From: Yijing Wang <wangyijing@huawei.com>
To: Bjorn Helgaas <bhelgaas@google.com>
Cc: <linux-pci@vger.kernel.org>, Yijing Wang <wangyijing@huawei.com>,
<stable@vger.kernel.org.#3.4+>
Subject: [PATCH] PCI/ASPM: Don't remove pcie_link_state until we stop the last device
Date: Thu, 30 Jul 2015 12:09:20 +0800 [thread overview]
Message-ID: <1438229360-370-1-git-send-email-wangyijing@huawei.com> (raw)
Now we stop the pci_bus->devices in reverse order, but in
pcie_aspm_exit_link_state(), we only would do something when
the device is the last one.
void pcie_aspm_exit_link_state(struct pci_dev *pdev)
{
...
if (!list_is_last(&pdev->bus_list, &parent->subordinate->devices))
goto out;
...
}
So if we have the following pcie tree, system may crash.
[b7-bd]--+-02.0-[bb-bd]--+-00.0-[bc-bd]----01.0-[bd]----00.0 PLX Technology, Inc. Device 0002
+-00.1 PLX Technology, Inc. Device 0002
+-00.2 PLX Technology, Inc. Device 0002
+-00.3 PLX Technology, Inc. Device 0002
\-00.4 PLX Technology, Inc. Device 0002
In this case, we would stop bb:00.4 before bb:00.0, so when we touch bb:00.4,
we would call pcie_aspm_exit_link_state(), and free the pcie_link_state.
So when we want to stop bd:00.0 and free related pcie_link_state,
it would try to access the parent pcie_link_state which has been freed.
Part crash call trace:
BUG: unable to handle kernel NULL pointer dereference at 0000000000000018
CPU 16 Pid: 33262, comm: IVS_PowerOn
RIP: 0010:[<ffffffffa0d7c14f>] [<ffffffffa0d7c14f>] pcie_config_aspm_link+0x3f/0x100
RSP: 0018:ffff8801bc577790 EFLAGS: 00010282
RAX: 0000000000000000 RBX: 0000000000000001 RCX: 000000000000e7e6
RDX: 000000000000e6e6 RSI: 00000000ffffc5ec RDI: 0000000000000246
RBP: ffff8801bc5777d0 R08: ffff88007b001000 R09: 00000000003fffff
...
Call Trace:
[<ffffffff8124a542>] pcie_config_aspm_path+0x32/0x60
[<ffffffffa0d7cc00>] pcie_aspm_exit_link_state+0x160/0x560
[<ffffffffa0d7c0bc>] pci_stop_bus_device+0x8c/0xe0
[<ffffffffa0d7c068>] pci_stop_bus_device+0x38/0xe0
[<ffffffffa0d7c068>] pci_stop_bus_device+0x38/0xe0
[<ffffffffa0d7c068>] pci_stop_bus_device+0x38/0xe0
[<ffffffffa0d7c068>] pci_stop_bus_device+0x38/0xe0
[<ffffffff8123eca1>] pci_stop_and_remove_bus_device+0x11/0x20
...
Signed-off-by: Yijing Wang <wangyijing@huawei.com>
CC: stable@vger.kernel.org #3.4+
---
drivers/pci/pcie/aspm.c | 3 ++-
1 files changed, 2 insertions(+), 1 deletions(-)
diff --git a/drivers/pci/pcie/aspm.c b/drivers/pci/pcie/aspm.c
index 317e355..c81f549 100644
--- a/drivers/pci/pcie/aspm.c
+++ b/drivers/pci/pcie/aspm.c
@@ -648,7 +648,8 @@ void pcie_aspm_exit_link_state(struct pci_dev *pdev)
* All PCIe functions are in one slot, remove one function will remove
* the whole slot, so just wait until we are the last function left.
*/
- if (!list_is_last(&pdev->bus_list, &parent->subordinate->devices))
+ if (!(pdev == list_first_entry(&parent->subordinate->devices,
+ struct pci_dev, bus_list)))
goto out;
link = parent->link_state;
--
1.7.1
next reply other threads:[~2015-07-30 4:16 UTC|newest]
Thread overview: 5+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-07-30 4:09 Yijing Wang [this message]
2015-08-29 12:20 ` [PATCH] PCI/ASPM: Don't remove pcie_link_state until we stop the last device Bjorn Helgaas
2015-08-31 1:24 ` wangyijing
2015-08-31 13:56 ` Bjorn Helgaas
2015-09-01 0:59 ` wangyijing
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=1438229360-370-1-git-send-email-wangyijing@huawei.com \
--to=wangyijing@huawei.com \
--cc=bhelgaas@google.com \
--cc=linux-pci@vger.kernel.org \
--cc=stable@vger.kernel.org.#3.4+ \
/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).