From 1449032be17abb69116dbc393f67ceb8bd034f92 Mon Sep 17 00:00:00 2001 From: Theodore Ts'o Date: Tue, 14 Dec 2010 15:27:50 -0500 Subject: [PATCH 1/2] ext4: Turn off multiple page-io submission by default Jon Nelson has found a test case which causes postgresql to fail with the error: psql:t.sql:4: ERROR: invalid page header in block 38269 of relation base/16384/16581 Under memory pressure, it looks like part of a file can end up getting replaced by zero's. Until we can figure out the cause, we'll roll back the change and use block_write_full_page() instead of ext4_bio_write_page(). The new, more efficient writing function can be used via the mount option mblk_io_submit, so we can test and fix the new page I/O code. To reproduce the problem, install postgres 8.4 or 9.0, and pin enough memory such that the system just at the end of triggering writeback before running the following sql script: begin; create temporary table foo as select x as a, ARRAY[x] as b FROM generate_series(1, 10000000 ) AS x; create index foo_a_idx on foo (a); create index foo_b_idx on foo USING GIN (b); rollback; If the temporary table is created on a hard drive partition which is encrypted using dm_crypt, then under memory pressure, approximately 30-40% of the time, pgsql will issue the above failure. This patch should fix this problem, and the problem will come back if the file system is mounted with the mblk_io_submit mount option. Reported-by: Jon Nelson Signed-off-by: "Theodore Ts'o" --- fs/ext4/ext4.h | 1 + fs/ext4/inode.c | 5 ++++- fs/ext4/super.c | 14 ++++++++++++-- 3 files changed, 17 insertions(+), 3 deletions(-) diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h index 6a5edea2d70b..94ce3d7a1c4b 100644 --- a/fs/ext4/ext4.h +++ b/fs/ext4/ext4.h @@ -910,6 +910,7 @@ struct ext4_inode_info { #define EXT4_MOUNT_JOURNAL_CHECKSUM 0x800000 /* Journal checksums */ #define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */ #define EXT4_MOUNT_I_VERSION 0x2000000 /* i_version support */ +#define EXT4_MOUNT_MBLK_IO_SUBMIT 0x4000000 /* multi-block io submits */ #define EXT4_MOUNT_DELALLOC 0x8000000 /* Delalloc support */ #define EXT4_MOUNT_DATA_ERR_ABORT 0x10000000 /* Abort on file data write */ #define EXT4_MOUNT_BLOCK_VALIDITY 0x20000000 /* Block validity checking */ diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c index bdbe69902207..e659597b690b 100644 --- a/fs/ext4/inode.c +++ b/fs/ext4/inode.c @@ -2125,9 +2125,12 @@ static int mpage_da_submit_io(struct mpage_da_data *mpd, */ if (unlikely(journal_data && PageChecked(page))) err = __ext4_journalled_writepage(page, len); - else + else if (test_opt(inode->i_sb, MBLK_IO_SUBMIT)) err = ext4_bio_write_page(&io_submit, page, len, mpd->wbc); + else + err = block_write_full_page(page, + noalloc_get_block_write, mpd->wbc); if (!err) mpd->pages_written++; diff --git a/fs/ext4/super.c b/fs/ext4/super.c index e32195d6aac3..fb15c9c0be74 100644 --- a/fs/ext4/super.c +++ b/fs/ext4/super.c @@ -1026,6 +1026,8 @@ static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs) !(def_mount_opts & EXT4_DEFM_NODELALLOC)) seq_puts(seq, ",nodelalloc"); + if (test_opt(sb, MBLK_IO_SUBMIT)) + seq_puts(seq, ",mblk_io_submit"); if (sbi->s_stripe) seq_printf(seq, ",stripe=%lu", sbi->s_stripe); /* @@ -1239,8 +1241,8 @@ enum { Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota, Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version, - Opt_stripe, Opt_delalloc, Opt_nodelalloc, - Opt_block_validity, Opt_noblock_validity, + Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit, + Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity, Opt_inode_readahead_blks, Opt_journal_ioprio, Opt_dioread_nolock, Opt_dioread_lock, Opt_discard, Opt_nodiscard, @@ -1304,6 +1306,8 @@ static const match_table_t tokens = { {Opt_resize, "resize"}, {Opt_delalloc, "delalloc"}, {Opt_nodelalloc, "nodelalloc"}, + {Opt_mblk_io_submit, "mblk_io_submit"}, + {Opt_nomblk_io_submit, "nomblk_io_submit"}, {Opt_block_validity, "block_validity"}, {Opt_noblock_validity, "noblock_validity"}, {Opt_inode_readahead_blks, "inode_readahead_blks=%u"}, @@ -1725,6 +1729,12 @@ set_qf_format: case Opt_nodelalloc: clear_opt(sbi->s_mount_opt, DELALLOC); break; + case Opt_mblk_io_submit: + set_opt(sbi->s_mount_opt, MBLK_IO_SUBMIT); + break; + case Opt_nomblk_io_submit: + clear_opt(sbi->s_mount_opt, MBLK_IO_SUBMIT); + break; case Opt_stripe: if (match_int(&args[0], &option)) return 0; From 6d5c3aa84b3e431f2d0fc39c73c867d1a4dd8cff Mon Sep 17 00:00:00 2001 From: Aaro Koskinen Date: Tue, 14 Dec 2010 21:45:31 -0500 Subject: [PATCH 2/2] ext4: fix typo which broke '..' detection in ext4_find_entry() There should be a check for the NUL character instead of '0'. Fortunately the only thing that cares about this is NFS serving, which is why we didn't notice this in the merge window testing. Reported-by: Phil Carmody Signed-off-by: Aaro Koskinen Signed-off-by: "Theodore Ts'o" --- fs/ext4/namei.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fs/ext4/namei.c b/fs/ext4/namei.c index 92203b8a099f..dc40e75cba88 100644 --- a/fs/ext4/namei.c +++ b/fs/ext4/namei.c @@ -872,7 +872,7 @@ static struct buffer_head * ext4_find_entry (struct inode *dir, if (namelen > EXT4_NAME_LEN) return NULL; if ((namelen <= 2) && (name[0] == '.') && - (name[1] == '.' || name[1] == '0')) { + (name[1] == '.' || name[1] == '\0')) { /* * "." or ".." will only be in the first block * NFS may look up ".."; "." should be handled by the VFS