Message-ID: <20101202175755.GA2119@xanatos>
Date:	Thu, 2 Dec 2010 09:57:55 -0800
From:	Sarah Sharp <sarah.a.sharp@...ux.intel.com>
To:	"VanCutsem, Geoffroy" <geoffroy.vancutsem@...el.com>,
	Alexander Viro <viro@...iv.linux.org.uk>
Cc:	linux-fsdevel@...r.kernel.org, linux-kernel@...r.kernel.org,
	linux-usb@...r.kernel.org
Subject: RE: Problem with USB on Lenovo W510

Hi Geoffroy,

None of the logs you forwarded seem to show a full kernel crash.  Is
there anyway you can use netconsole to capture the dmesg during a crash?
I have a tutorial here if you need guidance:

http://sarah.thesharps.us/2010-03-26-09-41

I'm especially interested in the case where you said you plugged in a
bus-powered hub and the machine crashed.


I do see a "hung process" message in one of the files.  I'm not a block
layer expert, so I'm forwarding it onto lkml and linux-fsdevel.

Background: Geoffroy is running 2.6.35.8 and trying to dd a file to a
USB flash drive.  The drive occasionally disconnects, and sometimes that
causes a "hung process" warning.  E.g.

[ 3600.458402] INFO: task udisks-daemon:1854 blocked for more than 120 seconds.
[ 3600.458409] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 3600.458414] udisks-daemon D 000000010004ce77     0  1854   1845 0x00000000
[ 3600.458423]  ffff88022dec7bb8 0000000000000086 ffff880200000000 00000000000158c0
[ 3600.458432]  ffff88022dec7fd8 00000000000158c0 ffff88022dec7fd8 ffff88021d5f5b80
[ 3600.458442]  00000000000158c0 00000000000158c0 ffff88022dec7fd8 00000000000158c0
[ 3600.458449] Call Trace:
[ 3600.458460]  [<ffffffff8115e510>] ? __pollwait+0x0/0xf0
[ 3600.458470]  [<ffffffff8157a6d7>] __mutex_lock_slowpath+0xf7/0x180
[ 3600.458477]  [<ffffffff8129a3f0>] ? exact_match+0x0/0x10
[ 3600.458484]  [<ffffffff8157a5bb>] mutex_lock+0x2b/0x50
[ 3600.458490]  [<ffffffff8117d3a8>] __blkdev_get+0x78/0x410
[ 3600.458495]  [<ffffffff8117d750>] blkdev_get+0x10/0x20
[ 3600.458500]  [<ffffffff8117d85a>] blkdev_open+0xfa/0x130
[ 3600.458505]  [<ffffffff8117d760>] ? blkdev_open+0x0/0x130
[ 3600.458512]  [<ffffffff8114b0f5>] __dentry_open+0xe5/0x330
[ 3600.458519]  [<ffffffff81258ebf>] ? security_inode_permission+0x1f/0x30
[ 3600.458525]  [<ffffffff8114b454>] nameidata_to_filp+0x54/0x70
[ 3600.458531]  [<ffffffff81159100>] do_last+0x3b0/0x6a0
[ 3600.458537]  [<ffffffff8115b39b>] do_filp_open+0x21b/0x660
[ 3600.458545]  [<ffffffff813a8386>] ? scsi_disk_put+0x46/0x60
[ 3600.458552]  [<ffffffff8116640a>] ? alloc_fd+0x10a/0x150
[ 3600.458557]  [<ffffffff8114aec9>] do_sys_open+0x69/0x140
[ 3600.458563]  [<ffffffff8114afe0>] sys_open+0x20/0x30
[ 3600.458571]  [<ffffffff8100a072>] system_call_fastpath+0x16/0x1b

Is this a known bug?

Geoffroy, the other thing you can try is upgrading to 2.6.35.9 or
2.6.36.1.  That probably won't cause the disconnects to go away, but it
might cause the crashes to stop.  If you continue to see disconnects,
it's probably something electrically wrong with your USB hosts.  You
might want to talk with your hardware vendor and/or get the laptop
replaced.

Sarah Sharp

----- Forwarded message from "VanCutsem, Geoffroy" <geoffroy.vancutsem@...el.com> -----

Date: Fri, 5 Nov 2010 04:38:09 -0700
From: "VanCutsem, Geoffroy" <geoffroy.vancutsem@...el.com>
To: "Sharp, Sarah A" <sarah.a.sharp@...el.com>
CC: "linux-usb@...r.kernel.org" <linux-usb@...r.kernel.org>
Subject: RE: Problem with USB on Lenovo W510

Hi Sarah,

I have ran more tests today (with debugging turned on) -- log files attached. The results are not always 100% consistent but at least I have seen the crash occuring on multiple occasions and with both the USB 2.0 and USB 3.0 ports. I am always testing with the exact same iso (to take that one variable out of the equation) but working with 2 different USB sticks. Here are the two models I am using (just in case it could be relevant):
- Kingston DataTraveler 1GB (DTI/1GB) -- log files which pertain to this key have the '-wk' extension because it's a white key.
- PQI Cool Drive U339 1GB -- log files which pertain to this key have the '-bk' extension because it's a blue key.

The blue key (PQI) was the one that seemed to trigger the crashes most frequently.

Just to clarify how I am running these tests:
- Using a 2.6.35.8 kernel from kernel.org with the standard 2.5.35 Ubuntu config file + USB debugging turned on (as per your request)
- I always reboot the machine before the next experiment if I see a crash.
- I always unmount the USB stick before attempting 'dd' (using 'sudo umount /media/my_usb_stick_name')
- I have not yet tried to physically unplug the drive in the middle of a 'dd' session but that's an easy experiment to do if it makes sense.

I hope these logs will help you understand what is going on on my system.

Thanks!
Geoffroy

> -----Original Message-----
> From: Sharp, Sarah A
> Sent: Thursday, November 04, 2010 7:40 PM
> To: VanCutsem, Geoffroy
> Cc: linux-usb@...r.kernel.org
> Subject: Re: Problem with USB on Lenovo W510
> 
> On Thu, Nov 04, 2010 at 09:32:07AM -0700, VanCutsem, Geoffroy wrote:
> > Hi Sarah,
> >
> > > -----Original Message-----
> > > From: Sharp, Sarah A
> > >
> > > There may be both EHCI and xHCI host controllers in the same
> system.
> > > xHCI handles all speeds of USB devices plugged into it.  The odd
> thing
> > > is your dmesg on the bug report you linked to shows that the xHCI
> > > driver
> > > is handling your storage device, but your lspci output doesn't show
> it!
> >
> > Wasn't it reflected by this entry:
> > 0f:00.0 USB Controller: NEC Corporation Device 0194 (rev 03) (prog-if
> 30)
> 
> Oh, sorry, I guess I missed that.
> 
> > I have done this today and the interesting thing is that with this
> vanilla 2.6.35.8 kernel (from kernel.org) and debugging turned on, the
> problem seems to have gone away. I have been able to 'dd' the exact
> same image onto a USB stick (tried twice, once from a USB3 port and
> once from the 'normal' port)... so perhaps this is a bug in
> Ubuntu/Canonical variant of 2.6.35?? I have attahched the relevant
> dmesg / lspci output to this email.
> 
> Well, I'm not so sure you actually triggered the bug.  The previous log
> showed the crash after the USB device disconnected in the middle of the
> dd transfer.  This log doesn't show a disconnect during a transfer
> under
> the USB 3.0 port; it only shows a disconnect when you switched the
> device over to the USB 2.0 port.
> 
> You can probably trigger the bug if you physically pull out the device
> while you're in the middle of the dd.  Can you try that with the
> 2.6.35.8 kernel?
> 
> Another possibility is the USB core and xHCI debugging is causing a
> delay that's hiding the bug.  You can also try turning off debugging on
> the 2.6.35.8 kernel and dding a iso over, if yanking out the device
> doesn't cause the bug.
> 
> Sarah Sharp





----- End forwarded message -----

View attachment "dmesg.USB2.log" of type "text/plain" (173575 bytes)

Powered by blists - more mailing lists