freebsd-dev/sys/ufs/ffs/ffs_vfsops.c

1412 lines
38 KiB
C
Raw Normal View History

1994-05-24 10:09:53 +00:00
/*
* Copyright (c) 1989, 1991, 1993, 1994
* The Regents of the University of California. All rights reserved.
*
* 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.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by the University of
* California, Berkeley and its contributors.
* 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.
*
* @(#)ffs_vfsops.c 8.31 (Berkeley) 5/20/95
1999-08-28 01:08:13 +00:00
* $FreeBSD$
1994-05-24 10:09:53 +00:00
*/
1996-01-05 18:31:58 +00:00
#include "opt_quota.h"
#include "opt_ufs.h"
1996-01-05 18:31:58 +00:00
1994-05-24 10:09:53 +00:00
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/namei.h>
#include <sys/proc.h>
#include <sys/kernel.h>
#include <sys/vnode.h>
#include <sys/mount.h>
#include <sys/bio.h>
1994-05-24 10:09:53 +00:00
#include <sys/buf.h>
#include <sys/conf.h>
#include <sys/fcntl.h>
1994-05-24 10:09:53 +00:00
#include <sys/disklabel.h>
#include <sys/malloc.h>
#include <sys/mutex.h>
#include <ufs/ufs/extattr.h>
1994-05-24 10:09:53 +00:00
#include <ufs/ufs/quota.h>
#include <ufs/ufs/ufsmount.h>
#include <ufs/ufs/inode.h>
#include <ufs/ufs/ufs_extern.h>
#include <ufs/ffs/fs.h>
#include <ufs/ffs/ffs_extern.h>
#include <vm/vm.h>
#include <vm/vm_page.h>
static MALLOC_DEFINE(M_FFSNODE, "FFS node", "FFS vnode private part");
1995-12-17 21:14:36 +00:00
static int ffs_sbupdate __P((struct ufsmount *, int));
int ffs_reload __P((struct mount *,struct ucred *,struct thread *));
1995-12-17 21:14:36 +00:00
static int ffs_oldfscompat __P((struct fs *));
static int ffs_init __P((struct vfsconf *));
1994-05-24 10:09:53 +00:00
1998-02-09 06:11:36 +00:00
static struct vfsops ufs_vfsops = {
1994-05-24 10:09:53 +00:00
ffs_mount,
ufs_start,
ffs_unmount,
ufs_root,
ufs_quotactl,
ffs_statfs,
ffs_sync,
ffs_vget,
ffs_fhtovp,
vfs_stdcheckexp,
1994-05-24 10:09:53 +00:00
ffs_vptofh,
ffs_init,
vfs_stduninit,
#ifdef UFS_EXTATTR
ufs_extattrctl,
#else
vfs_stdextattrctl,
#endif
1994-05-24 10:09:53 +00:00
};
VFS_SET(ufs_vfsops, ufs, 0);
1994-05-24 10:09:53 +00:00
/*
* ffs_mount
*
* Called when mounting local physical media
*
* PARAMETERS:
* mountroot
* mp mount point structure
* path NULL (flag for root mount!!!)
* data <unused>
* ndp <unused>
* p process (user credentials check [statfs])
*
* mount
* mp mount point structure
* path path to mount point
* data pointer to argument struct in user space
* ndp mount point namei() return (used for
* credentials on reload), reused to look
* up block device.
* p process (user credentials check)
*
* RETURNS: 0 Success
* !0 error number (errno.h)
*
* LOCK STATE:
1994-05-24 10:09:53 +00:00
*
* ENTRY
* mount point is locked
* EXIT
* mount point is locked
*
* NOTES:
* A NULL path can be used for a flag since the mount
* system call will fail with EFAULT in copyinstr in
* namei() if it is a genuine NULL from the user.
1994-05-24 10:09:53 +00:00
*/
Initial commit of IFS - a inode-namespaced FFS. Here is a short description: How it works: -- Basically ifs is a copy of ffs, overriding some vfs/vnops. (Yes, hack.) I didn't see the need in duplicating all of sys/ufs/ffs to get this off the ground. File creation is done through a special file - 'newfile' . When newfile is called, the system allocates and returns an inode. Note that newfile is done in a cloning fashion: fd = open("newfile", O_CREAT|O_RDWR, 0644); fstat(fd, &st); printf("new file is %d\n", (int)st.st_ino); Once you have created a file, you can open() and unlink() it by its returned inode number retrieved from the stat call, ie: fd = open("5", O_RDWR); The creation permissions depend entirely if you have write access to the root directory of the filesystem. To get the list of currently allocated inodes, VOP_READDIR has been added which returns a directory listing of those currently allocated. -- What this entails: * patching conf/files and conf/options to include IFS as a new compile option (and since ifs depends upon FFS, include the FFS routines) * An entry in i386/conf/NOTES indicating IFS exists and where to go for an explanation * Unstaticize a couple of routines in src/sys/ufs/ffs/ which the IFS routines require (ffs_mount() and ffs_reload()) * a new bunch of routines in src/sys/ufs/ifs/ which implement the IFS routines. IFS replaces some of the vfsops, and a handful of vnops - most notably are VFS_VGET(), VOP_LOOKUP(), VOP_UNLINK() and VOP_READDIR(). Any other directory operation is marked as invalid. What this results in: * an IFS partition's create permissions are controlled by the perm/ownership of the root mount point, just like a normal directory * Each inode has perm and ownership too * IFS does *NOT* mean an FFS partition can be opened per inode. This is a completely seperate filesystem here * Softupdates doesn't work with IFS, and really I don't think it needs it. Besides, fsck's are FAST. (Try it :-) * Inodes 0 and 1 aren't allocatable because they are special (dump/swap IIRC). Inode 2 isn't allocatable since UFS/FFS locks all inodes in the system against this particular inode, and unravelling THAT code isn't trivial. Therefore, useful inodes start at 3. Enjoy, and feedback is definitely appreciated!
2000-10-14 03:02:30 +00:00
int
ffs_mount(mp, path, data, ndp, td)
struct mount *mp; /* mount struct pointer*/
char *path; /* path to mount point*/
caddr_t data; /* arguments to FS specific mount*/
struct nameidata *ndp; /* mount point credentials*/
struct thread *td; /* process requesting mount*/
1994-05-24 10:09:53 +00:00
{
1998-06-04 17:21:39 +00:00
size_t size;
struct vnode *devvp;
struct ufs_args args;
struct ufsmount *ump = 0;
1994-05-24 10:09:53 +00:00
register struct fs *fs;
int error, flags;
mode_t accessmode;
1995-05-30 08:16:23 +00:00
1994-05-24 10:09:53 +00:00
/*
* Use NULL path to indicate we are mounting the root file system.
1994-05-24 10:09:53 +00:00
*/
if (path == NULL) {
if ((error = bdevvp(rootdev, &rootvp))) {
printf("ffs_mountroot: can't find rootvp\n");
return (error);
}
if ((error = ffs_mountfs(rootvp, mp, td, M_FFSNODE)) != 0)
return (error);
(void)VFS_STATFS(mp, &mp->mnt_stat, td);
return (0);
1994-05-24 10:09:53 +00:00
}
/*
* Mounting non-root file system or updating a file system
*/
if ((error = copyin(data, (caddr_t)&args, sizeof(struct ufs_args)))!= 0)
return (error);
1994-05-24 10:09:53 +00:00
/*
* If updating, check whether changing from read-only to
* read/write; if there is no device name, that's all we do.
*/
if (mp->mnt_flag & MNT_UPDATE) {
ump = VFSTOUFS(mp);
fs = ump->um_fs;
devvp = ump->um_devvp;
if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
if ((error = vn_start_write(NULL, &mp, V_WAIT)) != 0)
return (error);
/*
* Flush any dirty data.
*/
VFS_SYNC(mp, MNT_WAIT, td->td_proc->p_ucred, td);
/*
* Check for and optionally get rid of files open
* for writing.
*/
1994-05-24 10:09:53 +00:00
flags = WRITECLOSE;
if (mp->mnt_flag & MNT_FORCE)
flags |= FORCECLOSE;
if (mp->mnt_flag & MNT_SOFTDEP) {
error = softdep_flushfiles(mp, flags, td);
} else {
error = ffs_flushfiles(mp, flags, td);
}
if (error) {
vn_finished_write(mp);
return (error);
}
if (fs->fs_pendingblocks != 0 ||
fs->fs_pendinginodes != 0) {
printf("%s: update error: blocks %d files %d\n",
fs->fs_fsmnt, fs->fs_pendingblocks,
fs->fs_pendinginodes);
fs->fs_pendingblocks = 0;
fs->fs_pendinginodes = 0;
}
fs->fs_ronly = 1;
if ((fs->fs_flags & (FS_UNCLEAN | FS_NEEDSFSCK)) == 0)
fs->fs_clean = 1;
if ((error = ffs_sbupdate(ump, MNT_WAIT)) != 0) {
fs->fs_ronly = 0;
fs->fs_clean = 0;
vn_finished_write(mp);
return (error);
}
vn_finished_write(mp);
}
if ((mp->mnt_flag & MNT_RELOAD) &&
(error = ffs_reload(mp, ndp->ni_cnd.cn_cred, td)) != 0)
return (error);
if (fs->fs_ronly && (mp->mnt_kern_flag & MNTK_WANTRDWR)) {
/*
* If upgrade to read-write by non-root, then verify
* that user has necessary permissions on the device.
*/
if (suser_td(td)) {
vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
if ((error = VOP_ACCESS(devvp, VREAD | VWRITE,
td->td_ucred, td)) != 0) {
VOP_UNLOCK(devvp, 0, td);
return (error);
}
VOP_UNLOCK(devvp, 0, td);
}
fs->fs_flags &= ~FS_UNCLEAN;
if (fs->fs_clean == 0) {
fs->fs_flags |= FS_UNCLEAN;
if ((mp->mnt_flag & MNT_FORCE) ||
((fs->fs_flags & FS_NEEDSFSCK) == 0 &&
(fs->fs_flags & FS_DOSOFTDEP))) {
printf("WARNING: %s was not %s\n",
fs->fs_fsmnt, "properly dismounted");
} else {
printf(
"WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
fs->fs_fsmnt);
return (EPERM);
}
}
if ((error = vn_start_write(NULL, &mp, V_WAIT)) != 0)
return (error);
fs->fs_ronly = 0;
fs->fs_clean = 0;
if ((error = ffs_sbupdate(ump, MNT_WAIT)) != 0) {
vn_finished_write(mp);
return (error);
}
/* check to see if we need to start softdep */
if ((fs->fs_flags & FS_DOSOFTDEP) &&
(error = softdep_mount(devvp, mp, fs, td->td_ucred))){
vn_finished_write(mp);
return (error);
}
if (fs->fs_snapinum[0] != 0)
ffs_snapshot_mount(mp);
vn_finished_write(mp);
}
/*
* Soft updates is incompatible with "async",
* so if we are doing softupdates stop the user
* from setting the async flag in an update.
* Softdep_mount() clears it in an initial mount
* or ro->rw remount.
*/
if (mp->mnt_flag & MNT_SOFTDEP)
mp->mnt_flag &= ~MNT_ASYNC;
/*
* If not updating name, process export requests.
*/
if (args.fspec == 0)
return (vfs_export(mp, &args.export));
/*
* If this is a snapshot request, take the snapshot.
*/
if (mp->mnt_flag & MNT_SNAPSHOT)
return (ffs_snapshot(mp, args.fspec));
1994-05-24 10:09:53 +00:00
}
1994-05-24 10:09:53 +00:00
/*
* Not an update, or updating the name: look up the name
* and verify that it refers to a sensible block device.
*/
NDINIT(ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, td);
if ((error = namei(ndp)) != 0)
return (error);
NDFREE(ndp, NDF_ONLY_PNBUF);
1994-05-24 10:09:53 +00:00
devvp = ndp->ni_vp;
if (!vn_isdisk(devvp, &error)) {
vrele(devvp);
return (error);
}
/*
* If mount by non-root, then verify that user has necessary
* permissions on the device.
*/
if (suser_td(td)) {
accessmode = VREAD;
if ((mp->mnt_flag & MNT_RDONLY) == 0)
accessmode |= VWRITE;
vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
if ((error = VOP_ACCESS(devvp, accessmode, td->td_ucred, td))!= 0){
vput(devvp);
return (error);
}
VOP_UNLOCK(devvp, 0, td);
}
if (mp->mnt_flag & MNT_UPDATE) {
/*
* Update only
*
* If it's not the same vnode, or at least the same device
* then it's not correct.
*/
if (devvp != ump->um_devvp &&
devvp->v_rdev != ump->um_devvp->v_rdev)
error = EINVAL; /* needs translation */
vrele(devvp);
if (error)
return (error);
} else {
/*
* New mount
*
* We need the name for the mount point (also used for
* "last mounted on") copied in. If an error occurs,
* the mount point is discarded by the upper level code.
* Note that vfs_mount() populates f_mntonname for us.
*/
if ((error = ffs_mountfs(devvp, mp, td, M_FFSNODE)) != 0) {
vrele(devvp);
return (error);
}
1994-05-24 10:09:53 +00:00
}
/*
* Save "mounted from" device name info for mount point (NULL pad).
*/
copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, &size);
bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
/*
* Initialize filesystem stat information in mount struct.
*/
(void)VFS_STATFS(mp, &mp->mnt_stat, td);
return (0);
1994-05-24 10:09:53 +00:00
}
/*
* Reload all incore data for a filesystem (used after running fsck on
* the root filesystem and finding things to fix). The filesystem must
* be mounted read-only.
*
* Things to do to update the mount:
* 1) invalidate all cached meta-data.
* 2) re-read superblock from disk.
* 3) re-read summary information from disk.
* 4) invalidate all inactive vnodes.
* 5) invalidate all cached file data.
* 6) re-read inode data for all active vnodes.
*/
Initial commit of IFS - a inode-namespaced FFS. Here is a short description: How it works: -- Basically ifs is a copy of ffs, overriding some vfs/vnops. (Yes, hack.) I didn't see the need in duplicating all of sys/ufs/ffs to get this off the ground. File creation is done through a special file - 'newfile' . When newfile is called, the system allocates and returns an inode. Note that newfile is done in a cloning fashion: fd = open("newfile", O_CREAT|O_RDWR, 0644); fstat(fd, &st); printf("new file is %d\n", (int)st.st_ino); Once you have created a file, you can open() and unlink() it by its returned inode number retrieved from the stat call, ie: fd = open("5", O_RDWR); The creation permissions depend entirely if you have write access to the root directory of the filesystem. To get the list of currently allocated inodes, VOP_READDIR has been added which returns a directory listing of those currently allocated. -- What this entails: * patching conf/files and conf/options to include IFS as a new compile option (and since ifs depends upon FFS, include the FFS routines) * An entry in i386/conf/NOTES indicating IFS exists and where to go for an explanation * Unstaticize a couple of routines in src/sys/ufs/ffs/ which the IFS routines require (ffs_mount() and ffs_reload()) * a new bunch of routines in src/sys/ufs/ifs/ which implement the IFS routines. IFS replaces some of the vfsops, and a handful of vnops - most notably are VFS_VGET(), VOP_LOOKUP(), VOP_UNLINK() and VOP_READDIR(). Any other directory operation is marked as invalid. What this results in: * an IFS partition's create permissions are controlled by the perm/ownership of the root mount point, just like a normal directory * Each inode has perm and ownership too * IFS does *NOT* mean an FFS partition can be opened per inode. This is a completely seperate filesystem here * Softupdates doesn't work with IFS, and really I don't think it needs it. Besides, fsck's are FAST. (Try it :-) * Inodes 0 and 1 aren't allocatable because they are special (dump/swap IIRC). Inode 2 isn't allocatable since UFS/FFS locks all inodes in the system against this particular inode, and unravelling THAT code isn't trivial. Therefore, useful inodes start at 3. Enjoy, and feedback is definitely appreciated!
2000-10-14 03:02:30 +00:00
int
ffs_reload(mp, cred, td)
register struct mount *mp;
1994-05-24 10:09:53 +00:00
struct ucred *cred;
struct thread *td;
1994-05-24 10:09:53 +00:00
{
register struct vnode *vp, *nvp, *devvp;
struct inode *ip;
void *space;
1994-05-24 10:09:53 +00:00
struct buf *bp;
struct fs *fs, *newfs;
struct partinfo dpart;
dev_t dev;
1994-05-24 10:09:53 +00:00
int i, blks, size, error;
int32_t *lp;
1994-05-24 10:09:53 +00:00
if ((mp->mnt_flag & MNT_RDONLY) == 0)
1994-05-24 10:09:53 +00:00
return (EINVAL);
/*
* Step 1: invalidate all cached meta-data.
*/
devvp = VFSTOUFS(mp)->um_devvp;
vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
error = vinvalbuf(devvp, 0, cred, td, 0, 0);
VOP_UNLOCK(devvp, 0, td);
if (error)
1994-05-24 10:09:53 +00:00
panic("ffs_reload: dirty1");
dev = devvp->v_rdev;
/*
* Only VMIO the backing device if the backing device is a real
2001-05-29 21:21:53 +00:00
* block device.
*/
2001-05-29 21:21:53 +00:00
if (vn_isdisk(devvp, NULL)) {
vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
vfs_object_create(devvp, td, td->td_ucred);
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(&devvp->v_interlock);
VOP_UNLOCK(devvp, LK_INTERLOCK, td);
}
1994-05-24 10:09:53 +00:00
/*
* Step 2: re-read superblock from disk.
*/
if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, td) != 0)
size = DEV_BSIZE;
else
size = dpart.disklab->d_secsize;
if ((error = bread(devvp, (ufs_daddr_t)(SBOFF/size), SBSIZE, NOCRED,&bp)) != 0)
1994-05-24 10:09:53 +00:00
return (error);
newfs = (struct fs *)bp->b_data;
if (newfs->fs_magic != FS_MAGIC || newfs->fs_bsize > MAXBSIZE ||
newfs->fs_bsize < sizeof(struct fs)) {
brelse(bp);
return (EIO); /* XXX needs translation */
1994-05-24 10:09:53 +00:00
}
fs = VFSTOUFS(mp)->um_fs;
/*
* Copy pointer fields back into superblock before copying in XXX
* new superblock. These should really be in the ufsmount. XXX
* Note that important parameters (eg fs_ncg) are unchanged.
*/
newfs->fs_csp = fs->fs_csp;
newfs->fs_maxcluster = fs->fs_maxcluster;
newfs->fs_contigdirs = fs->fs_contigdirs;
newfs->fs_active = fs->fs_active;
bcopy(newfs, fs, (u_int)fs->fs_sbsize);
1994-05-24 10:09:53 +00:00
if (fs->fs_sbsize < SBSIZE)
bp->b_flags |= B_INVAL | B_NOCACHE;
1994-05-24 10:09:53 +00:00
brelse(bp);
mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
1994-05-24 10:09:53 +00:00
ffs_oldfscompat(fs);
/* An old fsck may have zeroed these fields, so recheck them. */
if (fs->fs_avgfilesize <= 0) /* XXX */
fs->fs_avgfilesize = AVFILESIZ; /* XXX */
if (fs->fs_avgfpdir <= 0) /* XXX */
fs->fs_avgfpdir = AFPDIR; /* XXX */
if (fs->fs_pendingblocks != 0 || fs->fs_pendinginodes != 0) {
printf("%s: reload pending error: blocks %d files %d\n",
fs->fs_fsmnt, fs->fs_pendingblocks, fs->fs_pendinginodes);
fs->fs_pendingblocks = 0;
fs->fs_pendinginodes = 0;
}
1994-05-24 10:09:53 +00:00
/*
* Step 3: re-read summary information from disk.
*/
blks = howmany(fs->fs_cssize, fs->fs_fsize);
space = fs->fs_csp;
1994-05-24 10:09:53 +00:00
for (i = 0; i < blks; i += fs->fs_frag) {
size = fs->fs_bsize;
if (i + fs->fs_frag > blks)
size = (blks - i) * fs->fs_fsize;
error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
NOCRED, &bp);
if (error)
1994-05-24 10:09:53 +00:00
return (error);
bcopy(bp->b_data, space, (u_int)size);
space = (char *)space + size;
1994-05-24 10:09:53 +00:00
brelse(bp);
}
/*
* We no longer know anything about clusters per cylinder group.
*/
if (fs->fs_contigsumsize > 0) {
lp = fs->fs_maxcluster;
for (i = 0; i < fs->fs_ncg; i++)
*lp++ = fs->fs_contigsumsize;
}
1994-05-24 10:09:53 +00:00
loop:
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(&mntvnode_mtx);
for (vp = TAILQ_FIRST(&mp->mnt_nvnodelist); vp != NULL; vp = nvp) {
if (vp->v_mount != mp) {
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_unlock(&mntvnode_mtx);
goto loop;
}
nvp = TAILQ_NEXT(vp, v_nmntvnodes);
mtx_unlock(&mntvnode_mtx);
1994-05-24 10:09:53 +00:00
/*
* Step 4: invalidate all inactive vnodes.
*/
if (vrecycle(vp, NULL, td))
goto loop;
1994-05-24 10:09:53 +00:00
/*
* Step 5: invalidate all cached file data.
*/
mtx_lock(&vp->v_interlock);
if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, td)) {
1994-05-24 10:09:53 +00:00
goto loop;
}
if (vinvalbuf(vp, 0, cred, td, 0, 0))
1994-05-24 10:09:53 +00:00
panic("ffs_reload: dirty2");
/*
* Step 6: re-read inode data for all active vnodes.
*/
ip = VTOI(vp);
error =
1994-05-24 10:09:53 +00:00
bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
(int)fs->fs_bsize, NOCRED, &bp);
if (error) {
1994-05-24 10:09:53 +00:00
vput(vp);
return (error);
}
ip->i_din = *((struct dinode *)bp->b_data +
ino_to_fsbo(fs, ip->i_number));
ip->i_effnlink = ip->i_nlink;
1994-05-24 10:09:53 +00:00
brelse(bp);
vput(vp);
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(&mntvnode_mtx);
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(&mntvnode_mtx);
1994-05-24 10:09:53 +00:00
return (0);
}
#include <sys/sysctl.h>
int bigcgs = 0;
SYSCTL_INT(_debug, OID_AUTO, bigcgs, CTLFLAG_RW, &bigcgs, 0, "");
1994-05-24 10:09:53 +00:00
/*
* Common code for mount and mountroot
*/
int
ffs_mountfs(devvp, mp, td, malloctype)
1994-05-24 10:09:53 +00:00
register struct vnode *devvp;
struct mount *mp;
struct thread *td;
struct malloc_type *malloctype;
1994-05-24 10:09:53 +00:00
{
register struct ufsmount *ump;
struct buf *bp;
register struct fs *fs;
dev_t dev;
1994-05-24 10:09:53 +00:00
struct partinfo dpart;
void *space;
int error, i, blks, size, ronly;
int32_t *lp;
struct ucred *cred;
u_int64_t maxfilesize; /* XXX */
1998-06-04 17:21:39 +00:00
size_t strsize;
int ncount;
1994-05-24 10:09:53 +00:00
dev = devvp->v_rdev;
cred = td ? td->td_ucred : NOCRED;
1994-05-24 10:09:53 +00:00
/*
* Disallow multiple mounts of the same device.
* Disallow mounting of a device that is currently in use
* (except for root, which might share swap device for miniroot).
* Flush out any old buffers remaining from a previous use.
*/
error = vfs_mountedon(devvp);
if (error)
1994-05-24 10:09:53 +00:00
return (error);
ncount = vcount(devvp);
This mega-commit is meant to fix numerous interrelated problems. There has been some bitrot and incorrect assumptions in the vfs_bio code. These problems have manifest themselves worse on NFS type filesystems, but can still affect local filesystems under certain circumstances. Most of the problems have involved mmap consistancy, and as a side-effect broke the vfs.ioopt code. This code might have been committed seperately, but almost everything is interrelated. 1) Allow (pmap_object_init_pt) prefaulting of buffer-busy pages that are fully valid. 2) Rather than deactivating erroneously read initial (header) pages in kern_exec, we now free them. 3) Fix the rundown of non-VMIO buffers that are in an inconsistent (missing vp) state. 4) Fix the disassociation of pages from buffers in brelse. The previous code had rotted and was faulty in a couple of important circumstances. 5) Remove a gratuitious buffer wakeup in vfs_vmio_release. 6) Remove a crufty and currently unused cluster mechanism for VBLK files in vfs_bio_awrite. When the code is functional, I'll add back a cleaner version. 7) The page busy count wakeups assocated with the buffer cache usage were incorrectly cleaned up in a previous commit by me. Revert to the original, correct version, but with a cleaner implementation. 8) The cluster read code now tries to keep data associated with buffers more aggressively (without breaking the heuristics) when it is presumed that the read data (buffers) will be soon needed. 9) Change to filesystem lockmgr locks so that they use LK_NOPAUSE. The delay loop waiting is not useful for filesystem locks, due to the length of the time intervals. 10) Correct and clean-up spec_getpages. 11) Implement a fully functional nfs_getpages, nfs_putpages. 12) Fix nfs_write so that modifications are coherent with the NFS data on the server disk (at least as well as NFS seems to allow.) 13) Properly support MS_INVALIDATE on NFS. 14) Properly pass down MS_INVALIDATE to lower levels of the VM code from vm_map_clean. 15) Better support the notion of pages being busy but valid, so that fewer in-transit waits occur. (use p->busy more for pageouts instead of PG_BUSY.) Since the page is fully valid, it is still usable for reads. 16) It is possible (in error) for cached pages to be busy. Make the page allocation code handle that case correctly. (It should probably be a printf or panic, but I want the system to handle coding errors robustly. I'll probably add a printf.) 17) Correct the design and usage of vm_page_sleep. It didn't handle consistancy problems very well, so make the design a little less lofty. After vm_page_sleep, if it ever blocked, it is still important to relookup the page (if the object generation count changed), and verify it's status (always.) 18) In vm_pageout.c, vm_pageout_clean had rotted, so clean that up. 19) Push the page busy for writes and VM_PROT_READ into vm_pageout_flush. 20) Fix vm_pager_put_pages and it's descendents to support an int flag instead of a boolean, so that we can pass down the invalidate bit.
1998-03-07 21:37:31 +00:00
if (ncount > 1 && devvp != rootvp)
1994-05-24 10:09:53 +00:00
return (EBUSY);
vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
error = vinvalbuf(devvp, V_SAVE, cred, td, 0, 0);
VOP_UNLOCK(devvp, 0, td);
if (error)
1994-05-24 10:09:53 +00:00
return (error);
/*
* Only VMIO the backing device if the backing device is a real
2001-05-29 21:21:53 +00:00
* block device.
* Note that it is optional that the backing device be VMIOed. This
* increases the opportunity for metadata caching.
*/
2001-05-29 21:21:53 +00:00
if (vn_isdisk(devvp, NULL)) {
vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
vfs_object_create(devvp, td, cred);
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(&devvp->v_interlock);
VOP_UNLOCK(devvp, LK_INTERLOCK, td);
}
1994-05-24 10:09:53 +00:00
ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, td);
VOP_UNLOCK(devvp, 0, td);
if (error)
1994-05-24 10:09:53 +00:00
return (error);
if (devvp->v_rdev->si_iosize_max > mp->mnt_iosize_max)
mp->mnt_iosize_max = devvp->v_rdev->si_iosize_max;
if (mp->mnt_iosize_max > MAXPHYS)
mp->mnt_iosize_max = MAXPHYS;
if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, cred, td) != 0)
1994-05-24 10:09:53 +00:00
size = DEV_BSIZE;
else
1994-05-24 10:09:53 +00:00
size = dpart.disklab->d_secsize;
bp = NULL;
ump = NULL;
if ((error = bread(devvp, SBLOCK, SBSIZE, cred, &bp)) != 0)
1994-05-24 10:09:53 +00:00
goto out;
fs = (struct fs *)bp->b_data;
if (fs->fs_magic != FS_MAGIC || fs->fs_bsize > MAXBSIZE ||
fs->fs_bsize < sizeof(struct fs)) {
error = EINVAL; /* XXX needs translation */
goto out;
}
fs->fs_fmod = 0;
fs->fs_flags &= ~FS_UNCLEAN;
if (fs->fs_clean == 0) {
fs->fs_flags |= FS_UNCLEAN;
if (ronly || (mp->mnt_flag & MNT_FORCE) ||
((fs->fs_flags & FS_NEEDSFSCK) == 0 &&
(fs->fs_flags & FS_DOSOFTDEP))) {
printf(
"WARNING: %s was not properly dismounted\n",
fs->fs_fsmnt);
} else {
printf(
"WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
fs->fs_fsmnt);
error = EPERM;
goto out;
}
if (fs->fs_pendingblocks != 0 || fs->fs_pendinginodes != 0) {
printf("%s: lost blocks %d files %d\n", fs->fs_fsmnt,
fs->fs_pendingblocks, fs->fs_pendinginodes);
fs->fs_pendingblocks = 0;
fs->fs_pendinginodes = 0;
}
}
if (fs->fs_pendingblocks != 0 || fs->fs_pendinginodes != 0) {
printf("%s: mount pending error: blocks %d files %d\n",
fs->fs_fsmnt, fs->fs_pendingblocks, fs->fs_pendinginodes);
fs->fs_pendingblocks = 0;
fs->fs_pendinginodes = 0;
}
/* XXX updating 4.2 FFS superblocks trashes rotational layout tables */
if (fs->fs_postblformat == FS_42POSTBLFMT && !ronly) {
error = EROFS; /* needs translation */
goto out;
}
ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK | M_ZERO);
ump->um_malloctype = malloctype;
ump->um_i_effnlink_valid = 1;
1994-05-24 10:09:53 +00:00
ump->um_fs = malloc((u_long)fs->fs_sbsize, M_UFSMNT,
M_WAITOK);
ump->um_blkatoff = ffs_blkatoff;
ump->um_truncate = ffs_truncate;
ump->um_update = ffs_update;
ump->um_valloc = ffs_valloc;
ump->um_vfree = ffs_vfree;
ump->um_balloc = ffs_balloc;
1994-05-24 10:09:53 +00:00
bcopy(bp->b_data, ump->um_fs, (u_int)fs->fs_sbsize);
if (fs->fs_sbsize < SBSIZE)
bp->b_flags |= B_INVAL | B_NOCACHE;
1994-05-24 10:09:53 +00:00
brelse(bp);
bp = NULL;
fs = ump->um_fs;
fs->fs_ronly = ronly;
size = fs->fs_cssize;
blks = howmany(size, fs->fs_fsize);
if (fs->fs_contigsumsize > 0)
size += fs->fs_ncg * sizeof(int32_t);
Directory layout preference improvements from Grigoriy Orlov <gluk@ptci.ru>. His description of the problem and solution follow. My own tests show speedups on typical filesystem intensive workloads of 5% to 12% which is very impressive considering the small amount of code change involved. ------ One day I noticed that some file operations run much faster on small file systems then on big ones. I've looked at the ffs algorithms, thought about them, and redesigned the dirpref algorithm. First I want to describe the results of my tests. These results are old and I have improved the algorithm after these tests were done. Nevertheless they show how big the perfomance speedup may be. I have done two file/directory intensive tests on a two OpenBSD systems with old and new dirpref algorithm. The first test is "tar -xzf ports.tar.gz", the second is "rm -rf ports". The ports.tar.gz file is the ports collection from the OpenBSD 2.8 release. It contains 6596 directories and 13868 files. The test systems are: 1. Celeron-450, 128Mb, two IDE drives, the system at wd0, file system for test is at wd1. Size of test file system is 8 Gb, number of cg=991, size of cg is 8m, block size = 8k, fragment size = 1k OpenBSD-current from Dec 2000 with BUFCACHEPERCENT=35 2. PIII-600, 128Mb, two IBM DTLA-307045 IDE drives at i815e, the system at wd0, file system for test is at wd1. Size of test file system is 40 Gb, number of cg=5324, size of cg is 8m, block size = 8k, fragment size = 1k OpenBSD-current from Dec 2000 with BUFCACHEPERCENT=50 You can get more info about the test systems and methods at: http://www.ptci.ru/gluk/dirpref/old/dirpref.html Test Results tar -xzf ports.tar.gz rm -rf ports mode old dirpref new dirpref speedup old dirprefnew dirpref speedup First system normal 667 472 1.41 477 331 1.44 async 285 144 1.98 130 14 9.29 sync 768 616 1.25 477 334 1.43 softdep 413 252 1.64 241 38 6.34 Second system normal 329 81 4.06 263.5 93.5 2.81 async 302 25.7 11.75 112 2.26 49.56 sync 281 57.0 4.93 263 90.5 2.9 softdep 341 40.6 8.4 284 4.76 59.66 "old dirpref" and "new dirpref" columns give a test time in seconds. speedup - speed increasement in times, ie. old dirpref / new dirpref. ------ Algorithm description The old dirpref algorithm is described in comments: /* * Find a cylinder to place a directory. * * The policy implemented by this algorithm is to select from * among those cylinder groups with above the average number of * free inodes, the one with the smallest number of directories. */ A new directory is allocated in a different cylinder groups than its parent directory resulting in a directory tree that is spreaded across all the cylinder groups. This spreading out results in a non-optimal access to the directories and files. When we have a small filesystem it is not a problem but when the filesystem is big then perfomance degradation becomes very apparent. What I mean by a big file system ? 1. A big filesystem is a filesystem which occupy 20-30 or more percent of total drive space, i.e. first and last cylinder are physically located relatively far from each other. 2. It has a relatively large number of cylinder groups, for example more cylinder groups than 50% of the buffers in the buffer cache. The first results in long access times, while the second results in many buffers being used by metadata operations. Such operations use cylinder group blocks and on-disk inode blocks. The cylinder group block (fs->fs_cblkno) contains struct cg, inode and block bit maps. It is 2k in size for the default filesystem parameters. If new and parent directories are located in different cylinder groups then the system performs more input/output operations and uses more buffers. On filesystems with many cylinder groups, lots of cache buffers are used for metadata operations. My solution for this problem is very simple. I allocate many directories in one cylinder group. I also do some things, so that the new allocation method does not cause excessive fragmentation and all directory inodes will not be located at a location far from its file's inodes and data. The algorithm is: /* * Find a cylinder group to place a directory. * * The policy implemented by this algorithm is to allocate a * directory inode in the same cylinder group as its parent * directory, but also to reserve space for its files inodes * and data. Restrict the number of directories which may be * allocated one after another in the same cylinder group * without intervening allocation of files. * * If we allocate a first level directory then force allocation * in another cylinder group. */ My early versions of dirpref give me a good results for a wide range of file operations and different filesystem capacities except one case: those applications that create their entire directory structure first and only later fill this structure with files. My solution for such and similar cases is to limit a number of directories which may be created one after another in the same cylinder group without intervening file creations. For this purpose, I allocate an array of counters at mount time. This array is linked to the superblock fs->fs_contigdirs[cg]. Each time a directory is created the counter increases and each time a file is created the counter decreases. A 60Gb filesystem with 8mb/cg requires 10kb of memory for the counters array. The maxcontigdirs is a maximum number of directories which may be created without an intervening file creation. I found in my tests that the best performance occurs when I restrict the number of directories in one cylinder group such that all its files may be located in the same cylinder group. There may be some deterioration in performance if all the file inodes are in the same cylinder group as its containing directory, but their data partially resides in a different cylinder group. The maxcontigdirs value is calculated to try to prevent this condition. Since there is no way to know how many files and directories will be allocated later I added two optimization parameters in superblock/tunefs. They are: int32_t fs_avgfilesize; /* expected average file size */ int32_t fs_avgfpdir; /* expected # of files per directory */ These parameters have reasonable defaults but may be tweeked for special uses of a filesystem. They are only necessary in rare cases like better tuning a filesystem being used to store a squid cache. I have been using this algorithm for about 3 months. I have done a lot of testing on filesystems with different capacities, average filesize, average number of files per directory, and so on. I think this algorithm has no negative impact on filesystem perfomance. It works better than the default one in all cases. The new dirpref will greatly improve untarring/removing/coping of big directories, decrease load on cvs servers and much more. The new dirpref doesn't speedup a compilation process, but also doesn't slow it down. Obtained from: Grigoriy Orlov <gluk@ptci.ru>
2001-04-10 08:38:59 +00:00
size += fs->fs_ncg * sizeof(u_int8_t);
space = malloc((u_long)size, M_UFSMNT, M_WAITOK);
fs->fs_csp = space;
1994-05-24 10:09:53 +00:00
for (i = 0; i < blks; i += fs->fs_frag) {
size = fs->fs_bsize;
if (i + fs->fs_frag > blks)
size = (blks - i) * fs->fs_fsize;
if ((error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
cred, &bp)) != 0) {
free(fs->fs_csp, M_UFSMNT);
1994-05-24 10:09:53 +00:00
goto out;
}
bcopy(bp->b_data, space, (u_int)size);
space = (char *)space + size;
1994-05-24 10:09:53 +00:00
brelse(bp);
bp = NULL;
}
if (fs->fs_contigsumsize > 0) {
fs->fs_maxcluster = lp = space;
for (i = 0; i < fs->fs_ncg; i++)
*lp++ = fs->fs_contigsumsize;
space = lp;
}
Directory layout preference improvements from Grigoriy Orlov <gluk@ptci.ru>. His description of the problem and solution follow. My own tests show speedups on typical filesystem intensive workloads of 5% to 12% which is very impressive considering the small amount of code change involved. ------ One day I noticed that some file operations run much faster on small file systems then on big ones. I've looked at the ffs algorithms, thought about them, and redesigned the dirpref algorithm. First I want to describe the results of my tests. These results are old and I have improved the algorithm after these tests were done. Nevertheless they show how big the perfomance speedup may be. I have done two file/directory intensive tests on a two OpenBSD systems with old and new dirpref algorithm. The first test is "tar -xzf ports.tar.gz", the second is "rm -rf ports". The ports.tar.gz file is the ports collection from the OpenBSD 2.8 release. It contains 6596 directories and 13868 files. The test systems are: 1. Celeron-450, 128Mb, two IDE drives, the system at wd0, file system for test is at wd1. Size of test file system is 8 Gb, number of cg=991, size of cg is 8m, block size = 8k, fragment size = 1k OpenBSD-current from Dec 2000 with BUFCACHEPERCENT=35 2. PIII-600, 128Mb, two IBM DTLA-307045 IDE drives at i815e, the system at wd0, file system for test is at wd1. Size of test file system is 40 Gb, number of cg=5324, size of cg is 8m, block size = 8k, fragment size = 1k OpenBSD-current from Dec 2000 with BUFCACHEPERCENT=50 You can get more info about the test systems and methods at: http://www.ptci.ru/gluk/dirpref/old/dirpref.html Test Results tar -xzf ports.tar.gz rm -rf ports mode old dirpref new dirpref speedup old dirprefnew dirpref speedup First system normal 667 472 1.41 477 331 1.44 async 285 144 1.98 130 14 9.29 sync 768 616 1.25 477 334 1.43 softdep 413 252 1.64 241 38 6.34 Second system normal 329 81 4.06 263.5 93.5 2.81 async 302 25.7 11.75 112 2.26 49.56 sync 281 57.0 4.93 263 90.5 2.9 softdep 341 40.6 8.4 284 4.76 59.66 "old dirpref" and "new dirpref" columns give a test time in seconds. speedup - speed increasement in times, ie. old dirpref / new dirpref. ------ Algorithm description The old dirpref algorithm is described in comments: /* * Find a cylinder to place a directory. * * The policy implemented by this algorithm is to select from * among those cylinder groups with above the average number of * free inodes, the one with the smallest number of directories. */ A new directory is allocated in a different cylinder groups than its parent directory resulting in a directory tree that is spreaded across all the cylinder groups. This spreading out results in a non-optimal access to the directories and files. When we have a small filesystem it is not a problem but when the filesystem is big then perfomance degradation becomes very apparent. What I mean by a big file system ? 1. A big filesystem is a filesystem which occupy 20-30 or more percent of total drive space, i.e. first and last cylinder are physically located relatively far from each other. 2. It has a relatively large number of cylinder groups, for example more cylinder groups than 50% of the buffers in the buffer cache. The first results in long access times, while the second results in many buffers being used by metadata operations. Such operations use cylinder group blocks and on-disk inode blocks. The cylinder group block (fs->fs_cblkno) contains struct cg, inode and block bit maps. It is 2k in size for the default filesystem parameters. If new and parent directories are located in different cylinder groups then the system performs more input/output operations and uses more buffers. On filesystems with many cylinder groups, lots of cache buffers are used for metadata operations. My solution for this problem is very simple. I allocate many directories in one cylinder group. I also do some things, so that the new allocation method does not cause excessive fragmentation and all directory inodes will not be located at a location far from its file's inodes and data. The algorithm is: /* * Find a cylinder group to place a directory. * * The policy implemented by this algorithm is to allocate a * directory inode in the same cylinder group as its parent * directory, but also to reserve space for its files inodes * and data. Restrict the number of directories which may be * allocated one after another in the same cylinder group * without intervening allocation of files. * * If we allocate a first level directory then force allocation * in another cylinder group. */ My early versions of dirpref give me a good results for a wide range of file operations and different filesystem capacities except one case: those applications that create their entire directory structure first and only later fill this structure with files. My solution for such and similar cases is to limit a number of directories which may be created one after another in the same cylinder group without intervening file creations. For this purpose, I allocate an array of counters at mount time. This array is linked to the superblock fs->fs_contigdirs[cg]. Each time a directory is created the counter increases and each time a file is created the counter decreases. A 60Gb filesystem with 8mb/cg requires 10kb of memory for the counters array. The maxcontigdirs is a maximum number of directories which may be created without an intervening file creation. I found in my tests that the best performance occurs when I restrict the number of directories in one cylinder group such that all its files may be located in the same cylinder group. There may be some deterioration in performance if all the file inodes are in the same cylinder group as its containing directory, but their data partially resides in a different cylinder group. The maxcontigdirs value is calculated to try to prevent this condition. Since there is no way to know how many files and directories will be allocated later I added two optimization parameters in superblock/tunefs. They are: int32_t fs_avgfilesize; /* expected average file size */ int32_t fs_avgfpdir; /* expected # of files per directory */ These parameters have reasonable defaults but may be tweeked for special uses of a filesystem. They are only necessary in rare cases like better tuning a filesystem being used to store a squid cache. I have been using this algorithm for about 3 months. I have done a lot of testing on filesystems with different capacities, average filesize, average number of files per directory, and so on. I think this algorithm has no negative impact on filesystem perfomance. It works better than the default one in all cases. The new dirpref will greatly improve untarring/removing/coping of big directories, decrease load on cvs servers and much more. The new dirpref doesn't speedup a compilation process, but also doesn't slow it down. Obtained from: Grigoriy Orlov <gluk@ptci.ru>
2001-04-10 08:38:59 +00:00
size = fs->fs_ncg * sizeof(u_int8_t);
fs->fs_contigdirs = (u_int8_t *)space;
bzero(fs->fs_contigdirs, size);
fs->fs_active = NULL;
Directory layout preference improvements from Grigoriy Orlov <gluk@ptci.ru>. His description of the problem and solution follow. My own tests show speedups on typical filesystem intensive workloads of 5% to 12% which is very impressive considering the small amount of code change involved. ------ One day I noticed that some file operations run much faster on small file systems then on big ones. I've looked at the ffs algorithms, thought about them, and redesigned the dirpref algorithm. First I want to describe the results of my tests. These results are old and I have improved the algorithm after these tests were done. Nevertheless they show how big the perfomance speedup may be. I have done two file/directory intensive tests on a two OpenBSD systems with old and new dirpref algorithm. The first test is "tar -xzf ports.tar.gz", the second is "rm -rf ports". The ports.tar.gz file is the ports collection from the OpenBSD 2.8 release. It contains 6596 directories and 13868 files. The test systems are: 1. Celeron-450, 128Mb, two IDE drives, the system at wd0, file system for test is at wd1. Size of test file system is 8 Gb, number of cg=991, size of cg is 8m, block size = 8k, fragment size = 1k OpenBSD-current from Dec 2000 with BUFCACHEPERCENT=35 2. PIII-600, 128Mb, two IBM DTLA-307045 IDE drives at i815e, the system at wd0, file system for test is at wd1. Size of test file system is 40 Gb, number of cg=5324, size of cg is 8m, block size = 8k, fragment size = 1k OpenBSD-current from Dec 2000 with BUFCACHEPERCENT=50 You can get more info about the test systems and methods at: http://www.ptci.ru/gluk/dirpref/old/dirpref.html Test Results tar -xzf ports.tar.gz rm -rf ports mode old dirpref new dirpref speedup old dirprefnew dirpref speedup First system normal 667 472 1.41 477 331 1.44 async 285 144 1.98 130 14 9.29 sync 768 616 1.25 477 334 1.43 softdep 413 252 1.64 241 38 6.34 Second system normal 329 81 4.06 263.5 93.5 2.81 async 302 25.7 11.75 112 2.26 49.56 sync 281 57.0 4.93 263 90.5 2.9 softdep 341 40.6 8.4 284 4.76 59.66 "old dirpref" and "new dirpref" columns give a test time in seconds. speedup - speed increasement in times, ie. old dirpref / new dirpref. ------ Algorithm description The old dirpref algorithm is described in comments: /* * Find a cylinder to place a directory. * * The policy implemented by this algorithm is to select from * among those cylinder groups with above the average number of * free inodes, the one with the smallest number of directories. */ A new directory is allocated in a different cylinder groups than its parent directory resulting in a directory tree that is spreaded across all the cylinder groups. This spreading out results in a non-optimal access to the directories and files. When we have a small filesystem it is not a problem but when the filesystem is big then perfomance degradation becomes very apparent. What I mean by a big file system ? 1. A big filesystem is a filesystem which occupy 20-30 or more percent of total drive space, i.e. first and last cylinder are physically located relatively far from each other. 2. It has a relatively large number of cylinder groups, for example more cylinder groups than 50% of the buffers in the buffer cache. The first results in long access times, while the second results in many buffers being used by metadata operations. Such operations use cylinder group blocks and on-disk inode blocks. The cylinder group block (fs->fs_cblkno) contains struct cg, inode and block bit maps. It is 2k in size for the default filesystem parameters. If new and parent directories are located in different cylinder groups then the system performs more input/output operations and uses more buffers. On filesystems with many cylinder groups, lots of cache buffers are used for metadata operations. My solution for this problem is very simple. I allocate many directories in one cylinder group. I also do some things, so that the new allocation method does not cause excessive fragmentation and all directory inodes will not be located at a location far from its file's inodes and data. The algorithm is: /* * Find a cylinder group to place a directory. * * The policy implemented by this algorithm is to allocate a * directory inode in the same cylinder group as its parent * directory, but also to reserve space for its files inodes * and data. Restrict the number of directories which may be * allocated one after another in the same cylinder group * without intervening allocation of files. * * If we allocate a first level directory then force allocation * in another cylinder group. */ My early versions of dirpref give me a good results for a wide range of file operations and different filesystem capacities except one case: those applications that create their entire directory structure first and only later fill this structure with files. My solution for such and similar cases is to limit a number of directories which may be created one after another in the same cylinder group without intervening file creations. For this purpose, I allocate an array of counters at mount time. This array is linked to the superblock fs->fs_contigdirs[cg]. Each time a directory is created the counter increases and each time a file is created the counter decreases. A 60Gb filesystem with 8mb/cg requires 10kb of memory for the counters array. The maxcontigdirs is a maximum number of directories which may be created without an intervening file creation. I found in my tests that the best performance occurs when I restrict the number of directories in one cylinder group such that all its files may be located in the same cylinder group. There may be some deterioration in performance if all the file inodes are in the same cylinder group as its containing directory, but their data partially resides in a different cylinder group. The maxcontigdirs value is calculated to try to prevent this condition. Since there is no way to know how many files and directories will be allocated later I added two optimization parameters in superblock/tunefs. They are: int32_t fs_avgfilesize; /* expected average file size */ int32_t fs_avgfpdir; /* expected # of files per directory */ These parameters have reasonable defaults but may be tweeked for special uses of a filesystem. They are only necessary in rare cases like better tuning a filesystem being used to store a squid cache. I have been using this algorithm for about 3 months. I have done a lot of testing on filesystems with different capacities, average filesize, average number of files per directory, and so on. I think this algorithm has no negative impact on filesystem perfomance. It works better than the default one in all cases. The new dirpref will greatly improve untarring/removing/coping of big directories, decrease load on cvs servers and much more. The new dirpref doesn't speedup a compilation process, but also doesn't slow it down. Obtained from: Grigoriy Orlov <gluk@ptci.ru>
2001-04-10 08:38:59 +00:00
/* Compatibility for old filesystems XXX */
if (fs->fs_avgfilesize <= 0) /* XXX */
fs->fs_avgfilesize = AVFILESIZ; /* XXX */
if (fs->fs_avgfpdir <= 0) /* XXX */
fs->fs_avgfpdir = AFPDIR; /* XXX */
1994-05-24 10:09:53 +00:00
mp->mnt_data = (qaddr_t)ump;
mp->mnt_stat.f_fsid.val[0] = fs->fs_id[0];
mp->mnt_stat.f_fsid.val[1] = fs->fs_id[1];
if (fs->fs_id[0] == 0 || fs->fs_id[1] == 0 ||
vfs_getvfs(&mp->mnt_stat.f_fsid))
vfs_getnewfsid(mp);
1994-05-24 10:09:53 +00:00
mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
mp->mnt_flag |= MNT_LOCAL;
1994-05-24 10:09:53 +00:00
ump->um_mountp = mp;
ump->um_dev = dev;
ump->um_devvp = devvp;
ump->um_nindir = fs->fs_nindir;
ump->um_bptrtodb = fs->fs_fsbtodb;
ump->um_seqinc = fs->fs_frag;
for (i = 0; i < MAXQUOTAS; i++)
ump->um_quotas[i] = NULLVP;
#ifdef UFS_EXTATTR
ufs_extattr_uepm_init(&ump->um_extattr);
#endif
devvp->v_rdev->si_mountpoint = mp;
1994-05-24 10:09:53 +00:00
ffs_oldfscompat(fs);
/*
* Set FS local "last mounted on" information (NULL pad)
*/
copystr( mp->mnt_stat.f_mntonname, /* mount point*/
fs->fs_fsmnt, /* copy area*/
sizeof(fs->fs_fsmnt) - 1, /* max size*/
&strsize); /* real size*/
bzero( fs->fs_fsmnt + strsize, sizeof(fs->fs_fsmnt) - strsize);
if( mp->mnt_flag & MNT_ROOTFS) {
/*
* Root mount; update timestamp in mount structure.
* this will be used by the common root mount code
* to update the system clock.
*/
mp->mnt_time = fs->fs_time;
}
ump->um_savedmaxfilesize = fs->fs_maxfilesize; /* XXX */
maxfilesize = (u_int64_t)0x40000000 * fs->fs_bsize - 1; /* XXX */
if (fs->fs_maxfilesize > maxfilesize) /* XXX */
fs->fs_maxfilesize = maxfilesize; /* XXX */
if (bigcgs) {
if (fs->fs_sparecon[0] <= 0)
fs->fs_sparecon[0] = fs->fs_cgsize;
fs->fs_cgsize = fs->fs_bsize;
}
if (ronly == 0) {
if ((fs->fs_flags & FS_DOSOFTDEP) &&
(error = softdep_mount(devvp, mp, fs, cred)) != 0) {
free(fs->fs_csp, M_UFSMNT);
goto out;
}
if (fs->fs_snapinum[0] != 0)
ffs_snapshot_mount(mp);
fs->fs_fmod = 1;
fs->fs_clean = 0;
(void) ffs_sbupdate(ump, MNT_WAIT);
}
#ifdef UFS_EXTATTR
#ifdef UFS_EXTATTR_AUTOSTART
/*
*
o Implement "options FFS_EXTATTR_AUTOSTART", which depends on "options FFS_EXTATTR". When extended attribute auto-starting is enabled, FFS will scan the .attribute directory off of the root of each file system, as it is mounted. If .attribute exists, EA support will be started for the file system. If there are files in the directory, FFS will attempt to start them as attribute backing files for attributes baring the same name. All attributes are started before access to the file system is permitted, so this permits race-free enabling of attributes. For attributes backing support for security features, such as ACLs, MAC, Capabilities, this is vital, as it prevents the file system attributes from getting out of sync as a result of file system operations between mount-time and the enabling of the extended attribute. The userland extattrctl tool will still function exactly as previously. Files must be placed directly in .attribute, which must be directly off of the file system root: symbolic links are not permitted. FFS_EXTATTR will continue to be able to function without FFS_EXTATTR_AUTOSTART for sites that do not want/require auto-starting. If you're using the UFS_ACL code available from www.TrustedBSD.org, using FFS_EXTATTR_AUTOSTART is recommended. o This support is implemented by adding an invocation of ufs_extattr_autostart() to ffs_mountfs(). In addition, several new supporting calls are introduced in ufs_extattr.c: ufs_extattr_autostart(): start EAs on the specified mount ufs_extattr_lookup(): given a directory and filename, return the vnode for the file. ufs_extattr_enable_with_open(): invoke ufs_extattr_enable() after doing the equililent of vn_open() on the passed file. ufs_extattr_iterate_directory(): iterate over a directory, invoking ufs_extattr_lookup() and ufs_extattr_enable_with_open() on each entry. o This feature is not widely tested, and therefore may contain bugs, caution is advised. Several changes are in the pipeline for this feature, including breaking out of EA namespaces into subdirectories of .attribute (this is waiting on the updated EA API), as well as a per-filesystem flag indicating whether or not EAs should be auto-started. This is required because administrators may not want .attribute auto-started on all file systems, especially if non-administrators have write access to the root of a file system. Obtained from: TrustedBSD Project
2001-03-14 05:32:31 +00:00
* Auto-starting does the following:
* - check for /.attribute in the fs, and extattr_start if so
* - for each file in .attribute, enable that file with
* an attribute of the same name.
* Not clear how to report errors -- probably eat them.
* This would all happen while the file system was busy/not
* available, so would effectively be "atomic".
*/
(void) ufs_extattr_autostart(mp, td);
#endif /* !UFS_EXTATTR_AUTOSTART */
#endif /* !UFS_EXTATTR */
1994-05-24 10:09:53 +00:00
return (0);
out:
devvp->v_rdev->si_mountpoint = NULL;
1994-05-24 10:09:53 +00:00
if (bp)
brelse(bp);
(void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, cred, td);
1994-05-24 10:09:53 +00:00
if (ump) {
free(ump->um_fs, M_UFSMNT);
free(ump, M_UFSMNT);
mp->mnt_data = (qaddr_t)0;
}
return (error);
}
/*
* Sanity checks for old file systems.
*
* XXX - goes away some day.
*/
1995-12-17 21:14:36 +00:00
static int
1994-05-24 10:09:53 +00:00
ffs_oldfscompat(fs)
struct fs *fs;
{
fs->fs_npsect = max(fs->fs_npsect, fs->fs_nsect); /* XXX */
fs->fs_interleave = max(fs->fs_interleave, 1); /* XXX */
if (fs->fs_postblformat == FS_42POSTBLFMT) /* XXX */
fs->fs_nrpos = 8; /* XXX */
if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
1995-11-20 12:25:37 +00:00
#if 0
int i; /* XXX */
u_int64_t sizepb = fs->fs_bsize; /* XXX */
/* XXX */
1994-05-24 10:09:53 +00:00
fs->fs_maxfilesize = fs->fs_bsize * NDADDR - 1; /* XXX */
for (i = 0; i < NIADDR; i++) { /* XXX */
sizepb *= NINDIR(fs); /* XXX */
fs->fs_maxfilesize += sizepb; /* XXX */
} /* XXX */
#endif
fs->fs_maxfilesize = (u_quad_t) 1LL << 39;
1994-05-24 10:09:53 +00:00
fs->fs_qbmask = ~fs->fs_bmask; /* XXX */
fs->fs_qfmask = ~fs->fs_fmask; /* XXX */
} /* XXX */
return (0);
}
/*
* unmount system call
*/
int
ffs_unmount(mp, mntflags, td)
1994-05-24 10:09:53 +00:00
struct mount *mp;
int mntflags;
struct thread *td;
1994-05-24 10:09:53 +00:00
{
register struct ufsmount *ump = VFSTOUFS(mp);
1994-05-24 10:09:53 +00:00
register struct fs *fs;
int error, flags;
1994-05-24 10:09:53 +00:00
flags = 0;
if (mntflags & MNT_FORCE) {
flags |= FORCECLOSE;
}
#ifdef UFS_EXTATTR
if ((error = ufs_extattr_stop(mp, td))) {
if (error != EOPNOTSUPP)
printf("ffs_unmount: ufs_extattr_stop returned %d\n",
error);
} else {
ufs_extattr_uepm_destroy(&ump->um_extattr);
}
#endif
if (mp->mnt_flag & MNT_SOFTDEP) {
if ((error = softdep_flushfiles(mp, flags, td)) != 0)
return (error);
} else {
if ((error = ffs_flushfiles(mp, flags, td)) != 0)
return (error);
}
1994-05-24 10:09:53 +00:00
fs = ump->um_fs;
if (bigcgs) {
fs->fs_cgsize = fs->fs_sparecon[0];
fs->fs_sparecon[0] = 0;
}
if (fs->fs_pendingblocks != 0 || fs->fs_pendinginodes != 0) {
printf("%s: unmount pending error: blocks %d files %d\n",
fs->fs_fsmnt, fs->fs_pendingblocks, fs->fs_pendinginodes);
fs->fs_pendingblocks = 0;
fs->fs_pendinginodes = 0;
}
if (fs->fs_ronly == 0) {
fs->fs_clean = fs->fs_flags & (FS_UNCLEAN|FS_NEEDSFSCK) ? 0 : 1;
error = ffs_sbupdate(ump, MNT_WAIT);
if (error) {
fs->fs_clean = 0;
return (error);
}
}
ump->um_devvp->v_rdev->si_mountpoint = NULL;
vinvalbuf(ump->um_devvp, V_SAVE, NOCRED, td, 0, 0);
error = VOP_CLOSE(ump->um_devvp, fs->fs_ronly ? FREAD : FREAD|FWRITE,
NOCRED, td);
vrele(ump->um_devvp);
free(fs->fs_csp, M_UFSMNT);
1994-05-24 10:09:53 +00:00
free(fs, M_UFSMNT);
free(ump, M_UFSMNT);
mp->mnt_data = (qaddr_t)0;
mp->mnt_flag &= ~MNT_LOCAL;
1994-05-24 10:09:53 +00:00
return (error);
}
/*
* Flush out all the files in a filesystem.
*/
int
ffs_flushfiles(mp, flags, td)
1994-05-24 10:09:53 +00:00
register struct mount *mp;
int flags;
struct thread *td;
1994-05-24 10:09:53 +00:00
{
register struct ufsmount *ump;
int error;
1994-05-24 10:09:53 +00:00
ump = VFSTOUFS(mp);
#ifdef QUOTA
if (mp->mnt_flag & MNT_QUOTA) {
int i;
error = vflush(mp, 0, SKIPSYSTEM|flags);
if (error)
1994-05-24 10:09:53 +00:00
return (error);
for (i = 0; i < MAXQUOTAS; i++) {
if (ump->um_quotas[i] == NULLVP)
continue;
quotaoff(td, mp, i);
1994-05-24 10:09:53 +00:00
}
/*
* Here we fall through to vflush again to ensure
* that we have gotten rid of all the system vnodes.
*/
}
#endif
if (ump->um_devvp->v_flag & VCOPYONWRITE) {
if ((error = vflush(mp, 0, SKIPSYSTEM | flags)) != 0)
return (error);
ffs_snapshot_unmount(mp);
/*
* Here we fall through to vflush again to ensure
* that we have gotten rid of all the system vnodes.
*/
}
/*
* Flush all the files.
*/
if ((error = vflush(mp, 0, flags)) != 0)
return (error);
/*
* Flush filesystem metadata.
*/
vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY, td);
error = VOP_FSYNC(ump->um_devvp, td->td_ucred, MNT_WAIT, td);
VOP_UNLOCK(ump->um_devvp, 0, td);
1994-05-24 10:09:53 +00:00
return (error);
}
/*
* Get file system statistics.
*/
int
ffs_statfs(mp, sbp, td)
1994-05-24 10:09:53 +00:00
struct mount *mp;
register struct statfs *sbp;
struct thread *td;
1994-05-24 10:09:53 +00:00
{
register struct ufsmount *ump;
register struct fs *fs;
ump = VFSTOUFS(mp);
fs = ump->um_fs;
if (fs->fs_magic != FS_MAGIC)
panic("ffs_statfs");
sbp->f_bsize = fs->fs_fsize;
sbp->f_iosize = fs->fs_bsize;
sbp->f_blocks = fs->fs_dsize;
sbp->f_bfree = fs->fs_cstotal.cs_nbfree * fs->fs_frag +
fs->fs_cstotal.cs_nffree + dbtofsb(fs, fs->fs_pendingblocks);
sbp->f_bavail = freespace(fs, fs->fs_minfree) +
dbtofsb(fs, fs->fs_pendingblocks);
1994-05-24 10:09:53 +00:00
sbp->f_files = fs->fs_ncg * fs->fs_ipg - ROOTINO;
sbp->f_ffree = fs->fs_cstotal.cs_nifree + fs->fs_pendinginodes;
1994-05-24 10:09:53 +00:00
if (sbp != &mp->mnt_stat) {
sbp->f_type = mp->mnt_vfc->vfc_typenum;
1994-05-24 10:09:53 +00:00
bcopy((caddr_t)mp->mnt_stat.f_mntonname,
(caddr_t)&sbp->f_mntonname[0], MNAMELEN);
bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
(caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
}
return (0);
}
/*
* Go through the disk queues to initiate sandbagged IO;
* go through the inodes to write those that have been modified;
* initiate the writing of the super block if it has been modified.
*
* Note: we are always called with the filesystem marked `MPBUSY'.
*/
int
ffs_sync(mp, waitfor, cred, td)
1994-05-24 10:09:53 +00:00
struct mount *mp;
int waitfor;
struct ucred *cred;
struct thread *td;
1994-05-24 10:09:53 +00:00
{
struct vnode *nvp, *vp, *devvp;
struct inode *ip;
struct ufsmount *ump = VFSTOUFS(mp);
struct fs *fs;
This patch corrects the first round of panics and hangs reported with the new snapshot code. Update addaliasu to correctly implement the semantics of the old checkalias function. When a device vnode first comes into existence, check to see if an anonymous vnode for the same device was created at boot time by bdevvp(). If so, adopt the bdevvp vnode rather than creating a new vnode for the device. This corrects a problem which caused the kernel to panic when taking a snapshot of the root filesystem. Change the calling convention of vn_write_suspend_wait() to be the same as vn_start_write(). Split out softdep_flushworklist() from softdep_flushfiles() so that it can be used to clear the work queue when suspending filesystem operations. Access to buffers becomes recursive so that snapshots can recursively traverse their indirect blocks using ffs_copyonwrite() when checking for the need for copy on write when flushing one of their own indirect blocks. This eliminates a deadlock between the syncer daemon and a process taking a snapshot. Ensure that softdep_process_worklist() can never block because of a snapshot being taken. This eliminates a problem with buffer starvation. Cleanup change in ffs_sync() which did not synchronously wait when MNT_WAIT was specified. The result was an unclean filesystem panic when doing forcible unmount with heavy filesystem I/O in progress. Return a zero'ed block when reading a block that was not in use at the time that a snapshot was taken. Normally, these blocks should never be read. However, the readahead code will occationally read them which can cause unexpected behavior. Clean up the debugging code that ensures that no blocks be written on a filesystem while it is suspended. Snapshots must explicitly label the blocks that they are writing during the suspension so that they do not cause a `write on suspended filesystem' panic. Reorganize ffs_copyonwrite() to eliminate a deadlock and also to prevent a race condition that would permit the same block to be copied twice. This change eliminates an unexpected soft updates inconsistency in fsck caused by the double allocation. Use bqrelse rather than brelse for buffers that will be needed soon again by the snapshot code. This improves snapshot performance.
2000-07-24 05:28:33 +00:00
int error, count, wait, lockreq, allerror = 0;
1994-05-24 10:09:53 +00:00
fs = ump->um_fs;
if (fs->fs_fmod != 0 && fs->fs_ronly != 0) { /* XXX */
printf("fs = %s\n", fs->fs_fsmnt);
panic("ffs_sync: rofs mod");
1994-05-24 10:09:53 +00:00
}
/*
* Write back each (modified) inode.
*/
This patch corrects the first round of panics and hangs reported with the new snapshot code. Update addaliasu to correctly implement the semantics of the old checkalias function. When a device vnode first comes into existence, check to see if an anonymous vnode for the same device was created at boot time by bdevvp(). If so, adopt the bdevvp vnode rather than creating a new vnode for the device. This corrects a problem which caused the kernel to panic when taking a snapshot of the root filesystem. Change the calling convention of vn_write_suspend_wait() to be the same as vn_start_write(). Split out softdep_flushworklist() from softdep_flushfiles() so that it can be used to clear the work queue when suspending filesystem operations. Access to buffers becomes recursive so that snapshots can recursively traverse their indirect blocks using ffs_copyonwrite() when checking for the need for copy on write when flushing one of their own indirect blocks. This eliminates a deadlock between the syncer daemon and a process taking a snapshot. Ensure that softdep_process_worklist() can never block because of a snapshot being taken. This eliminates a problem with buffer starvation. Cleanup change in ffs_sync() which did not synchronously wait when MNT_WAIT was specified. The result was an unclean filesystem panic when doing forcible unmount with heavy filesystem I/O in progress. Return a zero'ed block when reading a block that was not in use at the time that a snapshot was taken. Normally, these blocks should never be read. However, the readahead code will occationally read them which can cause unexpected behavior. Clean up the debugging code that ensures that no blocks be written on a filesystem while it is suspended. Snapshots must explicitly label the blocks that they are writing during the suspension so that they do not cause a `write on suspended filesystem' panic. Reorganize ffs_copyonwrite() to eliminate a deadlock and also to prevent a race condition that would permit the same block to be copied twice. This change eliminates an unexpected soft updates inconsistency in fsck caused by the double allocation. Use bqrelse rather than brelse for buffers that will be needed soon again by the snapshot code. This improves snapshot performance.
2000-07-24 05:28:33 +00:00
wait = 0;
lockreq = LK_EXCLUSIVE | LK_NOWAIT;
This patch corrects the first round of panics and hangs reported with the new snapshot code. Update addaliasu to correctly implement the semantics of the old checkalias function. When a device vnode first comes into existence, check to see if an anonymous vnode for the same device was created at boot time by bdevvp(). If so, adopt the bdevvp vnode rather than creating a new vnode for the device. This corrects a problem which caused the kernel to panic when taking a snapshot of the root filesystem. Change the calling convention of vn_write_suspend_wait() to be the same as vn_start_write(). Split out softdep_flushworklist() from softdep_flushfiles() so that it can be used to clear the work queue when suspending filesystem operations. Access to buffers becomes recursive so that snapshots can recursively traverse their indirect blocks using ffs_copyonwrite() when checking for the need for copy on write when flushing one of their own indirect blocks. This eliminates a deadlock between the syncer daemon and a process taking a snapshot. Ensure that softdep_process_worklist() can never block because of a snapshot being taken. This eliminates a problem with buffer starvation. Cleanup change in ffs_sync() which did not synchronously wait when MNT_WAIT was specified. The result was an unclean filesystem panic when doing forcible unmount with heavy filesystem I/O in progress. Return a zero'ed block when reading a block that was not in use at the time that a snapshot was taken. Normally, these blocks should never be read. However, the readahead code will occationally read them which can cause unexpected behavior. Clean up the debugging code that ensures that no blocks be written on a filesystem while it is suspended. Snapshots must explicitly label the blocks that they are writing during the suspension so that they do not cause a `write on suspended filesystem' panic. Reorganize ffs_copyonwrite() to eliminate a deadlock and also to prevent a race condition that would permit the same block to be copied twice. This change eliminates an unexpected soft updates inconsistency in fsck caused by the double allocation. Use bqrelse rather than brelse for buffers that will be needed soon again by the snapshot code. This improves snapshot performance.
2000-07-24 05:28:33 +00:00
if (waitfor == MNT_WAIT) {
wait = 1;
lockreq = LK_EXCLUSIVE;
This patch corrects the first round of panics and hangs reported with the new snapshot code. Update addaliasu to correctly implement the semantics of the old checkalias function. When a device vnode first comes into existence, check to see if an anonymous vnode for the same device was created at boot time by bdevvp(). If so, adopt the bdevvp vnode rather than creating a new vnode for the device. This corrects a problem which caused the kernel to panic when taking a snapshot of the root filesystem. Change the calling convention of vn_write_suspend_wait() to be the same as vn_start_write(). Split out softdep_flushworklist() from softdep_flushfiles() so that it can be used to clear the work queue when suspending filesystem operations. Access to buffers becomes recursive so that snapshots can recursively traverse their indirect blocks using ffs_copyonwrite() when checking for the need for copy on write when flushing one of their own indirect blocks. This eliminates a deadlock between the syncer daemon and a process taking a snapshot. Ensure that softdep_process_worklist() can never block because of a snapshot being taken. This eliminates a problem with buffer starvation. Cleanup change in ffs_sync() which did not synchronously wait when MNT_WAIT was specified. The result was an unclean filesystem panic when doing forcible unmount with heavy filesystem I/O in progress. Return a zero'ed block when reading a block that was not in use at the time that a snapshot was taken. Normally, these blocks should never be read. However, the readahead code will occationally read them which can cause unexpected behavior. Clean up the debugging code that ensures that no blocks be written on a filesystem while it is suspended. Snapshots must explicitly label the blocks that they are writing during the suspension so that they do not cause a `write on suspended filesystem' panic. Reorganize ffs_copyonwrite() to eliminate a deadlock and also to prevent a race condition that would permit the same block to be copied twice. This change eliminates an unexpected soft updates inconsistency in fsck caused by the double allocation. Use bqrelse rather than brelse for buffers that will be needed soon again by the snapshot code. This improves snapshot performance.
2000-07-24 05:28:33 +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(&mntvnode_mtx);
1994-05-24 10:09:53 +00:00
loop:
for (vp = TAILQ_FIRST(&mp->mnt_nvnodelist); vp != NULL; vp = nvp) {
1994-05-24 10:09:53 +00:00
/*
* If the vnode that we are about to sync is no longer
* associated with this mount point, start over.
*/
if (vp->v_mount != mp)
goto loop;
/*
* Depend on the mntvnode_slock to keep things stable enough
* for a quick test. Since there might be hundreds of
* thousands of vnodes, we cannot afford even a subroutine
* call unless there's a good chance that we have work to do.
*/
nvp = TAILQ_NEXT(vp, v_nmntvnodes);
1994-05-24 10:09:53 +00:00
ip = VTOI(vp);
if (vp->v_type == VNON || ((ip->i_flag &
(IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
TAILQ_EMPTY(&vp->v_dirtyblkhd))) {
1994-05-24 10:09:53 +00:00
continue;
}
if (vp->v_type != VCHR) {
mtx_unlock(&mntvnode_mtx);
if ((error = vget(vp, lockreq, 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_lock(&mntvnode_mtx);
if (error == ENOENT)
goto loop;
} else {
if ((error = VOP_FSYNC(vp, cred, waitfor, td)) != 0)
allerror = error;
VOP_UNLOCK(vp, 0, td);
vrele(vp);
mtx_lock(&mntvnode_mtx);
}
} else {
mtx_unlock(&mntvnode_mtx);
This patch corrects the first round of panics and hangs reported with the new snapshot code. Update addaliasu to correctly implement the semantics of the old checkalias function. When a device vnode first comes into existence, check to see if an anonymous vnode for the same device was created at boot time by bdevvp(). If so, adopt the bdevvp vnode rather than creating a new vnode for the device. This corrects a problem which caused the kernel to panic when taking a snapshot of the root filesystem. Change the calling convention of vn_write_suspend_wait() to be the same as vn_start_write(). Split out softdep_flushworklist() from softdep_flushfiles() so that it can be used to clear the work queue when suspending filesystem operations. Access to buffers becomes recursive so that snapshots can recursively traverse their indirect blocks using ffs_copyonwrite() when checking for the need for copy on write when flushing one of their own indirect blocks. This eliminates a deadlock between the syncer daemon and a process taking a snapshot. Ensure that softdep_process_worklist() can never block because of a snapshot being taken. This eliminates a problem with buffer starvation. Cleanup change in ffs_sync() which did not synchronously wait when MNT_WAIT was specified. The result was an unclean filesystem panic when doing forcible unmount with heavy filesystem I/O in progress. Return a zero'ed block when reading a block that was not in use at the time that a snapshot was taken. Normally, these blocks should never be read. However, the readahead code will occationally read them which can cause unexpected behavior. Clean up the debugging code that ensures that no blocks be written on a filesystem while it is suspended. Snapshots must explicitly label the blocks that they are writing during the suspension so that they do not cause a `write on suspended filesystem' panic. Reorganize ffs_copyonwrite() to eliminate a deadlock and also to prevent a race condition that would permit the same block to be copied twice. This change eliminates an unexpected soft updates inconsistency in fsck caused by the double allocation. Use bqrelse rather than brelse for buffers that will be needed soon again by the snapshot code. This improves snapshot performance.
2000-07-24 05:28:33 +00:00
UFS_UPDATE(vp, wait);
mtx_lock(&mntvnode_mtx);
}
if (TAILQ_NEXT(vp, v_nmntvnodes) != nvp)
goto loop;
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(&mntvnode_mtx);
1994-05-24 10:09:53 +00:00
/*
* Force stale file system control information to be flushed.
*/
This patch corrects the first round of panics and hangs reported with the new snapshot code. Update addaliasu to correctly implement the semantics of the old checkalias function. When a device vnode first comes into existence, check to see if an anonymous vnode for the same device was created at boot time by bdevvp(). If so, adopt the bdevvp vnode rather than creating a new vnode for the device. This corrects a problem which caused the kernel to panic when taking a snapshot of the root filesystem. Change the calling convention of vn_write_suspend_wait() to be the same as vn_start_write(). Split out softdep_flushworklist() from softdep_flushfiles() so that it can be used to clear the work queue when suspending filesystem operations. Access to buffers becomes recursive so that snapshots can recursively traverse their indirect blocks using ffs_copyonwrite() when checking for the need for copy on write when flushing one of their own indirect blocks. This eliminates a deadlock between the syncer daemon and a process taking a snapshot. Ensure that softdep_process_worklist() can never block because of a snapshot being taken. This eliminates a problem with buffer starvation. Cleanup change in ffs_sync() which did not synchronously wait when MNT_WAIT was specified. The result was an unclean filesystem panic when doing forcible unmount with heavy filesystem I/O in progress. Return a zero'ed block when reading a block that was not in use at the time that a snapshot was taken. Normally, these blocks should never be read. However, the readahead code will occationally read them which can cause unexpected behavior. Clean up the debugging code that ensures that no blocks be written on a filesystem while it is suspended. Snapshots must explicitly label the blocks that they are writing during the suspension so that they do not cause a `write on suspended filesystem' panic. Reorganize ffs_copyonwrite() to eliminate a deadlock and also to prevent a race condition that would permit the same block to be copied twice. This change eliminates an unexpected soft updates inconsistency in fsck caused by the double allocation. Use bqrelse rather than brelse for buffers that will be needed soon again by the snapshot code. This improves snapshot performance.
2000-07-24 05:28:33 +00:00
if (waitfor == MNT_WAIT) {
if ((error = softdep_flushworklist(ump->um_mountp, &count, td)))
This patch corrects the first round of panics and hangs reported with the new snapshot code. Update addaliasu to correctly implement the semantics of the old checkalias function. When a device vnode first comes into existence, check to see if an anonymous vnode for the same device was created at boot time by bdevvp(). If so, adopt the bdevvp vnode rather than creating a new vnode for the device. This corrects a problem which caused the kernel to panic when taking a snapshot of the root filesystem. Change the calling convention of vn_write_suspend_wait() to be the same as vn_start_write(). Split out softdep_flushworklist() from softdep_flushfiles() so that it can be used to clear the work queue when suspending filesystem operations. Access to buffers becomes recursive so that snapshots can recursively traverse their indirect blocks using ffs_copyonwrite() when checking for the need for copy on write when flushing one of their own indirect blocks. This eliminates a deadlock between the syncer daemon and a process taking a snapshot. Ensure that softdep_process_worklist() can never block because of a snapshot being taken. This eliminates a problem with buffer starvation. Cleanup change in ffs_sync() which did not synchronously wait when MNT_WAIT was specified. The result was an unclean filesystem panic when doing forcible unmount with heavy filesystem I/O in progress. Return a zero'ed block when reading a block that was not in use at the time that a snapshot was taken. Normally, these blocks should never be read. However, the readahead code will occationally read them which can cause unexpected behavior. Clean up the debugging code that ensures that no blocks be written on a filesystem while it is suspended. Snapshots must explicitly label the blocks that they are writing during the suspension so that they do not cause a `write on suspended filesystem' panic. Reorganize ffs_copyonwrite() to eliminate a deadlock and also to prevent a race condition that would permit the same block to be copied twice. This change eliminates an unexpected soft updates inconsistency in fsck caused by the double allocation. Use bqrelse rather than brelse for buffers that will be needed soon again by the snapshot code. This improves snapshot performance.
2000-07-24 05:28:33 +00:00
allerror = error;
/* Flushed work items may create new vnodes to clean */
if (count) {
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(&mntvnode_mtx);
This patch corrects the first round of panics and hangs reported with the new snapshot code. Update addaliasu to correctly implement the semantics of the old checkalias function. When a device vnode first comes into existence, check to see if an anonymous vnode for the same device was created at boot time by bdevvp(). If so, adopt the bdevvp vnode rather than creating a new vnode for the device. This corrects a problem which caused the kernel to panic when taking a snapshot of the root filesystem. Change the calling convention of vn_write_suspend_wait() to be the same as vn_start_write(). Split out softdep_flushworklist() from softdep_flushfiles() so that it can be used to clear the work queue when suspending filesystem operations. Access to buffers becomes recursive so that snapshots can recursively traverse their indirect blocks using ffs_copyonwrite() when checking for the need for copy on write when flushing one of their own indirect blocks. This eliminates a deadlock between the syncer daemon and a process taking a snapshot. Ensure that softdep_process_worklist() can never block because of a snapshot being taken. This eliminates a problem with buffer starvation. Cleanup change in ffs_sync() which did not synchronously wait when MNT_WAIT was specified. The result was an unclean filesystem panic when doing forcible unmount with heavy filesystem I/O in progress. Return a zero'ed block when reading a block that was not in use at the time that a snapshot was taken. Normally, these blocks should never be read. However, the readahead code will occationally read them which can cause unexpected behavior. Clean up the debugging code that ensures that no blocks be written on a filesystem while it is suspended. Snapshots must explicitly label the blocks that they are writing during the suspension so that they do not cause a `write on suspended filesystem' panic. Reorganize ffs_copyonwrite() to eliminate a deadlock and also to prevent a race condition that would permit the same block to be copied twice. This change eliminates an unexpected soft updates inconsistency in fsck caused by the double allocation. Use bqrelse rather than brelse for buffers that will be needed soon again by the snapshot code. This improves snapshot performance.
2000-07-24 05:28:33 +00:00
goto loop;
}
}
#ifdef QUOTA
qsync(mp);
#endif
devvp = ump->um_devvp;
mtx_lock(&devvp->v_interlock);
if (waitfor != MNT_LAZY &&
(devvp->v_numoutput > 0 || TAILQ_FIRST(&devvp->v_dirtyblkhd))) {
mtx_unlock(&devvp->v_interlock);
vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
if ((error = VOP_FSYNC(devvp, cred, waitfor, td)) != 0)
allerror = error;
VOP_UNLOCK(devvp, 0, td);
if (waitfor == MNT_WAIT) {
mtx_lock(&mntvnode_mtx);
goto loop;
}
} else
mtx_unlock(&devvp->v_interlock);
/*
* Write back modified superblock.
*/
if (fs->fs_fmod != 0 && (error = ffs_sbupdate(ump, waitfor)) != 0)
allerror = error;
1994-05-24 10:09:53 +00:00
return (allerror);
}
/*
* Look up a FFS dinode number to find its incore vnode, otherwise read it
* in from disk. If it is in core, wait for the lock bit to clear, then
* return the inode locked. Detection and handling of mount points must be
* done by the calling routine.
*/
1995-12-17 21:14:36 +00:00
static int ffs_inode_hash_lock;
/*
* ffs_inode_hash_lock is a variable to manage mutual exclusion
* of vnode allocation and intertion to the hash, especially to
* avoid holding more than one vnodes for the same inode in the
* hash table. ffs_inode_hash_lock must hence be tested-and-set
* or cleared atomically, accomplished by ffs_inode_hash_mtx.
*
* As vnode allocation may block during MALLOC() and zone
* allocation, we should also do msleep() to give away the CPU
* if anyone else is allocating a vnode. lockmgr is not suitable
* here because someone else may insert to the hash table the
* vnode we are trying to allocate during our sleep, in which
* case the hash table needs to be examined once again after
* waking up.
*/
static struct mtx ffs_inode_hash_mtx;
1994-05-24 10:09:53 +00:00
int
ffs_vget(mp, ino, vpp)
struct mount *mp;
ino_t ino;
struct vnode **vpp;
{
struct fs *fs;
struct inode *ip;
1994-05-24 10:09:53 +00:00
struct ufsmount *ump;
struct buf *bp;
struct vnode *vp;
dev_t dev;
int error, want_wakeup;
1994-05-24 10:09:53 +00:00
ump = VFSTOUFS(mp);
dev = ump->um_dev;
restart:
This mega-commit is meant to fix numerous interrelated problems. There has been some bitrot and incorrect assumptions in the vfs_bio code. These problems have manifest themselves worse on NFS type filesystems, but can still affect local filesystems under certain circumstances. Most of the problems have involved mmap consistancy, and as a side-effect broke the vfs.ioopt code. This code might have been committed seperately, but almost everything is interrelated. 1) Allow (pmap_object_init_pt) prefaulting of buffer-busy pages that are fully valid. 2) Rather than deactivating erroneously read initial (header) pages in kern_exec, we now free them. 3) Fix the rundown of non-VMIO buffers that are in an inconsistent (missing vp) state. 4) Fix the disassociation of pages from buffers in brelse. The previous code had rotted and was faulty in a couple of important circumstances. 5) Remove a gratuitious buffer wakeup in vfs_vmio_release. 6) Remove a crufty and currently unused cluster mechanism for VBLK files in vfs_bio_awrite. When the code is functional, I'll add back a cleaner version. 7) The page busy count wakeups assocated with the buffer cache usage were incorrectly cleaned up in a previous commit by me. Revert to the original, correct version, but with a cleaner implementation. 8) The cluster read code now tries to keep data associated with buffers more aggressively (without breaking the heuristics) when it is presumed that the read data (buffers) will be soon needed. 9) Change to filesystem lockmgr locks so that they use LK_NOPAUSE. The delay loop waiting is not useful for filesystem locks, due to the length of the time intervals. 10) Correct and clean-up spec_getpages. 11) Implement a fully functional nfs_getpages, nfs_putpages. 12) Fix nfs_write so that modifications are coherent with the NFS data on the server disk (at least as well as NFS seems to allow.) 13) Properly support MS_INVALIDATE on NFS. 14) Properly pass down MS_INVALIDATE to lower levels of the VM code from vm_map_clean. 15) Better support the notion of pages being busy but valid, so that fewer in-transit waits occur. (use p->busy more for pageouts instead of PG_BUSY.) Since the page is fully valid, it is still usable for reads. 16) It is possible (in error) for cached pages to be busy. Make the page allocation code handle that case correctly. (It should probably be a printf or panic, but I want the system to handle coding errors robustly. I'll probably add a printf.) 17) Correct the design and usage of vm_page_sleep. It didn't handle consistancy problems very well, so make the design a little less lofty. After vm_page_sleep, if it ever blocked, it is still important to relookup the page (if the object generation count changed), and verify it's status (always.) 18) In vm_pageout.c, vm_pageout_clean had rotted, so clean that up. 19) Push the page busy for writes and VM_PROT_READ into vm_pageout_flush. 20) Fix vm_pager_put_pages and it's descendents to support an int flag instead of a boolean, so that we can pass down the invalidate bit.
1998-03-07 21:37:31 +00:00
if ((*vpp = ufs_ihashget(dev, ino)) != NULL) {
1994-05-24 10:09:53 +00:00
return (0);
This mega-commit is meant to fix numerous interrelated problems. There has been some bitrot and incorrect assumptions in the vfs_bio code. These problems have manifest themselves worse on NFS type filesystems, but can still affect local filesystems under certain circumstances. Most of the problems have involved mmap consistancy, and as a side-effect broke the vfs.ioopt code. This code might have been committed seperately, but almost everything is interrelated. 1) Allow (pmap_object_init_pt) prefaulting of buffer-busy pages that are fully valid. 2) Rather than deactivating erroneously read initial (header) pages in kern_exec, we now free them. 3) Fix the rundown of non-VMIO buffers that are in an inconsistent (missing vp) state. 4) Fix the disassociation of pages from buffers in brelse. The previous code had rotted and was faulty in a couple of important circumstances. 5) Remove a gratuitious buffer wakeup in vfs_vmio_release. 6) Remove a crufty and currently unused cluster mechanism for VBLK files in vfs_bio_awrite. When the code is functional, I'll add back a cleaner version. 7) The page busy count wakeups assocated with the buffer cache usage were incorrectly cleaned up in a previous commit by me. Revert to the original, correct version, but with a cleaner implementation. 8) The cluster read code now tries to keep data associated with buffers more aggressively (without breaking the heuristics) when it is presumed that the read data (buffers) will be soon needed. 9) Change to filesystem lockmgr locks so that they use LK_NOPAUSE. The delay loop waiting is not useful for filesystem locks, due to the length of the time intervals. 10) Correct and clean-up spec_getpages. 11) Implement a fully functional nfs_getpages, nfs_putpages. 12) Fix nfs_write so that modifications are coherent with the NFS data on the server disk (at least as well as NFS seems to allow.) 13) Properly support MS_INVALIDATE on NFS. 14) Properly pass down MS_INVALIDATE to lower levels of the VM code from vm_map_clean. 15) Better support the notion of pages being busy but valid, so that fewer in-transit waits occur. (use p->busy more for pageouts instead of PG_BUSY.) Since the page is fully valid, it is still usable for reads. 16) It is possible (in error) for cached pages to be busy. Make the page allocation code handle that case correctly. (It should probably be a printf or panic, but I want the system to handle coding errors robustly. I'll probably add a printf.) 17) Correct the design and usage of vm_page_sleep. It didn't handle consistancy problems very well, so make the design a little less lofty. After vm_page_sleep, if it ever blocked, it is still important to relookup the page (if the object generation count changed), and verify it's status (always.) 18) In vm_pageout.c, vm_pageout_clean had rotted, so clean that up. 19) Push the page busy for writes and VM_PROT_READ into vm_pageout_flush. 20) Fix vm_pager_put_pages and it's descendents to support an int flag instead of a boolean, so that we can pass down the invalidate bit.
1998-03-07 21:37:31 +00:00
}
1994-05-24 10:09:53 +00:00
/*
* Lock out the creation of new entries in the FFS hash table in
* case getnewvnode() or MALLOC() blocks, otherwise a duplicate
* may occur!
*/
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(&ffs_inode_hash_mtx);
if (ffs_inode_hash_lock) {
while (ffs_inode_hash_lock) {
ffs_inode_hash_lock = -1;
msleep(&ffs_inode_hash_lock, &ffs_inode_hash_mtx, PVM, "ffsvgt", 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(&ffs_inode_hash_mtx);
goto restart;
}
ffs_inode_hash_lock = 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(&ffs_inode_hash_mtx);
/*
* If this MALLOC() is performed after the getnewvnode()
* it might block, leaving a vnode with a NULL v_data to be
* found by ffs_sync() if a sync happens to fire right then,
* which will cause a panic because ffs_sync() blindly
* dereferences vp->v_data (as well it should).
*/
MALLOC(ip, struct inode *, sizeof(struct inode),
ump->um_malloctype, M_WAITOK);
1994-05-24 10:09:53 +00:00
/* Allocate a new vnode/inode. */
error = getnewvnode(VT_UFS, mp, ffs_vnodeop_p, &vp);
if (error) {
/*
* Do not wake up processes while holding the mutex,
* otherwise the processes waken up immediately hit
* themselves into the mutex.
*/
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(&ffs_inode_hash_mtx);
want_wakeup = ffs_inode_hash_lock < 0;
ffs_inode_hash_lock = 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(&ffs_inode_hash_mtx);
if (want_wakeup)
wakeup(&ffs_inode_hash_lock);
1994-05-24 10:09:53 +00:00
*vpp = NULL;
FREE(ip, ump->um_malloctype);
1994-05-24 10:09:53 +00:00
return (error);
}
bzero((caddr_t)ip, sizeof(struct inode));
/*
* FFS supports lock sharing in the stack of vnodes
*/
vp->v_vnlock = &vp->v_lock;
lockinit(vp->v_vnlock, PINOD, "inode", VLKTIMEOUT, LK_CANRECURSE);
1994-05-24 10:09:53 +00:00
vp->v_data = ip;
ip->i_vnode = vp;
ip->i_fs = fs = ump->um_fs;
ip->i_dev = dev;
ip->i_number = ino;
#ifdef QUOTA
{
int i;
for (i = 0; i < MAXQUOTAS; i++)
ip->i_dquot[i] = NODQUOT;
}
1994-05-24 10:09:53 +00:00
#endif
/*
* Put it onto its hash chain and lock it so that other requests for
* this inode will block if they arrive while we are sleeping waiting
* for old data structures to be purged or for the contents of the
* disk portion of this inode to be read.
*/
ufs_ihashins(ip);
/*
* Do not wake up processes while holding the mutex,
* otherwise the processes waken up immediately hit
* themselves into the mutex.
*/
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(&ffs_inode_hash_mtx);
want_wakeup = ffs_inode_hash_lock < 0;
ffs_inode_hash_lock = 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(&ffs_inode_hash_mtx);
if (want_wakeup)
wakeup(&ffs_inode_hash_lock);
1994-05-24 10:09:53 +00:00
/* Read in the disk contents for the inode, copy into the inode. */
error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
(int)fs->fs_bsize, NOCRED, &bp);
if (error) {
1994-05-24 10:09:53 +00:00
/*
* The inode does not contain anything useful, so it would
* be misleading to leave it on its hash chain. With mode
* still zero, it will be unlinked and returned to the free
* list by vput().
*/
brelse(bp);
vput(vp);
1994-05-24 10:09:53 +00:00
*vpp = NULL;
return (error);
}
ip->i_din = *((struct dinode *)bp->b_data + ino_to_fsbo(fs, ino));
if (DOINGSOFTDEP(vp))
softdep_load_inodeblock(ip);
else
ip->i_effnlink = ip->i_nlink;
bqrelse(bp);
1994-05-24 10:09:53 +00:00
/*
* Initialize the vnode from the inode, check for aliases.
* Note that the underlying vnode may have changed.
*/
error = ufs_vinit(mp, ffs_specop_p, ffs_fifoop_p, &vp);
if (error) {
1994-05-24 10:09:53 +00:00
vput(vp);
*vpp = NULL;
return (error);
}
/*
* Finish inode initialization now that aliasing has been resolved.
*/
ip->i_devvp = ump->um_devvp;
VREF(ip->i_devvp);
/*
* Set up a generation number for this inode if it does not
* already have one. This should only happen on old filesystems.
*/
if (ip->i_gen == 0) {
ip->i_gen = random() / 2 + 1;
1994-05-24 10:09:53 +00:00
if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
ip->i_flag |= IN_MODIFIED;
}
/*
* Ensure that uid and gid are correct. This is a temporary
* fix until fsck has been changed to do the update.
*/
if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
ip->i_uid = ip->i_din.di_ouid; /* XXX */
ip->i_gid = ip->i_din.di_ogid; /* XXX */
} /* XXX */
*vpp = vp;
return (0);
}
/*
* File handle to vnode
*
* Have to be really careful about stale file handles:
* - check that the inode number is valid
* - call ffs_vget() to get the locked inode
* - check for an unallocated inode (i_mode == 0)
* - check that the given client host has export rights and return
* those rights via. exflagsp and credanonp
*/
int
ffs_fhtovp(mp, fhp, vpp)
1994-05-24 10:09:53 +00:00
register struct mount *mp;
struct fid *fhp;
struct vnode **vpp;
{
register struct ufid *ufhp;
struct fs *fs;
ufhp = (struct ufid *)fhp;
fs = VFSTOUFS(mp)->um_fs;
if (ufhp->ufid_ino < ROOTINO ||
ufhp->ufid_ino >= fs->fs_ncg * fs->fs_ipg)
return (ESTALE);
return (ufs_fhtovp(mp, ufhp, vpp));
1994-05-24 10:09:53 +00:00
}
/*
* Vnode pointer to File handle
*/
/* ARGSUSED */
int
1994-05-24 10:09:53 +00:00
ffs_vptofh(vp, fhp)
struct vnode *vp;
struct fid *fhp;
{
register struct inode *ip;
register struct ufid *ufhp;
ip = VTOI(vp);
ufhp = (struct ufid *)fhp;
ufhp->ufid_len = sizeof(struct ufid);
ufhp->ufid_ino = ip->i_number;
ufhp->ufid_gen = ip->i_gen;
return (0);
}
/*
* Initialize the filesystem; just use ufs_init.
*/
static int
ffs_init(vfsp)
struct vfsconf *vfsp;
{
softdep_initialize();
mtx_init(&ffs_inode_hash_mtx, "ifsvgt", MTX_DEF);
return (ufs_init(vfsp));
}
1994-05-24 10:09:53 +00:00
/*
* Write a superblock and associated information back to disk.
*/
1995-12-17 21:14:36 +00:00
static int
1994-05-24 10:09:53 +00:00
ffs_sbupdate(mp, waitfor)
struct ufsmount *mp;
int waitfor;
{
register struct fs *dfs, *fs = mp->um_fs;
1994-05-24 10:09:53 +00:00
register struct buf *bp;
int blks;
void *space;
int i, size, error, allerror = 0;
1994-05-24 10:09:53 +00:00
/*
* First write back the summary information.
*/
1994-05-24 10:09:53 +00:00
blks = howmany(fs->fs_cssize, fs->fs_fsize);
space = fs->fs_csp;
1994-05-24 10:09:53 +00:00
for (i = 0; i < blks; i += fs->fs_frag) {
size = fs->fs_bsize;
if (i + fs->fs_frag > blks)
size = (blks - i) * fs->fs_fsize;
bp = getblk(mp->um_devvp, fsbtodb(fs, fs->fs_csaddr + i),
size, 0, 0);
bcopy(space, bp->b_data, (u_int)size);
space = (char *)space + size;
if (waitfor != MNT_WAIT)
1994-05-24 10:09:53 +00:00
bawrite(bp);
else if ((error = bwrite(bp)) != 0)
allerror = error;
1994-05-24 10:09:53 +00:00
}
/*
* Now write back the superblock itself. If any errors occurred
* up to this point, then fail so that the superblock avoids
* being written out as clean.
*/
if (allerror)
return (allerror);
bp = getblk(mp->um_devvp, SBLOCK, (int)fs->fs_sbsize, 0, 0);
fs->fs_fmod = 0;
fs->fs_time = time_second;
bcopy((caddr_t)fs, bp->b_data, (u_int)fs->fs_sbsize);
/* Restore compatibility to old file systems. XXX */
dfs = (struct fs *)bp->b_data; /* XXX */
if (fs->fs_postblformat == FS_42POSTBLFMT) /* XXX */
dfs->fs_nrpos = -1; /* XXX */
if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
int32_t *lp, tmp; /* XXX */
/* XXX */
lp = (int32_t *)&dfs->fs_qbmask; /* XXX */
tmp = lp[4]; /* XXX */
for (i = 4; i > 0; i--) /* XXX */
lp[i] = lp[i-1]; /* XXX */
lp[0] = tmp; /* XXX */
} /* XXX */
dfs->fs_maxfilesize = mp->um_savedmaxfilesize; /* XXX */
if (waitfor != MNT_WAIT)
bawrite(bp);
else if ((error = bwrite(bp)) != 0)
allerror = error;
return (allerror);
1994-05-24 10:09:53 +00:00
}