Re: [gentoo-user] USB Problems

[Volker Armin Hemmann <volkerarmin@googlemail.com>]

On Sunday 12 June 2011 22:30:30 john wrote:
> On Sun, 12 Jun 2011 21:20:48 +0200
> 
> meino.cramer@gmx.de wrote:
> > john <jdm@arcticwolf.myzen.co.uk> [11-06-12 20:52]:
> > > On Sun, 12 Jun 2011 16:45:38 +0200
> > > 
> > > meino.cramer@gmx.de wrote:
> > > > john <jdm@arcticwolf.myzen.co.uk> [11-06-12 16:36]:
> > > > > Gents
> > > > > 
> > > > > Whenever I insert external USB device my machine locks and I
> > > > > have to do a hard reset (sys req does not even work). I have
> > > > > tried this in a graphical environment and without. Machine
> > > > > boots up fine with USB keyboard and mouse but will not boot
> > > > > up
> > > > > with usb storage (memory stick
> > > > > - tried several)
> > > > > 
> > > > > I have recently upgrade kernel to 2.6.38-r6 when this error
> > > > > started (approx). Hav also tried 2.6.38-r7 (which according
> > > > > to
> > > > > bugzilla should fix an issue similar to this) the error
> > > > > still
> > > > > happens.
> > > > > 
> > > > > Any suggestions on kernel config options, debugging would be
> > > > > appreciated. And how to trap full error message.
> > > > > 
> > > > > Error message alongs these lines
> > > > > 
> > > > > Kernel panic not syncing - fatal exception in iterrupt
> > > > > ksoftirqd/1 Not tainted
> > > > 
> > > > Hi John,
> > > > 
> > > > may be this could help to track th eerror more closely:
> > > > Recompile the kernel with CONFIG_USB_DEBUG set.
> > > > If not already done, enable CONFIG_MAGIC_SYSRQ and
> > > > read about the available commands in the according
> > > > docs under /usr/src/linux/Documentation/.
> > > > This is a overall life saver in many other cases too.
> > > > But it may be a security hole also -- you can disable
> > > > it in /proc/ later.
> > > > If possible change /etc/fstab to mount the partition
> > > > to which logging will go with the "sync" option so
> > > > every log entry will go directly onto the bare metal
> > > > instead of a RAM buffer.
> > > > 
> > > > Reboot into console mode (no X).
> > > > cd to the logging directory and
> > > > do a tail -f onto the logfile in question
> > > > 
> > > > Insert the "USB stick of death" ;) and
> > > > see, whether you can see anything in the
> > > > tailed log.
> > > > 
> > > > If not do the magic sysrq dance to sync hds,
> > > > remount readonly and finally reboot.
> > > > 
> > > > Take a closer look into the resulting logs.
> > > > 
> > > > Good luck!
> > > > 
> > > > Nevertheless have a nice weekend!
> > > > Best regards,
> > > > mcc
> > > 
> > > Thanks mmc,
> > > CONFIG_MAGIC_SYSRQ and CONFIG_USB_DEBUG set to y
> > > 
> > > When I plug in USB device SYSREQ does not work. I have tested when
> > > system is normal and all ok with SYSREQ so USB device is locking up
> > > system completely.
> > > 
> > > I have put sync in mount table but when device is inserted error log
> > > is not being produced (in /var/log/messages, etc). I may end up
> > > having to take a photo of error log(lol). Or there maybe some other
> > > way of catching error which Gentoo gurus may advise.
> > > 
> > > I have tried 3 memory sticks and they all lock system up. These work
> > > on another PC using Arch and work on this PC dual booting with
> > > Windows. I have a usb hard drive which works ok so it is a memory
> > > stick only issue!!!!????
> > > 
> > > Hmmm.
> > 
> > Hi John,
> > 
> > if sysreq does not work after inserting the usbstick then the kernel
> > stops working completly (the sysreq-code of the kernel is not
> > entered), which seems to be that something happening on very low
> > level (the problem not the sysreq-code of course... ;)
> > 
> > Just another shot into the dark:
> > Do you have modularized all usb-related stuff? This also may help to
> > let the usb-sticks work correctly.
> > 
> > Ah! Another thing: Insert the related modules into
> > /etc/conf.d/modules so that they get loaded in a very early
> > phase of the bootprocess.
> > 
> > With modules loaded and with no usb-sticks as root inspect the
> > according modules with modinfo -p <modulename> to see, whether
> > there a additional parameters you may pass to the modules to get
> > an extra of informations or - better - to fix the issue.
> > 
> > Modify under "Kernel hacking" in the kernel config the
> > "Default message log level" to see more.
> > 
> > Furthermore there is a switch "Kernel debugging" with sub-options.
> > May be turning on these may also help to get closer to all evil ;)
> > 
> > Last thing:
> > If there is nothing valueable stored on the usbsticks:
> > Try to reformat the usbsticks. Instead of putting a FAT32-fs on the
> > bare device, create a regular partition and create a Linux-fs
> > (reiserfs, ext2/3/4 or like that) on that partition.
> > 
> > This may not be good idea in general, but - if it works - this would
> > give a hint, where to search next.
> > 
> > Good luck!
> > 
> > Best regards,
> > mcc
> 
> Thanks Meino,
> Have recompliled USB as modules (ehci-hcd, ohci-hcd and usb-storage and
> load with /etc/conf.d/modules but still locks ups. Hope these are the
> right ones. I'll try scsi next.
> 
> Also tried fdisk Linux file system and format ext3. But all to no avail.
> Still locking up.
> 
> Have configure kernel hacking options as suggested.
> 
> Error message as follows:-
> 
> --------------------------------------------------------
> Bug: unable to handle kernel null pointer dereference at
> 0000000000000048
> 
> IP [<ffffffff811ddb8e>]elv_queue_empty+0x12/0x27
> 
> Oops 000 [#1] SMP
> 
> last sys file: /sys/devices/virtual/bdi/8:32/uevent
> 
> Pid: 1-, comm: ksoftirqd/1 Tainted: P 2.6.38-gentoo-r7 #4 MICRO-STAR
> MS-7388
> 
> RIP 0010[<ffffffff811ddb8e7>] elv-queue_empty+0x12/0x27
> --------------------------------------------------------
> 
> 
> From my very limited knowledge, I would suspect kernel problem (ksoft)
> and irq (software interrupt perhaps). Please put me right if I'm out of
> order here. I'm Linux noobie
> 
> 
> Think I'll drop back to an older kernel to see if I'm still getting the
> issue as this may elliminate hardware issue.
> 
> Unsure how to pass options to modules when loading. is this just done
> through /etc/conf.d/modules
> 
> I like the error messages though Oops, RIP (someone has a sense of
> humour)
> 
> 
> Thanks for you help
> 
> At least you learn while debgging

on lkml was just a reminder email of a bugzilla entry: panic on usb hdd.

So - it is a kernel bug. You will probably not find the culprit just building 
models.

Use vanilla sources, do a git bisect, report findings on lkml.
-- 
#163933