Aren
/
NetBSD
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
NetBSD/sys/ufs/lfs/lfs_bio.c

742 lines
20 KiB
C
Raw Normal View History

Tabify, and fix some comment alignment problems.
2003-02-20 07:27:23 +03:00
/* $NetBSD: lfs_bio.c,v 1.61 2003/02/20 04:27:23 perseant Exp $ */
New RCS ID's, take two. they're more aesthecially pleasant, and use 'NetBSD'
1994-06-29 10:39:25 +04:00
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
/*-
Add code to UBCify LFS. This is still behind "#ifdef LFS_UBC" for now (there are still some details to work out) but expect that to go away soon. To support these basic changes (creation of lfs_putpages, lfs_gop_write, mods to lfs_balloc) several other changes were made, to wit: * Create a writer daemon kernel thread whose purpose is to handle page writes for the pagedaemon, but which also takes over some of the functions of lfs_check(). This thread is started the first time an LFS is mounted. * Add a "flags" parameter to GOP_SIZE. Current values are GOP_SIZE_READ, meaning that the call should return the size of the in-core version of the file, and GOP_SIZE_WRITE, meaning that it should return the on-disk size. One of GOP_SIZE_READ or GOP_SIZE_WRITE must be specified. * Instead of using malloc(...M_WAITOK) for everything, reserve enough resources to get by and use malloc(...M_NOWAIT), using the reserves if necessary. Use the pool subsystem for structures small enough that this is feasible. This also obsoletes LFS_THROTTLE. And a few that are not strictly necessary: * Moves the LFS inode extensions off onto a separately allocated structure; getting closer to LFS as an LKM. "Welcome to 1.6O." * Unified GOP_ALLOC between FFS and LFS. * Update LFS copyright headers to correct values. * Actually cast to unsigned in lfs_shellsort, like the comment says. * Keep track of which segments were empty before the previous checkpoint; any segments that pass two checkpoints both dirty and empty can be summarily cleaned. Do this. Right now lfs_segclean still works, but this should be turned into an effectless compatibility syscall.
2003-02-18 02:48:08 +03:00
* Copyright (c) 1999, 2000, 2001, 2002, 2003 The NetBSD Foundation, Inc.
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
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:
Tabify, and fix some comment alignment problems.
2003-02-20 07:27:23 +03:00
* This product includes software developed by the NetBSD
* Foundation, Inc. and its contributors.
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
* 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.
*/
Update to 4.4-Lite fs code, with local changes.
1994-06-08 15:41:58 +04:00
/*
* Copyright (c) 1991, 1993
* 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.
*
Merge with Lite2 + local changes
1998-03-01 05:20:01 +03:00
* @(#)lfs_bio.c 8.10 (Berkeley) 6/10/95
Update to 4.4-Lite fs code, with local changes.
1994-06-08 15:41:58 +04:00
*/
add RCSID
2001-11-08 05:39:06 +03:00
#include <sys/cdefs.h>
Tabify, and fix some comment alignment problems.
2003-02-20 07:27:23 +03:00
__KERNEL_RCSID(0, "$NetBSD: lfs_bio.c,v 1.61 2003/02/20 04:27:23 perseant Exp $");
add RCSID
2001-11-08 05:39:06 +03:00
Update to 4.4-Lite fs code, with local changes.
1994-06-08 15:41:58 +04:00
#include <sys/param.h>
lfs prototypes
1996-02-10 01:28:45 +03:00
#include <sys/systm.h>
Update to 4.4-Lite fs code, with local changes.
1994-06-08 15:41:58 +04:00
#include <sys/proc.h>
#include <sys/buf.h>
#include <sys/vnode.h>
#include <sys/resourcevar.h>
#include <sys/mount.h>
#include <sys/kernel.h>
#include <ufs/ufs/inode.h>
#include <ufs/ufs/ufsmount.h>
lfs prototypes
1996-02-10 01:28:45 +03:00
#include <ufs/ufs/ufs_extern.h>
Update to 4.4-Lite fs code, with local changes.
1994-06-08 15:41:58 +04:00
#include <ufs/lfs/lfs.h>
#include <ufs/lfs/lfs_extern.h>
Add code to UBCify LFS. This is still behind "#ifdef LFS_UBC" for now (there are still some details to work out) but expect that to go away soon. To support these basic changes (creation of lfs_putpages, lfs_gop_write, mods to lfs_balloc) several other changes were made, to wit: * Create a writer daemon kernel thread whose purpose is to handle page writes for the pagedaemon, but which also takes over some of the functions of lfs_check(). This thread is started the first time an LFS is mounted. * Add a "flags" parameter to GOP_SIZE. Current values are GOP_SIZE_READ, meaning that the call should return the size of the in-core version of the file, and GOP_SIZE_WRITE, meaning that it should return the on-disk size. One of GOP_SIZE_READ or GOP_SIZE_WRITE must be specified. * Instead of using malloc(...M_WAITOK) for everything, reserve enough resources to get by and use malloc(...M_NOWAIT), using the reserves if necessary. Use the pool subsystem for structures small enough that this is feasible. This also obsoletes LFS_THROTTLE. And a few that are not strictly necessary: * Moves the LFS inode extensions off onto a separately allocated structure; getting closer to LFS as an LKM. "Welcome to 1.6O." * Unified GOP_ALLOC between FFS and LFS. * Update LFS copyright headers to correct values. * Actually cast to unsigned in lfs_shellsort, like the comment says. * Keep track of which segments were empty before the previous checkpoint; any segments that pass two checkpoints both dirty and empty can be summarily cleaned. Do this. Right now lfs_segclean still works, but this should be turned into an effectless compatibility syscall.
2003-02-18 02:48:08 +03:00
#include <uvm/uvm.h>
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
/* Macros to clear/set/test flags. */
# define SET(t, f) (t) |= (f)
# define CLR(t, f) (t) &= ~(f)
# define ISSET(t, f) ((t) & (f))
Update to 4.4-Lite fs code, with local changes.
1994-06-08 15:41:58 +04:00
/*
* LFS block write function.
*
* XXX
* No write cost accounting is done.
* This is almost certainly wrong for synchronous operations and NFS.
*/
Add code to UBCify LFS. This is still behind "#ifdef LFS_UBC" for now (there are still some details to work out) but expect that to go away soon. To support these basic changes (creation of lfs_putpages, lfs_gop_write, mods to lfs_balloc) several other changes were made, to wit: * Create a writer daemon kernel thread whose purpose is to handle page writes for the pagedaemon, but which also takes over some of the functions of lfs_check(). This thread is started the first time an LFS is mounted. * Add a "flags" parameter to GOP_SIZE. Current values are GOP_SIZE_READ, meaning that the call should return the size of the in-core version of the file, and GOP_SIZE_WRITE, meaning that it should return the on-disk size. One of GOP_SIZE_READ or GOP_SIZE_WRITE must be specified. * Instead of using malloc(...M_WAITOK) for everything, reserve enough resources to get by and use malloc(...M_NOWAIT), using the reserves if necessary. Use the pool subsystem for structures small enough that this is feasible. This also obsoletes LFS_THROTTLE. And a few that are not strictly necessary: * Moves the LFS inode extensions off onto a separately allocated structure; getting closer to LFS as an LKM. "Welcome to 1.6O." * Unified GOP_ALLOC between FFS and LFS. * Update LFS copyright headers to correct values. * Actually cast to unsigned in lfs_shellsort, like the comment says. * Keep track of which segments were empty before the previous checkpoint; any segments that pass two checkpoints both dirty and empty can be summarily cleaned. Do this. Right now lfs_segclean still works, but this should be turned into an effectless compatibility syscall.
2003-02-18 02:48:08 +03:00
int locked_queue_count = 0; /* Count of locked-down buffers. */
long locked_queue_bytes = 0L; /* Total size of locked buffers. */
Tabify, and fix some comment alignment problems.
2003-02-20 07:27:23 +03:00
int lfs_subsys_pages = 0L; /* Total number LFS-written pages */
int lfs_writing = 0; /* Set if already kicked off a writer
Update to 4.4-Lite fs code, with local changes.
1994-06-08 15:41:58 +04:00
because of buffer space */
Add code to UBCify LFS. This is still behind "#ifdef LFS_UBC" for now (there are still some details to work out) but expect that to go away soon. To support these basic changes (creation of lfs_putpages, lfs_gop_write, mods to lfs_balloc) several other changes were made, to wit: * Create a writer daemon kernel thread whose purpose is to handle page writes for the pagedaemon, but which also takes over some of the functions of lfs_check(). This thread is started the first time an LFS is mounted. * Add a "flags" parameter to GOP_SIZE. Current values are GOP_SIZE_READ, meaning that the call should return the size of the in-core version of the file, and GOP_SIZE_WRITE, meaning that it should return the on-disk size. One of GOP_SIZE_READ or GOP_SIZE_WRITE must be specified. * Instead of using malloc(...M_WAITOK) for everything, reserve enough resources to get by and use malloc(...M_NOWAIT), using the reserves if necessary. Use the pool subsystem for structures small enough that this is feasible. This also obsoletes LFS_THROTTLE. And a few that are not strictly necessary: * Moves the LFS inode extensions off onto a separately allocated structure; getting closer to LFS as an LKM. "Welcome to 1.6O." * Unified GOP_ALLOC between FFS and LFS. * Update LFS copyright headers to correct values. * Actually cast to unsigned in lfs_shellsort, like the comment says. * Keep track of which segments were empty before the previous checkpoint; any segments that pass two checkpoints both dirty and empty can be summarily cleaned. Do this. Right now lfs_segclean still works, but this should be turned into an effectless compatibility syscall.
2003-02-18 02:48:08 +03:00
struct simplelock lfs_subsys_lock; /* Lock on subsys_pages */
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
extern int lfs_dostats;
Add code to UBCify LFS. This is still behind "#ifdef LFS_UBC" for now (there are still some details to work out) but expect that to go away soon. To support these basic changes (creation of lfs_putpages, lfs_gop_write, mods to lfs_balloc) several other changes were made, to wit: * Create a writer daemon kernel thread whose purpose is to handle page writes for the pagedaemon, but which also takes over some of the functions of lfs_check(). This thread is started the first time an LFS is mounted. * Add a "flags" parameter to GOP_SIZE. Current values are GOP_SIZE_READ, meaning that the call should return the size of the in-core version of the file, and GOP_SIZE_WRITE, meaning that it should return the on-disk size. One of GOP_SIZE_READ or GOP_SIZE_WRITE must be specified. * Instead of using malloc(...M_WAITOK) for everything, reserve enough resources to get by and use malloc(...M_NOWAIT), using the reserves if necessary. Use the pool subsystem for structures small enough that this is feasible. This also obsoletes LFS_THROTTLE. And a few that are not strictly necessary: * Moves the LFS inode extensions off onto a separately allocated structure; getting closer to LFS as an LKM. "Welcome to 1.6O." * Unified GOP_ALLOC between FFS and LFS. * Update LFS copyright headers to correct values. * Actually cast to unsigned in lfs_shellsort, like the comment says. * Keep track of which segments were empty before the previous checkpoint; any segments that pass two checkpoints both dirty and empty can be summarily cleaned. Do this. Right now lfs_segclean still works, but this should be turned into an effectless compatibility syscall.
2003-02-18 02:48:08 +03:00
extern int lfs_do_flush;
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
- in lfs_reserve, reserve locked buffer count as well. - don't wait for locking buf in lfs_bwrite_ext to avoid deadlocks. - skip lfs_reserve when we're doing dirop. reserve more (for lfs_truncate) in set_dirop instead. this mostly solves PR 18972. (and hopefully PR 19196)
2002-12-26 16:37:18 +03:00
/*
* reserved number/bytes of locked buffers
*/
int locked_queue_rcount = 0;
long locked_queue_rbytes = 0L;
int lfs_fits_buf(struct lfs *, int, int);
- in lfs_reserve, vref vnodes that we're locking so that cleaner doesn't try to reclaim them. (workaround for deadlock noted in the comment in lfs_reserveavail) - in lfs_rename, mark vnodes which are being moved as well as directry vnodes.
2002-12-28 17:39:08 +03:00
int lfs_reservebuf(struct lfs *, struct vnode *vp, struct vnode *vp2,
int, int);
int lfs_reserveavail(struct lfs *, struct vnode *vp, struct vnode *vp2, int);
- in lfs_reserve, reserve locked buffer count as well. - don't wait for locking buf in lfs_bwrite_ext to avoid deadlocks. - skip lfs_reserve when we're doing dirop. reserve more (for lfs_truncate) in set_dirop instead. this mostly solves PR 18972. (and hopefully PR 19196)
2002-12-26 16:37:18 +03:00
int
lfs_fits_buf(struct lfs *fs, int n, int bytes)
{
int count_fit =
(locked_queue_count + locked_queue_rcount + n < LFS_WAIT_BUFS);
int bytes_fit =
(locked_queue_bytes + locked_queue_rbytes + bytes < LFS_WAIT_BYTES);
#ifdef DEBUG_LFS
if (!count_fit) {
printf("lfs_fits_buf: no fit count: %d + %d + %d >= %d\n",
locked_queue_count, locked_queue_rcount,
n, LFS_WAIT_BUFS);
}
if (!bytes_fit) {
printf("lfs_fits_buf: no fit bytes: %ld + %ld + %d >= %d\n",
locked_queue_bytes, locked_queue_rbytes,
bytes, LFS_WAIT_BYTES);
}
#endif /* DEBUG_LFS */
return (count_fit && bytes_fit);
}
- in lfs_reserve, vref vnodes that we're locking so that cleaner doesn't try to reclaim them. (workaround for deadlock noted in the comment in lfs_reserveavail) - in lfs_rename, mark vnodes which are being moved as well as directry vnodes.
2002-12-28 17:39:08 +03:00
/* ARGSUSED */
- in lfs_reserve, reserve locked buffer count as well. - don't wait for locking buf in lfs_bwrite_ext to avoid deadlocks. - skip lfs_reserve when we're doing dirop. reserve more (for lfs_truncate) in set_dirop instead. this mostly solves PR 18972. (and hopefully PR 19196)
2002-12-26 16:37:18 +03:00
int
- in lfs_reserve, vref vnodes that we're locking so that cleaner doesn't try to reclaim them. (workaround for deadlock noted in the comment in lfs_reserveavail) - in lfs_rename, mark vnodes which are being moved as well as directry vnodes.
2002-12-28 17:39:08 +03:00
lfs_reservebuf(struct lfs *fs, struct vnode *vp, struct vnode *vp2,
int n, int bytes)
- in lfs_reserve, reserve locked buffer count as well. - don't wait for locking buf in lfs_bwrite_ext to avoid deadlocks. - skip lfs_reserve when we're doing dirop. reserve more (for lfs_truncate) in set_dirop instead. this mostly solves PR 18972. (and hopefully PR 19196)
2002-12-26 16:37:18 +03:00
{
KASSERT(locked_queue_rcount >= 0);
KASSERT(locked_queue_rbytes >= 0);
while (n > 0 && !lfs_fits_buf(fs, n, bytes)) {
int error;
++fs->lfs_writer;
lfs_flush(fs, 0);
if (--fs->lfs_writer == 0)
wakeup(&fs->lfs_dirops);
error = tsleep(&locked_queue_count, PCATCH | PUSER,
"lfsresbuf", hz * LFS_BUFWAIT);
if (error && error != EWOULDBLOCK)
return error;
}
locked_queue_rcount += n;
locked_queue_rbytes += bytes;
KASSERT(locked_queue_rcount >= 0);
KASSERT(locked_queue_rbytes >= 0);
return 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
/*
* Try to reserve some blocks, prior to performing a sensitive operation that
* requires the vnode lock to be honored. If there is not enough space, give
* up the vnode lock temporarily and wait for the space to become available.
*
Fix miscalculation in lfs_fits found by Trevin Beattie <trevin@xmission.com>. Change some of the variable names from "nb", "db" to "fsb" to reflect their calling conventions.
2002-06-21 02:10:24 +04:00
* Called with vp locked. (Note nowever that if fsb < 0, vp is ignored.)
add a XXX comment to lfs_reserve. * it isn't safe to unlock vp here * because we're passing data using inode from namei. * (eg. i_offset)
2002-11-24 19:09:50 +03:00
*
more XXX comment.
2002-11-27 14:36:40 +03:00
* XXX YAMT - it isn't safe to unlock vp here
* because the node might be modified while we sleep.
* (eg. cached states like i_offset might be stale,
* the vnode might be truncated, etc..)
comment and assertions
2002-12-30 08:34:17 +03:00
* maybe we should have a way to restart the vnodeop (EVOPRESTART?)
* or rearrange vnodeop interface to leave vnode locking to file system
* specific code so that each file systems can have their own vnode locking and
* vnode re-using strategies.
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
*/
int
- in lfs_reserve, vref vnodes that we're locking so that cleaner doesn't try to reclaim them. (workaround for deadlock noted in the comment in lfs_reserveavail) - in lfs_rename, mark vnodes which are being moved as well as directry vnodes.
2002-12-28 17:39:08 +03:00
lfs_reserveavail(struct lfs *fs, struct vnode *vp, struct vnode *vp2, int fsb)
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
{
CLEANERINFO *cip;
struct buf *bp;
int error, slept;
slept = 0;
move check of lfs_unlockvp from lfs_reserveavail to lfs_reserve because lfs_reservebuf needs same check as well.
2002-12-30 08:31:53 +03:00
while (fsb > 0 && !lfs_fits(fs, fsb + fs->lfs_ravail)) {
#if 0 out vnode unlock/lock in lfs_reserve for now and add a comment about it. deadlock is better than corruption (or panic), IMO.
2002-12-17 18:23:37 +03:00
#if 0
/*
* XXX ideally, we should unlock vnodes here
* because we might sleep very long time.
*/
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
VOP_UNLOCK(vp, 0);
- in lfs_reserve, vref vnodes that we're locking so that cleaner doesn't try to reclaim them. (workaround for deadlock noted in the comment in lfs_reserveavail) - in lfs_rename, mark vnodes which are being moved as well as directry vnodes.
2002-12-28 17:39:08 +03:00
if (vp2 != NULL) {
VOP_UNLOCK(vp2, 0);
}
add a XXX comment. (description of possible deadlock)
2002-12-22 20:31:52 +03:00
#else
/*
* XXX since we'll sleep for cleaner with vnode lock holding,
comment and assertions
2002-12-30 08:34:17 +03:00
* deadlock will occur if cleaner tries to lock the vnode.
add a XXX comment. (description of possible deadlock)
2002-12-22 20:31:52 +03:00
* (eg. lfs_markv -> lfs_fastvget -> getnewvnode -> vclean)
*/
#if 0 out vnode unlock/lock in lfs_reserve for now and add a comment about it. deadlock is better than corruption (or panic), IMO.
2002-12-17 18:23:37 +03:00
#endif
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
Cast back to int32_t in LFS_EST_BFREE and LFS_EST_RSVD macros, for consistency with their arguments. Change the debugging printf in lfs_reserve to match, and enclose it in #ifdef DEBUG. Tested on alpha, arm32, sparc.
2000-09-13 04:07:56 +04:00
if (!slept) {
#ifdef DEBUG
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
printf("lfs_reserve: waiting for %ld (bfree = %d,"
Cast back to int32_t in LFS_EST_BFREE and LFS_EST_RSVD macros, for consistency with their arguments. Change the debugging printf in lfs_reserve to match, and enclose it in #ifdef DEBUG. Tested on alpha, arm32, sparc.
2000-09-13 04:07:56 +04:00
" est_bfree = %d)\n",
Fix miscalculation in lfs_fits found by Trevin Beattie <trevin@xmission.com>. Change some of the variable names from "nb", "db" to "fsb" to reflect their calling conventions.
2002-06-21 02:10:24 +04:00
fsb + fs->lfs_ravail, fs->lfs_bfree,
Cast back to int32_t in LFS_EST_BFREE and LFS_EST_RSVD macros, for consistency with their arguments. Change the debugging printf in lfs_reserve to match, and enclose it in #ifdef DEBUG. Tested on alpha, arm32, sparc.
2000-09-13 04:07:56 +04:00
LFS_EST_BFREE(fs));
#endif
}
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
++slept;
/* Wake up the cleaner */
LFS_CLEANERINFO(cip, fs, bp);
Do not needlessly dirty segment table blocks during lfs_segwrite, preventing needless disk activity when the filesystem is idle. (PR #10979.)
2000-11-12 10:58:36 +03:00
LFS_SYNC_CLEANERINFO(cip, fs, bp, 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
wakeup(&lfs_allclean_wakeup);
wakeup(&fs->lfs_nextseg);
error = tsleep(&fs->lfs_avail, PCATCH | PUSER, "lfs_reserve",
0);
#if 0 out vnode unlock/lock in lfs_reserve for now and add a comment about it. deadlock is better than corruption (or panic), IMO.
2002-12-17 18:23:37 +03:00
#if 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
vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX use lockstatus */
- in lfs_reserve, vref vnodes that we're locking so that cleaner doesn't try to reclaim them. (workaround for deadlock noted in the comment in lfs_reserveavail) - in lfs_rename, mark vnodes which are being moved as well as directry vnodes.
2002-12-28 17:39:08 +03:00
vn_lock(vp2, LK_EXCLUSIVE | LK_RETRY); /* XXX use lockstatus */
#if 0 out vnode unlock/lock in lfs_reserve for now and add a comment about it. deadlock is better than corruption (or panic), IMO.
2002-12-17 18:23:37 +03:00
#endif
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
if (error)
return error;
}
Fix miscalculation in lfs_fits found by Trevin Beattie <trevin@xmission.com>. Change some of the variable names from "nb", "db" to "fsb" to reflect their calling conventions.
2002-06-21 02:10:24 +04:00
#ifdef DEBUG
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
if (slept)
printf("lfs_reserve: woke up\n");
Fix miscalculation in lfs_fits found by Trevin Beattie <trevin@xmission.com>. Change some of the variable names from "nb", "db" to "fsb" to reflect their calling conventions.
2002-06-21 02:10:24 +04:00
#endif
fs->lfs_ravail += fsb;
- in lfs_reserve, reserve locked buffer count as well. - don't wait for locking buf in lfs_bwrite_ext to avoid deadlocks. - skip lfs_reserve when we're doing dirop. reserve more (for lfs_truncate) in set_dirop instead. this mostly solves PR 18972. (and hopefully PR 19196)
2002-12-26 16:37:18 +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
return 0;
}
- in lfs_reserve, reserve locked buffer count as well. - don't wait for locking buf in lfs_bwrite_ext to avoid deadlocks. - skip lfs_reserve when we're doing dirop. reserve more (for lfs_truncate) in set_dirop instead. this mostly solves PR 18972. (and hopefully PR 19196)
2002-12-26 16:37:18 +03:00
#ifdef DIAGNOSTIC
int lfs_rescount;
int lfs_rescountdirop;
#endif
int
- in lfs_reserve, vref vnodes that we're locking so that cleaner doesn't try to reclaim them. (workaround for deadlock noted in the comment in lfs_reserveavail) - in lfs_rename, mark vnodes which are being moved as well as directry vnodes.
2002-12-28 17:39:08 +03:00
lfs_reserve(struct lfs *fs, struct vnode *vp, struct vnode *vp2, int fsb)
- in lfs_reserve, reserve locked buffer count as well. - don't wait for locking buf in lfs_bwrite_ext to avoid deadlocks. - skip lfs_reserve when we're doing dirop. reserve more (for lfs_truncate) in set_dirop instead. this mostly solves PR 18972. (and hopefully PR 19196)
2002-12-26 16:37:18 +03:00
{
int error;
int cantwait;
- in lfs_reserve, vref vnodes that we're locking so that cleaner doesn't try to reclaim them. (workaround for deadlock noted in the comment in lfs_reserveavail) - in lfs_rename, mark vnodes which are being moved as well as directry vnodes.
2002-12-28 17:39:08 +03:00
KASSERT(fsb < 0 || VOP_ISLOCKED(vp));
KASSERT(vp2 == NULL || fsb < 0 || VOP_ISLOCKED(vp2));
comment and assertions
2002-12-30 08:34:17 +03:00
KASSERT(vp2 == NULL || !(VTOI(vp2)->i_flag & IN_ADIROP));
KASSERT(vp2 == NULL || vp2 != fs->lfs_unlockvp);
- in lfs_reserve, vref vnodes that we're locking so that cleaner doesn't try to reclaim them. (workaround for deadlock noted in the comment in lfs_reserveavail) - in lfs_rename, mark vnodes which are being moved as well as directry vnodes.
2002-12-28 17:39:08 +03:00
move check of lfs_unlockvp from lfs_reserveavail to lfs_reserve because lfs_reservebuf needs same check as well.
2002-12-30 08:31:53 +03:00
cantwait = (VTOI(vp)->i_flag & IN_ADIROP) || fs->lfs_unlockvp == vp;
- in lfs_reserve, reserve locked buffer count as well. - don't wait for locking buf in lfs_bwrite_ext to avoid deadlocks. - skip lfs_reserve when we're doing dirop. reserve more (for lfs_truncate) in set_dirop instead. this mostly solves PR 18972. (and hopefully PR 19196)
2002-12-26 16:37:18 +03:00
#ifdef DIAGNOSTIC
if (cantwait) {
if (fsb > 0)
lfs_rescountdirop++;
else if (fsb < 0)
lfs_rescountdirop--;
if (lfs_rescountdirop < 0)
panic("lfs_rescountdirop");
}
else {
if (fsb > 0)
lfs_rescount++;
else if (fsb < 0)
lfs_rescount--;
if (lfs_rescount < 0)
panic("lfs_rescount");
}
#endif
if (cantwait)
return 0;
fix vref/vunref mismatch.
2002-12-29 17:08:12 +03:00
/*
* XXX
* vref vnodes here so that cleaner doesn't try to reuse them.
comment and assertions
2002-12-30 08:34:17 +03:00
* (see XXX comment in lfs_reserveavail)
fix vref/vunref mismatch.
2002-12-29 17:08:12 +03:00
*/
lfs_vref(vp);
if (vp2 != NULL) {
lfs_vref(vp2);
}
- in lfs_reserve, vref vnodes that we're locking so that cleaner doesn't try to reclaim them. (workaround for deadlock noted in the comment in lfs_reserveavail) - in lfs_rename, mark vnodes which are being moved as well as directry vnodes.
2002-12-28 17:39:08 +03:00
error = lfs_reserveavail(fs, vp, vp2, fsb);
- in lfs_reserve, reserve locked buffer count as well. - don't wait for locking buf in lfs_bwrite_ext to avoid deadlocks. - skip lfs_reserve when we're doing dirop. reserve more (for lfs_truncate) in set_dirop instead. this mostly solves PR 18972. (and hopefully PR 19196)
2002-12-26 16:37:18 +03:00
if (error)
- in lfs_reserve, vref vnodes that we're locking so that cleaner doesn't try to reclaim them. (workaround for deadlock noted in the comment in lfs_reserveavail) - in lfs_rename, mark vnodes which are being moved as well as directry vnodes.
2002-12-28 17:39:08 +03:00
goto done;
- in lfs_reserve, reserve locked buffer count as well. - don't wait for locking buf in lfs_bwrite_ext to avoid deadlocks. - skip lfs_reserve when we're doing dirop. reserve more (for lfs_truncate) in set_dirop instead. this mostly solves PR 18972. (and hopefully PR 19196)
2002-12-26 16:37:18 +03:00
/*
* XXX just a guess. should be more precise.
*/
- in lfs_reserve, vref vnodes that we're locking so that cleaner doesn't try to reclaim them. (workaround for deadlock noted in the comment in lfs_reserveavail) - in lfs_rename, mark vnodes which are being moved as well as directry vnodes.
2002-12-28 17:39:08 +03:00
error = lfs_reservebuf(fs, vp, vp2,
fragstoblks(fs, fsb), fsbtob(fs, fsb));
- in lfs_reserve, reserve locked buffer count as well. - don't wait for locking buf in lfs_bwrite_ext to avoid deadlocks. - skip lfs_reserve when we're doing dirop. reserve more (for lfs_truncate) in set_dirop instead. this mostly solves PR 18972. (and hopefully PR 19196)
2002-12-26 16:37:18 +03:00
if (error)
- in lfs_reserve, vref vnodes that we're locking so that cleaner doesn't try to reclaim them. (workaround for deadlock noted in the comment in lfs_reserveavail) - in lfs_rename, mark vnodes which are being moved as well as directry vnodes.
2002-12-28 17:39:08 +03:00
lfs_reserveavail(fs, vp, vp2, -fsb);
done:
lfs_vunref(vp);
if (vp2 != NULL) {
lfs_vunref(vp2);
}
- in lfs_reserve, reserve locked buffer count as well. - don't wait for locking buf in lfs_bwrite_ext to avoid deadlocks. - skip lfs_reserve when we're doing dirop. reserve more (for lfs_truncate) in set_dirop instead. this mostly solves PR 18972. (and hopefully PR 19196)
2002-12-26 16:37:18 +03:00
return error;
}
Update to 4.4-Lite fs code, with local changes.
1994-06-08 15:41:58 +04:00
int
Merge the short-lived perseant-lfsv2 branch into the trunk. Kernels and tools understand both v1 and v2 filesystems; newfs_lfs generates v2 by default. Changes for the v2 layout include: - Segments of non-PO2 size and arbitrary block offset, so these can be matched to convenient physical characteristics of the partition (e.g., stripe or track size and offset). - Address by fragment instead of by disk sector, paving the way for non-512-byte-sector devices. In theory fragments can be as large as you like, though in reality they must be smaller than MAXBSIZE in size. - Use serial number and filesystem identifier to ensure that roll-forward doesn't get old data and think it's new. Roll-forward is enabled for v2 filesystems, though not for v1 filesystems by default. - The inode free list is now a tailq, paving the way for undelete (undelete is not yet implemented, but can be without further non-backwards-compatible changes to disk structures). - Inode atime information is kept in the Ifile, instead of on the inode; that is, the inode is never written *just* because atime was changed. Because of this the inodes remain near the file data on the disk, rather than wandering all over as the disk is read repeatedly. This speeds up repeated reads by a small but noticeable amount. Other changes of note include: - The ifile written by newfs_lfs can now be of arbitrary length, it is no longer restricted to a single indirect block. - Fixed an old bug where ctime was changed every time a vnode was created. I need to look more closely to make sure that the times are only updated during write(2) and friends, not after-the-fact during a segment write, and certainly not by the cleaner.
2001-07-14 00:30:18 +04:00
lfs_bwrite(void *v)
lfs prototypes
1996-02-10 01:28:45 +03:00
{
Update to 4.4-Lite fs code, with local changes.
1994-06-08 15:41:58 +04:00
struct vop_bwrite_args /* {
struct buf *a_bp;
lfs prototypes
1996-02-10 01:28:45 +03:00
} */ *ap = v;
Remove register declarations.
2000-03-30 16:41:09 +04:00
struct buf *bp = ap->a_bp;
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
#ifdef DIAGNOSTIC
Tabify, and fix some comment alignment problems.
2003-02-20 07:27:23 +03:00
if (VTOI(bp->b_vp)->i_lfs->lfs_ronly == 0 && (bp->b_flags & B_ASYNC)) {
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
panic("bawrite LFS buffer");
In lfs_bwrite, don't mark buffers dirty if lfs is mounted read-only. (Previously buffers could be marked dirty by the cleaner, and possibly by other means.) Also check for softdep mount in vfs_shutdown before trying to bawrite buffers, since other filesystems don't need it and lfs doesn't bawrite. (This fragment reviewed by fvdl.) Partially addresses PR#8964.
1999-12-15 10:10:32 +03:00
}
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
#endif /* DIAGNOSTIC */
return lfs_bwrite_ext(bp,0);
}
/*
Fix miscalculation in lfs_fits found by Trevin Beattie <trevin@xmission.com>. Change some of the variable names from "nb", "db" to "fsb" to reflect their calling conventions.
2002-06-21 02:10:24 +04:00
* Determine if there is enough room currently available to write fsb
* blocks. We need enough blocks for the new blocks, the current
* inode blocks (including potentially the ifile inode), a summary block,
* and the segment usage table, plus an ifile block.
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
*/
Correct accounting of lfs_avail, locked_queue_count, and locked_queue_bytes. (PR #11468). In the case of fragment allocation, check to see if enough space is available before extending a fragment already scheduled for writing. The locked_queue_* variables indicate the number of buffer headers and bytes, respectively, that are unavailable to getnewbuf() because they are locked up waiting for LFS to flush them; make sure that that is actually what we're counting, i.e., never count malloced buffers, and always use b_bufsize instead of b_bcount. If DEBUG is defined, the periodic calls to lfs_countlocked will now complain if either counter is incorrect. (In the future lfs_countlocked will not need to be called at all if DEBUG is not defined.)
2000-11-17 22:14:41 +03:00
int
Merge the short-lived perseant-lfsv2 branch into the trunk. Kernels and tools understand both v1 and v2 filesystems; newfs_lfs generates v2 by default. Changes for the v2 layout include: - Segments of non-PO2 size and arbitrary block offset, so these can be matched to convenient physical characteristics of the partition (e.g., stripe or track size and offset). - Address by fragment instead of by disk sector, paving the way for non-512-byte-sector devices. In theory fragments can be as large as you like, though in reality they must be smaller than MAXBSIZE in size. - Use serial number and filesystem identifier to ensure that roll-forward doesn't get old data and think it's new. Roll-forward is enabled for v2 filesystems, though not for v1 filesystems by default. - The inode free list is now a tailq, paving the way for undelete (undelete is not yet implemented, but can be without further non-backwards-compatible changes to disk structures). - Inode atime information is kept in the Ifile, instead of on the inode; that is, the inode is never written *just* because atime was changed. Because of this the inodes remain near the file data on the disk, rather than wandering all over as the disk is read repeatedly. This speeds up repeated reads by a small but noticeable amount. Other changes of note include: - The ifile written by newfs_lfs can now be of arbitrary length, it is no longer restricted to a single indirect block. - Fixed an old bug where ctime was changed every time a vnode was created. I need to look more closely to make sure that the times are only updated during write(2) and friends, not after-the-fact during a segment write, and certainly not by the cleaner.
2001-07-14 00:30:18 +04:00
lfs_fits(struct lfs *fs, int fsb)
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
{
Clean up accounting of lfs_uinodes (dirty but unwritten inodes). Make lfs_uinodes a signed quantity for debugging purposes, and set it to zero as fs mount time. Enclose setting/clearing of the dirty flags (IN_MODIFIED, IN_ACCESSED, IN_CLEANING) in macros, and use those macros everywhere. Make LFS_ITIMES use these macros; updated the ITIMES macro in inode.h to know about this. Make ufs_getattr use ITIMES instead of FFS_ITIMES.
2000-07-06 02:25:43 +04:00
int needed;
Merge the short-lived perseant-lfsv2 branch into the trunk. Kernels and tools understand both v1 and v2 filesystems; newfs_lfs generates v2 by default. Changes for the v2 layout include: - Segments of non-PO2 size and arbitrary block offset, so these can be matched to convenient physical characteristics of the partition (e.g., stripe or track size and offset). - Address by fragment instead of by disk sector, paving the way for non-512-byte-sector devices. In theory fragments can be as large as you like, though in reality they must be smaller than MAXBSIZE in size. - Use serial number and filesystem identifier to ensure that roll-forward doesn't get old data and think it's new. Roll-forward is enabled for v2 filesystems, though not for v1 filesystems by default. - The inode free list is now a tailq, paving the way for undelete (undelete is not yet implemented, but can be without further non-backwards-compatible changes to disk structures). - Inode atime information is kept in the Ifile, instead of on the inode; that is, the inode is never written *just* because atime was changed. Because of this the inodes remain near the file data on the disk, rather than wandering all over as the disk is read repeatedly. This speeds up repeated reads by a small but noticeable amount. Other changes of note include: - The ifile written by newfs_lfs can now be of arbitrary length, it is no longer restricted to a single indirect block. - Fixed an old bug where ctime was changed every time a vnode was created. I need to look more closely to make sure that the times are only updated during write(2) and friends, not after-the-fact during a segment write, and certainly not by the cleaner.
2001-07-14 00:30:18 +04:00
needed = fsb + btofsb(fs, fs->lfs_sumsize) +
Fix miscalculation in lfs_fits found by Trevin Beattie <trevin@xmission.com>. Change some of the variable names from "nb", "db" to "fsb" to reflect their calling conventions.
2002-06-21 02:10:24 +04:00
((howmany(fs->lfs_uinodes + 1, INOPB(fs)) + fs->lfs_segtabsz +
1) << (fs->lfs_blktodb - fs->lfs_fsbtodb));
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
Clean up accounting of lfs_uinodes (dirty but unwritten inodes). Make lfs_uinodes a signed quantity for debugging purposes, and set it to zero as fs mount time. Enclose setting/clearing of the dirty flags (IN_MODIFIED, IN_ACCESSED, IN_CLEANING) in macros, and use those macros everywhere. Make LFS_ITIMES use these macros; updated the ITIMES macro in inode.h to know about this. Make ufs_getattr use ITIMES instead of FFS_ITIMES.
2000-07-06 02:25:43 +04:00
if (needed >= fs->lfs_avail) {
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
#ifdef DEBUG
Merge the short-lived perseant-lfsv2 branch into the trunk. Kernels and tools understand both v1 and v2 filesystems; newfs_lfs generates v2 by default. Changes for the v2 layout include: - Segments of non-PO2 size and arbitrary block offset, so these can be matched to convenient physical characteristics of the partition (e.g., stripe or track size and offset). - Address by fragment instead of by disk sector, paving the way for non-512-byte-sector devices. In theory fragments can be as large as you like, though in reality they must be smaller than MAXBSIZE in size. - Use serial number and filesystem identifier to ensure that roll-forward doesn't get old data and think it's new. Roll-forward is enabled for v2 filesystems, though not for v1 filesystems by default. - The inode free list is now a tailq, paving the way for undelete (undelete is not yet implemented, but can be without further non-backwards-compatible changes to disk structures). - Inode atime information is kept in the Ifile, instead of on the inode; that is, the inode is never written *just* because atime was changed. Because of this the inodes remain near the file data on the disk, rather than wandering all over as the disk is read repeatedly. This speeds up repeated reads by a small but noticeable amount. Other changes of note include: - The ifile written by newfs_lfs can now be of arbitrary length, it is no longer restricted to a single indirect block. - Fixed an old bug where ctime was changed every time a vnode was created. I need to look more closely to make sure that the times are only updated during write(2) and friends, not after-the-fact during a segment write, and certainly not by the cleaner.
2001-07-14 00:30:18 +04:00
printf("lfs_fits: no fit: fsb = %d, uinodes = %d, "
Clean up accounting of lfs_uinodes (dirty but unwritten inodes). Make lfs_uinodes a signed quantity for debugging purposes, and set it to zero as fs mount time. Enclose setting/clearing of the dirty flags (IN_MODIFIED, IN_ACCESSED, IN_CLEANING) in macros, and use those macros everywhere. Make LFS_ITIMES use these macros; updated the ITIMES macro in inode.h to know about this. Make ufs_getattr use ITIMES instead of FFS_ITIMES.
2000-07-06 02:25:43 +04:00
"needed = %d, avail = %d\n",
Merge the short-lived perseant-lfsv2 branch into the trunk. Kernels and tools understand both v1 and v2 filesystems; newfs_lfs generates v2 by default. Changes for the v2 layout include: - Segments of non-PO2 size and arbitrary block offset, so these can be matched to convenient physical characteristics of the partition (e.g., stripe or track size and offset). - Address by fragment instead of by disk sector, paving the way for non-512-byte-sector devices. In theory fragments can be as large as you like, though in reality they must be smaller than MAXBSIZE in size. - Use serial number and filesystem identifier to ensure that roll-forward doesn't get old data and think it's new. Roll-forward is enabled for v2 filesystems, though not for v1 filesystems by default. - The inode free list is now a tailq, paving the way for undelete (undelete is not yet implemented, but can be without further non-backwards-compatible changes to disk structures). - Inode atime information is kept in the Ifile, instead of on the inode; that is, the inode is never written *just* because atime was changed. Because of this the inodes remain near the file data on the disk, rather than wandering all over as the disk is read repeatedly. This speeds up repeated reads by a small but noticeable amount. Other changes of note include: - The ifile written by newfs_lfs can now be of arbitrary length, it is no longer restricted to a single indirect block. - Fixed an old bug where ctime was changed every time a vnode was created. I need to look more closely to make sure that the times are only updated during write(2) and friends, not after-the-fact during a segment write, and certainly not by the cleaner.
2001-07-14 00:30:18 +04:00
fsb, fs->lfs_uinodes, needed, fs->lfs_avail);
Clean up accounting of lfs_uinodes (dirty but unwritten inodes). Make lfs_uinodes a signed quantity for debugging purposes, and set it to zero as fs mount time. Enclose setting/clearing of the dirty flags (IN_MODIFIED, IN_ACCESSED, IN_CLEANING) in macros, and use those macros everywhere. Make LFS_ITIMES use these macros; updated the ITIMES macro in inode.h to know about this. Make ufs_getattr use ITIMES instead of FFS_ITIMES.
2000-07-06 02:25:43 +04:00
#endif
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
return 0;
}
return 1;
}
Correct accounting of lfs_avail, locked_queue_count, and locked_queue_bytes. (PR #11468). In the case of fragment allocation, check to see if enough space is available before extending a fragment already scheduled for writing. The locked_queue_* variables indicate the number of buffer headers and bytes, respectively, that are unavailable to getnewbuf() because they are locked up waiting for LFS to flush them; make sure that that is actually what we're counting, i.e., never count malloced buffers, and always use b_bufsize instead of b_bcount. If DEBUG is defined, the periodic calls to lfs_countlocked will now complain if either counter is incorrect. (In the future lfs_countlocked will not need to be called at all if DEBUG is not defined.)
2000-11-17 22:14:41 +03:00
int
Fix miscalculation in lfs_fits found by Trevin Beattie <trevin@xmission.com>. Change some of the variable names from "nb", "db" to "fsb" to reflect their calling conventions.
2002-06-21 02:10:24 +04:00
lfs_availwait(struct lfs *fs, int fsb)
Correct accounting of lfs_avail, locked_queue_count, and locked_queue_bytes. (PR #11468). In the case of fragment allocation, check to see if enough space is available before extending a fragment already scheduled for writing. The locked_queue_* variables indicate the number of buffer headers and bytes, respectively, that are unavailable to getnewbuf() because they are locked up waiting for LFS to flush them; make sure that that is actually what we're counting, i.e., never count malloced buffers, and always use b_bufsize instead of b_bcount. If DEBUG is defined, the periodic calls to lfs_countlocked will now complain if either counter is incorrect. (In the future lfs_countlocked will not need to be called at all if DEBUG is not defined.)
2000-11-17 22:14:41 +03:00
{
int error;
CLEANERINFO *cip;
struct buf *cbp;
Fix miscalculation in lfs_fits found by Trevin Beattie <trevin@xmission.com>. Change some of the variable names from "nb", "db" to "fsb" to reflect their calling conventions.
2002-06-21 02:10:24 +04:00
while (!lfs_fits(fs, fsb)) {
Correct accounting of lfs_avail, locked_queue_count, and locked_queue_bytes. (PR #11468). In the case of fragment allocation, check to see if enough space is available before extending a fragment already scheduled for writing. The locked_queue_* variables indicate the number of buffer headers and bytes, respectively, that are unavailable to getnewbuf() because they are locked up waiting for LFS to flush them; make sure that that is actually what we're counting, i.e., never count malloced buffers, and always use b_bufsize instead of b_bcount. If DEBUG is defined, the periodic calls to lfs_countlocked will now complain if either counter is incorrect. (In the future lfs_countlocked will not need to be called at all if DEBUG is not defined.)
2000-11-17 22:14:41 +03:00
/*
* Out of space, need cleaner to run.
* Update the cleaner info, then wake it up.
* Note the cleanerinfo block is on the ifile
* so it CANT_WAIT.
*/
LFS_CLEANERINFO(cip, fs, cbp);
LFS_SYNC_CLEANERINFO(cip, fs, cbp, 0);
printf("lfs_availwait: out of available space, "
"waiting on cleaner\n");
wakeup(&lfs_allclean_wakeup);
wakeup(&fs->lfs_nextseg);
#ifdef DIAGNOSTIC
if (fs->lfs_seglock && fs->lfs_lockpid == curproc->p_pid)
panic("lfs_availwait: deadlock");
#endif
error = tsleep(&fs->lfs_avail, PCATCH | PUSER, "cleaner", 0);
if (error)
return (error);
}
return 0;
}
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
int
Merge the short-lived perseant-lfsv2 branch into the trunk. Kernels and tools understand both v1 and v2 filesystems; newfs_lfs generates v2 by default. Changes for the v2 layout include: - Segments of non-PO2 size and arbitrary block offset, so these can be matched to convenient physical characteristics of the partition (e.g., stripe or track size and offset). - Address by fragment instead of by disk sector, paving the way for non-512-byte-sector devices. In theory fragments can be as large as you like, though in reality they must be smaller than MAXBSIZE in size. - Use serial number and filesystem identifier to ensure that roll-forward doesn't get old data and think it's new. Roll-forward is enabled for v2 filesystems, though not for v1 filesystems by default. - The inode free list is now a tailq, paving the way for undelete (undelete is not yet implemented, but can be without further non-backwards-compatible changes to disk structures). - Inode atime information is kept in the Ifile, instead of on the inode; that is, the inode is never written *just* because atime was changed. Because of this the inodes remain near the file data on the disk, rather than wandering all over as the disk is read repeatedly. This speeds up repeated reads by a small but noticeable amount. Other changes of note include: - The ifile written by newfs_lfs can now be of arbitrary length, it is no longer restricted to a single indirect block. - Fixed an old bug where ctime was changed every time a vnode was created. I need to look more closely to make sure that the times are only updated during write(2) and friends, not after-the-fact during a segment write, and certainly not by the cleaner.
2001-07-14 00:30:18 +04:00
lfs_bwrite_ext(struct buf *bp, int flags)
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
{
Update to 4.4-Lite fs code, with local changes.
1994-06-08 15:41:58 +04:00
struct lfs *fs;
struct inode *ip;
- in lfs_reserve, reserve locked buffer count as well. - don't wait for locking buf in lfs_bwrite_ext to avoid deadlocks. - skip lfs_reserve when we're doing dirop. reserve more (for lfs_truncate) in set_dirop instead. this mostly solves PR 18972. (and hopefully PR 19196)
2002-12-26 16:37:18 +03:00
int fsb, s;
- in lfs_bwrite_ext, if we're cleaner, mark inode IN_CLEANING rather then IN_MODIFIED. otherwise cleaned (indirect) blocks belongs to the inode isn't written until next sync. - add assertions.
2002-12-14 16:41:25 +03:00
KASSERT(bp->b_flags & B_BUSY);
Add code to UBCify LFS. This is still behind "#ifdef LFS_UBC" for now (there are still some details to work out) but expect that to go away soon. To support these basic changes (creation of lfs_putpages, lfs_gop_write, mods to lfs_balloc) several other changes were made, to wit: * Create a writer daemon kernel thread whose purpose is to handle page writes for the pagedaemon, but which also takes over some of the functions of lfs_check(). This thread is started the first time an LFS is mounted. * Add a "flags" parameter to GOP_SIZE. Current values are GOP_SIZE_READ, meaning that the call should return the size of the in-core version of the file, and GOP_SIZE_WRITE, meaning that it should return the on-disk size. One of GOP_SIZE_READ or GOP_SIZE_WRITE must be specified. * Instead of using malloc(...M_WAITOK) for everything, reserve enough resources to get by and use malloc(...M_NOWAIT), using the reserves if necessary. Use the pool subsystem for structures small enough that this is feasible. This also obsoletes LFS_THROTTLE. And a few that are not strictly necessary: * Moves the LFS inode extensions off onto a separately allocated structure; getting closer to LFS as an LKM. "Welcome to 1.6O." * Unified GOP_ALLOC between FFS and LFS. * Update LFS copyright headers to correct values. * Actually cast to unsigned in lfs_shellsort, like the comment says. * Keep track of which segments were empty before the previous checkpoint; any segments that pass two checkpoints both dirty and empty can be summarily cleaned. Do this. Right now lfs_segclean still works, but this should be turned into an effectless compatibility syscall.
2003-02-18 02:48:08 +03:00
KASSERT(flags & BW_CLEAN || !LFS_IS_MALLOC_BUF(bp));
- in lfs_bwrite_ext, if we're cleaner, mark inode IN_CLEANING rather then IN_MODIFIED. otherwise cleaned (indirect) blocks belongs to the inode isn't written until next sync. - add assertions.
2002-12-14 16:41:25 +03:00
In lfs_bwrite, don't mark buffers dirty if lfs is mounted read-only. (Previously buffers could be marked dirty by the cleaner, and possibly by other means.) Also check for softdep mount in vfs_shutdown before trying to bawrite buffers, since other filesystems don't need it and lfs doesn't bawrite. (This fragment reviewed by fvdl.) Partially addresses PR#8964.
1999-12-15 10:10:32 +03:00
/*
* Don't write *any* blocks if we're mounted read-only.
* In particular the cleaner can't write blocks either.
*/
Tabify, and fix some comment alignment problems.
2003-02-20 07:27:23 +03:00
if (VTOI(bp->b_vp)->i_lfs->lfs_ronly) {
Correct accounting of lfs_avail, locked_queue_count, and locked_queue_bytes. (PR #11468). In the case of fragment allocation, check to see if enough space is available before extending a fragment already scheduled for writing. The locked_queue_* variables indicate the number of buffer headers and bytes, respectively, that are unavailable to getnewbuf() because they are locked up waiting for LFS to flush them; make sure that that is actually what we're counting, i.e., never count malloced buffers, and always use b_bufsize instead of b_bcount. If DEBUG is defined, the periodic calls to lfs_countlocked will now complain if either counter is incorrect. (In the future lfs_countlocked will not need to be called at all if DEBUG is not defined.)
2000-11-17 22:14:41 +03:00
bp->b_flags &= ~(B_DELWRI | B_READ | B_ERROR);
LFS_UNLOCK_BUF(bp);
Add code to UBCify LFS. This is still behind "#ifdef LFS_UBC" for now (there are still some details to work out) but expect that to go away soon. To support these basic changes (creation of lfs_putpages, lfs_gop_write, mods to lfs_balloc) several other changes were made, to wit: * Create a writer daemon kernel thread whose purpose is to handle page writes for the pagedaemon, but which also takes over some of the functions of lfs_check(). This thread is started the first time an LFS is mounted. * Add a "flags" parameter to GOP_SIZE. Current values are GOP_SIZE_READ, meaning that the call should return the size of the in-core version of the file, and GOP_SIZE_WRITE, meaning that it should return the on-disk size. One of GOP_SIZE_READ or GOP_SIZE_WRITE must be specified. * Instead of using malloc(...M_WAITOK) for everything, reserve enough resources to get by and use malloc(...M_NOWAIT), using the reserves if necessary. Use the pool subsystem for structures small enough that this is feasible. This also obsoletes LFS_THROTTLE. And a few that are not strictly necessary: * Moves the LFS inode extensions off onto a separately allocated structure; getting closer to LFS as an LKM. "Welcome to 1.6O." * Unified GOP_ALLOC between FFS and LFS. * Update LFS copyright headers to correct values. * Actually cast to unsigned in lfs_shellsort, like the comment says. * Keep track of which segments were empty before the previous checkpoint; any segments that pass two checkpoints both dirty and empty can be summarily cleaned. Do this. Right now lfs_segclean still works, but this should be turned into an effectless compatibility syscall.
2003-02-18 02:48:08 +03:00
if (LFS_IS_MALLOC_BUF(bp))
In lfs_bwrite, don't mark buffers dirty if lfs is mounted read-only. (Previously buffers could be marked dirty by the cleaner, and possibly by other means.) Also check for softdep mount in vfs_shutdown before trying to bawrite buffers, since other filesystems don't need it and lfs doesn't bawrite. (This fragment reviewed by fvdl.) Partially addresses PR#8964.
1999-12-15 10:10:32 +03:00
bp->b_flags &= ~B_BUSY;
else
brelse(bp);
return EROFS;
}
Update to 4.4-Lite fs code, with local changes.
1994-06-08 15:41:58 +04:00
/*
* Set the delayed write flag and use reassignbuf to move the buffer
* from the clean list to the dirty one.
*
* Set the B_LOCKED flag and unlock the buffer, causing brelse to move
* the buffer onto the LOCKED free list. This is necessary, otherwise
* getnewbuf() would try to reclaim the buffers using bawrite, which
* isn't going to work.
*
* XXX we don't let meta-data writes run out of space because they can
* come from the segment writer. We need to make sure that there is
* enough space reserved so that there's room to write meta-data
* blocks.
*/
if (!(bp->b_flags & B_LOCKED)) {
fs = VFSTOUFS(bp->b_vp->v_mount)->um_lfs;
Merge the short-lived perseant-lfsv2 branch into the trunk. Kernels and tools understand both v1 and v2 filesystems; newfs_lfs generates v2 by default. Changes for the v2 layout include: - Segments of non-PO2 size and arbitrary block offset, so these can be matched to convenient physical characteristics of the partition (e.g., stripe or track size and offset). - Address by fragment instead of by disk sector, paving the way for non-512-byte-sector devices. In theory fragments can be as large as you like, though in reality they must be smaller than MAXBSIZE in size. - Use serial number and filesystem identifier to ensure that roll-forward doesn't get old data and think it's new. Roll-forward is enabled for v2 filesystems, though not for v1 filesystems by default. - The inode free list is now a tailq, paving the way for undelete (undelete is not yet implemented, but can be without further non-backwards-compatible changes to disk structures). - Inode atime information is kept in the Ifile, instead of on the inode; that is, the inode is never written *just* because atime was changed. Because of this the inodes remain near the file data on the disk, rather than wandering all over as the disk is read repeatedly. This speeds up repeated reads by a small but noticeable amount. Other changes of note include: - The ifile written by newfs_lfs can now be of arbitrary length, it is no longer restricted to a single indirect block. - Fixed an old bug where ctime was changed every time a vnode was created. I need to look more closely to make sure that the times are only updated during write(2) and friends, not after-the-fact during a segment write, and certainly not by the cleaner.
2001-07-14 00:30:18 +04:00
fsb = fragstofsb(fs, numfrags(fs, bp->b_bcount));
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
ip = VTOI(bp->b_vp);
- in lfs_bwrite_ext, if we're cleaner, mark inode IN_CLEANING rather then IN_MODIFIED. otherwise cleaned (indirect) blocks belongs to the inode isn't written until next sync. - add assertions.
2002-12-14 16:41:25 +03:00
if (flags & BW_CLEAN) {
Clean up accounting of lfs_uinodes (dirty but unwritten inodes). Make lfs_uinodes a signed quantity for debugging purposes, and set it to zero as fs mount time. Enclose setting/clearing of the dirty flags (IN_MODIFIED, IN_ACCESSED, IN_CLEANING) in macros, and use those macros everywhere. Make LFS_ITIMES use these macros; updated the ITIMES macro in inode.h to know about this. Make ufs_getattr use ITIMES instead of FFS_ITIMES.
2000-07-06 02:25:43 +04:00
LFS_SET_UINO(ip, IN_CLEANING);
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
} else {
Merge the short-lived perseant-lfsv2 branch into the trunk. Kernels and tools understand both v1 and v2 filesystems; newfs_lfs generates v2 by default. Changes for the v2 layout include: - Segments of non-PO2 size and arbitrary block offset, so these can be matched to convenient physical characteristics of the partition (e.g., stripe or track size and offset). - Address by fragment instead of by disk sector, paving the way for non-512-byte-sector devices. In theory fragments can be as large as you like, though in reality they must be smaller than MAXBSIZE in size. - Use serial number and filesystem identifier to ensure that roll-forward doesn't get old data and think it's new. Roll-forward is enabled for v2 filesystems, though not for v1 filesystems by default. - The inode free list is now a tailq, paving the way for undelete (undelete is not yet implemented, but can be without further non-backwards-compatible changes to disk structures). - Inode atime information is kept in the Ifile, instead of on the inode; that is, the inode is never written *just* because atime was changed. Because of this the inodes remain near the file data on the disk, rather than wandering all over as the disk is read repeatedly. This speeds up repeated reads by a small but noticeable amount. Other changes of note include: - The ifile written by newfs_lfs can now be of arbitrary length, it is no longer restricted to a single indirect block. - Fixed an old bug where ctime was changed every time a vnode was created. I need to look more closely to make sure that the times are only updated during write(2) and friends, not after-the-fact during a segment write, and certainly not by the cleaner.
2001-07-14 00:30:18 +04:00
LFS_SET_UINO(ip, IN_MODIFIED);
if (bp->b_lblkno >= 0)
LFS_SET_UINO(ip, IN_UPDATE);
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
}
Merge the short-lived perseant-lfsv2 branch into the trunk. Kernels and tools understand both v1 and v2 filesystems; newfs_lfs generates v2 by default. Changes for the v2 layout include: - Segments of non-PO2 size and arbitrary block offset, so these can be matched to convenient physical characteristics of the partition (e.g., stripe or track size and offset). - Address by fragment instead of by disk sector, paving the way for non-512-byte-sector devices. In theory fragments can be as large as you like, though in reality they must be smaller than MAXBSIZE in size. - Use serial number and filesystem identifier to ensure that roll-forward doesn't get old data and think it's new. Roll-forward is enabled for v2 filesystems, though not for v1 filesystems by default. - The inode free list is now a tailq, paving the way for undelete (undelete is not yet implemented, but can be without further non-backwards-compatible changes to disk structures). - Inode atime information is kept in the Ifile, instead of on the inode; that is, the inode is never written *just* because atime was changed. Because of this the inodes remain near the file data on the disk, rather than wandering all over as the disk is read repeatedly. This speeds up repeated reads by a small but noticeable amount. Other changes of note include: - The ifile written by newfs_lfs can now be of arbitrary length, it is no longer restricted to a single indirect block. - Fixed an old bug where ctime was changed every time a vnode was created. I need to look more closely to make sure that the times are only updated during write(2) and friends, not after-the-fact during a segment write, and certainly not by the cleaner.
2001-07-14 00:30:18 +04:00
fs->lfs_avail -= fsb;
Correct accounting of lfs_avail, locked_queue_count, and locked_queue_bytes. (PR #11468). In the case of fragment allocation, check to see if enough space is available before extending a fragment already scheduled for writing. The locked_queue_* variables indicate the number of buffer headers and bytes, respectively, that are unavailable to getnewbuf() because they are locked up waiting for LFS to flush them; make sure that that is actually what we're counting, i.e., never count malloced buffers, and always use b_bufsize instead of b_bcount. If DEBUG is defined, the periodic calls to lfs_countlocked will now complain if either counter is incorrect. (In the future lfs_countlocked will not need to be called at all if DEBUG is not defined.)
2000-11-17 22:14:41 +03:00
bp->b_flags |= B_DELWRI;
LFS_LOCK_BUF(bp);
Phase one of my three-phase plan to make LFS play nice with UBC, and bug-fixes I found while making sure there weren't any new ones. * Make the write clusters keep track of the buffers whose blocks they contain. This should make it possible to (1) write clusters using a page mapping instead of malloc, if desired, and (2) schedule blocks for rewriting (somewhere else) if a write error occurs. Code is present to use pagemove() to construct the clusters but that is untested and will go away anyway in favor of page mapping. * DEBUG now keeps a log of Ifile writes, so that any lingering instances of the "dirty bufs" problem can be properly debugged. * Keep track of whether the Ifile has been dirtied by various routines that can be called by lfs_segwrite, and loop on that until it is clean, for a checkpoint. Checkpoints need to be squeaky clean. * Warn the user (once) if the Ifile grows larger than is reasonable for their buffer cache. Both lfs_mountfs and lfs_unmount check since the Ifile can grow. * If an inode is not found in a disk block, try rereading the block, under the assumption that the block was copied to a cluster and then freed. * Protect WRITEINPROG() with splbio() to fix a hang in lfs_update.
2002-05-15 00:03:53 +04:00
bp->b_flags &= ~(B_READ | B_DONE | B_ERROR);
Correct accounting of lfs_avail, locked_queue_count, and locked_queue_bytes. (PR #11468). In the case of fragment allocation, check to see if enough space is available before extending a fragment already scheduled for writing. The locked_queue_* variables indicate the number of buffer headers and bytes, respectively, that are unavailable to getnewbuf() because they are locked up waiting for LFS to flush them; make sure that that is actually what we're counting, i.e., never count malloced buffers, and always use b_bufsize instead of b_bcount. If DEBUG is defined, the periodic calls to lfs_countlocked will now complain if either counter is incorrect. (In the future lfs_countlocked will not need to be called at all if DEBUG is not defined.)
2000-11-17 22:14:41 +03:00
s = splbio();
Update to 4.4-Lite fs code, with local changes.
1994-06-08 15:41:58 +04:00
reassignbuf(bp, bp->b_vp);
splx(s);
}
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
add spaces for KNF. confirmed to produce identical objects.
2001-11-24 00:44:25 +03:00
if (bp->b_flags & B_CALL)
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
bp->b_flags &= ~B_BUSY;
else
brelse(bp);
Update to 4.4-Lite fs code, with local changes.
1994-06-08 15:41:58 +04:00
return (0);
}
Fixes associated with filling an LFS: Change the space computation to appear to change the size of the *disk* rather than the *bytes used* when more segment summaries and inode blocks are written. Try to estimate the amount of space that these will take up when more files are written, so the disk size doesn't change too much. Regularize error returns from lfs_valloc, lfs_balloc, lfs_truncate: they now fail entirely, rather than succeeding half-way and leaving the fs in an inconsistent state. Rewrite lfs_truncate, mostly stealing from ffs_truncate. The old lfs_truncate had difficulty truncating a large file to a non-zero size (indirect blocks were not handled appropriately). Unmark VDIROP on fvp after ufs_remove, ufs_rmdir, so these can be reclaimed immediately: this vnode would not be written to disk again anyway if the removal succeeded, and if it failed, no directory operation occurred. ufs_makeinode and ufs_mkdir now remove IN_ADIROP on error.
2000-06-28 00:57:11 +04:00
void
Merge the short-lived perseant-lfsv2 branch into the trunk. Kernels and tools understand both v1 and v2 filesystems; newfs_lfs generates v2 by default. Changes for the v2 layout include: - Segments of non-PO2 size and arbitrary block offset, so these can be matched to convenient physical characteristics of the partition (e.g., stripe or track size and offset). - Address by fragment instead of by disk sector, paving the way for non-512-byte-sector devices. In theory fragments can be as large as you like, though in reality they must be smaller than MAXBSIZE in size. - Use serial number and filesystem identifier to ensure that roll-forward doesn't get old data and think it's new. Roll-forward is enabled for v2 filesystems, though not for v1 filesystems by default. - The inode free list is now a tailq, paving the way for undelete (undelete is not yet implemented, but can be without further non-backwards-compatible changes to disk structures). - Inode atime information is kept in the Ifile, instead of on the inode; that is, the inode is never written *just* because atime was changed. Because of this the inodes remain near the file data on the disk, rather than wandering all over as the disk is read repeatedly. This speeds up repeated reads by a small but noticeable amount. Other changes of note include: - The ifile written by newfs_lfs can now be of arbitrary length, it is no longer restricted to a single indirect block. - Fixed an old bug where ctime was changed every time a vnode was created. I need to look more closely to make sure that the times are only updated during write(2) and friends, not after-the-fact during a segment write, and certainly not by the cleaner.
2001-07-14 00:30:18 +04:00
lfs_flush_fs(struct lfs *fs, int flags)
Fixed lfs_update (and related functions) so that calls from lfs_fsync will DTRT with vnodes marked VDIROP. In particular, the message "flushing VDIROP" will no longer appear, and the filesystem will remain stable in the event of a crash. This was particularly a problem with NFS-exported LFSes, since fsync was called on every file close.
1999-06-01 07:00:40 +04:00
{
Add code to UBCify LFS. This is still behind "#ifdef LFS_UBC" for now (there are still some details to work out) but expect that to go away soon. To support these basic changes (creation of lfs_putpages, lfs_gop_write, mods to lfs_balloc) several other changes were made, to wit: * Create a writer daemon kernel thread whose purpose is to handle page writes for the pagedaemon, but which also takes over some of the functions of lfs_check(). This thread is started the first time an LFS is mounted. * Add a "flags" parameter to GOP_SIZE. Current values are GOP_SIZE_READ, meaning that the call should return the size of the in-core version of the file, and GOP_SIZE_WRITE, meaning that it should return the on-disk size. One of GOP_SIZE_READ or GOP_SIZE_WRITE must be specified. * Instead of using malloc(...M_WAITOK) for everything, reserve enough resources to get by and use malloc(...M_NOWAIT), using the reserves if necessary. Use the pool subsystem for structures small enough that this is feasible. This also obsoletes LFS_THROTTLE. And a few that are not strictly necessary: * Moves the LFS inode extensions off onto a separately allocated structure; getting closer to LFS as an LKM. "Welcome to 1.6O." * Unified GOP_ALLOC between FFS and LFS. * Update LFS copyright headers to correct values. * Actually cast to unsigned in lfs_shellsort, like the comment says. * Keep track of which segments were empty before the previous checkpoint; any segments that pass two checkpoints both dirty and empty can be summarily cleaned. Do this. Right now lfs_segclean still works, but this should be turned into an effectless compatibility syscall.
2003-02-18 02:48:08 +03:00
if (fs->lfs_ronly)
return;
Fixed lfs_update (and related functions) so that calls from lfs_fsync will DTRT with vnodes marked VDIROP. In particular, the message "flushing VDIROP" will no longer appear, and the filesystem will remain stable in the event of a crash. This was particularly a problem with NFS-exported LFSes, since fsync was called on every file close.
1999-06-01 07:00:40 +04:00
Add code to UBCify LFS. This is still behind "#ifdef LFS_UBC" for now (there are still some details to work out) but expect that to go away soon. To support these basic changes (creation of lfs_putpages, lfs_gop_write, mods to lfs_balloc) several other changes were made, to wit: * Create a writer daemon kernel thread whose purpose is to handle page writes for the pagedaemon, but which also takes over some of the functions of lfs_check(). This thread is started the first time an LFS is mounted. * Add a "flags" parameter to GOP_SIZE. Current values are GOP_SIZE_READ, meaning that the call should return the size of the in-core version of the file, and GOP_SIZE_WRITE, meaning that it should return the on-disk size. One of GOP_SIZE_READ or GOP_SIZE_WRITE must be specified. * Instead of using malloc(...M_WAITOK) for everything, reserve enough resources to get by and use malloc(...M_NOWAIT), using the reserves if necessary. Use the pool subsystem for structures small enough that this is feasible. This also obsoletes LFS_THROTTLE. And a few that are not strictly necessary: * Moves the LFS inode extensions off onto a separately allocated structure; getting closer to LFS as an LKM. "Welcome to 1.6O." * Unified GOP_ALLOC between FFS and LFS. * Update LFS copyright headers to correct values. * Actually cast to unsigned in lfs_shellsort, like the comment says. * Keep track of which segments were empty before the previous checkpoint; any segments that pass two checkpoints both dirty and empty can be summarily cleaned. Do this. Right now lfs_segclean still works, but this should be turned into an effectless compatibility syscall.
2003-02-18 02:48:08 +03:00
/* disallow dirops during flush */
fs->lfs_writer++;
Fixed lfs_update (and related functions) so that calls from lfs_fsync will DTRT with vnodes marked VDIROP. In particular, the message "flushing VDIROP" will no longer appear, and the filesystem will remain stable in the event of a crash. This was particularly a problem with NFS-exported LFSes, since fsync was called on every file close.
1999-06-01 07:00:40 +04:00
Add code to UBCify LFS. This is still behind "#ifdef LFS_UBC" for now (there are still some details to work out) but expect that to go away soon. To support these basic changes (creation of lfs_putpages, lfs_gop_write, mods to lfs_balloc) several other changes were made, to wit: * Create a writer daemon kernel thread whose purpose is to handle page writes for the pagedaemon, but which also takes over some of the functions of lfs_check(). This thread is started the first time an LFS is mounted. * Add a "flags" parameter to GOP_SIZE. Current values are GOP_SIZE_READ, meaning that the call should return the size of the in-core version of the file, and GOP_SIZE_WRITE, meaning that it should return the on-disk size. One of GOP_SIZE_READ or GOP_SIZE_WRITE must be specified. * Instead of using malloc(...M_WAITOK) for everything, reserve enough resources to get by and use malloc(...M_NOWAIT), using the reserves if necessary. Use the pool subsystem for structures small enough that this is feasible. This also obsoletes LFS_THROTTLE. And a few that are not strictly necessary: * Moves the LFS inode extensions off onto a separately allocated structure; getting closer to LFS as an LKM. "Welcome to 1.6O." * Unified GOP_ALLOC between FFS and LFS. * Update LFS copyright headers to correct values. * Actually cast to unsigned in lfs_shellsort, like the comment says. * Keep track of which segments were empty before the previous checkpoint; any segments that pass two checkpoints both dirty and empty can be summarily cleaned. Do this. Right now lfs_segclean still works, but this should be turned into an effectless compatibility syscall.
2003-02-18 02:48:08 +03:00
/* drain dirops */
while (fs->lfs_dirops > 0) {
++fs->lfs_diropwait;
tsleep(&fs->lfs_writer, PRIBIO+1, "fldirop", 0);
--fs->lfs_diropwait;
Fixed lfs_update (and related functions) so that calls from lfs_fsync will DTRT with vnodes marked VDIROP. In particular, the message "flushing VDIROP" will no longer appear, and the filesystem will remain stable in the event of a crash. This was particularly a problem with NFS-exported LFSes, since fsync was called on every file close.
1999-06-01 07:00:40 +04:00
}
Add code to UBCify LFS. This is still behind "#ifdef LFS_UBC" for now (there are still some details to work out) but expect that to go away soon. To support these basic changes (creation of lfs_putpages, lfs_gop_write, mods to lfs_balloc) several other changes were made, to wit: * Create a writer daemon kernel thread whose purpose is to handle page writes for the pagedaemon, but which also takes over some of the functions of lfs_check(). This thread is started the first time an LFS is mounted. * Add a "flags" parameter to GOP_SIZE. Current values are GOP_SIZE_READ, meaning that the call should return the size of the in-core version of the file, and GOP_SIZE_WRITE, meaning that it should return the on-disk size. One of GOP_SIZE_READ or GOP_SIZE_WRITE must be specified. * Instead of using malloc(...M_WAITOK) for everything, reserve enough resources to get by and use malloc(...M_NOWAIT), using the reserves if necessary. Use the pool subsystem for structures small enough that this is feasible. This also obsoletes LFS_THROTTLE. And a few that are not strictly necessary: * Moves the LFS inode extensions off onto a separately allocated structure; getting closer to LFS as an LKM. "Welcome to 1.6O." * Unified GOP_ALLOC between FFS and LFS. * Update LFS copyright headers to correct values. * Actually cast to unsigned in lfs_shellsort, like the comment says. * Keep track of which segments were empty before the previous checkpoint; any segments that pass two checkpoints both dirty and empty can be summarily cleaned. Do this. Right now lfs_segclean still works, but this should be turned into an effectless compatibility syscall.
2003-02-18 02:48:08 +03:00
if (lfs_dostats)
++lfs_stats.flush_invoked;
lfs_segwrite(fs->lfs_ivnode->v_mount, flags);
/* allow dirops again */
if (--fs->lfs_writer == 0)
wakeup(&fs->lfs_dirops);
Fixed lfs_update (and related functions) so that calls from lfs_fsync will DTRT with vnodes marked VDIROP. In particular, the message "flushing VDIROP" will no longer appear, and the filesystem will remain stable in the event of a crash. This was particularly a problem with NFS-exported LFSes, since fsync was called on every file close.
1999-06-01 07:00:40 +04:00
}
Update to 4.4-Lite fs code, with local changes.
1994-06-08 15:41:58 +04:00
/*
* XXX
* This routine flushes buffers out of the B_LOCKED queue when LFS has too
* many locked down. Eventually the pageout daemon will simply call LFS
* when pages need to be reclaimed. Note, we have one static count of locked
* buffers, so we can't have more than a single file system. To make this
* work for multiple file systems, put the count into the mount structure.
*/
void
Merge the short-lived perseant-lfsv2 branch into the trunk. Kernels and tools understand both v1 and v2 filesystems; newfs_lfs generates v2 by default. Changes for the v2 layout include: - Segments of non-PO2 size and arbitrary block offset, so these can be matched to convenient physical characteristics of the partition (e.g., stripe or track size and offset). - Address by fragment instead of by disk sector, paving the way for non-512-byte-sector devices. In theory fragments can be as large as you like, though in reality they must be smaller than MAXBSIZE in size. - Use serial number and filesystem identifier to ensure that roll-forward doesn't get old data and think it's new. Roll-forward is enabled for v2 filesystems, though not for v1 filesystems by default. - The inode free list is now a tailq, paving the way for undelete (undelete is not yet implemented, but can be without further non-backwards-compatible changes to disk structures). - Inode atime information is kept in the Ifile, instead of on the inode; that is, the inode is never written *just* because atime was changed. Because of this the inodes remain near the file data on the disk, rather than wandering all over as the disk is read repeatedly. This speeds up repeated reads by a small but noticeable amount. Other changes of note include: - The ifile written by newfs_lfs can now be of arbitrary length, it is no longer restricted to a single indirect block. - Fixed an old bug where ctime was changed every time a vnode was created. I need to look more closely to make sure that the times are only updated during write(2) and friends, not after-the-fact during a segment write, and certainly not by the cleaner.
2001-07-14 00:30:18 +04:00
lfs_flush(struct lfs *fs, int flags)
Update to 4.4-Lite fs code, with local changes.
1994-06-08 15:41:58 +04:00
{
Remove register declarations.
2000-03-30 16:41:09 +04:00
struct mount *mp, *nmp;
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
add spaces for KNF. confirmed to produce identical objects.
2001-11-24 00:44:25 +03:00
if (lfs_dostats)
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
++lfs_stats.write_exceeded;
add spaces for KNF. confirmed to produce identical objects.
2001-11-24 00:44:25 +03:00
if (lfs_writing && flags == 0) {/* XXX flags */
In lfs_bwrite, don't mark buffers dirty if lfs is mounted read-only. (Previously buffers could be marked dirty by the cleaner, and possibly by other means.) Also check for softdep mount in vfs_shutdown before trying to bawrite buffers, since other filesystems don't need it and lfs doesn't bawrite. (This fragment reviewed by fvdl.) Partially addresses PR#8964.
1999-12-15 10:10:32 +03:00
#ifdef DEBUG_LFS
printf("lfs_flush: not flushing because another flush is active\n");
#endif
Update to 4.4-Lite fs code, with local changes.
1994-06-08 15:41:58 +04:00
return;
In lfs_bwrite, don't mark buffers dirty if lfs is mounted read-only. (Previously buffers could be marked dirty by the cleaner, and possibly by other means.) Also check for softdep mount in vfs_shutdown before trying to bawrite buffers, since other filesystems don't need it and lfs doesn't bawrite. (This fragment reviewed by fvdl.) Partially addresses PR#8964.
1999-12-15 10:10:32 +03:00
}
workaround for "another flush is..." infinity loop in writerd. if we're writerd, sleep in lfs_flush until another writer goes away instead of busy loop in writed.
2003-02-19 15:49:10 +03:00
/* XXX MP */
while (lfs_writing && (flags & SEGM_WRITERD))
ltsleep(&lfs_writing, PRIBIO + 1, "lfsflush", 0, 0);
Update to 4.4-Lite fs code, with local changes.
1994-06-08 15:41:58 +04:00
lfs_writing = 1;
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
Add code to UBCify LFS. This is still behind "#ifdef LFS_UBC" for now (there are still some details to work out) but expect that to go away soon. To support these basic changes (creation of lfs_putpages, lfs_gop_write, mods to lfs_balloc) several other changes were made, to wit: * Create a writer daemon kernel thread whose purpose is to handle page writes for the pagedaemon, but which also takes over some of the functions of lfs_check(). This thread is started the first time an LFS is mounted. * Add a "flags" parameter to GOP_SIZE. Current values are GOP_SIZE_READ, meaning that the call should return the size of the in-core version of the file, and GOP_SIZE_WRITE, meaning that it should return the on-disk size. One of GOP_SIZE_READ or GOP_SIZE_WRITE must be specified. * Instead of using malloc(...M_WAITOK) for everything, reserve enough resources to get by and use malloc(...M_NOWAIT), using the reserves if necessary. Use the pool subsystem for structures small enough that this is feasible. This also obsoletes LFS_THROTTLE. And a few that are not strictly necessary: * Moves the LFS inode extensions off onto a separately allocated structure; getting closer to LFS as an LKM. "Welcome to 1.6O." * Unified GOP_ALLOC between FFS and LFS. * Update LFS copyright headers to correct values. * Actually cast to unsigned in lfs_shellsort, like the comment says. * Keep track of which segments were empty before the previous checkpoint; any segments that pass two checkpoints both dirty and empty can be summarily cleaned. Do this. Right now lfs_segclean still works, but this should be turned into an effectless compatibility syscall.
2003-02-18 02:48:08 +03:00
lfs_subsys_pages = 0; /* XXXUBC need a better way to count this */
wakeup(&lfs_subsys_pages);
Merge with Lite2 + local changes
1998-03-01 05:20:01 +03:00
simple_lock(&mountlist_slock);
for (mp = mountlist.cqh_first; mp != (void *)&mountlist; mp = nmp) {
if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) {
nmp = mp->mnt_list.cqe_next;
continue;
}
add spaces for KNF. confirmed to produce identical objects.
2001-11-24 00:44:25 +03:00
if (strncmp(&mp->mnt_stat.f_fstypename[0], MOUNT_LFS, MFSNAMELEN) == 0)
Fixes associated with filling an LFS: Change the space computation to appear to change the size of the *disk* rather than the *bytes used* when more segment summaries and inode blocks are written. Try to estimate the amount of space that these will take up when more files are written, so the disk size doesn't change too much. Regularize error returns from lfs_valloc, lfs_balloc, lfs_truncate: they now fail entirely, rather than succeeding half-way and leaving the fs in an inconsistent state. Rewrite lfs_truncate, mostly stealing from ffs_truncate. The old lfs_truncate had difficulty truncating a large file to a non-zero size (indirect blocks were not handled appropriately). Unmark VDIROP on fvp after ufs_remove, ufs_rmdir, so these can be reclaimed immediately: this vnode would not be written to disk again anyway if the removal succeeded, and if it failed, no directory operation occurred. ufs_makeinode and ufs_mkdir now remove IN_ADIROP on error.
2000-06-28 00:57:11 +04:00
lfs_flush_fs(((struct ufsmount *)mp->mnt_data)->ufsmount_u.lfs, flags);
Merge with Lite2 + local changes
1998-03-01 05:20:01 +03:00
simple_lock(&mountlist_slock);
nmp = mp->mnt_list.cqe_next;
vfs_unbusy(mp);
Update to 4.4-Lite fs code, with local changes.
1994-06-08 15:41:58 +04:00
}
Merge with Lite2 + local changes
1998-03-01 05:20:01 +03:00
simple_unlock(&mountlist_slock);
Phase one of my three-phase plan to make LFS play nice with UBC, and bug-fixes I found while making sure there weren't any new ones. * Make the write clusters keep track of the buffers whose blocks they contain. This should make it possible to (1) write clusters using a page mapping instead of malloc, if desired, and (2) schedule blocks for rewriting (somewhere else) if a write error occurs. Code is present to use pagemove() to construct the clusters but that is untested and will go away anyway in favor of page mapping. * DEBUG now keeps a log of Ifile writes, so that any lingering instances of the "dirty bufs" problem can be properly debugged. * Keep track of whether the Ifile has been dirtied by various routines that can be called by lfs_segwrite, and loop on that until it is clean, for a checkpoint. Checkpoints need to be squeaky clean. * Warn the user (once) if the Ifile grows larger than is reasonable for their buffer cache. Both lfs_mountfs and lfs_unmount check since the Ifile can grow. * If an inode is not found in a disk block, try rereading the block, under the assumption that the block was copied to a cluster and then freed. * Protect WRITEINPROG() with splbio() to fix a hang in lfs_update.
2002-05-15 00:03:53 +04:00
LFS_DEBUG_COUNTLOCKED("flush");
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
Update to 4.4-Lite fs code, with local changes.
1994-06-08 15:41:58 +04:00
lfs_writing = 0;
workaround for "another flush is..." infinity loop in writerd. if we're writerd, sleep in lfs_flush until another writer goes away instead of busy loop in writed.
2003-02-19 15:49:10 +03:00
wakeup(&lfs_writing);
Update to 4.4-Lite fs code, with local changes.
1994-06-08 15:41:58 +04:00
}
Include the space taken by inodes in the count made by lfs_check(); make VOP_SETATTR call lfs_check. This prevents large numbers of inode changes (say, at the end of tar(1)) from filling the buffer cache.
2002-02-11 05:47:29 +03:00
#define INOCOUNT(fs) howmany((fs)->lfs_uinodes, INOPB(fs))
#define INOBYTES(fs) ((fs)->lfs_uinodes * DINODE_SIZE)
Update to 4.4-Lite fs code, with local changes.
1994-06-08 15:41:58 +04:00
int
Bump daddr_t to 64 bits. Replace it with int32_t in all places where it was used on-disk, so that on-disk formats remain the same. Remove ufs_daddr_t and ufs_lbn_t for the time being.
2003-01-25 00:55:02 +03:00
lfs_check(struct vnode *vp, daddr_t blkno, int flags)
Update to 4.4-Lite fs code, with local changes.
1994-06-08 15:41:58 +04:00
{
int error;
Disallow threshold-initiated cache flush when dirops are active. Also, make SET_ENDOP use lfs_check instead of inlining most of it.
1999-04-12 04:36:47 +04:00
struct lfs *fs;
Prevent dirops from getting around lfs_check and wedging the buffer cache. All the dirop vnops now mark the inodes with a new flag, IN_ADIROP, which is removed as soon as the dirop is done (as opposed to VDIROP which stays until the file is written). To address one issue raised in PR#9357.
2000-05-27 04:19:52 +04:00
struct inode *ip;
Address ufs_hashlock/ufs_ihashins protocol bug, discovered while doing a post-mortem of a production machine. Also, take the active dirop count off of the fs and make it global (since it is measuring a global resource) and tie the threshold value LFS_MAXDIROP to desiredvnodes.
1999-11-06 23:33:05 +03:00
extern int lfs_dirvcount;
Disallow threshold-initiated cache flush when dirops are active. Also, make SET_ENDOP use lfs_check instead of inlining most of it.
1999-04-12 04:36:47 +04:00
Update to 4.4-Lite fs code, with local changes.
1994-06-08 15:41:58 +04:00
error = 0;
Prevent dirops from getting around lfs_check and wedging the buffer cache. All the dirop vnops now mark the inodes with a new flag, IN_ADIROP, which is removed as soon as the dirop is done (as opposed to VDIROP which stays until the file is written). To address one issue raised in PR#9357.
2000-05-27 04:19:52 +04:00
ip = VTOI(vp);
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
Update to 4.4-Lite fs code, with local changes.
1994-06-08 15:41:58 +04:00
/* If out of buffers, wait on writer */
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
/* XXX KS - if it's the Ifile, we're probably the cleaner! */
Prevent dirops from getting around lfs_check and wedging the buffer cache. All the dirop vnops now mark the inodes with a new flag, IN_ADIROP, which is removed as soon as the dirop is done (as opposed to VDIROP which stays until the file is written). To address one issue raised in PR#9357.
2000-05-27 04:19:52 +04:00
if (ip->i_number == LFS_IFILE_INUM)
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
return 0;
Prevent dirops from getting around lfs_check and wedging the buffer cache. All the dirop vnops now mark the inodes with a new flag, IN_ADIROP, which is removed as soon as the dirop is done (as opposed to VDIROP which stays until the file is written). To address one issue raised in PR#9357.
2000-05-27 04:19:52 +04:00
/* If we're being called from inside a dirop, don't sleep */
if (ip->i_flag & IN_ADIROP)
Disallow threshold-initiated cache flush when dirops are active. Also, make SET_ENDOP use lfs_check instead of inlining most of it.
1999-04-12 04:36:47 +04:00
return 0;
Prevent dirops from getting around lfs_check and wedging the buffer cache. All the dirop vnops now mark the inodes with a new flag, IN_ADIROP, which is removed as soon as the dirop is done (as opposed to VDIROP which stays until the file is written). To address one issue raised in PR#9357.
2000-05-27 04:19:52 +04:00
fs = ip->i_lfs;
/*
* If we would flush below, but dirops are active, sleep.
* Note that a dirop cannot ever reach this code!
*/
while (fs->lfs_dirops > 0 &&
Include the space taken by inodes in the count made by lfs_check(); make VOP_SETATTR call lfs_check. This prevents large numbers of inode changes (say, at the end of tar(1)) from filling the buffer cache.
2002-02-11 05:47:29 +03:00
(locked_queue_count + INOCOUNT(fs) > LFS_MAX_BUFS ||
Tabify, and fix some comment alignment problems.
2003-02-20 07:27:23 +03:00
locked_queue_bytes + INOBYTES(fs) > LFS_MAX_BYTES ||
Add code to UBCify LFS. This is still behind "#ifdef LFS_UBC" for now (there are still some details to work out) but expect that to go away soon. To support these basic changes (creation of lfs_putpages, lfs_gop_write, mods to lfs_balloc) several other changes were made, to wit: * Create a writer daemon kernel thread whose purpose is to handle page writes for the pagedaemon, but which also takes over some of the functions of lfs_check(). This thread is started the first time an LFS is mounted. * Add a "flags" parameter to GOP_SIZE. Current values are GOP_SIZE_READ, meaning that the call should return the size of the in-core version of the file, and GOP_SIZE_WRITE, meaning that it should return the on-disk size. One of GOP_SIZE_READ or GOP_SIZE_WRITE must be specified. * Instead of using malloc(...M_WAITOK) for everything, reserve enough resources to get by and use malloc(...M_NOWAIT), using the reserves if necessary. Use the pool subsystem for structures small enough that this is feasible. This also obsoletes LFS_THROTTLE. And a few that are not strictly necessary: * Moves the LFS inode extensions off onto a separately allocated structure; getting closer to LFS as an LKM. "Welcome to 1.6O." * Unified GOP_ALLOC between FFS and LFS. * Update LFS copyright headers to correct values. * Actually cast to unsigned in lfs_shellsort, like the comment says. * Keep track of which segments were empty before the previous checkpoint; any segments that pass two checkpoints both dirty and empty can be summarily cleaned. Do this. Right now lfs_segclean still works, but this should be turned into an effectless compatibility syscall.
2003-02-18 02:48:08 +03:00
lfs_subsys_pages > LFS_MAX_PAGES ||
Tabify, and fix some comment alignment problems.
2003-02-20 07:27:23 +03:00
lfs_dirvcount > LFS_MAX_DIROP || fs->lfs_diropwait > 0))
Prevent dirops from getting around lfs_check and wedging the buffer cache. All the dirop vnops now mark the inodes with a new flag, IN_ADIROP, which is removed as soon as the dirop is done (as opposed to VDIROP which stays until the file is written). To address one issue raised in PR#9357.
2000-05-27 04:19:52 +04:00
{
++fs->lfs_diropwait;
tsleep(&fs->lfs_writer, PRIBIO+1, "bufdirop", 0);
--fs->lfs_diropwait;
}
Add code to UBCify LFS. This is still behind "#ifdef LFS_UBC" for now (there are still some details to work out) but expect that to go away soon. To support these basic changes (creation of lfs_putpages, lfs_gop_write, mods to lfs_balloc) several other changes were made, to wit: * Create a writer daemon kernel thread whose purpose is to handle page writes for the pagedaemon, but which also takes over some of the functions of lfs_check(). This thread is started the first time an LFS is mounted. * Add a "flags" parameter to GOP_SIZE. Current values are GOP_SIZE_READ, meaning that the call should return the size of the in-core version of the file, and GOP_SIZE_WRITE, meaning that it should return the on-disk size. One of GOP_SIZE_READ or GOP_SIZE_WRITE must be specified. * Instead of using malloc(...M_WAITOK) for everything, reserve enough resources to get by and use malloc(...M_NOWAIT), using the reserves if necessary. Use the pool subsystem for structures small enough that this is feasible. This also obsoletes LFS_THROTTLE. And a few that are not strictly necessary: * Moves the LFS inode extensions off onto a separately allocated structure; getting closer to LFS as an LKM. "Welcome to 1.6O." * Unified GOP_ALLOC between FFS and LFS. * Update LFS copyright headers to correct values. * Actually cast to unsigned in lfs_shellsort, like the comment says. * Keep track of which segments were empty before the previous checkpoint; any segments that pass two checkpoints both dirty and empty can be summarily cleaned. Do this. Right now lfs_segclean still works, but this should be turned into an effectless compatibility syscall.
2003-02-18 02:48:08 +03:00
#ifdef DEBUG_LFS_FLUSH
if (locked_queue_count + INOCOUNT(fs) > LFS_MAX_BUFS)
printf("lqc = %d, max %d\n", locked_queue_count + INOCOUNT(fs),
LFS_MAX_BUFS);
if (locked_queue_bytes + INOBYTES(fs) > LFS_MAX_BYTES)
printf("lqb = %ld, max %d\n", locked_queue_bytes + INOBYTES(fs),
LFS_MAX_BYTES);
if (lfs_subsys_pages > LFS_MAX_PAGES)
printf("lssp = %d, max %d\n", lfs_subsys_pages, LFS_MAX_PAGES);
if (lfs_dirvcount > LFS_MAX_DIROP)
printf("ldvc = %d, max %d\n", lfs_dirvcount, LFS_MAX_DIROP);
if (fs->lfs_diropwait > 0)
printf("ldvw = %d\n", fs->lfs_diropwait);
#endif
Include the space taken by inodes in the count made by lfs_check(); make VOP_SETATTR call lfs_check. This prevents large numbers of inode changes (say, at the end of tar(1)) from filling the buffer cache.
2002-02-11 05:47:29 +03:00
if (locked_queue_count + INOCOUNT(fs) > LFS_MAX_BUFS ||
locked_queue_bytes + INOBYTES(fs) > LFS_MAX_BYTES ||
Add code to UBCify LFS. This is still behind "#ifdef LFS_UBC" for now (there are still some details to work out) but expect that to go away soon. To support these basic changes (creation of lfs_putpages, lfs_gop_write, mods to lfs_balloc) several other changes were made, to wit: * Create a writer daemon kernel thread whose purpose is to handle page writes for the pagedaemon, but which also takes over some of the functions of lfs_check(). This thread is started the first time an LFS is mounted. * Add a "flags" parameter to GOP_SIZE. Current values are GOP_SIZE_READ, meaning that the call should return the size of the in-core version of the file, and GOP_SIZE_WRITE, meaning that it should return the on-disk size. One of GOP_SIZE_READ or GOP_SIZE_WRITE must be specified. * Instead of using malloc(...M_WAITOK) for everything, reserve enough resources to get by and use malloc(...M_NOWAIT), using the reserves if necessary. Use the pool subsystem for structures small enough that this is feasible. This also obsoletes LFS_THROTTLE. And a few that are not strictly necessary: * Moves the LFS inode extensions off onto a separately allocated structure; getting closer to LFS as an LKM. "Welcome to 1.6O." * Unified GOP_ALLOC between FFS and LFS. * Update LFS copyright headers to correct values. * Actually cast to unsigned in lfs_shellsort, like the comment says. * Keep track of which segments were empty before the previous checkpoint; any segments that pass two checkpoints both dirty and empty can be summarily cleaned. Do this. Right now lfs_segclean still works, but this should be turned into an effectless compatibility syscall.
2003-02-18 02:48:08 +03:00
lfs_subsys_pages > LFS_MAX_PAGES ||
lfs_dirvcount > LFS_MAX_DIROP || fs->lfs_diropwait > 0)
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
{
Disallow threshold-initiated cache flush when dirops are active. Also, make SET_ENDOP use lfs_check instead of inlining most of it.
1999-04-12 04:36:47 +04:00
lfs_flush(fs, flags);
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
}
Fixed lfs_update (and related functions) so that calls from lfs_fsync will DTRT with vnodes marked VDIROP. In particular, the message "flushing VDIROP" will no longer appear, and the filesystem will remain stable in the event of a crash. This was particularly a problem with NFS-exported LFSes, since fsync was called on every file close.
1999-06-01 07:00:40 +04:00
Add code to UBCify LFS. This is still behind "#ifdef LFS_UBC" for now (there are still some details to work out) but expect that to go away soon. To support these basic changes (creation of lfs_putpages, lfs_gop_write, mods to lfs_balloc) several other changes were made, to wit: * Create a writer daemon kernel thread whose purpose is to handle page writes for the pagedaemon, but which also takes over some of the functions of lfs_check(). This thread is started the first time an LFS is mounted. * Add a "flags" parameter to GOP_SIZE. Current values are GOP_SIZE_READ, meaning that the call should return the size of the in-core version of the file, and GOP_SIZE_WRITE, meaning that it should return the on-disk size. One of GOP_SIZE_READ or GOP_SIZE_WRITE must be specified. * Instead of using malloc(...M_WAITOK) for everything, reserve enough resources to get by and use malloc(...M_NOWAIT), using the reserves if necessary. Use the pool subsystem for structures small enough that this is feasible. This also obsoletes LFS_THROTTLE. And a few that are not strictly necessary: * Moves the LFS inode extensions off onto a separately allocated structure; getting closer to LFS as an LKM. "Welcome to 1.6O." * Unified GOP_ALLOC between FFS and LFS. * Update LFS copyright headers to correct values. * Actually cast to unsigned in lfs_shellsort, like the comment says. * Keep track of which segments were empty before the previous checkpoint; any segments that pass two checkpoints both dirty and empty can be summarily cleaned. Do this. Right now lfs_segclean still works, but this should be turned into an effectless compatibility syscall.
2003-02-18 02:48:08 +03:00
while (locked_queue_count + INOCOUNT(fs) > LFS_WAIT_BUFS ||
locked_queue_bytes + INOBYTES(fs) > LFS_WAIT_BYTES ||
lfs_subsys_pages > LFS_WAIT_PAGES ||
lfs_dirvcount > LFS_MAX_DIROP)
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
{
add spaces for KNF. confirmed to produce identical objects.
2001-11-24 00:44:25 +03:00
if (lfs_dostats)
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
++lfs_stats.wait_exceeded;
Merge the short-lived perseant-lfsv2 branch into the trunk. Kernels and tools understand both v1 and v2 filesystems; newfs_lfs generates v2 by default. Changes for the v2 layout include: - Segments of non-PO2 size and arbitrary block offset, so these can be matched to convenient physical characteristics of the partition (e.g., stripe or track size and offset). - Address by fragment instead of by disk sector, paving the way for non-512-byte-sector devices. In theory fragments can be as large as you like, though in reality they must be smaller than MAXBSIZE in size. - Use serial number and filesystem identifier to ensure that roll-forward doesn't get old data and think it's new. Roll-forward is enabled for v2 filesystems, though not for v1 filesystems by default. - The inode free list is now a tailq, paving the way for undelete (undelete is not yet implemented, but can be without further non-backwards-compatible changes to disk structures). - Inode atime information is kept in the Ifile, instead of on the inode; that is, the inode is never written *just* because atime was changed. Because of this the inodes remain near the file data on the disk, rather than wandering all over as the disk is read repeatedly. This speeds up repeated reads by a small but noticeable amount. Other changes of note include: - The ifile written by newfs_lfs can now be of arbitrary length, it is no longer restricted to a single indirect block. - Fixed an old bug where ctime was changed every time a vnode was created. I need to look more closely to make sure that the times are only updated during write(2) and friends, not after-the-fact during a segment write, and certainly not by the cleaner.
2001-07-14 00:30:18 +04:00
#ifdef DEBUG_LFS
In lfs_bwrite, don't mark buffers dirty if lfs is mounted read-only. (Previously buffers could be marked dirty by the cleaner, and possibly by other means.) Also check for softdep mount in vfs_shutdown before trying to bawrite buffers, since other filesystems don't need it and lfs doesn't bawrite. (This fragment reviewed by fvdl.) Partially addresses PR#8964.
1999-12-15 10:10:32 +03:00
printf("lfs_check: waiting: count=%d, bytes=%ld\n",
locked_queue_count, locked_queue_bytes);
#endif
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
error = tsleep(&locked_queue_count, PCATCH | PUSER,
"buffers", hz * LFS_BUFWAIT);
If LFS_DO_ROLLFORWARD is defined, roll forward from the older checkpoint on mount, through the newer checkpoint and on through any newer partial-segments that may have been written but not checkpointed because of an intervening crash. LFS_DO_ROLLFORWARD is not defined by default.
2000-11-27 06:33:57 +03:00
if (error != EWOULDBLOCK)
break;
Change the way LFS does block accounting, from trying to infer from the buffer cache flags, to marking the inode and/or indirect blocks with a special disk address UNWRITTEN==-2 when a block is accounted for. (This address is never written to disk, but only used in-core. This is essentially the same method of block accounting as on the UBC branch, where the buffer headers don't exist.) Make sure that truncation is handled properly, especially in the case of holey files. Fixes PR#9994.
2000-05-06 00:59:20 +04:00
/*
* lfs_flush might not flush all the buffers, if some of the
If LFS_DO_ROLLFORWARD is defined, roll forward from the older checkpoint on mount, through the newer checkpoint and on through any newer partial-segments that may have been written but not checkpointed because of an intervening crash. LFS_DO_ROLLFORWARD is not defined by default.
2000-11-27 06:33:57 +03:00
* inodes were locked or if most of them were Ifile blocks
Tabify, and fix some comment alignment problems.
2003-02-20 07:27:23 +03:00
* and we weren't asked to checkpoint. Try flushing again
If LFS_DO_ROLLFORWARD is defined, roll forward from the older checkpoint on mount, through the newer checkpoint and on through any newer partial-segments that may have been written but not checkpointed because of an intervening crash. LFS_DO_ROLLFORWARD is not defined by default.
2000-11-27 06:33:57 +03:00
* to keep us from blocking indefinitely.
Change the way LFS does block accounting, from trying to infer from the buffer cache flags, to marking the inode and/or indirect blocks with a special disk address UNWRITTEN==-2 when a block is accounted for. (This address is never written to disk, but only used in-core. This is essentially the same method of block accounting as on the UBC branch, where the buffer headers don't exist.) Make sure that truncation is handled properly, especially in the case of holey files. Fixes PR#9994.
2000-05-06 00:59:20 +04:00
*/
Include the space taken by inodes in the count made by lfs_check(); make VOP_SETATTR call lfs_check. This prevents large numbers of inode changes (say, at the end of tar(1)) from filling the buffer cache.
2002-02-11 05:47:29 +03:00
if (locked_queue_count + INOCOUNT(fs) > LFS_MAX_BUFS ||
locked_queue_bytes + INOBYTES(fs) > LFS_MAX_BYTES)
Change the way LFS does block accounting, from trying to infer from the buffer cache flags, to marking the inode and/or indirect blocks with a special disk address UNWRITTEN==-2 when a block is accounted for. (This address is never written to disk, but only used in-core. This is essentially the same method of block accounting as on the UBC branch, where the buffer headers don't exist.) Make sure that truncation is handled properly, especially in the case of holey files. Fixes PR#9994.
2000-05-06 00:59:20 +04:00
{
If LFS_DO_ROLLFORWARD is defined, roll forward from the older checkpoint on mount, through the newer checkpoint and on through any newer partial-segments that may have been written but not checkpointed because of an intervening crash. LFS_DO_ROLLFORWARD is not defined by default.
2000-11-27 06:33:57 +03:00
lfs_flush(fs, flags | SEGM_CKP);
Change the way LFS does block accounting, from trying to infer from the buffer cache flags, to marking the inode and/or indirect blocks with a special disk address UNWRITTEN==-2 when a block is accounted for. (This address is never written to disk, but only used in-core. This is essentially the same method of block accounting as on the UBC branch, where the buffer headers don't exist.) Make sure that truncation is handled properly, especially in the case of holey files. Fixes PR#9994.
2000-05-06 00:59:20 +04:00
}
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
}
return (error);
}
/*
* Allocate a new buffer header.
*/
Get rid of some old unnecessary code that cleared B_NEEDCOMMIT from buffers in lfs_writeseg (possibly after they had been freed). If MALLOCLOG is defined, make lfs_newbuf and lfs_freebuf pass along the caller's file and line to _malloc and _free.
2000-12-03 08:56:27 +03:00
struct buf *
Add code to UBCify LFS. This is still behind "#ifdef LFS_UBC" for now (there are still some details to work out) but expect that to go away soon. To support these basic changes (creation of lfs_putpages, lfs_gop_write, mods to lfs_balloc) several other changes were made, to wit: * Create a writer daemon kernel thread whose purpose is to handle page writes for the pagedaemon, but which also takes over some of the functions of lfs_check(). This thread is started the first time an LFS is mounted. * Add a "flags" parameter to GOP_SIZE. Current values are GOP_SIZE_READ, meaning that the call should return the size of the in-core version of the file, and GOP_SIZE_WRITE, meaning that it should return the on-disk size. One of GOP_SIZE_READ or GOP_SIZE_WRITE must be specified. * Instead of using malloc(...M_WAITOK) for everything, reserve enough resources to get by and use malloc(...M_NOWAIT), using the reserves if necessary. Use the pool subsystem for structures small enough that this is feasible. This also obsoletes LFS_THROTTLE. And a few that are not strictly necessary: * Moves the LFS inode extensions off onto a separately allocated structure; getting closer to LFS as an LKM. "Welcome to 1.6O." * Unified GOP_ALLOC between FFS and LFS. * Update LFS copyright headers to correct values. * Actually cast to unsigned in lfs_shellsort, like the comment says. * Keep track of which segments were empty before the previous checkpoint; any segments that pass two checkpoints both dirty and empty can be summarily cleaned. Do this. Right now lfs_segclean still works, but this should be turned into an effectless compatibility syscall.
2003-02-18 02:48:08 +03:00
lfs_newbuf(struct lfs *fs, struct vnode *vp, daddr_t daddr, size_t size, int type)
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
{
struct buf *bp;
size_t nbytes;
int s;
Merge the short-lived perseant-lfsv2 branch into the trunk. Kernels and tools understand both v1 and v2 filesystems; newfs_lfs generates v2 by default. Changes for the v2 layout include: - Segments of non-PO2 size and arbitrary block offset, so these can be matched to convenient physical characteristics of the partition (e.g., stripe or track size and offset). - Address by fragment instead of by disk sector, paving the way for non-512-byte-sector devices. In theory fragments can be as large as you like, though in reality they must be smaller than MAXBSIZE in size. - Use serial number and filesystem identifier to ensure that roll-forward doesn't get old data and think it's new. Roll-forward is enabled for v2 filesystems, though not for v1 filesystems by default. - The inode free list is now a tailq, paving the way for undelete (undelete is not yet implemented, but can be without further non-backwards-compatible changes to disk structures). - Inode atime information is kept in the Ifile, instead of on the inode; that is, the inode is never written *just* because atime was changed. Because of this the inodes remain near the file data on the disk, rather than wandering all over as the disk is read repeatedly. This speeds up repeated reads by a small but noticeable amount. Other changes of note include: - The ifile written by newfs_lfs can now be of arbitrary length, it is no longer restricted to a single indirect block. - Fixed an old bug where ctime was changed every time a vnode was created. I need to look more closely to make sure that the times are only updated during write(2) and friends, not after-the-fact during a segment write, and certainly not by the cleaner.
2001-07-14 00:30:18 +04:00
nbytes = roundup(size, fsbtob(fs, 1));
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
Add code to UBCify LFS. This is still behind "#ifdef LFS_UBC" for now (there are still some details to work out) but expect that to go away soon. To support these basic changes (creation of lfs_putpages, lfs_gop_write, mods to lfs_balloc) several other changes were made, to wit: * Create a writer daemon kernel thread whose purpose is to handle page writes for the pagedaemon, but which also takes over some of the functions of lfs_check(). This thread is started the first time an LFS is mounted. * Add a "flags" parameter to GOP_SIZE. Current values are GOP_SIZE_READ, meaning that the call should return the size of the in-core version of the file, and GOP_SIZE_WRITE, meaning that it should return the on-disk size. One of GOP_SIZE_READ or GOP_SIZE_WRITE must be specified. * Instead of using malloc(...M_WAITOK) for everything, reserve enough resources to get by and use malloc(...M_NOWAIT), using the reserves if necessary. Use the pool subsystem for structures small enough that this is feasible. This also obsoletes LFS_THROTTLE. And a few that are not strictly necessary: * Moves the LFS inode extensions off onto a separately allocated structure; getting closer to LFS as an LKM. "Welcome to 1.6O." * Unified GOP_ALLOC between FFS and LFS. * Update LFS copyright headers to correct values. * Actually cast to unsigned in lfs_shellsort, like the comment says. * Keep track of which segments were empty before the previous checkpoint; any segments that pass two checkpoints both dirty and empty can be summarily cleaned. Do this. Right now lfs_segclean still works, but this should be turned into an effectless compatibility syscall.
2003-02-18 02:48:08 +03:00
s = splbio();
bp = pool_get(&bufpool, PR_WAITOK);
splx(s);
memset(bp, 0, sizeof(struct buf));
init b_interlock.
2003-02-19 15:01:17 +03:00
simple_lock_init(&bp->b_interlock);
add spaces for KNF. confirmed to produce identical objects.
2001-11-24 00:44:25 +03:00
if (nbytes) {
Add code to UBCify LFS. This is still behind "#ifdef LFS_UBC" for now (there are still some details to work out) but expect that to go away soon. To support these basic changes (creation of lfs_putpages, lfs_gop_write, mods to lfs_balloc) several other changes were made, to wit: * Create a writer daemon kernel thread whose purpose is to handle page writes for the pagedaemon, but which also takes over some of the functions of lfs_check(). This thread is started the first time an LFS is mounted. * Add a "flags" parameter to GOP_SIZE. Current values are GOP_SIZE_READ, meaning that the call should return the size of the in-core version of the file, and GOP_SIZE_WRITE, meaning that it should return the on-disk size. One of GOP_SIZE_READ or GOP_SIZE_WRITE must be specified. * Instead of using malloc(...M_WAITOK) for everything, reserve enough resources to get by and use malloc(...M_NOWAIT), using the reserves if necessary. Use the pool subsystem for structures small enough that this is feasible. This also obsoletes LFS_THROTTLE. And a few that are not strictly necessary: * Moves the LFS inode extensions off onto a separately allocated structure; getting closer to LFS as an LKM. "Welcome to 1.6O." * Unified GOP_ALLOC between FFS and LFS. * Update LFS copyright headers to correct values. * Actually cast to unsigned in lfs_shellsort, like the comment says. * Keep track of which segments were empty before the previous checkpoint; any segments that pass two checkpoints both dirty and empty can be summarily cleaned. Do this. Right now lfs_segclean still works, but this should be turned into an effectless compatibility syscall.
2003-02-18 02:48:08 +03:00
bp->b_data = lfs_malloc(fs, nbytes, type);
/* memset(bp->b_data, 0, nbytes); */
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
}
#ifdef DIAGNOSTIC
add spaces for KNF. confirmed to produce identical objects.
2001-11-24 00:44:25 +03:00
if (vp == NULL)
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
panic("vp is NULL in lfs_newbuf");
add spaces for KNF. confirmed to produce identical objects.
2001-11-24 00:44:25 +03:00
if (bp == NULL)
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
panic("bp is NULL after malloc in lfs_newbuf");
Update to 4.4-Lite fs code, with local changes.
1994-06-08 15:41:58 +04:00
#endif
Make the buffer cache code MP-safe.
2003-02-06 00:38:38 +03:00
simple_lock_init(&bp->b_interlock);
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
s = splbio();
bgetvp(vp, bp);
splx(s);
Make the buffer cache code MP-safe.
2003-02-06 00:38:38 +03:00
Phase one of my three-phase plan to make LFS play nice with UBC, and bug-fixes I found while making sure there weren't any new ones. * Make the write clusters keep track of the buffers whose blocks they contain. This should make it possible to (1) write clusters using a page mapping instead of malloc, if desired, and (2) schedule blocks for rewriting (somewhere else) if a write error occurs. Code is present to use pagemove() to construct the clusters but that is untested and will go away anyway in favor of page mapping. * DEBUG now keeps a log of Ifile writes, so that any lingering instances of the "dirty bufs" problem can be properly debugged. * Keep track of whether the Ifile has been dirtied by various routines that can be called by lfs_segwrite, and loop on that until it is clean, for a checkpoint. Checkpoints need to be squeaky clean. * Warn the user (once) if the Ifile grows larger than is reasonable for their buffer cache. Both lfs_mountfs and lfs_unmount check since the Ifile can grow. * If an inode is not found in a disk block, try rereading the block, under the assumption that the block was copied to a cluster and then freed. * Protect WRITEINPROG() with splbio() to fix a hang in lfs_update.
2002-05-15 00:03:53 +04:00
bp->b_saveaddr = (caddr_t)fs;
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
bp->b_bufsize = size;
bp->b_bcount = size;
bp->b_lblkno = daddr;
bp->b_blkno = daddr;
bp->b_error = 0;
bp->b_resid = 0;
bp->b_iodone = lfs_callback;
bp->b_flags |= B_BUSY | B_CALL | B_NOCACHE;
return (bp);
}
void
Add code to UBCify LFS. This is still behind "#ifdef LFS_UBC" for now (there are still some details to work out) but expect that to go away soon. To support these basic changes (creation of lfs_putpages, lfs_gop_write, mods to lfs_balloc) several other changes were made, to wit: * Create a writer daemon kernel thread whose purpose is to handle page writes for the pagedaemon, but which also takes over some of the functions of lfs_check(). This thread is started the first time an LFS is mounted. * Add a "flags" parameter to GOP_SIZE. Current values are GOP_SIZE_READ, meaning that the call should return the size of the in-core version of the file, and GOP_SIZE_WRITE, meaning that it should return the on-disk size. One of GOP_SIZE_READ or GOP_SIZE_WRITE must be specified. * Instead of using malloc(...M_WAITOK) for everything, reserve enough resources to get by and use malloc(...M_NOWAIT), using the reserves if necessary. Use the pool subsystem for structures small enough that this is feasible. This also obsoletes LFS_THROTTLE. And a few that are not strictly necessary: * Moves the LFS inode extensions off onto a separately allocated structure; getting closer to LFS as an LKM. "Welcome to 1.6O." * Unified GOP_ALLOC between FFS and LFS. * Update LFS copyright headers to correct values. * Actually cast to unsigned in lfs_shellsort, like the comment says. * Keep track of which segments were empty before the previous checkpoint; any segments that pass two checkpoints both dirty and empty can be summarily cleaned. Do this. Right now lfs_segclean still works, but this should be turned into an effectless compatibility syscall.
2003-02-18 02:48:08 +03:00
lfs_freebuf(struct lfs *fs, struct buf *bp)
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
{
int s;
s = splbio();
add spaces for KNF. confirmed to produce identical objects.
2001-11-24 00:44:25 +03:00
if (bp->b_vp)
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
brelvp(bp);
if (!(bp->b_flags & B_INVAL)) { /* B_INVAL indicates a "fake" buffer */
Add code to UBCify LFS. This is still behind "#ifdef LFS_UBC" for now (there are still some details to work out) but expect that to go away soon. To support these basic changes (creation of lfs_putpages, lfs_gop_write, mods to lfs_balloc) several other changes were made, to wit: * Create a writer daemon kernel thread whose purpose is to handle page writes for the pagedaemon, but which also takes over some of the functions of lfs_check(). This thread is started the first time an LFS is mounted. * Add a "flags" parameter to GOP_SIZE. Current values are GOP_SIZE_READ, meaning that the call should return the size of the in-core version of the file, and GOP_SIZE_WRITE, meaning that it should return the on-disk size. One of GOP_SIZE_READ or GOP_SIZE_WRITE must be specified. * Instead of using malloc(...M_WAITOK) for everything, reserve enough resources to get by and use malloc(...M_NOWAIT), using the reserves if necessary. Use the pool subsystem for structures small enough that this is feasible. This also obsoletes LFS_THROTTLE. And a few that are not strictly necessary: * Moves the LFS inode extensions off onto a separately allocated structure; getting closer to LFS as an LKM. "Welcome to 1.6O." * Unified GOP_ALLOC between FFS and LFS. * Update LFS copyright headers to correct values. * Actually cast to unsigned in lfs_shellsort, like the comment says. * Keep track of which segments were empty before the previous checkpoint; any segments that pass two checkpoints both dirty and empty can be summarily cleaned. Do this. Right now lfs_segclean still works, but this should be turned into an effectless compatibility syscall.
2003-02-18 02:48:08 +03:00
lfs_free(fs, bp->b_data, LFS_NB_UNKNOWN);
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
bp->b_data = NULL;
Update to 4.4-Lite fs code, with local changes.
1994-06-08 15:41:58 +04:00
}
Add code to UBCify LFS. This is still behind "#ifdef LFS_UBC" for now (there are still some details to work out) but expect that to go away soon. To support these basic changes (creation of lfs_putpages, lfs_gop_write, mods to lfs_balloc) several other changes were made, to wit: * Create a writer daemon kernel thread whose purpose is to handle page writes for the pagedaemon, but which also takes over some of the functions of lfs_check(). This thread is started the first time an LFS is mounted. * Add a "flags" parameter to GOP_SIZE. Current values are GOP_SIZE_READ, meaning that the call should return the size of the in-core version of the file, and GOP_SIZE_WRITE, meaning that it should return the on-disk size. One of GOP_SIZE_READ or GOP_SIZE_WRITE must be specified. * Instead of using malloc(...M_WAITOK) for everything, reserve enough resources to get by and use malloc(...M_NOWAIT), using the reserves if necessary. Use the pool subsystem for structures small enough that this is feasible. This also obsoletes LFS_THROTTLE. And a few that are not strictly necessary: * Moves the LFS inode extensions off onto a separately allocated structure; getting closer to LFS as an LKM. "Welcome to 1.6O." * Unified GOP_ALLOC between FFS and LFS. * Update LFS copyright headers to correct values. * Actually cast to unsigned in lfs_shellsort, like the comment says. * Keep track of which segments were empty before the previous checkpoint; any segments that pass two checkpoints both dirty and empty can be summarily cleaned. Do this. Right now lfs_segclean still works, but this should be turned into an effectless compatibility syscall.
2003-02-18 02:48:08 +03:00
pool_put(&bufpool, bp);
splx(s);
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
}
Update to 4.4-Lite fs code, with local changes.
1994-06-08 15:41:58 +04:00
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
/*
* 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];
/*
* Return a count of buffers on the "locked" queue.
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
* Don't count malloced buffers, since they don't detract from the total.
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
*/
void
Phase one of my three-phase plan to make LFS play nice with UBC, and bug-fixes I found while making sure there weren't any new ones. * Make the write clusters keep track of the buffers whose blocks they contain. This should make it possible to (1) write clusters using a page mapping instead of malloc, if desired, and (2) schedule blocks for rewriting (somewhere else) if a write error occurs. Code is present to use pagemove() to construct the clusters but that is untested and will go away anyway in favor of page mapping. * DEBUG now keeps a log of Ifile writes, so that any lingering instances of the "dirty bufs" problem can be properly debugged. * Keep track of whether the Ifile has been dirtied by various routines that can be called by lfs_segwrite, and loop on that until it is clean, for a checkpoint. Checkpoints need to be squeaky clean. * Warn the user (once) if the Ifile grows larger than is reasonable for their buffer cache. Both lfs_mountfs and lfs_unmount check since the Ifile can grow. * If an inode is not found in a disk block, try rereading the block, under the assumption that the block was copied to a cluster and then freed. * Protect WRITEINPROG() with splbio() to fix a hang in lfs_update.
2002-05-15 00:03:53 +04:00
lfs_countlocked(int *count, long *bytes, char *msg)
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
{
Remove register declarations.
2000-03-30 16:41:09 +04:00
struct buf *bp;
int n = 0;
long int size = 0L;
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
for (bp = bufqueues[BQ_LOCKED].tqh_first; bp;
bp = bp->b_freelist.tqe_next) {
Add code to UBCify LFS. This is still behind "#ifdef LFS_UBC" for now (there are still some details to work out) but expect that to go away soon. To support these basic changes (creation of lfs_putpages, lfs_gop_write, mods to lfs_balloc) several other changes were made, to wit: * Create a writer daemon kernel thread whose purpose is to handle page writes for the pagedaemon, but which also takes over some of the functions of lfs_check(). This thread is started the first time an LFS is mounted. * Add a "flags" parameter to GOP_SIZE. Current values are GOP_SIZE_READ, meaning that the call should return the size of the in-core version of the file, and GOP_SIZE_WRITE, meaning that it should return the on-disk size. One of GOP_SIZE_READ or GOP_SIZE_WRITE must be specified. * Instead of using malloc(...M_WAITOK) for everything, reserve enough resources to get by and use malloc(...M_NOWAIT), using the reserves if necessary. Use the pool subsystem for structures small enough that this is feasible. This also obsoletes LFS_THROTTLE. And a few that are not strictly necessary: * Moves the LFS inode extensions off onto a separately allocated structure; getting closer to LFS as an LKM. "Welcome to 1.6O." * Unified GOP_ALLOC between FFS and LFS. * Update LFS copyright headers to correct values. * Actually cast to unsigned in lfs_shellsort, like the comment says. * Keep track of which segments were empty before the previous checkpoint; any segments that pass two checkpoints both dirty and empty can be summarily cleaned. Do this. Right now lfs_segclean still works, but this should be turned into an effectless compatibility syscall.
2003-02-18 02:48:08 +03:00
if (bp->b_flags & B_CALL)
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
continue;
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
n++;
size += bp->b_bufsize;
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
#ifdef DEBUG_LOCKED_LIST
if (n > nbuf)
panic("lfs_countlocked: this can't happen: more"
" buffers locked than exist");
#endif
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
}
Phase one of my three-phase plan to make LFS play nice with UBC, and bug-fixes I found while making sure there weren't any new ones. * Make the write clusters keep track of the buffers whose blocks they contain. This should make it possible to (1) write clusters using a page mapping instead of malloc, if desired, and (2) schedule blocks for rewriting (somewhere else) if a write error occurs. Code is present to use pagemove() to construct the clusters but that is untested and will go away anyway in favor of page mapping. * DEBUG now keeps a log of Ifile writes, so that any lingering instances of the "dirty bufs" problem can be properly debugged. * Keep track of whether the Ifile has been dirtied by various routines that can be called by lfs_segwrite, and loop on that until it is clean, for a checkpoint. Checkpoints need to be squeaky clean. * Warn the user (once) if the Ifile grows larger than is reasonable for their buffer cache. Both lfs_mountfs and lfs_unmount check since the Ifile can grow. * If an inode is not found in a disk block, try rereading the block, under the assumption that the block was copied to a cluster and then freed. * Protect WRITEINPROG() with splbio() to fix a hang in lfs_update.
2002-05-15 00:03:53 +04:00
#ifdef DEBUG_LOCKED_LIST
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
/* Theoretically this function never really does anything */
if (n != *count)
Phase one of my three-phase plan to make LFS play nice with UBC, and bug-fixes I found while making sure there weren't any new ones. * Make the write clusters keep track of the buffers whose blocks they contain. This should make it possible to (1) write clusters using a page mapping instead of malloc, if desired, and (2) schedule blocks for rewriting (somewhere else) if a write error occurs. Code is present to use pagemove() to construct the clusters but that is untested and will go away anyway in favor of page mapping. * DEBUG now keeps a log of Ifile writes, so that any lingering instances of the "dirty bufs" problem can be properly debugged. * Keep track of whether the Ifile has been dirtied by various routines that can be called by lfs_segwrite, and loop on that until it is clean, for a checkpoint. Checkpoints need to be squeaky clean. * Warn the user (once) if the Ifile grows larger than is reasonable for their buffer cache. Both lfs_mountfs and lfs_unmount check since the Ifile can grow. * If an inode is not found in a disk block, try rereading the block, under the assumption that the block was copied to a cluster and then freed. * Protect WRITEINPROG() with splbio() to fix a hang in lfs_update.
2002-05-15 00:03:53 +04:00
printf("lfs_countlocked: %s: adjusted buf count from %d to %d\n",
msg, *count, n);
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
if (size != *bytes)
Phase one of my three-phase plan to make LFS play nice with UBC, and bug-fixes I found while making sure there weren't any new ones. * Make the write clusters keep track of the buffers whose blocks they contain. This should make it possible to (1) write clusters using a page mapping instead of malloc, if desired, and (2) schedule blocks for rewriting (somewhere else) if a write error occurs. Code is present to use pagemove() to construct the clusters but that is untested and will go away anyway in favor of page mapping. * DEBUG now keeps a log of Ifile writes, so that any lingering instances of the "dirty bufs" problem can be properly debugged. * Keep track of whether the Ifile has been dirtied by various routines that can be called by lfs_segwrite, and loop on that until it is clean, for a checkpoint. Checkpoints need to be squeaky clean. * Warn the user (once) if the Ifile grows larger than is reasonable for their buffer cache. Both lfs_mountfs and lfs_unmount check since the Ifile can grow. * If an inode is not found in a disk block, try rereading the block, under the assumption that the block was copied to a cluster and then freed. * Protect WRITEINPROG() with splbio() to fix a hang in lfs_update.
2002-05-15 00:03:53 +04:00
printf("lfs_countlocked: %s: adjusted byte count from %ld to %ld\n",
msg, *bytes, size);
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
#endif
New sources should leave the LFS in a more-or-less working state. Changes include: - DIROP segregation is enabled, and greater care is taken to make sure that a checkpoint completes. Fsck is not needed to remount the filesystem. - Several checks to make sure that the LFS subsystem does not overuse various resources (memory, in particular). - The cleaner routines, lfs_markv in particular, are completely rewritten. A buffer overflow is removed. Greater care is taken to ensure that inodes come from where lfs_cleanerd say they come from (so we know nothing has changed since lfs_bmapv was called). - Fragment allocation is fixed, so that writes beyond end-of-file do the right thing.
1999-03-10 03:20:00 +03:00
*count = n;
*bytes = size;
return;
Update to 4.4-Lite fs code, with local changes.
1994-06-08 15:41:58 +04:00
}