[powerpc][next-20200701] Hung task timeouts during regression test runs

[Ming Lei]

On Thu, Jul 02, 2020 at 04:53:04PM +0530, Sachin Sant wrote:
> Starting with linux-next 20200701 release I am observing automated regressions
> tests taking longer time to complete. A test which took 10 minutes with next-20200630
> took more than 60 minutes against next-20200701. 
> 
> Following hung task timeout messages were seen during these runs
> 
> [ 1718.848351]       Not tainted 5.8.0-rc3-next-20200701-autotest #1
> [ 1718.848356] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
> [ 1718.848362] NetworkManager  D    0  2626      1 0x00040080
> [ 1718.848367] Call Trace:
> [ 1718.848374] [c0000008b0f6b8f0] [c000000000c6d558] schedule+0x78/0x130 (unreliable)
> [ 1718.848382] [c0000008b0f6bad0] [c00000000001b070] __switch_to+0x2e0/0x480
> [ 1718.848388] [c0000008b0f6bb30] [c000000000c6ce9c] __schedule+0x2cc/0x910
> [ 1718.848394] [c0000008b0f6bc10] [c000000000c6d558] schedule+0x78/0x130
> [ 1718.848401] [c0000008b0f6bc40] [c0000000005d5a64] jbd2_log_wait_commit+0xd4/0x1a0
> [ 1718.848408] [c0000008b0f6bcc0] [c00000000055fb6c] ext4_sync_file+0x1cc/0x480
> [ 1718.848415] [c0000008b0f6bd20] [c000000000493530] vfs_fsync_range+0x70/0xf0
> [ 1718.848421] [c0000008b0f6bd60] [c000000000493638] do_fsync+0x58/0xd0
> [ 1718.848427] [c0000008b0f6bda0] [c0000000004936d8] sys_fsync+0x28/0x40
> [ 1718.848433] [c0000008b0f6bdc0] [c000000000035e28] system_call_exception+0xf8/0x1c0
> [ 1718.848440] [c0000008b0f6be20] [c00000000000ca70] system_call_common+0xf0/0x278
> 
> Comparing next-20200630 with next-20200701 one possible candidate seems to
> be following commit:
> 
> commit 37f4a24c2469a10a4c16c641671bd766e276cf9f
>     blk-mq: centralise related handling into blk_mq_get_driver_tag
> 
> Reverting this commit allows the test to complete in 10 minutes.

Hello,

Thanks for the report.

Please try the following fix:

https://lore.kernel.org/linux-block/20200702062041.GC2452799@T590/raw


Thanks,
Ming