Re: Oops NFS Locking in 2.4.x

[Olivier Molteni <olivier@molteni.net>]

Trond Myklebust wrote:

>
> Looks like 2 processes are trying to free the same lock. The problem
> is that both processes can call filp_close() at the same
> time (by calling sys_close()).
>
> The bug boils down to:
>
>    -  locks_unlock_delete() assumes that the BKL (kernel_lock()) is
>       sufficient to protect against *thisfl_p from disappearing
>       beneath it due to some second process.
> BUT
>    -  The call to lock() in locks_unlock_delete() sleeps when the
>       underlying filesystem is NFS, hence 2 processes can race despite
>       the BKL assumption.
>
> Cheers,
>    Trond

Hi (2),

While searching on the sourceforge nfs list, I saw your message 4583160
posted on 10/31/2000 08:14:42 and concerning "[NFS] [PATCH] fix deadlocks
+ blocking in 2.4.0 pre6/7 knfsd locking..."
You give a patch modifying fs/loks.c wher, among other things, you were
added the new function locks_unlock_delete().

Do you know if all the patch had been included in the kernel or if some
part that deals with deadlocks were omited for some reasons ?
Do you think this corrections failed to definitively fix the deadlocks
problems ?

You are talking about semaphores in your post, do you think it's a
possible way to follow in order to fix the situation ?


Cheers,
Olivier.