REGRESSION on linux-next (next-20250814)

REGRESSION on linux-next (next-20250814)

[Borah, Chaitanya Kumar]

Hello Julian,

Hope you are doing well. I am Chaitanya from the linux graphics team in 
Intel.

This mail is regarding a regression we are seeing in our CI runs[1] on
linux-next repository.

Since the version next-20250814 [2], we are seeing the following regression

`````````````````````````````````````````````````````````````````````````````````
<4>[   25.645493] ======================================================
<4>[   25.645497] WARNING: possible circular locking dependency detected
<5>[   25.645501] 6.17.0-rc2-next-20250818-next-20250818-g3ac864c2d9bb+ 
#1 Not tainted
<4>[   25.645506] ------------------------------------------------------
<4>[   25.645509] swapper/0/1 is trying to acquire lock:
<5>[   25.645513] ffffffff83488270 (cpu_hotplug_lock){++++}-{0:0}, at: 
static_key_slow_inc+0x12/0x30
<4>[   25.645526]
                   but task is already holding lock:
<5>[   25.645529] ffff8881063fce30 
(&q->q_usage_counter(io)){++++}-{0:0}, at: 
blk_mq_freeze_queue_nomemsave+0x12/0x30
<4>[   25.645540]
                   which lock already depends on the new lock.

<4>[   25.645545]
                   the existing dependency chain (in reverse order) is:
<5>[   25.645549]
                   -> #2 (&q->q_usage_counter(io)){++++}-{0:0}:
<5>[   25.645554]        blk_alloc_queue+0x324/0x360
<5>[   25.645560]        blk_mq_alloc_queue+0x6a/0xe0
<5>[   25.645564]        __blk_mq_alloc_disk+0x19/0x70
<5>[   25.645567]        loop_add+0x240/0x430
<5>[   25.645573]        loop_init+0xcd/0x190
<5>[   25.645576]        do_one_initcall+0x5d/0x3f0
<5>[   25.645581]        kernel_init_freeable+0x3cd/0x6a0
<5>[   25.645586]        kernel_init+0x1b/0x200
<5>[   25.645591]        ret_from_fork+0x26c/0x2e0
<5>[   25.645597]        ret_from_fork_asm+0x1a/0x30
`````````````````````````````````````````````````````````````````````````````````
Details log can be found in [3].

After bisecting the tree, the following patch [4] seems to be the first 
"bad" commit

`````````````````````````````````````````````````````````````````````````````````````````````````````````
commit 8f5845e0743bf3512b71b3cb8afe06c192d6acc4
Author: Julian Sun sunjunchao2870@gmail.com
Date:   Tue Aug 12 23:42:57 2025 +0800

     block: restore default wbt enablement
`````````````````````````````````````````````````````````````````````````````````````````````````````````

We also verified that if we revert the patch the issue is not seen.

Could you please check why the patch causes this regression and provide 
a fix if necessary?

Thank you.

Regards

Chaitanya

[1]
https://intel-gfx-ci.01.org/tree/linux-next/combined-alt.html?
[2] 
https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/commit/?h=next-20250814
[3] 
https://intel-gfx-ci.01.org/tree/linux-next/next-20250818/bat-twl-1/boot0.txt
[4] 
https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/commit/?h=next-20250818&id=8f5845e0743bf3512b71b3cb8afe06c192d6acc4

Re: REGRESSION on linux-next (next-20250814)

[Yu Kuai]

Hi,

在 2025/08/18 16:35, Borah, Chaitanya Kumar 写道:
> Hello Julian,
> 
> Hope you are doing well. I am Chaitanya from the linux graphics team in 
> Intel.
> 
> This mail is regarding a regression we are seeing in our CI runs[1] on
> linux-next repository.
> 
> Since the version next-20250814 [2], we are seeing the following regression
> 
> ````````````````````````````````````````````````````````````````````````````````` 
> 
> <4>[   25.645493] ======================================================
> <4>[   25.645497] WARNING: possible circular locking dependency detected
> <5>[   25.645501] 6.17.0-rc2-next-20250818-next-20250818-g3ac864c2d9bb+ 
> #1 Not tainted
> <4>[   25.645506] ------------------------------------------------------
> <4>[   25.645509] swapper/0/1 is trying to acquire lock:
> <5>[   25.645513] ffffffff83488270 (cpu_hotplug_lock){++++}-{0:0}, at: 
> static_key_slow_inc+0x12/0x30
> <4>[   25.645526]
>                    but task is already holding lock:
> <5>[   25.645529] ffff8881063fce30 
> (&q->q_usage_counter(io)){++++}-{0:0}, at: 
> blk_mq_freeze_queue_nomemsave+0x12/0x30
> <4>[   25.645540]
>                    which lock already depends on the new lock.
> 
> <4>[   25.645545]
>                    the existing dependency chain (in reverse order) is:
> <5>[   25.645549]
>                    -> #2 (&q->q_usage_counter(io)){++++}-{0:0}:
> <5>[   25.645554]        blk_alloc_queue+0x324/0x360
> <5>[   25.645560]        blk_mq_alloc_queue+0x6a/0xe0
> <5>[   25.645564]        __blk_mq_alloc_disk+0x19/0x70
> <5>[   25.645567]        loop_add+0x240/0x430
> <5>[   25.645573]        loop_init+0xcd/0x190
> <5>[   25.645576]        do_one_initcall+0x5d/0x3f0
> <5>[   25.645581]        kernel_init_freeable+0x3cd/0x6a0
> <5>[   25.645586]        kernel_init+0x1b/0x200
> <5>[   25.645591]        ret_from_fork+0x26c/0x2e0
> <5>[   25.645597]        ret_from_fork_asm+0x1a/0x30
> ````````````````````````````````````````````````````````````````````````````````` 
> 
> Details log can be found in [3].

The bisecting is because this patch actually enable wbt now, while this
patch is not the root cause.

Following is a fix by Nilay, you can have a test if your problem can be
fixed.

https://lore.kernel.org/all/20250814082612.500845-1-nilay@linux.ibm.com/

Thanks,
Kuai

> 
> After bisecting the tree, the following patch [4] seems to be the first 
> "bad" commit
> 
> ````````````````````````````````````````````````````````````````````````````````````````````````````````` 
> 
> commit 8f5845e0743bf3512b71b3cb8afe06c192d6acc4
> Author: Julian Sun sunjunchao2870@gmail.com
> Date:   Tue Aug 12 23:42:57 2025 +0800
> 
>      block: restore default wbt enablement
> ````````````````````````````````````````````````````````````````````````````````````````````````````````` 
> 
> 
> We also verified that if we revert the patch the issue is not seen.
> 
> Could you please check why the patch causes this regression and provide 
> a fix if necessary?
> 
> Thank you.
> 
> Regards
> 
> Chaitanya
> 
> [1]
> https://intel-gfx-ci.01.org/tree/linux-next/combined-alt.html?
> [2] 
> https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/commit/?h=next-20250814 
> 
> [3] 
> https://intel-gfx-ci.01.org/tree/linux-next/next-20250818/bat-twl-1/boot0.txt 
> 
> [4] 
> https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/commit/?h=next-20250818&id=8f5845e0743bf3512b71b3cb8afe06c192d6acc4 
> 
> 
> 
> .
> 

Re: REGRESSION on linux-next (next-20250814)

[Borah, Chaitanya Kumar]

On 8/18/2025 2:46 PM, Yu Kuai wrote:
> 
> The bisecting is because this patch actually enable wbt now, while this
> patch is not the root cause.
> 
> Following is a fix by Nilay, you can have a test if your problem can be
> fixed.
> 
> https://lore.kernel.org/all/20250814082612.500845-1-nilay@linux.ibm.com/

Thank you, Kuai, for the reply. The patches works for us.
We will wait for them to get merged.

Regards

Chaitanya