Re: [PATCH] UBI: Fix possible deadlock in erase_worker()

[Richard Weinberger <richard@nod.at>]

Am 17.09.2014 10:28, schrieb Artem Bityutskiy:
> On Tue, 2014-09-16 at 09:48 +0200, Richard Weinberger wrote:
>> If sync_erase() failes with EINTR, ENOMEM, EAGAIN or
>> EBUSY erase_worker() re-schedules the failed work.
>> This will lead to a deadlock because erase_worker() is called
>> with work_sem held in read mode. And schedule_erase() will take
>> this lock again.
> 
> IIRC, the assumption was that the R/W semaphore may be taken in read
> mode many times, so it wouldn't hurt to do:
> 
> down_read()
> down_read()
> up_read()
> up_read()

Hmm, are you sure that this is legal?

Quoting rwsem.h:
/*
 * nested locking. NOTE: rwsems are not allowed to recurse
 * (which occurs if the same task tries to acquire the same
 * lock instance multiple times), but multiple locks of the
 * same lock class might be taken, if the order of the locks
 * is always the same. This ordering rule can be expressed
 * to lockdep via the _nested() APIs, but enumerating the
 * subclasses that are used. (If the nesting relationship is
 * static then another method for expressing nested locking is
 * the explicit definition of lock class keys and the use of
 * lockdep_set_class() at lock initialization time.
 * See Documentation/lockdep-design.txt for more details.)
 */

In this case the same task is taking the same lock multiple times,
which is not allowed according to rwsem.h.

Thanks,
//richard