linux-kernel - Re: [PATCH] ksmbd: use F_SETLK to force vfs_file

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <b9e08f35-66b5-b627-fe7e-f1907745be8a@virtuozzo.com>
Date:   Wed, 22 Dec 2021 10:40:23 +0300
From:   Vasily Averin <vvs@...tuozzo.com>
To:     Namjae Jeon <linkinjeon@...nel.org>
Cc:     Sergey Senozhatsky <senozhatsky@...omium.org>,
        Steve French <sfrench@...ba.org>,
        Hyunchul Lee <hyc.lee@...il.com>, kernel@...nvz.org,
        linux-cifs@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] ksmbd: use F_SETLK to force vfs_file_lock() to return
 asynchronously

On 22.12.2021 09:51, Vasily Averin wrote:
> On 22.12.2021 08:25, Namjae Jeon wrote:
>> 2021-12-22 13:32 GMT+09:00, Vasily Averin <vvs@...tuozzo.com>:
>>> On 22.12.2021 05:50, Namjae Jeon wrote:
>>>> 2021-12-21 22:08 GMT+09:00, Vasily Averin <vvs@...tuozzo.com>:
>>>>> On 21.12.2021 15:02, Namjae Jeon wrote:
>>>>>> 2021-12-19 18:34 GMT+09:00, Vasily Averin <vvs@...tuozzo.com>:
>>>>>>> To avoid possible deadlock ksmbd should process locks asynchronously.
>>>>>>> Callers expecting vfs_file_locks() to return asynchronously should
>>>>>>> only
>>>>>>> use F_SETLK, not F_SETLKW.
>>>>>> Should I check this patch instead of
>>>>>> [PATCH] ksmbd: force "fail immediately" flag on fs with its own ->lock
>>>>>> ?
>>>>>
>>>>> no, these patches are independent and both ones are required.
>>>>> current patch fixes incorrect kernel thread behaviour:
>>>>> kernel threads should not use F_SETLKW for locking requests.
>>>> How does this patch work? posix_lock_file in vfs_lock_file() does not use
>>>> cmd.
>>>> And your patch still leaves FL_SLEEP.
>>>
>>> "use F_SETLK, not F_SETLKW" was copy-pasted from requirement described in
>>> comment above vfs_lock_file().
>>>
>>> posix_lock_file() is not used in all ->lock() functions, and use F_SETLKW
>>> forces some of affected filesystem use blocking locks:
>> What I'm saying is that when we apply "ksmbd: force "fail immediately"
>> flag on fs with its own ->lock ", this patch is meaningless. How is
>> ->lock() with F_SETLKW called?
> 
> I got your point finally,
> yes, you are right, now this cannot happen.
> However I'm going to fix all affected filesystems and then revert
> "ksmbd: force "fail immediately" flag on fs with its own ->lock"
> When this happen and ksmbd will still use IS_SETLKW it will trigger the problems described below.

I've created
https://bugzilla.kernel.org/show_bug.cgi?id=215383