Bug 69830 - [NVC1] Kernel 3.11 and newer hang at X11 start (bisected)
Summary: [NVC1] Kernel 3.11 and newer hang at X11 start (bisected)
Status: RESOLVED DUPLICATE of bug 69203
Alias: None
Product: Mesa
Classification: Unclassified
Component: Drivers/DRI/nouveau (show other bugs)
Version: unspecified
Hardware: x86-64 (AMD64) Linux (All)
: medium major
Assignee: Nouveau Project
QA Contact:
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2013-09-26 02:58 UTC by Alan Ott
Modified: 2013-09-27 14:51 UTC (History)
0 users

See Also:
i915 platform:
i915 features:


Attachments
Xorg log (33.65 KB, text/plain)
2013-09-26 02:58 UTC, Alan Ott
Details
Boot log (see "GPU lockup") (80.29 KB, text/plain)
2013-09-26 03:01 UTC, Alan Ott
Details

Description Alan Ott 2013-09-26 02:58:01 UTC
Created attachment 86607 [details]
Xorg log

After updating from Linux kernel 3.9 to 3.12-rc2, X would no longer start. I bisected the problem (starting at 4a10c2ac2f368583138b774ca41fac4207911983) down to commit 30f4e0870d1726f31aa59804337cfd5e0a3f2ec7.

The bisect went like this:

a0fd4ec8f1ac1d966d33d1a18205b72830f9b24f - good
8ff860564076f99aeb4f6d35dcb6ea008624e64e - bad
d005f51eb93d71cd40ebd11dd377453fa8c8a42a - bad
56fbd2b65446d4fb4df7770c49a70d563b7569c9 - bad
791dc143ed2c441f5202d8721609d94dce9fcf88 - good
5ee86c4190f9e19a9e13906389069c73d7f75bfb - weird lines on top (called bad)
c03ff9e8fa5fc0186158b99a89f613325ff352cf - blank screen then monitor off (called bad)
30f4e0870d1726f31aa59804337cfd5e0a3f2ec7 - bad

The two commits above with comments (5ee and c03) didn't make it all the way to X11 (they hung before X11 was started, or so it seemed), but git bisect reported at the end that 30f4e0870d1726f31aa59804337cfd5e0a3f2ec7 was the first bad commit (and 5ee and c03 are both after 30f4e (the first bad one)).

Logs for a bad run are attached.
Comment 1 Alan Ott 2013-09-26 03:01:21 UTC
Created attachment 86608 [details]
Boot log (see "GPU lockup")
Comment 2 Ilia Mirkin 2013-09-26 03:46:43 UTC
A summary for casual observers: 30f4e0870d17 verified bad, 30f4e0870d17^ verified good.
Comment 3 Ilia Mirkin 2013-09-26 22:28:12 UTC
Alan, given the latest update on the other bug, I'm fairly convinced that it's the same issue that you're having. Feel free to split them back up if you think otherwise.

*** This bug has been marked as a duplicate of bug 69203 ***
Comment 4 Alan Ott 2013-09-26 22:33:02 UTC
The behavior I saw was a bit different than bug 62093.

I saw 30f4e08 being bad (x11 hang) and 30f4e08^ as being a good boot. That may just be different hardware. I can check and make sure I'm correct and got the builds I expected.
Comment 5 Ilia Mirkin 2013-09-26 22:35:08 UTC
If you check the last comment of that bug, that's what the OP found as well. (Unless I'm misunderstanding.)
Comment 6 Alan Ott 2013-09-26 22:42:43 UTC
(In reply to comment #5)
> If you check the last comment of that bug, that's what the OP found as well.
> (Unless I'm misunderstanding.)

On that other bug, the lists of commits don't match my git history. I don't have 791dc14 in the same place (it's inside a merge so I'll have to look closer). Also 30f4e08 causes an X11 hang for me, not a boot failure. If the user in the other bug has stuff out of order for some reason (using a tree other than mainline?) then it's probably the same issue.

These being the same issue is likely, but the record should show that on my system, with the mainline kernel, 30f4e08 is the first bad one and doesn't hang the system at boot; it just fails to start x11.
Comment 7 Ilia Mirkin 2013-09-26 22:47:07 UTC
https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=30f4e08

Note that the parent is 791dc143ed2c441f5202d8721609d94dce9fcf88. So I don't know what tree you're on, but you should get on something closer to mainline :)

Anyways, duly noted that you don't get a hang until you start X. Could be a fluke, or dumb luck, or some distro setup difference. I don't think it's a coincidence that a lot of people are having trouble with precisely NVC1 on 3.11+ though.
Comment 8 Alan Ott 2013-09-27 14:51:46 UTC
I'm on mainline. As evidence, noice the SHA1 of my HEAD:

alan@ato:~/work/linux-3.12$ git log |head
commit 4a10c2ac2f368583138b774ca41fac4207911983
Author: Linus Torvalds <torvalds@linux-foundation.org>
Date:   Mon Sep 23 15:41:09 2013 -0700

    Linux 3.12-rc2

commit 9d23108df359e572a0dca0b631bfee9f5e0fa9ea
Merge: e04a0a5 6174081
Author: Linus Torvalds <torvalds@linux-foundation.org>
Date:   Mon Sep 23 12:53:07 2013 -0700

See 3.12-rc2 has the same SHA1:
https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=4a10c2ac2f368583138b774ca41fac4207911983


The gitk shows that while 791dc is the parent of 30f, there was a lot of merge activity in between, so I think we're looking at the same thing. Sorry to confuse the matter.


Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.