Received-SPF: pass (google.com: domain of linux-wireless+bounces-4093-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:40f1:3f00::1 as permitted sender) client-ip=2604:1380:40f1:3f00::1;
Message-ID: <6be142ea-6e35-4f1e-8763-6f4455f51135@candelatech.com>
Date: Tue, 27 Feb 2024 05:51:22 -0800
Precedence: bulk
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
Subject: Re: Kernel deadlock in 6.7.5 + hacks, maybe debugfs related.
Content-Language: en-MW
To: Johannes Berg <johannes@sipsolutions.net>,
 linux-wireless <linux-wireless@vger.kernel.org>
References: <1c9fa9e5-09f1-0522-fdbc-dbcef4d255ca@candelatech.com>
 <bdc7c51e3b6afaa7ac79efa15c6d1d41e9e918b5.camel@sipsolutions.net>
 <95d56b9a-4b91-476f-bab3-7af0f28ff469@candelatech.com>
 <0f8a0451e547f318febcbe36823e1f9914cc684b.camel@sipsolutions.net>
From: Ben Greear <greearb@candelatech.com>
Organization: Candela Technologies
In-Reply-To: <0f8a0451e547f318febcbe36823e1f9914cc684b.camel@sipsolutions.net>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit

On 2/27/24 05:47, Johannes Berg wrote:
> 
>> Feb 26 06:01:45 ct523c-0b0b kernel: task:ip              state:D stack:0     pid:28125 tgid:28125 ppid:3604   flags:0x00004002
>> Feb 26 06:01:45 ct523c-0b0b kernel: Call Trace:
>> Feb 26 06:01:45 ct523c-0b0b kernel:  <TASK>
>> Feb 26 06:01:45 ct523c-0b0b kernel:  __schedule+0x42c/0xde0
>> Feb 26 06:01:45 ct523c-0b0b kernel:  schedule+0x3c/0x120
>> Feb 26 06:01:45 ct523c-0b0b kernel:  schedule_timeout+0x19c/0x1b0
>> Feb 26 06:01:45 ct523c-0b0b kernel:  ? mark_held_locks+0x49/0x70
>> Feb 26 06:01:45 ct523c-0b0b kernel:  __wait_for_common+0xba/0x1d0
>> Feb 26 06:01:45 ct523c-0b0b kernel:  ? usleep_range_state+0xb0/0xb0
>> Feb 26 06:01:45 ct523c-0b0b kernel:  remove_one+0x6b/0x100
> 
> Can you say where this remove_one+0x6b is?
> 
> I feel it's probably this:
> 
>         if (!refcount_dec_and_test(&fsd->active_users)) {
>                 wait_for_completion(&fsd->active_users_drained);
> 
> which ... looking at it, seems wrong?


(gdb) l *(remove_one+0x6b)
0xffffffff815c257b is in remove_one (/home/greearb/git/linux-6.7.dev.y/fs/debugfs/inode.c:757).
752			return;
753	
754		/* if we hit zero, just wait for all to finish */
755		if (!refcount_dec_and_test(&fsd->active_users)) {
756			wait_for_completion(&fsd->active_users_drained);
757			return;
758		}
759	
760		/* if we didn't hit zero, try to cancel any we can */
761		while (refcount_read(&fsd->active_users)) {
(gdb)

> 
> _Completely_ untested:

We can test it.

Thanks,
Ben

> 
> diff --git a/fs/debugfs/inode.c b/fs/debugfs/inode.c
> index 034a617cb1a5..fb636478c54d 100644
> --- a/fs/debugfs/inode.c
> +++ b/fs/debugfs/inode.c
> @@ -751,13 +751,19 @@ static void __debugfs_file_removed(struct dentry *dentry)
>   	if ((unsigned long)fsd & DEBUGFS_FSDATA_IS_REAL_FOPS_BIT)
>   		return;
>   
> -	/* if we hit zero, just wait for all to finish */
> -	if (!refcount_dec_and_test(&fsd->active_users)) {
> -		wait_for_completion(&fsd->active_users_drained);
> -		return;
> -	}
> +	/*
> +	 * Now that debugfs_file_get() no longer sees a valid entry,
> +	 * decrement the refcount to remove the initial reference.
> +	 */
> +	refcount_dec(&fsd->active_users);
>   
> -	/* if we didn't hit zero, try to cancel any we can */
> +	/*
> +	 * As long as it's not zero, try to cancel any cancellations,
> +	 * new incoming ones will wake up the completion as we might
> +	 * have raced: debugfs_file_get() had already been done, but
> +	 * debugfs_enter_cancellation() hadn't, by the time we got
> +	 * to this point here.
> +	 */
>   	while (refcount_read(&fsd->active_users)) {
>   		struct debugfs_cancellation *c;
>   
> 
> 
> 
> johannes
> 

-- 
Ben Greear <greearb@candelatech.com>
Candela Technologies Inc  http://www.candelatech.com