2000-12-03 10:34:49 +03:00
|
|
|
/* $NetBSD: lfs_syscalls.c,v 1.56 2000/12/03 07:34:49 perseant Exp $ */
|
1994-06-29 10:39:25 +04:00
|
|
|
|
1999-03-10 03:20:00 +03:00
|
|
|
/*-
|
Various bug-fixes to LFS, to wit:
Kernel:
* Add runtime quantity lfs_ravail, the number of disk-blocks reserved
for writing. Writes to the filesystem first reserve a maximum amount
of blocks before their write is allowed to proceed; after the blocks
are allocated the reserved total is reduced by a corresponding amount.
If the lfs_reserve function cannot immediately reserve the requested
number of blocks, the inode is unlocked, and the thread sleeps until
the cleaner has made enough space available for the blocks to be
reserved. In this way large files can be written to the filesystem
(or, smaller files can be written to a nearly-full but thoroughly
clean filesystem) and the cleaner can still function properly.
* Remove explicit switching on dlfs_minfreeseg from the kernel code; it
is now merely a fs-creation parameter used to compute dlfs_avail and
dlfs_bfree (and used by fsck_lfs(8) to check their accuracy). Its
former role is better assumed by a properly computed dlfs_avail.
* Bounds-check inode numbers submitted through lfs_bmapv and lfs_markv.
This prevents a panic, but, if the cleaner is feeding the filesystem
the wrong data, you are still in a world of hurt.
* Cleanup: remove explicit references of DEV_BSIZE in favor of
btodb()/dbtob().
lfs_cleanerd:
* Make -n mean "send N segments' blocks through a single call to
lfs_markv". Previously it had meant "clean N segments though N calls
to lfs_markv, before looking again to see if more need to be cleaned".
The new behavior gives better packing of direct data on disk with as
little metadata as possible, largely alleviating the problem that the
cleaner can consume more disk through inefficient use of metadata than
it frees by moving dirty data away from clean "holes" to produce
entirely clean segments.
* Make -b mean "read as many segments as necessary to write N segments
of dirty data back to disk", rather than its former meaning of "read
as many segments as necessary to free N segments worth of space". The
new meaning, combined with the new -n behavior described above,
further aids in cleaning storage efficiency as entire segments can be
written at once, using as few blocks as possible for segment summaries
and inode blocks.
* Make the cleaner take note of segments which could not be cleaned due
to error, and not attempt to clean them until they are entirely free
of dirty blocks. This prevents the case in which a cleanerd running
with -n 1 and without -b (formerly the default) would spin trying
repeatedly to clean a corrupt segment, while the remaining space
filled and deadlocked the filesystem.
* Update the lfs_cleanerd manual page to describe all the options,
including the changes mentioned here (in particular, the -b and -n
flags were previously undocumented).
fsck_lfs:
* Check, and optionally fix, lfs_avail (to an exact figure) and
lfs_bfree (within a margin of error) in pass 5.
newfs_lfs:
* Reduce the default dlfs_minfreeseg to 1/20 of the total segments.
* Add a warning if the sgs disklabel field is 16 (the default for FFS'
cpg, but not usually desirable for LFS' sgs: 5--8 is a better range).
* Change the calculation of lfs_avail and lfs_bfree, corresponding to
the kernel changes mentioned above.
mount_lfs:
* Add -N and -b options to pass corresponding -n and -b options to
lfs_cleanerd.
* Default to calling lfs_cleanerd with "-b -n 4".
[All of these changes were largely tested in the 1.5 branch, with the
idea that they (along with previous un-pulled-up work) could be applied
to the branch while it was still in ALPHA2; however my test system has
experienced corruption on another filesystem (/dev/console has gone
missing :^), and, while I believe this unrelated to the LFS changes, I
cannot with good conscience request that the changes be pulled up.]
2000-09-09 08:49:54 +04:00
|
|
|
* Copyright (c) 1999, 2000 The NetBSD Foundation, Inc.
|
1999-03-10 03:20:00 +03:00
|
|
|
* All rights reserved.
|
|
|
|
|
*
|
|
|
|
|
* This code is derived from software contributed to The NetBSD Foundation
|
|
|
|
|
* by Konrad E. Schroder <perseant@hhhh.org>.
|
|
|
|
|
*
|
|
|
|
|
* 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 NetBSD
|
|
|
|
|
* Foundation, Inc. and its contributors.
|
|
|
|
|
* 4. Neither the name of The NetBSD Foundation 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 NETBSD FOUNDATION, INC. 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 FOUNDATION 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.
|
|
|
|
|
*/
|
1994-06-08 15:41:58 +04:00
|
|
|
/*-
|
|
|
|
|
* Copyright (c) 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.
|
|
|
|
|
*
|
1998-03-01 05:20:01 +03:00
|
|
|
* @(#)lfs_syscalls.c 8.10 (Berkeley) 5/14/95
|
1994-06-08 15:41:58 +04:00
|
|
|
*/
|
|
|
|
|
|
2000-11-30 18:57:35 +03:00
|
|
|
#define LFS /* for prototypes in syscallargs.h */
|
1998-02-19 03:54:39 +03:00
|
|
|
|
1994-06-08 15:41:58 +04:00
|
|
|
#include <sys/param.h>
|
1994-10-20 07:20:55 +03:00
|
|
|
#include <sys/systm.h>
|
1994-06-08 15:41:58 +04:00
|
|
|
#include <sys/proc.h>
|
|
|
|
|
#include <sys/buf.h>
|
|
|
|
|
#include <sys/mount.h>
|
|
|
|
|
#include <sys/vnode.h>
|
|
|
|
|
#include <sys/malloc.h>
|
|
|
|
|
#include <sys/kernel.h>
|
|
|
|
|
|
1994-10-20 07:20:55 +03:00
|
|
|
#include <sys/syscallargs.h>
|
|
|
|
|
|
1994-06-08 15:41:58 +04:00
|
|
|
#include <ufs/ufs/quota.h>
|
|
|
|
|
#include <ufs/ufs/inode.h>
|
|
|
|
|
#include <ufs/ufs/ufsmount.h>
|
|
|
|
|
#include <ufs/ufs/ufs_extern.h>
|
|
|
|
|
|
|
|
|
|
#include <ufs/lfs/lfs.h>
|
|
|
|
|
#include <ufs/lfs/lfs_extern.h>
|
1996-02-10 01:28:45 +03:00
|
|
|
|
1999-03-10 03:20:00 +03:00
|
|
|
/* Flags for return from lfs_fastvget */
|
|
|
|
|
#define FVG_UNLOCK 0x01 /* Needs to be unlocked */
|
|
|
|
|
#define FVG_PUT 0x02 /* Needs to be vput() */
|
1994-06-08 15:41:58 +04:00
|
|
|
|
|
|
|
|
struct buf *lfs_fakebuf __P((struct vnode *, int, size_t, caddr_t));
|
2000-07-01 00:45:38 +04:00
|
|
|
int lfs_fasthashget __P((dev_t, ino_t, int *, struct vnode **));
|
1994-06-08 15:41:58 +04:00
|
|
|
|
1998-03-01 05:20:01 +03:00
|
|
|
int debug_cleaner = 0;
|
|
|
|
|
int clean_vnlocked = 0;
|
|
|
|
|
int clean_inlocked = 0;
|
1999-03-10 03:20:00 +03:00
|
|
|
int verbose_debug = 0;
|
|
|
|
|
|
|
|
|
|
pid_t lfs_cleaner_pid = 0;
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Definitions for the buffer free lists.
|
|
|
|
|
*/
|
|
|
|
|
#define BQUEUES 4 /* number of free buffer queues */
|
|
|
|
|
|
|
|
|
|
#define BQ_LOCKED 0 /* super-blocks &c */
|
|
|
|
|
#define BQ_LRU 1 /* lru, useful buffers */
|
|
|
|
|
#define BQ_AGE 2 /* rubbish */
|
|
|
|
|
#define BQ_EMPTY 3 /* buffer headers with no memory */
|
|
|
|
|
|
|
|
|
|
extern TAILQ_HEAD(bqueues, buf) bufqueues[BQUEUES];
|
|
|
|
|
|
|
|
|
|
#define LFS_FORCE_WRITE UNASSIGNED
|
|
|
|
|
|
|
|
|
|
#define LFS_VREF_THRESHOLD 128
|
1998-03-01 05:20:01 +03:00
|
|
|
|
1994-06-08 15:41:58 +04:00
|
|
|
/*
|
1999-06-09 08:52:11 +04:00
|
|
|
* sys_lfs_markv:
|
1994-06-08 15:41:58 +04:00
|
|
|
*
|
|
|
|
|
* This will mark inodes and blocks dirty, so they are written into the log.
|
|
|
|
|
* It will block until all the blocks have been written. The segment create
|
|
|
|
|
* time passed in the block_info and inode_info structures is used to decide
|
|
|
|
|
* if the data is valid for each block (in case some process dirtied a block
|
|
|
|
|
* or inode that is being cleaned between the determination that a block is
|
|
|
|
|
* live and the lfs_markv call).
|
|
|
|
|
*
|
|
|
|
|
* 0 on success
|
|
|
|
|
* -1/errno is return on error.
|
|
|
|
|
*/
|
|
|
|
|
int
|
1999-06-09 08:52:11 +04:00
|
|
|
sys_lfs_markv(p, v, retval)
|
1999-03-10 03:20:00 +03:00
|
|
|
struct proc *p;
|
|
|
|
|
void *v;
|
|
|
|
|
register_t *retval;
|
1995-09-22 03:39:20 +04:00
|
|
|
{
|
1999-06-09 16:18:19 +04:00
|
|
|
struct sys_lfs_markv_args /* {
|
1994-10-20 07:20:55 +03:00
|
|
|
syscallarg(fsid_t *) fsidp;
|
|
|
|
|
syscallarg(struct block_info *) blkiov;
|
|
|
|
|
syscallarg(int) blkcnt;
|
1999-06-09 16:18:19 +04:00
|
|
|
} */ *uap = v;
|
1994-06-08 15:41:58 +04:00
|
|
|
BLOCK_INFO *blkp;
|
|
|
|
|
IFILE *ifp;
|
1999-03-10 03:20:00 +03:00
|
|
|
struct buf *bp, *nbp;
|
1996-02-10 01:28:45 +03:00
|
|
|
struct inode *ip = NULL;
|
1994-06-08 15:41:58 +04:00
|
|
|
struct lfs *fs;
|
|
|
|
|
struct mount *mntp;
|
|
|
|
|
struct vnode *vp;
|
1999-03-10 03:20:00 +03:00
|
|
|
#ifdef DEBUG_LFS
|
|
|
|
|
int vputc=0, iwritten=0;
|
|
|
|
|
#endif
|
1994-06-08 15:41:58 +04:00
|
|
|
fsid_t fsid;
|
|
|
|
|
void *start;
|
|
|
|
|
ino_t lastino;
|
1998-03-01 05:20:01 +03:00
|
|
|
ufs_daddr_t b_daddr, v_daddr;
|
1999-03-10 03:20:00 +03:00
|
|
|
int origcnt, cnt, error, lfs_fastvget_unlock;
|
|
|
|
|
int do_again=0;
|
|
|
|
|
int s;
|
|
|
|
|
#ifdef CHECK_COPYIN
|
|
|
|
|
int i;
|
|
|
|
|
#endif /* CHECK_COPYIN */
|
|
|
|
|
#ifdef LFS_TRACK_IOS
|
|
|
|
|
int j;
|
|
|
|
|
#endif
|
|
|
|
|
int numlocked=0, numrefed=0;
|
Various bug-fixes to LFS, to wit:
Kernel:
* Add runtime quantity lfs_ravail, the number of disk-blocks reserved
for writing. Writes to the filesystem first reserve a maximum amount
of blocks before their write is allowed to proceed; after the blocks
are allocated the reserved total is reduced by a corresponding amount.
If the lfs_reserve function cannot immediately reserve the requested
number of blocks, the inode is unlocked, and the thread sleeps until
the cleaner has made enough space available for the blocks to be
reserved. In this way large files can be written to the filesystem
(or, smaller files can be written to a nearly-full but thoroughly
clean filesystem) and the cleaner can still function properly.
* Remove explicit switching on dlfs_minfreeseg from the kernel code; it
is now merely a fs-creation parameter used to compute dlfs_avail and
dlfs_bfree (and used by fsck_lfs(8) to check their accuracy). Its
former role is better assumed by a properly computed dlfs_avail.
* Bounds-check inode numbers submitted through lfs_bmapv and lfs_markv.
This prevents a panic, but, if the cleaner is feeding the filesystem
the wrong data, you are still in a world of hurt.
* Cleanup: remove explicit references of DEV_BSIZE in favor of
btodb()/dbtob().
lfs_cleanerd:
* Make -n mean "send N segments' blocks through a single call to
lfs_markv". Previously it had meant "clean N segments though N calls
to lfs_markv, before looking again to see if more need to be cleaned".
The new behavior gives better packing of direct data on disk with as
little metadata as possible, largely alleviating the problem that the
cleaner can consume more disk through inefficient use of metadata than
it frees by moving dirty data away from clean "holes" to produce
entirely clean segments.
* Make -b mean "read as many segments as necessary to write N segments
of dirty data back to disk", rather than its former meaning of "read
as many segments as necessary to free N segments worth of space". The
new meaning, combined with the new -n behavior described above,
further aids in cleaning storage efficiency as entire segments can be
written at once, using as few blocks as possible for segment summaries
and inode blocks.
* Make the cleaner take note of segments which could not be cleaned due
to error, and not attempt to clean them until they are entirely free
of dirty blocks. This prevents the case in which a cleanerd running
with -n 1 and without -b (formerly the default) would spin trying
repeatedly to clean a corrupt segment, while the remaining space
filled and deadlocked the filesystem.
* Update the lfs_cleanerd manual page to describe all the options,
including the changes mentioned here (in particular, the -b and -n
flags were previously undocumented).
fsck_lfs:
* Check, and optionally fix, lfs_avail (to an exact figure) and
lfs_bfree (within a margin of error) in pass 5.
newfs_lfs:
* Reduce the default dlfs_minfreeseg to 1/20 of the total segments.
* Add a warning if the sgs disklabel field is 16 (the default for FFS'
cpg, but not usually desirable for LFS' sgs: 5--8 is a better range).
* Change the calculation of lfs_avail and lfs_bfree, corresponding to
the kernel changes mentioned above.
mount_lfs:
* Add -N and -b options to pass corresponding -n and -b options to
lfs_cleanerd.
* Default to calling lfs_cleanerd with "-b -n 4".
[All of these changes were largely tested in the 1.5 branch, with the
idea that they (along with previous un-pulled-up work) could be applied
to the branch while it was still in ALPHA2; however my test system has
experienced corruption on another filesystem (/dev/console has gone
missing :^), and, while I believe this unrelated to the LFS changes, I
cannot with good conscience request that the changes be pulled up.]
2000-09-09 08:49:54 +04:00
|
|
|
ino_t maxino;
|
1994-06-08 15:41:58 +04:00
|
|
|
|
1996-02-10 01:28:45 +03:00
|
|
|
if ((error = copyin(SCARG(uap, fsidp), &fsid, sizeof(fsid_t))) != 0)
|
1994-06-08 15:41:58 +04:00
|
|
|
return (error);
|
1999-03-10 03:20:00 +03:00
|
|
|
|
1998-03-01 05:20:01 +03:00
|
|
|
if ((mntp = vfs_getvfs(&fsid)) == NULL)
|
2000-11-23 01:11:34 +03:00
|
|
|
return (ENOENT);
|
1994-06-08 15:41:58 +04:00
|
|
|
|
1999-03-10 03:20:00 +03:00
|
|
|
fs = VFSTOUFS(mntp)->um_lfs;
|
|
|
|
|
|
|
|
|
|
if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
|
|
|
|
|
return (error);
|
Various bug-fixes to LFS, to wit:
Kernel:
* Add runtime quantity lfs_ravail, the number of disk-blocks reserved
for writing. Writes to the filesystem first reserve a maximum amount
of blocks before their write is allowed to proceed; after the blocks
are allocated the reserved total is reduced by a corresponding amount.
If the lfs_reserve function cannot immediately reserve the requested
number of blocks, the inode is unlocked, and the thread sleeps until
the cleaner has made enough space available for the blocks to be
reserved. In this way large files can be written to the filesystem
(or, smaller files can be written to a nearly-full but thoroughly
clean filesystem) and the cleaner can still function properly.
* Remove explicit switching on dlfs_minfreeseg from the kernel code; it
is now merely a fs-creation parameter used to compute dlfs_avail and
dlfs_bfree (and used by fsck_lfs(8) to check their accuracy). Its
former role is better assumed by a properly computed dlfs_avail.
* Bounds-check inode numbers submitted through lfs_bmapv and lfs_markv.
This prevents a panic, but, if the cleaner is feeding the filesystem
the wrong data, you are still in a world of hurt.
* Cleanup: remove explicit references of DEV_BSIZE in favor of
btodb()/dbtob().
lfs_cleanerd:
* Make -n mean "send N segments' blocks through a single call to
lfs_markv". Previously it had meant "clean N segments though N calls
to lfs_markv, before looking again to see if more need to be cleaned".
The new behavior gives better packing of direct data on disk with as
little metadata as possible, largely alleviating the problem that the
cleaner can consume more disk through inefficient use of metadata than
it frees by moving dirty data away from clean "holes" to produce
entirely clean segments.
* Make -b mean "read as many segments as necessary to write N segments
of dirty data back to disk", rather than its former meaning of "read
as many segments as necessary to free N segments worth of space". The
new meaning, combined with the new -n behavior described above,
further aids in cleaning storage efficiency as entire segments can be
written at once, using as few blocks as possible for segment summaries
and inode blocks.
* Make the cleaner take note of segments which could not be cleaned due
to error, and not attempt to clean them until they are entirely free
of dirty blocks. This prevents the case in which a cleanerd running
with -n 1 and without -b (formerly the default) would spin trying
repeatedly to clean a corrupt segment, while the remaining space
filled and deadlocked the filesystem.
* Update the lfs_cleanerd manual page to describe all the options,
including the changes mentioned here (in particular, the -b and -n
flags were previously undocumented).
fsck_lfs:
* Check, and optionally fix, lfs_avail (to an exact figure) and
lfs_bfree (within a margin of error) in pass 5.
newfs_lfs:
* Reduce the default dlfs_minfreeseg to 1/20 of the total segments.
* Add a warning if the sgs disklabel field is 16 (the default for FFS'
cpg, but not usually desirable for LFS' sgs: 5--8 is a better range).
* Change the calculation of lfs_avail and lfs_bfree, corresponding to
the kernel changes mentioned above.
mount_lfs:
* Add -N and -b options to pass corresponding -n and -b options to
lfs_cleanerd.
* Default to calling lfs_cleanerd with "-b -n 4".
[All of these changes were largely tested in the 1.5 branch, with the
idea that they (along with previous un-pulled-up work) could be applied
to the branch while it was still in ALPHA2; however my test system has
experienced corruption on another filesystem (/dev/console has gone
missing :^), and, while I believe this unrelated to the LFS changes, I
cannot with good conscience request that the changes be pulled up.]
2000-09-09 08:49:54 +04:00
|
|
|
|
2000-11-23 01:11:34 +03:00
|
|
|
|
Various bug-fixes to LFS, to wit:
Kernel:
* Add runtime quantity lfs_ravail, the number of disk-blocks reserved
for writing. Writes to the filesystem first reserve a maximum amount
of blocks before their write is allowed to proceed; after the blocks
are allocated the reserved total is reduced by a corresponding amount.
If the lfs_reserve function cannot immediately reserve the requested
number of blocks, the inode is unlocked, and the thread sleeps until
the cleaner has made enough space available for the blocks to be
reserved. In this way large files can be written to the filesystem
(or, smaller files can be written to a nearly-full but thoroughly
clean filesystem) and the cleaner can still function properly.
* Remove explicit switching on dlfs_minfreeseg from the kernel code; it
is now merely a fs-creation parameter used to compute dlfs_avail and
dlfs_bfree (and used by fsck_lfs(8) to check their accuracy). Its
former role is better assumed by a properly computed dlfs_avail.
* Bounds-check inode numbers submitted through lfs_bmapv and lfs_markv.
This prevents a panic, but, if the cleaner is feeding the filesystem
the wrong data, you are still in a world of hurt.
* Cleanup: remove explicit references of DEV_BSIZE in favor of
btodb()/dbtob().
lfs_cleanerd:
* Make -n mean "send N segments' blocks through a single call to
lfs_markv". Previously it had meant "clean N segments though N calls
to lfs_markv, before looking again to see if more need to be cleaned".
The new behavior gives better packing of direct data on disk with as
little metadata as possible, largely alleviating the problem that the
cleaner can consume more disk through inefficient use of metadata than
it frees by moving dirty data away from clean "holes" to produce
entirely clean segments.
* Make -b mean "read as many segments as necessary to write N segments
of dirty data back to disk", rather than its former meaning of "read
as many segments as necessary to free N segments worth of space". The
new meaning, combined with the new -n behavior described above,
further aids in cleaning storage efficiency as entire segments can be
written at once, using as few blocks as possible for segment summaries
and inode blocks.
* Make the cleaner take note of segments which could not be cleaned due
to error, and not attempt to clean them until they are entirely free
of dirty blocks. This prevents the case in which a cleanerd running
with -n 1 and without -b (formerly the default) would spin trying
repeatedly to clean a corrupt segment, while the remaining space
filled and deadlocked the filesystem.
* Update the lfs_cleanerd manual page to describe all the options,
including the changes mentioned here (in particular, the -b and -n
flags were previously undocumented).
fsck_lfs:
* Check, and optionally fix, lfs_avail (to an exact figure) and
lfs_bfree (within a margin of error) in pass 5.
newfs_lfs:
* Reduce the default dlfs_minfreeseg to 1/20 of the total segments.
* Add a warning if the sgs disklabel field is 16 (the default for FFS'
cpg, but not usually desirable for LFS' sgs: 5--8 is a better range).
* Change the calculation of lfs_avail and lfs_bfree, corresponding to
the kernel changes mentioned above.
mount_lfs:
* Add -N and -b options to pass corresponding -n and -b options to
lfs_cleanerd.
* Default to calling lfs_cleanerd with "-b -n 4".
[All of these changes were largely tested in the 1.5 branch, with the
idea that they (along with previous un-pulled-up work) could be applied
to the branch while it was still in ALPHA2; however my test system has
experienced corruption on another filesystem (/dev/console has gone
missing :^), and, while I believe this unrelated to the LFS changes, I
cannot with good conscience request that the changes be pulled up.]
2000-09-09 08:49:54 +04:00
|
|
|
maxino = (dbtofsb(fs, VTOI(fs->lfs_ivnode)->i_ffs_blocks) -
|
|
|
|
|
fs->lfs_cleansz - fs->lfs_segtabsz) * fs->lfs_ifpb;
|
|
|
|
|
|
1999-03-10 03:20:00 +03:00
|
|
|
origcnt = cnt = SCARG(uap, blkcnt);
|
1994-06-08 15:41:58 +04:00
|
|
|
start = malloc(cnt * sizeof(BLOCK_INFO), M_SEGMENT, M_WAITOK);
|
1996-02-10 01:28:45 +03:00
|
|
|
error = copyin(SCARG(uap, blkiov), start, cnt * sizeof(BLOCK_INFO));
|
|
|
|
|
if (error)
|
1994-06-08 15:41:58 +04:00
|
|
|
goto err1;
|
1999-03-10 03:20:00 +03:00
|
|
|
|
2000-11-23 01:11:34 +03:00
|
|
|
if ((error = vfs_busy(mntp, LK_NOWAIT, NULL)) != 0)
|
|
|
|
|
return (error);
|
|
|
|
|
|
1999-03-10 03:20:00 +03:00
|
|
|
/*
|
|
|
|
|
* This seglock is just to prevent the fact that we might have to sleep
|
|
|
|
|
* from allowing the possibility that our blocks might become
|
|
|
|
|
* invalid.
|
|
|
|
|
*
|
|
|
|
|
* It is also important to note here that unless we specify SEGM_CKP,
|
|
|
|
|
* any Ifile blocks that we might be asked to clean will never get
|
|
|
|
|
* to the disk.
|
|
|
|
|
*/
|
|
|
|
|
lfs_seglock(fs, SEGM_SYNC|SEGM_CLEAN|SEGM_CKP);
|
|
|
|
|
|
1994-06-08 15:41:58 +04:00
|
|
|
/* Mark blocks/inodes dirty. */
|
|
|
|
|
error = 0;
|
|
|
|
|
|
1999-03-10 03:20:00 +03:00
|
|
|
#ifdef DEBUG_LFS
|
|
|
|
|
/* Run through and count the inodes */
|
|
|
|
|
lastino = LFS_UNUSED_INUM;
|
|
|
|
|
for(blkp = start; cnt--; ++blkp) {
|
|
|
|
|
if(lastino != blkp->bi_inode) {
|
|
|
|
|
lastino = blkp->bi_inode;
|
|
|
|
|
vputc++;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
cnt = origcnt;
|
|
|
|
|
printf("[%d/",vputc);
|
|
|
|
|
iwritten=0;
|
|
|
|
|
#endif /* DEBUG_LFS */
|
|
|
|
|
/* these were inside the initialization for the for loop */
|
|
|
|
|
v_daddr = LFS_UNUSED_DADDR;
|
|
|
|
|
lastino = LFS_UNUSED_INUM;
|
|
|
|
|
for (blkp = start; cnt--; ++blkp)
|
|
|
|
|
{
|
1999-04-12 04:40:06 +04:00
|
|
|
if(blkp->bi_daddr == LFS_FORCE_WRITE)
|
|
|
|
|
printf("lfs_markv: warning: force-writing ino %d lbn %d\n",
|
|
|
|
|
blkp->bi_inode, blkp->bi_lbn);
|
1999-03-10 03:20:00 +03:00
|
|
|
#ifdef LFS_TRACK_IOS
|
|
|
|
|
/*
|
|
|
|
|
* If there is I/O on this segment that is not yet complete,
|
|
|
|
|
* the cleaner probably does not have the right information.
|
|
|
|
|
* Send it packing.
|
|
|
|
|
*/
|
|
|
|
|
for(j=0;j<LFS_THROTTLE;j++) {
|
|
|
|
|
if(fs->lfs_pending[j] != LFS_UNUSED_DADDR
|
|
|
|
|
&& datosn(fs,fs->lfs_pending[j])==datosn(fs,blkp->bi_daddr)
|
|
|
|
|
&& blkp->bi_daddr != LFS_FORCE_WRITE)
|
|
|
|
|
{
|
|
|
|
|
printf("lfs_markv: attempt to clean pending segment? (#%d)\n",
|
|
|
|
|
datosn(fs, fs->lfs_pending[j]));
|
|
|
|
|
/* free(start,M_SEGMENT); */
|
|
|
|
|
/* return (EBUSY); */
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
#endif /* LFS_TRACK_IOS */
|
Various bug-fixes to LFS, to wit:
Kernel:
* Add runtime quantity lfs_ravail, the number of disk-blocks reserved
for writing. Writes to the filesystem first reserve a maximum amount
of blocks before their write is allowed to proceed; after the blocks
are allocated the reserved total is reduced by a corresponding amount.
If the lfs_reserve function cannot immediately reserve the requested
number of blocks, the inode is unlocked, and the thread sleeps until
the cleaner has made enough space available for the blocks to be
reserved. In this way large files can be written to the filesystem
(or, smaller files can be written to a nearly-full but thoroughly
clean filesystem) and the cleaner can still function properly.
* Remove explicit switching on dlfs_minfreeseg from the kernel code; it
is now merely a fs-creation parameter used to compute dlfs_avail and
dlfs_bfree (and used by fsck_lfs(8) to check their accuracy). Its
former role is better assumed by a properly computed dlfs_avail.
* Bounds-check inode numbers submitted through lfs_bmapv and lfs_markv.
This prevents a panic, but, if the cleaner is feeding the filesystem
the wrong data, you are still in a world of hurt.
* Cleanup: remove explicit references of DEV_BSIZE in favor of
btodb()/dbtob().
lfs_cleanerd:
* Make -n mean "send N segments' blocks through a single call to
lfs_markv". Previously it had meant "clean N segments though N calls
to lfs_markv, before looking again to see if more need to be cleaned".
The new behavior gives better packing of direct data on disk with as
little metadata as possible, largely alleviating the problem that the
cleaner can consume more disk through inefficient use of metadata than
it frees by moving dirty data away from clean "holes" to produce
entirely clean segments.
* Make -b mean "read as many segments as necessary to write N segments
of dirty data back to disk", rather than its former meaning of "read
as many segments as necessary to free N segments worth of space". The
new meaning, combined with the new -n behavior described above,
further aids in cleaning storage efficiency as entire segments can be
written at once, using as few blocks as possible for segment summaries
and inode blocks.
* Make the cleaner take note of segments which could not be cleaned due
to error, and not attempt to clean them until they are entirely free
of dirty blocks. This prevents the case in which a cleanerd running
with -n 1 and without -b (formerly the default) would spin trying
repeatedly to clean a corrupt segment, while the remaining space
filled and deadlocked the filesystem.
* Update the lfs_cleanerd manual page to describe all the options,
including the changes mentioned here (in particular, the -b and -n
flags were previously undocumented).
fsck_lfs:
* Check, and optionally fix, lfs_avail (to an exact figure) and
lfs_bfree (within a margin of error) in pass 5.
newfs_lfs:
* Reduce the default dlfs_minfreeseg to 1/20 of the total segments.
* Add a warning if the sgs disklabel field is 16 (the default for FFS'
cpg, but not usually desirable for LFS' sgs: 5--8 is a better range).
* Change the calculation of lfs_avail and lfs_bfree, corresponding to
the kernel changes mentioned above.
mount_lfs:
* Add -N and -b options to pass corresponding -n and -b options to
lfs_cleanerd.
* Default to calling lfs_cleanerd with "-b -n 4".
[All of these changes were largely tested in the 1.5 branch, with the
idea that they (along with previous un-pulled-up work) could be applied
to the branch while it was still in ALPHA2; however my test system has
experienced corruption on another filesystem (/dev/console has gone
missing :^), and, while I believe this unrelated to the LFS changes, I
cannot with good conscience request that the changes be pulled up.]
2000-09-09 08:49:54 +04:00
|
|
|
/* Bounds-check incoming data, avoid panic for failed VGET */
|
|
|
|
|
if (blkp->bi_inode <= 0 || blkp->bi_inode >= maxino) {
|
|
|
|
|
error = EINVAL;
|
|
|
|
|
goto again;
|
|
|
|
|
}
|
1994-06-08 15:41:58 +04:00
|
|
|
/*
|
|
|
|
|
* Get the IFILE entry (only once) and see if the file still
|
|
|
|
|
* exists.
|
|
|
|
|
*/
|
|
|
|
|
if (lastino != blkp->bi_inode) {
|
1999-03-10 03:20:00 +03:00
|
|
|
/*
|
|
|
|
|
* Finish the old file, if there was one. The presence
|
|
|
|
|
* of a usable vnode in vp is signaled by a valid v_daddr.
|
|
|
|
|
*/
|
1999-11-12 19:56:48 +03:00
|
|
|
if(v_daddr != LFS_UNUSED_DADDR) {
|
1999-03-10 03:20:00 +03:00
|
|
|
#ifdef DEBUG_LFS
|
1999-04-14 23:37:28 +04:00
|
|
|
if(ip->i_flag & (IN_MODIFIED|IN_CLEANING))
|
1999-03-10 03:20:00 +03:00
|
|
|
iwritten++;
|
|
|
|
|
#endif
|
|
|
|
|
if(lfs_fastvget_unlock) {
|
2000-06-28 00:57:11 +04:00
|
|
|
VOP_UNLOCK(vp, 0);
|
1999-03-10 03:20:00 +03:00
|
|
|
numlocked--;
|
1994-06-08 15:41:58 +04:00
|
|
|
}
|
|
|
|
|
lfs_vunref(vp);
|
1999-03-10 03:20:00 +03:00
|
|
|
numrefed--;
|
1994-06-08 15:41:58 +04:00
|
|
|
}
|
|
|
|
|
|
1999-03-10 03:20:00 +03:00
|
|
|
/*
|
|
|
|
|
* Start a new file
|
|
|
|
|
*/
|
1994-06-08 15:41:58 +04:00
|
|
|
lastino = blkp->bi_inode;
|
|
|
|
|
if (blkp->bi_inode == LFS_IFILE_INUM)
|
|
|
|
|
v_daddr = fs->lfs_idaddr;
|
|
|
|
|
else {
|
|
|
|
|
LFS_IENTRY(ifp, fs, blkp->bi_inode, bp);
|
1999-03-10 03:20:00 +03:00
|
|
|
/* XXX fix for force write */
|
1994-06-08 15:41:58 +04:00
|
|
|
v_daddr = ifp->if_daddr;
|
|
|
|
|
brelse(bp);
|
|
|
|
|
}
|
1999-03-10 03:20:00 +03:00
|
|
|
/* Don't force-write the ifile */
|
|
|
|
|
if (blkp->bi_inode == LFS_IFILE_INUM
|
|
|
|
|
&& blkp->bi_daddr == LFS_FORCE_WRITE)
|
|
|
|
|
{
|
|
|
|
|
continue;
|
|
|
|
|
}
|
1999-11-12 19:56:48 +03:00
|
|
|
if (v_daddr == LFS_UNUSED_DADDR
|
|
|
|
|
&& blkp->bi_daddr != LFS_FORCE_WRITE)
|
1999-03-10 03:20:00 +03:00
|
|
|
{
|
1994-06-08 15:41:58 +04:00
|
|
|
continue;
|
1999-03-10 03:20:00 +03:00
|
|
|
}
|
1994-06-08 15:41:58 +04:00
|
|
|
|
|
|
|
|
/* Get the vnode/inode. */
|
1999-03-10 03:20:00 +03:00
|
|
|
error=lfs_fastvget(mntp, blkp->bi_inode, v_daddr,
|
|
|
|
|
&vp,
|
|
|
|
|
(blkp->bi_lbn==LFS_UNUSED_LBN
|
|
|
|
|
? blkp->bi_bp
|
|
|
|
|
: NULL),
|
|
|
|
|
&lfs_fastvget_unlock);
|
|
|
|
|
if(lfs_fastvget_unlock)
|
|
|
|
|
numlocked++;
|
|
|
|
|
|
|
|
|
|
if(!error) {
|
|
|
|
|
numrefed++;
|
|
|
|
|
}
|
|
|
|
|
if(error) {
|
2000-06-28 00:57:11 +04:00
|
|
|
#ifdef DEBUG_LFS
|
1999-04-12 04:40:06 +04:00
|
|
|
printf("lfs_markv: lfs_fastvget failed with %d (ino %d, segment %d)\n",
|
1999-03-10 03:20:00 +03:00
|
|
|
error, blkp->bi_inode,
|
|
|
|
|
datosn(fs, blkp->bi_daddr));
|
2000-06-28 00:57:11 +04:00
|
|
|
#endif /* DEBUG_LFS */
|
1999-03-10 03:20:00 +03:00
|
|
|
/*
|
|
|
|
|
* If we got EAGAIN, that means that the
|
|
|
|
|
* Inode was locked. This is
|
|
|
|
|
* recoverable: just clean the rest of
|
|
|
|
|
* this segment, and let the cleaner try
|
|
|
|
|
* again with another. (When the
|
|
|
|
|
* cleaner runs again, this segment will
|
|
|
|
|
* sort high on the list, since it is
|
|
|
|
|
* now almost entirely empty.) But, we
|
|
|
|
|
* still set v_daddr = LFS_UNUSED_ADDR
|
|
|
|
|
* so as not to test this over and over
|
|
|
|
|
* again.
|
|
|
|
|
*/
|
|
|
|
|
if(error == EAGAIN) {
|
|
|
|
|
error = 0;
|
|
|
|
|
do_again++;
|
|
|
|
|
}
|
1994-06-08 15:41:58 +04:00
|
|
|
#ifdef DIAGNOSTIC
|
1999-03-10 03:20:00 +03:00
|
|
|
else if(error != ENOENT)
|
|
|
|
|
panic("lfs_markv VFS_VGET FAILED");
|
1994-06-08 15:41:58 +04:00
|
|
|
#endif
|
1999-03-10 03:20:00 +03:00
|
|
|
/* lastino = LFS_UNUSED_INUM; */
|
1994-06-08 15:41:58 +04:00
|
|
|
v_daddr = LFS_UNUSED_DADDR;
|
1999-03-10 03:20:00 +03:00
|
|
|
vp = NULL;
|
|
|
|
|
ip = NULL;
|
1994-06-08 15:41:58 +04:00
|
|
|
continue;
|
|
|
|
|
}
|
|
|
|
|
ip = VTOI(vp);
|
1999-03-10 03:20:00 +03:00
|
|
|
} else if (v_daddr == LFS_UNUSED_DADDR) {
|
|
|
|
|
/*
|
|
|
|
|
* This can only happen if the vnode is dead (or
|
|
|
|
|
* in any case we can't get it...e.g., it is
|
|
|
|
|
* inlocked). Keep going.
|
|
|
|
|
*/
|
1994-06-08 15:41:58 +04:00
|
|
|
continue;
|
1999-03-10 03:20:00 +03:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* Past this point we are guaranteed that vp, ip are valid. */
|
1994-06-08 15:41:58 +04:00
|
|
|
|
|
|
|
|
/* If this BLOCK_INFO didn't contain a block, keep going. */
|
1999-03-10 03:20:00 +03:00
|
|
|
if (blkp->bi_lbn == LFS_UNUSED_LBN) {
|
|
|
|
|
/* XXX need to make sure that the inode gets written in this case */
|
|
|
|
|
/* XXX but only write the inode if it's the right one */
|
2000-11-23 01:11:34 +03:00
|
|
|
if (blkp->bi_inode != LFS_IFILE_INUM) {
|
|
|
|
|
LFS_IENTRY(ifp, fs, blkp->bi_inode, bp);
|
|
|
|
|
if(ifp->if_daddr == blkp->bi_daddr
|
1999-03-10 03:20:00 +03:00
|
|
|
|| blkp->bi_daddr == LFS_FORCE_WRITE)
|
|
|
|
|
{
|
2000-07-06 02:25:43 +04:00
|
|
|
LFS_SET_UINO(ip, IN_CLEANING);
|
1999-03-10 03:20:00 +03:00
|
|
|
}
|
2000-11-23 01:11:34 +03:00
|
|
|
brelse(bp);
|
|
|
|
|
}
|
1994-06-08 15:41:58 +04:00
|
|
|
continue;
|
1999-03-10 03:20:00 +03:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
b_daddr = 0;
|
|
|
|
|
if(blkp->bi_daddr != LFS_FORCE_WRITE) {
|
|
|
|
|
if (VOP_BMAP(vp, blkp->bi_lbn, NULL, &b_daddr, NULL) ||
|
|
|
|
|
b_daddr != blkp->bi_daddr)
|
|
|
|
|
{
|
|
|
|
|
if(datosn(fs,b_daddr)
|
|
|
|
|
== datosn(fs,blkp->bi_daddr))
|
|
|
|
|
{
|
1999-04-12 04:40:06 +04:00
|
|
|
printf("lfs_markv: wrong da same seg: %x vs %x\n",
|
1999-03-10 03:20:00 +03:00
|
|
|
blkp->bi_daddr, b_daddr);
|
|
|
|
|
}
|
|
|
|
|
continue;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
/*
|
|
|
|
|
* If we got to here, then we are keeping the block. If
|
|
|
|
|
* it is an indirect block, we want to actually put it
|
|
|
|
|
* in the buffer cache so that it can be updated in the
|
|
|
|
|
* finish_meta section. If it's not, we need to
|
|
|
|
|
* allocate a fake buffer so that writeseg can perform
|
|
|
|
|
* the copyin and write the buffer.
|
|
|
|
|
*/
|
|
|
|
|
/*
|
|
|
|
|
* XXX - if the block we are reading has been *extended* since
|
|
|
|
|
* it was written to disk, then we risk throwing away
|
|
|
|
|
* the extension in bread()/getblk(). Check the size
|
|
|
|
|
* here.
|
|
|
|
|
*/
|
|
|
|
|
if(blkp->bi_size < fs->lfs_bsize) {
|
|
|
|
|
s = splbio();
|
|
|
|
|
bp = incore(vp, blkp->bi_lbn);
|
|
|
|
|
if(bp && bp->b_bcount > blkp->bi_size) {
|
|
|
|
|
printf("lfs_markv: %ld > %d (fixed)\n",
|
|
|
|
|
bp->b_bcount, blkp->bi_size);
|
|
|
|
|
blkp->bi_size = bp->b_bcount;
|
|
|
|
|
}
|
|
|
|
|
splx(s);
|
|
|
|
|
}
|
2000-01-15 00:41:11 +03:00
|
|
|
if (ip->i_number != LFS_IFILE_INUM && blkp->bi_lbn >= 0) {
|
|
|
|
|
/* Data Block */
|
1999-03-26 00:39:18 +03:00
|
|
|
bp = lfs_fakebuf(vp, blkp->bi_lbn,
|
|
|
|
|
blkp->bi_size, blkp->bi_bp);
|
|
|
|
|
/* Pretend we used bread() to get it */
|
|
|
|
|
bp->b_blkno = blkp->bi_daddr;
|
2000-01-15 00:41:11 +03:00
|
|
|
} else {
|
|
|
|
|
/* Indirect block */
|
1999-03-10 03:20:00 +03:00
|
|
|
bp = getblk(vp, blkp->bi_lbn, blkp->bi_size, 0, 0);
|
|
|
|
|
if (!(bp->b_flags & (B_DONE|B_DELWRI))) { /* B_CACHE */
|
|
|
|
|
/*
|
|
|
|
|
* The block in question was not found
|
|
|
|
|
* in the cache; i.e., the block that
|
|
|
|
|
* getblk() returned is empty. So, we
|
|
|
|
|
* can (and should) copy in the
|
|
|
|
|
* contents, because we've already
|
|
|
|
|
* determined that this was the right
|
|
|
|
|
* version of this block on disk.
|
|
|
|
|
*
|
|
|
|
|
* And, it can't have changed underneath
|
|
|
|
|
* us, because we have the segment lock.
|
|
|
|
|
*/
|
|
|
|
|
error = copyin(blkp->bi_bp, bp->b_data, blkp->bi_size);
|
|
|
|
|
if(error)
|
|
|
|
|
goto err2;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
if ((error = lfs_bwrite_ext(bp,BW_CLEAN)) != 0)
|
|
|
|
|
goto err2;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Finish the old file, if there was one
|
|
|
|
|
*/
|
|
|
|
|
if(v_daddr != LFS_UNUSED_DADDR) {
|
|
|
|
|
#ifdef DEBUG_LFS
|
|
|
|
|
if(ip->i_flag & (IN_MODIFIED|IN_CLEANING))
|
|
|
|
|
iwritten++;
|
|
|
|
|
#endif
|
|
|
|
|
if(lfs_fastvget_unlock) {
|
2000-06-28 00:57:11 +04:00
|
|
|
VOP_UNLOCK(vp, 0);
|
1999-03-10 03:20:00 +03:00
|
|
|
numlocked--;
|
1994-06-08 15:41:58 +04:00
|
|
|
}
|
1999-03-10 03:20:00 +03:00
|
|
|
lfs_vunref(vp);
|
|
|
|
|
numrefed--;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* The last write has to be SEGM_SYNC, because of calling semantics.
|
|
|
|
|
* It also has to be SEGM_CKP, because otherwise we could write
|
|
|
|
|
* over the newly cleaned data contained in a checkpoint, and then
|
|
|
|
|
* we'd be unhappy at recovery time.
|
|
|
|
|
*/
|
|
|
|
|
lfs_segwrite(mntp, SEGM_SYNC|SEGM_CLEAN|SEGM_CKP);
|
|
|
|
|
free(start, M_SEGMENT);
|
|
|
|
|
|
1994-06-08 15:41:58 +04:00
|
|
|
lfs_segunlock(fs);
|
|
|
|
|
|
1999-03-10 03:20:00 +03:00
|
|
|
#ifdef DEBUG_LFS
|
|
|
|
|
printf("%d]",iwritten);
|
|
|
|
|
if(numlocked != 0 || numrefed != 0) {
|
|
|
|
|
panic("lfs_markv: numlocked=%d numrefed=%d", numlocked, numrefed);
|
|
|
|
|
}
|
|
|
|
|
#endif
|
|
|
|
|
|
2000-11-23 01:11:34 +03:00
|
|
|
vfs_unbusy(mntp);
|
1999-03-10 03:20:00 +03:00
|
|
|
if(error)
|
|
|
|
|
return (error);
|
|
|
|
|
else if(do_again)
|
|
|
|
|
return EAGAIN;
|
1994-06-08 15:41:58 +04:00
|
|
|
|
1999-03-10 03:20:00 +03:00
|
|
|
return 0;
|
|
|
|
|
|
|
|
|
|
err2:
|
1999-04-12 04:40:06 +04:00
|
|
|
printf("lfs_markv err2\n");
|
2000-11-23 01:11:34 +03:00
|
|
|
if(lfs_fastvget_unlock) {
|
|
|
|
|
VOP_UNLOCK(vp, 0);
|
|
|
|
|
--numlocked;
|
|
|
|
|
}
|
1999-03-10 03:20:00 +03:00
|
|
|
lfs_vunref(vp);
|
2000-11-23 01:11:34 +03:00
|
|
|
--numrefed;
|
|
|
|
|
|
1999-03-10 03:20:00 +03:00
|
|
|
/* Free up fakebuffers -- have to take these from the LOCKED list */
|
|
|
|
|
again:
|
1999-11-24 02:52:40 +03:00
|
|
|
s = splbio();
|
1999-03-10 03:20:00 +03:00
|
|
|
for(bp = bufqueues[BQ_LOCKED].tqh_first; bp; bp=nbp) {
|
|
|
|
|
nbp = bp->b_freelist.tqe_next;
|
|
|
|
|
if(bp->b_flags & B_CALL) {
|
|
|
|
|
if(bp->b_flags & B_BUSY) { /* not bloody likely */
|
|
|
|
|
bp->b_flags |= B_WANTED;
|
|
|
|
|
tsleep(bp, PRIBIO+1, "markv", 0);
|
|
|
|
|
splx(s);
|
|
|
|
|
goto again;
|
|
|
|
|
}
|
2000-11-23 01:11:34 +03:00
|
|
|
if(bp->b_flags & B_DELWRI)
|
|
|
|
|
fs->lfs_avail += btodb(bp->b_bcount);
|
1999-03-10 03:20:00 +03:00
|
|
|
bremfree(bp);
|
|
|
|
|
splx(s);
|
|
|
|
|
brelse(bp);
|
1999-11-24 02:52:40 +03:00
|
|
|
s = splbio();
|
1999-03-10 03:20:00 +03:00
|
|
|
}
|
|
|
|
|
}
|
1999-11-24 02:52:40 +03:00
|
|
|
splx(s);
|
1999-03-10 03:20:00 +03:00
|
|
|
free(start, M_SEGMENT);
|
1994-06-08 15:41:58 +04:00
|
|
|
lfs_segunlock(fs);
|
2000-11-23 01:11:34 +03:00
|
|
|
vfs_unbusy(mntp);
|
|
|
|
|
#ifdef DEBUG_LFS
|
|
|
|
|
if(numlocked != 0 || numrefed != 0) {
|
|
|
|
|
panic("lfs_markv: numlocked=%d numrefed=%d", numlocked, numrefed);
|
|
|
|
|
}
|
|
|
|
|
#endif
|
|
|
|
|
|
1999-03-10 03:20:00 +03:00
|
|
|
return (error);
|
|
|
|
|
|
|
|
|
|
err1:
|
1999-04-12 04:40:06 +04:00
|
|
|
printf("lfs_markv err1\n");
|
1994-06-08 15:41:58 +04:00
|
|
|
free(start, M_SEGMENT);
|
|
|
|
|
return (error);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
1999-06-09 08:52:11 +04:00
|
|
|
* sys_lfs_bmapv:
|
1994-06-08 15:41:58 +04:00
|
|
|
*
|
|
|
|
|
* This will fill in the current disk address for arrays of blocks.
|
|
|
|
|
*
|
|
|
|
|
* 0 on success
|
|
|
|
|
* -1/errno is return on error.
|
|
|
|
|
*/
|
1999-03-10 03:20:00 +03:00
|
|
|
|
1994-06-08 15:41:58 +04:00
|
|
|
int
|
1999-06-09 08:52:11 +04:00
|
|
|
sys_lfs_bmapv(p, v, retval)
|
1994-06-08 15:41:58 +04:00
|
|
|
struct proc *p;
|
1995-09-22 03:39:20 +04:00
|
|
|
void *v;
|
|
|
|
|
register_t *retval;
|
|
|
|
|
{
|
1999-06-09 16:18:19 +04:00
|
|
|
struct sys_lfs_bmapv_args /* {
|
|
|
|
|
syscallarg(fsid_t *) fsidp;
|
|
|
|
|
syscallarg(struct block_info *) blkiov;
|
|
|
|
|
syscallarg(int) blkcnt;
|
|
|
|
|
} */ *uap = v;
|
1994-06-08 15:41:58 +04:00
|
|
|
BLOCK_INFO *blkp;
|
1999-03-10 03:20:00 +03:00
|
|
|
IFILE *ifp;
|
|
|
|
|
struct buf *bp;
|
|
|
|
|
struct inode *ip = NULL;
|
|
|
|
|
struct lfs *fs;
|
1994-06-08 15:41:58 +04:00
|
|
|
struct mount *mntp;
|
1998-03-01 05:20:01 +03:00
|
|
|
struct ufsmount *ump;
|
1994-06-08 15:41:58 +04:00
|
|
|
struct vnode *vp;
|
|
|
|
|
fsid_t fsid;
|
|
|
|
|
void *start;
|
1999-03-10 03:20:00 +03:00
|
|
|
ino_t lastino;
|
|
|
|
|
ufs_daddr_t v_daddr;
|
|
|
|
|
int origcnt, cnt, error, need_unlock=0;
|
|
|
|
|
int numlocked=0, numrefed=0;
|
|
|
|
|
#ifdef LFS_TRACK_IOS
|
|
|
|
|
int j;
|
|
|
|
|
#endif
|
1994-06-08 15:41:58 +04:00
|
|
|
|
1999-03-10 03:20:00 +03:00
|
|
|
lfs_cleaner_pid = p->p_pid;
|
|
|
|
|
|
1996-02-10 01:28:45 +03:00
|
|
|
if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
|
1994-06-08 15:41:58 +04:00
|
|
|
return (error);
|
1999-03-10 03:20:00 +03:00
|
|
|
|
|
|
|
|
if ((error = copyin(SCARG(uap, fsidp), &fsid, sizeof(fsid_t))) != 0)
|
1994-06-08 15:41:58 +04:00
|
|
|
return (error);
|
1998-03-01 05:20:01 +03:00
|
|
|
if ((mntp = vfs_getvfs(&fsid)) == NULL)
|
2000-11-23 01:11:34 +03:00
|
|
|
return (ENOENT);
|
1999-03-10 03:20:00 +03:00
|
|
|
|
|
|
|
|
ump = VFSTOUFS(mntp);
|
2000-11-23 01:11:34 +03:00
|
|
|
if ((error = vfs_busy(mntp, LK_NOWAIT, NULL)) != 0)
|
|
|
|
|
return (error);
|
1999-03-10 03:20:00 +03:00
|
|
|
|
|
|
|
|
origcnt = cnt = SCARG(uap, blkcnt);
|
|
|
|
|
start = malloc(cnt * sizeof(BLOCK_INFO), M_SEGMENT, M_WAITOK);
|
|
|
|
|
error = copyin(SCARG(uap, blkiov), start, cnt * sizeof(BLOCK_INFO));
|
1996-02-10 01:28:45 +03:00
|
|
|
if (error) {
|
1999-03-10 03:20:00 +03:00
|
|
|
free(start, M_SEGMENT);
|
2000-11-23 01:11:34 +03:00
|
|
|
vfs_unbusy(mntp);
|
1994-06-08 15:41:58 +04:00
|
|
|
return (error);
|
|
|
|
|
}
|
1999-03-10 03:20:00 +03:00
|
|
|
|
|
|
|
|
fs = VFSTOUFS(mntp)->um_lfs;
|
|
|
|
|
|
|
|
|
|
error = 0;
|
|
|
|
|
|
|
|
|
|
/* these were inside the initialization for the for loop */
|
|
|
|
|
v_daddr = LFS_UNUSED_DADDR;
|
|
|
|
|
lastino = LFS_UNUSED_INUM;
|
|
|
|
|
for (blkp = start; cnt--; ++blkp)
|
|
|
|
|
{
|
|
|
|
|
#ifdef DEBUG
|
|
|
|
|
if (datosn(fs, fs->lfs_curseg) == datosn(fs, blkp->bi_daddr)) {
|
1999-04-12 04:40:06 +04:00
|
|
|
printf("lfs_bmapv: attempt to clean current segment? (#%d)\n",
|
1999-03-10 03:20:00 +03:00
|
|
|
datosn(fs, fs->lfs_curseg));
|
|
|
|
|
free(start,M_SEGMENT);
|
2000-11-23 01:11:34 +03:00
|
|
|
vfs_unbusy(mntp);
|
1999-03-10 03:20:00 +03:00
|
|
|
return (EBUSY);
|
|
|
|
|
}
|
|
|
|
|
#endif /* DEBUG */
|
|
|
|
|
#ifdef LFS_TRACK_IOS
|
|
|
|
|
/*
|
|
|
|
|
* If there is I/O on this segment that is not yet complete,
|
|
|
|
|
* the cleaner probably does not have the right information.
|
|
|
|
|
* Send it packing.
|
|
|
|
|
*/
|
|
|
|
|
for(j=0;j<LFS_THROTTLE;j++) {
|
|
|
|
|
if(fs->lfs_pending[j] != LFS_UNUSED_DADDR
|
|
|
|
|
&& datosn(fs,fs->lfs_pending[j])==datosn(fs,blkp->bi_daddr))
|
|
|
|
|
{
|
|
|
|
|
printf("lfs_bmapv: attempt to clean pending segment? (#%d)\n",
|
|
|
|
|
datosn(fs, fs->lfs_pending[j]));
|
|
|
|
|
free(start,M_SEGMENT);
|
2000-11-23 01:11:34 +03:00
|
|
|
vfs_unbusy(mntp);
|
1999-03-10 03:20:00 +03:00
|
|
|
return (EBUSY);
|
|
|
|
|
}
|
|
|
|
|
}
|
1994-06-08 15:41:58 +04:00
|
|
|
|
1999-03-10 03:20:00 +03:00
|
|
|
#endif /* LFS_TRACK_IOS */
|
1998-03-01 05:20:01 +03:00
|
|
|
/*
|
1999-03-10 03:20:00 +03:00
|
|
|
* Get the IFILE entry (only once) and see if the file still
|
|
|
|
|
* exists.
|
1998-03-01 05:20:01 +03:00
|
|
|
*/
|
1999-03-10 03:20:00 +03:00
|
|
|
if (lastino != blkp->bi_inode) {
|
|
|
|
|
/*
|
|
|
|
|
* Finish the old file, if there was one. The presence
|
|
|
|
|
* of a usable vnode in vp is signaled by a valid
|
|
|
|
|
* v_daddr.
|
|
|
|
|
*/
|
|
|
|
|
if(v_daddr != LFS_UNUSED_DADDR) {
|
|
|
|
|
if(need_unlock) {
|
2000-06-28 00:57:11 +04:00
|
|
|
VOP_UNLOCK(vp, 0);
|
1999-03-10 03:20:00 +03:00
|
|
|
numlocked--;
|
|
|
|
|
}
|
|
|
|
|
lfs_vunref(vp);
|
|
|
|
|
numrefed--;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Start a new file
|
|
|
|
|
*/
|
|
|
|
|
lastino = blkp->bi_inode;
|
|
|
|
|
if (blkp->bi_inode == LFS_IFILE_INUM)
|
|
|
|
|
v_daddr = fs->lfs_idaddr;
|
|
|
|
|
else {
|
|
|
|
|
LFS_IENTRY(ifp, fs, blkp->bi_inode, bp);
|
|
|
|
|
v_daddr = ifp->if_daddr;
|
|
|
|
|
brelse(bp);
|
|
|
|
|
}
|
|
|
|
|
if (v_daddr == LFS_UNUSED_DADDR) {
|
|
|
|
|
blkp->bi_daddr = LFS_UNUSED_DADDR;
|
|
|
|
|
continue;
|
|
|
|
|
}
|
|
|
|
|
/*
|
|
|
|
|
* A regular call to VFS_VGET could deadlock
|
|
|
|
|
* here. Instead, we try an unlocked access.
|
|
|
|
|
*/
|
|
|
|
|
vp = ufs_ihashlookup(ump->um_dev, blkp->bi_inode);
|
1999-03-26 00:54:10 +03:00
|
|
|
if (vp != NULL && !(vp->v_flag & VXLOCK)) {
|
1999-03-10 03:20:00 +03:00
|
|
|
ip = VTOI(vp);
|
2000-06-22 22:11:45 +04:00
|
|
|
if (lfs_vref(vp)) {
|
|
|
|
|
v_daddr = LFS_UNUSED_DADDR;
|
|
|
|
|
need_unlock = 0;
|
|
|
|
|
continue;
|
|
|
|
|
}
|
2000-06-28 00:57:11 +04:00
|
|
|
numrefed++;
|
1999-03-10 03:20:00 +03:00
|
|
|
if(VOP_ISLOCKED(vp)) {
|
2000-06-28 00:57:11 +04:00
|
|
|
#ifdef DEBUG_LFS
|
|
|
|
|
printf("lfs_bmapv: inode %d inlocked\n",ip->i_number);
|
|
|
|
|
#endif
|
|
|
|
|
v_daddr = LFS_UNUSED_DADDR;
|
1999-03-10 03:20:00 +03:00
|
|
|
need_unlock = 0;
|
2000-06-28 00:57:11 +04:00
|
|
|
lfs_vunref(vp);
|
|
|
|
|
--numrefed;
|
|
|
|
|
continue;
|
1999-03-10 03:20:00 +03:00
|
|
|
} else {
|
2000-06-28 00:57:11 +04:00
|
|
|
vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
|
1999-03-10 03:20:00 +03:00
|
|
|
need_unlock = FVG_UNLOCK;
|
|
|
|
|
numlocked++;
|
|
|
|
|
}
|
|
|
|
|
} else {
|
|
|
|
|
error = VFS_VGET(mntp, blkp->bi_inode, &vp);
|
|
|
|
|
if(error) {
|
1999-03-26 00:54:10 +03:00
|
|
|
#ifdef DEBUG_LFS
|
1999-03-26 01:26:52 +03:00
|
|
|
printf("lfs_bmapv: vget of ino %d failed with %d",blkp->bi_inode,error);
|
1999-03-26 00:54:10 +03:00
|
|
|
#endif
|
2000-06-28 00:57:11 +04:00
|
|
|
v_daddr = LFS_UNUSED_DADDR;
|
|
|
|
|
need_unlock = 0;
|
1999-03-10 03:20:00 +03:00
|
|
|
continue;
|
|
|
|
|
} else {
|
|
|
|
|
need_unlock = FVG_PUT;
|
|
|
|
|
numlocked++;
|
|
|
|
|
numrefed++;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
ip = VTOI(vp);
|
|
|
|
|
} else if (v_daddr == LFS_UNUSED_DADDR) {
|
|
|
|
|
/*
|
|
|
|
|
* This can only happen if the vnode is dead.
|
|
|
|
|
* Keep going. Note that we DO NOT set the
|
|
|
|
|
* bi_addr to anything -- if we failed to get
|
|
|
|
|
* the vnode, for example, we want to assume
|
|
|
|
|
* conservatively that all of its blocks *are*
|
|
|
|
|
* located in the segment in question.
|
|
|
|
|
* lfs_markv will throw them out if we are
|
|
|
|
|
* wrong.
|
|
|
|
|
*/
|
|
|
|
|
/* blkp->bi_daddr = LFS_UNUSED_DADDR; */
|
|
|
|
|
continue;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* Past this point we are guaranteed that vp, ip are valid. */
|
|
|
|
|
|
|
|
|
|
if(blkp->bi_lbn == LFS_UNUSED_LBN) {
|
|
|
|
|
/*
|
|
|
|
|
* We just want the inode address, which is
|
|
|
|
|
* conveniently in v_daddr.
|
|
|
|
|
*/
|
|
|
|
|
blkp->bi_daddr = v_daddr;
|
|
|
|
|
} else {
|
|
|
|
|
error = VOP_BMAP(vp, blkp->bi_lbn, NULL,
|
|
|
|
|
&(blkp->bi_daddr), NULL);
|
|
|
|
|
if(error)
|
|
|
|
|
{
|
|
|
|
|
blkp->bi_daddr = LFS_UNUSED_DADDR;
|
|
|
|
|
continue;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Finish the old file, if there was one. The presence
|
|
|
|
|
* of a usable vnode in vp is signaled by a valid v_daddr.
|
|
|
|
|
*/
|
|
|
|
|
if(v_daddr != LFS_UNUSED_DADDR) {
|
|
|
|
|
if(need_unlock) {
|
2000-06-28 00:57:11 +04:00
|
|
|
VOP_UNLOCK(vp, 0);
|
1999-03-10 03:20:00 +03:00
|
|
|
numlocked--;
|
1994-06-08 15:41:58 +04:00
|
|
|
}
|
1999-03-10 03:20:00 +03:00
|
|
|
lfs_vunref(vp);
|
|
|
|
|
numrefed--;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if(numlocked != 0 || numrefed != 0) {
|
|
|
|
|
panic("lfs_bmapv: numlocked=%d numrefed=%d", numlocked,
|
|
|
|
|
numrefed);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
copyout(start, SCARG(uap, blkiov), origcnt * sizeof(BLOCK_INFO));
|
1994-06-08 15:41:58 +04:00
|
|
|
free(start, M_SEGMENT);
|
2000-11-23 01:11:34 +03:00
|
|
|
vfs_unbusy(mntp);
|
1999-03-10 03:20:00 +03:00
|
|
|
|
|
|
|
|
return 0;
|
1994-06-08 15:41:58 +04:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
1999-06-09 08:52:11 +04:00
|
|
|
* sys_lfs_segclean:
|
1994-06-08 15:41:58 +04:00
|
|
|
*
|
|
|
|
|
* Mark the segment clean.
|
|
|
|
|
*
|
|
|
|
|
* 0 on success
|
|
|
|
|
* -1/errno is return on error.
|
|
|
|
|
*/
|
|
|
|
|
int
|
1999-06-09 08:52:11 +04:00
|
|
|
sys_lfs_segclean(p, v, retval)
|
1994-06-08 15:41:58 +04:00
|
|
|
struct proc *p;
|
1995-09-22 03:39:20 +04:00
|
|
|
void *v;
|
|
|
|
|
register_t *retval;
|
|
|
|
|
{
|
1999-06-09 16:18:19 +04:00
|
|
|
struct sys_lfs_segclean_args /* {
|
|
|
|
|
syscallarg(fsid_t *) fsidp;
|
|
|
|
|
syscallarg(u_long) segment;
|
|
|
|
|
} */ *uap = v;
|
1994-06-08 15:41:58 +04:00
|
|
|
CLEANERINFO *cip;
|
|
|
|
|
SEGUSE *sup;
|
|
|
|
|
struct buf *bp;
|
|
|
|
|
struct mount *mntp;
|
|
|
|
|
struct lfs *fs;
|
|
|
|
|
fsid_t fsid;
|
|
|
|
|
int error;
|
1999-03-10 03:20:00 +03:00
|
|
|
|
1996-02-10 01:28:45 +03:00
|
|
|
if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
|
1994-06-08 15:41:58 +04:00
|
|
|
return (error);
|
1999-03-10 03:20:00 +03:00
|
|
|
|
1996-02-10 01:28:45 +03:00
|
|
|
if ((error = copyin(SCARG(uap, fsidp), &fsid, sizeof(fsid_t))) != 0)
|
1994-06-08 15:41:58 +04:00
|
|
|
return (error);
|
1998-03-01 05:20:01 +03:00
|
|
|
if ((mntp = vfs_getvfs(&fsid)) == NULL)
|
2000-11-23 01:11:34 +03:00
|
|
|
return (ENOENT);
|
1999-03-10 03:20:00 +03:00
|
|
|
|
1994-06-08 15:41:58 +04:00
|
|
|
fs = VFSTOUFS(mntp)->um_lfs;
|
1999-03-10 03:20:00 +03:00
|
|
|
|
1994-10-20 07:20:55 +03:00
|
|
|
if (datosn(fs, fs->lfs_curseg) == SCARG(uap, segment))
|
1994-06-08 15:41:58 +04:00
|
|
|
return (EBUSY);
|
1999-03-10 03:20:00 +03:00
|
|
|
|
2000-11-23 01:11:34 +03:00
|
|
|
if ((error = vfs_busy(mntp, LK_NOWAIT, NULL)) != 0)
|
|
|
|
|
return (error);
|
1994-10-20 07:20:55 +03:00
|
|
|
LFS_SEGENTRY(sup, fs, SCARG(uap, segment), bp);
|
1994-06-08 15:41:58 +04:00
|
|
|
if (sup->su_flags & SEGUSE_ACTIVE) {
|
|
|
|
|
brelse(bp);
|
2000-11-23 01:11:34 +03:00
|
|
|
vfs_unbusy(mntp);
|
1994-06-08 15:41:58 +04:00
|
|
|
return (EBUSY);
|
|
|
|
|
}
|
2000-10-20 21:48:05 +04:00
|
|
|
if (!(sup->su_flags & SEGUSE_DIRTY)) {
|
|
|
|
|
brelse(bp);
|
2000-11-23 01:11:34 +03:00
|
|
|
vfs_unbusy(mntp);
|
2000-10-20 21:48:05 +04:00
|
|
|
return (EALREADY);
|
|
|
|
|
}
|
1999-03-10 03:20:00 +03:00
|
|
|
|
2000-07-05 02:30:37 +04:00
|
|
|
fs->lfs_avail += fsbtodb(fs, fs->lfs_ssize);
|
|
|
|
|
if (sup->su_flags & SEGUSE_SUPERBLOCK)
|
|
|
|
|
fs->lfs_avail -= btodb(LFS_SBPAD);
|
Various bug-fixes to LFS, to wit:
Kernel:
* Add runtime quantity lfs_ravail, the number of disk-blocks reserved
for writing. Writes to the filesystem first reserve a maximum amount
of blocks before their write is allowed to proceed; after the blocks
are allocated the reserved total is reduced by a corresponding amount.
If the lfs_reserve function cannot immediately reserve the requested
number of blocks, the inode is unlocked, and the thread sleeps until
the cleaner has made enough space available for the blocks to be
reserved. In this way large files can be written to the filesystem
(or, smaller files can be written to a nearly-full but thoroughly
clean filesystem) and the cleaner can still function properly.
* Remove explicit switching on dlfs_minfreeseg from the kernel code; it
is now merely a fs-creation parameter used to compute dlfs_avail and
dlfs_bfree (and used by fsck_lfs(8) to check their accuracy). Its
former role is better assumed by a properly computed dlfs_avail.
* Bounds-check inode numbers submitted through lfs_bmapv and lfs_markv.
This prevents a panic, but, if the cleaner is feeding the filesystem
the wrong data, you are still in a world of hurt.
* Cleanup: remove explicit references of DEV_BSIZE in favor of
btodb()/dbtob().
lfs_cleanerd:
* Make -n mean "send N segments' blocks through a single call to
lfs_markv". Previously it had meant "clean N segments though N calls
to lfs_markv, before looking again to see if more need to be cleaned".
The new behavior gives better packing of direct data on disk with as
little metadata as possible, largely alleviating the problem that the
cleaner can consume more disk through inefficient use of metadata than
it frees by moving dirty data away from clean "holes" to produce
entirely clean segments.
* Make -b mean "read as many segments as necessary to write N segments
of dirty data back to disk", rather than its former meaning of "read
as many segments as necessary to free N segments worth of space". The
new meaning, combined with the new -n behavior described above,
further aids in cleaning storage efficiency as entire segments can be
written at once, using as few blocks as possible for segment summaries
and inode blocks.
* Make the cleaner take note of segments which could not be cleaned due
to error, and not attempt to clean them until they are entirely free
of dirty blocks. This prevents the case in which a cleanerd running
with -n 1 and without -b (formerly the default) would spin trying
repeatedly to clean a corrupt segment, while the remaining space
filled and deadlocked the filesystem.
* Update the lfs_cleanerd manual page to describe all the options,
including the changes mentioned here (in particular, the -b and -n
flags were previously undocumented).
fsck_lfs:
* Check, and optionally fix, lfs_avail (to an exact figure) and
lfs_bfree (within a margin of error) in pass 5.
newfs_lfs:
* Reduce the default dlfs_minfreeseg to 1/20 of the total segments.
* Add a warning if the sgs disklabel field is 16 (the default for FFS'
cpg, but not usually desirable for LFS' sgs: 5--8 is a better range).
* Change the calculation of lfs_avail and lfs_bfree, corresponding to
the kernel changes mentioned above.
mount_lfs:
* Add -N and -b options to pass corresponding -n and -b options to
lfs_cleanerd.
* Default to calling lfs_cleanerd with "-b -n 4".
[All of these changes were largely tested in the 1.5 branch, with the
idea that they (along with previous un-pulled-up work) could be applied
to the branch while it was still in ALPHA2; however my test system has
experienced corruption on another filesystem (/dev/console has gone
missing :^), and, while I believe this unrelated to the LFS changes, I
cannot with good conscience request that the changes be pulled up.]
2000-09-09 08:49:54 +04:00
|
|
|
fs->lfs_bfree += sup->su_nsums * btodb(LFS_SUMMARY_SIZE) +
|
|
|
|
|
fsbtodb(fs, sup->su_ninos);
|
|
|
|
|
fs->lfs_dmeta -= sup->su_nsums * btodb(LFS_SUMMARY_SIZE) +
|
|
|
|
|
fsbtodb(fs, sup->su_ninos);
|
2000-06-28 00:57:11 +04:00
|
|
|
if (fs->lfs_dmeta < 0)
|
|
|
|
|
fs->lfs_dmeta = 0;
|
1994-06-08 15:41:58 +04:00
|
|
|
sup->su_flags &= ~SEGUSE_DIRTY;
|
1999-03-10 03:20:00 +03:00
|
|
|
(void) VOP_BWRITE(bp);
|
|
|
|
|
|
1994-06-08 15:41:58 +04:00
|
|
|
LFS_CLEANERINFO(cip, fs, bp);
|
|
|
|
|
++cip->clean;
|
|
|
|
|
--cip->dirty;
|
1999-03-10 03:20:00 +03:00
|
|
|
fs->lfs_nclean = cip->clean;
|
Various bug-fixes to LFS, to wit:
Kernel:
* Add runtime quantity lfs_ravail, the number of disk-blocks reserved
for writing. Writes to the filesystem first reserve a maximum amount
of blocks before their write is allowed to proceed; after the blocks
are allocated the reserved total is reduced by a corresponding amount.
If the lfs_reserve function cannot immediately reserve the requested
number of blocks, the inode is unlocked, and the thread sleeps until
the cleaner has made enough space available for the blocks to be
reserved. In this way large files can be written to the filesystem
(or, smaller files can be written to a nearly-full but thoroughly
clean filesystem) and the cleaner can still function properly.
* Remove explicit switching on dlfs_minfreeseg from the kernel code; it
is now merely a fs-creation parameter used to compute dlfs_avail and
dlfs_bfree (and used by fsck_lfs(8) to check their accuracy). Its
former role is better assumed by a properly computed dlfs_avail.
* Bounds-check inode numbers submitted through lfs_bmapv and lfs_markv.
This prevents a panic, but, if the cleaner is feeding the filesystem
the wrong data, you are still in a world of hurt.
* Cleanup: remove explicit references of DEV_BSIZE in favor of
btodb()/dbtob().
lfs_cleanerd:
* Make -n mean "send N segments' blocks through a single call to
lfs_markv". Previously it had meant "clean N segments though N calls
to lfs_markv, before looking again to see if more need to be cleaned".
The new behavior gives better packing of direct data on disk with as
little metadata as possible, largely alleviating the problem that the
cleaner can consume more disk through inefficient use of metadata than
it frees by moving dirty data away from clean "holes" to produce
entirely clean segments.
* Make -b mean "read as many segments as necessary to write N segments
of dirty data back to disk", rather than its former meaning of "read
as many segments as necessary to free N segments worth of space". The
new meaning, combined with the new -n behavior described above,
further aids in cleaning storage efficiency as entire segments can be
written at once, using as few blocks as possible for segment summaries
and inode blocks.
* Make the cleaner take note of segments which could not be cleaned due
to error, and not attempt to clean them until they are entirely free
of dirty blocks. This prevents the case in which a cleanerd running
with -n 1 and without -b (formerly the default) would spin trying
repeatedly to clean a corrupt segment, while the remaining space
filled and deadlocked the filesystem.
* Update the lfs_cleanerd manual page to describe all the options,
including the changes mentioned here (in particular, the -b and -n
flags were previously undocumented).
fsck_lfs:
* Check, and optionally fix, lfs_avail (to an exact figure) and
lfs_bfree (within a margin of error) in pass 5.
newfs_lfs:
* Reduce the default dlfs_minfreeseg to 1/20 of the total segments.
* Add a warning if the sgs disklabel field is 16 (the default for FFS'
cpg, but not usually desirable for LFS' sgs: 5--8 is a better range).
* Change the calculation of lfs_avail and lfs_bfree, corresponding to
the kernel changes mentioned above.
mount_lfs:
* Add -N and -b options to pass corresponding -n and -b options to
lfs_cleanerd.
* Default to calling lfs_cleanerd with "-b -n 4".
[All of these changes were largely tested in the 1.5 branch, with the
idea that they (along with previous un-pulled-up work) could be applied
to the branch while it was still in ALPHA2; however my test system has
experienced corruption on another filesystem (/dev/console has gone
missing :^), and, while I believe this unrelated to the LFS changes, I
cannot with good conscience request that the changes be pulled up.]
2000-09-09 08:49:54 +04:00
|
|
|
cip->bfree = fs->lfs_bfree;
|
|
|
|
|
cip->avail = fs->lfs_avail - fs->lfs_ravail;
|
1994-06-08 15:41:58 +04:00
|
|
|
(void) VOP_BWRITE(bp);
|
|
|
|
|
wakeup(&fs->lfs_avail);
|
2000-11-23 01:11:34 +03:00
|
|
|
vfs_unbusy(mntp);
|
1999-03-10 03:20:00 +03:00
|
|
|
|
1994-06-08 15:41:58 +04:00
|
|
|
return (0);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
1999-06-09 08:52:11 +04:00
|
|
|
* sys_lfs_segwait:
|
1994-06-08 15:41:58 +04:00
|
|
|
*
|
|
|
|
|
* This will block until a segment in file system fsid is written. A timeout
|
|
|
|
|
* in milliseconds may be specified which will awake the cleaner automatically.
|
|
|
|
|
* An fsid of -1 means any file system, and a timeout of 0 means forever.
|
|
|
|
|
*
|
|
|
|
|
* 0 on success
|
|
|
|
|
* 1 on timeout
|
|
|
|
|
* -1/errno is return on error.
|
|
|
|
|
*/
|
|
|
|
|
int
|
1999-06-09 08:52:11 +04:00
|
|
|
sys_lfs_segwait(p, v, retval)
|
1994-06-08 15:41:58 +04:00
|
|
|
struct proc *p;
|
1995-09-22 03:39:20 +04:00
|
|
|
void *v;
|
|
|
|
|
register_t *retval;
|
|
|
|
|
{
|
1999-06-09 16:18:19 +04:00
|
|
|
struct sys_lfs_segwait_args /* {
|
|
|
|
|
syscallarg(fsid_t *) fsidp;
|
|
|
|
|
syscallarg(struct timeval *) tv;
|
|
|
|
|
} */ *uap = v;
|
1994-06-08 15:41:58 +04:00
|
|
|
extern int lfs_allclean_wakeup;
|
|
|
|
|
struct mount *mntp;
|
|
|
|
|
struct timeval atv;
|
|
|
|
|
fsid_t fsid;
|
|
|
|
|
void *addr;
|
|
|
|
|
u_long timeout;
|
|
|
|
|
int error, s;
|
1999-03-10 03:20:00 +03:00
|
|
|
|
1996-02-10 01:28:45 +03:00
|
|
|
if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) {
|
1994-06-08 15:41:58 +04:00
|
|
|
return (error);
|
|
|
|
|
}
|
1996-02-10 01:28:45 +03:00
|
|
|
if ((error = copyin(SCARG(uap, fsidp), &fsid, sizeof(fsid_t))) != 0)
|
1994-06-08 15:41:58 +04:00
|
|
|
return (error);
|
1998-03-01 05:20:01 +03:00
|
|
|
if ((mntp = vfs_getvfs(&fsid)) == NULL)
|
1994-06-08 15:41:58 +04:00
|
|
|
addr = &lfs_allclean_wakeup;
|
|
|
|
|
else
|
|
|
|
|
addr = &VFSTOUFS(mntp)->um_lfs->lfs_nextseg;
|
1999-03-10 03:20:00 +03:00
|
|
|
|
1994-10-20 07:20:55 +03:00
|
|
|
if (SCARG(uap, tv)) {
|
1996-02-10 01:28:45 +03:00
|
|
|
error = copyin(SCARG(uap, tv), &atv, sizeof(struct timeval));
|
|
|
|
|
if (error)
|
1994-06-08 15:41:58 +04:00
|
|
|
return (error);
|
|
|
|
|
if (itimerfix(&atv))
|
|
|
|
|
return (EINVAL);
|
2000-07-13 21:35:03 +04:00
|
|
|
/*
|
|
|
|
|
* XXX THIS COULD SLEEP FOREVER IF TIMEOUT IS {0,0}!
|
|
|
|
|
* XXX IS THAT WHAT IS INTENDED?
|
|
|
|
|
*/
|
1994-06-08 15:41:58 +04:00
|
|
|
s = splclock();
|
1995-03-21 16:33:34 +03:00
|
|
|
timeradd(&atv, &time, &atv);
|
1994-06-08 15:41:58 +04:00
|
|
|
timeout = hzto(&atv);
|
|
|
|
|
splx(s);
|
|
|
|
|
} else
|
|
|
|
|
timeout = 0;
|
1999-03-10 03:20:00 +03:00
|
|
|
|
1994-06-08 15:41:58 +04:00
|
|
|
error = tsleep(addr, PCATCH | PUSER, "segment", timeout);
|
|
|
|
|
return (error == ERESTART ? EINTR : 0);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* VFS_VGET call specialized for the cleaner. The cleaner already knows the
|
|
|
|
|
* daddr from the ifile, so don't look it up again. If the cleaner is
|
|
|
|
|
* processing IINFO structures, it may have the ondisk inode already, so
|
|
|
|
|
* don't go retrieving it again.
|
1999-03-10 03:20:00 +03:00
|
|
|
*
|
|
|
|
|
* If we find the vnode on the hash chain, then it may be locked by another
|
|
|
|
|
* process; so we set (*need_unlock) to zero.
|
|
|
|
|
*
|
|
|
|
|
* If we don't, we call ufs_ihashins, which locks the inode, and we set
|
|
|
|
|
* (*need_unlock) to non-zero.
|
|
|
|
|
*
|
|
|
|
|
* In either case we lfs_vref, and it is the caller's responsibility to
|
|
|
|
|
* lfs_vunref and VOP_UNLOCK (if necessary) when finished.
|
1994-06-08 15:41:58 +04:00
|
|
|
*/
|
1999-03-10 03:20:00 +03:00
|
|
|
extern struct lock ufs_hashlock;
|
|
|
|
|
|
2000-07-01 00:45:38 +04:00
|
|
|
int
|
|
|
|
|
lfs_fasthashget(dev, ino, need_unlock, vpp)
|
|
|
|
|
dev_t dev;
|
|
|
|
|
ino_t ino;
|
|
|
|
|
int *need_unlock;
|
|
|
|
|
struct vnode **vpp;
|
|
|
|
|
{
|
|
|
|
|
struct inode *ip;
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* This is playing fast and loose. Someone may have the inode
|
|
|
|
|
* locked, in which case they are going to be distinctly unhappy
|
|
|
|
|
* if we trash something.
|
|
|
|
|
*/
|
|
|
|
|
if ((*vpp = ufs_ihashlookup(dev, ino)) != NULL) {
|
|
|
|
|
if ((*vpp)->v_flag & VXLOCK) {
|
2000-07-05 02:30:37 +04:00
|
|
|
printf("lfs_fastvget: vnode VXLOCKed for ino %d\n",
|
|
|
|
|
ino);
|
2000-07-01 00:45:38 +04:00
|
|
|
clean_vnlocked++;
|
|
|
|
|
#ifdef LFS_EAGAIN_FAIL
|
|
|
|
|
return EAGAIN;
|
|
|
|
|
#endif
|
|
|
|
|
}
|
|
|
|
|
ip = VTOI(*vpp);
|
|
|
|
|
if (lfs_vref(*vpp)) {
|
|
|
|
|
clean_inlocked++;
|
|
|
|
|
return EAGAIN;
|
|
|
|
|
}
|
|
|
|
|
if (VOP_ISLOCKED(*vpp)) {
|
|
|
|
|
#ifdef DEBUG_LFS
|
|
|
|
|
printf("lfs_fastvget: ino %d inlocked by pid %d\n",
|
2000-07-05 02:30:37 +04:00
|
|
|
ip->i_number, (*vpp)->v_lock.lk_lockholder);
|
2000-07-01 00:45:38 +04:00
|
|
|
#endif
|
|
|
|
|
clean_inlocked++;
|
|
|
|
|
#ifdef LFS_EAGAIN_FAIL
|
|
|
|
|
lfs_vunref(*vpp);
|
|
|
|
|
return EAGAIN;
|
|
|
|
|
#endif /* LFS_EAGAIN_FAIL */
|
|
|
|
|
} else {
|
|
|
|
|
vn_lock(*vpp, LK_EXCLUSIVE | LK_RETRY);
|
|
|
|
|
*need_unlock |= FVG_UNLOCK;
|
|
|
|
|
}
|
|
|
|
|
} else
|
|
|
|
|
*vpp = NULL;
|
|
|
|
|
|
|
|
|
|
return (0);
|
|
|
|
|
}
|
|
|
|
|
|
1994-06-08 15:41:58 +04:00
|
|
|
int
|
1999-03-10 03:20:00 +03:00
|
|
|
lfs_fastvget(mp, ino, daddr, vpp, dinp, need_unlock)
|
1994-06-08 15:41:58 +04:00
|
|
|
struct mount *mp;
|
|
|
|
|
ino_t ino;
|
1998-03-01 05:20:01 +03:00
|
|
|
ufs_daddr_t daddr;
|
1994-06-08 15:41:58 +04:00
|
|
|
struct vnode **vpp;
|
|
|
|
|
struct dinode *dinp;
|
1999-03-10 03:20:00 +03:00
|
|
|
int *need_unlock;
|
1994-06-08 15:41:58 +04:00
|
|
|
{
|
2000-03-30 16:41:09 +04:00
|
|
|
struct inode *ip;
|
1994-06-08 15:41:58 +04:00
|
|
|
struct vnode *vp;
|
|
|
|
|
struct ufsmount *ump;
|
|
|
|
|
dev_t dev;
|
|
|
|
|
int error;
|
1999-03-10 03:20:00 +03:00
|
|
|
struct buf *bp;
|
|
|
|
|
|
1994-06-08 15:41:58 +04:00
|
|
|
ump = VFSTOUFS(mp);
|
|
|
|
|
dev = ump->um_dev;
|
1999-03-10 03:20:00 +03:00
|
|
|
*need_unlock = 0;
|
2000-07-01 00:45:38 +04:00
|
|
|
|
2000-11-27 06:33:57 +03:00
|
|
|
/*
|
|
|
|
|
* Wait until the filesystem is fully mounted before allowing vget
|
|
|
|
|
* to complete. This prevents possible problems with roll-forward.
|
|
|
|
|
*/
|
|
|
|
|
while(ump->um_lfs->lfs_flags & LFS_NOTYET) {
|
|
|
|
|
tsleep(&ump->um_lfs->lfs_flags, PRIBIO+1, "lfs_fnotyet", 0);
|
|
|
|
|
}
|
|
|
|
|
/*
|
|
|
|
|
* This is playing fast and loose. Someone may have the inode
|
|
|
|
|
* locked, in which case they are going to be distinctly unhappy
|
|
|
|
|
* if we trash something.
|
|
|
|
|
*/
|
|
|
|
|
|
2000-07-01 00:45:38 +04:00
|
|
|
error = lfs_fasthashget(dev, ino, need_unlock, vpp);
|
|
|
|
|
if (error != 0 || *vpp != NULL)
|
|
|
|
|
return (error);
|
|
|
|
|
|
2000-07-03 22:22:10 +04:00
|
|
|
if ((error = getnewvnode(VT_LFS, mp, lfs_vnodeop_p, &vp)) != 0) {
|
2000-07-01 00:45:38 +04:00
|
|
|
*vpp = NULL;
|
|
|
|
|
return (error);
|
|
|
|
|
}
|
|
|
|
|
|
1999-03-10 03:20:00 +03:00
|
|
|
do {
|
2000-07-01 00:45:38 +04:00
|
|
|
error = lfs_fasthashget(dev, ino, need_unlock, vpp);
|
|
|
|
|
if (error != 0 || *vpp != NULL) {
|
|
|
|
|
ungetnewvnode(vp);
|
|
|
|
|
return (error);
|
1999-03-10 03:20:00 +03:00
|
|
|
}
|
|
|
|
|
} while (lockmgr(&ufs_hashlock, LK_EXCLUSIVE|LK_SLEEPFAIL, 0));
|
1994-06-08 15:41:58 +04:00
|
|
|
|
|
|
|
|
/* Allocate new vnode/inode. */
|
2000-07-01 00:45:38 +04:00
|
|
|
lfs_vcreate(mp, ino, vp);
|
|
|
|
|
|
1994-06-08 15:41:58 +04:00
|
|
|
/*
|
|
|
|
|
* 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.
|
|
|
|
|
*/
|
|
|
|
|
ip = VTOI(vp);
|
|
|
|
|
ufs_ihashins(ip);
|
1999-03-10 03:20:00 +03:00
|
|
|
lockmgr(&ufs_hashlock, LK_RELEASE, 0);
|
|
|
|
|
|
1994-06-08 15:41:58 +04:00
|
|
|
/*
|
|
|
|
|
* XXX
|
|
|
|
|
* This may not need to be here, logically it should go down with
|
|
|
|
|
* the i_devvp initialization.
|
|
|
|
|
* Ask Kirk.
|
|
|
|
|
*/
|
|
|
|
|
ip->i_lfs = ump->um_lfs;
|
|
|
|
|
|
|
|
|
|
/* Read in the disk contents for the inode, copy into the inode. */
|
1996-02-10 01:28:45 +03:00
|
|
|
if (dinp) {
|
1998-10-23 04:32:35 +04:00
|
|
|
error = copyin(dinp, &ip->i_din.ffs_din, DINODE_SIZE);
|
1999-03-10 03:20:00 +03:00
|
|
|
if (error) {
|
1999-03-26 00:54:10 +03:00
|
|
|
printf("lfs_fastvget: dinode copyin failed for ino %d\n", ino);
|
1999-03-10 03:20:00 +03:00
|
|
|
ufs_ihashrem(ip);
|
|
|
|
|
|
|
|
|
|
/* Unlock and discard unneeded inode. */
|
1999-07-08 05:05:58 +04:00
|
|
|
lockmgr(&vp->v_lock, LK_RELEASE, &vp->v_interlock);
|
1999-03-10 03:20:00 +03:00
|
|
|
lfs_vunref(vp);
|
|
|
|
|
*vpp = NULL;
|
1994-06-08 15:41:58 +04:00
|
|
|
return (error);
|
1999-03-10 03:20:00 +03:00
|
|
|
}
|
|
|
|
|
if(ip->i_number != ino)
|
|
|
|
|
panic("lfs_fastvget: I was fed the wrong inode!");
|
|
|
|
|
} else {
|
1996-02-10 01:28:45 +03:00
|
|
|
error = bread(ump->um_devvp, daddr,
|
|
|
|
|
(int)ump->um_lfs->lfs_bsize, NOCRED, &bp);
|
|
|
|
|
if (error) {
|
1999-04-12 04:40:06 +04:00
|
|
|
printf("lfs_fastvget: bread failed with %d\n",error);
|
1994-06-08 15:41:58 +04:00
|
|
|
/*
|
|
|
|
|
* The inode does not contain anything useful, so it
|
|
|
|
|
* would be misleading to leave it on its hash chain.
|
|
|
|
|
* Iput() will return it to the free list.
|
|
|
|
|
*/
|
|
|
|
|
ufs_ihashrem(ip);
|
1999-03-10 03:20:00 +03:00
|
|
|
|
1994-06-08 15:41:58 +04:00
|
|
|
/* Unlock and discard unneeded inode. */
|
1999-07-08 05:05:58 +04:00
|
|
|
lockmgr(&vp->v_lock, LK_RELEASE, &vp->v_interlock);
|
1994-06-08 15:41:58 +04:00
|
|
|
lfs_vunref(vp);
|
|
|
|
|
brelse(bp);
|
|
|
|
|
*vpp = NULL;
|
|
|
|
|
return (error);
|
|
|
|
|
}
|
1997-06-11 14:09:37 +04:00
|
|
|
ip->i_din.ffs_din =
|
2000-01-19 03:03:04 +03:00
|
|
|
*lfs_ifind(ump->um_lfs, ino, bp);
|
1994-06-08 15:41:58 +04:00
|
|
|
brelse(bp);
|
|
|
|
|
}
|
1999-11-21 22:25:31 +03:00
|
|
|
ip->i_ffs_effnlink = ip->i_ffs_nlink;
|
2000-10-21 17:53:25 +04:00
|
|
|
ip->i_lfs_effnblks = ip->i_ffs_blocks;
|
1994-06-08 15:41:58 +04:00
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Initialize the vnode from the inode, check for aliases. In all
|
|
|
|
|
* cases re-init ip, the underlying vnode/inode may have changed.
|
|
|
|
|
*/
|
1998-06-25 00:58:44 +04:00
|
|
|
error = ufs_vinit(mp, lfs_specop_p, lfs_fifoop_p, &vp);
|
1996-02-10 01:28:45 +03:00
|
|
|
if (error) {
|
1999-04-12 04:11:01 +04:00
|
|
|
/* This CANNOT happen (see ufs_vinit) */
|
|
|
|
|
printf("lfs_fastvget: ufs_vinit returned %d for ino %d\n", error, ino);
|
1999-07-08 05:05:58 +04:00
|
|
|
lockmgr(&vp->v_lock, LK_RELEASE, &vp->v_interlock);
|
1994-06-08 15:41:58 +04:00
|
|
|
lfs_vunref(vp);
|
|
|
|
|
*vpp = NULL;
|
|
|
|
|
return (error);
|
|
|
|
|
}
|
1999-03-10 03:20:00 +03:00
|
|
|
#ifdef DEBUG_LFS
|
|
|
|
|
if(vp->v_type == VNON) {
|
|
|
|
|
printf("lfs_fastvget: ino %d is type VNON! (ifmt=%o, dinp=%p)\n",
|
|
|
|
|
ip->i_number, (ip->i_ffs_mode & IFMT)>>12, dinp);
|
|
|
|
|
lfs_dump_dinode(&ip->i_din.ffs_din);
|
|
|
|
|
#ifdef DDB
|
|
|
|
|
Debugger();
|
|
|
|
|
#endif
|
|
|
|
|
}
|
|
|
|
|
#endif /* DEBUG_LFS */
|
1994-06-08 15:41:58 +04:00
|
|
|
/*
|
|
|
|
|
* Finish inode initialization now that aliasing has been resolved.
|
|
|
|
|
*/
|
|
|
|
|
ip->i_devvp = ump->um_devvp;
|
|
|
|
|
VREF(ip->i_devvp);
|
|
|
|
|
*vpp = vp;
|
1999-03-10 03:20:00 +03:00
|
|
|
*need_unlock |= FVG_PUT;
|
|
|
|
|
|
2000-12-03 10:34:49 +03:00
|
|
|
uvm_vnp_setsize(vp, ip->i_ffs_size);
|
|
|
|
|
|
1994-06-08 15:41:58 +04:00
|
|
|
return (0);
|
|
|
|
|
}
|
1999-03-10 03:20:00 +03:00
|
|
|
|
1994-06-08 15:41:58 +04:00
|
|
|
struct buf *
|
|
|
|
|
lfs_fakebuf(vp, lbn, size, uaddr)
|
|
|
|
|
struct vnode *vp;
|
|
|
|
|
int lbn;
|
|
|
|
|
size_t size;
|
|
|
|
|
caddr_t uaddr;
|
|
|
|
|
{
|
|
|
|
|
struct buf *bp;
|
1999-03-26 01:26:52 +03:00
|
|
|
int error;
|
1999-03-10 03:20:00 +03:00
|
|
|
|
1999-03-26 01:26:52 +03:00
|
|
|
#ifndef ALLOW_VFLUSH_CORRUPTION
|
|
|
|
|
bp = lfs_newbuf(vp, lbn, size);
|
|
|
|
|
error = copyin(uaddr, bp->b_data, size);
|
|
|
|
|
if(error) {
|
|
|
|
|
lfs_freebuf(bp);
|
|
|
|
|
return NULL;
|
1999-03-10 03:20:00 +03:00
|
|
|
}
|
1999-03-26 01:26:52 +03:00
|
|
|
#else
|
1994-06-08 15:41:58 +04:00
|
|
|
bp = lfs_newbuf(vp, lbn, 0);
|
1999-03-26 01:26:52 +03:00
|
|
|
bp->b_flags |= B_INVAL;
|
1994-06-08 15:41:58 +04:00
|
|
|
bp->b_saveaddr = uaddr;
|
1999-03-26 01:26:52 +03:00
|
|
|
#endif
|
|
|
|
|
|
1994-06-08 15:41:58 +04:00
|
|
|
bp->b_bufsize = size;
|
|
|
|
|
bp->b_bcount = size;
|
|
|
|
|
return (bp);
|
|
|
|
|
}
|