[arch-general] Kernel 3.2.1 woes
Hi folks, just a heads up (and maybe request for feedback whether anybody has seen similar issues with this kernel upgrade). The recent upgrade from kernel 3.1.9 to 3.2.1 produced two graphics-related issues: On my dual-Opteron Asus board with an NVIDIA GeForce GTX 460, the X server fails to start and says that it cannot initialize the device at PCI 4:0:0 (i.e., my NVIDIA card). Downgrading to 3.1.9 fixes this issue. On my Core i7 Dell Latitude 6510 laptop using the on-chip graphics chip and the 1920x1080 laptop screen, everything is fine until I suspend the machine to RAM. When resuming, I get a black screen and not even rebooting brings the screen back to life. I have to turn the machine off and then back on. In contrast, my Core i7 desktop at home with an NVIDIA Quadro NVS 450 works just fine with the 3.2.1 kernel. I'm in the process of bisecting these issues (currently working on the laptop issue). I remember that back in the kernel 2.6.3x days, it took a long time for the Dell 6510 with 1920x1080 display and Intel graphics to boot without a black screen at all, and I seem to recall that this was because the patch that made it work apparently didn't do what it was supposed to do "right". Maybe they now found a way to do it right, only it doesn't work, at least after a suspend. Note that, even though I get a black screen after suspending, the machine is otherwise alive, as I can blindly switch to a TTY and reboot the machine cleanly from there. Bisecting is proving a bit painful because quite a few intermediate versions I'm testing result in a straight kernel panic after suspend. Cheers, Norbert -- () ascii ribbon campaign - against html e-mail /\ www.asciiribbon.org - against proprietary attachments
2012/1/23 Norbert Zeh <nzeh@cs.dal.ca>
Hi folks,
just a heads up (and maybe request for feedback whether anybody has seen similar issues with this kernel upgrade). The recent upgrade from kernel 3.1.9 to 3.2.1 produced two graphics-related issues:
On my dual-Opteron Asus board with an NVIDIA GeForce GTX 460, the X server fails to start and says that it cannot initialize the device at PCI 4:0:0 (i.e., my NVIDIA card). Downgrading to 3.1.9 fixes this issue.
On my Core i7 Dell Latitude 6510 laptop using the on-chip graphics chip and the 1920x1080 laptop screen, everything is fine until I suspend the machine to RAM. When resuming, I get a black screen and not even rebooting brings the screen back to life. I have to turn the machine off and then back on.
In contrast, my Core i7 desktop at home with an NVIDIA Quadro NVS 450 works just fine with the 3.2.1 kernel.
I'm in the process of bisecting these issues (currently working on the laptop issue). I remember that back in the kernel 2.6.3x days, it took a long time for the Dell 6510 with 1920x1080 display and Intel graphics to boot without a black screen at all, and I seem to recall that this was because the patch that made it work apparently didn't do what it was supposed to do "right". Maybe they now found a way to do it right, only it doesn't work, at least after a suspend.
Note that, even though I get a black screen after suspending, the machine is otherwise alive, as I can blindly switch to a TTY and reboot the machine cleanly from there. Bisecting is proving a bit painful because quite a few intermediate versions I'm testing result in a straight kernel panic after suspend.
Cheers, Norbert
--
() ascii ribbon campaign - against html e-mail /\ www.asciiribbon.org - against proprietary attachments
Erh... Did you blacklist nouveau on your laptop or not ? Just a guess of course, cause it looks like a kmod bug I saw when it was in testing... -- Frederic Bezies fredbezies@gmail.com
fredbezies [2012.01.23 0736 +0100]:
2012/1/23 Norbert Zeh <nzeh@cs.dal.ca>
Hi folks,
just a heads up (and maybe request for feedback whether anybody has seen similar issues with this kernel upgrade). The recent upgrade from kernel 3.1.9 to 3.2.1 produced two graphics-related issues:
On my dual-Opteron Asus board with an NVIDIA GeForce GTX 460, the X server fails to start and says that it cannot initialize the device at PCI 4:0:0 (i.e., my NVIDIA card). Downgrading to 3.1.9 fixes this issue.
On my Core i7 Dell Latitude 6510 laptop using the on-chip graphics chip and the 1920x1080 laptop screen, everything is fine until I suspend the machine to RAM. When resuming, I get a black screen and not even rebooting brings the screen back to life. I have to turn the machine off and then back on.
In contrast, my Core i7 desktop at home with an NVIDIA Quadro NVS 450 works just fine with the 3.2.1 kernel.
I'm in the process of bisecting these issues (currently working on the laptop issue). I remember that back in the kernel 2.6.3x days, it took a long time for the Dell 6510 with 1920x1080 display and Intel graphics to boot without a black screen at all, and I seem to recall that this was because the patch that made it work apparently didn't do what it was supposed to do "right". Maybe they now found a way to do it right, only it doesn't work, at least after a suspend.
Note that, even though I get a black screen after suspending, the machine is otherwise alive, as I can blindly switch to a TTY and reboot the machine cleanly from there. Bisecting is proving a bit painful because quite a few intermediate versions I'm testing result in a straight kernel panic after suspend.
Cheers, Norbert
--
() ascii ribbon campaign - against html e-mail /\ www.asciiribbon.org - against proprietary attachments
Erh... Did you blacklist nouveau on your laptop or not ? Just a guess of course, cause it looks like a kmod bug I saw when it was in testing...
No. I may be wrong, but I don't think this is needed here because the laptop has an Intel graphics chip. On the two NVIDIA boxes, nouveau is blacklisted. I really think this is a kernel regression because everything works as it should with 3.1.9 but not with 3.2.1. Apart from switching the kernel, I didn't touch any settings and the systems are fully up to date. - Norbert
-- Frederic Bezies fredbezies@gmail.com
-- () ascii ribbon campaign - against html e-mail /\ www.asciiribbon.org - against proprietary attachments
On Sunday 22 Jan 2012 21:39:31 Norbert Zeh wrote:
On my Core i7 Dell Latitude 6510 laptop using the on-chip graphics chip and the 1920x1080 laptop screen, everything is fine until I suspend the machine to RAM. When resuming, I get a black screen and not even rebooting brings the screen back to life. I have to turn the machine off and then back on.
Everything's working fine on my Latitude E5520 (HD3000 on-board GPU) with Linux 3.2.1-1. No problems resuming from suspend. Paul
于 2012年01月23日 17:49, Paul Gideon Dann 写道:
On Sunday 22 Jan 2012 21:39:31 Norbert Zeh wrote:
On my Core i7 Dell Latitude 6510 laptop using the on-chip graphics chip and the 1920x1080 laptop screen, everything is fine until I suspend the machine to RAM. When resuming, I get a black screen and not even rebooting brings the screen back to life. I have to turn the machine off and then back on. Everything's working fine on my Latitude E5520 (HD3000 on-board GPU) with Linux 3.2.1-1. No problems resuming from suspend.
Paul Dell Latitude D630 work fine in Kernel 3.2.1-1
I can't reproduce either bug on 3.2.1, latest NVIDIA drivers, GTX 460 1 GB. Are you sure the binary NVIDIA driver got rebuilt?
Michael Holmes [2012.01.23 1352 +0000]:
I can't reproduce either bug on 3.2.1, latest NVIDIA drivers, GTX 460 1 GB. Are you sure the binary NVIDIA driver got rebuilt?
I'm not sure, and this is something I should certainly try. However, along with the upgrade from 3.1.9 to 3.2.1 came an nvidia package upgrade from 290.10-1 to 290.10-2. Since the only difference is the arch package release number and the fact that 290.10-1 requires a kernel version < 3.2 and 290.10-2 requires a kernel version >= 3.2, I would think the 290.10-2 package is simply a rebuilt of this package against the new kernel. As I said, I don't see any issues on my Core i7 box with NVIDIA graphics, and the setups of these two boxes in terms of installed packages and configuration is pretty much identical. So I really think there's something going on with this dual-Opteron board at the kernel level, but I should certainly rule out the simple fixes first. Cheers, Norbert
On 01/23/2012 03:10 PM, Norbert Zeh wrote:
I can't reproduce either bug on 3.2.1, latest NVIDIA drivers, GTX 460 1 GB. Are you sure the binary NVIDIA driver got rebuilt? I'm not sure, and this is something I should certainly try. However, along with
Michael Holmes [2012.01.23 1352 +0000]: the upgrade from 3.1.9 to 3.2.1 came an nvidia package upgrade from 290.10-1 to 290.10-2. Since the only difference is the arch package release number and the fact that 290.10-1 requires a kernel version< 3.2 and 290.10-2 requires a kernel version>= 3.2, I would think the 290.10-2 package is simply a rebuilt of this package against the new kernel.
As I said, I don't see any issues on my Core i7 box with NVIDIA graphics, and the setups of these two boxes in terms of installed packages and configuration is pretty much identical. So I really think there's something going on with this dual-Opteron board at the kernel level, but I should certainly rule out the simple fixes first.
Cheers, Norbert @Norbert Do you have any shutdown or reboot issues? I have a similar setup to you only laptop, but if I run the nvidia driver X either gives a red screen when I try and shutdown or reboot, or just hangs with a cursor blinking. Just curious.
Thanks
Don Juan [2012.01.23 1521 -0800]:
@Norbert Do you have any shutdown or reboot issues? I have a similar setup to you only laptop, but if I run the nvidia driver X either gives a red screen when I try and shutdown or reboot, or just hangs with a cursor blinking. Just curious.
No, I don't have any such issues...well, in part I don't even get that far. Laptop/intel graphics: Boots/reboots/halts all fine, as long as I don't suspend to RAM. After suspend to RAM, screen backlight is on, but screen remains black until turning the machine off. Apart from the screen issue, the machine seems to behave fine. Opteron desktop/NVIDIA graphics: Boots/reboots/halts all fine. With 3.1.9 kernel, everything works as expected. With 3.2.1 kernel, I don't even get as far as starting X. i7 desktop/NVIDIA graphics: Everything works as expected with both kernel versions. Cheers, Norbert
On Monday 23 Jan 2012 21:53:25 Norbert Zeh wrote:
Laptop/intel graphics: Boots/reboots/halts all fine, as long as I don't suspend to RAM. After suspend to RAM, screen backlight is on, but screen remains black until turning the machine off. Apart from the screen issue, the machine seems to behave fine.
I get exactly the same behaviour on my laptop (Core i7, integrated Intel graphics). Did you figure it out? Pete.
Peter Lewis [2012.01.25 0919 +0000]:
On Monday 23 Jan 2012 21:53:25 Norbert Zeh wrote:
Laptop/intel graphics: Boots/reboots/halts all fine, as long as I don't suspend to RAM. After suspend to RAM, screen backlight is on, but screen remains black until turning the machine off. Apart from the screen issue, the machine seems to behave fine.
I get exactly the same behaviour on my laptop (Core i7, integrated Intel graphics).
Did you figure it out?
It's a time-consuming process. First result I got is that somewhere along the way, a commit changed processor initialization after resuming. It tries to avoid some unnecessary work. The problem is: with the "unnecessary work" (before this commit) it worked; after it, I get a kernel panic. Right now I know a point in the commit sequence where I get a kernel panic and a point where I get the black screen issue. I don't know whether there was a stage in between where there is no panic and no black screen. I have about 300 commits left to bisect, and I'll see whether I can finish bisecting this tonight (recompiling and testing the kernel about log 300 = 9 times). I don't have the expertise to fix this, but I want to know exactly where the bug was introduced, in order to provide useful information in my bug report to the kernel developers. Overall, I have the feeling that the 3.2.X kernel line introduced a number of regressions and is of much lower quality than the 3.1.X line. With 3.1.X, everything worked flawlessly. In addition to my graphics woes with 3.2.1, I also found that my intel wireless card starts misbehaving at times with 3.2.1, which it never did/does with any of the 3.1.X kernels. Cheers, Norbert -- () ascii ribbon campaign - against html e-mail /\ www.asciiribbon.org - against proprietary attachments
Have you figured this out? I have an issue that is similar, that may be related. When using an external display, my laptop screen is no longer usable since upgrading to 3.2.x from 3.1.x. I've tried using xrandr to turn it on with no luck, the mouse will go off the screen as if it is moved over to that display but the display will not come on. This problem happened right after the upgrade to 3.2.x for me.
Chris Cooper [2012.01.27 1649 +0000]:
Have you figured this out? I have an issue that is similar, that may be related.
When using an external display, my laptop screen is no longer usable since upgrading to 3.2.x from 3.1.x. I've tried using xrandr to turn it on with no luck, the mouse will go off the screen as if it is moved over to that display but the display will not come on. This problem happened right after the upgrade to 3.2.x for me.
Not figured out yet. Alas, as most people, I have to work for my money, and kernel bisecting is not part of this job. So I'll have to push this off by another week. Sorry. Cheers, Norbert -- () ascii ribbon campaign - against html e-mail /\ www.asciiribbon.org - against proprietary attachments
Chris Cooper [2012.01.27 1649 +0000]:
Have you figured this out? I have an issue that is similar, that may be related.
When using an external display, my laptop screen is no longer usable since upgrading to 3.2.x from 3.1.x. I've tried using xrandr to turn it on with no luck, the mouse will go off the screen as if it is moved over to that display but the display will not come on. This problem happened right after the upgrade to 3.2.x for me.
Norbert Zeh [2012.02.06 1138 -0400]:
Chris Cooper [2012.01.27 1649 +0000]:
Have you figured this out? I have an issue that is similar, that may be related.
When using an external display, my laptop screen is no longer usable since upgrading to 3.2.x from 3.1.x. I've tried using xrandr to turn it on with no luck, the mouse will go off the screen as if it is moved over to that display but the display will not come on. This problem happened right after the upgrade to 3.2.x for me.
Sorry for the previous empty message. Accidentally pressed "Send" before typing anything meaningful. The update on this is: I was too slow with bisecting the kernel on the Intel graphics issue. Somebody else did this already, reported the bug to LKML, and they are working on it. Can't quite find the reference to it right now, though. My NVIDIA issue remains even with the 3.2.4-1 kernel, while (as I said before) 3.1.9-2 works just fine. Given that the intel graphics issue is being worked on by kernel upstream, I'll start digging to see what I can come up with concerning this issue. Cheers, Norbert
participants (8)
-
Chris Cooper
-
Don Juan
-
fredbezies
-
Michael Holmes
-
Norbert Zeh
-
Paul Gideon Dann
-
Peter Lewis
-
朱格宏