kernel-fxtec-pro1x/fs/buffer.c

3383 lines
88 KiB
C
Raw Normal View History

/*
* linux/fs/buffer.c
*
* Copyright (C) 1991, 1992, 2002 Linus Torvalds
*/
/*
* Start bdflush() with kernel_thread not syscall - Paul Gortmaker, 12/95
*
* Removed a lot of unnecessary code and simplified things now that
* the buffer cache isn't our primary cache - Andrew Tridgell 12/96
*
* Speed up hash, lru, and free list operations. Use gfp() for allocating
* hash table, use SLAB cache for buffer heads. SMP threading. -DaveM
*
* Added 32k buffer block sizes - these are required older ARM systems. - RMK
*
* async buffer flushing, 1999 Andrea Arcangeli <andrea@suse.de>
*/
#include <linux/kernel.h>
#include <linux/syscalls.h>
#include <linux/fs.h>
#include <linux/mm.h>
#include <linux/percpu.h>
#include <linux/slab.h>
#include <linux/capability.h>
#include <linux/blkdev.h>
#include <linux/file.h>
#include <linux/quotaops.h>
#include <linux/highmem.h>
#include <linux/module.h>
#include <linux/writeback.h>
#include <linux/hash.h>
#include <linux/suspend.h>
#include <linux/buffer_head.h>
#include <linux/task_io_accounting_ops.h>
#include <linux/bio.h>
#include <linux/notifier.h>
#include <linux/cpu.h>
#include <linux/bitops.h>
#include <linux/mpage.h>
[PATCH] spinlock consolidation This patch (written by me and also containing many suggestions of Arjan van de Ven) does a major cleanup of the spinlock code. It does the following things: - consolidates and enhances the spinlock/rwlock debugging code - simplifies the asm/spinlock.h files - encapsulates the raw spinlock type and moves generic spinlock features (such as ->break_lock) into the generic code. - cleans up the spinlock code hierarchy to get rid of the spaghetti. Most notably there's now only a single variant of the debugging code, located in lib/spinlock_debug.c. (previously we had one SMP debugging variant per architecture, plus a separate generic one for UP builds) Also, i've enhanced the rwlock debugging facility, it will now track write-owners. There is new spinlock-owner/CPU-tracking on SMP builds too. All locks have lockup detection now, which will work for both soft and hard spin/rwlock lockups. The arch-level include files now only contain the minimally necessary subset of the spinlock code - all the rest that can be generalized now lives in the generic headers: include/asm-i386/spinlock_types.h | 16 include/asm-x86_64/spinlock_types.h | 16 I have also split up the various spinlock variants into separate files, making it easier to see which does what. The new layout is: SMP | UP ----------------------------|----------------------------------- asm/spinlock_types_smp.h | linux/spinlock_types_up.h linux/spinlock_types.h | linux/spinlock_types.h asm/spinlock_smp.h | linux/spinlock_up.h linux/spinlock_api_smp.h | linux/spinlock_api_up.h linux/spinlock.h | linux/spinlock.h /* * here's the role of the various spinlock/rwlock related include files: * * on SMP builds: * * asm/spinlock_types.h: contains the raw_spinlock_t/raw_rwlock_t and the * initializers * * linux/spinlock_types.h: * defines the generic type and initializers * * asm/spinlock.h: contains the __raw_spin_*()/etc. lowlevel * implementations, mostly inline assembly code * * (also included on UP-debug builds:) * * linux/spinlock_api_smp.h: * contains the prototypes for the _spin_*() APIs. * * linux/spinlock.h: builds the final spin_*() APIs. * * on UP builds: * * linux/spinlock_type_up.h: * contains the generic, simplified UP spinlock type. * (which is an empty structure on non-debug builds) * * linux/spinlock_types.h: * defines the generic type and initializers * * linux/spinlock_up.h: * contains the __raw_spin_*()/etc. version of UP * builds. (which are NOPs on non-debug, non-preempt * builds) * * (included on UP-non-debug builds:) * * linux/spinlock_api_up.h: * builds the _spin_*() APIs. * * linux/spinlock.h: builds the final spin_*() APIs. */ All SMP and UP architectures are converted by this patch. arm, i386, ia64, ppc, ppc64, s390/s390x, x64 was build-tested via crosscompilers. m32r, mips, sh, sparc, have not been tested yet, but should be mostly fine. From: Grant Grundler <grundler@parisc-linux.org> Booted and lightly tested on a500-44 (64-bit, SMP kernel, dual CPU). Builds 32-bit SMP kernel (not booted or tested). I did not try to build non-SMP kernels. That should be trivial to fix up later if necessary. I converted bit ops atomic_hash lock to raw_spinlock_t. Doing so avoids some ugly nesting of linux/*.h and asm/*.h files. Those particular locks are well tested and contained entirely inside arch specific code. I do NOT expect any new issues to arise with them. If someone does ever need to use debug/metrics with them, then they will need to unravel this hairball between spinlocks, atomic ops, and bit ops that exist only because parisc has exactly one atomic instruction: LDCW (load and clear word). From: "Luck, Tony" <tony.luck@intel.com> ia64 fix Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Arjan van de Ven <arjanv@infradead.org> Signed-off-by: Grant Grundler <grundler@parisc-linux.org> Cc: Matthew Wilcox <willy@debian.org> Signed-off-by: Hirokazu Takata <takata@linux-m32r.org> Signed-off-by: Mikael Pettersson <mikpe@csd.uu.se> Signed-off-by: Benoit Boissinot <benoit.boissinot@ens-lyon.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-09-10 01:25:56 -06:00
#include <linux/bit_spinlock.h>
static int fsync_buffers_list(spinlock_t *lock, struct list_head *list);
#define BH_ENTRY(list) list_entry((list), struct buffer_head, b_assoc_buffers)
inline void
init_buffer(struct buffer_head *bh, bh_end_io_t *handler, void *private)
{
bh->b_end_io = handler;
bh->b_private = private;
}
static int sync_buffer(void *word)
{
struct block_device *bd;
struct buffer_head *bh
= container_of(word, struct buffer_head, b_state);
smp_mb();
bd = bh->b_bdev;
if (bd)
blk_run_address_space(bd->bd_inode->i_mapping);
io_schedule();
return 0;
}
void __lock_buffer(struct buffer_head *bh)
{
wait_on_bit_lock(&bh->b_state, BH_Lock, sync_buffer,
TASK_UNINTERRUPTIBLE);
}
EXPORT_SYMBOL(__lock_buffer);
void unlock_buffer(struct buffer_head *bh)
{
clear_bit_unlock(BH_Lock, &bh->b_state);
smp_mb__after_clear_bit();
wake_up_bit(&bh->b_state, BH_Lock);
}
/*
* Block until a buffer comes unlocked. This doesn't stop it
* from becoming locked again - you have to lock it yourself
* if you want to preserve its state.
*/
void __wait_on_buffer(struct buffer_head * bh)
{
wait_on_bit(&bh->b_state, BH_Lock, sync_buffer, TASK_UNINTERRUPTIBLE);
}
static void
__clear_page_buffers(struct page *page)
{
ClearPagePrivate(page);
[PATCH] mm: split page table lock Christoph Lameter demonstrated very poor scalability on the SGI 512-way, with a many-threaded application which concurrently initializes different parts of a large anonymous area. This patch corrects that, by using a separate spinlock per page table page, to guard the page table entries in that page, instead of using the mm's single page_table_lock. (But even then, page_table_lock is still used to guard page table allocation, and anon_vma allocation.) In this implementation, the spinlock is tucked inside the struct page of the page table page: with a BUILD_BUG_ON in case it overflows - which it would in the case of 32-bit PA-RISC with spinlock debugging enabled. Splitting the lock is not quite for free: another cacheline access. Ideally, I suppose we would use split ptlock only for multi-threaded processes on multi-cpu machines; but deciding that dynamically would have its own costs. So for now enable it by config, at some number of cpus - since the Kconfig language doesn't support inequalities, let preprocessor compare that with NR_CPUS. But I don't think it's worth being user-configurable: for good testing of both split and unsplit configs, split now at 4 cpus, and perhaps change that to 8 later. There is a benefit even for singly threaded processes: kswapd can be attacking one part of the mm while another part is busy faulting. Signed-off-by: Hugh Dickins <hugh@veritas.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-10-29 19:16:40 -06:00
set_page_private(page, 0);
page_cache_release(page);
}
block: Supress Buffer I/O errors when SCSI REQ_QUIET flag set Allow the scsi request REQ_QUIET flag to be propagated to the buffer file system layer. The basic ideas is to pass the flag from the scsi request to the bio (block IO) and then to the buffer layer. The buffer layer can then suppress needless printks. This patch declutters the kernel log by removed the 40-50 (per lun) buffer io error messages seen during a boot in my multipath setup . It is a good chance any real errors will be missed in the "noise" it the logs without this patch. During boot I see blocks of messages like " __ratelimit: 211 callbacks suppressed Buffer I/O error on device sdm, logical block 5242879 Buffer I/O error on device sdm, logical block 5242879 Buffer I/O error on device sdm, logical block 5242847 Buffer I/O error on device sdm, logical block 1 Buffer I/O error on device sdm, logical block 5242878 Buffer I/O error on device sdm, logical block 5242879 Buffer I/O error on device sdm, logical block 5242879 Buffer I/O error on device sdm, logical block 5242879 Buffer I/O error on device sdm, logical block 5242879 Buffer I/O error on device sdm, logical block 5242872 " in my logs. My disk environment is multipath fiber channel using the SCSI_DH_RDAC code and multipathd. This topology includes an "active" and "ghost" path for each lun. IO's to the "ghost" path will never complete and the SCSI layer, via the scsi device handler rdac code, quick returns the IOs to theses paths and sets the REQ_QUIET scsi flag to suppress the scsi layer messages. I am wanting to extend the QUIET behavior to include the buffer file system layer to deal with these errors as well. I have been running this patch for a while now on several boxes without issue. A few runs of bonnie++ show no noticeable difference in performance in my setup. Thanks for John Stultz for the quiet_error finalization. Submitted-by: Keith Mannthey <kmannth@us.ibm.com> Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
2008-11-25 02:24:35 -07:00
static int quiet_error(struct buffer_head *bh)
{
if (!test_bit(BH_Quiet, &bh->b_state) && printk_ratelimit())
return 0;
return 1;
}
static void buffer_io_error(struct buffer_head *bh)
{
char b[BDEVNAME_SIZE];
printk(KERN_ERR "Buffer I/O error on device %s, logical block %Lu\n",
bdevname(bh->b_bdev, b),
(unsigned long long)bh->b_blocknr);
}
/*
* End-of-IO handler helper function which does not touch the bh after
* unlocking it.
* Note: unlock_buffer() sort-of does touch the bh after unlocking it, but
* a race there is benign: unlock_buffer() only use the bh's address for
* hashing after unlocking the buffer, so it doesn't actually touch the bh
* itself.
*/
static void __end_buffer_read_notouch(struct buffer_head *bh, int uptodate)
{
if (uptodate) {
set_buffer_uptodate(bh);
} else {
/* This happens, due to failed READA attempts. */
clear_buffer_uptodate(bh);
}
unlock_buffer(bh);
}
/*
* Default synchronous end-of-IO handler.. Just mark it up-to-date and
* unlock the buffer. This is what ll_rw_block uses too.
*/
void end_buffer_read_sync(struct buffer_head *bh, int uptodate)
{
__end_buffer_read_notouch(bh, uptodate);
put_bh(bh);
}
void end_buffer_write_sync(struct buffer_head *bh, int uptodate)
{
char b[BDEVNAME_SIZE];
if (uptodate) {
set_buffer_uptodate(bh);
} else {
block: Supress Buffer I/O errors when SCSI REQ_QUIET flag set Allow the scsi request REQ_QUIET flag to be propagated to the buffer file system layer. The basic ideas is to pass the flag from the scsi request to the bio (block IO) and then to the buffer layer. The buffer layer can then suppress needless printks. This patch declutters the kernel log by removed the 40-50 (per lun) buffer io error messages seen during a boot in my multipath setup . It is a good chance any real errors will be missed in the "noise" it the logs without this patch. During boot I see blocks of messages like " __ratelimit: 211 callbacks suppressed Buffer I/O error on device sdm, logical block 5242879 Buffer I/O error on device sdm, logical block 5242879 Buffer I/O error on device sdm, logical block 5242847 Buffer I/O error on device sdm, logical block 1 Buffer I/O error on device sdm, logical block 5242878 Buffer I/O error on device sdm, logical block 5242879 Buffer I/O error on device sdm, logical block 5242879 Buffer I/O error on device sdm, logical block 5242879 Buffer I/O error on device sdm, logical block 5242879 Buffer I/O error on device sdm, logical block 5242872 " in my logs. My disk environment is multipath fiber channel using the SCSI_DH_RDAC code and multipathd. This topology includes an "active" and "ghost" path for each lun. IO's to the "ghost" path will never complete and the SCSI layer, via the scsi device handler rdac code, quick returns the IOs to theses paths and sets the REQ_QUIET scsi flag to suppress the scsi layer messages. I am wanting to extend the QUIET behavior to include the buffer file system layer to deal with these errors as well. I have been running this patch for a while now on several boxes without issue. A few runs of bonnie++ show no noticeable difference in performance in my setup. Thanks for John Stultz for the quiet_error finalization. Submitted-by: Keith Mannthey <kmannth@us.ibm.com> Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
2008-11-25 02:24:35 -07:00
if (!buffer_eopnotsupp(bh) && !quiet_error(bh)) {
buffer_io_error(bh);
printk(KERN_WARNING "lost page write due to "
"I/O error on %s\n",
bdevname(bh->b_bdev, b));
}
set_buffer_write_io_error(bh);
clear_buffer_uptodate(bh);
}
unlock_buffer(bh);
put_bh(bh);
}
/*
* Write out and wait upon all the dirty data associated with a block
* device via its mapping. Does not take the superblock lock.
*/
int sync_blockdev(struct block_device *bdev)
{
int ret = 0;
if (bdev)
ret = filemap_write_and_wait(bdev->bd_inode->i_mapping);
return ret;
}
EXPORT_SYMBOL(sync_blockdev);
/*
* Write out and wait upon all dirty data associated with this
* device. Filesystem data as well as the underlying block
* device. Takes the superblock lock.
*/
int fsync_bdev(struct block_device *bdev)
{
struct super_block *sb = get_super(bdev);
if (sb) {
int res = fsync_super(sb);
drop_super(sb);
return res;
}
return sync_blockdev(bdev);
}
/**
* freeze_bdev -- lock a filesystem and force it into a consistent state
* @bdev: blockdevice to lock
*
* This takes the block device bd_mount_sem to make sure no new mounts
* happen on bdev until thaw_bdev() is called.
* If a superblock is found on this device, we take the s_umount semaphore
* on it to make sure nobody unmounts until the snapshot creation is done.
*/
struct super_block *freeze_bdev(struct block_device *bdev)
{
struct super_block *sb;
down(&bdev->bd_mount_sem);
sb = get_super(bdev);
if (sb && !(sb->s_flags & MS_RDONLY)) {
sb->s_frozen = SB_FREEZE_WRITE;
smp_wmb();
__fsync_super(sb);
sb->s_frozen = SB_FREEZE_TRANS;
smp_wmb();
sync_blockdev(sb->s_bdev);
filesystem freeze: add error handling of write_super_lockfs/unlockfs Currently, ext3 in mainline Linux doesn't have the freeze feature which suspends write requests. So, we cannot take a backup which keeps the filesystem's consistency with the storage device's features (snapshot and replication) while it is mounted. In many case, a commercial filesystem (e.g. VxFS) has the freeze feature and it would be used to get the consistent backup. If Linux's standard filesystem ext3 has the freeze feature, we can do it without a commercial filesystem. So I have implemented the ioctls of the freeze feature. I think we can take the consistent backup with the following steps. 1. Freeze the filesystem with the freeze ioctl. 2. Separate the replication volume or create the snapshot with the storage device's feature. 3. Unfreeze the filesystem with the unfreeze ioctl. 4. Take the backup from the separated replication volume or the snapshot. This patch: VFS: Changed the type of write_super_lockfs and unlockfs from "void" to "int" so that they can return an error. Rename write_super_lockfs and unlockfs of the super block operation freeze_fs and unfreeze_fs to avoid a confusion. ext3, ext4, xfs, gfs2, jfs: Changed the type of write_super_lockfs and unlockfs from "void" to "int" so that write_super_lockfs returns an error if needed, and unlockfs always returns 0. reiserfs: Changed the type of write_super_lockfs and unlockfs from "void" to "int" so that they always return 0 (success) to keep a current behavior. Signed-off-by: Takashi Sato <t-sato@yk.jp.nec.com> Signed-off-by: Masayuki Hamaguchi <m-hamaguchi@ys.jp.nec.com> Cc: <xfs-masters@oss.sgi.com> Cc: <linux-ext4@vger.kernel.org> Cc: Christoph Hellwig <hch@lst.de> Cc: Dave Kleikamp <shaggy@austin.ibm.com> Cc: Dave Chinner <david@fromorbit.com> Cc: Alasdair G Kergon <agk@redhat.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-09 17:40:58 -07:00
if (sb->s_op->freeze_fs)
sb->s_op->freeze_fs(sb);
}
sync_blockdev(bdev);
return sb; /* thaw_bdev releases s->s_umount and bd_mount_sem */
}
EXPORT_SYMBOL(freeze_bdev);
/**
* thaw_bdev -- unlock filesystem
* @bdev: blockdevice to unlock
* @sb: associated superblock
*
* Unlocks the filesystem and marks it writeable again after freeze_bdev().
*/
void thaw_bdev(struct block_device *bdev, struct super_block *sb)
{
if (sb) {
BUG_ON(sb->s_bdev != bdev);
filesystem freeze: add error handling of write_super_lockfs/unlockfs Currently, ext3 in mainline Linux doesn't have the freeze feature which suspends write requests. So, we cannot take a backup which keeps the filesystem's consistency with the storage device's features (snapshot and replication) while it is mounted. In many case, a commercial filesystem (e.g. VxFS) has the freeze feature and it would be used to get the consistent backup. If Linux's standard filesystem ext3 has the freeze feature, we can do it without a commercial filesystem. So I have implemented the ioctls of the freeze feature. I think we can take the consistent backup with the following steps. 1. Freeze the filesystem with the freeze ioctl. 2. Separate the replication volume or create the snapshot with the storage device's feature. 3. Unfreeze the filesystem with the unfreeze ioctl. 4. Take the backup from the separated replication volume or the snapshot. This patch: VFS: Changed the type of write_super_lockfs and unlockfs from "void" to "int" so that they can return an error. Rename write_super_lockfs and unlockfs of the super block operation freeze_fs and unfreeze_fs to avoid a confusion. ext3, ext4, xfs, gfs2, jfs: Changed the type of write_super_lockfs and unlockfs from "void" to "int" so that write_super_lockfs returns an error if needed, and unlockfs always returns 0. reiserfs: Changed the type of write_super_lockfs and unlockfs from "void" to "int" so that they always return 0 (success) to keep a current behavior. Signed-off-by: Takashi Sato <t-sato@yk.jp.nec.com> Signed-off-by: Masayuki Hamaguchi <m-hamaguchi@ys.jp.nec.com> Cc: <xfs-masters@oss.sgi.com> Cc: <linux-ext4@vger.kernel.org> Cc: Christoph Hellwig <hch@lst.de> Cc: Dave Kleikamp <shaggy@austin.ibm.com> Cc: Dave Chinner <david@fromorbit.com> Cc: Alasdair G Kergon <agk@redhat.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-09 17:40:58 -07:00
if (sb->s_op->unfreeze_fs)
sb->s_op->unfreeze_fs(sb);
sb->s_frozen = SB_UNFROZEN;
smp_wmb();
wake_up(&sb->s_wait_unfrozen);
drop_super(sb);
}
up(&bdev->bd_mount_sem);
}
EXPORT_SYMBOL(thaw_bdev);
/*
* Various filesystems appear to want __find_get_block to be non-blocking.
* But it's the page lock which protects the buffers. To get around this,
* we get exclusion from try_to_free_buffers with the blockdev mapping's
* private_lock.
*
* Hack idea: for the blockdev mapping, i_bufferlist_lock contention
* may be quite high. This code could TryLock the page, and if that
* succeeds, there is no need to take private_lock. (But if
* private_lock is contended then so is mapping->tree_lock).
*/
static struct buffer_head *
__find_get_block_slow(struct block_device *bdev, sector_t block)
{
struct inode *bd_inode = bdev->bd_inode;
struct address_space *bd_mapping = bd_inode->i_mapping;
struct buffer_head *ret = NULL;
pgoff_t index;
struct buffer_head *bh;
struct buffer_head *head;
struct page *page;
int all_mapped = 1;
index = block >> (PAGE_CACHE_SHIFT - bd_inode->i_blkbits);
page = find_get_page(bd_mapping, index);
if (!page)
goto out;
spin_lock(&bd_mapping->private_lock);
if (!page_has_buffers(page))
goto out_unlock;
head = page_buffers(page);
bh = head;
do {
if (bh->b_blocknr == block) {
ret = bh;
get_bh(bh);
goto out_unlock;
}
if (!buffer_mapped(bh))
all_mapped = 0;
bh = bh->b_this_page;
} while (bh != head);
/* we might be here because some of the buffers on this page are
* not mapped. This is due to various races between
* file io on the block device and getblk. It gets dealt with
* elsewhere, don't buffer_error if we had some unmapped buffers
*/
if (all_mapped) {
printk("__find_get_block_slow() failed. "
"block=%llu, b_blocknr=%llu\n",
(unsigned long long)block,
(unsigned long long)bh->b_blocknr);
printk("b_state=0x%08lx, b_size=%zu\n",
bh->b_state, bh->b_size);
printk("device blocksize: %d\n", 1 << bd_inode->i_blkbits);
}
out_unlock:
spin_unlock(&bd_mapping->private_lock);
page_cache_release(page);
out:
return ret;
}
/* If invalidate_buffers() will trash dirty buffers, it means some kind
of fs corruption is going on. Trashing dirty data always imply losing
information that was supposed to be just stored on the physical layer
by the user.
Thus invalidate_buffers in general usage is not allwowed to trash
dirty buffers. For example ioctl(FLSBLKBUF) expects dirty data to
be preserved. These buffers are simply skipped.
We also skip buffers which are still in use. For example this can
happen if a userspace program is reading the block device.
NOTE: In the case where the user removed a removable-media-disk even if
there's still dirty data not synced on disk (due a bug in the device driver
or due an error of the user), by not destroying the dirty buffers we could
generate corruption also on the next media inserted, thus a parameter is
necessary to handle this case in the most safe way possible (trying
to not corrupt also the new disk inserted with the data belonging to
the old now corrupted disk). Also for the ramdisk the natural thing
to do in order to release the ramdisk memory is to destroy dirty buffers.
These are two special cases. Normal usage imply the device driver
to issue a sync on the device (without waiting I/O completion) and
then an invalidate_buffers call that doesn't trash dirty buffers.
For handling cache coherency with the blkdev pagecache the 'update' case
is been introduced. It is needed to re-read from disk any pinned
buffer. NOTE: re-reading from disk is destructive so we can do it only
when we assume nobody is changing the buffercache under our I/O and when
we think the disk contains more recent information than the buffercache.
The update == 1 pass marks the buffers we need to update, the update == 2
pass does the actual I/O. */
void invalidate_bdev(struct block_device *bdev)
{
struct address_space *mapping = bdev->bd_inode->i_mapping;
if (mapping->nrpages == 0)
return;
invalidate_bh_lrus();
invalidate_mapping_pages(mapping, 0, -1);
}
/*
* Kick pdflush then try to free up some ZONE_NORMAL memory.
*/
static void free_more_memory(void)
{
struct zone *zone;
int nid;
wakeup_pdflush(1024);
yield();
for_each_online_node(nid) {
(void)first_zones_zonelist(node_zonelist(nid, GFP_NOFS),
gfp_zone(GFP_NOFS), NULL,
&zone);
if (zone)
try_to_free_pages(node_zonelist(nid, GFP_NOFS), 0,
GFP_NOFS);
}
}
/*
* I/O completion handler for block_read_full_page() - pages
* which come unlocked at the end of I/O.
*/
static void end_buffer_async_read(struct buffer_head *bh, int uptodate)
{
unsigned long flags;
struct buffer_head *first;
struct buffer_head *tmp;
struct page *page;
int page_uptodate = 1;
BUG_ON(!buffer_async_read(bh));
page = bh->b_page;
if (uptodate) {
set_buffer_uptodate(bh);
} else {
clear_buffer_uptodate(bh);
block: Supress Buffer I/O errors when SCSI REQ_QUIET flag set Allow the scsi request REQ_QUIET flag to be propagated to the buffer file system layer. The basic ideas is to pass the flag from the scsi request to the bio (block IO) and then to the buffer layer. The buffer layer can then suppress needless printks. This patch declutters the kernel log by removed the 40-50 (per lun) buffer io error messages seen during a boot in my multipath setup . It is a good chance any real errors will be missed in the "noise" it the logs without this patch. During boot I see blocks of messages like " __ratelimit: 211 callbacks suppressed Buffer I/O error on device sdm, logical block 5242879 Buffer I/O error on device sdm, logical block 5242879 Buffer I/O error on device sdm, logical block 5242847 Buffer I/O error on device sdm, logical block 1 Buffer I/O error on device sdm, logical block 5242878 Buffer I/O error on device sdm, logical block 5242879 Buffer I/O error on device sdm, logical block 5242879 Buffer I/O error on device sdm, logical block 5242879 Buffer I/O error on device sdm, logical block 5242879 Buffer I/O error on device sdm, logical block 5242872 " in my logs. My disk environment is multipath fiber channel using the SCSI_DH_RDAC code and multipathd. This topology includes an "active" and "ghost" path for each lun. IO's to the "ghost" path will never complete and the SCSI layer, via the scsi device handler rdac code, quick returns the IOs to theses paths and sets the REQ_QUIET scsi flag to suppress the scsi layer messages. I am wanting to extend the QUIET behavior to include the buffer file system layer to deal with these errors as well. I have been running this patch for a while now on several boxes without issue. A few runs of bonnie++ show no noticeable difference in performance in my setup. Thanks for John Stultz for the quiet_error finalization. Submitted-by: Keith Mannthey <kmannth@us.ibm.com> Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
2008-11-25 02:24:35 -07:00
if (!quiet_error(bh))
buffer_io_error(bh);
SetPageError(page);
}
/*
* Be _very_ careful from here on. Bad things can happen if
* two buffer heads end IO at almost the same time and both
* decide that the page is now completely done.
*/
first = page_buffers(page);
local_irq_save(flags);
bit_spin_lock(BH_Uptodate_Lock, &first->b_state);
clear_buffer_async_read(bh);
unlock_buffer(bh);
tmp = bh;
do {
if (!buffer_uptodate(tmp))
page_uptodate = 0;
if (buffer_async_read(tmp)) {
BUG_ON(!buffer_locked(tmp));
goto still_busy;
}
tmp = tmp->b_this_page;
} while (tmp != bh);
bit_spin_unlock(BH_Uptodate_Lock, &first->b_state);
local_irq_restore(flags);
/*
* If none of the buffers had errors and they are all
* uptodate then we can set the page uptodate.
*/
if (page_uptodate && !PageError(page))
SetPageUptodate(page);
unlock_page(page);
return;
still_busy:
bit_spin_unlock(BH_Uptodate_Lock, &first->b_state);
local_irq_restore(flags);
return;
}
/*
* Completion handler for block_write_full_page() - pages which are unlocked
* during I/O, and which have PageWriteback cleared upon I/O completion.
*/
static void end_buffer_async_write(struct buffer_head *bh, int uptodate)
{
char b[BDEVNAME_SIZE];
unsigned long flags;
struct buffer_head *first;
struct buffer_head *tmp;
struct page *page;
BUG_ON(!buffer_async_write(bh));
page = bh->b_page;
if (uptodate) {
set_buffer_uptodate(bh);
} else {
block: Supress Buffer I/O errors when SCSI REQ_QUIET flag set Allow the scsi request REQ_QUIET flag to be propagated to the buffer file system layer. The basic ideas is to pass the flag from the scsi request to the bio (block IO) and then to the buffer layer. The buffer layer can then suppress needless printks. This patch declutters the kernel log by removed the 40-50 (per lun) buffer io error messages seen during a boot in my multipath setup . It is a good chance any real errors will be missed in the "noise" it the logs without this patch. During boot I see blocks of messages like " __ratelimit: 211 callbacks suppressed Buffer I/O error on device sdm, logical block 5242879 Buffer I/O error on device sdm, logical block 5242879 Buffer I/O error on device sdm, logical block 5242847 Buffer I/O error on device sdm, logical block 1 Buffer I/O error on device sdm, logical block 5242878 Buffer I/O error on device sdm, logical block 5242879 Buffer I/O error on device sdm, logical block 5242879 Buffer I/O error on device sdm, logical block 5242879 Buffer I/O error on device sdm, logical block 5242879 Buffer I/O error on device sdm, logical block 5242872 " in my logs. My disk environment is multipath fiber channel using the SCSI_DH_RDAC code and multipathd. This topology includes an "active" and "ghost" path for each lun. IO's to the "ghost" path will never complete and the SCSI layer, via the scsi device handler rdac code, quick returns the IOs to theses paths and sets the REQ_QUIET scsi flag to suppress the scsi layer messages. I am wanting to extend the QUIET behavior to include the buffer file system layer to deal with these errors as well. I have been running this patch for a while now on several boxes without issue. A few runs of bonnie++ show no noticeable difference in performance in my setup. Thanks for John Stultz for the quiet_error finalization. Submitted-by: Keith Mannthey <kmannth@us.ibm.com> Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
2008-11-25 02:24:35 -07:00
if (!quiet_error(bh)) {
buffer_io_error(bh);
printk(KERN_WARNING "lost page write due to "
"I/O error on %s\n",
bdevname(bh->b_bdev, b));
}
set_bit(AS_EIO, &page->mapping->flags);
set_buffer_write_io_error(bh);
clear_buffer_uptodate(bh);
SetPageError(page);
}
first = page_buffers(page);
local_irq_save(flags);
bit_spin_lock(BH_Uptodate_Lock, &first->b_state);
clear_buffer_async_write(bh);
unlock_buffer(bh);
tmp = bh->b_this_page;
while (tmp != bh) {
if (buffer_async_write(tmp)) {
BUG_ON(!buffer_locked(tmp));
goto still_busy;
}
tmp = tmp->b_this_page;
}
bit_spin_unlock(BH_Uptodate_Lock, &first->b_state);
local_irq_restore(flags);
end_page_writeback(page);
return;
still_busy:
bit_spin_unlock(BH_Uptodate_Lock, &first->b_state);
local_irq_restore(flags);
return;
}
/*
* If a page's buffers are under async readin (end_buffer_async_read
* completion) then there is a possibility that another thread of
* control could lock one of the buffers after it has completed
* but while some of the other buffers have not completed. This
* locked buffer would confuse end_buffer_async_read() into not unlocking
* the page. So the absence of BH_Async_Read tells end_buffer_async_read()
* that this buffer is not under async I/O.
*
* The page comes unlocked when it has no locked buffer_async buffers
* left.
*
* PageLocked prevents anyone starting new async I/O reads any of
* the buffers.
*
* PageWriteback is used to prevent simultaneous writeout of the same
* page.
*
* PageLocked prevents anyone from starting writeback of a page which is
* under read I/O (PageWriteback is only ever set against a locked page).
*/
static void mark_buffer_async_read(struct buffer_head *bh)
{
bh->b_end_io = end_buffer_async_read;
set_buffer_async_read(bh);
}
void mark_buffer_async_write(struct buffer_head *bh)
{
bh->b_end_io = end_buffer_async_write;
set_buffer_async_write(bh);
}
EXPORT_SYMBOL(mark_buffer_async_write);
/*
* fs/buffer.c contains helper functions for buffer-backed address space's
* fsync functions. A common requirement for buffer-based filesystems is
* that certain data from the backing blockdev needs to be written out for
* a successful fsync(). For example, ext2 indirect blocks need to be
* written back and waited upon before fsync() returns.
*
* The functions mark_buffer_inode_dirty(), fsync_inode_buffers(),
* inode_has_buffers() and invalidate_inode_buffers() are provided for the
* management of a list of dependent buffers at ->i_mapping->private_list.
*
* Locking is a little subtle: try_to_free_buffers() will remove buffers
* from their controlling inode's queue when they are being freed. But
* try_to_free_buffers() will be operating against the *blockdev* mapping
* at the time, not against the S_ISREG file which depends on those buffers.
* So the locking for private_list is via the private_lock in the address_space
* which backs the buffers. Which is different from the address_space
* against which the buffers are listed. So for a particular address_space,
* mapping->private_lock does *not* protect mapping->private_list! In fact,
* mapping->private_list will always be protected by the backing blockdev's
* ->private_lock.
*
* Which introduces a requirement: all buffers on an address_space's
* ->private_list must be from the same address_space: the blockdev's.
*
* address_spaces which do not place buffers at ->private_list via these
* utility functions are free to use private_lock and private_list for
* whatever they want. The only requirement is that list_empty(private_list)
* be true at clear_inode() time.
*
* FIXME: clear_inode should not call invalidate_inode_buffers(). The
* filesystems should do that. invalidate_inode_buffers() should just go
* BUG_ON(!list_empty).
*
* FIXME: mark_buffer_dirty_inode() is a data-plane operation. It should
* take an address_space, not an inode. And it should be called
* mark_buffer_dirty_fsync() to clearly define why those buffers are being
* queued up.
*
* FIXME: mark_buffer_dirty_inode() doesn't need to add the buffer to the
* list if it is already on a list. Because if the buffer is on a list,
* it *must* already be on the right one. If not, the filesystem is being
* silly. This will save a ton of locking. But first we have to ensure
* that buffers are taken *off* the old inode's list when they are freed
* (presumably in truncate). That requires careful auditing of all
* filesystems (do it inside bforget()). It could also be done by bringing
* b_inode back.
*/
/*
* The buffer's backing address_space's private_lock must be held
*/
static void __remove_assoc_queue(struct buffer_head *bh)
{
list_del_init(&bh->b_assoc_buffers);
WARN_ON(!bh->b_assoc_map);
if (buffer_write_io_error(bh))
set_bit(AS_EIO, &bh->b_assoc_map->flags);
bh->b_assoc_map = NULL;
}
int inode_has_buffers(struct inode *inode)
{
return !list_empty(&inode->i_data.private_list);
}
/*
* osync is designed to support O_SYNC io. It waits synchronously for
* all already-submitted IO to complete, but does not queue any new
* writes to the disk.
*
* To do O_SYNC writes, just queue the buffer writes with ll_rw_block as
* you dirty the buffers, and then use osync_inode_buffers to wait for
* completion. Any other dirty buffers which are not yet queued for
* write will not be flushed to disk by the osync.
*/
static int osync_buffers_list(spinlock_t *lock, struct list_head *list)
{
struct buffer_head *bh;
struct list_head *p;
int err = 0;
spin_lock(lock);
repeat:
list_for_each_prev(p, list) {
bh = BH_ENTRY(p);
if (buffer_locked(bh)) {
get_bh(bh);
spin_unlock(lock);
wait_on_buffer(bh);
if (!buffer_uptodate(bh))
err = -EIO;
brelse(bh);
spin_lock(lock);
goto repeat;
}
}
spin_unlock(lock);
return err;
}
/**
* sync_mapping_buffers - write out & wait upon a mapping's "associated" buffers
* @mapping: the mapping which wants those buffers written
*
* Starts I/O against the buffers at mapping->private_list, and waits upon
* that I/O.
*
* Basically, this is a convenience function for fsync().
* @mapping is a file or directory which needs those buffers to be written for
* a successful fsync().
*/
int sync_mapping_buffers(struct address_space *mapping)
{
struct address_space *buffer_mapping = mapping->assoc_mapping;
if (buffer_mapping == NULL || list_empty(&mapping->private_list))
return 0;
return fsync_buffers_list(&buffer_mapping->private_lock,
&mapping->private_list);
}
EXPORT_SYMBOL(sync_mapping_buffers);
/*
* Called when we've recently written block `bblock', and it is known that
* `bblock' was for a buffer_boundary() buffer. This means that the block at
* `bblock + 1' is probably a dirty indirect block. Hunt it down and, if it's
* dirty, schedule it for IO. So that indirects merge nicely with their data.
*/
void write_boundary_block(struct block_device *bdev,
sector_t bblock, unsigned blocksize)
{
struct buffer_head *bh = __find_get_block(bdev, bblock + 1, blocksize);
if (bh) {
if (buffer_dirty(bh))
ll_rw_block(WRITE, 1, &bh);
put_bh(bh);
}
}
void mark_buffer_dirty_inode(struct buffer_head *bh, struct inode *inode)
{
struct address_space *mapping = inode->i_mapping;
struct address_space *buffer_mapping = bh->b_page->mapping;
mark_buffer_dirty(bh);
if (!mapping->assoc_mapping) {
mapping->assoc_mapping = buffer_mapping;
} else {
BUG_ON(mapping->assoc_mapping != buffer_mapping);
}
buffer_head: fix private_list handling There are two possible races in handling of private_list in buffer cache. 1) When fsync_buffers_list() processes a private_list, it clears b_assoc_mapping and moves buffer to its private list. Now drop_buffers() comes, sees a buffer is on list so it calls __remove_assoc_queue() which complains about b_assoc_mapping being cleared (as it cannot propagate possible IO error). This race has been actually observed in the wild. 2) When fsync_buffers_list() processes a private_list, mark_buffer_dirty_inode() can be called on bh which is already on the private list of fsync_buffers_list(). As buffer is on some list (note that the check is performed without private_lock), it is not readded to the mapping's private_list and after fsync_buffers_list() finishes, we have a dirty buffer which should be on private_list but it isn't. This race has not been reported, probably because most (but not all) callers of mark_buffer_dirty_inode() hold i_mutex and thus are serialized with fsync(). Fix these issues by not clearing b_assoc_map when fsync_buffers_list() moves buffer to a dedicated list and by reinserting buffer in private_list when it is found dirty after we have submitted buffer for IO. We also change the tests whether a buffer is on a private list from !list_empty(&bh->b_assoc_buffers) to bh->b_assoc_map so that they are single word reads and hence lockless checks are safe. Signed-off-by: Jan Kara <jack@suse.cz> Cc: Nick Piggin <npiggin@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 05:21:59 -07:00
if (!bh->b_assoc_map) {
spin_lock(&buffer_mapping->private_lock);
list_move_tail(&bh->b_assoc_buffers,
&mapping->private_list);
bh->b_assoc_map = mapping;
spin_unlock(&buffer_mapping->private_lock);
}
}
EXPORT_SYMBOL(mark_buffer_dirty_inode);
/*
* Mark the page dirty, and set it dirty in the radix tree, and mark the inode
* dirty.
*
* If warn is true, then emit a warning if the page is not uptodate and has
* not been truncated.
*/
static int __set_page_dirty(struct page *page,
struct address_space *mapping, int warn)
{
if (unlikely(!mapping))
return !TestSetPageDirty(page);
if (TestSetPageDirty(page))
return 0;
spin_lock_irq(&mapping->tree_lock);
if (page->mapping) { /* Race with truncate? */
WARN_ON_ONCE(warn && !PageUptodate(page));
if (mapping_cap_account_dirty(mapping)) {
__inc_zone_page_state(page, NR_FILE_DIRTY);
__inc_bdi_stat(mapping->backing_dev_info,
BDI_RECLAIMABLE);
task_io_account_write(PAGE_CACHE_SIZE);
}
radix_tree_tag_set(&mapping->page_tree,
page_index(page), PAGECACHE_TAG_DIRTY);
}
spin_unlock_irq(&mapping->tree_lock);
__mark_inode_dirty(mapping->host, I_DIRTY_PAGES);
return 1;
}
/*
* Add a page to the dirty page list.
*
* It is a sad fact of life that this function is called from several places
* deeply under spinlocking. It may not sleep.
*
* If the page has buffers, the uptodate buffers are set dirty, to preserve
* dirty-state coherency between the page and the buffers. It the page does
* not have buffers then when they are later attached they will all be set
* dirty.
*
* The buffers are dirtied before the page is dirtied. There's a small race
* window in which a writepage caller may see the page cleanness but not the
* buffer dirtiness. That's fine. If this code were to set the page dirty
* before the buffers, a concurrent writepage caller could clear the page dirty
* bit, see a bunch of clean buffers and we'd end up with dirty buffers/clean
* page on the dirty page list.
*
* We use private_lock to lock against try_to_free_buffers while using the
* page's buffer list. Also use this to protect against clean buffers being
* added to the page after it was set dirty.
*
* FIXME: may need to call ->reservepage here as well. That's rather up to the
* address_space though.
*/
int __set_page_dirty_buffers(struct page *page)
{
struct address_space *mapping = page_mapping(page);
if (unlikely(!mapping))
return !TestSetPageDirty(page);
spin_lock(&mapping->private_lock);
if (page_has_buffers(page)) {
struct buffer_head *head = page_buffers(page);
struct buffer_head *bh = head;
do {
set_buffer_dirty(bh);
bh = bh->b_this_page;
} while (bh != head);
}
spin_unlock(&mapping->private_lock);
return __set_page_dirty(page, mapping, 1);
}
EXPORT_SYMBOL(__set_page_dirty_buffers);
/*
* Write out and wait upon a list of buffers.
*
* We have conflicting pressures: we want to make sure that all
* initially dirty buffers get waited on, but that any subsequently
* dirtied buffers don't. After all, we don't want fsync to last
* forever if somebody is actively writing to the file.
*
* Do this in two main stages: first we copy dirty buffers to a
* temporary inode list, queueing the writes as we go. Then we clean
* up, waiting for those writes to complete.
*
* During this second stage, any subsequent updates to the file may end
* up refiling the buffer on the original inode's dirty list again, so
* there is a chance we will end up with a buffer queued for write but
* not yet completed on that list. So, as a final cleanup we go through
* the osync code to catch these locked, dirty buffers without requeuing
* any newly dirty buffers for write.
*/
static int fsync_buffers_list(spinlock_t *lock, struct list_head *list)
{
struct buffer_head *bh;
struct list_head tmp;
buffer_head: fix private_list handling There are two possible races in handling of private_list in buffer cache. 1) When fsync_buffers_list() processes a private_list, it clears b_assoc_mapping and moves buffer to its private list. Now drop_buffers() comes, sees a buffer is on list so it calls __remove_assoc_queue() which complains about b_assoc_mapping being cleared (as it cannot propagate possible IO error). This race has been actually observed in the wild. 2) When fsync_buffers_list() processes a private_list, mark_buffer_dirty_inode() can be called on bh which is already on the private list of fsync_buffers_list(). As buffer is on some list (note that the check is performed without private_lock), it is not readded to the mapping's private_list and after fsync_buffers_list() finishes, we have a dirty buffer which should be on private_list but it isn't. This race has not been reported, probably because most (but not all) callers of mark_buffer_dirty_inode() hold i_mutex and thus are serialized with fsync(). Fix these issues by not clearing b_assoc_map when fsync_buffers_list() moves buffer to a dedicated list and by reinserting buffer in private_list when it is found dirty after we have submitted buffer for IO. We also change the tests whether a buffer is on a private list from !list_empty(&bh->b_assoc_buffers) to bh->b_assoc_map so that they are single word reads and hence lockless checks are safe. Signed-off-by: Jan Kara <jack@suse.cz> Cc: Nick Piggin <npiggin@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 05:21:59 -07:00
struct address_space *mapping;
int err = 0, err2;
INIT_LIST_HEAD(&tmp);
spin_lock(lock);
while (!list_empty(list)) {
bh = BH_ENTRY(list->next);
buffer_head: fix private_list handling There are two possible races in handling of private_list in buffer cache. 1) When fsync_buffers_list() processes a private_list, it clears b_assoc_mapping and moves buffer to its private list. Now drop_buffers() comes, sees a buffer is on list so it calls __remove_assoc_queue() which complains about b_assoc_mapping being cleared (as it cannot propagate possible IO error). This race has been actually observed in the wild. 2) When fsync_buffers_list() processes a private_list, mark_buffer_dirty_inode() can be called on bh which is already on the private list of fsync_buffers_list(). As buffer is on some list (note that the check is performed without private_lock), it is not readded to the mapping's private_list and after fsync_buffers_list() finishes, we have a dirty buffer which should be on private_list but it isn't. This race has not been reported, probably because most (but not all) callers of mark_buffer_dirty_inode() hold i_mutex and thus are serialized with fsync(). Fix these issues by not clearing b_assoc_map when fsync_buffers_list() moves buffer to a dedicated list and by reinserting buffer in private_list when it is found dirty after we have submitted buffer for IO. We also change the tests whether a buffer is on a private list from !list_empty(&bh->b_assoc_buffers) to bh->b_assoc_map so that they are single word reads and hence lockless checks are safe. Signed-off-by: Jan Kara <jack@suse.cz> Cc: Nick Piggin <npiggin@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 05:21:59 -07:00
mapping = bh->b_assoc_map;
__remove_assoc_queue(bh);
buffer_head: fix private_list handling There are two possible races in handling of private_list in buffer cache. 1) When fsync_buffers_list() processes a private_list, it clears b_assoc_mapping and moves buffer to its private list. Now drop_buffers() comes, sees a buffer is on list so it calls __remove_assoc_queue() which complains about b_assoc_mapping being cleared (as it cannot propagate possible IO error). This race has been actually observed in the wild. 2) When fsync_buffers_list() processes a private_list, mark_buffer_dirty_inode() can be called on bh which is already on the private list of fsync_buffers_list(). As buffer is on some list (note that the check is performed without private_lock), it is not readded to the mapping's private_list and after fsync_buffers_list() finishes, we have a dirty buffer which should be on private_list but it isn't. This race has not been reported, probably because most (but not all) callers of mark_buffer_dirty_inode() hold i_mutex and thus are serialized with fsync(). Fix these issues by not clearing b_assoc_map when fsync_buffers_list() moves buffer to a dedicated list and by reinserting buffer in private_list when it is found dirty after we have submitted buffer for IO. We also change the tests whether a buffer is on a private list from !list_empty(&bh->b_assoc_buffers) to bh->b_assoc_map so that they are single word reads and hence lockless checks are safe. Signed-off-by: Jan Kara <jack@suse.cz> Cc: Nick Piggin <npiggin@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 05:21:59 -07:00
/* Avoid race with mark_buffer_dirty_inode() which does
* a lockless check and we rely on seeing the dirty bit */
smp_mb();
if (buffer_dirty(bh) || buffer_locked(bh)) {
list_add(&bh->b_assoc_buffers, &tmp);
buffer_head: fix private_list handling There are two possible races in handling of private_list in buffer cache. 1) When fsync_buffers_list() processes a private_list, it clears b_assoc_mapping and moves buffer to its private list. Now drop_buffers() comes, sees a buffer is on list so it calls __remove_assoc_queue() which complains about b_assoc_mapping being cleared (as it cannot propagate possible IO error). This race has been actually observed in the wild. 2) When fsync_buffers_list() processes a private_list, mark_buffer_dirty_inode() can be called on bh which is already on the private list of fsync_buffers_list(). As buffer is on some list (note that the check is performed without private_lock), it is not readded to the mapping's private_list and after fsync_buffers_list() finishes, we have a dirty buffer which should be on private_list but it isn't. This race has not been reported, probably because most (but not all) callers of mark_buffer_dirty_inode() hold i_mutex and thus are serialized with fsync(). Fix these issues by not clearing b_assoc_map when fsync_buffers_list() moves buffer to a dedicated list and by reinserting buffer in private_list when it is found dirty after we have submitted buffer for IO. We also change the tests whether a buffer is on a private list from !list_empty(&bh->b_assoc_buffers) to bh->b_assoc_map so that they are single word reads and hence lockless checks are safe. Signed-off-by: Jan Kara <jack@suse.cz> Cc: Nick Piggin <npiggin@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 05:21:59 -07:00
bh->b_assoc_map = mapping;
if (buffer_dirty(bh)) {
get_bh(bh);
spin_unlock(lock);
/*
* Ensure any pending I/O completes so that
* ll_rw_block() actually writes the current
* contents - it is a noop if I/O is still in
* flight on potentially older contents.
*/
Properly notify block layer of sync writes fsync_buffers_list() and sync_dirty_buffer() both issue async writes and then immediately wait on them. Conceptually, that makes them sync writes and we should treat them as such so that the IO schedulers can handle them appropriately. This patch fixes a write starvation issue that Lin Ming reported, where xx is stuck for more than 2 minutes because of a large number of synchronous IO in the system: INFO: task kjournald:20558 blocked for more than 120 seconds. "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. kjournald D ffff810010820978 6712 20558 2 ffff81022ddb1d10 0000000000000046 ffff81022e7baa10 ffffffff803ba6f2 ffff81022ecd0000 ffff8101e6dc9160 ffff81022ecd0348 000000008048b6cb 0000000000000086 ffff81022c4e8d30 0000000000000000 ffffffff80247537 Call Trace: [<ffffffff803ba6f2>] kobject_get+0x12/0x17 [<ffffffff80247537>] getnstimeofday+0x2f/0x83 [<ffffffff8029c1ac>] sync_buffer+0x0/0x3f [<ffffffff8066d195>] io_schedule+0x5d/0x9f [<ffffffff8029c1e7>] sync_buffer+0x3b/0x3f [<ffffffff8066d3f0>] __wait_on_bit+0x40/0x6f [<ffffffff8029c1ac>] sync_buffer+0x0/0x3f [<ffffffff8066d48b>] out_of_line_wait_on_bit+0x6c/0x78 [<ffffffff80243909>] wake_bit_function+0x0/0x23 [<ffffffff8029e3ad>] sync_dirty_buffer+0x98/0xcb [<ffffffff8030056b>] journal_commit_transaction+0x97d/0xcb6 [<ffffffff8023a676>] lock_timer_base+0x26/0x4b [<ffffffff8030300a>] kjournald+0xc1/0x1fb [<ffffffff802438db>] autoremove_wake_function+0x0/0x2e [<ffffffff80302f49>] kjournald+0x0/0x1fb [<ffffffff802437bb>] kthread+0x47/0x74 [<ffffffff8022de51>] schedule_tail+0x28/0x5d [<ffffffff8020cac8>] child_rip+0xa/0x12 [<ffffffff80243774>] kthread+0x0/0x74 [<ffffffff8020cabe>] child_rip+0x0/0x12 Lin Ming confirms that this patch fixes the issue. I've run tests with it for the past week and no ill effects have been observed, so I'm proposing it for inclusion into 2.6.26. Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
2008-07-01 01:07:34 -06:00
ll_rw_block(SWRITE_SYNC, 1, &bh);
brelse(bh);
spin_lock(lock);
}
}
}
while (!list_empty(&tmp)) {
bh = BH_ENTRY(tmp.prev);
get_bh(bh);
buffer_head: fix private_list handling There are two possible races in handling of private_list in buffer cache. 1) When fsync_buffers_list() processes a private_list, it clears b_assoc_mapping and moves buffer to its private list. Now drop_buffers() comes, sees a buffer is on list so it calls __remove_assoc_queue() which complains about b_assoc_mapping being cleared (as it cannot propagate possible IO error). This race has been actually observed in the wild. 2) When fsync_buffers_list() processes a private_list, mark_buffer_dirty_inode() can be called on bh which is already on the private list of fsync_buffers_list(). As buffer is on some list (note that the check is performed without private_lock), it is not readded to the mapping's private_list and after fsync_buffers_list() finishes, we have a dirty buffer which should be on private_list but it isn't. This race has not been reported, probably because most (but not all) callers of mark_buffer_dirty_inode() hold i_mutex and thus are serialized with fsync(). Fix these issues by not clearing b_assoc_map when fsync_buffers_list() moves buffer to a dedicated list and by reinserting buffer in private_list when it is found dirty after we have submitted buffer for IO. We also change the tests whether a buffer is on a private list from !list_empty(&bh->b_assoc_buffers) to bh->b_assoc_map so that they are single word reads and hence lockless checks are safe. Signed-off-by: Jan Kara <jack@suse.cz> Cc: Nick Piggin <npiggin@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 05:21:59 -07:00
mapping = bh->b_assoc_map;
__remove_assoc_queue(bh);
/* Avoid race with mark_buffer_dirty_inode() which does
* a lockless check and we rely on seeing the dirty bit */
smp_mb();
if (buffer_dirty(bh)) {
list_add(&bh->b_assoc_buffers,
&mapping->private_list);
buffer_head: fix private_list handling There are two possible races in handling of private_list in buffer cache. 1) When fsync_buffers_list() processes a private_list, it clears b_assoc_mapping and moves buffer to its private list. Now drop_buffers() comes, sees a buffer is on list so it calls __remove_assoc_queue() which complains about b_assoc_mapping being cleared (as it cannot propagate possible IO error). This race has been actually observed in the wild. 2) When fsync_buffers_list() processes a private_list, mark_buffer_dirty_inode() can be called on bh which is already on the private list of fsync_buffers_list(). As buffer is on some list (note that the check is performed without private_lock), it is not readded to the mapping's private_list and after fsync_buffers_list() finishes, we have a dirty buffer which should be on private_list but it isn't. This race has not been reported, probably because most (but not all) callers of mark_buffer_dirty_inode() hold i_mutex and thus are serialized with fsync(). Fix these issues by not clearing b_assoc_map when fsync_buffers_list() moves buffer to a dedicated list and by reinserting buffer in private_list when it is found dirty after we have submitted buffer for IO. We also change the tests whether a buffer is on a private list from !list_empty(&bh->b_assoc_buffers) to bh->b_assoc_map so that they are single word reads and hence lockless checks are safe. Signed-off-by: Jan Kara <jack@suse.cz> Cc: Nick Piggin <npiggin@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 05:21:59 -07:00
bh->b_assoc_map = mapping;
}
spin_unlock(lock);
wait_on_buffer(bh);
if (!buffer_uptodate(bh))
err = -EIO;
brelse(bh);
spin_lock(lock);
}
spin_unlock(lock);
err2 = osync_buffers_list(lock, list);
if (err)
return err;
else
return err2;
}
/*
* Invalidate any and all dirty buffers on a given inode. We are
* probably unmounting the fs, but that doesn't mean we have already
* done a sync(). Just drop the buffers from the inode list.
*
* NOTE: we take the inode's blockdev's mapping's private_lock. Which
* assumes that all the buffers are against the blockdev. Not true
* for reiserfs.
*/
void invalidate_inode_buffers(struct inode *inode)
{
if (inode_has_buffers(inode)) {
struct address_space *mapping = &inode->i_data;
struct list_head *list = &mapping->private_list;
struct address_space *buffer_mapping = mapping->assoc_mapping;
spin_lock(&buffer_mapping->private_lock);
while (!list_empty(list))
__remove_assoc_queue(BH_ENTRY(list->next));
spin_unlock(&buffer_mapping->private_lock);
}
}
EXPORT_SYMBOL(invalidate_inode_buffers);
/*
* Remove any clean buffers from the inode's buffer list. This is called
* when we're trying to free the inode itself. Those buffers can pin it.
*
* Returns true if all buffers were removed.
*/
int remove_inode_buffers(struct inode *inode)
{
int ret = 1;
if (inode_has_buffers(inode)) {
struct address_space *mapping = &inode->i_data;
struct list_head *list = &mapping->private_list;
struct address_space *buffer_mapping = mapping->assoc_mapping;
spin_lock(&buffer_mapping->private_lock);
while (!list_empty(list)) {
struct buffer_head *bh = BH_ENTRY(list->next);
if (buffer_dirty(bh)) {
ret = 0;
break;
}
__remove_assoc_queue(bh);
}
spin_unlock(&buffer_mapping->private_lock);
}
return ret;
}
/*
* Create the appropriate buffers when given a page for data area and
* the size of each buffer.. Use the bh->b_this_page linked list to
* follow the buffers created. Return NULL if unable to create more
* buffers.
*
* The retry flag is used to differentiate async IO (paging, swapping)
* which may not fail from ordinary buffer allocations.
*/
struct buffer_head *alloc_page_buffers(struct page *page, unsigned long size,
int retry)
{
struct buffer_head *bh, *head;
long offset;
try_again:
head = NULL;
offset = PAGE_SIZE;
while ((offset -= size) >= 0) {
bh = alloc_buffer_head(GFP_NOFS);
if (!bh)
goto no_grow;
bh->b_bdev = NULL;
bh->b_this_page = head;
bh->b_blocknr = -1;
head = bh;
bh->b_state = 0;
atomic_set(&bh->b_count, 0);
bh->b_private = NULL;
bh->b_size = size;
/* Link the buffer to its page */
set_bh_page(bh, page, offset);
init_buffer(bh, NULL, NULL);
}
return head;
/*
* In case anything failed, we just free everything we got.
*/
no_grow:
if (head) {
do {
bh = head;
head = head->b_this_page;
free_buffer_head(bh);
} while (head);
}
/*
* Return failure for non-async IO requests. Async IO requests
* are not allowed to fail, so we have to wait until buffer heads
* become available. But we don't want tasks sleeping with
* partially complete buffers, so all were released above.
*/
if (!retry)
return NULL;
/* We're _really_ low on memory. Now we just
* wait for old buffer heads to become free due to
* finishing IO. Since this is an async request and
* the reserve list is empty, we're sure there are
* async buffer heads in use.
*/
free_more_memory();
goto try_again;
}
EXPORT_SYMBOL_GPL(alloc_page_buffers);
static inline void
link_dev_buffers(struct page *page, struct buffer_head *head)
{
struct buffer_head *bh, *tail;
bh = head;
do {
tail = bh;
bh = bh->b_this_page;
} while (bh);
tail->b_this_page = head;
attach_page_buffers(page, head);
}
/*
* Initialise the state of a blockdev page's buffers.
*/
static void
init_page_buffers(struct page *page, struct block_device *bdev,
sector_t block, int size)
{
struct buffer_head *head = page_buffers(page);
struct buffer_head *bh = head;
int uptodate = PageUptodate(page);
do {
if (!buffer_mapped(bh)) {
init_buffer(bh, NULL, NULL);
bh->b_bdev = bdev;
bh->b_blocknr = block;
if (uptodate)
set_buffer_uptodate(bh);
set_buffer_mapped(bh);
}
block++;
bh = bh->b_this_page;
} while (bh != head);
}
/*
* Create the page-cache page that contains the requested block.
*
* This is user purely for blockdev mappings.
*/
static struct page *
grow_dev_page(struct block_device *bdev, sector_t block,
pgoff_t index, int size)
{
struct inode *inode = bdev->bd_inode;
struct page *page;
struct buffer_head *bh;
page = find_or_create_page(inode->i_mapping, index,
Add __GFP_MOVABLE for callers to flag allocations from high memory that may be migrated It is often known at allocation time whether a page may be migrated or not. This patch adds a flag called __GFP_MOVABLE and a new mask called GFP_HIGH_MOVABLE. Allocations using the __GFP_MOVABLE can be either migrated using the page migration mechanism or reclaimed by syncing with backing storage and discarding. An API function very similar to alloc_zeroed_user_highpage() is added for __GFP_MOVABLE allocations called alloc_zeroed_user_highpage_movable(). The flags used by alloc_zeroed_user_highpage() are not changed because it would change the semantics of an existing API. After this patch is applied there are no in-kernel users of alloc_zeroed_user_highpage() so it probably should be marked deprecated if this patch is merged. Note that this patch includes a minor cleanup to the use of __GFP_ZERO in shmem.c to keep all flag modifications to inode->mapping in the shmem_dir_alloc() helper function. This clean-up suggestion is courtesy of Hugh Dickens. Additional credit goes to Christoph Lameter and Linus Torvalds for shaping the concept. Credit to Hugh Dickens for catching issues with shmem swap vector and ramfs allocations. [akpm@linux-foundation.org: build fix] [hugh@veritas.com: __GFP_ZERO cleanup] Signed-off-by: Mel Gorman <mel@csn.ul.ie> Cc: Andy Whitcroft <apw@shadowen.org> Cc: Christoph Lameter <clameter@sgi.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-17 05:03:05 -06:00
(mapping_gfp_mask(inode->i_mapping) & ~__GFP_FS)|__GFP_MOVABLE);
if (!page)
return NULL;
BUG_ON(!PageLocked(page));
if (page_has_buffers(page)) {
bh = page_buffers(page);
if (bh->b_size == size) {
init_page_buffers(page, bdev, block, size);
return page;
}
if (!try_to_free_buffers(page))
goto failed;
}
/*
* Allocate some buffers for this page
*/
bh = alloc_page_buffers(page, size, 0);
if (!bh)
goto failed;
/*
* Link the page to the buffers and initialise them. Take the
* lock to be atomic wrt __find_get_block(), which does not
* run under the page lock.
*/
spin_lock(&inode->i_mapping->private_lock);
link_dev_buffers(page, bh);
init_page_buffers(page, bdev, block, size);
spin_unlock(&inode->i_mapping->private_lock);
return page;
failed:
BUG();
unlock_page(page);
page_cache_release(page);
return NULL;
}
/*
* Create buffers for the specified block device block's page. If
* that page was dirty, the buffers are set dirty also.
*/
static int
grow_buffers(struct block_device *bdev, sector_t block, int size)
{
struct page *page;
pgoff_t index;
int sizebits;
sizebits = -1;
do {
sizebits++;
} while ((size << sizebits) < PAGE_SIZE);
index = block >> sizebits;
/*
* Check for a block which wants to lie outside our maximum possible
* pagecache index. (this comparison is done using sector_t types).
*/
if (unlikely(index != block >> sizebits)) {
char b[BDEVNAME_SIZE];
printk(KERN_ERR "%s: requested out-of-range block %llu for "
"device %s\n",
__func__, (unsigned long long)block,
bdevname(bdev, b));
return -EIO;
}
block = index << sizebits;
/* Create a page with the proper size buffers.. */
page = grow_dev_page(bdev, block, index, size);
if (!page)
return 0;
unlock_page(page);
page_cache_release(page);
return 1;
}
static struct buffer_head *
__getblk_slow(struct block_device *bdev, sector_t block, int size)
{
/* Size must be multiple of hard sectorsize */
if (unlikely(size & (bdev_hardsect_size(bdev)-1) ||
(size < 512 || size > PAGE_SIZE))) {
printk(KERN_ERR "getblk(): invalid block size %d requested\n",
size);
printk(KERN_ERR "hardsect size: %d\n",
bdev_hardsect_size(bdev));
dump_stack();
return NULL;
}
for (;;) {
struct buffer_head * bh;
int ret;
bh = __find_get_block(bdev, block, size);
if (bh)
return bh;
ret = grow_buffers(bdev, block, size);
if (ret < 0)
return NULL;
if (ret == 0)
free_more_memory();
}
}
/*
* The relationship between dirty buffers and dirty pages:
*
* Whenever a page has any dirty buffers, the page's dirty bit is set, and
* the page is tagged dirty in its radix tree.
*
* At all times, the dirtiness of the buffers represents the dirtiness of
* subsections of the page. If the page has buffers, the page dirty bit is
* merely a hint about the true dirty state.
*
* When a page is set dirty in its entirety, all its buffers are marked dirty
* (if the page has buffers).
*
* When a buffer is marked dirty, its page is dirtied, but the page's other
* buffers are not.
*
* Also. When blockdev buffers are explicitly read with bread(), they
* individually become uptodate. But their backing page remains not
* uptodate - even if all of its buffers are uptodate. A subsequent
* block_read_full_page() against that page will discover all the uptodate
* buffers, will set the page uptodate and will perform no I/O.
*/
/**
* mark_buffer_dirty - mark a buffer_head as needing writeout
* @bh: the buffer_head to mark dirty
*
* mark_buffer_dirty() will set the dirty bit against the buffer, then set its
* backing page dirty, then tag the page as dirty in its address_space's radix
* tree and then attach the address_space's inode to its superblock's dirty
* inode list.
*
* mark_buffer_dirty() is atomic. It takes bh->b_page->mapping->private_lock,
* mapping->tree_lock and the global inode_lock.
*/
void mark_buffer_dirty(struct buffer_head *bh)
{
WARN_ON_ONCE(!buffer_uptodate(bh));
/*
* Very *carefully* optimize the it-is-already-dirty case.
*
* Don't let the final "is it dirty" escape to before we
* perhaps modified the buffer.
*/
if (buffer_dirty(bh)) {
smp_mb();
if (buffer_dirty(bh))
return;
}
if (!test_set_buffer_dirty(bh))
__set_page_dirty(bh->b_page, page_mapping(bh->b_page), 0);
}
/*
* Decrement a buffer_head's reference count. If all buffers against a page
* have zero reference count, are clean and unlocked, and if the page is clean
* and unlocked then try_to_free_buffers() may strip the buffers from the page
* in preparation for freeing it (sometimes, rarely, buffers are removed from
* a page but it ends up not being freed, and buffers may later be reattached).
*/
void __brelse(struct buffer_head * buf)
{
if (atomic_read(&buf->b_count)) {
put_bh(buf);
return;
}
WARN(1, KERN_ERR "VFS: brelse: Trying to free free buffer\n");
}
/*
* bforget() is like brelse(), except it discards any
* potentially dirty data.
*/
void __bforget(struct buffer_head *bh)
{
clear_buffer_dirty(bh);
buffer_head: fix private_list handling There are two possible races in handling of private_list in buffer cache. 1) When fsync_buffers_list() processes a private_list, it clears b_assoc_mapping and moves buffer to its private list. Now drop_buffers() comes, sees a buffer is on list so it calls __remove_assoc_queue() which complains about b_assoc_mapping being cleared (as it cannot propagate possible IO error). This race has been actually observed in the wild. 2) When fsync_buffers_list() processes a private_list, mark_buffer_dirty_inode() can be called on bh which is already on the private list of fsync_buffers_list(). As buffer is on some list (note that the check is performed without private_lock), it is not readded to the mapping's private_list and after fsync_buffers_list() finishes, we have a dirty buffer which should be on private_list but it isn't. This race has not been reported, probably because most (but not all) callers of mark_buffer_dirty_inode() hold i_mutex and thus are serialized with fsync(). Fix these issues by not clearing b_assoc_map when fsync_buffers_list() moves buffer to a dedicated list and by reinserting buffer in private_list when it is found dirty after we have submitted buffer for IO. We also change the tests whether a buffer is on a private list from !list_empty(&bh->b_assoc_buffers) to bh->b_assoc_map so that they are single word reads and hence lockless checks are safe. Signed-off-by: Jan Kara <jack@suse.cz> Cc: Nick Piggin <npiggin@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 05:21:59 -07:00
if (bh->b_assoc_map) {
struct address_space *buffer_mapping = bh->b_page->mapping;
spin_lock(&buffer_mapping->private_lock);
list_del_init(&bh->b_assoc_buffers);
bh->b_assoc_map = NULL;
spin_unlock(&buffer_mapping->private_lock);
}
__brelse(bh);
}
static struct buffer_head *__bread_slow(struct buffer_head *bh)
{
lock_buffer(bh);
if (buffer_uptodate(bh)) {
unlock_buffer(bh);
return bh;
} else {
get_bh(bh);
bh->b_end_io = end_buffer_read_sync;
submit_bh(READ, bh);
wait_on_buffer(bh);
if (buffer_uptodate(bh))
return bh;
}
brelse(bh);
return NULL;
}
/*
* Per-cpu buffer LRU implementation. To reduce the cost of __find_get_block().
* The bhs[] array is sorted - newest buffer is at bhs[0]. Buffers have their
* refcount elevated by one when they're in an LRU. A buffer can only appear
* once in a particular CPU's LRU. A single buffer can be present in multiple
* CPU's LRUs at the same time.
*
* This is a transparent caching front-end to sb_bread(), sb_getblk() and
* sb_find_get_block().
*
* The LRUs themselves only need locking against invalidate_bh_lrus. We use
* a local interrupt disable for that.
*/
#define BH_LRU_SIZE 8
struct bh_lru {
struct buffer_head *bhs[BH_LRU_SIZE];
};
static DEFINE_PER_CPU(struct bh_lru, bh_lrus) = {{ NULL }};
#ifdef CONFIG_SMP
#define bh_lru_lock() local_irq_disable()
#define bh_lru_unlock() local_irq_enable()
#else
#define bh_lru_lock() preempt_disable()
#define bh_lru_unlock() preempt_enable()
#endif
static inline void check_irqs_on(void)
{
#ifdef irqs_disabled
BUG_ON(irqs_disabled());
#endif
}
/*
* The LRU management algorithm is dopey-but-simple. Sorry.
*/
static void bh_lru_install(struct buffer_head *bh)
{
struct buffer_head *evictee = NULL;
struct bh_lru *lru;
check_irqs_on();
bh_lru_lock();
lru = &__get_cpu_var(bh_lrus);
if (lru->bhs[0] != bh) {
struct buffer_head *bhs[BH_LRU_SIZE];
int in;
int out = 0;
get_bh(bh);
bhs[out++] = bh;
for (in = 0; in < BH_LRU_SIZE; in++) {
struct buffer_head *bh2 = lru->bhs[in];
if (bh2 == bh) {
__brelse(bh2);
} else {
if (out >= BH_LRU_SIZE) {
BUG_ON(evictee != NULL);
evictee = bh2;
} else {
bhs[out++] = bh2;
}
}
}
while (out < BH_LRU_SIZE)
bhs[out++] = NULL;
memcpy(lru->bhs, bhs, sizeof(bhs));
}
bh_lru_unlock();
if (evictee)
__brelse(evictee);
}
/*
* Look up the bh in this cpu's LRU. If it's there, move it to the head.
*/
static struct buffer_head *
lookup_bh_lru(struct block_device *bdev, sector_t block, unsigned size)
{
struct buffer_head *ret = NULL;
struct bh_lru *lru;
unsigned int i;
check_irqs_on();
bh_lru_lock();
lru = &__get_cpu_var(bh_lrus);
for (i = 0; i < BH_LRU_SIZE; i++) {
struct buffer_head *bh = lru->bhs[i];
if (bh && bh->b_bdev == bdev &&
bh->b_blocknr == block && bh->b_size == size) {
if (i) {
while (i) {
lru->bhs[i] = lru->bhs[i - 1];
i--;
}
lru->bhs[0] = bh;
}
get_bh(bh);
ret = bh;
break;
}
}
bh_lru_unlock();
return ret;
}
/*
* Perform a pagecache lookup for the matching buffer. If it's there, refresh
* it in the LRU and mark it as accessed. If it is not present then return
* NULL
*/
struct buffer_head *
__find_get_block(struct block_device *bdev, sector_t block, unsigned size)
{
struct buffer_head *bh = lookup_bh_lru(bdev, block, size);
if (bh == NULL) {
bh = __find_get_block_slow(bdev, block);
if (bh)
bh_lru_install(bh);
}
if (bh)
touch_buffer(bh);
return bh;
}
EXPORT_SYMBOL(__find_get_block);
/*
* __getblk will locate (and, if necessary, create) the buffer_head
* which corresponds to the passed block_device, block and size. The
* returned buffer has its reference count incremented.
*
* __getblk() cannot fail - it just keeps trying. If you pass it an
* illegal block number, __getblk() will happily return a buffer_head
* which represents the non-existent block. Very weird.
*
* __getblk() will lock up the machine if grow_dev_page's try_to_free_buffers()
* attempt is failing. FIXME, perhaps?
*/
struct buffer_head *
__getblk(struct block_device *bdev, sector_t block, unsigned size)
{
struct buffer_head *bh = __find_get_block(bdev, block, size);
might_sleep();
if (bh == NULL)
bh = __getblk_slow(bdev, block, size);
return bh;
}
EXPORT_SYMBOL(__getblk);
/*
* Do async read-ahead on a buffer..
*/
void __breadahead(struct block_device *bdev, sector_t block, unsigned size)
{
struct buffer_head *bh = __getblk(bdev, block, size);
if (likely(bh)) {
ll_rw_block(READA, 1, &bh);
brelse(bh);
}
}
EXPORT_SYMBOL(__breadahead);
/**
* __bread() - reads a specified block and returns the bh
* @bdev: the block_device to read from
* @block: number of block
* @size: size (in bytes) to read
*
* Reads a specified block, and returns buffer head that contains it.
* It returns NULL if the block was unreadable.
*/
struct buffer_head *
__bread(struct block_device *bdev, sector_t block, unsigned size)
{
struct buffer_head *bh = __getblk(bdev, block, size);
if (likely(bh) && !buffer_uptodate(bh))
bh = __bread_slow(bh);
return bh;
}
EXPORT_SYMBOL(__bread);
/*
* invalidate_bh_lrus() is called rarely - but not only at unmount.
* This doesn't race because it runs in each cpu either in irq
* or with preempt disabled.
*/
static void invalidate_bh_lru(void *arg)
{
struct bh_lru *b = &get_cpu_var(bh_lrus);
int i;
for (i = 0; i < BH_LRU_SIZE; i++) {
brelse(b->bhs[i]);
b->bhs[i] = NULL;
}
put_cpu_var(bh_lrus);
}
void invalidate_bh_lrus(void)
{
on_each_cpu(invalidate_bh_lru, NULL, 1);
}
rewrite rd This is a rewrite of the ramdisk block device driver. The old one is really difficult because it effectively implements a block device which serves data out of its own buffer cache. It relies on the dirty bit being set, to pin its backing store in cache, however there are non trivial paths which can clear the dirty bit (eg. try_to_free_buffers()), which had recently lead to data corruption. And in general it is completely wrong for a block device driver to do this. The new one is more like a regular block device driver. It has no idea about vm/vfs stuff. It's backing store is similar to the buffer cache (a simple radix-tree of pages), but it doesn't know anything about page cache (the pages in the radix tree are not pagecache pages). There is one slight downside -- direct block device access and filesystem metadata access goes through an extra copy and gets stored in RAM twice. However, this downside is only slight, because the real buffercache of the device is now reclaimable (because we're not playing crazy games with it), so under memory intensive situations, footprint should effectively be the same -- maybe even a slight advantage to the new driver because it can also reclaim buffer heads. The fact that it now goes through all the regular vm/fs paths makes it much more useful for testing, too. text data bss dec hex filename 2837 849 384 4070 fe6 drivers/block/rd.o 3528 371 12 3911 f47 drivers/block/brd.o Text is larger, but data and bss are smaller, making total size smaller. A few other nice things about it: - Similar structure and layout to the new loop device handlinag. - Dynamic ramdisk creation. - Runtime flexible buffer head size (because it is no longer part of the ramdisk code). - Boot / load time flexible ramdisk size, which could easily be extended to a per-ramdisk runtime changeable size (eg. with an ioctl). - Can use highmem for the backing store. [akpm@linux-foundation.org: fix build] [byron.bbradley@gmail.com: make rd_size non-static] Signed-off-by: Nick Piggin <npiggin@suse.de> Signed-off-by: Byron Bradley <byron.bbradley@gmail.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 05:19:49 -07:00
EXPORT_SYMBOL_GPL(invalidate_bh_lrus);
void set_bh_page(struct buffer_head *bh,
struct page *page, unsigned long offset)
{
bh->b_page = page;
BUG_ON(offset >= PAGE_SIZE);
if (PageHighMem(page))
/*
* This catches illegal uses and preserves the offset:
*/
bh->b_data = (char *)(0 + offset);
else
bh->b_data = page_address(page) + offset;
}
EXPORT_SYMBOL(set_bh_page);
/*
* Called when truncating a buffer on a page completely.
*/
static void discard_buffer(struct buffer_head * bh)
{
lock_buffer(bh);
clear_buffer_dirty(bh);
bh->b_bdev = NULL;
clear_buffer_mapped(bh);
clear_buffer_req(bh);
clear_buffer_new(bh);
clear_buffer_delay(bh);
clear_buffer_unwritten(bh);
unlock_buffer(bh);
}
/**
* block_invalidatepage - invalidate part of all of a buffer-backed page
*
* @page: the page which is affected
* @offset: the index of the truncation point
*
* block_invalidatepage() is called when all or part of the page has become
* invalidatedby a truncate operation.
*
* block_invalidatepage() does not have to release all buffers, but it must
* ensure that no dirty buffer is left outside @offset and that no I/O
* is underway against any of the blocks which are outside the truncation
* point. Because the caller is about to free (and possibly reuse) those
* blocks on-disk.
*/
void block_invalidatepage(struct page *page, unsigned long offset)
{
struct buffer_head *head, *bh, *next;
unsigned int curr_off = 0;
BUG_ON(!PageLocked(page));
if (!page_has_buffers(page))
goto out;
head = page_buffers(page);
bh = head;
do {
unsigned int next_off = curr_off + bh->b_size;
next = bh->b_this_page;
/*
* is this block fully invalidated?
*/
if (offset <= curr_off)
discard_buffer(bh);
curr_off = next_off;
bh = next;
} while (bh != head);
/*
* We release buffers only if the entire page is being invalidated.
* The get_block cached value has been unconditionally invalidated,
* so real IO is not possible anymore.
*/
if (offset == 0)
try_to_release_page(page, 0);
out:
return;
}
EXPORT_SYMBOL(block_invalidatepage);
/*
* We attach and possibly dirty the buffers atomically wrt
* __set_page_dirty_buffers() via private_lock. try_to_free_buffers
* is already excluded via the page lock.
*/
void create_empty_buffers(struct page *page,
unsigned long blocksize, unsigned long b_state)
{
struct buffer_head *bh, *head, *tail;
head = alloc_page_buffers(page, blocksize, 1);
bh = head;
do {
bh->b_state |= b_state;
tail = bh;
bh = bh->b_this_page;
} while (bh);
tail->b_this_page = head;
spin_lock(&page->mapping->private_lock);
if (PageUptodate(page) || PageDirty(page)) {
bh = head;
do {
if (PageDirty(page))
set_buffer_dirty(bh);
if (PageUptodate(page))
set_buffer_uptodate(bh);
bh = bh->b_this_page;
} while (bh != head);
}
attach_page_buffers(page, head);
spin_unlock(&page->mapping->private_lock);
}
EXPORT_SYMBOL(create_empty_buffers);
/*
* We are taking a block for data and we don't want any output from any
* buffer-cache aliases starting from return from that function and
* until the moment when something will explicitly mark the buffer
* dirty (hopefully that will not happen until we will free that block ;-)
* We don't even need to mark it not-uptodate - nobody can expect
* anything from a newly allocated buffer anyway. We used to used
* unmap_buffer() for such invalidation, but that was wrong. We definitely
* don't want to mark the alias unmapped, for example - it would confuse
* anyone who might pick it with bread() afterwards...
*
* Also.. Note that bforget() doesn't lock the buffer. So there can
* be writeout I/O going on against recently-freed buffers. We don't
* wait on that I/O in bforget() - it's more efficient to wait on the I/O
* only if we really need to. That happens here.
*/
void unmap_underlying_metadata(struct block_device *bdev, sector_t block)
{
struct buffer_head *old_bh;
might_sleep();
old_bh = __find_get_block_slow(bdev, block);
if (old_bh) {
clear_buffer_dirty(old_bh);
wait_on_buffer(old_bh);
clear_buffer_req(old_bh);
__brelse(old_bh);
}
}
EXPORT_SYMBOL(unmap_underlying_metadata);
/*
* NOTE! All mapped/uptodate combinations are valid:
*
* Mapped Uptodate Meaning
*
* No No "unknown" - must do get_block()
* No Yes "hole" - zero-filled
* Yes No "allocated" - allocated on disk, not read in
* Yes Yes "valid" - allocated and up-to-date in memory.
*
* "Dirty" is valid only with the last case (mapped+uptodate).
*/
/*
* While block_write_full_page is writing back the dirty buffers under
* the page lock, whoever dirtied the buffers may decide to clean them
* again at any time. We handle that by only looking at the buffer
* state inside lock_buffer().
*
* If block_write_full_page() is called for regular writeback
* (wbc->sync_mode == WB_SYNC_NONE) then it will redirty a page which has a
* locked buffer. This only can happen if someone has written the buffer
* directly, with submit_bh(). At the address_space level PageWriteback
* prevents this contention from occurring.
*/
static int __block_write_full_page(struct inode *inode, struct page *page,
get_block_t *get_block, struct writeback_control *wbc)
{
int err;
sector_t block;
sector_t last_block;
struct buffer_head *bh, *head;
const unsigned blocksize = 1 << inode->i_blkbits;
int nr_underway = 0;
BUG_ON(!PageLocked(page));
last_block = (i_size_read(inode) - 1) >> inode->i_blkbits;
if (!page_has_buffers(page)) {
create_empty_buffers(page, blocksize,
(1 << BH_Dirty)|(1 << BH_Uptodate));
}
/*
* Be very careful. We have no exclusion from __set_page_dirty_buffers
* here, and the (potentially unmapped) buffers may become dirty at
* any time. If a buffer becomes dirty here after we've inspected it
* then we just miss that fact, and the page stays dirty.
*
* Buffers outside i_size may be dirtied by __set_page_dirty_buffers;
* handle that here by just cleaning them.
*/
[PATCH] fix possible PAGE_CACHE_SHIFT overflows We've had two instances recently of overflows when doing 64_bit_value = (32_bit_value << PAGE_CACHE_SHIFT) I did a tree-wide grep of `<<.*PAGE_CACHE_SHIFT' and this is the result. - afs_rxfs_fetch_descriptor.offset is of type off_t, which seems broken. - jfs and jffs are limited to 4GB anyway. - reiserfs map_block_for_writepage() takes an unsigned long for the block - it should take sector_t. (It'll fail for huge filesystems with blocksize<PAGE_CACHE_SIZE) - cramfs_read() needs to use sector_t (I think cramsfs is busted on large filesystems anyway) - affs is limited in file size anyway. - I generally didn't fix 32-bit overflows in directory operations. - arm's __flush_dcache_page() is peculiar. What if the page lies beyond 4G? - gss_wrap_req_priv() needs checking (snd_buf->page_base) Cc: Oleg Drokin <green@linuxhacker.ru> Cc: David Howells <dhowells@redhat.com> Cc: David Woodhouse <dwmw2@infradead.org> Cc: <reiserfs-dev@namesys.com> Cc: Christoph Hellwig <hch@lst.de> Cc: Anton Altaparmakov <aia21@cantab.net> Cc: Jeff Dike <jdike@addtoit.com> Cc: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it> Cc: Roman Zippel <zippel@linux-m68k.org> Cc: <linux-fsdevel@vger.kernel.org> Cc: Miklos Szeredi <miklos@szeredi.hu> Cc: Russell King <rmk@arm.linux.org.uk> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Cc: Neil Brown <neilb@cse.unsw.edu.au> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-01-08 02:03:05 -07:00
block = (sector_t)page->index << (PAGE_CACHE_SHIFT - inode->i_blkbits);
head = page_buffers(page);
bh = head;
/*
* Get all the dirty buffers mapped to disk addresses and
* handle any aliases from the underlying blockdev's mapping.
*/
do {
if (block > last_block) {
/*
* mapped buffers outside i_size will occur, because
* this page can be outside i_size when there is a
* truncate in progress.
*/
/*
* The buffer was zeroed by block_write_full_page()
*/
clear_buffer_dirty(bh);
set_buffer_uptodate(bh);
} else if ((!buffer_mapped(bh) || buffer_delay(bh)) &&
buffer_dirty(bh)) {
WARN_ON(bh->b_size != blocksize);
err = get_block(inode, block, bh, 1);
if (err)
goto recover;
clear_buffer_delay(bh);
if (buffer_new(bh)) {
/* blockdev mappings never come here */
clear_buffer_new(bh);
unmap_underlying_metadata(bh->b_bdev,
bh->b_blocknr);
}
}
bh = bh->b_this_page;
block++;
} while (bh != head);
do {
if (!buffer_mapped(bh))
continue;
/*
* If it's a fully non-blocking write attempt and we cannot
* lock the buffer then redirty the page. Note that this can
* potentially cause a busy-wait loop from pdflush and kswapd
* activity, but those code paths have their own higher-level
* throttling.
*/
if (wbc->sync_mode != WB_SYNC_NONE || !wbc->nonblocking) {
lock_buffer(bh);
} else if (!trylock_buffer(bh)) {
redirty_page_for_writepage(wbc, page);
continue;
}
if (test_clear_buffer_dirty(bh)) {
mark_buffer_async_write(bh);
} else {
unlock_buffer(bh);
}
} while ((bh = bh->b_this_page) != head);
/*
* The page and its buffers are protected by PageWriteback(), so we can
* drop the bh refcounts early.
*/
BUG_ON(PageWriteback(page));
set_page_writeback(page);
do {
struct buffer_head *next = bh->b_this_page;
if (buffer_async_write(bh)) {
submit_bh(WRITE, bh);
nr_underway++;
}
bh = next;
} while (bh != head);
unlock_page(page);
err = 0;
done:
if (nr_underway == 0) {
/*
* The page was marked dirty, but the buffers were
* clean. Someone wrote them back by hand with
* ll_rw_block/submit_bh. A rare case.
*/
end_page_writeback(page);
/*
* The page and buffer_heads can be released at any time from
* here on.
*/
}
return err;
recover:
/*
* ENOSPC, or some other error. We may already have added some
* blocks to the file, so we need to write these out to avoid
* exposing stale data.
* The page is currently locked and not marked for writeback
*/
bh = head;
/* Recovery: lock and submit the mapped buffers */
do {
if (buffer_mapped(bh) && buffer_dirty(bh) &&
!buffer_delay(bh)) {
lock_buffer(bh);
mark_buffer_async_write(bh);
} else {
/*
* The buffer may have been set dirty during
* attachment to a dirty page.
*/
clear_buffer_dirty(bh);
}
} while ((bh = bh->b_this_page) != head);
SetPageError(page);
BUG_ON(PageWriteback(page));
mapping_set_error(page->mapping, err);
set_page_writeback(page);
do {
struct buffer_head *next = bh->b_this_page;
if (buffer_async_write(bh)) {
clear_buffer_dirty(bh);
submit_bh(WRITE, bh);
nr_underway++;
}
bh = next;
} while (bh != head);
unlock_page(page);
goto done;
}
/*
* If a page has any new buffers, zero them out here, and mark them uptodate
* and dirty so they'll be written out (in order to prevent uninitialised
* block data from leaking). And clear the new bit.
*/
void page_zero_new_buffers(struct page *page, unsigned from, unsigned to)
{
unsigned int block_start, block_end;
struct buffer_head *head, *bh;
BUG_ON(!PageLocked(page));
if (!page_has_buffers(page))
return;
bh = head = page_buffers(page);
block_start = 0;
do {
block_end = block_start + bh->b_size;
if (buffer_new(bh)) {
if (block_end > from && block_start < to) {
if (!PageUptodate(page)) {
unsigned start, size;
start = max(from, block_start);
size = min(to, block_end) - start;
Pagecache zeroing: zero_user_segment, zero_user_segments and zero_user Simplify page cache zeroing of segments of pages through 3 functions zero_user_segments(page, start1, end1, start2, end2) Zeros two segments of the page. It takes the position where to start and end the zeroing which avoids length calculations and makes code clearer. zero_user_segment(page, start, end) Same for a single segment. zero_user(page, start, length) Length variant for the case where we know the length. We remove the zero_user_page macro. Issues: 1. Its a macro. Inline functions are preferable. 2. The KM_USER0 macro is only defined for HIGHMEM. Having to treat this special case everywhere makes the code needlessly complex. The parameter for zeroing is always KM_USER0 except in one single case that we open code. Avoiding KM_USER0 makes a lot of code not having to be dealing with the special casing for HIGHMEM anymore. Dealing with kmap is only necessary for HIGHMEM configurations. In those configurations we use KM_USER0 like we do for a series of other functions defined in highmem.h. Since KM_USER0 is depends on HIGHMEM the existing zero_user_page function could not be a macro. zero_user_* functions introduced here can be be inline because that constant is not used when these functions are called. Also extract the flushing of the caches to be outside of the kmap. [akpm@linux-foundation.org: fix nfs and ntfs build] [akpm@linux-foundation.org: fix ntfs build some more] Signed-off-by: Christoph Lameter <clameter@sgi.com> Cc: Steven French <sfrench@us.ibm.com> Cc: Michael Halcrow <mhalcrow@us.ibm.com> Cc: <linux-ext4@vger.kernel.org> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Cc: "J. Bruce Fields" <bfields@fieldses.org> Cc: Anton Altaparmakov <aia21@cantab.net> Cc: Mark Fasheh <mark.fasheh@oracle.com> Cc: David Chinner <dgc@sgi.com> Cc: Michael Halcrow <mhalcrow@us.ibm.com> Cc: Steven French <sfrench@us.ibm.com> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-04 23:28:29 -07:00
zero_user(page, start, size);
set_buffer_uptodate(bh);
}
clear_buffer_new(bh);
mark_buffer_dirty(bh);
}
}
block_start = block_end;
bh = bh->b_this_page;
} while (bh != head);
}
EXPORT_SYMBOL(page_zero_new_buffers);
static int __block_prepare_write(struct inode *inode, struct page *page,
unsigned from, unsigned to, get_block_t *get_block)
{
unsigned block_start, block_end;
sector_t block;
int err = 0;
unsigned blocksize, bbits;
struct buffer_head *bh, *head, *wait[2], **wait_bh=wait;
BUG_ON(!PageLocked(page));
BUG_ON(from > PAGE_CACHE_SIZE);
BUG_ON(to > PAGE_CACHE_SIZE);
BUG_ON(from > to);
blocksize = 1 << inode->i_blkbits;
if (!page_has_buffers(page))
create_empty_buffers(page, blocksize, 0);
head = page_buffers(page);
bbits = inode->i_blkbits;
block = (sector_t)page->index << (PAGE_CACHE_SHIFT - bbits);
for(bh = head, block_start = 0; bh != head || !block_start;
block++, block_start=block_end, bh = bh->b_this_page) {
block_end = block_start + blocksize;
if (block_end <= from || block_start >= to) {
if (PageUptodate(page)) {
if (!buffer_uptodate(bh))
set_buffer_uptodate(bh);
}
continue;
}
if (buffer_new(bh))
clear_buffer_new(bh);
if (!buffer_mapped(bh)) {
WARN_ON(bh->b_size != blocksize);
err = get_block(inode, block, bh, 1);
if (err)
break;
if (buffer_new(bh)) {
unmap_underlying_metadata(bh->b_bdev,
bh->b_blocknr);
if (PageUptodate(page)) {
clear_buffer_new(bh);
set_buffer_uptodate(bh);
mark_buffer_dirty(bh);
continue;
}
Pagecache zeroing: zero_user_segment, zero_user_segments and zero_user Simplify page cache zeroing of segments of pages through 3 functions zero_user_segments(page, start1, end1, start2, end2) Zeros two segments of the page. It takes the position where to start and end the zeroing which avoids length calculations and makes code clearer. zero_user_segment(page, start, end) Same for a single segment. zero_user(page, start, length) Length variant for the case where we know the length. We remove the zero_user_page macro. Issues: 1. Its a macro. Inline functions are preferable. 2. The KM_USER0 macro is only defined for HIGHMEM. Having to treat this special case everywhere makes the code needlessly complex. The parameter for zeroing is always KM_USER0 except in one single case that we open code. Avoiding KM_USER0 makes a lot of code not having to be dealing with the special casing for HIGHMEM anymore. Dealing with kmap is only necessary for HIGHMEM configurations. In those configurations we use KM_USER0 like we do for a series of other functions defined in highmem.h. Since KM_USER0 is depends on HIGHMEM the existing zero_user_page function could not be a macro. zero_user_* functions introduced here can be be inline because that constant is not used when these functions are called. Also extract the flushing of the caches to be outside of the kmap. [akpm@linux-foundation.org: fix nfs and ntfs build] [akpm@linux-foundation.org: fix ntfs build some more] Signed-off-by: Christoph Lameter <clameter@sgi.com> Cc: Steven French <sfrench@us.ibm.com> Cc: Michael Halcrow <mhalcrow@us.ibm.com> Cc: <linux-ext4@vger.kernel.org> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Cc: "J. Bruce Fields" <bfields@fieldses.org> Cc: Anton Altaparmakov <aia21@cantab.net> Cc: Mark Fasheh <mark.fasheh@oracle.com> Cc: David Chinner <dgc@sgi.com> Cc: Michael Halcrow <mhalcrow@us.ibm.com> Cc: Steven French <sfrench@us.ibm.com> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-04 23:28:29 -07:00
if (block_end > to || block_start < from)
zero_user_segments(page,
to, block_end,
block_start, from);
continue;
}
}
if (PageUptodate(page)) {
if (!buffer_uptodate(bh))
set_buffer_uptodate(bh);
continue;
}
if (!buffer_uptodate(bh) && !buffer_delay(bh) &&
!buffer_unwritten(bh) &&
(block_start < from || block_end > to)) {
ll_rw_block(READ, 1, &bh);
*wait_bh++=bh;
}
}
/*
* If we issued read requests - let them complete.
*/
while(wait_bh > wait) {
wait_on_buffer(*--wait_bh);
if (!buffer_uptodate(*wait_bh))
err = -EIO;
}
if (unlikely(err))
page_zero_new_buffers(page, from, to);
return err;
}
static int __block_commit_write(struct inode *inode, struct page *page,
unsigned from, unsigned to)
{
unsigned block_start, block_end;
int partial = 0;
unsigned blocksize;
struct buffer_head *bh, *head;
blocksize = 1 << inode->i_blkbits;
for(bh = head = page_buffers(page), block_start = 0;
bh != head || !block_start;
block_start=block_end, bh = bh->b_this_page) {
block_end = block_start + blocksize;
if (block_end <= from || block_start >= to) {
if (!buffer_uptodate(bh))
partial = 1;
} else {
set_buffer_uptodate(bh);
mark_buffer_dirty(bh);
}
clear_buffer_new(bh);
}
/*
* If this is a partial write which happened to make all buffers
* uptodate then we can optimize away a bogus readpage() for
* the next read(). Here we 'discover' whether the page went
* uptodate as a result of this (potentially partial) write.
*/
if (!partial)
SetPageUptodate(page);
return 0;
}
/*
* block_write_begin takes care of the basic task of block allocation and
* bringing partial write blocks uptodate first.
*
* If *pagep is not NULL, then block_write_begin uses the locked page
* at *pagep rather than allocating its own. In this case, the page will
* not be unlocked or deallocated on failure.
*/
int block_write_begin(struct file *file, struct address_space *mapping,
loff_t pos, unsigned len, unsigned flags,
struct page **pagep, void **fsdata,
get_block_t *get_block)
{
struct inode *inode = mapping->host;
int status = 0;
struct page *page;
pgoff_t index;
unsigned start, end;
int ownpage = 0;
index = pos >> PAGE_CACHE_SHIFT;
start = pos & (PAGE_CACHE_SIZE - 1);
end = start + len;
page = *pagep;
if (page == NULL) {
ownpage = 1;
fs: symlink write_begin allocation context fix With the write_begin/write_end aops, page_symlink was broken because it could no longer pass a GFP_NOFS type mask into the point where the allocations happened. They are done in write_begin, which would always assume that the filesystem can be entered from reclaim. This bug could cause filesystem deadlocks. The funny thing with having a gfp_t mask there is that it doesn't really allow the caller to arbitrarily tinker with the context in which it can be called. It couldn't ever be GFP_ATOMIC, for example, because it needs to take the page lock. The only thing any callers care about is __GFP_FS anyway, so turn that into a single flag. Add a new flag for write_begin, AOP_FLAG_NOFS. Filesystems can now act on this flag in their write_begin function. Change __grab_cache_page to accept a nofs argument as well, to honour that flag (while we're there, change the name to grab_cache_page_write_begin which is more instructive and does away with random leading underscores). This is really a more flexible way to go in the end anyway -- if a filesystem happens to want any extra allocations aside from the pagecache ones in ints write_begin function, it may now use GFP_KERNEL (rather than GFP_NOFS) for common case allocations (eg. ocfs2_alloc_write_ctxt, for a random example). [kosaki.motohiro@jp.fujitsu.com: fix ubifs] [kosaki.motohiro@jp.fujitsu.com: fix fuse] Signed-off-by: Nick Piggin <npiggin@suse.de> Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Cc: <stable@kernel.org> [2.6.28.x] Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> [ Cleaned up the calling convention: just pass in the AOP flags untouched to the grab_cache_page_write_begin() function. That just simplifies everybody, and may even allow future expansion of the logic. - Linus ] Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-04 13:00:53 -07:00
page = grab_cache_page_write_begin(mapping, index, flags);
if (!page) {
status = -ENOMEM;
goto out;
}
*pagep = page;
} else
BUG_ON(!PageLocked(page));
status = __block_prepare_write(inode, page, start, end, get_block);
if (unlikely(status)) {
ClearPageUptodate(page);
if (ownpage) {
unlock_page(page);
page_cache_release(page);
*pagep = NULL;
/*
* prepare_write() may have instantiated a few blocks
* outside i_size. Trim these off again. Don't need
* i_size_read because we hold i_mutex.
*/
if (pos + len > inode->i_size)
vmtruncate(inode, inode->i_size);
}
}
out:
return status;
}
EXPORT_SYMBOL(block_write_begin);
int block_write_end(struct file *file, struct address_space *mapping,
loff_t pos, unsigned len, unsigned copied,
struct page *page, void *fsdata)
{
struct inode *inode = mapping->host;
unsigned start;
start = pos & (PAGE_CACHE_SIZE - 1);
if (unlikely(copied < len)) {
/*
* The buffers that were written will now be uptodate, so we
* don't have to worry about a readpage reading them and
* overwriting a partial write. However if we have encountered
* a short write and only partially written into a buffer, it
* will not be marked uptodate, so a readpage might come in and
* destroy our partial write.
*
* Do the simplest thing, and just treat any short write to a
* non uptodate page as a zero-length write, and force the
* caller to redo the whole thing.
*/
if (!PageUptodate(page))
copied = 0;
page_zero_new_buffers(page, start+copied, start+len);
}
flush_dcache_page(page);
/* This could be a short (even 0-length) commit */
__block_commit_write(inode, page, start, start+copied);
return copied;
}
EXPORT_SYMBOL(block_write_end);
int generic_write_end(struct file *file, struct address_space *mapping,
loff_t pos, unsigned len, unsigned copied,
struct page *page, void *fsdata)
{
struct inode *inode = mapping->host;
int i_size_changed = 0;
copied = block_write_end(file, mapping, pos, len, copied, page, fsdata);
/*
* No need to use i_size_read() here, the i_size
* cannot change under us because we hold i_mutex.
*
* But it's important to update i_size while still holding page lock:
* page writeout could otherwise come in and zero beyond i_size.
*/
if (pos+copied > inode->i_size) {
i_size_write(inode, pos+copied);
i_size_changed = 1;
}
unlock_page(page);
page_cache_release(page);
/*
* Don't mark the inode dirty under page lock. First, it unnecessarily
* makes the holding time of page lock longer. Second, it forces lock
* ordering of page lock and transaction start for journaling
* filesystems.
*/
if (i_size_changed)
mark_inode_dirty(inode);
return copied;
}
EXPORT_SYMBOL(generic_write_end);
vfs: pagecache usage optimization for pagesize!=blocksize When we read some part of a file through pagecache, if there is a pagecache of corresponding index but this page is not uptodate, read IO is issued and this page will be uptodate. I think this is good for pagesize == blocksize environment but there is room for improvement on pagesize != blocksize environment. Because in this case a page can have multiple buffers and even if a page is not uptodate, some buffers can be uptodate. So I suggest that when all buffers which correspond to a part of a file that we want to read are uptodate, use this pagecache and copy data from this pagecache to user buffer even if a page is not uptodate. This can reduce read IO and improve system throughput. I wrote a benchmark program and got result number with this program. This benchmark do: 1: mount and open a test file. 2: create a 512MB file. 3: close a file and umount. 4: mount and again open a test file. 5: pwrite randomly 300000 times on a test file. offset is aligned by IO size(1024bytes). 6: measure time of preading randomly 100000 times on a test file. The result was: 2.6.26 330 sec 2.6.26-patched 226 sec Arch:i386 Filesystem:ext3 Blocksize:1024 bytes Memory: 1GB On ext3/4, a file is written through buffer/block. So random read/write mixed workloads or random read after random write workloads are optimized with this patch under pagesize != blocksize environment. This test result showed this. The benchmark program is as follows: #include <stdio.h> #include <sys/types.h> #include <sys/stat.h> #include <fcntl.h> #include <unistd.h> #include <time.h> #include <stdlib.h> #include <string.h> #include <sys/mount.h> #define LEN 1024 #define LOOP 1024*512 /* 512MB */ main(void) { unsigned long i, offset, filesize; int fd; char buf[LEN]; time_t t1, t2; if (mount("/dev/sda1", "/root/test1/", "ext3", 0, 0) < 0) { perror("cannot mount\n"); exit(1); } memset(buf, 0, LEN); fd = open("/root/test1/testfile", O_CREAT|O_RDWR|O_TRUNC); if (fd < 0) { perror("cannot open file\n"); exit(1); } for (i = 0; i < LOOP; i++) write(fd, buf, LEN); close(fd); if (umount("/root/test1/") < 0) { perror("cannot umount\n"); exit(1); } if (mount("/dev/sda1", "/root/test1/", "ext3", 0, 0) < 0) { perror("cannot mount\n"); exit(1); } fd = open("/root/test1/testfile", O_RDWR); if (fd < 0) { perror("cannot open file\n"); exit(1); } filesize = LEN * LOOP; for (i = 0; i < 300000; i++){ offset = (random() % filesize) & (~(LEN - 1)); pwrite(fd, buf, LEN, offset); } printf("start test\n"); time(&t1); for (i = 0; i < 100000; i++){ offset = (random() % filesize) & (~(LEN - 1)); pread(fd, buf, LEN, offset); } time(&t2); printf("%ld sec\n", t2-t1); close(fd); if (umount("/root/test1/") < 0) { perror("cannot umount\n"); exit(1); } } Signed-off-by: Hisashi Hifumi <hifumi.hisashi@oss.ntt.co.jp> Cc: Nick Piggin <nickpiggin@yahoo.com.au> Cc: Christoph Hellwig <hch@infradead.org> Cc: Jan Kara <jack@ucw.cz> Cc: <linux-ext4@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-28 16:46:36 -06:00
/*
* block_is_partially_uptodate checks whether buffers within a page are
* uptodate or not.
*
* Returns true if all buffers which correspond to a file portion
* we want to read are uptodate.
*/
int block_is_partially_uptodate(struct page *page, read_descriptor_t *desc,
unsigned long from)
{
struct inode *inode = page->mapping->host;
unsigned block_start, block_end, blocksize;
unsigned to;
struct buffer_head *bh, *head;
int ret = 1;
if (!page_has_buffers(page))
return 0;
blocksize = 1 << inode->i_blkbits;
to = min_t(unsigned, PAGE_CACHE_SIZE - from, desc->count);
to = from + to;
if (from < blocksize && to > PAGE_CACHE_SIZE - blocksize)
return 0;
head = page_buffers(page);
bh = head;
block_start = 0;
do {
block_end = block_start + blocksize;
if (block_end > from && block_start < to) {
if (!buffer_uptodate(bh)) {
ret = 0;
break;
}
if (block_end >= to)
break;
}
block_start = block_end;
bh = bh->b_this_page;
} while (bh != head);
return ret;
}
EXPORT_SYMBOL(block_is_partially_uptodate);
/*
* Generic "read page" function for block devices that have the normal
* get_block functionality. This is most of the block device filesystems.
* Reads the page asynchronously --- the unlock_buffer() and
* set/clear_buffer_uptodate() functions propagate buffer state into the
* page struct once IO has completed.
*/
int block_read_full_page(struct page *page, get_block_t *get_block)
{
struct inode *inode = page->mapping->host;
sector_t iblock, lblock;
struct buffer_head *bh, *head, *arr[MAX_BUF_PER_PAGE];
unsigned int blocksize;
int nr, i;
int fully_mapped = 1;
BUG_ON(!PageLocked(page));
blocksize = 1 << inode->i_blkbits;
if (!page_has_buffers(page))
create_empty_buffers(page, blocksize, 0);
head = page_buffers(page);
iblock = (sector_t)page->index << (PAGE_CACHE_SHIFT - inode->i_blkbits);
lblock = (i_size_read(inode)+blocksize-1) >> inode->i_blkbits;
bh = head;
nr = 0;
i = 0;
do {
if (buffer_uptodate(bh))
continue;
if (!buffer_mapped(bh)) {
int err = 0;
fully_mapped = 0;
if (iblock < lblock) {
WARN_ON(bh->b_size != blocksize);
err = get_block(inode, iblock, bh, 0);
if (err)
SetPageError(page);
}
if (!buffer_mapped(bh)) {
Pagecache zeroing: zero_user_segment, zero_user_segments and zero_user Simplify page cache zeroing of segments of pages through 3 functions zero_user_segments(page, start1, end1, start2, end2) Zeros two segments of the page. It takes the position where to start and end the zeroing which avoids length calculations and makes code clearer. zero_user_segment(page, start, end) Same for a single segment. zero_user(page, start, length) Length variant for the case where we know the length. We remove the zero_user_page macro. Issues: 1. Its a macro. Inline functions are preferable. 2. The KM_USER0 macro is only defined for HIGHMEM. Having to treat this special case everywhere makes the code needlessly complex. The parameter for zeroing is always KM_USER0 except in one single case that we open code. Avoiding KM_USER0 makes a lot of code not having to be dealing with the special casing for HIGHMEM anymore. Dealing with kmap is only necessary for HIGHMEM configurations. In those configurations we use KM_USER0 like we do for a series of other functions defined in highmem.h. Since KM_USER0 is depends on HIGHMEM the existing zero_user_page function could not be a macro. zero_user_* functions introduced here can be be inline because that constant is not used when these functions are called. Also extract the flushing of the caches to be outside of the kmap. [akpm@linux-foundation.org: fix nfs and ntfs build] [akpm@linux-foundation.org: fix ntfs build some more] Signed-off-by: Christoph Lameter <clameter@sgi.com> Cc: Steven French <sfrench@us.ibm.com> Cc: Michael Halcrow <mhalcrow@us.ibm.com> Cc: <linux-ext4@vger.kernel.org> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Cc: "J. Bruce Fields" <bfields@fieldses.org> Cc: Anton Altaparmakov <aia21@cantab.net> Cc: Mark Fasheh <mark.fasheh@oracle.com> Cc: David Chinner <dgc@sgi.com> Cc: Michael Halcrow <mhalcrow@us.ibm.com> Cc: Steven French <sfrench@us.ibm.com> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-04 23:28:29 -07:00
zero_user(page, i * blocksize, blocksize);
if (!err)
set_buffer_uptodate(bh);
continue;
}
/*
* get_block() might have updated the buffer
* synchronously
*/
if (buffer_uptodate(bh))
continue;
}
arr[nr++] = bh;
} while (i++, iblock++, (bh = bh->b_this_page) != head);
if (fully_mapped)
SetPageMappedToDisk(page);
if (!nr) {
/*
* All buffers are uptodate - we can set the page uptodate
* as well. But not if get_block() returned an error.
*/
if (!PageError(page))
SetPageUptodate(page);
unlock_page(page);
return 0;
}
/* Stage two: lock the buffers */
for (i = 0; i < nr; i++) {
bh = arr[i];
lock_buffer(bh);
mark_buffer_async_read(bh);
}
/*
* Stage 3: start the IO. Check for uptodateness
* inside the buffer lock in case another process reading
* the underlying blockdev brought it uptodate (the sct fix).
*/
for (i = 0; i < nr; i++) {
bh = arr[i];
if (buffer_uptodate(bh))
end_buffer_async_read(bh, 1);
else
submit_bh(READ, bh);
}
return 0;
}
/* utility function for filesystems that need to do work on expanding
* truncates. Uses filesystem pagecache writes to allow the filesystem to
* deal with the hole.
*/
int generic_cont_expand_simple(struct inode *inode, loff_t size)
{
struct address_space *mapping = inode->i_mapping;
struct page *page;
void *fsdata;
unsigned long limit;
int err;
err = -EFBIG;
limit = current->signal->rlim[RLIMIT_FSIZE].rlim_cur;
if (limit != RLIM_INFINITY && size > (loff_t)limit) {
send_sig(SIGXFSZ, current, 0);
goto out;
}
if (size > inode->i_sb->s_maxbytes)
goto out;
err = pagecache_write_begin(NULL, mapping, size, 0,
AOP_FLAG_UNINTERRUPTIBLE|AOP_FLAG_CONT_EXPAND,
&page, &fsdata);
if (err)
goto out;
err = pagecache_write_end(NULL, mapping, size, 0, 0, page, fsdata);
BUG_ON(err > 0);
out:
return err;
}
static int cont_expand_zero(struct file *file, struct address_space *mapping,
loff_t pos, loff_t *bytes)
{
struct inode *inode = mapping->host;
unsigned blocksize = 1 << inode->i_blkbits;
struct page *page;
void *fsdata;
pgoff_t index, curidx;
loff_t curpos;
unsigned zerofrom, offset, len;
int err = 0;
index = pos >> PAGE_CACHE_SHIFT;
offset = pos & ~PAGE_CACHE_MASK;
while (index > (curidx = (curpos = *bytes)>>PAGE_CACHE_SHIFT)) {
zerofrom = curpos & ~PAGE_CACHE_MASK;
if (zerofrom & (blocksize-1)) {
*bytes |= (blocksize-1);
(*bytes)++;
}
len = PAGE_CACHE_SIZE - zerofrom;
err = pagecache_write_begin(file, mapping, curpos, len,
AOP_FLAG_UNINTERRUPTIBLE,
&page, &fsdata);
if (err)
goto out;
Pagecache zeroing: zero_user_segment, zero_user_segments and zero_user Simplify page cache zeroing of segments of pages through 3 functions zero_user_segments(page, start1, end1, start2, end2) Zeros two segments of the page. It takes the position where to start and end the zeroing which avoids length calculations and makes code clearer. zero_user_segment(page, start, end) Same for a single segment. zero_user(page, start, length) Length variant for the case where we know the length. We remove the zero_user_page macro. Issues: 1. Its a macro. Inline functions are preferable. 2. The KM_USER0 macro is only defined for HIGHMEM. Having to treat this special case everywhere makes the code needlessly complex. The parameter for zeroing is always KM_USER0 except in one single case that we open code. Avoiding KM_USER0 makes a lot of code not having to be dealing with the special casing for HIGHMEM anymore. Dealing with kmap is only necessary for HIGHMEM configurations. In those configurations we use KM_USER0 like we do for a series of other functions defined in highmem.h. Since KM_USER0 is depends on HIGHMEM the existing zero_user_page function could not be a macro. zero_user_* functions introduced here can be be inline because that constant is not used when these functions are called. Also extract the flushing of the caches to be outside of the kmap. [akpm@linux-foundation.org: fix nfs and ntfs build] [akpm@linux-foundation.org: fix ntfs build some more] Signed-off-by: Christoph Lameter <clameter@sgi.com> Cc: Steven French <sfrench@us.ibm.com> Cc: Michael Halcrow <mhalcrow@us.ibm.com> Cc: <linux-ext4@vger.kernel.org> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Cc: "J. Bruce Fields" <bfields@fieldses.org> Cc: Anton Altaparmakov <aia21@cantab.net> Cc: Mark Fasheh <mark.fasheh@oracle.com> Cc: David Chinner <dgc@sgi.com> Cc: Michael Halcrow <mhalcrow@us.ibm.com> Cc: Steven French <sfrench@us.ibm.com> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-04 23:28:29 -07:00
zero_user(page, zerofrom, len);
err = pagecache_write_end(file, mapping, curpos, len, len,
page, fsdata);
if (err < 0)
goto out;
BUG_ON(err != len);
err = 0;
balance_dirty_pages_ratelimited(mapping);
}
/* page covers the boundary, find the boundary offset */
if (index == curidx) {
zerofrom = curpos & ~PAGE_CACHE_MASK;
/* if we will expand the thing last block will be filled */
if (offset <= zerofrom) {
goto out;
}
if (zerofrom & (blocksize-1)) {
*bytes |= (blocksize-1);
(*bytes)++;
}
len = offset - zerofrom;
err = pagecache_write_begin(file, mapping, curpos, len,
AOP_FLAG_UNINTERRUPTIBLE,
&page, &fsdata);
if (err)
goto out;
Pagecache zeroing: zero_user_segment, zero_user_segments and zero_user Simplify page cache zeroing of segments of pages through 3 functions zero_user_segments(page, start1, end1, start2, end2) Zeros two segments of the page. It takes the position where to start and end the zeroing which avoids length calculations and makes code clearer. zero_user_segment(page, start, end) Same for a single segment. zero_user(page, start, length) Length variant for the case where we know the length. We remove the zero_user_page macro. Issues: 1. Its a macro. Inline functions are preferable. 2. The KM_USER0 macro is only defined for HIGHMEM. Having to treat this special case everywhere makes the code needlessly complex. The parameter for zeroing is always KM_USER0 except in one single case that we open code. Avoiding KM_USER0 makes a lot of code not having to be dealing with the special casing for HIGHMEM anymore. Dealing with kmap is only necessary for HIGHMEM configurations. In those configurations we use KM_USER0 like we do for a series of other functions defined in highmem.h. Since KM_USER0 is depends on HIGHMEM the existing zero_user_page function could not be a macro. zero_user_* functions introduced here can be be inline because that constant is not used when these functions are called. Also extract the flushing of the caches to be outside of the kmap. [akpm@linux-foundation.org: fix nfs and ntfs build] [akpm@linux-foundation.org: fix ntfs build some more] Signed-off-by: Christoph Lameter <clameter@sgi.com> Cc: Steven French <sfrench@us.ibm.com> Cc: Michael Halcrow <mhalcrow@us.ibm.com> Cc: <linux-ext4@vger.kernel.org> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Cc: "J. Bruce Fields" <bfields@fieldses.org> Cc: Anton Altaparmakov <aia21@cantab.net> Cc: Mark Fasheh <mark.fasheh@oracle.com> Cc: David Chinner <dgc@sgi.com> Cc: Michael Halcrow <mhalcrow@us.ibm.com> Cc: Steven French <sfrench@us.ibm.com> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-04 23:28:29 -07:00
zero_user(page, zerofrom, len);
err = pagecache_write_end(file, mapping, curpos, len, len,
page, fsdata);
if (err < 0)
goto out;
BUG_ON(err != len);
err = 0;
}
out:
return err;
}
/*
* For moronic filesystems that do not allow holes in file.
* We may have to extend the file.
*/
int cont_write_begin(struct file *file, struct address_space *mapping,
loff_t pos, unsigned len, unsigned flags,
struct page **pagep, void **fsdata,
get_block_t *get_block, loff_t *bytes)
{
struct inode *inode = mapping->host;
unsigned blocksize = 1 << inode->i_blkbits;
unsigned zerofrom;
int err;
err = cont_expand_zero(file, mapping, pos, bytes);
if (err)
goto out;
zerofrom = *bytes & ~PAGE_CACHE_MASK;
if (pos+len > *bytes && zerofrom & (blocksize-1)) {
*bytes |= (blocksize-1);
(*bytes)++;
}
*pagep = NULL;
err = block_write_begin(file, mapping, pos, len,
flags, pagep, fsdata, get_block);
out:
return err;
}
int block_prepare_write(struct page *page, unsigned from, unsigned to,
get_block_t *get_block)
{
struct inode *inode = page->mapping->host;
int err = __block_prepare_write(inode, page, from, to, get_block);
if (err)
ClearPageUptodate(page);
return err;
}
int block_commit_write(struct page *page, unsigned from, unsigned to)
{
struct inode *inode = page->mapping->host;
__block_commit_write(inode,page,from,to);
return 0;
}
/*
* block_page_mkwrite() is not allowed to change the file size as it gets
* called from a page fault handler when a page is first dirtied. Hence we must
* be careful to check for EOF conditions here. We set the page up correctly
* for a written page which means we get ENOSPC checking when writing into
* holes and correct delalloc and unwritten extent mapping on filesystems that
* support these features.
*
* We are not allowed to take the i_mutex here so we have to play games to
* protect against truncate races as the page could now be beyond EOF. Because
* vmtruncate() writes the inode size before removing pages, once we have the
* page lock we can determine safely if the page is beyond EOF. If it is not
* beyond EOF, then the page is guaranteed safe against truncation until we
* unlock the page.
*/
int
block_page_mkwrite(struct vm_area_struct *vma, struct page *page,
get_block_t get_block)
{
struct inode *inode = vma->vm_file->f_path.dentry->d_inode;
unsigned long end;
loff_t size;
int ret = -EINVAL;
lock_page(page);
size = i_size_read(inode);
if ((page->mapping != inode->i_mapping) ||
(page_offset(page) > size)) {
/* page got truncated out from underneath us */
goto out_unlock;
}
/* page is wholly or partially inside EOF */
if (((page->index + 1) << PAGE_CACHE_SHIFT) > size)
end = size & ~PAGE_CACHE_MASK;
else
end = PAGE_CACHE_SIZE;
ret = block_prepare_write(page, 0, end, get_block);
if (!ret)
ret = block_commit_write(page, 0, end);
out_unlock:
unlock_page(page);
return ret;
}
/*
* nobh_write_begin()'s prereads are special: the buffer_heads are freed
* immediately, while under the page lock. So it needs a special end_io
* handler which does not touch the bh after unlocking it.
*/
static void end_buffer_read_nobh(struct buffer_head *bh, int uptodate)
{
__end_buffer_read_notouch(bh, uptodate);
}
/*
* Attach the singly-linked list of buffers created by nobh_write_begin, to
* the page (converting it to circular linked list and taking care of page
* dirty races).
*/
static void attach_nobh_buffers(struct page *page, struct buffer_head *head)
{
struct buffer_head *bh;
BUG_ON(!PageLocked(page));
spin_lock(&page->mapping->private_lock);
bh = head;
do {
if (PageDirty(page))
set_buffer_dirty(bh);
if (!bh->b_this_page)
bh->b_this_page = head;
bh = bh->b_this_page;
} while (bh != head);
attach_page_buffers(page, head);
spin_unlock(&page->mapping->private_lock);
}
/*
* On entry, the page is fully not uptodate.
* On exit the page is fully uptodate in the areas outside (from,to)
*/
int nobh_write_begin(struct file *file, struct address_space *mapping,
loff_t pos, unsigned len, unsigned flags,
struct page **pagep, void **fsdata,
get_block_t *get_block)
{
struct inode *inode = mapping->host;
const unsigned blkbits = inode->i_blkbits;
const unsigned blocksize = 1 << blkbits;
fs: fix nobh error handling nobh mode error handling is not just pretty slack, it's wrong. One cannot zero out the whole page to ensure new blocks are zeroed, because it just brings the whole page "uptodate" with zeroes even if that may not be the correct uptodate data. Also, other parts of the page may already contain dirty data which would get lost by zeroing it out. Thirdly, the writeback of zeroes to the new blocks will also erase existing blocks. All these conditions are pagecache and/or filesystem corruption. The problem comes about because we didn't keep track of which buffers actually are new or old. However it is not enough just to keep only this state, because at the point we start dirtying parts of the page (new blocks, with zeroes), the handling of IO errors becomes impossible without buffers because the page may only be partially uptodate, in which case the page flags allone cannot capture the state of the parts of the page. So allocate all buffers for the page upfront, but leave them unattached so that they don't pick up any other references and can be freed when we're done. If the error path is hit, then zero the new buffers as the regular buffer path does, then attach the buffers to the page so that it can actually be written out correctly and be subject to the normal IO error handling paths. As an upshot, we save 1K of kernel stack on ia64 or powerpc 64K page systems. Signed-off-by: Nick Piggin <npiggin@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 02:24:48 -06:00
struct buffer_head *head, *bh;
struct page *page;
pgoff_t index;
unsigned from, to;
unsigned block_in_page;
fs: fix nobh error handling nobh mode error handling is not just pretty slack, it's wrong. One cannot zero out the whole page to ensure new blocks are zeroed, because it just brings the whole page "uptodate" with zeroes even if that may not be the correct uptodate data. Also, other parts of the page may already contain dirty data which would get lost by zeroing it out. Thirdly, the writeback of zeroes to the new blocks will also erase existing blocks. All these conditions are pagecache and/or filesystem corruption. The problem comes about because we didn't keep track of which buffers actually are new or old. However it is not enough just to keep only this state, because at the point we start dirtying parts of the page (new blocks, with zeroes), the handling of IO errors becomes impossible without buffers because the page may only be partially uptodate, in which case the page flags allone cannot capture the state of the parts of the page. So allocate all buffers for the page upfront, but leave them unattached so that they don't pick up any other references and can be freed when we're done. If the error path is hit, then zero the new buffers as the regular buffer path does, then attach the buffers to the page so that it can actually be written out correctly and be subject to the normal IO error handling paths. As an upshot, we save 1K of kernel stack on ia64 or powerpc 64K page systems. Signed-off-by: Nick Piggin <npiggin@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 02:24:48 -06:00
unsigned block_start, block_end;
sector_t block_in_file;
int nr_reads = 0;
int ret = 0;
int is_mapped_to_disk = 1;
index = pos >> PAGE_CACHE_SHIFT;
from = pos & (PAGE_CACHE_SIZE - 1);
to = from + len;
fs: symlink write_begin allocation context fix With the write_begin/write_end aops, page_symlink was broken because it could no longer pass a GFP_NOFS type mask into the point where the allocations happened. They are done in write_begin, which would always assume that the filesystem can be entered from reclaim. This bug could cause filesystem deadlocks. The funny thing with having a gfp_t mask there is that it doesn't really allow the caller to arbitrarily tinker with the context in which it can be called. It couldn't ever be GFP_ATOMIC, for example, because it needs to take the page lock. The only thing any callers care about is __GFP_FS anyway, so turn that into a single flag. Add a new flag for write_begin, AOP_FLAG_NOFS. Filesystems can now act on this flag in their write_begin function. Change __grab_cache_page to accept a nofs argument as well, to honour that flag (while we're there, change the name to grab_cache_page_write_begin which is more instructive and does away with random leading underscores). This is really a more flexible way to go in the end anyway -- if a filesystem happens to want any extra allocations aside from the pagecache ones in ints write_begin function, it may now use GFP_KERNEL (rather than GFP_NOFS) for common case allocations (eg. ocfs2_alloc_write_ctxt, for a random example). [kosaki.motohiro@jp.fujitsu.com: fix ubifs] [kosaki.motohiro@jp.fujitsu.com: fix fuse] Signed-off-by: Nick Piggin <npiggin@suse.de> Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Cc: <stable@kernel.org> [2.6.28.x] Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> [ Cleaned up the calling convention: just pass in the AOP flags untouched to the grab_cache_page_write_begin() function. That just simplifies everybody, and may even allow future expansion of the logic. - Linus ] Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-04 13:00:53 -07:00
page = grab_cache_page_write_begin(mapping, index, flags);
if (!page)
return -ENOMEM;
*pagep = page;
*fsdata = NULL;
if (page_has_buffers(page)) {
unlock_page(page);
page_cache_release(page);
*pagep = NULL;
return block_write_begin(file, mapping, pos, len, flags, pagep,
fsdata, get_block);
}
fs: fix nobh error handling nobh mode error handling is not just pretty slack, it's wrong. One cannot zero out the whole page to ensure new blocks are zeroed, because it just brings the whole page "uptodate" with zeroes even if that may not be the correct uptodate data. Also, other parts of the page may already contain dirty data which would get lost by zeroing it out. Thirdly, the writeback of zeroes to the new blocks will also erase existing blocks. All these conditions are pagecache and/or filesystem corruption. The problem comes about because we didn't keep track of which buffers actually are new or old. However it is not enough just to keep only this state, because at the point we start dirtying parts of the page (new blocks, with zeroes), the handling of IO errors becomes impossible without buffers because the page may only be partially uptodate, in which case the page flags allone cannot capture the state of the parts of the page. So allocate all buffers for the page upfront, but leave them unattached so that they don't pick up any other references and can be freed when we're done. If the error path is hit, then zero the new buffers as the regular buffer path does, then attach the buffers to the page so that it can actually be written out correctly and be subject to the normal IO error handling paths. As an upshot, we save 1K of kernel stack on ia64 or powerpc 64K page systems. Signed-off-by: Nick Piggin <npiggin@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 02:24:48 -06:00
if (PageMappedToDisk(page))
return 0;
fs: fix nobh error handling nobh mode error handling is not just pretty slack, it's wrong. One cannot zero out the whole page to ensure new blocks are zeroed, because it just brings the whole page "uptodate" with zeroes even if that may not be the correct uptodate data. Also, other parts of the page may already contain dirty data which would get lost by zeroing it out. Thirdly, the writeback of zeroes to the new blocks will also erase existing blocks. All these conditions are pagecache and/or filesystem corruption. The problem comes about because we didn't keep track of which buffers actually are new or old. However it is not enough just to keep only this state, because at the point we start dirtying parts of the page (new blocks, with zeroes), the handling of IO errors becomes impossible without buffers because the page may only be partially uptodate, in which case the page flags allone cannot capture the state of the parts of the page. So allocate all buffers for the page upfront, but leave them unattached so that they don't pick up any other references and can be freed when we're done. If the error path is hit, then zero the new buffers as the regular buffer path does, then attach the buffers to the page so that it can actually be written out correctly and be subject to the normal IO error handling paths. As an upshot, we save 1K of kernel stack on ia64 or powerpc 64K page systems. Signed-off-by: Nick Piggin <npiggin@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 02:24:48 -06:00
/*
* Allocate buffers so that we can keep track of state, and potentially
* attach them to the page if an error occurs. In the common case of
* no error, they will just be freed again without ever being attached
* to the page (which is all OK, because we're under the page lock).
*
* Be careful: the buffer linked list is a NULL terminated one, rather
* than the circular one we're used to.
*/
head = alloc_page_buffers(page, blocksize, 0);
if (!head) {
ret = -ENOMEM;
goto out_release;
}
fs: fix nobh error handling nobh mode error handling is not just pretty slack, it's wrong. One cannot zero out the whole page to ensure new blocks are zeroed, because it just brings the whole page "uptodate" with zeroes even if that may not be the correct uptodate data. Also, other parts of the page may already contain dirty data which would get lost by zeroing it out. Thirdly, the writeback of zeroes to the new blocks will also erase existing blocks. All these conditions are pagecache and/or filesystem corruption. The problem comes about because we didn't keep track of which buffers actually are new or old. However it is not enough just to keep only this state, because at the point we start dirtying parts of the page (new blocks, with zeroes), the handling of IO errors becomes impossible without buffers because the page may only be partially uptodate, in which case the page flags allone cannot capture the state of the parts of the page. So allocate all buffers for the page upfront, but leave them unattached so that they don't pick up any other references and can be freed when we're done. If the error path is hit, then zero the new buffers as the regular buffer path does, then attach the buffers to the page so that it can actually be written out correctly and be subject to the normal IO error handling paths. As an upshot, we save 1K of kernel stack on ia64 or powerpc 64K page systems. Signed-off-by: Nick Piggin <npiggin@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 02:24:48 -06:00
block_in_file = (sector_t)page->index << (PAGE_CACHE_SHIFT - blkbits);
/*
* We loop across all blocks in the page, whether or not they are
* part of the affected region. This is so we can discover if the
* page is fully mapped-to-disk.
*/
fs: fix nobh error handling nobh mode error handling is not just pretty slack, it's wrong. One cannot zero out the whole page to ensure new blocks are zeroed, because it just brings the whole page "uptodate" with zeroes even if that may not be the correct uptodate data. Also, other parts of the page may already contain dirty data which would get lost by zeroing it out. Thirdly, the writeback of zeroes to the new blocks will also erase existing blocks. All these conditions are pagecache and/or filesystem corruption. The problem comes about because we didn't keep track of which buffers actually are new or old. However it is not enough just to keep only this state, because at the point we start dirtying parts of the page (new blocks, with zeroes), the handling of IO errors becomes impossible without buffers because the page may only be partially uptodate, in which case the page flags allone cannot capture the state of the parts of the page. So allocate all buffers for the page upfront, but leave them unattached so that they don't pick up any other references and can be freed when we're done. If the error path is hit, then zero the new buffers as the regular buffer path does, then attach the buffers to the page so that it can actually be written out correctly and be subject to the normal IO error handling paths. As an upshot, we save 1K of kernel stack on ia64 or powerpc 64K page systems. Signed-off-by: Nick Piggin <npiggin@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 02:24:48 -06:00
for (block_start = 0, block_in_page = 0, bh = head;
block_start < PAGE_CACHE_SIZE;
fs: fix nobh error handling nobh mode error handling is not just pretty slack, it's wrong. One cannot zero out the whole page to ensure new blocks are zeroed, because it just brings the whole page "uptodate" with zeroes even if that may not be the correct uptodate data. Also, other parts of the page may already contain dirty data which would get lost by zeroing it out. Thirdly, the writeback of zeroes to the new blocks will also erase existing blocks. All these conditions are pagecache and/or filesystem corruption. The problem comes about because we didn't keep track of which buffers actually are new or old. However it is not enough just to keep only this state, because at the point we start dirtying parts of the page (new blocks, with zeroes), the handling of IO errors becomes impossible without buffers because the page may only be partially uptodate, in which case the page flags allone cannot capture the state of the parts of the page. So allocate all buffers for the page upfront, but leave them unattached so that they don't pick up any other references and can be freed when we're done. If the error path is hit, then zero the new buffers as the regular buffer path does, then attach the buffers to the page so that it can actually be written out correctly and be subject to the normal IO error handling paths. As an upshot, we save 1K of kernel stack on ia64 or powerpc 64K page systems. Signed-off-by: Nick Piggin <npiggin@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 02:24:48 -06:00
block_in_page++, block_start += blocksize, bh = bh->b_this_page) {
int create;
fs: fix nobh error handling nobh mode error handling is not just pretty slack, it's wrong. One cannot zero out the whole page to ensure new blocks are zeroed, because it just brings the whole page "uptodate" with zeroes even if that may not be the correct uptodate data. Also, other parts of the page may already contain dirty data which would get lost by zeroing it out. Thirdly, the writeback of zeroes to the new blocks will also erase existing blocks. All these conditions are pagecache and/or filesystem corruption. The problem comes about because we didn't keep track of which buffers actually are new or old. However it is not enough just to keep only this state, because at the point we start dirtying parts of the page (new blocks, with zeroes), the handling of IO errors becomes impossible without buffers because the page may only be partially uptodate, in which case the page flags allone cannot capture the state of the parts of the page. So allocate all buffers for the page upfront, but leave them unattached so that they don't pick up any other references and can be freed when we're done. If the error path is hit, then zero the new buffers as the regular buffer path does, then attach the buffers to the page so that it can actually be written out correctly and be subject to the normal IO error handling paths. As an upshot, we save 1K of kernel stack on ia64 or powerpc 64K page systems. Signed-off-by: Nick Piggin <npiggin@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 02:24:48 -06:00
block_end = block_start + blocksize;
bh->b_state = 0;
create = 1;
if (block_start >= to)
create = 0;
ret = get_block(inode, block_in_file + block_in_page,
fs: fix nobh error handling nobh mode error handling is not just pretty slack, it's wrong. One cannot zero out the whole page to ensure new blocks are zeroed, because it just brings the whole page "uptodate" with zeroes even if that may not be the correct uptodate data. Also, other parts of the page may already contain dirty data which would get lost by zeroing it out. Thirdly, the writeback of zeroes to the new blocks will also erase existing blocks. All these conditions are pagecache and/or filesystem corruption. The problem comes about because we didn't keep track of which buffers actually are new or old. However it is not enough just to keep only this state, because at the point we start dirtying parts of the page (new blocks, with zeroes), the handling of IO errors becomes impossible without buffers because the page may only be partially uptodate, in which case the page flags allone cannot capture the state of the parts of the page. So allocate all buffers for the page upfront, but leave them unattached so that they don't pick up any other references and can be freed when we're done. If the error path is hit, then zero the new buffers as the regular buffer path does, then attach the buffers to the page so that it can actually be written out correctly and be subject to the normal IO error handling paths. As an upshot, we save 1K of kernel stack on ia64 or powerpc 64K page systems. Signed-off-by: Nick Piggin <npiggin@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 02:24:48 -06:00
bh, create);
if (ret)
goto failed;
fs: fix nobh error handling nobh mode error handling is not just pretty slack, it's wrong. One cannot zero out the whole page to ensure new blocks are zeroed, because it just brings the whole page "uptodate" with zeroes even if that may not be the correct uptodate data. Also, other parts of the page may already contain dirty data which would get lost by zeroing it out. Thirdly, the writeback of zeroes to the new blocks will also erase existing blocks. All these conditions are pagecache and/or filesystem corruption. The problem comes about because we didn't keep track of which buffers actually are new or old. However it is not enough just to keep only this state, because at the point we start dirtying parts of the page (new blocks, with zeroes), the handling of IO errors becomes impossible without buffers because the page may only be partially uptodate, in which case the page flags allone cannot capture the state of the parts of the page. So allocate all buffers for the page upfront, but leave them unattached so that they don't pick up any other references and can be freed when we're done. If the error path is hit, then zero the new buffers as the regular buffer path does, then attach the buffers to the page so that it can actually be written out correctly and be subject to the normal IO error handling paths. As an upshot, we save 1K of kernel stack on ia64 or powerpc 64K page systems. Signed-off-by: Nick Piggin <npiggin@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 02:24:48 -06:00
if (!buffer_mapped(bh))
is_mapped_to_disk = 0;
fs: fix nobh error handling nobh mode error handling is not just pretty slack, it's wrong. One cannot zero out the whole page to ensure new blocks are zeroed, because it just brings the whole page "uptodate" with zeroes even if that may not be the correct uptodate data. Also, other parts of the page may already contain dirty data which would get lost by zeroing it out. Thirdly, the writeback of zeroes to the new blocks will also erase existing blocks. All these conditions are pagecache and/or filesystem corruption. The problem comes about because we didn't keep track of which buffers actually are new or old. However it is not enough just to keep only this state, because at the point we start dirtying parts of the page (new blocks, with zeroes), the handling of IO errors becomes impossible without buffers because the page may only be partially uptodate, in which case the page flags allone cannot capture the state of the parts of the page. So allocate all buffers for the page upfront, but leave them unattached so that they don't pick up any other references and can be freed when we're done. If the error path is hit, then zero the new buffers as the regular buffer path does, then attach the buffers to the page so that it can actually be written out correctly and be subject to the normal IO error handling paths. As an upshot, we save 1K of kernel stack on ia64 or powerpc 64K page systems. Signed-off-by: Nick Piggin <npiggin@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 02:24:48 -06:00
if (buffer_new(bh))
unmap_underlying_metadata(bh->b_bdev, bh->b_blocknr);
if (PageUptodate(page)) {
set_buffer_uptodate(bh);
continue;
fs: fix nobh error handling nobh mode error handling is not just pretty slack, it's wrong. One cannot zero out the whole page to ensure new blocks are zeroed, because it just brings the whole page "uptodate" with zeroes even if that may not be the correct uptodate data. Also, other parts of the page may already contain dirty data which would get lost by zeroing it out. Thirdly, the writeback of zeroes to the new blocks will also erase existing blocks. All these conditions are pagecache and/or filesystem corruption. The problem comes about because we didn't keep track of which buffers actually are new or old. However it is not enough just to keep only this state, because at the point we start dirtying parts of the page (new blocks, with zeroes), the handling of IO errors becomes impossible without buffers because the page may only be partially uptodate, in which case the page flags allone cannot capture the state of the parts of the page. So allocate all buffers for the page upfront, but leave them unattached so that they don't pick up any other references and can be freed when we're done. If the error path is hit, then zero the new buffers as the regular buffer path does, then attach the buffers to the page so that it can actually be written out correctly and be subject to the normal IO error handling paths. As an upshot, we save 1K of kernel stack on ia64 or powerpc 64K page systems. Signed-off-by: Nick Piggin <npiggin@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 02:24:48 -06:00
}
if (buffer_new(bh) || !buffer_mapped(bh)) {
Pagecache zeroing: zero_user_segment, zero_user_segments and zero_user Simplify page cache zeroing of segments of pages through 3 functions zero_user_segments(page, start1, end1, start2, end2) Zeros two segments of the page. It takes the position where to start and end the zeroing which avoids length calculations and makes code clearer. zero_user_segment(page, start, end) Same for a single segment. zero_user(page, start, length) Length variant for the case where we know the length. We remove the zero_user_page macro. Issues: 1. Its a macro. Inline functions are preferable. 2. The KM_USER0 macro is only defined for HIGHMEM. Having to treat this special case everywhere makes the code needlessly complex. The parameter for zeroing is always KM_USER0 except in one single case that we open code. Avoiding KM_USER0 makes a lot of code not having to be dealing with the special casing for HIGHMEM anymore. Dealing with kmap is only necessary for HIGHMEM configurations. In those configurations we use KM_USER0 like we do for a series of other functions defined in highmem.h. Since KM_USER0 is depends on HIGHMEM the existing zero_user_page function could not be a macro. zero_user_* functions introduced here can be be inline because that constant is not used when these functions are called. Also extract the flushing of the caches to be outside of the kmap. [akpm@linux-foundation.org: fix nfs and ntfs build] [akpm@linux-foundation.org: fix ntfs build some more] Signed-off-by: Christoph Lameter <clameter@sgi.com> Cc: Steven French <sfrench@us.ibm.com> Cc: Michael Halcrow <mhalcrow@us.ibm.com> Cc: <linux-ext4@vger.kernel.org> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Cc: "J. Bruce Fields" <bfields@fieldses.org> Cc: Anton Altaparmakov <aia21@cantab.net> Cc: Mark Fasheh <mark.fasheh@oracle.com> Cc: David Chinner <dgc@sgi.com> Cc: Michael Halcrow <mhalcrow@us.ibm.com> Cc: Steven French <sfrench@us.ibm.com> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-04 23:28:29 -07:00
zero_user_segments(page, block_start, from,
to, block_end);
continue;
}
fs: fix nobh error handling nobh mode error handling is not just pretty slack, it's wrong. One cannot zero out the whole page to ensure new blocks are zeroed, because it just brings the whole page "uptodate" with zeroes even if that may not be the correct uptodate data. Also, other parts of the page may already contain dirty data which would get lost by zeroing it out. Thirdly, the writeback of zeroes to the new blocks will also erase existing blocks. All these conditions are pagecache and/or filesystem corruption. The problem comes about because we didn't keep track of which buffers actually are new or old. However it is not enough just to keep only this state, because at the point we start dirtying parts of the page (new blocks, with zeroes), the handling of IO errors becomes impossible without buffers because the page may only be partially uptodate, in which case the page flags allone cannot capture the state of the parts of the page. So allocate all buffers for the page upfront, but leave them unattached so that they don't pick up any other references and can be freed when we're done. If the error path is hit, then zero the new buffers as the regular buffer path does, then attach the buffers to the page so that it can actually be written out correctly and be subject to the normal IO error handling paths. As an upshot, we save 1K of kernel stack on ia64 or powerpc 64K page systems. Signed-off-by: Nick Piggin <npiggin@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 02:24:48 -06:00
if (buffer_uptodate(bh))
continue; /* reiserfs does this */
if (block_start < from || block_end > to) {
fs: fix nobh error handling nobh mode error handling is not just pretty slack, it's wrong. One cannot zero out the whole page to ensure new blocks are zeroed, because it just brings the whole page "uptodate" with zeroes even if that may not be the correct uptodate data. Also, other parts of the page may already contain dirty data which would get lost by zeroing it out. Thirdly, the writeback of zeroes to the new blocks will also erase existing blocks. All these conditions are pagecache and/or filesystem corruption. The problem comes about because we didn't keep track of which buffers actually are new or old. However it is not enough just to keep only this state, because at the point we start dirtying parts of the page (new blocks, with zeroes), the handling of IO errors becomes impossible without buffers because the page may only be partially uptodate, in which case the page flags allone cannot capture the state of the parts of the page. So allocate all buffers for the page upfront, but leave them unattached so that they don't pick up any other references and can be freed when we're done. If the error path is hit, then zero the new buffers as the regular buffer path does, then attach the buffers to the page so that it can actually be written out correctly and be subject to the normal IO error handling paths. As an upshot, we save 1K of kernel stack on ia64 or powerpc 64K page systems. Signed-off-by: Nick Piggin <npiggin@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 02:24:48 -06:00
lock_buffer(bh);
bh->b_end_io = end_buffer_read_nobh;
submit_bh(READ, bh);
nr_reads++;
}
}
if (nr_reads) {
/*
* The page is locked, so these buffers are protected from
* any VM or truncate activity. Hence we don't need to care
* for the buffer_head refcounts.
*/
fs: fix nobh error handling nobh mode error handling is not just pretty slack, it's wrong. One cannot zero out the whole page to ensure new blocks are zeroed, because it just brings the whole page "uptodate" with zeroes even if that may not be the correct uptodate data. Also, other parts of the page may already contain dirty data which would get lost by zeroing it out. Thirdly, the writeback of zeroes to the new blocks will also erase existing blocks. All these conditions are pagecache and/or filesystem corruption. The problem comes about because we didn't keep track of which buffers actually are new or old. However it is not enough just to keep only this state, because at the point we start dirtying parts of the page (new blocks, with zeroes), the handling of IO errors becomes impossible without buffers because the page may only be partially uptodate, in which case the page flags allone cannot capture the state of the parts of the page. So allocate all buffers for the page upfront, but leave them unattached so that they don't pick up any other references and can be freed when we're done. If the error path is hit, then zero the new buffers as the regular buffer path does, then attach the buffers to the page so that it can actually be written out correctly and be subject to the normal IO error handling paths. As an upshot, we save 1K of kernel stack on ia64 or powerpc 64K page systems. Signed-off-by: Nick Piggin <npiggin@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 02:24:48 -06:00
for (bh = head; bh; bh = bh->b_this_page) {
wait_on_buffer(bh);
if (!buffer_uptodate(bh))
ret = -EIO;
}
if (ret)
goto failed;
}
if (is_mapped_to_disk)
SetPageMappedToDisk(page);
*fsdata = head; /* to be released by nobh_write_end */
fs: fix nobh error handling nobh mode error handling is not just pretty slack, it's wrong. One cannot zero out the whole page to ensure new blocks are zeroed, because it just brings the whole page "uptodate" with zeroes even if that may not be the correct uptodate data. Also, other parts of the page may already contain dirty data which would get lost by zeroing it out. Thirdly, the writeback of zeroes to the new blocks will also erase existing blocks. All these conditions are pagecache and/or filesystem corruption. The problem comes about because we didn't keep track of which buffers actually are new or old. However it is not enough just to keep only this state, because at the point we start dirtying parts of the page (new blocks, with zeroes), the handling of IO errors becomes impossible without buffers because the page may only be partially uptodate, in which case the page flags allone cannot capture the state of the parts of the page. So allocate all buffers for the page upfront, but leave them unattached so that they don't pick up any other references and can be freed when we're done. If the error path is hit, then zero the new buffers as the regular buffer path does, then attach the buffers to the page so that it can actually be written out correctly and be subject to the normal IO error handling paths. As an upshot, we save 1K of kernel stack on ia64 or powerpc 64K page systems. Signed-off-by: Nick Piggin <npiggin@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 02:24:48 -06:00
return 0;
failed:
BUG_ON(!ret);
/*
fs: fix nobh error handling nobh mode error handling is not just pretty slack, it's wrong. One cannot zero out the whole page to ensure new blocks are zeroed, because it just brings the whole page "uptodate" with zeroes even if that may not be the correct uptodate data. Also, other parts of the page may already contain dirty data which would get lost by zeroing it out. Thirdly, the writeback of zeroes to the new blocks will also erase existing blocks. All these conditions are pagecache and/or filesystem corruption. The problem comes about because we didn't keep track of which buffers actually are new or old. However it is not enough just to keep only this state, because at the point we start dirtying parts of the page (new blocks, with zeroes), the handling of IO errors becomes impossible without buffers because the page may only be partially uptodate, in which case the page flags allone cannot capture the state of the parts of the page. So allocate all buffers for the page upfront, but leave them unattached so that they don't pick up any other references and can be freed when we're done. If the error path is hit, then zero the new buffers as the regular buffer path does, then attach the buffers to the page so that it can actually be written out correctly and be subject to the normal IO error handling paths. As an upshot, we save 1K of kernel stack on ia64 or powerpc 64K page systems. Signed-off-by: Nick Piggin <npiggin@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 02:24:48 -06:00
* Error recovery is a bit difficult. We need to zero out blocks that
* were newly allocated, and dirty them to ensure they get written out.
* Buffers need to be attached to the page at this point, otherwise
* the handling of potential IO errors during writeout would be hard
* (could try doing synchronous writeout, but what if that fails too?)
*/
attach_nobh_buffers(page, head);
page_zero_new_buffers(page, from, to);
fs: fix nobh error handling nobh mode error handling is not just pretty slack, it's wrong. One cannot zero out the whole page to ensure new blocks are zeroed, because it just brings the whole page "uptodate" with zeroes even if that may not be the correct uptodate data. Also, other parts of the page may already contain dirty data which would get lost by zeroing it out. Thirdly, the writeback of zeroes to the new blocks will also erase existing blocks. All these conditions are pagecache and/or filesystem corruption. The problem comes about because we didn't keep track of which buffers actually are new or old. However it is not enough just to keep only this state, because at the point we start dirtying parts of the page (new blocks, with zeroes), the handling of IO errors becomes impossible without buffers because the page may only be partially uptodate, in which case the page flags allone cannot capture the state of the parts of the page. So allocate all buffers for the page upfront, but leave them unattached so that they don't pick up any other references and can be freed when we're done. If the error path is hit, then zero the new buffers as the regular buffer path does, then attach the buffers to the page so that it can actually be written out correctly and be subject to the normal IO error handling paths. As an upshot, we save 1K of kernel stack on ia64 or powerpc 64K page systems. Signed-off-by: Nick Piggin <npiggin@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 02:24:48 -06:00
out_release:
unlock_page(page);
page_cache_release(page);
*pagep = NULL;
fs: fix nobh error handling nobh mode error handling is not just pretty slack, it's wrong. One cannot zero out the whole page to ensure new blocks are zeroed, because it just brings the whole page "uptodate" with zeroes even if that may not be the correct uptodate data. Also, other parts of the page may already contain dirty data which would get lost by zeroing it out. Thirdly, the writeback of zeroes to the new blocks will also erase existing blocks. All these conditions are pagecache and/or filesystem corruption. The problem comes about because we didn't keep track of which buffers actually are new or old. However it is not enough just to keep only this state, because at the point we start dirtying parts of the page (new blocks, with zeroes), the handling of IO errors becomes impossible without buffers because the page may only be partially uptodate, in which case the page flags allone cannot capture the state of the parts of the page. So allocate all buffers for the page upfront, but leave them unattached so that they don't pick up any other references and can be freed when we're done. If the error path is hit, then zero the new buffers as the regular buffer path does, then attach the buffers to the page so that it can actually be written out correctly and be subject to the normal IO error handling paths. As an upshot, we save 1K of kernel stack on ia64 or powerpc 64K page systems. Signed-off-by: Nick Piggin <npiggin@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 02:24:48 -06:00
if (pos + len > inode->i_size)
vmtruncate(inode, inode->i_size);
fs: fix nobh error handling nobh mode error handling is not just pretty slack, it's wrong. One cannot zero out the whole page to ensure new blocks are zeroed, because it just brings the whole page "uptodate" with zeroes even if that may not be the correct uptodate data. Also, other parts of the page may already contain dirty data which would get lost by zeroing it out. Thirdly, the writeback of zeroes to the new blocks will also erase existing blocks. All these conditions are pagecache and/or filesystem corruption. The problem comes about because we didn't keep track of which buffers actually are new or old. However it is not enough just to keep only this state, because at the point we start dirtying parts of the page (new blocks, with zeroes), the handling of IO errors becomes impossible without buffers because the page may only be partially uptodate, in which case the page flags allone cannot capture the state of the parts of the page. So allocate all buffers for the page upfront, but leave them unattached so that they don't pick up any other references and can be freed when we're done. If the error path is hit, then zero the new buffers as the regular buffer path does, then attach the buffers to the page so that it can actually be written out correctly and be subject to the normal IO error handling paths. As an upshot, we save 1K of kernel stack on ia64 or powerpc 64K page systems. Signed-off-by: Nick Piggin <npiggin@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 02:24:48 -06:00
return ret;
}
EXPORT_SYMBOL(nobh_write_begin);
int nobh_write_end(struct file *file, struct address_space *mapping,
loff_t pos, unsigned len, unsigned copied,
struct page *page, void *fsdata)
{
struct inode *inode = page->mapping->host;
struct buffer_head *head = fsdata;
struct buffer_head *bh;
vfs: fix data leak in nobh_write_end() Current nobh_write_end() implementation ignore partial writes(copied < len) case if page was fully mapped and simply mark page as Uptodate, which is totally wrong because area [pos+copied, pos+len) wasn't updated explicitly in previous write_begin call. It simply contains garbage from pagecache and result in data leakage. #TEST_CASE_BEGIN: ~~~~~~~~~~~~~~~~ In fact issue triggered by classical testcase open("/mnt/test", O_RDWR|O_CREAT|O_TRUNC, 0666) = 3 ftruncate(3, 409600) = 0 writev(3, [{"a", 1}, {NULL, 4095}], 2) = 1 ##TESTCASE_SOURCE: ~~~~~~~~~~~~~~~~~ #include <stdio.h> #include <stdlib.h> #include <fcntl.h> #include <sys/uio.h> #include <sys/mman.h> #include <errno.h> int main(int argc, char **argv) { int fd, ret; void* p; struct iovec iov[2]; fd = open(argv[1], O_RDWR|O_CREAT|O_TRUNC, 0666); ftruncate(fd, 409600); iov[0].iov_base="a"; iov[0].iov_len=1; iov[1].iov_base=NULL; iov[1].iov_len=4096; ret = writev(fd, iov, sizeof(iov)/sizeof(struct iovec)); printf("writev = %d, err = %d\n", ret, errno); return 0; } ##TESTCASE RESULT: ~~~~~~~~~~~~~~~~~~ [root@ts63 ~]# mount | grep mnt2 /dev/mapper/test on /mnt2 type ext2 (rw,nobh) [root@ts63 ~]# /tmp/writev /mnt2/test writev = 1, err = 0 [root@ts63 ~]# hexdump -C /mnt2/test 00000000 61 65 62 6f 6f 74 00 00 f0 b9 b4 59 3a 00 00 00 |aeboot.....Y:...| 00000010 20 00 00 00 00 00 00 00 21 00 00 00 00 00 00 00 | .......!.......| 00000020 df df df df df df df df df df df df df df df df |................| 00000030 3a 00 00 00 2a 00 00 00 21 00 00 00 00 00 00 00 |:...*...!.......| 00000040 60 c0 8c 00 00 00 00 00 40 4a 8d 00 00 00 00 00 |`.......@J......| 00000050 00 00 00 00 00 00 00 00 41 00 00 00 00 00 00 00 |........A.......| 00000060 74 69 6d 65 20 64 64 20 69 66 3d 2f 64 65 76 2f |time dd if=/dev/| 00000070 6c 6f 6f 70 30 20 20 6f 66 3d 2f 64 65 76 2f 6e |loop0 of=/dev/n| skip.. 00000f50 00 00 00 00 00 00 00 00 31 00 00 00 00 00 00 00 |........1.......| 00000f60 6d 6b 66 73 2e 65 78 74 33 20 2f 64 65 76 2f 76 |mkfs.ext3 /dev/v| 00000f70 7a 76 67 2f 74 65 73 74 20 2d 62 34 30 39 36 00 |zvg/test -b4096.| 00000f80 a0 fe 8c 00 00 00 00 00 21 00 00 00 00 00 00 00 |........!.......| 00000f90 23 31 32 30 35 39 35 30 34 30 34 00 3a 00 00 00 |#1205950404.:...| 00000fa0 20 00 8d 00 00 00 00 00 21 00 00 00 00 00 00 00 | .......!.......| 00000fb0 d0 cf 8c 00 00 00 00 00 10 d0 8c 00 00 00 00 00 |................| 00000fc0 00 00 00 00 00 00 00 00 41 00 00 00 00 00 00 00 |........A.......| 00000fd0 6d 6f 75 6e 74 20 2f 64 65 76 2f 76 7a 76 67 2f |mount /dev/vzvg/| 00000fe0 74 65 73 74 20 20 2f 76 7a 20 2d 6f 20 64 61 74 |test /vz -o dat| 00000ff0 61 3d 77 72 69 74 65 62 61 63 6b 00 00 00 00 00 |a=writeback.....| 00001000 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 |................| As you can see file's page contains garbage from pagecache instead of zeros. #TEST_CASE_END Attached patch: - Add sanity check BUG_ON in order to prevent incorrect usage by caller, This is function invariant because page can has buffers and in no zero *fadata pointer at the same time. - Always attach buffers to page is it is partial write case. - Always switch back to generic_write_end if page has buffers. This is reasonable because if page already has buffer then generic_write_begin was called previously. Signed-off-by: Dmitri Monakhov <dmonakhov@openvz.org> Reviewed-by: Nick Piggin <npiggin@suse.de> Cc: <stable@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-03-28 15:15:52 -06:00
BUG_ON(fsdata != NULL && page_has_buffers(page));
vfs: fix data leak in nobh_write_end() Current nobh_write_end() implementation ignore partial writes(copied < len) case if page was fully mapped and simply mark page as Uptodate, which is totally wrong because area [pos+copied, pos+len) wasn't updated explicitly in previous write_begin call. It simply contains garbage from pagecache and result in data leakage. #TEST_CASE_BEGIN: ~~~~~~~~~~~~~~~~ In fact issue triggered by classical testcase open("/mnt/test", O_RDWR|O_CREAT|O_TRUNC, 0666) = 3 ftruncate(3, 409600) = 0 writev(3, [{"a", 1}, {NULL, 4095}], 2) = 1 ##TESTCASE_SOURCE: ~~~~~~~~~~~~~~~~~ #include <stdio.h> #include <stdlib.h> #include <fcntl.h> #include <sys/uio.h> #include <sys/mman.h> #include <errno.h> int main(int argc, char **argv) { int fd, ret; void* p; struct iovec iov[2]; fd = open(argv[1], O_RDWR|O_CREAT|O_TRUNC, 0666); ftruncate(fd, 409600); iov[0].iov_base="a"; iov[0].iov_len=1; iov[1].iov_base=NULL; iov[1].iov_len=4096; ret = writev(fd, iov, sizeof(iov)/sizeof(struct iovec)); printf("writev = %d, err = %d\n", ret, errno); return 0; } ##TESTCASE RESULT: ~~~~~~~~~~~~~~~~~~ [root@ts63 ~]# mount | grep mnt2 /dev/mapper/test on /mnt2 type ext2 (rw,nobh) [root@ts63 ~]# /tmp/writev /mnt2/test writev = 1, err = 0 [root@ts63 ~]# hexdump -C /mnt2/test 00000000 61 65 62 6f 6f 74 00 00 f0 b9 b4 59 3a 00 00 00 |aeboot.....Y:...| 00000010 20 00 00 00 00 00 00 00 21 00 00 00 00 00 00 00 | .......!.......| 00000020 df df df df df df df df df df df df df df df df |................| 00000030 3a 00 00 00 2a 00 00 00 21 00 00 00 00 00 00 00 |:...*...!.......| 00000040 60 c0 8c 00 00 00 00 00 40 4a 8d 00 00 00 00 00 |`.......@J......| 00000050 00 00 00 00 00 00 00 00 41 00 00 00 00 00 00 00 |........A.......| 00000060 74 69 6d 65 20 64 64 20 69 66 3d 2f 64 65 76 2f |time dd if=/dev/| 00000070 6c 6f 6f 70 30 20 20 6f 66 3d 2f 64 65 76 2f 6e |loop0 of=/dev/n| skip.. 00000f50 00 00 00 00 00 00 00 00 31 00 00 00 00 00 00 00 |........1.......| 00000f60 6d 6b 66 73 2e 65 78 74 33 20 2f 64 65 76 2f 76 |mkfs.ext3 /dev/v| 00000f70 7a 76 67 2f 74 65 73 74 20 2d 62 34 30 39 36 00 |zvg/test -b4096.| 00000f80 a0 fe 8c 00 00 00 00 00 21 00 00 00 00 00 00 00 |........!.......| 00000f90 23 31 32 30 35 39 35 30 34 30 34 00 3a 00 00 00 |#1205950404.:...| 00000fa0 20 00 8d 00 00 00 00 00 21 00 00 00 00 00 00 00 | .......!.......| 00000fb0 d0 cf 8c 00 00 00 00 00 10 d0 8c 00 00 00 00 00 |................| 00000fc0 00 00 00 00 00 00 00 00 41 00 00 00 00 00 00 00 |........A.......| 00000fd0 6d 6f 75 6e 74 20 2f 64 65 76 2f 76 7a 76 67 2f |mount /dev/vzvg/| 00000fe0 74 65 73 74 20 20 2f 76 7a 20 2d 6f 20 64 61 74 |test /vz -o dat| 00000ff0 61 3d 77 72 69 74 65 62 61 63 6b 00 00 00 00 00 |a=writeback.....| 00001000 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 |................| As you can see file's page contains garbage from pagecache instead of zeros. #TEST_CASE_END Attached patch: - Add sanity check BUG_ON in order to prevent incorrect usage by caller, This is function invariant because page can has buffers and in no zero *fadata pointer at the same time. - Always attach buffers to page is it is partial write case. - Always switch back to generic_write_end if page has buffers. This is reasonable because if page already has buffer then generic_write_begin was called previously. Signed-off-by: Dmitri Monakhov <dmonakhov@openvz.org> Reviewed-by: Nick Piggin <npiggin@suse.de> Cc: <stable@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-03-28 15:15:52 -06:00
if (unlikely(copied < len) && !page_has_buffers(page))
attach_nobh_buffers(page, head);
if (page_has_buffers(page))
return generic_write_end(file, mapping, pos, len,
copied, page, fsdata);
fs: fix nobh error handling nobh mode error handling is not just pretty slack, it's wrong. One cannot zero out the whole page to ensure new blocks are zeroed, because it just brings the whole page "uptodate" with zeroes even if that may not be the correct uptodate data. Also, other parts of the page may already contain dirty data which would get lost by zeroing it out. Thirdly, the writeback of zeroes to the new blocks will also erase existing blocks. All these conditions are pagecache and/or filesystem corruption. The problem comes about because we didn't keep track of which buffers actually are new or old. However it is not enough just to keep only this state, because at the point we start dirtying parts of the page (new blocks, with zeroes), the handling of IO errors becomes impossible without buffers because the page may only be partially uptodate, in which case the page flags allone cannot capture the state of the parts of the page. So allocate all buffers for the page upfront, but leave them unattached so that they don't pick up any other references and can be freed when we're done. If the error path is hit, then zero the new buffers as the regular buffer path does, then attach the buffers to the page so that it can actually be written out correctly and be subject to the normal IO error handling paths. As an upshot, we save 1K of kernel stack on ia64 or powerpc 64K page systems. Signed-off-by: Nick Piggin <npiggin@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 02:24:48 -06:00
SetPageUptodate(page);
set_page_dirty(page);
if (pos+copied > inode->i_size) {
i_size_write(inode, pos+copied);
mark_inode_dirty(inode);
}
unlock_page(page);
page_cache_release(page);
while (head) {
bh = head;
head = head->b_this_page;
free_buffer_head(bh);
}
return copied;
}
EXPORT_SYMBOL(nobh_write_end);
/*
* nobh_writepage() - based on block_full_write_page() except
* that it tries to operate without attaching bufferheads to
* the page.
*/
int nobh_writepage(struct page *page, get_block_t *get_block,
struct writeback_control *wbc)
{
struct inode * const inode = page->mapping->host;
loff_t i_size = i_size_read(inode);
const pgoff_t end_index = i_size >> PAGE_CACHE_SHIFT;
unsigned offset;
int ret;
/* Is the page fully inside i_size? */
if (page->index < end_index)
goto out;
/* Is the page fully outside i_size? (truncate in progress) */
offset = i_size & (PAGE_CACHE_SIZE-1);
if (page->index >= end_index+1 || !offset) {
/*
* The page may have dirty, unmapped buffers. For example,
* they may have been added in ext3_writepage(). Make them
* freeable here, so the page does not leak.
*/
#if 0
/* Not really sure about this - do we need this ? */
if (page->mapping->a_ops->invalidatepage)
page->mapping->a_ops->invalidatepage(page, offset);
#endif
unlock_page(page);
return 0; /* don't care */
}
/*
* The page straddles i_size. It must be zeroed out on each and every
* writepage invocation because it may be mmapped. "A file is mapped
* in multiples of the page size. For a file that is not a multiple of
* the page size, the remaining memory is zeroed when mapped, and
* writes to that region are not written out to the file."
*/
Pagecache zeroing: zero_user_segment, zero_user_segments and zero_user Simplify page cache zeroing of segments of pages through 3 functions zero_user_segments(page, start1, end1, start2, end2) Zeros two segments of the page. It takes the position where to start and end the zeroing which avoids length calculations and makes code clearer. zero_user_segment(page, start, end) Same for a single segment. zero_user(page, start, length) Length variant for the case where we know the length. We remove the zero_user_page macro. Issues: 1. Its a macro. Inline functions are preferable. 2. The KM_USER0 macro is only defined for HIGHMEM. Having to treat this special case everywhere makes the code needlessly complex. The parameter for zeroing is always KM_USER0 except in one single case that we open code. Avoiding KM_USER0 makes a lot of code not having to be dealing with the special casing for HIGHMEM anymore. Dealing with kmap is only necessary for HIGHMEM configurations. In those configurations we use KM_USER0 like we do for a series of other functions defined in highmem.h. Since KM_USER0 is depends on HIGHMEM the existing zero_user_page function could not be a macro. zero_user_* functions introduced here can be be inline because that constant is not used when these functions are called. Also extract the flushing of the caches to be outside of the kmap. [akpm@linux-foundation.org: fix nfs and ntfs build] [akpm@linux-foundation.org: fix ntfs build some more] Signed-off-by: Christoph Lameter <clameter@sgi.com> Cc: Steven French <sfrench@us.ibm.com> Cc: Michael Halcrow <mhalcrow@us.ibm.com> Cc: <linux-ext4@vger.kernel.org> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Cc: "J. Bruce Fields" <bfields@fieldses.org> Cc: Anton Altaparmakov <aia21@cantab.net> Cc: Mark Fasheh <mark.fasheh@oracle.com> Cc: David Chinner <dgc@sgi.com> Cc: Michael Halcrow <mhalcrow@us.ibm.com> Cc: Steven French <sfrench@us.ibm.com> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-04 23:28:29 -07:00
zero_user_segment(page, offset, PAGE_CACHE_SIZE);
out:
ret = mpage_writepage(page, get_block, wbc);
if (ret == -EAGAIN)
ret = __block_write_full_page(inode, page, get_block, wbc);
return ret;
}
EXPORT_SYMBOL(nobh_writepage);
int nobh_truncate_page(struct address_space *mapping,
loff_t from, get_block_t *get_block)
{
pgoff_t index = from >> PAGE_CACHE_SHIFT;
unsigned offset = from & (PAGE_CACHE_SIZE-1);
unsigned blocksize;
sector_t iblock;
unsigned length, pos;
struct inode *inode = mapping->host;
struct page *page;
struct buffer_head map_bh;
int err;
blocksize = 1 << inode->i_blkbits;
length = offset & (blocksize - 1);
/* Block boundary? Nothing to do */
if (!length)
return 0;
length = blocksize - length;
iblock = (sector_t)index << (PAGE_CACHE_SHIFT - inode->i_blkbits);
page = grab_cache_page(mapping, index);
err = -ENOMEM;
if (!page)
goto out;
if (page_has_buffers(page)) {
has_buffers:
unlock_page(page);
page_cache_release(page);
return block_truncate_page(mapping, from, get_block);
}
/* Find the buffer that contains "offset" */
pos = blocksize;
while (offset >= pos) {
iblock++;
pos += blocksize;
}
err = get_block(inode, iblock, &map_bh, 0);
if (err)
goto unlock;
/* unmapped? It's a hole - nothing to do */
if (!buffer_mapped(&map_bh))
goto unlock;
/* Ok, it's mapped. Make sure it's up-to-date */
if (!PageUptodate(page)) {
err = mapping->a_ops->readpage(NULL, page);
if (err) {
page_cache_release(page);
goto out;
}
lock_page(page);
if (!PageUptodate(page)) {
err = -EIO;
goto unlock;
}
if (page_has_buffers(page))
goto has_buffers;
}
Pagecache zeroing: zero_user_segment, zero_user_segments and zero_user Simplify page cache zeroing of segments of pages through 3 functions zero_user_segments(page, start1, end1, start2, end2) Zeros two segments of the page. It takes the position where to start and end the zeroing which avoids length calculations and makes code clearer. zero_user_segment(page, start, end) Same for a single segment. zero_user(page, start, length) Length variant for the case where we know the length. We remove the zero_user_page macro. Issues: 1. Its a macro. Inline functions are preferable. 2. The KM_USER0 macro is only defined for HIGHMEM. Having to treat this special case everywhere makes the code needlessly complex. The parameter for zeroing is always KM_USER0 except in one single case that we open code. Avoiding KM_USER0 makes a lot of code not having to be dealing with the special casing for HIGHMEM anymore. Dealing with kmap is only necessary for HIGHMEM configurations. In those configurations we use KM_USER0 like we do for a series of other functions defined in highmem.h. Since KM_USER0 is depends on HIGHMEM the existing zero_user_page function could not be a macro. zero_user_* functions introduced here can be be inline because that constant is not used when these functions are called. Also extract the flushing of the caches to be outside of the kmap. [akpm@linux-foundation.org: fix nfs and ntfs build] [akpm@linux-foundation.org: fix ntfs build some more] Signed-off-by: Christoph Lameter <clameter@sgi.com> Cc: Steven French <sfrench@us.ibm.com> Cc: Michael Halcrow <mhalcrow@us.ibm.com> Cc: <linux-ext4@vger.kernel.org> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Cc: "J. Bruce Fields" <bfields@fieldses.org> Cc: Anton Altaparmakov <aia21@cantab.net> Cc: Mark Fasheh <mark.fasheh@oracle.com> Cc: David Chinner <dgc@sgi.com> Cc: Michael Halcrow <mhalcrow@us.ibm.com> Cc: Steven French <sfrench@us.ibm.com> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-04 23:28:29 -07:00
zero_user(page, offset, length);
set_page_dirty(page);
err = 0;
unlock:
unlock_page(page);
page_cache_release(page);
out:
return err;
}
EXPORT_SYMBOL(nobh_truncate_page);
int block_truncate_page(struct address_space *mapping,
loff_t from, get_block_t *get_block)
{
pgoff_t index = from >> PAGE_CACHE_SHIFT;
unsigned offset = from & (PAGE_CACHE_SIZE-1);
unsigned blocksize;
[PATCH] fix possible PAGE_CACHE_SHIFT overflows We've had two instances recently of overflows when doing 64_bit_value = (32_bit_value << PAGE_CACHE_SHIFT) I did a tree-wide grep of `<<.*PAGE_CACHE_SHIFT' and this is the result. - afs_rxfs_fetch_descriptor.offset is of type off_t, which seems broken. - jfs and jffs are limited to 4GB anyway. - reiserfs map_block_for_writepage() takes an unsigned long for the block - it should take sector_t. (It'll fail for huge filesystems with blocksize<PAGE_CACHE_SIZE) - cramfs_read() needs to use sector_t (I think cramsfs is busted on large filesystems anyway) - affs is limited in file size anyway. - I generally didn't fix 32-bit overflows in directory operations. - arm's __flush_dcache_page() is peculiar. What if the page lies beyond 4G? - gss_wrap_req_priv() needs checking (snd_buf->page_base) Cc: Oleg Drokin <green@linuxhacker.ru> Cc: David Howells <dhowells@redhat.com> Cc: David Woodhouse <dwmw2@infradead.org> Cc: <reiserfs-dev@namesys.com> Cc: Christoph Hellwig <hch@lst.de> Cc: Anton Altaparmakov <aia21@cantab.net> Cc: Jeff Dike <jdike@addtoit.com> Cc: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it> Cc: Roman Zippel <zippel@linux-m68k.org> Cc: <linux-fsdevel@vger.kernel.org> Cc: Miklos Szeredi <miklos@szeredi.hu> Cc: Russell King <rmk@arm.linux.org.uk> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Cc: Neil Brown <neilb@cse.unsw.edu.au> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-01-08 02:03:05 -07:00
sector_t iblock;
unsigned length, pos;
struct inode *inode = mapping->host;
struct page *page;
struct buffer_head *bh;
int err;
blocksize = 1 << inode->i_blkbits;
length = offset & (blocksize - 1);
/* Block boundary? Nothing to do */
if (!length)
return 0;
length = blocksize - length;
[PATCH] fix possible PAGE_CACHE_SHIFT overflows We've had two instances recently of overflows when doing 64_bit_value = (32_bit_value << PAGE_CACHE_SHIFT) I did a tree-wide grep of `<<.*PAGE_CACHE_SHIFT' and this is the result. - afs_rxfs_fetch_descriptor.offset is of type off_t, which seems broken. - jfs and jffs are limited to 4GB anyway. - reiserfs map_block_for_writepage() takes an unsigned long for the block - it should take sector_t. (It'll fail for huge filesystems with blocksize<PAGE_CACHE_SIZE) - cramfs_read() needs to use sector_t (I think cramsfs is busted on large filesystems anyway) - affs is limited in file size anyway. - I generally didn't fix 32-bit overflows in directory operations. - arm's __flush_dcache_page() is peculiar. What if the page lies beyond 4G? - gss_wrap_req_priv() needs checking (snd_buf->page_base) Cc: Oleg Drokin <green@linuxhacker.ru> Cc: David Howells <dhowells@redhat.com> Cc: David Woodhouse <dwmw2@infradead.org> Cc: <reiserfs-dev@namesys.com> Cc: Christoph Hellwig <hch@lst.de> Cc: Anton Altaparmakov <aia21@cantab.net> Cc: Jeff Dike <jdike@addtoit.com> Cc: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it> Cc: Roman Zippel <zippel@linux-m68k.org> Cc: <linux-fsdevel@vger.kernel.org> Cc: Miklos Szeredi <miklos@szeredi.hu> Cc: Russell King <rmk@arm.linux.org.uk> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Cc: Neil Brown <neilb@cse.unsw.edu.au> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-01-08 02:03:05 -07:00
iblock = (sector_t)index << (PAGE_CACHE_SHIFT - inode->i_blkbits);
page = grab_cache_page(mapping, index);
err = -ENOMEM;
if (!page)
goto out;
if (!page_has_buffers(page))
create_empty_buffers(page, blocksize, 0);
/* Find the buffer that contains "offset" */
bh = page_buffers(page);
pos = blocksize;
while (offset >= pos) {
bh = bh->b_this_page;
iblock++;
pos += blocksize;
}
err = 0;
if (!buffer_mapped(bh)) {
WARN_ON(bh->b_size != blocksize);
err = get_block(inode, iblock, bh, 0);
if (err)
goto unlock;
/* unmapped? It's a hole - nothing to do */
if (!buffer_mapped(bh))
goto unlock;
}
/* Ok, it's mapped. Make sure it's up-to-date */
if (PageUptodate(page))
set_buffer_uptodate(bh);
if (!buffer_uptodate(bh) && !buffer_delay(bh) && !buffer_unwritten(bh)) {
err = -EIO;
ll_rw_block(READ, 1, &bh);
wait_on_buffer(bh);
/* Uhhuh. Read error. Complain and punt. */
if (!buffer_uptodate(bh))
goto unlock;
}
Pagecache zeroing: zero_user_segment, zero_user_segments and zero_user Simplify page cache zeroing of segments of pages through 3 functions zero_user_segments(page, start1, end1, start2, end2) Zeros two segments of the page. It takes the position where to start and end the zeroing which avoids length calculations and makes code clearer. zero_user_segment(page, start, end) Same for a single segment. zero_user(page, start, length) Length variant for the case where we know the length. We remove the zero_user_page macro. Issues: 1. Its a macro. Inline functions are preferable. 2. The KM_USER0 macro is only defined for HIGHMEM. Having to treat this special case everywhere makes the code needlessly complex. The parameter for zeroing is always KM_USER0 except in one single case that we open code. Avoiding KM_USER0 makes a lot of code not having to be dealing with the special casing for HIGHMEM anymore. Dealing with kmap is only necessary for HIGHMEM configurations. In those configurations we use KM_USER0 like we do for a series of other functions defined in highmem.h. Since KM_USER0 is depends on HIGHMEM the existing zero_user_page function could not be a macro. zero_user_* functions introduced here can be be inline because that constant is not used when these functions are called. Also extract the flushing of the caches to be outside of the kmap. [akpm@linux-foundation.org: fix nfs and ntfs build] [akpm@linux-foundation.org: fix ntfs build some more] Signed-off-by: Christoph Lameter <clameter@sgi.com> Cc: Steven French <sfrench@us.ibm.com> Cc: Michael Halcrow <mhalcrow@us.ibm.com> Cc: <linux-ext4@vger.kernel.org> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Cc: "J. Bruce Fields" <bfields@fieldses.org> Cc: Anton Altaparmakov <aia21@cantab.net> Cc: Mark Fasheh <mark.fasheh@oracle.com> Cc: David Chinner <dgc@sgi.com> Cc: Michael Halcrow <mhalcrow@us.ibm.com> Cc: Steven French <sfrench@us.ibm.com> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-04 23:28:29 -07:00
zero_user(page, offset, length);
mark_buffer_dirty(bh);
err = 0;
unlock:
unlock_page(page);
page_cache_release(page);
out:
return err;
}
/*
* The generic ->writepage function for buffer-backed address_spaces
*/
int block_write_full_page(struct page *page, get_block_t *get_block,
struct writeback_control *wbc)
{
struct inode * const inode = page->mapping->host;
loff_t i_size = i_size_read(inode);
const pgoff_t end_index = i_size >> PAGE_CACHE_SHIFT;
unsigned offset;
/* Is the page fully inside i_size? */
if (page->index < end_index)
return __block_write_full_page(inode, page, get_block, wbc);
/* Is the page fully outside i_size? (truncate in progress) */
offset = i_size & (PAGE_CACHE_SIZE-1);
if (page->index >= end_index+1 || !offset) {
/*
* The page may have dirty, unmapped buffers. For example,
* they may have been added in ext3_writepage(). Make them
* freeable here, so the page does not leak.
*/
do_invalidatepage(page, 0);
unlock_page(page);
return 0; /* don't care */
}
/*
* The page straddles i_size. It must be zeroed out on each and every
* writepage invokation because it may be mmapped. "A file is mapped
* in multiples of the page size. For a file that is not a multiple of
* the page size, the remaining memory is zeroed when mapped, and
* writes to that region are not written out to the file."
*/
Pagecache zeroing: zero_user_segment, zero_user_segments and zero_user Simplify page cache zeroing of segments of pages through 3 functions zero_user_segments(page, start1, end1, start2, end2) Zeros two segments of the page. It takes the position where to start and end the zeroing which avoids length calculations and makes code clearer. zero_user_segment(page, start, end) Same for a single segment. zero_user(page, start, length) Length variant for the case where we know the length. We remove the zero_user_page macro. Issues: 1. Its a macro. Inline functions are preferable. 2. The KM_USER0 macro is only defined for HIGHMEM. Having to treat this special case everywhere makes the code needlessly complex. The parameter for zeroing is always KM_USER0 except in one single case that we open code. Avoiding KM_USER0 makes a lot of code not having to be dealing with the special casing for HIGHMEM anymore. Dealing with kmap is only necessary for HIGHMEM configurations. In those configurations we use KM_USER0 like we do for a series of other functions defined in highmem.h. Since KM_USER0 is depends on HIGHMEM the existing zero_user_page function could not be a macro. zero_user_* functions introduced here can be be inline because that constant is not used when these functions are called. Also extract the flushing of the caches to be outside of the kmap. [akpm@linux-foundation.org: fix nfs and ntfs build] [akpm@linux-foundation.org: fix ntfs build some more] Signed-off-by: Christoph Lameter <clameter@sgi.com> Cc: Steven French <sfrench@us.ibm.com> Cc: Michael Halcrow <mhalcrow@us.ibm.com> Cc: <linux-ext4@vger.kernel.org> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Cc: "J. Bruce Fields" <bfields@fieldses.org> Cc: Anton Altaparmakov <aia21@cantab.net> Cc: Mark Fasheh <mark.fasheh@oracle.com> Cc: David Chinner <dgc@sgi.com> Cc: Michael Halcrow <mhalcrow@us.ibm.com> Cc: Steven French <sfrench@us.ibm.com> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-04 23:28:29 -07:00
zero_user_segment(page, offset, PAGE_CACHE_SIZE);
return __block_write_full_page(inode, page, get_block, wbc);
}
sector_t generic_block_bmap(struct address_space *mapping, sector_t block,
get_block_t *get_block)
{
struct buffer_head tmp;
struct inode *inode = mapping->host;
tmp.b_state = 0;
tmp.b_blocknr = 0;
tmp.b_size = 1 << inode->i_blkbits;
get_block(inode, block, &tmp, 0);
return tmp.b_blocknr;
}
static void end_bio_bh_io_sync(struct bio *bio, int err)
{
struct buffer_head *bh = bio->bi_private;
if (err == -EOPNOTSUPP) {
set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
set_bit(BH_Eopnotsupp, &bh->b_state);
}
block: Supress Buffer I/O errors when SCSI REQ_QUIET flag set Allow the scsi request REQ_QUIET flag to be propagated to the buffer file system layer. The basic ideas is to pass the flag from the scsi request to the bio (block IO) and then to the buffer layer. The buffer layer can then suppress needless printks. This patch declutters the kernel log by removed the 40-50 (per lun) buffer io error messages seen during a boot in my multipath setup . It is a good chance any real errors will be missed in the "noise" it the logs without this patch. During boot I see blocks of messages like " __ratelimit: 211 callbacks suppressed Buffer I/O error on device sdm, logical block 5242879 Buffer I/O error on device sdm, logical block 5242879 Buffer I/O error on device sdm, logical block 5242847 Buffer I/O error on device sdm, logical block 1 Buffer I/O error on device sdm, logical block 5242878 Buffer I/O error on device sdm, logical block 5242879 Buffer I/O error on device sdm, logical block 5242879 Buffer I/O error on device sdm, logical block 5242879 Buffer I/O error on device sdm, logical block 5242879 Buffer I/O error on device sdm, logical block 5242872 " in my logs. My disk environment is multipath fiber channel using the SCSI_DH_RDAC code and multipathd. This topology includes an "active" and "ghost" path for each lun. IO's to the "ghost" path will never complete and the SCSI layer, via the scsi device handler rdac code, quick returns the IOs to theses paths and sets the REQ_QUIET scsi flag to suppress the scsi layer messages. I am wanting to extend the QUIET behavior to include the buffer file system layer to deal with these errors as well. I have been running this patch for a while now on several boxes without issue. A few runs of bonnie++ show no noticeable difference in performance in my setup. Thanks for John Stultz for the quiet_error finalization. Submitted-by: Keith Mannthey <kmannth@us.ibm.com> Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
2008-11-25 02:24:35 -07:00
if (unlikely (test_bit(BIO_QUIET,&bio->bi_flags)))
set_bit(BH_Quiet, &bh->b_state);
bh->b_end_io(bh, test_bit(BIO_UPTODATE, &bio->bi_flags));
bio_put(bio);
}
int submit_bh(int rw, struct buffer_head * bh)
{
struct bio *bio;
int ret = 0;
BUG_ON(!buffer_locked(bh));
BUG_ON(!buffer_mapped(bh));
BUG_ON(!bh->b_end_io);
/*
* Mask in barrier bit for a write (could be either a WRITE or a
* WRITE_SYNC
*/
if (buffer_ordered(bh) && (rw & WRITE))
rw |= WRITE_BARRIER;
/*
* Only clear out a write error when rewriting
*/
if (test_set_buffer_req(bh) && (rw & WRITE))
clear_buffer_write_io_error(bh);
/*
* from here on down, it's all bio -- do the initial mapping,
* submit_bio -> generic_make_request may further map this bio around
*/
bio = bio_alloc(GFP_NOIO, 1);
bio->bi_sector = bh->b_blocknr * (bh->b_size >> 9);
bio->bi_bdev = bh->b_bdev;
bio->bi_io_vec[0].bv_page = bh->b_page;
bio->bi_io_vec[0].bv_len = bh->b_size;
bio->bi_io_vec[0].bv_offset = bh_offset(bh);
bio->bi_vcnt = 1;
bio->bi_idx = 0;
bio->bi_size = bh->b_size;
bio->bi_end_io = end_bio_bh_io_sync;
bio->bi_private = bh;
bio_get(bio);
submit_bio(rw, bio);
if (bio_flagged(bio, BIO_EOPNOTSUPP))
ret = -EOPNOTSUPP;
bio_put(bio);
return ret;
}
/**
* ll_rw_block: low-level access to block devices (DEPRECATED)
* @rw: whether to %READ or %WRITE or %SWRITE or maybe %READA (readahead)
* @nr: number of &struct buffer_heads in the array
* @bhs: array of pointers to &struct buffer_head
*
* ll_rw_block() takes an array of pointers to &struct buffer_heads, and
* requests an I/O operation on them, either a %READ or a %WRITE. The third
* %SWRITE is like %WRITE only we make sure that the *current* data in buffers
* are sent to disk. The fourth %READA option is described in the documentation
* for generic_make_request() which ll_rw_block() calls.
*
* This function drops any buffer that it cannot get a lock on (with the
* BH_Lock state bit) unless SWRITE is required, any buffer that appears to be
* clean when doing a write request, and any buffer that appears to be
* up-to-date when doing read request. Further it marks as clean buffers that
* are processed for writing (the buffer cache won't assume that they are
* actually clean until the buffer gets unlocked).
*
* ll_rw_block sets b_end_io to simple completion handler that marks
* the buffer up-to-date (if approriate), unlocks the buffer and wakes
* any waiters.
*
* All of the buffers must be for the same device, and must also be a
* multiple of the current approved size for the device.
*/
void ll_rw_block(int rw, int nr, struct buffer_head *bhs[])
{
int i;
for (i = 0; i < nr; i++) {
struct buffer_head *bh = bhs[i];
Properly notify block layer of sync writes fsync_buffers_list() and sync_dirty_buffer() both issue async writes and then immediately wait on them. Conceptually, that makes them sync writes and we should treat them as such so that the IO schedulers can handle them appropriately. This patch fixes a write starvation issue that Lin Ming reported, where xx is stuck for more than 2 minutes because of a large number of synchronous IO in the system: INFO: task kjournald:20558 blocked for more than 120 seconds. "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. kjournald D ffff810010820978 6712 20558 2 ffff81022ddb1d10 0000000000000046 ffff81022e7baa10 ffffffff803ba6f2 ffff81022ecd0000 ffff8101e6dc9160 ffff81022ecd0348 000000008048b6cb 0000000000000086 ffff81022c4e8d30 0000000000000000 ffffffff80247537 Call Trace: [<ffffffff803ba6f2>] kobject_get+0x12/0x17 [<ffffffff80247537>] getnstimeofday+0x2f/0x83 [<ffffffff8029c1ac>] sync_buffer+0x0/0x3f [<ffffffff8066d195>] io_schedule+0x5d/0x9f [<ffffffff8029c1e7>] sync_buffer+0x3b/0x3f [<ffffffff8066d3f0>] __wait_on_bit+0x40/0x6f [<ffffffff8029c1ac>] sync_buffer+0x0/0x3f [<ffffffff8066d48b>] out_of_line_wait_on_bit+0x6c/0x78 [<ffffffff80243909>] wake_bit_function+0x0/0x23 [<ffffffff8029e3ad>] sync_dirty_buffer+0x98/0xcb [<ffffffff8030056b>] journal_commit_transaction+0x97d/0xcb6 [<ffffffff8023a676>] lock_timer_base+0x26/0x4b [<ffffffff8030300a>] kjournald+0xc1/0x1fb [<ffffffff802438db>] autoremove_wake_function+0x0/0x2e [<ffffffff80302f49>] kjournald+0x0/0x1fb [<ffffffff802437bb>] kthread+0x47/0x74 [<ffffffff8022de51>] schedule_tail+0x28/0x5d [<ffffffff8020cac8>] child_rip+0xa/0x12 [<ffffffff80243774>] kthread+0x0/0x74 [<ffffffff8020cabe>] child_rip+0x0/0x12 Lin Ming confirms that this patch fixes the issue. I've run tests with it for the past week and no ill effects have been observed, so I'm proposing it for inclusion into 2.6.26. Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
2008-07-01 01:07:34 -06:00
if (rw == SWRITE || rw == SWRITE_SYNC)
lock_buffer(bh);
else if (!trylock_buffer(bh))
continue;
Properly notify block layer of sync writes fsync_buffers_list() and sync_dirty_buffer() both issue async writes and then immediately wait on them. Conceptually, that makes them sync writes and we should treat them as such so that the IO schedulers can handle them appropriately. This patch fixes a write starvation issue that Lin Ming reported, where xx is stuck for more than 2 minutes because of a large number of synchronous IO in the system: INFO: task kjournald:20558 blocked for more than 120 seconds. "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. kjournald D ffff810010820978 6712 20558 2 ffff81022ddb1d10 0000000000000046 ffff81022e7baa10 ffffffff803ba6f2 ffff81022ecd0000 ffff8101e6dc9160 ffff81022ecd0348 000000008048b6cb 0000000000000086 ffff81022c4e8d30 0000000000000000 ffffffff80247537 Call Trace: [<ffffffff803ba6f2>] kobject_get+0x12/0x17 [<ffffffff80247537>] getnstimeofday+0x2f/0x83 [<ffffffff8029c1ac>] sync_buffer+0x0/0x3f [<ffffffff8066d195>] io_schedule+0x5d/0x9f [<ffffffff8029c1e7>] sync_buffer+0x3b/0x3f [<ffffffff8066d3f0>] __wait_on_bit+0x40/0x6f [<ffffffff8029c1ac>] sync_buffer+0x0/0x3f [<ffffffff8066d48b>] out_of_line_wait_on_bit+0x6c/0x78 [<ffffffff80243909>] wake_bit_function+0x0/0x23 [<ffffffff8029e3ad>] sync_dirty_buffer+0x98/0xcb [<ffffffff8030056b>] journal_commit_transaction+0x97d/0xcb6 [<ffffffff8023a676>] lock_timer_base+0x26/0x4b [<ffffffff8030300a>] kjournald+0xc1/0x1fb [<ffffffff802438db>] autoremove_wake_function+0x0/0x2e [<ffffffff80302f49>] kjournald+0x0/0x1fb [<ffffffff802437bb>] kthread+0x47/0x74 [<ffffffff8022de51>] schedule_tail+0x28/0x5d [<ffffffff8020cac8>] child_rip+0xa/0x12 [<ffffffff80243774>] kthread+0x0/0x74 [<ffffffff8020cabe>] child_rip+0x0/0x12 Lin Ming confirms that this patch fixes the issue. I've run tests with it for the past week and no ill effects have been observed, so I'm proposing it for inclusion into 2.6.26. Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
2008-07-01 01:07:34 -06:00
if (rw == WRITE || rw == SWRITE || rw == SWRITE_SYNC) {
if (test_clear_buffer_dirty(bh)) {
bh->b_end_io = end_buffer_write_sync;
get_bh(bh);
Properly notify block layer of sync writes fsync_buffers_list() and sync_dirty_buffer() both issue async writes and then immediately wait on them. Conceptually, that makes them sync writes and we should treat them as such so that the IO schedulers can handle them appropriately. This patch fixes a write starvation issue that Lin Ming reported, where xx is stuck for more than 2 minutes because of a large number of synchronous IO in the system: INFO: task kjournald:20558 blocked for more than 120 seconds. "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. kjournald D ffff810010820978 6712 20558 2 ffff81022ddb1d10 0000000000000046 ffff81022e7baa10 ffffffff803ba6f2 ffff81022ecd0000 ffff8101e6dc9160 ffff81022ecd0348 000000008048b6cb 0000000000000086 ffff81022c4e8d30 0000000000000000 ffffffff80247537 Call Trace: [<ffffffff803ba6f2>] kobject_get+0x12/0x17 [<ffffffff80247537>] getnstimeofday+0x2f/0x83 [<ffffffff8029c1ac>] sync_buffer+0x0/0x3f [<ffffffff8066d195>] io_schedule+0x5d/0x9f [<ffffffff8029c1e7>] sync_buffer+0x3b/0x3f [<ffffffff8066d3f0>] __wait_on_bit+0x40/0x6f [<ffffffff8029c1ac>] sync_buffer+0x0/0x3f [<ffffffff8066d48b>] out_of_line_wait_on_bit+0x6c/0x78 [<ffffffff80243909>] wake_bit_function+0x0/0x23 [<ffffffff8029e3ad>] sync_dirty_buffer+0x98/0xcb [<ffffffff8030056b>] journal_commit_transaction+0x97d/0xcb6 [<ffffffff8023a676>] lock_timer_base+0x26/0x4b [<ffffffff8030300a>] kjournald+0xc1/0x1fb [<ffffffff802438db>] autoremove_wake_function+0x0/0x2e [<ffffffff80302f49>] kjournald+0x0/0x1fb [<ffffffff802437bb>] kthread+0x47/0x74 [<ffffffff8022de51>] schedule_tail+0x28/0x5d [<ffffffff8020cac8>] child_rip+0xa/0x12 [<ffffffff80243774>] kthread+0x0/0x74 [<ffffffff8020cabe>] child_rip+0x0/0x12 Lin Ming confirms that this patch fixes the issue. I've run tests with it for the past week and no ill effects have been observed, so I'm proposing it for inclusion into 2.6.26. Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
2008-07-01 01:07:34 -06:00
if (rw == SWRITE_SYNC)
submit_bh(WRITE_SYNC, bh);
else
submit_bh(WRITE, bh);
continue;
}
} else {
if (!buffer_uptodate(bh)) {
bh->b_end_io = end_buffer_read_sync;
get_bh(bh);
submit_bh(rw, bh);
continue;
}
}
unlock_buffer(bh);
}
}
/*
* For a data-integrity writeout, we need to wait upon any in-progress I/O
* and then start new I/O and then wait upon it. The caller must have a ref on
* the buffer_head.
*/
int sync_dirty_buffer(struct buffer_head *bh)
{
int ret = 0;
WARN_ON(atomic_read(&bh->b_count) < 1);
lock_buffer(bh);
if (test_clear_buffer_dirty(bh)) {
get_bh(bh);
bh->b_end_io = end_buffer_write_sync;
Properly notify block layer of sync writes fsync_buffers_list() and sync_dirty_buffer() both issue async writes and then immediately wait on them. Conceptually, that makes them sync writes and we should treat them as such so that the IO schedulers can handle them appropriately. This patch fixes a write starvation issue that Lin Ming reported, where xx is stuck for more than 2 minutes because of a large number of synchronous IO in the system: INFO: task kjournald:20558 blocked for more than 120 seconds. "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. kjournald D ffff810010820978 6712 20558 2 ffff81022ddb1d10 0000000000000046 ffff81022e7baa10 ffffffff803ba6f2 ffff81022ecd0000 ffff8101e6dc9160 ffff81022ecd0348 000000008048b6cb 0000000000000086 ffff81022c4e8d30 0000000000000000 ffffffff80247537 Call Trace: [<ffffffff803ba6f2>] kobject_get+0x12/0x17 [<ffffffff80247537>] getnstimeofday+0x2f/0x83 [<ffffffff8029c1ac>] sync_buffer+0x0/0x3f [<ffffffff8066d195>] io_schedule+0x5d/0x9f [<ffffffff8029c1e7>] sync_buffer+0x3b/0x3f [<ffffffff8066d3f0>] __wait_on_bit+0x40/0x6f [<ffffffff8029c1ac>] sync_buffer+0x0/0x3f [<ffffffff8066d48b>] out_of_line_wait_on_bit+0x6c/0x78 [<ffffffff80243909>] wake_bit_function+0x0/0x23 [<ffffffff8029e3ad>] sync_dirty_buffer+0x98/0xcb [<ffffffff8030056b>] journal_commit_transaction+0x97d/0xcb6 [<ffffffff8023a676>] lock_timer_base+0x26/0x4b [<ffffffff8030300a>] kjournald+0xc1/0x1fb [<ffffffff802438db>] autoremove_wake_function+0x0/0x2e [<ffffffff80302f49>] kjournald+0x0/0x1fb [<ffffffff802437bb>] kthread+0x47/0x74 [<ffffffff8022de51>] schedule_tail+0x28/0x5d [<ffffffff8020cac8>] child_rip+0xa/0x12 [<ffffffff80243774>] kthread+0x0/0x74 [<ffffffff8020cabe>] child_rip+0x0/0x12 Lin Ming confirms that this patch fixes the issue. I've run tests with it for the past week and no ill effects have been observed, so I'm proposing it for inclusion into 2.6.26. Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
2008-07-01 01:07:34 -06:00
ret = submit_bh(WRITE_SYNC, bh);
wait_on_buffer(bh);
if (buffer_eopnotsupp(bh)) {
clear_buffer_eopnotsupp(bh);
ret = -EOPNOTSUPP;
}
if (!ret && !buffer_uptodate(bh))
ret = -EIO;
} else {
unlock_buffer(bh);
}
return ret;
}
/*
* try_to_free_buffers() checks if all the buffers on this particular page
* are unused, and releases them if so.
*
* Exclusion against try_to_free_buffers may be obtained by either
* locking the page or by holding its mapping's private_lock.
*
* If the page is dirty but all the buffers are clean then we need to
* be sure to mark the page clean as well. This is because the page
* may be against a block device, and a later reattachment of buffers
* to a dirty page will set *all* buffers dirty. Which would corrupt
* filesystem data on the same device.
*
* The same applies to regular filesystem pages: if all the buffers are
* clean then we set the page clean and proceed. To do that, we require
* total exclusion from __set_page_dirty_buffers(). That is obtained with
* private_lock.
*
* try_to_free_buffers() is non-blocking.
*/
static inline int buffer_busy(struct buffer_head *bh)
{
return atomic_read(&bh->b_count) |
(bh->b_state & ((1 << BH_Dirty) | (1 << BH_Lock)));
}
static int
drop_buffers(struct page *page, struct buffer_head **buffers_to_free)
{
struct buffer_head *head = page_buffers(page);
struct buffer_head *bh;
bh = head;
do {
if (buffer_write_io_error(bh) && page->mapping)
set_bit(AS_EIO, &page->mapping->flags);
if (buffer_busy(bh))
goto failed;
bh = bh->b_this_page;
} while (bh != head);
do {
struct buffer_head *next = bh->b_this_page;
buffer_head: fix private_list handling There are two possible races in handling of private_list in buffer cache. 1) When fsync_buffers_list() processes a private_list, it clears b_assoc_mapping and moves buffer to its private list. Now drop_buffers() comes, sees a buffer is on list so it calls __remove_assoc_queue() which complains about b_assoc_mapping being cleared (as it cannot propagate possible IO error). This race has been actually observed in the wild. 2) When fsync_buffers_list() processes a private_list, mark_buffer_dirty_inode() can be called on bh which is already on the private list of fsync_buffers_list(). As buffer is on some list (note that the check is performed without private_lock), it is not readded to the mapping's private_list and after fsync_buffers_list() finishes, we have a dirty buffer which should be on private_list but it isn't. This race has not been reported, probably because most (but not all) callers of mark_buffer_dirty_inode() hold i_mutex and thus are serialized with fsync(). Fix these issues by not clearing b_assoc_map when fsync_buffers_list() moves buffer to a dedicated list and by reinserting buffer in private_list when it is found dirty after we have submitted buffer for IO. We also change the tests whether a buffer is on a private list from !list_empty(&bh->b_assoc_buffers) to bh->b_assoc_map so that they are single word reads and hence lockless checks are safe. Signed-off-by: Jan Kara <jack@suse.cz> Cc: Nick Piggin <npiggin@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 05:21:59 -07:00
if (bh->b_assoc_map)
__remove_assoc_queue(bh);
bh = next;
} while (bh != head);
*buffers_to_free = head;
__clear_page_buffers(page);
return 1;
failed:
return 0;
}
int try_to_free_buffers(struct page *page)
{
struct address_space * const mapping = page->mapping;
struct buffer_head *buffers_to_free = NULL;
int ret = 0;
BUG_ON(!PageLocked(page));
if (PageWriteback(page))
return 0;
if (mapping == NULL) { /* can this still happen? */
ret = drop_buffers(page, &buffers_to_free);
goto out;
}
spin_lock(&mapping->private_lock);
ret = drop_buffers(page, &buffers_to_free);
/*
* If the filesystem writes its buffers by hand (eg ext3)
* then we can have clean buffers against a dirty page. We
* clean the page here; otherwise the VM will never notice
* that the filesystem did any IO at all.
*
* Also, during truncate, discard_buffer will have marked all
* the page's buffers clean. We discover that here and clean
* the page also.
*
* private_lock must be held over this entire operation in order
* to synchronise against __set_page_dirty_buffers and prevent the
* dirty bit from being lost.
*/
if (ret)
cancel_dirty_page(page, PAGE_CACHE_SIZE);
spin_unlock(&mapping->private_lock);
out:
if (buffers_to_free) {
struct buffer_head *bh = buffers_to_free;
do {
struct buffer_head *next = bh->b_this_page;
free_buffer_head(bh);
bh = next;
} while (bh != buffers_to_free);
}
return ret;
}
EXPORT_SYMBOL(try_to_free_buffers);
void block_sync_page(struct page *page)
{
struct address_space *mapping;
smp_mb();
mapping = page_mapping(page);
if (mapping)
blk_run_backing_dev(mapping->backing_dev_info, page);
}
/*
* There are no bdflush tunables left. But distributions are
* still running obsolete flush daemons, so we terminate them here.
*
* Use of bdflush() is deprecated and will be removed in a future kernel.
* The `pdflush' kernel threads fully replace bdflush daemons and this call.
*/
asmlinkage long sys_bdflush(int func, long data)
{
static int msg_count;
if (!capable(CAP_SYS_ADMIN))
return -EPERM;
if (msg_count < 5) {
msg_count++;
printk(KERN_INFO
"warning: process `%s' used the obsolete bdflush"
" system call\n", current->comm);
printk(KERN_INFO "Fix your initscripts?\n");
}
if (func == 1)
do_exit(0);
return 0;
}
/*
* Buffer-head allocation
*/
static struct kmem_cache *bh_cachep;
/*
* Once the number of bh's in the machine exceeds this level, we start
* stripping them in writeback.
*/
static int max_buffer_heads;
int buffer_heads_over_limit;
struct bh_accounting {
int nr; /* Number of live bh's */
int ratelimit; /* Limit cacheline bouncing */
};
static DEFINE_PER_CPU(struct bh_accounting, bh_accounting) = {0, 0};
static void recalc_bh_state(void)
{
int i;
int tot = 0;
if (__get_cpu_var(bh_accounting).ratelimit++ < 4096)
return;
__get_cpu_var(bh_accounting).ratelimit = 0;
for_each_online_cpu(i)
tot += per_cpu(bh_accounting, i).nr;
buffer_heads_over_limit = (tot > max_buffer_heads);
}
struct buffer_head *alloc_buffer_head(gfp_t gfp_flags)
{
struct buffer_head *ret = kmem_cache_alloc(bh_cachep, gfp_flags);
if (ret) {
INIT_LIST_HEAD(&ret->b_assoc_buffers);
get_cpu_var(bh_accounting).nr++;
recalc_bh_state();
put_cpu_var(bh_accounting);
}
return ret;
}
EXPORT_SYMBOL(alloc_buffer_head);
void free_buffer_head(struct buffer_head *bh)
{
BUG_ON(!list_empty(&bh->b_assoc_buffers));
kmem_cache_free(bh_cachep, bh);
get_cpu_var(bh_accounting).nr--;
recalc_bh_state();
put_cpu_var(bh_accounting);
}
EXPORT_SYMBOL(free_buffer_head);
static void buffer_exit_cpu(int cpu)
{
int i;
struct bh_lru *b = &per_cpu(bh_lrus, cpu);
for (i = 0; i < BH_LRU_SIZE; i++) {
brelse(b->bhs[i]);
b->bhs[i] = NULL;
}
get_cpu_var(bh_accounting).nr += per_cpu(bh_accounting, cpu).nr;
per_cpu(bh_accounting, cpu).nr = 0;
put_cpu_var(bh_accounting);
}
static int buffer_cpu_notify(struct notifier_block *self,
unsigned long action, void *hcpu)
{
if (action == CPU_DEAD || action == CPU_DEAD_FROZEN)
buffer_exit_cpu((unsigned long)hcpu);
return NOTIFY_OK;
}
/**
fs: fix kernel-doc notation warnings Fix kernel-doc notation warnings in fs/. Warning(mmotm-2008-0314-1449//fs/super.c:560): missing initial short description on line: * mark_files_ro Warning(mmotm-2008-0314-1449//fs/locks.c:1277): missing initial short description on line: * lease_get_mtime Warning(mmotm-2008-0314-1449//fs/locks.c:1277): missing initial short description on line: * lease_get_mtime Warning(mmotm-2008-0314-1449//fs/namei.c:1368): missing initial short description on line: * lookup_one_len: filesystem helper to lookup single pathname component Warning(mmotm-2008-0314-1449//fs/buffer.c:3221): missing initial short description on line: * bh_uptodate_or_lock: Test whether the buffer is uptodate Warning(mmotm-2008-0314-1449//fs/buffer.c:3240): missing initial short description on line: * bh_submit_read: Submit a locked buffer for reading Warning(mmotm-2008-0314-1449//fs/fs-writeback.c:30): missing initial short description on line: * writeback_acquire: attempt to get exclusive writeback access to a device Warning(mmotm-2008-0314-1449//fs/fs-writeback.c:47): missing initial short description on line: * writeback_in_progress: determine whether there is writeback in progress Warning(mmotm-2008-0314-1449//fs/fs-writeback.c:58): missing initial short description on line: * writeback_release: relinquish exclusive writeback access against a device. Warning(mmotm-2008-0314-1449//include/linux/jbd.h:351): contents before sections Warning(mmotm-2008-0314-1449//include/linux/jbd.h:561): contents before sections Warning(mmotm-2008-0314-1449//fs/jbd/transaction.c:1935): missing initial short description on line: * void journal_invalidatepage() Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-03-19 18:01:00 -06:00
* bh_uptodate_or_lock - Test whether the buffer is uptodate
* @bh: struct buffer_head
*
* Return true if the buffer is up-to-date and false,
* with the buffer locked, if not.
*/
int bh_uptodate_or_lock(struct buffer_head *bh)
{
if (!buffer_uptodate(bh)) {
lock_buffer(bh);
if (!buffer_uptodate(bh))
return 0;
unlock_buffer(bh);
}
return 1;
}
EXPORT_SYMBOL(bh_uptodate_or_lock);
/**
fs: fix kernel-doc notation warnings Fix kernel-doc notation warnings in fs/. Warning(mmotm-2008-0314-1449//fs/super.c:560): missing initial short description on line: * mark_files_ro Warning(mmotm-2008-0314-1449//fs/locks.c:1277): missing initial short description on line: * lease_get_mtime Warning(mmotm-2008-0314-1449//fs/locks.c:1277): missing initial short description on line: * lease_get_mtime Warning(mmotm-2008-0314-1449//fs/namei.c:1368): missing initial short description on line: * lookup_one_len: filesystem helper to lookup single pathname component Warning(mmotm-2008-0314-1449//fs/buffer.c:3221): missing initial short description on line: * bh_uptodate_or_lock: Test whether the buffer is uptodate Warning(mmotm-2008-0314-1449//fs/buffer.c:3240): missing initial short description on line: * bh_submit_read: Submit a locked buffer for reading Warning(mmotm-2008-0314-1449//fs/fs-writeback.c:30): missing initial short description on line: * writeback_acquire: attempt to get exclusive writeback access to a device Warning(mmotm-2008-0314-1449//fs/fs-writeback.c:47): missing initial short description on line: * writeback_in_progress: determine whether there is writeback in progress Warning(mmotm-2008-0314-1449//fs/fs-writeback.c:58): missing initial short description on line: * writeback_release: relinquish exclusive writeback access against a device. Warning(mmotm-2008-0314-1449//include/linux/jbd.h:351): contents before sections Warning(mmotm-2008-0314-1449//include/linux/jbd.h:561): contents before sections Warning(mmotm-2008-0314-1449//fs/jbd/transaction.c:1935): missing initial short description on line: * void journal_invalidatepage() Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-03-19 18:01:00 -06:00
* bh_submit_read - Submit a locked buffer for reading
* @bh: struct buffer_head
*
* Returns zero on success and -EIO on error.
*/
int bh_submit_read(struct buffer_head *bh)
{
BUG_ON(!buffer_locked(bh));
if (buffer_uptodate(bh)) {
unlock_buffer(bh);
return 0;
}
get_bh(bh);
bh->b_end_io = end_buffer_read_sync;
submit_bh(READ, bh);
wait_on_buffer(bh);
if (buffer_uptodate(bh))
return 0;
return -EIO;
}
EXPORT_SYMBOL(bh_submit_read);
static void
init_buffer_head(void *data)
{
struct buffer_head *bh = data;
memset(bh, 0, sizeof(*bh));
INIT_LIST_HEAD(&bh->b_assoc_buffers);
}
void __init buffer_init(void)
{
int nrpages;
bh_cachep = kmem_cache_create("buffer_head",
sizeof(struct buffer_head), 0,
(SLAB_RECLAIM_ACCOUNT|SLAB_PANIC|
SLAB_MEM_SPREAD),
init_buffer_head);
/*
* Limit the bh occupancy to 10% of ZONE_NORMAL
*/
nrpages = (nr_free_buffer_pages() * 10) / 100;
max_buffer_heads = nrpages * (PAGE_SIZE / sizeof(struct buffer_head));
hotcpu_notifier(buffer_cpu_notify, 0);
}
EXPORT_SYMBOL(__bforget);
EXPORT_SYMBOL(__brelse);
EXPORT_SYMBOL(__wait_on_buffer);
EXPORT_SYMBOL(block_commit_write);
EXPORT_SYMBOL(block_prepare_write);
EXPORT_SYMBOL(block_page_mkwrite);
EXPORT_SYMBOL(block_read_full_page);
EXPORT_SYMBOL(block_sync_page);
EXPORT_SYMBOL(block_truncate_page);
EXPORT_SYMBOL(block_write_full_page);
EXPORT_SYMBOL(cont_write_begin);
EXPORT_SYMBOL(end_buffer_read_sync);
EXPORT_SYMBOL(end_buffer_write_sync);
EXPORT_SYMBOL(file_fsync);
EXPORT_SYMBOL(fsync_bdev);
EXPORT_SYMBOL(generic_block_bmap);
EXPORT_SYMBOL(generic_cont_expand_simple);
EXPORT_SYMBOL(init_buffer);
EXPORT_SYMBOL(invalidate_bdev);
EXPORT_SYMBOL(ll_rw_block);
EXPORT_SYMBOL(mark_buffer_dirty);
EXPORT_SYMBOL(submit_bh);
EXPORT_SYMBOL(sync_dirty_buffer);
EXPORT_SYMBOL(unlock_buffer);