Bug 32092
| Summary: | Regression in zd1211rw | ||
|---|---|---|---|
| Product: | Drivers | Reporter: | jcallen |
| Component: | network-wireless | Assignee: | Jussi Kivilinna (jussi.kivilinna) |
| Status: | CLOSED CODE_FIX | ||
| Severity: | normal | CC: | florian, jussi.kivilinna, stf_xl |
| Priority: | P1 | ||
| Hardware: | All | ||
| OS: | Linux | ||
| Kernel Version: | 2.6.38 | Subsystem: | |
| Regression: | Yes | Bisected commit-id: | |
| Attachments: |
zd1211rw: make zd_usb_iowrite16v_async() handle bulk endpoints correctly
lsusb -vvv output dmesg output Remove bogus URB flag and add buffer len check to URB completion handler. kernel backtrace Fixes WARNING caused by zd_usb_reset_rx_idle_timer/del_timer_sync |
||
|
Description
jcallen
2011-03-29 05:23:58 UTC
Can you provide 'lsusb -vvv' on this device? Also 'dmesg' might help solving this. Created attachment 52442 [details]
zd1211rw: make zd_usb_iowrite16v_async() handle bulk endpoints correctly
Found 'lsusb -vvv' for this device from Internet and seems that this device has different type of endpoint than expected (bulk instead of interrupt).
Please test this patch, it handles both endpoint types now.
Created attachment 52522 [details]
lsusb -vvv output
Attached is the lsusb -vvv output for this device (just in case it differs from what you found on the internet).
I tried using the given patch, but got the same error. Looking more closely at the output, I see that the network scripts are actually failing with "cat: /sys/class/net/wlan0/carrier: Invalid argument" when attempting to check if /sys/class/net/wlan0/carrier has the value "1".
There does not appear to be any relevant dmesg output.
Ok, it is not about endpoint types after all. Can you compile zd1211rw with CONFIG_ZD1211RW_DEBUG=y for more dmesg output? Created attachment 52572 [details]
dmesg output
Created attachment 52612 [details]
Remove bogus URB flag and add buffer len check to URB completion handler.
I should have tested with USB debugging enabled, as it turns on more checks on URBs.
Patch removes the bogus flag introduced by sync->async patch. Old code had buffer len check that new code tried to handle with NO_SHORT flag.
Did you have chance to test this yet? Reproduced and tested fix myself. Created attachment 53272 [details] kernel backtrace (In reply to comment #7) > Created an attachment (id=52612) [details] > Remove bogus URB flag and add buffer len check to URB completion handler. > > I should have tested with USB debugging enabled, as it turns on more checks > on > URBs. > > Patch removes the bogus flag introduced by sync->async patch. Old code had > buffer len check that new code tried to handle with NO_SHORT flag. The new patch allows the network to start up properly, however this seems to have exposed another bug: my dmesg buffer immediately fills up with (apparently identical) backtraces as soon as I try to start up networking; I have attached a sample. I'm not sure if I should open a new bug report for this. Created attachment 53302 [details]
Fixes WARNING caused by zd_usb_reset_rx_idle_timer/del_timer_sync
I ran to same bug/WARNING and fix is going upstream. A patch referencing this bug report has been merged in v2.6.39-rc3: A patch referencing this bug report has been merged in v2.6.39-rc3: commit 2fc713b20469b2779fa89f582d2a9696b10031e7 Author: Jussi Kivilinna <jussi.kivilinna@mbnet.fi> Date: Fri Apr 1 19:34:08 2011 +0300 zd1211rw: remove URB_SHORT_NOT_OK flag in zd_usb_iowrite16v_async() The warning fix should be: commit 023535732f4db01af4921f20f058bc4561d9add7 Author: Jussi Kivilinna <jussi.kivilinna@mbnet.fi> Date: Fri Apr 1 19:34:16 2011 +0300 zd1211rw: reset rx idle timer from tasklet |