d/freebsd-nq
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
58aac12894
freebsd-nq/sys/kern/vfs_subr.c

3637 lines
87 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>
#include <sys/conf.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>
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>
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>
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 <sys/mac.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>
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>
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
Last major round (Unless Bruce thinks of somthing :-) of malloc changes. Distribute all but the most fundamental malloc types. This time I also remembered the trick to making things static: Put "static" in front of them. A couple of finer points by: bde
1997-10-12 20:26:33 +00:00
static MALLOC_DEFINE(M_NETADDR, "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);
Remove __P.
2002-03-19 21:25:46 +00:00
static void insmntque(struct vnode *vp, struct mount *mp);
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
static void vlruvp(struct vnode *vp);
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
static int flushbuflist(struct bufv *bufv, int flags, struct vnode *vp,
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);
- 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
static void vx_lock(struct vnode *vp);
static void vx_unlock(struct vnode *vp);
Make various vnode related functions static
2005-02-10 12:28:58 +00:00
static void vbusy(struct vnode *vp);
static void vdropl(struct vnode *vp);
static void vholdl(struct 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 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.
*/
int mpsafe_vfs = 0;
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()
* allocates a new vnode, never decreased.
*/
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,
};
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 vttoif_tab[9] = {
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,
S_IFSOCK, S_IFIFO, S_IFMT,
};
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
/*
* Minimum number of free vnodes. If there are fewer than this free vnodes,
* getnewvnode() will return a newly allocated vnode.
*/
Reduce the target number of vnodes on the freelist from desiredvnodes (usually a couple of thousand) to 25. The measured impact on cache-hits doesn't justify spending memory this way: Target number of free vnodes versus namecache hit rate in % during a make world: 10 98.5316 200 98.5479 500 98.5546 1000 98.5709 3000 98.6006 4000 98.6126
1997-09-25 16:17:57 +00:00
static u_long wantfreevnodes = 25;
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, "");
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 nameileafonly;
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
SYSCTL_INT(_vfs, OID_AUTO, nameileafonly, CTLFLAG_RW, &nameileafonly, 0, "");
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
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);
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 struct synclist *syncer_workitem_pending;
- 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;
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");
Gratuitous whitespace cleanup.
2002-06-06 15:46:38 +00:00
static int minvnodes;
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,
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
&minvnodes, 0, "Minimum number of 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
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
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
/* Hook for calling soft updates. */
Use indirect function pointer hooks instead of #ifdef SOFTUPDATES direct calls for the two places where the kernel calls into soft updates code. Set up the hooks in softdep_initialize() and NULL them out in softdep_uninitialize(). This change allows soft updates to function correctly when ufs is loaded as a module. Reviewed by: mckusick
2002-07-01 17:59:40 +00:00
int (*softdep_process_worklist_hook)(struct mount *);
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
*/
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
}
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
minvnodes = 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(&mountlist_mtx, "mountlist", NULL, MTX_DEF);
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,
NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
vnodepoll_zone = uma_zcreate("VNODEPOLL", sizeof (struct vpollinfo),
NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
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
*/
syncer_workitem_pending = hashinit(syncer_maxdelay, M_VNODE,
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_mask);
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);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
}
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
SYSINIT(vfs, SI_SUB_VFS, SI_ORDER_FIRST, vntblinit, NULL)
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
* unmounting. Interlock 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
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
vfs_busy(mp, flags, interlkp, 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
struct mount *mp;
int flags;
Convert lockmgr locks from using simple locks to using mutexes. Add lockdestroy() and appropriate invocations, which corresponds to lockinit() and must be called to clean up after a lockmgr lock is no longer needed.
2000-10-04 01:29:17 +00:00
struct mtx *interlkp;
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;
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 lkflags;
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);
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) {
- 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
if (flags & LK_NOWAIT) {
MNT_IUNLOCK(mp);
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
}
if (interlkp)
mtx_unlock(interlkp);
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;
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
/*
* Since all busy locks are shared except the exclusive
* lock granted when unmounting, the only place that a
* wakeup needs to be done is at the release of the
* exclusive lock at the end of dounmount.
*/
- 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(mp, MNT_MTX(mp), PVFS|PDROP, "vfs_busy", 0);
if (interlkp)
mtx_lock(interlkp);
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
}
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 (interlkp)
- 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
mtx_unlock(interlkp);
lkflags = LK_SHARED | LK_NOPAUSE | LK_INTERLOCK;
if (lockmgr(&mp->mnt_lock, lkflags, MNT_MTX(mp), 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
panic("vfs_busy: unexpected lock failure");
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
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
vfs_unbusy(mp, 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
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;
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
lockmgr(&mp->mnt_lock, LK_RELEASE, NULL, td);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
}
/*
* Lookup a mount point by filesystem identifier.
*/
struct mount *
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_getvfs(fsid)
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
fsid_t *fsid;
{
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
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]) {
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);
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
}
Introduce vfs_suser(), used to test if a user should have special privs for a mount.
2004-07-06 09:37:43 +00:00
/*
* Check if a user can access priveledged mount options.
*/
int
vfs_suser(struct mount *mp, struct thread *td)
{
int error;
if ((mp->mnt_flag & MNT_USER) == 0 ||
mp->mnt_cred->cr_uid != td->td_ucred->cr_uid) {
if ((error = suser(td)) != 0)
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
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_getnewfsid(mp)
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
struct mount *mp;
{
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;
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
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++;
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
if (vfs_getvfs(&tfsid) == NULL)
break;
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
vfs_timestamp(tsp)
struct timespec *tsp;
{
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
vattr_null(vap)
Remove register keyword
2005-01-28 12:39:10 +00:00
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;
trigger = cnt.v_page_count * 2 / usevnodes;
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
done = 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_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;
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
while (count && (vp = TAILQ_FIRST(&mp->mnt_nvnodelist)) != NULL) {
TAILQ_REMOVE(&mp->mnt_nvnodelist, vp, v_nmntvnodes);
TAILQ_INSERT_TAIL(&mp->mnt_nvnodelist, vp, v_nmntvnodes);
if (vp->v_type != VNON &&
vp->v_type != VBAD &&
- 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_TRYLOCK(vp)) {
if (VMIGHTFREE(vp) && /* critical path opt */
(vp->v_object == NULL ||
vp->v_object->resident_page_count < trigger)) {
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);
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
vgonel(vp, curthread);
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++;
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);
- 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
} else
VI_UNLOCK(vp);
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
}
--count;
}
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);
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;
}
/*
* 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;
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 done;
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;
Kill some XXXKSE's. vnlru/syncer are single threaded.
2004-03-29 22:45:33 +00:00
struct thread *td = FIRST_THREAD_IN_PROC(p);
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(&Giant);
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 (;;) {
kthread_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);
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
if (numvnodes - freevnodes <= desiredvnodes * 9 / 10) {
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) {
if (vfs_busy(mp, LK_NOWAIT, &mountlist_mtx, td)) {
nmp = TAILQ_NEXT(mp, mnt_list);
continue;
}
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);
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);
vfs_unbusy(mp, td);
}
mtx_unlock(&mountlist_mtx);
if (done == 0) {
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);
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
};
SYSINIT(vnlru, SI_SUB_KTHREAD_UPDATE, SI_ORDER_FIRST, kproc_start, &vnlru_kp)
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.
*/
- 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
/*
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
* Check to see if a free vnode can be recycled. If it can,
- 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
* recycle it and return it with the vnode interlock held.
- 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
{
struct thread *td = curthread;
vm_object_t object;
- 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
int error;
/* Don't recycle if we can't get the interlock */
- 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 (!VI_TRYLOCK(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 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
if (!VCANRECYCLE(vp)) {
VI_UNLOCK(vp);
return (EBUSY);
}
- 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.
*/
if (vn_lock(vp, LK_INTERLOCK | LK_EXCLUSIVE | LK_NOWAIT, td) != 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
return (EWOULDBLOCK);
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) {
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
VOP_UNLOCK(vp, 0, td);
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
}
- Make vn_lock() vget() and VOP_LOCK() all behave the same way WRT LK_INTERLOCK. The interlock will never be held on return from these functions even when there is an error. Errors typically only occur when the XLOCK is held which means this isn't the vnode we want anyway. Almost all users of these interfaces expected this behavior even though it was not provided before.
2002-08-22 07:44:45 +00:00
- 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
/*
* Don't recycle if we still have cached pages.
*/
Don't use VOP_GETVOBJECT, use vp->v_object directly.
2005-01-25 00:40:01 +00:00
object = vp->v_object;
if (object != NULL) {
Lock accesses to the vm_object's ref_count and resident_page_count.
2003-05-01 03:10:38 +00:00
VM_OBJECT_LOCK(object);
if (object->resident_page_count ||
object->ref_count) {
VM_OBJECT_UNLOCK(object);
error = EBUSY;
goto done;
}
VM_OBJECT_UNLOCK(object);
- 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
}
if (LIST_FIRST(&vp->v_cache_src)) {
/*
* note: nameileafonly sysctl is temporary,
* for debugging only, and will eventually be
* removed.
*/
if (nameileafonly > 0) {
/*
* Do not reuse namei-cached directory
* vnodes that have cached
* subdirectories.
*/
if (cache_leaf_test(vp) < 0) {
error = EISDIR;
goto done;
}
} else if (nameileafonly < 0 ||
vmiodirenable == 0) {
/*
* Do not reuse namei-cached directory
* vnodes if nameileafonly is -1 or
* if VMIO backing for directories is
* turned off (otherwise we reuse them
* too quickly).
*/
error = EBUSY;
goto done;
}
}
- 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
* mark it with XLOCK via vgonel() so that anyone who does find it
* will skip over it.
*/
VI_LOCK(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
if (!VCANRECYCLE(vp)) {
- 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
VI_UNLOCK(vp);
error = EBUSY;
goto done;
}
mtx_lock(&vnode_free_list_mtx);
TAILQ_REMOVE(&vnode_free_list, vp, v_freelist);
vp->v_iflag &= ~VI_FREE;
mtx_unlock(&vnode_free_list_mtx);
vp->v_iflag |= VI_DOOMED;
vtryrecycle: Don't rely on type VBAD alone to mean that we don't need to clean the vnode. If v_data is set, we still need to clean it. This code change should catch all incidents of the previous commit (INVARIANTS only).
2004-10-06 02:09:59 +00:00
if ((vp->v_type != VBAD) || (vp->v_data != NULL)) {
- 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
VOP_UNLOCK(vp, 0, td);
vgonel(vp, td);
} else
- 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
VOP_UNLOCK(vp, LK_INTERLOCK, td);
- 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
vn_finished_write(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
return (0);
done:
VOP_UNLOCK(vp, 0, td);
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);
- 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 (error);
}
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
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
getnewvnode(tag, mp, vops, vpp)
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
const char *tag;
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
struct mount *mp;
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
struct vop_vector *vops;
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
struct vnode **vpp;
{
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;
To avoid lock order reversals in getnewvnode(), the call to uma_zfree() must be delayed until the vnode interlock is released. Reported by: kris@ Approved by: re (jhb)
2002-12-08 05:06:50 +00:00
struct vpollinfo *pollinfo = 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
- 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
/*
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
* Try to reuse vnodes if we hit the max. This situation only
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
* occurs in certain large-memory (2G+) situations. We cannot
* attempt to directly reclaim vnodes due to nasty recursion
* problems.
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
*/
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
while (numvnodes - freevnodes > desiredvnodes) {
if (vnlruproc_sig == 0) {
vnlruproc_sig = 1; /* avoid unnecessary wakeups */
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);
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
/*
* Attempt to reuse a vnode already on the free list, allocating
* a new vnode if we can't find one or if we have not reached a
* good minimum for good LRU performance.
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
*/
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
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
if (freevnodes >= wantfreevnodes && numvnodes >= minvnodes) {
- 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
int error;
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
int count;
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
for (count = 0; count < freevnodes; count++) {
vp = TAILQ_FIRST(&vnode_free_list);
- 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
TAILQ_REMOVE(&vnode_free_list, vp, v_freelist);
- 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
TAILQ_INSERT_TAIL(&vnode_free_list, vp, v_freelist);
- 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
mtx_unlock(&vnode_free_list_mtx);
- 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
error = vtryrecycle(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
mtx_lock(&vnode_free_list_mtx);
When scanning the freelist looking for candidate vnodes to recycle, be sure to exit the loop with vp == NULL if no candidates are found. Formerly, this bug would cause the last vnode inspected to be used, even if it was not available. The result was a panic "vn_finished_write: neg cnt". Sponsored by: DARPA & NAI Labs.
2002-10-14 19:54:39 +00:00
if (error == 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
break;
When scanning the freelist looking for candidate vnodes to recycle, be sure to exit the loop with vp == NULL if no candidates are found. Formerly, this bug would cause the last vnode inspected to be used, even if it was not available. The result was a panic "vn_finished_write: neg cnt". Sponsored by: DARPA & NAI Labs.
2002-10-14 19:54:39 +00:00
vp = NULL;
Eliminated many redundant vm_map_lookup operations for vm_mmap. Speed up for vfs_bio -- addition of a routine bqrelse to greatly diminish overhead for merged cache. Efficiency improvement for vfs_cluster. It used to do alot of redundant calls to cluster_rbuild. Correct the ordering for vrele of .text and release of credentials. Use the selective tlb update for 486/586/P6. Numerous fixes to the size of objects allocated for files. Additionally, fixes in the various pagers. Fixes for proper positioning of vnode_pager_setsize in msdosfs and ext2fs. Fixes in the swap pager for exhausted resources. The pageout code will not as readily thrash. Change the page queue flags (PG_ACTIVE, PG_INACTIVE, PG_FREE, PG_CACHE) into page queue indices (PQ_ACTIVE, PQ_INACTIVE, PQ_FREE, PQ_CACHE), thereby improving efficiency of several routines. Eliminate even more unnecessary vm_page_protect operations. Significantly speed up process forks. Make vm_object_page_clean more efficient, thereby eliminating the pause that happens every 30seconds. Make sequential clustered writes B_ASYNC instead of B_DELWRI even in the case of filesystems mounted async. Fix a panic with busy pages when write clustering is done for non-VMIO buffers.
1996-01-19 04:00:31 +00:00
}
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
}
if (vp) {
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
freevnodes--;
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
bo = &vp->v_bufobj;
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(&vnode_free_list_mtx);
- 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
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
#ifdef INVARIANTS
All of this is cosmetic. prototypes, #includes, printfs and so on. Makes GCC a lot more silent.
1994-10-02 17:35:40 +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 (vp->v_data)
getnewvnode: Weaken the panic "cleaned vnode isn't" to a warning. Discussion: this panic (or waning) only occurs when the kernel is compiled with INVARIANTS. Otherwise the problem (which means that the vp->v_data field isn't NULL, and represents a coding error and possibly a memory leak) is silently ignored by setting it to NULL later on. Panicking here isn't very helpful: by this time, we can only find the symptoms. The panic occurs long after the reason for "not cleaning" has been forgotten; in the case in point, it was the result of severe file system corruption which left the v_type field set to VBAD. That issue will be addressed by a separate commit.
2004-10-06 02:06:11 +00:00
printf("cleaned vnode isn't, "
"address %p, inode %p\n",
vp, vp->v_data);
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
if (bo->bo_numoutput)
- Add a new assert in the getnewvnode(). Assert that the usecount is still 0 to detect getnewvnode() races. - Add the vnode address to a few panics near by to help in debugging. Sponsored by: Isilon Systems, Inc.
2005-02-08 23:27:10 +00:00
panic("%p: Clean vnode has pending I/O's", vp);
if (vp->v_usecount != 0)
panic("%p: Non-zero use count", vp);
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 (vp->v_writecount != 0)
- Add a new assert in the getnewvnode(). Assert that the usecount is still 0 to detect getnewvnode() races. - Add the vnode address to a few panics near by to help in debugging. Sponsored by: Isilon Systems, Inc.
2005-02-08 23:27:10 +00:00
panic("%p: Non-zero write count", vp);
All of this is cosmetic. prototypes, #includes, printfs and so on. Makes GCC a lot more silent.
1994-10-02 17:35:40 +00:00
}
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
#endif
To avoid lock order reversals in getnewvnode(), the call to uma_zfree() must be delayed until the vnode interlock is released. Reported by: kris@ Approved by: re (jhb)
2002-12-08 05:06:50 +00:00
if ((pollinfo = vp->v_pollinfo) != NULL) {
/*
* To avoid lock order reversals, the call to
* uma_zfree() must be delayed until the vnode
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
* interlock is released.
To avoid lock order reversals in getnewvnode(), the call to uma_zfree() must be delayed until the vnode interlock is released. Reported by: kris@ Approved by: re (jhb)
2002-12-08 05:06:50 +00:00
*/
Do not set a variable (vp->p_pollinfo) to NULL if we know it already has that value. Approved by: re
2002-11-27 16:45:54 +00:00
vp->v_pollinfo = NULL;
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
}
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
#ifdef MAC
mac_destroy_vnode(vp);
#endif
- 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 = 0;
vp->v_vflag = 0;
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
vp->v_lastw = 0;
vp->v_lasta = 0;
vp->v_cstart = 0;
vp->v_clen = 0;
vp->v_socket = 0;
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
lockdestroy(vp->v_vnlock);
lockinit(vp->v_vnlock, PVFS, tag, VLKTIMEOUT, LK_NOPAUSE);
Convert KASSERTS to VNASSERTS
2005-02-17 10:28:58 +00:00
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"));
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
} else {
- 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
numvnodes++;
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(&vnode_free_list_mtx);
- 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
Back out M_* changes, per decision of the TRB. Approved by: trb
2003-02-19 05:47:46 +00:00
vp = (struct vnode *) uma_zalloc(vnode_zone, M_WAITOK|M_ZERO);
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(&vp->v_interlock, "vnode interlock", NULL, MTX_DEF);
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
vp->v_dd = vp;
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
bo = &vp->v_bufobj;
Give the bufobj a private __bo_vnode for now to keep the syncer floating [1] At some point later the syncer will unlearn about vnodes and the filesystems method called by the syncer will know enough about what's in bo_private to do the right thing. [1] Ok, I know, but I couldn't resist the pun.
2004-10-29 09:33:32 +00:00
bo->__bo_vnode = vp;
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
bo->bo_mtx = &vp->v_interlock;
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
vp->v_vnlock = &vp->v_lock;
lockinit(vp->v_vnlock, PVFS, tag, VLKTIMEOUT, LK_NOPAUSE);
- 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
cache_purge(vp); /* Sets up v_id. */
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
LIST_INIT(&vp->v_cache_src);
TAILQ_INIT(&vp->v_cache_dst);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
}
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 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
TAILQ_INIT(&bo->bo_clean.bv_hd);
TAILQ_INIT(&bo->bo_dirty.bv_hd);
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;
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;
*vpp = vp;
vp->v_usecount = 1;
vp->v_data = 0;
To avoid lock order reversals in getnewvnode(), the call to uma_zfree() must be delayed until the vnode interlock is released. Reported by: kris@ Approved by: re (jhb)
2002-12-08 05:06:50 +00:00
if (pollinfo != 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
knlist_destroy(&pollinfo->vpi_selinfo.si_note);
To avoid lock order reversals in getnewvnode(), the call to uma_zfree() must be delayed until the vnode interlock is released. Reported by: kris@ Approved by: re (jhb)
2002-12-08 05:06:50 +00:00
mtx_destroy(&pollinfo->vpi_lock);
uma_zfree(vnodepoll_zone, pollinfo);
}
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
mac_init_vnode(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)
mac_associate_vnode_singlelabel(mp, vp);
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
else if (mp == NULL)
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
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);
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) {
Revision 1.496 would not boot on my system due to ffs_mount -> bdevvp -> getnewvnode(..., mp = NULL, ...) -> insmntqueue(vp, mp = NULL) -> KASSERT -> panic Make getnewvnode() only call insmntqueue() if the mountpoint parameter is not NULL.
2004-07-04 10:19:15 +00:00
insmntque(vp, mp);
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
bo->bo_bsize = mp->mnt_stat.f_iosize;
}
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
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
if (vp->v_mount == NULL)
return;
mp = vp->v_mount;
MNT_ILOCK(mp);
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--;
MNT_IUNLOCK(mp);
}
/*
* Insert into list of vnodes for the new mount point, if available.
*/
static void
insmntque(struct vnode *vp, struct mount *mp)
{
vp->v_mount = mp;
Convert KASSERTS to VNASSERTS
2005-02-17 10:28:58 +00:00
VNASSERT(mp != NULL, vp, ("Don't call insmntque(foo, NULL)"));
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(vp->v_mount);
TAILQ_INSERT_TAIL(&mp->mnt_nvnodelist, vp, v_nmntvnodes);
mp->mnt_nvnodelistsize++;
MNT_IUNLOCK(vp->v_mount);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
}
/*
* Flush out and invalidate all buffers associated with a vnode.
* Called with the underlying object locked.
*/
int
Eliminate unused and unnecessary "cred" argument from vinvalbuf()
2005-01-14 07:33:51 +00:00
vinvalbuf(vp, flags, td, slpflag, 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 vnode *vp;
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
int flags;
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;
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
int slpflag, slptimeo;
{
- 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;
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
struct bufobj *bo;
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
ASSERT_VOP_LOCKED(vp, "vinvalbuf");
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
bo = &vp->v_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);
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);
Add BO_SYNC() and add a default which uses the secret vnode pointer and VOP_FSYNC() for now.
2005-01-11 10:43:08 +00:00
if ((error = BO_SYNC(bo, MNT_WAIT, td)) != 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,
flags, vp, slpflag, slptimeo);
if (error == 0)
error = flushbuflist(&bo->bo_dirty,
flags, vp, 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);
vm_object_pip_wait(bo->bo_object, "vnvlbx");
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.
*/
Access vmobject via the bufobj instead of the vnode
2005-02-07 10:04:06 +00:00
if (bo->bo_object != NULL) {
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);
}
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
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
flushbuflist(bufv, flags, vp, slpflag, slptimeo)
struct bufv *bufv;
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
int flags;
struct vnode *vp;
int slpflag, slptimeo;
{
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;
More vnode -> bufobj migration.
2005-01-11 10:16:39 +00:00
struct bufobj *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
More vnode -> bufobj migration.
2005-01-11 10:16:39 +00:00
bo = &vp->v_bufobj;
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
}
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
}
/*
* 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
* check for it. Note that vfs_bio_awrite expects
Properly vector all bwrite() and BUF_WRITE() calls through the same path and s/BUF_WRITE()/bwrite()/ since it now does the same as bwrite().
2004-03-11 18:02:36 +00:00
* buffers to reside on a queue, while bwrite and
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
* brelse do not.
*/
if (((bp->b_flags & (B_DELWRI | B_INVAL)) == B_DELWRI) &&
(flags & V_SAVE)) {
if (bp->b_vp == vp) {
if (bp->b_flags & B_CLUSTEROK) {
vfs_bio_awrite(bp);
} else {
bremfree(bp);
bp->b_flags |= B_ASYNC;
Properly vector all bwrite() and BUF_WRITE() calls through the same path and s/BUF_WRITE()/bwrite()/ since it now does the same as bwrite().
2004-03-11 18:02:36 +00:00
bwrite(bp);
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
}
} else {
bremfree(bp);
Properly vector all bwrite() and BUF_WRITE() calls through the same path and s/BUF_WRITE()/bwrite()/ since it now does the same as bwrite().
2004-03-11 18:02:36 +00:00
(void) bwrite(bp);
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
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 (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
}
bremfree(bp);
bp->b_flags |= (B_INVAL | B_NOCACHE | B_RELBUF);
bp->b_flags &= ~B_ASYNC;
brelse(bp);
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
}
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
Polish vtruncbuf() to improve readability and style a bit.
2004-10-21 14:13:54 +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
/*
* 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:
- 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);
Polish vtruncbuf() to improve readability and style a bit.
2004-10-21 14:13:54 +00:00
bo = &vp->v_bufobj;
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,
VI_MTX(vp)) == ENOLCK)
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
}
Polish vtruncbuf() to improve readability and style a bit.
2004-10-21 14:13:54 +00:00
VI_LOCK(vp);
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,
VI_MTX(vp)) == ENOLCK)
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
}
Polish vtruncbuf() to improve readability and style a bit.
2004-10-21 14:13:54 +00:00
VI_LOCK(vp);
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,
VI_MTX(vp)) == ENOLCK) {
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);
VI_LOCK(vp);
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);
- 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_UNLOCK(vp);
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);
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->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);
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
{
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));
- Hold the vnode interlock across calls to bgetvp instead of acquiring it internally. This is required to stop multiple bufs from being associated with a single lblkno.
2003-03-02 06:05:23 +00:00
ASSERT_VI_LOCKED(vp, "bgetvp");
- 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
vholdl(vp);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
bp->b_vp = 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
bp->b_bufobj = &vp->v_bufobj;
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
/*
* Insert onto list for new vnode.
*/
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, &vp->v_bufobj, BX_VNCLEAN);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
}
/*
* Disassociate a buffer from 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
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);
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);
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
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
}
- 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
vdropl(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
bp->b_vp = NULL;
bp->b_bufobj = 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
BO_UNLOCK(bo);
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
{
- 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 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;
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++;
}
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;
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_INSERT_HEAD(&syncer_workitem_pending[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");
Reviewed by: Many at differnt times in differnt parts, including alan, john, me, luoqi, and kirk Submitted by: Matt Dillon <dillon@frebsd.org> This change implements a relatively sophisticated fix to getnewbuf(). There were two problems with getnewbuf(). First, the writerecursion can lead to a system stack overflow when you have NFS and/or VN devices in the system. Second, the free/dirty buffer accounting was completely broken. Not only did the nfs routines blow it trying to manually account for the buffer state, but the accounting that was done did not work well with the purpose of their existance: figuring out when getnewbuf() needs to sleep. The meat of the change is to kern/vfs_bio.c. The remaining diffs are all minor except for NFS, which includes both the fixes for bp interaction AND fixes for a 'biodone(): buffer already done' lockup. Sys/buf.h also contains a chaining structure which is not used by this patchset but is used by other patches that are coming soon. This patch deliniated by tags PRE_MAT_GETBUF and POST_MAT_GETBUF. (sorry for the missing T matt)
1999-03-12 02:24:58 +00:00
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
};
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
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
sync_vnode(struct bufobj *bo, struct thread *td)
{
struct vnode *vp;
struct mount *mp;
vp = bo->__bo_vnode; /* XXX */
if (VOP_ISLOCKED(vp, NULL) != 0)
return (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
if (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);
return (1);
}
vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
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);
Move the bit of the syncer which deals with vnodes into a separate function.
2004-11-14 15:24:38 +00:00
VOP_UNLOCK(vp, 0, td);
vn_finished_write(mp);
VI_LOCK(vp);
if ((bo->bo_flag & BO_ONWORKLST) != 0) {
/*
* 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.
*/
vn_syncer_add_to_worklist(bo, syncdelay);
}
vdropl(vp);
VI_UNLOCK(vp);
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)
{
- 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
struct synclist *next;
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 synclist *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;
Kill some XXXKSE's. vnlru/syncer are single threaded.
2004-03-29 22:45:33 +00:00
struct thread *td = FIRST_THREAD_IN_PROC(updateproc);
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 dummychan;
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
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(&Giant);
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;
starttime = time_second;
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
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 (;;) {
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
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);
kthread_suspend_check(td->td_proc);
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 &&
starttime != time_second) {
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
}
Eradicate the variable "time" from the kernel, using various measures. "time" wasn't a atomic variable, so splfoo() protection were needed around any access to it, unless you just wanted the seconds part. Most uses of time.tv_sec now uses the new variable time_second instead. gettime() changed to getmicrotime(0. Remove a couple of unneeded splfoo() protections, the new getmicrotime() is atomic, (until Bruce sets a breakpoint in it). A couple of places needed random data, so use read_random() instead of mucking about with time which isn't random. Add a new nfs_curusec() function. Mark a couple of bogosities involving the now disappeard time variable. Update ffs_update() to avoid the weird "== &time" checks, by fixing the one remaining call that passwd &time as args. Change profiling in ncr.c to use ticks instead of time. Resolution is the same. Add new function "tvtohz()" to avoid the bogus "splfoo(), add time, call hzto() which subtracts time" sequences. Reviewed by: bde
1998-03-30 09:56:58 +00:00
starttime = time_second;
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 {
slp = &syncer_workitem_pending[syncer_delayno];
syncer_delayno += 1;
if (syncer_delayno == syncer_maxdelay)
syncer_delayno = 0;
next = &syncer_workitem_pending[syncer_delayno];
/*
* If the worklist has wrapped since the
* it was emptied of all but syncer vnodes,
* 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) &&
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;
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
while ((bo = LIST_FIRST(slp)) != NULL) {
Move the bit of the syncer which deals with vnodes into a separate function.
2004-11-14 15:24:38 +00:00
error = sync_vnode(bo, td);
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
}
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--;
- 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
/*
* Do soft update processing.
*/
Use indirect function pointer hooks instead of #ifdef SOFTUPDATES direct calls for the two places where the kernel calls into soft updates code. Set up the hooks in softdep_initialize() and NULL them out in softdep_uninitialize(). This change allows soft updates to function correctly when ufs is loaded as a module. Reviewed by: mckusick
2002-07-01 17:59:40 +00:00
if (softdep_process_worklist_hook != NULL)
(*softdep_process_worklist_hook)(NULL);
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.
*/
- 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);
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 (rushjob > 0) {
rushjob -= 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_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
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
}
- 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
* Just sleep for a short period if time between
* 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)
tsleep(&dummychan, PPAUSE, "syncfnl",
hz / SYNCER_SHUTDOWN_SPEEDUP);
else if (time_second == starttime)
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
tsleep(&lbolt, PPAUSE, "syncer", 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
/*
* 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
speedup_syncer()
{
Completely redo thread states. Reviewed by: davidxu@freebsd.org
2002-09-11 08:13:56 +00:00
struct thread *td;
- 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
Completely redo thread states. Reviewed by: davidxu@freebsd.org
2002-09-11 08:13:56 +00:00
td = FIRST_THREAD_IN_PROC(updateproc);
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
sleepq_remove(td, &lbolt);
- 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);
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)
{
struct thread *td;
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
td = FIRST_THREAD_IN_PROC(updateproc);
sleepq_remove(td, &lbolt);
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);
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;
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.
*/
- 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
VI_LOCK(vp);
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);
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);
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
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
}
Eliminate unused second argument to reassignbuf() and simplify it accordingly.
2004-07-25 21:24:23 +00:00
VI_UNLOCK(vp);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
}
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
v_incr_usecount(struct vnode *vp, int delta)
{
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
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
vp->v_usecount += delta;
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();
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
vp->v_rdev->si_usecount += delta;
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
Correct a couple of spelling errors in comments.
1999-07-12 15:02:51 +00:00
* reference count and lock it. The vnode lock bit is set if the
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
* vnode is being eliminated in vgone. The process is awakened
* when the transition is completed, and an error returned to
* indicate that the vnode is no longer usable (possibly having
More s/file system/filesystem/g
2002-05-16 21:28:32 +00:00
* been changed to a new filesystem type).
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
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
vget(vp, flags, td)
Remove register keyword
2005-01-28 12:39:10 +00:00
struct vnode *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
int flags;
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;
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
/*
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 the vnode is in the process of being cleaned out for
* another use, we wait for the cleaning to finish and then
* return failure. Cleaning is determined by checking that
- 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
* the VI_XLOCK flag is set.
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +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
if ((flags & LK_INTERLOCK) == 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
VI_LOCK(vp);
v_vxproc was a bogus name for a thread (pointer).
2003-12-28 09:12:56 +00:00
if (vp->v_iflag & VI_XLOCK && vp->v_vxthread != curthread) {
- 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
if ((flags & LK_NOWAIT) == 0) {
vp->v_iflag |= VI_XWANT;
msleep(vp, VI_MTX(vp), PINOD | PDROP, "vget", 0);
- In vget() if LK_NOWAIT is specified we should return EBUSY and not ENOENT. Submitted by: Stephan Uphoff <ups@stups.com>
2003-12-16 17:08:27 +00:00
return (ENOENT);
}
VI_UNLOCK(vp);
return (EBUSY);
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
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
v_incr_usecount(vp, 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
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);
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 (flags & LK_TYPE_MASK) {
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 ((error = vn_lock(vp, flags | LK_INTERLOCK, td)) != 0) {
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
/*
* must expand vrele here because we do not want
* to call VOP_INACTIVE if the reference count
* drops back to zero since it was never really
* active. We must remove it from the free list
* before sleeping so that multiple processes do
* not try to recycle it.
*/
- 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);
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
v_incr_usecount(vp, -1);
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
if (VSHOULDFREE(vp))
vfree(vp);
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
else
vlruvp(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
VI_UNLOCK(vp);
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
}
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 (error);
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
}
- 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_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
- 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 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
v_incr_usecount(vp, 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
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
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
vrele(vp)
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
struct vnode *vp;
{
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"));
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"));
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)) {
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
v_incr_usecount(vp, -1);
- 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_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
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;
}
if (vp->v_usecount == 1) {
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
v_incr_usecount(vp, -1);
In the routines vrele() and vput(), we must lock the vnode and call VOP_INACTIVE before placing the vnode back on the free list. Otherwise there is a race condition on SMP machines between getnewvnode() locking the vnode to reclaim it and vrele() locking the vnode to inactivate it. This window of vulnerability becomes exaggerated in the presence of filesystems that have been suspended as the inactive routine may need to temporarily release the lock on the vnode to avoid deadlock with the syncer process.
2002-02-02 01:49:18 +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
* We must call VOP_INACTIVE with the node locked. Mark
* as VI_DOINGINACT to avoid recursion.
In the routines vrele() and vput(), we must lock the vnode and call VOP_INACTIVE before placing the vnode back on the free list. Otherwise there is a race condition on SMP machines between getnewvnode() locking the vnode to reclaim it and vrele() locking the vnode to inactivate it. This window of vulnerability becomes exaggerated in the presence of filesystems that have been suspended as the inactive routine may need to temporarily release the lock on the vnode to avoid deadlock with the syncer process.
2002-02-02 01:49:18 +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 (vn_lock(vp, LK_EXCLUSIVE | LK_INTERLOCK, td) == 0) {
VI_LOCK(vp);
Convert KASSERTS to VNASSERTS
2005-02-17 10:28:58 +00:00
VNASSERT((vp->v_iflag & VI_DOINGINACT) == 0, 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
("vrele: recursed on VI_DOINGINACT"));
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
vp->v_iflag |= VI_DOINGINACT;
VI_UNLOCK(vp);
In the routines vrele() and vput(), we must lock the vnode and call VOP_INACTIVE before placing the vnode back on the free list. Otherwise there is a race condition on SMP machines between getnewvnode() locking the vnode to reclaim it and vrele() locking the vnode to inactivate it. This window of vulnerability becomes exaggerated in the presence of filesystems that have been suspended as the inactive routine may need to temporarily release the lock on the vnode to avoid deadlock with the syncer process.
2002-02-02 01:49:18 +00:00
VOP_INACTIVE(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_LOCK(vp);
Convert KASSERTS to VNASSERTS
2005-02-17 10:28:58 +00:00
VNASSERT(vp->v_iflag & VI_DOINGINACT, vp,
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
("vrele: lost VI_DOINGINACT"));
vp->v_iflag &= ~VI_DOINGINACT;
Eliminate one case of VI_UNLOCK followed by an immediate VI_LOCK.
2003-09-19 19:13:54 +00:00
} else
VI_LOCK(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
if (VSHOULDFREE(vp))
vfree(vp);
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
else
vlruvp(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
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
} else {
#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
}
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
* re-aquiring the lock (as vrele() aquires 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
vput(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
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 */
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"));
- 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"));
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)) {
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
v_incr_usecount(vp, -1);
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
VOP_UNLOCK(vp, LK_INTERLOCK, td);
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
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 (vp->v_usecount == 1) {
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
v_incr_usecount(vp, -1);
In the routines vrele() and vput(), we must lock the vnode and call VOP_INACTIVE before placing the vnode back on the free list. Otherwise there is a race condition on SMP machines between getnewvnode() locking the vnode to reclaim it and vrele() locking the vnode to inactivate it. This window of vulnerability becomes exaggerated in the presence of filesystems that have been suspended as the inactive routine may need to temporarily release the lock on the vnode to avoid deadlock with the syncer process.
2002-02-02 01:49:18 +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
* We must call VOP_INACTIVE with the node locked, so
* we just need to release the vnode mutex. Mark as
* as VI_DOINGINACT to avoid recursion.
In the routines vrele() and vput(), we must lock the vnode and call VOP_INACTIVE before placing the vnode back on the free list. Otherwise there is a race condition on SMP machines between getnewvnode() locking the vnode to reclaim it and vrele() locking the vnode to inactivate it. This window of vulnerability becomes exaggerated in the presence of filesystems that have been suspended as the inactive routine may need to temporarily release the lock on the vnode to avoid deadlock with the syncer process.
2002-02-02 01:49:18 +00:00
*/
Convert KASSERTS to VNASSERTS
2005-02-17 10:28:58 +00:00
VNASSERT((vp->v_iflag & VI_DOINGINACT) == 0, 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
("vput: recursed on VI_DOINGINACT"));
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
vp->v_iflag |= VI_DOINGINACT;
- 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_UNLOCK(vp);
In the routines vrele() and vput(), we must lock the vnode and call VOP_INACTIVE before placing the vnode back on the free list. Otherwise there is a race condition on SMP machines between getnewvnode() locking the vnode to reclaim it and vrele() locking the vnode to inactivate it. This window of vulnerability becomes exaggerated in the presence of filesystems that have been suspended as the inactive routine may need to temporarily release the lock on the vnode to avoid deadlock with the syncer process.
2002-02-02 01:49:18 +00:00
VOP_INACTIVE(vp, td);
- 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);
Convert KASSERTS to VNASSERTS
2005-02-17 10:28:58 +00:00
VNASSERT(vp->v_iflag & VI_DOINGINACT, vp,
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
("vput: lost VI_DOINGINACT"));
vp->v_iflag &= ~VI_DOINGINACT;
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
if (VSHOULDFREE(vp))
vfree(vp);
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
else
vlruvp(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
VI_UNLOCK(vp);
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
} else {
#ifdef DIAGNOSTIC
vprint("vput: negative ref count", vp);
#endif
panic("vput: negative ref cnt");
}
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);
}
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
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
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);
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
Make various vnode related functions static
2005-02-10 12:28:58 +00:00
static void
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
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
if (vp->v_holdcnt <= 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
panic("vdrop: holdcnt");
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 (VSHOULDFREE(vp))
vfree(vp);
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
else
vlruvp(vp);
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
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
vflush(mp, rootrefs, flags, td)
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
struct mount *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
int rootrefs;
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
int flags;
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
struct thread *td;
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +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
struct vnode *vp, *nvp, *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
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.
*/
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
if ((error = VFS_ROOT(mp, &rootvp, td)) != 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
return (error);
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:
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_VNODE_FOREACH(vp, mp, nvp) {
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);
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);
- 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
error = vn_lock(vp, LK_INTERLOCK | LK_EXCLUSIVE, td);
if (error) {
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);
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)) {
- 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
VOP_UNLOCK(vp, 0, td);
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) {
error = VOP_GETATTR(vp, &vattr, td->td_ucred, td);
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_UNLOCK(vp, LK_INTERLOCK, td);
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);
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
- 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
VOP_UNLOCK(vp, 0, td);
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.
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
*/
if (vp->v_usecount == 0) {
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
vgonel(vp, td);
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;
}
Clean up the rundown of the object backing a vnode. This should fix NFS problems associated with forcible dismounts.
1996-10-17 02:49:35 +00:00
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 FORCECLOSE is set, forcibly close the vnode. For block
* or character devices, revert to an anonymous device. For
* all other files, just kill them.
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
*/
if (flags & FORCECLOSE) {
Convert KASSERTS to VNASSERTS
2005-02-17 10:28:58 +00:00
VNASSERT(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));
vgonel(vp, td);
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;
}
#ifdef DIAGNOSTIC
if (busyprt)
vprint("vflush: busy vnode", vp);
#endif
- 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_UNLOCK(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
busy++;
}
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) {
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
vgonel(rootvp, 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
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
}
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
if (busy)
return (EBUSY);
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);
}
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
/*
* This moves a now (likely recyclable) vnode to the end of the
* mountlist. XXX However, it is temporarily disabled until we
* can clean up ffs_sync() and friends, which have loop restart
* conditions which this code causes to operate O(N^2).
*/
static void
vlruvp(struct vnode *vp)
{
#if 0
struct mount *mp;
if ((mp = vp->v_mount) != NULL) {
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 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
TAILQ_REMOVE(&mp->mnt_nvnodelist, vp, v_nmntvnodes);
TAILQ_INSERT_TAIL(&mp->mnt_nvnodelist, vp, v_nmntvnodes);
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);
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
}
#endif
}
- 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
static void
vx_lock(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
- 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
ASSERT_VI_LOCKED(vp, "vx_lock");
/*
* Prevent the vnode from being recycled or brought into use while we
* clean it out.
*/
if (vp->v_iflag & VI_XLOCK)
Integrate vclean() into vgonel(). Various associated polishing.
2005-01-28 13:00:03 +00:00
panic("vx_lock: deadlock");
- 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
vp->v_iflag |= VI_XLOCK;
v_vxproc was a bogus name for a thread (pointer).
2003-12-28 09:12:56 +00:00
vp->v_vxthread = curthread;
- 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
}
static void
vx_unlock(struct vnode *vp)
{
ASSERT_VI_LOCKED(vp, "vx_unlock");
vp->v_iflag &= ~VI_XLOCK;
v_vxproc was a bogus name for a thread (pointer).
2003-12-28 09:12:56 +00:00
vp->v_vxthread = NULL;
- 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
if (vp->v_iflag & VI_XWANT) {
vp->v_iflag &= ~VI_XWANT;
wakeup(vp);
}
}
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.
* Release the passed interlock if the vnode will be recycled.
*/
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
{
VI_LOCK(vp);
if (vp->v_usecount == 0) {
vgonel(vp, td);
return (1);
}
VI_UNLOCK(vp);
return (0);
}
- 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)
{
struct thread *td = curthread; /* XXX */
VI_LOCK(vp);
vgonel(vp, td);
}
/*
* vgone, with the vp interlock held.
*/
void
vgonel(struct vnode *vp, struct thread *td)
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +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
int active;
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
/*
* If a vgone (or vclean) is already in progress,
* wait until it is done and return.
*/
ASSERT_VI_LOCKED(vp, "vgonel");
if (vp->v_iflag & VI_XLOCK) {
vp->v_iflag |= VI_XWANT;
msleep(vp, VI_MTX(vp), PINOD | PDROP, "vgone", 0);
return;
}
vx_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
* Check to see if the vnode is in use. If so we have to reference it
* before we clean it out so that its count cannot fall to zero and
* generate a race against ourselves to recycle it.
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
*/
While in the real world, I had a bad case of being swapped out for a lot of cycles. While waiting there I added a lot of the extra ()'s I have, (I have never used LISP to any extent). So I compiled the kernel with -Wall and shut up a lot of "suggest you add ()'s", removed a bunch of unused var's and added a couple of declarations here and there. Having a lap-top is highly recommended. My kernel still runs, yell at me if you kernel breaks.
1994-09-25 19:34:02 +00:00
if ((active = vp->v_usecount))
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
v_incr_usecount(vp, 1);
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
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
* Even if the count is zero, the VOP_INACTIVE routine may still
* have the object locked while it cleans it out. The VOP_LOCK
* ensures that the VOP_INACTIVE routine is done with its work.
* For active vnodes, it ensures that no other activity can
* occur while the underlying object is being cleaned out.
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
VOP_LOCK(vp, LK_DRAIN | LK_INTERLOCK, td);
Remove a window during running down a file vnode. Also, the OBJ_DEAD flag wasn't being respected during vref(), et. al. Note that this isn't the eventual fix for the locking problem. Fine grained SMP in the VM and VFS code will require (lots) more work.
1997-06-22 03:00:24 +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
* 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
*/
Eliminate the constant flags argument to vclean()
2005-01-24 22:22:02 +00:00
if (!TAILQ_EMPTY(&vp->v_bufobj.bo_dirty.bv_hd));
(void) vn_write_suspend_wait(vp, NULL, V_WAIT);
if (vinvalbuf(vp, V_SAVE, td, 0, 0) != 0)
vinvalbuf(vp, 0, td, 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
- Don't hold the vn lock while calling VOP_CLOSE in vclean().
2002-07-07 06:38:22 +00:00
/*
* Any other processes trying to obtain this lock must first
* wait for VXLOCK to clear, then call the new lock operation.
*/
VOP_UNLOCK(vp, 0, 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
* If purging an active vnode, it must be closed and
* deactivated before being reclaimed. Note that the
* VOP_INACTIVE will unlock the vnode.
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +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);
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);
if ((vp->v_iflag & VI_DOINGINACT) == 0) {
Convert KASSERTS to VNASSERTS
2005-02-17 10:28:58 +00:00
VNASSERT((vp->v_iflag & VI_DOINGINACT) == 0, 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
("vclean: recursed on VI_DOINGINACT"));
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
vp->v_iflag |= VI_DOINGINACT;
VI_UNLOCK(vp);
if (vn_lock(vp, LK_EXCLUSIVE | LK_NOWAIT, td) != 0)
panic("vclean: cannot relock.");
VOP_INACTIVE(vp, td);
VI_LOCK(vp);
Convert KASSERTS to VNASSERTS
2005-02-17 10:28:58 +00:00
VNASSERT(vp->v_iflag & VI_DOINGINACT, vp,
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
("vclean: lost VI_DOINGINACT"));
vp->v_iflag &= ~VI_DOINGINACT;
}
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))
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
panic("vclean: cannot reclaim");
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
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));
This patch fixes a locking bug that can result in deadlock if the codepath is followed. From the PR: vclean calls vrele leading to deadlock (if usecount > 0) vclean() calls vrele() if v_usecount of the node was higher than one. But before calling it, it sets the VXLOCK flag, which will make vn_lock called from vrele dead-lock. PR: kern/15117 Submitted by: Assar Westerlund <assar@stacken.kth.se> Reviewed by: rwatson Obtained from: NetBSD
2000-01-29 15:22:58 +00:00
if (active) {
/*
* Inline copy of vrele() since VOP_INACTIVE
* has already been called.
*/
- 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);
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
v_incr_usecount(vp, -1);
if (vp->v_usecount <= 0) {
- 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
#ifdef INVARIANTS
This patch fixes a locking bug that can result in deadlock if the codepath is followed. From the PR: vclean calls vrele leading to deadlock (if usecount > 0) vclean() calls vrele() if v_usecount of the node was higher than one. But before calling it, it sets the VXLOCK flag, which will make vn_lock called from vrele dead-lock. PR: kern/15117 Submitted by: Assar Westerlund <assar@stacken.kth.se> Reviewed by: rwatson Obtained from: NetBSD
2000-01-29 15:22:58 +00:00
if (vp->v_usecount < 0 || vp->v_writecount != 0) {
vprint("vclean: bad ref count", vp);
panic("vclean: ref cnt");
}
#endif
- 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
if (VSHOULDFREE(vp))
vfree(vp);
This patch fixes a locking bug that can result in deadlock if the codepath is followed. From the PR: vclean calls vrele leading to deadlock (if usecount > 0) vclean() calls vrele() if v_usecount of the node was higher than one. But before calling it, it sets the VXLOCK flag, which will make vn_lock called from vrele dead-lock. PR: kern/15117 Submitted by: Assar Westerlund <assar@stacken.kth.se> Reviewed by: rwatson Obtained from: NetBSD
2000-01-29 15:22:58 +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_UNLOCK(vp);
This patch fixes a locking bug that can result in deadlock if the codepath is followed. From the PR: vclean calls vrele leading to deadlock (if usecount > 0) vclean() calls vrele() if v_usecount of the node was higher than one. But before calling it, it sets the VXLOCK flag, which will make vn_lock called from vrele dead-lock. PR: kern/15117 Submitted by: Assar Westerlund <assar@stacken.kth.se> Reviewed by: rwatson Obtained from: NetBSD
2000-01-29 15:22:58 +00:00
}
- 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);
- 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);
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
if (VSHOULDFREE(vp))
vfree(vp);
Gratuitous whitespace cleanup.
2002-06-06 15:46:38 +00:00
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
/*
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
* Done with purge, reset to the standard lock and
* notify sleepers of the grim news.
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
*/
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";
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
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
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 special device, remove it from special device alias list
* if it is on one.
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
*/
- 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
VI_LOCK(vp);
Fix a LOR relating to freeing cdevs.
2004-10-01 06:33:39 +00:00
if (vp->v_type == VCHR && vp->v_rdev != NULL)
dev_rel(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
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 it is on the freelist and not already at the head,
Simplify and rationalise the management of the vnode free list (preparing the code to add snapshots).
2000-07-04 04:32:40 +00:00
* move it to the head of the list. The test of the
* VDOOMED flag and the reference count of zero is because
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
* it will be removed from the free list by getnewvnode,
* but will not have its reference count incremented until
* after calling vgone. If the reference count were
* incremented first, vgone would (incorrectly) try to
* close the previous instance of the underlying object.
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 (vp->v_usecount == 0 && !(vp->v_iflag & VI_DOOMED)) {
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);
- 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) {
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
TAILQ_REMOVE(&vnode_free_list, vp, v_freelist);
- 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
} else {
vp->v_iflag |= VI_FREE;
Update freevnodes when adding a vnode to the head of the free list.
1998-01-31 01:17:58 +00:00
freevnodes++;
- 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
}
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);
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(&vnode_free_list_mtx);
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
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
vp->v_type = VBAD;
- 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
vx_unlock(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
VI_UNLOCK(vp);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
}
/*
* Lookup a vnode by device number.
*/
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
Remove unused second arg to vfinddev(). Don't call addaliasu() on VBLK nodes.
2004-03-11 16:33:11 +00:00
vfinddev(dev, vpp)
Do the dreaded s/dev_t/struct cdev */ Bump __FreeBSD_version accordingly.
2004-06-16 09:47:26 +00:00
struct cdev *dev;
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
struct vnode **vpp;
{
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
struct vnode *vp;
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +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 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
SLIST_FOREACH(vp, &dev->si_hlist, v_specnext) {
Remove unused second arg to vfinddev(). Don't call addaliasu() on VBLK nodes.
2004-03-11 16:33:11 +00:00
*vpp = vp;
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();
Remove unused second arg to vfinddev(). Don't call addaliasu() on VBLK nodes.
2004-03-11 16:33:11 +00:00
return (1);
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +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_unlock();
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
return (0);
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
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
vcount(vp)
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
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
count_dev(dev)
Do the dreaded s/dev_t/struct cdev */ Bump __FreeBSD_version accordingly.
2004-06-16 09:47:26 +00:00
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[] =
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
{"VNON", "VREG", "VDIR", "VBLK", "VCHR", "VLNK", "VSOCK", "VFIFO", "VBAD"};
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;
Examine all occurrences of sprintf(), strcat(), and str[n]cpy() for possible buffer overflow problems. Replaced most sprintf()'s with snprintf(); for others cases, added terminating NUL bytes where appropriate, replaced constants like "16" with sizeof(), etc. These changes include several bug fixes, but most changes are for maintainability's sake. Any instance where it wasn't "immediately obvious" that a buffer overflow could not occur was made safer. Reviewed by: Bruce Evans <bde@zeta.org.au> Reviewed by: Matthew Dillon <dillon@apollo.backplane.com> Reviewed by: Mike Spengler <mks@networkcs.com>
1998-12-04 22:54:57 +00:00
char buf[96];
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)
strcat(buf, "|VV_ROOT");
if (vp->v_vflag & VV_TEXT)
strcat(buf, "|VV_TEXT");
if (vp->v_vflag & VV_SYSTEM)
strcat(buf, "|VV_SYSTEM");
if (vp->v_iflag & VI_XLOCK)
strcat(buf, "|VI_XLOCK");
if (vp->v_iflag & VI_XWANT)
strcat(buf, "|VI_XWANT");
if (vp->v_iflag & VI_DOOMED)
strcat(buf, "|VI_DOOMED");
if (vp->v_iflag & VI_FREE)
strcat(buf, "|VI_FREE");
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");
Change vprint() to vn_printf() which takes varargs. Add #define for vprint() to call vn_printf().
2005-01-24 13:58:08 +00:00
if (vp->v_object != NULL);
printf(" v_object %p\n", vp->v_object);
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);
Remove any VOP_PRINT that redundantly prints the tag. Move lockmgr_printinfo() into vprint() for everyone's benefit. Suggested by: bde
2002-09-18 20:42:04 +00:00
printf("\n");
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
Nitpicking and dusting performed on a train. Removes trivial warnings about unused variables, labels and other lint.
1998-10-25 17:44:59 +00:00
#include <ddb/ddb.h>
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.
*/
BSD 4.4 Lite Kernel Sources
1994-05-24 10:09:53 +00:00
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) {
Lock reporting and assertion changes. * lockstatus() and VOP_ISLOCKED() gets a new process argument and a new return value: LK_EXCLOTHER, when the lock is held exclusively by another process. * The ASSERT_VOP_(UN)LOCKED family is extended to use what this gives them * Extend the vnode_if.src format to allow more exact specification than locked/unlocked. This commit should not do any semantic changes unless you are using DEBUG_VFS_LOCKS. Discussed with: grog, mch, peter, phk Reviewed by: peter
1999-12-11 16:13:02 +00:00
if (VOP_ISLOCKED(vp, NULL))
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
}
}
#endif
- 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) {
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);
- 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) {
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 thread *td = req->td;
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) {
if (vfs_busy(mp, LK_NOWAIT, &mountlist_mtx, td))
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;
#define XV_COPY(field) xvn[n].xv_##field = vp->v_##field
XV_COPY(usecount);
XV_COPY(writecount);
XV_COPY(holdcnt);
XV_COPY(id);
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);
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
vfs_unbusy(mp, td);
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
vfs_unmountall()
{
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;
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 (curthread != NULL)
td = curthread;
Supply the correct process argument to dounmount when possible.
1998-05-17 19:38:55 +00:00
else
Fix a couple of style bugs introduced (or touched by) previous commit.
2002-02-07 23:06:26 +00:00
td = FIRST_THREAD_IN_PROC(initproc); /* XXX XXX proc0? */
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);
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
printf("unmount of %s failed (",
mp->mnt_stat.f_mntonname);
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 == 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)
{
Changes to support 1Tb filesizes. Pages are now named by an (object,index) pair instead of (object,offset) pair.
1995-12-11 04:58:34 +00:00
struct vnode *vp, *nvp;
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;
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
int tries;
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
tries = 5;
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);
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
loop:
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_FOREACH_SAFE(vp, &mp->mnt_nvnodelist, v_nmntvnodes, nvp) {
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 (vp->v_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
if (--tries > 0)
goto loop;
break;
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
VI_LOCK(vp);
- 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 (vp->v_iflag & VI_XLOCK) {
- 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);
If a file has been completely unlinked, stop automatically syncing the file. ffs will discard any pending dirty pages when it is closed, so we may as well not waste time trying to clean them. This doesn't stop other things from writing it out, eg: pageout, fsync(2) etc.
2001-08-27 06:09:56 +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
}
If a file has been completely unlinked, stop automatically syncing the file. ffs will discard any pending dirty pages when it is closed, so we may as well not waste time trying to clean them. This doesn't stop other things from writing it out, eg: pageout, fsync(2) etc.
2001-08-27 06:09:56 +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 ((vp->v_iflag & VI_OBJDIRTY) &&
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
(flags == MNT_WAIT || VOP_ISLOCKED(vp, NULL) == 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);
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
if (TAILQ_NEXT(vp, v_nmntvnodes) != nvp) {
if (--tries > 0)
goto loop;
break;
}
- 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.
*/
Simplify and rationalise the management of the vnode free list (preparing the code to add snapshots).
2000-07-04 04:32:40 +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
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);
Convert KASSERTS to VNASSERTS
2005-02-17 10:28:58 +00:00
VNASSERT((vp->v_iflag & VI_FREE) == 0, vp, ("vnode already free"));
- 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_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 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(&vnode_free_list_mtx);
- 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_AGE;
vp->v_iflag |= VI_FREE;
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
}
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
{
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, "vbusy");
Convert KASSERTS to VNASSERTS
2005-02-17 10:28:58 +00:00
VNASSERT((vp->v_iflag & VI_FREE) != 0, vp, ("vnode not free"));
- 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--;
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(&vnode_free_list_mtx);
- 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
- 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|VI_AGE);
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
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
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);
if (vp->v_pollinfo != NULL) {
uma_zfree(vnodepoll_zone, vi);
return;
}
vp->v_pollinfo = 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
mtx_init(&vp->v_pollinfo->vpi_lock, "vnode pollinfo", NULL, MTX_DEF);
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_init(&vp->v_pollinfo->vpi_selinfo.si_note,
&vp->v_pollinfo->vpi_lock);
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
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
vn_pollrecord(vp, td, 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
struct vnode *vp;
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;
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
short events;
{
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
Make v_addpollinfo() visible and non-inline. Have callers only call it as needed. Add necessary call in ufs_kqfilter(). Test-case found by: Andrew Gallatin <gallatin@cs.duke.edu>
2002-02-18 16:18:02 +00:00
if (vp->v_pollinfo == NULL)
Gratuitous whitespace cleanup.
2002-06-06 15:46:38 +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);
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
return events;
}
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);
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
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
/*
* 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 */
.vop_lock = vop_stdlock, /* lock */
.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
vfs_allocate_syncvnode(mp)
struct mount *mp;
{
struct vnode *vp;
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;
/*
* 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;
}
- 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);
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(&vp->v_bufobj,
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);
- 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_UNLOCK(vp);
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
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_fsync(ap)
struct vop_fsync_args /* {
struct vnode *a_vp;
struct ucred *a_cred;
int a_waitfor;
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 *a_td;
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
} */ *ap;
{
struct vnode *syncvp = ap->a_vp;
struct mount *mp = syncvp->v_mount;
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 = ap->a_td;
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
int error, asyncflag;
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);
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 (vfs_busy(mp, LK_EXCLUSIVE | LK_NOWAIT, &mountlist_mtx, td) != 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) {
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
vfs_unbusy(mp, td);
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);
}
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
asyncflag = mp->mnt_flag & MNT_ASYNC;
mp->mnt_flag &= ~MNT_ASYNC;
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);
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
error = VFS_SYNC(mp, MNT_LAZY, td);
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 (asyncflag)
mp->mnt_flag |= MNT_ASYNC;
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);
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
vfs_unbusy(mp, td);
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
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_inactive(ap)
struct vop_inactive_args /* {
struct vnode *a_vp;
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 *a_td;
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
} */ *ap;
{
- Backout the patch made in revision 1.75 of vfs_mount.c. The vputs here were hiding the real problem of the missing unlock in sync_inactive. - Add the missing unlock in sync_inactive. Submitted by: iedowse
2002-07-29 06:26:55 +00:00
VOP_UNLOCK(ap->a_vp, 0, ap->a_td);
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
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_reclaim(ap)
struct vop_reclaim_args /* {
struct vnode *a_vp;
} */ *ap;
{
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
- 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);
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;
- 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);
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--;
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
}
- 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_UNLOCK(vp);
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
Give vn_isdisk() a second argument where it can return a suitable errno. Suggested by: bde
2000-01-10 12:04:27 +00:00
vn_isdisk(vp, errp)
Introduce vn_isdisk(struct vnode *vp) function, and use it to test for diskness.
1999-08-25 12:24:39 +00:00
struct vnode *vp;
Give vn_isdisk() a second argument where it can return a suitable errno. Suggested by: bde
2000-01-10 12:04:27 +00:00
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
}
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
* Free data allocated by namei(); see namei(9) for details.
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
*/
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
void
NDFREE(ndp, flags)
struct nameidata *ndp;
Consistently use the BSD u_int and u_short instead of the SYSV uint and ushort. In most of these files, there was a mixture of both styles and this change just makes them self-consistent. Requested by: bde (kern_ktrace.c)
2003-08-07 15:04:27 +00:00
const u_int flags;
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
{
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 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
if (!(flags & NDF_NO_FREE_PNBUF) &&
(ndp->ni_cnd.cn_flags & HASBUF)) {
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
uma_zfree(namei_zone, ndp->ni_cnd.cn_pnbuf);
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
ndp->ni_cnd.cn_flags &= ~HASBUF;
}
if (!(flags & NDF_NO_DVP_UNLOCK) &&
(ndp->ni_cnd.cn_flags & LOCKPARENT) &&
ndp->ni_dvp != ndp->ni_vp)
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
VOP_UNLOCK(ndp->ni_dvp, 0, ndp->ni_cnd.cn_thread);
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
if (!(flags & NDF_NO_DVP_RELE) &&
(ndp->ni_cnd.cn_flags & (LOCKPARENT|WANTPARENT))) {
vrele(ndp->ni_dvp);
ndp->ni_dvp = NULL;
}
if (!(flags & NDF_NO_VP_UNLOCK) &&
(ndp->ni_cnd.cn_flags & LOCKLEAF) && ndp->ni_vp)
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
VOP_UNLOCK(ndp->ni_vp, 0, ndp->ni_cnd.cn_thread);
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
if (!(flags & NDF_NO_VP_RELE) &&
ndp->ni_vp) {
vrele(ndp->ni_vp);
ndp->ni_vp = NULL;
}
if (!(flags & NDF_NO_STARTDIR_RELE) &&
(ndp->ni_cnd.cn_flags & SAVESTART)) {
vrele(ndp->ni_startdir);
ndp->ni_startdir = NULL;
}
}
Centralize the canonical vop_access user/group/other check in vaccess(). Discussed with: bde
2000-08-20 08:36:26 +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.
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
Revert stuff which accidentally ended up in the previous commit.
2003-07-22 10:36:36 +00:00
vaccess(type, file_mode, file_uid, file_gid, acc_mode, cred, privused)
enum vtype type;
mode_t file_mode;
uid_t file_uid;
gid_t file_gid;
mode_t acc_mode;
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
{
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
mode_t dac_granted;
#ifdef CAPABILITIES
mode_t cap_granted;
#endif
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
if ((acc_mode & dac_granted) == acc_mode)
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
if ((acc_mode & dac_granted) == acc_mode)
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);
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 ((acc_mode & dac_granted) == acc_mode)
return (0);
privcheck:
Rename suser_cred()'s PRISON_ROOT flag to SUSER_ALLOWJAIL. This is somewhat clearer, but more importantly allows for a consistent naming scheme for suser_cred flags. The old name is still defined, but will be removed in a few days (unless I hear any complaints...) Discussed with: rwatson, scottl Requested by: jhb
2004-07-26 07:24:04 +00:00
if (!suser_cred(cred, SUSER_ALLOWJAIL)) {
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;
Centralize the canonical vop_access user/group/other check in vaccess(). Discussed with: bde
2000-08-20 08:36:26 +00:00
return (0);
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
}
#ifdef CAPABILITIES
/*
* Build a capability mask to determine if the set of capabilities
* satisfies the requirements when combined with the granted mask
* from above.
* For each capability, if the capability is required, bitwise
* or the request type onto the cap_granted mask.
*/
cap_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) {
/*
* For directories, use CAP_DAC_READ_SEARCH to satisfy
* VEXEC requests, instead of CAP_DAC_EXECUTE.
*/
if ((acc_mode & VEXEC) && ((dac_granted & VEXEC) == 0) &&
Rename suser_cred()'s PRISON_ROOT flag to SUSER_ALLOWJAIL. This is somewhat clearer, but more importantly allows for a consistent naming scheme for suser_cred flags. The old name is still defined, but will be removed in a few days (unless I hear any complaints...) Discussed with: rwatson, scottl Requested by: jhb
2004-07-26 07:24:04 +00:00
!cap_check(cred, NULL, CAP_DAC_READ_SEARCH, SUSER_ALLOWJAIL))
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
cap_granted |= VEXEC;
} else {
if ((acc_mode & VEXEC) && ((dac_granted & VEXEC) == 0) &&
Rename suser_cred()'s PRISON_ROOT flag to SUSER_ALLOWJAIL. This is somewhat clearer, but more importantly allows for a consistent naming scheme for suser_cred flags. The old name is still defined, but will be removed in a few days (unless I hear any complaints...) Discussed with: rwatson, scottl Requested by: jhb
2004-07-26 07:24:04 +00:00
!cap_check(cred, NULL, CAP_DAC_EXECUTE, SUSER_ALLOWJAIL))
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
cap_granted |= VEXEC;
}
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 ((acc_mode & VREAD) && ((dac_granted & VREAD) == 0) &&
Rename suser_cred()'s PRISON_ROOT flag to SUSER_ALLOWJAIL. This is somewhat clearer, but more importantly allows for a consistent naming scheme for suser_cred flags. The old name is still defined, but will be removed in a few days (unless I hear any complaints...) Discussed with: rwatson, scottl Requested by: jhb
2004-07-26 07:24:04 +00:00
!cap_check(cred, NULL, CAP_DAC_READ_SEARCH, SUSER_ALLOWJAIL))
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
cap_granted |= VREAD;
if ((acc_mode & VWRITE) && ((dac_granted & VWRITE) == 0) &&
Rename suser_cred()'s PRISON_ROOT flag to SUSER_ALLOWJAIL. This is somewhat clearer, but more importantly allows for a consistent naming scheme for suser_cred flags. The old name is still defined, but will be removed in a few days (unless I hear any complaints...) Discussed with: rwatson, scottl Requested by: jhb
2004-07-26 07:24:04 +00:00
!cap_check(cred, NULL, CAP_DAC_WRITE, SUSER_ALLOWJAIL))
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
cap_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
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
if ((acc_mode & VADMIN) && ((dac_granted & VADMIN) == 0) &&
Rename suser_cred()'s PRISON_ROOT flag to SUSER_ALLOWJAIL. This is somewhat clearer, but more importantly allows for a consistent naming scheme for suser_cred flags. The old name is still defined, but will be removed in a few days (unless I hear any complaints...) Discussed with: rwatson, scottl Requested by: jhb
2004-07-26 07:24:04 +00:00
!cap_check(cred, NULL, CAP_FOWNER, SUSER_ALLOWJAIL))
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
cap_granted |= VADMIN;
o Synchronize vaccess() capability access control checks with TrustedBSD tree. Obtained from: TrustedBSD Project
2000-09-06 12:18:24 +00:00
if ((acc_mode & (cap_granted | dac_granted)) == acc_mode) {
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);
}
#endif
Centralize the canonical vop_access user/group/other check in vaccess(). Discussed with: bde
2000-08-20 08:36:26 +00:00
o The move to using VADMIN under vaccess() resulted in some system calls returning EACCES instead of EPERM. This patch modifies vaccess() to return EPERM instead of EACCES if VADMIN is among the requested rights. This affects functions normally limited to the owners of a file, such as chmod(), as EPERM is the error indicating that privilege would allow the operation, rather than a chance in mandatory or discretionary rights. Reported by: bde
2001-01-23 04:15:19 +00:00
return ((acc_mode & 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
extattr_check_cred(struct vnode *vp, int attrnamespace,
struct ucred *cred, struct thread *td, int access)
{
/*
* 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);
/*
* Do not allow privileged processes in jail to directly
* manipulate system attributes.
*
* XXX What capability should apply here?
* Probably CAP_SYS_SETFFLAG.
*/
switch (attrnamespace) {
case EXTATTR_NAMESPACE_SYSTEM:
/* Potentially should be: return (EPERM); */
return (suser_cred(cred, 0));
case EXTATTR_NAMESPACE_USER:
return (VOP_ACCESS(vp, access, cred, td));
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.
*/
#define IGNORE_LOCK(vp) ((vp)->v_type == VCHR || (vp)->v_type == VBAD)
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)
Update for the KDB framework: o Call kdb_enter() instead of Debugger().
2004-07-10 21:47:53 +00:00
kdb_enter("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)
{
More style fixes. Obtained from: bde
2004-01-05 23:40:46 +00:00
if (vp && !IGNORE_LOCK(vp) && VOP_ISLOCKED(vp, NULL) == 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)
{
if (vp && !IGNORE_LOCK(vp) &&
VOP_ISLOCKED(vp, curthread) == LK_EXCLUSIVE)
vfs_badlock("is locked but should not be", str, vp);
}
#if 0
void
assert_vop_elocked(struct vnode *vp, const char *str)
{
if (vp && !IGNORE_LOCK(vp) &&
VOP_ISLOCKED(vp, curthread) != LK_EXCLUSIVE)
vfs_badlock("is not exclusive locked but should be", str, vp);
}
void
assert_vop_elocked_other(struct vnode *vp, const char *str)
{
if (vp && !IGNORE_LOCK(vp) &&
VOP_ISLOCKED(vp, curthread) != LK_EXCLOTHER)
vfs_badlock("is not exclusive locked by another thread",
str, vp);
}
void
assert_vop_slocked(struct vnode *vp, const char *str)
{
if (vp && !IGNORE_LOCK(vp) &&
VOP_ISLOCKED(vp, curthread) != LK_SHARED)
vfs_badlock("is not locked shared but should be", str, vp);
}
#endif /* 0 */
void
vop_rename_pre(void *ap)
{
struct vop_rename_args *a = ap;
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). */
if (a->a_tdvp != a->a_fdvp)
ASSERT_VOP_UNLOCKED(a->a_fdvp, "vop_rename: fdvp locked");
if (a->a_tvp != a->a_fvp)
ASSERT_VOP_UNLOCKED(a->a_fvp, "vop_rename: tvp locked");
/* 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");
}
void
vop_strategy_pre(void *ap)
{
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;
if (BUF_REFCNT(bp) < 1) {
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)
Update for the KDB framework: o Call kdb_enter() instead of Debugger().
2004-07-10 21:47:53 +00:00
kdb_enter("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
vop_lookup_pre(void *ap)
{
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");
}
void
vop_lookup_post(void *ap, int rc)
{
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 componentname *cnp;
struct vnode *dvp;
struct vnode *vp;
int flags;
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;
cnp = a->a_cnp;
vp = *(a->a_vpp);
flags = cnp->cn_flags;
ASSERT_VI_UNLOCKED(dvp, "VOP_LOOKUP");
More style fixes. Obtained from: bde
2004-01-05 23:40:46 +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
/*
* If this is the last path component for this lookup and LOCKPARENT
* is set, OR if there is an error the directory has to be locked.
*/
if ((flags & LOCKPARENT) && (flags & ISLASTCN))
ASSERT_VOP_LOCKED(dvp, "VOP_LOOKUP (LOCKPARENT)");
else if (rc != 0)
ASSERT_VOP_LOCKED(dvp, "VOP_LOOKUP (error)");
else if (dvp != vp)
ASSERT_VOP_UNLOCKED(dvp, "VOP_LOOKUP (dvp)");
if (flags & PDIRUNLOCK)
ASSERT_VOP_UNLOCKED(dvp, "VOP_LOOKUP (PDIRUNLOCK)");
}
void
vop_lock_pre(void *ap)
{
struct vop_lock_args *a = ap;
if ((a->a_flags & LK_INTERLOCK) == 0)
ASSERT_VI_UNLOCKED(a->a_vp, "VOP_LOCK");
else
ASSERT_VI_LOCKED(a->a_vp, "VOP_LOCK");
}
void
vop_lock_post(void *ap, int rc)
{
More style fixes. Obtained from: bde
2004-01-05 23:40:46 +00:00
struct vop_lock_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");
}
void
vop_unlock_pre(void *ap)
{
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");
}
void
vop_unlock_post(void *ap, int rc)
{
struct vop_unlock_args *a = ap;
if (a->a_flags & LK_INTERLOCK)
ASSERT_VI_UNLOCKED(a->a_vp, "VOP_UNLOCK");
}
#endif /* DEBUG_VFS_LOCKS */
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)
{
knlist_init(&fs_knlist, NULL);
}
/* 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) {
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);
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 (VFS_SYSCTL(mp, vc.vc_op, req));
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
}
fixup sysctl by fsid node
2004-07-08 06:11:36 +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));
}
Reference in New Issue Copy Permalink