linux-kernel - Re: linux-3.4-rc5: Oops in klist

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <4F9F0AF3.8060409@fold.natur.cuni.cz>
Date:	Mon, 30 Apr 2012 23:58:11 +0200
From:	Martin Mokrejs <mmokrejs@...d.natur.cuni.cz>
To:	linux-pci@...r.kernel.org, Yinghai Lu <yinghai@...nel.org>,
	Sarah Sharp <sarah.a.sharp@...ux.intel.com>,
	LKML <linux-kernel@...r.kernel.org>
Subject: Re: linux-3.4-rc5: Oops in klist_del()

Martin Mokrejs wrote:
> Hi,
>   I was removing an ExpressCard with a NEC chipset used by xhci_hcd and the kernel crashed
> on my Dell Vostro 3550 laptop. Its express slot chipset somehow does not report properly
> that a card was unplugged, as discussed earlier here on linux-pci under Subject
> "Re: 3.2.11: PCI Express card cannot be re-detected withing cca 60sec timeframe".
> 
>   Maybe this helps somebody to find the cause? Here is a bit of the attached Oops snapshot
> stacktrace:
> 
> klist_del
> fdevice_del
> device_unregister
> pci_stop_bus_device
> pci_stop_and_remove_bus_device
> devehp_unconfigure_device
> pciehp_disable_slot
> ? pciehp_disable_slot
> pciehp_power_thread
> process_one_work
> worker_thriad
> ? manage_workers.clone
> ? manage_workers.clone
> kthrerd
> ...

This is 100% reproducible. It happens with 3.4-rc5 and 3.3.4 kernels while not with 3.2.14.
I did not have the external enclosure earlier but repeated unplugs/re-plugs of the NEC-based
expresscard were/are just not enough to trigger the Oops. One needs the ASMedia2105 in addition.

To hit the Ooops, the disk has to be *RE-detected* and unplugged (so not its very first plugin).
Express card removal after the very first plugin says usually something about the surprise
removal. When I re-plug the disk, it gets detected, and it removal hits the Ooops. I am attaching
dmesg from 3.2.14 kernel which does not have the problem. You can see how many times I did remove
and re-plug the card. If I waited long ebough for xhci_hcd to timeout waiting for the unplugged
device, it says

xhci_hcd 0000:11:00.0: Timeout while waiting for configure endpoint command
xHCI xhci_drop_endpoint called for root hub
xHCI xhci_check_bandwidth called for root hub
xhci_hcd 0000:11:00.0: USB bus 5 deregistered
xhci_hcd 0000:11:00.0: PCI INT A disabled

and then, subsequent re-plug of the card with the disk attached results in detected SCSI device.
In contrast, if I re-plug the express card earlier express card slot change is reported but the
disc is not detected (something is blocked in xhci_hcd I think).

The above stacktrace (of actually the camera snapshot is not the first stacktrace, unfortunately).
I see there are 2-3 screen scrolled away. Hmm, but I do not have anything what could be used as
a remote console ...

Martin

View attachment "dmesg-3.2.14-replugs-of-NEC_uPD720200.txt" of type "text/plain" (94320 bytes)