ext4_da_update_reserve_space warning with kernel 3.10

ext4_da_update_reserve_space warning with kernel 3.10

[Vinson Lee]

Hi.

We're triggering a kernel warning in ext4_da_update_reserve_space on
kernels 3.4 and 3.10. We do not have a reproducible test case, but the
warning happens more frequently with a slow network.

EXT4-fs warning (device sdh1): ext4_da_update_reserve_space:362: ino
1310732, allocated 1 with only 0 reserved metadata blocks (releasing 1
blocks with reserved 14 data blocks)
------------[ cut here ]------------
WARNING: at fs/ext4/inode.c:363 ext4_da_update_reserve_space+0xe7/0x20c()
Modules linked in: netconsole configfs ipv6 ppdev parport_pc lp
parport tcp_diag inet_diag ipmi_si ipmi_devintf ipm
O_wdt iTCO_vendor_support acpi_cpufreq freq_table mperf coretemp
kvm_intel kvm crc32c_intel ghash_clmulni_intel mic
i2c_core lpc_ich mfd_core shpchp mlx4_en mlx4_core video mpt2sas
scsi_transport_sas raid_class
CPU: 0 PID: 50134 Comm: java Not tainted 3.10.50 #1
 0000000000000000 ffff880595dc1bb0 ffffffff814a4094 ffff880595dc1be8
 ffffffff8103cbf9 0000000000000000 ffff88039cdd1ef0 000000000000000e
 ffff880052527f60 ffff8807e9ad6000 ffff880595dc1bf8 ffffffff8103ccbf
Call Trace:
 [<ffffffff814a4094>] dump_stack+0x19/0x1b
 [<ffffffff8103cbf9>] warn_slowpath_common+0x65/0x7d
 [<ffffffff8103ccbf>] warn_slowpath_null+0x1a/0x1c
 [<ffffffff81190b7e>] ext4_da_update_reserve_space+0xe7/0x20c
 [<ffffffff811b30e7>] ext4_ext_map_blocks+0xae5/0xc09
 [<ffffffff81190ed5>] ext4_map_blocks+0x232/0x3a0
 [<ffffffff81192e28>] mpage_da_map_and_submit+0x8e/0x32c
 [<ffffffff8119375e>] ext4_da_writepages+0x2cf/0x483
 [<ffffffff810ed6d3>] do_writepages+0x1e/0x2c
 [<ffffffff810e5b3e>] __filemap_fdatawrite_range+0x55/0x57
 [<ffffffff810e7ef7>] SyS_fadvise64_64+0x163/0x1da
 [<ffffffff810e7f7c>] SyS_fadvise64+0xe/0x12
 [<ffffffff814aee46>] system_call_fastpath+0x1a/0x1f
---[ end trace bfdca050e80bbf21 ]---


This looks similar to a bug fixed by 3.12 commit "ext4: fix warning in
ext4_da_update_reserve_space()" [1]. However, this commit wasn't
backported to stable [2].

Cheers,
Vinson

[1] https://bugzilla.kernel.org/show_bug.cgi?id=63511
[2] http://lists.openwall.net/linux-ext4/2013/10/23/7

Re: ext4_da_update_reserve_space warning with kernel 3.10

[Jan Kara]

  Hello,

On Thu 18-12-14 17:23:22, Vinson Lee wrote:
> We're triggering a kernel warning in ext4_da_update_reserve_space on
> kernels 3.4 and 3.10. We do not have a reproducible test case, but the
> warning happens more frequently with a slow network.
> 
> EXT4-fs warning (device sdh1): ext4_da_update_reserve_space:362: ino
> 1310732, allocated 1 with only 0 reserved metadata blocks (releasing 1
> blocks with reserved 14 data blocks)
  Yes, it looks like the problem you referenced below. But we don't plan to
fix this in 3.10 or even 3.4 stable kernel. The warning is mostly harmless
and fixing it would require rather intrusive changes to be backported...

								Honza

> ------------[ cut here ]------------
> WARNING: at fs/ext4/inode.c:363 ext4_da_update_reserve_space+0xe7/0x20c()
> Modules linked in: netconsole configfs ipv6 ppdev parport_pc lp
> parport tcp_diag inet_diag ipmi_si ipmi_devintf ipm
> O_wdt iTCO_vendor_support acpi_cpufreq freq_table mperf coretemp
> kvm_intel kvm crc32c_intel ghash_clmulni_intel mic
> i2c_core lpc_ich mfd_core shpchp mlx4_en mlx4_core video mpt2sas
> scsi_transport_sas raid_class
> CPU: 0 PID: 50134 Comm: java Not tainted 3.10.50 #1
>  0000000000000000 ffff880595dc1bb0 ffffffff814a4094 ffff880595dc1be8
>  ffffffff8103cbf9 0000000000000000 ffff88039cdd1ef0 000000000000000e
>  ffff880052527f60 ffff8807e9ad6000 ffff880595dc1bf8 ffffffff8103ccbf
> Call Trace:
>  [<ffffffff814a4094>] dump_stack+0x19/0x1b
>  [<ffffffff8103cbf9>] warn_slowpath_common+0x65/0x7d
>  [<ffffffff8103ccbf>] warn_slowpath_null+0x1a/0x1c
>  [<ffffffff81190b7e>] ext4_da_update_reserve_space+0xe7/0x20c
>  [<ffffffff811b30e7>] ext4_ext_map_blocks+0xae5/0xc09
>  [<ffffffff81190ed5>] ext4_map_blocks+0x232/0x3a0
>  [<ffffffff81192e28>] mpage_da_map_and_submit+0x8e/0x32c
>  [<ffffffff8119375e>] ext4_da_writepages+0x2cf/0x483
>  [<ffffffff810ed6d3>] do_writepages+0x1e/0x2c
>  [<ffffffff810e5b3e>] __filemap_fdatawrite_range+0x55/0x57
>  [<ffffffff810e7ef7>] SyS_fadvise64_64+0x163/0x1da
>  [<ffffffff810e7f7c>] SyS_fadvise64+0xe/0x12
>  [<ffffffff814aee46>] system_call_fastpath+0x1a/0x1f
> ---[ end trace bfdca050e80bbf21 ]---
> 
> 
> This looks similar to a bug fixed by 3.12 commit "ext4: fix warning in
> ext4_da_update_reserve_space()" [1]. However, this commit wasn't
> backported to stable [2].
> 
> Cheers,
> Vinson
> 
> [1] https://bugzilla.kernel.org/show_bug.cgi?id=63511
> [2] http://lists.openwall.net/linux-ext4/2013/10/23/7
-- 
Jan Kara <jack@suse.cz>
SUSE Labs, CR

Re: ext4_da_update_reserve_space warning with kernel 3.10

[Theodore Ts'o]

On Thu, Dec 18, 2014 at 05:23:22PM -0800, Vinson Lee wrote:
> Hi.
> 
> We're triggering a kernel warning in ext4_da_update_reserve_space on
> kernels 3.4 and 3.10. We do not have a reproducible test case, but the
> warning happens more frequently with a slow network.
> 
> EXT4-fs warning (device sdh1): ext4_da_update_reserve_space:362: ino
> 1310732, allocated 1 with only 0 reserved metadata blocks (releasing 1
> blocks with reserved 14 data blocks)

Known problem, and fixed in newer kernels.  In fact, the warning is
harmless with the 3.10 kernels (since it includes commit 27dd4385422)
which prevents us from failing to be able to allocate metadata blocks.
In later kernels, as suggested by the commit description, we
eliminated this particular warning since it was no longer necessary.
As far as the 3.4 kernel is concerned, you could try backporting
27dd4385422, although there's a good chance enough stuff has changed
that the backport might not be at all trivial.

Regards,

					- Ted


> ------------[ cut here ]------------
> WARNING: at fs/ext4/inode.c:363 ext4_da_update_reserve_space+0xe7/0x20c()
> Modules linked in: netconsole configfs ipv6 ppdev parport_pc lp
> parport tcp_diag inet_diag ipmi_si ipmi_devintf ipm
> O_wdt iTCO_vendor_support acpi_cpufreq freq_table mperf coretemp
> kvm_intel kvm crc32c_intel ghash_clmulni_intel mic
> i2c_core lpc_ich mfd_core shpchp mlx4_en mlx4_core video mpt2sas
> scsi_transport_sas raid_class
> CPU: 0 PID: 50134 Comm: java Not tainted 3.10.50 #1
>  0000000000000000 ffff880595dc1bb0 ffffffff814a4094 ffff880595dc1be8
>  ffffffff8103cbf9 0000000000000000 ffff88039cdd1ef0 000000000000000e
>  ffff880052527f60 ffff8807e9ad6000 ffff880595dc1bf8 ffffffff8103ccbf
> Call Trace:
>  [<ffffffff814a4094>] dump_stack+0x19/0x1b
>  [<ffffffff8103cbf9>] warn_slowpath_common+0x65/0x7d
>  [<ffffffff8103ccbf>] warn_slowpath_null+0x1a/0x1c
>  [<ffffffff81190b7e>] ext4_da_update_reserve_space+0xe7/0x20c
>  [<ffffffff811b30e7>] ext4_ext_map_blocks+0xae5/0xc09
>  [<ffffffff81190ed5>] ext4_map_blocks+0x232/0x3a0
>  [<ffffffff81192e28>] mpage_da_map_and_submit+0x8e/0x32c
>  [<ffffffff8119375e>] ext4_da_writepages+0x2cf/0x483
>  [<ffffffff810ed6d3>] do_writepages+0x1e/0x2c
>  [<ffffffff810e5b3e>] __filemap_fdatawrite_range+0x55/0x57
>  [<ffffffff810e7ef7>] SyS_fadvise64_64+0x163/0x1da
>  [<ffffffff810e7f7c>] SyS_fadvise64+0xe/0x12
>  [<ffffffff814aee46>] system_call_fastpath+0x1a/0x1f
> ---[ end trace bfdca050e80bbf21 ]---
> 
> 
> This looks similar to a bug fixed by 3.12 commit "ext4: fix warning in
> ext4_da_update_reserve_space()" [1]. However, this commit wasn't
> backported to stable [2].
> 
> Cheers,
> Vinson
> 
> [1] https://bugzilla.kernel.org/show_bug.cgi?id=63511
> [2] http://lists.openwall.net/linux-ext4/2013/10/23/7