Re: possible deadlock in xfrm_policy_delete

[Herbert Xu]

On Thu, Sep 24, 2020 at 06:44:12AM +0200, Dmitry Vyukov wrote:
>
> FWIW one of the dups of this issue was bisected to:
> 
> commit 1909760f5fc3f123e47b4e24e0ccdc0fc8f3f106
> Author: Ahmed S. Darwish <a.darwish@xxxxxxxxxxxxx>
> Date:   Fri Sep 4 15:32:31 2020 +0000
> 
>     seqlock: PREEMPT_RT: Do not starve seqlock_t writers
> 
> Can it be related?

Whether there is a genuine problem or not, the lockdep report
that I quoted is simply non-sensical because it's considering
two distinct seqlocks to be the same lock.

I don't think the lockdep issue has anything to do with the commit
question because this commit is specific to seqlocks.  There is
another syzbot report in this pile that mixed the SCTP socket lock
with the TCP socket lock and those are not seqlocks.

It's almost as if when a spinlock is freed and reallocated lockdep
is not clearing the existing state.

Cheers,
-- 
Email: Herbert Xu <herbert@xxxxxxxxxxxxxxxxxxx>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt