Package: linux-image-3.9.6
Version: 3.9.6
Severity: normal
the "normal" reporting information is being removed because the report
is being generated from an alternative non-SMTP-networked system
here's what has to be done to solve the problem:
write_sys("/sys/bus/usb/devices/usb1/remove", "1")
write_sys("/sys/bus/usb/devices/usb2/remove", "1")
write_sys("/sys/devices/pci0000:00/0000:00:0f.5/remove", "1")
write_sys("/sys/devices/pci0000:00/0000:00:0f.4/remove", "1")
time.sleep(5)
write_sys("/sys/devices/pci0000:00/pci_bus/0000:00/rescan", "1")
time.sleep(5)
the two devices are listed from lspci, here:
00:0f.4 USB controller: Advanced Micro Devices [AMD] CS5536 [Geode companion]
OHC (rev 02)
00:0f.5 USB controller: Advanced Micro Devices [AMD] CS5536 [Geode companion]
EHC (rev 02)
which correspond to the two "remove"s above.
this is basically an incredibly drastic PCI USB host controller reset
which has to be done in between mountkernfs and udev. it's therefore
set up on an /etc/rc.S/S03 level script which carries out the resets
that *should* be being done by the kernel modules themselves in the
first place.
symptoms are as follows:
* put a USB 3G MiniPCIe modem into an alix6f2 LX800 board
* ls /dev/ttyACM0
* reboot (do NOT powercycle)
* log in again
* ls /dev/ttyACM0 - FAILS.
however if instead of a reboot a halt and power-cycle is carried out instead,
the 2nd ls /dev/ttyACM0 SUCCEEDS.
it *also* succeeds even without a powercycle if the above drastic PCI device
reset is carried out.
if however this is carried out *after* udev is allowed to pick things up
then we have a bit of a problem, which is that e.g. other USB devices
(such as USB memory sticks or USB-to-serial converters) suddenly find that
in the middle of loading the kernel module the device just... disappears
(because the USB hub it's connected to just got a kick in the nuts).
this has probably been a long-standing bug in the CS5536 PCI host controller
code. the reset should be being carried out at startup when the module
is loaded.
--
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
