter. - Remove the markercnt refcount as it is useless. This patch modifies VFS ABI and breaks KPI for vfs_busy() so manpages and __FreeBSD_version will be modified accordingly. Discussed with: kib Tested by: pho
2008-11-02 10:15:42 +00:00
if (flags & MBF_MNTLSTLOCK)
mtx_lock(&mountlist_mtx);
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
CTR1(KTR_VFS, "%s: failed busying after sleep", __func__);
This is the kernel Lite/2 commit. There are some requisite userland changes, so don't expect to be able to run the kernel as-is (very well) without the appropriate Lite/2 userland changes. The system boots and can mount UFS filesystems. Untested: ext2fs, msdosfs, NFS Known problems: Incorrect Berkeley ID strings in some files. Mount_std mounts will not work until the getfsent library routine is changed. Reviewed by: various people Submitted by: Jeffery Hsu <hsu@freebsd.org>
1997-02-10 02:22:35 +00:00
return (ENOENT);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
}
Improve VFS locking: - Implement real draining for vfs consumers by not relying on the mnt_lock and using instead a refcount in order to keep track of lock requesters. - Due to the change above, remove the mnt_lock lockmgr because it is now useless. - Due to the change above, vfs_busy() is no more linked to a lockmgr. Change so its KPI by removing the interlock argument and defining 2 new flags for it: MBF_NOWAIT which basically replaces the LK_NOWAIT of the old version (which was unlinked from the lockmgr alredy) and MBF_MNTLSTLOCK which provides the ability to drop the mountlist_mtx once the mnt interlock is held (ability still desired by most consumers). - The stub used into vfs_mount_destroy(), that allows to override the mnt_ref if running for more than 3 seconds, make it totally useless. Remove it as it was thought to work into older versions. If a problem of "refcount held never going away" should appear, we will need to fix properly instead than trust on such hackish solution. - Fix a bug where returning (with an error) from dounmount() was still leaving the MNTK_MWAIT flag on even if it the waiters were actually woken up. Just a place in vfs_mount_destroy() is left because it is going to recycle the structure in any case, so it doesn't matter. - Remove the markercnt refcount as it is useless. This patch modifies VFS ABI and breaks KPI for vfs_busy() so manpages and __FreeBSD_version will be modified accordingly. Discussed with: kib Tested by: pho
2008-11-02 10:15:42 +00:00
if (flags & MBF_MNTLSTLOCK)
mtx_unlock(&mountlist_mtx);
mp->mnt_lockref++;
MNT_IUNLOCK(mp);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
return (0);
}
/*
This is the kernel Lite/2 commit. There are some requisite userland changes, so don't expect to be able to run the kernel as-is (very well) without the appropriate Lite/2 userland changes. The system boots and can mount UFS filesystems. Untested: ext2fs, msdosfs, NFS Known problems: Incorrect Berkeley ID strings in some files. Mount_std mounts will not work until the getfsent library routine is changed. Reviewed by: various people Submitted by: Jeffery Hsu <hsu@freebsd.org>
1997-02-10 02:22:35 +00:00
* Free a busy filesystem.
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
*/
void
Decontextualize vfs_busy(), vfs_unbusy() and vfs_mount_alloc() functions. Manpages are updated accordingly. Tested by: Diego Sardina <siarodx at gmail dot com>
2008-08-31 14:26:08 +00:00
vfs_unbusy(struct mount *mp)
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
{
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
CTR2(KTR_VFS, "%s: mp %p", __func__, mp);
Improve VFS locking: - Implement real draining for vfs consumers by not relying on the mnt_lock and using instead a refcount in order to keep track of lock requesters. - Due to the change above, remove the mnt_lock lockmgr because it is now useless. - Due to the change above, vfs_busy() is no more linked to a lockmgr. Change so its KPI by removing the interlock argument and defining 2 new flags for it: MBF_NOWAIT which basically replaces the LK_NOWAIT of the old version (which was unlinked from the lockmgr alredy) and MBF_MNTLSTLOCK which provides the ability to drop the mountlist_mtx once the mnt interlock is held (ability still desired by most consumers). - The stub used into vfs_mount_destroy(), that allows to override the mnt_ref if running for more than 3 seconds, make it totally useless. Remove it as it was thought to work into older versions. If a problem of "refcount held never going away" should appear, we will need to fix properly instead than trust on such hackish solution. - Fix a bug where returning (with an error) from dounmount() was still leaving the MNTK_MWAIT flag on even if it the waiters were actually woken up. Just a place in vfs_mount_destroy() is left because it is going to recycle the structure in any case, so it doesn't matter. - Remove the markercnt refcount as it is useless. This patch modifies VFS ABI and breaks KPI for vfs_busy() so manpages and __FreeBSD_version will be modified accordingly. Discussed with: kib Tested by: pho
2008-11-02 10:15:42 +00:00
MNT_ILOCK(mp);
MNT_REL(mp);
Add KASSERTs to make it easier to debug problems like the one fixed in r188141. Reviewed by: kib,attilio Approved by: rwatson (mentor) Tested by: pho Sponsored by: FreeBSD Foundation
2009-02-06 18:16:01 +00:00
KASSERT(mp->mnt_lockref > 0, ("negative mnt_lockref"));
Improve VFS locking: - Implement real draining for vfs consumers by not relying on the mnt_lock and using instead a refcount in order to keep track of lock requesters. - Due to the change above, remove the mnt_lock lockmgr because it is now useless. - Due to the change above, vfs_busy() is no more linked to a lockmgr. Change so its KPI by removing the interlock argument and defining 2 new flags for it: MBF_NOWAIT which basically replaces the LK_NOWAIT of the old version (which was unlinked from the lockmgr alredy) and MBF_MNTLSTLOCK which provides the ability to drop the mountlist_mtx once the mnt interlock is held (ability still desired by most consumers). - The stub used into vfs_mount_destroy(), that allows to override the mnt_ref if running for more than 3 seconds, make it totally useless. Remove it as it was thought to work into older versions. If a problem of "refcount held never going away" should appear, we will need to fix properly instead than trust on such hackish solution. - Fix a bug where returning (with an error) from dounmount() was still leaving the MNTK_MWAIT flag on even if it the waiters were actually woken up. Just a place in vfs_mount_destroy() is left because it is going to recycle the structure in any case, so it doesn't matter. - Remove the markercnt refcount as it is useless. This patch modifies VFS ABI and breaks KPI for vfs_busy() so manpages and __FreeBSD_version will be modified accordingly. Discussed with: kib Tested by: pho
2008-11-02 10:15:42 +00:00
mp->mnt_lockref--;
if (mp->mnt_lockref == 0 && (mp->mnt_kern_flag & MNTK_DRAINING) != 0) {
MPASS(mp->mnt_kern_flag & MNTK_UNMOUNT);
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
CTR1(KTR_VFS, "%s: waking up waiters", __func__);
Improve VFS locking: - Implement real draining for vfs consumers by not relying on the mnt_lock and using instead a refcount in order to keep track of lock requesters. - Due to the change above, remove the mnt_lock lockmgr because it is now useless. - Due to the change above, vfs_busy() is no more linked to a lockmgr. Change so its KPI by removing the interlock argument and defining 2 new flags for it: MBF_NOWAIT which basically replaces the LK_NOWAIT of the old version (which was unlinked from the lockmgr alredy) and MBF_MNTLSTLOCK which provides the ability to drop the mountlist_mtx once the mnt interlock is held (ability still desired by most consumers). - The stub used into vfs_mount_destroy(), that allows to override the mnt_ref if running for more than 3 seconds, make it totally useless. Remove it as it was thought to work into older versions. If a problem of "refcount held never going away" should appear, we will need to fix properly instead than trust on such hackish solution. - Fix a bug where returning (with an error) from dounmount() was still leaving the MNTK_MWAIT flag on even if it the waiters were actually woken up. Just a place in vfs_mount_destroy() is left because it is going to recycle the structure in any case, so it doesn't matter. - Remove the markercnt refcount as it is useless. This patch modifies VFS ABI and breaks KPI for vfs_busy() so manpages and __FreeBSD_version will be modified accordingly. Discussed with: kib Tested by: pho
2008-11-02 10:15:42 +00:00
mp->mnt_kern_flag &= ~MNTK_DRAINING;
wakeup(&mp->mnt_lockref);
}
MNT_IUNLOCK(mp);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
}
/*
* Lookup a mount point by filesystem identifier.
*/
struct mount *
Convert remaining functions in vfs_subr.c from K&R prototypes to ANSI C prototypes, as the majority of new functions added have been in this style. Changing prototype style now results in gcc noticing that the implementation of vn_pollrecord() has a 'short' argument instead of 'int' as prototyped in vnode.h, so correct that definition. In practice this didn't matter as only poll flags in the lower 16 bits are used. MFC after: 1 week
2006-01-21 19:42:10 +00:00
vfs_getvfs(fsid_t *fsid)
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
{
Remove register keyword
2005-01-28 12:39:10 +00:00
struct mount *mp;
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
CTR2(KTR_VFS, "%s: fsid %p", __func__, fsid);
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_lock(&mountlist_mtx);
struct mountlist and struct mount.mnt_list have no business being a CIRCLEQ. Change them to TAILQ_HEAD and TAILQ_ENTRY respectively. This removes ugly mp != (void*)&mountlist comparisons. Requested by: phk Submitted by: Jake Burkholder jake@checker.org PR: 14967
1999-11-20 10:00:46 +00:00
TAILQ_FOREACH(mp, &mountlist, mnt_list) {
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
if (mp->mnt_stat.f_fsid.val[0] == fsid->val[0] &&
This is the kernel Lite/2 commit. There are some requisite userland changes, so don't expect to be able to run the kernel as-is (very well) without the appropriate Lite/2 userland changes. The system boots and can mount UFS filesystems. Untested: ext2fs, msdosfs, NFS Known problems: Incorrect Berkeley ID strings in some files. Mount_std mounts will not work until the getfsent library routine is changed. Reviewed by: various people Submitted by: Jeffery Hsu <hsu@freebsd.org>
1997-02-10 02:22:35 +00:00
mp->mnt_stat.f_fsid.val[1] == fsid->val[1]) {
- Hold a reference from the time vfs_busy starts until vfs_unbusy is called. - vfs_getvfs has to return a reference to prevent the returned mountpoint from changing identities. - Release references acquired via vfs_getvfs. Discussed with: tegge Tested by: kris Sponsored by: Isilon Systems, Inc.
2006-03-31 03:53:25 +00:00
vfs_ref(mp);
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_unlock(&mountlist_mtx);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
return (mp);
Fix some harmless mis-indents. Spotted by: FlexeLint
2002-10-01 15:48:31 +00:00
}
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
}
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_unlock(&mountlist_mtx);
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
CTR2(KTR_VFS, "%s: lookup failed for %p id", __func__, fsid);
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
return ((struct mount *) 0);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
}
In the nfsrv_fhtovp(), after the vfs_getvfs() function found the pointer to the fs, but before a vnode on the fs is locked, unmount may free fs structures, causing access to destroyed data and freed memory. Introduce a vfs_busymp() function that looks up and busies found fs while mountlist_mtx is held. Use it in nfsrv_fhtovp() and in the implementation of the handle syscalls. Two other uses of the vfs_getvfs() in the vfs_subr.c, namely in sysctl_vfs_ctl and vfs_getnewfsid seems to be ok. In particular, sysctl_vfs_ctl is protected by Giant by being a non-sleeping sysctl handler, that prevents Giant-locked unmount code to interfere with it. Noted by: tegge Reviewed by: dfr Tested by: pho MFC after: 1 month
2008-11-29 13:34:59 +00:00
/*
* Lookup a mount point by filesystem identifier, busying it before
* returning.
*/
struct mount *
vfs_busyfs(fsid_t *fsid)
{
struct mount *mp;
int error;
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
CTR2(KTR_VFS, "%s: fsid %p", __func__, fsid);
In the nfsrv_fhtovp(), after the vfs_getvfs() function found the pointer to the fs, but before a vnode on the fs is locked, unmount may free fs structures, causing access to destroyed data and freed memory. Introduce a vfs_busymp() function that looks up and busies found fs while mountlist_mtx is held. Use it in nfsrv_fhtovp() and in the implementation of the handle syscalls. Two other uses of the vfs_getvfs() in the vfs_subr.c, namely in sysctl_vfs_ctl and vfs_getnewfsid seems to be ok. In particular, sysctl_vfs_ctl is protected by Giant by being a non-sleeping sysctl handler, that prevents Giant-locked unmount code to interfere with it. Noted by: tegge Reviewed by: dfr Tested by: pho MFC after: 1 month
2008-11-29 13:34:59 +00:00
mtx_lock(&mountlist_mtx);
TAILQ_FOREACH(mp, &mountlist, mnt_list) {
if (mp->mnt_stat.f_fsid.val[0] == fsid->val[0] &&
mp->mnt_stat.f_fsid.val[1] == fsid->val[1]) {
error = vfs_busy(mp, MBF_MNTLSTLOCK);
if (error) {
mtx_unlock(&mountlist_mtx);
return (NULL);
}
return (mp);
}
}
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
CTR2(KTR_VFS, "%s: lookup failed for %p id", __func__, fsid);
In the nfsrv_fhtovp(), after the vfs_getvfs() function found the pointer to the fs, but before a vnode on the fs is locked, unmount may free fs structures, causing access to destroyed data and freed memory. Introduce a vfs_busymp() function that looks up and busies found fs while mountlist_mtx is held. Use it in nfsrv_fhtovp() and in the implementation of the handle syscalls. Two other uses of the vfs_getvfs() in the vfs_subr.c, namely in sysctl_vfs_ctl and vfs_getnewfsid seems to be ok. In particular, sysctl_vfs_ctl is protected by Giant by being a non-sleeping sysctl handler, that prevents Giant-locked unmount code to interfere with it. Noted by: tegge Reviewed by: dfr Tested by: pho MFC after: 1 month
2008-11-29 13:34:59 +00:00
mtx_unlock(&mountlist_mtx);
return ((struct mount *) 0);
}
Introduce vfs_suser(), used to test if a user should have special privs for a mount.
2004-07-06 09:37:43 +00:00
/*
Correct typos.
2007-04-10 15:22:40 +00:00
* Check if a user can access privileged mount options.
Introduce vfs_suser(), used to test if a user should have special privs for a mount.
2004-07-06 09:37:43 +00:00
*/
int
vfs_suser(struct mount *mp, struct thread *td)
{
int error;
Fix jails and jail-friendly file systems handling: - We need to allow for PRIV_VFS_MOUNT_OWNER inside a jail. - Move security checks to vfs_suser() and deny unmounting and updating for jailed root from different jails, etc. OK'ed by: rwatson
2007-04-13 23:54:22 +00:00
/*
* If the thread is jailed, but this is not a jail-friendly file
* system, deny immediately.
*/
Update ZFS from version 6 to 13 and bring some FreeBSD-specific changes. This bring huge amount of changes, I'll enumerate only user-visible changes: - Delegated Administration Allows regular users to perform ZFS operations, like file system creation, snapshot creation, etc. - L2ARC Level 2 cache for ZFS - allows to use additional disks for cache. Huge performance improvements mostly for random read of mostly static content. - slog Allow to use additional disks for ZFS Intent Log to speed up operations like fsync(2). - vfs.zfs.super_owner Allows regular users to perform privileged operations on files stored on ZFS file systems owned by him. Very careful with this one. - chflags(2) Not all the flags are supported. This still needs work. - ZFSBoot Support to boot off of ZFS pool. Not finished, AFAIK. Submitted by: dfr - Snapshot properties - New failure modes Before if write requested failed, system paniced. Now one can select from one of three failure modes: - panic - panic on write error - wait - wait for disk to reappear - continue - serve read requests if possible, block write requests - Refquota, refreservation properties Just quota and reservation properties, but don't count space consumed by children file systems, clones and snapshots. - Sparse volumes ZVOLs that don't reserve space in the pool. - External attributes Compatible with extattr(2). - NFSv4-ACLs Not sure about the status, might not be complete yet. Submitted by: trasz - Creation-time properties - Regression tests for zpool(8) command. Obtained from: OpenSolaris
2008-11-17 20:49:29 +00:00
if (!(mp->mnt_vfc->vfc_flags & VFCF_JAIL) && jailed(td->td_ucred))
Fix jails and jail-friendly file systems handling: - We need to allow for PRIV_VFS_MOUNT_OWNER inside a jail. - Move security checks to vfs_suser() and deny unmounting and updating for jailed root from different jails, etc. OK'ed by: rwatson
2007-04-13 23:54:22 +00:00
return (EPERM);
/*
* If the file system was mounted outside a jail and a jailed thread
* tries to access it, deny immediately.
*/
if (!jailed(mp->mnt_cred) && jailed(td->td_ucred))
return (EPERM);
/*
* If the file system was mounted inside different jail that the jail of
* the calling thread, deny immediately.
*/
if (jailed(mp->mnt_cred) && jailed(td->td_ucred) &&
mp->mnt_cred->cr_prison != td->td_ucred->cr_prison) {
return (EPERM);
}
Update ZFS from version 6 to 13 and bring some FreeBSD-specific changes. This bring huge amount of changes, I'll enumerate only user-visible changes: - Delegated Administration Allows regular users to perform ZFS operations, like file system creation, snapshot creation, etc. - L2ARC Level 2 cache for ZFS - allows to use additional disks for cache. Huge performance improvements mostly for random read of mostly static content. - slog Allow to use additional disks for ZFS Intent Log to speed up operations like fsync(2). - vfs.zfs.super_owner Allows regular users to perform privileged operations on files stored on ZFS file systems owned by him. Very careful with this one. - chflags(2) Not all the flags are supported. This still needs work. - ZFSBoot Support to boot off of ZFS pool. Not finished, AFAIK. Submitted by: dfr - Snapshot properties - New failure modes Before if write requested failed, system paniced. Now one can select from one of three failure modes: - panic - panic on write error - wait - wait for disk to reappear - continue - serve read requests if possible, block write requests - Refquota, refreservation properties Just quota and reservation properties, but don't count space consumed by children file systems, clones and snapshots. - Sparse volumes ZVOLs that don't reserve space in the pool. - External attributes Compatible with extattr(2). - NFSv4-ACLs Not sure about the status, might not be complete yet. Submitted by: trasz - Creation-time properties - Regression tests for zpool(8) command. Obtained from: OpenSolaris
2008-11-17 20:49:29 +00:00
/*
* If file system supports delegated administration, we don't check
* for the PRIV_VFS_MOUNT_OWNER privilege - it will be better verified
* by the file system itself.
* If this is not the user that did original mount, we check for
* the PRIV_VFS_MOUNT_OWNER privilege.
*/
if (!(mp->mnt_vfc->vfc_flags & VFCF_DELEGADMIN) &&
Introduce vfs_suser(), used to test if a user should have special privs for a mount.
2004-07-06 09:37:43 +00:00
mp->mnt_cred->cr_uid != td->td_ucred->cr_uid) {
Sweep kernel replacing suser(9) calls with priv(9) calls, assigning specific privilege names to a broad range of privileges. These may require some future tweaking. Sponsored by: nCircle Network Security, Inc. Obtained from: TrustedBSD Project Discussed on: arch@ Reviewed (at least in part) by: mlaier, jmg, pjd, bde, ceri, Alex Lyashkov <umka at sevcity dot net>, Skip Ford <skip dot ford at verizon dot net>, Antoine Brodin <antoine dot brodin at laposte dot net>
2006-11-06 13:42:10 +00:00
if ((error = priv_check(td, PRIV_VFS_MOUNT_OWNER)) != 0)
Introduce vfs_suser(), used to test if a user should have special privs for a mount.
2004-07-06 09:37:43 +00:00
return (error);
}
return (0);
}
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
/*
Don't try so hard to make the lower 16 bits of fsids unique. It tended to recycle full fsids after only 16 mount/unmount's. This is probably too often for exported fsids. Now we recycle the full fsids only after 2^16 mount/ umount's and only ensure uniqueness in the lower 16 bits if there have been <= 256 calls to vfs_getnewfsid() since the system started.
2000-03-14 14:19:49 +00:00
* Get a new unique fsid. Try to make its val[0] unique, since this value
* will be used to create fake device numbers for stat(). Also try (but
* not so hard) make its val[0] unique mod 2^16, since some emulators only
* support 16-bit device numbers. We end up with unique val[0]'s for the
* first 2^16 calls and unique val[0]'s mod 2^16 for the first 2^8 calls.
Fix BOOTP root FS mounts. Also cleanup vfs_getnewfsid() and collapse addaliasu() into addalias() (no operational change) and clarify comments relating to a trick that vclean() uses. The fix to BOOTP is yet another hack. Actually, rootfsid handling is already a major hack. The whole thing needs to be cleaned up. Reviewed by: David Greenman <dg@root.com>, Alan Cox <alc@cs.rice.edu>
1999-09-19 06:24:21 +00:00
*
Try harder to make the lower 16 bits of fsids unique. The vfs type number was packed very wastefully, giving perfect non-uniqeness in the lower 16 bits of fsids for filesystems with the same vfs type. This made linux_stat() return perfectly non-unique (broken) 16-bit st_dev's for nfs mount points, and effectively reduced mntid_base to 8 bits so that the vfs_getnewfsid() looped endlessly when there are already 256 mounted filesystems with the required vfs type. Approved by: jkh
2000-03-12 14:23:21 +00:00
* Keep in mind that several mounts may be running in parallel. Starting
Don't try so hard to make the lower 16 bits of fsids unique. It tended to recycle full fsids after only 16 mount/unmount's. This is probably too often for exported fsids. Now we recycle the full fsids only after 2^16 mount/ umount's and only ensure uniqueness in the lower 16 bits if there have been <= 256 calls to vfs_getnewfsid() since the system started.
2000-03-14 14:19:49 +00:00
* the search one past where the previous search terminated is both a
* micro-optimization and a defense against returning the same fsid to
* different mounts.
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
*/
void
Convert remaining functions in vfs_subr.c from K&R prototypes to ANSI C prototypes, as the majority of new functions added have been in this style. Changing prototype style now results in gcc noticing that the implementation of vn_pollrecord() has a 'short' argument instead of 'int' as prototyped in vnode.h, so correct that definition. In practice this didn't matter as only poll flags in the lower 16 bits are used. MFC after: 1 week
2006-01-21 19:42:10 +00:00
vfs_getnewfsid(struct mount *mp)
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
{
Don't try so hard to make the lower 16 bits of fsids unique. It tended to recycle full fsids after only 16 mount/unmount's. This is probably too often for exported fsids. Now we recycle the full fsids only after 2^16 mount/ umount's and only ensure uniqueness in the lower 16 bits if there have been <= 256 calls to vfs_getnewfsid() since the system started.
2000-03-14 14:19:49 +00:00
static u_int16_t mntid_base;
- Hold a reference from the time vfs_busy starts until vfs_unbusy is called. - vfs_getvfs has to return a reference to prevent the returned mountpoint from changing identities. - Release references acquired via vfs_getvfs. Discussed with: tegge Tested by: kris Sponsored by: Isilon Systems, Inc.
2006-03-31 03:53:25 +00:00
struct mount *nmp;
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
fsid_t tfsid;
Don't try so hard to make the lower 16 bits of fsids unique. It tended to recycle full fsids after only 16 mount/unmount's. This is probably too often for exported fsids. Now we recycle the full fsids only after 2^16 mount/ umount's and only ensure uniqueness in the lower 16 bits if there have been <= 256 calls to vfs_getnewfsid() since the system started.
2000-03-14 14:19:49 +00:00
int mtype;
Fix BOOTP root FS mounts. Also cleanup vfs_getnewfsid() and collapse addaliasu() into addalias() (no operational change) and clarify comments relating to a trick that vclean() uses. The fix to BOOTP is yet another hack. Actually, rootfsid handling is already a major hack. The whole thing needs to be cleaned up. Reviewed by: David Greenman <dg@root.com>, Alan Cox <alc@cs.rice.edu>
1999-09-19 06:24:21 +00:00
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
CTR2(KTR_VFS, "%s: mp %p", __func__, mp);
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_lock(&mntid_mtx);
This is the kernel Lite/2 commit. There are some requisite userland changes, so don't expect to be able to run the kernel as-is (very well) without the appropriate Lite/2 userland changes. The system boots and can mount UFS filesystems. Untested: ext2fs, msdosfs, NFS Known problems: Incorrect Berkeley ID strings in some files. Mount_std mounts will not work until the getfsent library routine is changed. Reviewed by: various people Submitted by: Jeffery Hsu <hsu@freebsd.org>
1997-02-10 02:22:35 +00:00
mtype = mp->mnt_vfc->vfc_typenum;
Try harder to make the lower 16 bits of fsids unique. The vfs type number was packed very wastefully, giving perfect non-uniqeness in the lower 16 bits of fsids for filesystems with the same vfs type. This made linux_stat() return perfectly non-unique (broken) 16-bit st_dev's for nfs mount points, and effectively reduced mntid_base to 8 bits so that the vfs_getnewfsid() looped endlessly when there are already 256 mounted filesystems with the required vfs type. Approved by: jkh
2000-03-12 14:23:21 +00:00
tfsid.val[1] = mtype;
Fix support for more than 256 simultaneous mounts. Theoretical limit is 2^16 mounts per fs type. Reported by: Troy Arie Cobb <tcobb@staff.circle.net> via phk Reviewed by: bde
2000-07-07 14:01:08 +00:00
mtype = (mtype & 0xFF) << 24;
Don't try so hard to make the lower 16 bits of fsids unique. It tended to recycle full fsids after only 16 mount/unmount's. This is probably too often for exported fsids. Now we recycle the full fsids only after 2^16 mount/ umount's and only ensure uniqueness in the lower 16 bits if there have been <= 256 calls to vfs_getnewfsid() since the system started.
2000-03-14 14:19:49 +00:00
for (;;) {
Second half of the dev_t cleanup. The big lines are: NODEV -> NULL NOUDEV -> NODEV udev_t -> dev_t udev2dev() -> findcdev() Various minor adjustments including handling of userland access to kernel space struct cdev etc.
2004-06-17 17:16:53 +00:00
tfsid.val[0] = makedev(255,
Fix support for more than 256 simultaneous mounts. Theoretical limit is 2^16 mounts per fs type. Reported by: Troy Arie Cobb <tcobb@staff.circle.net> via phk Reviewed by: bde
2000-07-07 14:01:08 +00:00
mtype | ((mntid_base & 0xFF00) << 8) | (mntid_base & 0xFF));
mntid_base++;
- Hold a reference from the time vfs_busy starts until vfs_unbusy is called. - vfs_getvfs has to return a reference to prevent the returned mountpoint from changing identities. - Release references acquired via vfs_getvfs. Discussed with: tegge Tested by: kris Sponsored by: Isilon Systems, Inc.
2006-03-31 03:53:25 +00:00
if ((nmp = vfs_getvfs(&tfsid)) == NULL)
Fix BOOTP root FS mounts. Also cleanup vfs_getnewfsid() and collapse addaliasu() into addalias() (no operational change) and clarify comments relating to a trick that vclean() uses. The fix to BOOTP is yet another hack. Actually, rootfsid handling is already a major hack. The whole thing needs to be cleaned up. Reviewed by: David Greenman <dg@root.com>, Alan Cox <alc@cs.rice.edu>
1999-09-19 06:24:21 +00:00
break;
- Hold a reference from the time vfs_busy starts until vfs_unbusy is called. - vfs_getvfs has to return a reference to prevent the returned mountpoint from changing identities. - Release references acquired via vfs_getvfs. Discussed with: tegge Tested by: kris Sponsored by: Isilon Systems, Inc.
2006-03-31 03:53:25 +00:00
vfs_rel(nmp);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
}
mp->mnt_stat.f_fsid.val[0] = tfsid.val[0];
Fix BOOTP root FS mounts. Also cleanup vfs_getnewfsid() and collapse addaliasu() into addalias() (no operational change) and clarify comments relating to a trick that vclean() uses. The fix to BOOTP is yet another hack. Actually, rootfsid handling is already a major hack. The whole thing needs to be cleaned up. Reviewed by: David Greenman <dg@root.com>, Alan Cox <alc@cs.rice.edu>
1999-09-19 06:24:21 +00:00
mp->mnt_stat.f_fsid.val[1] = tfsid.val[1];
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_unlock(&mntid_mtx);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
}
Support full-precision file timestamps. Until now, only the seconds have been maintained, and that is still the default. A new sysctl variable "vfs.timestamp_precision" can be used to enable higher levels of precision: 0 = seconds only; nanoseconds zeroed (default). 1 = seconds and nanoseconds, accurate within 1/HZ. 2 = seconds and nanoseconds, truncated to microseconds. >=3 = seconds and nanoseconds, maximum precision. Level 1 uses getnanotime(), which is fast but can be wrong by up to 1/HZ. Level 2 uses microtime(). It might be desirable for consistency with utimes() and friends, which take timeval structures rather than timespecs. Level 3 uses nanotime() for the higest precision. I benchmarked levels 0, 1, and 3 by copying a 550 MB tree with "cpio -pdu". There was almost negligible difference in the system times -- much less than 1%, and less than the variation among multiple runs at the same level. Bruce Evans dreamed up a torture test involving 1-byte reads with intervening fstat() calls, but the cpio test seems more realistic to me. This feature is currently implemented only for the UFS (FFS and MFS) filesystems. But I think it should be easy to support it in the others as well. An earlier version of this was reviewed by Bruce. He's not to blame for any breakage I've introduced since then. Reviewed by: bde (an earlier version of the code)
1999-08-22 00:15:16 +00:00
/*
* Knob to control the precision of file timestamps:
*
* 0 = seconds only; nanoseconds zeroed.
* 1 = seconds and nanoseconds, accurate within 1/HZ.
* 2 = seconds and nanoseconds, truncated to microseconds.
* >=3 = seconds and nanoseconds, maximum precision.
*/
enum { TSP_SEC, TSP_HZ, TSP_USEC, TSP_NSEC };
static int timestamp_precision = TSP_SEC;
SYSCTL_INT(_vfs, OID_AUTO, timestamp_precision, CTLFLAG_RW,
&timestamp_precision, 0, "");
/*
* Get a current timestamp.
*/
void
Convert remaining functions in vfs_subr.c from K&R prototypes to ANSI C prototypes, as the majority of new functions added have been in this style. Changing prototype style now results in gcc noticing that the implementation of vn_pollrecord() has a 'short' argument instead of 'int' as prototyped in vnode.h, so correct that definition. In practice this didn't matter as only poll flags in the lower 16 bits are used. MFC after: 1 week
2006-01-21 19:42:10 +00:00
vfs_timestamp(struct timespec *tsp)
Support full-precision file timestamps. Until now, only the seconds have been maintained, and that is still the default. A new sysctl variable "vfs.timestamp_precision" can be used to enable higher levels of precision: 0 = seconds only; nanoseconds zeroed (default). 1 = seconds and nanoseconds, accurate within 1/HZ. 2 = seconds and nanoseconds, truncated to microseconds. >=3 = seconds and nanoseconds, maximum precision. Level 1 uses getnanotime(), which is fast but can be wrong by up to 1/HZ. Level 2 uses microtime(). It might be desirable for consistency with utimes() and friends, which take timeval structures rather than timespecs. Level 3 uses nanotime() for the higest precision. I benchmarked levels 0, 1, and 3 by copying a 550 MB tree with "cpio -pdu". There was almost negligible difference in the system times -- much less than 1%, and less than the variation among multiple runs at the same level. Bruce Evans dreamed up a torture test involving 1-byte reads with intervening fstat() calls, but the cpio test seems more realistic to me. This feature is currently implemented only for the UFS (FFS and MFS) filesystems. But I think it should be easy to support it in the others as well. An earlier version of this was reviewed by Bruce. He's not to blame for any breakage I've introduced since then. Reviewed by: bde (an earlier version of the code)
1999-08-22 00:15:16 +00:00
{
struct timeval tv;
switch (timestamp_precision) {
case TSP_SEC:
tsp->tv_sec = time_second;
tsp->tv_nsec = 0;
break;
case TSP_HZ:
getnanotime(tsp);
break;
case TSP_USEC:
microtime(&tv);
TIMEVAL_TO_TIMESPEC(&tv, tsp);
break;
case TSP_NSEC:
default:
nanotime(tsp);
break;
}
}
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
/*
* Set vnode attributes to VNOVAL
*/
The big 4.4BSD Lite to FreeBSD 2.0.0 (Development) patch. Reviewed by: Rodney W. Grimes Submitted by: John Dyson and David Greenman
1994-05-25 09:21:21 +00:00
void
Convert remaining functions in vfs_subr.c from K&R prototypes to ANSI C prototypes, as the majority of new functions added have been in this style. Changing prototype style now results in gcc noticing that the implementation of vn_pollrecord() has a 'short' argument instead of 'int' as prototyped in vnode.h, so correct that definition. In practice this didn't matter as only poll flags in the lower 16 bits are used. MFC after: 1 week
2006-01-21 19:42:10 +00:00
vattr_null(struct vattr *vap)
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
{
vap->va_type = VNON;
The big 4.4BSD Lite to FreeBSD 2.0.0 (Development) patch. Reviewed by: Rodney W. Grimes Submitted by: John Dyson and David Greenman
1994-05-25 09:21:21 +00:00
vap->va_size = VNOVAL;
vap->va_bytes = VNOVAL;
Initialise all the fields separately in vattr_null since on the alpha they are not all the same width.
1998-07-12 16:45:39 +00:00
vap->va_mode = VNOVAL;
vap->va_nlink = VNOVAL;
vap->va_uid = VNOVAL;
vap->va_gid = VNOVAL;
vap->va_fsid = VNOVAL;
vap->va_fileid = VNOVAL;
vap->va_blocksize = VNOVAL;
vap->va_rdev = VNOVAL;
vap->va_atime.tv_sec = VNOVAL;
vap->va_atime.tv_nsec = VNOVAL;
vap->va_mtime.tv_sec = VNOVAL;
vap->va_mtime.tv_nsec = VNOVAL;
vap->va_ctime.tv_sec = VNOVAL;
vap->va_ctime.tv_nsec = VNOVAL;
Change utimes to set the file creation time (for filesystems that support creation times such as UFS2) to the value of the modification time if the value of the modification time is older than the current creation time. See utimes(2) for further details. Sponsored by: DARPA & NAI Labs.
2002-07-17 02:03:19 +00:00
vap->va_birthtime.tv_sec = VNOVAL;
vap->va_birthtime.tv_nsec = VNOVAL;
Initialise all the fields separately in vattr_null since on the alpha they are not all the same width.
1998-07-12 16:45:39 +00:00
vap->va_flags = VNOVAL;
vap->va_gen = VNOVAL;
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
vap->va_vaflags = 0;
}
Implement kern.maxvnodes. adjusting kern.maxvnodes now actually has a real effect. Optimize vfs_msync(). Avoid having to continually drop and re-obtain mutexes when scanning the vnode list. Improves looping case by 500%. Optimize ffs_sync(). Avoid having to continually drop and re-obtain mutexes when scanning the vnode list. This makes a couple of assumptions, which I believe are ok, in regards to vnode stability when the mount list mutex is held. Improves looping case by 500%. (more optimization work is needed on top of these fixes) MFC after: 1 week
2001-10-26 00:08:05 +00:00
/*
* This routine is called when we have too many vnodes. It attempts
* to free <count> vnodes and will potentially free vnodes that still
* have VM backing store (VM backing store is typically the cause
* of a vnode blowout so we want to do this). Therefore, this operation
* is not considered cheap.
*
* A number of conditions may prevent a vnode from being reclaimed.
* the buffer cache may have references on the vnode, a directory
* vnode may still have references due to the namei cache representing
* underlying files, or the vnode may be in active use. It is not
* desireable to reuse such vnodes. These conditions may cause the
* number of vnodes to reach some minimum value regardless of what
This is a forward port of Peter's vlrureclaim() fix, with some minor mods by me to make it more efficient. The original code had serious balancing problems and could also deadlock easily. This code relegates the vnode reclamation to its own kproc and relaxes the vnode reclamation requirements to better maintain kern.maxvnodes. This code still doesn't balance as well as it could, but it does a much better job then the original code. Approved by: re@freebsd.org Obtained from: ps, peter, dillon MFS Assuming: Assuming no problems crop up in Yahoo testing MFC after: 7 days
2001-12-18 20:48:54 +00:00
* you set kern.maxvnodes to. Do not set kern.maxvnodes too low.
Implement kern.maxvnodes. adjusting kern.maxvnodes now actually has a real effect. Optimize vfs_msync(). Avoid having to continually drop and re-obtain mutexes when scanning the vnode list. Improves looping case by 500%. Optimize ffs_sync(). Avoid having to continually drop and re-obtain mutexes when scanning the vnode list. This makes a couple of assumptions, which I believe are ok, in regards to vnode stability when the mount list mutex is held. Improves looping case by 500%. (more optimization work is needed on top of these fixes) MFC after: 1 week
2001-10-26 00:08:05 +00:00
*/
This is a forward port of Peter's vlrureclaim() fix, with some minor mods by me to make it more efficient. The original code had serious balancing problems and could also deadlock easily. This code relegates the vnode reclamation to its own kproc and relaxes the vnode reclamation requirements to better maintain kern.maxvnodes. This code still doesn't balance as well as it could, but it does a much better job then the original code. Approved by: re@freebsd.org Obtained from: ps, peter, dillon MFS Assuming: Assuming no problems crop up in Yahoo testing MFC after: 7 days
2001-12-18 20:48:54 +00:00
static int
Adjust the number of vnodes scanned by vlrureclaim() according to the size of the vnode list.
2003-03-26 22:15:58 +00:00
vlrureclaim(struct mount *mp)
Implement kern.maxvnodes. adjusting kern.maxvnodes now actually has a real effect. Optimize vfs_msync(). Avoid having to continually drop and re-obtain mutexes when scanning the vnode list. Improves looping case by 500%. Optimize ffs_sync(). Avoid having to continually drop and re-obtain mutexes when scanning the vnode list. This makes a couple of assumptions, which I believe are ok, in regards to vnode stability when the mount list mutex is held. Improves looping case by 500%. (more optimization work is needed on top of these fixes) MFC after: 1 week
2001-10-26 00:08:05 +00:00
{
struct vnode *vp;
This is a forward port of Peter's vlrureclaim() fix, with some minor mods by me to make it more efficient. The original code had serious balancing problems and could also deadlock easily. This code relegates the vnode reclamation to its own kproc and relaxes the vnode reclamation requirements to better maintain kern.maxvnodes. This code still doesn't balance as well as it could, but it does a much better job then the original code. Approved by: re@freebsd.org Obtained from: ps, peter, dillon MFS Assuming: Assuming no problems crop up in Yahoo testing MFC after: 7 days
2001-12-18 20:48:54 +00:00
int done;
Add vlruvp() routine - implements LRU operation for vnode recycling. We calculate a trigger point that both guarentees we will find a sufficient number of vnodes to recycle and prevents us from recycling vnodes with lots of resident pages. This particular section of code is designed to recycle vnodes, not do unnecessary frees of cached VM pages.
2002-01-10 18:31:53 +00:00
int trigger;
int usevnodes;
Adjust the number of vnodes scanned by vlrureclaim() according to the size of the vnode list.
2003-03-26 22:15:58 +00:00
int count;
Add vlruvp() routine - implements LRU operation for vnode recycling. We calculate a trigger point that both guarentees we will find a sufficient number of vnodes to recycle and prevents us from recycling vnodes with lots of resident pages. This particular section of code is designed to recycle vnodes, not do unnecessary frees of cached VM pages.
2002-01-10 18:31:53 +00:00
/*
* Calculate the trigger point, don't allow user
* screwups to blow us up. This prevents us from
* recycling vnodes with lots of resident pages. We
* aren't trying to free memory, we are trying to
* free vnodes.
*/
usevnodes = desiredvnodes;
if (usevnodes <= 0)
usevnodes = 1;
Revert VMCNT_* operations introduction. Probabilly, a general approach is not the better solution here, so we should solve the sched_lock protection problems separately. Requested by: alc Approved by: jeff (mentor)
2007-05-31 22:52:15 +00:00
trigger = cnt.v_page_count * 2 / usevnodes;
This is a forward port of Peter's vlrureclaim() fix, with some minor mods by me to make it more efficient. The original code had serious balancing problems and could also deadlock easily. This code relegates the vnode reclamation to its own kproc and relaxes the vnode reclamation requirements to better maintain kern.maxvnodes. This code still doesn't balance as well as it could, but it does a much better job then the original code. Approved by: re@freebsd.org Obtained from: ps, peter, dillon MFS Assuming: Assuming no problems crop up in Yahoo testing MFC after: 7 days
2001-12-18 20:48:54 +00:00
done = 0;
- Remove vx_lock, vx_unlock, vx_wait, etc. - Add a vn_start_write/vn_finished_write around vlrureclaim so we don't do writing ops without suspending. This could suspend the vlruproc which should not be a problem under normal circumstances. - Manually implement VMIGHTFREE in vlrureclaim as this was the only instance where it was used. - Acquire a lock before calling vgone() as it now requires it. - Move the acquisition of the vnode interlock from vtryrecycle() to getnewvnode() so that if it fails we don't drop and reacquire the vnode_free_list_mtx. - Check for a usecount or holdcount at the end of vtryrecycle() in case someone grabbed a ref while we were recycling. Abort the recycle, and on the final ref drop this vnode will be placed on the head of the free list. - Move the redundant VOP_INACTIVE protection code into the local vinactive() routine to avoid code bloat. - Keep the vnode lock held across calls to vgone() in several places. - vgonel() no longer uses XLOCK, instead callers must hold an exclusive vnode lock. The VI_DOOMED flag is set to allow other threads to detect a vnode which is no longer valid. This flag is set until the last reference is gone, and there are no chances for a new ref. vgonel() holds this lock across the entire function, which greatly simplifies logic. _ Only vfree() in one place in vgone() not three. - Adjust vget() to check the VI_DOOMED flag prior to waiting on the lock in the LK_NOWAIT case. In other cases, check after we have slept and acquired an exlusive lock. This will simulate the old vx_wait() behavior. Sponsored by: Isilon Systems, Inc.
2005-03-13 11:54:28 +00:00
vn_start_write(NULL, &mp, V_WAIT);
Remove mntvnode_mtx and replace it with per-mountpoint mutex. Introduce two new macros MNT_ILOCK(mp)/MNT_IUNLOCK(mp) to operate on this mutex transparently. Eventually new mutex will be protecting more fields in struct mount, not only vnode list. Discussed with: jeff
2003-11-05 04:30:08 +00:00
MNT_ILOCK(mp);
Adjust the number of vnodes scanned by vlrureclaim() according to the size of the vnode list.
2003-03-26 22:15:58 +00:00
count = mp->mnt_nvnodelistsize / 10 + 1;
Add marker vnodes to ensure that all vnodes associated with the mount point are iterated over when using MNT_VNODE_FOREACH. Reviewed by: truckman
2006-01-09 20:42:19 +00:00
while (count != 0) {
vp = TAILQ_FIRST(&mp->mnt_nvnodelist);
while (vp != NULL && vp->v_type == VMARKER)
vp = TAILQ_NEXT(vp, v_nmntvnodes);
if (vp == NULL)
break;
Implement kern.maxvnodes. adjusting kern.maxvnodes now actually has a real effect. Optimize vfs_msync(). Avoid having to continually drop and re-obtain mutexes when scanning the vnode list. Improves looping case by 500%. Optimize ffs_sync(). Avoid having to continually drop and re-obtain mutexes when scanning the vnode list. This makes a couple of assumptions, which I believe are ok, in regards to vnode stability when the mount list mutex is held. Improves looping case by 500%. (more optimization work is needed on top of these fixes) MFC after: 1 week
2001-10-26 00:08:05 +00:00
TAILQ_REMOVE(&mp->mnt_nvnodelist, vp, v_nmntvnodes);
TAILQ_INSERT_TAIL(&mp->mnt_nvnodelist, vp, v_nmntvnodes);
--count;
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
if (!VI_TRYLOCK(vp))
Back out the removal of LK_NOWAIT from the VOP_LOCK() call in vlrureclaim() in vfs_subr.c 1.636 because waiting for the vnode lock aggravates an existing race condition. It is also undesirable according to the commit log for 1.631. Fix the tiny race condition that remains by rechecking the vnode state after grabbing the vnode lock and grabbing the vnode interlock. Fix the problem of other threads being starved (which 1.636 attempted to fix by removing LK_NOWAIT) by calling uio_yield() periodically in vlrureclaim(). This should be more deterministic than hoping that VOP_LOCK() without LK_NOWAIT will block, which may not happen in this loop. Reviewed by: kan MFC after: 5 days
2005-08-23 03:44:06 +00:00
goto next_iter;
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
/*
* If it's been deconstructed already, it's still
* referenced, or it exceeds the trigger, skip it.
*/
Back out the removal of LK_NOWAIT from the VOP_LOCK() call in vlrureclaim() in vfs_subr.c 1.636 because waiting for the vnode lock aggravates an existing race condition. It is also undesirable according to the commit log for 1.631. Fix the tiny race condition that remains by rechecking the vnode state after grabbing the vnode lock and grabbing the vnode interlock. Fix the problem of other threads being starved (which 1.636 attempted to fix by removing LK_NOWAIT) by calling uio_yield() periodically in vlrureclaim(). This should be more deterministic than hoping that VOP_LOCK() without LK_NOWAIT will block, which may not happen in this loop. Reviewed by: kan MFC after: 5 days
2005-08-23 03:44:06 +00:00
if (vp->v_usecount || !LIST_EMPTY(&(vp)->v_cache_src) ||
(vp->v_iflag & VI_DOOMED) != 0 || (vp->v_object != NULL &&
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
vp->v_object->resident_page_count > trigger)) {
VI_UNLOCK(vp);
Back out the removal of LK_NOWAIT from the VOP_LOCK() call in vlrureclaim() in vfs_subr.c 1.636 because waiting for the vnode lock aggravates an existing race condition. It is also undesirable according to the commit log for 1.631. Fix the tiny race condition that remains by rechecking the vnode state after grabbing the vnode lock and grabbing the vnode interlock. Fix the problem of other threads being starved (which 1.636 attempted to fix by removing LK_NOWAIT) by calling uio_yield() periodically in vlrureclaim(). This should be more deterministic than hoping that VOP_LOCK() without LK_NOWAIT will block, which may not happen in this loop. Reviewed by: kan MFC after: 5 days
2005-08-23 03:44:06 +00:00
goto next_iter;
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
}
MNT_IUNLOCK(mp);
vholdl(vp);
VOP_LOCK1() (and so VOP_LOCK()) and VOP_UNLOCK() are only used in conjuction with 'thread' argument passing which is always curthread. Remove the unuseful extra-argument and pass explicitly curthread to lower layer functions, when necessary. KPI results broken by this change, which should affect several ports, so version bumping and manpage update will be further committed. Tested by: kris, pho, Diego Sardina <siarodx at gmail dot com>
2008-01-13 14:44:15 +00:00
if (VOP_LOCK(vp, LK_INTERLOCK|LK_EXCLUSIVE|LK_NOWAIT)) {
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
vdrop(vp);
Back out the removal of LK_NOWAIT from the VOP_LOCK() call in vlrureclaim() in vfs_subr.c 1.636 because waiting for the vnode lock aggravates an existing race condition. It is also undesirable according to the commit log for 1.631. Fix the tiny race condition that remains by rechecking the vnode state after grabbing the vnode lock and grabbing the vnode interlock. Fix the problem of other threads being starved (which 1.636 attempted to fix by removing LK_NOWAIT) by calling uio_yield() periodically in vlrureclaim(). This should be more deterministic than hoping that VOP_LOCK() without LK_NOWAIT will block, which may not happen in this loop. Reviewed by: kan MFC after: 5 days
2005-08-23 03:44:06 +00:00
goto next_iter_mntunlocked;
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
}
VI_LOCK(vp);
Back out the removal of LK_NOWAIT from the VOP_LOCK() call in vlrureclaim() in vfs_subr.c 1.636 because waiting for the vnode lock aggravates an existing race condition. It is also undesirable according to the commit log for 1.631. Fix the tiny race condition that remains by rechecking the vnode state after grabbing the vnode lock and grabbing the vnode interlock. Fix the problem of other threads being starved (which 1.636 attempted to fix by removing LK_NOWAIT) by calling uio_yield() periodically in vlrureclaim(). This should be more deterministic than hoping that VOP_LOCK() without LK_NOWAIT will block, which may not happen in this loop. Reviewed by: kan MFC after: 5 days
2005-08-23 03:44:06 +00:00
/*
* v_usecount may have been bumped after VOP_LOCK() dropped
* the vnode interlock and before it was locked again.
*
* It is not necessary to recheck VI_DOOMED because it can
* only be set by another thread that holds both the vnode
* lock and vnode interlock. If another thread has the
* vnode lock before we get to VOP_LOCK() and obtains the
* vnode interlock after VOP_LOCK() drops the vnode
* interlock, the other thread will be unable to drop the
* vnode lock before our VOP_LOCK() call fails.
*/
if (vp->v_usecount || !LIST_EMPTY(&(vp)->v_cache_src) ||
Minor style cleanups (mostly removal of trailing whitespaces).
2007-04-10 15:29:37 +00:00
(vp->v_object != NULL &&
Back out the removal of LK_NOWAIT from the VOP_LOCK() call in vlrureclaim() in vfs_subr.c 1.636 because waiting for the vnode lock aggravates an existing race condition. It is also undesirable according to the commit log for 1.631. Fix the tiny race condition that remains by rechecking the vnode state after grabbing the vnode lock and grabbing the vnode interlock. Fix the problem of other threads being starved (which 1.636 attempted to fix by removing LK_NOWAIT) by calling uio_yield() periodically in vlrureclaim(). This should be more deterministic than hoping that VOP_LOCK() without LK_NOWAIT will block, which may not happen in this loop. Reviewed by: kan MFC after: 5 days
2005-08-23 03:44:06 +00:00
vp->v_object->resident_page_count > trigger)) {
VOP_LOCK1() (and so VOP_LOCK()) and VOP_UNLOCK() are only used in conjuction with 'thread' argument passing which is always curthread. Remove the unuseful extra-argument and pass explicitly curthread to lower layer functions, when necessary. KPI results broken by this change, which should affect several ports, so version bumping and manpage update will be further committed. Tested by: kris, pho, Diego Sardina <siarodx at gmail dot com>
2008-01-13 14:44:15 +00:00
VOP_UNLOCK(vp, LK_INTERLOCK);
Back out the removal of LK_NOWAIT from the VOP_LOCK() call in vlrureclaim() in vfs_subr.c 1.636 because waiting for the vnode lock aggravates an existing race condition. It is also undesirable according to the commit log for 1.631. Fix the tiny race condition that remains by rechecking the vnode state after grabbing the vnode lock and grabbing the vnode interlock. Fix the problem of other threads being starved (which 1.636 attempted to fix by removing LK_NOWAIT) by calling uio_yield() periodically in vlrureclaim(). This should be more deterministic than hoping that VOP_LOCK() without LK_NOWAIT will block, which may not happen in this loop. Reviewed by: kan MFC after: 5 days
2005-08-23 03:44:06 +00:00
goto next_iter_mntunlocked;
}
KASSERT((vp->v_iflag & VI_DOOMED) == 0,
("VI_DOOMED unexpectedly detected in vlrureclaim()"));
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
vgonel(vp);
VOP_LOCK1() (and so VOP_LOCK()) and VOP_UNLOCK() are only used in conjuction with 'thread' argument passing which is always curthread. Remove the unuseful extra-argument and pass explicitly curthread to lower layer functions, when necessary. KPI results broken by this change, which should affect several ports, so version bumping and manpage update will be further committed. Tested by: kris, pho, Diego Sardina <siarodx at gmail dot com>
2008-01-13 14:44:15 +00:00
VOP_UNLOCK(vp, 0);
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
vdropl(vp);
done++;
Back out the removal of LK_NOWAIT from the VOP_LOCK() call in vlrureclaim() in vfs_subr.c 1.636 because waiting for the vnode lock aggravates an existing race condition. It is also undesirable according to the commit log for 1.631. Fix the tiny race condition that remains by rechecking the vnode state after grabbing the vnode lock and grabbing the vnode interlock. Fix the problem of other threads being starved (which 1.636 attempted to fix by removing LK_NOWAIT) by calling uio_yield() periodically in vlrureclaim(). This should be more deterministic than hoping that VOP_LOCK() without LK_NOWAIT will block, which may not happen in this loop. Reviewed by: kan MFC after: 5 days
2005-08-23 03:44:06 +00:00
next_iter_mntunlocked:
if ((count % 256) != 0)
goto relock_mnt;
goto yield;
next_iter:
if ((count % 256) != 0)
continue;
MNT_IUNLOCK(mp);
yield:
uio_yield();
relock_mnt:
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
MNT_ILOCK(mp);
Implement kern.maxvnodes. adjusting kern.maxvnodes now actually has a real effect. Optimize vfs_msync(). Avoid having to continually drop and re-obtain mutexes when scanning the vnode list. Improves looping case by 500%. Optimize ffs_sync(). Avoid having to continually drop and re-obtain mutexes when scanning the vnode list. This makes a couple of assumptions, which I believe are ok, in regards to vnode stability when the mount list mutex is held. Improves looping case by 500%. (more optimization work is needed on top of these fixes) MFC after: 1 week
2001-10-26 00:08:05 +00:00
}
Remove mntvnode_mtx and replace it with per-mountpoint mutex. Introduce two new macros MNT_ILOCK(mp)/MNT_IUNLOCK(mp) to operate on this mutex transparently. Eventually new mutex will be protecting more fields in struct mount, not only vnode list. Discussed with: jeff
2003-11-05 04:30:08 +00:00
MNT_IUNLOCK(mp);
- Remove vx_lock, vx_unlock, vx_wait, etc. - Add a vn_start_write/vn_finished_write around vlrureclaim so we don't do writing ops without suspending. This could suspend the vlruproc which should not be a problem under normal circumstances. - Manually implement VMIGHTFREE in vlrureclaim as this was the only instance where it was used. - Acquire a lock before calling vgone() as it now requires it. - Move the acquisition of the vnode interlock from vtryrecycle() to getnewvnode() so that if it fails we don't drop and reacquire the vnode_free_list_mtx. - Check for a usecount or holdcount at the end of vtryrecycle() in case someone grabbed a ref while we were recycling. Abort the recycle, and on the final ref drop this vnode will be placed on the head of the free list. - Move the redundant VOP_INACTIVE protection code into the local vinactive() routine to avoid code bloat. - Keep the vnode lock held across calls to vgone() in several places. - vgonel() no longer uses XLOCK, instead callers must hold an exclusive vnode lock. The VI_DOOMED flag is set to allow other threads to detect a vnode which is no longer valid. This flag is set until the last reference is gone, and there are no chances for a new ref. vgonel() holds this lock across the entire function, which greatly simplifies logic. _ Only vfree() in one place in vgone() not three. - Adjust vget() to check the VI_DOOMED flag prior to waiting on the lock in the LK_NOWAIT case. In other cases, check after we have slept and acquired an exlusive lock. This will simulate the old vx_wait() behavior. Sponsored by: Isilon Systems, Inc.
2005-03-13 11:54:28 +00:00
vn_finished_write(mp);
This is a forward port of Peter's vlrureclaim() fix, with some minor mods by me to make it more efficient. The original code had serious balancing problems and could also deadlock easily. This code relegates the vnode reclamation to its own kproc and relaxes the vnode reclamation requirements to better maintain kern.maxvnodes. This code still doesn't balance as well as it could, but it does a much better job then the original code. Approved by: re@freebsd.org Obtained from: ps, peter, dillon MFS Assuming: Assuming no problems crop up in Yahoo testing MFC after: 7 days
2001-12-18 20:48:54 +00:00
return done;
}
- Don't recycle vnodes anymore. Free them once they are dead. getnewvnode now always allocates a new vnode. - Define a new function, vnlru_free, which frees vnodes from the free list. It takes as a parameter the number of vnodes to free, which is wantfreevnodes - freevnodes when called from vnlru_proc or 1 when called from getnewvnode(). For now, getnewvnode() still tries to reclaim a free vnode before creating a new one when we are near the limit. - Define a function, vdestroy, which handles the actual release of memory and teardown of locks, etc. This could become a uma_dtor() routine. - Get rid of minvnodes. Now wantfreevnodes is 1/4th the max vnodes. This keeps more unreferenced vnodes around so that files which have only been stat'd are less likely to be kicked out of the system before we have a chance to read them, etc. These vnodes may still be freed via the normal vnlru_proc() routines which may some day become a real lru.
2005-03-25 05:34:39 +00:00
/*
* Attempt to keep the free list at wantfreevnodes length.
*/
static void
vnlru_free(int count)
{
struct vnode *vp;
- VFS_LOCK_GIANT when recycling a vnode via getnewvnode. We may be recycling for an unrelated filesystem. I really don't like potentially acquiring giant in the context of a giantless filesystem but there are reasonable objections to removing the recycling from this path. Sponsored by: Isilon Systems, Inc.
2006-04-04 06:46:10 +00:00
int vfslocked;
- Don't recycle vnodes anymore. Free them once they are dead. getnewvnode now always allocates a new vnode. - Define a new function, vnlru_free, which frees vnodes from the free list. It takes as a parameter the number of vnodes to free, which is wantfreevnodes - freevnodes when called from vnlru_proc or 1 when called from getnewvnode(). For now, getnewvnode() still tries to reclaim a free vnode before creating a new one when we are near the limit. - Define a function, vdestroy, which handles the actual release of memory and teardown of locks, etc. This could become a uma_dtor() routine. - Get rid of minvnodes. Now wantfreevnodes is 1/4th the max vnodes. This keeps more unreferenced vnodes around so that files which have only been stat'd are less likely to be kicked out of the system before we have a chance to read them, etc. These vnodes may still be freed via the normal vnlru_proc() routines which may some day become a real lru.
2005-03-25 05:34:39 +00:00
mtx_assert(&vnode_free_list_mtx, MA_OWNED);
for (; count > 0; count--) {
vp = TAILQ_FIRST(&vnode_free_list);
/*
* The list can be modified while the free_list_mtx
* has been dropped and vp could be NULL here.
*/
if (!vp)
break;
- Add KTR_VFS tracing to track the life of vnodes. Eventually KTR_VFS events could be added to cover other interesting details. - Add some VNASSERTs to discover places where we access vnodes after they have been uma_zfree'd before we try to free them again. - Add a few more VNASSERTs to vdestroy() to be certain that the vnode is really unused. Sponsored by: Isilon Systems, Inc.
2005-06-11 01:16:46 +00:00
VNASSERT(vp->v_op != NULL, vp,
("vnlru_free: vnode already reclaimed."));
- Don't recycle vnodes anymore. Free them once they are dead. getnewvnode now always allocates a new vnode. - Define a new function, vnlru_free, which frees vnodes from the free list. It takes as a parameter the number of vnodes to free, which is wantfreevnodes - freevnodes when called from vnlru_proc or 1 when called from getnewvnode(). For now, getnewvnode() still tries to reclaim a free vnode before creating a new one when we are near the limit. - Define a function, vdestroy, which handles the actual release of memory and teardown of locks, etc. This could become a uma_dtor() routine. - Get rid of minvnodes. Now wantfreevnodes is 1/4th the max vnodes. This keeps more unreferenced vnodes around so that files which have only been stat'd are less likely to be kicked out of the system before we have a chance to read them, etc. These vnodes may still be freed via the normal vnlru_proc() routines which may some day become a real lru.
2005-03-25 05:34:39 +00:00
TAILQ_REMOVE(&vnode_free_list, vp, v_freelist);
/*
* Don't recycle if we can't get the interlock.
*/
- In vnlru_free() remove the vnode from the free list before we call vtryrecycle(). We could sometimes get into situations where two threads could try to recycle the same vnode before this. - vtryrecycle() is now responsible for returning the vnode to the free list if it fails and someone else hasn't done it. - Make a new function vfreehead() which moves a vnode to the head of the free list and use it in vgone() to clean up that code a bit. Sponsored by: Isilon Systems, Inc. Reported by: pho, kkenn
2005-04-30 11:22:40 +00:00
if (!VI_TRYLOCK(vp)) {
TAILQ_INSERT_TAIL(&vnode_free_list, vp, v_freelist);
- Don't recycle vnodes anymore. Free them once they are dead. getnewvnode now always allocates a new vnode. - Define a new function, vnlru_free, which frees vnodes from the free list. It takes as a parameter the number of vnodes to free, which is wantfreevnodes - freevnodes when called from vnlru_proc or 1 when called from getnewvnode(). For now, getnewvnode() still tries to reclaim a free vnode before creating a new one when we are near the limit. - Define a function, vdestroy, which handles the actual release of memory and teardown of locks, etc. This could become a uma_dtor() routine. - Get rid of minvnodes. Now wantfreevnodes is 1/4th the max vnodes. This keeps more unreferenced vnodes around so that files which have only been stat'd are less likely to be kicked out of the system before we have a chance to read them, etc. These vnodes may still be freed via the normal vnlru_proc() routines which may some day become a real lru.
2005-03-25 05:34:39 +00:00
continue;
- In vnlru_free() remove the vnode from the free list before we call vtryrecycle(). We could sometimes get into situations where two threads could try to recycle the same vnode before this. - vtryrecycle() is now responsible for returning the vnode to the free list if it fails and someone else hasn't done it. - Make a new function vfreehead() which moves a vnode to the head of the free list and use it in vgone() to clean up that code a bit. Sponsored by: Isilon Systems, Inc. Reported by: pho, kkenn
2005-04-30 11:22:40 +00:00
}
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
VNASSERT(VCANRECYCLE(vp), vp,
("vp inconsistent on freelist"));
- Don't recycle vnodes anymore. Free them once they are dead. getnewvnode now always allocates a new vnode. - Define a new function, vnlru_free, which frees vnodes from the free list. It takes as a parameter the number of vnodes to free, which is wantfreevnodes - freevnodes when called from vnlru_proc or 1 when called from getnewvnode(). For now, getnewvnode() still tries to reclaim a free vnode before creating a new one when we are near the limit. - Define a function, vdestroy, which handles the actual release of memory and teardown of locks, etc. This could become a uma_dtor() routine. - Get rid of minvnodes. Now wantfreevnodes is 1/4th the max vnodes. This keeps more unreferenced vnodes around so that files which have only been stat'd are less likely to be kicked out of the system before we have a chance to read them, etc. These vnodes may still be freed via the normal vnlru_proc() routines which may some day become a real lru.
2005-03-25 05:34:39 +00:00
freevnodes--;
- In vnlru_free() remove the vnode from the free list before we call vtryrecycle(). We could sometimes get into situations where two threads could try to recycle the same vnode before this. - vtryrecycle() is now responsible for returning the vnode to the free list if it fails and someone else hasn't done it. - Make a new function vfreehead() which moves a vnode to the head of the free list and use it in vgone() to clean up that code a bit. Sponsored by: Isilon Systems, Inc. Reported by: pho, kkenn
2005-04-30 11:22:40 +00:00
vp->v_iflag &= ~VI_FREE;
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
vholdl(vp);
- Don't recycle vnodes anymore. Free them once they are dead. getnewvnode now always allocates a new vnode. - Define a new function, vnlru_free, which frees vnodes from the free list. It takes as a parameter the number of vnodes to free, which is wantfreevnodes - freevnodes when called from vnlru_proc or 1 when called from getnewvnode(). For now, getnewvnode() still tries to reclaim a free vnode before creating a new one when we are near the limit. - Define a function, vdestroy, which handles the actual release of memory and teardown of locks, etc. This could become a uma_dtor() routine. - Get rid of minvnodes. Now wantfreevnodes is 1/4th the max vnodes. This keeps more unreferenced vnodes around so that files which have only been stat'd are less likely to be kicked out of the system before we have a chance to read them, etc. These vnodes may still be freed via the normal vnlru_proc() routines which may some day become a real lru.
2005-03-25 05:34:39 +00:00
mtx_unlock(&vnode_free_list_mtx);
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
VI_UNLOCK(vp);
- VFS_LOCK_GIANT when recycling a vnode via getnewvnode. We may be recycling for an unrelated filesystem. I really don't like potentially acquiring giant in the context of a giantless filesystem but there are reasonable objections to removing the recycling from this path. Sponsored by: Isilon Systems, Inc.
2006-04-04 06:46:10 +00:00
vfslocked = VFS_LOCK_GIANT(vp->v_mount);
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
vtryrecycle(vp);
- VFS_LOCK_GIANT when recycling a vnode via getnewvnode. We may be recycling for an unrelated filesystem. I really don't like potentially acquiring giant in the context of a giantless filesystem but there are reasonable objections to removing the recycling from this path. Sponsored by: Isilon Systems, Inc.
2006-04-04 06:46:10 +00:00
VFS_UNLOCK_GIANT(vfslocked);
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
/*
* If the recycled succeeded this vdrop will actually free
* the vnode. If not it will simply place it back on
* the free list.
*/
vdrop(vp);
- Don't recycle vnodes anymore. Free them once they are dead. getnewvnode now always allocates a new vnode. - Define a new function, vnlru_free, which frees vnodes from the free list. It takes as a parameter the number of vnodes to free, which is wantfreevnodes - freevnodes when called from vnlru_proc or 1 when called from getnewvnode(). For now, getnewvnode() still tries to reclaim a free vnode before creating a new one when we are near the limit. - Define a function, vdestroy, which handles the actual release of memory and teardown of locks, etc. This could become a uma_dtor() routine. - Get rid of minvnodes. Now wantfreevnodes is 1/4th the max vnodes. This keeps more unreferenced vnodes around so that files which have only been stat'd are less likely to be kicked out of the system before we have a chance to read them, etc. These vnodes may still be freed via the normal vnlru_proc() routines which may some day become a real lru.
2005-03-25 05:34:39 +00:00
mtx_lock(&vnode_free_list_mtx);
}
}
This is a forward port of Peter's vlrureclaim() fix, with some minor mods by me to make it more efficient. The original code had serious balancing problems and could also deadlock easily. This code relegates the vnode reclamation to its own kproc and relaxes the vnode reclamation requirements to better maintain kern.maxvnodes. This code still doesn't balance as well as it could, but it does a much better job then the original code. Approved by: re@freebsd.org Obtained from: ps, peter, dillon MFS Assuming: Assuming no problems crop up in Yahoo testing MFC after: 7 days
2001-12-18 20:48:54 +00:00
/*
* Attempt to recycle vnodes in a context that is always safe to block.
More s/file system/filesystem/g
2002-05-16 21:28:32 +00:00
* Calling vlrurecycle() from the bowels of filesystem code has some
This is a forward port of Peter's vlrureclaim() fix, with some minor mods by me to make it more efficient. The original code had serious balancing problems and could also deadlock easily. This code relegates the vnode reclamation to its own kproc and relaxes the vnode reclamation requirements to better maintain kern.maxvnodes. This code still doesn't balance as well as it could, but it does a much better job then the original code. Approved by: re@freebsd.org Obtained from: ps, peter, dillon MFS Assuming: Assuming no problems crop up in Yahoo testing MFC after: 7 days
2001-12-18 20:48:54 +00:00
* interesting deadlock problems.
*/
static struct proc *vnlruproc;
static int vnlruproc_sig;
Gratuitous whitespace cleanup.
2002-06-06 15:46:38 +00:00
static void
This is a forward port of Peter's vlrureclaim() fix, with some minor mods by me to make it more efficient. The original code had serious balancing problems and could also deadlock easily. This code relegates the vnode reclamation to its own kproc and relaxes the vnode reclamation requirements to better maintain kern.maxvnodes. This code still doesn't balance as well as it could, but it does a much better job then the original code. Approved by: re@freebsd.org Obtained from: ps, peter, dillon MFS Assuming: Assuming no problems crop up in Yahoo testing MFC after: 7 days
2001-12-18 20:48:54 +00:00
vnlru_proc(void)
{
struct mount *mp, *nmp;
Push down Giant in the vlnru kproc main loop so that it is only acquired around calls to vlrureclaim() on non-MPSAFE filesystems. Specifically, vnlru no longer needs Giant for the common case of waking up and deciding there is nothing for it to do. MFC after: 2 weeks
2009-01-23 22:08:54 +00:00
int done, vfslocked;
This is a forward port of Peter's vlrureclaim() fix, with some minor mods by me to make it more efficient. The original code had serious balancing problems and could also deadlock easily. This code relegates the vnode reclamation to its own kproc and relaxes the vnode reclamation requirements to better maintain kern.maxvnodes. This code still doesn't balance as well as it could, but it does a much better job then the original code. Approved by: re@freebsd.org Obtained from: ps, peter, dillon MFS Assuming: Assuming no problems crop up in Yahoo testing MFC after: 7 days
2001-12-18 20:48:54 +00:00
struct proc *p = vnlruproc;
EVENTHANDLER_REGISTER(shutdown_pre_sync, kproc_shutdown, p,
Gratuitous whitespace cleanup.
2002-06-06 15:46:38 +00:00
SHUTDOWN_PRI_FIRST);
This is a forward port of Peter's vlrureclaim() fix, with some minor mods by me to make it more efficient. The original code had serious balancing problems and could also deadlock easily. This code relegates the vnode reclamation to its own kproc and relaxes the vnode reclamation requirements to better maintain kern.maxvnodes. This code still doesn't balance as well as it could, but it does a much better job then the original code. Approved by: re@freebsd.org Obtained from: ps, peter, dillon MFS Assuming: Assuming no problems crop up in Yahoo testing MFC after: 7 days
2001-12-18 20:48:54 +00:00
for (;;) {
Rename the kthread_xxx (e.g. kthread_create()) calls to kproc_xxx as they actually make whole processes. Thos makes way for us to add REAL kthread_create() and friends that actually make theads. it turns out that most of these calls actually end up being moved back to the thread version when it's added. but we need to make this cosmetic change first. I'd LOVE to do this rename in 7.0 so that we can eventually MFC the new kthread_xxx() calls.
2007-10-20 23:23:23 +00:00
kproc_suspend_check(p);
- Extend the vnode_free_list_mtx to cover numvnodes and freevnodes. This was done only some of the time before, and now it is uniformly applied.
2002-08-13 05:29:48 +00:00
mtx_lock(&vnode_free_list_mtx);
- Don't recycle vnodes anymore. Free them once they are dead. getnewvnode now always allocates a new vnode. - Define a new function, vnlru_free, which frees vnodes from the free list. It takes as a parameter the number of vnodes to free, which is wantfreevnodes - freevnodes when called from vnlru_proc or 1 when called from getnewvnode(). For now, getnewvnode() still tries to reclaim a free vnode before creating a new one when we are near the limit. - Define a function, vdestroy, which handles the actual release of memory and teardown of locks, etc. This could become a uma_dtor() routine. - Get rid of minvnodes. Now wantfreevnodes is 1/4th the max vnodes. This keeps more unreferenced vnodes around so that files which have only been stat'd are less likely to be kicked out of the system before we have a chance to read them, etc. These vnodes may still be freed via the normal vnlru_proc() routines which may some day become a real lru.
2005-03-25 05:34:39 +00:00
if (freevnodes > wantfreevnodes)
vnlru_free(freevnodes - wantfreevnodes);
if (numvnodes <= desiredvnodes * 9 / 10) {
This is a forward port of Peter's vlrureclaim() fix, with some minor mods by me to make it more efficient. The original code had serious balancing problems and could also deadlock easily. This code relegates the vnode reclamation to its own kproc and relaxes the vnode reclamation requirements to better maintain kern.maxvnodes. This code still doesn't balance as well as it could, but it does a much better job then the original code. Approved by: re@freebsd.org Obtained from: ps, peter, dillon MFS Assuming: Assuming no problems crop up in Yahoo testing MFC after: 7 days
2001-12-18 20:48:54 +00:00
vnlruproc_sig = 0;
Vnodes pull in 800-900 bytes these days, all things counted, so we need to treat desiredvnodes much more like a limit than as a vague concept. On a 2GB RAM machine where desired vnodes is 130k, we run out of kmem_map space when we hit about 190k vnodes. If we wake up the vnode washer in getnewvnode(), sleep until it is done, so that it has a chance to offer us a washed vnode. If we don't sleep here we'll just race ahead and allocate yet a vnode which will never get freed. In the vnodewasher, instead of doing 10 vnodes per mountpoint per rotation, do 10% of the vnodes distributed evenly across the mountpoints.
2002-12-29 10:39:05 +00:00
wakeup(&vnlruproc_sig);
- Add the tunable and sysctl for the mpsafevfs. It currently defaults to off. - Protect access to mnt_kern_flag with the mointpoint mutex. - Remove some KASSERTs which are not legal checks without the appropriate locks held. - Use VCANRECYCLE() rather than rolling several slightly different checks together. - Return from vtryrecycle() with a recycled vnode rather than a locked vnode. This simplifies some locking. - Remove several GIANT_REQUIRED lines. - Add a few KASSERTs to help with INACT debugging. Sponsored By: Isilon Systems, Inc.
2005-01-24 10:41:01 +00:00
msleep(vnlruproc, &vnode_free_list_mtx,
PVFS|PDROP, "vlruwt", hz);
This is a forward port of Peter's vlrureclaim() fix, with some minor mods by me to make it more efficient. The original code had serious balancing problems and could also deadlock easily. This code relegates the vnode reclamation to its own kproc and relaxes the vnode reclamation requirements to better maintain kern.maxvnodes. This code still doesn't balance as well as it could, but it does a much better job then the original code. Approved by: re@freebsd.org Obtained from: ps, peter, dillon MFS Assuming: Assuming no problems crop up in Yahoo testing MFC after: 7 days
2001-12-18 20:48:54 +00:00
continue;
}
- Extend the vnode_free_list_mtx to cover numvnodes and freevnodes. This was done only some of the time before, and now it is uniformly applied.
2002-08-13 05:29:48 +00:00
mtx_unlock(&vnode_free_list_mtx);
This is a forward port of Peter's vlrureclaim() fix, with some minor mods by me to make it more efficient. The original code had serious balancing problems and could also deadlock easily. This code relegates the vnode reclamation to its own kproc and relaxes the vnode reclamation requirements to better maintain kern.maxvnodes. This code still doesn't balance as well as it could, but it does a much better job then the original code. Approved by: re@freebsd.org Obtained from: ps, peter, dillon MFS Assuming: Assuming no problems crop up in Yahoo testing MFC after: 7 days
2001-12-18 20:48:54 +00:00
done = 0;
mtx_lock(&mountlist_mtx);
for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
Improve VFS locking: - Implement real draining for vfs consumers by not relying on the mnt_lock and using instead a refcount in order to keep track of lock requesters. - Due to the change above, remove the mnt_lock lockmgr because it is now useless. - Due to the change above, vfs_busy() is no more linked to a lockmgr. Change so its KPI by removing the interlock argument and defining 2 new flags for it: MBF_NOWAIT which basically replaces the LK_NOWAIT of the old version (which was unlinked from the lockmgr alredy) and MBF_MNTLSTLOCK which provides the ability to drop the mountlist_mtx once the mnt interlock is held (ability still desired by most consumers). - The stub used into vfs_mount_destroy(), that allows to override the mnt_ref if running for more than 3 seconds, make it totally useless. Remove it as it was thought to work into older versions. If a problem of "refcount held never going away" should appear, we will need to fix properly instead than trust on such hackish solution. - Fix a bug where returning (with an error) from dounmount() was still leaving the MNTK_MWAIT flag on even if it the waiters were actually woken up. Just a place in vfs_mount_destroy() is left because it is going to recycle the structure in any case, so it doesn't matter. - Remove the markercnt refcount as it is useless. This patch modifies VFS ABI and breaks KPI for vfs_busy() so manpages and __FreeBSD_version will be modified accordingly. Discussed with: kib Tested by: pho
2008-11-02 10:15:42 +00:00
if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
This is a forward port of Peter's vlrureclaim() fix, with some minor mods by me to make it more efficient. The original code had serious balancing problems and could also deadlock easily. This code relegates the vnode reclamation to its own kproc and relaxes the vnode reclamation requirements to better maintain kern.maxvnodes. This code still doesn't balance as well as it could, but it does a much better job then the original code. Approved by: re@freebsd.org Obtained from: ps, peter, dillon MFS Assuming: Assuming no problems crop up in Yahoo testing MFC after: 7 days
2001-12-18 20:48:54 +00:00
nmp = TAILQ_NEXT(mp, mnt_list);
continue;
}
Push down Giant in the vlnru kproc main loop so that it is only acquired around calls to vlrureclaim() on non-MPSAFE filesystems. Specifically, vnlru no longer needs Giant for the common case of waking up and deciding there is nothing for it to do. MFC after: 2 weeks
2009-01-23 22:08:54 +00:00
vfslocked = VFS_LOCK_GIANT(mp);
Adjust the number of vnodes scanned by vlrureclaim() according to the size of the vnode list.
2003-03-26 22:15:58 +00:00
done += vlrureclaim(mp);
Push down Giant in the vlnru kproc main loop so that it is only acquired around calls to vlrureclaim() on non-MPSAFE filesystems. Specifically, vnlru no longer needs Giant for the common case of waking up and deciding there is nothing for it to do. MFC after: 2 weeks
2009-01-23 22:08:54 +00:00
VFS_UNLOCK_GIANT(vfslocked);
This is a forward port of Peter's vlrureclaim() fix, with some minor mods by me to make it more efficient. The original code had serious balancing problems and could also deadlock easily. This code relegates the vnode reclamation to its own kproc and relaxes the vnode reclamation requirements to better maintain kern.maxvnodes. This code still doesn't balance as well as it could, but it does a much better job then the original code. Approved by: re@freebsd.org Obtained from: ps, peter, dillon MFS Assuming: Assuming no problems crop up in Yahoo testing MFC after: 7 days
2001-12-18 20:48:54 +00:00
mtx_lock(&mountlist_mtx);
nmp = TAILQ_NEXT(mp, mnt_list);
Decontextualize vfs_busy(), vfs_unbusy() and vfs_mount_alloc() functions. Manpages are updated accordingly. Tested by: Diego Sardina <siarodx at gmail dot com>
2008-08-31 14:26:08 +00:00
vfs_unbusy(mp);
This is a forward port of Peter's vlrureclaim() fix, with some minor mods by me to make it more efficient. The original code had serious balancing problems and could also deadlock easily. This code relegates the vnode reclamation to its own kproc and relaxes the vnode reclamation requirements to better maintain kern.maxvnodes. This code still doesn't balance as well as it could, but it does a much better job then the original code. Approved by: re@freebsd.org Obtained from: ps, peter, dillon MFS Assuming: Assuming no problems crop up in Yahoo testing MFC after: 7 days
2001-12-18 20:48:54 +00:00
}
mtx_unlock(&mountlist_mtx);
if (done == 0) {
When we are running low on vnodes, there is currently no way to ask other subsystems to release some vnodes. Implement backpressure based on vfs_lowvnodes event (similar to vm_lowmem for memory).
2007-04-13 08:38:48 +00:00
EVENTHANDLER_INVOKE(vfs_lowvnodes, desiredvnodes / 10);
Use a different mechanism to get the vnlru process to wake up and notice the shutdown request at reboot/halt time. Disable the printf 'vnlru process getting nowhere, pausing...' and instead export the count to the debug.vnlru_nowhere sysctl.
2001-12-19 01:31:12 +00:00
#if 0
/* These messages are temporary debugging aids */
if (vnlru_nowhere < 5)
printf("vnlru process getting nowhere..\n");
else if (vnlru_nowhere == 5)
printf("vnlru process messages stopped.\n");
#endif
vnlru_nowhere++;
Fix type-o in previous commit (tsleep was using wrong rendezvous point)
2001-12-25 01:23:25 +00:00
tsleep(vnlruproc, PPAUSE, "vlrup", hz * 3);
Minor style cleanups (mostly removal of trailing whitespaces).
2007-04-10 15:29:37 +00:00
} else
- Allow vnlru to drop giant if the filesystem does not require it. The vnlru proc is extremely inefficient, potentially iteration over tens of thousands of vnodes without blocking. Droping Giant allows other threads to preempt us although we should revisit the algorithm to fix the runtime problems especially since this may hold up all vnode allocations. - Remove the LK_NOWAIT from the VOP_LOCK in vlrureclaim. This provides a natural blocking point to help alleviate the situation described above although it may not technically be desirable. - yield after we make a pass on all mount points to prevent us from blocking other threads which require Giant. MFC after: 2 weeks
2005-07-20 01:43:27 +00:00
uio_yield();
This is a forward port of Peter's vlrureclaim() fix, with some minor mods by me to make it more efficient. The original code had serious balancing problems and could also deadlock easily. This code relegates the vnode reclamation to its own kproc and relaxes the vnode reclamation requirements to better maintain kern.maxvnodes. This code still doesn't balance as well as it could, but it does a much better job then the original code. Approved by: re@freebsd.org Obtained from: ps, peter, dillon MFS Assuming: Assuming no problems crop up in Yahoo testing MFC after: 7 days
2001-12-18 20:48:54 +00:00
}
Implement kern.maxvnodes. adjusting kern.maxvnodes now actually has a real effect. Optimize vfs_msync(). Avoid having to continually drop and re-obtain mutexes when scanning the vnode list. Improves looping case by 500%. Optimize ffs_sync(). Avoid having to continually drop and re-obtain mutexes when scanning the vnode list. This makes a couple of assumptions, which I believe are ok, in regards to vnode stability when the mount list mutex is held. Improves looping case by 500%. (more optimization work is needed on top of these fixes) MFC after: 1 week
2001-10-26 00:08:05 +00:00
}
This is a forward port of Peter's vlrureclaim() fix, with some minor mods by me to make it more efficient. The original code had serious balancing problems and could also deadlock easily. This code relegates the vnode reclamation to its own kproc and relaxes the vnode reclamation requirements to better maintain kern.maxvnodes. This code still doesn't balance as well as it could, but it does a much better job then the original code. Approved by: re@freebsd.org Obtained from: ps, peter, dillon MFS Assuming: Assuming no problems crop up in Yahoo testing MFC after: 7 days
2001-12-18 20:48:54 +00:00
static struct kproc_desc vnlru_kp = {
"vnlru",
vnlru_proc,
&vnlruproc
};
In keeping with style(9)'s recommendations on macros, use a ';' after each SYSINIT() macro invocation. This makes a number of lightweight C parsers much happier with the FreeBSD kernel source, including cflow's prcc and lxr. MFC after: 1 month Discussed with: imp, rink
2008-03-16 10:58:09 +00:00
SYSINIT(vnlru, SI_SUB_KTHREAD_UPDATE, SI_ORDER_FIRST, kproc_start,
&vnlru_kp);
This is a forward port of Peter's vlrureclaim() fix, with some minor mods by me to make it more efficient. The original code had serious balancing problems and could also deadlock easily. This code relegates the vnode reclamation to its own kproc and relaxes the vnode reclamation requirements to better maintain kern.maxvnodes. This code still doesn't balance as well as it could, but it does a much better job then the original code. Approved by: re@freebsd.org Obtained from: ps, peter, dillon MFS Assuming: Assuming no problems crop up in Yahoo testing MFC after: 7 days
2001-12-18 20:48:54 +00:00
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
/*
* Routines having to do with the management of the vnode table.
*/
Update ZFS from version 6 to 13 and bring some FreeBSD-specific changes. This bring huge amount of changes, I'll enumerate only user-visible changes: - Delegated Administration Allows regular users to perform ZFS operations, like file system creation, snapshot creation, etc. - L2ARC Level 2 cache for ZFS - allows to use additional disks for cache. Huge performance improvements mostly for random read of mostly static content. - slog Allow to use additional disks for ZFS Intent Log to speed up operations like fsync(2). - vfs.zfs.super_owner Allows regular users to perform privileged operations on files stored on ZFS file systems owned by him. Very careful with this one. - chflags(2) Not all the flags are supported. This still needs work. - ZFSBoot Support to boot off of ZFS pool. Not finished, AFAIK. Submitted by: dfr - Snapshot properties - New failure modes Before if write requested failed, system paniced. Now one can select from one of three failure modes: - panic - panic on write error - wait - wait for disk to reappear - continue - serve read requests if possible, block write requests - Refquota, refreservation properties Just quota and reservation properties, but don't count space consumed by children file systems, clones and snapshots. - Sparse volumes ZVOLs that don't reserve space in the pool. - External attributes Compatible with extattr(2). - NFSv4-ACLs Not sure about the status, might not be complete yet. Submitted by: trasz - Creation-time properties - Regression tests for zpool(8) command. Obtained from: OpenSolaris
2008-11-17 20:49:29 +00:00
void
- Don't recycle vnodes anymore. Free them once they are dead. getnewvnode now always allocates a new vnode. - Define a new function, vnlru_free, which frees vnodes from the free list. It takes as a parameter the number of vnodes to free, which is wantfreevnodes - freevnodes when called from vnlru_proc or 1 when called from getnewvnode(). For now, getnewvnode() still tries to reclaim a free vnode before creating a new one when we are near the limit. - Define a function, vdestroy, which handles the actual release of memory and teardown of locks, etc. This could become a uma_dtor() routine. - Get rid of minvnodes. Now wantfreevnodes is 1/4th the max vnodes. This keeps more unreferenced vnodes around so that files which have only been stat'd are less likely to be kicked out of the system before we have a chance to read them, etc. These vnodes may still be freed via the normal vnlru_proc() routines which may some day become a real lru.
2005-03-25 05:34:39 +00:00
vdestroy(struct vnode *vp)
{
struct bufobj *bo;
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
CTR2(KTR_VFS, "%s: vp %p", __func__, vp);
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
mtx_lock(&vnode_free_list_mtx);
numvnodes--;
mtx_unlock(&vnode_free_list_mtx);
- Don't recycle vnodes anymore. Free them once they are dead. getnewvnode now always allocates a new vnode. - Define a new function, vnlru_free, which frees vnodes from the free list. It takes as a parameter the number of vnodes to free, which is wantfreevnodes - freevnodes when called from vnlru_proc or 1 when called from getnewvnode(). For now, getnewvnode() still tries to reclaim a free vnode before creating a new one when we are near the limit. - Define a function, vdestroy, which handles the actual release of memory and teardown of locks, etc. This could become a uma_dtor() routine. - Get rid of minvnodes. Now wantfreevnodes is 1/4th the max vnodes. This keeps more unreferenced vnodes around so that files which have only been stat'd are less likely to be kicked out of the system before we have a chance to read them, etc. These vnodes may still be freed via the normal vnlru_proc() routines which may some day become a real lru.
2005-03-25 05:34:39 +00:00
bo = &vp->v_bufobj;
- Add KTR_VFS tracing to track the life of vnodes. Eventually KTR_VFS events could be added to cover other interesting details. - Add some VNASSERTs to discover places where we access vnodes after they have been uma_zfree'd before we try to free them again. - Add a few more VNASSERTs to vdestroy() to be certain that the vnode is really unused. Sponsored by: Isilon Systems, Inc.
2005-06-11 01:16:46 +00:00
VNASSERT((vp->v_iflag & VI_FREE) == 0, vp,
("cleaned vnode still on the free list."));
- Don't recycle vnodes anymore. Free them once they are dead. getnewvnode now always allocates a new vnode. - Define a new function, vnlru_free, which frees vnodes from the free list. It takes as a parameter the number of vnodes to free, which is wantfreevnodes - freevnodes when called from vnlru_proc or 1 when called from getnewvnode(). For now, getnewvnode() still tries to reclaim a free vnode before creating a new one when we are near the limit. - Define a function, vdestroy, which handles the actual release of memory and teardown of locks, etc. This could become a uma_dtor() routine. - Get rid of minvnodes. Now wantfreevnodes is 1/4th the max vnodes. This keeps more unreferenced vnodes around so that files which have only been stat'd are less likely to be kicked out of the system before we have a chance to read them, etc. These vnodes may still be freed via the normal vnlru_proc() routines which may some day become a real lru.
2005-03-25 05:34:39 +00:00
VNASSERT(vp->v_data == NULL, vp, ("cleaned vnode isn't"));
- Add KTR_VFS tracing to track the life of vnodes. Eventually KTR_VFS events could be added to cover other interesting details. - Add some VNASSERTs to discover places where we access vnodes after they have been uma_zfree'd before we try to free them again. - Add a few more VNASSERTs to vdestroy() to be certain that the vnode is really unused. Sponsored by: Isilon Systems, Inc.
2005-06-11 01:16:46 +00:00
VNASSERT(vp->v_holdcnt == 0, vp, ("Non-zero hold count"));
- Don't recycle vnodes anymore. Free them once they are dead. getnewvnode now always allocates a new vnode. - Define a new function, vnlru_free, which frees vnodes from the free list. It takes as a parameter the number of vnodes to free, which is wantfreevnodes - freevnodes when called from vnlru_proc or 1 when called from getnewvnode(). For now, getnewvnode() still tries to reclaim a free vnode before creating a new one when we are near the limit. - Define a function, vdestroy, which handles the actual release of memory and teardown of locks, etc. This could become a uma_dtor() routine. - Get rid of minvnodes. Now wantfreevnodes is 1/4th the max vnodes. This keeps more unreferenced vnodes around so that files which have only been stat'd are less likely to be kicked out of the system before we have a chance to read them, etc. These vnodes may still be freed via the normal vnlru_proc() routines which may some day become a real lru.
2005-03-25 05:34:39 +00:00
VNASSERT(vp->v_usecount == 0, vp, ("Non-zero use count"));
VNASSERT(vp->v_writecount == 0, vp, ("Non-zero write count"));
VNASSERT(bo->bo_numoutput == 0, vp, ("Clean vnode has pending I/O's"));
VNASSERT(bo->bo_clean.bv_cnt == 0, vp, ("cleanbufcnt not 0"));
VNASSERT(bo->bo_clean.bv_root == NULL, vp, ("cleanblkroot not NULL"));
VNASSERT(bo->bo_dirty.bv_cnt == 0, vp, ("dirtybufcnt not 0"));
VNASSERT(bo->bo_dirty.bv_root == NULL, vp, ("dirtyblkroot not NULL"));
- Assert that we're not in the name cache anymore in vdestroy(). Sponsored by: Isilon Systems, Inc.
2005-06-11 08:48:09 +00:00
VNASSERT(TAILQ_EMPTY(&vp->v_cache_dst), vp, ("vp has namecache dst"));
VNASSERT(LIST_EMPTY(&vp->v_cache_src), vp, ("vp has namecache src"));
- Unlock before we call mac_destroy_vnode to prevent a lock order reversal. Found by: trhodes
2005-08-03 05:36:50 +00:00
VI_UNLOCK(vp);
Back out my rev. 1.674. The better fix (rev. 1.637) is already in tree. Approved by: kan (mentor)
2006-07-05 16:33:25 +00:00
#ifdef MAC
Merge first in a series of TrustedBSD MAC Framework KPI changes from Mac OS X Leopard--rationalize naming for entry points to the following general forms: mac_<object>_<method/action> mac_<object>_check_<method/action> The previous naming scheme was inconsistent and mostly reversed from the new scheme. Also, make object types more consistent and remove spaces from object types that contain multiple parts ("posix_sem" -> "posixsem") to make mechanical parsing easier. Introduce a new "netinet" object type for certain IPv4/IPv6-related methods. Also simplify, slightly, some entry point names. All MAC policy modules will need to be recompiled, and modules not updates as part of this commit will need to be modified to conform to the new KPI. Sponsored by: SPARTA (original patches against Mac OS X) Obtained from: TrustedBSD Project, Apple Computer
2007-10-24 19:04:04 +00:00
mac_vnode_destroy(vp);
Back out my rev. 1.674. The better fix (rev. 1.637) is already in tree. Approved by: kan (mentor)
2006-07-05 16:33:25 +00:00
#endif
Protect check for v_pollinfo == NULL and assignment of the newly allocated vpollinfo with vnode interlock. Fully initialize vpollinfo before putting pointer to it into vp->v_pollinfo. Discussed with: dwhite Tested by: pho MFC after: 1 week
2008-10-28 12:08:36 +00:00
if (vp->v_pollinfo != NULL)
destroy_vpollinfo(vp->v_pollinfo);
- Add KTR_VFS tracing to track the life of vnodes. Eventually KTR_VFS events could be added to cover other interesting details. - Add some VNASSERTs to discover places where we access vnodes after they have been uma_zfree'd before we try to free them again. - Add a few more VNASSERTs to vdestroy() to be certain that the vnode is really unused. Sponsored by: Isilon Systems, Inc.
2005-06-11 01:16:46 +00:00
#ifdef INVARIANTS
/* XXX Elsewhere we can detect an already freed vnode via NULL v_op. */
vp->v_op = NULL;
#endif
- Don't recycle vnodes anymore. Free them once they are dead. getnewvnode now always allocates a new vnode. - Define a new function, vnlru_free, which frees vnodes from the free list. It takes as a parameter the number of vnodes to free, which is wantfreevnodes - freevnodes when called from vnlru_proc or 1 when called from getnewvnode(). For now, getnewvnode() still tries to reclaim a free vnode before creating a new one when we are near the limit. - Define a function, vdestroy, which handles the actual release of memory and teardown of locks, etc. This could become a uma_dtor() routine. - Get rid of minvnodes. Now wantfreevnodes is 1/4th the max vnodes. This keeps more unreferenced vnodes around so that files which have only been stat'd are less likely to be kicked out of the system before we have a chance to read them, etc. These vnodes may still be freed via the normal vnlru_proc() routines which may some day become a real lru.
2005-03-25 05:34:39 +00:00
lockdestroy(vp->v_vnlock);
mtx_destroy(&vp->v_interlock);
- Destroy the bo mtx when the vnode is destroyed.
2008-04-02 10:40:03 +00:00
mtx_destroy(BO_MTX(bo));
- Don't recycle vnodes anymore. Free them once they are dead. getnewvnode now always allocates a new vnode. - Define a new function, vnlru_free, which frees vnodes from the free list. It takes as a parameter the number of vnodes to free, which is wantfreevnodes - freevnodes when called from vnlru_proc or 1 when called from getnewvnode(). For now, getnewvnode() still tries to reclaim a free vnode before creating a new one when we are near the limit. - Define a function, vdestroy, which handles the actual release of memory and teardown of locks, etc. This could become a uma_dtor() routine. - Get rid of minvnodes. Now wantfreevnodes is 1/4th the max vnodes. This keeps more unreferenced vnodes around so that files which have only been stat'd are less likely to be kicked out of the system before we have a chance to read them, etc. These vnodes may still be freed via the normal vnlru_proc() routines which may some day become a real lru.
2005-03-25 05:34:39 +00:00
uma_zfree(vnode_zone, vp);
}
- Move some logic from getnewvnode() to a new function vcanrecycle() - Unlock the free list mutex around vcanrecycle to prevent a lock order reversal.
2002-08-05 10:15:56 +00:00
/*
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
* Try to recycle a freed vnode. We abort if anyone picks up a reference
* before we actually vgone(). This function must be called with the vnode
* held to prevent the vnode from being returned to the free list midway
* through vgone().
- Move some logic from getnewvnode() to a new function vcanrecycle() - Unlock the free list mutex around vcanrecycle to prevent a lock order reversal.
2002-08-05 10:15:56 +00:00
*/
static int
- Rename vcanrecycle() to vtryrecycle() to reflect its new role. - In vtryrecycle() try to vgonel the vnode if all of the previous checks passed. We won't vgonel if someone has either acquired a hold or usecount or started the vgone process elsewhere. This is because we may have been removed from the free list while we were inspecting the vnode for recycling. - The VI_TRYLOCK stops two threads from entering getnewvnode() and recycling the same vnode. To further reduce the likelyhood of this event, requeue the vnode on the tail of the list prior to calling vtryrecycle(). We can not actually remove the vnode from the list until we know that it's going to be recycled because other interlock holders may see the VI_FREE flag and try to remove it from the free list. - Kill a bogus XXX comment. If XLOCK is set we shouldn't wait for it regardless of MNT_WAIT because the vnode does not actually belong to this filesystem.
2003-10-05 05:35:41 +00:00
vtryrecycle(struct vnode *vp)
- Move some logic from getnewvnode() to a new function vcanrecycle() - Unlock the free list mutex around vcanrecycle to prevent a lock order reversal.
2002-08-05 10:15:56 +00:00
{
- Rename vcanrecycle() to vtryrecycle() to reflect its new role. - In vtryrecycle() try to vgonel the vnode if all of the previous checks passed. We won't vgonel if someone has either acquired a hold or usecount or started the vgone process elsewhere. This is because we may have been removed from the free list while we were inspecting the vnode for recycling. - The VI_TRYLOCK stops two threads from entering getnewvnode() and recycling the same vnode. To further reduce the likelyhood of this event, requeue the vnode on the tail of the list prior to calling vtryrecycle(). We can not actually remove the vnode from the list until we know that it's going to be recycled because other interlock holders may see the VI_FREE flag and try to remove it from the free list. - Kill a bogus XXX comment. If XLOCK is set we shouldn't wait for it regardless of MNT_WAIT because the vnode does not actually belong to this filesystem.
2003-10-05 05:35:41 +00:00
struct mount *vnmp;
- Move some logic from getnewvnode() to a new function vcanrecycle() - Unlock the free list mutex around vcanrecycle to prevent a lock order reversal.
2002-08-05 10:15:56 +00:00
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
CTR2(KTR_VFS, "%s: vp %p", __func__, vp);
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
VNASSERT(vp->v_holdcnt, vp,
("vtryrecycle: Recycling vp %p without a reference.", vp));
- In a Giantless world, the vn_lock() in vcanrecycle() could legitimately fail. Remove the panic from that case and document why it might fail. - Document the reason for calling cache_purge() on a newly created vnode. - In insmntque() order the operations so that we can call mtx_unlock() one fewer times. This makes the code somewhat clearer as well. - Add XXX comments in sched_sync() and vflush(). - In vget(), do not sleep while waiting for XLOCK to clear if LK_NOWAIT is set. - In vclean() we don't need to acquire a lock around a single TAILQ_FIRST call. It's ok if we race here, the vinvalbuf will just do nothing. - Increase the scope of the lock in vgonel() to reduce the number of lock operations that are performed.
2003-10-04 15:10:40 +00:00
/*
* This vnode may found and locked via some other list, if so we
* can't recycle it yet.
*/
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
if (VOP_LOCK(vp, LK_EXCLUSIVE | LK_NOWAIT) != 0) {
CTR2(KTR_VFS,
"%s: impossible to recycle, vp %p lock is already held",
__func__, vp);
- Move some logic from getnewvnode() to a new function vcanrecycle() - Unlock the free list mutex around vcanrecycle to prevent a lock order reversal.
2002-08-05 10:15:56 +00:00
return (EWOULDBLOCK);
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
}
When considering a vnode for reuse in getnewvnode, we call vcanrecycle to check a free vnode's availability. If it is available, vcanrecycle returns an error code of zero and the vnode in question locked. The getnewvnode routine then used to call vn_start_write with the V_NOWAIT flag. If the filesystem was suspended while taking a snapshot, the vn_start_write would fail but getnewvnode would fail to unlock the vnode, instead leaving it locked on the freelist. The result would be that the vnode would be locked forever and would eventually hang the system with a race to the root when it was attempted to recycle it. This fix moves the vn_start_write check into vcanrecycle where it will properly unlock the vnode if it is unavailable for recycling due to filesystem suspension. Sponsored by: DARPA & NAI Labs.
2002-10-11 01:04:14 +00:00
/*
* Don't recycle if its filesystem is being suspended.
*/
- Rename vcanrecycle() to vtryrecycle() to reflect its new role. - In vtryrecycle() try to vgonel the vnode if all of the previous checks passed. We won't vgonel if someone has either acquired a hold or usecount or started the vgone process elsewhere. This is because we may have been removed from the free list while we were inspecting the vnode for recycling. - The VI_TRYLOCK stops two threads from entering getnewvnode() and recycling the same vnode. To further reduce the likelyhood of this event, requeue the vnode on the tail of the list prior to calling vtryrecycle(). We can not actually remove the vnode from the list until we know that it's going to be recycled because other interlock holders may see the VI_FREE flag and try to remove it from the free list. - Kill a bogus XXX comment. If XLOCK is set we shouldn't wait for it regardless of MNT_WAIT because the vnode does not actually belong to this filesystem.
2003-10-05 05:35:41 +00:00
if (vn_start_write(vp, &vnmp, V_NOWAIT) != 0) {
VOP_LOCK1() (and so VOP_LOCK()) and VOP_UNLOCK() are only used in conjuction with 'thread' argument passing which is always curthread. Remove the unuseful extra-argument and pass explicitly curthread to lower layer functions, when necessary. KPI results broken by this change, which should affect several ports, so version bumping and manpage update will be further committed. Tested by: kris, pho, Diego Sardina <siarodx at gmail dot com>
2008-01-13 14:44:15 +00:00
VOP_UNLOCK(vp, 0);
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
CTR2(KTR_VFS,
"%s: impossible to recycle, cannot start the write for %p",
__func__, vp);
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
return (EBUSY);
When considering a vnode for reuse in getnewvnode, we call vcanrecycle to check a free vnode's availability. If it is available, vcanrecycle returns an error code of zero and the vnode in question locked. The getnewvnode routine then used to call vn_start_write with the V_NOWAIT flag. If the filesystem was suspended while taking a snapshot, the vn_start_write would fail but getnewvnode would fail to unlock the vnode, instead leaving it locked on the freelist. The result would be that the vnode would be locked forever and would eventually hang the system with a race to the root when it was attempted to recycle it. This fix moves the vn_start_write check into vcanrecycle where it will properly unlock the vnode if it is unavailable for recycling due to filesystem suspension. Sponsored by: DARPA & NAI Labs.
2002-10-11 01:04:14 +00:00
}
- Rename vcanrecycle() to vtryrecycle() to reflect its new role. - In vtryrecycle() try to vgonel the vnode if all of the previous checks passed. We won't vgonel if someone has either acquired a hold or usecount or started the vgone process elsewhere. This is because we may have been removed from the free list while we were inspecting the vnode for recycling. - The VI_TRYLOCK stops two threads from entering getnewvnode() and recycling the same vnode. To further reduce the likelyhood of this event, requeue the vnode on the tail of the list prior to calling vtryrecycle(). We can not actually remove the vnode from the list until we know that it's going to be recycled because other interlock holders may see the VI_FREE flag and try to remove it from the free list. - Kill a bogus XXX comment. If XLOCK is set we shouldn't wait for it regardless of MNT_WAIT because the vnode does not actually belong to this filesystem.
2003-10-05 05:35:41 +00:00
/*
* If we got this far, we need to acquire the interlock and see if
* anyone picked up this vnode from another list. If not, we will
- Now that there are no external users of vfree() make it static. - Move VSHOULDBUSY, VSHOULDFREE, and VTRYRECYCLE into vfs_subr.c so no one else attempts to grow a dependency on them. - Now that objects with pages hold the vnode we don't have to do unlocked checks for the page count in the vm object in VSHOULDFREE. These three macros could simply check for holdcnt state transitions to determine whether the vnode is on the free list already, but the extra safety the flag affords us is probably worth the minimal cost. - The leafonly sysctl and code have been dead for several years now, remove the sysctl and the code that employed it from vtryrecycle(). - vtryrecycle() also no longer has to check the object's page count as the object holds the vnode until it reaches 0. Sponsored by: Isilon Systems, Inc.
2005-03-15 14:38:16 +00:00
* mark it with DOOMED via vgonel() so that anyone who does find it
- Rename vcanrecycle() to vtryrecycle() to reflect its new role. - In vtryrecycle() try to vgonel the vnode if all of the previous checks passed. We won't vgonel if someone has either acquired a hold or usecount or started the vgone process elsewhere. This is because we may have been removed from the free list while we were inspecting the vnode for recycling. - The VI_TRYLOCK stops two threads from entering getnewvnode() and recycling the same vnode. To further reduce the likelyhood of this event, requeue the vnode on the tail of the list prior to calling vtryrecycle(). We can not actually remove the vnode from the list until we know that it's going to be recycled because other interlock holders may see the VI_FREE flag and try to remove it from the free list. - Kill a bogus XXX comment. If XLOCK is set we shouldn't wait for it regardless of MNT_WAIT because the vnode does not actually belong to this filesystem.
2003-10-05 05:35:41 +00:00
* will skip over it.
*/
VI_LOCK(vp);
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
if (vp->v_usecount) {
VOP_LOCK1() (and so VOP_LOCK()) and VOP_UNLOCK() are only used in conjuction with 'thread' argument passing which is always curthread. Remove the unuseful extra-argument and pass explicitly curthread to lower layer functions, when necessary. KPI results broken by this change, which should affect several ports, so version bumping and manpage update will be further committed. Tested by: kris, pho, Diego Sardina <siarodx at gmail dot com>
2008-01-13 14:44:15 +00:00
VOP_UNLOCK(vp, LK_INTERLOCK);
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
vn_finished_write(vnmp);
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
CTR2(KTR_VFS,
"%s: impossible to recycle, %p is already referenced",
__func__, vp);
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
return (EBUSY);
- In vnlru_free() remove the vnode from the free list before we call vtryrecycle(). We could sometimes get into situations where two threads could try to recycle the same vnode before this. - vtryrecycle() is now responsible for returning the vnode to the free list if it fails and someone else hasn't done it. - Make a new function vfreehead() which moves a vnode to the head of the free list and use it in vgone() to clean up that code a bit. Sponsored by: Isilon Systems, Inc. Reported by: pho, kkenn
2005-04-30 11:22:40 +00:00
}
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
if ((vp->v_iflag & VI_DOOMED) == 0)
vgonel(vp);
VOP_LOCK1() (and so VOP_LOCK()) and VOP_UNLOCK() are only used in conjuction with 'thread' argument passing which is always curthread. Remove the unuseful extra-argument and pass explicitly curthread to lower layer functions, when necessary. KPI results broken by this change, which should affect several ports, so version bumping and manpage update will be further committed. Tested by: kris, pho, Diego Sardina <siarodx at gmail dot com>
2008-01-13 14:44:15 +00:00
VOP_UNLOCK(vp, LK_INTERLOCK);
Always call vn_finished_write after vn_start_write was called. All occurences of 'goto done' after vn_start_write invocation were cleaning up incompletely.
2004-03-06 04:09:54 +00:00
vn_finished_write(vnmp);
- In vnlru_free() remove the vnode from the free list before we call vtryrecycle(). We could sometimes get into situations where two threads could try to recycle the same vnode before this. - vtryrecycle() is now responsible for returning the vnode to the free list if it fails and someone else hasn't done it. - Make a new function vfreehead() which moves a vnode to the head of the free list and use it in vgone() to clean up that code a bit. Sponsored by: Isilon Systems, Inc. Reported by: pho, kkenn
2005-04-30 11:22:40 +00:00
return (0);
- Move some logic from getnewvnode() to a new function vcanrecycle() - Unlock the free list mutex around vcanrecycle to prevent a lock order reversal.
2002-08-05 10:15:56 +00:00
}
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
/*
* Return the next vnode from the free list.
*/
The big 4.4BSD Lite to FreeBSD 2.0.0 (Development) patch. Reviewed by: Rodney W. Grimes Submitted by: John Dyson and David Greenman
1994-05-25 09:21:21 +00:00
int
Convert remaining functions in vfs_subr.c from K&R prototypes to ANSI C prototypes, as the majority of new functions added have been in this style. Changing prototype style now results in gcc noticing that the implementation of vn_pollrecord() has a 'short' argument instead of 'int' as prototyped in vnode.h, so correct that definition. In practice this didn't matter as only poll flags in the lower 16 bits are used. MFC after: 1 week
2006-01-21 19:42:10 +00:00
getnewvnode(const char *tag, struct mount *mp, struct vop_vector *vops,
struct vnode **vpp)
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
{
Simplify and rationalise the management of the vnode free list (preparing the code to add snapshots).
2000-07-04 04:32:40 +00:00
struct vnode *vp = NULL;
Add BO_* macros parallel to VI_* macros for manipulating the bo_mtx. Initialize the bo_mtx when we allocate a vnode i getnewvnode() For now we point to the vnodes interlock mutex, that retains the exact same locking sematics. Move v_numoutput from vnode to bufobj. Add renaming macro to postpone code sweep.
2004-10-21 14:42:31 +00:00
struct bufobj *bo;
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
CTR3(KTR_VFS, "%s: mp %p with tag %s", __func__, mp, tag);
- Extend the vnode_free_list_mtx to cover numvnodes and freevnodes. This was done only some of the time before, and now it is uniformly applied.
2002-08-13 05:29:48 +00:00
mtx_lock(&vnode_free_list_mtx);
Clean up and improve the namecache. 1. We always keep one 16th of the vnodes on the freelist, so that the namecache doesn't get trashed. It used to be that it wasn't a problem, but the only vnodes getting released these days are directories and things which Clean up and improve the namecache. 1. We always keep one 16th of the vnodes on the freelist, so that the namecache doesn't get trashed. It used to be that it wasn't a problem, but the only vnodes getting released these days are directories and things which gets forced out of the VM/cache. The latter is not numerous enough to keep the pool of vnodes needed for the namecache sufficiently big. 2. Purge invalid entries in the namecache as soon as we notice them. This avoids a stale entry pushing out a valid entry on the LRU list. 3. Speed up the lookup in the namecache by avoid a special case branch. 4. Make the cache purge routines do the thing they're supposed to, and in a decently efficient manner. 5. Make the size of the namecache follow the number of vnodes, so that we can always point to all the vnodes we have in core. 6. Readability has gone way up. 7. Added a "options NCH_STATISTICS" feature that will gather more detailed statistics on the performance of the namecache. Reviewed by: davidg (cvs is dumping core on me :-( )
1995-03-09 20:27:04 +00:00
/*
- Don't recycle vnodes anymore. Free them once they are dead. getnewvnode now always allocates a new vnode. - Define a new function, vnlru_free, which frees vnodes from the free list. It takes as a parameter the number of vnodes to free, which is wantfreevnodes - freevnodes when called from vnlru_proc or 1 when called from getnewvnode(). For now, getnewvnode() still tries to reclaim a free vnode before creating a new one when we are near the limit. - Define a function, vdestroy, which handles the actual release of memory and teardown of locks, etc. This could become a uma_dtor() routine. - Get rid of minvnodes. Now wantfreevnodes is 1/4th the max vnodes. This keeps more unreferenced vnodes around so that files which have only been stat'd are less likely to be kicked out of the system before we have a chance to read them, etc. These vnodes may still be freed via the normal vnlru_proc() routines which may some day become a real lru.
2005-03-25 05:34:39 +00:00
* Lend our context to reclaim vnodes if they've exceeded the max.
*/
if (freevnodes > wantfreevnodes)
vnlru_free(1);
/*
* Wait for available vnodes.
Implement kern.maxvnodes. adjusting kern.maxvnodes now actually has a real effect. Optimize vfs_msync(). Avoid having to continually drop and re-obtain mutexes when scanning the vnode list. Improves looping case by 500%. Optimize ffs_sync(). Avoid having to continually drop and re-obtain mutexes when scanning the vnode list. This makes a couple of assumptions, which I believe are ok, in regards to vnode stability when the mount list mutex is held. Improves looping case by 500%. (more optimization work is needed on top of these fixes) MFC after: 1 week
2001-10-26 00:08:05 +00:00
*/
- Instead of waiting forever to get a vnode in getnewvnode() wait for one to become available for one second and then return ENFILE. We can run out of vnodes, and there must be a hard limit because without one we can quickly run out of KVA on x86. Presently the system can deadlock if there are maxvnodes directories in the namecache. The original 4.x BSD behavior was to return ENFILE if we reached the max, but 4.x BSD did not have the vnlru proc so it was less profitable to wait.
2005-04-04 11:43:44 +00:00
if (numvnodes > desiredvnodes) {
getnewvnode() can be called with NULL mp. Found by: Coverity Prevent (tm) Coverity ID: 1521 Confirmed by: phk
2006-08-10 08:56:03 +00:00
if (mp != NULL && (mp->mnt_kern_flag & MNTK_SUSPEND)) {
Add a bandaid to avoid a deadlock in a situation, when we are trying to suspend a file system, but need to obtain a vnode. We may not be able to do it, because all vnodes could be already in use and other processes cannot release them, because they are waiting in "suspfs" state. In such situation, we allow to allocate a vnode anyway. This is a temporary fix - there is no backpressure to free vnodes allocated in those circumstances. MFC after: 1 week Reviewed by: tegge
2006-08-09 12:47:30 +00:00
/*
* File system is beeing suspended, we cannot risk a
* deadlock here, so allocate new vnode anyway.
*/
if (freevnodes > wantfreevnodes)
vnlru_free(freevnodes - wantfreevnodes);
goto alloc;
}
Vnodes pull in 800-900 bytes these days, all things counted, so we need to treat desiredvnodes much more like a limit than as a vague concept. On a 2GB RAM machine where desired vnodes is 130k, we run out of kmem_map space when we hit about 190k vnodes. If we wake up the vnode washer in getnewvnode(), sleep until it is done, so that it has a chance to offer us a washed vnode. If we don't sleep here we'll just race ahead and allocate yet a vnode which will never get freed. In the vnodewasher, instead of doing 10 vnodes per mountpoint per rotation, do 10% of the vnodes distributed evenly across the mountpoints.
2002-12-29 10:39:05 +00:00
if (vnlruproc_sig == 0) {
Minor style cleanups (mostly removal of trailing whitespaces).
2007-04-10 15:29:37 +00:00
vnlruproc_sig = 1; /* avoid unnecessary wakeups */
Vnodes pull in 800-900 bytes these days, all things counted, so we need to treat desiredvnodes much more like a limit than as a vague concept. On a 2GB RAM machine where desired vnodes is 130k, we run out of kmem_map space when we hit about 190k vnodes. If we wake up the vnode washer in getnewvnode(), sleep until it is done, so that it has a chance to offer us a washed vnode. If we don't sleep here we'll just race ahead and allocate yet a vnode which will never get freed. In the vnodewasher, instead of doing 10 vnodes per mountpoint per rotation, do 10% of the vnodes distributed evenly across the mountpoints.
2002-12-29 10:39:05 +00:00
wakeup(vnlruproc);
}
- Add the tunable and sysctl for the mpsafevfs. It currently defaults to off. - Protect access to mnt_kern_flag with the mointpoint mutex. - Remove some KASSERTs which are not legal checks without the appropriate locks held. - Use VCANRECYCLE() rather than rolling several slightly different checks together. - Return from vtryrecycle() with a recycled vnode rather than a locked vnode. This simplifies some locking. - Remove several GIANT_REQUIRED lines. - Add a few KASSERTs to help with INACT debugging. Sponsored By: Isilon Systems, Inc.
2005-01-24 10:41:01 +00:00
msleep(&vnlruproc_sig, &vnode_free_list_mtx, PVFS,
"vlruwk", hz);
- Disable code which allows getnewvnode() to fail. Many ffs_vget() callers do not correctly deal with failures. This presently risks deadlock problems if dependency processing is held up by failures to allocate a vnode, however, this is better than the situation with the failures. Sponsored by: Isilon Systems, Inc.
2005-04-22 00:57:05 +00:00
#if 0 /* XXX Not all VFS_VGET/ffs_vget callers check returns. */
- Add a missing unlock of the vnode_free_list_mtx. Spotted by: Antoine Brodin
2005-04-04 12:07:16 +00:00
if (numvnodes > desiredvnodes) {
mtx_unlock(&vnode_free_list_mtx);
- Instead of waiting forever to get a vnode in getnewvnode() wait for one to become available for one second and then return ENFILE. We can run out of vnodes, and there must be a hard limit because without one we can quickly run out of KVA on x86. Presently the system can deadlock if there are maxvnodes directories in the namecache. The original 4.x BSD behavior was to return ENFILE if we reached the max, but 4.x BSD did not have the vnlru proc so it was less profitable to wait.
2005-04-04 11:43:44 +00:00
return (ENFILE);
- Add a missing unlock of the vnode_free_list_mtx. Spotted by: Antoine Brodin
2005-04-04 12:07:16 +00:00
}
- Disable code which allows getnewvnode() to fail. Many ffs_vget() callers do not correctly deal with failures. This presently risks deadlock problems if dependency processing is held up by failures to allocate a vnode, however, this is better than the situation with the failures. Sponsored by: Isilon Systems, Inc.
2005-04-22 00:57:05 +00:00
#endif
This is a forward port of Peter's vlrureclaim() fix, with some minor mods by me to make it more efficient. The original code had serious balancing problems and could also deadlock easily. This code relegates the vnode reclamation to its own kproc and relaxes the vnode reclamation requirements to better maintain kern.maxvnodes. This code still doesn't balance as well as it could, but it does a much better job then the original code. Approved by: re@freebsd.org Obtained from: ps, peter, dillon MFS Assuming: Assuming no problems crop up in Yahoo testing MFC after: 7 days
2001-12-18 20:48:54 +00:00
}
Add a bandaid to avoid a deadlock in a situation, when we are trying to suspend a file system, but need to obtain a vnode. We may not be able to do it, because all vnodes could be already in use and other processes cannot release them, because they are waiting in "suspfs" state. In such situation, we allow to allocate a vnode anyway. This is a temporary fix - there is no backpressure to free vnodes allocated in those circumstances. MFC after: 1 week Reviewed by: tegge
2006-08-09 12:47:30 +00:00
alloc:
- Don't recycle vnodes anymore. Free them once they are dead. getnewvnode now always allocates a new vnode. - Define a new function, vnlru_free, which frees vnodes from the free list. It takes as a parameter the number of vnodes to free, which is wantfreevnodes - freevnodes when called from vnlru_proc or 1 when called from getnewvnode(). For now, getnewvnode() still tries to reclaim a free vnode before creating a new one when we are near the limit. - Define a function, vdestroy, which handles the actual release of memory and teardown of locks, etc. This could become a uma_dtor() routine. - Get rid of minvnodes. Now wantfreevnodes is 1/4th the max vnodes. This keeps more unreferenced vnodes around so that files which have only been stat'd are less likely to be kicked out of the system before we have a chance to read them, etc. These vnodes may still be freed via the normal vnlru_proc() routines which may some day become a real lru.
2005-03-25 05:34:39 +00:00
numvnodes++;
mtx_unlock(&vnode_free_list_mtx);
vp = (struct vnode *) uma_zalloc(vnode_zone, M_WAITOK|M_ZERO);
Implement kern.maxvnodes. adjusting kern.maxvnodes now actually has a real effect. Optimize vfs_msync(). Avoid having to continually drop and re-obtain mutexes when scanning the vnode list. Improves looping case by 500%. Optimize ffs_sync(). Avoid having to continually drop and re-obtain mutexes when scanning the vnode list. This makes a couple of assumptions, which I believe are ok, in regards to vnode stability when the mount list mutex is held. Improves looping case by 500%. (more optimization work is needed on top of these fixes) MFC after: 1 week
2001-10-26 00:08:05 +00:00
/*
- Don't recycle vnodes anymore. Free them once they are dead. getnewvnode now always allocates a new vnode. - Define a new function, vnlru_free, which frees vnodes from the free list. It takes as a parameter the number of vnodes to free, which is wantfreevnodes - freevnodes when called from vnlru_proc or 1 when called from getnewvnode(). For now, getnewvnode() still tries to reclaim a free vnode before creating a new one when we are near the limit. - Define a function, vdestroy, which handles the actual release of memory and teardown of locks, etc. This could become a uma_dtor() routine. - Get rid of minvnodes. Now wantfreevnodes is 1/4th the max vnodes. This keeps more unreferenced vnodes around so that files which have only been stat'd are less likely to be kicked out of the system before we have a chance to read them, etc. These vnodes may still be freed via the normal vnlru_proc() routines which may some day become a real lru.
2005-03-25 05:34:39 +00:00
* Setup locks.
Clean up and improve the namecache. 1. We always keep one 16th of the vnodes on the freelist, so that the namecache doesn't get trashed. It used to be that it wasn't a problem, but the only vnodes getting released these days are directories and things which Clean up and improve the namecache. 1. We always keep one 16th of the vnodes on the freelist, so that the namecache doesn't get trashed. It used to be that it wasn't a problem, but the only vnodes getting released these days are directories and things which gets forced out of the VM/cache. The latter is not numerous enough to keep the pool of vnodes needed for the namecache sufficiently big. 2. Purge invalid entries in the namecache as soon as we notice them. This avoids a stale entry pushing out a valid entry on the LRU list. 3. Speed up the lookup in the namecache by avoid a special case branch. 4. Make the cache purge routines do the thing they're supposed to, and in a decently efficient manner. 5. Make the size of the namecache follow the number of vnodes, so that we can always point to all the vnodes we have in core. 6. Readability has gone way up. 7. Added a "options NCH_STATISTICS" feature that will gather more detailed statistics on the performance of the namecache. Reviewed by: davidg (cvs is dumping core on me :-( )
1995-03-09 20:27:04 +00:00
*/
- Don't recycle vnodes anymore. Free them once they are dead. getnewvnode now always allocates a new vnode. - Define a new function, vnlru_free, which frees vnodes from the free list. It takes as a parameter the number of vnodes to free, which is wantfreevnodes - freevnodes when called from vnlru_proc or 1 when called from getnewvnode(). For now, getnewvnode() still tries to reclaim a free vnode before creating a new one when we are near the limit. - Define a function, vdestroy, which handles the actual release of memory and teardown of locks, etc. This could become a uma_dtor() routine. - Get rid of minvnodes. Now wantfreevnodes is 1/4th the max vnodes. This keeps more unreferenced vnodes around so that files which have only been stat'd are less likely to be kicked out of the system before we have a chance to read them, etc. These vnodes may still be freed via the normal vnlru_proc() routines which may some day become a real lru.
2005-03-25 05:34:39 +00:00
vp->v_vnlock = &vp->v_lock;
mtx_init(&vp->v_interlock, "vnode interlock", NULL, MTX_DEF);
- Disable vfs shared locks by default. They must be specifically enabled on filesystems which safely support them. It appears that many network filesystems specifically are not shared lock safe. Sponsored by: Isilon Systems, Inc.
2005-03-31 05:22:45 +00:00
/*
* By default, don't allow shared locks unless filesystems
* opt-in.
*/
lockinit(vp->v_vnlock, PVFS, tag, VLKTIMEOUT, LK_NOSHARE);
- Don't recycle vnodes anymore. Free them once they are dead. getnewvnode now always allocates a new vnode. - Define a new function, vnlru_free, which frees vnodes from the free list. It takes as a parameter the number of vnodes to free, which is wantfreevnodes - freevnodes when called from vnlru_proc or 1 when called from getnewvnode(). For now, getnewvnode() still tries to reclaim a free vnode before creating a new one when we are near the limit. - Define a function, vdestroy, which handles the actual release of memory and teardown of locks, etc. This could become a uma_dtor() routine. - Get rid of minvnodes. Now wantfreevnodes is 1/4th the max vnodes. This keeps more unreferenced vnodes around so that files which have only been stat'd are less likely to be kicked out of the system before we have a chance to read them, etc. These vnodes may still be freed via the normal vnlru_proc() routines which may some day become a real lru.
2005-03-25 05:34:39 +00:00
/*
* Initialize bufobj.
*/
bo = &vp->v_bufobj;
bo->__bo_vnode = vp;
- Complete part of the unfinished bufobj work by consistently using BO_LOCK/UNLOCK/MTX when manipulating the bufobj. - Create a new lock in the bufobj to lock bufobj fields independently. This leaves the vnode interlock as an 'identity' lock while the bufobj is an io lock. The bufobj lock is ordered before the vnode interlock and also before the mnt ilock. - Exploit this new lock order to simplify softdep_check_suspend(). - A few sync related functions are marked with a new XXX to note that we may not properly interlock against a non-zero bv_cnt when attempting to sync all vnodes on a mountlist. I do not believe this race is important. If I'm wrong this will make these locations easier to find. Reviewed by: kib (earlier diff) Tested by: kris, pho (earlier diff)
2008-03-22 09:15:16 +00:00
mtx_init(BO_MTX(bo), "bufobj interlock", NULL, MTX_DEF);
Move the buffer method vector (buf->b_op) to the bufobj. Extend it with a strategy method. Add bufstrategy() which do the usual VOP_SPECSTRATEGY/VOP_STRATEGY song and dance. Rename ibwrite to bufwrite(). Move the two NFS buf_ops to more sensible places, add bufstrategy to them. Add inlines for bwrite() and bstrategy() which calls through buf->b_bufobj->b_ops->b_{write,strategy}(). Replace almost all VOP_STRATEGY()/VOP_SPECSTRATEGY() calls with bstrategy().
2004-10-24 20:03:41 +00:00
bo->bo_ops = &buf_ops_bio;
Always initialize bo_private along with bo_ops in getnewvnode(). Spotted by: tegge
2004-11-03 21:09:23 +00:00
bo->bo_private = vp;
- Don't recycle vnodes anymore. Free them once they are dead. getnewvnode now always allocates a new vnode. - Define a new function, vnlru_free, which frees vnodes from the free list. It takes as a parameter the number of vnodes to free, which is wantfreevnodes - freevnodes when called from vnlru_proc or 1 when called from getnewvnode(). For now, getnewvnode() still tries to reclaim a free vnode before creating a new one when we are near the limit. - Define a function, vdestroy, which handles the actual release of memory and teardown of locks, etc. This could become a uma_dtor() routine. - Get rid of minvnodes. Now wantfreevnodes is 1/4th the max vnodes. This keeps more unreferenced vnodes around so that files which have only been stat'd are less likely to be kicked out of the system before we have a chance to read them, etc. These vnodes may still be freed via the normal vnlru_proc() routines which may some day become a real lru.
2005-03-25 05:34:39 +00:00
TAILQ_INIT(&bo->bo_clean.bv_hd);
TAILQ_INIT(&bo->bo_dirty.bv_hd);
/*
* Initialize namecache.
*/
LIST_INIT(&vp->v_cache_src);
TAILQ_INIT(&vp->v_cache_dst);
/*
* Finalize various vnode identity bits.
*/
Initialize v_rdev in getnewvnode() - it appears that some filesystems may not properly initialize this field in all cases, and this would result in very anti-social behavior (overwriting on some other random device/location). Submitted by: John Dyson
1995-02-27 06:50:08 +00:00
vp->v_type = VNON;
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
vp->v_tag = tag;
vp->v_op = vops;
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
v_incr_usecount(vp);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
vp->v_data = 0;
Move vnode MAC label initialization to after the release of the vnode interlock in getnewvnode() to avoid possible sleeps while holding the mutex. Note that the warning from Witness is a slight false positive since we know there will be no contention on the interlock since we haven't made the vnode available for use yet, but the theory is not a bad one. Obtained from: TrustedBSD Project Sponsored by: DARPA, Network Associates Laboratories
2002-09-30 20:51:48 +00:00
#ifdef MAC
Merge first in a series of TrustedBSD MAC Framework KPI changes from Mac OS X Leopard--rationalize naming for entry points to the following general forms: mac_<object>_<method/action> mac_<object>_check_<method/action> The previous naming scheme was inconsistent and mostly reversed from the new scheme. Also, make object types more consistent and remove spaces from object types that contain multiple parts ("posix_sem" -> "posixsem") to make mechanical parsing easier. Introduce a new "netinet" object type for certain IPv4/IPv6-related methods. Also simplify, slightly, some entry point names. All MAC policy modules will need to be recompiled, and modules not updates as part of this commit will need to be modified to conform to the new KPI. Sponsored by: SPARTA (original patches against Mac OS X) Obtained from: TrustedBSD Project, Apple Computer
2007-10-24 19:04:04 +00:00
mac_vnode_init(vp);
Slightly change the semantics of vnode labels for MAC: rather than "refreshing" the label on the vnode before use, just get the label right from inception. For single-label file systems, set the label in the generic VFS getnewvnode() code; for multi-label file systems, leave the labeling up to the file system. With UFS1/2, this means reading the extended attribute during vfs_vget() as the inode is pulled off disk, rather than hitting the extended attributes frequently during operations later, improving performance. This also corrects sematics for shared vnode locks, which were not previously present in the system. This chances the cache coherrency properties WRT out-of-band access to label data, but in an acceptable form. With UFS1, there is a small race condition during automatic extended attribute start -- this is not present with UFS2, and occurs because EAs aren't available at vnode inception. We'll introduce a work around for this shortly. Approved by: re Obtained from: TrustedBSD Project Sponsored by: DARPA, Network Associates Laboratories
2002-10-26 14:38:24 +00:00
if (mp != NULL && (mp->mnt_flag & MNT_MULTILABEL) == 0)
Merge first in a series of TrustedBSD MAC Framework KPI changes from Mac OS X Leopard--rationalize naming for entry points to the following general forms: mac_<object>_<method/action> mac_<object>_check_<method/action> The previous naming scheme was inconsistent and mostly reversed from the new scheme. Also, make object types more consistent and remove spaces from object types that contain multiple parts ("posix_sem" -> "posixsem") to make mechanical parsing easier. Introduce a new "netinet" object type for certain IPv4/IPv6-related methods. Also simplify, slightly, some entry point names. All MAC policy modules will need to be recompiled, and modules not updates as part of this commit will need to be modified to conform to the new KPI. Sponsored by: SPARTA (original patches against Mac OS X) Obtained from: TrustedBSD Project, Apple Computer
2007-10-24 19:04:04 +00:00
mac_vnode_associate_singlelabel(mp, vp);
Remove worrying printf warning on bootup when processing vnodes which have NULL mount-points. This is the case for special vnodes, such as the one used in nameiinit() which is used for crossing mount points in lookup() to avoid lock ordering issues. MFC after: 2 weeks Discussed with: rwatson, kib
2008-08-24 20:16:44 +00:00
else if (mp == NULL && vops != &dead_vnodeops)
When MAC is enabled, warn if getnewvnode() is asked to produce a vnode without a mountpoint. In this scenario, there's no useful source for a label on the vnode, since we can't query the mountpoint for the labeling strategy or default label.
2004-10-22 11:04:58 +00:00
printf("NULL mp in getnewvnode()\n");
Move vnode MAC label initialization to after the release of the vnode interlock in getnewvnode() to avoid possible sleeps while holding the mutex. Note that the warning from Witness is a slight false positive since we know there will be no contention on the interlock since we haven't made the vnode available for use yet, but the theory is not a bad one. Obtained from: TrustedBSD Project Sponsored by: DARPA, Network Associates Laboratories
2002-09-30 20:51:48 +00:00
#endif
Put the I/O block size in bufobj->bo_bsize. We keep si_bsize_phys around for now as that is the simplest way to pull the number out of disk device drivers in devfs_open(). The correct solution would be to do an ioctl(DIOCGSECTORSIZE), but the point is probably mooth when filesystems sit on GEOM, so don't bother for now.
2004-10-26 07:39:12 +00:00
if (mp != NULL) {
bo->bo_bsize = mp->mnt_stat.f_iosize;
Holding a vnode doesn't prevent v_mount from disappearing (when the vnode is inactivated), possibly leading to a NULL dereference when checking if the mount wants knotes to be activated in the VOP hooks. So, we add a new vnode flag VV_NOKNOTE that is only set in getnewvnode(), if necessary, and check it when activating knotes. Since the flags are not erased when a vnode is being held, we can safely read them. Reviewed by: kris@ MFC after: 3 days
2005-08-06 01:42:04 +00:00
if ((mp->mnt_kern_flag & MNTK_NOKNOTE) != 0)
vp->v_vflag |= VV_NOKNOTE;
Put the I/O block size in bufobj->bo_bsize. We keep si_bsize_phys around for now as that is the simplest way to pull the number out of disk device drivers in devfs_open(). The correct solution would be to do an ioctl(DIOCGSECTORSIZE), but the point is probably mooth when filesystems sit on GEOM, so don't bother for now.
2004-10-26 07:39:12 +00:00
}
Clean-up the vget mechanism by permanently attaching VM objects to vnodes, therefore vget doesn't need to do so anymore. Other minor improvements include the temp free vnode queue obeying the VAGE flag and a printf that warns of to-be-removed code being executed.
1998-02-23 06:59:52 +00:00
- Don't recycle vnodes anymore. Free them once they are dead. getnewvnode now always allocates a new vnode. - Define a new function, vnlru_free, which frees vnodes from the free list. It takes as a parameter the number of vnodes to free, which is wantfreevnodes - freevnodes when called from vnlru_proc or 1 when called from getnewvnode(). For now, getnewvnode() still tries to reclaim a free vnode before creating a new one when we are near the limit. - Define a function, vdestroy, which handles the actual release of memory and teardown of locks, etc. This could become a uma_dtor() routine. - Get rid of minvnodes. Now wantfreevnodes is 1/4th the max vnodes. This keeps more unreferenced vnodes around so that files which have only been stat'd are less likely to be kicked out of the system before we have a chance to read them, etc. These vnodes may still be freed via the normal vnlru_proc() routines which may some day become a real lru.
2005-03-25 05:34:39 +00:00
*vpp = vp;
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
return (0);
}
/*
When we traverse the vnodes on a mountpoint we need to look out for our cached 'next vnode' being removed from this mountpoint. If we find that it was recycled, we restart our traversal from the start of the list. Code to do that is in all local disk filesystems (and a few other places) and looks roughly like this: MNT_ILOCK(mp); loop: for (vp = TAILQ_FIRST(&mp...); (vp = nvp) != NULL; nvp = TAILQ_NEXT(vp,...)) { if (vp->v_mount != mp) goto loop; MNT_IUNLOCK(mp); ... MNT_ILOCK(mp); } MNT_IUNLOCK(mp); The code which takes vnodes off a mountpoint looks like this: MNT_ILOCK(vp->v_mount); ... TAILQ_REMOVE(&vp->v_mount->mnt_nvnodelist, vp, v_nmntvnodes); ... MNT_IUNLOCK(vp->v_mount); ... vp->v_mount = something; (Take a moment and try to spot the locking error before you read on.) On a SMP system, one CPU could have removed nvp from our mountlist but not yet gotten to assign a new value to vp->v_mount while another CPU simultaneously get to the top of the traversal loop where it finds that (vp->v_mount != mp) is not true despite the fact that the vnode has indeed been removed from our mountpoint. Fix: Introduce the macro MNT_VNODE_FOREACH() to traverse the list of vnodes on a mountpoint while taking into account that vnodes may be removed from the list as we go. This saves approx 65 lines of duplicated code. Split the insmntque() which potentially moves a vnode from one mount point to another into delmntque() and insmntque() which does just what the names say. Fix delmntque() to set vp->v_mount to NULL while holding the mountpoint lock.
2004-07-04 08:52:35 +00:00
* Delete from old mount point vnode list, if on one.
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
*/
Staticized.
1997-11-22 08:35:46 +00:00
static void
When we traverse the vnodes on a mountpoint we need to look out for our cached 'next vnode' being removed from this mountpoint. If we find that it was recycled, we restart our traversal from the start of the list. Code to do that is in all local disk filesystems (and a few other places) and looks roughly like this: MNT_ILOCK(mp); loop: for (vp = TAILQ_FIRST(&mp...); (vp = nvp) != NULL; nvp = TAILQ_NEXT(vp,...)) { if (vp->v_mount != mp) goto loop; MNT_IUNLOCK(mp); ... MNT_ILOCK(mp); } MNT_IUNLOCK(mp); The code which takes vnodes off a mountpoint looks like this: MNT_ILOCK(vp->v_mount); ... TAILQ_REMOVE(&vp->v_mount->mnt_nvnodelist, vp, v_nmntvnodes); ... MNT_IUNLOCK(vp->v_mount); ... vp->v_mount = something; (Take a moment and try to spot the locking error before you read on.) On a SMP system, one CPU could have removed nvp from our mountlist but not yet gotten to assign a new value to vp->v_mount while another CPU simultaneously get to the top of the traversal loop where it finds that (vp->v_mount != mp) is not true despite the fact that the vnode has indeed been removed from our mountpoint. Fix: Introduce the macro MNT_VNODE_FOREACH() to traverse the list of vnodes on a mountpoint while taking into account that vnodes may be removed from the list as we go. This saves approx 65 lines of duplicated code. Split the insmntque() which potentially moves a vnode from one mount point to another into delmntque() and insmntque() which does just what the names say. Fix delmntque() to set vp->v_mount to NULL while holding the mountpoint lock.
2004-07-04 08:52:35 +00:00
delmntque(struct vnode *vp)
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
{
When we traverse the vnodes on a mountpoint we need to look out for our cached 'next vnode' being removed from this mountpoint. If we find that it was recycled, we restart our traversal from the start of the list. Code to do that is in all local disk filesystems (and a few other places) and looks roughly like this: MNT_ILOCK(mp); loop: for (vp = TAILQ_FIRST(&mp...); (vp = nvp) != NULL; nvp = TAILQ_NEXT(vp,...)) { if (vp->v_mount != mp) goto loop; MNT_IUNLOCK(mp); ... MNT_ILOCK(mp); } MNT_IUNLOCK(mp); The code which takes vnodes off a mountpoint looks like this: MNT_ILOCK(vp->v_mount); ... TAILQ_REMOVE(&vp->v_mount->mnt_nvnodelist, vp, v_nmntvnodes); ... MNT_IUNLOCK(vp->v_mount); ... vp->v_mount = something; (Take a moment and try to spot the locking error before you read on.) On a SMP system, one CPU could have removed nvp from our mountlist but not yet gotten to assign a new value to vp->v_mount while another CPU simultaneously get to the top of the traversal loop where it finds that (vp->v_mount != mp) is not true despite the fact that the vnode has indeed been removed from our mountpoint. Fix: Introduce the macro MNT_VNODE_FOREACH() to traverse the list of vnodes on a mountpoint while taking into account that vnodes may be removed from the list as we go. This saves approx 65 lines of duplicated code. Split the insmntque() which potentially moves a vnode from one mount point to another into delmntque() and insmntque() which does just what the names say. Fix delmntque() to set vp->v_mount to NULL while holding the mountpoint lock.
2004-07-04 08:52:35 +00:00
struct mount *mp;
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
When we traverse the vnodes on a mountpoint we need to look out for our cached 'next vnode' being removed from this mountpoint. If we find that it was recycled, we restart our traversal from the start of the list. Code to do that is in all local disk filesystems (and a few other places) and looks roughly like this: MNT_ILOCK(mp); loop: for (vp = TAILQ_FIRST(&mp...); (vp = nvp) != NULL; nvp = TAILQ_NEXT(vp,...)) { if (vp->v_mount != mp) goto loop; MNT_IUNLOCK(mp); ... MNT_ILOCK(mp); } MNT_IUNLOCK(mp); The code which takes vnodes off a mountpoint looks like this: MNT_ILOCK(vp->v_mount); ... TAILQ_REMOVE(&vp->v_mount->mnt_nvnodelist, vp, v_nmntvnodes); ... MNT_IUNLOCK(vp->v_mount); ... vp->v_mount = something; (Take a moment and try to spot the locking error before you read on.) On a SMP system, one CPU could have removed nvp from our mountlist but not yet gotten to assign a new value to vp->v_mount while another CPU simultaneously get to the top of the traversal loop where it finds that (vp->v_mount != mp) is not true despite the fact that the vnode has indeed been removed from our mountpoint. Fix: Introduce the macro MNT_VNODE_FOREACH() to traverse the list of vnodes on a mountpoint while taking into account that vnodes may be removed from the list as we go. This saves approx 65 lines of duplicated code. Split the insmntque() which potentially moves a vnode from one mount point to another into delmntque() and insmntque() which does just what the names say. Fix delmntque() to set vp->v_mount to NULL while holding the mountpoint lock.
2004-07-04 08:52:35 +00:00
mp = vp->v_mount;
- Add a ref count to the mount structure. Sleep for up to 3 seconds in vfs_mount_destroy waiting for this ref to hit 0. We don't print an error if we are rebooting as the root mount always retains some refernces by init proc. - Acquire a mnt ref for every vnode allocated to a mount point. Drop this ref only once vdestroy() has been called and the mount has been freed. - No longer NULL the v_mount pointer in delmntque() so that we may release the ref after vgone() has been called. This allows us to guarantee that the mount point structure will be valid until the last vnode has lost its last ref. - Fix a few places that rely on checking v_mount to detect recycling. Sponsored by: Isilon Systems, Inc. MFC After: 1 week
2006-02-06 10:19:50 +00:00
if (mp == NULL)
return;
When we traverse the vnodes on a mountpoint we need to look out for our cached 'next vnode' being removed from this mountpoint. If we find that it was recycled, we restart our traversal from the start of the list. Code to do that is in all local disk filesystems (and a few other places) and looks roughly like this: MNT_ILOCK(mp); loop: for (vp = TAILQ_FIRST(&mp...); (vp = nvp) != NULL; nvp = TAILQ_NEXT(vp,...)) { if (vp->v_mount != mp) goto loop; MNT_IUNLOCK(mp); ... MNT_ILOCK(mp); } MNT_IUNLOCK(mp); The code which takes vnodes off a mountpoint looks like this: MNT_ILOCK(vp->v_mount); ... TAILQ_REMOVE(&vp->v_mount->mnt_nvnodelist, vp, v_nmntvnodes); ... MNT_IUNLOCK(vp->v_mount); ... vp->v_mount = something; (Take a moment and try to spot the locking error before you read on.) On a SMP system, one CPU could have removed nvp from our mountlist but not yet gotten to assign a new value to vp->v_mount while another CPU simultaneously get to the top of the traversal loop where it finds that (vp->v_mount != mp) is not true despite the fact that the vnode has indeed been removed from our mountpoint. Fix: Introduce the macro MNT_VNODE_FOREACH() to traverse the list of vnodes on a mountpoint while taking into account that vnodes may be removed from the list as we go. This saves approx 65 lines of duplicated code. Split the insmntque() which potentially moves a vnode from one mount point to another into delmntque() and insmntque() which does just what the names say. Fix delmntque() to set vp->v_mount to NULL while holding the mountpoint lock.
2004-07-04 08:52:35 +00:00
MNT_ILOCK(mp);
- Release the mount ref once the vnode has been recycled rather than once the last reference is dropped. I forgot that vnodes can stick around for a very long time until processes discover that they are dead. This means that a vnode reference is not sufficient to keep the mount referenced and even more code will be required to ref mount points. Discovered by: kris
2006-02-23 05:15:37 +00:00
vp->v_mount = NULL;
Convert KASSERTS to VNASSERTS
2005-02-17 10:28:58 +00:00
VNASSERT(mp->mnt_nvnodelistsize > 0, vp,
When we traverse the vnodes on a mountpoint we need to look out for our cached 'next vnode' being removed from this mountpoint. If we find that it was recycled, we restart our traversal from the start of the list. Code to do that is in all local disk filesystems (and a few other places) and looks roughly like this: MNT_ILOCK(mp); loop: for (vp = TAILQ_FIRST(&mp...); (vp = nvp) != NULL; nvp = TAILQ_NEXT(vp,...)) { if (vp->v_mount != mp) goto loop; MNT_IUNLOCK(mp); ... MNT_ILOCK(mp); } MNT_IUNLOCK(mp); The code which takes vnodes off a mountpoint looks like this: MNT_ILOCK(vp->v_mount); ... TAILQ_REMOVE(&vp->v_mount->mnt_nvnodelist, vp, v_nmntvnodes); ... MNT_IUNLOCK(vp->v_mount); ... vp->v_mount = something; (Take a moment and try to spot the locking error before you read on.) On a SMP system, one CPU could have removed nvp from our mountlist but not yet gotten to assign a new value to vp->v_mount while another CPU simultaneously get to the top of the traversal loop where it finds that (vp->v_mount != mp) is not true despite the fact that the vnode has indeed been removed from our mountpoint. Fix: Introduce the macro MNT_VNODE_FOREACH() to traverse the list of vnodes on a mountpoint while taking into account that vnodes may be removed from the list as we go. This saves approx 65 lines of duplicated code. Split the insmntque() which potentially moves a vnode from one mount point to another into delmntque() and insmntque() which does just what the names say. Fix delmntque() to set vp->v_mount to NULL while holding the mountpoint lock.
2004-07-04 08:52:35 +00:00
("bad mount point vnode list size"));
TAILQ_REMOVE(&mp->mnt_nvnodelist, vp, v_nmntvnodes);
mp->mnt_nvnodelistsize--;
- Release the mount ref once the vnode has been recycled rather than once the last reference is dropped. I forgot that vnodes can stick around for a very long time until processes discover that they are dead. This means that a vnode reference is not sufficient to keep the mount referenced and even more code will be required to ref mount points. Discovered by: kris
2006-02-23 05:15:37 +00:00
MNT_REL(mp);
When we traverse the vnodes on a mountpoint we need to look out for our cached 'next vnode' being removed from this mountpoint. If we find that it was recycled, we restart our traversal from the start of the list. Code to do that is in all local disk filesystems (and a few other places) and looks roughly like this: MNT_ILOCK(mp); loop: for (vp = TAILQ_FIRST(&mp...); (vp = nvp) != NULL; nvp = TAILQ_NEXT(vp,...)) { if (vp->v_mount != mp) goto loop; MNT_IUNLOCK(mp); ... MNT_ILOCK(mp); } MNT_IUNLOCK(mp); The code which takes vnodes off a mountpoint looks like this: MNT_ILOCK(vp->v_mount); ... TAILQ_REMOVE(&vp->v_mount->mnt_nvnodelist, vp, v_nmntvnodes); ... MNT_IUNLOCK(vp->v_mount); ... vp->v_mount = something; (Take a moment and try to spot the locking error before you read on.) On a SMP system, one CPU could have removed nvp from our mountlist but not yet gotten to assign a new value to vp->v_mount while another CPU simultaneously get to the top of the traversal loop where it finds that (vp->v_mount != mp) is not true despite the fact that the vnode has indeed been removed from our mountpoint. Fix: Introduce the macro MNT_VNODE_FOREACH() to traverse the list of vnodes on a mountpoint while taking into account that vnodes may be removed from the list as we go. This saves approx 65 lines of duplicated code. Split the insmntque() which potentially moves a vnode from one mount point to another into delmntque() and insmntque() which does just what the names say. Fix delmntque() to set vp->v_mount to NULL while holding the mountpoint lock.
2004-07-04 08:52:35 +00:00
MNT_IUNLOCK(mp);
}
Make insmntque() externally visibile and allow it to fail (e.g. during late stages of unmount). On failure, the vnode is recycled. Add insmntque1(), to allow for file system specific cleanup when recycling vnode on failure. Change getnewvnode() to no longer call insmntque(). Previously, embryonic vnodes were put onto the list of vnode belonging to a file system, which is unsafe for a file system marked MPSAFE. Change vfs_hash_insert() to no longer lock the vnode. The caller now has that responsibility. Change most file systems to lock the vnode and call insmntque() or insmntque1() after a new vnode has been sufficiently setup. Handle failed insmntque*() calls by propagating errors to callers, possibly after some file system specific cleanup. Approved by: re (kensmith) Reviewed by: kib In collaboration with: kib
2007-03-13 01:50:27 +00:00
static void
insmntque_stddtr(struct vnode *vp, void *dtr_arg)
{
vp->v_data = NULL;
vp->v_op = &dead_vnodeops;
/* XXX non mp-safe fs may still call insmntque with vnode
unlocked */
Axe the 'thread' argument from VOP_ISLOCKED() and lockstatus() as it is always curthread. As KPI gets broken by this patch, manpages and __FreeBSD_version will be updated by further commits. Tested by: Andrea Barberio <insomniac at slackware dot it>
2008-02-25 18:45:57 +00:00
if (!VOP_ISLOCKED(vp))
vn_lock() is currently only used with the 'curthread' passed as argument. Remove this argument and pass curthread directly to underlying VOP_LOCK1() VFS method. This modify makes the code cleaner and in particular remove an annoying dependence helping next lockmgr() cleanup. KPI results, obviously, changed. Manpage and FreeBSD_version will be updated through further commits. As a side note, would be valuable to say that next commits will address a similar cleanup about VFS methods, in particular vop_lock1 and vop_unlock. Tested by: Diego Sardina <siarodx at gmail dot com>, Andrea Di Pasquale <whyx dot it at gmail dot com>
2008-01-10 01:10:58 +00:00
vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
Make insmntque() externally visibile and allow it to fail (e.g. during late stages of unmount). On failure, the vnode is recycled. Add insmntque1(), to allow for file system specific cleanup when recycling vnode on failure. Change getnewvnode() to no longer call insmntque(). Previously, embryonic vnodes were put onto the list of vnode belonging to a file system, which is unsafe for a file system marked MPSAFE. Change vfs_hash_insert() to no longer lock the vnode. The caller now has that responsibility. Change most file systems to lock the vnode and call insmntque() or insmntque1() after a new vnode has been sufficiently setup. Handle failed insmntque*() calls by propagating errors to callers, possibly after some file system specific cleanup. Approved by: re (kensmith) Reviewed by: kib In collaboration with: kib
2007-03-13 01:50:27 +00:00
vgone(vp);
vput(vp);
}
When we traverse the vnodes on a mountpoint we need to look out for our cached 'next vnode' being removed from this mountpoint. If we find that it was recycled, we restart our traversal from the start of the list. Code to do that is in all local disk filesystems (and a few other places) and looks roughly like this: MNT_ILOCK(mp); loop: for (vp = TAILQ_FIRST(&mp...); (vp = nvp) != NULL; nvp = TAILQ_NEXT(vp,...)) { if (vp->v_mount != mp) goto loop; MNT_IUNLOCK(mp); ... MNT_ILOCK(mp); } MNT_IUNLOCK(mp); The code which takes vnodes off a mountpoint looks like this: MNT_ILOCK(vp->v_mount); ... TAILQ_REMOVE(&vp->v_mount->mnt_nvnodelist, vp, v_nmntvnodes); ... MNT_IUNLOCK(vp->v_mount); ... vp->v_mount = something; (Take a moment and try to spot the locking error before you read on.) On a SMP system, one CPU could have removed nvp from our mountlist but not yet gotten to assign a new value to vp->v_mount while another CPU simultaneously get to the top of the traversal loop where it finds that (vp->v_mount != mp) is not true despite the fact that the vnode has indeed been removed from our mountpoint. Fix: Introduce the macro MNT_VNODE_FOREACH() to traverse the list of vnodes on a mountpoint while taking into account that vnodes may be removed from the list as we go. This saves approx 65 lines of duplicated code. Split the insmntque() which potentially moves a vnode from one mount point to another into delmntque() and insmntque() which does just what the names say. Fix delmntque() to set vp->v_mount to NULL while holding the mountpoint lock.
2004-07-04 08:52:35 +00:00
/*
* Insert into list of vnodes for the new mount point, if available.
*/
Make insmntque() externally visibile and allow it to fail (e.g. during late stages of unmount). On failure, the vnode is recycled. Add insmntque1(), to allow for file system specific cleanup when recycling vnode on failure. Change getnewvnode() to no longer call insmntque(). Previously, embryonic vnodes were put onto the list of vnode belonging to a file system, which is unsafe for a file system marked MPSAFE. Change vfs_hash_insert() to no longer lock the vnode. The caller now has that responsibility. Change most file systems to lock the vnode and call insmntque() or insmntque1() after a new vnode has been sufficiently setup. Handle failed insmntque*() calls by propagating errors to callers, possibly after some file system specific cleanup. Approved by: re (kensmith) Reviewed by: kib In collaboration with: kib
2007-03-13 01:50:27 +00:00
int
insmntque1(struct vnode *vp, struct mount *mp,
void (*dtr)(struct vnode *, void *), void *dtr_arg)
When we traverse the vnodes on a mountpoint we need to look out for our cached 'next vnode' being removed from this mountpoint. If we find that it was recycled, we restart our traversal from the start of the list. Code to do that is in all local disk filesystems (and a few other places) and looks roughly like this: MNT_ILOCK(mp); loop: for (vp = TAILQ_FIRST(&mp...); (vp = nvp) != NULL; nvp = TAILQ_NEXT(vp,...)) { if (vp->v_mount != mp) goto loop; MNT_IUNLOCK(mp); ... MNT_ILOCK(mp); } MNT_IUNLOCK(mp); The code which takes vnodes off a mountpoint looks like this: MNT_ILOCK(vp->v_mount); ... TAILQ_REMOVE(&vp->v_mount->mnt_nvnodelist, vp, v_nmntvnodes); ... MNT_IUNLOCK(vp->v_mount); ... vp->v_mount = something; (Take a moment and try to spot the locking error before you read on.) On a SMP system, one CPU could have removed nvp from our mountlist but not yet gotten to assign a new value to vp->v_mount while another CPU simultaneously get to the top of the traversal loop where it finds that (vp->v_mount != mp) is not true despite the fact that the vnode has indeed been removed from our mountpoint. Fix: Introduce the macro MNT_VNODE_FOREACH() to traverse the list of vnodes on a mountpoint while taking into account that vnodes may be removed from the list as we go. This saves approx 65 lines of duplicated code. Split the insmntque() which potentially moves a vnode from one mount point to another into delmntque() and insmntque() which does just what the names say. Fix delmntque() to set vp->v_mount to NULL while holding the mountpoint lock.
2004-07-04 08:52:35 +00:00
{
Introduce the VV_FORCEINSMQ vnode flag. It instructs the insmnque() function to ignore the unmounting and forces insertion of the vnode into the mount vnode list. Change insmntque() to fail when forced unmount is in progress and VV_FORCEINSMQ is not specified. Add an assertion to the insmntque(), requiring the vnode to be exclusively locked for mp-safe filesystems. Use the VV_FORCEINSMQ for the creation of the syncvnode. Tested by: pho Reviewed by: tegge MFC after: 1 month
2008-08-28 09:08:15 +00:00
int locked;
When we traverse the vnodes on a mountpoint we need to look out for our cached 'next vnode' being removed from this mountpoint. If we find that it was recycled, we restart our traversal from the start of the list. Code to do that is in all local disk filesystems (and a few other places) and looks roughly like this: MNT_ILOCK(mp); loop: for (vp = TAILQ_FIRST(&mp...); (vp = nvp) != NULL; nvp = TAILQ_NEXT(vp,...)) { if (vp->v_mount != mp) goto loop; MNT_IUNLOCK(mp); ... MNT_ILOCK(mp); } MNT_IUNLOCK(mp); The code which takes vnodes off a mountpoint looks like this: MNT_ILOCK(vp->v_mount); ... TAILQ_REMOVE(&vp->v_mount->mnt_nvnodelist, vp, v_nmntvnodes); ... MNT_IUNLOCK(vp->v_mount); ... vp->v_mount = something; (Take a moment and try to spot the locking error before you read on.) On a SMP system, one CPU could have removed nvp from our mountlist but not yet gotten to assign a new value to vp->v_mount while another CPU simultaneously get to the top of the traversal loop where it finds that (vp->v_mount != mp) is not true despite the fact that the vnode has indeed been removed from our mountpoint. Fix: Introduce the macro MNT_VNODE_FOREACH() to traverse the list of vnodes on a mountpoint while taking into account that vnodes may be removed from the list as we go. This saves approx 65 lines of duplicated code. Split the insmntque() which potentially moves a vnode from one mount point to another into delmntque() and insmntque() which does just what the names say. Fix delmntque() to set vp->v_mount to NULL while holding the mountpoint lock.
2004-07-04 08:52:35 +00:00
Make insmntque() externally visibile and allow it to fail (e.g. during late stages of unmount). On failure, the vnode is recycled. Add insmntque1(), to allow for file system specific cleanup when recycling vnode on failure. Change getnewvnode() to no longer call insmntque(). Previously, embryonic vnodes were put onto the list of vnode belonging to a file system, which is unsafe for a file system marked MPSAFE. Change vfs_hash_insert() to no longer lock the vnode. The caller now has that responsibility. Change most file systems to lock the vnode and call insmntque() or insmntque1() after a new vnode has been sufficiently setup. Handle failed insmntque*() calls by propagating errors to callers, possibly after some file system specific cleanup. Approved by: re (kensmith) Reviewed by: kib In collaboration with: kib
2007-03-13 01:50:27 +00:00
KASSERT(vp->v_mount == NULL,
("insmntque: vnode already on per mount vnode list"));
Convert KASSERTS to VNASSERTS
2005-02-17 10:28:58 +00:00
VNASSERT(mp != NULL, vp, ("Don't call insmntque(foo, NULL)"));
Introduce the VV_FORCEINSMQ vnode flag. It instructs the insmnque() function to ignore the unmounting and forces insertion of the vnode into the mount vnode list. Change insmntque() to fail when forced unmount is in progress and VV_FORCEINSMQ is not specified. Add an assertion to the insmntque(), requiring the vnode to be exclusively locked for mp-safe filesystems. Use the VV_FORCEINSMQ for the creation of the syncvnode. Tested by: pho Reviewed by: tegge MFC after: 1 month
2008-08-28 09:08:15 +00:00
#ifdef DEBUG_VFS_LOCKS
if (!VFS_NEEDSGIANT(mp))
ASSERT_VOP_ELOCKED(vp,
"insmntque: mp-safe fs and non-locked vp");
#endif
- Add a ref count to the mount structure. Sleep for up to 3 seconds in vfs_mount_destroy waiting for this ref to hit 0. We don't print an error if we are rebooting as the root mount always retains some refernces by init proc. - Acquire a mnt ref for every vnode allocated to a mount point. Drop this ref only once vdestroy() has been called and the mount has been freed. - No longer NULL the v_mount pointer in delmntque() so that we may release the ref after vgone() has been called. This allows us to guarantee that the mount point structure will be valid until the last vnode has lost its last ref. - Fix a few places that rely on checking v_mount to detect recycling. Sponsored by: Isilon Systems, Inc. MFC After: 1 week
2006-02-06 10:19:50 +00:00
MNT_ILOCK(mp);
When restoring the mount after umount failed, the MNTK_UNMOUNT flag prevents insmntque() from placing reallocated syncer vnode on mount list, that causes panic in vfs_allocate_syncvnode(). Introduce MNTK_NOINSMNTQ flag, that marks the period when instmntque is not allowed to success, instead of MNTK_UNMOUNT. The MNTK_NOINSMNTQ is set and cleared simultaneously with MNTK_UNMOUNT, except on umount error path, where it is cleaned just before the syncer vnode is going to be allocated. Reported by: Peter Jeremy <peterjeremy optushome com au> Suggested by: tegge Approved by: re (rwatson)
2007-09-12 16:31:32 +00:00
if ((mp->mnt_kern_flag & MNTK_NOINSMNTQ) != 0 &&
Introduce the VV_FORCEINSMQ vnode flag. It instructs the insmnque() function to ignore the unmounting and forces insertion of the vnode into the mount vnode list. Change insmntque() to fail when forced unmount is in progress and VV_FORCEINSMQ is not specified. Add an assertion to the insmntque(), requiring the vnode to be exclusively locked for mp-safe filesystems. Use the VV_FORCEINSMQ for the creation of the syncvnode. Tested by: pho Reviewed by: tegge MFC after: 1 month
2008-08-28 09:08:15 +00:00
((mp->mnt_kern_flag & MNTK_UNMOUNTF) != 0 ||
mp->mnt_nvnodelistsize == 0)) {
locked = VOP_ISLOCKED(vp);
if (!locked || (locked == LK_EXCLUSIVE &&
(vp->v_vflag & VV_FORCEINSMQ) == 0)) {
MNT_IUNLOCK(mp);
if (dtr != NULL)
dtr(vp, dtr_arg);
return (EBUSY);
}
Make insmntque() externally visibile and allow it to fail (e.g. during late stages of unmount). On failure, the vnode is recycled. Add insmntque1(), to allow for file system specific cleanup when recycling vnode on failure. Change getnewvnode() to no longer call insmntque(). Previously, embryonic vnodes were put onto the list of vnode belonging to a file system, which is unsafe for a file system marked MPSAFE. Change vfs_hash_insert() to no longer lock the vnode. The caller now has that responsibility. Change most file systems to lock the vnode and call insmntque() or insmntque1() after a new vnode has been sufficiently setup. Handle failed insmntque*() calls by propagating errors to callers, possibly after some file system specific cleanup. Approved by: re (kensmith) Reviewed by: kib In collaboration with: kib
2007-03-13 01:50:27 +00:00
}
vp->v_mount = mp;
- Add a ref count to the mount structure. Sleep for up to 3 seconds in vfs_mount_destroy waiting for this ref to hit 0. We don't print an error if we are rebooting as the root mount always retains some refernces by init proc. - Acquire a mnt ref for every vnode allocated to a mount point. Drop this ref only once vdestroy() has been called and the mount has been freed. - No longer NULL the v_mount pointer in delmntque() so that we may release the ref after vgone() has been called. This allows us to guarantee that the mount point structure will be valid until the last vnode has lost its last ref. - Fix a few places that rely on checking v_mount to detect recycling. Sponsored by: Isilon Systems, Inc. MFC After: 1 week
2006-02-06 10:19:50 +00:00
MNT_REF(mp);
When we traverse the vnodes on a mountpoint we need to look out for our cached 'next vnode' being removed from this mountpoint. If we find that it was recycled, we restart our traversal from the start of the list. Code to do that is in all local disk filesystems (and a few other places) and looks roughly like this: MNT_ILOCK(mp); loop: for (vp = TAILQ_FIRST(&mp...); (vp = nvp) != NULL; nvp = TAILQ_NEXT(vp,...)) { if (vp->v_mount != mp) goto loop; MNT_IUNLOCK(mp); ... MNT_ILOCK(mp); } MNT_IUNLOCK(mp); The code which takes vnodes off a mountpoint looks like this: MNT_ILOCK(vp->v_mount); ... TAILQ_REMOVE(&vp->v_mount->mnt_nvnodelist, vp, v_nmntvnodes); ... MNT_IUNLOCK(vp->v_mount); ... vp->v_mount = something; (Take a moment and try to spot the locking error before you read on.) On a SMP system, one CPU could have removed nvp from our mountlist but not yet gotten to assign a new value to vp->v_mount while another CPU simultaneously get to the top of the traversal loop where it finds that (vp->v_mount != mp) is not true despite the fact that the vnode has indeed been removed from our mountpoint. Fix: Introduce the macro MNT_VNODE_FOREACH() to traverse the list of vnodes on a mountpoint while taking into account that vnodes may be removed from the list as we go. This saves approx 65 lines of duplicated code. Split the insmntque() which potentially moves a vnode from one mount point to another into delmntque() and insmntque() which does just what the names say. Fix delmntque() to set vp->v_mount to NULL while holding the mountpoint lock.
2004-07-04 08:52:35 +00:00
TAILQ_INSERT_TAIL(&mp->mnt_nvnodelist, vp, v_nmntvnodes);
Add marker vnodes to ensure that all vnodes associated with the mount point are iterated over when using MNT_VNODE_FOREACH. Reviewed by: truckman
2006-01-09 20:42:19 +00:00
VNASSERT(mp->mnt_nvnodelistsize >= 0, vp,
("neg mount point vnode list size"));
When we traverse the vnodes on a mountpoint we need to look out for our cached 'next vnode' being removed from this mountpoint. If we find that it was recycled, we restart our traversal from the start of the list. Code to do that is in all local disk filesystems (and a few other places) and looks roughly like this: MNT_ILOCK(mp); loop: for (vp = TAILQ_FIRST(&mp...); (vp = nvp) != NULL; nvp = TAILQ_NEXT(vp,...)) { if (vp->v_mount != mp) goto loop; MNT_IUNLOCK(mp); ... MNT_ILOCK(mp); } MNT_IUNLOCK(mp); The code which takes vnodes off a mountpoint looks like this: MNT_ILOCK(vp->v_mount); ... TAILQ_REMOVE(&vp->v_mount->mnt_nvnodelist, vp, v_nmntvnodes); ... MNT_IUNLOCK(vp->v_mount); ... vp->v_mount = something; (Take a moment and try to spot the locking error before you read on.) On a SMP system, one CPU could have removed nvp from our mountlist but not yet gotten to assign a new value to vp->v_mount while another CPU simultaneously get to the top of the traversal loop where it finds that (vp->v_mount != mp) is not true despite the fact that the vnode has indeed been removed from our mountpoint. Fix: Introduce the macro MNT_VNODE_FOREACH() to traverse the list of vnodes on a mountpoint while taking into account that vnodes may be removed from the list as we go. This saves approx 65 lines of duplicated code. Split the insmntque() which potentially moves a vnode from one mount point to another into delmntque() and insmntque() which does just what the names say. Fix delmntque() to set vp->v_mount to NULL while holding the mountpoint lock.
2004-07-04 08:52:35 +00:00
mp->mnt_nvnodelistsize++;
- Add a ref count to the mount structure. Sleep for up to 3 seconds in vfs_mount_destroy waiting for this ref to hit 0. We don't print an error if we are rebooting as the root mount always retains some refernces by init proc. - Acquire a mnt ref for every vnode allocated to a mount point. Drop this ref only once vdestroy() has been called and the mount has been freed. - No longer NULL the v_mount pointer in delmntque() so that we may release the ref after vgone() has been called. This allows us to guarantee that the mount point structure will be valid until the last vnode has lost its last ref. - Fix a few places that rely on checking v_mount to detect recycling. Sponsored by: Isilon Systems, Inc. MFC After: 1 week
2006-02-06 10:19:50 +00:00
MNT_IUNLOCK(mp);
Make insmntque() externally visibile and allow it to fail (e.g. during late stages of unmount). On failure, the vnode is recycled. Add insmntque1(), to allow for file system specific cleanup when recycling vnode on failure. Change getnewvnode() to no longer call insmntque(). Previously, embryonic vnodes were put onto the list of vnode belonging to a file system, which is unsafe for a file system marked MPSAFE. Change vfs_hash_insert() to no longer lock the vnode. The caller now has that responsibility. Change most file systems to lock the vnode and call insmntque() or insmntque1() after a new vnode has been sufficiently setup. Handle failed insmntque*() calls by propagating errors to callers, possibly after some file system specific cleanup. Approved by: re (kensmith) Reviewed by: kib In collaboration with: kib
2007-03-13 01:50:27 +00:00
return (0);
}
int
insmntque(struct vnode *vp, struct mount *mp)
{
return (insmntque1(vp, mp, insmntque_stddtr, NULL));
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
}
/*
Try to unbreak the vnode locking around vop_reclaim() (based mostly on patch from kan@). Pull bufobj_invalbuf() out of vinvalbuf() and make g_vfs call it on close. This is not yet a generally safe function, but for this very specific use it is safe. This solves the problem with buffers not being flushed by unmount or after failed mount attempts.
2005-02-19 11:44:57 +00:00
* Flush out and invalidate all buffers associated with a bufobj
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
* Called with the underlying object locked.
*/
int
Remove the struct thread unuseful argument from bufobj interface. In particular following functions KPI results modified: - bufobj_invalbuf() - bufsync() and BO_SYNC() "virtual method" of the buffer objects set. Main consumers of bufobj functions are affected by this change too and, in particular, functions which changed their KPI are: - vinvalbuf() - g_vfs_close() Due to the KPI breakage, __FreeBSD_version will be bumped in a later commit. As a side note, please consider just temporary the 'curthread' argument passing to VOP_SYNC() (in bufsync()) as it will be axed out ASAP Reviewed by: kib Tested by: Giovanni Trematerra <giovanni dot trematerra at gmail dot com>
2008-10-10 21:23:50 +00:00
bufobj_invalbuf(struct bufobj *bo, int flags, int slpflag, int slptimeo)
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
{
- Remove spls(). The locking that has replaced them is in place and they no longer serve as guidelines for future work.
2003-09-19 23:52:06 +00:00
int error;
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
Move the syncer linkage from vnode to bufobj. This is not quite a perfect separation: the syncer still think it knows that everything is a vnode.
2004-10-27 08:05:02 +00:00
BO_LOCK(bo);
Submitted by: Kirk McKusick <mckusick@McKusick.COM> Fix for potential hang when trying to reboot the system or to forcibly unmount a soft update enabled filesystem. FreeBSD already handled the reboot case differently, this is however a better fix.
1998-06-10 18:13:19 +00:00
if (flags & V_SAVE) {
Move the VI_BWAIT flag into no bo_flag element of bufobj and call it BO_WWAIT Add bufobj_wref(), bufobj_wdrop() and bufobj_wwait() to handle the write count on a bufobj. Bufobj_wdrop() replaces vwakeup(). Use these functions all relevant places except in ffs_softdep.c where the use if interlocked_sleep() makes this impossible. Rename b_vnbufs to b_bobufs now that we touch all the relevant files anyway.
2004-10-21 15:53:54 +00:00
error = bufobj_wwait(bo, slpflag, slptimeo);
if (error) {
More vnode -> bufobj migration.
2005-01-11 10:16:39 +00:00
BO_UNLOCK(bo);
Move the VI_BWAIT flag into no bo_flag element of bufobj and call it BO_WWAIT Add bufobj_wref(), bufobj_wdrop() and bufobj_wwait() to handle the write count on a bufobj. Bufobj_wdrop() replaces vwakeup(). Use these functions all relevant places except in ffs_softdep.c where the use if interlocked_sleep() makes this impossible. Rename b_vnbufs to b_bobufs now that we touch all the relevant files anyway.
2004-10-21 15:53:54 +00:00
return (error);
Submitted by: Kirk McKusick <mckusick@McKusick.COM> Fix for potential hang when trying to reboot the system or to forcibly unmount a soft update enabled filesystem. FreeBSD already handled the reboot case differently, this is however a better fix.
1998-06-10 18:13:19 +00:00
}
Move the VI_BWAIT flag into no bo_flag element of bufobj and call it BO_WWAIT Add bufobj_wref(), bufobj_wdrop() and bufobj_wwait() to handle the write count on a bufobj. Bufobj_wdrop() replaces vwakeup(). Use these functions all relevant places except in ffs_softdep.c where the use if interlocked_sleep() makes this impossible. Rename b_vnbufs to b_bobufs now that we touch all the relevant files anyway.
2004-10-21 15:53:54 +00:00
if (bo->bo_dirty.bv_cnt > 0) {
More vnode -> bufobj migration.
2005-01-11 10:16:39 +00:00
BO_UNLOCK(bo);
Remove the struct thread unuseful argument from bufobj interface. In particular following functions KPI results modified: - bufobj_invalbuf() - bufsync() and BO_SYNC() "virtual method" of the buffer objects set. Main consumers of bufobj functions are affected by this change too and, in particular, functions which changed their KPI are: - vinvalbuf() - g_vfs_close() Due to the KPI breakage, __FreeBSD_version will be bumped in a later commit. As a side note, please consider just temporary the 'curthread' argument passing to VOP_SYNC() (in bufsync()) as it will be axed out ASAP Reviewed by: kib Tested by: Giovanni Trematerra <giovanni dot trematerra at gmail dot com>
2008-10-10 21:23:50 +00:00
if ((error = BO_SYNC(bo, MNT_WAIT)) != 0)
Submitted by: Kirk McKusick <mckusick@McKusick.COM> Fix for potential hang when trying to reboot the system or to forcibly unmount a soft update enabled filesystem. FreeBSD already handled the reboot case differently, this is however a better fix.
1998-06-10 18:13:19 +00:00
return (error);
- Replace v_flag with v_iflag and v_vflag - v_vflag is protected by the vnode lock and is used when synchronization with VOP calls is needed. - v_iflag is protected by interlock and is used for dealing with vnode management issues. These flags include X/O LOCK, FREE, DOOMED, etc. - All accesses to v_iflag and v_vflag have either been locked or marked with mp_fixme's. - Many ASSERT_VOP_LOCKED calls have been added where the locking was not clear. - Many functions in vfs_subr.c were restructured to provide for stronger locking. Idea stolen from: BSD/OS
2002-08-04 10:29:36 +00:00
/*
* XXX We could save a lock/unlock if this was only
* enabled under INVARIANTS
*/
More vnode -> bufobj migration.
2005-01-11 10:16:39 +00:00
BO_LOCK(bo);
Move the VI_BWAIT flag into no bo_flag element of bufobj and call it BO_WWAIT Add bufobj_wref(), bufobj_wdrop() and bufobj_wwait() to handle the write count on a bufobj. Bufobj_wdrop() replaces vwakeup(). Use these functions all relevant places except in ffs_softdep.c where the use if interlocked_sleep() makes this impossible. Rename b_vnbufs to b_bobufs now that we touch all the relevant files anyway.
2004-10-21 15:53:54 +00:00
if (bo->bo_numoutput > 0 || bo->bo_dirty.bv_cnt > 0)
Submitted by: Kirk McKusick <mckusick@McKusick.COM> Fix for potential hang when trying to reboot the system or to forcibly unmount a soft update enabled filesystem. FreeBSD already handled the reboot case differently, this is however a better fix.
1998-06-10 18:13:19 +00:00
panic("vinvalbuf: dirty bufs");
}
Gratuitous whitespace cleanup.
2002-06-06 15:46:38 +00:00
}
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
/*
* If you alter this loop please notice that interlock is dropped and
* reacquired in flushbuflist. Special care is needed to ensure that
* no race conditions occur from this.
*/
Give flushbuflist() a struct bufv as first argument and avoid home-rolling TAILQ_FOREACH_SAFE(). Loose the error pointer argument and return any errors the normal way. Return EAGAIN for the case where more work needs to be done.
2005-01-11 10:01:54 +00:00
do {
error = flushbuflist(&bo->bo_clean,
Try to unbreak the vnode locking around vop_reclaim() (based mostly on patch from kan@). Pull bufobj_invalbuf() out of vinvalbuf() and make g_vfs call it on close. This is not yet a generally safe function, but for this very specific use it is safe. This solves the problem with buffers not being flushed by unmount or after failed mount attempts.
2005-02-19 11:44:57 +00:00
flags, bo, slpflag, slptimeo);
Give flushbuflist() a struct bufv as first argument and avoid home-rolling TAILQ_FOREACH_SAFE(). Loose the error pointer argument and return any errors the normal way. Return EAGAIN for the case where more work needs to be done.
2005-01-11 10:01:54 +00:00
if (error == 0)
error = flushbuflist(&bo->bo_dirty,
Try to unbreak the vnode locking around vop_reclaim() (based mostly on patch from kan@). Pull bufobj_invalbuf() out of vinvalbuf() and make g_vfs call it on close. This is not yet a generally safe function, but for this very specific use it is safe. This solves the problem with buffers not being flushed by unmount or after failed mount attempts.
2005-02-19 11:44:57 +00:00
flags, bo, slpflag, slptimeo);
Fix a bug I introduced in 1.561 which has caused considerable filesystem unhappiness lately. As far as I can tell, no files that have made it safely to disk have been endangered, but stuff in transit has been in peril. Pointy hat: phk
2005-01-16 21:09:39 +00:00
if (error != 0 && error != EAGAIN) {
Give flushbuflist() a struct bufv as first argument and avoid home-rolling TAILQ_FOREACH_SAFE(). Loose the error pointer argument and return any errors the normal way. Return EAGAIN for the case where more work needs to be done.
2005-01-11 10:01:54 +00:00
BO_UNLOCK(bo);
return (error);
Add support to UFS2 to provide storage for extended attributes. As this code is not actually used by any of the existing interfaces, it seems unlikely to break anything (famous last words). The internal kernel interface to manipulate these attributes is invoked using two new IO_ flags: IO_NORMAL and IO_EXT. These flags may be specified in the ioflags word of VOP_READ, VOP_WRITE, and VOP_TRUNCATE. Specifying IO_NORMAL means that you want to do I/O to the normal data part of the file and IO_EXT means that you want to do I/O to the extended attributes part of the file. IO_NORMAL and IO_EXT are mutually exclusive for VOP_READ and VOP_WRITE, but may be specified individually or together in the case of VOP_TRUNCATE. For example, when removing a file, VOP_TRUNCATE is called with both IO_NORMAL and IO_EXT set. For backward compatibility, if neither IO_NORMAL nor IO_EXT is set, then IO_NORMAL is assumed. Note that the BA_ and IO_ flags have been `merged' so that they may both be used in the same flags word. This merger is possible by assigning the IO_ flags to the low sixteen bits and the BA_ flags the high sixteen bits. This works because the high sixteen bits of the IO_ word is reserved for read-ahead and help with write clustering so will never be used for flags. This merge lets us get away from code of the form: if (ioflags & IO_SYNC) flags |= BA_SYNC; For the future, I have considered adding a new field to the vattr structure, va_extsize. This addition could then be exported through the stat structure to allow applications to find out the size of the extended attribute storage and also would provide a more standard interface for truncating them (via VOP_SETATTR rather than VOP_TRUNCATE). I am also contemplating adding a pathconf parameter (for concreteness, lets call it _PC_MAX_EXTSIZE) which would let an application determine the maximum size of the extended atribute storage. Sponsored by: DARPA & NAI Labs.
2002-07-19 07:29:39 +00:00
}
Give flushbuflist() a struct bufv as first argument and avoid home-rolling TAILQ_FOREACH_SAFE(). Loose the error pointer argument and return any errors the normal way. Return EAGAIN for the case where more work needs to be done.
2005-01-11 10:01:54 +00:00
} while (error != 0);
"bogus" fixes from 1.1.5 to work around some cache coherency problems.
1994-08-29 06:09:15 +00:00
vinvalbuf() was only waiting for write-I/O to complete. It really has to wait for both read AND write I/O to complete. Only NFS calls vinvalbuf() on an active vnode (when the server indicates that the file is stale), so this bug fix only effects NFS clients. MFC after: 3 days
2001-10-05 20:10:32 +00:00
/*
* Wait for I/O to complete. XXX needs cleaning up. The vnode can
* have write I/O in-progress but if there is a VM object then the
* VM object can also have read-I/O in-progress.
*/
do {
Move the VI_BWAIT flag into no bo_flag element of bufobj and call it BO_WWAIT Add bufobj_wref(), bufobj_wdrop() and bufobj_wwait() to handle the write count on a bufobj. Bufobj_wdrop() replaces vwakeup(). Use these functions all relevant places except in ffs_softdep.c where the use if interlocked_sleep() makes this impossible. Rename b_vnbufs to b_bobufs now that we touch all the relevant files anyway.
2004-10-21 15:53:54 +00:00
bufobj_wwait(bo, 0, 0);
Access vmobject via the bufobj instead of the vnode
2005-02-07 10:04:06 +00:00
BO_UNLOCK(bo);
if (bo->bo_object != NULL) {
VM_OBJECT_LOCK(bo->bo_object);
Try to unbreak the vnode locking around vop_reclaim() (based mostly on patch from kan@). Pull bufobj_invalbuf() out of vinvalbuf() and make g_vfs call it on close. This is not yet a generally safe function, but for this very specific use it is safe. This solves the problem with buffers not being flushed by unmount or after failed mount attempts.
2005-02-19 11:44:57 +00:00
vm_object_pip_wait(bo->bo_object, "bovlbx");
Access vmobject via the bufobj instead of the vnode
2005-02-07 10:04:06 +00:00
VM_OBJECT_UNLOCK(bo->bo_object);
vinvalbuf() was only waiting for write-I/O to complete. It really has to wait for both read AND write I/O to complete. Only NFS calls vinvalbuf() on an active vnode (when the server indicates that the file is stale), so this bug fix only effects NFS clients. MFC after: 3 days
2001-10-05 20:10:32 +00:00
}
Access vmobject via the bufobj instead of the vnode
2005-02-07 10:04:06 +00:00
BO_LOCK(bo);
Loose the v_dirty* and v_clean* alias macros. Check the count field where we just want to know the full/empty state, rather than using TAILQ_EMPTY() or TAILQ_FIRST().
2004-10-25 09:14:03 +00:00
} while (bo->bo_numoutput > 0);
Access vmobject via the bufobj instead of the vnode
2005-02-07 10:04:06 +00:00
BO_UNLOCK(bo);
Fixed splbio problems in vinvalbuf. Closes PR#2875, although fixed differently by me.
1997-03-05 04:54:54 +00:00
Don't attempt to sync pages in the V_SAVE case of vinvalbuf; doing so can lead to a deadlock. Just let the VM system deal with it.
1995-03-20 02:08:24 +00:00
/*
* Destroy the copy in the VM cache, too.
*/
FFS puts the extended attributes blocks at the negative blocks for the vnode, from -1 down. When vinvalbuf(vp, V_ALT) is done for the vnode, it incorrectly does vm_object_page_remove(0, 0), removing all pages from the underlying vm object, not only the pages that back the extended attributes data. Change vinvalbuf() to not remove any pages from the object when V_NORMAL or V_ALT are specified. Instead, the only in-tree caller in ffs_inode.c:ffs_truncate() that specifies V_ALT explicitely removes the corresponding page range. The V_NORMAL caller does vnode_pager_setsize(vp, 0) immediately after the call to vinvalbuf(V_NORMAL) already. Reported by: csjp Reviewed by: ups MFC after: 3 weeks
2009-01-20 11:27:45 +00:00
if (bo->bo_object != NULL && (flags & (V_ALT | V_NORMAL)) == 0) {
Access vmobject via the bufobj instead of the vnode
2005-02-07 10:04:06 +00:00
VM_OBJECT_LOCK(bo->bo_object);
vm_object_page_remove(bo->bo_object, 0, 0,
Remove the V_SAVEMETA flag, nothing uses it any more now that msdosfs and ext2fs call vtruncbuf() directly. This simplifies and cleans up vinvalbuf() a little.
1998-10-29 09:51:28 +00:00
(flags & V_SAVE) ? TRUE : FALSE);
Access vmobject via the bufobj instead of the vnode
2005-02-07 10:04:06 +00:00
VM_OBJECT_UNLOCK(bo->bo_object);
"bogus" fixes from 1.1.5 to work around some cache coherency problems.
1994-08-29 06:09:15 +00:00
}
Make our v_usecount vnode reference count work identically to the original BSD code. The association between the vnode and the vm_object no longer includes reference counts. The major difference is that vm_object's are no longer freed gratuitiously from the vnode, and so once an object is created for the vnode, it will last as long as the vnode does. When a vnode object reference count is incremented, then the underlying vnode reference count is incremented also. The two "objects" are now more intimately related, and so the interactions are now much less complex. When vnodes are now normally placed onto the free queue with an object still attached. The rundown of the object happens at vnode rundown time, and happens with exactly the same filesystem semantics of the original VFS code. There is absolutely no need for vnode_pager_uncache and other travesties like that anymore. A side-effect of these changes is that SMP locking should be much simpler, the I/O copyin/copyout optimizations work, NFS should be more ponderable, and further work on layered filesystems should be less frustrating, because of the totally coherent management of the vnode objects and vnodes. Please be careful with your system while running this code, but I would greatly appreciate feedback as soon a reasonably possible.
1998-01-06 05:26:17 +00:00
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
#ifdef INVARIANTS
More vnode -> bufobj migration.
2005-01-11 10:16:39 +00:00
BO_LOCK(bo);
Add support to UFS2 to provide storage for extended attributes. As this code is not actually used by any of the existing interfaces, it seems unlikely to break anything (famous last words). The internal kernel interface to manipulate these attributes is invoked using two new IO_ flags: IO_NORMAL and IO_EXT. These flags may be specified in the ioflags word of VOP_READ, VOP_WRITE, and VOP_TRUNCATE. Specifying IO_NORMAL means that you want to do I/O to the normal data part of the file and IO_EXT means that you want to do I/O to the extended attributes part of the file. IO_NORMAL and IO_EXT are mutually exclusive for VOP_READ and VOP_WRITE, but may be specified individually or together in the case of VOP_TRUNCATE. For example, when removing a file, VOP_TRUNCATE is called with both IO_NORMAL and IO_EXT set. For backward compatibility, if neither IO_NORMAL nor IO_EXT is set, then IO_NORMAL is assumed. Note that the BA_ and IO_ flags have been `merged' so that they may both be used in the same flags word. This merger is possible by assigning the IO_ flags to the low sixteen bits and the BA_ flags the high sixteen bits. This works because the high sixteen bits of the IO_ word is reserved for read-ahead and help with write clustering so will never be used for flags. This merge lets us get away from code of the form: if (ioflags & IO_SYNC) flags |= BA_SYNC; For the future, I have considered adding a new field to the vattr structure, va_extsize. This addition could then be exported through the stat structure to allow applications to find out the size of the extended attribute storage and also would provide a more standard interface for truncating them (via VOP_SETATTR rather than VOP_TRUNCATE). I am also contemplating adding a pathconf parameter (for concreteness, lets call it _PC_MAX_EXTSIZE) which would let an application determine the maximum size of the extended atribute storage. Sponsored by: DARPA & NAI Labs.
2002-07-19 07:29:39 +00:00
if ((flags & (V_ALT | V_NORMAL)) == 0 &&
More vnode -> bufobj migration.
2005-01-11 10:16:39 +00:00
(bo->bo_dirty.bv_cnt > 0 || bo->bo_clean.bv_cnt > 0))
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
panic("vinvalbuf: flush failed");
More vnode -> bufobj migration.
2005-01-11 10:16:39 +00:00
BO_UNLOCK(bo);
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
#endif
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
return (0);
}
Try to unbreak the vnode locking around vop_reclaim() (based mostly on patch from kan@). Pull bufobj_invalbuf() out of vinvalbuf() and make g_vfs call it on close. This is not yet a generally safe function, but for this very specific use it is safe. This solves the problem with buffers not being flushed by unmount or after failed mount attempts.
2005-02-19 11:44:57 +00:00
/*
* Flush out and invalidate all buffers associated with a vnode.
* Called with the underlying object locked.
*/
int
Remove the struct thread unuseful argument from bufobj interface. In particular following functions KPI results modified: - bufobj_invalbuf() - bufsync() and BO_SYNC() "virtual method" of the buffer objects set. Main consumers of bufobj functions are affected by this change too and, in particular, functions which changed their KPI are: - vinvalbuf() - g_vfs_close() Due to the KPI breakage, __FreeBSD_version will be bumped in a later commit. As a side note, please consider just temporary the 'curthread' argument passing to VOP_SYNC() (in bufsync()) as it will be axed out ASAP Reviewed by: kib Tested by: Giovanni Trematerra <giovanni dot trematerra at gmail dot com>
2008-10-10 21:23:50 +00:00
vinvalbuf(struct vnode *vp, int flags, int slpflag, int slptimeo)
Try to unbreak the vnode locking around vop_reclaim() (based mostly on patch from kan@). Pull bufobj_invalbuf() out of vinvalbuf() and make g_vfs call it on close. This is not yet a generally safe function, but for this very specific use it is safe. This solves the problem with buffers not being flushed by unmount or after failed mount attempts.
2005-02-19 11:44:57 +00:00
{
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
CTR3(KTR_VFS, "%s: vp %p with flags %d", __func__, vp, flags);
Try to unbreak the vnode locking around vop_reclaim() (based mostly on patch from kan@). Pull bufobj_invalbuf() out of vinvalbuf() and make g_vfs call it on close. This is not yet a generally safe function, but for this very specific use it is safe. This solves the problem with buffers not being flushed by unmount or after failed mount attempts.
2005-02-19 11:44:57 +00:00
ASSERT_VOP_LOCKED(vp, "vinvalbuf");
Remove the struct thread unuseful argument from bufobj interface. In particular following functions KPI results modified: - bufobj_invalbuf() - bufsync() and BO_SYNC() "virtual method" of the buffer objects set. Main consumers of bufobj functions are affected by this change too and, in particular, functions which changed their KPI are: - vinvalbuf() - g_vfs_close() Due to the KPI breakage, __FreeBSD_version will be bumped in a later commit. As a side note, please consider just temporary the 'curthread' argument passing to VOP_SYNC() (in bufsync()) as it will be axed out ASAP Reviewed by: kib Tested by: Giovanni Trematerra <giovanni dot trematerra at gmail dot com>
2008-10-10 21:23:50 +00:00
return (bufobj_invalbuf(&vp->v_bufobj, flags, slpflag, slptimeo));
Try to unbreak the vnode locking around vop_reclaim() (based mostly on patch from kan@). Pull bufobj_invalbuf() out of vinvalbuf() and make g_vfs call it on close. This is not yet a generally safe function, but for this very specific use it is safe. This solves the problem with buffers not being flushed by unmount or after failed mount attempts.
2005-02-19 11:44:57 +00:00
}
Add support to UFS2 to provide storage for extended attributes. As this code is not actually used by any of the existing interfaces, it seems unlikely to break anything (famous last words). The internal kernel interface to manipulate these attributes is invoked using two new IO_ flags: IO_NORMAL and IO_EXT. These flags may be specified in the ioflags word of VOP_READ, VOP_WRITE, and VOP_TRUNCATE. Specifying IO_NORMAL means that you want to do I/O to the normal data part of the file and IO_EXT means that you want to do I/O to the extended attributes part of the file. IO_NORMAL and IO_EXT are mutually exclusive for VOP_READ and VOP_WRITE, but may be specified individually or together in the case of VOP_TRUNCATE. For example, when removing a file, VOP_TRUNCATE is called with both IO_NORMAL and IO_EXT set. For backward compatibility, if neither IO_NORMAL nor IO_EXT is set, then IO_NORMAL is assumed. Note that the BA_ and IO_ flags have been `merged' so that they may both be used in the same flags word. This merger is possible by assigning the IO_ flags to the low sixteen bits and the BA_ flags the high sixteen bits. This works because the high sixteen bits of the IO_ word is reserved for read-ahead and help with write clustering so will never be used for flags. This merge lets us get away from code of the form: if (ioflags & IO_SYNC) flags |= BA_SYNC; For the future, I have considered adding a new field to the vattr structure, va_extsize. This addition could then be exported through the stat structure to allow applications to find out the size of the extended attribute storage and also would provide a more standard interface for truncating them (via VOP_SETATTR rather than VOP_TRUNCATE). I am also contemplating adding a pathconf parameter (for concreteness, lets call it _PC_MAX_EXTSIZE) which would let an application determine the maximum size of the extended atribute storage. Sponsored by: DARPA & NAI Labs.
2002-07-19 07:29:39 +00:00
/*
* Flush out buffers on the specified list.
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
*
Add support to UFS2 to provide storage for extended attributes. As this code is not actually used by any of the existing interfaces, it seems unlikely to break anything (famous last words). The internal kernel interface to manipulate these attributes is invoked using two new IO_ flags: IO_NORMAL and IO_EXT. These flags may be specified in the ioflags word of VOP_READ, VOP_WRITE, and VOP_TRUNCATE. Specifying IO_NORMAL means that you want to do I/O to the normal data part of the file and IO_EXT means that you want to do I/O to the extended attributes part of the file. IO_NORMAL and IO_EXT are mutually exclusive for VOP_READ and VOP_WRITE, but may be specified individually or together in the case of VOP_TRUNCATE. For example, when removing a file, VOP_TRUNCATE is called with both IO_NORMAL and IO_EXT set. For backward compatibility, if neither IO_NORMAL nor IO_EXT is set, then IO_NORMAL is assumed. Note that the BA_ and IO_ flags have been `merged' so that they may both be used in the same flags word. This merger is possible by assigning the IO_ flags to the low sixteen bits and the BA_ flags the high sixteen bits. This works because the high sixteen bits of the IO_ word is reserved for read-ahead and help with write clustering so will never be used for flags. This merge lets us get away from code of the form: if (ioflags & IO_SYNC) flags |= BA_SYNC; For the future, I have considered adding a new field to the vattr structure, va_extsize. This addition could then be exported through the stat structure to allow applications to find out the size of the extended attribute storage and also would provide a more standard interface for truncating them (via VOP_SETATTR rather than VOP_TRUNCATE). I am also contemplating adding a pathconf parameter (for concreteness, lets call it _PC_MAX_EXTSIZE) which would let an application determine the maximum size of the extended atribute storage. Sponsored by: DARPA & NAI Labs.
2002-07-19 07:29:39 +00:00
*/
static int
Convert remaining functions in vfs_subr.c from K&R prototypes to ANSI C prototypes, as the majority of new functions added have been in this style. Changing prototype style now results in gcc noticing that the implementation of vn_pollrecord() has a 'short' argument instead of 'int' as prototyped in vnode.h, so correct that definition. In practice this didn't matter as only poll flags in the lower 16 bits are used. MFC after: 1 week
2006-01-21 19:42:10 +00:00
flushbuflist( struct bufv *bufv, int flags, struct bufobj *bo, int slpflag,
int slptimeo)
Add support to UFS2 to provide storage for extended attributes. As this code is not actually used by any of the existing interfaces, it seems unlikely to break anything (famous last words). The internal kernel interface to manipulate these attributes is invoked using two new IO_ flags: IO_NORMAL and IO_EXT. These flags may be specified in the ioflags word of VOP_READ, VOP_WRITE, and VOP_TRUNCATE. Specifying IO_NORMAL means that you want to do I/O to the normal data part of the file and IO_EXT means that you want to do I/O to the extended attributes part of the file. IO_NORMAL and IO_EXT are mutually exclusive for VOP_READ and VOP_WRITE, but may be specified individually or together in the case of VOP_TRUNCATE. For example, when removing a file, VOP_TRUNCATE is called with both IO_NORMAL and IO_EXT set. For backward compatibility, if neither IO_NORMAL nor IO_EXT is set, then IO_NORMAL is assumed. Note that the BA_ and IO_ flags have been `merged' so that they may both be used in the same flags word. This merger is possible by assigning the IO_ flags to the low sixteen bits and the BA_ flags the high sixteen bits. This works because the high sixteen bits of the IO_ word is reserved for read-ahead and help with write clustering so will never be used for flags. This merge lets us get away from code of the form: if (ioflags & IO_SYNC) flags |= BA_SYNC; For the future, I have considered adding a new field to the vattr structure, va_extsize. This addition could then be exported through the stat structure to allow applications to find out the size of the extended attribute storage and also would provide a more standard interface for truncating them (via VOP_SETATTR rather than VOP_TRUNCATE). I am also contemplating adding a pathconf parameter (for concreteness, lets call it _PC_MAX_EXTSIZE) which would let an application determine the maximum size of the extended atribute storage. Sponsored by: DARPA & NAI Labs.
2002-07-19 07:29:39 +00:00
{
struct buf *bp, *nbp;
Fix a bug I introduced in 1.561 which has caused considerable filesystem unhappiness lately. As far as I can tell, no files that have made it safely to disk have been endangered, but stuff in transit has been in peril. Pointy hat: phk
2005-01-16 21:09:39 +00:00
int retval, error;
Break out of loop if next buffer pointer has become invalid while flushing current buffer. Reviewed by: kan
2005-09-16 18:28:12 +00:00
daddr_t lblkno;
b_xflags_t xflags;
Add support to UFS2 to provide storage for extended attributes. As this code is not actually used by any of the existing interfaces, it seems unlikely to break anything (famous last words). The internal kernel interface to manipulate these attributes is invoked using two new IO_ flags: IO_NORMAL and IO_EXT. These flags may be specified in the ioflags word of VOP_READ, VOP_WRITE, and VOP_TRUNCATE. Specifying IO_NORMAL means that you want to do I/O to the normal data part of the file and IO_EXT means that you want to do I/O to the extended attributes part of the file. IO_NORMAL and IO_EXT are mutually exclusive for VOP_READ and VOP_WRITE, but may be specified individually or together in the case of VOP_TRUNCATE. For example, when removing a file, VOP_TRUNCATE is called with both IO_NORMAL and IO_EXT set. For backward compatibility, if neither IO_NORMAL nor IO_EXT is set, then IO_NORMAL is assumed. Note that the BA_ and IO_ flags have been `merged' so that they may both be used in the same flags word. This merger is possible by assigning the IO_ flags to the low sixteen bits and the BA_ flags the high sixteen bits. This works because the high sixteen bits of the IO_ word is reserved for read-ahead and help with write clustering so will never be used for flags. This merge lets us get away from code of the form: if (ioflags & IO_SYNC) flags |= BA_SYNC; For the future, I have considered adding a new field to the vattr structure, va_extsize. This addition could then be exported through the stat structure to allow applications to find out the size of the extended attribute storage and also would provide a more standard interface for truncating them (via VOP_SETATTR rather than VOP_TRUNCATE). I am also contemplating adding a pathconf parameter (for concreteness, lets call it _PC_MAX_EXTSIZE) which would let an application determine the maximum size of the extended atribute storage. Sponsored by: DARPA & NAI Labs.
2002-07-19 07:29:39 +00:00
More vnode -> bufobj migration.
2005-01-11 10:16:39 +00:00
ASSERT_BO_LOCKED(bo);
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
Fix a bug I introduced in 1.561 which has caused considerable filesystem unhappiness lately. As far as I can tell, no files that have made it safely to disk have been endangered, but stuff in transit has been in peril. Pointy hat: phk
2005-01-16 21:09:39 +00:00
retval = 0;
Give flushbuflist() a struct bufv as first argument and avoid home-rolling TAILQ_FOREACH_SAFE(). Loose the error pointer argument and return any errors the normal way. Return EAGAIN for the case where more work needs to be done.
2005-01-11 10:01:54 +00:00
TAILQ_FOREACH_SAFE(bp, &bufv->bv_hd, b_bobufs, nbp) {
Add support to UFS2 to provide storage for extended attributes. As this code is not actually used by any of the existing interfaces, it seems unlikely to break anything (famous last words). The internal kernel interface to manipulate these attributes is invoked using two new IO_ flags: IO_NORMAL and IO_EXT. These flags may be specified in the ioflags word of VOP_READ, VOP_WRITE, and VOP_TRUNCATE. Specifying IO_NORMAL means that you want to do I/O to the normal data part of the file and IO_EXT means that you want to do I/O to the extended attributes part of the file. IO_NORMAL and IO_EXT are mutually exclusive for VOP_READ and VOP_WRITE, but may be specified individually or together in the case of VOP_TRUNCATE. For example, when removing a file, VOP_TRUNCATE is called with both IO_NORMAL and IO_EXT set. For backward compatibility, if neither IO_NORMAL nor IO_EXT is set, then IO_NORMAL is assumed. Note that the BA_ and IO_ flags have been `merged' so that they may both be used in the same flags word. This merger is possible by assigning the IO_ flags to the low sixteen bits and the BA_ flags the high sixteen bits. This works because the high sixteen bits of the IO_ word is reserved for read-ahead and help with write clustering so will never be used for flags. This merge lets us get away from code of the form: if (ioflags & IO_SYNC) flags |= BA_SYNC; For the future, I have considered adding a new field to the vattr structure, va_extsize. This addition could then be exported through the stat structure to allow applications to find out the size of the extended attribute storage and also would provide a more standard interface for truncating them (via VOP_SETATTR rather than VOP_TRUNCATE). I am also contemplating adding a pathconf parameter (for concreteness, lets call it _PC_MAX_EXTSIZE) which would let an application determine the maximum size of the extended atribute storage. Sponsored by: DARPA & NAI Labs.
2002-07-19 07:29:39 +00:00
if (((flags & V_NORMAL) && (bp->b_xflags & BX_ALTDATA)) ||
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
((flags & V_ALT) && (bp->b_xflags & BX_ALTDATA) == 0)) {
Add support to UFS2 to provide storage for extended attributes. As this code is not actually used by any of the existing interfaces, it seems unlikely to break anything (famous last words). The internal kernel interface to manipulate these attributes is invoked using two new IO_ flags: IO_NORMAL and IO_EXT. These flags may be specified in the ioflags word of VOP_READ, VOP_WRITE, and VOP_TRUNCATE. Specifying IO_NORMAL means that you want to do I/O to the normal data part of the file and IO_EXT means that you want to do I/O to the extended attributes part of the file. IO_NORMAL and IO_EXT are mutually exclusive for VOP_READ and VOP_WRITE, but may be specified individually or together in the case of VOP_TRUNCATE. For example, when removing a file, VOP_TRUNCATE is called with both IO_NORMAL and IO_EXT set. For backward compatibility, if neither IO_NORMAL nor IO_EXT is set, then IO_NORMAL is assumed. Note that the BA_ and IO_ flags have been `merged' so that they may both be used in the same flags word. This merger is possible by assigning the IO_ flags to the low sixteen bits and the BA_ flags the high sixteen bits. This works because the high sixteen bits of the IO_ word is reserved for read-ahead and help with write clustering so will never be used for flags. This merge lets us get away from code of the form: if (ioflags & IO_SYNC) flags |= BA_SYNC; For the future, I have considered adding a new field to the vattr structure, va_extsize. This addition could then be exported through the stat structure to allow applications to find out the size of the extended attribute storage and also would provide a more standard interface for truncating them (via VOP_SETATTR rather than VOP_TRUNCATE). I am also contemplating adding a pathconf parameter (for concreteness, lets call it _PC_MAX_EXTSIZE) which would let an application determine the maximum size of the extended atribute storage. Sponsored by: DARPA & NAI Labs.
2002-07-19 07:29:39 +00:00
continue;
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
}
Break out of loop if next buffer pointer has become invalid while flushing current buffer. Reviewed by: kan
2005-09-16 18:28:12 +00:00
lblkno = 0;
xflags = 0;
if (nbp != NULL) {
lblkno = nbp->b_lblkno;
xflags = nbp->b_xflags &
(BX_BKGRDMARKER | BX_VNDIRTY | BX_VNCLEAN);
}
Fix a bug I introduced in 1.561 which has caused considerable filesystem unhappiness lately. As far as I can tell, no files that have made it safely to disk have been endangered, but stuff in transit has been in peril. Pointy hat: phk
2005-01-16 21:09:39 +00:00
retval = EAGAIN;
- Add an interlock argument to BUF_LOCK and BUF_TIMELOCK. - Remove the buftimelock mutex and acquire the buf's interlock to protect these fields instead. - Hold the vnode interlock while locking bufs on the clean/dirty queues. This reduces some cases from one BUF_LOCK with a LK_NOWAIT and another BUF_LOCK with a LK_TIMEFAIL to a single lock. Reviewed by: arch, mckusick
2003-02-25 03:37:48 +00:00
error = BUF_TIMELOCK(bp,
More vnode -> bufobj migration.
2005-01-11 10:16:39 +00:00
LK_EXCLUSIVE | LK_SLEEPFAIL | LK_INTERLOCK, BO_MTX(bo),
- Add an interlock argument to BUF_LOCK and BUF_TIMELOCK. - Remove the buftimelock mutex and acquire the buf's interlock to protect these fields instead. - Hold the vnode interlock while locking bufs on the clean/dirty queues. This reduces some cases from one BUF_LOCK with a LK_NOWAIT and another BUF_LOCK with a LK_TIMEFAIL to a single lock. Reviewed by: arch, mckusick
2003-02-25 03:37:48 +00:00
"flushbuf", slpflag, slptimeo);
if (error) {
More vnode -> bufobj migration.
2005-01-11 10:16:39 +00:00
BO_LOCK(bo);
Give flushbuflist() a struct bufv as first argument and avoid home-rolling TAILQ_FOREACH_SAFE(). Loose the error pointer argument and return any errors the normal way. Return EAGAIN for the case where more work needs to be done.
2005-01-11 10:01:54 +00:00
return (error != ENOLCK ? error : EAGAIN);
Add support to UFS2 to provide storage for extended attributes. As this code is not actually used by any of the existing interfaces, it seems unlikely to break anything (famous last words). The internal kernel interface to manipulate these attributes is invoked using two new IO_ flags: IO_NORMAL and IO_EXT. These flags may be specified in the ioflags word of VOP_READ, VOP_WRITE, and VOP_TRUNCATE. Specifying IO_NORMAL means that you want to do I/O to the normal data part of the file and IO_EXT means that you want to do I/O to the extended attributes part of the file. IO_NORMAL and IO_EXT are mutually exclusive for VOP_READ and VOP_WRITE, but may be specified individually or together in the case of VOP_TRUNCATE. For example, when removing a file, VOP_TRUNCATE is called with both IO_NORMAL and IO_EXT set. For backward compatibility, if neither IO_NORMAL nor IO_EXT is set, then IO_NORMAL is assumed. Note that the BA_ and IO_ flags have been `merged' so that they may both be used in the same flags word. This merger is possible by assigning the IO_ flags to the low sixteen bits and the BA_ flags the high sixteen bits. This works because the high sixteen bits of the IO_ word is reserved for read-ahead and help with write clustering so will never be used for flags. This merge lets us get away from code of the form: if (ioflags & IO_SYNC) flags |= BA_SYNC; For the future, I have considered adding a new field to the vattr structure, va_extsize. This addition could then be exported through the stat structure to allow applications to find out the size of the extended attribute storage and also would provide a more standard interface for truncating them (via VOP_SETATTR rather than VOP_TRUNCATE). I am also contemplating adding a pathconf parameter (for concreteness, lets call it _PC_MAX_EXTSIZE) which would let an application determine the maximum size of the extended atribute storage. Sponsored by: DARPA & NAI Labs.
2002-07-19 07:29:39 +00:00
}
- Assert that the bufobj matches in flushbuflists. I still haven't gotten to root cause on exactly how this happens. - If the assert is disabled, we presently try to handle this case, but the BUF_UNLOCK was missing. Thus, if this condition ever hit we would leak a buf lock. Many thanks to Peter Holm for all his help in finding this bug. He really put more effort into it than I did.
2005-04-06 06:49:46 +00:00
KASSERT(bp->b_bufobj == bo,
Minor style cleanups (mostly removal of trailing whitespaces).
2007-04-10 15:29:37 +00:00
("bp %p wrong b_bufobj %p should be %p",
- In reassignbuf() add many asserts to validate the head and tail pointers of the clean and dirty lists. This is in an attempt to catch the wrong bufobj problem sooner. - In vgonel() don't acquire an extra reference in the active case, the vnode lock and VI_DOOMED protect us from recursively cleaning. - Also in vgonel() clean up some stale comments. Sponsored by: Isilon Systems, Inc. Approved by: re (blanket vfs)
2005-06-14 20:31:53 +00:00
bp, bp->b_bufobj, bo));
Try to unbreak the vnode locking around vop_reclaim() (based mostly on patch from kan@). Pull bufobj_invalbuf() out of vinvalbuf() and make g_vfs call it on close. This is not yet a generally safe function, but for this very specific use it is safe. This solves the problem with buffers not being flushed by unmount or after failed mount attempts.
2005-02-19 11:44:57 +00:00
if (bp->b_bufobj != bo) { /* XXX: necessary ? */
- Assert that the bufobj matches in flushbuflists. I still haven't gotten to root cause on exactly how this happens. - If the assert is disabled, we presently try to handle this case, but the BUF_UNLOCK was missing. Thus, if this condition ever hit we would leak a buf lock. Many thanks to Peter Holm for all his help in finding this bug. He really put more effort into it than I did.
2005-04-06 06:49:46 +00:00
BUF_UNLOCK(bp);
Try to unbreak the vnode locking around vop_reclaim() (based mostly on patch from kan@). Pull bufobj_invalbuf() out of vinvalbuf() and make g_vfs call it on close. This is not yet a generally safe function, but for this very specific use it is safe. This solves the problem with buffers not being flushed by unmount or after failed mount attempts.
2005-02-19 11:44:57 +00:00
BO_LOCK(bo);
return (EAGAIN);
}
Add support to UFS2 to provide storage for extended attributes. As this code is not actually used by any of the existing interfaces, it seems unlikely to break anything (famous last words). The internal kernel interface to manipulate these attributes is invoked using two new IO_ flags: IO_NORMAL and IO_EXT. These flags may be specified in the ioflags word of VOP_READ, VOP_WRITE, and VOP_TRUNCATE. Specifying IO_NORMAL means that you want to do I/O to the normal data part of the file and IO_EXT means that you want to do I/O to the extended attributes part of the file. IO_NORMAL and IO_EXT are mutually exclusive for VOP_READ and VOP_WRITE, but may be specified individually or together in the case of VOP_TRUNCATE. For example, when removing a file, VOP_TRUNCATE is called with both IO_NORMAL and IO_EXT set. For backward compatibility, if neither IO_NORMAL nor IO_EXT is set, then IO_NORMAL is assumed. Note that the BA_ and IO_ flags have been `merged' so that they may both be used in the same flags word. This merger is possible by assigning the IO_ flags to the low sixteen bits and the BA_ flags the high sixteen bits. This works because the high sixteen bits of the IO_ word is reserved for read-ahead and help with write clustering so will never be used for flags. This merge lets us get away from code of the form: if (ioflags & IO_SYNC) flags |= BA_SYNC; For the future, I have considered adding a new field to the vattr structure, va_extsize. This addition could then be exported through the stat structure to allow applications to find out the size of the extended attribute storage and also would provide a more standard interface for truncating them (via VOP_SETATTR rather than VOP_TRUNCATE). I am also contemplating adding a pathconf parameter (for concreteness, lets call it _PC_MAX_EXTSIZE) which would let an application determine the maximum size of the extended atribute storage. Sponsored by: DARPA & NAI Labs.
2002-07-19 07:29:39 +00:00
/*
* XXX Since there are no node locks for NFS, I
* believe there is a slight chance that a delayed
* write will occur while sleeping just above, so
Try to unbreak the vnode locking around vop_reclaim() (based mostly on patch from kan@). Pull bufobj_invalbuf() out of vinvalbuf() and make g_vfs call it on close. This is not yet a generally safe function, but for this very specific use it is safe. This solves the problem with buffers not being flushed by unmount or after failed mount attempts.
2005-02-19 11:44:57 +00:00
* check for it.
Add support to UFS2 to provide storage for extended attributes. As this code is not actually used by any of the existing interfaces, it seems unlikely to break anything (famous last words). The internal kernel interface to manipulate these attributes is invoked using two new IO_ flags: IO_NORMAL and IO_EXT. These flags may be specified in the ioflags word of VOP_READ, VOP_WRITE, and VOP_TRUNCATE. Specifying IO_NORMAL means that you want to do I/O to the normal data part of the file and IO_EXT means that you want to do I/O to the extended attributes part of the file. IO_NORMAL and IO_EXT are mutually exclusive for VOP_READ and VOP_WRITE, but may be specified individually or together in the case of VOP_TRUNCATE. For example, when removing a file, VOP_TRUNCATE is called with both IO_NORMAL and IO_EXT set. For backward compatibility, if neither IO_NORMAL nor IO_EXT is set, then IO_NORMAL is assumed. Note that the BA_ and IO_ flags have been `merged' so that they may both be used in the same flags word. This merger is possible by assigning the IO_ flags to the low sixteen bits and the BA_ flags the high sixteen bits. This works because the high sixteen bits of the IO_ word is reserved for read-ahead and help with write clustering so will never be used for flags. This merge lets us get away from code of the form: if (ioflags & IO_SYNC) flags |= BA_SYNC; For the future, I have considered adding a new field to the vattr structure, va_extsize. This addition could then be exported through the stat structure to allow applications to find out the size of the extended attribute storage and also would provide a more standard interface for truncating them (via VOP_SETATTR rather than VOP_TRUNCATE). I am also contemplating adding a pathconf parameter (for concreteness, lets call it _PC_MAX_EXTSIZE) which would let an application determine the maximum size of the extended atribute storage. Sponsored by: DARPA & NAI Labs.
2002-07-19 07:29:39 +00:00
*/
if (((bp->b_flags & (B_DELWRI | B_INVAL)) == B_DELWRI) &&
Try to unbreak the vnode locking around vop_reclaim() (based mostly on patch from kan@). Pull bufobj_invalbuf() out of vinvalbuf() and make g_vfs call it on close. This is not yet a generally safe function, but for this very specific use it is safe. This solves the problem with buffers not being flushed by unmount or after failed mount attempts.
2005-02-19 11:44:57 +00:00
(flags & V_SAVE)) {
bremfree(bp);
bp->b_flags |= B_ASYNC;
bwrite(bp);
More vnode -> bufobj migration.
2005-01-11 10:16:39 +00:00
BO_LOCK(bo);
Try to unbreak the vnode locking around vop_reclaim() (based mostly on patch from kan@). Pull bufobj_invalbuf() out of vinvalbuf() and make g_vfs call it on close. This is not yet a generally safe function, but for this very specific use it is safe. This solves the problem with buffers not being flushed by unmount or after failed mount attempts.
2005-02-19 11:44:57 +00:00
return (EAGAIN); /* XXX: why not loop ? */
Add support to UFS2 to provide storage for extended attributes. As this code is not actually used by any of the existing interfaces, it seems unlikely to break anything (famous last words). The internal kernel interface to manipulate these attributes is invoked using two new IO_ flags: IO_NORMAL and IO_EXT. These flags may be specified in the ioflags word of VOP_READ, VOP_WRITE, and VOP_TRUNCATE. Specifying IO_NORMAL means that you want to do I/O to the normal data part of the file and IO_EXT means that you want to do I/O to the extended attributes part of the file. IO_NORMAL and IO_EXT are mutually exclusive for VOP_READ and VOP_WRITE, but may be specified individually or together in the case of VOP_TRUNCATE. For example, when removing a file, VOP_TRUNCATE is called with both IO_NORMAL and IO_EXT set. For backward compatibility, if neither IO_NORMAL nor IO_EXT is set, then IO_NORMAL is assumed. Note that the BA_ and IO_ flags have been `merged' so that they may both be used in the same flags word. This merger is possible by assigning the IO_ flags to the low sixteen bits and the BA_ flags the high sixteen bits. This works because the high sixteen bits of the IO_ word is reserved for read-ahead and help with write clustering so will never be used for flags. This merge lets us get away from code of the form: if (ioflags & IO_SYNC) flags |= BA_SYNC; For the future, I have considered adding a new field to the vattr structure, va_extsize. This addition could then be exported through the stat structure to allow applications to find out the size of the extended attribute storage and also would provide a more standard interface for truncating them (via VOP_SETATTR rather than VOP_TRUNCATE). I am also contemplating adding a pathconf parameter (for concreteness, lets call it _PC_MAX_EXTSIZE) which would let an application determine the maximum size of the extended atribute storage. Sponsored by: DARPA & NAI Labs.
2002-07-19 07:29:39 +00:00
}
bremfree(bp);
Do not set B_NOCACHE on buffers when releasing them in flushbuflist(). If B_NOCACHE is set the pages of vm backed buffers will be invalidated. However clean buffers can be backed by dirty VM pages so invalidating them can lead to data loss. Add support for flush dirty page in the data invalidation function of some network file systems. This fixes data losses during vnode recycling (and other code paths using invalbuf(*,V_SAVE,*,*)) for data written using an mmaped file. Collaborative effort by: jhb@,mohans@,peter@,ps@,ups@ Reviewed by: tegge@ MFC after: 7 days
2006-05-25 01:00:35 +00:00
bp->b_flags |= (B_INVAL | B_RELBUF);
Add support to UFS2 to provide storage for extended attributes. As this code is not actually used by any of the existing interfaces, it seems unlikely to break anything (famous last words). The internal kernel interface to manipulate these attributes is invoked using two new IO_ flags: IO_NORMAL and IO_EXT. These flags may be specified in the ioflags word of VOP_READ, VOP_WRITE, and VOP_TRUNCATE. Specifying IO_NORMAL means that you want to do I/O to the normal data part of the file and IO_EXT means that you want to do I/O to the extended attributes part of the file. IO_NORMAL and IO_EXT are mutually exclusive for VOP_READ and VOP_WRITE, but may be specified individually or together in the case of VOP_TRUNCATE. For example, when removing a file, VOP_TRUNCATE is called with both IO_NORMAL and IO_EXT set. For backward compatibility, if neither IO_NORMAL nor IO_EXT is set, then IO_NORMAL is assumed. Note that the BA_ and IO_ flags have been `merged' so that they may both be used in the same flags word. This merger is possible by assigning the IO_ flags to the low sixteen bits and the BA_ flags the high sixteen bits. This works because the high sixteen bits of the IO_ word is reserved for read-ahead and help with write clustering so will never be used for flags. This merge lets us get away from code of the form: if (ioflags & IO_SYNC) flags |= BA_SYNC; For the future, I have considered adding a new field to the vattr structure, va_extsize. This addition could then be exported through the stat structure to allow applications to find out the size of the extended attribute storage and also would provide a more standard interface for truncating them (via VOP_SETATTR rather than VOP_TRUNCATE). I am also contemplating adding a pathconf parameter (for concreteness, lets call it _PC_MAX_EXTSIZE) which would let an application determine the maximum size of the extended atribute storage. Sponsored by: DARPA & NAI Labs.
2002-07-19 07:29:39 +00:00
bp->b_flags &= ~B_ASYNC;
brelse(bp);
More vnode -> bufobj migration.
2005-01-11 10:16:39 +00:00
BO_LOCK(bo);
Break out of loop if next buffer pointer has become invalid while flushing current buffer. Reviewed by: kan
2005-09-16 18:28:12 +00:00
if (nbp != NULL &&
Minor style cleanups (mostly removal of trailing whitespaces).
2007-04-10 15:29:37 +00:00
(nbp->b_bufobj != bo ||
Break out of loop if next buffer pointer has become invalid while flushing current buffer. Reviewed by: kan
2005-09-16 18:28:12 +00:00
nbp->b_lblkno != lblkno ||
(nbp->b_xflags &
(BX_BKGRDMARKER | BX_VNDIRTY | BX_VNCLEAN)) != xflags))
break; /* nbp invalid */
Add support to UFS2 to provide storage for extended attributes. As this code is not actually used by any of the existing interfaces, it seems unlikely to break anything (famous last words). The internal kernel interface to manipulate these attributes is invoked using two new IO_ flags: IO_NORMAL and IO_EXT. These flags may be specified in the ioflags word of VOP_READ, VOP_WRITE, and VOP_TRUNCATE. Specifying IO_NORMAL means that you want to do I/O to the normal data part of the file and IO_EXT means that you want to do I/O to the extended attributes part of the file. IO_NORMAL and IO_EXT are mutually exclusive for VOP_READ and VOP_WRITE, but may be specified individually or together in the case of VOP_TRUNCATE. For example, when removing a file, VOP_TRUNCATE is called with both IO_NORMAL and IO_EXT set. For backward compatibility, if neither IO_NORMAL nor IO_EXT is set, then IO_NORMAL is assumed. Note that the BA_ and IO_ flags have been `merged' so that they may both be used in the same flags word. This merger is possible by assigning the IO_ flags to the low sixteen bits and the BA_ flags the high sixteen bits. This works because the high sixteen bits of the IO_ word is reserved for read-ahead and help with write clustering so will never be used for flags. This merge lets us get away from code of the form: if (ioflags & IO_SYNC) flags |= BA_SYNC; For the future, I have considered adding a new field to the vattr structure, va_extsize. This addition could then be exported through the stat structure to allow applications to find out the size of the extended attribute storage and also would provide a more standard interface for truncating them (via VOP_SETATTR rather than VOP_TRUNCATE). I am also contemplating adding a pathconf parameter (for concreteness, lets call it _PC_MAX_EXTSIZE) which would let an application determine the maximum size of the extended atribute storage. Sponsored by: DARPA & NAI Labs.
2002-07-19 07:29:39 +00:00
}
Fix a bug I introduced in 1.561 which has caused considerable filesystem unhappiness lately. As far as I can tell, no files that have made it safely to disk have been endangered, but stuff in transit has been in peril. Pointy hat: phk
2005-01-16 21:09:39 +00:00
return (retval);
Add support to UFS2 to provide storage for extended attributes. As this code is not actually used by any of the existing interfaces, it seems unlikely to break anything (famous last words). The internal kernel interface to manipulate these attributes is invoked using two new IO_ flags: IO_NORMAL and IO_EXT. These flags may be specified in the ioflags word of VOP_READ, VOP_WRITE, and VOP_TRUNCATE. Specifying IO_NORMAL means that you want to do I/O to the normal data part of the file and IO_EXT means that you want to do I/O to the extended attributes part of the file. IO_NORMAL and IO_EXT are mutually exclusive for VOP_READ and VOP_WRITE, but may be specified individually or together in the case of VOP_TRUNCATE. For example, when removing a file, VOP_TRUNCATE is called with both IO_NORMAL and IO_EXT set. For backward compatibility, if neither IO_NORMAL nor IO_EXT is set, then IO_NORMAL is assumed. Note that the BA_ and IO_ flags have been `merged' so that they may both be used in the same flags word. This merger is possible by assigning the IO_ flags to the low sixteen bits and the BA_ flags the high sixteen bits. This works because the high sixteen bits of the IO_ word is reserved for read-ahead and help with write clustering so will never be used for flags. This merge lets us get away from code of the form: if (ioflags & IO_SYNC) flags |= BA_SYNC; For the future, I have considered adding a new field to the vattr structure, va_extsize. This addition could then be exported through the stat structure to allow applications to find out the size of the extended attribute storage and also would provide a more standard interface for truncating them (via VOP_SETATTR rather than VOP_TRUNCATE). I am also contemplating adding a pathconf parameter (for concreteness, lets call it _PC_MAX_EXTSIZE) which would let an application determine the maximum size of the extended atribute storage. Sponsored by: DARPA & NAI Labs.
2002-07-19 07:29:39 +00:00
}
Some VM improvements, including elimination of alot of Sig-11 problems. Tor Egge and others have helped with various VM bugs lately, but don't blame him -- blame me!!! pmap.c: 1) Create an object for kernel page table allocations. This fixes a bogus allocation method previously used for such, by grabbing pages from the kernel object, using bogus pindexes. (This was a code cleanup, and perhaps a minor system stability issue.) pmap.c: 2) Pre-set the modify and accessed bits when prudent. This will decrease bus traffic under certain circumstances. vfs_bio.c, vfs_cluster.c: 3) Rather than calculating the beginning virtual byte offset multiple times, stick the offset into the buffer header, so that the calculated offset can be reused. (Long long multiplies are often expensive, and this is a probably unmeasurable performance improvement, and code cleanup.) vfs_bio.c: 4) Handle write recursion more intelligently (but not perfectly) so that it is less likely to cause a system panic, and is also much more robust. vfs_bio.c: 5) getblk incorrectly wrote out blocks that are incorrectly sized. The problem is fixed, and writes blocks out ONLY when B_DELWRI is true. vfs_bio.c: 6) Check that already constituted buffers have fully valid pages. If not, then make sure that the B_CACHE bit is not set. (This was a major source of Sig-11 type problems.) vfs_bio.c: 7) Fix a potential system deadlock due to an incorrectly specified sleep priority while waiting for a buffer write operation. The change that I made opens the system up to serious problems, and we need to examine the issue of process sleep priorities. vfs_cluster.c, vfs_bio.c: 8) Make clustered reads work more correctly (and more completely) when buffers are already constituted, but not fully valid. (This was another system reliability issue.) vfs_subr.c, ffs_inode.c: 9) Create a vtruncbuf function, which is used by filesystems that can truncate files. The vinvalbuf forced a file sync type operation, while vtruncbuf only invalidates the buffers past the new end of file, and also invalidates the appropriate pages. (This was a system reliabiliy and performance issue.) 10) Modify FFS to use vtruncbuf. vm_object.c: 11) Make the object rundown mechanism for OBJT_VNODE type objects work more correctly. Included in that fix, create pager entries for the OBJT_DEAD pager type, so that paging requests that might slip in during race conditions are properly handled. (This was a system reliability issue.) vm_page.c: 12) Make some of the page validation routines be a little less picky about arguments passed to them. Also, support page invalidation change the object generation count so that we handle generation counts a little more robustly. vm_pageout.c: 13) Further reduce pageout daemon activity when the system doesn't need help from it. There should be no additional performance decrease even when the pageout daemon is running. (This was a significant performance issue.) vnode_pager.c: 14) Teach the vnode pager to handle race conditions during vnode deallocations.
1998-03-16 01:56:03 +00:00
/*
* Truncate a file's buffer and pages to a specified length. This
* is in lieu of the old vinvalbuf mechanism, which performed unneeded
* sync activity.
*/
int
Convert remaining functions in vfs_subr.c from K&R prototypes to ANSI C prototypes, as the majority of new functions added have been in this style. Changing prototype style now results in gcc noticing that the implementation of vn_pollrecord() has a 'short' argument instead of 'int' as prototyped in vnode.h, so correct that definition. In practice this didn't matter as only poll flags in the lower 16 bits are used. MFC after: 1 week
2006-01-21 19:42:10 +00:00
vtruncbuf(struct vnode *vp, struct ucred *cred, struct thread *td,
off_t length, int blksize)
Some VM improvements, including elimination of alot of Sig-11 problems. Tor Egge and others have helped with various VM bugs lately, but don't blame him -- blame me!!! pmap.c: 1) Create an object for kernel page table allocations. This fixes a bogus allocation method previously used for such, by grabbing pages from the kernel object, using bogus pindexes. (This was a code cleanup, and perhaps a minor system stability issue.) pmap.c: 2) Pre-set the modify and accessed bits when prudent. This will decrease bus traffic under certain circumstances. vfs_bio.c, vfs_cluster.c: 3) Rather than calculating the beginning virtual byte offset multiple times, stick the offset into the buffer header, so that the calculated offset can be reused. (Long long multiplies are often expensive, and this is a probably unmeasurable performance improvement, and code cleanup.) vfs_bio.c: 4) Handle write recursion more intelligently (but not perfectly) so that it is less likely to cause a system panic, and is also much more robust. vfs_bio.c: 5) getblk incorrectly wrote out blocks that are incorrectly sized. The problem is fixed, and writes blocks out ONLY when B_DELWRI is true. vfs_bio.c: 6) Check that already constituted buffers have fully valid pages. If not, then make sure that the B_CACHE bit is not set. (This was a major source of Sig-11 type problems.) vfs_bio.c: 7) Fix a potential system deadlock due to an incorrectly specified sleep priority while waiting for a buffer write operation. The change that I made opens the system up to serious problems, and we need to examine the issue of process sleep priorities. vfs_cluster.c, vfs_bio.c: 8) Make clustered reads work more correctly (and more completely) when buffers are already constituted, but not fully valid. (This was another system reliability issue.) vfs_subr.c, ffs_inode.c: 9) Create a vtruncbuf function, which is used by filesystems that can truncate files. The vinvalbuf forced a file sync type operation, while vtruncbuf only invalidates the buffers past the new end of file, and also invalidates the appropriate pages. (This was a system reliabiliy and performance issue.) 10) Modify FFS to use vtruncbuf. vm_object.c: 11) Make the object rundown mechanism for OBJT_VNODE type objects work more correctly. Included in that fix, create pager entries for the OBJT_DEAD pager type, so that paging requests that might slip in during race conditions are properly handled. (This was a system reliability issue.) vm_page.c: 12) Make some of the page validation routines be a little less picky about arguments passed to them. Also, support page invalidation change the object generation count so that we handle generation counts a little more robustly. vm_pageout.c: 13) Further reduce pageout daemon activity when the system doesn't need help from it. There should be no additional performance decrease even when the pageout daemon is running. (This was a significant performance issue.) vnode_pager.c: 14) Teach the vnode pager to handle race conditions during vnode deallocations.
1998-03-16 01:56:03 +00:00
{
Polish vtruncbuf() to improve readability and style a bit.
2004-10-21 14:13:54 +00:00
struct buf *bp, *nbp;
- Remove spls(). The locking that has replaced them is in place and they no longer serve as guidelines for future work.
2003-09-19 23:52:06 +00:00
int anyfreed;
Some VM improvements, including elimination of alot of Sig-11 problems. Tor Egge and others have helped with various VM bugs lately, but don't blame him -- blame me!!! pmap.c: 1) Create an object for kernel page table allocations. This fixes a bogus allocation method previously used for such, by grabbing pages from the kernel object, using bogus pindexes. (This was a code cleanup, and perhaps a minor system stability issue.) pmap.c: 2) Pre-set the modify and accessed bits when prudent. This will decrease bus traffic under certain circumstances. vfs_bio.c, vfs_cluster.c: 3) Rather than calculating the beginning virtual byte offset multiple times, stick the offset into the buffer header, so that the calculated offset can be reused. (Long long multiplies are often expensive, and this is a probably unmeasurable performance improvement, and code cleanup.) vfs_bio.c: 4) Handle write recursion more intelligently (but not perfectly) so that it is less likely to cause a system panic, and is also much more robust. vfs_bio.c: 5) getblk incorrectly wrote out blocks that are incorrectly sized. The problem is fixed, and writes blocks out ONLY when B_DELWRI is true. vfs_bio.c: 6) Check that already constituted buffers have fully valid pages. If not, then make sure that the B_CACHE bit is not set. (This was a major source of Sig-11 type problems.) vfs_bio.c: 7) Fix a potential system deadlock due to an incorrectly specified sleep priority while waiting for a buffer write operation. The change that I made opens the system up to serious problems, and we need to examine the issue of process sleep priorities. vfs_cluster.c, vfs_bio.c: 8) Make clustered reads work more correctly (and more completely) when buffers are already constituted, but not fully valid. (This was another system reliability issue.) vfs_subr.c, ffs_inode.c: 9) Create a vtruncbuf function, which is used by filesystems that can truncate files. The vinvalbuf forced a file sync type operation, while vtruncbuf only invalidates the buffers past the new end of file, and also invalidates the appropriate pages. (This was a system reliabiliy and performance issue.) 10) Modify FFS to use vtruncbuf. vm_object.c: 11) Make the object rundown mechanism for OBJT_VNODE type objects work more correctly. Included in that fix, create pager entries for the OBJT_DEAD pager type, so that paging requests that might slip in during race conditions are properly handled. (This was a system reliability issue.) vm_page.c: 12) Make some of the page validation routines be a little less picky about arguments passed to them. Also, support page invalidation change the object generation count so that we handle generation counts a little more robustly. vm_pageout.c: 13) Further reduce pageout daemon activity when the system doesn't need help from it. There should be no additional performance decrease even when the pageout daemon is running. (This was a significant performance issue.) vnode_pager.c: 14) Teach the vnode pager to handle race conditions during vnode deallocations.
1998-03-16 01:56:03 +00:00
int trunclbn;
Polish vtruncbuf() to improve readability and style a bit.
2004-10-21 14:13:54 +00:00
struct bufobj *bo;
Some VM improvements, including elimination of alot of Sig-11 problems. Tor Egge and others have helped with various VM bugs lately, but don't blame him -- blame me!!! pmap.c: 1) Create an object for kernel page table allocations. This fixes a bogus allocation method previously used for such, by grabbing pages from the kernel object, using bogus pindexes. (This was a code cleanup, and perhaps a minor system stability issue.) pmap.c: 2) Pre-set the modify and accessed bits when prudent. This will decrease bus traffic under certain circumstances. vfs_bio.c, vfs_cluster.c: 3) Rather than calculating the beginning virtual byte offset multiple times, stick the offset into the buffer header, so that the calculated offset can be reused. (Long long multiplies are often expensive, and this is a probably unmeasurable performance improvement, and code cleanup.) vfs_bio.c: 4) Handle write recursion more intelligently (but not perfectly) so that it is less likely to cause a system panic, and is also much more robust. vfs_bio.c: 5) getblk incorrectly wrote out blocks that are incorrectly sized. The problem is fixed, and writes blocks out ONLY when B_DELWRI is true. vfs_bio.c: 6) Check that already constituted buffers have fully valid pages. If not, then make sure that the B_CACHE bit is not set. (This was a major source of Sig-11 type problems.) vfs_bio.c: 7) Fix a potential system deadlock due to an incorrectly specified sleep priority while waiting for a buffer write operation. The change that I made opens the system up to serious problems, and we need to examine the issue of process sleep priorities. vfs_cluster.c, vfs_bio.c: 8) Make clustered reads work more correctly (and more completely) when buffers are already constituted, but not fully valid. (This was another system reliability issue.) vfs_subr.c, ffs_inode.c: 9) Create a vtruncbuf function, which is used by filesystems that can truncate files. The vinvalbuf forced a file sync type operation, while vtruncbuf only invalidates the buffers past the new end of file, and also invalidates the appropriate pages. (This was a system reliabiliy and performance issue.) 10) Modify FFS to use vtruncbuf. vm_object.c: 11) Make the object rundown mechanism for OBJT_VNODE type objects work more correctly. Included in that fix, create pager entries for the OBJT_DEAD pager type, so that paging requests that might slip in during race conditions are properly handled. (This was a system reliability issue.) vm_page.c: 12) Make some of the page validation routines be a little less picky about arguments passed to them. Also, support page invalidation change the object generation count so that we handle generation counts a little more robustly. vm_pageout.c: 13) Further reduce pageout daemon activity when the system doesn't need help from it. There should be no additional performance decrease even when the pageout daemon is running. (This was a significant performance issue.) vnode_pager.c: 14) Teach the vnode pager to handle race conditions during vnode deallocations.
1998-03-16 01:56:03 +00:00
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
CTR5(KTR_VFS, "%s: vp %p with cred %p and block %d:%ju", __func__,
vp, cred, blksize, (uintmax_t)length);
Some VM improvements, including elimination of alot of Sig-11 problems. Tor Egge and others have helped with various VM bugs lately, but don't blame him -- blame me!!! pmap.c: 1) Create an object for kernel page table allocations. This fixes a bogus allocation method previously used for such, by grabbing pages from the kernel object, using bogus pindexes. (This was a code cleanup, and perhaps a minor system stability issue.) pmap.c: 2) Pre-set the modify and accessed bits when prudent. This will decrease bus traffic under certain circumstances. vfs_bio.c, vfs_cluster.c: 3) Rather than calculating the beginning virtual byte offset multiple times, stick the offset into the buffer header, so that the calculated offset can be reused. (Long long multiplies are often expensive, and this is a probably unmeasurable performance improvement, and code cleanup.) vfs_bio.c: 4) Handle write recursion more intelligently (but not perfectly) so that it is less likely to cause a system panic, and is also much more robust. vfs_bio.c: 5) getblk incorrectly wrote out blocks that are incorrectly sized. The problem is fixed, and writes blocks out ONLY when B_DELWRI is true. vfs_bio.c: 6) Check that already constituted buffers have fully valid pages. If not, then make sure that the B_CACHE bit is not set. (This was a major source of Sig-11 type problems.) vfs_bio.c: 7) Fix a potential system deadlock due to an incorrectly specified sleep priority while waiting for a buffer write operation. The change that I made opens the system up to serious problems, and we need to examine the issue of process sleep priorities. vfs_cluster.c, vfs_bio.c: 8) Make clustered reads work more correctly (and more completely) when buffers are already constituted, but not fully valid. (This was another system reliability issue.) vfs_subr.c, ffs_inode.c: 9) Create a vtruncbuf function, which is used by filesystems that can truncate files. The vinvalbuf forced a file sync type operation, while vtruncbuf only invalidates the buffers past the new end of file, and also invalidates the appropriate pages. (This was a system reliabiliy and performance issue.) 10) Modify FFS to use vtruncbuf. vm_object.c: 11) Make the object rundown mechanism for OBJT_VNODE type objects work more correctly. Included in that fix, create pager entries for the OBJT_DEAD pager type, so that paging requests that might slip in during race conditions are properly handled. (This was a system reliability issue.) vm_page.c: 12) Make some of the page validation routines be a little less picky about arguments passed to them. Also, support page invalidation change the object generation count so that we handle generation counts a little more robustly. vm_pageout.c: 13) Further reduce pageout daemon activity when the system doesn't need help from it. There should be no additional performance decrease even when the pageout daemon is running. (This was a significant performance issue.) vnode_pager.c: 14) Teach the vnode pager to handle race conditions during vnode deallocations.
1998-03-16 01:56:03 +00:00
/*
* Round up to the *next* lbn.
*/
Fix an embarassing problem in vtruncbuf.
1998-03-19 18:46:58 +00:00
trunclbn = (length + blksize - 1) / blksize;
Some VM improvements, including elimination of alot of Sig-11 problems. Tor Egge and others have helped with various VM bugs lately, but don't blame him -- blame me!!! pmap.c: 1) Create an object for kernel page table allocations. This fixes a bogus allocation method previously used for such, by grabbing pages from the kernel object, using bogus pindexes. (This was a code cleanup, and perhaps a minor system stability issue.) pmap.c: 2) Pre-set the modify and accessed bits when prudent. This will decrease bus traffic under certain circumstances. vfs_bio.c, vfs_cluster.c: 3) Rather than calculating the beginning virtual byte offset multiple times, stick the offset into the buffer header, so that the calculated offset can be reused. (Long long multiplies are often expensive, and this is a probably unmeasurable performance improvement, and code cleanup.) vfs_bio.c: 4) Handle write recursion more intelligently (but not perfectly) so that it is less likely to cause a system panic, and is also much more robust. vfs_bio.c: 5) getblk incorrectly wrote out blocks that are incorrectly sized. The problem is fixed, and writes blocks out ONLY when B_DELWRI is true. vfs_bio.c: 6) Check that already constituted buffers have fully valid pages. If not, then make sure that the B_CACHE bit is not set. (This was a major source of Sig-11 type problems.) vfs_bio.c: 7) Fix a potential system deadlock due to an incorrectly specified sleep priority while waiting for a buffer write operation. The change that I made opens the system up to serious problems, and we need to examine the issue of process sleep priorities. vfs_cluster.c, vfs_bio.c: 8) Make clustered reads work more correctly (and more completely) when buffers are already constituted, but not fully valid. (This was another system reliability issue.) vfs_subr.c, ffs_inode.c: 9) Create a vtruncbuf function, which is used by filesystems that can truncate files. The vinvalbuf forced a file sync type operation, while vtruncbuf only invalidates the buffers past the new end of file, and also invalidates the appropriate pages. (This was a system reliabiliy and performance issue.) 10) Modify FFS to use vtruncbuf. vm_object.c: 11) Make the object rundown mechanism for OBJT_VNODE type objects work more correctly. Included in that fix, create pager entries for the OBJT_DEAD pager type, so that paging requests that might slip in during race conditions are properly handled. (This was a system reliability issue.) vm_page.c: 12) Make some of the page validation routines be a little less picky about arguments passed to them. Also, support page invalidation change the object generation count so that we handle generation counts a little more robustly. vm_pageout.c: 13) Further reduce pageout daemon activity when the system doesn't need help from it. There should be no additional performance decrease even when the pageout daemon is running. (This was a significant performance issue.) vnode_pager.c: 14) Teach the vnode pager to handle race conditions during vnode deallocations.
1998-03-16 01:56:03 +00:00
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
ASSERT_VOP_LOCKED(vp, "vtruncbuf");
Some VM improvements, including elimination of alot of Sig-11 problems. Tor Egge and others have helped with various VM bugs lately, but don't blame him -- blame me!!! pmap.c: 1) Create an object for kernel page table allocations. This fixes a bogus allocation method previously used for such, by grabbing pages from the kernel object, using bogus pindexes. (This was a code cleanup, and perhaps a minor system stability issue.) pmap.c: 2) Pre-set the modify and accessed bits when prudent. This will decrease bus traffic under certain circumstances. vfs_bio.c, vfs_cluster.c: 3) Rather than calculating the beginning virtual byte offset multiple times, stick the offset into the buffer header, so that the calculated offset can be reused. (Long long multiplies are often expensive, and this is a probably unmeasurable performance improvement, and code cleanup.) vfs_bio.c: 4) Handle write recursion more intelligently (but not perfectly) so that it is less likely to cause a system panic, and is also much more robust. vfs_bio.c: 5) getblk incorrectly wrote out blocks that are incorrectly sized. The problem is fixed, and writes blocks out ONLY when B_DELWRI is true. vfs_bio.c: 6) Check that already constituted buffers have fully valid pages. If not, then make sure that the B_CACHE bit is not set. (This was a major source of Sig-11 type problems.) vfs_bio.c: 7) Fix a potential system deadlock due to an incorrectly specified sleep priority while waiting for a buffer write operation. The change that I made opens the system up to serious problems, and we need to examine the issue of process sleep priorities. vfs_cluster.c, vfs_bio.c: 8) Make clustered reads work more correctly (and more completely) when buffers are already constituted, but not fully valid. (This was another system reliability issue.) vfs_subr.c, ffs_inode.c: 9) Create a vtruncbuf function, which is used by filesystems that can truncate files. The vinvalbuf forced a file sync type operation, while vtruncbuf only invalidates the buffers past the new end of file, and also invalidates the appropriate pages. (This was a system reliabiliy and performance issue.) 10) Modify FFS to use vtruncbuf. vm_object.c: 11) Make the object rundown mechanism for OBJT_VNODE type objects work more correctly. Included in that fix, create pager entries for the OBJT_DEAD pager type, so that paging requests that might slip in during race conditions are properly handled. (This was a system reliability issue.) vm_page.c: 12) Make some of the page validation routines be a little less picky about arguments passed to them. Also, support page invalidation change the object generation count so that we handle generation counts a little more robustly. vm_pageout.c: 13) Further reduce pageout daemon activity when the system doesn't need help from it. There should be no additional performance decrease even when the pageout daemon is running. (This was a significant performance issue.) vnode_pager.c: 14) Teach the vnode pager to handle race conditions during vnode deallocations.
1998-03-16 01:56:03 +00:00
restart:
Polish vtruncbuf() to improve readability and style a bit.
2004-10-21 14:13:54 +00:00
bo = &vp->v_bufobj;
- Complete part of the unfinished bufobj work by consistently using BO_LOCK/UNLOCK/MTX when manipulating the bufobj. - Create a new lock in the bufobj to lock bufobj fields independently. This leaves the vnode interlock as an 'identity' lock while the bufobj is an io lock. The bufobj lock is ordered before the vnode interlock and also before the mnt ilock. - Exploit this new lock order to simplify softdep_check_suspend(). - A few sync related functions are marked with a new XXX to note that we may not properly interlock against a non-zero bv_cnt when attempting to sync all vnodes on a mountlist. I do not believe this race is important. If I'm wrong this will make these locations easier to find. Reviewed by: kib (earlier diff) Tested by: kris, pho (earlier diff)
2008-03-22 09:15:16 +00:00
BO_LOCK(bo);
Some VM improvements, including elimination of alot of Sig-11 problems. Tor Egge and others have helped with various VM bugs lately, but don't blame him -- blame me!!! pmap.c: 1) Create an object for kernel page table allocations. This fixes a bogus allocation method previously used for such, by grabbing pages from the kernel object, using bogus pindexes. (This was a code cleanup, and perhaps a minor system stability issue.) pmap.c: 2) Pre-set the modify and accessed bits when prudent. This will decrease bus traffic under certain circumstances. vfs_bio.c, vfs_cluster.c: 3) Rather than calculating the beginning virtual byte offset multiple times, stick the offset into the buffer header, so that the calculated offset can be reused. (Long long multiplies are often expensive, and this is a probably unmeasurable performance improvement, and code cleanup.) vfs_bio.c: 4) Handle write recursion more intelligently (but not perfectly) so that it is less likely to cause a system panic, and is also much more robust. vfs_bio.c: 5) getblk incorrectly wrote out blocks that are incorrectly sized. The problem is fixed, and writes blocks out ONLY when B_DELWRI is true. vfs_bio.c: 6) Check that already constituted buffers have fully valid pages. If not, then make sure that the B_CACHE bit is not set. (This was a major source of Sig-11 type problems.) vfs_bio.c: 7) Fix a potential system deadlock due to an incorrectly specified sleep priority while waiting for a buffer write operation. The change that I made opens the system up to serious problems, and we need to examine the issue of process sleep priorities. vfs_cluster.c, vfs_bio.c: 8) Make clustered reads work more correctly (and more completely) when buffers are already constituted, but not fully valid. (This was another system reliability issue.) vfs_subr.c, ffs_inode.c: 9) Create a vtruncbuf function, which is used by filesystems that can truncate files. The vinvalbuf forced a file sync type operation, while vtruncbuf only invalidates the buffers past the new end of file, and also invalidates the appropriate pages. (This was a system reliabiliy and performance issue.) 10) Modify FFS to use vtruncbuf. vm_object.c: 11) Make the object rundown mechanism for OBJT_VNODE type objects work more correctly. Included in that fix, create pager entries for the OBJT_DEAD pager type, so that paging requests that might slip in during race conditions are properly handled. (This was a system reliability issue.) vm_page.c: 12) Make some of the page validation routines be a little less picky about arguments passed to them. Also, support page invalidation change the object generation count so that we handle generation counts a little more robustly. vm_pageout.c: 13) Further reduce pageout daemon activity when the system doesn't need help from it. There should be no additional performance decrease even when the pageout daemon is running. (This was a significant performance issue.) vnode_pager.c: 14) Teach the vnode pager to handle race conditions during vnode deallocations.
1998-03-16 01:56:03 +00:00
anyfreed = 1;
for (;anyfreed;) {
anyfreed = 0;
Move the VI_BWAIT flag into no bo_flag element of bufobj and call it BO_WWAIT Add bufobj_wref(), bufobj_wdrop() and bufobj_wwait() to handle the write count on a bufobj. Bufobj_wdrop() replaces vwakeup(). Use these functions all relevant places except in ffs_softdep.c where the use if interlocked_sleep() makes this impossible. Rename b_vnbufs to b_bobufs now that we touch all the relevant files anyway.
2004-10-21 15:53:54 +00:00
TAILQ_FOREACH_SAFE(bp, &bo->bo_clean.bv_hd, b_bobufs, nbp) {
Polish vtruncbuf() to improve readability and style a bit.
2004-10-21 14:13:54 +00:00
if (bp->b_lblkno < trunclbn)
continue;
if (BUF_LOCK(bp,
LK_EXCLUSIVE | LK_SLEEPFAIL | LK_INTERLOCK,
- Complete part of the unfinished bufobj work by consistently using BO_LOCK/UNLOCK/MTX when manipulating the bufobj. - Create a new lock in the bufobj to lock bufobj fields independently. This leaves the vnode interlock as an 'identity' lock while the bufobj is an io lock. The bufobj lock is ordered before the vnode interlock and also before the mnt ilock. - Exploit this new lock order to simplify softdep_check_suspend(). - A few sync related functions are marked with a new XXX to note that we may not properly interlock against a non-zero bv_cnt when attempting to sync all vnodes on a mountlist. I do not believe this race is important. If I'm wrong this will make these locations easier to find. Reviewed by: kib (earlier diff) Tested by: kris, pho (earlier diff)
2008-03-22 09:15:16 +00:00
BO_MTX(bo)) == ENOLCK)
Polish vtruncbuf() to improve readability and style a bit.
2004-10-21 14:13:54 +00:00
goto restart;
bremfree(bp);
bp->b_flags |= (B_INVAL | B_RELBUF);
bp->b_flags &= ~B_ASYNC;
brelse(bp);
anyfreed = 1;
if (nbp != NULL &&
(((nbp->b_xflags & BX_VNCLEAN) == 0) ||
(nbp->b_vp != vp) ||
(nbp->b_flags & B_DELWRI))) {
goto restart;
Some VM improvements, including elimination of alot of Sig-11 problems. Tor Egge and others have helped with various VM bugs lately, but don't blame him -- blame me!!! pmap.c: 1) Create an object for kernel page table allocations. This fixes a bogus allocation method previously used for such, by grabbing pages from the kernel object, using bogus pindexes. (This was a code cleanup, and perhaps a minor system stability issue.) pmap.c: 2) Pre-set the modify and accessed bits when prudent. This will decrease bus traffic under certain circumstances. vfs_bio.c, vfs_cluster.c: 3) Rather than calculating the beginning virtual byte offset multiple times, stick the offset into the buffer header, so that the calculated offset can be reused. (Long long multiplies are often expensive, and this is a probably unmeasurable performance improvement, and code cleanup.) vfs_bio.c: 4) Handle write recursion more intelligently (but not perfectly) so that it is less likely to cause a system panic, and is also much more robust. vfs_bio.c: 5) getblk incorrectly wrote out blocks that are incorrectly sized. The problem is fixed, and writes blocks out ONLY when B_DELWRI is true. vfs_bio.c: 6) Check that already constituted buffers have fully valid pages. If not, then make sure that the B_CACHE bit is not set. (This was a major source of Sig-11 type problems.) vfs_bio.c: 7) Fix a potential system deadlock due to an incorrectly specified sleep priority while waiting for a buffer write operation. The change that I made opens the system up to serious problems, and we need to examine the issue of process sleep priorities. vfs_cluster.c, vfs_bio.c: 8) Make clustered reads work more correctly (and more completely) when buffers are already constituted, but not fully valid. (This was another system reliability issue.) vfs_subr.c, ffs_inode.c: 9) Create a vtruncbuf function, which is used by filesystems that can truncate files. The vinvalbuf forced a file sync type operation, while vtruncbuf only invalidates the buffers past the new end of file, and also invalidates the appropriate pages. (This was a system reliabiliy and performance issue.) 10) Modify FFS to use vtruncbuf. vm_object.c: 11) Make the object rundown mechanism for OBJT_VNODE type objects work more correctly. Included in that fix, create pager entries for the OBJT_DEAD pager type, so that paging requests that might slip in during race conditions are properly handled. (This was a system reliability issue.) vm_page.c: 12) Make some of the page validation routines be a little less picky about arguments passed to them. Also, support page invalidation change the object generation count so that we handle generation counts a little more robustly. vm_pageout.c: 13) Further reduce pageout daemon activity when the system doesn't need help from it. There should be no additional performance decrease even when the pageout daemon is running. (This was a significant performance issue.) vnode_pager.c: 14) Teach the vnode pager to handle race conditions during vnode deallocations.
1998-03-16 01:56:03 +00:00
}
- Complete part of the unfinished bufobj work by consistently using BO_LOCK/UNLOCK/MTX when manipulating the bufobj. - Create a new lock in the bufobj to lock bufobj fields independently. This leaves the vnode interlock as an 'identity' lock while the bufobj is an io lock. The bufobj lock is ordered before the vnode interlock and also before the mnt ilock. - Exploit this new lock order to simplify softdep_check_suspend(). - A few sync related functions are marked with a new XXX to note that we may not properly interlock against a non-zero bv_cnt when attempting to sync all vnodes on a mountlist. I do not believe this race is important. If I'm wrong this will make these locations easier to find. Reviewed by: kib (earlier diff) Tested by: kris, pho (earlier diff)
2008-03-22 09:15:16 +00:00
BO_LOCK(bo);
Some VM improvements, including elimination of alot of Sig-11 problems. Tor Egge and others have helped with various VM bugs lately, but don't blame him -- blame me!!! pmap.c: 1) Create an object for kernel page table allocations. This fixes a bogus allocation method previously used for such, by grabbing pages from the kernel object, using bogus pindexes. (This was a code cleanup, and perhaps a minor system stability issue.) pmap.c: 2) Pre-set the modify and accessed bits when prudent. This will decrease bus traffic under certain circumstances. vfs_bio.c, vfs_cluster.c: 3) Rather than calculating the beginning virtual byte offset multiple times, stick the offset into the buffer header, so that the calculated offset can be reused. (Long long multiplies are often expensive, and this is a probably unmeasurable performance improvement, and code cleanup.) vfs_bio.c: 4) Handle write recursion more intelligently (but not perfectly) so that it is less likely to cause a system panic, and is also much more robust. vfs_bio.c: 5) getblk incorrectly wrote out blocks that are incorrectly sized. The problem is fixed, and writes blocks out ONLY when B_DELWRI is true. vfs_bio.c: 6) Check that already constituted buffers have fully valid pages. If not, then make sure that the B_CACHE bit is not set. (This was a major source of Sig-11 type problems.) vfs_bio.c: 7) Fix a potential system deadlock due to an incorrectly specified sleep priority while waiting for a buffer write operation. The change that I made opens the system up to serious problems, and we need to examine the issue of process sleep priorities. vfs_cluster.c, vfs_bio.c: 8) Make clustered reads work more correctly (and more completely) when buffers are already constituted, but not fully valid. (This was another system reliability issue.) vfs_subr.c, ffs_inode.c: 9) Create a vtruncbuf function, which is used by filesystems that can truncate files. The vinvalbuf forced a file sync type operation, while vtruncbuf only invalidates the buffers past the new end of file, and also invalidates the appropriate pages. (This was a system reliabiliy and performance issue.) 10) Modify FFS to use vtruncbuf. vm_object.c: 11) Make the object rundown mechanism for OBJT_VNODE type objects work more correctly. Included in that fix, create pager entries for the OBJT_DEAD pager type, so that paging requests that might slip in during race conditions are properly handled. (This was a system reliability issue.) vm_page.c: 12) Make some of the page validation routines be a little less picky about arguments passed to them. Also, support page invalidation change the object generation count so that we handle generation counts a little more robustly. vm_pageout.c: 13) Further reduce pageout daemon activity when the system doesn't need help from it. There should be no additional performance decrease even when the pageout daemon is running. (This was a significant performance issue.) vnode_pager.c: 14) Teach the vnode pager to handle race conditions during vnode deallocations.
1998-03-16 01:56:03 +00:00
}
Move the VI_BWAIT flag into no bo_flag element of bufobj and call it BO_WWAIT Add bufobj_wref(), bufobj_wdrop() and bufobj_wwait() to handle the write count on a bufobj. Bufobj_wdrop() replaces vwakeup(). Use these functions all relevant places except in ffs_softdep.c where the use if interlocked_sleep() makes this impossible. Rename b_vnbufs to b_bobufs now that we touch all the relevant files anyway.
2004-10-21 15:53:54 +00:00
TAILQ_FOREACH_SAFE(bp, &bo->bo_dirty.bv_hd, b_bobufs, nbp) {
Polish vtruncbuf() to improve readability and style a bit.
2004-10-21 14:13:54 +00:00
if (bp->b_lblkno < trunclbn)
continue;
if (BUF_LOCK(bp,
LK_EXCLUSIVE | LK_SLEEPFAIL | LK_INTERLOCK,
- Complete part of the unfinished bufobj work by consistently using BO_LOCK/UNLOCK/MTX when manipulating the bufobj. - Create a new lock in the bufobj to lock bufobj fields independently. This leaves the vnode interlock as an 'identity' lock while the bufobj is an io lock. The bufobj lock is ordered before the vnode interlock and also before the mnt ilock. - Exploit this new lock order to simplify softdep_check_suspend(). - A few sync related functions are marked with a new XXX to note that we may not properly interlock against a non-zero bv_cnt when attempting to sync all vnodes on a mountlist. I do not believe this race is important. If I'm wrong this will make these locations easier to find. Reviewed by: kib (earlier diff) Tested by: kris, pho (earlier diff)
2008-03-22 09:15:16 +00:00
BO_MTX(bo)) == ENOLCK)
Polish vtruncbuf() to improve readability and style a bit.
2004-10-21 14:13:54 +00:00
goto restart;
bremfree(bp);
bp->b_flags |= (B_INVAL | B_RELBUF);
bp->b_flags &= ~B_ASYNC;
brelse(bp);
anyfreed = 1;
if (nbp != NULL &&
(((nbp->b_xflags & BX_VNDIRTY) == 0) ||
(nbp->b_vp != vp) ||
(nbp->b_flags & B_DELWRI) == 0)) {
goto restart;
Some VM improvements, including elimination of alot of Sig-11 problems. Tor Egge and others have helped with various VM bugs lately, but don't blame him -- blame me!!! pmap.c: 1) Create an object for kernel page table allocations. This fixes a bogus allocation method previously used for such, by grabbing pages from the kernel object, using bogus pindexes. (This was a code cleanup, and perhaps a minor system stability issue.) pmap.c: 2) Pre-set the modify and accessed bits when prudent. This will decrease bus traffic under certain circumstances. vfs_bio.c, vfs_cluster.c: 3) Rather than calculating the beginning virtual byte offset multiple times, stick the offset into the buffer header, so that the calculated offset can be reused. (Long long multiplies are often expensive, and this is a probably unmeasurable performance improvement, and code cleanup.) vfs_bio.c: 4) Handle write recursion more intelligently (but not perfectly) so that it is less likely to cause a system panic, and is also much more robust. vfs_bio.c: 5) getblk incorrectly wrote out blocks that are incorrectly sized. The problem is fixed, and writes blocks out ONLY when B_DELWRI is true. vfs_bio.c: 6) Check that already constituted buffers have fully valid pages. If not, then make sure that the B_CACHE bit is not set. (This was a major source of Sig-11 type problems.) vfs_bio.c: 7) Fix a potential system deadlock due to an incorrectly specified sleep priority while waiting for a buffer write operation. The change that I made opens the system up to serious problems, and we need to examine the issue of process sleep priorities. vfs_cluster.c, vfs_bio.c: 8) Make clustered reads work more correctly (and more completely) when buffers are already constituted, but not fully valid. (This was another system reliability issue.) vfs_subr.c, ffs_inode.c: 9) Create a vtruncbuf function, which is used by filesystems that can truncate files. The vinvalbuf forced a file sync type operation, while vtruncbuf only invalidates the buffers past the new end of file, and also invalidates the appropriate pages. (This was a system reliabiliy and performance issue.) 10) Modify FFS to use vtruncbuf. vm_object.c: 11) Make the object rundown mechanism for OBJT_VNODE type objects work more correctly. Included in that fix, create pager entries for the OBJT_DEAD pager type, so that paging requests that might slip in during race conditions are properly handled. (This was a system reliability issue.) vm_page.c: 12) Make some of the page validation routines be a little less picky about arguments passed to them. Also, support page invalidation change the object generation count so that we handle generation counts a little more robustly. vm_pageout.c: 13) Further reduce pageout daemon activity when the system doesn't need help from it. There should be no additional performance decrease even when the pageout daemon is running. (This was a significant performance issue.) vnode_pager.c: 14) Teach the vnode pager to handle race conditions during vnode deallocations.
1998-03-16 01:56:03 +00:00
}
- Complete part of the unfinished bufobj work by consistently using BO_LOCK/UNLOCK/MTX when manipulating the bufobj. - Create a new lock in the bufobj to lock bufobj fields independently. This leaves the vnode interlock as an 'identity' lock while the bufobj is an io lock. The bufobj lock is ordered before the vnode interlock and also before the mnt ilock. - Exploit this new lock order to simplify softdep_check_suspend(). - A few sync related functions are marked with a new XXX to note that we may not properly interlock against a non-zero bv_cnt when attempting to sync all vnodes on a mountlist. I do not believe this race is important. If I'm wrong this will make these locations easier to find. Reviewed by: kib (earlier diff) Tested by: kris, pho (earlier diff)
2008-03-22 09:15:16 +00:00
BO_LOCK(bo);
Some VM improvements, including elimination of alot of Sig-11 problems. Tor Egge and others have helped with various VM bugs lately, but don't blame him -- blame me!!! pmap.c: 1) Create an object for kernel page table allocations. This fixes a bogus allocation method previously used for such, by grabbing pages from the kernel object, using bogus pindexes. (This was a code cleanup, and perhaps a minor system stability issue.) pmap.c: 2) Pre-set the modify and accessed bits when prudent. This will decrease bus traffic under certain circumstances. vfs_bio.c, vfs_cluster.c: 3) Rather than calculating the beginning virtual byte offset multiple times, stick the offset into the buffer header, so that the calculated offset can be reused. (Long long multiplies are often expensive, and this is a probably unmeasurable performance improvement, and code cleanup.) vfs_bio.c: 4) Handle write recursion more intelligently (but not perfectly) so that it is less likely to cause a system panic, and is also much more robust. vfs_bio.c: 5) getblk incorrectly wrote out blocks that are incorrectly sized. The problem is fixed, and writes blocks out ONLY when B_DELWRI is true. vfs_bio.c: 6) Check that already constituted buffers have fully valid pages. If not, then make sure that the B_CACHE bit is not set. (This was a major source of Sig-11 type problems.) vfs_bio.c: 7) Fix a potential system deadlock due to an incorrectly specified sleep priority while waiting for a buffer write operation. The change that I made opens the system up to serious problems, and we need to examine the issue of process sleep priorities. vfs_cluster.c, vfs_bio.c: 8) Make clustered reads work more correctly (and more completely) when buffers are already constituted, but not fully valid. (This was another system reliability issue.) vfs_subr.c, ffs_inode.c: 9) Create a vtruncbuf function, which is used by filesystems that can truncate files. The vinvalbuf forced a file sync type operation, while vtruncbuf only invalidates the buffers past the new end of file, and also invalidates the appropriate pages. (This was a system reliabiliy and performance issue.) 10) Modify FFS to use vtruncbuf. vm_object.c: 11) Make the object rundown mechanism for OBJT_VNODE type objects work more correctly. Included in that fix, create pager entries for the OBJT_DEAD pager type, so that paging requests that might slip in during race conditions are properly handled. (This was a system reliability issue.) vm_page.c: 12) Make some of the page validation routines be a little less picky about arguments passed to them. Also, support page invalidation change the object generation count so that we handle generation counts a little more robustly. vm_pageout.c: 13) Further reduce pageout daemon activity when the system doesn't need help from it. There should be no additional performance decrease even when the pageout daemon is running. (This was a significant performance issue.) vnode_pager.c: 14) Teach the vnode pager to handle race conditions during vnode deallocations.
1998-03-16 01:56:03 +00:00
}
}
Correct a severely evil bug in the vtruncbuf code. It didn't cause me any problems until after the previous commit. This problem then caused a severe case of creeping crud on my diskdrive, and hosed my system so bad, that I needed to do a complete reinstall. Sorry!!! I assume that others have manifest this bug.
1998-03-17 06:30:52 +00:00
In kern_physio.c fix tsleep priority messup. In vfs_bio.c, remove b_generation count usage, remove redundant reassignbuf, remove redundant spl(s), manage page PG_ZERO flags more correctly, utilize in invalid value for b_offset until it is properly initialized. Add asserts for #ifdef DIAGNOSTIC, when b_offset is improperly used. when a process is not performing I/O, and just waiting on a buffer generally, make the sleep priority low. only check page validity in getblk for B_VMIO buffers. In vfs_cluster, add b_offset asserts, correct pointer calculation for clustered reads. Improve readability of certain parts of the code. Remove redundant spl(s). In vfs_subr, correct usage of vfs_bio_awrite (From Andrew Gallatin <gallatin@cs.duke.edu>). More vtruncbuf problems fixed.
1998-03-19 22:48:16 +00:00
if (length > 0) {
restartsync:
Move the VI_BWAIT flag into no bo_flag element of bufobj and call it BO_WWAIT Add bufobj_wref(), bufobj_wdrop() and bufobj_wwait() to handle the write count on a bufobj. Bufobj_wdrop() replaces vwakeup(). Use these functions all relevant places except in ffs_softdep.c where the use if interlocked_sleep() makes this impossible. Rename b_vnbufs to b_bobufs now that we touch all the relevant files anyway.
2004-10-21 15:53:54 +00:00
TAILQ_FOREACH_SAFE(bp, &bo->bo_dirty.bv_hd, b_bobufs, nbp) {
- Remove a dead check for bp->b_vp == vp in vtruncbuf(). This has not been possible for some time. - Lock the buf before accessing fields. This should very rarely be locked. - Assert that B_DELWRI is set after we acquire the buf. This should always be the case now.
2003-03-13 07:22:53 +00:00
if (bp->b_lblkno > 0)
continue;
/*
* Since we hold the vnode lock this should only
* fail if we're racing with the buf daemon.
*/
if (BUF_LOCK(bp,
LK_EXCLUSIVE | LK_SLEEPFAIL | LK_INTERLOCK,
- Pass BO_MTX(bo) to lockmgr in vtruncbuf, we don't own the vnode interlock here anymore. Reported by: kris
2008-03-23 01:42:19 +00:00
BO_MTX(bo)) == ENOLCK) {
- Remove a dead check for bp->b_vp == vp in vtruncbuf(). This has not been possible for some time. - Lock the buf before accessing fields. This should very rarely be locked. - Assert that B_DELWRI is set after we acquire the buf. This should always be the case now.
2003-03-13 07:22:53 +00:00
goto restart;
Correct a severely evil bug in the vtruncbuf code. It didn't cause me any problems until after the previous commit. This problem then caused a severe case of creeping crud on my diskdrive, and hosed my system so bad, that I needed to do a complete reinstall. Sorry!!! I assume that others have manifest this bug.
1998-03-17 06:30:52 +00:00
}
Convert KASSERTS to VNASSERTS
2005-02-17 10:28:58 +00:00
VNASSERT((bp->b_flags & B_DELWRI), vp,
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
("buf(%p) on dirty queue without DELWRI", bp));
- Remove a dead check for bp->b_vp == vp in vtruncbuf(). This has not been possible for some time. - Lock the buf before accessing fields. This should very rarely be locked. - Assert that B_DELWRI is set after we acquire the buf. This should always be the case now.
2003-03-13 07:22:53 +00:00
bremfree(bp);
bawrite(bp);
- Complete part of the unfinished bufobj work by consistently using BO_LOCK/UNLOCK/MTX when manipulating the bufobj. - Create a new lock in the bufobj to lock bufobj fields independently. This leaves the vnode interlock as an 'identity' lock while the bufobj is an io lock. The bufobj lock is ordered before the vnode interlock and also before the mnt ilock. - Exploit this new lock order to simplify softdep_check_suspend(). - A few sync related functions are marked with a new XXX to note that we may not properly interlock against a non-zero bv_cnt when attempting to sync all vnodes on a mountlist. I do not believe this race is important. If I'm wrong this will make these locations easier to find. Reviewed by: kib (earlier diff) Tested by: kris, pho (earlier diff)
2008-03-22 09:15:16 +00:00
BO_LOCK(bo);
- Remove a dead check for bp->b_vp == vp in vtruncbuf(). This has not been possible for some time. - Lock the buf before accessing fields. This should very rarely be locked. - Assert that B_DELWRI is set after we acquire the buf. This should always be the case now.
2003-03-13 07:22:53 +00:00
goto restartsync;
Correct a severely evil bug in the vtruncbuf code. It didn't cause me any problems until after the previous commit. This problem then caused a severe case of creeping crud on my diskdrive, and hosed my system so bad, that I needed to do a complete reinstall. Sorry!!! I assume that others have manifest this bug.
1998-03-17 06:30:52 +00:00
}
}
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
Move the VI_BWAIT flag into no bo_flag element of bufobj and call it BO_WWAIT Add bufobj_wref(), bufobj_wdrop() and bufobj_wwait() to handle the write count on a bufobj. Bufobj_wdrop() replaces vwakeup(). Use these functions all relevant places except in ffs_softdep.c where the use if interlocked_sleep() makes this impossible. Rename b_vnbufs to b_bobufs now that we touch all the relevant files anyway.
2004-10-21 15:53:54 +00:00
bufobj_wwait(bo, 0, 0);
- Complete part of the unfinished bufobj work by consistently using BO_LOCK/UNLOCK/MTX when manipulating the bufobj. - Create a new lock in the bufobj to lock bufobj fields independently. This leaves the vnode interlock as an 'identity' lock while the bufobj is an io lock. The bufobj lock is ordered before the vnode interlock and also before the mnt ilock. - Exploit this new lock order to simplify softdep_check_suspend(). - A few sync related functions are marked with a new XXX to note that we may not properly interlock against a non-zero bv_cnt when attempting to sync all vnodes on a mountlist. I do not believe this race is important. If I'm wrong this will make these locations easier to find. Reviewed by: kib (earlier diff) Tested by: kris, pho (earlier diff)
2008-03-22 09:15:16 +00:00
BO_UNLOCK(bo);
Some VM improvements, including elimination of alot of Sig-11 problems. Tor Egge and others have helped with various VM bugs lately, but don't blame him -- blame me!!! pmap.c: 1) Create an object for kernel page table allocations. This fixes a bogus allocation method previously used for such, by grabbing pages from the kernel object, using bogus pindexes. (This was a code cleanup, and perhaps a minor system stability issue.) pmap.c: 2) Pre-set the modify and accessed bits when prudent. This will decrease bus traffic under certain circumstances. vfs_bio.c, vfs_cluster.c: 3) Rather than calculating the beginning virtual byte offset multiple times, stick the offset into the buffer header, so that the calculated offset can be reused. (Long long multiplies are often expensive, and this is a probably unmeasurable performance improvement, and code cleanup.) vfs_bio.c: 4) Handle write recursion more intelligently (but not perfectly) so that it is less likely to cause a system panic, and is also much more robust. vfs_bio.c: 5) getblk incorrectly wrote out blocks that are incorrectly sized. The problem is fixed, and writes blocks out ONLY when B_DELWRI is true. vfs_bio.c: 6) Check that already constituted buffers have fully valid pages. If not, then make sure that the B_CACHE bit is not set. (This was a major source of Sig-11 type problems.) vfs_bio.c: 7) Fix a potential system deadlock due to an incorrectly specified sleep priority while waiting for a buffer write operation. The change that I made opens the system up to serious problems, and we need to examine the issue of process sleep priorities. vfs_cluster.c, vfs_bio.c: 8) Make clustered reads work more correctly (and more completely) when buffers are already constituted, but not fully valid. (This was another system reliability issue.) vfs_subr.c, ffs_inode.c: 9) Create a vtruncbuf function, which is used by filesystems that can truncate files. The vinvalbuf forced a file sync type operation, while vtruncbuf only invalidates the buffers past the new end of file, and also invalidates the appropriate pages. (This was a system reliabiliy and performance issue.) 10) Modify FFS to use vtruncbuf. vm_object.c: 11) Make the object rundown mechanism for OBJT_VNODE type objects work more correctly. Included in that fix, create pager entries for the OBJT_DEAD pager type, so that paging requests that might slip in during race conditions are properly handled. (This was a system reliability issue.) vm_page.c: 12) Make some of the page validation routines be a little less picky about arguments passed to them. Also, support page invalidation change the object generation count so that we handle generation counts a little more robustly. vm_pageout.c: 13) Further reduce pageout daemon activity when the system doesn't need help from it. There should be no additional performance decrease even when the pageout daemon is running. (This was a significant performance issue.) vnode_pager.c: 14) Teach the vnode pager to handle race conditions during vnode deallocations.
1998-03-16 01:56:03 +00:00
vnode_pager_setsize(vp, length);
return (0);
}
Replace the global buffer hash table with per-vnode splay trees using a methodology similar to the vm_map_entry splay and the VM splay that Alan Cox is working on. Extensive testing has appeared to have shown no increase in overhead. Disadvantages Dirties more cache lines during lookups. Not as fast as a hash table lookup (but still N log N and optimal when there is locality of reference). Advantages vnode->v_dirtyblkhd is now perfectly sorted, making fsync/sync/filesystem syncer operate more efficiently. I get to rip out all the old hacks (some of which were mine) that tried to keep the v_dirtyblkhd tailq sorted. The per-vnode splay tree should be easier to lock / SMPng pushdown on vnodes will be easier. This commit along with another that Alan is working on for the VM page global hash table will allow me to implement ranged fsync(), optimize server-side nfs commit rpcs, and implement partial syncs by the filesystem syncer (aka filesystem syncer would detect that someone is trying to get the vnode lock, remembers its place, and skip to the next vnode). Note that the buffer cache splay is somewhat more complex then other splays due to special handling of background bitmap writes (multiple buffers with the same lblkno in the same vnode), and B_INVAL discontinuities between the old hash table and the existence of the buffer on the v_cleanblkhd list. Suggested by: alc
2002-07-10 17:02:32 +00:00
/*
* buf_splay() - splay tree core for the clean/dirty list of buffers in
* a vnode.
*
* NOTE: We have to deal with the special case of a background bitmap
* buffer, a situation where two buffers will have the same logical
* block offset. We want (1) only the foreground buffer to be accessed
* in a lookup and (2) must differentiate between the foreground and
* background buffer in the splay tree algorithm because the splay
* tree cannot normally handle multiple entities with the same 'index'.
* We accomplish this by adding differentiating flags to the splay tree's
* numerical domain.
*/
static
struct buf *
buf_splay(daddr_t lblkno, b_xflags_t xflags, struct buf *root)
{
struct buf dummy;
struct buf *lefttreemax, *righttreemin, *y;
if (root == NULL)
return (NULL);
lefttreemax = righttreemin = &dummy;
for (;;) {
if (lblkno < root->b_lblkno ||
(lblkno == root->b_lblkno &&
(xflags & BX_BKGRDMARKER) < (root->b_xflags & BX_BKGRDMARKER))) {
if ((y = root->b_left) == NULL)
break;
if (lblkno < y->b_lblkno) {
/* Rotate right. */
root->b_left = y->b_right;
y->b_right = root;
root = y;
if ((y = root->b_left) == NULL)
break;
}
/* Link into the new root's right tree. */
righttreemin->b_left = root;
righttreemin = root;
} else if (lblkno > root->b_lblkno ||
(lblkno == root->b_lblkno &&
(xflags & BX_BKGRDMARKER) > (root->b_xflags & BX_BKGRDMARKER))) {
if ((y = root->b_right) == NULL)
break;
if (lblkno > y->b_lblkno) {
/* Rotate left. */
root->b_right = y->b_left;
y->b_left = root;
root = y;
if ((y = root->b_right) == NULL)
break;
}
/* Link into the new root's left tree. */
lefttreemax->b_right = root;
lefttreemax = root;
} else {
break;
}
root = y;
}
/* Assemble the new root. */
lefttreemax->b_right = root->b_left;
righttreemin->b_left = root->b_right;
root->b_left = dummy.b_right;
root->b_right = dummy.b_left;
return (root);
}
Add b_bufobj to struct buf which eventually will eliminate the need for b_vp. Initialize b_bufobj for all buffers. Make incore() and gbincore() take a bufobj instead of a vnode. Make inmem() local to vfs_bio.c Change a lot of VI_[UN]LOCK(bp->b_vp) to BO_[UN]LOCK(bp->b_bufobj) also VI_MTX() to BO_MTX(), Make buf_vlist_add() take a bufobj instead of a vnode. Eliminate other uses of bp->b_vp where bp->b_bufobj will do. Various minor polishing: remove "register", turn panic into KASSERT, use new function declarations, TAILQ_FOREACH_SAFE() etc.
2004-10-22 08:47:20 +00:00
static void
Replace the global buffer hash table with per-vnode splay trees using a methodology similar to the vm_map_entry splay and the VM splay that Alan Cox is working on. Extensive testing has appeared to have shown no increase in overhead. Disadvantages Dirties more cache lines during lookups. Not as fast as a hash table lookup (but still N log N and optimal when there is locality of reference). Advantages vnode->v_dirtyblkhd is now perfectly sorted, making fsync/sync/filesystem syncer operate more efficiently. I get to rip out all the old hacks (some of which were mine) that tried to keep the v_dirtyblkhd tailq sorted. The per-vnode splay tree should be easier to lock / SMPng pushdown on vnodes will be easier. This commit along with another that Alan is working on for the VM page global hash table will allow me to implement ranged fsync(), optimize server-side nfs commit rpcs, and implement partial syncs by the filesystem syncer (aka filesystem syncer would detect that someone is trying to get the vnode lock, remembers its place, and skip to the next vnode). Note that the buffer cache splay is somewhat more complex then other splays due to special handling of background bitmap writes (multiple buffers with the same lblkno in the same vnode), and B_INVAL discontinuities between the old hash table and the existence of the buffer on the v_cleanblkhd list. Suggested by: alc
2002-07-10 17:02:32 +00:00
buf_vlist_remove(struct buf *bp)
{
struct buf *root;
Simplify buf_vlist_remove(). Now that we have encapsulated the splaytree related information into a structure we can eliminate the half of this function.
2004-10-21 13:48:50 +00:00
struct bufv *bv;
Replace the global buffer hash table with per-vnode splay trees using a methodology similar to the vm_map_entry splay and the VM splay that Alan Cox is working on. Extensive testing has appeared to have shown no increase in overhead. Disadvantages Dirties more cache lines during lookups. Not as fast as a hash table lookup (but still N log N and optimal when there is locality of reference). Advantages vnode->v_dirtyblkhd is now perfectly sorted, making fsync/sync/filesystem syncer operate more efficiently. I get to rip out all the old hacks (some of which were mine) that tried to keep the v_dirtyblkhd tailq sorted. The per-vnode splay tree should be easier to lock / SMPng pushdown on vnodes will be easier. This commit along with another that Alan is working on for the VM page global hash table will allow me to implement ranged fsync(), optimize server-side nfs commit rpcs, and implement partial syncs by the filesystem syncer (aka filesystem syncer would detect that someone is trying to get the vnode lock, remembers its place, and skip to the next vnode). Note that the buffer cache splay is somewhat more complex then other splays due to special handling of background bitmap writes (multiple buffers with the same lblkno in the same vnode), and B_INVAL discontinuities between the old hash table and the existence of the buffer on the v_cleanblkhd list. Suggested by: alc
2002-07-10 17:02:32 +00:00
Add b_bufobj to struct buf which eventually will eliminate the need for b_vp. Initialize b_bufobj for all buffers. Make incore() and gbincore() take a bufobj instead of a vnode. Make inmem() local to vfs_bio.c Change a lot of VI_[UN]LOCK(bp->b_vp) to BO_[UN]LOCK(bp->b_bufobj) also VI_MTX() to BO_MTX(), Make buf_vlist_add() take a bufobj instead of a vnode. Eliminate other uses of bp->b_vp where bp->b_bufobj will do. Various minor polishing: remove "register", turn panic into KASSERT, use new function declarations, TAILQ_FOREACH_SAFE() etc.
2004-10-22 08:47:20 +00:00
KASSERT(bp->b_bufobj != NULL, ("No b_bufobj %p", bp));
ASSERT_BO_LOCKED(bp->b_bufobj);
- Try to catch the wrong bufobj panics a little earlier. I believe they are actually caused by a buf with both VNCLEAN and VNDIRTY set. In the traces it is clear that the buf is removed from the dirty queue while it is actually on the clean queue which leaves the tail pointer set. Assert that both flags are not set in buf_vlist_add and buf_vlist_remove. Sponsored by: Isilon Systems, Inc. Approved by: re (blanket vfs)
2005-06-18 18:17:03 +00:00
KASSERT((bp->b_xflags & (BX_VNDIRTY|BX_VNCLEAN)) !=
(BX_VNDIRTY|BX_VNCLEAN),
("buf_vlist_remove: Buf %p is on two lists", bp));
Minor style cleanups (mostly removal of trailing whitespaces).
2007-04-10 15:29:37 +00:00
if (bp->b_xflags & BX_VNDIRTY)
Add b_bufobj to struct buf which eventually will eliminate the need for b_vp. Initialize b_bufobj for all buffers. Make incore() and gbincore() take a bufobj instead of a vnode. Make inmem() local to vfs_bio.c Change a lot of VI_[UN]LOCK(bp->b_vp) to BO_[UN]LOCK(bp->b_bufobj) also VI_MTX() to BO_MTX(), Make buf_vlist_add() take a bufobj instead of a vnode. Eliminate other uses of bp->b_vp where bp->b_bufobj will do. Various minor polishing: remove "register", turn panic into KASSERT, use new function declarations, TAILQ_FOREACH_SAFE() etc.
2004-10-22 08:47:20 +00:00
bv = &bp->b_bufobj->bo_dirty;
Simplify buf_vlist_remove(). Now that we have encapsulated the splaytree related information into a structure we can eliminate the half of this function.
2004-10-21 13:48:50 +00:00
else
Add b_bufobj to struct buf which eventually will eliminate the need for b_vp. Initialize b_bufobj for all buffers. Make incore() and gbincore() take a bufobj instead of a vnode. Make inmem() local to vfs_bio.c Change a lot of VI_[UN]LOCK(bp->b_vp) to BO_[UN]LOCK(bp->b_bufobj) also VI_MTX() to BO_MTX(), Make buf_vlist_add() take a bufobj instead of a vnode. Eliminate other uses of bp->b_vp where bp->b_bufobj will do. Various minor polishing: remove "register", turn panic into KASSERT, use new function declarations, TAILQ_FOREACH_SAFE() etc.
2004-10-22 08:47:20 +00:00
bv = &bp->b_bufobj->bo_clean;
Simplify buf_vlist_remove(). Now that we have encapsulated the splaytree related information into a structure we can eliminate the half of this function.
2004-10-21 13:48:50 +00:00
if (bp != bv->bv_root) {
root = buf_splay(bp->b_lblkno, bp->b_xflags, bv->bv_root);
KASSERT(root == bp, ("splay lookup failed in remove"));
}
if (bp->b_left == NULL) {
root = bp->b_right;
Replace the global buffer hash table with per-vnode splay trees using a methodology similar to the vm_map_entry splay and the VM splay that Alan Cox is working on. Extensive testing has appeared to have shown no increase in overhead. Disadvantages Dirties more cache lines during lookups. Not as fast as a hash table lookup (but still N log N and optimal when there is locality of reference). Advantages vnode->v_dirtyblkhd is now perfectly sorted, making fsync/sync/filesystem syncer operate more efficiently. I get to rip out all the old hacks (some of which were mine) that tried to keep the v_dirtyblkhd tailq sorted. The per-vnode splay tree should be easier to lock / SMPng pushdown on vnodes will be easier. This commit along with another that Alan is working on for the VM page global hash table will allow me to implement ranged fsync(), optimize server-side nfs commit rpcs, and implement partial syncs by the filesystem syncer (aka filesystem syncer would detect that someone is trying to get the vnode lock, remembers its place, and skip to the next vnode). Note that the buffer cache splay is somewhat more complex then other splays due to special handling of background bitmap writes (multiple buffers with the same lblkno in the same vnode), and B_INVAL discontinuities between the old hash table and the existence of the buffer on the v_cleanblkhd list. Suggested by: alc
2002-07-10 17:02:32 +00:00
} else {
Simplify buf_vlist_remove(). Now that we have encapsulated the splaytree related information into a structure we can eliminate the half of this function.
2004-10-21 13:48:50 +00:00
root = buf_splay(bp->b_lblkno, bp->b_xflags, bp->b_left);
root->b_right = bp->b_right;
Replace the global buffer hash table with per-vnode splay trees using a methodology similar to the vm_map_entry splay and the VM splay that Alan Cox is working on. Extensive testing has appeared to have shown no increase in overhead. Disadvantages Dirties more cache lines during lookups. Not as fast as a hash table lookup (but still N log N and optimal when there is locality of reference). Advantages vnode->v_dirtyblkhd is now perfectly sorted, making fsync/sync/filesystem syncer operate more efficiently. I get to rip out all the old hacks (some of which were mine) that tried to keep the v_dirtyblkhd tailq sorted. The per-vnode splay tree should be easier to lock / SMPng pushdown on vnodes will be easier. This commit along with another that Alan is working on for the VM page global hash table will allow me to implement ranged fsync(), optimize server-side nfs commit rpcs, and implement partial syncs by the filesystem syncer (aka filesystem syncer would detect that someone is trying to get the vnode lock, remembers its place, and skip to the next vnode). Note that the buffer cache splay is somewhat more complex then other splays due to special handling of background bitmap writes (multiple buffers with the same lblkno in the same vnode), and B_INVAL discontinuities between the old hash table and the existence of the buffer on the v_cleanblkhd list. Suggested by: alc
2002-07-10 17:02:32 +00:00
}
Simplify buf_vlist_remove(). Now that we have encapsulated the splaytree related information into a structure we can eliminate the half of this function.
2004-10-21 13:48:50 +00:00
bv->bv_root = root;
Move the VI_BWAIT flag into no bo_flag element of bufobj and call it BO_WWAIT Add bufobj_wref(), bufobj_wdrop() and bufobj_wwait() to handle the write count on a bufobj. Bufobj_wdrop() replaces vwakeup(). Use these functions all relevant places except in ffs_softdep.c where the use if interlocked_sleep() makes this impossible. Rename b_vnbufs to b_bobufs now that we touch all the relevant files anyway.
2004-10-21 15:53:54 +00:00
TAILQ_REMOVE(&bv->bv_hd, bp, b_bobufs);
Simplify buf_vlist_remove(). Now that we have encapsulated the splaytree related information into a structure we can eliminate the half of this function.
2004-10-21 13:48:50 +00:00
bv->bv_cnt--;
Replace the global buffer hash table with per-vnode splay trees using a methodology similar to the vm_map_entry splay and the VM splay that Alan Cox is working on. Extensive testing has appeared to have shown no increase in overhead. Disadvantages Dirties more cache lines during lookups. Not as fast as a hash table lookup (but still N log N and optimal when there is locality of reference). Advantages vnode->v_dirtyblkhd is now perfectly sorted, making fsync/sync/filesystem syncer operate more efficiently. I get to rip out all the old hacks (some of which were mine) that tried to keep the v_dirtyblkhd tailq sorted. The per-vnode splay tree should be easier to lock / SMPng pushdown on vnodes will be easier. This commit along with another that Alan is working on for the VM page global hash table will allow me to implement ranged fsync(), optimize server-side nfs commit rpcs, and implement partial syncs by the filesystem syncer (aka filesystem syncer would detect that someone is trying to get the vnode lock, remembers its place, and skip to the next vnode). Note that the buffer cache splay is somewhat more complex then other splays due to special handling of background bitmap writes (multiple buffers with the same lblkno in the same vnode), and B_INVAL discontinuities between the old hash table and the existence of the buffer on the v_cleanblkhd list. Suggested by: alc
2002-07-10 17:02:32 +00:00
bp->b_xflags &= ~(BX_VNDIRTY | BX_VNCLEAN);
}
/*
* Add the buffer to the sorted clean or dirty block list using a
* splay tree algorithm.
*
* NOTE: xflags is passed as a constant, optimizing this inline function!
*/
Add b_bufobj to struct buf which eventually will eliminate the need for b_vp. Initialize b_bufobj for all buffers. Make incore() and gbincore() take a bufobj instead of a vnode. Make inmem() local to vfs_bio.c Change a lot of VI_[UN]LOCK(bp->b_vp) to BO_[UN]LOCK(bp->b_bufobj) also VI_MTX() to BO_MTX(), Make buf_vlist_add() take a bufobj instead of a vnode. Eliminate other uses of bp->b_vp where bp->b_bufobj will do. Various minor polishing: remove "register", turn panic into KASSERT, use new function declarations, TAILQ_FOREACH_SAFE() etc.
2004-10-22 08:47:20 +00:00
static void
buf_vlist_add(struct buf *bp, struct bufobj *bo, b_xflags_t xflags)
Replace the global buffer hash table with per-vnode splay trees using a methodology similar to the vm_map_entry splay and the VM splay that Alan Cox is working on. Extensive testing has appeared to have shown no increase in overhead. Disadvantages Dirties more cache lines during lookups. Not as fast as a hash table lookup (but still N log N and optimal when there is locality of reference). Advantages vnode->v_dirtyblkhd is now perfectly sorted, making fsync/sync/filesystem syncer operate more efficiently. I get to rip out all the old hacks (some of which were mine) that tried to keep the v_dirtyblkhd tailq sorted. The per-vnode splay tree should be easier to lock / SMPng pushdown on vnodes will be easier. This commit along with another that Alan is working on for the VM page global hash table will allow me to implement ranged fsync(), optimize server-side nfs commit rpcs, and implement partial syncs by the filesystem syncer (aka filesystem syncer would detect that someone is trying to get the vnode lock, remembers its place, and skip to the next vnode). Note that the buffer cache splay is somewhat more complex then other splays due to special handling of background bitmap writes (multiple buffers with the same lblkno in the same vnode), and B_INVAL discontinuities between the old hash table and the existence of the buffer on the v_cleanblkhd list. Suggested by: alc
2002-07-10 17:02:32 +00:00
{
struct buf *root;
Add b_bufobj to struct buf which eventually will eliminate the need for b_vp. Initialize b_bufobj for all buffers. Make incore() and gbincore() take a bufobj instead of a vnode. Make inmem() local to vfs_bio.c Change a lot of VI_[UN]LOCK(bp->b_vp) to BO_[UN]LOCK(bp->b_bufobj) also VI_MTX() to BO_MTX(), Make buf_vlist_add() take a bufobj instead of a vnode. Eliminate other uses of bp->b_vp where bp->b_bufobj will do. Various minor polishing: remove "register", turn panic into KASSERT, use new function declarations, TAILQ_FOREACH_SAFE() etc.
2004-10-22 08:47:20 +00:00
struct bufv *bv;
Replace the global buffer hash table with per-vnode splay trees using a methodology similar to the vm_map_entry splay and the VM splay that Alan Cox is working on. Extensive testing has appeared to have shown no increase in overhead. Disadvantages Dirties more cache lines during lookups. Not as fast as a hash table lookup (but still N log N and optimal when there is locality of reference). Advantages vnode->v_dirtyblkhd is now perfectly sorted, making fsync/sync/filesystem syncer operate more efficiently. I get to rip out all the old hacks (some of which were mine) that tried to keep the v_dirtyblkhd tailq sorted. The per-vnode splay tree should be easier to lock / SMPng pushdown on vnodes will be easier. This commit along with another that Alan is working on for the VM page global hash table will allow me to implement ranged fsync(), optimize server-side nfs commit rpcs, and implement partial syncs by the filesystem syncer (aka filesystem syncer would detect that someone is trying to get the vnode lock, remembers its place, and skip to the next vnode). Note that the buffer cache splay is somewhat more complex then other splays due to special handling of background bitmap writes (multiple buffers with the same lblkno in the same vnode), and B_INVAL discontinuities between the old hash table and the existence of the buffer on the v_cleanblkhd list. Suggested by: alc
2002-07-10 17:02:32 +00:00
Add b_bufobj to struct buf which eventually will eliminate the need for b_vp. Initialize b_bufobj for all buffers. Make incore() and gbincore() take a bufobj instead of a vnode. Make inmem() local to vfs_bio.c Change a lot of VI_[UN]LOCK(bp->b_vp) to BO_[UN]LOCK(bp->b_bufobj) also VI_MTX() to BO_MTX(), Make buf_vlist_add() take a bufobj instead of a vnode. Eliminate other uses of bp->b_vp where bp->b_bufobj will do. Various minor polishing: remove "register", turn panic into KASSERT, use new function declarations, TAILQ_FOREACH_SAFE() etc.
2004-10-22 08:47:20 +00:00
ASSERT_BO_LOCKED(bo);
- Try to catch the wrong bufobj panics a little earlier. I believe they are actually caused by a buf with both VNCLEAN and VNDIRTY set. In the traces it is clear that the buf is removed from the dirty queue while it is actually on the clean queue which leaves the tail pointer set. Assert that both flags are not set in buf_vlist_add and buf_vlist_remove. Sponsored by: Isilon Systems, Inc. Approved by: re (blanket vfs)
2005-06-18 18:17:03 +00:00
KASSERT((bp->b_xflags & (BX_VNDIRTY|BX_VNCLEAN)) == 0,
("buf_vlist_add: Buf %p has existing xflags %d", bp, bp->b_xflags));
Replace the global buffer hash table with per-vnode splay trees using a methodology similar to the vm_map_entry splay and the VM splay that Alan Cox is working on. Extensive testing has appeared to have shown no increase in overhead. Disadvantages Dirties more cache lines during lookups. Not as fast as a hash table lookup (but still N log N and optimal when there is locality of reference). Advantages vnode->v_dirtyblkhd is now perfectly sorted, making fsync/sync/filesystem syncer operate more efficiently. I get to rip out all the old hacks (some of which were mine) that tried to keep the v_dirtyblkhd tailq sorted. The per-vnode splay tree should be easier to lock / SMPng pushdown on vnodes will be easier. This commit along with another that Alan is working on for the VM page global hash table will allow me to implement ranged fsync(), optimize server-side nfs commit rpcs, and implement partial syncs by the filesystem syncer (aka filesystem syncer would detect that someone is trying to get the vnode lock, remembers its place, and skip to the next vnode). Note that the buffer cache splay is somewhat more complex then other splays due to special handling of background bitmap writes (multiple buffers with the same lblkno in the same vnode), and B_INVAL discontinuities between the old hash table and the existence of the buffer on the v_cleanblkhd list. Suggested by: alc
2002-07-10 17:02:32 +00:00
bp->b_xflags |= xflags;
Add b_bufobj to struct buf which eventually will eliminate the need for b_vp. Initialize b_bufobj for all buffers. Make incore() and gbincore() take a bufobj instead of a vnode. Make inmem() local to vfs_bio.c Change a lot of VI_[UN]LOCK(bp->b_vp) to BO_[UN]LOCK(bp->b_bufobj) also VI_MTX() to BO_MTX(), Make buf_vlist_add() take a bufobj instead of a vnode. Eliminate other uses of bp->b_vp where bp->b_bufobj will do. Various minor polishing: remove "register", turn panic into KASSERT, use new function declarations, TAILQ_FOREACH_SAFE() etc.
2004-10-22 08:47:20 +00:00
if (xflags & BX_VNDIRTY)
bv = &bo->bo_dirty;
else
bv = &bo->bo_clean;
root = buf_splay(bp->b_lblkno, bp->b_xflags, bv->bv_root);
if (root == NULL) {
bp->b_left = NULL;
bp->b_right = NULL;
TAILQ_INSERT_TAIL(&bv->bv_hd, bp, b_bobufs);
} else if (bp->b_lblkno < root->b_lblkno ||
(bp->b_lblkno == root->b_lblkno &&
(bp->b_xflags & BX_BKGRDMARKER) < (root->b_xflags & BX_BKGRDMARKER))) {
bp->b_left = root->b_left;
bp->b_right = root;
root->b_left = NULL;
TAILQ_INSERT_BEFORE(root, bp, b_bobufs);
Replace the global buffer hash table with per-vnode splay trees using a methodology similar to the vm_map_entry splay and the VM splay that Alan Cox is working on. Extensive testing has appeared to have shown no increase in overhead. Disadvantages Dirties more cache lines during lookups. Not as fast as a hash table lookup (but still N log N and optimal when there is locality of reference). Advantages vnode->v_dirtyblkhd is now perfectly sorted, making fsync/sync/filesystem syncer operate more efficiently. I get to rip out all the old hacks (some of which were mine) that tried to keep the v_dirtyblkhd tailq sorted. The per-vnode splay tree should be easier to lock / SMPng pushdown on vnodes will be easier. This commit along with another that Alan is working on for the VM page global hash table will allow me to implement ranged fsync(), optimize server-side nfs commit rpcs, and implement partial syncs by the filesystem syncer (aka filesystem syncer would detect that someone is trying to get the vnode lock, remembers its place, and skip to the next vnode). Note that the buffer cache splay is somewhat more complex then other splays due to special handling of background bitmap writes (multiple buffers with the same lblkno in the same vnode), and B_INVAL discontinuities between the old hash table and the existence of the buffer on the v_cleanblkhd list. Suggested by: alc
2002-07-10 17:02:32 +00:00
} else {
Add b_bufobj to struct buf which eventually will eliminate the need for b_vp. Initialize b_bufobj for all buffers. Make incore() and gbincore() take a bufobj instead of a vnode. Make inmem() local to vfs_bio.c Change a lot of VI_[UN]LOCK(bp->b_vp) to BO_[UN]LOCK(bp->b_bufobj) also VI_MTX() to BO_MTX(), Make buf_vlist_add() take a bufobj instead of a vnode. Eliminate other uses of bp->b_vp where bp->b_bufobj will do. Various minor polishing: remove "register", turn panic into KASSERT, use new function declarations, TAILQ_FOREACH_SAFE() etc.
2004-10-22 08:47:20 +00:00
bp->b_right = root->b_right;
bp->b_left = root;
root->b_right = NULL;
TAILQ_INSERT_AFTER(&bv->bv_hd, root, bp, b_bobufs);
Replace the global buffer hash table with per-vnode splay trees using a methodology similar to the vm_map_entry splay and the VM splay that Alan Cox is working on. Extensive testing has appeared to have shown no increase in overhead. Disadvantages Dirties more cache lines during lookups. Not as fast as a hash table lookup (but still N log N and optimal when there is locality of reference). Advantages vnode->v_dirtyblkhd is now perfectly sorted, making fsync/sync/filesystem syncer operate more efficiently. I get to rip out all the old hacks (some of which were mine) that tried to keep the v_dirtyblkhd tailq sorted. The per-vnode splay tree should be easier to lock / SMPng pushdown on vnodes will be easier. This commit along with another that Alan is working on for the VM page global hash table will allow me to implement ranged fsync(), optimize server-side nfs commit rpcs, and implement partial syncs by the filesystem syncer (aka filesystem syncer would detect that someone is trying to get the vnode lock, remembers its place, and skip to the next vnode). Note that the buffer cache splay is somewhat more complex then other splays due to special handling of background bitmap writes (multiple buffers with the same lblkno in the same vnode), and B_INVAL discontinuities between the old hash table and the existence of the buffer on the v_cleanblkhd list. Suggested by: alc
2002-07-10 17:02:32 +00:00
}
Add b_bufobj to struct buf which eventually will eliminate the need for b_vp. Initialize b_bufobj for all buffers. Make incore() and gbincore() take a bufobj instead of a vnode. Make inmem() local to vfs_bio.c Change a lot of VI_[UN]LOCK(bp->b_vp) to BO_[UN]LOCK(bp->b_bufobj) also VI_MTX() to BO_MTX(), Make buf_vlist_add() take a bufobj instead of a vnode. Eliminate other uses of bp->b_vp where bp->b_bufobj will do. Various minor polishing: remove "register", turn panic into KASSERT, use new function declarations, TAILQ_FOREACH_SAFE() etc.
2004-10-22 08:47:20 +00:00
bv->bv_cnt++;
bv->bv_root = bp;
Replace the global buffer hash table with per-vnode splay trees using a methodology similar to the vm_map_entry splay and the VM splay that Alan Cox is working on. Extensive testing has appeared to have shown no increase in overhead. Disadvantages Dirties more cache lines during lookups. Not as fast as a hash table lookup (but still N log N and optimal when there is locality of reference). Advantages vnode->v_dirtyblkhd is now perfectly sorted, making fsync/sync/filesystem syncer operate more efficiently. I get to rip out all the old hacks (some of which were mine) that tried to keep the v_dirtyblkhd tailq sorted. The per-vnode splay tree should be easier to lock / SMPng pushdown on vnodes will be easier. This commit along with another that Alan is working on for the VM page global hash table will allow me to implement ranged fsync(), optimize server-side nfs commit rpcs, and implement partial syncs by the filesystem syncer (aka filesystem syncer would detect that someone is trying to get the vnode lock, remembers its place, and skip to the next vnode). Note that the buffer cache splay is somewhat more complex then other splays due to special handling of background bitmap writes (multiple buffers with the same lblkno in the same vnode), and B_INVAL discontinuities between the old hash table and the existence of the buffer on the v_cleanblkhd list. Suggested by: alc
2002-07-10 17:02:32 +00:00
}
/*
* Lookup a buffer using the splay tree. Note that we specifically avoid
* shadow buffers used in background bitmap writes.
*
* This code isn't quite efficient as it could be because we are maintaining
* two sorted lists and do not know which list the block resides in.
Optimize the use of splay in gbincore(). During a "make buildworld" the desired buffer is found at one of the roots more than 60% of the time. Thus, checking both roots before performing either splay eliminates unnecessary splays on the first tree splayed. Approved by: re (jhb)
2003-05-13 04:36:02 +00:00
*
* During a "make buildworld" the desired buffer is found at one of
* the roots more than 60% of the time. Thus, checking both roots
* before performing either splay eliminates unnecessary splays on the
* first tree splayed.
Replace the global buffer hash table with per-vnode splay trees using a methodology similar to the vm_map_entry splay and the VM splay that Alan Cox is working on. Extensive testing has appeared to have shown no increase in overhead. Disadvantages Dirties more cache lines during lookups. Not as fast as a hash table lookup (but still N log N and optimal when there is locality of reference). Advantages vnode->v_dirtyblkhd is now perfectly sorted, making fsync/sync/filesystem syncer operate more efficiently. I get to rip out all the old hacks (some of which were mine) that tried to keep the v_dirtyblkhd tailq sorted. The per-vnode splay tree should be easier to lock / SMPng pushdown on vnodes will be easier. This commit along with another that Alan is working on for the VM page global hash table will allow me to implement ranged fsync(), optimize server-side nfs commit rpcs, and implement partial syncs by the filesystem syncer (aka filesystem syncer would detect that someone is trying to get the vnode lock, remembers its place, and skip to the next vnode). Note that the buffer cache splay is somewhat more complex then other splays due to special handling of background bitmap writes (multiple buffers with the same lblkno in the same vnode), and B_INVAL discontinuities between the old hash table and the existence of the buffer on the v_cleanblkhd list. Suggested by: alc
2002-07-10 17:02:32 +00:00
*/
struct buf *
Add b_bufobj to struct buf which eventually will eliminate the need for b_vp. Initialize b_bufobj for all buffers. Make incore() and gbincore() take a bufobj instead of a vnode. Make inmem() local to vfs_bio.c Change a lot of VI_[UN]LOCK(bp->b_vp) to BO_[UN]LOCK(bp->b_bufobj) also VI_MTX() to BO_MTX(), Make buf_vlist_add() take a bufobj instead of a vnode. Eliminate other uses of bp->b_vp where bp->b_bufobj will do. Various minor polishing: remove "register", turn panic into KASSERT, use new function declarations, TAILQ_FOREACH_SAFE() etc.
2004-10-22 08:47:20 +00:00
gbincore(struct bufobj *bo, daddr_t lblkno)
Replace the global buffer hash table with per-vnode splay trees using a methodology similar to the vm_map_entry splay and the VM splay that Alan Cox is working on. Extensive testing has appeared to have shown no increase in overhead. Disadvantages Dirties more cache lines during lookups. Not as fast as a hash table lookup (but still N log N and optimal when there is locality of reference). Advantages vnode->v_dirtyblkhd is now perfectly sorted, making fsync/sync/filesystem syncer operate more efficiently. I get to rip out all the old hacks (some of which were mine) that tried to keep the v_dirtyblkhd tailq sorted. The per-vnode splay tree should be easier to lock / SMPng pushdown on vnodes will be easier. This commit along with another that Alan is working on for the VM page global hash table will allow me to implement ranged fsync(), optimize server-side nfs commit rpcs, and implement partial syncs by the filesystem syncer (aka filesystem syncer would detect that someone is trying to get the vnode lock, remembers its place, and skip to the next vnode). Note that the buffer cache splay is somewhat more complex then other splays due to special handling of background bitmap writes (multiple buffers with the same lblkno in the same vnode), and B_INVAL discontinuities between the old hash table and the existence of the buffer on the v_cleanblkhd list. Suggested by: alc
2002-07-10 17:02:32 +00:00
{
struct buf *bp;
Add b_bufobj to struct buf which eventually will eliminate the need for b_vp. Initialize b_bufobj for all buffers. Make incore() and gbincore() take a bufobj instead of a vnode. Make inmem() local to vfs_bio.c Change a lot of VI_[UN]LOCK(bp->b_vp) to BO_[UN]LOCK(bp->b_bufobj) also VI_MTX() to BO_MTX(), Make buf_vlist_add() take a bufobj instead of a vnode. Eliminate other uses of bp->b_vp where bp->b_bufobj will do. Various minor polishing: remove "register", turn panic into KASSERT, use new function declarations, TAILQ_FOREACH_SAFE() etc.
2004-10-22 08:47:20 +00:00
ASSERT_BO_LOCKED(bo);
if ((bp = bo->bo_clean.bv_root) != NULL &&
Optimize the use of splay in gbincore(). During a "make buildworld" the desired buffer is found at one of the roots more than 60% of the time. Thus, checking both roots before performing either splay eliminates unnecessary splays on the first tree splayed. Approved by: re (jhb)
2003-05-13 04:36:02 +00:00
bp->b_lblkno == lblkno && !(bp->b_xflags & BX_BKGRDMARKER))
return (bp);
Add b_bufobj to struct buf which eventually will eliminate the need for b_vp. Initialize b_bufobj for all buffers. Make incore() and gbincore() take a bufobj instead of a vnode. Make inmem() local to vfs_bio.c Change a lot of VI_[UN]LOCK(bp->b_vp) to BO_[UN]LOCK(bp->b_bufobj) also VI_MTX() to BO_MTX(), Make buf_vlist_add() take a bufobj instead of a vnode. Eliminate other uses of bp->b_vp where bp->b_bufobj will do. Various minor polishing: remove "register", turn panic into KASSERT, use new function declarations, TAILQ_FOREACH_SAFE() etc.
2004-10-22 08:47:20 +00:00
if ((bp = bo->bo_dirty.bv_root) != NULL &&
Optimize the use of splay in gbincore(). During a "make buildworld" the desired buffer is found at one of the roots more than 60% of the time. Thus, checking both roots before performing either splay eliminates unnecessary splays on the first tree splayed. Approved by: re (jhb)
2003-05-13 04:36:02 +00:00
bp->b_lblkno == lblkno && !(bp->b_xflags & BX_BKGRDMARKER))
return (bp);
Add b_bufobj to struct buf which eventually will eliminate the need for b_vp. Initialize b_bufobj for all buffers. Make incore() and gbincore() take a bufobj instead of a vnode. Make inmem() local to vfs_bio.c Change a lot of VI_[UN]LOCK(bp->b_vp) to BO_[UN]LOCK(bp->b_bufobj) also VI_MTX() to BO_MTX(), Make buf_vlist_add() take a bufobj instead of a vnode. Eliminate other uses of bp->b_vp where bp->b_bufobj will do. Various minor polishing: remove "register", turn panic into KASSERT, use new function declarations, TAILQ_FOREACH_SAFE() etc.
2004-10-22 08:47:20 +00:00
if ((bp = bo->bo_clean.bv_root) != NULL) {
bo->bo_clean.bv_root = bp = buf_splay(lblkno, 0, bp);
Optimize the use of splay in gbincore(). During a "make buildworld" the desired buffer is found at one of the roots more than 60% of the time. Thus, checking both roots before performing either splay eliminates unnecessary splays on the first tree splayed. Approved by: re (jhb)
2003-05-13 04:36:02 +00:00
if (bp->b_lblkno == lblkno && !(bp->b_xflags & BX_BKGRDMARKER))
return (bp);
}
Add b_bufobj to struct buf which eventually will eliminate the need for b_vp. Initialize b_bufobj for all buffers. Make incore() and gbincore() take a bufobj instead of a vnode. Make inmem() local to vfs_bio.c Change a lot of VI_[UN]LOCK(bp->b_vp) to BO_[UN]LOCK(bp->b_bufobj) also VI_MTX() to BO_MTX(), Make buf_vlist_add() take a bufobj instead of a vnode. Eliminate other uses of bp->b_vp where bp->b_bufobj will do. Various minor polishing: remove "register", turn panic into KASSERT, use new function declarations, TAILQ_FOREACH_SAFE() etc.
2004-10-22 08:47:20 +00:00
if ((bp = bo->bo_dirty.bv_root) != NULL) {
bo->bo_dirty.bv_root = bp = buf_splay(lblkno, 0, bp);
Optimize the use of splay in gbincore(). During a "make buildworld" the desired buffer is found at one of the roots more than 60% of the time. Thus, checking both roots before performing either splay eliminates unnecessary splays on the first tree splayed. Approved by: re (jhb)
2003-05-13 04:36:02 +00:00
if (bp->b_lblkno == lblkno && !(bp->b_xflags & BX_BKGRDMARKER))
return (bp);
}
return (NULL);
Replace the global buffer hash table with per-vnode splay trees using a methodology similar to the vm_map_entry splay and the VM splay that Alan Cox is working on. Extensive testing has appeared to have shown no increase in overhead. Disadvantages Dirties more cache lines during lookups. Not as fast as a hash table lookup (but still N log N and optimal when there is locality of reference). Advantages vnode->v_dirtyblkhd is now perfectly sorted, making fsync/sync/filesystem syncer operate more efficiently. I get to rip out all the old hacks (some of which were mine) that tried to keep the v_dirtyblkhd tailq sorted. The per-vnode splay tree should be easier to lock / SMPng pushdown on vnodes will be easier. This commit along with another that Alan is working on for the VM page global hash table will allow me to implement ranged fsync(), optimize server-side nfs commit rpcs, and implement partial syncs by the filesystem syncer (aka filesystem syncer would detect that someone is trying to get the vnode lock, remembers its place, and skip to the next vnode). Note that the buffer cache splay is somewhat more complex then other splays due to special handling of background bitmap writes (multiple buffers with the same lblkno in the same vnode), and B_INVAL discontinuities between the old hash table and the existence of the buffer on the v_cleanblkhd list. Suggested by: alc
2002-07-10 17:02:32 +00:00
}
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
/*
* Associate a buffer with a vnode.
*/
The big 4.4BSD Lite to FreeBSD 2.0.0 (Development) patch. Reviewed by: Rodney W. Grimes Submitted by: John Dyson and David Greenman
1994-05-25 09:21:21 +00:00
void
Add b_bufobj to struct buf which eventually will eliminate the need for b_vp. Initialize b_bufobj for all buffers. Make incore() and gbincore() take a bufobj instead of a vnode. Make inmem() local to vfs_bio.c Change a lot of VI_[UN]LOCK(bp->b_vp) to BO_[UN]LOCK(bp->b_bufobj) also VI_MTX() to BO_MTX(), Make buf_vlist_add() take a bufobj instead of a vnode. Eliminate other uses of bp->b_vp where bp->b_bufobj will do. Various minor polishing: remove "register", turn panic into KASSERT, use new function declarations, TAILQ_FOREACH_SAFE() etc.
2004-10-22 08:47:20 +00:00
bgetvp(struct vnode *vp, struct buf *bp)
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
{
- Complete part of the unfinished bufobj work by consistently using BO_LOCK/UNLOCK/MTX when manipulating the bufobj. - Create a new lock in the bufobj to lock bufobj fields independently. This leaves the vnode interlock as an 'identity' lock while the bufobj is an io lock. The bufobj lock is ordered before the vnode interlock and also before the mnt ilock. - Exploit this new lock order to simplify softdep_check_suspend(). - A few sync related functions are marked with a new XXX to note that we may not properly interlock against a non-zero bv_cnt when attempting to sync all vnodes on a mountlist. I do not believe this race is important. If I'm wrong this will make these locations easier to find. Reviewed by: kib (earlier diff) Tested by: kris, pho (earlier diff)
2008-03-22 09:15:16 +00:00
struct bufobj *bo;
Convert KASSERTS to VNASSERTS
2005-02-17 10:28:58 +00:00
- Complete part of the unfinished bufobj work by consistently using BO_LOCK/UNLOCK/MTX when manipulating the bufobj. - Create a new lock in the bufobj to lock bufobj fields independently. This leaves the vnode interlock as an 'identity' lock while the bufobj is an io lock. The bufobj lock is ordered before the vnode interlock and also before the mnt ilock. - Exploit this new lock order to simplify softdep_check_suspend(). - A few sync related functions are marked with a new XXX to note that we may not properly interlock against a non-zero bv_cnt when attempting to sync all vnodes on a mountlist. I do not believe this race is important. If I'm wrong this will make these locations easier to find. Reviewed by: kib (earlier diff) Tested by: kris, pho (earlier diff)
2008-03-22 09:15:16 +00:00
bo = &vp->v_bufobj;
ASSERT_BO_LOCKED(bo);
Convert KASSERTS to VNASSERTS
2005-02-17 10:28:58 +00:00
VNASSERT(bp->b_vp == NULL, bp->b_vp, ("bgetvp: not free"));
KNFize, by bde.
1999-01-10 01:58:29 +00:00
- Add the tunable and sysctl for the mpsafevfs. It currently defaults to off. - Protect access to mnt_kern_flag with the mointpoint mutex. - Remove some KASSERTs which are not legal checks without the appropriate locks held. - Use VCANRECYCLE() rather than rolling several slightly different checks together. - Return from vtryrecycle() with a recycled vnode rather than a locked vnode. This simplifies some locking. - Remove several GIANT_REQUIRED lines. - Add a few KASSERTs to help with INACT debugging. Sponsored By: Isilon Systems, Inc.
2005-01-24 10:41:01 +00:00
CTR3(KTR_BUF, "bgetvp(%p) vp %p flags %X", bp, vp, bp->b_flags);
Convert KASSERTS to VNASSERTS
2005-02-17 10:28:58 +00:00
VNASSERT((bp->b_xflags & (BX_VNDIRTY|BX_VNCLEAN)) == 0, vp,
Replace the global buffer hash table with per-vnode splay trees using a methodology similar to the vm_map_entry splay and the VM splay that Alan Cox is working on. Extensive testing has appeared to have shown no increase in overhead. Disadvantages Dirties more cache lines during lookups. Not as fast as a hash table lookup (but still N log N and optimal when there is locality of reference). Advantages vnode->v_dirtyblkhd is now perfectly sorted, making fsync/sync/filesystem syncer operate more efficiently. I get to rip out all the old hacks (some of which were mine) that tried to keep the v_dirtyblkhd tailq sorted. The per-vnode splay tree should be easier to lock / SMPng pushdown on vnodes will be easier. This commit along with another that Alan is working on for the VM page global hash table will allow me to implement ranged fsync(), optimize server-side nfs commit rpcs, and implement partial syncs by the filesystem syncer (aka filesystem syncer would detect that someone is trying to get the vnode lock, remembers its place, and skip to the next vnode). Note that the buffer cache splay is somewhat more complex then other splays due to special handling of background bitmap writes (multiple buffers with the same lblkno in the same vnode), and B_INVAL discontinuities between the old hash table and the existence of the buffer on the v_cleanblkhd list. Suggested by: alc
2002-07-10 17:02:32 +00:00
("bgetvp: bp already attached! %p", bp));
- Complete part of the unfinished bufobj work by consistently using BO_LOCK/UNLOCK/MTX when manipulating the bufobj. - Create a new lock in the bufobj to lock bufobj fields independently. This leaves the vnode interlock as an 'identity' lock while the bufobj is an io lock. The bufobj lock is ordered before the vnode interlock and also before the mnt ilock. - Exploit this new lock order to simplify softdep_check_suspend(). - A few sync related functions are marked with a new XXX to note that we may not properly interlock against a non-zero bv_cnt when attempting to sync all vnodes on a mountlist. I do not believe this race is important. If I'm wrong this will make these locations easier to find. Reviewed by: kib (earlier diff) Tested by: kris, pho (earlier diff)
2008-03-22 09:15:16 +00:00
vhold(vp);
if (VFS_NEEDSGIANT(vp->v_mount) || bo->bo_flag & BO_NEEDSGIANT)
- Add the B_NEEDSGIANT flag which is only set if the vnode that owns a buf requires Giant. It is set in bgetvp and cleared in brelvp. - Create QUEUE_DIRTY_GIANT for dirty buffers that require giant. - In the buf daemon, only grab giant when processing QUEUE_DIRTY_GIANT and only if we think there are buffers in that queue. Sponsored by: Isilon Systems, Inc.
2006-03-31 02:56:30 +00:00
bp->b_flags |= B_NEEDSGIANT;
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
bp->b_vp = vp;
- Complete part of the unfinished bufobj work by consistently using BO_LOCK/UNLOCK/MTX when manipulating the bufobj. - Create a new lock in the bufobj to lock bufobj fields independently. This leaves the vnode interlock as an 'identity' lock while the bufobj is an io lock. The bufobj lock is ordered before the vnode interlock and also before the mnt ilock. - Exploit this new lock order to simplify softdep_check_suspend(). - A few sync related functions are marked with a new XXX to note that we may not properly interlock against a non-zero bv_cnt when attempting to sync all vnodes on a mountlist. I do not believe this race is important. If I'm wrong this will make these locations easier to find. Reviewed by: kib (earlier diff) Tested by: kris, pho (earlier diff)
2008-03-22 09:15:16 +00:00
bp->b_bufobj = bo;
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
/*
* Insert onto list for new vnode.
*/
- Complete part of the unfinished bufobj work by consistently using BO_LOCK/UNLOCK/MTX when manipulating the bufobj. - Create a new lock in the bufobj to lock bufobj fields independently. This leaves the vnode interlock as an 'identity' lock while the bufobj is an io lock. The bufobj lock is ordered before the vnode interlock and also before the mnt ilock. - Exploit this new lock order to simplify softdep_check_suspend(). - A few sync related functions are marked with a new XXX to note that we may not properly interlock against a non-zero bv_cnt when attempting to sync all vnodes on a mountlist. I do not believe this race is important. If I'm wrong this will make these locations easier to find. Reviewed by: kib (earlier diff) Tested by: kris, pho (earlier diff)
2008-03-22 09:15:16 +00:00
buf_vlist_add(bp, bo, BX_VNCLEAN);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
}
/*
* Disassociate a buffer from a vnode.
*/
b_waiters cannot be adequately protected by the interlock because it is dropped after the call to lockmgr() so just revert this approach using something similar to the precedent one: BUF_LOCKWAITERS() just checks if there are waiters (not the actual number of them) and it is based on newly introduced lockmgr_waiters() which returns if the lockmgr has waiters or not. The name has been choosen differently by old lockwaiters() in order to not confuse them. KPI results enriched by this commit so __FreeBSD_version bumping and manpage update will be happening soon. 'struct buf' also changes, so kernel ABI is disturbed. Bug found by: jeff Approved by: jeff, kib
2008-03-28 12:30:12 +00:00
void
Add b_bufobj to struct buf which eventually will eliminate the need for b_vp. Initialize b_bufobj for all buffers. Make incore() and gbincore() take a bufobj instead of a vnode. Make inmem() local to vfs_bio.c Change a lot of VI_[UN]LOCK(bp->b_vp) to BO_[UN]LOCK(bp->b_bufobj) also VI_MTX() to BO_MTX(), Make buf_vlist_add() take a bufobj instead of a vnode. Eliminate other uses of bp->b_vp where bp->b_bufobj will do. Various minor polishing: remove "register", turn panic into KASSERT, use new function declarations, TAILQ_FOREACH_SAFE() etc.
2004-10-22 08:47:20 +00:00
brelvp(struct buf *bp)
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
{
Add b_bufobj to struct buf which eventually will eliminate the need for b_vp. Initialize b_bufobj for all buffers. Make incore() and gbincore() take a bufobj instead of a vnode. Make inmem() local to vfs_bio.c Change a lot of VI_[UN]LOCK(bp->b_vp) to BO_[UN]LOCK(bp->b_bufobj) also VI_MTX() to BO_MTX(), Make buf_vlist_add() take a bufobj instead of a vnode. Eliminate other uses of bp->b_vp where bp->b_bufobj will do. Various minor polishing: remove "register", turn panic into KASSERT, use new function declarations, TAILQ_FOREACH_SAFE() etc.
2004-10-22 08:47:20 +00:00
struct bufobj *bo;
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
struct vnode *vp;
- Add the tunable and sysctl for the mpsafevfs. It currently defaults to off. - Protect access to mnt_kern_flag with the mointpoint mutex. - Remove some KASSERTs which are not legal checks without the appropriate locks held. - Use VCANRECYCLE() rather than rolling several slightly different checks together. - Return from vtryrecycle() with a recycled vnode rather than a locked vnode. This simplifies some locking. - Remove several GIANT_REQUIRED lines. - Add a few KASSERTs to help with INACT debugging. Sponsored By: Isilon Systems, Inc.
2005-01-24 10:41:01 +00:00
CTR3(KTR_BUF, "brelvp(%p) vp %p flags %X", bp, bp->b_vp, bp->b_flags);
Split DIAGNOSTIC -> DIAGNOSTIC, INVARIANTS, and INVARIANT_SUPPORT as discussed on -hackers. Introduce 'KASSERT(assertion, ("panic message", args))' for simple check + panic. Reviewed by: msmith
1999-01-08 17:31:30 +00:00
KASSERT(bp->b_vp != NULL, ("brelvp: NULL"));
Lots of improvements, including restructring the caching and management of vnodes and objects. There are some metadata performance improvements that come along with this. There are also a few prototypes added when the need is noticed. Changes include: 1) Cleaning up vref, vget. 2) Removal of the object cache. 3) Nuke vnode_pager_uncache and friends, because they aren't needed anymore. 4) Correct some missing LK_RETRY's in vn_lock. 5) Correct the page range in the code for msync. Be gentle, and please give me feedback asap.
1997-12-29 00:25:11 +00:00
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
/*
* Delete from old vnode list, if on one.
*/
Move the syncer linkage from vnode to bufobj. This is not quite a perfect separation: the syncer still think it knows that everything is a vnode.
2004-10-27 08:05:02 +00:00
vp = bp->b_vp; /* XXX */
Add b_bufobj to struct buf which eventually will eliminate the need for b_vp. Initialize b_bufobj for all buffers. Make incore() and gbincore() take a bufobj instead of a vnode. Make inmem() local to vfs_bio.c Change a lot of VI_[UN]LOCK(bp->b_vp) to BO_[UN]LOCK(bp->b_bufobj) also VI_MTX() to BO_MTX(), Make buf_vlist_add() take a bufobj instead of a vnode. Eliminate other uses of bp->b_vp where bp->b_bufobj will do. Various minor polishing: remove "register", turn panic into KASSERT, use new function declarations, TAILQ_FOREACH_SAFE() etc.
2004-10-22 08:47:20 +00:00
bo = bp->b_bufobj;
Move the syncer linkage from vnode to bufobj. This is not quite a perfect separation: the syncer still think it knows that everything is a vnode.
2004-10-27 08:05:02 +00:00
BO_LOCK(bo);
Replace the global buffer hash table with per-vnode splay trees using a methodology similar to the vm_map_entry splay and the VM splay that Alan Cox is working on. Extensive testing has appeared to have shown no increase in overhead. Disadvantages Dirties more cache lines during lookups. Not as fast as a hash table lookup (but still N log N and optimal when there is locality of reference). Advantages vnode->v_dirtyblkhd is now perfectly sorted, making fsync/sync/filesystem syncer operate more efficiently. I get to rip out all the old hacks (some of which were mine) that tried to keep the v_dirtyblkhd tailq sorted. The per-vnode splay tree should be easier to lock / SMPng pushdown on vnodes will be easier. This commit along with another that Alan is working on for the VM page global hash table will allow me to implement ranged fsync(), optimize server-side nfs commit rpcs, and implement partial syncs by the filesystem syncer (aka filesystem syncer would detect that someone is trying to get the vnode lock, remembers its place, and skip to the next vnode). Note that the buffer cache splay is somewhat more complex then other splays due to special handling of background bitmap writes (multiple buffers with the same lblkno in the same vnode), and B_INVAL discontinuities between the old hash table and the existence of the buffer on the v_cleanblkhd list. Suggested by: alc
2002-07-10 17:02:32 +00:00
if (bp->b_xflags & (BX_VNDIRTY | BX_VNCLEAN))
buf_vlist_remove(bp);
- All buffers should either be clean or dirty. If neither of these flags are set when we attempt to remove a buffer from a queue we should panic. Hopefully this will catch the source of the wrong bufobj panics. Sponsored by: Isilon Systems, Inc.
2005-05-01 12:00:36 +00:00
else
panic("brelvp: Buffer %p not on queue.", bp);
Move the syncer linkage from vnode to bufobj. This is not quite a perfect separation: the syncer still think it knows that everything is a vnode.
2004-10-27 08:05:02 +00:00
if ((bo->bo_flag & BO_ONWORKLST) && bo->bo_dirty.bv_cnt == 0) {
bo->bo_flag &= ~BO_ONWORKLST;
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
mtx_lock(&sync_mtx);
Move the syncer linkage from vnode to bufobj. This is not quite a perfect separation: the syncer still think it knows that everything is a vnode.
2004-10-27 08:05:02 +00:00
LIST_REMOVE(bo, bo_synclist);
Minor style cleanups (mostly removal of trailing whitespaces).
2007-04-10 15:29:37 +00:00
syncer_worklist_len--;
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
mtx_unlock(&sync_mtx);
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
}
- Add the B_NEEDSGIANT flag which is only set if the vnode that owns a buf requires Giant. It is set in bgetvp and cleared in brelvp. - Create QUEUE_DIRTY_GIANT for dirty buffers that require giant. - In the buf daemon, only grab giant when processing QUEUE_DIRTY_GIANT and only if we think there are buffers in that queue. Sponsored by: Isilon Systems, Inc.
2006-03-31 02:56:30 +00:00
bp->b_flags &= ~B_NEEDSGIANT;
Add b_bufobj to struct buf which eventually will eliminate the need for b_vp. Initialize b_bufobj for all buffers. Make incore() and gbincore() take a bufobj instead of a vnode. Make inmem() local to vfs_bio.c Change a lot of VI_[UN]LOCK(bp->b_vp) to BO_[UN]LOCK(bp->b_bufobj) also VI_MTX() to BO_MTX(), Make buf_vlist_add() take a bufobj instead of a vnode. Eliminate other uses of bp->b_vp where bp->b_bufobj will do. Various minor polishing: remove "register", turn panic into KASSERT, use new function declarations, TAILQ_FOREACH_SAFE() etc.
2004-10-22 08:47:20 +00:00
bp->b_vp = NULL;
bp->b_bufobj = NULL;
- Complete part of the unfinished bufobj work by consistently using BO_LOCK/UNLOCK/MTX when manipulating the bufobj. - Create a new lock in the bufobj to lock bufobj fields independently. This leaves the vnode interlock as an 'identity' lock while the bufobj is an io lock. The bufobj lock is ordered before the vnode interlock and also before the mnt ilock. - Exploit this new lock order to simplify softdep_check_suspend(). - A few sync related functions are marked with a new XXX to note that we may not properly interlock against a non-zero bv_cnt when attempting to sync all vnodes on a mountlist. I do not believe this race is important. If I'm wrong this will make these locations easier to find. Reviewed by: kib (earlier diff) Tested by: kris, pho (earlier diff)
2008-03-22 09:15:16 +00:00
BO_UNLOCK(bo);
vdrop(vp);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
}
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
/*
* Add an item to the syncer work queue.
*/
Protect vn worklist and vn->v_{clean,dirty}blkhd at splbio(). Get rid of extra LIST_REMOVE() Reviewed by: hsu@FreeBSD.ORG (Jeffrey Hsu), mckusick@McKusick.COM Submitted by: hsu@FreeBSD.ORG (Jeffrey Hsu), dillon@backplane.com ( Matthew Dillon )
1999-02-19 17:36:58 +00:00
static void
Move the syncer linkage from vnode to bufobj. This is not quite a perfect separation: the syncer still think it knows that everything is a vnode.
2004-10-27 08:05:02 +00:00
vn_syncer_add_to_worklist(struct bufobj *bo, int delay)
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
{
sync_vnode() has some messy code about locking in order to deal with mount fs needing Giant to be held when processing bufobjs. Use a different subqueue for pending workitems on filesystems requiring Giant. This simplifies the code notably and also reduces the number of Giant acquisitions (and the whole processing cost). Suggested by: jeff Reviewed by: kib Tested by: pho
2008-05-04 13:54:55 +00:00
int queue, slot;
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
Move the syncer linkage from vnode to bufobj. This is not quite a perfect separation: the syncer still think it knows that everything is a vnode.
2004-10-27 08:05:02 +00:00
ASSERT_BO_LOCKED(bo);
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
mtx_lock(&sync_mtx);
Move the syncer linkage from vnode to bufobj. This is not quite a perfect separation: the syncer still think it knows that everything is a vnode.
2004-10-27 08:05:02 +00:00
if (bo->bo_flag & BO_ONWORKLST)
LIST_REMOVE(bo, bo_synclist);
When shutting down the syncer kernel thread, first tell it to run faster and iterate to over its work list a few times in an attempt to empty the work list before the syncer terminates. This leaves fewer dirty blocks to be written at the "syncing disks" stage and keeps the the "giving up on N buffers" problem from being triggered by the presence of a large soft updates work list at system shutdown time. The downside is that the syncer takes noticeably longer to terminate. Tested by: "Arjan van Leeuwen" <avleeuwen AT piwebs DOT com> Approved by: mckusick
2004-07-01 23:59:19 +00:00
else {
Move the syncer linkage from vnode to bufobj. This is not quite a perfect separation: the syncer still think it knows that everything is a vnode.
2004-10-27 08:05:02 +00:00
bo->bo_flag |= BO_ONWORKLST;
Minor style cleanups (mostly removal of trailing whitespaces).
2007-04-10 15:29:37 +00:00
syncer_worklist_len++;
When shutting down the syncer kernel thread, first tell it to run faster and iterate to over its work list a few times in an attempt to empty the work list before the syncer terminates. This leaves fewer dirty blocks to be written at the "syncing disks" stage and keeps the the "giving up on N buffers" problem from being triggered by the presence of a large soft updates work list at system shutdown time. The downside is that the syncer takes noticeably longer to terminate. Tested by: "Arjan van Leeuwen" <avleeuwen AT piwebs DOT com> Approved by: mckusick
2004-07-01 23:59:19 +00:00
}
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
if (delay > syncer_maxdelay - 2)
delay = syncer_maxdelay - 2;
slot = (syncer_delayno + delay) & syncer_mask;
sync_vnode() has some messy code about locking in order to deal with mount fs needing Giant to be held when processing bufobjs. Use a different subqueue for pending workitems on filesystems requiring Giant. This simplifies the code notably and also reduces the number of Giant acquisitions (and the whole processing cost). Suggested by: jeff Reviewed by: kib Tested by: pho
2008-05-04 13:54:55 +00:00
queue = VFS_NEEDSGIANT(bo->__bo_vnode->v_mount) ? WI_GIANTQ :
WI_MPSAFEQ;
LIST_INSERT_HEAD(&syncer_workitem_pending[queue][slot], bo,
bo_synclist);
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
mtx_unlock(&sync_mtx);
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
}
Rework syncer termination code: Speed up the syncer when shutting down by sleeping for a shorter period of time instead of cranking up rushjob and using the normal one second sleep. Skip empty worklist slots when shutting down to avoid lengthy intervals of inactivity. Give I/O more time to complete between steps by not speeding the syncer quite as much. Terminate the syncer after one full pass through the worklist plus one second with the worklist containing nothing but syncer vnodes. Print an indication of shutdown progress to the console. Add a sysctl, vfs.worklist_len, to allow the size of the syncer worklist to be monitored.
2004-07-05 01:07:33 +00:00
static int
sysctl_vfs_worklist_len(SYSCTL_HANDLER_ARGS)
{
int error, len;
mtx_lock(&sync_mtx);
len = syncer_worklist_len - sync_vnode_count;
mtx_unlock(&sync_mtx);
error = SYSCTL_OUT(req, &len, sizeof(len));
return (error);
}
SYSCTL_PROC(_vfs, OID_AUTO, worklist_len, CTLTYPE_INT | CTLFLAG_RD, NULL, 0,
sysctl_vfs_worklist_len, "I", "Syncer thread worklist length");
Un-staticize runningbufwakeup() and staticize updateproc. Add a new private thread flag to indicate that the thread should not sleep if runningbufspace is too large. Set this flag on the bufdaemon and syncer threads so that they skip the waitrunningbufspace() call in bufwrite() rather than than checking the proc pointer vs. the known proc pointers for these two threads. A way of preventing these threads from being starved for I/O but still placing limits on their outstanding I/O would be desirable. Set this flag in ffs_copyonwrite() to prevent bufwrite() calls from blocking on the runningbufspace check while holding snaplk. This prevents snaplk from being held for an arbitrarily long period of time if runningbufspace is high and greatly reduces the contention for snaplk. The disadvantage is that ffs_copyonwrite() can start a large amount of I/O if there are a large number of snapshots, which could cause a deadlock in other parts of the code. Call runningbufwakeup() in ffs_copyonwrite() to decrement runningbufspace before attempting to grab snaplk so that I/O requests waiting on snaplk are not counted in runningbufspace as being in-progress. Increment runningbufspace again before actually launching the original I/O request. Prior to the above two changes, the system could deadlock if enough I/O requests were blocked by snaplk to prevent runningbufspace from falling below lorunningspace and one of the bawrite() calls in ffs_copyonwrite() blocked in waitrunningbufspace() while holding snaplk. See <http://www.holm.cc/stress/log/cons143.html>
2005-09-30 01:30:01 +00:00
static struct proc *updateproc;
Remove __P.
2002-03-19 21:25:46 +00:00
static void sched_sync(void);
Slight reorganization of kernel thread/process creation. Instead of using SYSINIT_KT() etc (which is a static, compile-time procedure), use a NetBSD-style kthread_create() interface. kproc_start is still available as a SYSINIT() hook. This allowed simplification of chunks of the sysinit code in the process. This kthread_create() is our old kproc_start internals, with the SYSINIT_KT fork hooks grafted in and tweaked to work the same as the NetBSD one. One thing I'd like to do shortly is get rid of nfsiod as a user initiated process. It makes sense for the nfs client code to create them on the fly as needed up to a user settable limit. This means that nfsiod doesn't need to be in /sbin and is always "available". This is a fair bit easier to do outside of the SYSINIT_KT() framework.
1999-07-01 13:21:46 +00:00
static struct kproc_desc up_kp = {
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
"syncer",
sched_sync,
&updateproc
};
In keeping with style(9)'s recommendations on macros, use a ';' after each SYSINIT() macro invocation. This makes a number of lightweight C parsers much happier with the FreeBSD kernel source, including cflow's prcc and lxr. MFC after: 1 month Discussed with: imp, rink
2008-03-16 10:58:09 +00:00
SYSINIT(syncer, SI_SUB_KTHREAD_UPDATE, SI_ORDER_FIRST, kproc_start, &up_kp);
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
Move the bit of the syncer which deals with vnodes into a separate function.
2004-11-14 15:24:38 +00:00
static int
Use curthread instead of the FIRST_THREAD_IN_PROC for vnlru and syncer, when applicable. Aquire Giant slightly later for vnlru. In the syncer, aquire the Giant only when a vnode belongs to the non-MPsafe fs. In both speedup_syncer() and syncer_shutdown(), remove the syncer thread from the lbolt sleep queue after the syncer state is modified, not before. Herded by: attilio Tested by: Peter Holm Reviewed by: ups MFC after: 1 week
2007-12-05 09:34:04 +00:00
sync_vnode(struct synclist *slp, struct bufobj **bo, struct thread *td)
Move the bit of the syncer which deals with vnodes into a separate function.
2004-11-14 15:24:38 +00:00
{
struct vnode *vp;
struct mount *mp;
Use curthread instead of the FIRST_THREAD_IN_PROC for vnlru and syncer, when applicable. Aquire Giant slightly later for vnlru. In the syncer, aquire the Giant only when a vnode belongs to the non-MPsafe fs. In both speedup_syncer() and syncer_shutdown(), remove the syncer thread from the lbolt sleep queue after the syncer state is modified, not before. Herded by: attilio Tested by: Peter Holm Reviewed by: ups MFC after: 1 week
2007-12-05 09:34:04 +00:00
*bo = LIST_FIRST(slp);
sync_vnode() has some messy code about locking in order to deal with mount fs needing Giant to be held when processing bufobjs. Use a different subqueue for pending workitems on filesystems requiring Giant. This simplifies the code notably and also reduces the number of Giant acquisitions (and the whole processing cost). Suggested by: jeff Reviewed by: kib Tested by: pho
2008-05-04 13:54:55 +00:00
if (*bo == NULL)
Use curthread instead of the FIRST_THREAD_IN_PROC for vnlru and syncer, when applicable. Aquire Giant slightly later for vnlru. In the syncer, aquire the Giant only when a vnode belongs to the non-MPsafe fs. In both speedup_syncer() and syncer_shutdown(), remove the syncer thread from the lbolt sleep queue after the syncer state is modified, not before. Herded by: attilio Tested by: Peter Holm Reviewed by: ups MFC after: 1 week
2007-12-05 09:34:04 +00:00
return (0);
vp = (*bo)->__bo_vnode; /* XXX */
sync_vnode() has some messy code about locking in order to deal with mount fs needing Giant to be held when processing bufobjs. Use a different subqueue for pending workitems on filesystems requiring Giant. This simplifies the code notably and also reduces the number of Giant acquisitions (and the whole processing cost). Suggested by: jeff Reviewed by: kib Tested by: pho
2008-05-04 13:54:55 +00:00
if (VOP_ISLOCKED(vp) != 0 || VI_TRYLOCK(vp) == 0)
Move the bit of the syncer which deals with vnodes into a separate function.
2004-11-14 15:24:38 +00:00
return (1);
/*
* We use vhold in case the vnode does not
* successfully sync. vhold prevents the vnode from
* going away when we unlock the sync_mtx so that
* we can acquire the vnode interlock.
*/
vholdl(vp);
mtx_unlock(&sync_mtx);
- Add the tunable and sysctl for the mpsafevfs. It currently defaults to off. - Protect access to mnt_kern_flag with the mointpoint mutex. - Remove some KASSERTs which are not legal checks without the appropriate locks held. - Use VCANRECYCLE() rather than rolling several slightly different checks together. - Return from vtryrecycle() with a recycled vnode rather than a locked vnode. This simplifies some locking. - Remove several GIANT_REQUIRED lines. - Add a few KASSERTs to help with INACT debugging. Sponsored By: Isilon Systems, Inc.
2005-01-24 10:41:01 +00:00
VI_UNLOCK(vp);
if (vn_start_write(vp, &mp, V_NOWAIT) != 0) {
vdrop(vp);
mtx_lock(&sync_mtx);
- Only return 1 from sync_vnode() in cases where the vnode is still at the head of the sync list. This prevents sched_sync() from re-queueing a vnode which may have been freed already. Discussed with: kib
2008-03-23 01:44:28 +00:00
return (*bo == LIST_FIRST(slp));
- Add the tunable and sysctl for the mpsafevfs. It currently defaults to off. - Protect access to mnt_kern_flag with the mointpoint mutex. - Remove some KASSERTs which are not legal checks without the appropriate locks held. - Use VCANRECYCLE() rather than rolling several slightly different checks together. - Return from vtryrecycle() with a recycled vnode rather than a locked vnode. This simplifies some locking. - Remove several GIANT_REQUIRED lines. - Add a few KASSERTs to help with INACT debugging. Sponsored By: Isilon Systems, Inc.
2005-01-24 10:41:01 +00:00
}
vn_lock() is currently only used with the 'curthread' passed as argument. Remove this argument and pass curthread directly to underlying VOP_LOCK1() VFS method. This modify makes the code cleaner and in particular remove an annoying dependence helping next lockmgr() cleanup. KPI results, obviously, changed. Manpage and FreeBSD_version will be updated through further commits. As a side note, would be valuable to say that next commits will address a similar cleanup about VFS methods, in particular vop_lock1 and vop_unlock. Tested by: Diego Sardina <siarodx at gmail dot com>, Andrea Di Pasquale <whyx dot it at gmail dot com>
2008-01-10 01:10:58 +00:00
vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
Remove the unused credential argument from VOP_FSYNC() and VFS_SYNC(). I'm not sure why a credential was added to these in the first place, it is not used anywhere and it doesn't make much sense: The credentials for syncing a file (ability to write to the file) should be checked at the system call level. Credentials for syncing one or more filesystems ("none") should be checked at the system call level as well. If the filesystem implementation needs a particular credential to carry out the syncing it would logically have to the cached mount credential, or a credential cached along with any delayed write data. Discussed with: rwatson
2005-01-11 07:36:22 +00:00
(void) VOP_FSYNC(vp, MNT_LAZY, td);
VOP_LOCK1() (and so VOP_LOCK()) and VOP_UNLOCK() are only used in conjuction with 'thread' argument passing which is always curthread. Remove the unuseful extra-argument and pass explicitly curthread to lower layer functions, when necessary. KPI results broken by this change, which should affect several ports, so version bumping and manpage update will be further committed. Tested by: kris, pho, Diego Sardina <siarodx at gmail dot com>
2008-01-13 14:44:15 +00:00
VOP_UNLOCK(vp, 0);
Move the bit of the syncer which deals with vnodes into a separate function.
2004-11-14 15:24:38 +00:00
vn_finished_write(mp);
- Complete part of the unfinished bufobj work by consistently using BO_LOCK/UNLOCK/MTX when manipulating the bufobj. - Create a new lock in the bufobj to lock bufobj fields independently. This leaves the vnode interlock as an 'identity' lock while the bufobj is an io lock. The bufobj lock is ordered before the vnode interlock and also before the mnt ilock. - Exploit this new lock order to simplify softdep_check_suspend(). - A few sync related functions are marked with a new XXX to note that we may not properly interlock against a non-zero bv_cnt when attempting to sync all vnodes on a mountlist. I do not believe this race is important. If I'm wrong this will make these locations easier to find. Reviewed by: kib (earlier diff) Tested by: kris, pho (earlier diff)
2008-03-22 09:15:16 +00:00
BO_LOCK(*bo);
Use curthread instead of the FIRST_THREAD_IN_PROC for vnlru and syncer, when applicable. Aquire Giant slightly later for vnlru. In the syncer, aquire the Giant only when a vnode belongs to the non-MPsafe fs. In both speedup_syncer() and syncer_shutdown(), remove the syncer thread from the lbolt sleep queue after the syncer state is modified, not before. Herded by: attilio Tested by: Peter Holm Reviewed by: ups MFC after: 1 week
2007-12-05 09:34:04 +00:00
if (((*bo)->bo_flag & BO_ONWORKLST) != 0) {
Move the bit of the syncer which deals with vnodes into a separate function.
2004-11-14 15:24:38 +00:00
/*
* Put us back on the worklist. The worklist
* routine will remove us from our current
* position and then add us back in at a later
* position.
*/
Use curthread instead of the FIRST_THREAD_IN_PROC for vnlru and syncer, when applicable. Aquire Giant slightly later for vnlru. In the syncer, aquire the Giant only when a vnode belongs to the non-MPsafe fs. In both speedup_syncer() and syncer_shutdown(), remove the syncer thread from the lbolt sleep queue after the syncer state is modified, not before. Herded by: attilio Tested by: Peter Holm Reviewed by: ups MFC after: 1 week
2007-12-05 09:34:04 +00:00
vn_syncer_add_to_worklist(*bo, syncdelay);
Move the bit of the syncer which deals with vnodes into a separate function.
2004-11-14 15:24:38 +00:00
}
- Complete part of the unfinished bufobj work by consistently using BO_LOCK/UNLOCK/MTX when manipulating the bufobj. - Create a new lock in the bufobj to lock bufobj fields independently. This leaves the vnode interlock as an 'identity' lock while the bufobj is an io lock. The bufobj lock is ordered before the vnode interlock and also before the mnt ilock. - Exploit this new lock order to simplify softdep_check_suspend(). - A few sync related functions are marked with a new XXX to note that we may not properly interlock against a non-zero bv_cnt when attempting to sync all vnodes on a mountlist. I do not believe this race is important. If I'm wrong this will make these locations easier to find. Reviewed by: kib (earlier diff) Tested by: kris, pho (earlier diff)
2008-03-22 09:15:16 +00:00
BO_UNLOCK(*bo);
vdrop(vp);
Move the bit of the syncer which deals with vnodes into a separate function.
2004-11-14 15:24:38 +00:00
mtx_lock(&sync_mtx);
return (0);
}
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
/*
* System filesystem synchronizer daemon.
*/
Be consistent about "static" functions: if the function is marked static in its prototype, mark it static at the definition too. Inspired by: FlexeLint warning #512
2002-09-28 17:15:38 +00:00
static void
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
sched_sync(void)
{
sync_vnode() has some messy code about locking in order to deal with mount fs needing Giant to be held when processing bufobjs. Use a different subqueue for pending workitems on filesystems requiring Giant. This simplifies the code notably and also reduces the number of Giant acquisitions (and the whole processing cost). Suggested by: jeff Reviewed by: kib Tested by: pho
2008-05-04 13:54:55 +00:00
struct synclist *gnext, *next;
struct synclist *gslp, *slp;
Move the syncer linkage from vnode to bufobj. This is not quite a perfect separation: the syncer still think it knows that everything is a vnode.
2004-10-27 08:05:02 +00:00
struct bufobj *bo;
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
long starttime;
Use curthread instead of the FIRST_THREAD_IN_PROC for vnlru and syncer, when applicable. Aquire Giant slightly later for vnlru. In the syncer, aquire the Giant only when a vnode belongs to the non-MPsafe fs. In both speedup_syncer() and syncer_shutdown(), remove the syncer thread from the lbolt sleep queue after the syncer state is modified, not before. Herded by: attilio Tested by: Peter Holm Reviewed by: ups MFC after: 1 week
2007-12-05 09:34:04 +00:00
struct thread *td = curthread;
Rework syncer termination code: Speed up the syncer when shutting down by sleeping for a shorter period of time instead of cranking up rushjob and using the normal one second sleep. Skip empty worklist slots when shutting down to avoid lengthy intervals of inactivity. Give I/O more time to complete between steps by not speeding the syncer quite as much. Terminate the syncer after one full pass through the worklist plus one second with the worklist containing nothing but syncer vnodes. Print an indication of shutdown progress to the console. Add a sysctl, vfs.worklist_len, to allow the size of the syncer worklist to be monitored.
2004-07-05 01:07:33 +00:00
int last_work_seen;
int net_worklist_len;
int syncer_final_iter;
Tidy up system shutdown.
2004-07-15 04:29:48 +00:00
int first_printf;
Move the bit of the syncer which deals with vnodes into a separate function.
2004-11-14 15:24:38 +00:00
int error;
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
Unconditionally set last_work_seen while in the SYNCER_RUNNING state so that last_work_seen has a reasonable value at the transition to the SYNCER_SHUTTING_DOWN state, even if net_worklist_len happened to be zero at the time. Initialize last_work_seen to zero as a safety measure in case the syncer never ran in the SYNCER_RUNNING state. Tested by: phk
2004-07-05 21:32:01 +00:00
last_work_seen = 0;
Rework syncer termination code: Speed up the syncer when shutting down by sleeping for a shorter period of time instead of cranking up rushjob and using the normal one second sleep. Skip empty worklist slots when shutting down to avoid lengthy intervals of inactivity. Give I/O more time to complete between steps by not speeding the syncer quite as much. Terminate the syncer after one full pass through the worklist plus one second with the worklist containing nothing but syncer vnodes. Print an indication of shutdown progress to the console. Add a sysctl, vfs.worklist_len, to allow the size of the syncer worklist to be monitored.
2004-07-05 01:07:33 +00:00
syncer_final_iter = 0;
Tidy up system shutdown.
2004-07-15 04:29:48 +00:00
first_printf = 1;
Rework syncer termination code: Speed up the syncer when shutting down by sleeping for a shorter period of time instead of cranking up rushjob and using the normal one second sleep. Skip empty worklist slots when shutting down to avoid lengthy intervals of inactivity. Give I/O more time to complete between steps by not speeding the syncer quite as much. Terminate the syncer after one full pass through the worklist plus one second with the worklist containing nothing but syncer vnodes. Print an indication of shutdown progress to the console. Add a sysctl, vfs.worklist_len, to allow the size of the syncer worklist to be monitored.
2004-07-05 01:07:33 +00:00
syncer_state = SYNCER_RUNNING;
use monotonic `time_uptime' instead of `time_second' Approved by: anholt (mentor) Discussed on: arch
2005-09-12 15:31:28 +00:00
starttime = time_uptime;
Un-staticize runningbufwakeup() and staticize updateproc. Add a new private thread flag to indicate that the thread should not sleep if runningbufspace is too large. Set this flag on the bufdaemon and syncer threads so that they skip the waitrunningbufspace() call in bufwrite() rather than than checking the proc pointer vs. the known proc pointers for these two threads. A way of preventing these threads from being starved for I/O but still placing limits on their outstanding I/O would be desirable. Set this flag in ffs_copyonwrite() to prevent bufwrite() calls from blocking on the runningbufspace check while holding snaplk. This prevents snaplk from being held for an arbitrarily long period of time if runningbufspace is high and greatly reduces the contention for snaplk. The disadvantage is that ffs_copyonwrite() can start a large amount of I/O if there are a large number of snapshots, which could cause a deadlock in other parts of the code. Call runningbufwakeup() in ffs_copyonwrite() to decrement runningbufspace before attempting to grab snaplk so that I/O requests waiting on snaplk are not counted in runningbufspace as being in-progress. Increment runningbufspace again before actually launching the original I/O request. Prior to the above two changes, the system could deadlock if enough I/O requests were blocked by snaplk to prevent runningbufspace from falling below lorunningspace and one of the bawrite() calls in ffs_copyonwrite() blocked in waitrunningbufspace() while holding snaplk. See <http://www.holm.cc/stress/log/cons143.html>
2005-09-30 01:30:01 +00:00
td->td_pflags |= TDP_NORUNNINGBUF;
Major update to the way synchronization is done in the kernel. Highlights include: * Mutual exclusion is used instead of spl*(). See mutex(9). (Note: The alpha port is still in transition and currently uses both.) * Per-CPU idle processes. * Interrupts are run in their own separate kernel threads and can be preempted (i386 only). Partially contributed by: BSDi (BSD/OS) Submissions by (at least): cp, dfr, dillon, grog, jake, jhb, sheldonh
2000-09-07 01:33:02 +00:00
When shutting down the syncer kernel thread, first tell it to run faster and iterate to over its work list a few times in an attempt to empty the work list before the syncer terminates. This leaves fewer dirty blocks to be written at the "syncing disks" stage and keeps the the "giving up on N buffers" problem from being triggered by the presence of a large soft updates work list at system shutdown time. The downside is that the syncer takes noticeably longer to terminate. Tested by: "Arjan van Leeuwen" <avleeuwen AT piwebs DOT com> Approved by: mckusick
2004-07-01 23:59:19 +00:00
EVENTHANDLER_REGISTER(shutdown_pre_sync, syncer_shutdown, td->td_proc,
Gratuitous whitespace cleanup.
2002-06-06 15:46:38 +00:00
SHUTDOWN_PRI_LAST);
Introduce a mechanism to suspend/resume system processes. Suspend syncer and bufdaemon prior to disk sync during system shutdown.
2000-01-07 08:36:44 +00:00
Simplify operations with sync_mtx in sched_sync(): - Don't drop the lock just to reacquire it again to check rushjob, this only wastes time. - Use msleep() to drop the mutex while sleeping instead of explicitly unlocking around tsleep. Reviewed by: pjd
2006-11-07 19:45:05 +00:00
mtx_lock(&sync_mtx);
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
for (;;) {
Rework syncer termination code: Speed up the syncer when shutting down by sleeping for a shorter period of time instead of cranking up rushjob and using the normal one second sleep. Skip empty worklist slots when shutting down to avoid lengthy intervals of inactivity. Give I/O more time to complete between steps by not speeding the syncer quite as much. Terminate the syncer after one full pass through the worklist plus one second with the worklist containing nothing but syncer vnodes. Print an indication of shutdown progress to the console. Add a sysctl, vfs.worklist_len, to allow the size of the syncer worklist to be monitored.
2004-07-05 01:07:33 +00:00
if (syncer_state == SYNCER_FINAL_DELAY &&
syncer_final_iter == 0) {
When shutting down the syncer kernel thread, first tell it to run faster and iterate to over its work list a few times in an attempt to empty the work list before the syncer terminates. This leaves fewer dirty blocks to be written at the "syncing disks" stage and keeps the the "giving up on N buffers" problem from being triggered by the presence of a large soft updates work list at system shutdown time. The downside is that the syncer takes noticeably longer to terminate. Tested by: "Arjan van Leeuwen" <avleeuwen AT piwebs DOT com> Approved by: mckusick
2004-07-01 23:59:19 +00:00
mtx_unlock(&sync_mtx);
Rename the kthread_xxx (e.g. kthread_create()) calls to kproc_xxx as they actually make whole processes. Thos makes way for us to add REAL kthread_create() and friends that actually make theads. it turns out that most of these calls actually end up being moved back to the thread version when it's added. but we need to make this cosmetic change first. I'd LOVE to do this rename in 7.0 so that we can eventually MFC the new kthread_xxx() calls.
2007-10-20 23:23:23 +00:00
kproc_suspend_check(td->td_proc);
When shutting down the syncer kernel thread, first tell it to run faster and iterate to over its work list a few times in an attempt to empty the work list before the syncer terminates. This leaves fewer dirty blocks to be written at the "syncing disks" stage and keeps the the "giving up on N buffers" problem from being triggered by the presence of a large soft updates work list at system shutdown time. The downside is that the syncer takes noticeably longer to terminate. Tested by: "Arjan van Leeuwen" <avleeuwen AT piwebs DOT com> Approved by: mckusick
2004-07-01 23:59:19 +00:00
mtx_lock(&sync_mtx);
}
Rework syncer termination code: Speed up the syncer when shutting down by sleeping for a shorter period of time instead of cranking up rushjob and using the normal one second sleep. Skip empty worklist slots when shutting down to avoid lengthy intervals of inactivity. Give I/O more time to complete between steps by not speeding the syncer quite as much. Terminate the syncer after one full pass through the worklist plus one second with the worklist containing nothing but syncer vnodes. Print an indication of shutdown progress to the console. Add a sysctl, vfs.worklist_len, to allow the size of the syncer worklist to be monitored.
2004-07-05 01:07:33 +00:00
net_worklist_len = syncer_worklist_len - sync_vnode_count;
Tidy up system shutdown.
2004-07-15 04:29:48 +00:00
if (syncer_state != SYNCER_RUNNING &&
use monotonic `time_uptime' instead of `time_second' Approved by: anholt (mentor) Discussed on: arch
2005-09-12 15:31:28 +00:00
starttime != time_uptime) {
Tidy up system shutdown.
2004-07-15 04:29:48 +00:00
if (first_printf) {
Skip the syncing disks loop if there are no dirty buffers. Remove a variable used to flag the initial printf. Submitted by: truckman (earlier version)
2004-08-10 01:32:05 +00:00
printf("\nSyncing disks, vnodes remaining...");
Tidy up system shutdown.
2004-07-15 04:29:48 +00:00
first_printf = 0;
}
Rework syncer termination code: Speed up the syncer when shutting down by sleeping for a shorter period of time instead of cranking up rushjob and using the normal one second sleep. Skip empty worklist slots when shutting down to avoid lengthy intervals of inactivity. Give I/O more time to complete between steps by not speeding the syncer quite as much. Terminate the syncer after one full pass through the worklist plus one second with the worklist containing nothing but syncer vnodes. Print an indication of shutdown progress to the console. Add a sysctl, vfs.worklist_len, to allow the size of the syncer worklist to be monitored.
2004-07-05 01:07:33 +00:00
printf("%d ", net_worklist_len);
Tidy up system shutdown.
2004-07-15 04:29:48 +00:00
}
use monotonic `time_uptime' instead of `time_second' Approved by: anholt (mentor) Discussed on: arch
2005-09-12 15:31:28 +00:00
starttime = time_uptime;
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
/*
Protect vn worklist and vn->v_{clean,dirty}blkhd at splbio(). Get rid of extra LIST_REMOVE() Reviewed by: hsu@FreeBSD.ORG (Jeffrey Hsu), mckusick@McKusick.COM Submitted by: hsu@FreeBSD.ORG (Jeffrey Hsu), dillon@backplane.com ( Matthew Dillon )
1999-02-19 17:36:58 +00:00
* Push files whose dirty time has expired. Be careful
* of interrupt race on slp queue.
Rework syncer termination code: Speed up the syncer when shutting down by sleeping for a shorter period of time instead of cranking up rushjob and using the normal one second sleep. Skip empty worklist slots when shutting down to avoid lengthy intervals of inactivity. Give I/O more time to complete between steps by not speeding the syncer quite as much. Terminate the syncer after one full pass through the worklist plus one second with the worklist containing nothing but syncer vnodes. Print an indication of shutdown progress to the console. Add a sysctl, vfs.worklist_len, to allow the size of the syncer worklist to be monitored.
2004-07-05 01:07:33 +00:00
*
* Skip over empty worklist slots when shutting down.
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
*/
Rework syncer termination code: Speed up the syncer when shutting down by sleeping for a shorter period of time instead of cranking up rushjob and using the normal one second sleep. Skip empty worklist slots when shutting down to avoid lengthy intervals of inactivity. Give I/O more time to complete between steps by not speeding the syncer quite as much. Terminate the syncer after one full pass through the worklist plus one second with the worklist containing nothing but syncer vnodes. Print an indication of shutdown progress to the console. Add a sysctl, vfs.worklist_len, to allow the size of the syncer worklist to be monitored.
2004-07-05 01:07:33 +00:00
do {
sync_vnode() has some messy code about locking in order to deal with mount fs needing Giant to be held when processing bufobjs. Use a different subqueue for pending workitems on filesystems requiring Giant. This simplifies the code notably and also reduces the number of Giant acquisitions (and the whole processing cost). Suggested by: jeff Reviewed by: kib Tested by: pho
2008-05-04 13:54:55 +00:00
slp = &syncer_workitem_pending[WI_MPSAFEQ][syncer_delayno];
gslp = &syncer_workitem_pending[WI_GIANTQ][syncer_delayno];
Rework syncer termination code: Speed up the syncer when shutting down by sleeping for a shorter period of time instead of cranking up rushjob and using the normal one second sleep. Skip empty worklist slots when shutting down to avoid lengthy intervals of inactivity. Give I/O more time to complete between steps by not speeding the syncer quite as much. Terminate the syncer after one full pass through the worklist plus one second with the worklist containing nothing but syncer vnodes. Print an indication of shutdown progress to the console. Add a sysctl, vfs.worklist_len, to allow the size of the syncer worklist to be monitored.
2004-07-05 01:07:33 +00:00
syncer_delayno += 1;
if (syncer_delayno == syncer_maxdelay)
syncer_delayno = 0;
sync_vnode() has some messy code about locking in order to deal with mount fs needing Giant to be held when processing bufobjs. Use a different subqueue for pending workitems on filesystems requiring Giant. This simplifies the code notably and also reduces the number of Giant acquisitions (and the whole processing cost). Suggested by: jeff Reviewed by: kib Tested by: pho
2008-05-04 13:54:55 +00:00
next = &syncer_workitem_pending[WI_MPSAFEQ][syncer_delayno];
gnext = &syncer_workitem_pending[WI_GIANTQ][syncer_delayno];
Rework syncer termination code: Speed up the syncer when shutting down by sleeping for a shorter period of time instead of cranking up rushjob and using the normal one second sleep. Skip empty worklist slots when shutting down to avoid lengthy intervals of inactivity. Give I/O more time to complete between steps by not speeding the syncer quite as much. Terminate the syncer after one full pass through the worklist plus one second with the worklist containing nothing but syncer vnodes. Print an indication of shutdown progress to the console. Add a sysctl, vfs.worklist_len, to allow the size of the syncer worklist to be monitored.
2004-07-05 01:07:33 +00:00
/*
* If the worklist has wrapped since the
Minor style cleanups (mostly removal of trailing whitespaces).
2007-04-10 15:29:37 +00:00
* it was emptied of all but syncer vnodes,
Rework syncer termination code: Speed up the syncer when shutting down by sleeping for a shorter period of time instead of cranking up rushjob and using the normal one second sleep. Skip empty worklist slots when shutting down to avoid lengthy intervals of inactivity. Give I/O more time to complete between steps by not speeding the syncer quite as much. Terminate the syncer after one full pass through the worklist plus one second with the worklist containing nothing but syncer vnodes. Print an indication of shutdown progress to the console. Add a sysctl, vfs.worklist_len, to allow the size of the syncer worklist to be monitored.
2004-07-05 01:07:33 +00:00
* switch to the FINAL_DELAY state and run
* for one more second.
*/
if (syncer_state == SYNCER_SHUTTING_DOWN &&
net_worklist_len == 0 &&
last_work_seen == syncer_delayno) {
syncer_state = SYNCER_FINAL_DELAY;
syncer_final_iter = SYNCER_SHUTDOWN_SPEEDUP;
}
} while (syncer_state != SYNCER_RUNNING && LIST_EMPTY(slp) &&
sync_vnode() has some messy code about locking in order to deal with mount fs needing Giant to be held when processing bufobjs. Use a different subqueue for pending workitems on filesystems requiring Giant. This simplifies the code notably and also reduces the number of Giant acquisitions (and the whole processing cost). Suggested by: jeff Reviewed by: kib Tested by: pho
2008-05-04 13:54:55 +00:00
LIST_EMPTY(gslp) && syncer_worklist_len > 0);
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
Rework syncer termination code: Speed up the syncer when shutting down by sleeping for a shorter period of time instead of cranking up rushjob and using the normal one second sleep. Skip empty worklist slots when shutting down to avoid lengthy intervals of inactivity. Give I/O more time to complete between steps by not speeding the syncer quite as much. Terminate the syncer after one full pass through the worklist plus one second with the worklist containing nothing but syncer vnodes. Print an indication of shutdown progress to the console. Add a sysctl, vfs.worklist_len, to allow the size of the syncer worklist to be monitored.
2004-07-05 01:07:33 +00:00
/*
* Keep track of the last time there was anything
* on the worklist other than syncer vnodes.
* Return to the SHUTTING_DOWN state if any
* new work appears.
*/
Unconditionally set last_work_seen while in the SYNCER_RUNNING state so that last_work_seen has a reasonable value at the transition to the SYNCER_SHUTTING_DOWN state, even if net_worklist_len happened to be zero at the time. Initialize last_work_seen to zero as a safety measure in case the syncer never ran in the SYNCER_RUNNING state. Tested by: phk
2004-07-05 21:32:01 +00:00
if (net_worklist_len > 0 || syncer_state == SYNCER_RUNNING)
Rework syncer termination code: Speed up the syncer when shutting down by sleeping for a shorter period of time instead of cranking up rushjob and using the normal one second sleep. Skip empty worklist slots when shutting down to avoid lengthy intervals of inactivity. Give I/O more time to complete between steps by not speeding the syncer quite as much. Terminate the syncer after one full pass through the worklist plus one second with the worklist containing nothing but syncer vnodes. Print an indication of shutdown progress to the console. Add a sysctl, vfs.worklist_len, to allow the size of the syncer worklist to be monitored.
2004-07-05 01:07:33 +00:00
last_work_seen = syncer_delayno;
Unconditionally set last_work_seen while in the SYNCER_RUNNING state so that last_work_seen has a reasonable value at the transition to the SYNCER_SHUTTING_DOWN state, even if net_worklist_len happened to be zero at the time. Initialize last_work_seen to zero as a safety measure in case the syncer never ran in the SYNCER_RUNNING state. Tested by: phk
2004-07-05 21:32:01 +00:00
if (net_worklist_len > 0 && syncer_state == SYNCER_FINAL_DELAY)
syncer_state = SYNCER_SHUTTING_DOWN;
Use curthread instead of the FIRST_THREAD_IN_PROC for vnlru and syncer, when applicable. Aquire Giant slightly later for vnlru. In the syncer, aquire the Giant only when a vnode belongs to the non-MPsafe fs. In both speedup_syncer() and syncer_shutdown(), remove the syncer thread from the lbolt sleep queue after the syncer state is modified, not before. Herded by: attilio Tested by: Peter Holm Reviewed by: ups MFC after: 1 week
2007-12-05 09:34:04 +00:00
while (!LIST_EMPTY(slp)) {
error = sync_vnode(slp, &bo, td);
Move the bit of the syncer which deals with vnodes into a separate function.
2004-11-14 15:24:38 +00:00
if (error == 1) {
Move the syncer linkage from vnode to bufobj. This is not quite a perfect separation: the syncer still think it knows that everything is a vnode.
2004-10-27 08:05:02 +00:00
LIST_REMOVE(bo, bo_synclist);
LIST_INSERT_HEAD(next, bo, bo_synclist);
- In sched_sync() test our preconditions prior to dropping the sync_mtx. This is so that we may grab the interlock while still holding the sync_mtx. We have to VI_TRYLOCK() because in all other cases the lock order runs the other way. - If we don't meet any of the preconditions, reinsert the vp into the list for the next second. - We don't need to panic if we fail to sync here because each FSYNC function handles this case. Removing this redundant code also simplifies locking.
2003-10-04 18:03:53 +00:00
continue;
}
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
}
sync_vnode() has some messy code about locking in order to deal with mount fs needing Giant to be held when processing bufobjs. Use a different subqueue for pending workitems on filesystems requiring Giant. This simplifies the code notably and also reduces the number of Giant acquisitions (and the whole processing cost). Suggested by: jeff Reviewed by: kib Tested by: pho
2008-05-04 13:54:55 +00:00
if (!LIST_EMPTY(gslp)) {
mtx_unlock(&sync_mtx);
mtx_lock(&Giant);
mtx_lock(&sync_mtx);
while (!LIST_EMPTY(gslp)) {
error = sync_vnode(gslp, &bo, td);
if (error == 1) {
LIST_REMOVE(bo, bo_synclist);
LIST_INSERT_HEAD(gnext, bo,
bo_synclist);
continue;
}
}
mtx_unlock(&Giant);
}
Rework syncer termination code: Speed up the syncer when shutting down by sleeping for a shorter period of time instead of cranking up rushjob and using the normal one second sleep. Skip empty worklist slots when shutting down to avoid lengthy intervals of inactivity. Give I/O more time to complete between steps by not speeding the syncer quite as much. Terminate the syncer after one full pass through the worklist plus one second with the worklist containing nothing but syncer vnodes. Print an indication of shutdown progress to the console. Add a sysctl, vfs.worklist_len, to allow the size of the syncer worklist to be monitored.
2004-07-05 01:07:33 +00:00
if (syncer_state == SYNCER_FINAL_DELAY && syncer_final_iter > 0)
syncer_final_iter--;
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
/*
* The variable rushjob allows the kernel to speed up the
* processing of the filesystem syncer process. A rushjob
* value of N tells the filesystem syncer to process the next
* N seconds worth of work on its queue ASAP. Currently rushjob
* is used by the soft update code to speed up the filesystem
* syncer process when the incore state is getting so far
* ahead of the disk that the kernel memory pool is being
* threatened with exhaustion.
*/
if (rushjob > 0) {
rushjob -= 1;
continue;
Rework syncer termination code: Speed up the syncer when shutting down by sleeping for a shorter period of time instead of cranking up rushjob and using the normal one second sleep. Skip empty worklist slots when shutting down to avoid lengthy intervals of inactivity. Give I/O more time to complete between steps by not speeding the syncer quite as much. Terminate the syncer after one full pass through the worklist plus one second with the worklist containing nothing but syncer vnodes. Print an indication of shutdown progress to the console. Add a sysctl, vfs.worklist_len, to allow the size of the syncer worklist to be monitored.
2004-07-05 01:07:33 +00:00
}
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
/*
Fix comment typo and function declaration.
2006-11-07 19:07:33 +00:00
* Just sleep for a short period of time between
Rework syncer termination code: Speed up the syncer when shutting down by sleeping for a shorter period of time instead of cranking up rushjob and using the normal one second sleep. Skip empty worklist slots when shutting down to avoid lengthy intervals of inactivity. Give I/O more time to complete between steps by not speeding the syncer quite as much. Terminate the syncer after one full pass through the worklist plus one second with the worklist containing nothing but syncer vnodes. Print an indication of shutdown progress to the console. Add a sysctl, vfs.worklist_len, to allow the size of the syncer worklist to be monitored.
2004-07-05 01:07:33 +00:00
* iterations when shutting down to allow some I/O
* to happen.
*
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
* If it has taken us less than a second to process the
* current work, then wait. Otherwise start right over
* again. We can still lose time if any single round
* takes more than two seconds, but it does not really
* matter as we are just trying to generally pace the
* filesystem activity.
*/
Rework syncer termination code: Speed up the syncer when shutting down by sleeping for a shorter period of time instead of cranking up rushjob and using the normal one second sleep. Skip empty worklist slots when shutting down to avoid lengthy intervals of inactivity. Give I/O more time to complete between steps by not speeding the syncer quite as much. Terminate the syncer after one full pass through the worklist plus one second with the worklist containing nothing but syncer vnodes. Print an indication of shutdown progress to the console. Add a sysctl, vfs.worklist_len, to allow the size of the syncer worklist to be monitored.
2004-07-05 01:07:33 +00:00
if (syncer_state != SYNCER_RUNNING)
Remove the use of lbolt from the VFS syncer. It seems we only use `lbolt' inside the VFS syncer and the TTY layer now. Because I'm planning to replace the TTY layer next month, there's no reason to keep `lbolt' if it's only used in a single thread inside the kernel. Because the syncer code wanted to wake up the syncer thread before the timeout, it called sleepq_remove(). Because we now just use a condvar(9) with a timeout value of `hz', we can wake it up using cv_broadcast() without waking up any unrelated threads. Reviewed by: phk
2008-07-30 12:39:18 +00:00
cv_timedwait(&sync_wakeup, &sync_mtx,
Rework syncer termination code: Speed up the syncer when shutting down by sleeping for a shorter period of time instead of cranking up rushjob and using the normal one second sleep. Skip empty worklist slots when shutting down to avoid lengthy intervals of inactivity. Give I/O more time to complete between steps by not speeding the syncer quite as much. Terminate the syncer after one full pass through the worklist plus one second with the worklist containing nothing but syncer vnodes. Print an indication of shutdown progress to the console. Add a sysctl, vfs.worklist_len, to allow the size of the syncer worklist to be monitored.
2004-07-05 01:07:33 +00:00
hz / SYNCER_SHUTDOWN_SPEEDUP);
use monotonic `time_uptime' instead of `time_second' Approved by: anholt (mentor) Discussed on: arch
2005-09-12 15:31:28 +00:00
else if (time_uptime == starttime)
Remove the use of lbolt from the VFS syncer. It seems we only use `lbolt' inside the VFS syncer and the TTY layer now. Because I'm planning to replace the TTY layer next month, there's no reason to keep `lbolt' if it's only used in a single thread inside the kernel. Because the syncer code wanted to wake up the syncer thread before the timeout, it called sleepq_remove(). Because we now just use a condvar(9) with a timeout value of `hz', we can wake it up using cv_broadcast() without waking up any unrelated threads. Reviewed by: phk
2008-07-30 12:39:18 +00:00
cv_timedwait(&sync_wakeup, &sync_mtx, hz);
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
}
}
Get rid of the global variable rushjob and replace it with a function in kern/vfs_subr.c named speedup_syncer() which handles the speedup request. Change the various clients of rushjob to use the new function.
1999-06-15 23:37:29 +00:00
/*
* Request the syncer daemon to speed up its work.
* We never push it to speed up more than half of its
* normal turn time, otherwise it could take over the cpu.
*/
int
Fix comment typo and function declaration.
2006-11-07 19:07:33 +00:00
speedup_syncer(void)
Get rid of the global variable rushjob and replace it with a function in kern/vfs_subr.c named speedup_syncer() which handles the speedup request. Change the various clients of rushjob to use the new function.
1999-06-15 23:37:29 +00:00
{
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
int ret = 0;
Get rid of the global variable rushjob and replace it with a function in kern/vfs_subr.c named speedup_syncer() which handles the speedup request. Change the various clients of rushjob to use the new function.
1999-06-15 23:37:29 +00:00
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
mtx_lock(&sync_mtx);
Get rid of the global variable rushjob and replace it with a function in kern/vfs_subr.c named speedup_syncer() which handles the speedup request. Change the various clients of rushjob to use the new function.
1999-06-15 23:37:29 +00:00
if (rushjob < syncdelay / 2) {
rushjob += 1;
stat_rush_requests += 1;
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
ret = 1;
Get rid of the global variable rushjob and replace it with a function in kern/vfs_subr.c named speedup_syncer() which handles the speedup request. Change the various clients of rushjob to use the new function.
1999-06-15 23:37:29 +00:00
}
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
mtx_unlock(&sync_mtx);
Remove the use of lbolt from the VFS syncer. It seems we only use `lbolt' inside the VFS syncer and the TTY layer now. Because I'm planning to replace the TTY layer next month, there's no reason to keep `lbolt' if it's only used in a single thread inside the kernel. Because the syncer code wanted to wake up the syncer thread before the timeout, it called sleepq_remove(). Because we now just use a condvar(9) with a timeout value of `hz', we can wake it up using cv_broadcast() without waking up any unrelated threads. Reviewed by: phk
2008-07-30 12:39:18 +00:00
cv_broadcast(&sync_wakeup);
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
return (ret);
Get rid of the global variable rushjob and replace it with a function in kern/vfs_subr.c named speedup_syncer() which handles the speedup request. Change the various clients of rushjob to use the new function.
1999-06-15 23:37:29 +00:00
}
When shutting down the syncer kernel thread, first tell it to run faster and iterate to over its work list a few times in an attempt to empty the work list before the syncer terminates. This leaves fewer dirty blocks to be written at the "syncing disks" stage and keeps the the "giving up on N buffers" problem from being triggered by the presence of a large soft updates work list at system shutdown time. The downside is that the syncer takes noticeably longer to terminate. Tested by: "Arjan van Leeuwen" <avleeuwen AT piwebs DOT com> Approved by: mckusick
2004-07-01 23:59:19 +00:00
/*
* Tell the syncer to speed up its work and run though its work
* list several times, then tell it to shut down.
*/
static void
syncer_shutdown(void *arg, int howto)
{
Don't attempt to trigger the syncer thread final sync code in the shutdown_pre_sync state if the RB_NOSYNC flag is set. This is the likely cause of hangs after a system panic that are keeping crash dumps from being done. This is a MFC candidate for RELENG_5. MFC after: 3 days
2004-08-20 19:21:47 +00:00
if (howto & RB_NOSYNC)
return;
When shutting down the syncer kernel thread, first tell it to run faster and iterate to over its work list a few times in an attempt to empty the work list before the syncer terminates. This leaves fewer dirty blocks to be written at the "syncing disks" stage and keeps the the "giving up on N buffers" problem from being triggered by the presence of a large soft updates work list at system shutdown time. The downside is that the syncer takes noticeably longer to terminate. Tested by: "Arjan van Leeuwen" <avleeuwen AT piwebs DOT com> Approved by: mckusick
2004-07-01 23:59:19 +00:00
mtx_lock(&sync_mtx);
Rework syncer termination code: Speed up the syncer when shutting down by sleeping for a shorter period of time instead of cranking up rushjob and using the normal one second sleep. Skip empty worklist slots when shutting down to avoid lengthy intervals of inactivity. Give I/O more time to complete between steps by not speeding the syncer quite as much. Terminate the syncer after one full pass through the worklist plus one second with the worklist containing nothing but syncer vnodes. Print an indication of shutdown progress to the console. Add a sysctl, vfs.worklist_len, to allow the size of the syncer worklist to be monitored.
2004-07-05 01:07:33 +00:00
syncer_state = SYNCER_SHUTTING_DOWN;
rushjob = 0;
When shutting down the syncer kernel thread, first tell it to run faster and iterate to over its work list a few times in an attempt to empty the work list before the syncer terminates. This leaves fewer dirty blocks to be written at the "syncing disks" stage and keeps the the "giving up on N buffers" problem from being triggered by the presence of a large soft updates work list at system shutdown time. The downside is that the syncer takes noticeably longer to terminate. Tested by: "Arjan van Leeuwen" <avleeuwen AT piwebs DOT com> Approved by: mckusick
2004-07-01 23:59:19 +00:00
mtx_unlock(&sync_mtx);
Remove the use of lbolt from the VFS syncer. It seems we only use `lbolt' inside the VFS syncer and the TTY layer now. Because I'm planning to replace the TTY layer next month, there's no reason to keep `lbolt' if it's only used in a single thread inside the kernel. Because the syncer code wanted to wake up the syncer thread before the timeout, it called sleepq_remove(). Because we now just use a condvar(9) with a timeout value of `hz', we can wake it up using cv_broadcast() without waking up any unrelated threads. Reviewed by: phk
2008-07-30 12:39:18 +00:00
cv_broadcast(&sync_wakeup);
When shutting down the syncer kernel thread, first tell it to run faster and iterate to over its work list a few times in an attempt to empty the work list before the syncer terminates. This leaves fewer dirty blocks to be written at the "syncing disks" stage and keeps the the "giving up on N buffers" problem from being triggered by the presence of a large soft updates work list at system shutdown time. The downside is that the syncer takes noticeably longer to terminate. Tested by: "Arjan van Leeuwen" <avleeuwen AT piwebs DOT com> Approved by: mckusick
2004-07-01 23:59:19 +00:00
kproc_shutdown(arg, howto);
}
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
/*
* Reassign a buffer from one vnode to another.
* Used to assign file specific control information
* (indirect blocks) to the vnode to which they belong.
*/
The big 4.4BSD Lite to FreeBSD 2.0.0 (Development) patch. Reviewed by: Rodney W. Grimes Submitted by: John Dyson and David Greenman
1994-05-25 09:21:21 +00:00
void
Eliminate unused second argument to reassignbuf() and simplify it accordingly.
2004-07-25 21:24:23 +00:00
reassignbuf(struct buf *bp)
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
{
- In reassignbuf() don't unlock vp and lock newvp if they are the same. Doing so creates a race where the buf is on neither list. - Only vfree() in an error case in vclean() if VSHOULDFREE() thinks we should. - Convert the error case in vclean() to INVARIANTS from DIAGNOSTIC as this really should not happen and is fast to check.
2003-09-20 00:21:48 +00:00
struct vnode *vp;
Add b_bufobj to struct buf which eventually will eliminate the need for b_vp. Initialize b_bufobj for all buffers. Make incore() and gbincore() take a bufobj instead of a vnode. Make inmem() local to vfs_bio.c Change a lot of VI_[UN]LOCK(bp->b_vp) to BO_[UN]LOCK(bp->b_bufobj) also VI_MTX() to BO_MTX(), Make buf_vlist_add() take a bufobj instead of a vnode. Eliminate other uses of bp->b_vp where bp->b_bufobj will do. Various minor polishing: remove "register", turn panic into KASSERT, use new function declarations, TAILQ_FOREACH_SAFE() etc.
2004-10-22 08:47:20 +00:00
struct bufobj *bo;
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
int delay;
- In reassignbuf() add many asserts to validate the head and tail pointers of the clean and dirty lists. This is in an attempt to catch the wrong bufobj problem sooner. - In vgonel() don't acquire an extra reference in the active case, the vnode lock and VI_DOOMED protect us from recursively cleaning. - Also in vgonel() clean up some stale comments. Sponsored by: Isilon Systems, Inc. Approved by: re (blanket vfs)
2005-06-14 20:31:53 +00:00
#ifdef INVARIANTS
struct bufv *bv;
#endif
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
- In reassignbuf() don't unlock vp and lock newvp if they are the same. Doing so creates a race where the buf is on neither list. - Only vfree() in an error case in vclean() if VSHOULDFREE() thinks we should. - Convert the error case in vclean() to INVARIANTS from DIAGNOSTIC as this really should not happen and is fast to check.
2003-09-20 00:21:48 +00:00
vp = bp->b_vp;
Add b_bufobj to struct buf which eventually will eliminate the need for b_vp. Initialize b_bufobj for all buffers. Make incore() and gbincore() take a bufobj instead of a vnode. Make inmem() local to vfs_bio.c Change a lot of VI_[UN]LOCK(bp->b_vp) to BO_[UN]LOCK(bp->b_bufobj) also VI_MTX() to BO_MTX(), Make buf_vlist_add() take a bufobj instead of a vnode. Eliminate other uses of bp->b_vp where bp->b_bufobj will do. Various minor polishing: remove "register", turn panic into KASSERT, use new function declarations, TAILQ_FOREACH_SAFE() etc.
2004-10-22 08:47:20 +00:00
bo = bp->b_bufobj;
The buffer queue mechanism has been reformulated. Instead of having QUEUE_AGE, QUEUE_LRU, and QUEUE_EMPTY we instead have QUEUE_CLEAN, QUEUE_DIRTY, QUEUE_EMPTY, and QUEUE_EMPTYKVA. With this patch clean and dirty buffers have been separated. Empty buffers with KVM assignments have been separated from truely empty buffers. getnewbuf() has been rewritten and now operates in a 100% optimal fashion. That is, it is able to find precisely the right kind of buffer it needs to allocate a new buffer, defragment KVM, or to free-up an existing buffer when the buffer cache is full (which is a steady-state situation for the buffer cache). Buffer flushing has been reorganized. Previously buffers were flushed in the context of whatever process hit the conditions forcing buffer flushing to occur. This resulted in processes blocking on conditions unrelated to what they were doing. This also resulted in inappropriate VFS stacking chains due to multiple processes getting stuck trying to flush dirty buffers or due to a single process getting into a situation where it might attempt to flush buffers recursively - a situation that was only partially fixed in prior commits. We have added a new daemon called the buf_daemon which is responsible for flushing dirty buffers when the number of dirty buffers exceeds the vfs.hidirtybuffers limit. This daemon attempts to dynamically adjust the rate at which dirty buffers are flushed such that getnewbuf() calls (almost) never block. The number of nbufs and amount of buffer space is now scaled past the 8MB limit that was previously imposed for systems with over 64MB of memory, and the vfs.{lo,hi}dirtybuffers limits have been relaxed somewhat. The number of physical buffers has been increased with the intention that we will manage physical I/O differently in the future. reassignbuf previously attempted to keep the dirtyblkhd list sorted which could result in non-deterministic operation under certain conditions, such as when a large number of dirty buffers are being managed. This algorithm has been changed. reassignbuf now keeps buffers locally sorted if it can do so cheaply, and otherwise gives up and adds buffers to the head of the dirtyblkhd list. The new algorithm is deterministic but not perfect. The new algorithm greatly reduces problems that previously occured when write_behind was turned off in the system. The P_FLSINPROG proc->p_flag bit has been replaced by the more descriptive P_BUFEXHAUST bit. This bit allows processes working with filesystem buffers to use available emergency reserves. Normal processes do not set this bit and are not allowed to dig into emergency reserves. The purpose of this bit is to avoid low-memory deadlocks. A small race condition was fixed in getpbuf() in vm/vm_pager.c. Submitted by: Matthew Dillon <dillon@apollo.backplane.com> Reviewed by: Kirk McKusick <mckusick@mckusick.com>
1999-07-04 00:25:38 +00:00
++reassignbufcalls;
Certain vnode buffer list operations were not being spl protected, and they needed to be. Brelse for example can be called at interrupt level, and the buffer list operations were not being protected from it.
1996-08-15 06:45:01 +00:00
- Add the tunable and sysctl for the mpsafevfs. It currently defaults to off. - Protect access to mnt_kern_flag with the mointpoint mutex. - Remove some KASSERTs which are not legal checks without the appropriate locks held. - Use VCANRECYCLE() rather than rolling several slightly different checks together. - Return from vtryrecycle() with a recycled vnode rather than a locked vnode. This simplifies some locking. - Remove several GIANT_REQUIRED lines. - Add a few KASSERTs to help with INACT debugging. Sponsored By: Isilon Systems, Inc.
2005-01-24 10:41:01 +00:00
CTR3(KTR_BUF, "reassignbuf(%p) vp %p flags %X",
bp, bp->b_vp, bp->b_flags);
This is a rather large commit that encompasses the new swapper, changes to the VM system to support the new swapper, VM bug fixes, several VM optimizations, and some additional revamping of the VM code. The specific bug fixes will be documented with additional forced commits. This commit is somewhat rough in regards to code cleanup issues. Reviewed by: "John S. Dyson" <root@dyson.iquest.net>, "David Greenman" <dg@root.com>
1999-01-21 08:29:12 +00:00
/*
* B_PAGING flagged buffers cannot be reassigned because their vp
* is not fully linked in.
*/
if (bp->b_flags & B_PAGING)
panic("cannot reassign paging buffer");
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
/*
* Delete from old vnode list, if on one.
*/
- Complete part of the unfinished bufobj work by consistently using BO_LOCK/UNLOCK/MTX when manipulating the bufobj. - Create a new lock in the bufobj to lock bufobj fields independently. This leaves the vnode interlock as an 'identity' lock while the bufobj is an io lock. The bufobj lock is ordered before the vnode interlock and also before the mnt ilock. - Exploit this new lock order to simplify softdep_check_suspend(). - A few sync related functions are marked with a new XXX to note that we may not properly interlock against a non-zero bv_cnt when attempting to sync all vnodes on a mountlist. I do not believe this race is important. If I'm wrong this will make these locations easier to find. Reviewed by: kib (earlier diff) Tested by: kris, pho (earlier diff)
2008-03-22 09:15:16 +00:00
BO_LOCK(bo);
Eliminate unused second argument to reassignbuf() and simplify it accordingly.
2004-07-25 21:24:23 +00:00
if (bp->b_xflags & (BX_VNDIRTY | BX_VNCLEAN))
Replace the global buffer hash table with per-vnode splay trees using a methodology similar to the vm_map_entry splay and the VM splay that Alan Cox is working on. Extensive testing has appeared to have shown no increase in overhead. Disadvantages Dirties more cache lines during lookups. Not as fast as a hash table lookup (but still N log N and optimal when there is locality of reference). Advantages vnode->v_dirtyblkhd is now perfectly sorted, making fsync/sync/filesystem syncer operate more efficiently. I get to rip out all the old hacks (some of which were mine) that tried to keep the v_dirtyblkhd tailq sorted. The per-vnode splay tree should be easier to lock / SMPng pushdown on vnodes will be easier. This commit along with another that Alan is working on for the VM page global hash table will allow me to implement ranged fsync(), optimize server-side nfs commit rpcs, and implement partial syncs by the filesystem syncer (aka filesystem syncer would detect that someone is trying to get the vnode lock, remembers its place, and skip to the next vnode). Note that the buffer cache splay is somewhat more complex then other splays due to special handling of background bitmap writes (multiple buffers with the same lblkno in the same vnode), and B_INVAL discontinuities between the old hash table and the existence of the buffer on the v_cleanblkhd list. Suggested by: alc
2002-07-10 17:02:32 +00:00
buf_vlist_remove(bp);
- All buffers should either be clean or dirty. If neither of these flags are set when we attempt to remove a buffer from a queue we should panic. Hopefully this will catch the source of the wrong bufobj panics. Sponsored by: Isilon Systems, Inc.
2005-05-01 12:00:36 +00:00
else
panic("reassignbuf: Buffer %p not on queue.", bp);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
/*
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
* If dirty, put on list of dirty buffers; otherwise insert onto list
* of clean buffers.
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
*/
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
if (bp->b_flags & B_DELWRI) {
Move the syncer linkage from vnode to bufobj. This is not quite a perfect separation: the syncer still think it knows that everything is a vnode.
2004-10-27 08:05:02 +00:00
if ((bo->bo_flag & BO_ONWORKLST) == 0) {
Eliminate unused second argument to reassignbuf() and simplify it accordingly.
2004-07-25 21:24:23 +00:00
switch (vp->v_type) {
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
case VDIR:
Get rid of the global variable rushjob and replace it with a function in kern/vfs_subr.c named speedup_syncer() which handles the speedup request. Change the various clients of rushjob to use the new function.
1999-06-15 23:37:29 +00:00
delay = dirdelay;
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
break;
Convert various pieces of code to use vn_isdisk() rather than checking for vp->v_type == VBLK. In ccd: we don't need to call VOP_GETATTR to find the type of a vnode. Reviewed by: sos
1999-11-22 10:33:55 +00:00
case VCHR:
Remove a left over from userland buffer-cache access to disks.
2004-06-14 14:25:03 +00:00
delay = metadelay;
break;
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
default:
Get rid of the global variable rushjob and replace it with a function in kern/vfs_subr.c named speedup_syncer() which handles the speedup request. Change the various clients of rushjob to use the new function.
1999-06-15 23:37:29 +00:00
delay = filedelay;
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
}
Move the syncer linkage from vnode to bufobj. This is not quite a perfect separation: the syncer still think it knows that everything is a vnode.
2004-10-27 08:05:02 +00:00
vn_syncer_add_to_worklist(bo, delay);
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
}
Add b_bufobj to struct buf which eventually will eliminate the need for b_vp. Initialize b_bufobj for all buffers. Make incore() and gbincore() take a bufobj instead of a vnode. Make inmem() local to vfs_bio.c Change a lot of VI_[UN]LOCK(bp->b_vp) to BO_[UN]LOCK(bp->b_bufobj) also VI_MTX() to BO_MTX(), Make buf_vlist_add() take a bufobj instead of a vnode. Eliminate other uses of bp->b_vp where bp->b_bufobj will do. Various minor polishing: remove "register", turn panic into KASSERT, use new function declarations, TAILQ_FOREACH_SAFE() etc.
2004-10-22 08:47:20 +00:00
buf_vlist_add(bp, bo, BX_VNDIRTY);
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
} else {
Add b_bufobj to struct buf which eventually will eliminate the need for b_vp. Initialize b_bufobj for all buffers. Make incore() and gbincore() take a bufobj instead of a vnode. Make inmem() local to vfs_bio.c Change a lot of VI_[UN]LOCK(bp->b_vp) to BO_[UN]LOCK(bp->b_bufobj) also VI_MTX() to BO_MTX(), Make buf_vlist_add() take a bufobj instead of a vnode. Eliminate other uses of bp->b_vp where bp->b_bufobj will do. Various minor polishing: remove "register", turn panic into KASSERT, use new function declarations, TAILQ_FOREACH_SAFE() etc.
2004-10-22 08:47:20 +00:00
buf_vlist_add(bp, bo, BX_VNCLEAN);
Replace the global buffer hash table with per-vnode splay trees using a methodology similar to the vm_map_entry splay and the VM splay that Alan Cox is working on. Extensive testing has appeared to have shown no increase in overhead. Disadvantages Dirties more cache lines during lookups. Not as fast as a hash table lookup (but still N log N and optimal when there is locality of reference). Advantages vnode->v_dirtyblkhd is now perfectly sorted, making fsync/sync/filesystem syncer operate more efficiently. I get to rip out all the old hacks (some of which were mine) that tried to keep the v_dirtyblkhd tailq sorted. The per-vnode splay tree should be easier to lock / SMPng pushdown on vnodes will be easier. This commit along with another that Alan is working on for the VM page global hash table will allow me to implement ranged fsync(), optimize server-side nfs commit rpcs, and implement partial syncs by the filesystem syncer (aka filesystem syncer would detect that someone is trying to get the vnode lock, remembers its place, and skip to the next vnode). Note that the buffer cache splay is somewhat more complex then other splays due to special handling of background bitmap writes (multiple buffers with the same lblkno in the same vnode), and B_INVAL discontinuities between the old hash table and the existence of the buffer on the v_cleanblkhd list. Suggested by: alc
2002-07-10 17:02:32 +00:00
Move the syncer linkage from vnode to bufobj. This is not quite a perfect separation: the syncer still think it knows that everything is a vnode.
2004-10-27 08:05:02 +00:00
if ((bo->bo_flag & BO_ONWORKLST) && bo->bo_dirty.bv_cnt == 0) {
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
mtx_lock(&sync_mtx);
Move the syncer linkage from vnode to bufobj. This is not quite a perfect separation: the syncer still think it knows that everything is a vnode.
2004-10-27 08:05:02 +00:00
LIST_REMOVE(bo, bo_synclist);
Minor style cleanups (mostly removal of trailing whitespaces).
2007-04-10 15:29:37 +00:00
syncer_worklist_len--;
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
mtx_unlock(&sync_mtx);
Move the syncer linkage from vnode to bufobj. This is not quite a perfect separation: the syncer still think it knows that everything is a vnode.
2004-10-27 08:05:02 +00:00
bo->bo_flag &= ~BO_ONWORKLST;
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
}
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
}
- In reassignbuf() add many asserts to validate the head and tail pointers of the clean and dirty lists. This is in an attempt to catch the wrong bufobj problem sooner. - In vgonel() don't acquire an extra reference in the active case, the vnode lock and VI_DOOMED protect us from recursively cleaning. - Also in vgonel() clean up some stale comments. Sponsored by: Isilon Systems, Inc. Approved by: re (blanket vfs)
2005-06-14 20:31:53 +00:00
#ifdef INVARIANTS
bv = &bo->bo_clean;
bp = TAILQ_FIRST(&bv->bv_hd);
KASSERT(bp == NULL || bp->b_bufobj == bo,
("bp %p wrong b_bufobj %p should be %p", bp, bp->b_bufobj, bo));
bp = TAILQ_LAST(&bv->bv_hd, buflists);
KASSERT(bp == NULL || bp->b_bufobj == bo,
("bp %p wrong b_bufobj %p should be %p", bp, bp->b_bufobj, bo));
bv = &bo->bo_dirty;
bp = TAILQ_FIRST(&bv->bv_hd);
KASSERT(bp == NULL || bp->b_bufobj == bo,
("bp %p wrong b_bufobj %p should be %p", bp, bp->b_bufobj, bo));
bp = TAILQ_LAST(&bv->bv_hd, buflists);
KASSERT(bp == NULL || bp->b_bufobj == bo,
("bp %p wrong b_bufobj %p should be %p", bp, bp->b_bufobj, bo));
#endif
- Complete part of the unfinished bufobj work by consistently using BO_LOCK/UNLOCK/MTX when manipulating the bufobj. - Create a new lock in the bufobj to lock bufobj fields independently. This leaves the vnode interlock as an 'identity' lock while the bufobj is an io lock. The bufobj lock is ordered before the vnode interlock and also before the mnt ilock. - Exploit this new lock order to simplify softdep_check_suspend(). - A few sync related functions are marked with a new XXX to note that we may not properly interlock against a non-zero bv_cnt when attempting to sync all vnodes on a mountlist. I do not believe this race is important. If I'm wrong this will make these locations easier to find. Reviewed by: kib (earlier diff) Tested by: kris, pho (earlier diff)
2008-03-22 09:15:16 +00:00
BO_UNLOCK(bo);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
}
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
/*
* Increment the use and hold counts on the vnode, taking care to reference
* the driver's usecount if this is a chardev. The vholdl() will remove
* the vnode from the free list if it is presently free. Requires the
* vnode interlock and returns with it held.
*/
static void
v_incr_usecount(struct vnode *vp)
{
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
CTR2(KTR_VFS, "%s: vp %p", __func__, vp);
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
vp->v_usecount++;
if (vp->v_type == VCHR && vp->v_rdev != NULL) {
dev_lock();
vp->v_rdev->si_usecount++;
dev_unlock();
}
vholdl(vp);
}
- Solve a race where we could lose a call to VOP_INACTIVE. If vget() waiting on a lock held the last usecount ref on a vnode and the lock failed we would not call INACTIVE. Solve this by only holding a holdcnt to prevent the vnode from disappearing while we wait on vn_lock. Other callers may now VOP_INACTIVE while we are waiting on the lock, however this race is acceptable, while losing INACTIVE is not. Discussed with: kan, pjd Tested by: kkenn Sponsored by: Isilon Systems, Inc. MFC After: 1 week
2006-02-01 00:30:05 +00:00
/*
* Turn a holdcnt into a use+holdcnt such that only one call to
* v_decr_usecount is needed.
*/
static void
v_upgrade_usecount(struct vnode *vp)
{
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
CTR2(KTR_VFS, "%s: vp %p", __func__, vp);
- Solve a race where we could lose a call to VOP_INACTIVE. If vget() waiting on a lock held the last usecount ref on a vnode and the lock failed we would not call INACTIVE. Solve this by only holding a holdcnt to prevent the vnode from disappearing while we wait on vn_lock. Other callers may now VOP_INACTIVE while we are waiting on the lock, however this race is acceptable, while losing INACTIVE is not. Discussed with: kan, pjd Tested by: kkenn Sponsored by: Isilon Systems, Inc. MFC After: 1 week
2006-02-01 00:30:05 +00:00
vp->v_usecount++;
if (vp->v_type == VCHR && vp->v_rdev != NULL) {
dev_lock();
vp->v_rdev->si_usecount++;
dev_unlock();
}
}
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
/*
* Decrement the vnode use and hold count along with the driver's usecount
* if this is a chardev. The vdropl() below releases the vnode interlock
* as it may free the vnode.
*/
static void
v_decr_usecount(struct vnode *vp)
{
ASSERT_VI_LOCKED(vp, __FUNCTION__);
VNASSERT(vp->v_usecount > 0, vp,
("v_decr_usecount: negative usecount"));
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
CTR2(KTR_VFS, "%s: vp %p", __func__, vp);
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
vp->v_usecount--;
if (vp->v_type == VCHR && vp->v_rdev != NULL) {
dev_lock();
vp->v_rdev->si_usecount--;
dev_unlock();
}
vdropl(vp);
}
/*
* Decrement only the use count and driver use count. This is intended to
* be paired with a follow on vdropl() to release the remaining hold count.
* In this way we may vgone() a vnode with a 0 usecount without risk of
* having it end up on a free list because the hold count is kept above 0.
*/
Fix the spechash lock order reversal by keeping an updated sum of v_usecount in the dev_t which vcount() can return without locking any vnodes. Seen by: jhb
2002-10-24 19:38:56 +00:00
static void
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
v_decr_useonly(struct vnode *vp)
Fix the spechash lock order reversal by keeping an updated sum of v_usecount in the dev_t which vcount() can return without locking any vnodes. Seen by: jhb
2002-10-24 19:38:56 +00:00
{
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
ASSERT_VI_LOCKED(vp, __FUNCTION__);
VNASSERT(vp->v_usecount > 0, vp,
("v_decr_useonly: negative usecount"));
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
CTR2(KTR_VFS, "%s: vp %p", __func__, vp);
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
vp->v_usecount--;
In vrele() we can actually have a VCHR with v_rdev == NULL if we came from the bottom of addaliasu(). Don't panic.
2002-10-25 07:58:25 +00:00
if (vp->v_type == VCHR && vp->v_rdev != NULL) {
Do not refcount the cdevsw, but rather maintain a cdev->si_threadcount of the number of threads which are inside whatever is behind the cdevsw for this particular cdev. Make the device mutex visible through dev_lock() and dev_unlock(). We may want finer granularity later. Replace spechash_mtx use with dev_lock()/dev_unlock().
2004-09-23 07:17:41 +00:00
dev_lock();
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
vp->v_rdev->si_usecount--;
Do not refcount the cdevsw, but rather maintain a cdev->si_threadcount of the number of threads which are inside whatever is behind the cdevsw for this particular cdev. Make the device mutex visible through dev_lock() and dev_unlock(). We may want finer granularity later. Replace spechash_mtx use with dev_lock()/dev_unlock().
2004-09-23 07:17:41 +00:00
dev_unlock();
Fix the spechash lock order reversal by keeping an updated sum of v_usecount in the dev_t which vcount() can return without locking any vnodes. Seen by: jhb
2002-10-24 19:38:56 +00:00
}
}
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
/*
* Grab a particular vnode from the free list, increment its
- Greatly simplify vget() by removing the guarantee that any new references to a vnode with VI_OWEINACT set will force the vinactive() call. The kernel makes no guarantees about which reference was the last to close a file or when the actual inactive processing will happen. The previous code was designed to preserve existing semantics in the face of shared locks, however, this was unnecessary. Discussed with: mckusick
2008-03-24 04:22:58 +00:00
* reference count and lock it. VI_DOOMED is set if the vnode
* is being destroyed. Only callers who specify LK_RETRY will
* see doomed vnodes. If inactive processing was delayed in
* vput try to do it here.
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
*/
The big 4.4BSD Lite to FreeBSD 2.0.0 (Development) patch. Reviewed by: Rodney W. Grimes Submitted by: John Dyson and David Greenman
1994-05-25 09:21:21 +00:00
int
- Solve a race where we could lose a call to VOP_INACTIVE. If vget() waiting on a lock held the last usecount ref on a vnode and the lock failed we would not call INACTIVE. Solve this by only holding a holdcnt to prevent the vnode from disappearing while we wait on vn_lock. Other callers may now VOP_INACTIVE while we are waiting on the lock, however this race is acceptable, while losing INACTIVE is not. Discussed with: kan, pjd Tested by: kkenn Sponsored by: Isilon Systems, Inc. MFC After: 1 week
2006-02-01 00:30:05 +00:00
vget(struct vnode *vp, int flags, struct thread *td)
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
{
This is the kernel Lite/2 commit. There are some requisite userland changes, so don't expect to be able to run the kernel as-is (very well) without the appropriate Lite/2 userland changes. The system boots and can mount UFS filesystems. Untested: ext2fs, msdosfs, NFS Known problems: Incorrect Berkeley ID strings in some files. Mount_std mounts will not work until the getfsent library routine is changed. Reviewed by: various people Submitted by: Jeffery Hsu <hsu@freebsd.org>
1997-02-10 02:22:35 +00:00
int error;
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
- Remove vx_lock, vx_unlock, vx_wait, etc. - Add a vn_start_write/vn_finished_write around vlrureclaim so we don't do writing ops without suspending. This could suspend the vlruproc which should not be a problem under normal circumstances. - Manually implement VMIGHTFREE in vlrureclaim as this was the only instance where it was used. - Acquire a lock before calling vgone() as it now requires it. - Move the acquisition of the vnode interlock from vtryrecycle() to getnewvnode() so that if it fails we don't drop and reacquire the vnode_free_list_mtx. - Check for a usecount or holdcount at the end of vtryrecycle() in case someone grabbed a ref while we were recycling. Abort the recycle, and on the final ref drop this vnode will be placed on the head of the free list. - Move the redundant VOP_INACTIVE protection code into the local vinactive() routine to avoid code bloat. - Keep the vnode lock held across calls to vgone() in several places. - vgonel() no longer uses XLOCK, instead callers must hold an exclusive vnode lock. The VI_DOOMED flag is set to allow other threads to detect a vnode which is no longer valid. This flag is set until the last reference is gone, and there are no chances for a new ref. vgonel() holds this lock across the entire function, which greatly simplifies logic. _ Only vfree() in one place in vgone() not three. - Adjust vget() to check the VI_DOOMED flag prior to waiting on the lock in the LK_NOWAIT case. In other cases, check after we have slept and acquired an exlusive lock. This will simulate the old vx_wait() behavior. Sponsored by: Isilon Systems, Inc.
2005-03-13 11:54:28 +00:00
error = 0;
- Solve a race where we could lose a call to VOP_INACTIVE. If vget() waiting on a lock held the last usecount ref on a vnode and the lock failed we would not call INACTIVE. Solve this by only holding a holdcnt to prevent the vnode from disappearing while we wait on vn_lock. Other callers may now VOP_INACTIVE while we are waiting on the lock, however this race is acceptable, while losing INACTIVE is not. Discussed with: kan, pjd Tested by: kkenn Sponsored by: Isilon Systems, Inc. MFC After: 1 week
2006-02-01 00:30:05 +00:00
VFS_ASSERT_GIANT(vp->v_mount);
- Disallow XFS mounting in write mode. The write support never worked really and there is no need to maintain it. - Fix vn_get() in order to let it call vget(9) with a valid locking request. vget(9) returns the vnode locked in order to prevent recycling, but in this case internal XFS locks alredy prevent it from happening, so it is safe to drop the vnode lock before to return by vn_get(). - Add a VNASSERT() in vget(9) in order to catch malformed locking requests. Discussed with: kan, kib Tested by: Lothar Braun <lothar at lobraun dot de>
2008-07-21 23:01:09 +00:00
VNASSERT((flags & LK_TYPE_MASK) != 0, vp,
("vget: invalid lock operation"));
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
CTR3(KTR_VFS, "%s: vp %p with flags %d", __func__, vp, flags);
- Remove vx_lock, vx_unlock, vx_wait, etc. - Add a vn_start_write/vn_finished_write around vlrureclaim so we don't do writing ops without suspending. This could suspend the vlruproc which should not be a problem under normal circumstances. - Manually implement VMIGHTFREE in vlrureclaim as this was the only instance where it was used. - Acquire a lock before calling vgone() as it now requires it. - Move the acquisition of the vnode interlock from vtryrecycle() to getnewvnode() so that if it fails we don't drop and reacquire the vnode_free_list_mtx. - Check for a usecount or holdcount at the end of vtryrecycle() in case someone grabbed a ref while we were recycling. Abort the recycle, and on the final ref drop this vnode will be placed on the head of the free list. - Move the redundant VOP_INACTIVE protection code into the local vinactive() routine to avoid code bloat. - Keep the vnode lock held across calls to vgone() in several places. - vgonel() no longer uses XLOCK, instead callers must hold an exclusive vnode lock. The VI_DOOMED flag is set to allow other threads to detect a vnode which is no longer valid. This flag is set until the last reference is gone, and there are no chances for a new ref. vgonel() holds this lock across the entire function, which greatly simplifies logic. _ Only vfree() in one place in vgone() not three. - Adjust vget() to check the VI_DOOMED flag prior to waiting on the lock in the LK_NOWAIT case. In other cases, check after we have slept and acquired an exlusive lock. This will simulate the old vx_wait() behavior. Sponsored by: Isilon Systems, Inc.
2005-03-13 11:54:28 +00:00
if ((flags & LK_INTERLOCK) == 0)
VI_LOCK(vp);
- Solve a race where we could lose a call to VOP_INACTIVE. If vget() waiting on a lock held the last usecount ref on a vnode and the lock failed we would not call INACTIVE. Solve this by only holding a holdcnt to prevent the vnode from disappearing while we wait on vn_lock. Other callers may now VOP_INACTIVE while we are waiting on the lock, however this race is acceptable, while losing INACTIVE is not. Discussed with: kan, pjd Tested by: kkenn Sponsored by: Isilon Systems, Inc. MFC After: 1 week
2006-02-01 00:30:05 +00:00
vholdl(vp);
vn_lock() is currently only used with the 'curthread' passed as argument. Remove this argument and pass curthread directly to underlying VOP_LOCK1() VFS method. This modify makes the code cleaner and in particular remove an annoying dependence helping next lockmgr() cleanup. KPI results, obviously, changed. Manpage and FreeBSD_version will be updated through further commits. As a side note, would be valuable to say that next commits will address a similar cleanup about VFS methods, in particular vop_lock1 and vop_unlock. Tested by: Diego Sardina <siarodx at gmail dot com>, Andrea Di Pasquale <whyx dot it at gmail dot com>
2008-01-10 01:10:58 +00:00
if ((error = vn_lock(vp, flags | LK_INTERLOCK)) != 0) {
- Solve a race where we could lose a call to VOP_INACTIVE. If vget() waiting on a lock held the last usecount ref on a vnode and the lock failed we would not call INACTIVE. Solve this by only holding a holdcnt to prevent the vnode from disappearing while we wait on vn_lock. Other callers may now VOP_INACTIVE while we are waiting on the lock, however this race is acceptable, while losing INACTIVE is not. Discussed with: kan, pjd Tested by: kkenn Sponsored by: Isilon Systems, Inc. MFC After: 1 week
2006-02-01 00:30:05 +00:00
vdrop(vp);
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
CTR2(KTR_VFS, "%s: impossible to lock vnode %p", __func__,
vp);
- Change the VOP_LOCK UPGRADE in vput() to do a LK_NOWAIT to avoid a potential lock order reversal. Also, don't unlock the vnode if this fails, lockmgr has already unlocked it for us. - Restructure vget() now that vn_lock() does all of VI_DOOMED checking for us and also handles the case where there is no real lock type. - If VI_OWEINACT is set, we need to upgrade the lock request to EXCLUSIVE so that we can call inactive. It's not legal to vget a vnode that hasn't had INACTIVE called yet. Sponsored by: Isilon Systems, Inc.
2005-04-11 09:28:32 +00:00
return (error);
- Remove vx_lock, vx_unlock, vx_wait, etc. - Add a vn_start_write/vn_finished_write around vlrureclaim so we don't do writing ops without suspending. This could suspend the vlruproc which should not be a problem under normal circumstances. - Manually implement VMIGHTFREE in vlrureclaim as this was the only instance where it was used. - Acquire a lock before calling vgone() as it now requires it. - Move the acquisition of the vnode interlock from vtryrecycle() to getnewvnode() so that if it fails we don't drop and reacquire the vnode_free_list_mtx. - Check for a usecount or holdcount at the end of vtryrecycle() in case someone grabbed a ref while we were recycling. Abort the recycle, and on the final ref drop this vnode will be placed on the head of the free list. - Move the redundant VOP_INACTIVE protection code into the local vinactive() routine to avoid code bloat. - Keep the vnode lock held across calls to vgone() in several places. - vgonel() no longer uses XLOCK, instead callers must hold an exclusive vnode lock. The VI_DOOMED flag is set to allow other threads to detect a vnode which is no longer valid. This flag is set until the last reference is gone, and there are no chances for a new ref. vgonel() holds this lock across the entire function, which greatly simplifies logic. _ Only vfree() in one place in vgone() not three. - Adjust vget() to check the VI_DOOMED flag prior to waiting on the lock in the LK_NOWAIT case. In other cases, check after we have slept and acquired an exlusive lock. This will simulate the old vx_wait() behavior. Sponsored by: Isilon Systems, Inc.
2005-03-13 11:54:28 +00:00
}
- Greatly simplify vget() by removing the guarantee that any new references to a vnode with VI_OWEINACT set will force the vinactive() call. The kernel makes no guarantees about which reference was the last to close a file or when the actual inactive processing will happen. The previous code was designed to preserve existing semantics in the face of shared locks, however, this was unnecessary. Discussed with: mckusick
2008-03-24 04:22:58 +00:00
if (vp->v_iflag & VI_DOOMED && (flags & LK_RETRY) == 0)
panic("vget: vn_lock failed to return ENOENT\n");
- Solve a race where we could lose a call to VOP_INACTIVE. If vget() waiting on a lock held the last usecount ref on a vnode and the lock failed we would not call INACTIVE. Solve this by only holding a holdcnt to prevent the vnode from disappearing while we wait on vn_lock. Other callers may now VOP_INACTIVE while we are waiting on the lock, however this race is acceptable, while losing INACTIVE is not. Discussed with: kan, pjd Tested by: kkenn Sponsored by: Isilon Systems, Inc. MFC After: 1 week
2006-02-01 00:30:05 +00:00
VI_LOCK(vp);
/* Upgrade our holdcnt to a usecount. */
v_upgrade_usecount(vp);
- Greatly simplify vget() by removing the guarantee that any new references to a vnode with VI_OWEINACT set will force the vinactive() call. The kernel makes no guarantees about which reference was the last to close a file or when the actual inactive processing will happen. The previous code was designed to preserve existing semantics in the face of shared locks, however, this was unnecessary. Discussed with: mckusick
2008-03-24 04:22:58 +00:00
/*
* We don't guarantee that any particular close will
* trigger inactive processing so just make a best effort
* here at preventing a reference to a removed file. If
* we don't succeed no harm is done.
*/
if (vp->v_iflag & VI_OWEINACT) {
if (VOP_ISLOCKED(vp) == LK_EXCLUSIVE &&
(flags & LK_NOWAIT) == 0)
- Change the VOP_LOCK UPGRADE in vput() to do a LK_NOWAIT to avoid a potential lock order reversal. Also, don't unlock the vnode if this fails, lockmgr has already unlocked it for us. - Restructure vget() now that vn_lock() does all of VI_DOOMED checking for us and also handles the case where there is no real lock type. - If VI_OWEINACT is set, we need to upgrade the lock request to EXCLUSIVE so that we can call inactive. It's not legal to vget a vnode that hasn't had INACTIVE called yet. Sponsored by: Isilon Systems, Inc.
2005-04-11 09:28:32 +00:00
vinactive(vp, td);
- Greatly simplify vget() by removing the guarantee that any new references to a vnode with VI_OWEINACT set will force the vinactive() call. The kernel makes no guarantees about which reference was the last to close a file or when the actual inactive processing will happen. The previous code was designed to preserve existing semantics in the face of shared locks, however, this was unnecessary. Discussed with: mckusick
2008-03-24 04:22:58 +00:00
vp->v_iflag &= ~VI_OWEINACT;
}
VI_UNLOCK(vp);
This is the kernel Lite/2 commit. There are some requisite userland changes, so don't expect to be able to run the kernel as-is (very well) without the appropriate Lite/2 userland changes. The system boots and can mount UFS filesystems. Untested: ext2fs, msdosfs, NFS Known problems: Incorrect Berkeley ID strings in some files. Mount_std mounts will not work until the getfsent library routine is changed. Reviewed by: various people Submitted by: Jeffery Hsu <hsu@freebsd.org>
1997-02-10 02:22:35 +00:00
return (0);
}
Lots of improvements, including restructring the caching and management of vnodes and objects. There are some metadata performance improvements that come along with this. There are also a few prototypes added when the need is noticed. Changes include: 1) Cleaning up vref, vget. 2) Removal of the object cache. 3) Nuke vnode_pager_uncache and friends, because they aren't needed anymore. 4) Correct some missing LK_RETRY's in vn_lock. 5) Correct the page range in the code for msync. Be gentle, and please give me feedback asap.
1997-12-29 00:25:11 +00:00
Gratuitous whitespace cleanup.
2002-06-06 15:46:38 +00:00
/*
Style fixes based on comments by bde
2000-10-05 18:22:46 +00:00
* Increase the reference count of a vnode.
Style fixes: * Add lots of comments * Convert a couple of assertions to KASSERT() * Minimal whitespace & misapplied {} fixes * Convert #if 0 to #if COMPILING_LINT for code we presently do not support, but want to keep available. Reviewed by: adrian, markm
2000-09-22 12:22:36 +00:00
*/
Add the vnode interlock back around vref.
1997-12-29 16:54:03 +00:00
void
vref(struct vnode *vp)
{
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
CTR2(KTR_VFS, "%s: vp %p", __func__, vp);
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
VI_LOCK(vp);
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
v_incr_usecount(vp);
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
VI_UNLOCK(vp);
Add the vnode interlock back around vref.
1997-12-29 16:54:03 +00:00
}
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
/*
* Return reference count of a vnode.
*
* The results of this call are only guaranteed when some mechanism other
* than the VI lock is used to stop other processes from gaining references
* to the vnode. This may be the case if the caller holds the only reference.
* This is also useful when stale data is acceptable as race conditions may
* be accounted for by some other means.
*/
int
vrefcnt(struct vnode *vp)
{
int usecnt;
VI_LOCK(vp);
usecnt = vp->v_usecount;
VI_UNLOCK(vp);
return (usecnt);
}
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
/*
Lots of improvements, including restructring the caching and management of vnodes and objects. There are some metadata performance improvements that come along with this. There are also a few prototypes added when the need is noticed. Changes include: 1) Cleaning up vref, vget. 2) Removal of the object cache. 3) Nuke vnode_pager_uncache and friends, because they aren't needed anymore. 4) Correct some missing LK_RETRY's in vn_lock. 5) Correct the page range in the code for msync. Be gentle, and please give me feedback asap.
1997-12-29 00:25:11 +00:00
* Vnode put/release.
* If count drops to zero, call inactive routine and return to freelist.
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
*/
The big 4.4BSD Lite to FreeBSD 2.0.0 (Development) patch. Reviewed by: Rodney W. Grimes Submitted by: John Dyson and David Greenman
1994-05-25 09:21:21 +00:00
void
Convert remaining functions in vfs_subr.c from K&R prototypes to ANSI C prototypes, as the majority of new functions added have been in this style. Changing prototype style now results in gcc noticing that the implementation of vn_pollrecord() has a 'short' argument instead of 'int' as prototyped in vnode.h, so correct that definition. In practice this didn't matter as only poll flags in the lower 16 bits are used. MFC after: 1 week
2006-01-21 19:42:10 +00:00
vrele(struct vnode *vp)
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
{
KSE Milestone 2 Note ALL MODULES MUST BE RECOMPILED make the kernel aware that there are smaller units of scheduling than the process. (but only allow one thread per process at this time). This is functionally equivalent to teh previousl -current except that there is a thread associated with each process. Sorry john! (your next MFC will be a doosie!) Reviewed by: peter@freebsd.org, dillon@freebsd.org X-MFC after: ha ha ha ha
2001-09-12 08:38:13 +00:00
struct thread *td = curthread; /* XXX */
Lots of improvements, including restructring the caching and management of vnodes and objects. There are some metadata performance improvements that come along with this. There are also a few prototypes added when the need is noticed. Changes include: 1) Cleaning up vref, vget. 2) Removal of the object cache. 3) Nuke vnode_pager_uncache and friends, because they aren't needed anymore. 4) Correct some missing LK_RETRY's in vn_lock. 5) Correct the page range in the code for msync. Be gentle, and please give me feedback asap.
1997-12-29 00:25:11 +00:00
KNFize, by bde.
1999-01-10 01:58:29 +00:00
KASSERT(vp != NULL, ("vrele: null vp"));
- Solve a race where we could lose a call to VOP_INACTIVE. If vget() waiting on a lock held the last usecount ref on a vnode and the lock failed we would not call INACTIVE. Solve this by only holding a holdcnt to prevent the vnode from disappearing while we wait on vn_lock. Other callers may now VOP_INACTIVE while we are waiting on the lock, however this race is acceptable, while losing INACTIVE is not. Discussed with: kan, pjd Tested by: kkenn Sponsored by: Isilon Systems, Inc. MFC After: 1 week
2006-02-01 00:30:05 +00:00
VFS_ASSERT_GIANT(vp->v_mount);
Split DIAGNOSTIC -> DIAGNOSTIC, INVARIANTS, and INVARIANT_SUPPORT as discussed on -hackers. Introduce 'KASSERT(assertion, ("panic message", args))' for simple check + panic. Reviewed by: msmith
1999-01-08 17:31:30 +00:00
- Replace v_flag with v_iflag and v_vflag - v_vflag is protected by the vnode lock and is used when synchronization with VOP calls is needed. - v_iflag is protected by interlock and is used for dealing with vnode management issues. These flags include X/O LOCK, FREE, DOOMED, etc. - All accesses to v_iflag and v_vflag have either been locked or marked with mp_fixme's. - Many ASSERT_VOP_LOCKED calls have been added where the locking was not clear. - Many functions in vfs_subr.c were restructured to provide for stronger locking. Idea stolen from: BSD/OS
2002-08-04 10:29:36 +00:00
VI_LOCK(vp);
Even though this looks like it, this is not a complex code change. The interface into the "VMIO" system has changed to be more consistant and robust. Essentially, it is now no longer necessary to call vn_open to get merged VM/Buffer cache operation, and exceptional conditions such as merged operation of VBLK devices is simpler and more correct. This code corrects a potentially large set of problems including the problems with ktrace output and loaded systems, file create/deletes, etc. Most of the changes to NFS are cosmetic and name changes, eliminating a layer of subroutine calls. The direct calls to vput/vrele have been re-instituted for better cross platform compatibility. Reviewed by: davidg
1996-08-21 21:56:23 +00:00
In vrele() and vput(), avoid triggering the confusing "missed vn_close" KASSERT when vp->v_usecount is zero or negative. In this case, the "v*: negative ref cnt" panic that follows is much more appropriate. Reviewed by: mckusick
2001-05-11 20:42:41 +00:00
/* Skip this v_writecount check if we're going to panic below. */
Convert KASSERTS to VNASSERTS
2005-02-17 10:28:58 +00:00
VNASSERT(vp->v_writecount < vp->v_usecount || vp->v_usecount < 1, vp,
In vrele() and vput(), avoid triggering the confusing "missed vn_close" KASSERT when vp->v_usecount is zero or negative. In this case, the "v*: negative ref cnt" panic that follows is much more appropriate. Reviewed by: mckusick
2001-05-11 20:42:41 +00:00
("vrele: missed vn_close"));
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
CTR2(KTR_VFS, "%s: vp %p", __func__, vp);
Move KASSERTs which checks value of v_usecount after vnode locking, so it will not produce wrong alarms.
2000-10-02 09:57:06 +00:00
Add a new vnode flag VI_DOINGINACT to indicate that a VOP_INACTIVE call is in progress on the vnode. When vput() or vrele() sees a 1->0 reference count transition, it now return without any further action if this flag is set. This flag is necessary to avoid recursion into VOP_INACTIVE if the filesystem inactive routine causes the reference count to increase and then drop back to zero. It is also used to guarantee that an unlocked vnode will not be recycled while blocked in VOP_INACTIVE(). There are at least two cases where the recursion can occur: one is that the softupdates code called by ufs_inactive() via ffs_truncate() can call vput() on the vnode. This has been reported by many people as "lockmgr: draining against myself" panics. The other case is that nfs_inactive() can call vget() and then vrele() on the vnode to clean up a sillyrename file. Reviewed by: mckusick (an older version of the patch)
2002-12-29 18:30:49 +00:00
if (vp->v_usecount > 1 || ((vp->v_iflag & VI_DOINGINACT) &&
vp->v_usecount == 1)) {
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
v_decr_usecount(vp);
Make our v_usecount vnode reference count work identically to the original BSD code. The association between the vnode and the vm_object no longer includes reference counts. The major difference is that vm_object's are no longer freed gratuitiously from the vnode, and so once an object is created for the vnode, it will last as long as the vnode does. When a vnode object reference count is incremented, then the underlying vnode reference count is incremented also. The two "objects" are now more intimately related, and so the interactions are now much less complex. When vnodes are now normally placed onto the free queue with an object still attached. The rundown of the object happens at vnode rundown time, and happens with exactly the same filesystem semantics of the original VFS code. There is absolutely no need for vnode_pager_uncache and other travesties like that anymore. A side-effect of these changes is that SMP locking should be much simpler, the I/O copyin/copyout optimizations work, NFS should be more ponderable, and further work on layered filesystems should be less frustrating, because of the totally coherent management of the vnode objects and vnodes. Please be careful with your system while running this code, but I would greatly appreciate feedback as soon a reasonably possible.
1998-01-06 05:26:17 +00:00
return;
}
- Slightly rearrange vrele() to move the common case in one indentation level. Sponsored by: Isilon Systems, Inc.
2005-03-14 07:16:55 +00:00
if (vp->v_usecount != 1) {
Lots of improvements, including restructring the caching and management of vnodes and objects. There are some metadata performance improvements that come along with this. There are also a few prototypes added when the need is noticed. Changes include: 1) Cleaning up vref, vget. 2) Removal of the object cache. 3) Nuke vnode_pager_uncache and friends, because they aren't needed anymore. 4) Correct some missing LK_RETRY's in vn_lock. 5) Correct the page range in the code for msync. Be gentle, and please give me feedback asap.
1997-12-29 00:25:11 +00:00
#ifdef DIAGNOSTIC
vprint("vrele: negative ref count", vp);
#endif
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
VI_UNLOCK(vp);
Lots of improvements, including restructring the caching and management of vnodes and objects. There are some metadata performance improvements that come along with this. There are also a few prototypes added when the need is noticed. Changes include: 1) Cleaning up vref, vget. 2) Removal of the object cache. 3) Nuke vnode_pager_uncache and friends, because they aren't needed anymore. 4) Correct some missing LK_RETRY's in vn_lock. 5) Correct the page range in the code for msync. Be gentle, and please give me feedback asap.
1997-12-29 00:25:11 +00:00
panic("vrele: negative ref cnt");
Even though this looks like it, this is not a complex code change. The interface into the "VMIO" system has changed to be more consistant and robust. Essentially, it is now no longer necessary to call vn_open to get merged VM/Buffer cache operation, and exceptional conditions such as merged operation of VBLK devices is simpler and more correct. This code corrects a potentially large set of problems including the problems with ktrace output and loaded systems, file create/deletes, etc. Most of the changes to NFS are cosmetic and name changes, eliminating a layer of subroutine calls. The direct calls to vput/vrele have been re-instituted for better cross platform compatibility. Reviewed by: davidg
1996-08-21 21:56:23 +00:00
}
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
CTR2(KTR_VFS, "%s: return vnode %p to the freelist", __func__, vp);
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
/*
* We want to hold the vnode until the inactive finishes to
* prevent vgone() races. We drop the use count here and the
* hold count below when we're done.
*/
v_decr_useonly(vp);
- Slightly rearrange vrele() to move the common case in one indentation level. Sponsored by: Isilon Systems, Inc.
2005-03-14 07:16:55 +00:00
/*
* We must call VOP_INACTIVE with the node locked. Mark
* as VI_DOINGINACT to avoid recursion.
*/
Use vn_start_secondary_write() and vn_finished_secondary_write() as a replacement for vn_write_suspend_wait() to better account for secondary write processing. Close race where secondary writes could be started after ffs_sync() returned but before the file system was marked as suspended. Detect if secondary writes or softdep processing occurred during vnode sync loop in ffs_sync() and retry the loop if needed.
2006-03-08 23:43:39 +00:00
vp->v_iflag |= VI_OWEINACT;
vn_lock() is currently only used with the 'curthread' passed as argument. Remove this argument and pass curthread directly to underlying VOP_LOCK1() VFS method. This modify makes the code cleaner and in particular remove an annoying dependence helping next lockmgr() cleanup. KPI results, obviously, changed. Manpage and FreeBSD_version will be updated through further commits. As a side note, would be valuable to say that next commits will address a similar cleanup about VFS methods, in particular vop_lock1 and vop_unlock. Tested by: Diego Sardina <siarodx at gmail dot com>, Andrea Di Pasquale <whyx dot it at gmail dot com>
2008-01-10 01:10:58 +00:00
if (vn_lock(vp, LK_EXCLUSIVE | LK_INTERLOCK) == 0) {
- Slightly rearrange vrele() to move the common case in one indentation level. Sponsored by: Isilon Systems, Inc.
2005-03-14 07:16:55 +00:00
VI_LOCK(vp);
Use vn_start_secondary_write() and vn_finished_secondary_write() as a replacement for vn_write_suspend_wait() to better account for secondary write processing. Close race where secondary writes could be started after ffs_sync() returned but before the file system was marked as suspended. Detect if secondary writes or softdep processing occurred during vnode sync loop in ffs_sync() and retry the loop if needed.
2006-03-08 23:43:39 +00:00
if (vp->v_usecount > 0)
vp->v_iflag &= ~VI_OWEINACT;
if (vp->v_iflag & VI_OWEINACT)
vinactive(vp, td);
VOP_LOCK1() (and so VOP_LOCK()) and VOP_UNLOCK() are only used in conjuction with 'thread' argument passing which is always curthread. Remove the unuseful extra-argument and pass explicitly curthread to lower layer functions, when necessary. KPI results broken by this change, which should affect several ports, so version bumping and manpage update will be further committed. Tested by: kris, pho, Diego Sardina <siarodx at gmail dot com>
2008-01-13 14:44:15 +00:00
VOP_UNLOCK(vp, 0);
Use vn_start_secondary_write() and vn_finished_secondary_write() as a replacement for vn_write_suspend_wait() to better account for secondary write processing. Close race where secondary writes could be started after ffs_sync() returned but before the file system was marked as suspended. Detect if secondary writes or softdep processing occurred during vnode sync loop in ffs_sync() and retry the loop if needed.
2006-03-08 23:43:39 +00:00
} else {
- Slightly rearrange vrele() to move the common case in one indentation level. Sponsored by: Isilon Systems, Inc.
2005-03-14 07:16:55 +00:00
VI_LOCK(vp);
Use vn_start_secondary_write() and vn_finished_secondary_write() as a replacement for vn_write_suspend_wait() to better account for secondary write processing. Close race where secondary writes could be started after ffs_sync() returned but before the file system was marked as suspended. Detect if secondary writes or softdep processing occurred during vnode sync loop in ffs_sync() and retry the loop if needed.
2006-03-08 23:43:39 +00:00
if (vp->v_usecount > 0)
vp->v_iflag &= ~VI_OWEINACT;
}
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
vdropl(vp);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
}
Gratuitous whitespace cleanup.
2002-06-06 15:46:38 +00:00
/*
Style fixes based on comments by bde
2000-10-05 18:22:46 +00:00
* Release an already locked vnode. This give the same effects as
* unlock+vrele(), but takes less time and avoids releasing and
Universally adopt most conventional spelling of acquire.
2007-05-27 20:50:23 +00:00
* re-aquiring the lock (as vrele() acquires the lock internally.)
Style fixes: * Add lots of comments * Convert a couple of assertions to KASSERT() * Minimal whitespace & misapplied {} fixes * Convert #if 0 to #if COMPILING_LINT for code we presently do not support, but want to keep available. Reviewed by: adrian, markm
2000-09-22 12:22:36 +00:00
*/
Lots of improvements, including restructring the caching and management of vnodes and objects. There are some metadata performance improvements that come along with this. There are also a few prototypes added when the need is noticed. Changes include: 1) Cleaning up vref, vget. 2) Removal of the object cache. 3) Nuke vnode_pager_uncache and friends, because they aren't needed anymore. 4) Correct some missing LK_RETRY's in vn_lock. 5) Correct the page range in the code for msync. Be gentle, and please give me feedback asap.
1997-12-29 00:25:11 +00:00
void
Convert remaining functions in vfs_subr.c from K&R prototypes to ANSI C prototypes, as the majority of new functions added have been in this style. Changing prototype style now results in gcc noticing that the implementation of vn_pollrecord() has a 'short' argument instead of 'int' as prototyped in vnode.h, so correct that definition. In practice this didn't matter as only poll flags in the lower 16 bits are used. MFC after: 1 week
2006-01-21 19:42:10 +00:00
vput(struct vnode *vp)
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
{
KSE Milestone 2 Note ALL MODULES MUST BE RECOMPILED make the kernel aware that there are smaller units of scheduling than the process. (but only allow one thread per process at this time). This is functionally equivalent to teh previousl -current except that there is a thread associated with each process. Sorry john! (your next MFC will be a doosie!) Reviewed by: peter@freebsd.org, dillon@freebsd.org X-MFC after: ha ha ha ha
2001-09-12 08:38:13 +00:00
struct thread *td = curthread; /* XXX */
- Dont clear OWEINACT in vbusy(), we still owe an inactive call if someone vhold()s us. - Avoid an extra mutex acquire and release in the common case of vgonel() by checking for OWEINACT at the start of the function. - Fix the case where we set OWEINACT in vput(). LK_EXCLUPGRADE drops our shared lock if it fails. Sponsored by: Isilon Systems, Inc.
2005-03-29 10:02:48 +00:00
int error;
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
Split DIAGNOSTIC -> DIAGNOSTIC, INVARIANTS, and INVARIANT_SUPPORT as discussed on -hackers. Introduce 'KASSERT(assertion, ("panic message", args))' for simple check + panic. Reviewed by: msmith
1999-01-08 17:31:30 +00:00
KASSERT(vp != NULL, ("vput: null vp"));
- Move code that should probably be an assert above the main body of vrele so that we can decrease the indentation of the real work and make things slightly more clear. Sponsored by: Isilon Systems, Inc.
2005-03-28 11:18:47 +00:00
ASSERT_VOP_LOCKED(vp, "vput");
- Solve a race where we could lose a call to VOP_INACTIVE. If vget() waiting on a lock held the last usecount ref on a vnode and the lock failed we would not call INACTIVE. Solve this by only holding a holdcnt to prevent the vnode from disappearing while we wait on vn_lock. Other callers may now VOP_INACTIVE while we are waiting on the lock, however this race is acceptable, while losing INACTIVE is not. Discussed with: kan, pjd Tested by: kkenn Sponsored by: Isilon Systems, Inc. MFC After: 1 week
2006-02-01 00:30:05 +00:00
VFS_ASSERT_GIANT(vp->v_mount);
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
CTR2(KTR_VFS, "%s: vp %p", __func__, vp);
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
VI_LOCK(vp);
In vrele() and vput(), avoid triggering the confusing "missed vn_close" KASSERT when vp->v_usecount is zero or negative. In this case, the "v*: negative ref cnt" panic that follows is much more appropriate. Reviewed by: mckusick
2001-05-11 20:42:41 +00:00
/* Skip this v_writecount check if we're going to panic below. */
Convert KASSERTS to VNASSERTS
2005-02-17 10:28:58 +00:00
VNASSERT(vp->v_writecount < vp->v_usecount || vp->v_usecount < 1, vp,
In vrele() and vput(), avoid triggering the confusing "missed vn_close" KASSERT when vp->v_usecount is zero or negative. In this case, the "v*: negative ref cnt" panic that follows is much more appropriate. Reviewed by: mckusick
2001-05-11 20:42:41 +00:00
("vput: missed vn_close"));
- Dont clear OWEINACT in vbusy(), we still owe an inactive call if someone vhold()s us. - Avoid an extra mutex acquire and release in the common case of vgonel() by checking for OWEINACT at the start of the function. - Fix the case where we set OWEINACT in vput(). LK_EXCLUPGRADE drops our shared lock if it fails. Sponsored by: Isilon Systems, Inc.
2005-03-29 10:02:48 +00:00
error = 0;
Move KASSERTs which checks value of v_usecount after vnode locking, so it will not produce wrong alarms.
2000-10-02 09:57:06 +00:00
Add a new vnode flag VI_DOINGINACT to indicate that a VOP_INACTIVE call is in progress on the vnode. When vput() or vrele() sees a 1->0 reference count transition, it now return without any further action if this flag is set. This flag is necessary to avoid recursion into VOP_INACTIVE if the filesystem inactive routine causes the reference count to increase and then drop back to zero. It is also used to guarantee that an unlocked vnode will not be recycled while blocked in VOP_INACTIVE(). There are at least two cases where the recursion can occur: one is that the softupdates code called by ufs_inactive() via ffs_truncate() can call vput() on the vnode. This has been reported by many people as "lockmgr: draining against myself" panics. The other case is that nfs_inactive() can call vget() and then vrele() on the vnode to clean up a sillyrename file. Reviewed by: mckusick (an older version of the patch)
2002-12-29 18:30:49 +00:00
if (vp->v_usecount > 1 || ((vp->v_iflag & VI_DOINGINACT) &&
vp->v_usecount == 1)) {
VOP_LOCK1() (and so VOP_LOCK()) and VOP_UNLOCK() are only used in conjuction with 'thread' argument passing which is always curthread. Remove the unuseful extra-argument and pass explicitly curthread to lower layer functions, when necessary. KPI results broken by this change, which should affect several ports, so version bumping and manpage update will be further committed. Tested by: kris, pho, Diego Sardina <siarodx at gmail dot com>
2008-01-13 14:44:15 +00:00
VOP_UNLOCK(vp, 0);
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
v_decr_usecount(vp);
Make our v_usecount vnode reference count work identically to the original BSD code. The association between the vnode and the vm_object no longer includes reference counts. The major difference is that vm_object's are no longer freed gratuitiously from the vnode, and so once an object is created for the vnode, it will last as long as the vnode does. When a vnode object reference count is incremented, then the underlying vnode reference count is incremented also. The two "objects" are now more intimately related, and so the interactions are now much less complex. When vnodes are now normally placed onto the free queue with an object still attached. The rundown of the object happens at vnode rundown time, and happens with exactly the same filesystem semantics of the original VFS code. There is absolutely no need for vnode_pager_uncache and other travesties like that anymore. A side-effect of these changes is that SMP locking should be much simpler, the I/O copyin/copyout optimizations work, NFS should be more ponderable, and further work on layered filesystems should be less frustrating, because of the totally coherent management of the vnode objects and vnodes. Please be careful with your system while running this code, but I would greatly appreciate feedback as soon a reasonably possible.
1998-01-06 05:26:17 +00:00
return;
}
Even though this looks like it, this is not a complex code change. The interface into the "VMIO" system has changed to be more consistant and robust. Essentially, it is now no longer necessary to call vn_open to get merged VM/Buffer cache operation, and exceptional conditions such as merged operation of VBLK devices is simpler and more correct. This code corrects a potentially large set of problems including the problems with ktrace output and loaded systems, file create/deletes, etc. Most of the changes to NFS are cosmetic and name changes, eliminating a layer of subroutine calls. The direct calls to vput/vrele have been re-instituted for better cross platform compatibility. Reviewed by: davidg
1996-08-21 21:56:23 +00:00
- Move code that should probably be an assert above the main body of vrele so that we can decrease the indentation of the real work and make things slightly more clear. Sponsored by: Isilon Systems, Inc.
2005-03-28 11:18:47 +00:00
if (vp->v_usecount != 1) {
Lots of improvements, including restructring the caching and management of vnodes and objects. There are some metadata performance improvements that come along with this. There are also a few prototypes added when the need is noticed. Changes include: 1) Cleaning up vref, vget. 2) Removal of the object cache. 3) Nuke vnode_pager_uncache and friends, because they aren't needed anymore. 4) Correct some missing LK_RETRY's in vn_lock. 5) Correct the page range in the code for msync. Be gentle, and please give me feedback asap.
1997-12-29 00:25:11 +00:00
#ifdef DIAGNOSTIC
vprint("vput: negative ref count", vp);
#endif
panic("vput: negative ref cnt");
}
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
CTR2(KTR_VFS, "%s: return to freelist the vnode %p", __func__, vp);
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
/*
* We want to hold the vnode until the inactive finishes to
* prevent vgone() races. We drop the use count here and the
* hold count below when we're done.
*/
v_decr_useonly(vp);
- Dont clear OWEINACT in vbusy(), we still owe an inactive call if someone vhold()s us. - Avoid an extra mutex acquire and release in the common case of vgonel() by checking for OWEINACT at the start of the function. - Fix the case where we set OWEINACT in vput(). LK_EXCLUPGRADE drops our shared lock if it fails. Sponsored by: Isilon Systems, Inc.
2005-03-29 10:02:48 +00:00
vp->v_iflag |= VI_OWEINACT;
Axe the 'thread' argument from VOP_ISLOCKED() and lockstatus() as it is always curthread. As KPI gets broken by this patch, manpages and __FreeBSD_version will be updated by further commits. Tested by: Andrea Barberio <insomniac at slackware dot it>
2008-02-25 18:45:57 +00:00
if (VOP_ISLOCKED(vp) != LK_EXCLUSIVE) {
VOP_LOCK1() (and so VOP_LOCK()) and VOP_UNLOCK() are only used in conjuction with 'thread' argument passing which is always curthread. Remove the unuseful extra-argument and pass explicitly curthread to lower layer functions, when necessary. KPI results broken by this change, which should affect several ports, so version bumping and manpage update will be further committed. Tested by: kris, pho, Diego Sardina <siarodx at gmail dot com>
2008-01-13 14:44:15 +00:00
error = VOP_LOCK(vp, LK_UPGRADE|LK_INTERLOCK|LK_NOWAIT);
- Dont clear OWEINACT in vbusy(), we still owe an inactive call if someone vhold()s us. - Avoid an extra mutex acquire and release in the common case of vgonel() by checking for OWEINACT at the start of the function. - Fix the case where we set OWEINACT in vput(). LK_EXCLUPGRADE drops our shared lock if it fails. Sponsored by: Isilon Systems, Inc.
2005-03-29 10:02:48 +00:00
VI_LOCK(vp);
Use vn_start_secondary_write() and vn_finished_secondary_write() as a replacement for vn_write_suspend_wait() to better account for secondary write processing. Close race where secondary writes could be started after ffs_sync() returned but before the file system was marked as suspended. Detect if secondary writes or softdep processing occurred during vnode sync loop in ffs_sync() and retry the loop if needed.
2006-03-08 23:43:39 +00:00
if (error) {
if (vp->v_usecount > 0)
vp->v_iflag &= ~VI_OWEINACT;
- Change the VOP_LOCK UPGRADE in vput() to do a LK_NOWAIT to avoid a potential lock order reversal. Also, don't unlock the vnode if this fails, lockmgr has already unlocked it for us. - Restructure vget() now that vn_lock() does all of VI_DOOMED checking for us and also handles the case where there is no real lock type. - If VI_OWEINACT is set, we need to upgrade the lock request to EXCLUSIVE so that we can call inactive. It's not legal to vget a vnode that hasn't had INACTIVE called yet. Sponsored by: Isilon Systems, Inc.
2005-04-11 09:28:32 +00:00
goto done;
Use vn_start_secondary_write() and vn_finished_secondary_write() as a replacement for vn_write_suspend_wait() to better account for secondary write processing. Close race where secondary writes could be started after ffs_sync() returned but before the file system was marked as suspended. Detect if secondary writes or softdep processing occurred during vnode sync loop in ffs_sync() and retry the loop if needed.
2006-03-08 23:43:39 +00:00
}
- Dont clear OWEINACT in vbusy(), we still owe an inactive call if someone vhold()s us. - Avoid an extra mutex acquire and release in the common case of vgonel() by checking for OWEINACT at the start of the function. - Fix the case where we set OWEINACT in vput(). LK_EXCLUPGRADE drops our shared lock if it fails. Sponsored by: Isilon Systems, Inc.
2005-03-29 10:02:48 +00:00
}
Use vn_start_secondary_write() and vn_finished_secondary_write() as a replacement for vn_write_suspend_wait() to better account for secondary write processing. Close race where secondary writes could be started after ffs_sync() returned but before the file system was marked as suspended. Detect if secondary writes or softdep processing occurred during vnode sync loop in ffs_sync() and retry the loop if needed.
2006-03-08 23:43:39 +00:00
if (vp->v_usecount > 0)
vp->v_iflag &= ~VI_OWEINACT;
- Change the VOP_LOCK UPGRADE in vput() to do a LK_NOWAIT to avoid a potential lock order reversal. Also, don't unlock the vnode if this fails, lockmgr has already unlocked it for us. - Restructure vget() now that vn_lock() does all of VI_DOOMED checking for us and also handles the case where there is no real lock type. - If VI_OWEINACT is set, we need to upgrade the lock request to EXCLUSIVE so that we can call inactive. It's not legal to vget a vnode that hasn't had INACTIVE called yet. Sponsored by: Isilon Systems, Inc.
2005-04-11 09:28:32 +00:00
if (vp->v_iflag & VI_OWEINACT)
- Move code that should probably be an assert above the main body of vrele so that we can decrease the indentation of the real work and make things slightly more clear. Sponsored by: Isilon Systems, Inc.
2005-03-28 11:18:47 +00:00
vinactive(vp, td);
VOP_LOCK1() (and so VOP_LOCK()) and VOP_UNLOCK() are only used in conjuction with 'thread' argument passing which is always curthread. Remove the unuseful extra-argument and pass explicitly curthread to lower layer functions, when necessary. KPI results broken by this change, which should affect several ports, so version bumping and manpage update will be further committed. Tested by: kris, pho, Diego Sardina <siarodx at gmail dot com>
2008-01-13 14:44:15 +00:00
VOP_UNLOCK(vp, 0);
- Change the VOP_LOCK UPGRADE in vput() to do a LK_NOWAIT to avoid a potential lock order reversal. Also, don't unlock the vnode if this fails, lockmgr has already unlocked it for us. - Restructure vget() now that vn_lock() does all of VI_DOOMED checking for us and also handles the case where there is no real lock type. - If VI_OWEINACT is set, we need to upgrade the lock request to EXCLUSIVE so that we can call inactive. It's not legal to vget a vnode that hasn't had INACTIVE called yet. Sponsored by: Isilon Systems, Inc.
2005-04-11 09:28:32 +00:00
done:
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
vdropl(vp);
Fix the simple_lock problem with the physical I/O buffer code, and also fix the missing simple_unlock in vrele, and improve vrele/vput by merging them into one routine. BDE pointed these problems out.
1997-02-27 02:57:03 +00:00
}
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
/*
Change the 0xdeadb hack to a flag called VDOOMED. Introduce VFREE which indicates that vnode is on freelist. Rename vholdrele() to vdrop(). Create vfree() and vbusy() to add/delete vnode from freelist. Add vfree()/vbusy() to keep (v_holdcnt != 0 || v_usecount != 0) vnodes off the freelist. Generalize vhold()/v_holdcnt to mean "do not recycle". Fix reassignbuf()s lack of use of vhold(). Use vhold() instead of checking v_cache_src list. Remove vtouch(), the vnodes are always vget'ed soon enough after for it to have any measuable effect. Add sysctl debug.freevnodes to keep track of things. Move cache_purge() up in getnewvnodes to avoid race. Decrement v_usecount after VOP_INACTIVE(), put a vhold() on it during VOP_INACTIVE() Unmacroize vhold()/vdrop() Print out VDOOMED and VFREE flags (XXX: should use %b) Reviewed by: dyson
1997-08-31 07:32:39 +00:00
* Somebody doesn't want the vnode recycled.
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
*/
The big 4.4BSD Lite to FreeBSD 2.0.0 (Development) patch. Reviewed by: Rodney W. Grimes Submitted by: John Dyson and David Greenman
1994-05-25 09:21:21 +00:00
void
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
vhold(struct vnode *vp)
{
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
VI_LOCK(vp);
vholdl(vp);
VI_UNLOCK(vp);
}
- Expose vholdl() so it may be used outside of vfs_subr.c
2005-03-15 13:43:10 +00:00
void
Add b_bufobj to struct buf which eventually will eliminate the need for b_vp. Initialize b_bufobj for all buffers. Make incore() and gbincore() take a bufobj instead of a vnode. Make inmem() local to vfs_bio.c Change a lot of VI_[UN]LOCK(bp->b_vp) to BO_[UN]LOCK(bp->b_bufobj) also VI_MTX() to BO_MTX(), Make buf_vlist_add() take a bufobj instead of a vnode. Eliminate other uses of bp->b_vp where bp->b_bufobj will do. Various minor polishing: remove "register", turn panic into KASSERT, use new function declarations, TAILQ_FOREACH_SAFE() etc.
2004-10-22 08:47:20 +00:00
vholdl(struct vnode *vp)
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
{
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
CTR2(KTR_VFS, "%s: vp %p", __func__, vp);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
vp->v_holdcnt++;
Change the 0xdeadb hack to a flag called VDOOMED. Introduce VFREE which indicates that vnode is on freelist. Rename vholdrele() to vdrop(). Create vfree() and vbusy() to add/delete vnode from freelist. Add vfree()/vbusy() to keep (v_holdcnt != 0 || v_usecount != 0) vnodes off the freelist. Generalize vhold()/v_holdcnt to mean "do not recycle". Fix reassignbuf()s lack of use of vhold(). Use vhold() instead of checking v_cache_src list. Remove vtouch(), the vnodes are always vget'ed soon enough after for it to have any measuable effect. Add sysctl debug.freevnodes to keep track of things. Move cache_purge() up in getnewvnodes to avoid race. Decrement v_usecount after VOP_INACTIVE(), put a vhold() on it during VOP_INACTIVE() Unmacroize vhold()/vdrop() Print out VDOOMED and VFREE flags (XXX: should use %b) Reviewed by: dyson
1997-08-31 07:32:39 +00:00
if (VSHOULDBUSY(vp))
vbusy(vp);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
}
/*
Style fixes based on comments by bde
2000-10-05 18:22:46 +00:00
* Note that there is one less who cares about this vnode. vdrop() is the
* opposite of vhold().
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
*/
The big 4.4BSD Lite to FreeBSD 2.0.0 (Development) patch. Reviewed by: Rodney W. Grimes Submitted by: John Dyson and David Greenman
1994-05-25 09:21:21 +00:00
void
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
vdrop(struct vnode *vp)
{
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
VI_LOCK(vp);
vdropl(vp);
}
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
/*
* Drop the hold count of the vnode. If this is the last reference to
* the vnode we will free it if it has been vgone'd otherwise it is
* placed on the free list.
*/
Make vdropl() public; zfs needs it. There is also plenty of existing file system code (mostly *_reclaim()) which look like this: VOP_LOCK(vp); /* examine vp */ VOP_UNLOCK(vp); vdrop(vp); This can now be rewritten to: VOP_LOCK(vp); /* examine vp */ vdropl(vp); /* will unlock vp */ MFC after: 1 week
2007-03-31 23:57:17 +00:00
void
Make various vnode related functions static
2005-02-10 12:28:58 +00:00
vdropl(struct vnode *vp)
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
{
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
Now that the vdropl() function is public, assert that the vnode interlock is held.
2007-04-01 10:45:32 +00:00
ASSERT_VI_LOCKED(vp, "vdropl");
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
CTR2(KTR_VFS, "%s: vp %p", __func__, vp);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
if (vp->v_holdcnt <= 0)
- Increment the holdcnt once for each usecount reference. This allows us to use only the holdcnt to determine whether a vnode may be recycled, simplifying the V* macros as well as vtryrecycle(), etc. Sponsored by: Isilon Systems, Inc.
2005-03-14 09:25:19 +00:00
panic("vdrop: holdcnt %d", vp->v_holdcnt);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
vp->v_holdcnt--;
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
if (vp->v_holdcnt == 0) {
if (vp->v_iflag & VI_DOOMED) {
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
CTR2(KTR_VFS, "%s: destroying the vnode %p", __func__,
vp);
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
vdestroy(vp);
return;
} else
vfree(vp);
}
VI_UNLOCK(vp);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
}
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
/*
* Call VOP_INACTIVE on the vnode and manage the DOINGINACT and OWEINACT
* flags. DOINGINACT prevents us from recursing in calls to vinactive.
* OWEINACT tracks whether a vnode missed a call to inactive due to a
* failed lock upgrade.
*/
- Remove vx_lock, vx_unlock, vx_wait, etc. - Add a vn_start_write/vn_finished_write around vlrureclaim so we don't do writing ops without suspending. This could suspend the vlruproc which should not be a problem under normal circumstances. - Manually implement VMIGHTFREE in vlrureclaim as this was the only instance where it was used. - Acquire a lock before calling vgone() as it now requires it. - Move the acquisition of the vnode interlock from vtryrecycle() to getnewvnode() so that if it fails we don't drop and reacquire the vnode_free_list_mtx. - Check for a usecount or holdcount at the end of vtryrecycle() in case someone grabbed a ref while we were recycling. Abort the recycle, and on the final ref drop this vnode will be placed on the head of the free list. - Move the redundant VOP_INACTIVE protection code into the local vinactive() routine to avoid code bloat. - Keep the vnode lock held across calls to vgone() in several places. - vgonel() no longer uses XLOCK, instead callers must hold an exclusive vnode lock. The VI_DOOMED flag is set to allow other threads to detect a vnode which is no longer valid. This flag is set until the last reference is gone, and there are no chances for a new ref. vgonel() holds this lock across the entire function, which greatly simplifies logic. _ Only vfree() in one place in vgone() not three. - Adjust vget() to check the VI_DOOMED flag prior to waiting on the lock in the LK_NOWAIT case. In other cases, check after we have slept and acquired an exlusive lock. This will simulate the old vx_wait() behavior. Sponsored by: Isilon Systems, Inc.
2005-03-13 11:54:28 +00:00
static void
vinactive(struct vnode *vp, struct thread *td)
{
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
Assert for exclusive vnode lock in vinactive(), vrecycle() and vgonel() functions. Reviewed by: kib
2008-07-27 11:48:15 +00:00
ASSERT_VOP_ELOCKED(vp, "vinactive");
- Remove vx_lock, vx_unlock, vx_wait, etc. - Add a vn_start_write/vn_finished_write around vlrureclaim so we don't do writing ops without suspending. This could suspend the vlruproc which should not be a problem under normal circumstances. - Manually implement VMIGHTFREE in vlrureclaim as this was the only instance where it was used. - Acquire a lock before calling vgone() as it now requires it. - Move the acquisition of the vnode interlock from vtryrecycle() to getnewvnode() so that if it fails we don't drop and reacquire the vnode_free_list_mtx. - Check for a usecount or holdcount at the end of vtryrecycle() in case someone grabbed a ref while we were recycling. Abort the recycle, and on the final ref drop this vnode will be placed on the head of the free list. - Move the redundant VOP_INACTIVE protection code into the local vinactive() routine to avoid code bloat. - Keep the vnode lock held across calls to vgone() in several places. - vgonel() no longer uses XLOCK, instead callers must hold an exclusive vnode lock. The VI_DOOMED flag is set to allow other threads to detect a vnode which is no longer valid. This flag is set until the last reference is gone, and there are no chances for a new ref. vgonel() holds this lock across the entire function, which greatly simplifies logic. _ Only vfree() in one place in vgone() not three. - Adjust vget() to check the VI_DOOMED flag prior to waiting on the lock in the LK_NOWAIT case. In other cases, check after we have slept and acquired an exlusive lock. This will simulate the old vx_wait() behavior. Sponsored by: Isilon Systems, Inc.
2005-03-13 11:54:28 +00:00
ASSERT_VI_LOCKED(vp, "vinactive");
VNASSERT((vp->v_iflag & VI_DOINGINACT) == 0, vp,
("vinactive: recursed on VI_DOINGINACT"));
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
CTR2(KTR_VFS, "%s: vp %p", __func__, vp);
- Remove vx_lock, vx_unlock, vx_wait, etc. - Add a vn_start_write/vn_finished_write around vlrureclaim so we don't do writing ops without suspending. This could suspend the vlruproc which should not be a problem under normal circumstances. - Manually implement VMIGHTFREE in vlrureclaim as this was the only instance where it was used. - Acquire a lock before calling vgone() as it now requires it. - Move the acquisition of the vnode interlock from vtryrecycle() to getnewvnode() so that if it fails we don't drop and reacquire the vnode_free_list_mtx. - Check for a usecount or holdcount at the end of vtryrecycle() in case someone grabbed a ref while we were recycling. Abort the recycle, and on the final ref drop this vnode will be placed on the head of the free list. - Move the redundant VOP_INACTIVE protection code into the local vinactive() routine to avoid code bloat. - Keep the vnode lock held across calls to vgone() in several places. - vgonel() no longer uses XLOCK, instead callers must hold an exclusive vnode lock. The VI_DOOMED flag is set to allow other threads to detect a vnode which is no longer valid. This flag is set until the last reference is gone, and there are no chances for a new ref. vgonel() holds this lock across the entire function, which greatly simplifies logic. _ Only vfree() in one place in vgone() not three. - Adjust vget() to check the VI_DOOMED flag prior to waiting on the lock in the LK_NOWAIT case. In other cases, check after we have slept and acquired an exlusive lock. This will simulate the old vx_wait() behavior. Sponsored by: Isilon Systems, Inc.
2005-03-13 11:54:28 +00:00
vp->v_iflag |= VI_DOINGINACT;
- Clear OWEINACT prior to calling VOP_INACTIVE to remove the possibility of a vget causing another call to INACTIVE before we're finished.
2005-06-07 22:05:32 +00:00
vp->v_iflag &= ~VI_OWEINACT;
- Remove vx_lock, vx_unlock, vx_wait, etc. - Add a vn_start_write/vn_finished_write around vlrureclaim so we don't do writing ops without suspending. This could suspend the vlruproc which should not be a problem under normal circumstances. - Manually implement VMIGHTFREE in vlrureclaim as this was the only instance where it was used. - Acquire a lock before calling vgone() as it now requires it. - Move the acquisition of the vnode interlock from vtryrecycle() to getnewvnode() so that if it fails we don't drop and reacquire the vnode_free_list_mtx. - Check for a usecount or holdcount at the end of vtryrecycle() in case someone grabbed a ref while we were recycling. Abort the recycle, and on the final ref drop this vnode will be placed on the head of the free list. - Move the redundant VOP_INACTIVE protection code into the local vinactive() routine to avoid code bloat. - Keep the vnode lock held across calls to vgone() in several places. - vgonel() no longer uses XLOCK, instead callers must hold an exclusive vnode lock. The VI_DOOMED flag is set to allow other threads to detect a vnode which is no longer valid. This flag is set until the last reference is gone, and there are no chances for a new ref. vgonel() holds this lock across the entire function, which greatly simplifies logic. _ Only vfree() in one place in vgone() not three. - Adjust vget() to check the VI_DOOMED flag prior to waiting on the lock in the LK_NOWAIT case. In other cases, check after we have slept and acquired an exlusive lock. This will simulate the old vx_wait() behavior. Sponsored by: Isilon Systems, Inc.
2005-03-13 11:54:28 +00:00
VI_UNLOCK(vp);
VOP_INACTIVE(vp, td);
VI_LOCK(vp);
VNASSERT(vp->v_iflag & VI_DOINGINACT, vp,
("vinactive: lost VI_DOINGINACT"));
- Clear OWEINACT prior to calling VOP_INACTIVE to remove the possibility of a vget causing another call to INACTIVE before we're finished.
2005-06-07 22:05:32 +00:00
vp->v_iflag &= ~VI_DOINGINACT;
- Remove vx_lock, vx_unlock, vx_wait, etc. - Add a vn_start_write/vn_finished_write around vlrureclaim so we don't do writing ops without suspending. This could suspend the vlruproc which should not be a problem under normal circumstances. - Manually implement VMIGHTFREE in vlrureclaim as this was the only instance where it was used. - Acquire a lock before calling vgone() as it now requires it. - Move the acquisition of the vnode interlock from vtryrecycle() to getnewvnode() so that if it fails we don't drop and reacquire the vnode_free_list_mtx. - Check for a usecount or holdcount at the end of vtryrecycle() in case someone grabbed a ref while we were recycling. Abort the recycle, and on the final ref drop this vnode will be placed on the head of the free list. - Move the redundant VOP_INACTIVE protection code into the local vinactive() routine to avoid code bloat. - Keep the vnode lock held across calls to vgone() in several places. - vgonel() no longer uses XLOCK, instead callers must hold an exclusive vnode lock. The VI_DOOMED flag is set to allow other threads to detect a vnode which is no longer valid. This flag is set until the last reference is gone, and there are no chances for a new ref. vgonel() holds this lock across the entire function, which greatly simplifies logic. _ Only vfree() in one place in vgone() not three. - Adjust vget() to check the VI_DOOMED flag prior to waiting on the lock in the LK_NOWAIT case. In other cases, check after we have slept and acquired an exlusive lock. This will simulate the old vx_wait() behavior. Sponsored by: Isilon Systems, Inc.
2005-03-13 11:54:28 +00:00
}
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
/*
* Remove any vnodes in the vnode table belonging to mount point mp.
*
Change the second argument of vflush() to an integer that specifies the number of references on the filesystem root vnode to be both expected and released. Many filesystems hold an extra reference on the filesystem root vnode, which must be accounted for when determining if the filesystem is busy and then released if it isn't busy. The old `skipvp' approach required individual filesystem xxx_unmount functions to re-implement much of vflush()'s logic to deal with the root vnode. All 9 filesystems that hold an extra reference on the root vnode got the logic wrong in the case of forced unmounts, so `umount -f' would always fail if there were any extra root vnode references. Fix this issue centrally in vflush(), now that we can. This commit also fixes a vnode reference leak in devfs, which could result in idle devfs filesystems that refuse to unmount. Reviewed by: phk, bp
2001-05-16 18:04:37 +00:00
* If FORCECLOSE is not specified, there should not be any active ones,
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
* return error if any are found (nb: this is a user error, not a
Change the second argument of vflush() to an integer that specifies the number of references on the filesystem root vnode to be both expected and released. Many filesystems hold an extra reference on the filesystem root vnode, which must be accounted for when determining if the filesystem is busy and then released if it isn't busy. The old `skipvp' approach required individual filesystem xxx_unmount functions to re-implement much of vflush()'s logic to deal with the root vnode. All 9 filesystems that hold an extra reference on the root vnode got the logic wrong in the case of forced unmounts, so `umount -f' would always fail if there were any extra root vnode references. Fix this issue centrally in vflush(), now that we can. This commit also fixes a vnode reference leak in devfs, which could result in idle devfs filesystems that refuse to unmount. Reviewed by: phk, bp
2001-05-16 18:04:37 +00:00
* system error). If FORCECLOSE is specified, detach any active vnodes
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
* that are found.
Change the second argument of vflush() to an integer that specifies the number of references on the filesystem root vnode to be both expected and released. Many filesystems hold an extra reference on the filesystem root vnode, which must be accounted for when determining if the filesystem is busy and then released if it isn't busy. The old `skipvp' approach required individual filesystem xxx_unmount functions to re-implement much of vflush()'s logic to deal with the root vnode. All 9 filesystems that hold an extra reference on the root vnode got the logic wrong in the case of forced unmounts, so `umount -f' would always fail if there were any extra root vnode references. Fix this issue centrally in vflush(), now that we can. This commit also fixes a vnode reference leak in devfs, which could result in idle devfs filesystems that refuse to unmount. Reviewed by: phk, bp
2001-05-16 18:04:37 +00:00
*
* If WRITECLOSE is set, only flush out regular file vnodes open for
* writing.
*
- Replace v_flag with v_iflag and v_vflag - v_vflag is protected by the vnode lock and is used when synchronization with VOP calls is needed. - v_iflag is protected by interlock and is used for dealing with vnode management issues. These flags include X/O LOCK, FREE, DOOMED, etc. - All accesses to v_iflag and v_vflag have either been locked or marked with mp_fixme's. - Many ASSERT_VOP_LOCKED calls have been added where the locking was not clear. - Many functions in vfs_subr.c were restructured to provide for stronger locking. Idea stolen from: BSD/OS
2002-08-04 10:29:36 +00:00
* SKIPSYSTEM causes any vnodes marked VV_SYSTEM to be skipped.
Change the second argument of vflush() to an integer that specifies the number of references on the filesystem root vnode to be both expected and released. Many filesystems hold an extra reference on the filesystem root vnode, which must be accounted for when determining if the filesystem is busy and then released if it isn't busy. The old `skipvp' approach required individual filesystem xxx_unmount functions to re-implement much of vflush()'s logic to deal with the root vnode. All 9 filesystems that hold an extra reference on the root vnode got the logic wrong in the case of forced unmounts, so `umount -f' would always fail if there were any extra root vnode references. Fix this issue centrally in vflush(), now that we can. This commit also fixes a vnode reference leak in devfs, which could result in idle devfs filesystems that refuse to unmount. Reviewed by: phk, bp
2001-05-16 18:04:37 +00:00
*
* `rootrefs' specifies the base reference count for the root vnode
* of this filesystem. The root vnode is considered busy if its
Make VFS_ROOT() and vflush() take a thread argument. This is to allow filesystems to decide based on the passed thread which vnode to return. Several filesystems used curthread, they now use the passed thread.
2004-07-12 08:14:09 +00:00
* v_usecount exceeds this value. On a successful return, vflush(, td)
Change the second argument of vflush() to an integer that specifies the number of references on the filesystem root vnode to be both expected and released. Many filesystems hold an extra reference on the filesystem root vnode, which must be accounted for when determining if the filesystem is busy and then released if it isn't busy. The old `skipvp' approach required individual filesystem xxx_unmount functions to re-implement much of vflush()'s logic to deal with the root vnode. All 9 filesystems that hold an extra reference on the root vnode got the logic wrong in the case of forced unmounts, so `umount -f' would always fail if there were any extra root vnode references. Fix this issue centrally in vflush(), now that we can. This commit also fixes a vnode reference leak in devfs, which could result in idle devfs filesystems that refuse to unmount. Reviewed by: phk, bp
2001-05-16 18:04:37 +00:00
* will call vrele() on the root vnode exactly rootrefs times.
* If the SKIPSYSTEM or WRITECLOSE flags are specified, rootrefs must
* be zero.
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
*/
#ifdef DIAGNOSTIC
Staticize. Unstaticize a function in scsi/scsi_base that was used, with an undocumented option. My last count on the LINT kernel shows: Total symbols: 3647 unref symbols: 463 undef symbols: 4 1 ref symbols: 1751 2 ref symbols: 485 Approaching the pain threshold now.
1995-12-17 21:23:44 +00:00
static int busyprt = 0; /* print out busy vnodes */
Use OID_AUTO instead of magic number for the Lite2 sysctl debug.busyprt. Removed declaration of vfs_unmountroot() again. Staticized vgonel().
1997-04-01 13:05:34 +00:00
SYSCTL_INT(_debug, OID_AUTO, busyprt, CTLFLAG_RW, &busyprt, 0, "");
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
#endif
The big 4.4BSD Lite to FreeBSD 2.0.0 (Development) patch. Reviewed by: Rodney W. Grimes Submitted by: John Dyson and David Greenman
1994-05-25 09:21:21 +00:00
int
Convert remaining functions in vfs_subr.c from K&R prototypes to ANSI C prototypes, as the majority of new functions added have been in this style. Changing prototype style now results in gcc noticing that the implementation of vn_pollrecord() has a 'short' argument instead of 'int' as prototyped in vnode.h, so correct that definition. In practice this didn't matter as only poll flags in the lower 16 bits are used. MFC after: 1 week
2006-01-21 19:42:10 +00:00
vflush( struct mount *mp, int rootrefs, int flags, struct thread *td)
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
{
Add marker vnodes to ensure that all vnodes associated with the mount point are iterated over when using MNT_VNODE_FOREACH. Reviewed by: truckman
2006-01-09 20:42:19 +00:00
struct vnode *vp, *mvp, *rootvp = NULL;
When downgrading a filesystem from read-write to read-only, operations involving file removal or file update were not always being fully committed to disk. The result was lost files or corrupted file data. This change ensures that the filesystem is properly synced to disk before the filesystem is down-graded. This delta also fixes a long standing bug in which a file open for reading has been unlinked. When the last open reference to the file is closed, the inode is reclaimed by the filesystem. Previously, if the filesystem had been down-graded to read-only, the inode could not be reclaimed, and thus was lost and had to be later recovered by fsck. With this change, such files are found at the time of the down-grade. Normally they will result in the filesystem down-grade failing with `device busy'. If a forcible down-grade is done, then the affected files will be revoked causing the inode to be released and the open file descriptors to begin failing on attempts to read. Submitted by: "Sam Leffler" <sam@errno.com>
2002-01-15 07:17:12 +00:00
struct vattr vattr;
Change the second argument of vflush() to an integer that specifies the number of references on the filesystem root vnode to be both expected and released. Many filesystems hold an extra reference on the filesystem root vnode, which must be accounted for when determining if the filesystem is busy and then released if it isn't busy. The old `skipvp' approach required individual filesystem xxx_unmount functions to re-implement much of vflush()'s logic to deal with the root vnode. All 9 filesystems that hold an extra reference on the root vnode got the logic wrong in the case of forced unmounts, so `umount -f' would always fail if there were any extra root vnode references. Fix this issue centrally in vflush(), now that we can. This commit also fixes a vnode reference leak in devfs, which could result in idle devfs filesystems that refuse to unmount. Reviewed by: phk, bp
2001-05-16 18:04:37 +00:00
int busy = 0, error;
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
CTR4(KTR_VFS, "%s: mp %p with rootrefs %d and flags %d", __func__, mp,
rootrefs, flags);
Change the second argument of vflush() to an integer that specifies the number of references on the filesystem root vnode to be both expected and released. Many filesystems hold an extra reference on the filesystem root vnode, which must be accounted for when determining if the filesystem is busy and then released if it isn't busy. The old `skipvp' approach required individual filesystem xxx_unmount functions to re-implement much of vflush()'s logic to deal with the root vnode. All 9 filesystems that hold an extra reference on the root vnode got the logic wrong in the case of forced unmounts, so `umount -f' would always fail if there were any extra root vnode references. Fix this issue centrally in vflush(), now that we can. This commit also fixes a vnode reference leak in devfs, which could result in idle devfs filesystems that refuse to unmount. Reviewed by: phk, bp
2001-05-16 18:04:37 +00:00
if (rootrefs > 0) {
KASSERT((flags & (SKIPSYSTEM | WRITECLOSE)) == 0,
("vflush: bad args"));
/*
* Get the filesystem root vnode. We can vput() it
* immediately, since with rootrefs > 0, it won't go away.
*/
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
if ((error = VFS_ROOT(mp, LK_EXCLUSIVE, &rootvp, td)) != 0) {
CTR2(KTR_VFS, "%s: vfs_root lookup failed with %d",
__func__, error);
Change the second argument of vflush() to an integer that specifies the number of references on the filesystem root vnode to be both expected and released. Many filesystems hold an extra reference on the filesystem root vnode, which must be accounted for when determining if the filesystem is busy and then released if it isn't busy. The old `skipvp' approach required individual filesystem xxx_unmount functions to re-implement much of vflush()'s logic to deal with the root vnode. All 9 filesystems that hold an extra reference on the root vnode got the logic wrong in the case of forced unmounts, so `umount -f' would always fail if there were any extra root vnode references. Fix this issue centrally in vflush(), now that we can. This commit also fixes a vnode reference leak in devfs, which could result in idle devfs filesystems that refuse to unmount. Reviewed by: phk, bp
2001-05-16 18:04:37 +00:00
return (error);
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
}
Change the second argument of vflush() to an integer that specifies the number of references on the filesystem root vnode to be both expected and released. Many filesystems hold an extra reference on the filesystem root vnode, which must be accounted for when determining if the filesystem is busy and then released if it isn't busy. The old `skipvp' approach required individual filesystem xxx_unmount functions to re-implement much of vflush()'s logic to deal with the root vnode. All 9 filesystems that hold an extra reference on the root vnode got the logic wrong in the case of forced unmounts, so `umount -f' would always fail if there were any extra root vnode references. Fix this issue centrally in vflush(), now that we can. This commit also fixes a vnode reference leak in devfs, which could result in idle devfs filesystems that refuse to unmount. Reviewed by: phk, bp
2001-05-16 18:04:37 +00:00
vput(rootvp);
- Don't hold the vn lock while calling VOP_CLOSE in vclean().
2002-07-07 06:38:22 +00:00
Change the second argument of vflush() to an integer that specifies the number of references on the filesystem root vnode to be both expected and released. Many filesystems hold an extra reference on the filesystem root vnode, which must be accounted for when determining if the filesystem is busy and then released if it isn't busy. The old `skipvp' approach required individual filesystem xxx_unmount functions to re-implement much of vflush()'s logic to deal with the root vnode. All 9 filesystems that hold an extra reference on the root vnode got the logic wrong in the case of forced unmounts, so `umount -f' would always fail if there were any extra root vnode references. Fix this issue centrally in vflush(), now that we can. This commit also fixes a vnode reference leak in devfs, which could result in idle devfs filesystems that refuse to unmount. Reviewed by: phk, bp
2001-05-16 18:04:37 +00:00
}
Remove mntvnode_mtx and replace it with per-mountpoint mutex. Introduce two new macros MNT_ILOCK(mp)/MNT_IUNLOCK(mp) to operate on this mutex transparently. Eventually new mutex will be protecting more fields in struct mount, not only vnode list. Discussed with: jeff
2003-11-05 04:30:08 +00:00
MNT_ILOCK(mp);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
loop:
Add marker vnodes to ensure that all vnodes associated with the mount point are iterated over when using MNT_VNODE_FOREACH. Reviewed by: truckman
2006-01-09 20:42:19 +00:00
MNT_VNODE_FOREACH(vp, mp, mvp) {
This is the kernel Lite/2 commit. There are some requisite userland changes, so don't expect to be able to run the kernel as-is (very well) without the appropriate Lite/2 userland changes. The system boots and can mount UFS filesystems. Untested: ext2fs, msdosfs, NFS Known problems: Incorrect Berkeley ID strings in some files. Mount_std mounts will not work until the getfsent library routine is changed. Reviewed by: various people Submitted by: Jeffery Hsu <hsu@freebsd.org>
1997-02-10 02:22:35 +00:00
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
VI_LOCK(vp);
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
vholdl(vp);
Remove mntvnode_mtx and replace it with per-mountpoint mutex. Introduce two new macros MNT_ILOCK(mp)/MNT_IUNLOCK(mp) to operate on this mutex transparently. Eventually new mutex will be protecting more fields in struct mount, not only vnode list. Discussed with: jeff
2003-11-05 04:30:08 +00:00
MNT_IUNLOCK(mp);
vn_lock() is currently only used with the 'curthread' passed as argument. Remove this argument and pass curthread directly to underlying VOP_LOCK1() VFS method. This modify makes the code cleaner and in particular remove an annoying dependence helping next lockmgr() cleanup. KPI results, obviously, changed. Manpage and FreeBSD_version will be updated through further commits. As a side note, would be valuable to say that next commits will address a similar cleanup about VFS methods, in particular vop_lock1 and vop_unlock. Tested by: Diego Sardina <siarodx at gmail dot com>, Andrea Di Pasquale <whyx dot it at gmail dot com>
2008-01-10 01:10:58 +00:00
error = vn_lock(vp, LK_INTERLOCK | LK_EXCLUSIVE);
- Fix an XXX. Check the error of vn_lock() in vflush(). Don't specify LK_RETRY either, we don't want this vnode if it turns into another. - Remove the code that checks the mount point after acquiring the lock we are guaranteed to either fail or get the vnode that we wanted.
2003-10-05 07:12:38 +00:00
if (error) {
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
vdrop(vp);
Remove mntvnode_mtx and replace it with per-mountpoint mutex. Introduce two new macros MNT_ILOCK(mp)/MNT_IUNLOCK(mp) to operate on this mutex transparently. Eventually new mutex will be protecting more fields in struct mount, not only vnode list. Discussed with: jeff
2003-11-05 04:30:08 +00:00
MNT_ILOCK(mp);
Add marker vnodes to ensure that all vnodes associated with the mount point are iterated over when using MNT_VNODE_FOREACH. Reviewed by: truckman
2006-01-09 20:42:19 +00:00
MNT_VNODE_FOREACH_ABORT_ILOCKED(mp, mvp);
Detect that a vnode has been reclaimed while vflush() was waiting to lock the vnode and restart the loop. Vflush() is vulnerable since it does not hold a reference to the vnode and it holds no other locks while waiting for the vnode lock. The vnode will no longer be on the list when the loop is restarted. Approved by: re (rwatson)
2003-05-16 19:46:51 +00:00
goto loop;
}
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
/*
- Replace v_flag with v_iflag and v_vflag - v_vflag is protected by the vnode lock and is used when synchronization with VOP calls is needed. - v_iflag is protected by interlock and is used for dealing with vnode management issues. These flags include X/O LOCK, FREE, DOOMED, etc. - All accesses to v_iflag and v_vflag have either been locked or marked with mp_fixme's. - Many ASSERT_VOP_LOCKED calls have been added where the locking was not clear. - Many functions in vfs_subr.c were restructured to provide for stronger locking. Idea stolen from: BSD/OS
2002-08-04 10:29:36 +00:00
* Skip over a vnodes marked VV_SYSTEM.
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
*/
- Replace v_flag with v_iflag and v_vflag - v_vflag is protected by the vnode lock and is used when synchronization with VOP calls is needed. - v_iflag is protected by interlock and is used for dealing with vnode management issues. These flags include X/O LOCK, FREE, DOOMED, etc. - All accesses to v_iflag and v_vflag have either been locked or marked with mp_fixme's. - Many ASSERT_VOP_LOCKED calls have been added where the locking was not clear. - Many functions in vfs_subr.c were restructured to provide for stronger locking. Idea stolen from: BSD/OS
2002-08-04 10:29:36 +00:00
if ((flags & SKIPSYSTEM) && (vp->v_vflag & VV_SYSTEM)) {
VOP_LOCK1() (and so VOP_LOCK()) and VOP_UNLOCK() are only used in conjuction with 'thread' argument passing which is always curthread. Remove the unuseful extra-argument and pass explicitly curthread to lower layer functions, when necessary. KPI results broken by this change, which should affect several ports, so version bumping and manpage update will be further committed. Tested by: kris, pho, Diego Sardina <siarodx at gmail dot com>
2008-01-13 14:44:15 +00:00
VOP_UNLOCK(vp, 0);
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
vdrop(vp);
Remove mntvnode_mtx and replace it with per-mountpoint mutex. Introduce two new macros MNT_ILOCK(mp)/MNT_IUNLOCK(mp) to operate on this mutex transparently. Eventually new mutex will be protecting more fields in struct mount, not only vnode list. Discussed with: jeff
2003-11-05 04:30:08 +00:00
MNT_ILOCK(mp);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
continue;
This is the kernel Lite/2 commit. There are some requisite userland changes, so don't expect to be able to run the kernel as-is (very well) without the appropriate Lite/2 userland changes. The system boots and can mount UFS filesystems. Untested: ext2fs, msdosfs, NFS Known problems: Incorrect Berkeley ID strings in some files. Mount_std mounts will not work until the getfsent library routine is changed. Reviewed by: various people Submitted by: Jeffery Hsu <hsu@freebsd.org>
1997-02-10 02:22:35 +00:00
}
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
/*
When downgrading a filesystem from read-write to read-only, operations involving file removal or file update were not always being fully committed to disk. The result was lost files or corrupted file data. This change ensures that the filesystem is properly synced to disk before the filesystem is down-graded. This delta also fixes a long standing bug in which a file open for reading has been unlinked. When the last open reference to the file is closed, the inode is reclaimed by the filesystem. Previously, if the filesystem had been down-graded to read-only, the inode could not be reclaimed, and thus was lost and had to be later recovered by fsck. With this change, such files are found at the time of the down-grade. Normally they will result in the filesystem down-grade failing with `device busy'. If a forcible down-grade is done, then the affected files will be revoked causing the inode to be released and the open file descriptors to begin failing on attempts to read. Submitted by: "Sam Leffler" <sam@errno.com>
2002-01-15 07:17:12 +00:00
* If WRITECLOSE is set, flush out unlinked but still open
* files (even if open only for reading) and regular file
Gratuitous whitespace cleanup.
2002-06-06 15:46:38 +00:00
* vnodes open for writing.
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
*/
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
if (flags & WRITECLOSE) {
Decontextualize the couplet VOP_GETATTR / VOP_SETATTR as the passed thread was always curthread and totally unuseful. Tested by: Giovanni Trematerra <giovanni dot trematerra at gmail dot com>
2008-08-28 15:23:18 +00:00
error = VOP_GETATTR(vp, &vattr, td->td_ucred);
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
VI_LOCK(vp);
- Fix interlock handling in vn_lock(). Previously, vn_lock() could return with interlock held in error conditions when the caller did not specify LK_INTERLOCK. - Add several comments to vn_lock() describing the rational behind the code flow since it was not immediately obvious.
2002-08-22 06:51:06 +00:00
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
if ((vp->v_type == VNON ||
(error == 0 && vattr.va_nlink > 0)) &&
(vp->v_writecount == 0 || vp->v_type != VREG)) {
VOP_LOCK1() (and so VOP_LOCK()) and VOP_UNLOCK() are only used in conjuction with 'thread' argument passing which is always curthread. Remove the unuseful extra-argument and pass explicitly curthread to lower layer functions, when necessary. KPI results broken by this change, which should affect several ports, so version bumping and manpage update will be further committed. Tested by: kris, pho, Diego Sardina <siarodx at gmail dot com>
2008-01-13 14:44:15 +00:00
VOP_UNLOCK(vp, 0);
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
vdropl(vp);
Remove mntvnode_mtx and replace it with per-mountpoint mutex. Introduce two new macros MNT_ILOCK(mp)/MNT_IUNLOCK(mp) to operate on this mutex transparently. Eventually new mutex will be protecting more fields in struct mount, not only vnode list. Discussed with: jeff
2003-11-05 04:30:08 +00:00
MNT_ILOCK(mp);
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
continue;
}
} else
VI_LOCK(vp);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
/*
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
* With v_usecount == 0, all we need to do is clear out the
* vnode data structures and we are done.
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
*
* If FORCECLOSE is set, forcibly close the vnode.
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
*/
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
if (vp->v_usecount == 0 || (flags & FORCECLOSE)) {
VNASSERT(vp->v_usecount == 0 ||
(vp->v_type != VCHR && vp->v_type != VBLK), vp,
Since we do not support forceful unmount of DEVFS we can do away with the partially implemented vnode-readoption code in vgonechrl().
2005-01-04 08:49:14 +00:00
("device VNODE %p is FORCECLOSED", vp));
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
vgonel(vp);
} else {
busy++;
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
#ifdef DIAGNOSTIC
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
if (busyprt)
vprint("vflush: busy vnode", vp);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
#endif
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
}
VOP_LOCK1() (and so VOP_LOCK()) and VOP_UNLOCK() are only used in conjuction with 'thread' argument passing which is always curthread. Remove the unuseful extra-argument and pass explicitly curthread to lower layer functions, when necessary. KPI results broken by this change, which should affect several ports, so version bumping and manpage update will be further committed. Tested by: kris, pho, Diego Sardina <siarodx at gmail dot com>
2008-01-13 14:44:15 +00:00
VOP_UNLOCK(vp, 0);
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
vdropl(vp);
Remove mntvnode_mtx and replace it with per-mountpoint mutex. Introduce two new macros MNT_ILOCK(mp)/MNT_IUNLOCK(mp) to operate on this mutex transparently. Eventually new mutex will be protecting more fields in struct mount, not only vnode list. Discussed with: jeff
2003-11-05 04:30:08 +00:00
MNT_ILOCK(mp);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
}
Remove mntvnode_mtx and replace it with per-mountpoint mutex. Introduce two new macros MNT_ILOCK(mp)/MNT_IUNLOCK(mp) to operate on this mutex transparently. Eventually new mutex will be protecting more fields in struct mount, not only vnode list. Discussed with: jeff
2003-11-05 04:30:08 +00:00
MNT_IUNLOCK(mp);
Change the second argument of vflush() to an integer that specifies the number of references on the filesystem root vnode to be both expected and released. Many filesystems hold an extra reference on the filesystem root vnode, which must be accounted for when determining if the filesystem is busy and then released if it isn't busy. The old `skipvp' approach required individual filesystem xxx_unmount functions to re-implement much of vflush()'s logic to deal with the root vnode. All 9 filesystems that hold an extra reference on the root vnode got the logic wrong in the case of forced unmounts, so `umount -f' would always fail if there were any extra root vnode references. Fix this issue centrally in vflush(), now that we can. This commit also fixes a vnode reference leak in devfs, which could result in idle devfs filesystems that refuse to unmount. Reviewed by: phk, bp
2001-05-16 18:04:37 +00:00
if (rootrefs > 0 && (flags & FORCECLOSE) == 0) {
/*
* If just the root vnode is busy, and if its refcount
* is equal to `rootrefs', then go ahead and kill it.
*/
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
VI_LOCK(rootvp);
Change the second argument of vflush() to an integer that specifies the number of references on the filesystem root vnode to be both expected and released. Many filesystems hold an extra reference on the filesystem root vnode, which must be accounted for when determining if the filesystem is busy and then released if it isn't busy. The old `skipvp' approach required individual filesystem xxx_unmount functions to re-implement much of vflush()'s logic to deal with the root vnode. All 9 filesystems that hold an extra reference on the root vnode got the logic wrong in the case of forced unmounts, so `umount -f' would always fail if there were any extra root vnode references. Fix this issue centrally in vflush(), now that we can. This commit also fixes a vnode reference leak in devfs, which could result in idle devfs filesystems that refuse to unmount. Reviewed by: phk, bp
2001-05-16 18:04:37 +00:00
KASSERT(busy > 0, ("vflush: not busy"));
Convert KASSERTS to VNASSERTS
2005-02-17 10:28:58 +00:00
VNASSERT(rootvp->v_usecount >= rootrefs, rootvp,
Dump the actual bad values when this assertion is tripped.
2004-07-12 04:13:38 +00:00
("vflush: usecount %d < rootrefs %d",
rootvp->v_usecount, rootrefs));
Change the second argument of vflush() to an integer that specifies the number of references on the filesystem root vnode to be both expected and released. Many filesystems hold an extra reference on the filesystem root vnode, which must be accounted for when determining if the filesystem is busy and then released if it isn't busy. The old `skipvp' approach required individual filesystem xxx_unmount functions to re-implement much of vflush()'s logic to deal with the root vnode. All 9 filesystems that hold an extra reference on the root vnode got the logic wrong in the case of forced unmounts, so `umount -f' would always fail if there were any extra root vnode references. Fix this issue centrally in vflush(), now that we can. This commit also fixes a vnode reference leak in devfs, which could result in idle devfs filesystems that refuse to unmount. Reviewed by: phk, bp
2001-05-16 18:04:37 +00:00
if (busy == 1 && rootvp->v_usecount == rootrefs) {
VOP_LOCK1() (and so VOP_LOCK()) and VOP_UNLOCK() are only used in conjuction with 'thread' argument passing which is always curthread. Remove the unuseful extra-argument and pass explicitly curthread to lower layer functions, when necessary. KPI results broken by this change, which should affect several ports, so version bumping and manpage update will be further committed. Tested by: kris, pho, Diego Sardina <siarodx at gmail dot com>
2008-01-13 14:44:15 +00:00
VOP_LOCK(rootvp, LK_EXCLUSIVE|LK_INTERLOCK);
- Remove vx_lock, vx_unlock, vx_wait, etc. - Add a vn_start_write/vn_finished_write around vlrureclaim so we don't do writing ops without suspending. This could suspend the vlruproc which should not be a problem under normal circumstances. - Manually implement VMIGHTFREE in vlrureclaim as this was the only instance where it was used. - Acquire a lock before calling vgone() as it now requires it. - Move the acquisition of the vnode interlock from vtryrecycle() to getnewvnode() so that if it fails we don't drop and reacquire the vnode_free_list_mtx. - Check for a usecount or holdcount at the end of vtryrecycle() in case someone grabbed a ref while we were recycling. Abort the recycle, and on the final ref drop this vnode will be placed on the head of the free list. - Move the redundant VOP_INACTIVE protection code into the local vinactive() routine to avoid code bloat. - Keep the vnode lock held across calls to vgone() in several places. - vgonel() no longer uses XLOCK, instead callers must hold an exclusive vnode lock. The VI_DOOMED flag is set to allow other threads to detect a vnode which is no longer valid. This flag is set until the last reference is gone, and there are no chances for a new ref. vgonel() holds this lock across the entire function, which greatly simplifies logic. _ Only vfree() in one place in vgone() not three. - Adjust vget() to check the VI_DOOMED flag prior to waiting on the lock in the LK_NOWAIT case. In other cases, check after we have slept and acquired an exlusive lock. This will simulate the old vx_wait() behavior. Sponsored by: Isilon Systems, Inc.
2005-03-13 11:54:28 +00:00
vgone(rootvp);
VOP_LOCK1() (and so VOP_LOCK()) and VOP_UNLOCK() are only used in conjuction with 'thread' argument passing which is always curthread. Remove the unuseful extra-argument and pass explicitly curthread to lower layer functions, when necessary. KPI results broken by this change, which should affect several ports, so version bumping and manpage update will be further committed. Tested by: kris, pho, Diego Sardina <siarodx at gmail dot com>
2008-01-13 14:44:15 +00:00
VOP_UNLOCK(rootvp, 0);
Change the second argument of vflush() to an integer that specifies the number of references on the filesystem root vnode to be both expected and released. Many filesystems hold an extra reference on the filesystem root vnode, which must be accounted for when determining if the filesystem is busy and then released if it isn't busy. The old `skipvp' approach required individual filesystem xxx_unmount functions to re-implement much of vflush()'s logic to deal with the root vnode. All 9 filesystems that hold an extra reference on the root vnode got the logic wrong in the case of forced unmounts, so `umount -f' would always fail if there were any extra root vnode references. Fix this issue centrally in vflush(), now that we can. This commit also fixes a vnode reference leak in devfs, which could result in idle devfs filesystems that refuse to unmount. Reviewed by: phk, bp
2001-05-16 18:04:37 +00:00
busy = 0;
} else
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
VI_UNLOCK(rootvp);
Change the second argument of vflush() to an integer that specifies the number of references on the filesystem root vnode to be both expected and released. Many filesystems hold an extra reference on the filesystem root vnode, which must be accounted for when determining if the filesystem is busy and then released if it isn't busy. The old `skipvp' approach required individual filesystem xxx_unmount functions to re-implement much of vflush()'s logic to deal with the root vnode. All 9 filesystems that hold an extra reference on the root vnode got the logic wrong in the case of forced unmounts, so `umount -f' would always fail if there were any extra root vnode references. Fix this issue centrally in vflush(), now that we can. This commit also fixes a vnode reference leak in devfs, which could result in idle devfs filesystems that refuse to unmount. Reviewed by: phk, bp
2001-05-16 18:04:37 +00:00
}
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
if (busy) {
CTR2(KTR_VFS, "%s: failing as %d vnodes are busy", __func__,
busy);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
return (EBUSY);
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
}
Change the second argument of vflush() to an integer that specifies the number of references on the filesystem root vnode to be both expected and released. Many filesystems hold an extra reference on the filesystem root vnode, which must be accounted for when determining if the filesystem is busy and then released if it isn't busy. The old `skipvp' approach required individual filesystem xxx_unmount functions to re-implement much of vflush()'s logic to deal with the root vnode. All 9 filesystems that hold an extra reference on the root vnode got the logic wrong in the case of forced unmounts, so `umount -f' would always fail if there were any extra root vnode references. Fix this issue centrally in vflush(), now that we can. This commit also fixes a vnode reference leak in devfs, which could result in idle devfs filesystems that refuse to unmount. Reviewed by: phk, bp
2001-05-16 18:04:37 +00:00
for (; rootrefs > 0; rootrefs--)
vrele(rootvp);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
return (0);
}
Integrate vclean() into vgonel(). Various associated polishing.
2005-01-28 13:00:03 +00:00
/*
* Recycle an unused vnode to the front of the free list.
*/
int
Remove unused argument to vrecycle()
2005-01-28 13:08:21 +00:00
vrecycle(struct vnode *vp, struct thread *td)
Integrate vclean() into vgonel(). Various associated polishing.
2005-01-28 13:00:03 +00:00
{
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
int recycled;
Integrate vclean() into vgonel(). Various associated polishing.
2005-01-28 13:00:03 +00:00
Assert for exclusive vnode lock in vinactive(), vrecycle() and vgonel() functions. Reviewed by: kib
2008-07-27 11:48:15 +00:00
ASSERT_VOP_ELOCKED(vp, "vrecycle");
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
CTR2(KTR_VFS, "%s: vp %p", __func__, vp);
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
recycled = 0;
Integrate vclean() into vgonel(). Various associated polishing.
2005-01-28 13:00:03 +00:00
VI_LOCK(vp);
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
if (vp->v_usecount == 0) {
recycled = 1;
vgonel(vp);
Integrate vclean() into vgonel(). Various associated polishing.
2005-01-28 13:00:03 +00:00
}
VI_UNLOCK(vp);
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
return (recycled);
Integrate vclean() into vgonel(). Various associated polishing.
2005-01-28 13:00:03 +00:00
}
- Move the xlock 'locking' code into vx_lock() and vx_unlock(). - Create a new function, vgonechrl(), which performs vgone for an in-use character device. Move the code from vflush() that did this into vgonechrl(). - Hold the xlock across the entirety of vgonel() and vgonechrl() so that at no point will an invalid vnode exist on any list without XLOCK set. - Move the xlock code out of vclean() now that it is in the vgone*() functions.
2003-10-05 00:02:41 +00:00
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
/*
Integrate vclean() into vgonel(). Various associated polishing.
2005-01-28 13:00:03 +00:00
* Eliminate all activity associated with a vnode
* in preparation for reuse.
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
*/
Integrate vclean() into vgonel(). Various associated polishing.
2005-01-28 13:00:03 +00:00
void
vgone(struct vnode *vp)
{
VI_LOCK(vp);
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
vgonel(vp);
VI_UNLOCK(vp);
Integrate vclean() into vgonel(). Various associated polishing.
2005-01-28 13:00:03 +00:00
}
/*
* vgone, with the vp interlock held.
*/
void
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
vgonel(struct vnode *vp)
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
{
- Don't make vgonel() globally visible, we want to change its prototype anyway and it's not used outside of vfs_subr.c. - Change vgonel() to accept a parameter which determines whether or not we'll put the vnode on the free list when we're done. - Use the new vgonel() parameter rather than VI_DOOMED to signal our intentions in vtryrecycle(). - In vgonel() return if VI_DOOMED is already set, this vnode has already been reclaimed. Sponsored by: Isilon Systems, Inc.
2005-06-13 06:26:55 +00:00
struct thread *td;
- Dont clear OWEINACT in vbusy(), we still owe an inactive call if someone vhold()s us. - Avoid an extra mutex acquire and release in the common case of vgonel() by checking for OWEINACT at the start of the function. - Fix the case where we set OWEINACT in vput(). LK_EXCLUPGRADE drops our shared lock if it fails. Sponsored by: Isilon Systems, Inc.
2005-03-29 10:02:48 +00:00
int oweinact;
Make our v_usecount vnode reference count work identically to the original BSD code. The association between the vnode and the vm_object no longer includes reference counts. The major difference is that vm_object's are no longer freed gratuitiously from the vnode, and so once an object is created for the vnode, it will last as long as the vnode does. When a vnode object reference count is incremented, then the underlying vnode reference count is incremented also. The two "objects" are now more intimately related, and so the interactions are now much less complex. When vnodes are now normally placed onto the free queue with an object still attached. The rundown of the object happens at vnode rundown time, and happens with exactly the same filesystem semantics of the original VFS code. There is absolutely no need for vnode_pager_uncache and other travesties like that anymore. A side-effect of these changes is that SMP locking should be much simpler, the I/O copyin/copyout optimizations work, NFS should be more ponderable, and further work on layered filesystems should be less frustrating, because of the totally coherent management of the vnode objects and vnodes. Please be careful with your system while running this code, but I would greatly appreciate feedback as soon a reasonably possible.
1998-01-06 05:26:17 +00:00
int active;
Use vn_start_secondary_write() and vn_finished_secondary_write() as a replacement for vn_write_suspend_wait() to better account for secondary write processing. Close race where secondary writes could be started after ffs_sync() returned but before the file system was marked as suspended. Detect if secondary writes or softdep processing occurred during vnode sync loop in ffs_sync() and retry the loop if needed.
2006-03-08 23:43:39 +00:00
struct mount *mp;
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
Assert for exclusive vnode lock in vinactive(), vrecycle() and vgonel() functions. Reviewed by: kib
2008-07-27 11:48:15 +00:00
ASSERT_VOP_ELOCKED(vp, "vgonel");
Integrate vclean() into vgonel(). Various associated polishing.
2005-01-28 13:00:03 +00:00
ASSERT_VI_LOCKED(vp, "vgonel");
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
VNASSERT(vp->v_holdcnt, vp,
("vgonel: vp %p has no reference.", vp));
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
CTR2(KTR_VFS, "%s: vp %p", __func__, vp);
- In reassignbuf() add many asserts to validate the head and tail pointers of the clean and dirty lists. This is in an attempt to catch the wrong bufobj problem sooner. - In vgonel() don't acquire an extra reference in the active case, the vnode lock and VI_DOOMED protect us from recursively cleaning. - Also in vgonel() clean up some stale comments. Sponsored by: Isilon Systems, Inc. Approved by: re (blanket vfs)
2005-06-14 20:31:53 +00:00
td = curthread;
Integrate vclean() into vgonel(). Various associated polishing.
2005-01-28 13:00:03 +00:00
- Don't make vgonel() globally visible, we want to change its prototype anyway and it's not used outside of vfs_subr.c. - Change vgonel() to accept a parameter which determines whether or not we'll put the vnode on the free list when we're done. - Use the new vgonel() parameter rather than VI_DOOMED to signal our intentions in vtryrecycle(). - In vgonel() return if VI_DOOMED is already set, this vnode has already been reclaimed. Sponsored by: Isilon Systems, Inc.
2005-06-13 06:26:55 +00:00
/*
* Don't vgonel if we're already doomed.
*/
Do not drop the vnode interlock if vdropl is called on already doomed vnode. vdropl callers expect it to return with interlock still being held. MFC after: 2 days
2005-08-10 11:46:03 +00:00
if (vp->v_iflag & VI_DOOMED)
- Don't make vgonel() globally visible, we want to change its prototype anyway and it's not used outside of vfs_subr.c. - Change vgonel() to accept a parameter which determines whether or not we'll put the vnode on the free list when we're done. - Use the new vgonel() parameter rather than VI_DOOMED to signal our intentions in vtryrecycle(). - In vgonel() return if VI_DOOMED is already set, this vnode has already been reclaimed. Sponsored by: Isilon Systems, Inc.
2005-06-13 06:26:55 +00:00
return;
- In reassignbuf() add many asserts to validate the head and tail pointers of the clean and dirty lists. This is in an attempt to catch the wrong bufobj problem sooner. - In vgonel() don't acquire an extra reference in the active case, the vnode lock and VI_DOOMED protect us from recursively cleaning. - Also in vgonel() clean up some stale comments. Sponsored by: Isilon Systems, Inc. Approved by: re (blanket vfs)
2005-06-14 20:31:53 +00:00
vp->v_iflag |= VI_DOOMED;
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
/*
- In reassignbuf() add many asserts to validate the head and tail pointers of the clean and dirty lists. This is in an attempt to catch the wrong bufobj problem sooner. - In vgonel() don't acquire an extra reference in the active case, the vnode lock and VI_DOOMED protect us from recursively cleaning. - Also in vgonel() clean up some stale comments. Sponsored by: Isilon Systems, Inc. Approved by: re (blanket vfs)
2005-06-14 20:31:53 +00:00
* Check to see if the vnode is in use. If so, we have to call
* VOP_CLOSE() and VOP_INACTIVE().
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
*/
- In reassignbuf() add many asserts to validate the head and tail pointers of the clean and dirty lists. This is in an attempt to catch the wrong bufobj problem sooner. - In vgonel() don't acquire an extra reference in the active case, the vnode lock and VI_DOOMED protect us from recursively cleaning. - Also in vgonel() clean up some stale comments. Sponsored by: Isilon Systems, Inc. Approved by: re (blanket vfs)
2005-06-14 20:31:53 +00:00
active = vp->v_usecount;
- Dont clear OWEINACT in vbusy(), we still owe an inactive call if someone vhold()s us. - Avoid an extra mutex acquire and release in the common case of vgonel() by checking for OWEINACT at the start of the function. - Fix the case where we set OWEINACT in vput(). LK_EXCLUPGRADE drops our shared lock if it fails. Sponsored by: Isilon Systems, Inc.
2005-03-29 10:02:48 +00:00
oweinact = (vp->v_iflag & VI_OWEINACT);
- Remove vx_lock, vx_unlock, vx_wait, etc. - Add a vn_start_write/vn_finished_write around vlrureclaim so we don't do writing ops without suspending. This could suspend the vlruproc which should not be a problem under normal circumstances. - Manually implement VMIGHTFREE in vlrureclaim as this was the only instance where it was used. - Acquire a lock before calling vgone() as it now requires it. - Move the acquisition of the vnode interlock from vtryrecycle() to getnewvnode() so that if it fails we don't drop and reacquire the vnode_free_list_mtx. - Check for a usecount or holdcount at the end of vtryrecycle() in case someone grabbed a ref while we were recycling. Abort the recycle, and on the final ref drop this vnode will be placed on the head of the free list. - Move the redundant VOP_INACTIVE protection code into the local vinactive() routine to avoid code bloat. - Keep the vnode lock held across calls to vgone() in several places. - vgonel() no longer uses XLOCK, instead callers must hold an exclusive vnode lock. The VI_DOOMED flag is set to allow other threads to detect a vnode which is no longer valid. This flag is set until the last reference is gone, and there are no chances for a new ref. vgonel() holds this lock across the entire function, which greatly simplifies logic. _ Only vfree() in one place in vgone() not three. - Adjust vget() to check the VI_DOOMED flag prior to waiting on the lock in the LK_NOWAIT case. In other cases, check after we have slept and acquired an exlusive lock. This will simulate the old vx_wait() behavior. Sponsored by: Isilon Systems, Inc.
2005-03-13 11:54:28 +00:00
VI_UNLOCK(vp);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
/*
This is the kernel Lite/2 commit. There are some requisite userland changes, so don't expect to be able to run the kernel as-is (very well) without the appropriate Lite/2 userland changes. The system boots and can mount UFS filesystems. Untested: ext2fs, msdosfs, NFS Known problems: Incorrect Berkeley ID strings in some files. Mount_std mounts will not work until the getfsent library routine is changed. Reviewed by: various people Submitted by: Jeffery Hsu <hsu@freebsd.org>
1997-02-10 02:22:35 +00:00
* Clean out any buffers associated with the vnode.
If a buffer flush fails when trying to reclaim a vnode, it is too late to save the vnode, so just toss any remaining unwritten buffers rather than leaving them lying around to make trouble in the future.
2000-07-04 03:23:29 +00:00
* If the flush fails, just toss the buffers.
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
*/
Use vn_start_secondary_write() and vn_finished_secondary_write() as a replacement for vn_write_suspend_wait() to better account for secondary write processing. Close race where secondary writes could be started after ffs_sync() returned but before the file system was marked as suspended. Detect if secondary writes or softdep processing occurred during vnode sync loop in ffs_sync() and retry the loop if needed.
2006-03-08 23:43:39 +00:00
mp = NULL;
Remove another ';' after if(). Also spotted by: bz
2005-03-27 07:53:13 +00:00
if (!TAILQ_EMPTY(&vp->v_bufobj.bo_dirty.bv_hd))
Use vn_start_secondary_write() and vn_finished_secondary_write() as a replacement for vn_write_suspend_wait() to better account for secondary write processing. Close race where secondary writes could be started after ffs_sync() returned but before the file system was marked as suspended. Detect if secondary writes or softdep processing occurred during vnode sync loop in ffs_sync() and retry the loop if needed.
2006-03-08 23:43:39 +00:00
(void) vn_start_secondary_write(vp, &mp, V_WAIT);
Remove the struct thread unuseful argument from bufobj interface. In particular following functions KPI results modified: - bufobj_invalbuf() - bufsync() and BO_SYNC() "virtual method" of the buffer objects set. Main consumers of bufobj functions are affected by this change too and, in particular, functions which changed their KPI are: - vinvalbuf() - g_vfs_close() Due to the KPI breakage, __FreeBSD_version will be bumped in a later commit. As a side note, please consider just temporary the 'curthread' argument passing to VOP_SYNC() (in bufsync()) as it will be axed out ASAP Reviewed by: kib Tested by: Giovanni Trematerra <giovanni dot trematerra at gmail dot com>
2008-10-10 21:23:50 +00:00
if (vinvalbuf(vp, V_SAVE, 0, 0) != 0)
vinvalbuf(vp, 0, 0, 0);
If a buffer flush fails when trying to reclaim a vnode, it is too late to save the vnode, so just toss any remaining unwritten buffers rather than leaving them lying around to make trouble in the future.
2000-07-04 03:23:29 +00:00
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
/*
This is the kernel Lite/2 commit. There are some requisite userland changes, so don't expect to be able to run the kernel as-is (very well) without the appropriate Lite/2 userland changes. The system boots and can mount UFS filesystems. Untested: ext2fs, msdosfs, NFS Known problems: Incorrect Berkeley ID strings in some files. Mount_std mounts will not work until the getfsent library routine is changed. Reviewed by: various people Submitted by: Jeffery Hsu <hsu@freebsd.org>
1997-02-10 02:22:35 +00:00
* If purging an active vnode, it must be closed and
- Remove vx_lock, vx_unlock, vx_wait, etc. - Add a vn_start_write/vn_finished_write around vlrureclaim so we don't do writing ops without suspending. This could suspend the vlruproc which should not be a problem under normal circumstances. - Manually implement VMIGHTFREE in vlrureclaim as this was the only instance where it was used. - Acquire a lock before calling vgone() as it now requires it. - Move the acquisition of the vnode interlock from vtryrecycle() to getnewvnode() so that if it fails we don't drop and reacquire the vnode_free_list_mtx. - Check for a usecount or holdcount at the end of vtryrecycle() in case someone grabbed a ref while we were recycling. Abort the recycle, and on the final ref drop this vnode will be placed on the head of the free list. - Move the redundant VOP_INACTIVE protection code into the local vinactive() routine to avoid code bloat. - Keep the vnode lock held across calls to vgone() in several places. - vgonel() no longer uses XLOCK, instead callers must hold an exclusive vnode lock. The VI_DOOMED flag is set to allow other threads to detect a vnode which is no longer valid. This flag is set until the last reference is gone, and there are no chances for a new ref. vgonel() holds this lock across the entire function, which greatly simplifies logic. _ Only vfree() in one place in vgone() not three. - Adjust vget() to check the VI_DOOMED flag prior to waiting on the lock in the LK_NOWAIT case. In other cases, check after we have slept and acquired an exlusive lock. This will simulate the old vx_wait() behavior. Sponsored by: Isilon Systems, Inc.
2005-03-13 11:54:28 +00:00
* deactivated before being reclaimed.
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
*/
- Dont clear OWEINACT in vbusy(), we still owe an inactive call if someone vhold()s us. - Avoid an extra mutex acquire and release in the common case of vgonel() by checking for OWEINACT at the start of the function. - Fix the case where we set OWEINACT in vput(). LK_EXCLUPGRADE drops our shared lock if it fails. Sponsored by: Isilon Systems, Inc.
2005-03-29 10:02:48 +00:00
if (active)
Eliminate the constant flags argument to vclean()
2005-01-24 22:22:02 +00:00
VOP_CLOSE(vp, FNONBLOCK, NOCRED, td);
- Dont clear OWEINACT in vbusy(), we still owe an inactive call if someone vhold()s us. - Avoid an extra mutex acquire and release in the common case of vgonel() by checking for OWEINACT at the start of the function. - Fix the case where we set OWEINACT in vput(). LK_EXCLUPGRADE drops our shared lock if it fails. Sponsored by: Isilon Systems, Inc.
2005-03-29 10:02:48 +00:00
if (oweinact || active) {
Add a new vnode flag VI_DOINGINACT to indicate that a VOP_INACTIVE call is in progress on the vnode. When vput() or vrele() sees a 1->0 reference count transition, it now return without any further action if this flag is set. This flag is necessary to avoid recursion into VOP_INACTIVE if the filesystem inactive routine causes the reference count to increase and then drop back to zero. It is also used to guarantee that an unlocked vnode will not be recycled while blocked in VOP_INACTIVE(). There are at least two cases where the recursion can occur: one is that the softupdates code called by ufs_inactive() via ffs_truncate() can call vput() on the vnode. This has been reported by many people as "lockmgr: draining against myself" panics. The other case is that nfs_inactive() can call vget() and then vrele() on the vnode to clean up a sillyrename file. Reviewed by: mckusick (an older version of the patch)
2002-12-29 18:30:49 +00:00
VI_LOCK(vp);
- Remove vx_lock, vx_unlock, vx_wait, etc. - Add a vn_start_write/vn_finished_write around vlrureclaim so we don't do writing ops without suspending. This could suspend the vlruproc which should not be a problem under normal circumstances. - Manually implement VMIGHTFREE in vlrureclaim as this was the only instance where it was used. - Acquire a lock before calling vgone() as it now requires it. - Move the acquisition of the vnode interlock from vtryrecycle() to getnewvnode() so that if it fails we don't drop and reacquire the vnode_free_list_mtx. - Check for a usecount or holdcount at the end of vtryrecycle() in case someone grabbed a ref while we were recycling. Abort the recycle, and on the final ref drop this vnode will be placed on the head of the free list. - Move the redundant VOP_INACTIVE protection code into the local vinactive() routine to avoid code bloat. - Keep the vnode lock held across calls to vgone() in several places. - vgonel() no longer uses XLOCK, instead callers must hold an exclusive vnode lock. The VI_DOOMED flag is set to allow other threads to detect a vnode which is no longer valid. This flag is set until the last reference is gone, and there are no chances for a new ref. vgonel() holds this lock across the entire function, which greatly simplifies logic. _ Only vfree() in one place in vgone() not three. - Adjust vget() to check the VI_DOOMED flag prior to waiting on the lock in the LK_NOWAIT case. In other cases, check after we have slept and acquired an exlusive lock. This will simulate the old vx_wait() behavior. Sponsored by: Isilon Systems, Inc.
2005-03-13 11:54:28 +00:00
if ((vp->v_iflag & VI_DOINGINACT) == 0)
vinactive(vp, td);
Add a new vnode flag VI_DOINGINACT to indicate that a VOP_INACTIVE call is in progress on the vnode. When vput() or vrele() sees a 1->0 reference count transition, it now return without any further action if this flag is set. This flag is necessary to avoid recursion into VOP_INACTIVE if the filesystem inactive routine causes the reference count to increase and then drop back to zero. It is also used to guarantee that an unlocked vnode will not be recycled while blocked in VOP_INACTIVE(). There are at least two cases where the recursion can occur: one is that the softupdates code called by ufs_inactive() via ffs_truncate() can call vput() on the vnode. This has been reported by many people as "lockmgr: draining against myself" panics. The other case is that nfs_inactive() can call vget() and then vrele() on the vnode to clean up a sillyrename file. Reviewed by: mckusick (an older version of the patch)
2002-12-29 18:30:49 +00:00
VI_UNLOCK(vp);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
}
/*
* Reclaim the vnode.
*/
KSE Milestone 2 Note ALL MODULES MUST BE RECOMPILED make the kernel aware that there are smaller units of scheduling than the process. (but only allow one thread per process at this time). This is functionally equivalent to teh previousl -current except that there is a thread associated with each process. Sorry john! (your next MFC will be a doosie!) Reviewed by: peter@freebsd.org, dillon@freebsd.org X-MFC after: ha ha ha ha
2001-09-12 08:38:13 +00:00
if (VOP_RECLAIM(vp, td))
- Remove vx_lock, vx_unlock, vx_wait, etc. - Add a vn_start_write/vn_finished_write around vlrureclaim so we don't do writing ops without suspending. This could suspend the vlruproc which should not be a problem under normal circumstances. - Manually implement VMIGHTFREE in vlrureclaim as this was the only instance where it was used. - Acquire a lock before calling vgone() as it now requires it. - Move the acquisition of the vnode interlock from vtryrecycle() to getnewvnode() so that if it fails we don't drop and reacquire the vnode_free_list_mtx. - Check for a usecount or holdcount at the end of vtryrecycle() in case someone grabbed a ref while we were recycling. Abort the recycle, and on the final ref drop this vnode will be placed on the head of the free list. - Move the redundant VOP_INACTIVE protection code into the local vinactive() routine to avoid code bloat. - Keep the vnode lock held across calls to vgone() in several places. - vgonel() no longer uses XLOCK, instead callers must hold an exclusive vnode lock. The VI_DOOMED flag is set to allow other threads to detect a vnode which is no longer valid. This flag is set until the last reference is gone, and there are no chances for a new ref. vgonel() holds this lock across the entire function, which greatly simplifies logic. _ Only vfree() in one place in vgone() not three. - Adjust vget() to check the VI_DOOMED flag prior to waiting on the lock in the LK_NOWAIT case. In other cases, check after we have slept and acquired an exlusive lock. This will simulate the old vx_wait() behavior. Sponsored by: Isilon Systems, Inc.
2005-03-13 11:54:28 +00:00
panic("vgone: cannot reclaim");
Use vn_start_secondary_write() and vn_finished_secondary_write() as a replacement for vn_write_suspend_wait() to better account for secondary write processing. Close race where secondary writes could be started after ffs_sync() returned but before the file system was marked as suspended. Detect if secondary writes or softdep processing occurred during vnode sync loop in ffs_sync() and retry the loop if needed.
2006-03-08 23:43:39 +00:00
if (mp != NULL)
vn_finished_secondary_write(mp);
Convert KASSERTS to VNASSERTS
2005-02-17 10:28:58 +00:00
VNASSERT(vp->v_object == NULL, vp,
Don't call VOP_DESTROYVOBJECT(), trust that VOP_RECLAIM() did what was necessary.
2005-02-07 07:48:03 +00:00
("vop_reclaim left v_object vp=%p, tag=%s", vp, vp->v_tag));
Move the head of byte-level advisory lock list from the filesystem-specific vnode data to the struct vnode. Provide the default implementation for the vop_advlock and vop_advlockasync. Purge the locks on the vnode reclaim by using the lf_purgelocks(). The default implementation is augmented for the nfs and smbfs. In the nfs_advlock, push the Giant inside the nfs_dolock. Before the change, the vop_advlock and vop_advlockasync have taken the unlocked vnode and dereferenced the fs-private inode data, racing with with the vnode reclamation due to forced unmount. Now, the vop_getattr under the shared vnode lock is used to obtain the inode size, and later, in the lf_advlockasync, after locking the vnode interlock, the VI_DOOMED flag is checked to prevent an operation on the doomed vnode. The implementation of the lf_purgelocks() is submitted by dfr. Reported by: kris Tested by: kris, pho Discussed with: jeff, dfr MFC after: 2 weeks
2008-04-16 11:33:32 +00:00
/*
* Clear the advisory locks and wake up waiting threads.
*/
lf_purgelocks(vp, &(vp->v_lockf));
- Don't cache_purge() in getnewvnode. It's done in vclean(). With this purge, the purge in vclean, and the filesystems purge, we had 3 purges per vnode. - Move the insmntque(vp, 0) to vclean() so that we may remove it from the two vgone() functions and reduce the number of lock operations required.
2003-10-05 02:48:04 +00:00
/*
* Delete from old mount point vnode list.
*/
When we traverse the vnodes on a mountpoint we need to look out for our cached 'next vnode' being removed from this mountpoint. If we find that it was recycled, we restart our traversal from the start of the list. Code to do that is in all local disk filesystems (and a few other places) and looks roughly like this: MNT_ILOCK(mp); loop: for (vp = TAILQ_FIRST(&mp...); (vp = nvp) != NULL; nvp = TAILQ_NEXT(vp,...)) { if (vp->v_mount != mp) goto loop; MNT_IUNLOCK(mp); ... MNT_ILOCK(mp); } MNT_IUNLOCK(mp); The code which takes vnodes off a mountpoint looks like this: MNT_ILOCK(vp->v_mount); ... TAILQ_REMOVE(&vp->v_mount->mnt_nvnodelist, vp, v_nmntvnodes); ... MNT_IUNLOCK(vp->v_mount); ... vp->v_mount = something; (Take a moment and try to spot the locking error before you read on.) On a SMP system, one CPU could have removed nvp from our mountlist but not yet gotten to assign a new value to vp->v_mount while another CPU simultaneously get to the top of the traversal loop where it finds that (vp->v_mount != mp) is not true despite the fact that the vnode has indeed been removed from our mountpoint. Fix: Introduce the macro MNT_VNODE_FOREACH() to traverse the list of vnodes on a mountpoint while taking into account that vnodes may be removed from the list as we go. This saves approx 65 lines of duplicated code. Split the insmntque() which potentially moves a vnode from one mount point to another into delmntque() and insmntque() which does just what the names say. Fix delmntque() to set vp->v_mount to NULL while holding the mountpoint lock.
2004-07-04 08:52:35 +00:00
delmntque(vp);
This is the kernel Lite/2 commit. There are some requisite userland changes, so don't expect to be able to run the kernel as-is (very well) without the appropriate Lite/2 userland changes. The system boots and can mount UFS filesystems. Untested: ext2fs, msdosfs, NFS Known problems: Incorrect Berkeley ID strings in some files. Mount_std mounts will not work until the getfsent library routine is changed. Reviewed by: various people Submitted by: Jeffery Hsu <hsu@freebsd.org>
1997-02-10 02:22:35 +00:00
cache_purge(vp);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
/*
- In reassignbuf() add many asserts to validate the head and tail pointers of the clean and dirty lists. This is in an attempt to catch the wrong bufobj problem sooner. - In vgonel() don't acquire an extra reference in the active case, the vnode lock and VI_DOOMED protect us from recursively cleaning. - Also in vgonel() clean up some stale comments. Sponsored by: Isilon Systems, Inc. Approved by: re (blanket vfs)
2005-06-14 20:31:53 +00:00
* Done with purge, reset to the standard lock and invalidate
* the vnode.
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
*/
- In reassignbuf() add many asserts to validate the head and tail pointers of the clean and dirty lists. This is in an attempt to catch the wrong bufobj problem sooner. - In vgonel() don't acquire an extra reference in the active case, the vnode lock and VI_DOOMED protect us from recursively cleaning. - Also in vgonel() clean up some stale comments. Sponsored by: Isilon Systems, Inc. Approved by: re (blanket vfs)
2005-06-14 20:31:53 +00:00
VI_LOCK(vp);
Unconditionally reset vp->v_vnlock back to the default in the vclean() function (e.g., vp->v_vnlock = &vp->v_lock) rather than requiring filesystems that use alternate locks to do so in their vop_reclaim functions. This change is a further cleanup of the vop_stdlock interface. Submitted by: Poul-Henning Kamp <phk@critter.freebsd.dk> Sponsored by: DARPA & NAI Labs.
2002-10-14 19:44:51 +00:00
vp->v_vnlock = &vp->v_lock;
Back when VOP_* was introduced, we did not have new-style struct initializations but we did have lofty goals and big ideals. Adjust to more contemporary circumstances and gain type checking. Replace the entire vop_t frobbing thing with properly typed structures. The only casualty is that we can not add a new VOP_ method with a loadable module. History has not given us reason to belive this would ever be feasible in the the first place. Eliminate in toto VOCALL(), vop_t, VNODEOP_SET() etc. Give coda correct prototypes and function definitions for all vop_()s. Generate a bit more data from the vnode_if.src file: a struct vop_vector and protype typedefs for all vop methods. Add a new vop_bypass() and make vop_default be a pointer to another struct vop_vector. Remove a lot of vfs_init since vop_vector is ready to use from the compiler. Cast various vop_mumble() to void * with uppercase name, for instance VOP_PANIC, VOP_NULL etc. Implement VCALL() by making vdesc_offset the offsetof() the relevant function pointer in vop_vector. This is disgusting but since the code is generated by a script comparatively safe. The alternative for nullfs etc. would be much worse. Fix up all vnode method vectors to remove casts so they become typesafe. (The bulk of this is generated by scripts)
2004-12-01 23:16:38 +00:00
vp->v_op = &dead_vnodeops;
Remove all use of vnode->v_tag, replacing with appropriate substitutes. v_tag is now const char * and should only be used for debugging. Additionally: 1. All users of VT_NTS now check vfsconf->vf_type VFCF_NETWORK 2. The user of VT_PROCFS now checks for the new flag VV_PROCDEP, which is propagated by pseudofs to all child vnodes if the fs sets PFS_PROCDEP. Suggested by: phk Reviewed by: bde, rwatson (earlier version)
2002-09-14 09:02:28 +00:00
vp->v_tag = "none";
- Remove vx_lock, vx_unlock, vx_wait, etc. - Add a vn_start_write/vn_finished_write around vlrureclaim so we don't do writing ops without suspending. This could suspend the vlruproc which should not be a problem under normal circumstances. - Manually implement VMIGHTFREE in vlrureclaim as this was the only instance where it was used. - Acquire a lock before calling vgone() as it now requires it. - Move the acquisition of the vnode interlock from vtryrecycle() to getnewvnode() so that if it fails we don't drop and reacquire the vnode_free_list_mtx. - Check for a usecount or holdcount at the end of vtryrecycle() in case someone grabbed a ref while we were recycling. Abort the recycle, and on the final ref drop this vnode will be placed on the head of the free list. - Move the redundant VOP_INACTIVE protection code into the local vinactive() routine to avoid code bloat. - Keep the vnode lock held across calls to vgone() in several places. - vgonel() no longer uses XLOCK, instead callers must hold an exclusive vnode lock. The VI_DOOMED flag is set to allow other threads to detect a vnode which is no longer valid. This flag is set until the last reference is gone, and there are no chances for a new ref. vgonel() holds this lock across the entire function, which greatly simplifies logic. _ Only vfree() in one place in vgone() not three. - Adjust vget() to check the VI_DOOMED flag prior to waiting on the lock in the LK_NOWAIT case. In other cases, check after we have slept and acquired an exlusive lock. This will simulate the old vx_wait() behavior. Sponsored by: Isilon Systems, Inc.
2005-03-13 11:54:28 +00:00
vp->v_type = VBAD;
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
}
/*
* Calculate the total number of references to a special device.
*/
The big 4.4BSD Lite to FreeBSD 2.0.0 (Development) patch. Reviewed by: Rodney W. Grimes Submitted by: John Dyson and David Greenman
1994-05-25 09:21:21 +00:00
int
Convert remaining functions in vfs_subr.c from K&R prototypes to ANSI C prototypes, as the majority of new functions added have been in this style. Changing prototype style now results in gcc noticing that the implementation of vn_pollrecord() has a 'short' argument instead of 'int' as prototyped in vnode.h, so correct that definition. In practice this didn't matter as only poll flags in the lower 16 bits are used. MFC after: 1 week
2006-01-21 19:42:10 +00:00
vcount(struct vnode *vp)
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
{
int count;
Do not refcount the cdevsw, but rather maintain a cdev->si_threadcount of the number of threads which are inside whatever is behind the cdevsw for this particular cdev. Make the device mutex visible through dev_lock() and dev_unlock(). We may want finer granularity later. Replace spechash_mtx use with dev_lock()/dev_unlock().
2004-09-23 07:17:41 +00:00
dev_lock();
Fix the spechash lock order reversal by keeping an updated sum of v_usecount in the dev_t which vcount() can return without locking any vnodes. Seen by: jhb
2002-10-24 19:38:56 +00:00
count = vp->v_rdev->si_usecount;
Do not refcount the cdevsw, but rather maintain a cdev->si_threadcount of the number of threads which are inside whatever is behind the cdevsw for this particular cdev. Make the device mutex visible through dev_lock() and dev_unlock(). We may want finer granularity later. Replace spechash_mtx use with dev_lock()/dev_unlock().
2004-09-23 07:17:41 +00:00
dev_unlock();
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
return (count);
}
Simplify the handling of VCHR and VBLK vnodes using the new dev_t: Make the alias list a SLIST. Drop the "fast recycling" optimization of vnodes (including the returning of a prexisting but stale vnode from checkalias). It doesn't buy us anything now that we don't hardlimit vnodes anymore. Rename checkalias2() and checkalias() to addalias() and addaliasu() - which takes dev_t and udev_t arg respectively. Make the revoke syscalls use vcount() instead of VALIASED. Remove VALIASED flag, we don't need it now and it is faster to traverse the much shorter lists than to maintain the flag. vfs_mountedon() can check the dev_t directly, all the vnodes point to the same one. Print the devicename in specfs/vprint(). Remove a couple of stale LFS vnode flags. Remove unimplemented/unused LK_DRAINED;
1999-08-26 14:53:31 +00:00
Do refcounting of open devices (more) correctly. count_dev funtion by phk.
2000-02-07 23:05:40 +00:00
/*
Do the dreaded s/dev_t/struct cdev */ Bump __FreeBSD_version accordingly.
2004-06-16 09:47:26 +00:00
* Same as above, but using the struct cdev *as argument
Do refcounting of open devices (more) correctly. count_dev funtion by phk.
2000-02-07 23:05:40 +00:00
*/
int
Convert remaining functions in vfs_subr.c from K&R prototypes to ANSI C prototypes, as the majority of new functions added have been in this style. Changing prototype style now results in gcc noticing that the implementation of vn_pollrecord() has a 'short' argument instead of 'int' as prototyped in vnode.h, so correct that definition. In practice this didn't matter as only poll flags in the lower 16 bits are used. MFC after: 1 week
2006-01-21 19:42:10 +00:00
count_dev(struct cdev *dev)
Do refcounting of open devices (more) correctly. count_dev funtion by phk.
2000-02-07 23:05:40 +00:00
{
Simplify count_dev()
2003-10-17 11:56:48 +00:00
int count;
Do refcounting of open devices (more) correctly. count_dev funtion by phk.
2000-02-07 23:05:40 +00:00
Do not refcount the cdevsw, but rather maintain a cdev->si_threadcount of the number of threads which are inside whatever is behind the cdevsw for this particular cdev. Make the device mutex visible through dev_lock() and dev_unlock(). We may want finer granularity later. Replace spechash_mtx use with dev_lock()/dev_unlock().
2004-09-23 07:17:41 +00:00
dev_lock();
Simplify count_dev()
2003-10-17 11:56:48 +00:00
count = dev->si_usecount;
Do not refcount the cdevsw, but rather maintain a cdev->si_threadcount of the number of threads which are inside whatever is behind the cdevsw for this particular cdev. Make the device mutex visible through dev_lock() and dev_unlock(). We may want finer granularity later. Replace spechash_mtx use with dev_lock()/dev_unlock().
2004-09-23 07:17:41 +00:00
dev_unlock();
Simplify count_dev()
2003-10-17 11:56:48 +00:00
return(count);
Do refcounting of open devices (more) correctly. count_dev funtion by phk.
2000-02-07 23:05:40 +00:00
}
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
/*
* Print out a description of a vnode.
*/
static char *typename[] =
Add marker vnodes to ensure that all vnodes associated with the mount point are iterated over when using MNT_VNODE_FOREACH. Reviewed by: truckman
2006-01-09 20:42:19 +00:00
{"VNON", "VREG", "VDIR", "VBLK", "VCHR", "VLNK", "VSOCK", "VFIFO", "VBAD",
"VMARKER"};
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
The big 4.4BSD Lite to FreeBSD 2.0.0 (Development) patch. Reviewed by: Rodney W. Grimes Submitted by: John Dyson and David Greenman
1994-05-25 09:21:21 +00:00
void
Change vprint() to vn_printf() which takes varargs. Add #define for vprint() to call vn_printf().
2005-01-24 13:58:08 +00:00
vn_printf(struct vnode *vp, const char *fmt, ...)
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
{
Change vprint() to vn_printf() which takes varargs. Add #define for vprint() to call vn_printf().
2005-01-24 13:58:08 +00:00
va_list ap;
Improve vn_printf() by: - adding missing vnode flags, - printing unknown flags as numbers, - using strlcat() instead of strcat(). Approved by: re (bmah)
2007-08-13 21:23:30 +00:00
char buf[256], buf2[16];
u_long flags;
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
Change vprint() to vn_printf() which takes varargs. Add #define for vprint() to call vn_printf().
2005-01-24 13:58:08 +00:00
va_start(ap, fmt);
vprintf(fmt, ap);
va_end(ap);
printf("%p: ", (void *)vp);
printf("tag %s, type %s\n", vp->v_tag, typename[vp->v_type]);
printf(" usecount %d, writecount %d, refcount %d mountedhere %p\n",
Improve vprint() a little bit: break long lines, reduce indent and tell if the VI_LOCK() is held.
2004-12-03 12:09:34 +00:00
vp->v_usecount, vp->v_writecount, vp->v_holdcnt, vp->v_mountedhere);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
buf[0] = '\0';
Change vprint() to vn_printf() which takes varargs. Add #define for vprint() to call vn_printf().
2005-01-24 13:58:08 +00:00
buf[1] = '\0';
- Replace v_flag with v_iflag and v_vflag - v_vflag is protected by the vnode lock and is used when synchronization with VOP calls is needed. - v_iflag is protected by interlock and is used for dealing with vnode management issues. These flags include X/O LOCK, FREE, DOOMED, etc. - All accesses to v_iflag and v_vflag have either been locked or marked with mp_fixme's. - Many ASSERT_VOP_LOCKED calls have been added where the locking was not clear. - Many functions in vfs_subr.c were restructured to provide for stronger locking. Idea stolen from: BSD/OS
2002-08-04 10:29:36 +00:00
if (vp->v_vflag & VV_ROOT)
Improve vn_printf() by: - adding missing vnode flags, - printing unknown flags as numbers, - using strlcat() instead of strcat(). Approved by: re (bmah)
2007-08-13 21:23:30 +00:00
strlcat(buf, "|VV_ROOT", sizeof(buf));
if (vp->v_vflag & VV_ISTTY)
strlcat(buf, "|VV_ISTTY", sizeof(buf));
if (vp->v_vflag & VV_NOSYNC)
strlcat(buf, "|VV_NOSYNC", sizeof(buf));
if (vp->v_vflag & VV_CACHEDLABEL)
strlcat(buf, "|VV_CACHEDLABEL", sizeof(buf));
- Replace v_flag with v_iflag and v_vflag - v_vflag is protected by the vnode lock and is used when synchronization with VOP calls is needed. - v_iflag is protected by interlock and is used for dealing with vnode management issues. These flags include X/O LOCK, FREE, DOOMED, etc. - All accesses to v_iflag and v_vflag have either been locked or marked with mp_fixme's. - Many ASSERT_VOP_LOCKED calls have been added where the locking was not clear. - Many functions in vfs_subr.c were restructured to provide for stronger locking. Idea stolen from: BSD/OS
2002-08-04 10:29:36 +00:00
if (vp->v_vflag & VV_TEXT)
Improve vn_printf() by: - adding missing vnode flags, - printing unknown flags as numbers, - using strlcat() instead of strcat(). Approved by: re (bmah)
2007-08-13 21:23:30 +00:00
strlcat(buf, "|VV_TEXT", sizeof(buf));
if (vp->v_vflag & VV_COPYONWRITE)
strlcat(buf, "|VV_COPYONWRITE", sizeof(buf));
- Replace v_flag with v_iflag and v_vflag - v_vflag is protected by the vnode lock and is used when synchronization with VOP calls is needed. - v_iflag is protected by interlock and is used for dealing with vnode management issues. These flags include X/O LOCK, FREE, DOOMED, etc. - All accesses to v_iflag and v_vflag have either been locked or marked with mp_fixme's. - Many ASSERT_VOP_LOCKED calls have been added where the locking was not clear. - Many functions in vfs_subr.c were restructured to provide for stronger locking. Idea stolen from: BSD/OS
2002-08-04 10:29:36 +00:00
if (vp->v_vflag & VV_SYSTEM)
Improve vn_printf() by: - adding missing vnode flags, - printing unknown flags as numbers, - using strlcat() instead of strcat(). Approved by: re (bmah)
2007-08-13 21:23:30 +00:00
strlcat(buf, "|VV_SYSTEM", sizeof(buf));
if (vp->v_vflag & VV_PROCDEP)
strlcat(buf, "|VV_PROCDEP", sizeof(buf));
if (vp->v_vflag & VV_NOKNOTE)
strlcat(buf, "|VV_NOKNOTE", sizeof(buf));
Add gjournal specific code to the UFS file system: - Add FS_GJOURNAL flag which enables gjournal support on a file system. - Add cg_unrefs field to the cylinder group structure which holds number of unreferenced (orphaned) inodes in the given cylinder group. - Add fs_unrefs field to the super block structure which holds total number of unreferenced (orphaned) inodes. - When file or a directory is orphaned (last reference is removed, but object is still open), increase fs_unrefs and cg_unrefs fields, which is a hint for fsck in which cylinder groups looks for such (orphaned) objects. - When file is last closed, decrease {fs,cg}_unrefs fields. - Add VV_DELETED vnode flag which points at orphaned objects. Sponsored by: home.pl
2006-10-31 21:48:54 +00:00
if (vp->v_vflag & VV_DELETED)
Improve vn_printf() by: - adding missing vnode flags, - printing unknown flags as numbers, - using strlcat() instead of strcat(). Approved by: re (bmah)
2007-08-13 21:23:30 +00:00
strlcat(buf, "|VV_DELETED", sizeof(buf));
if (vp->v_vflag & VV_MD)
strlcat(buf, "|VV_MD", sizeof(buf));
flags = vp->v_vflag & ~(VV_ROOT | VV_ISTTY | VV_NOSYNC |
VV_CACHEDLABEL | VV_TEXT | VV_COPYONWRITE | VV_SYSTEM | VV_PROCDEP |
VV_NOKNOTE | VV_DELETED | VV_MD);
if (flags != 0) {
snprintf(buf2, sizeof(buf2), "|VV(0x%lx)", flags);
strlcat(buf, buf2, sizeof(buf));
}
if (vp->v_iflag & VI_MOUNT)
strlcat(buf, "|VI_MOUNT", sizeof(buf));
if (vp->v_iflag & VI_AGE)
strlcat(buf, "|VI_AGE", sizeof(buf));
- Replace v_flag with v_iflag and v_vflag - v_vflag is protected by the vnode lock and is used when synchronization with VOP calls is needed. - v_iflag is protected by interlock and is used for dealing with vnode management issues. These flags include X/O LOCK, FREE, DOOMED, etc. - All accesses to v_iflag and v_vflag have either been locked or marked with mp_fixme's. - Many ASSERT_VOP_LOCKED calls have been added where the locking was not clear. - Many functions in vfs_subr.c were restructured to provide for stronger locking. Idea stolen from: BSD/OS
2002-08-04 10:29:36 +00:00
if (vp->v_iflag & VI_DOOMED)
Improve vn_printf() by: - adding missing vnode flags, - printing unknown flags as numbers, - using strlcat() instead of strcat(). Approved by: re (bmah)
2007-08-13 21:23:30 +00:00
strlcat(buf, "|VI_DOOMED", sizeof(buf));
- Replace v_flag with v_iflag and v_vflag - v_vflag is protected by the vnode lock and is used when synchronization with VOP calls is needed. - v_iflag is protected by interlock and is used for dealing with vnode management issues. These flags include X/O LOCK, FREE, DOOMED, etc. - All accesses to v_iflag and v_vflag have either been locked or marked with mp_fixme's. - Many ASSERT_VOP_LOCKED calls have been added where the locking was not clear. - Many functions in vfs_subr.c were restructured to provide for stronger locking. Idea stolen from: BSD/OS
2002-08-04 10:29:36 +00:00
if (vp->v_iflag & VI_FREE)
Improve vn_printf() by: - adding missing vnode flags, - printing unknown flags as numbers, - using strlcat() instead of strcat(). Approved by: re (bmah)
2007-08-13 21:23:30 +00:00
strlcat(buf, "|VI_FREE", sizeof(buf));
if (vp->v_iflag & VI_OBJDIRTY)
strlcat(buf, "|VI_OBJDIRTY", sizeof(buf));
if (vp->v_iflag & VI_DOINGINACT)
strlcat(buf, "|VI_DOINGINACT", sizeof(buf));
if (vp->v_iflag & VI_OWEINACT)
strlcat(buf, "|VI_OWEINACT", sizeof(buf));
flags = vp->v_iflag & ~(VI_MOUNT | VI_AGE | VI_DOOMED | VI_FREE |
VI_OBJDIRTY | VI_DOINGINACT | VI_OWEINACT);
if (flags != 0) {
snprintf(buf2, sizeof(buf2), "|VI(0x%lx)", flags);
strlcat(buf, buf2, sizeof(buf));
}
Change vprint() to vn_printf() which takes varargs. Add #define for vprint() to call vn_printf().
2005-01-24 13:58:08 +00:00
printf(" flags (%s)\n", buf + 1);
Improve vprint() a little bit: break long lines, reduce indent and tell if the VI_LOCK() is held.
2004-12-03 12:09:34 +00:00
if (mtx_owned(VI_MTX(vp)))
printf(" VI_LOCKed");
Remove extra ; at end of if(). Found by: bz
2005-03-27 07:52:12 +00:00
if (vp->v_object != NULL)
- Remove vx_lock, vx_unlock, vx_wait, etc. - Add a vn_start_write/vn_finished_write around vlrureclaim so we don't do writing ops without suspending. This could suspend the vlruproc which should not be a problem under normal circumstances. - Manually implement VMIGHTFREE in vlrureclaim as this was the only instance where it was used. - Acquire a lock before calling vgone() as it now requires it. - Move the acquisition of the vnode interlock from vtryrecycle() to getnewvnode() so that if it fails we don't drop and reacquire the vnode_free_list_mtx. - Check for a usecount or holdcount at the end of vtryrecycle() in case someone grabbed a ref while we were recycling. Abort the recycle, and on the final ref drop this vnode will be placed on the head of the free list. - Move the redundant VOP_INACTIVE protection code into the local vinactive() routine to avoid code bloat. - Keep the vnode lock held across calls to vgone() in several places. - vgonel() no longer uses XLOCK, instead callers must hold an exclusive vnode lock. The VI_DOOMED flag is set to allow other threads to detect a vnode which is no longer valid. This flag is set until the last reference is gone, and there are no chances for a new ref. vgonel() holds this lock across the entire function, which greatly simplifies logic. _ Only vfree() in one place in vgone() not three. - Adjust vget() to check the VI_DOOMED flag prior to waiting on the lock in the LK_NOWAIT case. In other cases, check after we have slept and acquired an exlusive lock. This will simulate the old vx_wait() behavior. Sponsored by: Isilon Systems, Inc.
2005-03-13 11:54:28 +00:00
printf(" v_object %p ref %d pages %d\n",
vp->v_object, vp->v_object->ref_count,
vp->v_object->resident_page_count);
Change vprint() to vn_printf() which takes varargs. Add #define for vprint() to call vn_printf().
2005-01-24 13:58:08 +00:00
printf(" ");
Regularize the vop_stdlock'ing protocol across all the filesystems that use it. Specifically, vop_stdlock uses the lock pointed to by vp->v_vnlock. By default, getnewvnode sets up vp->v_vnlock to reference vp->v_lock. Filesystems that wish to use the default do not need to allocate a lock at the front of their node structure (as some still did) or do a lockinit. They can simply start using vn_lock/VOP_UNLOCK. Filesystems that wish to manage their own locks, but still use the vop_stdlock functions (such as nullfs) can simply replace vp->v_vnlock with a pointer to the lock that they wish to have used for the vnode. Such filesystems are responsible for setting the vp->v_vnlock back to the default in their vop_reclaim routine (e.g., vp->v_vnlock = &vp->v_lock). In theory, this set of changes cleans up the existing filesystem lock interface and should have no function change to the existing locking scheme. Sponsored by: DARPA & NAI Labs.
2002-10-14 03:20:36 +00:00
lockmgr_printinfo(vp->v_vnlock);
Finish cleanup of vprint() which was begun with changing v_tag to a string. Remove extraneous uses of vop_null, instead defering to the default op. Rename vnode type "vfs" to the more descriptive "syncer". Fix formatting for various filesystems that use vop_print.
2003-03-03 19:15:40 +00:00
if (vp->v_data != NULL)
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
VOP_PRINT(vp);
}
Changed #ifdef around printlockedvnodes() from DEBUG to DDB.
1995-04-16 11:33:33 +00:00
#ifdef DDB
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
/*
* List all of the locked vnodes in the system.
* Called when debugging the kernel.
*/
Rename the db command lockedvnodes to lockedvnods so that it fits on the help screen and one doens't think we have a lockedvnodesmap command.
2002-06-29 04:45:09 +00:00
DB_SHOW_COMMAND(lockedvnods, lockedvnodes)
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
{
Merged some missing locking from Lite2: - getnewvnode() and vref() were missing one simple_unlock() each. - the Lite2 locking changes weren't merged at all in printlockedvnodes() or sysctl_vnode(). Merging these undid some KNF style regressions.
1997-02-25 19:33:23 +00:00
struct mount *mp, *nmp;
struct vnode *vp;
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
Remove bogus locking from DDB's "show lockedvnods" command: using synchronization primitives from inside DDB is generally a bad idea, and in this case it frequently results in panics due to DDB commands being executed from the sio fast interrupt context on a serial console. Replace the locking with a note that a lack of locking means that DDB may get see inconsistent views of the mount and vnode lists, which could also result in a panic. More frequently, though, this avoids a panic than causes it. Discussed with ages ago: bde Approved by: re (scottl)
2003-05-12 14:37:47 +00:00
/*
* Note: because this is DDB, we can't obey the locking semantics
* for these structures, which means we could catch an inconsistent
* state and dereference a nasty pointer. Not much to be done
* about that.
*/
In "show lockedvnods" DDB command, use db_printf() rather than printf() so that the results end up in the DDB output stream rather than the console output stream. This should likely also be done for the vprint() function it calls. MFC after: 3 months
2007-12-28 00:47:31 +00:00
db_printf("Locked vnodes\n");
struct mountlist and struct mount.mnt_list have no business being a CIRCLEQ. Change them to TAILQ_HEAD and TAILQ_ENTRY respectively. This removes ugly mp != (void*)&mountlist comparisons. Requested by: phk Submitted by: Jake Burkholder jake@checker.org PR: 14967
1999-11-20 10:00:46 +00:00
for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
Remove bogus locking from DDB's "show lockedvnods" command: using synchronization primitives from inside DDB is generally a bad idea, and in this case it frequently results in panics due to DDB commands being executed from the sio fast interrupt context on a serial console. Replace the locking with a note that a lack of locking means that DDB may get see inconsistent views of the mount and vnode lists, which could also result in a panic. More frequently, though, this avoids a panic than causes it. Discussed with ages ago: bde Approved by: re (scottl)
2003-05-12 14:37:47 +00:00
nmp = TAILQ_NEXT(mp, mnt_list);
Change the vnode list under the mount point from a LIST to a TAILQ in preparation for an implementation of limiting code for kern.maxvnodes. MFC after: 3 days
2001-10-23 01:21:29 +00:00
TAILQ_FOREACH(vp, &mp->mnt_nvnodelist, v_nmntvnodes) {
Conver all explicit instances to VOP_ISLOCKED(arg, NULL) into VOP_ISLOCKED(arg, curthread). Now, VOP_ISLOCKED() and lockstatus() should only acquire curthread as argument; this will lead in axing the additional argument from both functions, making the code cleaner. Reviewed by: jeff, kib
2008-02-08 21:45:47 +00:00
if (vp->v_type != VMARKER &&
Axe the 'thread' argument from VOP_ISLOCKED() and lockstatus() as it is always curthread. As KPI gets broken by this patch, manpages and __FreeBSD_version will be updated by further commits. Tested by: Andrea Barberio <insomniac at slackware dot it>
2008-02-25 18:45:57 +00:00
VOP_ISLOCKED(vp))
Don't pass NULL to vprint()
2005-02-10 08:55:08 +00:00
vprint("", vp);
Merged some missing locking from Lite2: - getnewvnode() and vref() were missing one simple_unlock() each. - the Lite2 locking changes weren't merged at all in printlockedvnodes() or sysctl_vnode(). Merging these undid some KNF style regressions.
1997-02-25 19:33:23 +00:00
}
struct mountlist and struct mount.mnt_list have no business being a CIRCLEQ. Change them to TAILQ_HEAD and TAILQ_ENTRY respectively. This removes ugly mp != (void*)&mountlist comparisons. Requested by: phk Submitted by: Jake Burkholder jake@checker.org PR: 14967
1999-11-20 10:00:46 +00:00
nmp = TAILQ_NEXT(mp, mnt_list);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
}
}
Add 'show vnode <addr>' DDB command.
2006-09-04 22:15:44 +00:00
/*
* Show details about the given vnode.
*/
DB_SHOW_COMMAND(vnode, db_show_vnode)
{
struct vnode *vp;
if (!have_addr)
return;
vp = (struct vnode *)addr;
vn_printf(vp, "vnode ");
}
Implement 'show mount' command in DDB. Without argument, it prints short info about all currently mounted file systems. When an address is given as an argument, prints detailed info about the given mount point. MFC after: 2 weeks
2008-04-26 13:04:48 +00:00
/*
* Show details about the given mount point.
*/
DB_SHOW_COMMAND(mount, db_show_mount)
{
struct mount *mp;
struct statfs *sp;
struct vnode *vp;
char buf[512];
u_int flags;
if (!have_addr) {
/* No address given, print short info about all mount points. */
TAILQ_FOREACH(mp, &mountlist, mnt_list) {
db_printf("%p %s on %s (%s)\n", mp,
mp->mnt_stat.f_mntfromname,
mp->mnt_stat.f_mntonname,
mp->mnt_stat.f_fstypename);
Be more friendly for DDB pager. Educated by: jhb's BSDCan presentation
2008-05-18 21:08:12 +00:00
if (db_pager_quit)
break;
Implement 'show mount' command in DDB. Without argument, it prints short info about all currently mounted file systems. When an address is given as an argument, prints detailed info about the given mount point. MFC after: 2 weeks
2008-04-26 13:04:48 +00:00
}
db_printf("\nMore info: show mount <addr>\n");
return;
}
mp = (struct mount *)addr;
db_printf("%p %s on %s (%s)\n", mp, mp->mnt_stat.f_mntfromname,
mp->mnt_stat.f_mntonname, mp->mnt_stat.f_fstypename);
buf[0] = '\0';
flags = mp->mnt_flag;
#define MNT_FLAG(flag) do { \
if (flags & (flag)) { \
if (buf[0] != '\0') \
strlcat(buf, ", ", sizeof(buf)); \
strlcat(buf, (#flag) + 4, sizeof(buf)); \
flags &= ~(flag); \
} \
} while (0)
MNT_FLAG(MNT_RDONLY);
MNT_FLAG(MNT_SYNCHRONOUS);
MNT_FLAG(MNT_NOEXEC);
MNT_FLAG(MNT_NOSUID);
MNT_FLAG(MNT_UNION);
MNT_FLAG(MNT_ASYNC);
MNT_FLAG(MNT_SUIDDIR);
MNT_FLAG(MNT_SOFTDEP);
MNT_FLAG(MNT_NOSYMFOLLOW);
MNT_FLAG(MNT_GJOURNAL);
MNT_FLAG(MNT_MULTILABEL);
MNT_FLAG(MNT_ACLS);
MNT_FLAG(MNT_NOATIME);
MNT_FLAG(MNT_NOCLUSTERR);
MNT_FLAG(MNT_NOCLUSTERW);
MNT_FLAG(MNT_EXRDONLY);
MNT_FLAG(MNT_EXPORTED);
MNT_FLAG(MNT_DEFEXPORTED);
MNT_FLAG(MNT_EXPORTANON);
MNT_FLAG(MNT_EXKERB);
MNT_FLAG(MNT_EXPUBLIC);
MNT_FLAG(MNT_LOCAL);
MNT_FLAG(MNT_QUOTA);
MNT_FLAG(MNT_ROOTFS);
MNT_FLAG(MNT_USER);
MNT_FLAG(MNT_IGNORE);
MNT_FLAG(MNT_UPDATE);
MNT_FLAG(MNT_DELEXPORT);
MNT_FLAG(MNT_RELOAD);
MNT_FLAG(MNT_FORCE);
MNT_FLAG(MNT_SNAPSHOT);
MNT_FLAG(MNT_BYFSID);
#undef MNT_FLAG
if (flags != 0) {
if (buf[0] != '\0')
strlcat(buf, ", ", sizeof(buf));
snprintf(buf + strlen(buf), sizeof(buf) - strlen(buf),
"0x%08x", flags);
}
db_printf(" mnt_flag = %s\n", buf);
buf[0] = '\0';
flags = mp->mnt_kern_flag;
#define MNT_KERN_FLAG(flag) do { \
if (flags & (flag)) { \
if (buf[0] != '\0') \
strlcat(buf, ", ", sizeof(buf)); \
strlcat(buf, (#flag) + 5, sizeof(buf)); \
flags &= ~(flag); \
} \
} while (0)
MNT_KERN_FLAG(MNTK_UNMOUNTF);
MNT_KERN_FLAG(MNTK_ASYNC);
MNT_KERN_FLAG(MNTK_SOFTDEP);
MNT_KERN_FLAG(MNTK_NOINSMNTQ);
MNT_KERN_FLAG(MNTK_UNMOUNT);
MNT_KERN_FLAG(MNTK_MWAIT);
MNT_KERN_FLAG(MNTK_SUSPEND);
MNT_KERN_FLAG(MNTK_SUSPEND2);
MNT_KERN_FLAG(MNTK_SUSPENDED);
MNT_KERN_FLAG(MNTK_MPSAFE);
MNT_KERN_FLAG(MNTK_NOKNOTE);
MNT_KERN_FLAG(MNTK_LOOKUP_SHARED);
#undef MNT_KERN_FLAG
if (flags != 0) {
if (buf[0] != '\0')
strlcat(buf, ", ", sizeof(buf));
snprintf(buf + strlen(buf), sizeof(buf) - strlen(buf),
"0x%08x", flags);
}
db_printf(" mnt_kern_flag = %s\n", buf);
sp = &mp->mnt_stat;
db_printf(" mnt_stat = { version=%u type=%u flags=0x%016jx "
"bsize=%ju iosize=%ju blocks=%ju bfree=%ju bavail=%jd files=%ju "
"ffree=%jd syncwrites=%ju asyncwrites=%ju syncreads=%ju "
"asyncreads=%ju namemax=%u owner=%u fsid=[%d, %d] }\n",
(u_int)sp->f_version, (u_int)sp->f_type, (uintmax_t)sp->f_flags,
(uintmax_t)sp->f_bsize, (uintmax_t)sp->f_iosize,
(uintmax_t)sp->f_blocks, (uintmax_t)sp->f_bfree,
(intmax_t)sp->f_bavail, (uintmax_t)sp->f_files,
(intmax_t)sp->f_ffree, (uintmax_t)sp->f_syncwrites,
(uintmax_t)sp->f_asyncwrites, (uintmax_t)sp->f_syncreads,
(uintmax_t)sp->f_asyncreads, (u_int)sp->f_namemax,
(u_int)sp->f_owner, (int)sp->f_fsid.val[0], (int)sp->f_fsid.val[1]);
db_printf(" mnt_cred = { uid=%u ruid=%u",
(u_int)mp->mnt_cred->cr_uid, (u_int)mp->mnt_cred->cr_ruid);
if (mp->mnt_cred->cr_prison != NULL)
db_printf(", jail=%d", mp->mnt_cred->cr_prison->pr_id);
db_printf(" }\n");
db_printf(" mnt_ref = %d\n", mp->mnt_ref);
db_printf(" mnt_gen = %d\n", mp->mnt_gen);
db_printf(" mnt_nvnodelistsize = %d\n", mp->mnt_nvnodelistsize);
db_printf(" mnt_writeopcount = %d\n", mp->mnt_writeopcount);
db_printf(" mnt_noasync = %u\n", mp->mnt_noasync);
db_printf(" mnt_maxsymlinklen = %d\n", mp->mnt_maxsymlinklen);
db_printf(" mnt_iosize_max = %d\n", mp->mnt_iosize_max);
db_printf(" mnt_hashseed = %u\n", mp->mnt_hashseed);
db_printf(" mnt_secondary_writes = %d\n", mp->mnt_secondary_writes);
db_printf(" mnt_secondary_accwrites = %d\n",
mp->mnt_secondary_accwrites);
db_printf(" mnt_gjprovider = %s\n",
mp->mnt_gjprovider != NULL ? mp->mnt_gjprovider : "NULL");
db_printf("\n");
TAILQ_FOREACH(vp, &mp->mnt_nvnodelist, v_nmntvnodes) {
Be more friendly for DDB pager. Educated by: jhb's BSDCan presentation
2008-05-18 21:08:12 +00:00
if (vp->v_type != VMARKER) {
Implement 'show mount' command in DDB. Without argument, it prints short info about all currently mounted file systems. When an address is given as an argument, prints detailed info about the given mount point. MFC after: 2 weeks
2008-04-26 13:04:48 +00:00
vn_printf(vp, "vnode ");
Be more friendly for DDB pager. Educated by: jhb's BSDCan presentation
2008-05-18 21:08:12 +00:00
if (db_pager_quit)
break;
}
Implement 'show mount' command in DDB. Without argument, it prints short info about all currently mounted file systems. When an address is given as an argument, prints detailed info about the given mount point. MFC after: 2 weeks
2008-04-26 13:04:48 +00:00
}
}
Add 'show vnode <addr>' DDB command.
2006-09-04 22:15:44 +00:00
#endif /* DDB */
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
- Introduce a new struct xvfsconf, the userland version of struct vfsconf. - Make getvfsbyname() take a struct xvfsconf *. - Convert several consumers of getvfsbyname() to use struct xvfsconf. - Correct the getvfsbyname.3 manpage. - Create a new vfs.conflist sysctl to dump all the struct xvfsconf in the kernel, and rewrite getvfsbyname() to use this instead of the weird existing API. - Convert some {set,get,end}vfsent() consumers to use the new vfs.conflist sysctl. - Convert a vfsload() call in nfsiod.c to kldload() and remove the useless vfsisloadable() and endvfsent() calls. - Add a warning printf() in vfs_sysctl() to tell people they are using an old userland. After these changes, it's possible to modify struct vfsconf without breaking the binary compatibility. Please note that these changes don't break this compatibility either. When bp will have updated mount_smbfs(8) with the patch I sent him, there will be no more consumers of the {set,get,end}vfsent(), vfsisloadable() and vfsload() API, and I will promptly delete it.
2002-08-10 20:19:04 +00:00
/*
* Fill in a struct xvfsconf based on a struct vfsconf.
*/
static void
vfsconf2x(struct vfsconf *vfsp, struct xvfsconf *xvfsp)
{
strcpy(xvfsp->vfc_name, vfsp->vfc_name);
xvfsp->vfc_typenum = vfsp->vfc_typenum;
xvfsp->vfc_refcount = vfsp->vfc_refcount;
xvfsp->vfc_flags = vfsp->vfc_flags;
/*
* These are unused in userland, we keep them
* to not break binary compatibility.
*/
xvfsp->vfc_vfsops = NULL;
xvfsp->vfc_next = NULL;
}
Put deprecated sysctl code inside BURN_BRIDGES.
2004-04-11 21:09:22 +00:00
/*
* Top level filesystem related information gathering.
*/
- Introduce a new struct xvfsconf, the userland version of struct vfsconf. - Make getvfsbyname() take a struct xvfsconf *. - Convert several consumers of getvfsbyname() to use struct xvfsconf. - Correct the getvfsbyname.3 manpage. - Create a new vfs.conflist sysctl to dump all the struct xvfsconf in the kernel, and rewrite getvfsbyname() to use this instead of the weird existing API. - Convert some {set,get,end}vfsent() consumers to use the new vfs.conflist sysctl. - Convert a vfsload() call in nfsiod.c to kldload() and remove the useless vfsisloadable() and endvfsent() calls. - Add a warning printf() in vfs_sysctl() to tell people they are using an old userland. After these changes, it's possible to modify struct vfsconf without breaking the binary compatibility. Please note that these changes don't break this compatibility either. When bp will have updated mount_smbfs(8) with the patch I sent him, there will be no more consumers of the {set,get,end}vfsent(), vfsisloadable() and vfsload() API, and I will promptly delete it.
2002-08-10 20:19:04 +00:00
static int
sysctl_vfs_conflist(SYSCTL_HANDLER_ARGS)
{
struct vfsconf *vfsp;
Convert the vfsconf list to a TAILQ. Introduce vfs_byname() function to find things on it. Staticize vfs_nmount() function under the name vfs_donmount(). Various cleanups.
2004-07-27 22:32:01 +00:00
struct xvfsconf xvfsp;
int error;
- Introduce a new struct xvfsconf, the userland version of struct vfsconf. - Make getvfsbyname() take a struct xvfsconf *. - Convert several consumers of getvfsbyname() to use struct xvfsconf. - Correct the getvfsbyname.3 manpage. - Create a new vfs.conflist sysctl to dump all the struct xvfsconf in the kernel, and rewrite getvfsbyname() to use this instead of the weird existing API. - Convert some {set,get,end}vfsent() consumers to use the new vfs.conflist sysctl. - Convert a vfsload() call in nfsiod.c to kldload() and remove the useless vfsisloadable() and endvfsent() calls. - Add a warning printf() in vfs_sysctl() to tell people they are using an old userland. After these changes, it's possible to modify struct vfsconf without breaking the binary compatibility. Please note that these changes don't break this compatibility either. When bp will have updated mount_smbfs(8) with the patch I sent him, there will be no more consumers of the {set,get,end}vfsent(), vfsisloadable() and vfsload() API, and I will promptly delete it.
2002-08-10 20:19:04 +00:00
Convert the vfsconf list to a TAILQ. Introduce vfs_byname() function to find things on it. Staticize vfs_nmount() function under the name vfs_donmount(). Various cleanups.
2004-07-27 22:32:01 +00:00
error = 0;
TAILQ_FOREACH(vfsp, &vfsconf, vfc_list) {
If we are going to 1. Copy a NULL-terminated string into a fixed-length buffer, and 2. copyout that buffer to userland, we really ought to 0. Zero the entire buffer first. Security: FreeBSD-SA-05:08.kmem
2005-05-06 02:50:00 +00:00
bzero(&xvfsp, sizeof(xvfsp));
Convert the vfsconf list to a TAILQ. Introduce vfs_byname() function to find things on it. Staticize vfs_nmount() function under the name vfs_donmount(). Various cleanups.
2004-07-27 22:32:01 +00:00
vfsconf2x(vfsp, &xvfsp);
error = SYSCTL_OUT(req, &xvfsp, sizeof xvfsp);
if (error)
break;
- Introduce a new struct xvfsconf, the userland version of struct vfsconf. - Make getvfsbyname() take a struct xvfsconf *. - Convert several consumers of getvfsbyname() to use struct xvfsconf. - Correct the getvfsbyname.3 manpage. - Create a new vfs.conflist sysctl to dump all the struct xvfsconf in the kernel, and rewrite getvfsbyname() to use this instead of the weird existing API. - Convert some {set,get,end}vfsent() consumers to use the new vfs.conflist sysctl. - Convert a vfsload() call in nfsiod.c to kldload() and remove the useless vfsisloadable() and endvfsent() calls. - Add a warning printf() in vfs_sysctl() to tell people they are using an old userland. After these changes, it's possible to modify struct vfsconf without breaking the binary compatibility. Please note that these changes don't break this compatibility either. When bp will have updated mount_smbfs(8) with the patch I sent him, there will be no more consumers of the {set,get,end}vfsent(), vfsisloadable() and vfsload() API, and I will promptly delete it.
2002-08-10 20:19:04 +00:00
}
return (error);
}
SYSCTL_PROC(_vfs, OID_AUTO, conflist, CTLFLAG_RD, NULL, 0, sysctl_vfs_conflist,
"S,xvfsconf", "List of all configured filesystems");
Put deprecated sysctl code inside BURN_BRIDGES.
2004-04-11 21:09:22 +00:00
#ifndef BURN_BRIDGES
Remove __P.
2002-03-19 21:25:46 +00:00
static int sysctl_ovfs_conf(SYSCTL_HANDLER_ARGS);
Merged Lite2's vfs_sysctl(). It doesn't fit very well into FreeBSD's (phk's) sysctl framework, and I needed special code to disambiguate the VFS_GENERIC node from the VFS_VFSCONF leaf, so I only converted the leaves to the FreeBSD framework. The error handling isn't quite right. CSRGS's sysctls seem to return ENOTDIR too much and FreeBSD's sysctls don't agree with the man page.
1997-03-03 12:58:20 +00:00
Attach vfs_sysctl() one level lower so that only the levels below VFS_GENERIC aren't done in the FreeBSD way. The previous commit broke the nfs sysctls.
1997-03-04 18:31:56 +00:00
static int
Previous commit changing SYSCTL_HANDLER_ARGS violated KNF. Pointed out by: bde
2000-07-04 11:25:35 +00:00
vfs_sysctl(SYSCTL_HANDLER_ARGS)
Moved vfs sysctls to where Lite2 put them. No code changes yet.
1997-03-02 11:06:22 +00:00
{
Attach vfs_sysctl() one level lower so that only the levels below VFS_GENERIC aren't done in the FreeBSD way. The previous commit broke the nfs sysctls.
1997-03-04 18:31:56 +00:00
int *name = (int *)arg1 - 1; /* XXX */
u_int namelen = arg2 + 1; /* XXX */
Moved vfs sysctls to where Lite2 put them. No code changes yet.
1997-03-02 11:06:22 +00:00
struct vfsconf *vfsp;
- Introduce a new struct xvfsconf, the userland version of struct vfsconf. - Make getvfsbyname() take a struct xvfsconf *. - Convert several consumers of getvfsbyname() to use struct xvfsconf. - Correct the getvfsbyname.3 manpage. - Create a new vfs.conflist sysctl to dump all the struct xvfsconf in the kernel, and rewrite getvfsbyname() to use this instead of the weird existing API. - Convert some {set,get,end}vfsent() consumers to use the new vfs.conflist sysctl. - Convert a vfsload() call in nfsiod.c to kldload() and remove the useless vfsisloadable() and endvfsent() calls. - Add a warning printf() in vfs_sysctl() to tell people they are using an old userland. After these changes, it's possible to modify struct vfsconf without breaking the binary compatibility. Please note that these changes don't break this compatibility either. When bp will have updated mount_smbfs(8) with the patch I sent him, there will be no more consumers of the {set,get,end}vfsent(), vfsisloadable() and vfsload() API, and I will promptly delete it.
2002-08-10 20:19:04 +00:00
struct xvfsconf xvfsp;
printf("WARNING: userland calling deprecated sysctl, "
"please rebuild world\n");
Moved vfs sysctls to where Lite2 put them. No code changes yet.
1997-03-02 11:06:22 +00:00
Oops, the previous revision unconfigured too much pre-Lite2 compatibilty cruft. At least lsvfs(1) was broken.
1998-08-29 13:13:10 +00:00
#if 1 || defined(COMPAT_PRELITE2)
Merged Lite2's vfs_sysctl(). It doesn't fit very well into FreeBSD's (phk's) sysctl framework, and I needed special code to disambiguate the VFS_GENERIC node from the VFS_VFSCONF leaf, so I only converted the leaves to the FreeBSD framework. The error handling isn't quite right. CSRGS's sysctls seem to return ENOTDIR too much and FreeBSD's sysctls don't agree with the man page.
1997-03-03 12:58:20 +00:00
/* Resolve ambiguity between VFS_VFSCONF and VFS_GENERIC. */
Attach vfs_sysctl() one level lower so that only the levels below VFS_GENERIC aren't done in the FreeBSD way. The previous commit broke the nfs sysctls.
1997-03-04 18:31:56 +00:00
if (namelen == 1)
Merged Lite2's vfs_sysctl(). It doesn't fit very well into FreeBSD's (phk's) sysctl framework, and I needed special code to disambiguate the VFS_GENERIC node from the VFS_VFSCONF leaf, so I only converted the leaves to the FreeBSD framework. The error handling isn't quite right. CSRGS's sysctls seem to return ENOTDIR too much and FreeBSD's sysctls don't agree with the man page.
1997-03-03 12:58:20 +00:00
return (sysctl_ovfs_conf(oidp, arg1, arg2, req));
#endif
Moved vfs sysctls to where Lite2 put them. No code changes yet.
1997-03-02 11:06:22 +00:00
Merged Lite2's vfs_sysctl(). It doesn't fit very well into FreeBSD's (phk's) sysctl framework, and I needed special code to disambiguate the VFS_GENERIC node from the VFS_VFSCONF leaf, so I only converted the leaves to the FreeBSD framework. The error handling isn't quite right. CSRGS's sysctls seem to return ENOTDIR too much and FreeBSD's sysctls don't agree with the man page.
1997-03-03 12:58:20 +00:00
switch (name[1]) {
case VFS_MAXTYPENUM:
if (namelen != 2)
return (ENOTDIR);
return (SYSCTL_OUT(req, &maxvfsconf, sizeof(int)));
case VFS_CONF:
if (namelen != 3)
return (ENOTDIR); /* overloaded */
Convert the vfsconf list to a TAILQ. Introduce vfs_byname() function to find things on it. Staticize vfs_nmount() function under the name vfs_donmount(). Various cleanups.
2004-07-27 22:32:01 +00:00
TAILQ_FOREACH(vfsp, &vfsconf, vfc_list)
Merged Lite2's vfs_sysctl(). It doesn't fit very well into FreeBSD's (phk's) sysctl framework, and I needed special code to disambiguate the VFS_GENERIC node from the VFS_VFSCONF leaf, so I only converted the leaves to the FreeBSD framework. The error handling isn't quite right. CSRGS's sysctls seem to return ENOTDIR too much and FreeBSD's sysctls don't agree with the man page.
1997-03-03 12:58:20 +00:00
if (vfsp->vfc_typenum == name[2])
break;
if (vfsp == NULL)
return (EOPNOTSUPP);
If we are going to 1. Copy a NULL-terminated string into a fixed-length buffer, and 2. copyout that buffer to userland, we really ought to 0. Zero the entire buffer first. Security: FreeBSD-SA-05:08.kmem
2005-05-06 02:50:00 +00:00
bzero(&xvfsp, sizeof(xvfsp));
- Introduce a new struct xvfsconf, the userland version of struct vfsconf. - Make getvfsbyname() take a struct xvfsconf *. - Convert several consumers of getvfsbyname() to use struct xvfsconf. - Correct the getvfsbyname.3 manpage. - Create a new vfs.conflist sysctl to dump all the struct xvfsconf in the kernel, and rewrite getvfsbyname() to use this instead of the weird existing API. - Convert some {set,get,end}vfsent() consumers to use the new vfs.conflist sysctl. - Convert a vfsload() call in nfsiod.c to kldload() and remove the useless vfsisloadable() and endvfsent() calls. - Add a warning printf() in vfs_sysctl() to tell people they are using an old userland. After these changes, it's possible to modify struct vfsconf without breaking the binary compatibility. Please note that these changes don't break this compatibility either. When bp will have updated mount_smbfs(8) with the patch I sent him, there will be no more consumers of the {set,get,end}vfsent(), vfsisloadable() and vfsload() API, and I will promptly delete it.
2002-08-10 20:19:04 +00:00
vfsconf2x(vfsp, &xvfsp);
return (SYSCTL_OUT(req, &xvfsp, sizeof(xvfsp)));
Merged Lite2's vfs_sysctl(). It doesn't fit very well into FreeBSD's (phk's) sysctl framework, and I needed special code to disambiguate the VFS_GENERIC node from the VFS_VFSCONF leaf, so I only converted the leaves to the FreeBSD framework. The error handling isn't quite right. CSRGS's sysctls seem to return ENOTDIR too much and FreeBSD's sysctls don't agree with the man page.
1997-03-03 12:58:20 +00:00
}
return (EOPNOTSUPP);
}
Moved vfs sysctls to where Lite2 put them. No code changes yet.
1997-03-02 11:06:22 +00:00
Make various vnode related functions static
2005-02-10 12:28:58 +00:00
static SYSCTL_NODE(_vfs, VFS_GENERIC, generic, CTLFLAG_RD | CTLFLAG_SKIP,
vfs_sysctl, "Generic filesystem");
Attach vfs_sysctl() one level lower so that only the levels below VFS_GENERIC aren't done in the FreeBSD way. The previous commit broke the nfs sysctls.
1997-03-04 18:31:56 +00:00
Oops, the previous revision unconfigured too much pre-Lite2 compatibilty cruft. At least lsvfs(1) was broken.
1998-08-29 13:13:10 +00:00
#if 1 || defined(COMPAT_PRELITE2)
Moved vfs sysctls to where Lite2 put them. No code changes yet.
1997-03-02 11:06:22 +00:00
static int
Previous commit changing SYSCTL_HANDLER_ARGS violated KNF. Pointed out by: bde
2000-07-04 11:25:35 +00:00
sysctl_ovfs_conf(SYSCTL_HANDLER_ARGS)
Moved vfs sysctls to where Lite2 put them. No code changes yet.
1997-03-02 11:06:22 +00:00
{
int error;
struct vfsconf *vfsp;
Merged Lite2's vfs_sysctl(). It doesn't fit very well into FreeBSD's (phk's) sysctl framework, and I needed special code to disambiguate the VFS_GENERIC node from the VFS_VFSCONF leaf, so I only converted the leaves to the FreeBSD framework. The error handling isn't quite right. CSRGS's sysctls seem to return ENOTDIR too much and FreeBSD's sysctls don't agree with the man page.
1997-03-03 12:58:20 +00:00
struct ovfsconf ovfs;
Moved vfs sysctls to where Lite2 put them. No code changes yet.
1997-03-02 11:06:22 +00:00
Convert the vfsconf list to a TAILQ. Introduce vfs_byname() function to find things on it. Staticize vfs_nmount() function under the name vfs_donmount(). Various cleanups.
2004-07-27 22:32:01 +00:00
TAILQ_FOREACH(vfsp, &vfsconf, vfc_list) {
If we are going to 1. Copy a NULL-terminated string into a fixed-length buffer, and 2. copyout that buffer to userland, we really ought to 0. Zero the entire buffer first. Security: FreeBSD-SA-05:08.kmem
2005-05-06 02:50:00 +00:00
bzero(&ovfs, sizeof(ovfs));
Moved vfs sysctls to where Lite2 put them. No code changes yet.
1997-03-02 11:06:22 +00:00
ovfs.vfc_vfsops = vfsp->vfc_vfsops; /* XXX used as flag */
strcpy(ovfs.vfc_name, vfsp->vfc_name);
ovfs.vfc_index = vfsp->vfc_typenum;
ovfs.vfc_refcount = vfsp->vfc_refcount;
ovfs.vfc_flags = vfsp->vfc_flags;
error = SYSCTL_OUT(req, &ovfs, sizeof ovfs);
if (error)
return error;
}
return 0;
}
Oops, the previous revision unconfigured too much pre-Lite2 compatibilty cruft. At least lsvfs(1) was broken.
1998-08-29 13:13:10 +00:00
#endif /* 1 || COMPAT_PRELITE2 */
Put deprecated sysctl code inside BURN_BRIDGES.
2004-04-11 21:09:22 +00:00
#endif /* !BURN_BRIDGES */
Moved vfs sysctls to where Lite2 put them. No code changes yet.
1997-03-02 11:06:22 +00:00
Initialize v_cachedid to -1 in getnewvnode(). Reintroduce the kern.vnode sysctl and make it export xvnodes rather than vnodes. Sponsored by: DARPA, NAI Labs
2002-07-31 12:24:35 +00:00
#define KINFO_VNODESLOP 10
Bracket the kern.vnode sysctl in #ifdef notyet because it results in massive locking issues on diskless systems. It is also not clear that this sysctl is non-dangerous in its requirements for locked down memory on large RAM systems.
2003-02-23 18:09:05 +00:00
#ifdef notyet
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
/*
* Dump vnode list (via sysctl).
*/
/* ARGSUSED */
Mega commit for sysctl. Convert the remaining sysctl stuff to the new way of doing things. the devconf stuff is the reason for the large number of files. Cleaned up some compiler warnings while I were there.
1995-11-20 12:42:39 +00:00
static int
Previous commit changing SYSCTL_HANDLER_ARGS violated KNF. Pointed out by: bde
2000-07-04 11:25:35 +00:00
sysctl_vnode(SYSCTL_HANDLER_ARGS)
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
{
Initialize v_cachedid to -1 in getnewvnode(). Reintroduce the kern.vnode sysctl and make it export xvnodes rather than vnodes. Sponsored by: DARPA, NAI Labs
2002-07-31 12:24:35 +00:00
struct xvnode *xvn;
struct mount *mp;
struct vnode *vp;
int error, len, n;
Mega commit for sysctl. Convert the remaining sysctl stuff to the new way of doing things. the devconf stuff is the reason for the large number of files. Cleaned up some compiler warnings while I were there.
1995-11-20 12:42:39 +00:00
- Extend the vnode_free_list_mtx to cover numvnodes and freevnodes. This was done only some of the time before, and now it is uniformly applied.
2002-08-13 05:29:48 +00:00
/*
* Stale numvnodes access is not fatal here.
*/
Mega commit for sysctl. Convert the remaining sysctl stuff to the new way of doing things. the devconf stuff is the reason for the large number of files. Cleaned up some compiler warnings while I were there.
1995-11-20 12:42:39 +00:00
req->lock = 0;
Initialize v_cachedid to -1 in getnewvnode(). Reintroduce the kern.vnode sysctl and make it export xvnodes rather than vnodes. Sponsored by: DARPA, NAI Labs
2002-07-31 12:24:35 +00:00
len = (numvnodes + KINFO_VNODESLOP) * sizeof *xvn;
if (!req->oldptr)
/* Make an estimate */
return (SYSCTL_OUT(req, 0, len));
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
Split the mlock() kernel code into two parts, mlock(), which unpacks the syscall arguments and does the suser() permission check, and kern_mlock(), which does the resource limit checking and calls vm_map_wire(). Split munlock() in a similar way. Enable the RLIMIT_MEMLOCK checking code in kern_mlock(). Replace calls to vslock() and vsunlock() in the sysctl code with calls to kern_mlock() and kern_munlock() so that the sysctl code will obey the wired memory limits. Nuke the vslock() and vsunlock() implementations, which are no longer used. Add a member to struct sysctl_req to track the amount of memory that is wired to handle the request. Modify sysctl_wire_old_buffer() to return an error if its call to kern_mlock() fails. Only wire the minimum of the length specified in the sysctl request and the length specified in its argument list. It is recommended that sysctl handlers that use sysctl_wire_old_buffer() should specify reasonable estimates for the amount of data they want to return so that only the minimum amount of memory is wired no matter what length has been specified by the request. Modify the callers of sysctl_wire_old_buffer() to look for the error return. Modify sysctl_old_user to obey the wired buffer length and clean up its implementation. Reviewed by: bms
2004-02-26 00:27:04 +00:00
error = sysctl_wire_old_buffer(req, 0);
if (error != 0)
return (error);
Back out M_* changes, per decision of the TRB. Approved by: trb
2003-02-19 05:47:46 +00:00
xvn = malloc(len, M_TEMP, M_ZERO | M_WAITOK);
Initialize v_cachedid to -1 in getnewvnode(). Reintroduce the kern.vnode sysctl and make it export xvnodes rather than vnodes. Sponsored by: DARPA, NAI Labs
2002-07-31 12:24:35 +00:00
n = 0;
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_lock(&mountlist_mtx);
Initialize v_cachedid to -1 in getnewvnode(). Reintroduce the kern.vnode sysctl and make it export xvnodes rather than vnodes. Sponsored by: DARPA, NAI Labs
2002-07-31 12:24:35 +00:00
TAILQ_FOREACH(mp, &mountlist, mnt_list) {
Improve VFS locking: - Implement real draining for vfs consumers by not relying on the mnt_lock and using instead a refcount in order to keep track of lock requesters. - Due to the change above, remove the mnt_lock lockmgr because it is now useless. - Due to the change above, vfs_busy() is no more linked to a lockmgr. Change so its KPI by removing the interlock argument and defining 2 new flags for it: MBF_NOWAIT which basically replaces the LK_NOWAIT of the old version (which was unlinked from the lockmgr alredy) and MBF_MNTLSTLOCK which provides the ability to drop the mountlist_mtx once the mnt interlock is held (ability still desired by most consumers). - The stub used into vfs_mount_destroy(), that allows to override the mnt_ref if running for more than 3 seconds, make it totally useless. Remove it as it was thought to work into older versions. If a problem of "refcount held never going away" should appear, we will need to fix properly instead than trust on such hackish solution. - Fix a bug where returning (with an error) from dounmount() was still leaving the MNTK_MWAIT flag on even if it the waiters were actually woken up. Just a place in vfs_mount_destroy() is left because it is going to recycle the structure in any case, so it doesn't matter. - Remove the markercnt refcount as it is useless. This patch modifies VFS ABI and breaks KPI for vfs_busy() so manpages and __FreeBSD_version will be modified accordingly. Discussed with: kib Tested by: pho
2008-11-02 10:15:42 +00:00
if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK))
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
continue;
Remove mntvnode_mtx and replace it with per-mountpoint mutex. Introduce two new macros MNT_ILOCK(mp)/MNT_IUNLOCK(mp) to operate on this mutex transparently. Eventually new mutex will be protecting more fields in struct mount, not only vnode list. Discussed with: jeff
2003-11-05 04:30:08 +00:00
MNT_ILOCK(mp);
Initialize v_cachedid to -1 in getnewvnode(). Reintroduce the kern.vnode sysctl and make it export xvnodes rather than vnodes. Sponsored by: DARPA, NAI Labs
2002-07-31 12:24:35 +00:00
TAILQ_FOREACH(vp, &mp->mnt_nvnodelist, v_nmntvnodes) {
if (n == len)
break;
vref(vp);
xvn[n].xv_size = sizeof *xvn;
xvn[n].xv_vnode = vp;
Eliminate v_id and v_ddid. The name cache now holds references to vnodes whose names it caches, so we no longer need a `generation number' to tell us if a referenced vnode is invalid. Replace the use of the parent's v_id in the hash function with the address of the parent vnode. Tested by: Peter Holm Glanced at by: jeff, phk
2005-03-30 03:01:36 +00:00
xvn[n].xv_id = 0; /* XXX compat */
Initialize v_cachedid to -1 in getnewvnode(). Reintroduce the kern.vnode sysctl and make it export xvnodes rather than vnodes. Sponsored by: DARPA, NAI Labs
2002-07-31 12:24:35 +00:00
#define XV_COPY(field) xvn[n].xv_##field = vp->v_##field
XV_COPY(usecount);
XV_COPY(writecount);
XV_COPY(holdcnt);
XV_COPY(mount);
XV_COPY(numoutput);
XV_COPY(type);
#undef XV_COPY
- Replace v_flag with v_iflag and v_vflag - v_vflag is protected by the vnode lock and is used when synchronization with VOP calls is needed. - v_iflag is protected by interlock and is used for dealing with vnode management issues. These flags include X/O LOCK, FREE, DOOMED, etc. - All accesses to v_iflag and v_vflag have either been locked or marked with mp_fixme's. - Many ASSERT_VOP_LOCKED calls have been added where the locking was not clear. - Many functions in vfs_subr.c were restructured to provide for stronger locking. Idea stolen from: BSD/OS
2002-08-04 10:29:36 +00:00
xvn[n].xv_flag = vp->v_vflag;
Initialize v_cachedid to -1 in getnewvnode(). Reintroduce the kern.vnode sysctl and make it export xvnodes rather than vnodes. Sponsored by: DARPA, NAI Labs
2002-07-31 12:24:35 +00:00
switch (vp->v_type) {
case VREG:
case VDIR:
case VLNK:
break;
case VBLK:
case VCHR:
if (vp->v_rdev == NULL) {
vrele(vp);
continue;
}
xvn[n].xv_dev = dev2udev(vp->v_rdev);
break;
case VSOCK:
xvn[n].xv_socket = vp->v_socket;
break;
case VFIFO:
xvn[n].xv_fifo = vp->v_fifoinfo;
break;
case VNON:
case VBAD:
default:
/* shouldn't happen? */
vrele(vp);
continue;
}
vrele(vp);
++n;
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
}
Remove mntvnode_mtx and replace it with per-mountpoint mutex. Introduce two new macros MNT_ILOCK(mp)/MNT_IUNLOCK(mp) to operate on this mutex transparently. Eventually new mutex will be protecting more fields in struct mount, not only vnode list. Discussed with: jeff
2003-11-05 04:30:08 +00:00
MNT_IUNLOCK(mp);
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_lock(&mountlist_mtx);
Decontextualize vfs_busy(), vfs_unbusy() and vfs_mount_alloc() functions. Manpages are updated accordingly. Tested by: Diego Sardina <siarodx at gmail dot com>
2008-08-31 14:26:08 +00:00
vfs_unbusy(mp);
Initialize v_cachedid to -1 in getnewvnode(). Reintroduce the kern.vnode sysctl and make it export xvnodes rather than vnodes. Sponsored by: DARPA, NAI Labs
2002-07-31 12:24:35 +00:00
if (n == len)
break;
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
}
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_unlock(&mountlist_mtx);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
Initialize v_cachedid to -1 in getnewvnode(). Reintroduce the kern.vnode sysctl and make it export xvnodes rather than vnodes. Sponsored by: DARPA, NAI Labs
2002-07-31 12:24:35 +00:00
error = SYSCTL_OUT(req, xvn, n * sizeof *xvn);
free(xvn, M_TEMP);
return (error);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
}
A couple of minor tweaks to the sysctl stuff.
1995-12-06 13:27:39 +00:00
SYSCTL_PROC(_kern, KERN_VNODE, vnode, CTLTYPE_OPAQUE|CTLFLAG_RD,
Nit in previous commit: the correct sysctl type is "S,xvnode"
2002-07-31 12:25:28 +00:00
0, 0, sysctl_vnode, "S,xvnode", "");
Bracket the kern.vnode sysctl in #ifdef notyet because it results in massive locking issues on diskless systems. It is also not clear that this sysctl is non-dangerous in its requirements for locked down memory on large RAM systems.
2003-02-23 18:09:05 +00:00
#endif
Mega commit for sysctl. Convert the remaining sysctl stuff to the new way of doing things. the devconf stuff is the reason for the large number of files. Cleaned up some compiler warnings while I were there.
1995-11-20 12:42:39 +00:00
This is the kernel Lite/2 commit. There are some requisite userland changes, so don't expect to be able to run the kernel as-is (very well) without the appropriate Lite/2 userland changes. The system boots and can mount UFS filesystems. Untested: ext2fs, msdosfs, NFS Known problems: Incorrect Berkeley ID strings in some files. Mount_std mounts will not work until the getfsent library routine is changed. Reviewed by: various people Submitted by: Jeffery Hsu <hsu@freebsd.org>
1997-02-10 02:22:35 +00:00
/*
Fixed unmounting of the root fs. vfs_unmountroot() wasn't fully updated to do Lite2 locking and vfs_unmountall() wasn't as simple as the Lite2 version.
1997-02-26 15:35:42 +00:00
* Unmount all filesystems. The list is traversed in reverse order
* of mounting to avoid dependencies.
This is the kernel Lite/2 commit. There are some requisite userland changes, so don't expect to be able to run the kernel as-is (very well) without the appropriate Lite/2 userland changes. The system boots and can mount UFS filesystems. Untested: ext2fs, msdosfs, NFS Known problems: Incorrect Berkeley ID strings in some files. Mount_std mounts will not work until the getfsent library routine is changed. Reviewed by: various people Submitted by: Jeffery Hsu <hsu@freebsd.org>
1997-02-10 02:22:35 +00:00
*/
void
Convert remaining functions in vfs_subr.c from K&R prototypes to ANSI C prototypes, as the majority of new functions added have been in this style. Changing prototype style now results in gcc noticing that the implementation of vn_pollrecord() has a 'short' argument instead of 'int' as prototyped in vnode.h, so correct that definition. In practice this didn't matter as only poll flags in the lower 16 bits are used. MFC after: 1 week
2006-01-21 19:42:10 +00:00
vfs_unmountall(void)
This is the kernel Lite/2 commit. There are some requisite userland changes, so don't expect to be able to run the kernel as-is (very well) without the appropriate Lite/2 userland changes. The system boots and can mount UFS filesystems. Untested: ext2fs, msdosfs, NFS Known problems: Incorrect Berkeley ID strings in some files. Mount_std mounts will not work until the getfsent library routine is changed. Reviewed by: various people Submitted by: Jeffery Hsu <hsu@freebsd.org>
1997-02-10 02:22:35 +00:00
{
struct mountlist and struct mount.mnt_list have no business being a CIRCLEQ. Change them to TAILQ_HEAD and TAILQ_ENTRY respectively. This removes ugly mp != (void*)&mountlist comparisons. Requested by: phk Submitted by: Jake Burkholder jake@checker.org PR: 14967
1999-11-20 10:00:46 +00:00
struct mount *mp;
KSE Milestone 2 Note ALL MODULES MUST BE RECOMPILED make the kernel aware that there are smaller units of scheduling than the process. (but only allow one thread per process at this time). This is functionally equivalent to teh previousl -current except that there is a thread associated with each process. Sorry john! (your next MFC will be a doosie!) Reviewed by: peter@freebsd.org, dillon@freebsd.org X-MFC after: ha ha ha ha
2001-09-12 08:38:13 +00:00
struct thread *td;
This is the kernel Lite/2 commit. There are some requisite userland changes, so don't expect to be able to run the kernel as-is (very well) without the appropriate Lite/2 userland changes. The system boots and can mount UFS filesystems. Untested: ext2fs, msdosfs, NFS Known problems: Incorrect Berkeley ID strings in some files. Mount_std mounts will not work until the getfsent library routine is changed. Reviewed by: various people Submitted by: Jeffery Hsu <hsu@freebsd.org>
1997-02-10 02:22:35 +00:00
int error;
Since it is not possible for curthread to be NULL in this context, drop the check+initialization for a straight initialization. Also assert that curthread will never be NULL just to be sure. Discussed with: rwatson, peter MFC after: 1 week
2005-05-02 02:07:55 +00:00
KASSERT(curthread != NULL, ("vfs_unmountall: NULL curthread"));
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
CTR1(KTR_VFS, "%s: unmounting all filesystems", __func__);
Since it is not possible for curthread to be NULL in this context, drop the check+initialization for a straight initialization. Also assert that curthread will never be NULL just to be sure. Discussed with: rwatson, peter MFC after: 1 week
2005-05-02 02:07:55 +00:00
td = curthread;
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
Fixed unmounting of the root fs. vfs_unmountroot() wasn't fully updated to do Lite2 locking and vfs_unmountall() wasn't as simple as the Lite2 version.
1997-02-26 15:35:42 +00:00
/*
* Since this only runs when rebooting, it is not interlocked.
*/
struct mountlist and struct mount.mnt_list have no business being a CIRCLEQ. Change them to TAILQ_HEAD and TAILQ_ENTRY respectively. This removes ugly mp != (void*)&mountlist comparisons. Requested by: phk Submitted by: Jake Burkholder jake@checker.org PR: 14967
1999-11-20 10:00:46 +00:00
while(!TAILQ_EMPTY(&mountlist)) {
mp = TAILQ_LAST(&mountlist, mntlist);
KSE Milestone 2 Note ALL MODULES MUST BE RECOMPILED make the kernel aware that there are smaller units of scheduling than the process. (but only allow one thread per process at this time). This is functionally equivalent to teh previousl -current except that there is a thread associated with each process. Sorry john! (your next MFC will be a doosie!) Reviewed by: peter@freebsd.org, dillon@freebsd.org X-MFC after: ha ha ha ha
2001-09-12 08:38:13 +00:00
error = dounmount(mp, MNT_FORCE, td);
This is the kernel Lite/2 commit. There are some requisite userland changes, so don't expect to be able to run the kernel as-is (very well) without the appropriate Lite/2 userland changes. The system boots and can mount UFS filesystems. Untested: ext2fs, msdosfs, NFS Known problems: Incorrect Berkeley ID strings in some files. Mount_std mounts will not work until the getfsent library routine is changed. Reviewed by: various people Submitted by: Jeffery Hsu <hsu@freebsd.org>
1997-02-10 02:22:35 +00:00
if (error) {
struct mountlist and struct mount.mnt_list have no business being a CIRCLEQ. Change them to TAILQ_HEAD and TAILQ_ENTRY respectively. This removes ugly mp != (void*)&mountlist comparisons. Requested by: phk Submitted by: Jake Burkholder jake@checker.org PR: 14967
1999-11-20 10:00:46 +00:00
TAILQ_REMOVE(&mountlist, mp, mnt_list);
Silence "busy" warnings when unmounting devfs at system shutdown. This is a workaround for non-symetric teardown of the file systems at shutdown with respect to the mount order at boot. The proper long term fix is to properly detach devfs from the root mount before unmounting each, and should be implemented, but since the problem is non-harmful, this temporary band-aid will prevent false positive bug reports and unnecessary error output for 6.0-RELEASE. MFC after: 3 days Tested by: pav, pjd
2005-08-20 17:12:47 +00:00
/*
* XXX: Due to the way in which we mount the root
* file system off of devfs, devfs will generate a
* "busy" warning when we try to unmount it before
* the root. Don't print a warning as a result in
* order to avoid false positive errors that may
* cause needless upset.
*/
if (strcmp(mp->mnt_vfc->vfc_name, "devfs") != 0) {
printf("unmount of %s failed (",
mp->mnt_stat.f_mntonname);
if (error == EBUSY)
printf("BUSY)\n");
else
printf("%d)\n", error);
}
struct mountlist and struct mount.mnt_list have no business being a CIRCLEQ. Change them to TAILQ_HEAD and TAILQ_ENTRY respectively. This removes ugly mp != (void*)&mountlist comparisons. Requested by: phk Submitted by: Jake Burkholder jake@checker.org PR: 14967
1999-11-20 10:00:46 +00:00
} else {
/* The unmount has removed mp from the mountlist */
This is the kernel Lite/2 commit. There are some requisite userland changes, so don't expect to be able to run the kernel as-is (very well) without the appropriate Lite/2 userland changes. The system boots and can mount UFS filesystems. Untested: ext2fs, msdosfs, NFS Known problems: Incorrect Berkeley ID strings in some files. Mount_std mounts will not work until the getfsent library routine is changed. Reviewed by: various people Submitted by: Jeffery Hsu <hsu@freebsd.org>
1997-02-10 02:22:35 +00:00
}
}
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
}
Changes to fix the following bugs: 1) Files weren't properly synced on filesystems other than UFS. In some cases, this lead to lost data. Most likely would be noticed on NFS. The fix is to make the VM page sync/object_clean general rather than in each filesystem. 2) Mixing regular and mmaped file I/O on NFS was very broken. It caused chunks of files to end up as zeroes rather than the intended contents. The fix was to fix several race conditions and to kludge up the "b_dirtyoff" and "b_dirtyend" that NFS relies upon - paying attention to page modifications that occurred via the mmapping. Reviewed by: David Greenman Submitted by: John Dyson
1995-05-21 21:39:31 +00:00
/*
* perform msync on all vnodes under a mount point
* the mount point must be locked.
*/
void
Change the vnode list under the mount point from a LIST to a TAILQ in preparation for an implementation of limiting code for kern.maxvnodes. MFC after: 3 days
2001-10-23 01:21:29 +00:00
vfs_msync(struct mount *mp, int flags)
{
Add marker vnodes to ensure that all vnodes associated with the mount point are iterated over when using MNT_VNODE_FOREACH. Reviewed by: truckman
2006-01-09 20:42:19 +00:00
struct vnode *vp, *mvp;
In vfs_msync(), test to see if the vnode being examined is "interesting" (ie: it has a vm_object attached and is marked as OBJ_MIGHTBEDIRTY) before attempting to lock it. This should reduce the cpu hit that is incurred when doing a sync(2) and when the syncer process is doing the 30-second writeback of dirty mmap() data to disk. Skip this speedup if we are doing an unmount() to be sure to get everything - we can afford to occasionally miss a msync while the system is running, but not at unmount. I'm not sure about the VXLOCK and MNT_WAIT case, it seems a bit odd to skip doing a page_clean at unmount time just because a vnode is VXLOCKed, but that's what was being done before...
1998-04-18 06:26:16 +00:00
struct vm_object *obj;
Make our v_usecount vnode reference count work identically to the original BSD code. The association between the vnode and the vm_object no longer includes reference counts. The major difference is that vm_object's are no longer freed gratuitiously from the vnode, and so once an object is created for the vnode, it will last as long as the vnode does. When a vnode object reference count is incremented, then the underlying vnode reference count is incremented also. The two "objects" are now more intimately related, and so the interactions are now much less complex. When vnodes are now normally placed onto the free queue with an object still attached. The rundown of the object happens at vnode rundown time, and happens with exactly the same filesystem semantics of the original VFS code. There is absolutely no need for vnode_pager_uncache and other travesties like that anymore. A side-effect of these changes is that SMP locking should be much simpler, the I/O copyin/copyout optimizations work, NFS should be more ponderable, and further work on layered filesystems should be less frustrating, because of the totally coherent management of the vnode objects and vnodes. Please be careful with your system while running this code, but I would greatly appreciate feedback as soon a reasonably possible.
1998-01-06 05:26:17 +00:00
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
CTR2(KTR_VFS, "%s: mp %p", __func__, mp);
Remove mntvnode_mtx and replace it with per-mountpoint mutex. Introduce two new macros MNT_ILOCK(mp)/MNT_IUNLOCK(mp) to operate on this mutex transparently. Eventually new mutex will be protecting more fields in struct mount, not only vnode list. Discussed with: jeff
2003-11-05 04:30:08 +00:00
MNT_ILOCK(mp);
Add marker vnodes to ensure that all vnodes associated with the mount point are iterated over when using MNT_VNODE_FOREACH. Reviewed by: truckman
2006-01-09 20:42:19 +00:00
MNT_VNODE_FOREACH(vp, mp, mvp) {
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
VI_LOCK(vp);
- Replace v_flag with v_iflag and v_vflag - v_vflag is protected by the vnode lock and is used when synchronization with VOP calls is needed. - v_iflag is protected by interlock and is used for dealing with vnode management issues. These flags include X/O LOCK, FREE, DOOMED, etc. - All accesses to v_iflag and v_vflag have either been locked or marked with mp_fixme's. - Many ASSERT_VOP_LOCKED calls have been added where the locking was not clear. - Many functions in vfs_subr.c were restructured to provide for stronger locking. Idea stolen from: BSD/OS
2002-08-04 10:29:36 +00:00
if ((vp->v_iflag & VI_OBJDIRTY) &&
Axe the 'thread' argument from VOP_ISLOCKED() and lockstatus() as it is always curthread. As KPI gets broken by this patch, manpages and __FreeBSD_version will be updated by further commits. Tested by: Andrea Barberio <insomniac at slackware dot it>
2008-02-25 18:45:57 +00:00
(flags == MNT_WAIT || VOP_ISLOCKED(vp) == 0)) {
Remove mntvnode_mtx and replace it with per-mountpoint mutex. Introduce two new macros MNT_ILOCK(mp)/MNT_IUNLOCK(mp) to operate on this mutex transparently. Eventually new mutex will be protecting more fields in struct mount, not only vnode list. Discussed with: jeff
2003-11-05 04:30:08 +00:00
MNT_IUNLOCK(mp);
Make our v_usecount vnode reference count work identically to the original BSD code. The association between the vnode and the vm_object no longer includes reference counts. The major difference is that vm_object's are no longer freed gratuitiously from the vnode, and so once an object is created for the vnode, it will last as long as the vnode does. When a vnode object reference count is incremented, then the underlying vnode reference count is incremented also. The two "objects" are now more intimately related, and so the interactions are now much less complex. When vnodes are now normally placed onto the free queue with an object still attached. The rundown of the object happens at vnode rundown time, and happens with exactly the same filesystem semantics of the original VFS code. There is absolutely no need for vnode_pager_uncache and other travesties like that anymore. A side-effect of these changes is that SMP locking should be much simpler, the I/O copyin/copyout optimizations work, NFS should be more ponderable, and further work on layered filesystems should be less frustrating, because of the totally coherent management of the vnode objects and vnodes. Please be careful with your system while running this code, but I would greatly appreciate feedback as soon a reasonably possible.
1998-01-06 05:26:17 +00:00
if (!vget(vp,
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
LK_EXCLUSIVE | LK_RETRY | LK_INTERLOCK,
curthread)) {
if (vp->v_vflag & VV_NOSYNC) { /* unlinked */
vput(vp);
Remove mntvnode_mtx and replace it with per-mountpoint mutex. Introduce two new macros MNT_ILOCK(mp)/MNT_IUNLOCK(mp) to operate on this mutex transparently. Eventually new mutex will be protecting more fields in struct mount, not only vnode list. Discussed with: jeff
2003-11-05 04:30:08 +00:00
MNT_ILOCK(mp);
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
continue;
}
Don't use VOP_GETVOBJECT, use vp->v_object directly.
2005-01-25 00:40:01 +00:00
obj = vp->v_object;
if (obj != NULL) {
- Acquire the vm_object's lock when performing vm_object_page_clean(). - Add a parameter to vm_pageout_flush() that tells vm_pageout_flush() whether its caller has locked the vm_object. (This is a temporary measure to bootstrap vm_object locking.)
2003-04-24 04:31:25 +00:00
VM_OBJECT_LOCK(obj);
Introduce a global lock for the vm subsystem (vm_mtx). vm_mtx does not recurse and is required for most low level vm operations. faults can not be taken without holding Giant. Memory subsystems can now call the base page allocators safely. Almost all atomic ops were removed as they are covered under the vm mutex. Alpha and ia64 now need to catch up to i386's trap handlers. FFS and NFS have been tested, other filesystems will need minor changes (grabbing the vm lock when twiddling page properties). Reviewed (partially) by: jake, jhb
2001-05-19 01:28:09 +00:00
vm_object_page_clean(obj, 0, 0,
flags == MNT_WAIT ?
OBJPC_SYNC : OBJPC_NOSYNC);
- Acquire the vm_object's lock when performing vm_object_page_clean(). - Add a parameter to vm_pageout_flush() that tells vm_pageout_flush() whether its caller has locked the vm_object. (This is a temporary measure to bootstrap vm_object locking.)
2003-04-24 04:31:25 +00:00
VM_OBJECT_UNLOCK(obj);
Make our v_usecount vnode reference count work identically to the original BSD code. The association between the vnode and the vm_object no longer includes reference counts. The major difference is that vm_object's are no longer freed gratuitiously from the vnode, and so once an object is created for the vnode, it will last as long as the vnode does. When a vnode object reference count is incremented, then the underlying vnode reference count is incremented also. The two "objects" are now more intimately related, and so the interactions are now much less complex. When vnodes are now normally placed onto the free queue with an object still attached. The rundown of the object happens at vnode rundown time, and happens with exactly the same filesystem semantics of the original VFS code. There is absolutely no need for vnode_pager_uncache and other travesties like that anymore. A side-effect of these changes is that SMP locking should be much simpler, the I/O copyin/copyout optimizations work, NFS should be more ponderable, and further work on layered filesystems should be less frustrating, because of the totally coherent management of the vnode objects and vnodes. Please be careful with your system while running this code, but I would greatly appreciate feedback as soon a reasonably possible.
1998-01-06 05:26:17 +00:00
}
vput(vp);
}
Remove mntvnode_mtx and replace it with per-mountpoint mutex. Introduce two new macros MNT_ILOCK(mp)/MNT_IUNLOCK(mp) to operate on this mutex transparently. Eventually new mutex will be protecting more fields in struct mount, not only vnode list. Discussed with: jeff
2003-11-05 04:30:08 +00:00
MNT_ILOCK(mp);
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
} else
VI_UNLOCK(vp);
Changes to fix the following bugs: 1) Files weren't properly synced on filesystems other than UFS. In some cases, this lead to lost data. Most likely would be noticed on NFS. The fix is to make the VM page sync/object_clean general rather than in each filesystem. 2) Mixing regular and mmaped file I/O on NFS was very broken. It caused chunks of files to end up as zeroes rather than the intended contents. The fix was to fix several race conditions and to kludge up the "b_dirtyoff" and "b_dirtyend" that NFS relies upon - paying attention to page modifications that occurred via the mmapping. Reviewed by: David Greenman Submitted by: John Dyson
1995-05-21 21:39:31 +00:00
}
Remove mntvnode_mtx and replace it with per-mountpoint mutex. Introduce two new macros MNT_ILOCK(mp)/MNT_IUNLOCK(mp) to operate on this mutex transparently. Eventually new mutex will be protecting more fields in struct mount, not only vnode list. Discussed with: jeff
2003-11-05 04:30:08 +00:00
MNT_IUNLOCK(mp);
Changes to fix the following bugs: 1) Files weren't properly synced on filesystems other than UFS. In some cases, this lead to lost data. Most likely would be noticed on NFS. The fix is to make the VM page sync/object_clean general rather than in each filesystem. 2) Mixing regular and mmaped file I/O on NFS was very broken. It caused chunks of files to end up as zeroes rather than the intended contents. The fix was to fix several race conditions and to kludge up the "b_dirtyoff" and "b_dirtyend" that NFS relies upon - paying attention to page modifications that occurred via the mmapping. Reviewed by: David Greenman Submitted by: John Dyson
1995-05-21 21:39:31 +00:00
}
Even though this looks like it, this is not a complex code change. The interface into the "VMIO" system has changed to be more consistant and robust. Essentially, it is now no longer necessary to call vn_open to get merged VM/Buffer cache operation, and exceptional conditions such as merged operation of VBLK devices is simpler and more correct. This code corrects a potentially large set of problems including the problems with ktrace output and loaded systems, file create/deletes, etc. Most of the changes to NFS are cosmetic and name changes, eliminating a layer of subroutine calls. The direct calls to vput/vrele have been re-instituted for better cross platform compatibility. Reviewed by: davidg
1996-08-21 21:56:23 +00:00
Style fixes: * Add lots of comments * Convert a couple of assertions to KASSERT() * Minimal whitespace & misapplied {} fixes * Convert #if 0 to #if COMPILING_LINT for code we presently do not support, but want to keep available. Reviewed by: adrian, markm
2000-09-22 12:22:36 +00:00
/*
* Mark a vnode as free, putting it up for recycling.
*/
- Now that there are no external users of vfree() make it static. - Move VSHOULDBUSY, VSHOULDFREE, and VTRYRECYCLE into vfs_subr.c so no one else attempts to grow a dependency on them. - Now that objects with pages hold the vnode we don't have to do unlocked checks for the page count in the vm object in VSHOULDFREE. These three macros could simply check for holdcnt state transitions to determine whether the vnode is on the free list already, but the extra safety the flag affords us is probably worth the minimal cost. - The leafonly sysctl and code have been dead for several years now, remove the sysctl and the code that employed it from vtryrecycle(). - vtryrecycle() also no longer has to check the object's page count as the object holds the vnode until it reaches 0. Sponsored by: Isilon Systems, Inc.
2005-03-15 14:38:16 +00:00
static void
Add b_bufobj to struct buf which eventually will eliminate the need for b_vp. Initialize b_bufobj for all buffers. Make incore() and gbincore() take a bufobj instead of a vnode. Make inmem() local to vfs_bio.c Change a lot of VI_[UN]LOCK(bp->b_vp) to BO_[UN]LOCK(bp->b_bufobj) also VI_MTX() to BO_MTX(), Make buf_vlist_add() take a bufobj instead of a vnode. Eliminate other uses of bp->b_vp where bp->b_bufobj will do. Various minor polishing: remove "register", turn panic into KASSERT, use new function declarations, TAILQ_FOREACH_SAFE() etc.
2004-10-22 08:47:20 +00:00
vfree(struct vnode *vp)
1. Add a {pointer, v_id} pair to the vnode to store the reference to the ".." vnode. This is cheaper storagewise than keeping it in the namecache, and it makes more sense since it's a 1:1 mapping. 2. Also handle the case of "." more intelligently rather than stuff the namecache with pointless entries. 3. Add two lists to the vnode and hang namecache entries which go from or to this vnode. When cleaning a vnode, delete all namecache entries it invalidates. 4. Never reuse namecache enties, malloc new ones when we need it, free old ones when they die. No longer a hard limit on how many we can have. 5. Remove the upper limit on namelength of namecache entries. 6. Make a global list for negative namecache entries, limit their number to a sysctl'able (debug.ncnegfactor) fraction of the total namecache. Currently the default fraction is 1/16th. (Suggestions for better default wanted!) 7. Assign v_id correctly in the face of 32bit rollover. 8. Remove the LRU list for namecache entries, not needed. Remove the #ifdef NCH_STATISTICS stuff, it's not needed either. 9. Use the vnode freelist as a true LRU list, also for namecache accesses. 10. Reuse vnodes more aggresively but also more selectively, if we can't reuse, malloc a new one. There is no longer a hard limit on their number, they grow to the point where we don't reuse potentially usable vnodes. A vnode will not get recycled if still has pages in core or if it is the source of namecache entries (Yes, this does indeed work :-) "." and ".." are not namecache entries any longer...) 11. Do not overload the v_id field in namecache entries with whiteout information, use a char sized flags field instead, so we can get rid of the vpid and v_id fields from the namecache struct. Since we're linked to the vnodes and purged when they're cleaned, we don't have to check the v_id any more. 12. NFS knew about the limitation on name length in the namecache, it shouldn't and doesn't now. Bugs: The namecache statistics no longer includes the hits for ".." and "." hits. Performance impact: Generally in the +/- 0.5% for "normal" workstations, but I hope this will allow the system to be selftuning over a bigger range of "special" applications. The case where RAM is available but unused for cache because we don't have any vnodes should be gone. Future work: Straighten out the namecache statistics. "desiredvnodes" is still used to (bogusly ?) size hash tables in the filesystems. I have still to find a way to safely free unused vnodes back so their number can shrink when not needed. There is a few uses of the v_id field left in the filesystems, scheduled for demolition at a later time. Maybe a one slot cache for unused namecache entries should be implemented to decrease the malloc/free frequency.
1997-05-04 09:17:38 +00:00
{
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
- Move ASSERT_VOP_*LOCK* functionality into functions in vfs_subr.c - Make the VI asserts more orthogonal to the rest of the asserts by using a new, common vfs_badlock() function and adding a 'str' arg. - Adjust generated ASSERTS to match the new prototype. - Adjust explicit ASSERTS to match the new prototype.
2002-09-26 04:48:44 +00:00
ASSERT_VI_LOCKED(vp, "vfree");
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_lock(&vnode_free_list_mtx);
- Add KTR_VFS tracing to track the life of vnodes. Eventually KTR_VFS events could be added to cover other interesting details. - Add some VNASSERTs to discover places where we access vnodes after they have been uma_zfree'd before we try to free them again. - Add a few more VNASSERTs to vdestroy() to be certain that the vnode is really unused. Sponsored by: Isilon Systems, Inc.
2005-06-11 01:16:46 +00:00
VNASSERT(vp->v_op != NULL, vp, ("vfree: vnode already reclaimed."));
Convert KASSERTS to VNASSERTS
2005-02-17 10:28:58 +00:00
VNASSERT((vp->v_iflag & VI_FREE) == 0, vp, ("vnode already free"));
- Remove vx_lock, vx_unlock, vx_wait, etc. - Add a vn_start_write/vn_finished_write around vlrureclaim so we don't do writing ops without suspending. This could suspend the vlruproc which should not be a problem under normal circumstances. - Manually implement VMIGHTFREE in vlrureclaim as this was the only instance where it was used. - Acquire a lock before calling vgone() as it now requires it. - Move the acquisition of the vnode interlock from vtryrecycle() to getnewvnode() so that if it fails we don't drop and reacquire the vnode_free_list_mtx. - Check for a usecount or holdcount at the end of vtryrecycle() in case someone grabbed a ref while we were recycling. Abort the recycle, and on the final ref drop this vnode will be placed on the head of the free list. - Move the redundant VOP_INACTIVE protection code into the local vinactive() routine to avoid code bloat. - Keep the vnode lock held across calls to vgone() in several places. - vgonel() no longer uses XLOCK, instead callers must hold an exclusive vnode lock. The VI_DOOMED flag is set to allow other threads to detect a vnode which is no longer valid. This flag is set until the last reference is gone, and there are no chances for a new ref. vgonel() holds this lock across the entire function, which greatly simplifies logic. _ Only vfree() in one place in vgone() not three. - Adjust vget() to check the VI_DOOMED flag prior to waiting on the lock in the LK_NOWAIT case. In other cases, check after we have slept and acquired an exlusive lock. This will simulate the old vx_wait() behavior. Sponsored by: Isilon Systems, Inc.
2005-03-13 11:54:28 +00:00
VNASSERT(VSHOULDFREE(vp), vp, ("vfree: freeing when we shouldn't"));
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
VNASSERT((vp->v_iflag & VI_DOOMED) == 0, vp,
("vfree: Freeing doomed vnode"));
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
CTR2(KTR_VFS, "%s: vp %p", __func__, vp);
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
if (vp->v_iflag & VI_AGE) {
Change the 0xdeadb hack to a flag called VDOOMED. Introduce VFREE which indicates that vnode is on freelist. Rename vholdrele() to vdrop(). Create vfree() and vbusy() to add/delete vnode from freelist. Add vfree()/vbusy() to keep (v_holdcnt != 0 || v_usecount != 0) vnodes off the freelist. Generalize vhold()/v_holdcnt to mean "do not recycle". Fix reassignbuf()s lack of use of vhold(). Use vhold() instead of checking v_cache_src list. Remove vtouch(), the vnodes are always vget'ed soon enough after for it to have any measuable effect. Add sysctl debug.freevnodes to keep track of things. Move cache_purge() up in getnewvnodes to avoid race. Decrement v_usecount after VOP_INACTIVE(), put a vhold() on it during VOP_INACTIVE() Unmacroize vhold()/vdrop() Print out VDOOMED and VFREE flags (XXX: should use %b) Reviewed by: dyson
1997-08-31 07:32:39 +00:00
TAILQ_INSERT_HEAD(&vnode_free_list, vp, v_freelist);
} else {
TAILQ_INSERT_TAIL(&vnode_free_list, vp, v_freelist);
1. Add a {pointer, v_id} pair to the vnode to store the reference to the ".." vnode. This is cheaper storagewise than keeping it in the namecache, and it makes more sense since it's a 1:1 mapping. 2. Also handle the case of "." more intelligently rather than stuff the namecache with pointless entries. 3. Add two lists to the vnode and hang namecache entries which go from or to this vnode. When cleaning a vnode, delete all namecache entries it invalidates. 4. Never reuse namecache enties, malloc new ones when we need it, free old ones when they die. No longer a hard limit on how many we can have. 5. Remove the upper limit on namelength of namecache entries. 6. Make a global list for negative namecache entries, limit their number to a sysctl'able (debug.ncnegfactor) fraction of the total namecache. Currently the default fraction is 1/16th. (Suggestions for better default wanted!) 7. Assign v_id correctly in the face of 32bit rollover. 8. Remove the LRU list for namecache entries, not needed. Remove the #ifdef NCH_STATISTICS stuff, it's not needed either. 9. Use the vnode freelist as a true LRU list, also for namecache accesses. 10. Reuse vnodes more aggresively but also more selectively, if we can't reuse, malloc a new one. There is no longer a hard limit on their number, they grow to the point where we don't reuse potentially usable vnodes. A vnode will not get recycled if still has pages in core or if it is the source of namecache entries (Yes, this does indeed work :-) "." and ".." are not namecache entries any longer...) 11. Do not overload the v_id field in namecache entries with whiteout information, use a char sized flags field instead, so we can get rid of the vpid and v_id fields from the namecache struct. Since we're linked to the vnodes and purged when they're cleaned, we don't have to check the v_id any more. 12. NFS knew about the limitation on name length in the namecache, it shouldn't and doesn't now. Bugs: The namecache statistics no longer includes the hits for ".." and "." hits. Performance impact: Generally in the +/- 0.5% for "normal" workstations, but I hope this will allow the system to be selftuning over a bigger range of "special" applications. The case where RAM is available but unused for cache because we don't have any vnodes should be gone. Future work: Straighten out the namecache statistics. "desiredvnodes" is still used to (bogusly ?) size hash tables in the filesystems. I have still to find a way to safely free unused vnodes back so their number can shrink when not needed. There is a few uses of the v_id field left in the filesystems, scheduled for demolition at a later time. Maybe a one slot cache for unused namecache entries should be implemented to decrease the malloc/free frequency.
1997-05-04 09:17:38 +00:00
}
Change the 0xdeadb hack to a flag called VDOOMED. Introduce VFREE which indicates that vnode is on freelist. Rename vholdrele() to vdrop(). Create vfree() and vbusy() to add/delete vnode from freelist. Add vfree()/vbusy() to keep (v_holdcnt != 0 || v_usecount != 0) vnodes off the freelist. Generalize vhold()/v_holdcnt to mean "do not recycle". Fix reassignbuf()s lack of use of vhold(). Use vhold() instead of checking v_cache_src list. Remove vtouch(), the vnodes are always vget'ed soon enough after for it to have any measuable effect. Add sysctl debug.freevnodes to keep track of things. Move cache_purge() up in getnewvnodes to avoid race. Decrement v_usecount after VOP_INACTIVE(), put a vhold() on it during VOP_INACTIVE() Unmacroize vhold()/vdrop() Print out VDOOMED and VFREE flags (XXX: should use %b) Reviewed by: dyson
1997-08-31 07:32:39 +00:00
freevnodes++;
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
vp->v_iflag &= ~VI_AGE;
- Replace v_flag with v_iflag and v_vflag - v_vflag is protected by the vnode lock and is used when synchronization with VOP calls is needed. - v_iflag is protected by interlock and is used for dealing with vnode management issues. These flags include X/O LOCK, FREE, DOOMED, etc. - All accesses to v_iflag and v_vflag have either been locked or marked with mp_fixme's. - Many ASSERT_VOP_LOCKED calls have been added where the locking was not clear. - Many functions in vfs_subr.c were restructured to provide for stronger locking. Idea stolen from: BSD/OS
2002-08-04 10:29:36 +00:00
vp->v_iflag |= VI_FREE;
- In vnlru_free() remove the vnode from the free list before we call vtryrecycle(). We could sometimes get into situations where two threads could try to recycle the same vnode before this. - vtryrecycle() is now responsible for returning the vnode to the free list if it fails and someone else hasn't done it. - Make a new function vfreehead() which moves a vnode to the head of the free list and use it in vgone() to clean up that code a bit. Sponsored by: Isilon Systems, Inc. Reported by: pho, kkenn
2005-04-30 11:22:40 +00:00
mtx_unlock(&vnode_free_list_mtx);
}
Gratuitous whitespace cleanup.
2002-06-06 15:46:38 +00:00
/*
Style fixes: * Add lots of comments * Convert a couple of assertions to KASSERT() * Minimal whitespace & misapplied {} fixes * Convert #if 0 to #if COMPILING_LINT for code we presently do not support, but want to keep available. Reviewed by: adrian, markm
2000-09-22 12:22:36 +00:00
* Opposite of vfree() - mark a vnode as in use.
*/
Make various vnode related functions static
2005-02-10 12:28:58 +00:00
static void
Add b_bufobj to struct buf which eventually will eliminate the need for b_vp. Initialize b_bufobj for all buffers. Make incore() and gbincore() take a bufobj instead of a vnode. Make inmem() local to vfs_bio.c Change a lot of VI_[UN]LOCK(bp->b_vp) to BO_[UN]LOCK(bp->b_bufobj) also VI_MTX() to BO_MTX(), Make buf_vlist_add() take a bufobj instead of a vnode. Eliminate other uses of bp->b_vp where bp->b_bufobj will do. Various minor polishing: remove "register", turn panic into KASSERT, use new function declarations, TAILQ_FOREACH_SAFE() etc.
2004-10-22 08:47:20 +00:00
vbusy(struct vnode *vp)
Change the 0xdeadb hack to a flag called VDOOMED. Introduce VFREE which indicates that vnode is on freelist. Rename vholdrele() to vdrop(). Create vfree() and vbusy() to add/delete vnode from freelist. Add vfree()/vbusy() to keep (v_holdcnt != 0 || v_usecount != 0) vnodes off the freelist. Generalize vhold()/v_holdcnt to mean "do not recycle". Fix reassignbuf()s lack of use of vhold(). Use vhold() instead of checking v_cache_src list. Remove vtouch(), the vnodes are always vget'ed soon enough after for it to have any measuable effect. Add sysctl debug.freevnodes to keep track of things. Move cache_purge() up in getnewvnodes to avoid race. Decrement v_usecount after VOP_INACTIVE(), put a vhold() on it during VOP_INACTIVE() Unmacroize vhold()/vdrop() Print out VDOOMED and VFREE flags (XXX: should use %b) Reviewed by: dyson
1997-08-31 07:32:39 +00:00
{
- Move ASSERT_VOP_*LOCK* functionality into functions in vfs_subr.c - Make the VI asserts more orthogonal to the rest of the asserts by using a new, common vfs_badlock() function and adding a 'str' arg. - Adjust generated ASSERTS to match the new prototype. - Adjust explicit ASSERTS to match the new prototype.
2002-09-26 04:48:44 +00:00
ASSERT_VI_LOCKED(vp, "vbusy");
Convert KASSERTS to VNASSERTS
2005-02-17 10:28:58 +00:00
VNASSERT((vp->v_iflag & VI_FREE) != 0, vp, ("vnode not free"));
- Add KTR_VFS tracing to track the life of vnodes. Eventually KTR_VFS events could be added to cover other interesting details. - Add some VNASSERTs to discover places where we access vnodes after they have been uma_zfree'd before we try to free them again. - Add a few more VNASSERTs to vdestroy() to be certain that the vnode is really unused. Sponsored by: Isilon Systems, Inc.
2005-06-11 01:16:46 +00:00
VNASSERT(vp->v_op != NULL, vp, ("vbusy: vnode already reclaimed."));
Add more KTR_VFS logging point in order to have a more effective tracing. Reviewed by: brueffer, kib Tested by: Gianni Trematerra <giovanni D trematerra A gmail D com>
2009-02-05 15:03:35 +00:00
CTR2(KTR_VFS, "%s: vp %p", __func__, vp);
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
mtx_lock(&vnode_free_list_mtx);
Simplify and rationalise the management of the vnode free list (preparing the code to add snapshots).
2000-07-04 04:32:40 +00:00
TAILQ_REMOVE(&vnode_free_list, vp, v_freelist);
freevnodes--;
- Dont clear OWEINACT in vbusy(), we still owe an inactive call if someone vhold()s us. - Avoid an extra mutex acquire and release in the common case of vgonel() by checking for OWEINACT at the start of the function. - Fix the case where we set OWEINACT in vput(). LK_EXCLUPGRADE drops our shared lock if it fails. Sponsored by: Isilon Systems, Inc.
2005-03-29 10:02:48 +00:00
vp->v_iflag &= ~(VI_FREE|VI_AGE);
- Change holdcnt use around vnode recycling. We now always keep a holdcnt ref while we're calling vgone(). This prevents transient refs from re-adding us to the free list. Previously, a vfree() triggered via vinvalbuf() getting rid of all of a vnode's pages could place a partially destructed vnode on the free list where vtryrecycle() could find it. The first call to vtryrecycle would hang up on the vnode lock, but when it failed it would place a now dead vnode onto the free list, and another call to vtryrecycle() would free an already free vnode. There were many complications of having a zero ref count while freeing which can now go away. - Change vdropl() to release the interlock before returning. All callers now respect this, so vdropl() directly frees VI_DOOMED vnodes once the last ref is dropped. This means that we'll never have VI_DOOMED vnodes on the free list. - Seperate v_incr_usecount() into v_incr_usecount(), v_decr_usecount() and v_decr_useonly(). The incr/decr split is so that incr usecount can return with the interlock still held while decr drops the interlock so it can call vdropl() which will potentially free the vnode. The calling function can't drop the lock of an already free'd node. v_decr_useonly() drops a usecount without droping the hold count. This is done so the usecount reaches zero in vput() before we recycle, however the holdcount is still 1 which prevents any new references from placing the vnode back on the free list. - Fix vnlrureclaim() to vhold the vnode since it doesn't do a vget(). We wouldn't want vnlrureclaim() to bump the usecount since this has different semantics. Also change vnlrureclaim() to do a NOWAIT on the vn_lock. When this function runs we're usually in a desperate situation and we wouldn't want to wait for any specific vnode to be released. - Fix a bunch of misc comments to reflect the new behavior. - Add vhold() and vdrop() to vflush() for the same reasons that we do in vlrureclaim(). Previously we held no reference and a vnode could have been freed while we were waiting on the lock. - Get rid of vlruvp() and vfreehead(). Neither are used. vlruvp() should really be rethought before it's reintroduced. - vgonel() always returns with the vnode locked now and never puts the vnode back on a free list. The vnode will be freed as soon as the last reference is released. Sponsored by: Isilon Systems, Inc. Debugging help from: Kris Kennaway, Peter Holm Approved by: re (blanket vfs)
2005-06-16 04:41:42 +00:00
mtx_unlock(&vnode_free_list_mtx);
1. Add a {pointer, v_id} pair to the vnode to store the reference to the ".." vnode. This is cheaper storagewise than keeping it in the namecache, and it makes more sense since it's a 1:1 mapping. 2. Also handle the case of "." more intelligently rather than stuff the namecache with pointless entries. 3. Add two lists to the vnode and hang namecache entries which go from or to this vnode. When cleaning a vnode, delete all namecache entries it invalidates. 4. Never reuse namecache enties, malloc new ones when we need it, free old ones when they die. No longer a hard limit on how many we can have. 5. Remove the upper limit on namelength of namecache entries. 6. Make a global list for negative namecache entries, limit their number to a sysctl'able (debug.ncnegfactor) fraction of the total namecache. Currently the default fraction is 1/16th. (Suggestions for better default wanted!) 7. Assign v_id correctly in the face of 32bit rollover. 8. Remove the LRU list for namecache entries, not needed. Remove the #ifdef NCH_STATISTICS stuff, it's not needed either. 9. Use the vnode freelist as a true LRU list, also for namecache accesses. 10. Reuse vnodes more aggresively but also more selectively, if we can't reuse, malloc a new one. There is no longer a hard limit on their number, they grow to the point where we don't reuse potentially usable vnodes. A vnode will not get recycled if still has pages in core or if it is the source of namecache entries (Yes, this does indeed work :-) "." and ".." are not namecache entries any longer...) 11. Do not overload the v_id field in namecache entries with whiteout information, use a char sized flags field instead, so we can get rid of the vpid and v_id fields from the namecache struct. Since we're linked to the vnodes and purged when they're cleaned, we don't have to check the v_id any more. 12. NFS knew about the limitation on name length in the namecache, it shouldn't and doesn't now. Bugs: The namecache statistics no longer includes the hits for ".." and "." hits. Performance impact: Generally in the +/- 0.5% for "normal" workstations, but I hope this will allow the system to be selftuning over a bigger range of "special" applications. The case where RAM is available but unused for cache because we don't have any vnodes should be gone. Future work: Straighten out the namecache statistics. "desiredvnodes" is still used to (bogusly ?) size hash tables in the filesystems. I have still to find a way to safely free unused vnodes back so their number can shrink when not needed. There is a few uses of the v_id field left in the filesystems, scheduled for demolition at a later time. Maybe a one slot cache for unused namecache entries should be implemented to decrease the malloc/free frequency.
1997-05-04 09:17:38 +00:00
}
Add support for poll(2) on files. vop_nopoll() now returns POLLNVAL if one of the new poll types is requested; hopefully this will not break any existing code. (This is done so that programs have a dependable way of determining whether a filesystem supports the extended poll types or not.) The new poll types added are: POLLWRITE - file contents may have been modified POLLNLINK - file was linked, unlinked, or renamed POLLATTRIB - file's attributes may have been changed POLLEXTEND - file was extended Note that the internal operation of poll() means that it is impossible for two processes to reliably poll for the same event (this could be fixed but may not be worth it), so it is not possible to rewrite `tail -f' to use poll at this time.
1997-12-15 03:09:59 +00:00
Protect check for v_pollinfo == NULL and assignment of the newly allocated vpollinfo with vnode interlock. Fully initialize vpollinfo before putting pointer to it into vp->v_pollinfo. Discussed with: dwhite Tested by: pho MFC after: 1 week
2008-10-28 12:08:36 +00:00
static void
destroy_vpollinfo(struct vpollinfo *vi)
{
knlist_destroy(&vi->vpi_selinfo.si_note);
mtx_destroy(&vi->vpi_lock);
uma_zfree(vnodepoll_zone, vi);
}
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
/*
* Initalize per-vnode helper structure to hold poll-related state.
*/
void
v_addpollinfo(struct vnode *vp)
{
In v_addpollinfo(), we allocate storage to back vp->v_pollinfo. However, we may sleep when doing so; check that we didn't race with another thread allocating storage for the vnode after allocation is made to a local pointer, and only update the vnode pointer if it's still NULL. Otherwise, accept that another thread got there first, and release the local storage. Discussed with: jmg
2004-08-11 01:27:53 +00:00
struct vpollinfo *vi;
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
Protect check for v_pollinfo == NULL and assignment of the newly allocated vpollinfo with vnode interlock. Fully initialize vpollinfo before putting pointer to it into vp->v_pollinfo. Discussed with: dwhite Tested by: pho MFC after: 1 week
2008-10-28 12:08:36 +00:00
if (vp->v_pollinfo != NULL)
return;
In v_addpollinfo(), we allocate storage to back vp->v_pollinfo. However, we may sleep when doing so; check that we didn't race with another thread allocating storage for the vnode after allocation is made to a local pointer, and only update the vnode pointer if it's still NULL. Otherwise, accept that another thread got there first, and release the local storage. Discussed with: jmg
2004-08-11 01:27:53 +00:00
vi = uma_zalloc(vnodepoll_zone, M_WAITOK);
Protect check for v_pollinfo == NULL and assignment of the newly allocated vpollinfo with vnode interlock. Fully initialize vpollinfo before putting pointer to it into vp->v_pollinfo. Discussed with: dwhite Tested by: pho MFC after: 1 week
2008-10-28 12:08:36 +00:00
mtx_init(&vi->vpi_lock, "vnode pollinfo", NULL, MTX_DEF);
knlist_init(&vi->vpi_selinfo.si_note, vp, vfs_knllock,
vfs_knlunlock, vfs_knllocked);
VI_LOCK(vp);
In v_addpollinfo(), we allocate storage to back vp->v_pollinfo. However, we may sleep when doing so; check that we didn't race with another thread allocating storage for the vnode after allocation is made to a local pointer, and only update the vnode pointer if it's still NULL. Otherwise, accept that another thread got there first, and release the local storage. Discussed with: jmg
2004-08-11 01:27:53 +00:00
if (vp->v_pollinfo != NULL) {
Protect check for v_pollinfo == NULL and assignment of the newly allocated vpollinfo with vnode interlock. Fully initialize vpollinfo before putting pointer to it into vp->v_pollinfo. Discussed with: dwhite Tested by: pho MFC after: 1 week
2008-10-28 12:08:36 +00:00
VI_UNLOCK(vp);
destroy_vpollinfo(vi);
In v_addpollinfo(), we allocate storage to back vp->v_pollinfo. However, we may sleep when doing so; check that we didn't race with another thread allocating storage for the vnode after allocation is made to a local pointer, and only update the vnode pointer if it's still NULL. Otherwise, accept that another thread got there first, and release the local storage. Discussed with: jmg
2004-08-11 01:27:53 +00:00
return;
}
vp->v_pollinfo = vi;
Protect check for v_pollinfo == NULL and assignment of the newly allocated vpollinfo with vnode interlock. Fully initialize vpollinfo before putting pointer to it into vp->v_pollinfo. Discussed with: dwhite Tested by: pho MFC after: 1 week
2008-10-28 12:08:36 +00:00
VI_UNLOCK(vp);
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
}
Add support for poll(2) on files. vop_nopoll() now returns POLLNVAL if one of the new poll types is requested; hopefully this will not break any existing code. (This is done so that programs have a dependable way of determining whether a filesystem supports the extended poll types or not.) The new poll types added are: POLLWRITE - file contents may have been modified POLLNLINK - file was linked, unlinked, or renamed POLLATTRIB - file's attributes may have been changed POLLEXTEND - file was extended Note that the internal operation of poll() means that it is impossible for two processes to reliably poll for the same event (this could be fixed but may not be worth it), so it is not possible to rewrite `tail -f' to use poll at this time.
1997-12-15 03:09:59 +00:00
/*
* Record a process's interest in events which might happen to
* a vnode. Because poll uses the historic select-style interface
* internally, this routine serves as both the ``check for any
* pending events'' and the ``record my interest in future events''
* functions. (These are done together, while the lock is held,
* to avoid race conditions.)
*/
int
Convert remaining functions in vfs_subr.c from K&R prototypes to ANSI C prototypes, as the majority of new functions added have been in this style. Changing prototype style now results in gcc noticing that the implementation of vn_pollrecord() has a 'short' argument instead of 'int' as prototyped in vnode.h, so correct that definition. In practice this didn't matter as only poll flags in the lower 16 bits are used. MFC after: 1 week
2006-01-21 19:42:10 +00:00
vn_pollrecord(struct vnode *vp, struct thread *td, int events)
Add support for poll(2) on files. vop_nopoll() now returns POLLNVAL if one of the new poll types is requested; hopefully this will not break any existing code. (This is done so that programs have a dependable way of determining whether a filesystem supports the extended poll types or not.) The new poll types added are: POLLWRITE - file contents may have been modified POLLNLINK - file was linked, unlinked, or renamed POLLATTRIB - file's attributes may have been changed POLLEXTEND - file was extended Note that the internal operation of poll() means that it is impossible for two processes to reliably poll for the same event (this could be fixed but may not be worth it), so it is not possible to rewrite `tail -f' to use poll at this time.
1997-12-15 03:09:59 +00:00
{
Move the stuff related to select and poll out of struct vnode. The use of the zone allocator may or may not be overkill. There is an XXX: over in ufs/ufs/ufs_vnops.c that jlemon may need to revisit. This shaves about 60 bytes of struct vnode which on my laptop means 600k less RAM used for vnodes.
2002-02-17 21:15:36 +00:00
Protect check for v_pollinfo == NULL and assignment of the newly allocated vpollinfo with vnode interlock. Fully initialize vpollinfo before putting pointer to it into vp->v_pollinfo. Discussed with: dwhite Tested by: pho MFC after: 1 week
2008-10-28 12:08:36 +00:00
v_addpollinfo(vp);
Move the stuff related to select and poll out of struct vnode. The use of the zone allocator may or may not be overkill. There is an XXX: over in ufs/ufs/ufs_vnops.c that jlemon may need to revisit. This shaves about 60 bytes of struct vnode which on my laptop means 600k less RAM used for vnodes.
2002-02-17 21:15:36 +00:00
mtx_lock(&vp->v_pollinfo->vpi_lock);
if (vp->v_pollinfo->vpi_revents & events) {
Add support for poll(2) on files. vop_nopoll() now returns POLLNVAL if one of the new poll types is requested; hopefully this will not break any existing code. (This is done so that programs have a dependable way of determining whether a filesystem supports the extended poll types or not.) The new poll types added are: POLLWRITE - file contents may have been modified POLLNLINK - file was linked, unlinked, or renamed POLLATTRIB - file's attributes may have been changed POLLEXTEND - file was extended Note that the internal operation of poll() means that it is impossible for two processes to reliably poll for the same event (this could be fixed but may not be worth it), so it is not possible to rewrite `tail -f' to use poll at this time.
1997-12-15 03:09:59 +00:00
/*
* This leaves events we are not interested
* in available for the other process which
* which presumably had requested them
* (otherwise they would never have been
* recorded).
*/
Move the stuff related to select and poll out of struct vnode. The use of the zone allocator may or may not be overkill. There is an XXX: over in ufs/ufs/ufs_vnops.c that jlemon may need to revisit. This shaves about 60 bytes of struct vnode which on my laptop means 600k less RAM used for vnodes.
2002-02-17 21:15:36 +00:00
events &= vp->v_pollinfo->vpi_revents;
vp->v_pollinfo->vpi_revents &= ~events;
Add support for poll(2) on files. vop_nopoll() now returns POLLNVAL if one of the new poll types is requested; hopefully this will not break any existing code. (This is done so that programs have a dependable way of determining whether a filesystem supports the extended poll types or not.) The new poll types added are: POLLWRITE - file contents may have been modified POLLNLINK - file was linked, unlinked, or renamed POLLATTRIB - file's attributes may have been changed POLLEXTEND - file was extended Note that the internal operation of poll() means that it is impossible for two processes to reliably poll for the same event (this could be fixed but may not be worth it), so it is not possible to rewrite `tail -f' to use poll at this time.
1997-12-15 03:09:59 +00:00
Move the stuff related to select and poll out of struct vnode. The use of the zone allocator may or may not be overkill. There is an XXX: over in ufs/ufs/ufs_vnops.c that jlemon may need to revisit. This shaves about 60 bytes of struct vnode which on my laptop means 600k less RAM used for vnodes.
2002-02-17 21:15:36 +00:00
mtx_unlock(&vp->v_pollinfo->vpi_lock);
Style return statements in vn_pollrecord().
2008-10-28 12:22:33 +00:00
return (events);
Add support for poll(2) on files. vop_nopoll() now returns POLLNVAL if one of the new poll types is requested; hopefully this will not break any existing code. (This is done so that programs have a dependable way of determining whether a filesystem supports the extended poll types or not.) The new poll types added are: POLLWRITE - file contents may have been modified POLLNLINK - file was linked, unlinked, or renamed POLLATTRIB - file's attributes may have been changed POLLEXTEND - file was extended Note that the internal operation of poll() means that it is impossible for two processes to reliably poll for the same event (this could be fixed but may not be worth it), so it is not possible to rewrite `tail -f' to use poll at this time.
1997-12-15 03:09:59 +00:00
}
Move the stuff related to select and poll out of struct vnode. The use of the zone allocator may or may not be overkill. There is an XXX: over in ufs/ufs/ufs_vnops.c that jlemon may need to revisit. This shaves about 60 bytes of struct vnode which on my laptop means 600k less RAM used for vnodes.
2002-02-17 21:15:36 +00:00
vp->v_pollinfo->vpi_events |= events;
selrecord(td, &vp->v_pollinfo->vpi_selinfo);
mtx_unlock(&vp->v_pollinfo->vpi_lock);
Style return statements in vn_pollrecord().
2008-10-28 12:22:33 +00:00
return (0);
Add support for poll(2) on files. vop_nopoll() now returns POLLNVAL if one of the new poll types is requested; hopefully this will not break any existing code. (This is done so that programs have a dependable way of determining whether a filesystem supports the extended poll types or not.) The new poll types added are: POLLWRITE - file contents may have been modified POLLNLINK - file was linked, unlinked, or renamed POLLATTRIB - file's attributes may have been changed POLLEXTEND - file was extended Note that the internal operation of poll() means that it is impossible for two processes to reliably poll for the same event (this could be fixed but may not be worth it), so it is not possible to rewrite `tail -f' to use poll at this time.
1997-12-15 03:09:59 +00:00
}
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
/*
* Routine to create and manage a filesystem syncer vnode.
*/
Remove __P.
2002-03-19 21:25:46 +00:00
#define sync_close ((int (*)(struct vop_close_args *))nullop)
static int sync_fsync(struct vop_fsync_args *);
static int sync_inactive(struct vop_inactive_args *);
static int sync_reclaim(struct vop_reclaim_args *);
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
Back when VOP_* was introduced, we did not have new-style struct initializations but we did have lofty goals and big ideals. Adjust to more contemporary circumstances and gain type checking. Replace the entire vop_t frobbing thing with properly typed structures. The only casualty is that we can not add a new VOP_ method with a loadable module. History has not given us reason to belive this would ever be feasible in the the first place. Eliminate in toto VOCALL(), vop_t, VNODEOP_SET() etc. Give coda correct prototypes and function definitions for all vop_()s. Generate a bit more data from the vnode_if.src file: a struct vop_vector and protype typedefs for all vop methods. Add a new vop_bypass() and make vop_default be a pointer to another struct vop_vector. Remove a lot of vfs_init since vop_vector is ready to use from the compiler. Cast various vop_mumble() to void * with uppercase name, for instance VOP_PANIC, VOP_NULL etc. Implement VCALL() by making vdesc_offset the offsetof() the relevant function pointer in vop_vector. This is disgusting but since the code is generated by a script comparatively safe. The alternative for nullfs etc. would be much worse. Fix up all vnode method vectors to remove casts so they become typesafe. (The bulk of this is generated by scripts)
2004-12-01 23:16:38 +00:00
static struct vop_vector sync_vnodeops = {
.vop_bypass = VOP_EOPNOTSUPP,
.vop_close = sync_close, /* close */
.vop_fsync = sync_fsync, /* fsync */
.vop_inactive = sync_inactive, /* inactive */
.vop_reclaim = sync_reclaim, /* reclaim */
Since renaming of vop_lock to _vop_lock, pre- and post-condition function calls are no more generated for vop_lock. Rename _vop_lock to vop_lock1 to satisfy tools/vnode_if.awk assumption about vop naming conventions. This restores pre/post-condition calls.
2007-05-18 13:02:13 +00:00
.vop_lock1 = vop_stdlock, /* lock */
Back when VOP_* was introduced, we did not have new-style struct initializations but we did have lofty goals and big ideals. Adjust to more contemporary circumstances and gain type checking. Replace the entire vop_t frobbing thing with properly typed structures. The only casualty is that we can not add a new VOP_ method with a loadable module. History has not given us reason to belive this would ever be feasible in the the first place. Eliminate in toto VOCALL(), vop_t, VNODEOP_SET() etc. Give coda correct prototypes and function definitions for all vop_()s. Generate a bit more data from the vnode_if.src file: a struct vop_vector and protype typedefs for all vop methods. Add a new vop_bypass() and make vop_default be a pointer to another struct vop_vector. Remove a lot of vfs_init since vop_vector is ready to use from the compiler. Cast various vop_mumble() to void * with uppercase name, for instance VOP_PANIC, VOP_NULL etc. Implement VCALL() by making vdesc_offset the offsetof() the relevant function pointer in vop_vector. This is disgusting but since the code is generated by a script comparatively safe. The alternative for nullfs etc. would be much worse. Fix up all vnode method vectors to remove casts so they become typesafe. (The bulk of this is generated by scripts)
2004-12-01 23:16:38 +00:00
.vop_unlock = vop_stdunlock, /* unlock */
.vop_islocked = vop_stdislocked, /* islocked */
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
};
/*
* Create a new filesystem syncer vnode for the specified mount point.
*/
int
Convert remaining functions in vfs_subr.c from K&R prototypes to ANSI C prototypes, as the majority of new functions added have been in this style. Changing prototype style now results in gcc noticing that the implementation of vn_pollrecord() has a 'short' argument instead of 'int' as prototyped in vnode.h, so correct that definition. In practice this didn't matter as only poll flags in the lower 16 bits are used. MFC after: 1 week
2006-01-21 19:42:10 +00:00
vfs_allocate_syncvnode(struct mount *mp)
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
{
struct vnode *vp;
- Complete part of the unfinished bufobj work by consistently using BO_LOCK/UNLOCK/MTX when manipulating the bufobj. - Create a new lock in the bufobj to lock bufobj fields independently. This leaves the vnode interlock as an 'identity' lock while the bufobj is an io lock. The bufobj lock is ordered before the vnode interlock and also before the mnt ilock. - Exploit this new lock order to simplify softdep_check_suspend(). - A few sync related functions are marked with a new XXX to note that we may not properly interlock against a non-zero bv_cnt when attempting to sync all vnodes on a mountlist. I do not believe this race is important. If I'm wrong this will make these locations easier to find. Reviewed by: kib (earlier diff) Tested by: kris, pho (earlier diff)
2008-03-22 09:15:16 +00:00
struct bufobj *bo;
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
static long start, incr, next;
int error;
/* Allocate a new vnode */
Back when VOP_* was introduced, we did not have new-style struct initializations but we did have lofty goals and big ideals. Adjust to more contemporary circumstances and gain type checking. Replace the entire vop_t frobbing thing with properly typed structures. The only casualty is that we can not add a new VOP_ method with a loadable module. History has not given us reason to belive this would ever be feasible in the the first place. Eliminate in toto VOCALL(), vop_t, VNODEOP_SET() etc. Give coda correct prototypes and function definitions for all vop_()s. Generate a bit more data from the vnode_if.src file: a struct vop_vector and protype typedefs for all vop methods. Add a new vop_bypass() and make vop_default be a pointer to another struct vop_vector. Remove a lot of vfs_init since vop_vector is ready to use from the compiler. Cast various vop_mumble() to void * with uppercase name, for instance VOP_PANIC, VOP_NULL etc. Implement VCALL() by making vdesc_offset the offsetof() the relevant function pointer in vop_vector. This is disgusting but since the code is generated by a script comparatively safe. The alternative for nullfs etc. would be much worse. Fix up all vnode method vectors to remove casts so they become typesafe. (The bulk of this is generated by scripts)
2004-12-01 23:16:38 +00:00
if ((error = getnewvnode("syncer", mp, &sync_vnodeops, &vp)) != 0) {
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
mp->mnt_syncer = NULL;
return (error);
}
vp->v_type = VNON;
Introduce the VV_FORCEINSMQ vnode flag. It instructs the insmnque() function to ignore the unmounting and forces insertion of the vnode into the mount vnode list. Change insmntque() to fail when forced unmount is in progress and VV_FORCEINSMQ is not specified. Add an assertion to the insmntque(), requiring the vnode to be exclusively locked for mp-safe filesystems. Use the VV_FORCEINSMQ for the creation of the syncvnode. Tested by: pho Reviewed by: tegge MFC after: 1 month
2008-08-28 09:08:15 +00:00
vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
vp->v_vflag |= VV_FORCEINSMQ;
Make insmntque() externally visibile and allow it to fail (e.g. during late stages of unmount). On failure, the vnode is recycled. Add insmntque1(), to allow for file system specific cleanup when recycling vnode on failure. Change getnewvnode() to no longer call insmntque(). Previously, embryonic vnodes were put onto the list of vnode belonging to a file system, which is unsafe for a file system marked MPSAFE. Change vfs_hash_insert() to no longer lock the vnode. The caller now has that responsibility. Change most file systems to lock the vnode and call insmntque() or insmntque1() after a new vnode has been sufficiently setup. Handle failed insmntque*() calls by propagating errors to callers, possibly after some file system specific cleanup. Approved by: re (kensmith) Reviewed by: kib In collaboration with: kib
2007-03-13 01:50:27 +00:00
error = insmntque(vp, mp);
if (error != 0)
panic("vfs_allocate_syncvnode: insmntque failed");
Introduce the VV_FORCEINSMQ vnode flag. It instructs the insmnque() function to ignore the unmounting and forces insertion of the vnode into the mount vnode list. Change insmntque() to fail when forced unmount is in progress and VV_FORCEINSMQ is not specified. Add an assertion to the insmntque(), requiring the vnode to be exclusively locked for mp-safe filesystems. Use the VV_FORCEINSMQ for the creation of the syncvnode. Tested by: pho Reviewed by: tegge MFC after: 1 month
2008-08-28 09:08:15 +00:00
vp->v_vflag &= ~VV_FORCEINSMQ;
VOP_UNLOCK(vp, 0);
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
/*
* Place the vnode onto the syncer worklist. We attempt to
* scatter them about on the list so that they will go off
* at evenly distributed times even if all the filesystems
* are mounted at once.
*/
next += incr;
if (next == 0 || next > syncer_maxdelay) {
start /= 2;
incr /= 2;
if (start == 0) {
start = syncer_maxdelay / 2;
incr = syncer_maxdelay;
}
next = start;
}
- Complete part of the unfinished bufobj work by consistently using BO_LOCK/UNLOCK/MTX when manipulating the bufobj. - Create a new lock in the bufobj to lock bufobj fields independently. This leaves the vnode interlock as an 'identity' lock while the bufobj is an io lock. The bufobj lock is ordered before the vnode interlock and also before the mnt ilock. - Exploit this new lock order to simplify softdep_check_suspend(). - A few sync related functions are marked with a new XXX to note that we may not properly interlock against a non-zero bv_cnt when attempting to sync all vnodes on a mountlist. I do not believe this race is important. If I'm wrong this will make these locations easier to find. Reviewed by: kib (earlier diff) Tested by: kris, pho (earlier diff)
2008-03-22 09:15:16 +00:00
bo = &vp->v_bufobj;
BO_LOCK(bo);
vn_syncer_add_to_worklist(bo, syncdelay > 0 ? next % syncdelay : 0);
When shutting down the syncer kernel thread, first tell it to run faster and iterate to over its work list a few times in an attempt to empty the work list before the syncer terminates. This leaves fewer dirty blocks to be written at the "syncing disks" stage and keeps the the "giving up on N buffers" problem from being triggered by the presence of a large soft updates work list at system shutdown time. The downside is that the syncer takes noticeably longer to terminate. Tested by: "Arjan van Leeuwen" <avleeuwen AT piwebs DOT com> Approved by: mckusick
2004-07-01 23:59:19 +00:00
/* XXX - vn_syncer_add_to_worklist() also grabs and drops sync_mtx. */
mtx_lock(&sync_mtx);
sync_vnode_count++;
mtx_unlock(&sync_mtx);
- Complete part of the unfinished bufobj work by consistently using BO_LOCK/UNLOCK/MTX when manipulating the bufobj. - Create a new lock in the bufobj to lock bufobj fields independently. This leaves the vnode interlock as an 'identity' lock while the bufobj is an io lock. The bufobj lock is ordered before the vnode interlock and also before the mnt ilock. - Exploit this new lock order to simplify softdep_check_suspend(). - A few sync related functions are marked with a new XXX to note that we may not properly interlock against a non-zero bv_cnt when attempting to sync all vnodes on a mountlist. I do not believe this race is important. If I'm wrong this will make these locations easier to find. Reviewed by: kib (earlier diff) Tested by: kris, pho (earlier diff)
2008-03-22 09:15:16 +00:00
BO_UNLOCK(bo);
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
mp->mnt_syncer = vp;
return (0);
}
/*
* Do a lazy sync of the filesystem.
*/
Staticize.
1998-12-21 23:38:33 +00:00
static int
Convert remaining functions in vfs_subr.c from K&R prototypes to ANSI C prototypes, as the majority of new functions added have been in this style. Changing prototype style now results in gcc noticing that the implementation of vn_pollrecord() has a 'short' argument instead of 'int' as prototyped in vnode.h, so correct that definition. In practice this didn't matter as only poll flags in the lower 16 bits are used. MFC after: 1 week
2006-01-21 19:42:10 +00:00
sync_fsync(struct vop_fsync_args *ap)
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
{
struct vnode *syncvp = ap->a_vp;
struct mount *mp = syncvp->v_mount;
Add mnt_noasync counter to better handle interleaved calls to nmount(), sync() and sync_fsync() without losing MNT_ASYNC. Add MNTK_ASYNC flag which is set only when MNT_ASYNC is set and mnt_noasync is zero, and check that flag instead of MNT_ASYNC before initiating async io.
2006-09-26 04:15:59 +00:00
int error;
Move the syncer linkage from vnode to bufobj. This is not quite a perfect separation: the syncer still think it knows that everything is a vnode.
2004-10-27 08:05:02 +00:00
struct bufobj *bo;
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
/*
* We only need to do something if this is a lazy evaluation.
*/
if (ap->a_waitfor != MNT_LAZY)
return (0);
/*
* Move ourselves to the back of the sync list.
*/
Move the syncer linkage from vnode to bufobj. This is not quite a perfect separation: the syncer still think it knows that everything is a vnode.
2004-10-27 08:05:02 +00:00
bo = &syncvp->v_bufobj;
BO_LOCK(bo);
vn_syncer_add_to_worklist(bo, syncdelay);
BO_UNLOCK(bo);
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
/*
* Walk the list of vnodes pushing all that are dirty and
* not already on the sync list.
*/
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_lock(&mountlist_mtx);
Improve VFS locking: - Implement real draining for vfs consumers by not relying on the mnt_lock and using instead a refcount in order to keep track of lock requesters. - Due to the change above, remove the mnt_lock lockmgr because it is now useless. - Due to the change above, vfs_busy() is no more linked to a lockmgr. Change so its KPI by removing the interlock argument and defining 2 new flags for it: MBF_NOWAIT which basically replaces the LK_NOWAIT of the old version (which was unlinked from the lockmgr alredy) and MBF_MNTLSTLOCK which provides the ability to drop the mountlist_mtx once the mnt interlock is held (ability still desired by most consumers). - The stub used into vfs_mount_destroy(), that allows to override the mnt_ref if running for more than 3 seconds, make it totally useless. Remove it as it was thought to work into older versions. If a problem of "refcount held never going away" should appear, we will need to fix properly instead than trust on such hackish solution. - Fix a bug where returning (with an error) from dounmount() was still leaving the MNTK_MWAIT flag on even if it the waiters were actually woken up. Just a place in vfs_mount_destroy() is left because it is going to recycle the structure in any case, so it doesn't matter. - Remove the markercnt refcount as it is useless. This patch modifies VFS ABI and breaks KPI for vfs_busy() so manpages and __FreeBSD_version will be modified accordingly. Discussed with: kib Tested by: pho
2008-11-02 10:15:42 +00:00
if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) {
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_unlock(&mountlist_mtx);
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
return (0);
Unlock mountlist_slock if the mount point was busy (unmount in progress) during the attempt at lazy fsync.
1998-04-15 18:37:49 +00:00
}
Add snapshots to the fast filesystem. Most of the changes support the gating of system calls that cause modifications to the underlying filesystem. The gating can be enabled by any filesystem that needs to consistently suspend operations by adding the vop_stdgetwritemount to their set of vnops. Once gating is enabled, the function vfs_write_suspend stops all new write operations to a filesystem, allows any filesystem modifying system calls already in progress to complete, then sync's the filesystem to disk and returns. The function vfs_write_resume allows the suspended write operations to begin again. Gating is not added by default for all filesystems as for SMP systems it adds two extra locks to such critical kernel paths as the write system call. Thus, gating should only be added as needed. Details on the use and current status of snapshots in FFS can be found in /sys/ufs/ffs/README.snapshot so for brevity and timelyness is not included here. Unless and until you create a snapshot file, these changes should have no effect on your system (famous last words).
2000-07-11 22:07:57 +00:00
if (vn_start_write(NULL, &mp, V_NOWAIT) != 0) {
Decontextualize vfs_busy(), vfs_unbusy() and vfs_mount_alloc() functions. Manpages are updated accordingly. Tested by: Diego Sardina <siarodx at gmail dot com>
2008-08-31 14:26:08 +00:00
vfs_unbusy(mp);
Add snapshots to the fast filesystem. Most of the changes support the gating of system calls that cause modifications to the underlying filesystem. The gating can be enabled by any filesystem that needs to consistently suspend operations by adding the vop_stdgetwritemount to their set of vnops. Once gating is enabled, the function vfs_write_suspend stops all new write operations to a filesystem, allows any filesystem modifying system calls already in progress to complete, then sync's the filesystem to disk and returns. The function vfs_write_resume allows the suspended write operations to begin again. Gating is not added by default for all filesystems as for SMP systems it adds two extra locks to such critical kernel paths as the write system call. Thus, gating should only be added as needed. Details on the use and current status of snapshots in FFS can be found in /sys/ufs/ffs/README.snapshot so for brevity and timelyness is not included here. Unless and until you create a snapshot file, these changes should have no effect on your system (famous last words).
2000-07-11 22:07:57 +00:00
return (0);
}
Use mount interlock to protect all changes to mnt_flag and mnt_kern_flag. This eliminates a race where MNT_UPDATE flag could be lost when nmount() raced against sync(), sync_fsync() or quotactl().
2006-09-26 04:12:49 +00:00
MNT_ILOCK(mp);
Add mnt_noasync counter to better handle interleaved calls to nmount(), sync() and sync_fsync() without losing MNT_ASYNC. Add MNTK_ASYNC flag which is set only when MNT_ASYNC is set and mnt_noasync is zero, and check that flag instead of MNT_ASYNC before initiating async io.
2006-09-26 04:15:59 +00:00
mp->mnt_noasync++;
mp->mnt_kern_flag &= ~MNTK_ASYNC;
Use mount interlock to protect all changes to mnt_flag and mnt_kern_flag. This eliminates a race where MNT_UPDATE flag could be lost when nmount() raced against sync(), sync_fsync() or quotactl().
2006-09-26 04:12:49 +00:00
MNT_IUNLOCK(mp);
When the softdep conversion took place, the periodic vfs_msync() from update got lost. This is responsible for ensuring that dirty mmap() pages get periodically written to disk. Without it, long time mmap's might not have their dirty pages written out at all of the system crashes or isn't cleanly shut down. This could be nasty if you've got a long-running writing via mmap(), dirty pages used to get written to disk within 30 seconds or so.
1998-04-16 03:31:26 +00:00
vfs_msync(mp, MNT_NOWAIT);
Decontextualize vfs_busy(), vfs_unbusy() and vfs_mount_alloc() functions. Manpages are updated accordingly. Tested by: Diego Sardina <siarodx at gmail dot com>
2008-08-31 14:26:08 +00:00
error = VFS_SYNC(mp, MNT_LAZY, ap->a_td);
Use mount interlock to protect all changes to mnt_flag and mnt_kern_flag. This eliminates a race where MNT_UPDATE flag could be lost when nmount() raced against sync(), sync_fsync() or quotactl().
2006-09-26 04:12:49 +00:00
MNT_ILOCK(mp);
Add mnt_noasync counter to better handle interleaved calls to nmount(), sync() and sync_fsync() without losing MNT_ASYNC. Add MNTK_ASYNC flag which is set only when MNT_ASYNC is set and mnt_noasync is zero, and check that flag instead of MNT_ASYNC before initiating async io.
2006-09-26 04:15:59 +00:00
mp->mnt_noasync--;
if ((mp->mnt_flag & MNT_ASYNC) != 0 && mp->mnt_noasync == 0)
mp->mnt_kern_flag |= MNTK_ASYNC;
Use mount interlock to protect all changes to mnt_flag and mnt_kern_flag. This eliminates a race where MNT_UPDATE flag could be lost when nmount() raced against sync(), sync_fsync() or quotactl().
2006-09-26 04:12:49 +00:00
MNT_IUNLOCK(mp);
Add snapshots to the fast filesystem. Most of the changes support the gating of system calls that cause modifications to the underlying filesystem. The gating can be enabled by any filesystem that needs to consistently suspend operations by adding the vop_stdgetwritemount to their set of vnops. Once gating is enabled, the function vfs_write_suspend stops all new write operations to a filesystem, allows any filesystem modifying system calls already in progress to complete, then sync's the filesystem to disk and returns. The function vfs_write_resume allows the suspended write operations to begin again. Gating is not added by default for all filesystems as for SMP systems it adds two extra locks to such critical kernel paths as the write system call. Thus, gating should only be added as needed. Details on the use and current status of snapshots in FFS can be found in /sys/ufs/ffs/README.snapshot so for brevity and timelyness is not included here. Unless and until you create a snapshot file, these changes should have no effect on your system (famous last words).
2000-07-11 22:07:57 +00:00
vn_finished_write(mp);
Decontextualize vfs_busy(), vfs_unbusy() and vfs_mount_alloc() functions. Manpages are updated accordingly. Tested by: Diego Sardina <siarodx at gmail dot com>
2008-08-31 14:26:08 +00:00
vfs_unbusy(mp);
Within ufs, the ffs_sync and ffs_fsync functions did not always check for and/or report I/O errors. The result is that a VFS_SYNC or VOP_FSYNC called with MNT_WAIT could loop infinitely on ufs in the presence of a hard error writing a disk sector or in a filesystem full condition. This patch ensures that I/O errors will always be checked and returned. This patch also ensures that every call to VFS_SYNC or VOP_FSYNC with MNT_WAIT set checks for and takes appropriate action when an error is returned. Sponsored by: DARPA & NAI Labs.
2002-10-25 00:20:37 +00:00
return (error);
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
}
/*
* The syncer vnode is no referenced.
*/
Staticize.
1998-12-21 23:38:33 +00:00
static int
Convert remaining functions in vfs_subr.c from K&R prototypes to ANSI C prototypes, as the majority of new functions added have been in this style. Changing prototype style now results in gcc noticing that the implementation of vn_pollrecord() has a 'short' argument instead of 'int' as prototyped in vnode.h, so correct that definition. In practice this didn't matter as only poll flags in the lower 16 bits are used. MFC after: 1 week
2006-01-21 19:42:10 +00:00
sync_inactive(struct vop_inactive_args *ap)
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
{
vgone(ap->a_vp);
return (0);
}
/*
* The syncer vnode is no longer needed and is being decommissioned.
Protect vn worklist and vn->v_{clean,dirty}blkhd at splbio(). Get rid of extra LIST_REMOVE() Reviewed by: hsu@FreeBSD.ORG (Jeffrey Hsu), mckusick@McKusick.COM Submitted by: hsu@FreeBSD.ORG (Jeffrey Hsu), dillon@backplane.com ( Matthew Dillon )
1999-02-19 17:36:58 +00:00
*
- Remove spls(). The locking that has replaced them is in place and they no longer serve as guidelines for future work.
2003-09-19 23:52:06 +00:00
* Modifications to the worklist must be protected by sync_mtx.
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
*/
Staticize.
1998-12-21 23:38:33 +00:00
static int
Convert remaining functions in vfs_subr.c from K&R prototypes to ANSI C prototypes, as the majority of new functions added have been in this style. Changing prototype style now results in gcc noticing that the implementation of vn_pollrecord() has a 'short' argument instead of 'int' as prototyped in vnode.h, so correct that definition. In practice this didn't matter as only poll flags in the lower 16 bits are used. MFC after: 1 week
2006-01-21 19:42:10 +00:00
sync_reclaim(struct vop_reclaim_args *ap)
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
{
struct vnode *vp = ap->a_vp;
Move the syncer linkage from vnode to bufobj. This is not quite a perfect separation: the syncer still think it knows that everything is a vnode.
2004-10-27 08:05:02 +00:00
struct bufobj *bo;
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
Move the syncer linkage from vnode to bufobj. This is not quite a perfect separation: the syncer still think it knows that everything is a vnode.
2004-10-27 08:05:02 +00:00
bo = &vp->v_bufobj;
- Complete part of the unfinished bufobj work by consistently using BO_LOCK/UNLOCK/MTX when manipulating the bufobj. - Create a new lock in the bufobj to lock bufobj fields independently. This leaves the vnode interlock as an 'identity' lock while the bufobj is an io lock. The bufobj lock is ordered before the vnode interlock and also before the mnt ilock. - Exploit this new lock order to simplify softdep_check_suspend(). - A few sync related functions are marked with a new XXX to note that we may not properly interlock against a non-zero bv_cnt when attempting to sync all vnodes on a mountlist. I do not believe this race is important. If I'm wrong this will make these locations easier to find. Reviewed by: kib (earlier diff) Tested by: kris, pho (earlier diff)
2008-03-22 09:15:16 +00:00
BO_LOCK(bo);
- Remove spls(). The locking that has replaced them is in place and they no longer serve as guidelines for future work.
2003-09-19 23:52:06 +00:00
vp->v_mount->mnt_syncer = NULL;
Move the syncer linkage from vnode to bufobj. This is not quite a perfect separation: the syncer still think it knows that everything is a vnode.
2004-10-27 08:05:02 +00:00
if (bo->bo_flag & BO_ONWORKLST) {
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
mtx_lock(&sync_mtx);
Move the syncer linkage from vnode to bufobj. This is not quite a perfect separation: the syncer still think it knows that everything is a vnode.
2004-10-27 08:05:02 +00:00
LIST_REMOVE(bo, bo_synclist);
Minor style cleanups (mostly removal of trailing whitespaces).
2007-04-10 15:29:37 +00:00
syncer_worklist_len--;
When shutting down the syncer kernel thread, first tell it to run faster and iterate to over its work list a few times in an attempt to empty the work list before the syncer terminates. This leaves fewer dirty blocks to be written at the "syncing disks" stage and keeps the the "giving up on N buffers" problem from being triggered by the presence of a large soft updates work list at system shutdown time. The downside is that the syncer takes noticeably longer to terminate. Tested by: "Arjan van Leeuwen" <avleeuwen AT piwebs DOT com> Approved by: mckusick
2004-07-01 23:59:19 +00:00
sync_vnode_count--;
- Lock down the syncer with sync_mtx. - Enable vfs_badlock_mutex by default. - Assert that the vp is locked in VOP_UNLOCK. - Use standard interlock macros in remaining code. - Correct a race in getnewvnode(). - Lock access to v_numoutput with interlock. - Lock access to buf lists and splay tree with interlock. - Add VOP and VI asserts. - Lock b_vnbufs with the vnode interlock. - Add vrefcnt() for callers who want to retreive the vnode ref without holding a lock. Add a comment that describes when this is safe. - Add vholdl() and vdropl() so that callers who already own the interlock can avoid race conditions and unnecessary unlocking. - Move the VOP_GETATTR() in vflush() into the WRITECLOSE conditional case. - Hold the interlock before droping the mntlist_mtx in vflush() to avoid a race. - Fix locking in vfs_msync().
2002-09-25 02:22:21 +00:00
mtx_unlock(&sync_mtx);
Move the syncer linkage from vnode to bufobj. This is not quite a perfect separation: the syncer still think it knows that everything is a vnode.
2004-10-27 08:05:02 +00:00
bo->bo_flag &= ~BO_ONWORKLST;
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
}
- Complete part of the unfinished bufobj work by consistently using BO_LOCK/UNLOCK/MTX when manipulating the bufobj. - Create a new lock in the bufobj to lock bufobj fields independently. This leaves the vnode interlock as an 'identity' lock while the bufobj is an io lock. The bufobj lock is ordered before the vnode interlock and also before the mnt ilock. - Exploit this new lock order to simplify softdep_check_suspend(). - A few sync related functions are marked with a new XXX to note that we may not properly interlock against a non-zero bv_cnt when attempting to sync all vnodes on a mountlist. I do not believe this race is important. If I'm wrong this will make these locations easier to find. Reviewed by: kib (earlier diff) Tested by: kris, pho (earlier diff)
2008-03-22 09:15:16 +00:00
BO_UNLOCK(bo);
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman) Submitted by: Kirk McKusick (mcKusick@mckusick.com) Obtained from: WHistle development tree
1998-03-08 09:59:44 +00:00
return (0);
}
Introduce vn_isdisk(struct vnode *vp) function, and use it to test for diskness.
1999-08-25 12:24:39 +00:00
/*
* Check if vnode represents a disk device
*/
int
Convert remaining functions in vfs_subr.c from K&R prototypes to ANSI C prototypes, as the majority of new functions added have been in this style. Changing prototype style now results in gcc noticing that the implementation of vn_pollrecord() has a 'short' argument instead of 'int' as prototyped in vnode.h, so correct that definition. In practice this didn't matter as only poll flags in the lower 16 bits are used. MFC after: 1 week
2006-01-21 19:42:10 +00:00
vn_isdisk(struct vnode *vp, int *errp)
Introduce vn_isdisk(struct vnode *vp) function, and use it to test for diskness.
1999-08-25 12:24:39 +00:00
{
Simplify vn_isdisk() a bit.
2003-10-12 14:04:39 +00:00
int error;
Move extern declaration of dead_vnodeop_p to a .h file. Remove race condition in vn_isdisk().
2000-09-05 21:09:56 +00:00
Simplify vn_isdisk() a bit.
2003-10-12 14:04:39 +00:00
error = 0;
Hold dev_lock and check for NULL devsw pointer when we determine if a vnode is a disk.
2004-09-24 06:16:08 +00:00
dev_lock();
Simplify vn_isdisk() a bit.
2003-10-12 14:04:39 +00:00
if (vp->v_type != VCHR)
error = ENOTBLK;
else if (vp->v_rdev == NULL)
error = ENXIO;
Hold dev_lock and check for NULL devsw pointer when we determine if a vnode is a disk.
2004-09-24 06:16:08 +00:00
else if (vp->v_rdev->si_devsw == NULL)
error = ENXIO;
else if (!(vp->v_rdev->si_devsw->d_flags & D_DISK))
Simplify vn_isdisk() a bit.
2003-10-12 14:04:39 +00:00
error = ENOTBLK;
Hold dev_lock and check for NULL devsw pointer when we determine if a vnode is a disk.
2004-09-24 06:16:08 +00:00
dev_unlock();
Give vn_isdisk() a second argument where it can return a suitable errno. Suggested by: bde
2000-01-10 12:04:27 +00:00
if (errp != NULL)
Simplify vn_isdisk() a bit.
2003-10-12 14:04:39 +00:00
*errp = error;
return (error == 0);
Introduce vn_isdisk(struct vnode *vp) function, and use it to test for diskness.
1999-08-25 12:24:39 +00:00
}
o Add additional comment describing vaccess() behavior. Requested by: eivind Reviewed by: eivind, adrian
2000-09-20 17:18:12 +00:00
/*
More s/file system/filesystem/g
2002-05-16 21:28:32 +00:00
* Common filesystem object access control check routine. Accepts a
o Add additional comment describing vaccess() behavior. Requested by: eivind Reviewed by: eivind, adrian
2000-09-20 17:18:12 +00:00
* vnode's type, "mode", uid and gid, requested access mode, credentials,
* and optional call-by-reference privused argument allowing vaccess()
* to indicate to the caller whether privilege was used to satisfy the
Note that the privilege indicating flag to vaccess() originally used by the process accounting system is now deprecated.
2002-07-31 02:05:12 +00:00
* request (obsoleted). Returns 0 on success, or an errno on failure.
Improve commenting of vaccess(), making sure to be clear that the ifdef capabilities code is there for reference and never actually used. Slight style tweak.
2006-08-06 10:43:35 +00:00
*
* The ifdef'd CAPABILITIES version is here for reference, but is not
* actually used.
o Add additional comment describing vaccess() behavior. Requested by: eivind Reviewed by: eivind, adrian
2000-09-20 17:18:12 +00:00
*/
Centralize the canonical vop_access user/group/other check in vaccess(). Discussed with: bde
2000-08-20 08:36:26 +00:00
int
Convert remaining functions in vfs_subr.c from K&R prototypes to ANSI C prototypes, as the majority of new functions added have been in this style. Changing prototype style now results in gcc noticing that the implementation of vn_pollrecord() has a 'short' argument instead of 'int' as prototyped in vnode.h, so correct that definition. In practice this didn't matter as only poll flags in the lower 16 bits are used. MFC after: 1 week
2006-01-21 19:42:10 +00:00
vaccess(enum vtype type, mode_t file_mode, uid_t file_uid, gid_t file_gid,
Introduce accmode_t. This is required for NFSv4 ACLs - it will be neccessary to add more V* constants, and the variables changed by this patch were often being assigned to mode_t variables, which is 16 bit. Approved by: rwatson (mentor)
2008-10-28 13:44:11 +00:00
accmode_t accmode, struct ucred *cred, int *privused)
Centralize the canonical vop_access user/group/other check in vaccess(). Discussed with: bde
2000-08-20 08:36:26 +00:00
{
Introduce accmode_t. This is required for NFSv4 ACLs - it will be neccessary to add more V* constants, and the variables changed by this patch were often being assigned to mode_t variables, which is 16 bit. Approved by: rwatson (mentor)
2008-10-28 13:44:11 +00:00
accmode_t dac_granted;
accmode_t priv_granted;
Centralize the canonical vop_access user/group/other check in vaccess(). Discussed with: bde
2000-08-20 08:36:26 +00:00
/*
o Restructure vaccess() so as to check for DAC permission to modify the object before falling back on privilege. Make vaccess() accept an additional optional argument, privused, to determine whether privilege was required for vaccess() to return 0. Add commented out capability checks for reference. Rename some variables to make it more clear which modes/uids/etc are associated with the object, and which with the access mode. o Update file system use of vaccess() to pass NULL as the optional privused argument. Once additional patches are applied, suser() will no longer set ASU, so privused will permit passing of privilege information up the stack to the caller. Reviewed by: bde, green, phk, -security, others Obtained from: TrustedBSD Project
2000-08-29 14:45:49 +00:00
* Look for a normal, non-privileged way to access the file/directory
* as requested. If it exists, go with that.
Centralize the canonical vop_access user/group/other check in vaccess(). Discussed with: bde
2000-08-20 08:36:26 +00:00
*/
o Restructure vaccess() so as to check for DAC permission to modify the object before falling back on privilege. Make vaccess() accept an additional optional argument, privused, to determine whether privilege was required for vaccess() to return 0. Add commented out capability checks for reference. Rename some variables to make it more clear which modes/uids/etc are associated with the object, and which with the access mode. o Update file system use of vaccess() to pass NULL as the optional privused argument. Once additional patches are applied, suser() will no longer set ASU, so privused will permit passing of privilege information up the stack to the caller. Reviewed by: bde, green, phk, -security, others Obtained from: TrustedBSD Project
2000-08-29 14:45:49 +00:00
if (privused != NULL)
*privused = 0;
dac_granted = 0;
/* Check the owner. */
if (cred->cr_uid == file_uid) {
o Introduce new VOP_ACCESS() flag VADMIN, allowing file systems to perform "administrative" authorization checks. In most cases, the VADMIN test checks to make sure the credential effective uid is the same as the file owner. o Modify vaccess() to set VADMIN as an available right if the uid is appropriate. o Modify references to uid-based access control operations such that they now always invoke VOP_ACCESS() instead of using hard-coded policy checks. o This allows alternative UFS policies to be implemented by replacing only ufs_access() (such as mandatory system policies). o VOP_ACCESS() requires the caller to hold an exclusive vnode lock on the vnode: I believe that new invocations of VOP_ACCESS() are always called with the lock held. o Some direct checks of the uid remain, largely associated with the QUOTA and SUIDDIR code. Reviewed by: eivind Obtained from: TrustedBSD Project
2000-10-19 07:53:59 +00:00
dac_granted |= VADMIN;
o Restructure vaccess() so as to check for DAC permission to modify the object before falling back on privilege. Make vaccess() accept an additional optional argument, privused, to determine whether privilege was required for vaccess() to return 0. Add commented out capability checks for reference. Rename some variables to make it more clear which modes/uids/etc are associated with the object, and which with the access mode. o Update file system use of vaccess() to pass NULL as the optional privused argument. Once additional patches are applied, suser() will no longer set ASU, so privused will permit passing of privilege information up the stack to the caller. Reviewed by: bde, green, phk, -security, others Obtained from: TrustedBSD Project
2000-08-29 14:45:49 +00:00
if (file_mode & S_IXUSR)
dac_granted |= VEXEC;
if (file_mode & S_IRUSR)
dac_granted |= VREAD;
if (file_mode & S_IWUSR)
Teach discretionary access control methods for files about VAPPEND and VALLPERM. Obtained from: TrustedBSD Project Sponsored by: DARPA, NAI Labs
2002-07-22 03:57:07 +00:00
dac_granted |= (VWRITE | VAPPEND);
o Restructure vaccess() so as to check for DAC permission to modify the object before falling back on privilege. Make vaccess() accept an additional optional argument, privused, to determine whether privilege was required for vaccess() to return 0. Add commented out capability checks for reference. Rename some variables to make it more clear which modes/uids/etc are associated with the object, and which with the access mode. o Update file system use of vaccess() to pass NULL as the optional privused argument. Once additional patches are applied, suser() will no longer set ASU, so privused will permit passing of privilege information up the stack to the caller. Reviewed by: bde, green, phk, -security, others Obtained from: TrustedBSD Project
2000-08-29 14:45:49 +00:00
Introduce accmode_t. This is required for NFSv4 ACLs - it will be neccessary to add more V* constants, and the variables changed by this patch were often being assigned to mode_t variables, which is 16 bit. Approved by: rwatson (mentor)
2008-10-28 13:44:11 +00:00
if ((accmode & dac_granted) == accmode)
o Restructure vaccess() so as to check for DAC permission to modify the object before falling back on privilege. Make vaccess() accept an additional optional argument, privused, to determine whether privilege was required for vaccess() to return 0. Add commented out capability checks for reference. Rename some variables to make it more clear which modes/uids/etc are associated with the object, and which with the access mode. o Update file system use of vaccess() to pass NULL as the optional privused argument. Once additional patches are applied, suser() will no longer set ASU, so privused will permit passing of privilege information up the stack to the caller. Reviewed by: bde, green, phk, -security, others Obtained from: TrustedBSD Project
2000-08-29 14:45:49 +00:00
return (0);
goto privcheck;
}
/* Otherwise, check the groups (first match) */
if (groupmember(file_gid, cred)) {
if (file_mode & S_IXGRP)
dac_granted |= VEXEC;
if (file_mode & S_IRGRP)
dac_granted |= VREAD;
if (file_mode & S_IWGRP)
Teach discretionary access control methods for files about VAPPEND and VALLPERM. Obtained from: TrustedBSD Project Sponsored by: DARPA, NAI Labs
2002-07-22 03:57:07 +00:00
dac_granted |= (VWRITE | VAPPEND);
o Restructure vaccess() so as to check for DAC permission to modify the object before falling back on privilege. Make vaccess() accept an additional optional argument, privused, to determine whether privilege was required for vaccess() to return 0. Add commented out capability checks for reference. Rename some variables to make it more clear which modes/uids/etc are associated with the object, and which with the access mode. o Update file system use of vaccess() to pass NULL as the optional privused argument. Once additional patches are applied, suser() will no longer set ASU, so privused will permit passing of privilege information up the stack to the caller. Reviewed by: bde, green, phk, -security, others Obtained from: TrustedBSD Project
2000-08-29 14:45:49 +00:00
Introduce accmode_t. This is required for NFSv4 ACLs - it will be neccessary to add more V* constants, and the variables changed by this patch were often being assigned to mode_t variables, which is 16 bit. Approved by: rwatson (mentor)
2008-10-28 13:44:11 +00:00
if ((accmode & dac_granted) == accmode)
o Restructure vaccess() so as to check for DAC permission to modify the object before falling back on privilege. Make vaccess() accept an additional optional argument, privused, to determine whether privilege was required for vaccess() to return 0. Add commented out capability checks for reference. Rename some variables to make it more clear which modes/uids/etc are associated with the object, and which with the access mode. o Update file system use of vaccess() to pass NULL as the optional privused argument. Once additional patches are applied, suser() will no longer set ASU, so privused will permit passing of privilege information up the stack to the caller. Reviewed by: bde, green, phk, -security, others Obtained from: TrustedBSD Project
2000-08-29 14:45:49 +00:00
return (0);
goto privcheck;
}
/* Otherwise, check everyone else. */
if (file_mode & S_IXOTH)
dac_granted |= VEXEC;
if (file_mode & S_IROTH)
dac_granted |= VREAD;
if (file_mode & S_IWOTH)
Teach discretionary access control methods for files about VAPPEND and VALLPERM. Obtained from: TrustedBSD Project Sponsored by: DARPA, NAI Labs
2002-07-22 03:57:07 +00:00
dac_granted |= (VWRITE | VAPPEND);
Introduce accmode_t. This is required for NFSv4 ACLs - it will be neccessary to add more V* constants, and the variables changed by this patch were often being assigned to mode_t variables, which is 16 bit. Approved by: rwatson (mentor)
2008-10-28 13:44:11 +00:00
if ((accmode & dac_granted) == accmode)
o Restructure vaccess() so as to check for DAC permission to modify the object before falling back on privilege. Make vaccess() accept an additional optional argument, privused, to determine whether privilege was required for vaccess() to return 0. Add commented out capability checks for reference. Rename some variables to make it more clear which modes/uids/etc are associated with the object, and which with the access mode. o Update file system use of vaccess() to pass NULL as the optional privused argument. Once additional patches are applied, suser() will no longer set ASU, so privused will permit passing of privilege information up the stack to the caller. Reviewed by: bde, green, phk, -security, others Obtained from: TrustedBSD Project
2000-08-29 14:45:49 +00:00
return (0);
privcheck:
/*
Sweep kernel replacing suser(9) calls with priv(9) calls, assigning specific privilege names to a broad range of privileges. These may require some future tweaking. Sponsored by: nCircle Network Security, Inc. Obtained from: TrustedBSD Project Discussed on: arch@ Reviewed (at least in part) by: mlaier, jmg, pjd, bde, ceri, Alex Lyashkov <umka at sevcity dot net>, Skip Ford <skip dot ford at verizon dot net>, Antoine Brodin <antoine dot brodin at laposte dot net>
2006-11-06 13:42:10 +00:00
* Build a privilege mask to determine if the set of privileges
o Restructure vaccess() so as to check for DAC permission to modify the object before falling back on privilege. Make vaccess() accept an additional optional argument, privused, to determine whether privilege was required for vaccess() to return 0. Add commented out capability checks for reference. Rename some variables to make it more clear which modes/uids/etc are associated with the object, and which with the access mode. o Update file system use of vaccess() to pass NULL as the optional privused argument. Once additional patches are applied, suser() will no longer set ASU, so privused will permit passing of privilege information up the stack to the caller. Reviewed by: bde, green, phk, -security, others Obtained from: TrustedBSD Project
2000-08-29 14:45:49 +00:00
* satisfies the requirements when combined with the granted mask
Sweep kernel replacing suser(9) calls with priv(9) calls, assigning specific privilege names to a broad range of privileges. These may require some future tweaking. Sponsored by: nCircle Network Security, Inc. Obtained from: TrustedBSD Project Discussed on: arch@ Reviewed (at least in part) by: mlaier, jmg, pjd, bde, ceri, Alex Lyashkov <umka at sevcity dot net>, Skip Ford <skip dot ford at verizon dot net>, Antoine Brodin <antoine dot brodin at laposte dot net>
2006-11-06 13:42:10 +00:00
* from above. For each privilege, if the privilege is required,
* bitwise or the request type onto the priv_granted mask.
o Restructure vaccess() so as to check for DAC permission to modify the object before falling back on privilege. Make vaccess() accept an additional optional argument, privused, to determine whether privilege was required for vaccess() to return 0. Add commented out capability checks for reference. Rename some variables to make it more clear which modes/uids/etc are associated with the object, and which with the access mode. o Update file system use of vaccess() to pass NULL as the optional privused argument. Once additional patches are applied, suser() will no longer set ASU, so privused will permit passing of privilege information up the stack to the caller. Reviewed by: bde, green, phk, -security, others Obtained from: TrustedBSD Project
2000-08-29 14:45:49 +00:00
*/
Sweep kernel replacing suser(9) calls with priv(9) calls, assigning specific privilege names to a broad range of privileges. These may require some future tweaking. Sponsored by: nCircle Network Security, Inc. Obtained from: TrustedBSD Project Discussed on: arch@ Reviewed (at least in part) by: mlaier, jmg, pjd, bde, ceri, Alex Lyashkov <umka at sevcity dot net>, Skip Ford <skip dot ford at verizon dot net>, Antoine Brodin <antoine dot brodin at laposte dot net>
2006-11-06 13:42:10 +00:00
priv_granted = 0;
Merge from POSIX.1e Capabilities development tree: o POSIX.1e capabilities authorize overriding of VEXEC for VDIR based on CAP_DAC_READ_SEARCH, but of !VDIR based on CAP_DAC_EXECUTE. Add appropriate conditionals to vaccess() to take that into account. o Synchronization cap_check_xxx() -> cap_check() change. Obtained from: TrustedBSD Project
2001-11-02 15:16:59 +00:00
if (type == VDIR) {
/*
Sweep kernel replacing suser(9) calls with priv(9) calls, assigning specific privilege names to a broad range of privileges. These may require some future tweaking. Sponsored by: nCircle Network Security, Inc. Obtained from: TrustedBSD Project Discussed on: arch@ Reviewed (at least in part) by: mlaier, jmg, pjd, bde, ceri, Alex Lyashkov <umka at sevcity dot net>, Skip Ford <skip dot ford at verizon dot net>, Antoine Brodin <antoine dot brodin at laposte dot net>
2006-11-06 13:42:10 +00:00
* For directories, use PRIV_VFS_LOOKUP to satisfy VEXEC
* requests, instead of PRIV_VFS_EXEC.
Merge from POSIX.1e Capabilities development tree: o POSIX.1e capabilities authorize overriding of VEXEC for VDIR based on CAP_DAC_READ_SEARCH, but of !VDIR based on CAP_DAC_EXECUTE. Add appropriate conditionals to vaccess() to take that into account. o Synchronization cap_check_xxx() -> cap_check() change. Obtained from: TrustedBSD Project
2001-11-02 15:16:59 +00:00
*/
Introduce accmode_t. This is required for NFSv4 ACLs - it will be neccessary to add more V* constants, and the variables changed by this patch were often being assigned to mode_t variables, which is 16 bit. Approved by: rwatson (mentor)
2008-10-28 13:44:11 +00:00
if ((accmode & VEXEC) && ((dac_granted & VEXEC) == 0) &&
Eliminate now-unused SUSER_ALLOWJAIL arguments to priv_check_cred(); in some cases, move to priv_check() if it was an operation on a thread and no other flags were present. Eliminate caller-side jail exception checking (also now-unused); jail privilege exception code now goes solely in kern_jail.c. We can't yet eliminate suser() due to some cases in the KAME code where a privilege check is performed and then used in many different deferred paths. Do, however, move those prototypes to priv.h. Reviewed by: csjp Obtained from: TrustedBSD Project
2007-06-12 00:12:01 +00:00
!priv_check_cred(cred, PRIV_VFS_LOOKUP, 0))
Sweep kernel replacing suser(9) calls with priv(9) calls, assigning specific privilege names to a broad range of privileges. These may require some future tweaking. Sponsored by: nCircle Network Security, Inc. Obtained from: TrustedBSD Project Discussed on: arch@ Reviewed (at least in part) by: mlaier, jmg, pjd, bde, ceri, Alex Lyashkov <umka at sevcity dot net>, Skip Ford <skip dot ford at verizon dot net>, Antoine Brodin <antoine dot brodin at laposte dot net>
2006-11-06 13:42:10 +00:00
priv_granted |= VEXEC;
Merge from POSIX.1e Capabilities development tree: o POSIX.1e capabilities authorize overriding of VEXEC for VDIR based on CAP_DAC_READ_SEARCH, but of !VDIR based on CAP_DAC_EXECUTE. Add appropriate conditionals to vaccess() to take that into account. o Synchronization cap_check_xxx() -> cap_check() change. Obtained from: TrustedBSD Project
2001-11-02 15:16:59 +00:00
} else {
Introduce accmode_t. This is required for NFSv4 ACLs - it will be neccessary to add more V* constants, and the variables changed by this patch were often being assigned to mode_t variables, which is 16 bit. Approved by: rwatson (mentor)
2008-10-28 13:44:11 +00:00
if ((accmode & VEXEC) && ((dac_granted & VEXEC) == 0) &&
Eliminate now-unused SUSER_ALLOWJAIL arguments to priv_check_cred(); in some cases, move to priv_check() if it was an operation on a thread and no other flags were present. Eliminate caller-side jail exception checking (also now-unused); jail privilege exception code now goes solely in kern_jail.c. We can't yet eliminate suser() due to some cases in the KAME code where a privilege check is performed and then used in many different deferred paths. Do, however, move those prototypes to priv.h. Reviewed by: csjp Obtained from: TrustedBSD Project
2007-06-12 00:12:01 +00:00
!priv_check_cred(cred, PRIV_VFS_EXEC, 0))
Sweep kernel replacing suser(9) calls with priv(9) calls, assigning specific privilege names to a broad range of privileges. These may require some future tweaking. Sponsored by: nCircle Network Security, Inc. Obtained from: TrustedBSD Project Discussed on: arch@ Reviewed (at least in part) by: mlaier, jmg, pjd, bde, ceri, Alex Lyashkov <umka at sevcity dot net>, Skip Ford <skip dot ford at verizon dot net>, Antoine Brodin <antoine dot brodin at laposte dot net>
2006-11-06 13:42:10 +00:00
priv_granted |= VEXEC;
Merge from POSIX.1e Capabilities development tree: o POSIX.1e capabilities authorize overriding of VEXEC for VDIR based on CAP_DAC_READ_SEARCH, but of !VDIR based on CAP_DAC_EXECUTE. Add appropriate conditionals to vaccess() to take that into account. o Synchronization cap_check_xxx() -> cap_check() change. Obtained from: TrustedBSD Project
2001-11-02 15:16:59 +00:00
}
o Restructure vaccess() so as to check for DAC permission to modify the object before falling back on privilege. Make vaccess() accept an additional optional argument, privused, to determine whether privilege was required for vaccess() to return 0. Add commented out capability checks for reference. Rename some variables to make it more clear which modes/uids/etc are associated with the object, and which with the access mode. o Update file system use of vaccess() to pass NULL as the optional privused argument. Once additional patches are applied, suser() will no longer set ASU, so privused will permit passing of privilege information up the stack to the caller. Reviewed by: bde, green, phk, -security, others Obtained from: TrustedBSD Project
2000-08-29 14:45:49 +00:00
Introduce accmode_t. This is required for NFSv4 ACLs - it will be neccessary to add more V* constants, and the variables changed by this patch were often being assigned to mode_t variables, which is 16 bit. Approved by: rwatson (mentor)
2008-10-28 13:44:11 +00:00
if ((accmode & VREAD) && ((dac_granted & VREAD) == 0) &&
Eliminate now-unused SUSER_ALLOWJAIL arguments to priv_check_cred(); in some cases, move to priv_check() if it was an operation on a thread and no other flags were present. Eliminate caller-side jail exception checking (also now-unused); jail privilege exception code now goes solely in kern_jail.c. We can't yet eliminate suser() due to some cases in the KAME code where a privilege check is performed and then used in many different deferred paths. Do, however, move those prototypes to priv.h. Reviewed by: csjp Obtained from: TrustedBSD Project
2007-06-12 00:12:01 +00:00
!priv_check_cred(cred, PRIV_VFS_READ, 0))
Sweep kernel replacing suser(9) calls with priv(9) calls, assigning specific privilege names to a broad range of privileges. These may require some future tweaking. Sponsored by: nCircle Network Security, Inc. Obtained from: TrustedBSD Project Discussed on: arch@ Reviewed (at least in part) by: mlaier, jmg, pjd, bde, ceri, Alex Lyashkov <umka at sevcity dot net>, Skip Ford <skip dot ford at verizon dot net>, Antoine Brodin <antoine dot brodin at laposte dot net>
2006-11-06 13:42:10 +00:00
priv_granted |= VREAD;
o Restructure vaccess() so as to check for DAC permission to modify the object before falling back on privilege. Make vaccess() accept an additional optional argument, privused, to determine whether privilege was required for vaccess() to return 0. Add commented out capability checks for reference. Rename some variables to make it more clear which modes/uids/etc are associated with the object, and which with the access mode. o Update file system use of vaccess() to pass NULL as the optional privused argument. Once additional patches are applied, suser() will no longer set ASU, so privused will permit passing of privilege information up the stack to the caller. Reviewed by: bde, green, phk, -security, others Obtained from: TrustedBSD Project
2000-08-29 14:45:49 +00:00
Introduce accmode_t. This is required for NFSv4 ACLs - it will be neccessary to add more V* constants, and the variables changed by this patch were often being assigned to mode_t variables, which is 16 bit. Approved by: rwatson (mentor)
2008-10-28 13:44:11 +00:00
if ((accmode & VWRITE) && ((dac_granted & VWRITE) == 0) &&
Eliminate now-unused SUSER_ALLOWJAIL arguments to priv_check_cred(); in some cases, move to priv_check() if it was an operation on a thread and no other flags were present. Eliminate caller-side jail exception checking (also now-unused); jail privilege exception code now goes solely in kern_jail.c. We can't yet eliminate suser() due to some cases in the KAME code where a privilege check is performed and then used in many different deferred paths. Do, however, move those prototypes to priv.h. Reviewed by: csjp Obtained from: TrustedBSD Project
2007-06-12 00:12:01 +00:00
!priv_check_cred(cred, PRIV_VFS_WRITE, 0))
Sweep kernel replacing suser(9) calls with priv(9) calls, assigning specific privilege names to a broad range of privileges. These may require some future tweaking. Sponsored by: nCircle Network Security, Inc. Obtained from: TrustedBSD Project Discussed on: arch@ Reviewed (at least in part) by: mlaier, jmg, pjd, bde, ceri, Alex Lyashkov <umka at sevcity dot net>, Skip Ford <skip dot ford at verizon dot net>, Antoine Brodin <antoine dot brodin at laposte dot net>
2006-11-06 13:42:10 +00:00
priv_granted |= (VWRITE | VAPPEND);
o Restructure vaccess() so as to check for DAC permission to modify the object before falling back on privilege. Make vaccess() accept an additional optional argument, privused, to determine whether privilege was required for vaccess() to return 0. Add commented out capability checks for reference. Rename some variables to make it more clear which modes/uids/etc are associated with the object, and which with the access mode. o Update file system use of vaccess() to pass NULL as the optional privused argument. Once additional patches are applied, suser() will no longer set ASU, so privused will permit passing of privilege information up the stack to the caller. Reviewed by: bde, green, phk, -security, others Obtained from: TrustedBSD Project
2000-08-29 14:45:49 +00:00
Introduce accmode_t. This is required for NFSv4 ACLs - it will be neccessary to add more V* constants, and the variables changed by this patch were often being assigned to mode_t variables, which is 16 bit. Approved by: rwatson (mentor)
2008-10-28 13:44:11 +00:00
if ((accmode & VADMIN) && ((dac_granted & VADMIN) == 0) &&
Eliminate now-unused SUSER_ALLOWJAIL arguments to priv_check_cred(); in some cases, move to priv_check() if it was an operation on a thread and no other flags were present. Eliminate caller-side jail exception checking (also now-unused); jail privilege exception code now goes solely in kern_jail.c. We can't yet eliminate suser() due to some cases in the KAME code where a privilege check is performed and then used in many different deferred paths. Do, however, move those prototypes to priv.h. Reviewed by: csjp Obtained from: TrustedBSD Project
2007-06-12 00:12:01 +00:00
!priv_check_cred(cred, PRIV_VFS_ADMIN, 0))
Sweep kernel replacing suser(9) calls with priv(9) calls, assigning specific privilege names to a broad range of privileges. These may require some future tweaking. Sponsored by: nCircle Network Security, Inc. Obtained from: TrustedBSD Project Discussed on: arch@ Reviewed (at least in part) by: mlaier, jmg, pjd, bde, ceri, Alex Lyashkov <umka at sevcity dot net>, Skip Ford <skip dot ford at verizon dot net>, Antoine Brodin <antoine dot brodin at laposte dot net>
2006-11-06 13:42:10 +00:00
priv_granted |= VADMIN;
o Introduce new VOP_ACCESS() flag VADMIN, allowing file systems to perform "administrative" authorization checks. In most cases, the VADMIN test checks to make sure the credential effective uid is the same as the file owner. o Modify vaccess() to set VADMIN as an available right if the uid is appropriate. o Modify references to uid-based access control operations such that they now always invoke VOP_ACCESS() instead of using hard-coded policy checks. o This allows alternative UFS policies to be implemented by replacing only ufs_access() (such as mandatory system policies). o VOP_ACCESS() requires the caller to hold an exclusive vnode lock on the vnode: I believe that new invocations of VOP_ACCESS() are always called with the lock held. o Some direct checks of the uid remain, largely associated with the QUOTA and SUIDDIR code. Reviewed by: eivind Obtained from: TrustedBSD Project
2000-10-19 07:53:59 +00:00
Introduce accmode_t. This is required for NFSv4 ACLs - it will be neccessary to add more V* constants, and the variables changed by this patch were often being assigned to mode_t variables, which is 16 bit. Approved by: rwatson (mentor)
2008-10-28 13:44:11 +00:00
if ((accmode & (priv_granted | dac_granted)) == accmode) {
o Restructure vaccess() so as to check for DAC permission to modify the object before falling back on privilege. Make vaccess() accept an additional optional argument, privused, to determine whether privilege was required for vaccess() to return 0. Add commented out capability checks for reference. Rename some variables to make it more clear which modes/uids/etc are associated with the object, and which with the access mode. o Update file system use of vaccess() to pass NULL as the optional privused argument. Once additional patches are applied, suser() will no longer set ASU, so privused will permit passing of privilege information up the stack to the caller. Reviewed by: bde, green, phk, -security, others Obtained from: TrustedBSD Project
2000-08-29 14:45:49 +00:00
/* XXX audit: privilege used */
if (privused != NULL)
*privused = 1;
return (0);
}
Centralize the canonical vop_access user/group/other check in vaccess(). Discussed with: bde
2000-08-20 08:36:26 +00:00
Introduce accmode_t. This is required for NFSv4 ACLs - it will be neccessary to add more V* constants, and the variables changed by this patch were often being assigned to mode_t variables, which is 16 bit. Approved by: rwatson (mentor)
2008-10-28 13:44:11 +00:00
return ((accmode & VADMIN) ? EPERM : EACCES);
Centralize the canonical vop_access user/group/other check in vaccess(). Discussed with: bde
2000-08-20 08:36:26 +00:00
}
Introduce new extattr_check_cred() function which implements the canonical crential washing for extended attributes. Sponsored by: DARPA & NAI Labs.
2002-09-05 20:38:57 +00:00
/*
* Credential check based on process requesting service, and per-attribute
* permissions.
*/
int
Convert remaining functions in vfs_subr.c from K&R prototypes to ANSI C prototypes, as the majority of new functions added have been in this style. Changing prototype style now results in gcc noticing that the implementation of vn_pollrecord() has a 'short' argument instead of 'int' as prototyped in vnode.h, so correct that definition. In practice this didn't matter as only poll flags in the lower 16 bits are used. MFC after: 1 week
2006-01-21 19:42:10 +00:00
extattr_check_cred(struct vnode *vp, int attrnamespace, struct ucred *cred,
Introduce accmode_t. This is required for NFSv4 ACLs - it will be neccessary to add more V* constants, and the variables changed by this patch were often being assigned to mode_t variables, which is 16 bit. Approved by: rwatson (mentor)
2008-10-28 13:44:11 +00:00
struct thread *td, accmode_t accmode)
Introduce new extattr_check_cred() function which implements the canonical crential washing for extended attributes. Sponsored by: DARPA & NAI Labs.
2002-09-05 20:38:57 +00:00
{
/*
* Kernel-invoked always succeeds.
*/
Fix an inherited style bug: compare with NOCRED instead of NULL. Sponsored by: DARPA & NAI Labs.
2002-09-05 20:46:19 +00:00
if (cred == NOCRED)
Introduce new extattr_check_cred() function which implements the canonical crential washing for extended attributes. Sponsored by: DARPA & NAI Labs.
2002-09-05 20:38:57 +00:00
return (0);
/*
Sweep kernel replacing suser(9) calls with priv(9) calls, assigning specific privilege names to a broad range of privileges. These may require some future tweaking. Sponsored by: nCircle Network Security, Inc. Obtained from: TrustedBSD Project Discussed on: arch@ Reviewed (at least in part) by: mlaier, jmg, pjd, bde, ceri, Alex Lyashkov <umka at sevcity dot net>, Skip Ford <skip dot ford at verizon dot net>, Antoine Brodin <antoine dot brodin at laposte dot net>
2006-11-06 13:42:10 +00:00
* Do not allow privileged processes in jail to directly manipulate
* system attributes.
Introduce new extattr_check_cred() function which implements the canonical crential washing for extended attributes. Sponsored by: DARPA & NAI Labs.
2002-09-05 20:38:57 +00:00
*/
switch (attrnamespace) {
case EXTATTR_NAMESPACE_SYSTEM:
/* Potentially should be: return (EPERM); */
Sweep kernel replacing suser(9) calls with priv(9) calls, assigning specific privilege names to a broad range of privileges. These may require some future tweaking. Sponsored by: nCircle Network Security, Inc. Obtained from: TrustedBSD Project Discussed on: arch@ Reviewed (at least in part) by: mlaier, jmg, pjd, bde, ceri, Alex Lyashkov <umka at sevcity dot net>, Skip Ford <skip dot ford at verizon dot net>, Antoine Brodin <antoine dot brodin at laposte dot net>
2006-11-06 13:42:10 +00:00
return (priv_check_cred(cred, PRIV_VFS_EXTATTR_SYSTEM, 0));
Introduce new extattr_check_cred() function which implements the canonical crential washing for extended attributes. Sponsored by: DARPA & NAI Labs.
2002-09-05 20:38:57 +00:00
case EXTATTR_NAMESPACE_USER:
Introduce accmode_t. This is required for NFSv4 ACLs - it will be neccessary to add more V* constants, and the variables changed by this patch were often being assigned to mode_t variables, which is 16 bit. Approved by: rwatson (mentor)
2008-10-28 13:44:11 +00:00
return (VOP_ACCESS(vp, accmode, cred, td));
Introduce new extattr_check_cred() function which implements the canonical crential washing for extended attributes. Sponsored by: DARPA & NAI Labs.
2002-09-05 20:38:57 +00:00
default:
return (EPERM);
}
}
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
#ifdef DEBUG_VFS_LOCKS
/*
* This only exists to supress warnings from unlocked specfs accesses. It is
* no longer ok to have an unlocked VFS.
*/
- Move vp test for beeing NULL under IGNORE_LOCK(). - Check if panicstr isn't set, if it is ignore the lock. This helps to avoid confusion, because lockmgr is a no-op when panicstr isn't NULL, so asserting anything at this point doesn't make sense and can just race with other panic. Discussed with: kib
2008-07-27 11:46:42 +00:00
#define IGNORE_LOCK(vp) (panicstr != NULL || (vp) == NULL || \
(vp)->v_type == VCHR || (vp)->v_type == VBAD)
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
int vfs_badlock_ddb = 1; /* Drop into debugger on violation. */
put several of the options for DEBUG_VFS_LOCKS under control of sysctls.
2004-07-21 07:13:14 +00:00
SYSCTL_INT(_debug, OID_AUTO, vfs_badlock_ddb, CTLFLAG_RW, &vfs_badlock_ddb, 0, "");
More style fixes. Obtained from: bde
2004-01-05 23:40:46 +00:00
int vfs_badlock_mutex = 1; /* Check for interlock across VOPs. */
put several of the options for DEBUG_VFS_LOCKS under control of sysctls.
2004-07-21 07:13:14 +00:00
SYSCTL_INT(_debug, OID_AUTO, vfs_badlock_mutex, CTLFLAG_RW, &vfs_badlock_mutex, 0, "");
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
int vfs_badlock_print = 1; /* Print lock violations. */
put several of the options for DEBUG_VFS_LOCKS under control of sysctls.
2004-07-21 07:13:14 +00:00
SYSCTL_INT(_debug, OID_AUTO, vfs_badlock_print, CTLFLAG_RW, &vfs_badlock_print, 0, "");
#ifdef KDB
int vfs_badlock_backtrace = 1; /* Print backtrace at lock violations. */
SYSCTL_INT(_debug, OID_AUTO, vfs_badlock_backtrace, CTLFLAG_RW, &vfs_badlock_backtrace, 0, "");
#endif
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
static void
vfs_badlock(const char *msg, const char *str, struct vnode *vp)
{
put several of the options for DEBUG_VFS_LOCKS under control of sysctls.
2004-07-21 07:13:14 +00:00
#ifdef KDB
if (vfs_badlock_backtrace)
kdb_backtrace();
#endif
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
if (vfs_badlock_print)
printf("%s: %p %s\n", str, (void *)vp, msg);
if (vfs_badlock_ddb)
Add a new 'why' argument to kdb_enter(), and a set of constants to use for that argument. This will allow DDB to detect the broad category of reason why the debugger has been entered, which it can use for the purposes of deciding which DDB script to run. Assign approximate why values to all current consumers of the kdb_enter() interface.
2007-12-25 17:52:02 +00:00
kdb_enter(KDB_WHY_VFSLOCK, "lock violation");
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
}
void
assert_vi_locked(struct vnode *vp, const char *str)
{
if (vfs_badlock_mutex && !mtx_owned(VI_MTX(vp)))
vfs_badlock("interlock is not locked but should be", str, vp);
}
void
assert_vi_unlocked(struct vnode *vp, const char *str)
{
if (vfs_badlock_mutex && mtx_owned(VI_MTX(vp)))
vfs_badlock("interlock is locked but should not be", str, vp);
}
void
assert_vop_locked(struct vnode *vp, const char *str)
{
- Move vp test for beeing NULL under IGNORE_LOCK(). - Check if panicstr isn't set, if it is ignore the lock. This helps to avoid confusion, because lockmgr is a no-op when panicstr isn't NULL, so asserting anything at this point doesn't make sense and can just race with other panic. Discussed with: kib
2008-07-27 11:46:42 +00:00
if (!IGNORE_LOCK(vp) && VOP_ISLOCKED(vp) == 0)
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
vfs_badlock("is not locked but should be", str, vp);
}
void
assert_vop_unlocked(struct vnode *vp, const char *str)
{
- Move vp test for beeing NULL under IGNORE_LOCK(). - Check if panicstr isn't set, if it is ignore the lock. This helps to avoid confusion, because lockmgr is a no-op when panicstr isn't NULL, so asserting anything at this point doesn't make sense and can just race with other panic. Discussed with: kib
2008-07-27 11:46:42 +00:00
if (!IGNORE_LOCK(vp) && VOP_ISLOCKED(vp) == LK_EXCLUSIVE)
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
vfs_badlock("is locked but should not be", str, vp);
}
void
assert_vop_elocked(struct vnode *vp, const char *str)
{
- Move vp test for beeing NULL under IGNORE_LOCK(). - Check if panicstr isn't set, if it is ignore the lock. This helps to avoid confusion, because lockmgr is a no-op when panicstr isn't NULL, so asserting anything at this point doesn't make sense and can just race with other panic. Discussed with: kib
2008-07-27 11:46:42 +00:00
if (!IGNORE_LOCK(vp) && VOP_ISLOCKED(vp) != LK_EXCLUSIVE)
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
vfs_badlock("is not exclusive locked but should be", str, vp);
}
- Enable ASSERT_VOP_ELOCKED and assert_vop_elocked() now that vnode_if.awk uses it. Sponsored by: Isilon Systems, Inc.
2005-04-11 15:17:06 +00:00
#if 0
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
void
assert_vop_elocked_other(struct vnode *vp, const char *str)
{
- Move vp test for beeing NULL under IGNORE_LOCK(). - Check if panicstr isn't set, if it is ignore the lock. This helps to avoid confusion, because lockmgr is a no-op when panicstr isn't NULL, so asserting anything at this point doesn't make sense and can just race with other panic. Discussed with: kib
2008-07-27 11:46:42 +00:00
if (!IGNORE_LOCK(vp) && VOP_ISLOCKED(vp) != LK_EXCLOTHER)
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
vfs_badlock("is not exclusive locked by another thread",
str, vp);
}
void
assert_vop_slocked(struct vnode *vp, const char *str)
{
- Move vp test for beeing NULL under IGNORE_LOCK(). - Check if panicstr isn't set, if it is ignore the lock. This helps to avoid confusion, because lockmgr is a no-op when panicstr isn't NULL, so asserting anything at this point doesn't make sense and can just race with other panic. Discussed with: kib
2008-07-27 11:46:42 +00:00
if (!IGNORE_LOCK(vp) && VOP_ISLOCKED(vp) != LK_SHARED)
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
vfs_badlock("is not locked shared but should be", str, vp);
}
#endif /* 0 */
Allow EVFILT_VNODE events to work on every filesystem type, not just UFS by: - Making the pre and post hooks for the VOP functions work even when DEBUG_VFS_LOCKS is not defined. - Moving the KNOTE activations into the corresponding VOP hooks. - Creating a MNTK_NOKNOTE flag for the mnt_kern_flag field of struct mount that permits filesystems to disable the new behavior. - Creating a default VOP_KQFILTER function: vfs_kqfilter() My benchmarks have not revealed any performance degradation. Reviewed by: jeff, bde Approved by: rwatson, jmg (kqueue changes), grehan (mentor)
2005-06-09 20:20:31 +00:00
#endif /* DEBUG_VFS_LOCKS */
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
void
vop_rename_pre(void *ap)
{
struct vop_rename_args *a = ap;
Allow EVFILT_VNODE events to work on every filesystem type, not just UFS by: - Making the pre and post hooks for the VOP functions work even when DEBUG_VFS_LOCKS is not defined. - Moving the KNOTE activations into the corresponding VOP hooks. - Creating a MNTK_NOKNOTE flag for the mnt_kern_flag field of struct mount that permits filesystems to disable the new behavior. - Creating a default VOP_KQFILTER function: vfs_kqfilter() My benchmarks have not revealed any performance degradation. Reviewed by: jeff, bde Approved by: rwatson, jmg (kqueue changes), grehan (mentor)
2005-06-09 20:20:31 +00:00
#ifdef DEBUG_VFS_LOCKS
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
if (a->a_tvp)
ASSERT_VI_UNLOCKED(a->a_tvp, "VOP_RENAME");
ASSERT_VI_UNLOCKED(a->a_tdvp, "VOP_RENAME");
ASSERT_VI_UNLOCKED(a->a_fvp, "VOP_RENAME");
ASSERT_VI_UNLOCKED(a->a_fdvp, "VOP_RENAME");
/* Check the source (from). */
- Correct an assert in vop_rename_pre. fdvp may be locked if it is either the target directory or file. This case should fail in the filesystem anyway and perhaps kern_rename() should catch it. Sponsored by: Isilon Systems, Inc.
2006-03-19 20:14:46 +00:00
if (a->a_tdvp != a->a_fdvp && a->a_tvp != a->a_fdvp)
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
ASSERT_VOP_UNLOCKED(a->a_fdvp, "vop_rename: fdvp locked");
if (a->a_tvp != a->a_fvp)
Typo, 'from' vnode is locked here, not 'to' vnode.
2006-11-04 23:57:02 +00:00
ASSERT_VOP_UNLOCKED(a->a_fvp, "vop_rename: fvp locked");
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
/* Check the target. */
if (a->a_tvp)
ASSERT_VOP_LOCKED(a->a_tvp, "vop_rename: tvp not locked");
ASSERT_VOP_LOCKED(a->a_tdvp, "vop_rename: tdvp not locked");
Allow EVFILT_VNODE events to work on every filesystem type, not just UFS by: - Making the pre and post hooks for the VOP functions work even when DEBUG_VFS_LOCKS is not defined. - Moving the KNOTE activations into the corresponding VOP hooks. - Creating a MNTK_NOKNOTE flag for the mnt_kern_flag field of struct mount that permits filesystems to disable the new behavior. - Creating a default VOP_KQFILTER function: vfs_kqfilter() My benchmarks have not revealed any performance degradation. Reviewed by: jeff, bde Approved by: rwatson, jmg (kqueue changes), grehan (mentor)
2005-06-09 20:20:31 +00:00
#endif
if (a->a_tdvp != a->a_fdvp)
Fix a typo in vop_rename_pre() where we ended up using vholdl() instead of vhold(), even though the vnode interlock is unlocked. MFC after: 3 days
2005-08-28 23:00:11 +00:00
vhold(a->a_fdvp);
Allow EVFILT_VNODE events to work on every filesystem type, not just UFS by: - Making the pre and post hooks for the VOP functions work even when DEBUG_VFS_LOCKS is not defined. - Moving the KNOTE activations into the corresponding VOP hooks. - Creating a MNTK_NOKNOTE flag for the mnt_kern_flag field of struct mount that permits filesystems to disable the new behavior. - Creating a default VOP_KQFILTER function: vfs_kqfilter() My benchmarks have not revealed any performance degradation. Reviewed by: jeff, bde Approved by: rwatson, jmg (kqueue changes), grehan (mentor)
2005-06-09 20:20:31 +00:00
if (a->a_tvp != a->a_fvp)
vhold(a->a_fvp);
vhold(a->a_tdvp);
if (a->a_tvp)
vhold(a->a_tvp);
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
}
void
vop_strategy_pre(void *ap)
{
Allow EVFILT_VNODE events to work on every filesystem type, not just UFS by: - Making the pre and post hooks for the VOP functions work even when DEBUG_VFS_LOCKS is not defined. - Moving the KNOTE activations into the corresponding VOP hooks. - Creating a MNTK_NOKNOTE flag for the mnt_kern_flag field of struct mount that permits filesystems to disable the new behavior. - Creating a default VOP_KQFILTER function: vfs_kqfilter() My benchmarks have not revealed any performance degradation. Reviewed by: jeff, bde Approved by: rwatson, jmg (kqueue changes), grehan (mentor)
2005-06-09 20:20:31 +00:00
#ifdef DEBUG_VFS_LOCKS
More style fixes. Obtained from: bde
2004-01-05 23:40:46 +00:00
struct vop_strategy_args *a;
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
struct buf *bp;
More style fixes. Obtained from: bde
2004-01-05 23:40:46 +00:00
a = ap;
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
bp = a->a_bp;
/*
* Cluster ops lock their component buffers but not the IO container.
*/
if ((bp->b_flags & B_CLUSTER) != 0)
return;
- Introduce the function lockmgr_recursed() which returns true if the lockmgr lkp, when held in exclusive mode, is recursed - Introduce the function BUF_RECURSED() which does the same for bufobj locks based on the top of lockmgr_recursed() - Introduce the function BUF_ISLOCKED() which works like the counterpart VOP_ISLOCKED(9), showing the state of lockmgr linked with the bufobj BUF_RECURSED() and BUF_ISLOCKED() entirely replace the usage of bogus BUF_REFCNT() in a more explicative and SMP-compliant way. This allows us to axe out BUF_REFCNT() and leaving the function lockcount() totally unused in our stock kernel. Further commits will axe lockcount() as well as part of lockmgr() cleanup. KPI results, obviously, broken so further commits will update manpages and freebsd version. Tested by: kris (on UFS and NFS)
2008-01-19 17:36:23 +00:00
if (!BUF_ISLOCKED(bp)) {
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
if (vfs_badlock_print)
printf(
More style fixes. Obtained from: bde
2004-01-05 23:40:46 +00:00
"VOP_STRATEGY: bp is not locked but should be\n");
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
if (vfs_badlock_ddb)
Add a new 'why' argument to kdb_enter(), and a set of constants to use for that argument. This will allow DDB to detect the broad category of reason why the debugger has been entered, which it can use for the purposes of deciding which DDB script to run. Assign approximate why values to all current consumers of the kdb_enter() interface.
2007-12-25 17:52:02 +00:00
kdb_enter(KDB_WHY_VFSLOCK, "lock violation");
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
}
Allow EVFILT_VNODE events to work on every filesystem type, not just UFS by: - Making the pre and post hooks for the VOP functions work even when DEBUG_VFS_LOCKS is not defined. - Moving the KNOTE activations into the corresponding VOP hooks. - Creating a MNTK_NOKNOTE flag for the mnt_kern_flag field of struct mount that permits filesystems to disable the new behavior. - Creating a default VOP_KQFILTER function: vfs_kqfilter() My benchmarks have not revealed any performance degradation. Reviewed by: jeff, bde Approved by: rwatson, jmg (kqueue changes), grehan (mentor)
2005-06-09 20:20:31 +00:00
#endif
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
}
void
vop_lookup_pre(void *ap)
{
Allow EVFILT_VNODE events to work on every filesystem type, not just UFS by: - Making the pre and post hooks for the VOP functions work even when DEBUG_VFS_LOCKS is not defined. - Moving the KNOTE activations into the corresponding VOP hooks. - Creating a MNTK_NOKNOTE flag for the mnt_kern_flag field of struct mount that permits filesystems to disable the new behavior. - Creating a default VOP_KQFILTER function: vfs_kqfilter() My benchmarks have not revealed any performance degradation. Reviewed by: jeff, bde Approved by: rwatson, jmg (kqueue changes), grehan (mentor)
2005-06-09 20:20:31 +00:00
#ifdef DEBUG_VFS_LOCKS
More style fixes. Obtained from: bde
2004-01-05 23:40:46 +00:00
struct vop_lookup_args *a;
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
struct vnode *dvp;
More style fixes. Obtained from: bde
2004-01-05 23:40:46 +00:00
a = ap;
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
dvp = a->a_dvp;
ASSERT_VI_UNLOCKED(dvp, "VOP_LOOKUP");
ASSERT_VOP_LOCKED(dvp, "VOP_LOOKUP");
Allow EVFILT_VNODE events to work on every filesystem type, not just UFS by: - Making the pre and post hooks for the VOP functions work even when DEBUG_VFS_LOCKS is not defined. - Moving the KNOTE activations into the corresponding VOP hooks. - Creating a MNTK_NOKNOTE flag for the mnt_kern_flag field of struct mount that permits filesystems to disable the new behavior. - Creating a default VOP_KQFILTER function: vfs_kqfilter() My benchmarks have not revealed any performance degradation. Reviewed by: jeff, bde Approved by: rwatson, jmg (kqueue changes), grehan (mentor)
2005-06-09 20:20:31 +00:00
#endif
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
}
void
vop_lookup_post(void *ap, int rc)
{
Allow EVFILT_VNODE events to work on every filesystem type, not just UFS by: - Making the pre and post hooks for the VOP functions work even when DEBUG_VFS_LOCKS is not defined. - Moving the KNOTE activations into the corresponding VOP hooks. - Creating a MNTK_NOKNOTE flag for the mnt_kern_flag field of struct mount that permits filesystems to disable the new behavior. - Creating a default VOP_KQFILTER function: vfs_kqfilter() My benchmarks have not revealed any performance degradation. Reviewed by: jeff, bde Approved by: rwatson, jmg (kqueue changes), grehan (mentor)
2005-06-09 20:20:31 +00:00
#ifdef DEBUG_VFS_LOCKS
More style fixes. Obtained from: bde
2004-01-05 23:40:46 +00:00
struct vop_lookup_args *a;
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
struct vnode *dvp;
struct vnode *vp;
More style fixes. Obtained from: bde
2004-01-05 23:40:46 +00:00
a = ap;
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
dvp = a->a_dvp;
vp = *(a->a_vpp);
ASSERT_VI_UNLOCKED(dvp, "VOP_LOOKUP");
- Change vop_lookup_post assertions to reflect recent vfs_lookup changes. Sponsored by: Isilon Systems, Inc.
2005-04-13 10:57:53 +00:00
ASSERT_VOP_LOCKED(dvp, "VOP_LOOKUP");
- Adjust asserts in vop_lookup_post() to match the new post PDIRUNLOCK vfs. Sponsored by: Isilon Systems, Inc.
2005-03-28 09:25:25 +00:00
if (!rc)
ASSERT_VOP_LOCKED(vp, "VOP_LOOKUP (child)");
Allow EVFILT_VNODE events to work on every filesystem type, not just UFS by: - Making the pre and post hooks for the VOP functions work even when DEBUG_VFS_LOCKS is not defined. - Moving the KNOTE activations into the corresponding VOP hooks. - Creating a MNTK_NOKNOTE flag for the mnt_kern_flag field of struct mount that permits filesystems to disable the new behavior. - Creating a default VOP_KQFILTER function: vfs_kqfilter() My benchmarks have not revealed any performance degradation. Reviewed by: jeff, bde Approved by: rwatson, jmg (kqueue changes), grehan (mentor)
2005-06-09 20:20:31 +00:00
#endif
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
}
void
vop_lock_pre(void *ap)
{
Allow EVFILT_VNODE events to work on every filesystem type, not just UFS by: - Making the pre and post hooks for the VOP functions work even when DEBUG_VFS_LOCKS is not defined. - Moving the KNOTE activations into the corresponding VOP hooks. - Creating a MNTK_NOKNOTE flag for the mnt_kern_flag field of struct mount that permits filesystems to disable the new behavior. - Creating a default VOP_KQFILTER function: vfs_kqfilter() My benchmarks have not revealed any performance degradation. Reviewed by: jeff, bde Approved by: rwatson, jmg (kqueue changes), grehan (mentor)
2005-06-09 20:20:31 +00:00
#ifdef DEBUG_VFS_LOCKS
Since renaming of vop_lock to _vop_lock, pre- and post-condition function calls are no more generated for vop_lock. Rename _vop_lock to vop_lock1 to satisfy tools/vnode_if.awk assumption about vop naming conventions. This restores pre/post-condition calls.
2007-05-18 13:02:13 +00:00
struct vop_lock1_args *a = ap;
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
if ((a->a_flags & LK_INTERLOCK) == 0)
ASSERT_VI_UNLOCKED(a->a_vp, "VOP_LOCK");
else
ASSERT_VI_LOCKED(a->a_vp, "VOP_LOCK");
Allow EVFILT_VNODE events to work on every filesystem type, not just UFS by: - Making the pre and post hooks for the VOP functions work even when DEBUG_VFS_LOCKS is not defined. - Moving the KNOTE activations into the corresponding VOP hooks. - Creating a MNTK_NOKNOTE flag for the mnt_kern_flag field of struct mount that permits filesystems to disable the new behavior. - Creating a default VOP_KQFILTER function: vfs_kqfilter() My benchmarks have not revealed any performance degradation. Reviewed by: jeff, bde Approved by: rwatson, jmg (kqueue changes), grehan (mentor)
2005-06-09 20:20:31 +00:00
#endif
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
}
void
vop_lock_post(void *ap, int rc)
{
Allow EVFILT_VNODE events to work on every filesystem type, not just UFS by: - Making the pre and post hooks for the VOP functions work even when DEBUG_VFS_LOCKS is not defined. - Moving the KNOTE activations into the corresponding VOP hooks. - Creating a MNTK_NOKNOTE flag for the mnt_kern_flag field of struct mount that permits filesystems to disable the new behavior. - Creating a default VOP_KQFILTER function: vfs_kqfilter() My benchmarks have not revealed any performance degradation. Reviewed by: jeff, bde Approved by: rwatson, jmg (kqueue changes), grehan (mentor)
2005-06-09 20:20:31 +00:00
#ifdef DEBUG_VFS_LOCKS
Since renaming of vop_lock to _vop_lock, pre- and post-condition function calls are no more generated for vop_lock. Rename _vop_lock to vop_lock1 to satisfy tools/vnode_if.awk assumption about vop naming conventions. This restores pre/post-condition calls.
2007-05-18 13:02:13 +00:00
struct vop_lock1_args *a = ap;
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
ASSERT_VI_UNLOCKED(a->a_vp, "VOP_LOCK");
if (rc == 0)
ASSERT_VOP_LOCKED(a->a_vp, "VOP_LOCK");
Allow EVFILT_VNODE events to work on every filesystem type, not just UFS by: - Making the pre and post hooks for the VOP functions work even when DEBUG_VFS_LOCKS is not defined. - Moving the KNOTE activations into the corresponding VOP hooks. - Creating a MNTK_NOKNOTE flag for the mnt_kern_flag field of struct mount that permits filesystems to disable the new behavior. - Creating a default VOP_KQFILTER function: vfs_kqfilter() My benchmarks have not revealed any performance degradation. Reviewed by: jeff, bde Approved by: rwatson, jmg (kqueue changes), grehan (mentor)
2005-06-09 20:20:31 +00:00
#endif
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
}
void
vop_unlock_pre(void *ap)
{
Allow EVFILT_VNODE events to work on every filesystem type, not just UFS by: - Making the pre and post hooks for the VOP functions work even when DEBUG_VFS_LOCKS is not defined. - Moving the KNOTE activations into the corresponding VOP hooks. - Creating a MNTK_NOKNOTE flag for the mnt_kern_flag field of struct mount that permits filesystems to disable the new behavior. - Creating a default VOP_KQFILTER function: vfs_kqfilter() My benchmarks have not revealed any performance degradation. Reviewed by: jeff, bde Approved by: rwatson, jmg (kqueue changes), grehan (mentor)
2005-06-09 20:20:31 +00:00
#ifdef DEBUG_VFS_LOCKS
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
struct vop_unlock_args *a = ap;
if (a->a_flags & LK_INTERLOCK)
ASSERT_VI_LOCKED(a->a_vp, "VOP_UNLOCK");
ASSERT_VOP_LOCKED(a->a_vp, "VOP_UNLOCK");
Allow EVFILT_VNODE events to work on every filesystem type, not just UFS by: - Making the pre and post hooks for the VOP functions work even when DEBUG_VFS_LOCKS is not defined. - Moving the KNOTE activations into the corresponding VOP hooks. - Creating a MNTK_NOKNOTE flag for the mnt_kern_flag field of struct mount that permits filesystems to disable the new behavior. - Creating a default VOP_KQFILTER function: vfs_kqfilter() My benchmarks have not revealed any performance degradation. Reviewed by: jeff, bde Approved by: rwatson, jmg (kqueue changes), grehan (mentor)
2005-06-09 20:20:31 +00:00
#endif
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
}
void
vop_unlock_post(void *ap, int rc)
{
Allow EVFILT_VNODE events to work on every filesystem type, not just UFS by: - Making the pre and post hooks for the VOP functions work even when DEBUG_VFS_LOCKS is not defined. - Moving the KNOTE activations into the corresponding VOP hooks. - Creating a MNTK_NOKNOTE flag for the mnt_kern_flag field of struct mount that permits filesystems to disable the new behavior. - Creating a default VOP_KQFILTER function: vfs_kqfilter() My benchmarks have not revealed any performance degradation. Reviewed by: jeff, bde Approved by: rwatson, jmg (kqueue changes), grehan (mentor)
2005-06-09 20:20:31 +00:00
#ifdef DEBUG_VFS_LOCKS
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
struct vop_unlock_args *a = ap;
if (a->a_flags & LK_INTERLOCK)
ASSERT_VI_UNLOCKED(a->a_vp, "VOP_UNLOCK");
Allow EVFILT_VNODE events to work on every filesystem type, not just UFS by: - Making the pre and post hooks for the VOP functions work even when DEBUG_VFS_LOCKS is not defined. - Moving the KNOTE activations into the corresponding VOP hooks. - Creating a MNTK_NOKNOTE flag for the mnt_kern_flag field of struct mount that permits filesystems to disable the new behavior. - Creating a default VOP_KQFILTER function: vfs_kqfilter() My benchmarks have not revealed any performance degradation. Reviewed by: jeff, bde Approved by: rwatson, jmg (kqueue changes), grehan (mentor)
2005-06-09 20:20:31 +00:00
#endif
}
void
vop_create_post(void *ap, int rc)
{
struct vop_create_args *a = ap;
if (!rc)
Minor style cleanups (mostly removal of trailing whitespaces).
2007-04-10 15:29:37 +00:00
VFS_KNOTE_LOCKED(a->a_dvp, NOTE_WRITE);
Allow EVFILT_VNODE events to work on every filesystem type, not just UFS by: - Making the pre and post hooks for the VOP functions work even when DEBUG_VFS_LOCKS is not defined. - Moving the KNOTE activations into the corresponding VOP hooks. - Creating a MNTK_NOKNOTE flag for the mnt_kern_flag field of struct mount that permits filesystems to disable the new behavior. - Creating a default VOP_KQFILTER function: vfs_kqfilter() My benchmarks have not revealed any performance degradation. Reviewed by: jeff, bde Approved by: rwatson, jmg (kqueue changes), grehan (mentor)
2005-06-09 20:20:31 +00:00
}
void
vop_link_post(void *ap, int rc)
{
struct vop_link_args *a = ap;
Minor style cleanups (mostly removal of trailing whitespaces).
2007-04-10 15:29:37 +00:00
Allow EVFILT_VNODE events to work on every filesystem type, not just UFS by: - Making the pre and post hooks for the VOP functions work even when DEBUG_VFS_LOCKS is not defined. - Moving the KNOTE activations into the corresponding VOP hooks. - Creating a MNTK_NOKNOTE flag for the mnt_kern_flag field of struct mount that permits filesystems to disable the new behavior. - Creating a default VOP_KQFILTER function: vfs_kqfilter() My benchmarks have not revealed any performance degradation. Reviewed by: jeff, bde Approved by: rwatson, jmg (kqueue changes), grehan (mentor)
2005-06-09 20:20:31 +00:00
if (!rc) {
Minor style cleanups (mostly removal of trailing whitespaces).
2007-04-10 15:29:37 +00:00
VFS_KNOTE_LOCKED(a->a_vp, NOTE_LINK);
Fix the recent panics/LORs/hangs created by my kqueue commit by: - Introducing the possibility of using locks different than mutexes for the knlist locking. In order to do this, we add three arguments to knlist_init() to specify the functions to use to lock, unlock and check if the lock is owned. If these arguments are NULL, we assume mtx_lock, mtx_unlock and mtx_owned, respectively. - Using the vnode lock for the knlist locking, when doing kqueue operations on a vnode. This way, we don't have to lock the vnode while holding a mutex, in filt_vfsread. Reviewed by: jmg Approved by: re (scottl), scottl (mentor override) Pointyhat to: ssouhlal Will be happy: everyone
2005-07-01 16:28:32 +00:00
VFS_KNOTE_LOCKED(a->a_tdvp, NOTE_WRITE);
Allow EVFILT_VNODE events to work on every filesystem type, not just UFS by: - Making the pre and post hooks for the VOP functions work even when DEBUG_VFS_LOCKS is not defined. - Moving the KNOTE activations into the corresponding VOP hooks. - Creating a MNTK_NOKNOTE flag for the mnt_kern_flag field of struct mount that permits filesystems to disable the new behavior. - Creating a default VOP_KQFILTER function: vfs_kqfilter() My benchmarks have not revealed any performance degradation. Reviewed by: jeff, bde Approved by: rwatson, jmg (kqueue changes), grehan (mentor)
2005-06-09 20:20:31 +00:00
}
}
void
vop_mkdir_post(void *ap, int rc)
{
struct vop_mkdir_args *a = ap;
if (!rc)
Fix the recent panics/LORs/hangs created by my kqueue commit by: - Introducing the possibility of using locks different than mutexes for the knlist locking. In order to do this, we add three arguments to knlist_init() to specify the functions to use to lock, unlock and check if the lock is owned. If these arguments are NULL, we assume mtx_lock, mtx_unlock and mtx_owned, respectively. - Using the vnode lock for the knlist locking, when doing kqueue operations on a vnode. This way, we don't have to lock the vnode while holding a mutex, in filt_vfsread. Reviewed by: jmg Approved by: re (scottl), scottl (mentor override) Pointyhat to: ssouhlal Will be happy: everyone
2005-07-01 16:28:32 +00:00
VFS_KNOTE_LOCKED(a->a_dvp, NOTE_WRITE | NOTE_LINK);
Allow EVFILT_VNODE events to work on every filesystem type, not just UFS by: - Making the pre and post hooks for the VOP functions work even when DEBUG_VFS_LOCKS is not defined. - Moving the KNOTE activations into the corresponding VOP hooks. - Creating a MNTK_NOKNOTE flag for the mnt_kern_flag field of struct mount that permits filesystems to disable the new behavior. - Creating a default VOP_KQFILTER function: vfs_kqfilter() My benchmarks have not revealed any performance degradation. Reviewed by: jeff, bde Approved by: rwatson, jmg (kqueue changes), grehan (mentor)
2005-06-09 20:20:31 +00:00
}
void
vop_mknod_post(void *ap, int rc)
{
struct vop_mknod_args *a = ap;
if (!rc)
Fix the recent panics/LORs/hangs created by my kqueue commit by: - Introducing the possibility of using locks different than mutexes for the knlist locking. In order to do this, we add three arguments to knlist_init() to specify the functions to use to lock, unlock and check if the lock is owned. If these arguments are NULL, we assume mtx_lock, mtx_unlock and mtx_owned, respectively. - Using the vnode lock for the knlist locking, when doing kqueue operations on a vnode. This way, we don't have to lock the vnode while holding a mutex, in filt_vfsread. Reviewed by: jmg Approved by: re (scottl), scottl (mentor override) Pointyhat to: ssouhlal Will be happy: everyone
2005-07-01 16:28:32 +00:00
VFS_KNOTE_LOCKED(a->a_dvp, NOTE_WRITE);
Allow EVFILT_VNODE events to work on every filesystem type, not just UFS by: - Making the pre and post hooks for the VOP functions work even when DEBUG_VFS_LOCKS is not defined. - Moving the KNOTE activations into the corresponding VOP hooks. - Creating a MNTK_NOKNOTE flag for the mnt_kern_flag field of struct mount that permits filesystems to disable the new behavior. - Creating a default VOP_KQFILTER function: vfs_kqfilter() My benchmarks have not revealed any performance degradation. Reviewed by: jeff, bde Approved by: rwatson, jmg (kqueue changes), grehan (mentor)
2005-06-09 20:20:31 +00:00
}
void
vop_remove_post(void *ap, int rc)
{
struct vop_remove_args *a = ap;
if (!rc) {
Fix the recent panics/LORs/hangs created by my kqueue commit by: - Introducing the possibility of using locks different than mutexes for the knlist locking. In order to do this, we add three arguments to knlist_init() to specify the functions to use to lock, unlock and check if the lock is owned. If these arguments are NULL, we assume mtx_lock, mtx_unlock and mtx_owned, respectively. - Using the vnode lock for the knlist locking, when doing kqueue operations on a vnode. This way, we don't have to lock the vnode while holding a mutex, in filt_vfsread. Reviewed by: jmg Approved by: re (scottl), scottl (mentor override) Pointyhat to: ssouhlal Will be happy: everyone
2005-07-01 16:28:32 +00:00
VFS_KNOTE_LOCKED(a->a_dvp, NOTE_WRITE);
VFS_KNOTE_LOCKED(a->a_vp, NOTE_DELETE);
Allow EVFILT_VNODE events to work on every filesystem type, not just UFS by: - Making the pre and post hooks for the VOP functions work even when DEBUG_VFS_LOCKS is not defined. - Moving the KNOTE activations into the corresponding VOP hooks. - Creating a MNTK_NOKNOTE flag for the mnt_kern_flag field of struct mount that permits filesystems to disable the new behavior. - Creating a default VOP_KQFILTER function: vfs_kqfilter() My benchmarks have not revealed any performance degradation. Reviewed by: jeff, bde Approved by: rwatson, jmg (kqueue changes), grehan (mentor)
2005-06-09 20:20:31 +00:00
}
}
void
vop_rename_post(void *ap, int rc)
{
struct vop_rename_args *a = ap;
if (!rc) {
Fix the recent panics/LORs/hangs created by my kqueue commit by: - Introducing the possibility of using locks different than mutexes for the knlist locking. In order to do this, we add three arguments to knlist_init() to specify the functions to use to lock, unlock and check if the lock is owned. If these arguments are NULL, we assume mtx_lock, mtx_unlock and mtx_owned, respectively. - Using the vnode lock for the knlist locking, when doing kqueue operations on a vnode. This way, we don't have to lock the vnode while holding a mutex, in filt_vfsread. Reviewed by: jmg Approved by: re (scottl), scottl (mentor override) Pointyhat to: ssouhlal Will be happy: everyone
2005-07-01 16:28:32 +00:00
VFS_KNOTE_UNLOCKED(a->a_fdvp, NOTE_WRITE);
VFS_KNOTE_UNLOCKED(a->a_tdvp, NOTE_WRITE);
VFS_KNOTE_UNLOCKED(a->a_fvp, NOTE_RENAME);
Allow EVFILT_VNODE events to work on every filesystem type, not just UFS by: - Making the pre and post hooks for the VOP functions work even when DEBUG_VFS_LOCKS is not defined. - Moving the KNOTE activations into the corresponding VOP hooks. - Creating a MNTK_NOKNOTE flag for the mnt_kern_flag field of struct mount that permits filesystems to disable the new behavior. - Creating a default VOP_KQFILTER function: vfs_kqfilter() My benchmarks have not revealed any performance degradation. Reviewed by: jeff, bde Approved by: rwatson, jmg (kqueue changes), grehan (mentor)
2005-06-09 20:20:31 +00:00
if (a->a_tvp)
Fix the recent panics/LORs/hangs created by my kqueue commit by: - Introducing the possibility of using locks different than mutexes for the knlist locking. In order to do this, we add three arguments to knlist_init() to specify the functions to use to lock, unlock and check if the lock is owned. If these arguments are NULL, we assume mtx_lock, mtx_unlock and mtx_owned, respectively. - Using the vnode lock for the knlist locking, when doing kqueue operations on a vnode. This way, we don't have to lock the vnode while holding a mutex, in filt_vfsread. Reviewed by: jmg Approved by: re (scottl), scottl (mentor override) Pointyhat to: ssouhlal Will be happy: everyone
2005-07-01 16:28:32 +00:00
VFS_KNOTE_UNLOCKED(a->a_tvp, NOTE_DELETE);
Allow EVFILT_VNODE events to work on every filesystem type, not just UFS by: - Making the pre and post hooks for the VOP functions work even when DEBUG_VFS_LOCKS is not defined. - Moving the KNOTE activations into the corresponding VOP hooks. - Creating a MNTK_NOKNOTE flag for the mnt_kern_flag field of struct mount that permits filesystems to disable the new behavior. - Creating a default VOP_KQFILTER function: vfs_kqfilter() My benchmarks have not revealed any performance degradation. Reviewed by: jeff, bde Approved by: rwatson, jmg (kqueue changes), grehan (mentor)
2005-06-09 20:20:31 +00:00
}
if (a->a_tdvp != a->a_fdvp)
vdrop(a->a_fdvp);
if (a->a_tvp != a->a_fvp)
vdrop(a->a_fvp);
vdrop(a->a_tdvp);
if (a->a_tvp)
vdrop(a->a_tvp);
}
void
vop_rmdir_post(void *ap, int rc)
{
struct vop_rmdir_args *a = ap;
if (!rc) {
Fix the recent panics/LORs/hangs created by my kqueue commit by: - Introducing the possibility of using locks different than mutexes for the knlist locking. In order to do this, we add three arguments to knlist_init() to specify the functions to use to lock, unlock and check if the lock is owned. If these arguments are NULL, we assume mtx_lock, mtx_unlock and mtx_owned, respectively. - Using the vnode lock for the knlist locking, when doing kqueue operations on a vnode. This way, we don't have to lock the vnode while holding a mutex, in filt_vfsread. Reviewed by: jmg Approved by: re (scottl), scottl (mentor override) Pointyhat to: ssouhlal Will be happy: everyone
2005-07-01 16:28:32 +00:00
VFS_KNOTE_LOCKED(a->a_dvp, NOTE_WRITE | NOTE_LINK);
VFS_KNOTE_LOCKED(a->a_vp, NOTE_DELETE);
Allow EVFILT_VNODE events to work on every filesystem type, not just UFS by: - Making the pre and post hooks for the VOP functions work even when DEBUG_VFS_LOCKS is not defined. - Moving the KNOTE activations into the corresponding VOP hooks. - Creating a MNTK_NOKNOTE flag for the mnt_kern_flag field of struct mount that permits filesystems to disable the new behavior. - Creating a default VOP_KQFILTER function: vfs_kqfilter() My benchmarks have not revealed any performance degradation. Reviewed by: jeff, bde Approved by: rwatson, jmg (kqueue changes), grehan (mentor)
2005-06-09 20:20:31 +00:00
}
}
void
vop_setattr_post(void *ap, int rc)
{
struct vop_setattr_args *a = ap;
if (!rc)
Fix the recent panics/LORs/hangs created by my kqueue commit by: - Introducing the possibility of using locks different than mutexes for the knlist locking. In order to do this, we add three arguments to knlist_init() to specify the functions to use to lock, unlock and check if the lock is owned. If these arguments are NULL, we assume mtx_lock, mtx_unlock and mtx_owned, respectively. - Using the vnode lock for the knlist locking, when doing kqueue operations on a vnode. This way, we don't have to lock the vnode while holding a mutex, in filt_vfsread. Reviewed by: jmg Approved by: re (scottl), scottl (mentor override) Pointyhat to: ssouhlal Will be happy: everyone
2005-07-01 16:28:32 +00:00
VFS_KNOTE_LOCKED(a->a_vp, NOTE_ATTRIB);
Allow EVFILT_VNODE events to work on every filesystem type, not just UFS by: - Making the pre and post hooks for the VOP functions work even when DEBUG_VFS_LOCKS is not defined. - Moving the KNOTE activations into the corresponding VOP hooks. - Creating a MNTK_NOKNOTE flag for the mnt_kern_flag field of struct mount that permits filesystems to disable the new behavior. - Creating a default VOP_KQFILTER function: vfs_kqfilter() My benchmarks have not revealed any performance degradation. Reviewed by: jeff, bde Approved by: rwatson, jmg (kqueue changes), grehan (mentor)
2005-06-09 20:20:31 +00:00
}
void
vop_symlink_post(void *ap, int rc)
{
struct vop_symlink_args *a = ap;
Minor style cleanups (mostly removal of trailing whitespaces).
2007-04-10 15:29:37 +00:00
Allow EVFILT_VNODE events to work on every filesystem type, not just UFS by: - Making the pre and post hooks for the VOP functions work even when DEBUG_VFS_LOCKS is not defined. - Moving the KNOTE activations into the corresponding VOP hooks. - Creating a MNTK_NOKNOTE flag for the mnt_kern_flag field of struct mount that permits filesystems to disable the new behavior. - Creating a default VOP_KQFILTER function: vfs_kqfilter() My benchmarks have not revealed any performance degradation. Reviewed by: jeff, bde Approved by: rwatson, jmg (kqueue changes), grehan (mentor)
2005-06-09 20:20:31 +00:00
if (!rc)
Fix the recent panics/LORs/hangs created by my kqueue commit by: - Introducing the possibility of using locks different than mutexes for the knlist locking. In order to do this, we add three arguments to knlist_init() to specify the functions to use to lock, unlock and check if the lock is owned. If these arguments are NULL, we assume mtx_lock, mtx_unlock and mtx_owned, respectively. - Using the vnode lock for the knlist locking, when doing kqueue operations on a vnode. This way, we don't have to lock the vnode while holding a mutex, in filt_vfsread. Reviewed by: jmg Approved by: re (scottl), scottl (mentor override) Pointyhat to: ssouhlal Will be happy: everyone
2005-07-01 16:28:32 +00:00
VFS_KNOTE_LOCKED(a->a_dvp, NOTE_WRITE);
style(9): Add empty line before first code line in functions with no local variables. Properly terminate comment sentences. Indent lines which are longer that 80 characters. Move v_addpollinfo closer to the rest of poll-related functions. Move DEBUG_VFS_LOCKS ifdefed block to the end of file. Obtained from: bde (partly)
2004-01-05 19:04:29 +00:00
}
Introduce a new kevent filter. EVFILT_FS that will be used to signal generic filesystem events to userspace. Currently only mount and unmount of filesystems are signalled. Soon to be added, up/down status of NFS. Introduce a sysctl node used to route requests to/from filesystems based on filesystem ids. Introduce a new vfsop, vfs_sysctl(mp, req) that is used as the callback/ entrypoint by the sysctl code to change individual filesystems.
2004-07-04 10:52:54 +00:00
Add locking to the kqueue subsystem. This also makes the kqueue subsystem a more complete subsystem, and removes the knowlege of how things are implemented from the drivers. Include locking around filter ops, so a module like aio will know when not to be unloaded if there are outstanding knotes using it's filter ops. Currently, it uses the MTX_DUPOK even though it is not always safe to aquire duplicate locks. Witness currently doesn't support the ability to discover if a dup lock is ok (in some cases). Reviewed by: green, rwatson (both earlier versions)
2004-08-15 06:24:42 +00:00
static struct knlist fs_knlist;
static void
vfs_event_init(void *arg)
{
Fix the recent panics/LORs/hangs created by my kqueue commit by: - Introducing the possibility of using locks different than mutexes for the knlist locking. In order to do this, we add three arguments to knlist_init() to specify the functions to use to lock, unlock and check if the lock is owned. If these arguments are NULL, we assume mtx_lock, mtx_unlock and mtx_owned, respectively. - Using the vnode lock for the knlist locking, when doing kqueue operations on a vnode. This way, we don't have to lock the vnode while holding a mutex, in filt_vfsread. Reviewed by: jmg Approved by: re (scottl), scottl (mentor override) Pointyhat to: ssouhlal Will be happy: everyone
2005-07-01 16:28:32 +00:00
knlist_init(&fs_knlist, NULL, NULL, NULL, NULL);
Add locking to the kqueue subsystem. This also makes the kqueue subsystem a more complete subsystem, and removes the knowlege of how things are implemented from the drivers. Include locking around filter ops, so a module like aio will know when not to be unloaded if there are outstanding knotes using it's filter ops. Currently, it uses the MTX_DUPOK even though it is not always safe to aquire duplicate locks. Witness currently doesn't support the ability to discover if a dup lock is ok (in some cases). Reviewed by: green, rwatson (both earlier versions)
2004-08-15 06:24:42 +00:00
}
/* XXX - correct order? */
SYSINIT(vfs_knlist, SI_SUB_VFS, SI_ORDER_ANY, vfs_event_init, NULL);
Introduce a new kevent filter. EVFILT_FS that will be used to signal generic filesystem events to userspace. Currently only mount and unmount of filesystems are signalled. Soon to be added, up/down status of NFS. Introduce a sysctl node used to route requests to/from filesystems based on filesystem ids. Introduce a new vfsop, vfs_sysctl(mp, req) that is used as the callback/ entrypoint by the sysctl code to change individual filesystems.
2004-07-04 10:52:54 +00:00
void
NFS mobility PHASE I, II & III (phase VI, and V pending): Rebind the client socket when we experience a timeout. This fixes the case where our IP changes for some reason. Signal a VFS event when NFS transitions from up to down and vice versa. Add a placeholder vfs_sysctl where we will put status reporting shortly. Also: Make down NFS mounts return EIO instead of EINTR when there is a soft timeout or force unmount in progress.
2004-07-06 09:12:03 +00:00
vfs_event_signal(fsid_t *fsid, u_int32_t event, intptr_t data __unused)
Introduce a new kevent filter. EVFILT_FS that will be used to signal generic filesystem events to userspace. Currently only mount and unmount of filesystems are signalled. Soon to be added, up/down status of NFS. Introduce a sysctl node used to route requests to/from filesystems based on filesystem ids. Introduce a new vfsop, vfs_sysctl(mp, req) that is used as the callback/ entrypoint by the sysctl code to change individual filesystems.
2004-07-04 10:52:54 +00:00
{
Add locking to the kqueue subsystem. This also makes the kqueue subsystem a more complete subsystem, and removes the knowlege of how things are implemented from the drivers. Include locking around filter ops, so a module like aio will know when not to be unloaded if there are outstanding knotes using it's filter ops. Currently, it uses the MTX_DUPOK even though it is not always safe to aquire duplicate locks. Witness currently doesn't support the ability to discover if a dup lock is ok (in some cases). Reviewed by: green, rwatson (both earlier versions)
2004-08-15 06:24:42 +00:00
KNOTE_UNLOCKED(&fs_knlist, event);
Introduce a new kevent filter. EVFILT_FS that will be used to signal generic filesystem events to userspace. Currently only mount and unmount of filesystems are signalled. Soon to be added, up/down status of NFS. Introduce a sysctl node used to route requests to/from filesystems based on filesystem ids. Introduce a new vfsop, vfs_sysctl(mp, req) that is used as the callback/ entrypoint by the sysctl code to change individual filesystems.
2004-07-04 10:52:54 +00:00
}
static int filt_fsattach(struct knote *kn);
static void filt_fsdetach(struct knote *kn);
static int filt_fsevent(struct knote *kn, long hint);
struct filterops fs_filtops =
{ 0, filt_fsattach, filt_fsdetach, filt_fsevent };
static int
filt_fsattach(struct knote *kn)
{
kn->kn_flags |= EV_CLEAR;
Add locking to the kqueue subsystem. This also makes the kqueue subsystem a more complete subsystem, and removes the knowlege of how things are implemented from the drivers. Include locking around filter ops, so a module like aio will know when not to be unloaded if there are outstanding knotes using it's filter ops. Currently, it uses the MTX_DUPOK even though it is not always safe to aquire duplicate locks. Witness currently doesn't support the ability to discover if a dup lock is ok (in some cases). Reviewed by: green, rwatson (both earlier versions)
2004-08-15 06:24:42 +00:00
knlist_add(&fs_knlist, kn, 0);
Introduce a new kevent filter. EVFILT_FS that will be used to signal generic filesystem events to userspace. Currently only mount and unmount of filesystems are signalled. Soon to be added, up/down status of NFS. Introduce a sysctl node used to route requests to/from filesystems based on filesystem ids. Introduce a new vfsop, vfs_sysctl(mp, req) that is used as the callback/ entrypoint by the sysctl code to change individual filesystems.
2004-07-04 10:52:54 +00:00
return (0);
}
static void
filt_fsdetach(struct knote *kn)
{
Add locking to the kqueue subsystem. This also makes the kqueue subsystem a more complete subsystem, and removes the knowlege of how things are implemented from the drivers. Include locking around filter ops, so a module like aio will know when not to be unloaded if there are outstanding knotes using it's filter ops. Currently, it uses the MTX_DUPOK even though it is not always safe to aquire duplicate locks. Witness currently doesn't support the ability to discover if a dup lock is ok (in some cases). Reviewed by: green, rwatson (both earlier versions)
2004-08-15 06:24:42 +00:00
knlist_remove(&fs_knlist, kn, 0);
Introduce a new kevent filter. EVFILT_FS that will be used to signal generic filesystem events to userspace. Currently only mount and unmount of filesystems are signalled. Soon to be added, up/down status of NFS. Introduce a sysctl node used to route requests to/from filesystems based on filesystem ids. Introduce a new vfsop, vfs_sysctl(mp, req) that is used as the callback/ entrypoint by the sysctl code to change individual filesystems.
2004-07-04 10:52:54 +00:00
}
static int
filt_fsevent(struct knote *kn, long hint)
{
kn->kn_fflags |= hint;
return (kn->kn_fflags != 0);
}
static int
sysctl_vfs_ctl(SYSCTL_HANDLER_ARGS)
{
struct vfsidctl vc;
int error;
struct mount *mp;
error = SYSCTL_IN(req, &vc, sizeof(vc));
if (error)
return (error);
Pass the operation in with the fsidctl. Remove some fsidctls that we will not be using. Correct prototypes for fs sysctls.
2004-07-04 20:21:58 +00:00
if (vc.vc_vers != VFS_CTL_VERS1)
return (EINVAL);
Introduce a new kevent filter. EVFILT_FS that will be used to signal generic filesystem events to userspace. Currently only mount and unmount of filesystems are signalled. Soon to be added, up/down status of NFS. Introduce a sysctl node used to route requests to/from filesystems based on filesystem ids. Introduce a new vfsop, vfs_sysctl(mp, req) that is used as the callback/ entrypoint by the sysctl code to change individual filesystems.
2004-07-04 10:52:54 +00:00
mp = vfs_getvfs(&vc.vc_fsid);
if (mp == NULL)
return (ENOENT);
Pass the operation in with the fsidctl. Remove some fsidctls that we will not be using. Correct prototypes for fs sysctls.
2004-07-04 20:21:58 +00:00
/* ensure that a specific sysctl goes to the right filesystem. */
if (strcmp(vc.vc_fstypename, "*") != 0 &&
strcmp(vc.vc_fstypename, mp->mnt_vfc->vfc_name) != 0) {
- Hold a reference from the time vfs_busy starts until vfs_unbusy is called. - vfs_getvfs has to return a reference to prevent the returned mountpoint from changing identities. - Release references acquired via vfs_getvfs. Discussed with: tegge Tested by: kris Sponsored by: Isilon Systems, Inc.
2006-03-31 03:53:25 +00:00
vfs_rel(mp);
Pass the operation in with the fsidctl. Remove some fsidctls that we will not be using. Correct prototypes for fs sysctls.
2004-07-04 20:21:58 +00:00
return (EINVAL);
}
Introduce a new kevent filter. EVFILT_FS that will be used to signal generic filesystem events to userspace. Currently only mount and unmount of filesystems are signalled. Soon to be added, up/down status of NFS. Introduce a sysctl node used to route requests to/from filesystems based on filesystem ids. Introduce a new vfsop, vfs_sysctl(mp, req) that is used as the callback/ entrypoint by the sysctl code to change individual filesystems.
2004-07-04 10:52:54 +00:00
VCTLTOREQ(&vc, req);
- Hold a reference from the time vfs_busy starts until vfs_unbusy is called. - vfs_getvfs has to return a reference to prevent the returned mountpoint from changing identities. - Release references acquired via vfs_getvfs. Discussed with: tegge Tested by: kris Sponsored by: Isilon Systems, Inc.
2006-03-31 03:53:25 +00:00
error = VFS_SYSCTL(mp, vc.vc_op, req);
vfs_rel(mp);
return (error);
Introduce a new kevent filter. EVFILT_FS that will be used to signal generic filesystem events to userspace. Currently only mount and unmount of filesystems are signalled. Soon to be added, up/down status of NFS. Introduce a sysctl node used to route requests to/from filesystems based on filesystem ids. Introduce a new vfsop, vfs_sysctl(mp, req) that is used as the callback/ entrypoint by the sysctl code to change individual filesystems.
2004-07-04 10:52:54 +00:00
}
Minor style cleanups (mostly removal of trailing whitespaces).
2007-04-10 15:29:37 +00:00
SYSCTL_PROC(_vfs, OID_AUTO, ctl, CTLFLAG_WR, NULL, 0, sysctl_vfs_ctl, "",
"Sysctl by fsid");
Create simple function init_va_filerev() for initializing a va_filerev field. Replace three instances of longhaired initialization va_filerev fields. Added XXX comment wondering why we don't use random bits instead of uptime of the system for this purpose.
2004-09-07 09:17:05 +00:00
/*
* Function to initialize a va_filerev field sensibly.
* XXX: Wouldn't a random number make a lot more sense ??
*/
u_quad_t
init_va_filerev(void)
{
struct bintime bt;
getbinuptime(&bt);
return (((u_quad_t)bt.sec << 32LL) | (bt.frac >> 32LL));
}
Allow EVFILT_VNODE events to work on every filesystem type, not just UFS by: - Making the pre and post hooks for the VOP functions work even when DEBUG_VFS_LOCKS is not defined. - Moving the KNOTE activations into the corresponding VOP hooks. - Creating a MNTK_NOKNOTE flag for the mnt_kern_flag field of struct mount that permits filesystems to disable the new behavior. - Creating a default VOP_KQFILTER function: vfs_kqfilter() My benchmarks have not revealed any performance degradation. Reviewed by: jeff, bde Approved by: rwatson, jmg (kqueue changes), grehan (mentor)
2005-06-09 20:20:31 +00:00
static int filt_vfsread(struct knote *kn, long hint);
static int filt_vfswrite(struct knote *kn, long hint);
static int filt_vfsvnode(struct knote *kn, long hint);
static void filt_vfsdetach(struct knote *kn);
static struct filterops vfsread_filtops =
{ 1, NULL, filt_vfsdetach, filt_vfsread };
static struct filterops vfswrite_filtops =
{ 1, NULL, filt_vfsdetach, filt_vfswrite };
static struct filterops vfsvnode_filtops =
{ 1, NULL, filt_vfsdetach, filt_vfsvnode };
Fix the recent panics/LORs/hangs created by my kqueue commit by: - Introducing the possibility of using locks different than mutexes for the knlist locking. In order to do this, we add three arguments to knlist_init() to specify the functions to use to lock, unlock and check if the lock is owned. If these arguments are NULL, we assume mtx_lock, mtx_unlock and mtx_owned, respectively. - Using the vnode lock for the knlist locking, when doing kqueue operations on a vnode. This way, we don't have to lock the vnode while holding a mutex, in filt_vfsread. Reviewed by: jmg Approved by: re (scottl), scottl (mentor override) Pointyhat to: ssouhlal Will be happy: everyone
2005-07-01 16:28:32 +00:00
static void
vfs_knllock(void *arg)
{
struct vnode *vp = arg;
vn_lock() is currently only used with the 'curthread' passed as argument. Remove this argument and pass curthread directly to underlying VOP_LOCK1() VFS method. This modify makes the code cleaner and in particular remove an annoying dependence helping next lockmgr() cleanup. KPI results, obviously, changed. Manpage and FreeBSD_version will be updated through further commits. As a side note, would be valuable to say that next commits will address a similar cleanup about VFS methods, in particular vop_lock1 and vop_unlock. Tested by: Diego Sardina <siarodx at gmail dot com>, Andrea Di Pasquale <whyx dot it at gmail dot com>
2008-01-10 01:10:58 +00:00
vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
Fix the recent panics/LORs/hangs created by my kqueue commit by: - Introducing the possibility of using locks different than mutexes for the knlist locking. In order to do this, we add three arguments to knlist_init() to specify the functions to use to lock, unlock and check if the lock is owned. If these arguments are NULL, we assume mtx_lock, mtx_unlock and mtx_owned, respectively. - Using the vnode lock for the knlist locking, when doing kqueue operations on a vnode. This way, we don't have to lock the vnode while holding a mutex, in filt_vfsread. Reviewed by: jmg Approved by: re (scottl), scottl (mentor override) Pointyhat to: ssouhlal Will be happy: everyone
2005-07-01 16:28:32 +00:00
}
static void
vfs_knlunlock(void *arg)
{
struct vnode *vp = arg;
VOP_LOCK1() (and so VOP_LOCK()) and VOP_UNLOCK() are only used in conjuction with 'thread' argument passing which is always curthread. Remove the unuseful extra-argument and pass explicitly curthread to lower layer functions, when necessary. KPI results broken by this change, which should affect several ports, so version bumping and manpage update will be further committed. Tested by: kris, pho, Diego Sardina <siarodx at gmail dot com>
2008-01-13 14:44:15 +00:00
VOP_UNLOCK(vp, 0);
Fix the recent panics/LORs/hangs created by my kqueue commit by: - Introducing the possibility of using locks different than mutexes for the knlist locking. In order to do this, we add three arguments to knlist_init() to specify the functions to use to lock, unlock and check if the lock is owned. If these arguments are NULL, we assume mtx_lock, mtx_unlock and mtx_owned, respectively. - Using the vnode lock for the knlist locking, when doing kqueue operations on a vnode. This way, we don't have to lock the vnode while holding a mutex, in filt_vfsread. Reviewed by: jmg Approved by: re (scottl), scottl (mentor override) Pointyhat to: ssouhlal Will be happy: everyone
2005-07-01 16:28:32 +00:00
}
static int
vfs_knllocked(void *arg)
{
struct vnode *vp = arg;
Axe the 'thread' argument from VOP_ISLOCKED() and lockstatus() as it is always curthread. As KPI gets broken by this patch, manpages and __FreeBSD_version will be updated by further commits. Tested by: Andrea Barberio <insomniac at slackware dot it>
2008-02-25 18:45:57 +00:00
return (VOP_ISLOCKED(vp) == LK_EXCLUSIVE);
Fix the recent panics/LORs/hangs created by my kqueue commit by: - Introducing the possibility of using locks different than mutexes for the knlist locking. In order to do this, we add three arguments to knlist_init() to specify the functions to use to lock, unlock and check if the lock is owned. If these arguments are NULL, we assume mtx_lock, mtx_unlock and mtx_owned, respectively. - Using the vnode lock for the knlist locking, when doing kqueue operations on a vnode. This way, we don't have to lock the vnode while holding a mutex, in filt_vfsread. Reviewed by: jmg Approved by: re (scottl), scottl (mentor override) Pointyhat to: ssouhlal Will be happy: everyone
2005-07-01 16:28:32 +00:00
}
Allow EVFILT_VNODE events to work on every filesystem type, not just UFS by: - Making the pre and post hooks for the VOP functions work even when DEBUG_VFS_LOCKS is not defined. - Moving the KNOTE activations into the corresponding VOP hooks. - Creating a MNTK_NOKNOTE flag for the mnt_kern_flag field of struct mount that permits filesystems to disable the new behavior. - Creating a default VOP_KQFILTER function: vfs_kqfilter() My benchmarks have not revealed any performance degradation. Reviewed by: jeff, bde Approved by: rwatson, jmg (kqueue changes), grehan (mentor)
2005-06-09 20:20:31 +00:00
int
vfs_kqfilter(struct vop_kqfilter_args *ap)
{
struct vnode *vp = ap->a_vp;
struct knote *kn = ap->a_kn;
Minor style cleanups (mostly removal of trailing whitespaces).
2007-04-10 15:29:37 +00:00
struct knlist *knl;
Allow EVFILT_VNODE events to work on every filesystem type, not just UFS by: - Making the pre and post hooks for the VOP functions work even when DEBUG_VFS_LOCKS is not defined. - Moving the KNOTE activations into the corresponding VOP hooks. - Creating a MNTK_NOKNOTE flag for the mnt_kern_flag field of struct mount that permits filesystems to disable the new behavior. - Creating a default VOP_KQFILTER function: vfs_kqfilter() My benchmarks have not revealed any performance degradation. Reviewed by: jeff, bde Approved by: rwatson, jmg (kqueue changes), grehan (mentor)
2005-06-09 20:20:31 +00:00
switch (kn->kn_filter) {
case EVFILT_READ:
kn->kn_fop = &vfsread_filtops;
break;
case EVFILT_WRITE:
kn->kn_fop = &vfswrite_filtops;
break;
case EVFILT_VNODE:
kn->kn_fop = &vfsvnode_filtops;
break;
default:
In vfs_kqfilter(), return EINVAL instead of 1 (EPERM) when an unsupported kqueue filter type is requested on a vnode. MFC after: 3 days
2005-09-12 19:22:37 +00:00
return (EINVAL);
Allow EVFILT_VNODE events to work on every filesystem type, not just UFS by: - Making the pre and post hooks for the VOP functions work even when DEBUG_VFS_LOCKS is not defined. - Moving the KNOTE activations into the corresponding VOP hooks. - Creating a MNTK_NOKNOTE flag for the mnt_kern_flag field of struct mount that permits filesystems to disable the new behavior. - Creating a default VOP_KQFILTER function: vfs_kqfilter() My benchmarks have not revealed any performance degradation. Reviewed by: jeff, bde Approved by: rwatson, jmg (kqueue changes), grehan (mentor)
2005-06-09 20:20:31 +00:00
}
kn->kn_hook = (caddr_t)vp;
Protect check for v_pollinfo == NULL and assignment of the newly allocated vpollinfo with vnode interlock. Fully initialize vpollinfo before putting pointer to it into vp->v_pollinfo. Discussed with: dwhite Tested by: pho MFC after: 1 week
2008-10-28 12:08:36 +00:00
v_addpollinfo(vp);
Allow EVFILT_VNODE events to work on every filesystem type, not just UFS by: - Making the pre and post hooks for the VOP functions work even when DEBUG_VFS_LOCKS is not defined. - Moving the KNOTE activations into the corresponding VOP hooks. - Creating a MNTK_NOKNOTE flag for the mnt_kern_flag field of struct mount that permits filesystems to disable the new behavior. - Creating a default VOP_KQFILTER function: vfs_kqfilter() My benchmarks have not revealed any performance degradation. Reviewed by: jeff, bde Approved by: rwatson, jmg (kqueue changes), grehan (mentor)
2005-06-09 20:20:31 +00:00
if (vp->v_pollinfo == NULL)
return (ENOMEM);
Fix the recent panics/LORs/hangs created by my kqueue commit by: - Introducing the possibility of using locks different than mutexes for the knlist locking. In order to do this, we add three arguments to knlist_init() to specify the functions to use to lock, unlock and check if the lock is owned. If these arguments are NULL, we assume mtx_lock, mtx_unlock and mtx_owned, respectively. - Using the vnode lock for the knlist locking, when doing kqueue operations on a vnode. This way, we don't have to lock the vnode while holding a mutex, in filt_vfsread. Reviewed by: jmg Approved by: re (scottl), scottl (mentor override) Pointyhat to: ssouhlal Will be happy: everyone
2005-07-01 16:28:32 +00:00
knl = &vp->v_pollinfo->vpi_selinfo.si_note;
knlist_add(knl, kn, 0);
Allow EVFILT_VNODE events to work on every filesystem type, not just UFS by: - Making the pre and post hooks for the VOP functions work even when DEBUG_VFS_LOCKS is not defined. - Moving the KNOTE activations into the corresponding VOP hooks. - Creating a MNTK_NOKNOTE flag for the mnt_kern_flag field of struct mount that permits filesystems to disable the new behavior. - Creating a default VOP_KQFILTER function: vfs_kqfilter() My benchmarks have not revealed any performance degradation. Reviewed by: jeff, bde Approved by: rwatson, jmg (kqueue changes), grehan (mentor)
2005-06-09 20:20:31 +00:00
return (0);
}
/*
* Detach knote from vnode
*/
static void
filt_vfsdetach(struct knote *kn)
{
struct vnode *vp = (struct vnode *)kn->kn_hook;
KASSERT(vp->v_pollinfo != NULL, ("Missing v_pollinfo"));
knlist_remove(&vp->v_pollinfo->vpi_selinfo.si_note, kn, 0);
}
/*ARGSUSED*/
static int
filt_vfsread(struct knote *kn, long hint)
{
struct vnode *vp = (struct vnode *)kn->kn_hook;
struct vattr va;
/*
* filesystem is gone, so set the EOF flag and schedule
* the knote for deletion.
*/
if (hint == NOTE_REVOKE) {
kn->kn_flags |= (EV_EOF | EV_ONESHOT);
return (1);
}
Decontextualize the couplet VOP_GETATTR / VOP_SETATTR as the passed thread was always curthread and totally unuseful. Tested by: Giovanni Trematerra <giovanni dot trematerra at gmail dot com>
2008-08-28 15:23:18 +00:00
if (VOP_GETATTR(vp, &va, curthread->td_ucred))
Allow EVFILT_VNODE events to work on every filesystem type, not just UFS by: - Making the pre and post hooks for the VOP functions work even when DEBUG_VFS_LOCKS is not defined. - Moving the KNOTE activations into the corresponding VOP hooks. - Creating a MNTK_NOKNOTE flag for the mnt_kern_flag field of struct mount that permits filesystems to disable the new behavior. - Creating a default VOP_KQFILTER function: vfs_kqfilter() My benchmarks have not revealed any performance degradation. Reviewed by: jeff, bde Approved by: rwatson, jmg (kqueue changes), grehan (mentor)
2005-06-09 20:20:31 +00:00
return (0);
kn->kn_data = va.va_size - kn->kn_fp->f_offset;
return (kn->kn_data != 0);
}
/*ARGSUSED*/
static int
filt_vfswrite(struct knote *kn, long hint)
{
/*
* filesystem is gone, so set the EOF flag and schedule
* the knote for deletion.
*/
if (hint == NOTE_REVOKE)
kn->kn_flags |= (EV_EOF | EV_ONESHOT);
kn->kn_data = 0;
return (1);
}
static int
filt_vfsvnode(struct knote *kn, long hint)
{
if (kn->kn_sfflags & hint)
kn->kn_fflags |= hint;
if (hint == NOTE_REVOKE) {
kn->kn_flags |= EV_EOF;
return (1);
}
return (kn->kn_fflags != 0);
}
Introduce vfs_read_dirent() which can help VOP_READDIR() implementations by handling all the cookie stuff.
2005-09-12 08:46:07 +00:00
int
vfs_read_dirent(struct vop_readdir_args *ap, struct dirent *dp, off_t off)
{
int error;
if (dp->d_reclen > ap->a_uio->uio_resid)
return (ENAMETOOLONG);
error = uiomove(dp, dp->d_reclen, ap->a_uio);
if (error) {
if (ap->a_ncookies != NULL) {
if (ap->a_cookies != NULL)
free(ap->a_cookies, M_TEMP);
ap->a_cookies = NULL;
*ap->a_ncookies = 0;
}
return (error);
}
if (ap->a_ncookies == NULL)
return (0);
This is a workaround for a complicated issue involving VFS cookies and devfs. The PR and patch have the details. The ultimate fix requires architectural changes and clarifications to the VFS API, but this will prevent the system from panicking when someone does "ls /dev" while running in a shell under the linuxulator. This issue affects HEAD and RELENG_6 only. PR: 88249 Submitted by: "Devon H. O'Dell" <dodell@ixsystems.com> MFC after: 3 days
2005-11-09 22:03:50 +00:00
KASSERT(ap->a_cookies,
("NULL ap->a_cookies value with non-NULL ap->a_ncookies!"));
Introduce vfs_read_dirent() which can help VOP_READDIR() implementations by handling all the cookie stuff.
2005-09-12 08:46:07 +00:00
*ap->a_cookies = realloc(*ap->a_cookies,
(*ap->a_ncookies + 1) * sizeof(u_long), M_TEMP, M_WAITOK | M_ZERO);
(*ap->a_cookies)[*ap->a_ncookies] = off;
return (0);
}
Move execve's access time update functionality into a new vfs_mark_atime() function, and use the new function for performing efficient atime updates in mmap(). Reviewed by: bde MFC after: 2 weeks
2005-10-12 06:56:00 +00:00
/*
* Mark for update the access time of the file if the filesystem
Tweak the wording for vfs_mark_atime() since the I/O it is avoiding by not updating va_atime via VOP_SETATTR() isn't always synchronous. For some filesystems it is asynchronous. Suggested by: bde
2009-01-23 22:13:00 +00:00
* supports VOP_MARKATIME. This functionality is used by execve and
* mmap, so we want to avoid the I/O implied by directly setting
* va_atime for the sake of efficiency.
Move execve's access time update functionality into a new vfs_mark_atime() function, and use the new function for performing efficient atime updates in mmap(). Reviewed by: bde MFC after: 2 weeks
2005-10-12 06:56:00 +00:00
*/
void
Decontextualize the couplet VOP_GETATTR / VOP_SETATTR as the passed thread was always curthread and totally unuseful. Tested by: Giovanni Trematerra <giovanni dot trematerra at gmail dot com>
2008-08-28 15:23:18 +00:00
vfs_mark_atime(struct vnode *vp, struct ucred *cred)
Move execve's access time update functionality into a new vfs_mark_atime() function, and use the new function for performing efficient atime updates in mmap(). Reviewed by: bde MFC after: 2 weeks
2005-10-12 06:56:00 +00:00
{
Fix a few style bogons. Submitted by: bde
2009-01-21 20:08:17 +00:00
if ((vp->v_mount->mnt_flag & (MNT_NOATIME | MNT_RDONLY)) == 0)
Move the VA_MARKATIME flag for VOP_SETATTR() out into its own VOP: VOP_MARKATIME() since unlike the rest of VOP_SETATTR(), VA_MARKATIME can be performed while holding a shared vnode lock (the same functionality is done internally by VOP_READ which can run with a shared vnode lock). Add missing locking of the vnode interlock to the ufs implementation and remove a special note and test from the NFS client about not supporting the feature. Inspired by: ups Tested by: pho
2009-01-21 14:42:00 +00:00
(void)VOP_MARKATIME(vp);
Move execve's access time update functionality into a new vfs_mark_atime() function, and use the new function for performing efficient atime updates in mmap(). Reviewed by: bde MFC after: 2 weeks
2005-10-12 06:56:00 +00:00
}
Reference in New Issue Copy Permalink