On Mon, 27 Feb 2012, Adrien Decostre wrote: > On Mon, Feb 27, 2012 at 4:23 PM, Alan Stern <stern@xxxxxxxxxxxxxxxxxxx> wrote: > > On Mon, 27 Feb 2012, Adrien Decostre wrote: > > > >> Dear all, > >> > >> I am currently doing some tests to communicate with an USB HID device > >> by means of the HIDAPI and libusb libraries on a Linux x64 > >> distribution. > >> One of these tests consists basically of successive enumeration cycles: > >> -��opening the hid device > >> -��exchanging some data between the host and the device > >> -��at some point, closing the communication with the hid device > >> -��wait a few second and reiterate the above steps > >> > >> This test runs correctly a few times and then suddenly crashes. > >> Apparently, this crash happens in a random way. > >> > >> The strange point is that after this crash, the file system in RAM is > >> "corrupted": some folders like �etc� have disappeared and I have to > >> reboot the kernel. > > > > I have no idea what could be causing that corruption, unless the > > filesystems in question reside on a USB drive. > > For one of the 2 PCs, this is indeed the case. I must further check > for the other one. > > >> The crash has been seen on 2 different machines: > >> - on x64 Ubuntu 2.6.35 based kernel (Ubuntu 10.10) > >> - on x64 Angstr�.6.39 based kernel (the log given below have been > >> obtained on this machine). On this last machine, the crash happens > >> only on the 2nd usb bus. When the test usb device is connected to the > >> 1st bus usb, I have not seen any problem up to now. > >> > >> When looking to the dmesg output, the crash seems to be caused by the > >> kworker process. The crash log starts with a crash of the ehci driver > >> (see below for a complete dmesg log): > >> ehci_hcd 0000:00:13.2: force halt; handshake ffffc90000070024 00004000 > >> 00000000 -> -110 > >> ehci_hcd 0000:00:13.2: HC died; cleaning up > >> usb 2-4: usbfs: usb_submit_urb returned -110 > >> usb 2-3: USB disconnect, device number 2 > >> > >> In the past days, I have been looking for a possible bug in my test > >> application or in the HIDAPI / libusb libraries (cf. > >> http://libusb.6.n5.nabble.com/Application-using-HIDAPI-and-libusb-causes-file-system-corruption-td5508578.html) > >> but I am now wondering if this could be related to a known issue in > >> the kernel. > > > > It's a known bug in the EHCI hardware in quite a few different > > platforms. �Here's an experimental patch to work around it: > > > > � � � �http://marc.info/?l=linux-usb&m=132440828606642&w=2 > > > > You may have to do some hand editing to get that patch to apply, > > because you're using some old kernels. > > Thanks a lot for this information. > I did quick test today and it apparently solves the issue. I will test > it more extensively in the coming days. I have to warn you that the patch really is only experimental, and it is far from the finished version that will eventually get merged into the kernel -- in fact, I haven't even started to write the finished version yet. On the other hand, it doesn't hurt to have more people testing the experimental patch to verify that the idea behind it really does work in practice. In addition to crashes, be on the watch for data loss (dropouts) of any kind. Alan Stern -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
