linux-sgx.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Jarkko Sakkinen <jarkko@kernel.org>
To: Dave Hansen <dave.hansen@intel.com>
Cc: Sean Christopherson <seanjc@google.com>,
	linux-sgx@vger.kernel.org, kai.huang@intel.com,
	haitao.huang@intel.com, stable@vger.kernel.org,
	Haitao Huang <haitao.huang@linux.intel.com>,
	Thomas Gleixner <tglx@linutronix.de>,
	Ingo Molnar <mingo@redhat.com>, Borislav Petkov <bp@alien8.de>,
	x86@kernel.org, "H. Peter Anvin" <hpa@zytor.com>,
	Jethro Beekman <jethro@fortanix.com>
Subject: Re: [PATCH v4] x86/sgx: Fix the call order of synchronize_srcu() in sgx_release()
Date: Thu, 21 Jan 2021 14:55:17 +0200	[thread overview]
Message-ID: <YAl5tR5uvDqv/Is6@suppilovahvero.lan> (raw)
In-Reply-To: <042741ff-1436-f3f2-df23-a524d1d7026d@intel.com>

On Wed, Jan 20, 2021 at 05:19:44PM -0800, Dave Hansen wrote:
> On 1/20/21 4:29 PM, Jarkko Sakkinen wrote:
> > On Wed, Jan 20, 2021 at 09:35:28AM -0800, Sean Christopherson wrote:
> >> On Fri, Jan 15, 2021, jarkko@kernel.org wrote:
> >>> From: Jarkko Sakkinen <jarkko@kernel.org>
> >>>
> >>> The most trivial example of a race condition can be demonstrated with this
> >>> example where mm_list contains just one entry:
> >>>
> >>> CPU A                   CPU B
> >>> sgx_release()
> >>>                         sgx_mmu_notifier_release()
> >>>                         list_del_rcu()
> >>> sgx_encl_release()
> >>>                         synchronize_srcu()
> >>> cleanup_srcu_struct()
> >>>
> >>> To fix this, call synchronize_srcu() before checking whether mm_list is
> >>> empty in sgx_release().
> >> Why haven't you included the splat that Haitao provided?  That would go a long
> >> way to helping answer Boris' question about exactly what is broken...
> > I've lost the klog.
> 
> Haitao said he thought it was this:
> 
> > void cleanup_srcu_struct(struct srcu_struct *ssp)
> > {
> ...
> >         if (WARN_ON(srcu_readers_active(ssp)))
> >                 return; /* Just leak it! */
> 
> Which would indicate that an 'encl' kref reached 0 while some other
> thread was inside a
> 
>         idx = srcu_read_lock(&encl->srcu);
> 	...
> 	srcu_read_unlock(&encl->srcu, idx);
> 
> critical section.  A quick audit didn't turn up any obvious suspects,
> though.
> 
> If that *is* it, it might be nice to try to catch the culprit at
> srcu_read_{un}lock() time.  If there's ever a 0 refcount at those sites,
> it would be nice to spew a warning:
> 
>         idx = srcu_read_lock(&encl->srcu);
> 	WARN_ON(!atomic_read(&encl->refcount->refcount);
> 	...
> 	WARN_ON(!atomic_read(&encl->refcount->refcount);
> 	srcu_read_unlock(&encl->srcu, idx);
> 
> at each site.

The root cause is fully known already and audited.

An mm_list entry is kept up until the process exits *or* when
VFS close happens, and sgx_release() executes and removes it.

It's entirely possible that MMU notifier callback registered
by a process happens while sgx_release() is executing, which
causes list_del_rcu() to happen, unnoticed by sgx_release().

If that empties the list, cleanup_srcu_struct() is executed
unsynchronized in the middle a unfinished grace period.

/Jarkko

  reply	other threads:[~2021-01-21 12:56 UTC|newest]

Thread overview: 17+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-01-15  1:46 [PATCH v4] x86/sgx: Fix the call order of synchronize_srcu() in sgx_release() jarkko
2021-01-15  7:18 ` Borislav Petkov
2021-01-16  5:12   ` Jarkko Sakkinen
2021-01-18 18:57     ` Borislav Petkov
2021-01-20 14:43       ` Jarkko Sakkinen
2021-01-20 17:34         ` Dave Hansen
2021-01-21  0:26           ` Jarkko Sakkinen
2021-01-22 18:20             ` Haitao Huang
2021-01-20 17:35 ` Sean Christopherson
2021-01-21  0:29   ` Jarkko Sakkinen
2021-01-21  1:19     ` Dave Hansen
2021-01-21 12:55       ` Jarkko Sakkinen [this message]
2021-01-21 18:19         ` Dave Hansen
2021-01-22 16:56   ` Dave Hansen
2021-01-23  8:58     ` Jarkko Sakkinen
2021-01-25 15:49       ` Dave Hansen
2021-01-27 17:31         ` Jarkko Sakkinen

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=YAl5tR5uvDqv/Is6@suppilovahvero.lan \
    --to=jarkko@kernel.org \
    --cc=bp@alien8.de \
    --cc=dave.hansen@intel.com \
    --cc=haitao.huang@intel.com \
    --cc=haitao.huang@linux.intel.com \
    --cc=hpa@zytor.com \
    --cc=jethro@fortanix.com \
    --cc=kai.huang@intel.com \
    --cc=linux-sgx@vger.kernel.org \
    --cc=mingo@redhat.com \
    --cc=seanjc@google.com \
    --cc=stable@vger.kernel.org \
    --cc=tglx@linutronix.de \
    --cc=x86@kernel.org \
    /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).