linux/fs/jbd2/commit.c

1139 lines
34 KiB
C
Raw Normal View History

/*
* linux/fs/jbd2/commit.c
*
* Written by Stephen C. Tweedie <sct@redhat.com>, 1998
*
* Copyright 1998 Red Hat corp --- All Rights Reserved
*
* This file is part of the Linux kernel and is made available under
* the terms of the GNU General Public License, version 2, or at your
* option, any later version, incorporated herein by reference.
*
* Journal commit routines for the generic filesystem journaling code;
* part of the ext2fs journaling system.
*/
#include <linux/time.h>
#include <linux/fs.h>
#include <linux/jbd2.h>
#include <linux/errno.h>
#include <linux/slab.h>
#include <linux/mm.h>
#include <linux/pagemap.h>
jbd2: jbd2 stats through procfs The patch below updates the jbd stats patch to 2.6.20/jbd2. The initial patch was posted by Alex Tomas in December 2005 (http://marc.info/?l=linux-ext4&m=113538565128617&w=2). It provides statistics via procfs such as transaction lifetime and size. Sometimes, investigating performance problems, i find useful to have stats from jbd about transaction's lifetime, size, etc. here is a patch for review and inclusion probably. for example, stats after creation of 3M files in htree directory: [root@bob ~]# cat /proc/fs/jbd/sda/history R/C tid wait run lock flush log hndls block inlog ctime write drop close R 261 8260 2720 0 0 750 9892 8170 8187 C 259 750 0 4885 1 R 262 20 2200 10 0 770 9836 8170 8187 R 263 30 2200 10 0 3070 9812 8170 8187 R 264 0 5000 10 0 1340 0 0 0 C 261 8240 3212 4957 0 R 265 8260 1470 0 0 4640 9854 8170 8187 R 266 0 5000 10 0 1460 0 0 0 C 262 8210 2989 4868 0 R 267 8230 1490 10 0 4440 9875 8171 8188 R 268 0 5000 10 0 1260 0 0 0 C 263 7710 2937 4908 0 R 269 7730 1470 10 0 3330 9841 8170 8187 R 270 0 5000 10 0 830 0 0 0 C 265 8140 3234 4898 0 C 267 720 0 4849 1 R 271 8630 2740 20 0 740 9819 8170 8187 C 269 800 0 4214 1 R 272 40 2170 10 0 830 9716 8170 8187 R 273 40 2280 0 0 3530 9799 8170 8187 R 274 0 5000 10 0 990 0 0 0 where, R - line for transaction's life from T_RUNNING to T_FINISHED C - line for transaction's checkpointing tid - transaction's id wait - for how long we were waiting for new transaction to start (the longest period journal_start() took in this transaction) run - real transaction's lifetime (from T_RUNNING to T_LOCKED lock - how long we were waiting for all handles to close (time the transaction was in T_LOCKED) flush - how long it took to flush all data (data=ordered) log - how long it took to write the transaction to the log hndls - how many handles got to the transaction block - how many blocks got to the transaction inlog - how many blocks are written to the log (block + descriptors) ctime - how long it took to checkpoint the transaction write - how many blocks have been written during checkpointing drop - how many blocks have been dropped during checkpointing close - how many running transactions have been closed to checkpoint this one all times are in msec. [root@bob ~]# cat /proc/fs/jbd/sda/info 280 transaction, each upto 8192 blocks average: 1633ms waiting for transaction 3616ms running transaction 5ms transaction was being locked 1ms flushing data (in ordered mode) 1799ms logging transaction 11781 handles per transaction 5629 blocks per transaction 5641 logged blocks per transaction Signed-off-by: Johann Lombardi <johann.lombardi@bull.net> Signed-off-by: Mariusz Kozlowski <m.kozlowski@tuxland.pl> Signed-off-by: Mingming Cao <cmm@us.ibm.com> Signed-off-by: Eric Sandeen <sandeen@redhat.com>
2008-01-29 04:58:27 +00:00
#include <linux/jiffies.h>
#include <linux/crc32.h>
#include <linux/writeback.h>
#include <linux/backing-dev.h>
#include <linux/bio.h>
#include <linux/blkdev.h>
#include <linux/bitops.h>
#include <trace/events/jbd2.h>
/*
* IO end handler for temporary buffer_heads handling writes to the journal.
*/
static void journal_end_buffer_io_sync(struct buffer_head *bh, int uptodate)
{
struct buffer_head *orig_bh = bh->b_private;
BUFFER_TRACE(bh, "");
if (uptodate)
set_buffer_uptodate(bh);
else
clear_buffer_uptodate(bh);
if (orig_bh) {
clear_bit_unlock(BH_Shadow, &orig_bh->b_state);
smp_mb__after_atomic();
wake_up_bit(&orig_bh->b_state, BH_Shadow);
}
unlock_buffer(bh);
}
/*
* When an ext4 file is truncated, it is possible that some pages are not
* successfully freed, because they are attached to a committing transaction.
* After the transaction commits, these pages are left on the LRU, with no
* ->mapping, and with attached buffers. These pages are trivially reclaimable
* by the VM, but their apparent absence upsets the VM accounting, and it makes
* the numbers in /proc/meminfo look odd.
*
* So here, we have a buffer which has just come off the forget list. Look to
* see if we can strip all buffers from the backing page.
*
* Called under lock_journal(), and possibly under journal_datalist_lock. The
* caller provided us with a ref against the buffer, and we drop that here.
*/
static void release_buffer_page(struct buffer_head *bh)
{
struct page *page;
if (buffer_dirty(bh))
goto nope;
if (atomic_read(&bh->b_count) != 1)
goto nope;
page = bh->b_page;
if (!page)
goto nope;
if (page->mapping)
goto nope;
/* OK, it's a truncated page */
if (!trylock_page(page))
goto nope;
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time ago with promise that one day it will be possible to implement page cache with bigger chunks than PAGE_SIZE. This promise never materialized. And unlikely will. We have many places where PAGE_CACHE_SIZE assumed to be equal to PAGE_SIZE. And it's constant source of confusion on whether PAGE_CACHE_* or PAGE_* constant should be used in a particular case, especially on the border between fs and mm. Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much breakage to be doable. Let's stop pretending that pages in page cache are special. They are not. The changes are pretty straight-forward: - <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN}; - page_cache_get() -> get_page(); - page_cache_release() -> put_page(); This patch contains automated changes generated with coccinelle using script below. For some reason, coccinelle doesn't patch header files. I've called spatch for them manually. The only adjustment after coccinelle is revert of changes to PAGE_CAHCE_ALIGN definition: we are going to drop it later. There are few places in the code where coccinelle didn't reach. I'll fix them manually in a separate patch. Comments and documentation also will be addressed with the separate patch. virtual patch @@ expression E; @@ - E << (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ expression E; @@ - E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ @@ - PAGE_CACHE_SHIFT + PAGE_SHIFT @@ @@ - PAGE_CACHE_SIZE + PAGE_SIZE @@ @@ - PAGE_CACHE_MASK + PAGE_MASK @@ expression E; @@ - PAGE_CACHE_ALIGN(E) + PAGE_ALIGN(E) @@ expression E; @@ - page_cache_get(E) + get_page(E) @@ expression E; @@ - page_cache_release(E) + put_page(E) Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 12:29:47 +00:00
get_page(page);
__brelse(bh);
try_to_free_buffers(page);
unlock_page(page);
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time ago with promise that one day it will be possible to implement page cache with bigger chunks than PAGE_SIZE. This promise never materialized. And unlikely will. We have many places where PAGE_CACHE_SIZE assumed to be equal to PAGE_SIZE. And it's constant source of confusion on whether PAGE_CACHE_* or PAGE_* constant should be used in a particular case, especially on the border between fs and mm. Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much breakage to be doable. Let's stop pretending that pages in page cache are special. They are not. The changes are pretty straight-forward: - <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN}; - page_cache_get() -> get_page(); - page_cache_release() -> put_page(); This patch contains automated changes generated with coccinelle using script below. For some reason, coccinelle doesn't patch header files. I've called spatch for them manually. The only adjustment after coccinelle is revert of changes to PAGE_CAHCE_ALIGN definition: we are going to drop it later. There are few places in the code where coccinelle didn't reach. I'll fix them manually in a separate patch. Comments and documentation also will be addressed with the separate patch. virtual patch @@ expression E; @@ - E << (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ expression E; @@ - E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ @@ - PAGE_CACHE_SHIFT + PAGE_SHIFT @@ @@ - PAGE_CACHE_SIZE + PAGE_SIZE @@ @@ - PAGE_CACHE_MASK + PAGE_MASK @@ expression E; @@ - PAGE_CACHE_ALIGN(E) + PAGE_ALIGN(E) @@ expression E; @@ - page_cache_get(E) + get_page(E) @@ expression E; @@ - page_cache_release(E) + put_page(E) Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 12:29:47 +00:00
put_page(page);
return;
nope:
__brelse(bh);
}
static void jbd2_commit_block_csum_set(journal_t *j, struct buffer_head *bh)
{
struct commit_header *h;
__u32 csum;
if (!jbd2_journal_has_csum_v2or3(j))
return;
h = (struct commit_header *)(bh->b_data);
h->h_chksum_type = 0;
h->h_chksum_size = 0;
h->h_chksum[0] = 0;
csum = jbd2_chksum(j, j->j_csum_seed, bh->b_data, j->j_blocksize);
h->h_chksum[0] = cpu_to_be32(csum);
}
/*
* Done it all: now submit the commit record. We should have
* cleaned up our previous buffers by now, so if we are in abort
* mode we can now just skip the rest of the journal write
* entirely.
*
* Returns 1 if the journal needs to be aborted or 0 on success
*/
static int journal_submit_commit_record(journal_t *journal,
transaction_t *commit_transaction,
struct buffer_head **cbh,
__u32 crc32_sum)
{
struct commit_header *tmp;
struct buffer_head *bh;
int ret;
struct timespec64 now = current_kernel_time64();
*cbh = NULL;
if (is_journal_aborted(journal))
return 0;
bh = jbd2_journal_get_descriptor_buffer(commit_transaction,
JBD2_COMMIT_BLOCK);
if (!bh)
return 1;
tmp = (struct commit_header *)bh->b_data;
tmp->h_commit_sec = cpu_to_be64(now.tv_sec);
tmp->h_commit_nsec = cpu_to_be32(now.tv_nsec);
if (jbd2_has_feature_checksum(journal)) {
tmp->h_chksum_type = JBD2_CRC32_CHKSUM;
tmp->h_chksum_size = JBD2_CRC32_CHKSUM_SIZE;
tmp->h_chksum[0] = cpu_to_be32(crc32_sum);
}
jbd2_commit_block_csum_set(journal, bh);
BUFFER_TRACE(bh, "submit commit block");
lock_buffer(bh);
clear_buffer_dirty(bh);
set_buffer_uptodate(bh);
bh->b_end_io = journal_end_buffer_io_sync;
if (journal->j_flags & JBD2_BARRIER &&
!jbd2_has_feature_async_commit(journal))
ret = submit_bh(REQ_OP_WRITE, WRITE_SYNC | WRITE_FLUSH_FUA, bh);
else
ret = submit_bh(REQ_OP_WRITE, WRITE_SYNC, bh);
*cbh = bh;
return ret;
}
/*
* This function along with journal_submit_commit_record
* allows to write the commit record asynchronously.
*/
static int journal_wait_on_commit_record(journal_t *journal,
struct buffer_head *bh)
{
int ret = 0;
clear_buffer_dirty(bh);
wait_on_buffer(bh);
if (unlikely(!buffer_uptodate(bh)))
ret = -EIO;
put_bh(bh); /* One for getblk() */
return ret;
}
/*
* write the filemap data using writepage() address_space_operations.
* We don't do block allocation here even for delalloc. We don't
* use writepages() because with dealyed allocation we may be doing
* block allocation in writepages().
*/
static int journal_submit_inode_data_buffers(struct address_space *mapping)
{
int ret;
struct writeback_control wbc = {
.sync_mode = WB_SYNC_ALL,
.nr_to_write = mapping->nrpages * 2,
.range_start = 0,
.range_end = i_size_read(mapping->host),
};
ret = generic_writepages(mapping, &wbc);
return ret;
}
/*
* Submit all the data buffers of inode associated with the transaction to
* disk.
*
* We are in a committing transaction. Therefore no new inode can be added to
* our inode list. We use JI_COMMIT_RUNNING flag to protect inode we currently
* operate on from being released while we write out pages.
*/
static int journal_submit_data_buffers(journal_t *journal,
transaction_t *commit_transaction)
{
struct jbd2_inode *jinode;
int err, ret = 0;
struct address_space *mapping;
spin_lock(&journal->j_list_lock);
list_for_each_entry(jinode, &commit_transaction->t_inode_list, i_list) {
if (!(jinode->i_flags & JI_WRITE_DATA))
continue;
mapping = jinode->i_vfs_inode->i_mapping;
jinode->i_flags |= JI_COMMIT_RUNNING;
spin_unlock(&journal->j_list_lock);
/*
* submit the inode data buffers. We use writepage
* instead of writepages. Because writepages can do
* block allocation with delalloc. We need to write
* only allocated blocks here.
*/
trace_jbd2_submit_inode_data(jinode->i_vfs_inode);
err = journal_submit_inode_data_buffers(mapping);
if (!ret)
ret = err;
spin_lock(&journal->j_list_lock);
J_ASSERT(jinode->i_transaction == commit_transaction);
jinode->i_flags &= ~JI_COMMIT_RUNNING;
smp_mb();
wake_up_bit(&jinode->i_flags, __JI_COMMIT_RUNNING);
}
spin_unlock(&journal->j_list_lock);
return ret;
}
/*
* Wait for data submitted for writeout, refile inodes to proper
* transaction if needed.
*
*/
static int journal_finish_inode_data_buffers(journal_t *journal,
transaction_t *commit_transaction)
{
struct jbd2_inode *jinode, *next_i;
int err, ret = 0;
/* For locking, see the comment in journal_submit_data_buffers() */
spin_lock(&journal->j_list_lock);
list_for_each_entry(jinode, &commit_transaction->t_inode_list, i_list) {
if (!(jinode->i_flags & JI_WAIT_DATA))
continue;
jinode->i_flags |= JI_COMMIT_RUNNING;
spin_unlock(&journal->j_list_lock);
err = filemap_fdatawait(jinode->i_vfs_inode->i_mapping);
if (err) {
/*
* Because AS_EIO is cleared by
* filemap_fdatawait_range(), set it again so
* that user process can get -EIO from fsync().
*/
mapping_set_error(jinode->i_vfs_inode->i_mapping, -EIO);
if (!ret)
ret = err;
}
spin_lock(&journal->j_list_lock);
jinode->i_flags &= ~JI_COMMIT_RUNNING;
smp_mb();
wake_up_bit(&jinode->i_flags, __JI_COMMIT_RUNNING);
}
/* Now refile inode to proper lists */
list_for_each_entry_safe(jinode, next_i,
&commit_transaction->t_inode_list, i_list) {
list_del(&jinode->i_list);
if (jinode->i_next_transaction) {
jinode->i_transaction = jinode->i_next_transaction;
jinode->i_next_transaction = NULL;
list_add(&jinode->i_list,
&jinode->i_transaction->t_inode_list);
} else {
jinode->i_transaction = NULL;
}
}
spin_unlock(&journal->j_list_lock);
return ret;
}
static __u32 jbd2_checksum_data(__u32 crc32_sum, struct buffer_head *bh)
{
struct page *page = bh->b_page;
char *addr;
__u32 checksum;
addr = kmap_atomic(page);
checksum = crc32_be(crc32_sum,
(void *)(addr + offset_in_page(bh->b_data)), bh->b_size);
kunmap_atomic(addr);
return checksum;
}
static void write_tag_block(journal_t *j, journal_block_tag_t *tag,
unsigned long long block)
{
tag->t_blocknr = cpu_to_be32(block & (u32)~0);
if (jbd2_has_feature_64bit(j))
tag->t_blocknr_high = cpu_to_be32((block >> 31) >> 1);
}
static void jbd2_block_tag_csum_set(journal_t *j, journal_block_tag_t *tag,
struct buffer_head *bh, __u32 sequence)
{
journal_block_tag3_t *tag3 = (journal_block_tag3_t *)tag;
struct page *page = bh->b_page;
__u8 *addr;
__u32 csum32;
__be32 seq;
if (!jbd2_journal_has_csum_v2or3(j))
return;
seq = cpu_to_be32(sequence);
addr = kmap_atomic(page);
csum32 = jbd2_chksum(j, j->j_csum_seed, (__u8 *)&seq, sizeof(seq));
csum32 = jbd2_chksum(j, csum32, addr + offset_in_page(bh->b_data),
bh->b_size);
kunmap_atomic(addr);
if (jbd2_has_feature_csum3(j))
tag3->t_checksum = cpu_to_be32(csum32);
else
tag->t_checksum = cpu_to_be16(csum32);
}
/*
* jbd2_journal_commit_transaction
*
* The primary function for committing a transaction to the log. This
* function is called by the journal thread to begin a complete commit.
*/
void jbd2_journal_commit_transaction(journal_t *journal)
{
jbd2: jbd2 stats through procfs The patch below updates the jbd stats patch to 2.6.20/jbd2. The initial patch was posted by Alex Tomas in December 2005 (http://marc.info/?l=linux-ext4&m=113538565128617&w=2). It provides statistics via procfs such as transaction lifetime and size. Sometimes, investigating performance problems, i find useful to have stats from jbd about transaction's lifetime, size, etc. here is a patch for review and inclusion probably. for example, stats after creation of 3M files in htree directory: [root@bob ~]# cat /proc/fs/jbd/sda/history R/C tid wait run lock flush log hndls block inlog ctime write drop close R 261 8260 2720 0 0 750 9892 8170 8187 C 259 750 0 4885 1 R 262 20 2200 10 0 770 9836 8170 8187 R 263 30 2200 10 0 3070 9812 8170 8187 R 264 0 5000 10 0 1340 0 0 0 C 261 8240 3212 4957 0 R 265 8260 1470 0 0 4640 9854 8170 8187 R 266 0 5000 10 0 1460 0 0 0 C 262 8210 2989 4868 0 R 267 8230 1490 10 0 4440 9875 8171 8188 R 268 0 5000 10 0 1260 0 0 0 C 263 7710 2937 4908 0 R 269 7730 1470 10 0 3330 9841 8170 8187 R 270 0 5000 10 0 830 0 0 0 C 265 8140 3234 4898 0 C 267 720 0 4849 1 R 271 8630 2740 20 0 740 9819 8170 8187 C 269 800 0 4214 1 R 272 40 2170 10 0 830 9716 8170 8187 R 273 40 2280 0 0 3530 9799 8170 8187 R 274 0 5000 10 0 990 0 0 0 where, R - line for transaction's life from T_RUNNING to T_FINISHED C - line for transaction's checkpointing tid - transaction's id wait - for how long we were waiting for new transaction to start (the longest period journal_start() took in this transaction) run - real transaction's lifetime (from T_RUNNING to T_LOCKED lock - how long we were waiting for all handles to close (time the transaction was in T_LOCKED) flush - how long it took to flush all data (data=ordered) log - how long it took to write the transaction to the log hndls - how many handles got to the transaction block - how many blocks got to the transaction inlog - how many blocks are written to the log (block + descriptors) ctime - how long it took to checkpoint the transaction write - how many blocks have been written during checkpointing drop - how many blocks have been dropped during checkpointing close - how many running transactions have been closed to checkpoint this one all times are in msec. [root@bob ~]# cat /proc/fs/jbd/sda/info 280 transaction, each upto 8192 blocks average: 1633ms waiting for transaction 3616ms running transaction 5ms transaction was being locked 1ms flushing data (in ordered mode) 1799ms logging transaction 11781 handles per transaction 5629 blocks per transaction 5641 logged blocks per transaction Signed-off-by: Johann Lombardi <johann.lombardi@bull.net> Signed-off-by: Mariusz Kozlowski <m.kozlowski@tuxland.pl> Signed-off-by: Mingming Cao <cmm@us.ibm.com> Signed-off-by: Eric Sandeen <sandeen@redhat.com>
2008-01-29 04:58:27 +00:00
struct transaction_stats_s stats;
transaction_t *commit_transaction;
struct journal_head *jh;
struct buffer_head *descriptor;
struct buffer_head **wbuf = journal->j_wbuf;
int bufs;
int flags;
int err;
unsigned long long blocknr;
ktime_t start_time;
u64 commit_time;
char *tagp = NULL;
journal_block_tag_t *tag = NULL;
int space_left = 0;
int first_tag = 0;
int tag_flag;
jbd2: fix race between jbd2_journal_remove_checkpoint and ->j_commit_callback The following race is possible: [kjournald2] other_task jbd2_journal_commit_transaction() j_state = T_FINISHED; spin_unlock(&journal->j_list_lock); ->jbd2_journal_remove_checkpoint() ->jbd2_journal_free_transaction(); ->kmem_cache_free(transaction) ->j_commit_callback(journal, transaction); -> USE_AFTER_FREE WARNING: at lib/list_debug.c:62 __list_del_entry+0x1c0/0x250() Hardware name: list_del corruption. prev->next should be ffff88019a4ec198, but was 6b6b6b6b6b6b6b6b Modules linked in: cpufreq_ondemand acpi_cpufreq freq_table mperf coretemp kvm_intel kvm crc32c_intel ghash_clmulni_intel microcode sg xhci_hcd button sd_mod crc_t10dif aesni_intel ablk_helper cryptd lrw aes_x86_64 xts gf128mul ahci libahci pata_acpi ata_generic dm_mirror dm_region_hash dm_log dm_mod Pid: 16400, comm: jbd2/dm-1-8 Tainted: G W 3.8.0-rc3+ #107 Call Trace: [<ffffffff8106fb0d>] warn_slowpath_common+0xad/0xf0 [<ffffffff8106fc06>] warn_slowpath_fmt+0x46/0x50 [<ffffffff813637e9>] ? ext4_journal_commit_callback+0x99/0xc0 [<ffffffff8148cae0>] __list_del_entry+0x1c0/0x250 [<ffffffff813637bf>] ext4_journal_commit_callback+0x6f/0xc0 [<ffffffff813ca336>] jbd2_journal_commit_transaction+0x23a6/0x2570 [<ffffffff8108aa42>] ? try_to_del_timer_sync+0x82/0xa0 [<ffffffff8108b491>] ? del_timer_sync+0x91/0x1e0 [<ffffffff813d3ecf>] kjournald2+0x19f/0x6a0 [<ffffffff810ad630>] ? wake_up_bit+0x40/0x40 [<ffffffff813d3d30>] ? bit_spin_lock+0x80/0x80 [<ffffffff810ac6be>] kthread+0x10e/0x120 [<ffffffff810ac5b0>] ? __init_kthread_worker+0x70/0x70 [<ffffffff818ff6ac>] ret_from_fork+0x7c/0xb0 [<ffffffff810ac5b0>] ? __init_kthread_worker+0x70/0x70 In order to demonstrace this issue one should mount ext4 with mount -o discard option on SSD disk. This makes callback longer and race window becomes wider. In order to fix this we should mark transaction as finished only after callbacks have completed Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu> Cc: stable@vger.kernel.org
2013-04-04 02:06:52 +00:00
int i;
int tag_bytes = journal_tag_bytes(journal);
struct buffer_head *cbh = NULL; /* For transactional checksums */
__u32 crc32_sum = ~0;
struct blk_plug plug;
/* Tail of the journal */
unsigned long first_block;
tid_t first_tid;
int update_tail;
int csum_size = 0;
LIST_HEAD(io_bufs);
LIST_HEAD(log_bufs);
if (jbd2_journal_has_csum_v2or3(journal))
csum_size = sizeof(struct jbd2_journal_block_tail);
/*
* First job: lock down the current transaction and wait for
* all outstanding updates to complete.
*/
/* Do we need to erase the effects of a prior jbd2_journal_flush? */
if (journal->j_flags & JBD2_FLUSHED) {
jbd_debug(3, "super block updated\n");
mutex_lock(&journal->j_checkpoint_mutex);
/*
* We hold j_checkpoint_mutex so tail cannot change under us.
* We don't need any special data guarantees for writing sb
* since journal is empty and it is ok for write to be
* flushed only with transaction commit.
*/
jbd2_journal_update_sb_log_tail(journal,
journal->j_tail_sequence,
journal->j_tail,
WRITE_SYNC);
mutex_unlock(&journal->j_checkpoint_mutex);
} else {
jbd_debug(3, "superblock not updated\n");
}
J_ASSERT(journal->j_running_transaction != NULL);
J_ASSERT(journal->j_committing_transaction == NULL);
commit_transaction = journal->j_running_transaction;
trace_jbd2_start_commit(journal, commit_transaction);
jbd_debug(1, "JBD2: starting commit of transaction %d\n",
commit_transaction->t_tid);
write_lock(&journal->j_state_lock);
J_ASSERT(commit_transaction->t_state == T_RUNNING);
commit_transaction->t_state = T_LOCKED;
trace_jbd2_commit_locking(journal, commit_transaction);
stats.run.rs_wait = commit_transaction->t_max_wait;
stats.run.rs_request_delay = 0;
stats.run.rs_locked = jiffies;
if (commit_transaction->t_requested)
stats.run.rs_request_delay =
jbd2_time_diff(commit_transaction->t_requested,
stats.run.rs_locked);
stats.run.rs_running = jbd2_time_diff(commit_transaction->t_start,
stats.run.rs_locked);
jbd2: jbd2 stats through procfs The patch below updates the jbd stats patch to 2.6.20/jbd2. The initial patch was posted by Alex Tomas in December 2005 (http://marc.info/?l=linux-ext4&m=113538565128617&w=2). It provides statistics via procfs such as transaction lifetime and size. Sometimes, investigating performance problems, i find useful to have stats from jbd about transaction's lifetime, size, etc. here is a patch for review and inclusion probably. for example, stats after creation of 3M files in htree directory: [root@bob ~]# cat /proc/fs/jbd/sda/history R/C tid wait run lock flush log hndls block inlog ctime write drop close R 261 8260 2720 0 0 750 9892 8170 8187 C 259 750 0 4885 1 R 262 20 2200 10 0 770 9836 8170 8187 R 263 30 2200 10 0 3070 9812 8170 8187 R 264 0 5000 10 0 1340 0 0 0 C 261 8240 3212 4957 0 R 265 8260 1470 0 0 4640 9854 8170 8187 R 266 0 5000 10 0 1460 0 0 0 C 262 8210 2989 4868 0 R 267 8230 1490 10 0 4440 9875 8171 8188 R 268 0 5000 10 0 1260 0 0 0 C 263 7710 2937 4908 0 R 269 7730 1470 10 0 3330 9841 8170 8187 R 270 0 5000 10 0 830 0 0 0 C 265 8140 3234 4898 0 C 267 720 0 4849 1 R 271 8630 2740 20 0 740 9819 8170 8187 C 269 800 0 4214 1 R 272 40 2170 10 0 830 9716 8170 8187 R 273 40 2280 0 0 3530 9799 8170 8187 R 274 0 5000 10 0 990 0 0 0 where, R - line for transaction's life from T_RUNNING to T_FINISHED C - line for transaction's checkpointing tid - transaction's id wait - for how long we were waiting for new transaction to start (the longest period journal_start() took in this transaction) run - real transaction's lifetime (from T_RUNNING to T_LOCKED lock - how long we were waiting for all handles to close (time the transaction was in T_LOCKED) flush - how long it took to flush all data (data=ordered) log - how long it took to write the transaction to the log hndls - how many handles got to the transaction block - how many blocks got to the transaction inlog - how many blocks are written to the log (block + descriptors) ctime - how long it took to checkpoint the transaction write - how many blocks have been written during checkpointing drop - how many blocks have been dropped during checkpointing close - how many running transactions have been closed to checkpoint this one all times are in msec. [root@bob ~]# cat /proc/fs/jbd/sda/info 280 transaction, each upto 8192 blocks average: 1633ms waiting for transaction 3616ms running transaction 5ms transaction was being locked 1ms flushing data (in ordered mode) 1799ms logging transaction 11781 handles per transaction 5629 blocks per transaction 5641 logged blocks per transaction Signed-off-by: Johann Lombardi <johann.lombardi@bull.net> Signed-off-by: Mariusz Kozlowski <m.kozlowski@tuxland.pl> Signed-off-by: Mingming Cao <cmm@us.ibm.com> Signed-off-by: Eric Sandeen <sandeen@redhat.com>
2008-01-29 04:58:27 +00:00
spin_lock(&commit_transaction->t_handle_lock);
while (atomic_read(&commit_transaction->t_updates)) {
DEFINE_WAIT(wait);
prepare_to_wait(&journal->j_wait_updates, &wait,
TASK_UNINTERRUPTIBLE);
if (atomic_read(&commit_transaction->t_updates)) {
spin_unlock(&commit_transaction->t_handle_lock);
write_unlock(&journal->j_state_lock);
schedule();
write_lock(&journal->j_state_lock);
spin_lock(&commit_transaction->t_handle_lock);
}
finish_wait(&journal->j_wait_updates, &wait);
}
spin_unlock(&commit_transaction->t_handle_lock);
J_ASSERT (atomic_read(&commit_transaction->t_outstanding_credits) <=
journal->j_max_transaction_buffers);
/*
* First thing we are allowed to do is to discard any remaining
* BJ_Reserved buffers. Note, it is _not_ permissible to assume
* that there are no such buffers: if a large filesystem
* operation like a truncate needs to split itself over multiple
* transactions, then it may try to do a jbd2_journal_restart() while
* there are still BJ_Reserved buffers outstanding. These must
* be released cleanly from the current transaction.
*
* In this case, the filesystem must still reserve write access
* again before modifying the buffer in the new transaction, but
* we do not require it to remember exactly which old buffers it
* has reserved. This is consistent with the existing behaviour
* that multiple jbd2_journal_get_write_access() calls to the same
* buffer are perfectly permissible.
*/
while (commit_transaction->t_reserved_list) {
jh = commit_transaction->t_reserved_list;
JBUFFER_TRACE(jh, "reserved, unused: refile");
/*
* A jbd2_journal_get_undo_access()+jbd2_journal_release_buffer() may
* leave undo-committed data.
*/
if (jh->b_committed_data) {
struct buffer_head *bh = jh2bh(jh);
jbd_lock_bh_state(bh);
jbd2_free(jh->b_committed_data, bh->b_size);
jh->b_committed_data = NULL;
jbd_unlock_bh_state(bh);
}
jbd2_journal_refile_buffer(journal, jh);
}
/*
* Now try to drop any written-back buffers from the journal's
* checkpoint lists. We do this *before* commit because it potentially
* frees some memory
*/
spin_lock(&journal->j_list_lock);
__jbd2_journal_clean_checkpoint_list(journal, false);
spin_unlock(&journal->j_list_lock);
jbd_debug(3, "JBD2: commit phase 1\n");
/*
* Clear revoked flag to reflect there is no revoked buffers
* in the next transaction which is going to be started.
*/
jbd2_clear_buffer_revoked_flags(journal);
/*
* Switch to a new revoke table.
*/
jbd2_journal_switch_revoke_table(journal);
/*
* Reserved credits cannot be claimed anymore, free them
*/
atomic_sub(atomic_read(&journal->j_reserved_credits),
&commit_transaction->t_outstanding_credits);
trace_jbd2_commit_flushing(journal, commit_transaction);
stats.run.rs_flushing = jiffies;
stats.run.rs_locked = jbd2_time_diff(stats.run.rs_locked,
stats.run.rs_flushing);
jbd2: jbd2 stats through procfs The patch below updates the jbd stats patch to 2.6.20/jbd2. The initial patch was posted by Alex Tomas in December 2005 (http://marc.info/?l=linux-ext4&m=113538565128617&w=2). It provides statistics via procfs such as transaction lifetime and size. Sometimes, investigating performance problems, i find useful to have stats from jbd about transaction's lifetime, size, etc. here is a patch for review and inclusion probably. for example, stats after creation of 3M files in htree directory: [root@bob ~]# cat /proc/fs/jbd/sda/history R/C tid wait run lock flush log hndls block inlog ctime write drop close R 261 8260 2720 0 0 750 9892 8170 8187 C 259 750 0 4885 1 R 262 20 2200 10 0 770 9836 8170 8187 R 263 30 2200 10 0 3070 9812 8170 8187 R 264 0 5000 10 0 1340 0 0 0 C 261 8240 3212 4957 0 R 265 8260 1470 0 0 4640 9854 8170 8187 R 266 0 5000 10 0 1460 0 0 0 C 262 8210 2989 4868 0 R 267 8230 1490 10 0 4440 9875 8171 8188 R 268 0 5000 10 0 1260 0 0 0 C 263 7710 2937 4908 0 R 269 7730 1470 10 0 3330 9841 8170 8187 R 270 0 5000 10 0 830 0 0 0 C 265 8140 3234 4898 0 C 267 720 0 4849 1 R 271 8630 2740 20 0 740 9819 8170 8187 C 269 800 0 4214 1 R 272 40 2170 10 0 830 9716 8170 8187 R 273 40 2280 0 0 3530 9799 8170 8187 R 274 0 5000 10 0 990 0 0 0 where, R - line for transaction's life from T_RUNNING to T_FINISHED C - line for transaction's checkpointing tid - transaction's id wait - for how long we were waiting for new transaction to start (the longest period journal_start() took in this transaction) run - real transaction's lifetime (from T_RUNNING to T_LOCKED lock - how long we were waiting for all handles to close (time the transaction was in T_LOCKED) flush - how long it took to flush all data (data=ordered) log - how long it took to write the transaction to the log hndls - how many handles got to the transaction block - how many blocks got to the transaction inlog - how many blocks are written to the log (block + descriptors) ctime - how long it took to checkpoint the transaction write - how many blocks have been written during checkpointing drop - how many blocks have been dropped during checkpointing close - how many running transactions have been closed to checkpoint this one all times are in msec. [root@bob ~]# cat /proc/fs/jbd/sda/info 280 transaction, each upto 8192 blocks average: 1633ms waiting for transaction 3616ms running transaction 5ms transaction was being locked 1ms flushing data (in ordered mode) 1799ms logging transaction 11781 handles per transaction 5629 blocks per transaction 5641 logged blocks per transaction Signed-off-by: Johann Lombardi <johann.lombardi@bull.net> Signed-off-by: Mariusz Kozlowski <m.kozlowski@tuxland.pl> Signed-off-by: Mingming Cao <cmm@us.ibm.com> Signed-off-by: Eric Sandeen <sandeen@redhat.com>
2008-01-29 04:58:27 +00:00
commit_transaction->t_state = T_FLUSH;
journal->j_committing_transaction = commit_transaction;
journal->j_running_transaction = NULL;
start_time = ktime_get();
commit_transaction->t_log_start = journal->j_head;
wake_up(&journal->j_wait_transaction_locked);
write_unlock(&journal->j_state_lock);
jbd_debug(3, "JBD2: commit phase 2a\n");
/*
* Now start flushing things to disk, in the order they appear
* on the transaction lists. Data blocks go first.
*/
err = journal_submit_data_buffers(journal, commit_transaction);
if (err)
jbd2_journal_abort(journal, err);
blk_start_plug(&plug);
jbd2_journal_write_revoke_records(commit_transaction, &log_bufs);
jbd_debug(3, "JBD2: commit phase 2b\n");
/*
* Way to go: we have now written out all of the data for a
* transaction! Now comes the tricky part: we need to write out
* metadata. Loop over the transaction's entire buffer list:
*/
write_lock(&journal->j_state_lock);
commit_transaction->t_state = T_COMMIT;
write_unlock(&journal->j_state_lock);
trace_jbd2_commit_logging(journal, commit_transaction);
stats.run.rs_logging = jiffies;
stats.run.rs_flushing = jbd2_time_diff(stats.run.rs_flushing,
stats.run.rs_logging);
stats.run.rs_blocks =
atomic_read(&commit_transaction->t_outstanding_credits);
stats.run.rs_blocks_logged = 0;
jbd2: jbd2 stats through procfs The patch below updates the jbd stats patch to 2.6.20/jbd2. The initial patch was posted by Alex Tomas in December 2005 (http://marc.info/?l=linux-ext4&m=113538565128617&w=2). It provides statistics via procfs such as transaction lifetime and size. Sometimes, investigating performance problems, i find useful to have stats from jbd about transaction's lifetime, size, etc. here is a patch for review and inclusion probably. for example, stats after creation of 3M files in htree directory: [root@bob ~]# cat /proc/fs/jbd/sda/history R/C tid wait run lock flush log hndls block inlog ctime write drop close R 261 8260 2720 0 0 750 9892 8170 8187 C 259 750 0 4885 1 R 262 20 2200 10 0 770 9836 8170 8187 R 263 30 2200 10 0 3070 9812 8170 8187 R 264 0 5000 10 0 1340 0 0 0 C 261 8240 3212 4957 0 R 265 8260 1470 0 0 4640 9854 8170 8187 R 266 0 5000 10 0 1460 0 0 0 C 262 8210 2989 4868 0 R 267 8230 1490 10 0 4440 9875 8171 8188 R 268 0 5000 10 0 1260 0 0 0 C 263 7710 2937 4908 0 R 269 7730 1470 10 0 3330 9841 8170 8187 R 270 0 5000 10 0 830 0 0 0 C 265 8140 3234 4898 0 C 267 720 0 4849 1 R 271 8630 2740 20 0 740 9819 8170 8187 C 269 800 0 4214 1 R 272 40 2170 10 0 830 9716 8170 8187 R 273 40 2280 0 0 3530 9799 8170 8187 R 274 0 5000 10 0 990 0 0 0 where, R - line for transaction's life from T_RUNNING to T_FINISHED C - line for transaction's checkpointing tid - transaction's id wait - for how long we were waiting for new transaction to start (the longest period journal_start() took in this transaction) run - real transaction's lifetime (from T_RUNNING to T_LOCKED lock - how long we were waiting for all handles to close (time the transaction was in T_LOCKED) flush - how long it took to flush all data (data=ordered) log - how long it took to write the transaction to the log hndls - how many handles got to the transaction block - how many blocks got to the transaction inlog - how many blocks are written to the log (block + descriptors) ctime - how long it took to checkpoint the transaction write - how many blocks have been written during checkpointing drop - how many blocks have been dropped during checkpointing close - how many running transactions have been closed to checkpoint this one all times are in msec. [root@bob ~]# cat /proc/fs/jbd/sda/info 280 transaction, each upto 8192 blocks average: 1633ms waiting for transaction 3616ms running transaction 5ms transaction was being locked 1ms flushing data (in ordered mode) 1799ms logging transaction 11781 handles per transaction 5629 blocks per transaction 5641 logged blocks per transaction Signed-off-by: Johann Lombardi <johann.lombardi@bull.net> Signed-off-by: Mariusz Kozlowski <m.kozlowski@tuxland.pl> Signed-off-by: Mingming Cao <cmm@us.ibm.com> Signed-off-by: Eric Sandeen <sandeen@redhat.com>
2008-01-29 04:58:27 +00:00
jbd2: fix possible journal overflow issues There are several cases where the running transaction can get buffers added to its BJ_Metadata list which it never dirtied, which makes its t_nr_buffers counter end up larger than its t_outstanding_credits counter. This will cause issues when starting new transactions as while we are logging buffers we decrement t_outstanding_buffers, so when t_outstanding_buffers goes negative, we will report that we need less space in the journal than we actually need, so transactions will be started even though there may not be enough room for them. In the worst case scenario (which admittedly is almost impossible to reproduce) this will result in the journal running out of space. The fix is to only refile buffers from the committing transaction to the running transactions BJ_Modified list when b_modified is set on that journal, which is the only way to be sure if the running transaction has modified that buffer. This patch also fixes an accounting error in journal_forget, it is possible that we can call journal_forget on a buffer without having modified it, only gotten write access to it, so instead of freeing a credit, we only do so if the buffer was modified. The assert will help catch if this problem occurs. Without these two patches I could hit this assert within minutes of running postmark, with them this issue no longer arises. Cc: <linux-ext4@vger.kernel.org> Cc: Jan Kara <jack@ucw.cz> Signed-off-by: Josef Bacik <jbacik@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2008-04-17 14:38:59 +00:00
J_ASSERT(commit_transaction->t_nr_buffers <=
atomic_read(&commit_transaction->t_outstanding_credits));
jbd2: fix possible journal overflow issues There are several cases where the running transaction can get buffers added to its BJ_Metadata list which it never dirtied, which makes its t_nr_buffers counter end up larger than its t_outstanding_credits counter. This will cause issues when starting new transactions as while we are logging buffers we decrement t_outstanding_buffers, so when t_outstanding_buffers goes negative, we will report that we need less space in the journal than we actually need, so transactions will be started even though there may not be enough room for them. In the worst case scenario (which admittedly is almost impossible to reproduce) this will result in the journal running out of space. The fix is to only refile buffers from the committing transaction to the running transactions BJ_Modified list when b_modified is set on that journal, which is the only way to be sure if the running transaction has modified that buffer. This patch also fixes an accounting error in journal_forget, it is possible that we can call journal_forget on a buffer without having modified it, only gotten write access to it, so instead of freeing a credit, we only do so if the buffer was modified. The assert will help catch if this problem occurs. Without these two patches I could hit this assert within minutes of running postmark, with them this issue no longer arises. Cc: <linux-ext4@vger.kernel.org> Cc: Jan Kara <jack@ucw.cz> Signed-off-by: Josef Bacik <jbacik@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2008-04-17 14:38:59 +00:00
err = 0;
bufs = 0;
descriptor = NULL;
while (commit_transaction->t_buffers) {
/* Find the next buffer to be journaled... */
jh = commit_transaction->t_buffers;
/* If we're in abort mode, we just un-journal the buffer and
release it. */
if (is_journal_aborted(journal)) {
clear_buffer_jbddirty(jh2bh(jh));
JBUFFER_TRACE(jh, "journal is aborting: refile");
jbd2: Add buffer triggers Filesystems often to do compute intensive operation on some metadata. If this operation is repeated many times, it can be very expensive. It would be much nicer if the operation could be performed once before a buffer goes to disk. This adds triggers to jbd2 buffer heads. Just before writing a metadata buffer to the journal, jbd2 will optionally call a commit trigger associated with the buffer. If the journal is aborted, an abort trigger will be called on any dirty buffers as they are dropped from pending transactions. ocfs2 will use this feature. Initially I tried to come up with a more generic trigger that could be used for non-buffer-related events like transaction completion. It doesn't tie nicely, because the information a buffer trigger needs (specific to a journal_head) isn't the same as what a transaction trigger needs (specific to a tranaction_t or perhaps journal_t). So I implemented a buffer set, with the understanding that journal/transaction wide triggers should be implemented separately. There is only one trigger set allowed per buffer. I can't think of any reason to attach more than one set. Contrast this with a journal or transaction in which multiple places may want to watch the entire transaction separately. The trigger sets are considered static allocation from the jbd2 perspective. ocfs2 will just have one trigger set per block type, setting the same set on every bh of the same type. Signed-off-by: Joel Becker <joel.becker@oracle.com> Cc: "Theodore Ts'o" <tytso@mit.edu> Cc: <linux-ext4@vger.kernel.org> Signed-off-by: Mark Fasheh <mfasheh@suse.com>
2008-09-11 22:35:47 +00:00
jbd2_buffer_abort_trigger(jh,
jh->b_frozen_data ?
jh->b_frozen_triggers :
jh->b_triggers);
jbd2_journal_refile_buffer(journal, jh);
/* If that was the last one, we need to clean up
* any descriptor buffers which may have been
* already allocated, even if we are now
* aborting. */
if (!commit_transaction->t_buffers)
goto start_journal_io;
continue;
}
/* Make sure we have a descriptor block in which to
record the metadata buffer. */
if (!descriptor) {
J_ASSERT (bufs == 0);
jbd_debug(4, "JBD2: get descriptor\n");
descriptor = jbd2_journal_get_descriptor_buffer(
commit_transaction,
JBD2_DESCRIPTOR_BLOCK);
if (!descriptor) {
jbd2_journal_abort(journal, -EIO);
continue;
}
jbd_debug(4, "JBD2: got buffer %llu (%p)\n",
(unsigned long long)descriptor->b_blocknr,
descriptor->b_data);
tagp = &descriptor->b_data[sizeof(journal_header_t)];
space_left = descriptor->b_size -
sizeof(journal_header_t);
first_tag = 1;
set_buffer_jwrite(descriptor);
set_buffer_dirty(descriptor);
wbuf[bufs++] = descriptor;
/* Record it so that we can wait for IO
completion later */
BUFFER_TRACE(descriptor, "ph3: file as descriptor");
jbd2_file_log_bh(&log_bufs, descriptor);
}
/* Where is the buffer to be written? */
err = jbd2_journal_next_log_block(journal, &blocknr);
/* If the block mapping failed, just abandon the buffer
and repeat this loop: we'll fall into the
refile-on-abort condition above. */
if (err) {
jbd2_journal_abort(journal, err);
continue;
}
/*
* start_this_handle() uses t_outstanding_credits to determine
* the free space in the log, but this counter is changed
* by jbd2_journal_next_log_block() also.
*/
atomic_dec(&commit_transaction->t_outstanding_credits);
/* Bump b_count to prevent truncate from stumbling over
the shadowed buffer! @@@ This can go if we ever get
rid of the shadow pairing of buffers. */
atomic_inc(&jh2bh(jh)->b_count);
/*
* Make a temporary IO buffer with which to write it out
* (this will requeue the metadata buffer to BJ_Shadow).
*/
set_bit(BH_JWrite, &jh2bh(jh)->b_state);
JBUFFER_TRACE(jh, "ph3: write metadata");
flags = jbd2_journal_write_metadata_buffer(commit_transaction,
jh, &wbuf[bufs], blocknr);
if (flags < 0) {
jbd2_journal_abort(journal, flags);
continue;
}
jbd2_file_log_bh(&io_bufs, wbuf[bufs]);
/* Record the new block's tag in the current descriptor
buffer */
tag_flag = 0;
if (flags & 1)
tag_flag |= JBD2_FLAG_ESCAPE;
if (!first_tag)
tag_flag |= JBD2_FLAG_SAME_UUID;
tag = (journal_block_tag_t *) tagp;
write_tag_block(journal, tag, jh2bh(jh)->b_blocknr);
tag->t_flags = cpu_to_be16(tag_flag);
jbd2_block_tag_csum_set(journal, tag, wbuf[bufs],
commit_transaction->t_tid);
tagp += tag_bytes;
space_left -= tag_bytes;
bufs++;
if (first_tag) {
memcpy (tagp, journal->j_uuid, 16);
tagp += 16;
space_left -= 16;
first_tag = 0;
}
/* If there's no more to do, or if the descriptor is full,
let the IO rip! */
if (bufs == journal->j_wbufsize ||
commit_transaction->t_buffers == NULL ||
space_left < tag_bytes + 16 + csum_size) {
jbd_debug(4, "JBD2: Submit %d IOs\n", bufs);
/* Write an end-of-descriptor marker before
submitting the IOs. "tag" still points to
the last tag we set up. */
tag->t_flags |= cpu_to_be16(JBD2_FLAG_LAST_TAG);
jbd2_descriptor_block_csum_set(journal, descriptor);
start_journal_io:
for (i = 0; i < bufs; i++) {
struct buffer_head *bh = wbuf[i];
/*
* Compute checksum.
*/
if (jbd2_has_feature_checksum(journal)) {
crc32_sum =
jbd2_checksum_data(crc32_sum, bh);
}
lock_buffer(bh);
clear_buffer_dirty(bh);
set_buffer_uptodate(bh);
bh->b_end_io = journal_end_buffer_io_sync;
submit_bh(REQ_OP_WRITE, WRITE_SYNC, bh);
}
cond_resched();
stats.run.rs_blocks_logged += bufs;
/* Force a new descriptor to be generated next
time round the loop. */
descriptor = NULL;
bufs = 0;
}
}
err = journal_finish_inode_data_buffers(journal, commit_transaction);
if (err) {
printk(KERN_WARNING
"JBD2: Detected IO errors while flushing file data "
"on %s\n", journal->j_devname);
if (journal->j_flags & JBD2_ABORT_ON_SYNCDATA_ERR)
jbd2_journal_abort(journal, err);
err = 0;
}
/*
* Get current oldest transaction in the log before we issue flush
* to the filesystem device. After the flush we can be sure that
* blocks of all older transactions are checkpointed to persistent
* storage and we will be safe to update journal start in the
* superblock with the numbers we get here.
*/
update_tail =
jbd2_journal_get_log_tail(journal, &first_tid, &first_block);
write_lock(&journal->j_state_lock);
if (update_tail) {
long freed = first_block - journal->j_tail;
if (first_block < journal->j_tail)
freed += journal->j_last - journal->j_first;
/* Update tail only if we free significant amount of space */
if (freed < journal->j_maxlen / 4)
update_tail = 0;
}
J_ASSERT(commit_transaction->t_state == T_COMMIT);
commit_transaction->t_state = T_COMMIT_DFLUSH;
write_unlock(&journal->j_state_lock);
/*
* If the journal is not located on the file system device,
* then we must flush the file system device before we issue
* the commit record
*/
if (commit_transaction->t_need_data_flush &&
(journal->j_fs_dev != journal->j_dev) &&
(journal->j_flags & JBD2_BARRIER))
blkdev_issue_flush(journal->j_fs_dev, GFP_NOFS, NULL);
/* Done it all: now write the commit record asynchronously. */
if (jbd2_has_feature_async_commit(journal)) {
err = journal_submit_commit_record(journal, commit_transaction,
&cbh, crc32_sum);
if (err)
__jbd2_journal_abort_hard(journal);
}
blk_finish_plug(&plug);
/* Lo and behold: we have just managed to send a transaction to
the log. Before we can commit it, wait for the IO so far to
complete. Control buffers being written are on the
transaction's t_log_list queue, and metadata buffers are on
the io_bufs list.
Wait for the buffers in reverse order. That way we are
less likely to be woken up until all IOs have completed, and
so we incur less scheduling load.
*/
jbd_debug(3, "JBD2: commit phase 3\n");
while (!list_empty(&io_bufs)) {
struct buffer_head *bh = list_entry(io_bufs.prev,
struct buffer_head,
b_assoc_buffers);
wait_on_buffer(bh);
cond_resched();
if (unlikely(!buffer_uptodate(bh)))
err = -EIO;
jbd2_unfile_log_bh(bh);
/*
* The list contains temporary buffer heads created by
* jbd2_journal_write_metadata_buffer().
*/
BUFFER_TRACE(bh, "dumping temporary bh");
__brelse(bh);
J_ASSERT_BH(bh, atomic_read(&bh->b_count) == 0);
free_buffer_head(bh);
/* We also have to refile the corresponding shadowed buffer */
jh = commit_transaction->t_shadow_list->b_tprev;
bh = jh2bh(jh);
clear_buffer_jwrite(bh);
J_ASSERT_BH(bh, buffer_jbddirty(bh));
J_ASSERT_BH(bh, !buffer_shadow(bh));
/* The metadata is now released for reuse, but we need
to remember it against this transaction so that when
we finally commit, we can do any checkpointing
required. */
JBUFFER_TRACE(jh, "file as BJ_Forget");
jbd2_journal_file_buffer(jh, commit_transaction, BJ_Forget);
JBUFFER_TRACE(jh, "brelse shadowed buffer");
__brelse(bh);
}
J_ASSERT (commit_transaction->t_shadow_list == NULL);
jbd_debug(3, "JBD2: commit phase 4\n");
/* Here we wait for the revoke record and descriptor record buffers */
while (!list_empty(&log_bufs)) {
struct buffer_head *bh;
bh = list_entry(log_bufs.prev, struct buffer_head, b_assoc_buffers);
wait_on_buffer(bh);
cond_resched();
if (unlikely(!buffer_uptodate(bh)))
err = -EIO;
BUFFER_TRACE(bh, "ph5: control buffer writeout done: unfile");
clear_buffer_jwrite(bh);
jbd2_unfile_log_bh(bh);
__brelse(bh); /* One for getblk */
/* AKPM: bforget here */
}
if (err)
jbd2_journal_abort(journal, err);
jbd_debug(3, "JBD2: commit phase 5\n");
write_lock(&journal->j_state_lock);
J_ASSERT(commit_transaction->t_state == T_COMMIT_DFLUSH);
commit_transaction->t_state = T_COMMIT_JFLUSH;
write_unlock(&journal->j_state_lock);
if (!jbd2_has_feature_async_commit(journal)) {
err = journal_submit_commit_record(journal, commit_transaction,
&cbh, crc32_sum);
if (err)
__jbd2_journal_abort_hard(journal);
}
if (cbh)
err = journal_wait_on_commit_record(journal, cbh);
if (jbd2_has_feature_async_commit(journal) &&
journal->j_flags & JBD2_BARRIER) {
blkdev_issue_flush(journal->j_dev, GFP_NOFS, NULL);
}
if (err)
jbd2_journal_abort(journal, err);
/*
* Now disk caches for filesystem device are flushed so we are safe to
* erase checkpointed transactions from the log by updating journal
* superblock.
*/
if (update_tail)
jbd2_update_log_tail(journal, first_tid, first_block);
/* End of a transaction! Finally, we can do checkpoint
processing: any buffers committed as a result of this
transaction can be removed from any checkpoint list it was on
before. */
jbd_debug(3, "JBD2: commit phase 6\n");
J_ASSERT(list_empty(&commit_transaction->t_inode_list));
J_ASSERT(commit_transaction->t_buffers == NULL);
J_ASSERT(commit_transaction->t_checkpoint_list == NULL);
J_ASSERT(commit_transaction->t_shadow_list == NULL);
restart_loop:
/*
* As there are other places (journal_unmap_buffer()) adding buffers
* to this list we have to be careful and hold the j_list_lock.
*/
spin_lock(&journal->j_list_lock);
while (commit_transaction->t_forget) {
transaction_t *cp_transaction;
struct buffer_head *bh;
jbd2: Fix oops in jbd2_journal_remove_journal_head() jbd2_journal_remove_journal_head() can oops when trying to access journal_head returned by bh2jh(). This is caused for example by the following race: TASK1 TASK2 jbd2_journal_commit_transaction() ... processing t_forget list __jbd2_journal_refile_buffer(jh); if (!jh->b_transaction) { jbd_unlock_bh_state(bh); jbd2_journal_try_to_free_buffers() jbd2_journal_grab_journal_head(bh) jbd_lock_bh_state(bh) __journal_try_to_free_buffer() jbd2_journal_put_journal_head(jh) jbd2_journal_remove_journal_head(bh); jbd2_journal_put_journal_head() in TASK2 sees that b_jcount == 0 and buffer is not part of any transaction and thus frees journal_head before TASK1 gets to doing so. Note that even buffer_head can be released by try_to_free_buffers() after jbd2_journal_put_journal_head() which adds even larger opportunity for oops (but I didn't see this happen in reality). Fix the problem by making transactions hold their own journal_head reference (in b_jcount). That way we don't have to remove journal_head explicitely via jbd2_journal_remove_journal_head() and instead just remove journal_head when b_jcount drops to zero. The result of this is that [__]jbd2_journal_refile_buffer(), [__]jbd2_journal_unfile_buffer(), and __jdb2_journal_remove_checkpoint() can free journal_head which needs modification of a few callers. Also we have to be careful because once journal_head is removed, buffer_head might be freed as well. So we have to get our own buffer_head reference where it matters. Signed-off-by: Jan Kara <jack@suse.cz> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-06-13 19:38:22 +00:00
int try_to_free = 0;
jh = commit_transaction->t_forget;
spin_unlock(&journal->j_list_lock);
bh = jh2bh(jh);
jbd2: Fix oops in jbd2_journal_remove_journal_head() jbd2_journal_remove_journal_head() can oops when trying to access journal_head returned by bh2jh(). This is caused for example by the following race: TASK1 TASK2 jbd2_journal_commit_transaction() ... processing t_forget list __jbd2_journal_refile_buffer(jh); if (!jh->b_transaction) { jbd_unlock_bh_state(bh); jbd2_journal_try_to_free_buffers() jbd2_journal_grab_journal_head(bh) jbd_lock_bh_state(bh) __journal_try_to_free_buffer() jbd2_journal_put_journal_head(jh) jbd2_journal_remove_journal_head(bh); jbd2_journal_put_journal_head() in TASK2 sees that b_jcount == 0 and buffer is not part of any transaction and thus frees journal_head before TASK1 gets to doing so. Note that even buffer_head can be released by try_to_free_buffers() after jbd2_journal_put_journal_head() which adds even larger opportunity for oops (but I didn't see this happen in reality). Fix the problem by making transactions hold their own journal_head reference (in b_jcount). That way we don't have to remove journal_head explicitely via jbd2_journal_remove_journal_head() and instead just remove journal_head when b_jcount drops to zero. The result of this is that [__]jbd2_journal_refile_buffer(), [__]jbd2_journal_unfile_buffer(), and __jdb2_journal_remove_checkpoint() can free journal_head which needs modification of a few callers. Also we have to be careful because once journal_head is removed, buffer_head might be freed as well. So we have to get our own buffer_head reference where it matters. Signed-off-by: Jan Kara <jack@suse.cz> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-06-13 19:38:22 +00:00
/*
* Get a reference so that bh cannot be freed before we are
* done with it.
*/
get_bh(bh);
jbd_lock_bh_state(bh);
J_ASSERT_JH(jh, jh->b_transaction == commit_transaction);
/*
* If there is undo-protected committed data against
* this buffer, then we can remove it now. If it is a
* buffer needing such protection, the old frozen_data
* field now points to a committed version of the
* buffer, so rotate that field to the new committed
* data.
*
* Otherwise, we can just throw away the frozen data now.
jbd2: Add buffer triggers Filesystems often to do compute intensive operation on some metadata. If this operation is repeated many times, it can be very expensive. It would be much nicer if the operation could be performed once before a buffer goes to disk. This adds triggers to jbd2 buffer heads. Just before writing a metadata buffer to the journal, jbd2 will optionally call a commit trigger associated with the buffer. If the journal is aborted, an abort trigger will be called on any dirty buffers as they are dropped from pending transactions. ocfs2 will use this feature. Initially I tried to come up with a more generic trigger that could be used for non-buffer-related events like transaction completion. It doesn't tie nicely, because the information a buffer trigger needs (specific to a journal_head) isn't the same as what a transaction trigger needs (specific to a tranaction_t or perhaps journal_t). So I implemented a buffer set, with the understanding that journal/transaction wide triggers should be implemented separately. There is only one trigger set allowed per buffer. I can't think of any reason to attach more than one set. Contrast this with a journal or transaction in which multiple places may want to watch the entire transaction separately. The trigger sets are considered static allocation from the jbd2 perspective. ocfs2 will just have one trigger set per block type, setting the same set on every bh of the same type. Signed-off-by: Joel Becker <joel.becker@oracle.com> Cc: "Theodore Ts'o" <tytso@mit.edu> Cc: <linux-ext4@vger.kernel.org> Signed-off-by: Mark Fasheh <mfasheh@suse.com>
2008-09-11 22:35:47 +00:00
*
* We also know that the frozen data has already fired
* its triggers if they exist, so we can clear that too.
*/
if (jh->b_committed_data) {
jbd2_free(jh->b_committed_data, bh->b_size);
jh->b_committed_data = NULL;
if (jh->b_frozen_data) {
jh->b_committed_data = jh->b_frozen_data;
jh->b_frozen_data = NULL;
jbd2: Add buffer triggers Filesystems often to do compute intensive operation on some metadata. If this operation is repeated many times, it can be very expensive. It would be much nicer if the operation could be performed once before a buffer goes to disk. This adds triggers to jbd2 buffer heads. Just before writing a metadata buffer to the journal, jbd2 will optionally call a commit trigger associated with the buffer. If the journal is aborted, an abort trigger will be called on any dirty buffers as they are dropped from pending transactions. ocfs2 will use this feature. Initially I tried to come up with a more generic trigger that could be used for non-buffer-related events like transaction completion. It doesn't tie nicely, because the information a buffer trigger needs (specific to a journal_head) isn't the same as what a transaction trigger needs (specific to a tranaction_t or perhaps journal_t). So I implemented a buffer set, with the understanding that journal/transaction wide triggers should be implemented separately. There is only one trigger set allowed per buffer. I can't think of any reason to attach more than one set. Contrast this with a journal or transaction in which multiple places may want to watch the entire transaction separately. The trigger sets are considered static allocation from the jbd2 perspective. ocfs2 will just have one trigger set per block type, setting the same set on every bh of the same type. Signed-off-by: Joel Becker <joel.becker@oracle.com> Cc: "Theodore Ts'o" <tytso@mit.edu> Cc: <linux-ext4@vger.kernel.org> Signed-off-by: Mark Fasheh <mfasheh@suse.com>
2008-09-11 22:35:47 +00:00
jh->b_frozen_triggers = NULL;
}
} else if (jh->b_frozen_data) {
jbd2_free(jh->b_frozen_data, bh->b_size);
jh->b_frozen_data = NULL;
jbd2: Add buffer triggers Filesystems often to do compute intensive operation on some metadata. If this operation is repeated many times, it can be very expensive. It would be much nicer if the operation could be performed once before a buffer goes to disk. This adds triggers to jbd2 buffer heads. Just before writing a metadata buffer to the journal, jbd2 will optionally call a commit trigger associated with the buffer. If the journal is aborted, an abort trigger will be called on any dirty buffers as they are dropped from pending transactions. ocfs2 will use this feature. Initially I tried to come up with a more generic trigger that could be used for non-buffer-related events like transaction completion. It doesn't tie nicely, because the information a buffer trigger needs (specific to a journal_head) isn't the same as what a transaction trigger needs (specific to a tranaction_t or perhaps journal_t). So I implemented a buffer set, with the understanding that journal/transaction wide triggers should be implemented separately. There is only one trigger set allowed per buffer. I can't think of any reason to attach more than one set. Contrast this with a journal or transaction in which multiple places may want to watch the entire transaction separately. The trigger sets are considered static allocation from the jbd2 perspective. ocfs2 will just have one trigger set per block type, setting the same set on every bh of the same type. Signed-off-by: Joel Becker <joel.becker@oracle.com> Cc: "Theodore Ts'o" <tytso@mit.edu> Cc: <linux-ext4@vger.kernel.org> Signed-off-by: Mark Fasheh <mfasheh@suse.com>
2008-09-11 22:35:47 +00:00
jh->b_frozen_triggers = NULL;
}
spin_lock(&journal->j_list_lock);
cp_transaction = jh->b_cp_transaction;
if (cp_transaction) {
JBUFFER_TRACE(jh, "remove from old cp transaction");
jbd2: jbd2 stats through procfs The patch below updates the jbd stats patch to 2.6.20/jbd2. The initial patch was posted by Alex Tomas in December 2005 (http://marc.info/?l=linux-ext4&m=113538565128617&w=2). It provides statistics via procfs such as transaction lifetime and size. Sometimes, investigating performance problems, i find useful to have stats from jbd about transaction's lifetime, size, etc. here is a patch for review and inclusion probably. for example, stats after creation of 3M files in htree directory: [root@bob ~]# cat /proc/fs/jbd/sda/history R/C tid wait run lock flush log hndls block inlog ctime write drop close R 261 8260 2720 0 0 750 9892 8170 8187 C 259 750 0 4885 1 R 262 20 2200 10 0 770 9836 8170 8187 R 263 30 2200 10 0 3070 9812 8170 8187 R 264 0 5000 10 0 1340 0 0 0 C 261 8240 3212 4957 0 R 265 8260 1470 0 0 4640 9854 8170 8187 R 266 0 5000 10 0 1460 0 0 0 C 262 8210 2989 4868 0 R 267 8230 1490 10 0 4440 9875 8171 8188 R 268 0 5000 10 0 1260 0 0 0 C 263 7710 2937 4908 0 R 269 7730 1470 10 0 3330 9841 8170 8187 R 270 0 5000 10 0 830 0 0 0 C 265 8140 3234 4898 0 C 267 720 0 4849 1 R 271 8630 2740 20 0 740 9819 8170 8187 C 269 800 0 4214 1 R 272 40 2170 10 0 830 9716 8170 8187 R 273 40 2280 0 0 3530 9799 8170 8187 R 274 0 5000 10 0 990 0 0 0 where, R - line for transaction's life from T_RUNNING to T_FINISHED C - line for transaction's checkpointing tid - transaction's id wait - for how long we were waiting for new transaction to start (the longest period journal_start() took in this transaction) run - real transaction's lifetime (from T_RUNNING to T_LOCKED lock - how long we were waiting for all handles to close (time the transaction was in T_LOCKED) flush - how long it took to flush all data (data=ordered) log - how long it took to write the transaction to the log hndls - how many handles got to the transaction block - how many blocks got to the transaction inlog - how many blocks are written to the log (block + descriptors) ctime - how long it took to checkpoint the transaction write - how many blocks have been written during checkpointing drop - how many blocks have been dropped during checkpointing close - how many running transactions have been closed to checkpoint this one all times are in msec. [root@bob ~]# cat /proc/fs/jbd/sda/info 280 transaction, each upto 8192 blocks average: 1633ms waiting for transaction 3616ms running transaction 5ms transaction was being locked 1ms flushing data (in ordered mode) 1799ms logging transaction 11781 handles per transaction 5629 blocks per transaction 5641 logged blocks per transaction Signed-off-by: Johann Lombardi <johann.lombardi@bull.net> Signed-off-by: Mariusz Kozlowski <m.kozlowski@tuxland.pl> Signed-off-by: Mingming Cao <cmm@us.ibm.com> Signed-off-by: Eric Sandeen <sandeen@redhat.com>
2008-01-29 04:58:27 +00:00
cp_transaction->t_chp_stats.cs_dropped++;
__jbd2_journal_remove_checkpoint(jh);
}
/* Only re-checkpoint the buffer_head if it is marked
* dirty. If the buffer was added to the BJ_Forget list
* by jbd2_journal_forget, it may no longer be dirty and
* there's no point in keeping a checkpoint record for
* it. */
/*
* A buffer which has been freed while still being journaled by
* a previous transaction.
*/
if (buffer_freed(bh)) {
/*
* If the running transaction is the one containing
* "add to orphan" operation (b_next_transaction !=
* NULL), we have to wait for that transaction to
* commit before we can really get rid of the buffer.
* So just clear b_modified to not confuse transaction
* credit accounting and refile the buffer to
* BJ_Forget of the running transaction. If the just
* committed transaction contains "add to orphan"
* operation, we can completely invalidate the buffer
* now. We are rather through in that since the
* buffer may be still accessible when blocksize <
* pagesize and it is attached to the last partial
* page.
*/
jh->b_modified = 0;
if (!jh->b_next_transaction) {
clear_buffer_freed(bh);
clear_buffer_jbddirty(bh);
clear_buffer_mapped(bh);
clear_buffer_new(bh);
clear_buffer_req(bh);
bh->b_bdev = NULL;
}
}
if (buffer_jbddirty(bh)) {
JBUFFER_TRACE(jh, "add to new checkpointing trans");
__jbd2_journal_insert_checkpoint(jh, commit_transaction);
if (is_journal_aborted(journal))
clear_buffer_jbddirty(bh);
} else {
J_ASSERT_BH(bh, !buffer_dirty(bh));
jbd2: Fix oops in jbd2_journal_remove_journal_head() jbd2_journal_remove_journal_head() can oops when trying to access journal_head returned by bh2jh(). This is caused for example by the following race: TASK1 TASK2 jbd2_journal_commit_transaction() ... processing t_forget list __jbd2_journal_refile_buffer(jh); if (!jh->b_transaction) { jbd_unlock_bh_state(bh); jbd2_journal_try_to_free_buffers() jbd2_journal_grab_journal_head(bh) jbd_lock_bh_state(bh) __journal_try_to_free_buffer() jbd2_journal_put_journal_head(jh) jbd2_journal_remove_journal_head(bh); jbd2_journal_put_journal_head() in TASK2 sees that b_jcount == 0 and buffer is not part of any transaction and thus frees journal_head before TASK1 gets to doing so. Note that even buffer_head can be released by try_to_free_buffers() after jbd2_journal_put_journal_head() which adds even larger opportunity for oops (but I didn't see this happen in reality). Fix the problem by making transactions hold their own journal_head reference (in b_jcount). That way we don't have to remove journal_head explicitely via jbd2_journal_remove_journal_head() and instead just remove journal_head when b_jcount drops to zero. The result of this is that [__]jbd2_journal_refile_buffer(), [__]jbd2_journal_unfile_buffer(), and __jdb2_journal_remove_checkpoint() can free journal_head which needs modification of a few callers. Also we have to be careful because once journal_head is removed, buffer_head might be freed as well. So we have to get our own buffer_head reference where it matters. Signed-off-by: Jan Kara <jack@suse.cz> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-06-13 19:38:22 +00:00
/*
* The buffer on BJ_Forget list and not jbddirty means
* it has been freed by this transaction and hence it
* could not have been reallocated until this
* transaction has committed. *BUT* it could be
* reallocated once we have written all the data to
* disk and before we process the buffer on BJ_Forget
jbd2: Fix oops in jbd2_journal_remove_journal_head() jbd2_journal_remove_journal_head() can oops when trying to access journal_head returned by bh2jh(). This is caused for example by the following race: TASK1 TASK2 jbd2_journal_commit_transaction() ... processing t_forget list __jbd2_journal_refile_buffer(jh); if (!jh->b_transaction) { jbd_unlock_bh_state(bh); jbd2_journal_try_to_free_buffers() jbd2_journal_grab_journal_head(bh) jbd_lock_bh_state(bh) __journal_try_to_free_buffer() jbd2_journal_put_journal_head(jh) jbd2_journal_remove_journal_head(bh); jbd2_journal_put_journal_head() in TASK2 sees that b_jcount == 0 and buffer is not part of any transaction and thus frees journal_head before TASK1 gets to doing so. Note that even buffer_head can be released by try_to_free_buffers() after jbd2_journal_put_journal_head() which adds even larger opportunity for oops (but I didn't see this happen in reality). Fix the problem by making transactions hold their own journal_head reference (in b_jcount). That way we don't have to remove journal_head explicitely via jbd2_journal_remove_journal_head() and instead just remove journal_head when b_jcount drops to zero. The result of this is that [__]jbd2_journal_refile_buffer(), [__]jbd2_journal_unfile_buffer(), and __jdb2_journal_remove_checkpoint() can free journal_head which needs modification of a few callers. Also we have to be careful because once journal_head is removed, buffer_head might be freed as well. So we have to get our own buffer_head reference where it matters. Signed-off-by: Jan Kara <jack@suse.cz> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-06-13 19:38:22 +00:00
* list.
*/
if (!jh->b_next_transaction)
try_to_free = 1;
}
jbd2: Fix oops in jbd2_journal_remove_journal_head() jbd2_journal_remove_journal_head() can oops when trying to access journal_head returned by bh2jh(). This is caused for example by the following race: TASK1 TASK2 jbd2_journal_commit_transaction() ... processing t_forget list __jbd2_journal_refile_buffer(jh); if (!jh->b_transaction) { jbd_unlock_bh_state(bh); jbd2_journal_try_to_free_buffers() jbd2_journal_grab_journal_head(bh) jbd_lock_bh_state(bh) __journal_try_to_free_buffer() jbd2_journal_put_journal_head(jh) jbd2_journal_remove_journal_head(bh); jbd2_journal_put_journal_head() in TASK2 sees that b_jcount == 0 and buffer is not part of any transaction and thus frees journal_head before TASK1 gets to doing so. Note that even buffer_head can be released by try_to_free_buffers() after jbd2_journal_put_journal_head() which adds even larger opportunity for oops (but I didn't see this happen in reality). Fix the problem by making transactions hold their own journal_head reference (in b_jcount). That way we don't have to remove journal_head explicitely via jbd2_journal_remove_journal_head() and instead just remove journal_head when b_jcount drops to zero. The result of this is that [__]jbd2_journal_refile_buffer(), [__]jbd2_journal_unfile_buffer(), and __jdb2_journal_remove_checkpoint() can free journal_head which needs modification of a few callers. Also we have to be careful because once journal_head is removed, buffer_head might be freed as well. So we have to get our own buffer_head reference where it matters. Signed-off-by: Jan Kara <jack@suse.cz> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-06-13 19:38:22 +00:00
JBUFFER_TRACE(jh, "refile or unfile buffer");
__jbd2_journal_refile_buffer(jh);
jbd_unlock_bh_state(bh);
if (try_to_free)
release_buffer_page(bh); /* Drops bh reference */
else
__brelse(bh);
cond_resched_lock(&journal->j_list_lock);
}
spin_unlock(&journal->j_list_lock);
/*
* This is a bit sleazy. We use j_list_lock to protect transition
* of a transaction into T_FINISHED state and calling
* __jbd2_journal_drop_transaction(). Otherwise we could race with
* other checkpointing code processing the transaction...
*/
write_lock(&journal->j_state_lock);
spin_lock(&journal->j_list_lock);
/*
* Now recheck if some buffers did not get attached to the transaction
* while the lock was dropped...
*/
if (commit_transaction->t_forget) {
spin_unlock(&journal->j_list_lock);
write_unlock(&journal->j_state_lock);
goto restart_loop;
}
/* Add the transaction to the checkpoint list
* __journal_remove_checkpoint() can not destroy transaction
* under us because it is not marked as T_FINISHED yet */
if (journal->j_checkpoint_transactions == NULL) {
journal->j_checkpoint_transactions = commit_transaction;
commit_transaction->t_cpnext = commit_transaction;
commit_transaction->t_cpprev = commit_transaction;
} else {
commit_transaction->t_cpnext =
journal->j_checkpoint_transactions;
commit_transaction->t_cpprev =
commit_transaction->t_cpnext->t_cpprev;
commit_transaction->t_cpnext->t_cpprev =
commit_transaction;
commit_transaction->t_cpprev->t_cpnext =
commit_transaction;
}
spin_unlock(&journal->j_list_lock);
/* Done with this transaction! */
jbd_debug(3, "JBD2: commit phase 7\n");
J_ASSERT(commit_transaction->t_state == T_COMMIT_JFLUSH);
jbd2: jbd2 stats through procfs The patch below updates the jbd stats patch to 2.6.20/jbd2. The initial patch was posted by Alex Tomas in December 2005 (http://marc.info/?l=linux-ext4&m=113538565128617&w=2). It provides statistics via procfs such as transaction lifetime and size. Sometimes, investigating performance problems, i find useful to have stats from jbd about transaction's lifetime, size, etc. here is a patch for review and inclusion probably. for example, stats after creation of 3M files in htree directory: [root@bob ~]# cat /proc/fs/jbd/sda/history R/C tid wait run lock flush log hndls block inlog ctime write drop close R 261 8260 2720 0 0 750 9892 8170 8187 C 259 750 0 4885 1 R 262 20 2200 10 0 770 9836 8170 8187 R 263 30 2200 10 0 3070 9812 8170 8187 R 264 0 5000 10 0 1340 0 0 0 C 261 8240 3212 4957 0 R 265 8260 1470 0 0 4640 9854 8170 8187 R 266 0 5000 10 0 1460 0 0 0 C 262 8210 2989 4868 0 R 267 8230 1490 10 0 4440 9875 8171 8188 R 268 0 5000 10 0 1260 0 0 0 C 263 7710 2937 4908 0 R 269 7730 1470 10 0 3330 9841 8170 8187 R 270 0 5000 10 0 830 0 0 0 C 265 8140 3234 4898 0 C 267 720 0 4849 1 R 271 8630 2740 20 0 740 9819 8170 8187 C 269 800 0 4214 1 R 272 40 2170 10 0 830 9716 8170 8187 R 273 40 2280 0 0 3530 9799 8170 8187 R 274 0 5000 10 0 990 0 0 0 where, R - line for transaction's life from T_RUNNING to T_FINISHED C - line for transaction's checkpointing tid - transaction's id wait - for how long we were waiting for new transaction to start (the longest period journal_start() took in this transaction) run - real transaction's lifetime (from T_RUNNING to T_LOCKED lock - how long we were waiting for all handles to close (time the transaction was in T_LOCKED) flush - how long it took to flush all data (data=ordered) log - how long it took to write the transaction to the log hndls - how many handles got to the transaction block - how many blocks got to the transaction inlog - how many blocks are written to the log (block + descriptors) ctime - how long it took to checkpoint the transaction write - how many blocks have been written during checkpointing drop - how many blocks have been dropped during checkpointing close - how many running transactions have been closed to checkpoint this one all times are in msec. [root@bob ~]# cat /proc/fs/jbd/sda/info 280 transaction, each upto 8192 blocks average: 1633ms waiting for transaction 3616ms running transaction 5ms transaction was being locked 1ms flushing data (in ordered mode) 1799ms logging transaction 11781 handles per transaction 5629 blocks per transaction 5641 logged blocks per transaction Signed-off-by: Johann Lombardi <johann.lombardi@bull.net> Signed-off-by: Mariusz Kozlowski <m.kozlowski@tuxland.pl> Signed-off-by: Mingming Cao <cmm@us.ibm.com> Signed-off-by: Eric Sandeen <sandeen@redhat.com>
2008-01-29 04:58:27 +00:00
commit_transaction->t_start = jiffies;
stats.run.rs_logging = jbd2_time_diff(stats.run.rs_logging,
commit_transaction->t_start);
jbd2: jbd2 stats through procfs The patch below updates the jbd stats patch to 2.6.20/jbd2. The initial patch was posted by Alex Tomas in December 2005 (http://marc.info/?l=linux-ext4&m=113538565128617&w=2). It provides statistics via procfs such as transaction lifetime and size. Sometimes, investigating performance problems, i find useful to have stats from jbd about transaction's lifetime, size, etc. here is a patch for review and inclusion probably. for example, stats after creation of 3M files in htree directory: [root@bob ~]# cat /proc/fs/jbd/sda/history R/C tid wait run lock flush log hndls block inlog ctime write drop close R 261 8260 2720 0 0 750 9892 8170 8187 C 259 750 0 4885 1 R 262 20 2200 10 0 770 9836 8170 8187 R 263 30 2200 10 0 3070 9812 8170 8187 R 264 0 5000 10 0 1340 0 0 0 C 261 8240 3212 4957 0 R 265 8260 1470 0 0 4640 9854 8170 8187 R 266 0 5000 10 0 1460 0 0 0 C 262 8210 2989 4868 0 R 267 8230 1490 10 0 4440 9875 8171 8188 R 268 0 5000 10 0 1260 0 0 0 C 263 7710 2937 4908 0 R 269 7730 1470 10 0 3330 9841 8170 8187 R 270 0 5000 10 0 830 0 0 0 C 265 8140 3234 4898 0 C 267 720 0 4849 1 R 271 8630 2740 20 0 740 9819 8170 8187 C 269 800 0 4214 1 R 272 40 2170 10 0 830 9716 8170 8187 R 273 40 2280 0 0 3530 9799 8170 8187 R 274 0 5000 10 0 990 0 0 0 where, R - line for transaction's life from T_RUNNING to T_FINISHED C - line for transaction's checkpointing tid - transaction's id wait - for how long we were waiting for new transaction to start (the longest period journal_start() took in this transaction) run - real transaction's lifetime (from T_RUNNING to T_LOCKED lock - how long we were waiting for all handles to close (time the transaction was in T_LOCKED) flush - how long it took to flush all data (data=ordered) log - how long it took to write the transaction to the log hndls - how many handles got to the transaction block - how many blocks got to the transaction inlog - how many blocks are written to the log (block + descriptors) ctime - how long it took to checkpoint the transaction write - how many blocks have been written during checkpointing drop - how many blocks have been dropped during checkpointing close - how many running transactions have been closed to checkpoint this one all times are in msec. [root@bob ~]# cat /proc/fs/jbd/sda/info 280 transaction, each upto 8192 blocks average: 1633ms waiting for transaction 3616ms running transaction 5ms transaction was being locked 1ms flushing data (in ordered mode) 1799ms logging transaction 11781 handles per transaction 5629 blocks per transaction 5641 logged blocks per transaction Signed-off-by: Johann Lombardi <johann.lombardi@bull.net> Signed-off-by: Mariusz Kozlowski <m.kozlowski@tuxland.pl> Signed-off-by: Mingming Cao <cmm@us.ibm.com> Signed-off-by: Eric Sandeen <sandeen@redhat.com>
2008-01-29 04:58:27 +00:00
/*
* File the transaction statistics
jbd2: jbd2 stats through procfs The patch below updates the jbd stats patch to 2.6.20/jbd2. The initial patch was posted by Alex Tomas in December 2005 (http://marc.info/?l=linux-ext4&m=113538565128617&w=2). It provides statistics via procfs such as transaction lifetime and size. Sometimes, investigating performance problems, i find useful to have stats from jbd about transaction's lifetime, size, etc. here is a patch for review and inclusion probably. for example, stats after creation of 3M files in htree directory: [root@bob ~]# cat /proc/fs/jbd/sda/history R/C tid wait run lock flush log hndls block inlog ctime write drop close R 261 8260 2720 0 0 750 9892 8170 8187 C 259 750 0 4885 1 R 262 20 2200 10 0 770 9836 8170 8187 R 263 30 2200 10 0 3070 9812 8170 8187 R 264 0 5000 10 0 1340 0 0 0 C 261 8240 3212 4957 0 R 265 8260 1470 0 0 4640 9854 8170 8187 R 266 0 5000 10 0 1460 0 0 0 C 262 8210 2989 4868 0 R 267 8230 1490 10 0 4440 9875 8171 8188 R 268 0 5000 10 0 1260 0 0 0 C 263 7710 2937 4908 0 R 269 7730 1470 10 0 3330 9841 8170 8187 R 270 0 5000 10 0 830 0 0 0 C 265 8140 3234 4898 0 C 267 720 0 4849 1 R 271 8630 2740 20 0 740 9819 8170 8187 C 269 800 0 4214 1 R 272 40 2170 10 0 830 9716 8170 8187 R 273 40 2280 0 0 3530 9799 8170 8187 R 274 0 5000 10 0 990 0 0 0 where, R - line for transaction's life from T_RUNNING to T_FINISHED C - line for transaction's checkpointing tid - transaction's id wait - for how long we were waiting for new transaction to start (the longest period journal_start() took in this transaction) run - real transaction's lifetime (from T_RUNNING to T_LOCKED lock - how long we were waiting for all handles to close (time the transaction was in T_LOCKED) flush - how long it took to flush all data (data=ordered) log - how long it took to write the transaction to the log hndls - how many handles got to the transaction block - how many blocks got to the transaction inlog - how many blocks are written to the log (block + descriptors) ctime - how long it took to checkpoint the transaction write - how many blocks have been written during checkpointing drop - how many blocks have been dropped during checkpointing close - how many running transactions have been closed to checkpoint this one all times are in msec. [root@bob ~]# cat /proc/fs/jbd/sda/info 280 transaction, each upto 8192 blocks average: 1633ms waiting for transaction 3616ms running transaction 5ms transaction was being locked 1ms flushing data (in ordered mode) 1799ms logging transaction 11781 handles per transaction 5629 blocks per transaction 5641 logged blocks per transaction Signed-off-by: Johann Lombardi <johann.lombardi@bull.net> Signed-off-by: Mariusz Kozlowski <m.kozlowski@tuxland.pl> Signed-off-by: Mingming Cao <cmm@us.ibm.com> Signed-off-by: Eric Sandeen <sandeen@redhat.com>
2008-01-29 04:58:27 +00:00
*/
stats.ts_tid = commit_transaction->t_tid;
stats.run.rs_handle_count =
atomic_read(&commit_transaction->t_handle_count);
trace_jbd2_run_stats(journal->j_fs_dev->bd_dev,
commit_transaction->t_tid, &stats.run);
stats.ts_requested = (commit_transaction->t_requested) ? 1 : 0;
jbd2: jbd2 stats through procfs The patch below updates the jbd stats patch to 2.6.20/jbd2. The initial patch was posted by Alex Tomas in December 2005 (http://marc.info/?l=linux-ext4&m=113538565128617&w=2). It provides statistics via procfs such as transaction lifetime and size. Sometimes, investigating performance problems, i find useful to have stats from jbd about transaction's lifetime, size, etc. here is a patch for review and inclusion probably. for example, stats after creation of 3M files in htree directory: [root@bob ~]# cat /proc/fs/jbd/sda/history R/C tid wait run lock flush log hndls block inlog ctime write drop close R 261 8260 2720 0 0 750 9892 8170 8187 C 259 750 0 4885 1 R 262 20 2200 10 0 770 9836 8170 8187 R 263 30 2200 10 0 3070 9812 8170 8187 R 264 0 5000 10 0 1340 0 0 0 C 261 8240 3212 4957 0 R 265 8260 1470 0 0 4640 9854 8170 8187 R 266 0 5000 10 0 1460 0 0 0 C 262 8210 2989 4868 0 R 267 8230 1490 10 0 4440 9875 8171 8188 R 268 0 5000 10 0 1260 0 0 0 C 263 7710 2937 4908 0 R 269 7730 1470 10 0 3330 9841 8170 8187 R 270 0 5000 10 0 830 0 0 0 C 265 8140 3234 4898 0 C 267 720 0 4849 1 R 271 8630 2740 20 0 740 9819 8170 8187 C 269 800 0 4214 1 R 272 40 2170 10 0 830 9716 8170 8187 R 273 40 2280 0 0 3530 9799 8170 8187 R 274 0 5000 10 0 990 0 0 0 where, R - line for transaction's life from T_RUNNING to T_FINISHED C - line for transaction's checkpointing tid - transaction's id wait - for how long we were waiting for new transaction to start (the longest period journal_start() took in this transaction) run - real transaction's lifetime (from T_RUNNING to T_LOCKED lock - how long we were waiting for all handles to close (time the transaction was in T_LOCKED) flush - how long it took to flush all data (data=ordered) log - how long it took to write the transaction to the log hndls - how many handles got to the transaction block - how many blocks got to the transaction inlog - how many blocks are written to the log (block + descriptors) ctime - how long it took to checkpoint the transaction write - how many blocks have been written during checkpointing drop - how many blocks have been dropped during checkpointing close - how many running transactions have been closed to checkpoint this one all times are in msec. [root@bob ~]# cat /proc/fs/jbd/sda/info 280 transaction, each upto 8192 blocks average: 1633ms waiting for transaction 3616ms running transaction 5ms transaction was being locked 1ms flushing data (in ordered mode) 1799ms logging transaction 11781 handles per transaction 5629 blocks per transaction 5641 logged blocks per transaction Signed-off-by: Johann Lombardi <johann.lombardi@bull.net> Signed-off-by: Mariusz Kozlowski <m.kozlowski@tuxland.pl> Signed-off-by: Mingming Cao <cmm@us.ibm.com> Signed-off-by: Eric Sandeen <sandeen@redhat.com>
2008-01-29 04:58:27 +00:00
jbd2: fix race between jbd2_journal_remove_checkpoint and ->j_commit_callback The following race is possible: [kjournald2] other_task jbd2_journal_commit_transaction() j_state = T_FINISHED; spin_unlock(&journal->j_list_lock); ->jbd2_journal_remove_checkpoint() ->jbd2_journal_free_transaction(); ->kmem_cache_free(transaction) ->j_commit_callback(journal, transaction); -> USE_AFTER_FREE WARNING: at lib/list_debug.c:62 __list_del_entry+0x1c0/0x250() Hardware name: list_del corruption. prev->next should be ffff88019a4ec198, but was 6b6b6b6b6b6b6b6b Modules linked in: cpufreq_ondemand acpi_cpufreq freq_table mperf coretemp kvm_intel kvm crc32c_intel ghash_clmulni_intel microcode sg xhci_hcd button sd_mod crc_t10dif aesni_intel ablk_helper cryptd lrw aes_x86_64 xts gf128mul ahci libahci pata_acpi ata_generic dm_mirror dm_region_hash dm_log dm_mod Pid: 16400, comm: jbd2/dm-1-8 Tainted: G W 3.8.0-rc3+ #107 Call Trace: [<ffffffff8106fb0d>] warn_slowpath_common+0xad/0xf0 [<ffffffff8106fc06>] warn_slowpath_fmt+0x46/0x50 [<ffffffff813637e9>] ? ext4_journal_commit_callback+0x99/0xc0 [<ffffffff8148cae0>] __list_del_entry+0x1c0/0x250 [<ffffffff813637bf>] ext4_journal_commit_callback+0x6f/0xc0 [<ffffffff813ca336>] jbd2_journal_commit_transaction+0x23a6/0x2570 [<ffffffff8108aa42>] ? try_to_del_timer_sync+0x82/0xa0 [<ffffffff8108b491>] ? del_timer_sync+0x91/0x1e0 [<ffffffff813d3ecf>] kjournald2+0x19f/0x6a0 [<ffffffff810ad630>] ? wake_up_bit+0x40/0x40 [<ffffffff813d3d30>] ? bit_spin_lock+0x80/0x80 [<ffffffff810ac6be>] kthread+0x10e/0x120 [<ffffffff810ac5b0>] ? __init_kthread_worker+0x70/0x70 [<ffffffff818ff6ac>] ret_from_fork+0x7c/0xb0 [<ffffffff810ac5b0>] ? __init_kthread_worker+0x70/0x70 In order to demonstrace this issue one should mount ext4 with mount -o discard option on SSD disk. This makes callback longer and race window becomes wider. In order to fix this we should mark transaction as finished only after callbacks have completed Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu> Cc: stable@vger.kernel.org
2013-04-04 02:06:52 +00:00
commit_transaction->t_state = T_COMMIT_CALLBACK;
J_ASSERT(commit_transaction == journal->j_committing_transaction);
journal->j_commit_sequence = commit_transaction->t_tid;
journal->j_committing_transaction = NULL;
commit_time = ktime_to_ns(ktime_sub(ktime_get(), start_time));
/*
* weight the commit time higher than the average time so we don't
* react too strongly to vast changes in the commit time
*/
if (likely(journal->j_average_commit_time))
journal->j_average_commit_time = (commit_time +
journal->j_average_commit_time*3) / 4;
else
journal->j_average_commit_time = commit_time;
jbd2: fix race between jbd2_journal_remove_checkpoint and ->j_commit_callback The following race is possible: [kjournald2] other_task jbd2_journal_commit_transaction() j_state = T_FINISHED; spin_unlock(&journal->j_list_lock); ->jbd2_journal_remove_checkpoint() ->jbd2_journal_free_transaction(); ->kmem_cache_free(transaction) ->j_commit_callback(journal, transaction); -> USE_AFTER_FREE WARNING: at lib/list_debug.c:62 __list_del_entry+0x1c0/0x250() Hardware name: list_del corruption. prev->next should be ffff88019a4ec198, but was 6b6b6b6b6b6b6b6b Modules linked in: cpufreq_ondemand acpi_cpufreq freq_table mperf coretemp kvm_intel kvm crc32c_intel ghash_clmulni_intel microcode sg xhci_hcd button sd_mod crc_t10dif aesni_intel ablk_helper cryptd lrw aes_x86_64 xts gf128mul ahci libahci pata_acpi ata_generic dm_mirror dm_region_hash dm_log dm_mod Pid: 16400, comm: jbd2/dm-1-8 Tainted: G W 3.8.0-rc3+ #107 Call Trace: [<ffffffff8106fb0d>] warn_slowpath_common+0xad/0xf0 [<ffffffff8106fc06>] warn_slowpath_fmt+0x46/0x50 [<ffffffff813637e9>] ? ext4_journal_commit_callback+0x99/0xc0 [<ffffffff8148cae0>] __list_del_entry+0x1c0/0x250 [<ffffffff813637bf>] ext4_journal_commit_callback+0x6f/0xc0 [<ffffffff813ca336>] jbd2_journal_commit_transaction+0x23a6/0x2570 [<ffffffff8108aa42>] ? try_to_del_timer_sync+0x82/0xa0 [<ffffffff8108b491>] ? del_timer_sync+0x91/0x1e0 [<ffffffff813d3ecf>] kjournald2+0x19f/0x6a0 [<ffffffff810ad630>] ? wake_up_bit+0x40/0x40 [<ffffffff813d3d30>] ? bit_spin_lock+0x80/0x80 [<ffffffff810ac6be>] kthread+0x10e/0x120 [<ffffffff810ac5b0>] ? __init_kthread_worker+0x70/0x70 [<ffffffff818ff6ac>] ret_from_fork+0x7c/0xb0 [<ffffffff810ac5b0>] ? __init_kthread_worker+0x70/0x70 In order to demonstrace this issue one should mount ext4 with mount -o discard option on SSD disk. This makes callback longer and race window becomes wider. In order to fix this we should mark transaction as finished only after callbacks have completed Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu> Cc: stable@vger.kernel.org
2013-04-04 02:06:52 +00:00
write_unlock(&journal->j_state_lock);
if (journal->j_commit_callback)
journal->j_commit_callback(journal, commit_transaction);
trace_jbd2_end_commit(journal, commit_transaction);
jbd_debug(1, "JBD2: commit %d complete, head %d\n",
journal->j_commit_sequence, journal->j_tail_sequence);
jbd2: fix race between jbd2_journal_remove_checkpoint and ->j_commit_callback The following race is possible: [kjournald2] other_task jbd2_journal_commit_transaction() j_state = T_FINISHED; spin_unlock(&journal->j_list_lock); ->jbd2_journal_remove_checkpoint() ->jbd2_journal_free_transaction(); ->kmem_cache_free(transaction) ->j_commit_callback(journal, transaction); -> USE_AFTER_FREE WARNING: at lib/list_debug.c:62 __list_del_entry+0x1c0/0x250() Hardware name: list_del corruption. prev->next should be ffff88019a4ec198, but was 6b6b6b6b6b6b6b6b Modules linked in: cpufreq_ondemand acpi_cpufreq freq_table mperf coretemp kvm_intel kvm crc32c_intel ghash_clmulni_intel microcode sg xhci_hcd button sd_mod crc_t10dif aesni_intel ablk_helper cryptd lrw aes_x86_64 xts gf128mul ahci libahci pata_acpi ata_generic dm_mirror dm_region_hash dm_log dm_mod Pid: 16400, comm: jbd2/dm-1-8 Tainted: G W 3.8.0-rc3+ #107 Call Trace: [<ffffffff8106fb0d>] warn_slowpath_common+0xad/0xf0 [<ffffffff8106fc06>] warn_slowpath_fmt+0x46/0x50 [<ffffffff813637e9>] ? ext4_journal_commit_callback+0x99/0xc0 [<ffffffff8148cae0>] __list_del_entry+0x1c0/0x250 [<ffffffff813637bf>] ext4_journal_commit_callback+0x6f/0xc0 [<ffffffff813ca336>] jbd2_journal_commit_transaction+0x23a6/0x2570 [<ffffffff8108aa42>] ? try_to_del_timer_sync+0x82/0xa0 [<ffffffff8108b491>] ? del_timer_sync+0x91/0x1e0 [<ffffffff813d3ecf>] kjournald2+0x19f/0x6a0 [<ffffffff810ad630>] ? wake_up_bit+0x40/0x40 [<ffffffff813d3d30>] ? bit_spin_lock+0x80/0x80 [<ffffffff810ac6be>] kthread+0x10e/0x120 [<ffffffff810ac5b0>] ? __init_kthread_worker+0x70/0x70 [<ffffffff818ff6ac>] ret_from_fork+0x7c/0xb0 [<ffffffff810ac5b0>] ? __init_kthread_worker+0x70/0x70 In order to demonstrace this issue one should mount ext4 with mount -o discard option on SSD disk. This makes callback longer and race window becomes wider. In order to fix this we should mark transaction as finished only after callbacks have completed Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu> Cc: stable@vger.kernel.org
2013-04-04 02:06:52 +00:00
write_lock(&journal->j_state_lock);
spin_lock(&journal->j_list_lock);
commit_transaction->t_state = T_FINISHED;
/* Check if the transaction can be dropped now that we are finished */
jbd2: fix race between jbd2_journal_remove_checkpoint and ->j_commit_callback The following race is possible: [kjournald2] other_task jbd2_journal_commit_transaction() j_state = T_FINISHED; spin_unlock(&journal->j_list_lock); ->jbd2_journal_remove_checkpoint() ->jbd2_journal_free_transaction(); ->kmem_cache_free(transaction) ->j_commit_callback(journal, transaction); -> USE_AFTER_FREE WARNING: at lib/list_debug.c:62 __list_del_entry+0x1c0/0x250() Hardware name: list_del corruption. prev->next should be ffff88019a4ec198, but was 6b6b6b6b6b6b6b6b Modules linked in: cpufreq_ondemand acpi_cpufreq freq_table mperf coretemp kvm_intel kvm crc32c_intel ghash_clmulni_intel microcode sg xhci_hcd button sd_mod crc_t10dif aesni_intel ablk_helper cryptd lrw aes_x86_64 xts gf128mul ahci libahci pata_acpi ata_generic dm_mirror dm_region_hash dm_log dm_mod Pid: 16400, comm: jbd2/dm-1-8 Tainted: G W 3.8.0-rc3+ #107 Call Trace: [<ffffffff8106fb0d>] warn_slowpath_common+0xad/0xf0 [<ffffffff8106fc06>] warn_slowpath_fmt+0x46/0x50 [<ffffffff813637e9>] ? ext4_journal_commit_callback+0x99/0xc0 [<ffffffff8148cae0>] __list_del_entry+0x1c0/0x250 [<ffffffff813637bf>] ext4_journal_commit_callback+0x6f/0xc0 [<ffffffff813ca336>] jbd2_journal_commit_transaction+0x23a6/0x2570 [<ffffffff8108aa42>] ? try_to_del_timer_sync+0x82/0xa0 [<ffffffff8108b491>] ? del_timer_sync+0x91/0x1e0 [<ffffffff813d3ecf>] kjournald2+0x19f/0x6a0 [<ffffffff810ad630>] ? wake_up_bit+0x40/0x40 [<ffffffff813d3d30>] ? bit_spin_lock+0x80/0x80 [<ffffffff810ac6be>] kthread+0x10e/0x120 [<ffffffff810ac5b0>] ? __init_kthread_worker+0x70/0x70 [<ffffffff818ff6ac>] ret_from_fork+0x7c/0xb0 [<ffffffff810ac5b0>] ? __init_kthread_worker+0x70/0x70 In order to demonstrace this issue one should mount ext4 with mount -o discard option on SSD disk. This makes callback longer and race window becomes wider. In order to fix this we should mark transaction as finished only after callbacks have completed Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu> Cc: stable@vger.kernel.org
2013-04-04 02:06:52 +00:00
if (commit_transaction->t_checkpoint_list == NULL &&
commit_transaction->t_checkpoint_io_list == NULL) {
__jbd2_journal_drop_transaction(journal, commit_transaction);
jbd2_journal_free_transaction(commit_transaction);
}
spin_unlock(&journal->j_list_lock);
write_unlock(&journal->j_state_lock);
wake_up(&journal->j_wait_done_commit);
/*
* Calculate overall stats
*/
spin_lock(&journal->j_history_lock);
journal->j_stats.ts_tid++;
journal->j_stats.ts_requested += stats.ts_requested;
journal->j_stats.run.rs_wait += stats.run.rs_wait;
journal->j_stats.run.rs_request_delay += stats.run.rs_request_delay;
journal->j_stats.run.rs_running += stats.run.rs_running;
journal->j_stats.run.rs_locked += stats.run.rs_locked;
journal->j_stats.run.rs_flushing += stats.run.rs_flushing;
journal->j_stats.run.rs_logging += stats.run.rs_logging;
journal->j_stats.run.rs_handle_count += stats.run.rs_handle_count;
journal->j_stats.run.rs_blocks += stats.run.rs_blocks;
journal->j_stats.run.rs_blocks_logged += stats.run.rs_blocks_logged;
spin_unlock(&journal->j_history_lock);
}