From: Daniel Jordan <daniel.m.jordan@oracle.com>
To: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Eric Biggers <ebiggers@kernel.org>,
Linux Crypto Mailing List <linux-crypto@vger.kernel.org>,
Steffen Klassert <steffen.klassert@secunet.com>,
Daniel Jordan <daniel.m.jordan@oracle.com>
Subject: Re: [v2 PATCH] crypto: pcrypt - Avoid deadlock by using per-instance padata queues
Date: Tue, 26 Nov 2019 00:32:38 -0500 [thread overview]
Message-ID: <20191126053238.yxhtfbt5okcjycuy@ca-dmjordan1.us.oracle.com> (raw)
In-Reply-To: <20191119185827.nerskpvddkcsih25@gondor.apana.org.au>
On Wed, Nov 20, 2019 at 02:58:27AM +0800, Herbert Xu wrote:
> /* Replace the internal control structure with a new one. */
> -static void padata_replace(struct padata_instance *pinst,
> - struct parallel_data *pd_new)
> +static int padata_replace_one(struct padata_shell *ps)
> {
> - struct parallel_data *pd_old = pinst->pd;
> + struct parallel_data *pd_old = rcu_dereference_protected(ps->pd, 1);
> + struct parallel_data *pd_new;
> int notification_mask = 0;
>
> - pinst->flags |= PADATA_RESET;
> -
> - rcu_assign_pointer(pinst->pd, pd_new);
> + pd_new = padata_alloc_pd(ps);
> + if (!pd_new)
> + return -ENOMEM;
>
> - synchronize_rcu();
> + rcu_assign_pointer(ps->pd, pd_new);
>
> if (!cpumask_equal(pd_old->cpumask.pcpu, pd_new->cpumask.pcpu))
> notification_mask |= PADATA_CPU_PARALLEL;
> @@ -510,10 +515,25 @@ static void padata_replace(struct padata_instance *pinst,
> if (atomic_dec_and_test(&pd_old->refcnt))
> padata_free_pd(pd_old);
>
> + return notification_mask;
> +}
> +
> +static void padata_replace(struct padata_instance *pinst)
> +{
> + int notification_mask = 0;
> + struct padata_shell *ps;
> +
> + pinst->flags |= PADATA_RESET;
> +
> + list_for_each_entry(ps, &pinst->pslist, list)
> + notification_mask |= padata_replace_one(ps);
> +
> + synchronize_rcu();
> +
> if (notification_mask)
> blocking_notifier_call_chain(&pinst->cpumask_change_notifier,
> notification_mask,
> - &pd_new->cpumask);
> + &pinst->cpumask);
>
> pinst->flags &= ~PADATA_RESET;
> }
I think it's possible for a task in padata_do_parallel() racing with another in
padata_replace() to use a pd after free. The synchronize_rcu() comes after the
pd_old->refcnt's are dec'd.
padata_do_parallel()
rcu_dereference_bh(ps->pd)
// doesn't see PADATA_RESET set
padata_replace()
pinst->flags |= PADATA_RESET
padata_replace_one()
rcu_assign_pointer(ps->pd, pd_new)
atomic_dec_and_test(&pd_old->refcnt)
padata_free_pd()
atomic_inc(&pd->refcnt) // too late
If I'm not missing something, one way out is adding a list_head to
parallel_data for remembering the old pd's on a local list in padata_replace()
so that this function can loop over it and drop the refs after
synchronize_rcu(). A padata_do_parallel() call will have then had a chance to
take a ref on the pd it's using.
And, not this patch, but with the removal of flushing it seems there's no need
for PADATA_RESET, so it and its EBUSY error can go away.
next prev parent reply other threads:[~2019-11-26 5:32 UTC|newest]
Thread overview: 11+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-11-19 13:05 [PATCH] crypto: pcrypt - Avoid deadlock by using per-instance padata queues Herbert Xu
2019-11-19 17:37 ` Eric Biggers
2019-11-19 18:58 ` [v2 PATCH] " Herbert Xu
2019-11-22 16:28 ` Daniel Jordan
2019-11-26 5:32 ` Daniel Jordan [this message]
2019-11-26 7:58 ` [v3 " Herbert Xu
2019-11-27 19:14 ` Eric Biggers
2019-11-29 8:40 ` [PATCH] crypto: pcrypt - Do not clear MAY_SLEEP flag in original request Herbert Xu
2019-11-29 19:24 ` Eric Biggers
2019-11-27 23:38 ` [v3 PATCH] crypto: pcrypt - Avoid deadlock by using per-instance padata queues Daniel Jordan
2019-11-29 19:25 ` Eric Biggers
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20191126053238.yxhtfbt5okcjycuy@ca-dmjordan1.us.oracle.com \
--to=daniel.m.jordan@oracle.com \
--cc=ebiggers@kernel.org \
--cc=herbert@gondor.apana.org.au \
--cc=linux-crypto@vger.kernel.org \
--cc=steffen.klassert@secunet.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).