Bug 14200
| Summary: | kernel oops with latest tg3 changes | ||
|---|---|---|---|
| Product: | Drivers | Reporter: | Daniel Vetter (daniel) |
| Component: | Network | Assignee: | drivers_network (drivers_network) |
| Status: | CLOSED CODE_FIX | ||
| Severity: | normal | CC: | mcarlson, rjw |
| Priority: | P1 | ||
| Hardware: | All | ||
| OS: | Linux | ||
| Kernel Version: | 2.6.31-git | Subsystem: | |
| Regression: | Yes | Bisected commit-id: | |
| Bug Depends on: | |||
| Bug Blocks: | 14230 | ||
| Attachments: |
dmesg of working kernel 2.6.31-01217-g483e3cd
Fix return ring size breakage |
||
|
Description
Daniel Vetter
2009-09-21 14:33:07 UTC
Created attachment 23130 [details]
dmesg of working kernel 2.6.31-01217-g483e3cd
Adding Matt Carlson because he's the author of all relevant patches. I scanned over those patches. So far I don't see the problem. I tried reproducing the problem with a 5703 B0 here. I failed there too. I'll be interested to see what the bisection yields. Please be advised that there is a bug in one of the later patches in this patchset that may reduce the effectiveness of the bisection. The good news is that the bug happens in the last patch that might be relevant to the 5703. Hopefully the bisection will point to an earlier patch, but if not, the fix is small and should apply cleanly to any other bisection point. On Mon, Sep 21, 2009 at 05:45:49PM +0000, bugzilla-daemon@bugzilla.kernel.org wrote: > --- Comment #3 from Matt Carlson <mcarlson@broadcom.com> 2009-09-21 17:45:48 > --- > I scanned over those patches. So far I don't see the problem. I tried > reproducing the problem with a 5703 B0 here. I failed there too. I'll be > interested to see what the bisection yields. Thanks for the fast turnaround. > Please be advised that there is a bug in one of the later patches in this > patchset that may reduce the effectiveness of the bisection. The good news > is > that the bug happens in the last patch that might be relevant to the 5703. > Hopefully the bisection will point to an earlier patch, but if not, the fix > is > small and should apply cleanly to any other bisection point. Bisecting points at t f6eb9b1fc1411d22c073f5264e5630a541d0f7df Author: Matt Carlson <mcarlson@broadcom.com> Date: Tue Sep 1 13:19:53 2009 +0000 tg3: Add 5717 asic rev This patch adds the 5717 asic rev. Signed-off-by: Matt Carlson <mcarlson@broadcom.com> Reviewed-by: Benjamin Li <benli@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net> I've doubled-tested both first-bad and last-good. I've also tried to revert this on top of latest -linus, but failed due to conflicts. -Daniel Created attachment 23132 [details]
Fix return ring size breakage
Drat. That was the problematic patch. If you were to apply this patch, does the bisection still point to this patch as the culprit?
FYI, this patch represents the last patch that was integrated and is already included in the current upstream kernel (and the snapshot you are working from).
Are you using jumbo frames by any chance? I've tested your patch and this fixed the problem. I've also checked the serial console output and it looks like I'm hitting another problem. Bad luck while bisecting pointed then to this unrelated tg3 problem. I'm closing this bug now. Thanks for your help in tracking this down. -Daniel I don't think we are done though. That patch has been (or should have been) applied to the tip of your tree. That is where you discovered the bug though, right? If so, then the real bug should still be lurking between the current bisection point and the tip of the tree (a few patches away). > --- Comment #8 from Matt Carlson <mcarlson@broadcom.com> 2009-09-23 17:57:56
> ---
> I don't think we are done though. That patch has been (or should have been)
> applied to the tip of your tree. That is where you discovered the bug
> though,
> right? If so, then the real bug should still be lurking between the current
> bisection point and the tip of the tree (a few patches away).
As I've said, there is still another problem. But I looked at the traces
via serial console, and it is _definitely_ something else (somewhere in
the block layer). Thanks to your patch, I can make sure that I track down
the right problem when bisect again. I was fooled only because I checked
the serial console only after bisecting a few revisions which lead me into
thinking that the original problem was tg3 related. It is not.
I'll post a link here to the new bug report, as soon as I find time to
bisect it and prepare a report.
Thanks for your help, Daniel
Handled-By : Matt Carlson <mcarlson@broadcom.com> Patch : http://bugzilla.kernel.org/attachment.cgi?id=23132 Fixed by commit 5ea1c50662d447de344812054175d7151783ea25. For reference, as promised, the new bug report (with my real problem): http://bugzilla.kernel.org/show_bug.cgi?id=14290 |