From: "Rafael J. Wysocki" <rjw@rjwysocki.net>
To: Yinghai Lu <yinghai@kernel.org>
Cc: Bjorn Helgaas <bhelgaas@google.com>,
"Rafael J. Wysocki" <rafael.j.wysocki@intel.com>,
Gu Zheng <guz.fnst@cn.fujitsu.com>,
Guo Chao <yan@linux.vnet.ibm.com>,
"linux-pci@vger.kernel.org" <linux-pci@vger.kernel.org>,
"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
Mika Westerberg <mika.westerberg@linux.intel.com>
Subject: Re: [PATCH v2 04/10] PCI: Destroy pci dev only once
Date: Sun, 01 Dec 2013 02:24:33 +0100 [thread overview]
Message-ID: <3611889.9a0SszCTpv@vostro.rjw.lan> (raw)
In-Reply-To: <CAE9FiQUG-WV26o5MtM=QPfN1kEDGs4Li2wdGB88T==yrxGum_w@mail.gmail.com>
On Saturday, November 30, 2013 02:27:15 PM Yinghai Lu wrote:
> On Sat, Nov 30, 2013 at 1:37 PM, Rafael J. Wysocki <rjw@rjwysocki.net> wrote:
> > On Saturday, November 30, 2013 01:31:33 AM Rafael J. Wysocki wrote:
> >> On Saturday, November 30, 2013 12:45:55 AM Rafael J. Wysocki wrote:
> >> > On Saturday, November 30, 2013 12:38:26 AM Rafael J. Wysocki wrote:
> >> > > On Tuesday, November 26, 2013 06:26:54 PM Yinghai Lu wrote:
> >> > > > On Tue, Nov 26, 2013 at 5:24 PM, Rafael J. Wysocki <rjw@rjwysocki.net> wrote:
> >> > > > >
> >> > > > > So assume pci_destroy_dev() is called twice in parallel for the same dev
> >> > > > > by two different threads. Thread 1 does the atomic_inc_and_test() and
> >> > > > > finds that it is OK to do the device_del() and put_device() which causes
> >> > > > > the device object to be freed. Then thread 2 does the atomic_inc_and_test()
> >> > > > > on the already freed device object and crashes the kernel.
> >> > > > >
> >> > > > thread2 should still hold one extra reference.
> >> > > > that is in
> >> > > > device_schedule_callback
> >> > > > ==> sysfs_schedule_callback
> >> > > > ==> kobject_get(kobj)
> >> > > >
> >> > > > pci_destroy_dev for thread2 is called at this point.
> >> > > >
> >> > > > and that reference will be released from
> >> > > > sysfs_schedule_callback
> >> > > > ==> kobject_put()...
> >> > >
> >> > > Well, that would be the case if thread 2 was started by device_schedule_callback(),
> >> > > but again, for example, it may be trim_stale_devices() started by acpiphp_check_bridge()
> >> > > that doesn't hold extra references to the pci_dev. [Well, that piece of code
> >> > > is racy anyway, because it walks bus->devices without locking. Which is my
> >> > > fault too, because I overlooked that. Shame, shame.]
> >> > >
>
> can you add extra reference to that path?
hotplug_event_work()
hotplug_event()
acpiphp_check_bridge()
trim_stale_devices()
pci_stop_and_remove_bus_device()
Yes, it should hold a reference to dev, but adding it there doesn't really help,
because there are list walks over &bus->devices in acpiphp_check_bridge() and
trim_stale_devices() that are racy with respect to pci_stop_and_remove_bus_device()
run from device_schedule_callback().
> >> > > Perhaps we can do something like the (untested) patch below (in addition to the
> >> > > $subject patch). Do you see any immediate problems with it?
> >> >
> >> > Ah, I see one. It will break pci_stop_bus_device() and pci_remove_bus_device().
> >> > So much for being clever.
> >> >
> >> > Moreover, it looks like those two routines above are racy too for the same
> >> > reason?
> >>
> >> The (still untested) patch below is what I have come up with for now. The
> >> is_gone flag is now only operated under pci_remove_rescan_mutex, so it need
> >> not be atomic. Of course, whoever calls pci_stop_and_remove_bus_device()
> >> (the "locked" one) should hold a ref to the device being removed to avoid
> >> use-after-free (the callers need to be audited for that).
>
> if you can use device_schedule_...,
No, I can't. I need to hold acpi_scan_lock taken in hotplug_event_work()
throughout all bus trimming/scanning and I need to protect list walks over
&bus->devices too.
> should have hold reference may be
> atomic would be better than lock/unlock everywhere?
The locking is necessary not only for the device removal itself, but also for
the safety of the &bus->devices list walks.
Besides, remove_callback() in remove.c already holds pci_remove_rescan_mutex
around pci_stop_and_remove_bus_device() and I don't see how it would be safe
to run pci_stop_and_remove_bus_device() without holding that mutex from
anywhere else.
For one example, pci_stop_and_remove_bus_device() that is not run under
pci_remove_rescan_mutex can race with the stuff called under that mutex
in dev_bus_rescan_store() (and elsewhere in pci-sysfs.c).
So either pci_remove_rescan_mutex is useless and should be dropped, or
it is there for a purpose, in which case it needs to be used around
pci_stop_and_remove_bus_device() everywhere. There's no other possibility
and to my eyes that mutex is necessary.
Thanks,
Rafael
next prev parent reply other threads:[~2013-12-01 1:11 UTC|newest]
Thread overview: 69+ messages / expand[flat|nested] mbox.gz Atom feed top
2013-11-26 1:28 [PATCH v2 00/10] PCI: Double removing fix and allocate 64bit mmio pref Yinghai Lu
2013-11-26 1:28 ` [PATCH v2 01/10] PCI: Use device_release_driver in pci_stop_root_bus Yinghai Lu
2013-11-27 1:09 ` Rafael J. Wysocki
2013-11-26 1:28 ` [PATCH v2 02/10] PCI: Move back pci_proc_attach_devices calling Yinghai Lu
2013-11-26 1:28 ` [PATCH v2 03/10] PCI: Move resources and bus_list releasing to pci_release_dev Yinghai Lu
2013-11-27 1:15 ` Rafael J. Wysocki
2013-11-27 2:15 ` Yinghai Lu
2013-11-26 1:28 ` [PATCH v2 04/10] PCI: Destroy pci dev only once Yinghai Lu
2013-11-26 3:38 ` Bjorn Helgaas
2013-11-26 19:34 ` Yinghai Lu
2013-11-26 20:13 ` Yinghai Lu
2013-11-27 1:24 ` Rafael J. Wysocki
2013-11-27 2:26 ` Yinghai Lu
2013-11-29 23:38 ` Rafael J. Wysocki
2013-11-29 23:45 ` Rafael J. Wysocki
2013-11-30 0:31 ` Rafael J. Wysocki
2013-11-30 21:37 ` Rafael J. Wysocki
2013-11-30 22:27 ` Yinghai Lu
2013-12-01 1:24 ` Rafael J. Wysocki [this message]
2013-12-02 1:29 ` Rafael J. Wysocki
2013-12-02 14:49 ` Rafael J. Wysocki
2013-12-05 22:40 ` Bjorn Helgaas
2013-12-06 1:21 ` Rafael J. Wysocki
2013-12-06 6:29 ` Yinghai Lu
2014-01-10 14:20 ` [PATCH 0/9] PCI: Eliminate race conditions between hotplug and sysfs rescan/remove (Was: Re: [PATCH v2 04/10] PCI: Destroy pci dev only once) Rafael J. Wysocki
2014-01-10 14:22 ` [PATCH 1/9] PCI: Global rescan-remove lock Rafael J. Wysocki
2014-01-10 14:23 ` [PATCH 2/9] ACPI / PCI: Use global PCI rescan-remove locking in PCI root hotplug Rafael J. Wysocki
2014-01-10 14:24 ` [PATCH 3/9] ACPI / hotplug / PCI: Use global PCI rescan-remove locking Rafael J. Wysocki
2014-01-10 14:25 ` [PATCH 4/9] PCMCIA / cardbus: " Rafael J. Wysocki
2014-01-10 14:26 ` [PATCH 5/9] PCI / hotplug: " Rafael J. Wysocki
2014-01-10 14:27 ` [PATCH 6/9] platform / x86: " Rafael J. Wysocki
2014-01-10 14:27 ` [PATCH 7/9] MPT / PCI: Use pci_stop_and_remove_bus_device_locked() Rafael J. Wysocki
2014-01-10 14:28 ` [PATCH 8/9] powerpc / eeh_driver: Use global PCI rescan-remove locking Rafael J. Wysocki
2014-01-15 13:36 ` [Update][PATCH " Rafael J. Wysocki
2014-01-15 17:38 ` Bjorn Helgaas
2014-01-10 14:29 ` [PATCH 9/9] Xen / PCI: " Rafael J. Wysocki
2014-01-15 18:02 ` [PATCH 0/9] PCI: Eliminate race conditions between hotplug and sysfs rescan/remove (Was: Re: [PATCH v2 04/10] PCI: Destroy pci dev only once) Bjorn Helgaas
2013-12-06 6:52 ` [PATCH v2 04/10] PCI: Destroy pci dev only once Yinghai Lu
2013-12-07 1:27 ` Rafael J. Wysocki
2013-12-08 3:31 ` Yinghai Lu
2013-12-08 3:50 ` Greg Kroah-Hartman
2013-12-09 15:24 ` Ethan Zhao
2013-12-09 19:08 ` Greg Kroah-Hartman
2013-12-10 7:43 ` Ethan Zhao
2014-01-13 1:03 ` [PATCH] PCI / remove: Check parent kobject in pci_destroy_dev() (was: Re: [PATCH v2 04/10] PCI: Destroy pci dev only once) Rafael J. Wysocki
2013-11-27 1:17 ` [PATCH v2 04/10] PCI: Destroy pci dev only once Rafael J. Wysocki
2013-11-26 1:28 ` [PATCH v2 05/10] PCI: pcibus address to resource converting take bus directly Yinghai Lu
2013-11-26 1:28 ` [PATCH v2 06/10] PCI: Add pcibios_bus_addr_to_res() Yinghai Lu
2013-11-26 1:28 ` [PATCH v2 07/10] PCI: Try to allocate mem64 above 4G at first Yinghai Lu
2013-11-26 4:15 ` Bjorn Helgaas
2013-11-26 20:14 ` Yinghai Lu
2013-11-26 1:28 ` [PATCH v2 08/10] PCI: Try best to allocate pref mmio 64bit above 4g Yinghai Lu
2013-11-26 4:17 ` Bjorn Helgaas
2013-11-26 6:59 ` Guo Chao
2013-11-26 17:53 ` Bjorn Helgaas
2013-11-26 22:00 ` Yinghai Lu
2013-11-26 22:01 ` Bjorn Helgaas
2013-11-27 0:33 ` Yinghai Lu
2013-11-26 1:28 ` [PATCH v2 09/10] PCI: Sort pci root bus resources list Yinghai Lu
2013-11-26 4:18 ` Bjorn Helgaas
2013-11-26 1:28 ` [PATCH v2 10/10] intel-gtt: Read 64bit for gmar_bus_addr Yinghai Lu
2013-11-26 3:46 ` Bjorn Helgaas
2013-11-26 19:35 ` Yinghai Lu
2013-12-11 18:48 ` Bjorn Helgaas
2013-12-11 19:58 ` Yinghai Lu
2013-12-21 0:27 ` Bjorn Helgaas
2013-12-21 1:19 ` Yinghai Lu
2013-12-21 18:50 ` Bjorn Helgaas
2013-12-23 22:33 ` Bjorn Helgaas
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=3611889.9a0SszCTpv@vostro.rjw.lan \
--to=rjw@rjwysocki.net \
--cc=bhelgaas@google.com \
--cc=guz.fnst@cn.fujitsu.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-pci@vger.kernel.org \
--cc=mika.westerberg@linux.intel.com \
--cc=rafael.j.wysocki@intel.com \
--cc=yan@linux.vnet.ibm.com \
--cc=yinghai@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
all inboxes | Powered by JetHome®