Bugzilla – Bug 105009
Oops when removing rtl8150 from USB-port
Last modified: 2006-09-12 13:02:23 UTC
- conntect rtl8150 - configure network device - ping -f remote_machine - disconnect rtl8150 -> Oops The Oops is attaches as screenshot, because the machine doesn't log it to /var/log/messages
Created attachment 46201 [details] Oops
It hits a BUG at softirq.c:266. Looks like the rtl8150 tasklet not getting killed when the device is removed? There ought to be a tasklet_kill somewhere. Greg, can you please have a look at this one? Thanks!
BTW - why does the taint flags show F for forced module load, Andreas?
I will try to make up a patch for this, can you test a patch? Oh, and what kind of device is this? I'll try to find one to test with.
Yes, I will test a patch as soon as I get one. The devive is a USB-network-card. You can buy it from http://www.reichelt.de for about 10 euro I think.
Created attachment 46608 [details] Patch to try to fix oops. Can you try this patch out to see if it solves the problem or not? I've only compile tested it as I don't have the hardware for this device.
I'm compiling the modules.
The patch fixes the problem, but the kernel oopses oopses with a different call trace. I will attach again a screenshor.
Created attachment 46615 [details] new oops
Ugh, not good, I'll ping the ehci author/maintainer about it...
What's the status on this bug? Andreas, did you see the oops in beta4, too? Greg, did you get a response from the ehci author/maintainer?
Yes, beta4 Oopses too.
I never got a response from the EHCI maintainer, and no one else is reporting the problem. I've pushed the proposed patch upstream, as it was correct. As for further followups, I don't have this device to test with, but it looks like Intel might have one for me to use, I'll let you know what happens with that. Also, does 10.0 still show this problem? 2.6.14?
The problem still exists in 10.1 beta1.
Created attachment 64416 [details] Oops with 10.1 beta1
Re: comment #15, that looks like a different problem. Did that happen when you unplugged the device? Also, what about Beta3?
I tried beta3 and kernel-smp-2.6.16_rc2_git5-3 fron factory. The problem still exists. Yes, it happens when unplugging a configured device.
Created attachment 67827 [details] 10.1beta3
This doesn't look like a USB problem anymore, but rather a subdomain issue. Reassigning to the apparmor people.
Greg. Don't think it is our problem. 1) Comment #18. Yes we're in the call trace but it looks foobared to me, there is no path from subdomain_file_permission to seq_read. 2) Comment #15, apparmor isn't even loaded! 3) Failure in a softirq seems the common case (to me). If you find something more compelling than the word subdomain in the trace, feel free to kick it back over to us.
Any chance 10.1 RC2 helps out here? I really have no idea, sorry...
Same problem with RC3
I'm guessing that the final 10.1 kernel still has this problem? Any way you can test the 2.6.17-rc4 kernel from kernel.org to see if this is fixed there? Without this device, solving this is going to be difficult :(
Closing due to inactivity. If you have the needed information, please reopen with it.