linux-kernel - Re: Linux 3.8-rc1 - another regression on USB :-(

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]

Date:	Fri, 28 Dec 2012 18:12:29 -0500
From:	Woody Suwalski <terraluna977@...il.com>
To:	Greg Kroah-Hartman <gregkh@...uxfoundation.org>
CC:	Linus Torvalds <torvalds@...ux-foundation.org>,
	USB list <linux-usb@...r.kernel.org>,
	Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: Linux 3.8-rc1 - another regression on USB :-(

Woody Suwalski wrote:
> Greg Kroah-Hartman wrote:
>> On Sun, Dec 23, 2012 at 02:35:49PM -0800, Linus Torvalds wrote:
>>> Woody,
>>>   Any chance you can bisect this? It's not going to be hugely pleasant
>>> (with 11k+ commits in between 3.7 and 3.8-rc1 you'll have to compile
>>> and test at least 14 kernels), but it would help enormously. Of
>>> course, maybe some USB person can guess what would cause the device to
>>> go offline..
>>>
>>> Added Greg and the linux-usb mailing list to the participants list:
>>> the images are in the original email on lkml, but there isn't anything
>>> particularly interesting there, it really just seems to be an
>>> unexpected and spurious USB disconnect, resulting in "USB disconnect,
>>> device number 2" followed by "Rejecting I/O to offline device".
>> I don't see any images on lkml, sorry.
>>
>> The kernel log for when the disconnect happened would be great to get.
>>
>> The kernel can't cause a device to disconnect, that's an electrical
>> thing usually, is this perchance a flaky device/connection?  Or has it
>> always worked on older kernels?
>>
>> What host controller is being used here (xhci, ehci, etc.?)
>>
>> thanks,
>>
>> greg k-h
> I backtrack - it is not directly a USB problem.
> If I boot the image into a single mode, and then verify reading the 
> squashfs as a compressed file or the files in the mounted uncompressed 
> image - there are no errors on read.
> Yet if i let it boot into X - it does not - gets bogged with disk 
> access errors.
> And then there is no access to a USB key anymore.
> Will need to investigate further..
>
> Sorry for the alarm, Greg...
>
Couple more observations:
I see the same problem on a Thinkpad T500, and on a Eeepc with GMA500, 
so it is not just i915 systems issue.
There is no guarantee that the single mode will boot, or that it will be 
sane.
Errors seem to appear in the same general range of sectors, but not in 
the same place - more like timing - problem appears when udev / X loads 
extra stuff.
Even if single mode works, the access to USB keys is dog slow - below 
1M/s - normally it is in 30M/s range.

So the 3.8 kernel is sick, and USB may be just a first subsystem to be 
hit by a problem.

IRQ handling? DMA?
Config is more or less the same config converted via "make oldconfig" 
from 3.7 and earlier kernels.
However since nobody else seems to have noticed problems, I will keep 
digging...

Woody




View attachment "config-3.8" of type "text/plain" (101460 bytes)