Latest working kernel version: none, the system is new Earliest failing kernel version: 2.6.25.15 Distribution: Gentoo Hardware Environment: Asus M2A-VM with AMD 690G/ATI SB600 chipset Software Environment: Gnome 2.22.2 Problem Description: the ehci-hcd driver randomly fails with loss of high-speed USB functionality. The problem appears more often during copy of big file (>1Gb) to Mass Storage Device, but can happen also during normal operation. When it happens, for several minutes no USB peripheral is available, then the ohci-hci takes over and MSC is available at full-speed only. The problem happens in the attached syslog, where a file transfer to the USB was in progress, suddenly at 14:04:12 the USB mouse stops functioning. The bug was initially reported to the gentoo-amd64 mailing list and is viewable at http://www.nabble.com/USB-2.0-falling-back-to-full-speed-td19387197.html Steps to reproduce: insert USB key, start large file copy to the key, after some time the copy hangs and USB mouse stops responding
Created attachment 17901 [details] syslog captured when problem happens The problem starts at time 14:04:12. The USB was compiled with CONFIG_USB_DEBUG.
Created attachment 17902 [details] cat /proc/cpuinfo output
Created attachment 17903 [details] cat /proc/iomem output
Created attachment 17904 [details] cat /proc/ioports output
Created attachment 17905 [details] lspci -vvv output
Created attachment 17906 [details] lsusb output The lsusb was taken after the problem happened. There is a USB key attached to the Bus #2, but it is not shown after the problem happens. The problem happens from the USB ports located on the front of the case but also from the ports on the back of the case.
Created attachment 17907 [details] cat /proc/scsi/scsi output
Created attachment 17908 [details] /usr/src/linux/scripts/ver_linux output
Created attachment 17909 [details] kernel config file
This is a known problem, there is a bug in the chipset :( Can you try out 2.6.27-rc6, it should be resolved there.
Thanks. I will try it next Moday and come back to you, now I am on a low bandwidth connection.
2.6.27-rc6 does not fix the problem. In the attached log, as soon as I plug in a USB key I loose the USB mouse. The syslog shows several attmpts to start the ehci_hcd, then again it falls back to ohci-hcd. # uname -a Linux louie 2.6.27-rc6 #1 SMP Tue Sep 23 18:14:29 CEST 2008 x86_64 AMD Athlon(tm) 64 X2 Dual Core Processor 5600+ AuthenticAMD GNU/Linux
Created attachment 17954 [details] syslog with kernel 2.6.27-rc6 The USB key was insterted at time 18:23:30. At this time, USB mouse stops responding.
Any other test I can make? I'd like to help getting this fixed, otherwise this brand new motherboard is not much useful for me.
On Wed, Sep 24, 2008 at 01:09:16PM -0700, bugme-daemon@bugzilla.kernel.org wrote: > Any other test I can make? I'd like to help getting this fixed, otherwise > this > brand new motherboard is not much useful for me. Hm, I don't know, have you tried contacting AMD? They have been helpful in resolving the other issue that this chipset had.
I contacted AMD and Asus, in the meanwhile I maybe found the cause. I have a Philips LCD monitor with a USB hub embedded, it is visible in lsusb.txt as: Bus 001 Device 002: ID 04b4:6560 Cypress Semiconductor Corp. CY7C65640 USB-2.0 "TetraHub" I was using this hub for the USB mouse and HP scanner (not for the mass storage devices). I detached the monitor's USB connector and attached the mouse directly to the MB connectors leaving the scanner unconnected for now. I tried several file copy to USB key and HD and it seems to work fine, at high-speed, on 2.6.27-rc6. I'll do some more testing and post the results. If you need some test or detail on the monitor's hub let me know.
Created attachment 18187 [details] dmesg log during the problematic behavior I use USB mouse and keyboard plugged into a small 4-port USB hub, which describes itself as "ID 05e3:0608 Genesys Logic, Inc. USB-2.0 4-Port HUB". When I plug in a USB flash stick (directly into a port on the motherboard, not into the hub), either keyboard, or mouse, or both become completely frozen (keyboard, sometimes, with some key pressed down and repeating). I am attaching the log of dmesg messages taken during a case of buggy behavior. In this log: - At 288s, I plug in the USB stick; not copying anything to or from the stick (not even mounting it); - Some time after that, the keyboard stops working; - At 435s, I unplug the keyboard from USB hub; - At 441s, I replug the keyboard back into USB hub; keyboard works; - At about 600s, keyboard again stops responding completely, meanwhile the system thinks that "right arrow" key is being held down; - At 633s, I unplug the keyboard from USB hub; - At 639s, I replug the keyboard back into USB hub; keyboard works; - At 658s, I unplug the flash stick from the PC. After removing the flash stick, keyboard and mouse work absolutely fine for many hours. The problem is reproducible with any flash stick simply plugged into any port on the motherboard (no data exchange at high speeds or no data exchange at all seem to be required to trigger the bug). (The used kernel version is 2.6.26.)
After pluggin in all the peripherals directly into the ports on the motherboard, the USB operation seems to be much more (or completely?) stable. So it's probably related to SB600's general handling of USB hubs, because the original submitter's problem was also solved by removing the hub. However, it should be noted that in Microsoft Windows the same hardware works for me perfectly even *with* the USB hub. Also, I think this bug might be related to another one: http://bugzilla.kernel.org/show_bug.cgi?id=11456
i can confirm, that removing the usb hub solves my problems as well. unfortunately this is no real option.
Attaching devices through an USB hub continues to cause the described issue as of 2.6.27.3. Operation without the hub confirmed to be completely stable.
I have the same hub and SB600, will try removing it later to see if that solves the problem.
I see the status on this bug is "Needinfo" -- what other info can we provide? I would also like to help get this fixed. I can confirm this bug exists in at least these kernels on Fedora 9: kernel-2.6.26.5-45.fc9.x86_64 kernel-2.6.26.6-79.fc9.x86_64 kernel-2.6.27.5-37.fc9.x86_64 Either of these hubs will cause it: ID 05e3:0608 Genesys Logic, Inc. USB-2.0 4-Port HUB 0409:005a NEC Corp. HighSpeed Hub See also my bug report at the Fedora project: https://bugzilla.redhat.com/show_bug.cgi?id=472326
This might be the fix for this problem: http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=b09bc6cbae4dd3a2d35722668ef2c502a7b8b093 However, it only applies itself on SB700, maybe because the patch author probably did not know that SB600 are affected too. I did not test it yet.
Created attachment 18965 [details] Add the SB600 device as needing the USB hang workaround And indeed, after applying the attached patch on top of 2.6.28-rc6, I could do a high-speed transfer from an USB flash and have no problems with other devices. This is with the Genesys USB 2.0 4-port hub plugged in. $ dd if=/dev/sdb of=/dev/null bs=1M 3911+1 records in 3911+1 records out 4102028288 bytes (4.1 GB) copied, 153.569 s, 26.7 MB/s Before the patch, this would lock up the USB keyboard and/or mouse within a couple of seconds. Perhaps the workaround needs more refinement with regard to which devices exactly need it to be applied. The original author of the solution hardcoded 4396 (SB700 USB 2.0) specifically with 3a and 3b revisions of SMBus. I added the device 4386 (SB600 USB 2.0) without any SMBus revision checking. This might be too broad, I don't know. Feel free to further test/modify/improve it.
Thanks for reporting this bug and providing the testing patch, we will check this issue on SB600 and submit one formal workaround patch if necessary.
I can confirm that Roman Mamedov's patch for the SB600 works to fix this problem when applied to the 2.6.27 kernel, after first applying Andiry Xu's patch for the SB700.
The patch for SB600 has been accepted by Greg, here is the temp link, which will be merged to linus source tree soon: http://www.kernel.org/pub/linux/kernel/people/gregkh/gregkh-2.6/patches/usb.current/usb-fix-sb600-usb-subsystem-hang-bug.patch Thanks,
Roman's fix (id=18965) works for me on the original hw setup (Hub embedded in Philips monitor connected to the SB600-based motherboard). The behavior is now that during a file transfer to a key directly attached to the motherboard there is an interruption after several seconds, the syslog reports a reset of the peripheral, and then the transfer resumes. This is on 2.6.28-rc6 kernel.
Fixed by 0a99e8ac430a27825bd055719765fd0d65cd797f