[Ocfs2-devel] [PATCH] ocfs2: update inode size after zeroed the hole
Junxiao Bi
junxiao.bi at oracle.com
Thu Jul 11 18:25:07 PDT 2013
Hi Younger,
Thanks for looking at this. See the comment below.
On 07/11/2013 08:42 PM, Younger Liu wrote:
> On 2013/7/10 14:47, Junxiao Bi wrote:
>> Sometimes, this patch will cause a call trace, see following, this code
>> snippet is referenced from ocfs2_write_end(), I am not familiar with
>> jbd, anybody know why?
>>
>>
>> JBD: sda7-523: jh->b_next_transaction (58645, (null), 0) !=
>> transaction (ffff88014fa75ec0, 59276)
>> ------------[ cut here ]------------
>> WARNING: at fs/jbd2/transaction.c:1237
>> jbd2_journal_dirty_metadata+0xf6/0x250()
>> Modules linked in: ebtable_nat ebtables nf_conntrack_ipv4 nf_defrag_ipv4
>> xt_state nf_conntrack ipt_REJECT iptable_filter ip_tables bridge stp llc
>> autofs4 hidp rfcomm bluetooth rfkill lockd sunrpc cpufreq_ondemand
>> ip6t_REJECT ip6table_filter ip6_tables be2iscsi iscsi_boot_sysfs ib_iser
>> rdma_cm ib_cm iw_cm ib_sa ib_mad ib_core ib_addr iscsi_tcp bnx2i cnic
>> uio ipv6 cxgb3i libcxgbi cxgb3 mdio libiscsi_tcp libiscsi
>> scsi_transport_iscsi sbs sbshc hed acpi_pad acpi_ipmi ipmi_msghandler
>> parport_pc lp parport i915 snd_hda_codec_hdmi snd_hda_codec_realtek
>> snd_hda_intel snd_hda_codec snd_hwdep snd_seq_dummy snd_seq_oss
>> drm_kms_helper snd_seq_midi_event drm snd_seq sg snd_seq_device
>> snd_pcm_oss snd_mixer_oss e1000e iTCO_wdt snd_pcm iTCO_vendor_support
>> i2c_algo_bit sr_mod cdrom snd_timer snd i2c_i801 ptp dcdbas soundcore
>> pcspkr lpc_ich pps_core acpi_cpufreq serio_raw i2c_core video
>> snd_page_alloc mfd_core freq_table mperf dm_snapshot(F) dm_zero(F)
>> dm_mirror(F) dm_region_hash(F) dm_log(F) dm_mod(F) shpchp(F) ahci(F)
>> libahci(F) sd_mod(F) crc_t10dif(F) ext3(F) jbd(F) mbcache(F)
>> CPU: 2 PID: 6940 Comm: tar Tainted: PF 3.10.0-rc6 #2
>> Hardware name: Dell Inc. OptiPlex 790/0PG55N, BIOS A11 12/30/2011
>> ffffffff8184f02f ffff880214c29808 ffffffff8132e369 ffff880214c29848
>> ffffffff81041e10 0000000000001000 ffff88014f8ff600 00000000ffffffea
>> ffff88014fa75ec0 ffff88021e275000 ffff88020a0b4338 ffff880214c29858
>> Call Trace:
>> [<ffffffff8132e369>] dump_stack+0x19/0x20
>> [<ffffffff81041e10>] warn_slowpath_common+0x70/0xa0
>> [<ffffffff81041e5a>] warn_slowpath_null+0x1a/0x20
>> [<ffffffff811e0eb6>] jbd2_journal_dirty_metadata+0xf6/0x250
>> [<ffffffff812304ef>] ocfs2_journal_dirty+0x2f/0x70
>> [<ffffffff81224571>] ocfs2_write_zero_page+0x301/0x3e0
>> [<ffffffff812246e8>] ocfs2_zero_extend_range+0x98/0x160
>> [<ffffffff81224872>] ocfs2_zero_extend+0xc2/0x220
>> [<ffffffff812049cc>] ocfs2_zero_tail+0x2c/0x40
>> [<ffffffff81207fbb>] ocfs2_write_begin_nolock+0xfb/0xc80
>> [<ffffffff81229220>] ? ocfs2_read_inode_block+0x10/0x20
>> [<ffffffff81217d8a>] ? ocfs2_assign_bh+0x2a/0xb0
>> [<ffffffff8121c5a9>] ? ocfs2_inode_lock_full_nested+0xa9/0x470
>> [<ffffffff81208c76>] ocfs2_write_begin+0x136/0x200
>> [<ffffffff8110c444>] generic_perform_write+0xd4/0x1e0
>> [<ffffffff8110c5b5>] generic_file_buffered_write+0x65/0xa0
>> [<ffffffff812266b1>] ocfs2_file_aio_write+0x691/0x7b0
>> [<ffffffff81163ea7>] do_sync_write+0x97/0xe0
>> [<ffffffff811647a5>] vfs_write+0xe5/0x1e0
>> [<ffffffff81164ae1>] SyS_write+0x61/0xa0
>> [<ffffffff81604cc2>] system_call_fastpath+0x16/0x1b
>> ---[ end trace edb3aa856f6f3cde ]---
>>
> Could you put out the issue code? and tell us which line of code trigger the bug.
> From "fs/jbd2/transaction.c:1237", I could not find the details.
The issue is cause by the posted patch, without the patch, no this issue.
I think I found the cause, that was I should call ocfs2_journal_access_di()
before call ocfs2_journal_dirty(). This is asked by
jbd2_journal_dirty_metadata()
which is called by ocfs2_journal_dirty(). I will posted the v2 patch.
Thanks,
Junxiao.
>
>> On 07/10/2013 09:03 AM, Junxiao Bi wrote:
>>> fs-writeback will release the dirty pages without page lock
>>> whose offset are over inode size, the release happens at
>>> block_write_full_page_endio(). If not update, dirty pages
>>> in file holes may be released before flushed to the disk,
>>> then file holes will contain some non-zero data, this will
>>> cause sparse file cksum error.
>>>
>>> Signed-off-by: Junxiao Bi <junxiao.bi at oracle.com>
>>> ---
>>> fs/ocfs2/file.c | 25 ++++++++++++++++++++-----
>>> 1 file changed, 20 insertions(+), 5 deletions(-)
>>>
>>> diff --git a/fs/ocfs2/file.c b/fs/ocfs2/file.c
>>> index ff54014..d1264ef 100644
>>> --- a/fs/ocfs2/file.c
>>> +++ b/fs/ocfs2/file.c
>>> @@ -756,7 +756,7 @@ out:
>>> * to be too fragile to do exactly what we need without us having to
>>> * worry about recursive locking in ->write_begin() and ->write_end(). */
>>> static int ocfs2_write_zero_page(struct inode *inode, u64 abs_from,
>>> - u64 abs_to)
>>> + u64 abs_to, struct buffer_head *di_bh)
>>> {
>>> struct address_space *mapping = inode->i_mapping;
>>> struct page *page;
>>> @@ -764,6 +764,7 @@ static int ocfs2_write_zero_page(struct inode *inode, u64 abs_from,
>>> handle_t *handle = NULL;
>>> int ret = 0;
>>> unsigned zero_from, zero_to, block_start, block_end;
>>> + struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
>>>
>>> BUG_ON(abs_from >= abs_to);
>>> BUG_ON(abs_to > (((u64)index + 1) << PAGE_CACHE_SHIFT));
>>> @@ -823,8 +824,22 @@ static int ocfs2_write_zero_page(struct inode *inode, u64 abs_from,
>>> ret = 0;
>>> }
>>>
>>> - if (handle)
>>> + if (handle) {
>>> + /*
>>> + * fs-writeback will release the dirty pages without page lock
>>> + * whose offset are over inode size, the release happens at
>>> + * block_write_full_page_endio().
>>> + */
>>> + i_size_write(inode, abs_to);
>>> + inode->i_blocks = ocfs2_inode_sector_count(inode);
>>> + di->i_size = cpu_to_le64((u64)i_size_read(inode));
>>> + inode->i_mtime = inode->i_ctime = CURRENT_TIME;
>>> + di->i_mtime = di->i_ctime = cpu_to_le64(inode->i_mtime.tv_sec);
>>> + di->i_ctime_nsec = cpu_to_le32(inode->i_mtime.tv_nsec);
>>> + di->i_mtime_nsec = di->i_ctime_nsec;
>>> + ocfs2_journal_dirty(handle, di_bh);
>>> ocfs2_commit_trans(OCFS2_SB(inode->i_sb), handle);
>>> + }
>>>
>>> out_unlock:
>>> unlock_page(page);
>>> @@ -920,7 +935,7 @@ out:
>>> * has made sure that the entire range needs zeroing.
>>> */
>>> static int ocfs2_zero_extend_range(struct inode *inode, u64 range_start,
>>> - u64 range_end)
>>> + u64 range_end, struct buffer_head *di_bh)
>>> {
>>> int rc = 0;
>>> u64 next_pos;
>>> @@ -936,7 +951,7 @@ static int ocfs2_zero_extend_range(struct inode *inode, u64 range_start,
>>> next_pos = (zero_pos & PAGE_CACHE_MASK) + PAGE_CACHE_SIZE;
>>> if (next_pos > range_end)
>>> next_pos = range_end;
>>> - rc = ocfs2_write_zero_page(inode, zero_pos, next_pos);
>>> + rc = ocfs2_write_zero_page(inode, zero_pos, next_pos, di_bh);
>>> if (rc < 0) {
>>> mlog_errno(rc);
>>> break;
>>> @@ -982,7 +997,7 @@ int ocfs2_zero_extend(struct inode *inode, struct buffer_head *di_bh,
>>> range_end = zero_to_size;
>>>
>>> ret = ocfs2_zero_extend_range(inode, range_start,
>>> - range_end);
>>> + range_end, di_bh);
>>> if (ret) {
>>> mlog_errno(ret);
>>> break;
>>
>> _______________________________________________
>> Ocfs2-devel mailing list
>> Ocfs2-devel at oss.oracle.com
>> https://oss.oracle.com/mailman/listinfo/ocfs2-devel
>>
>> .
>>
>
More information about the Ocfs2-devel
mailing list