Bugzilla – Bug 155761
SONY Vaio PCG-GRT995MP does not resume-from-RAM unless 'noapic' is given.
Last modified: 2008-03-18 20:03:08 UTC
A Sony Vaio PCG-GRT995MP worked fine wrt suspend2ram in 10.0. It does not, however, in 10.1beta6. Adding "nolapic" fixes this. Without nolapic, it hangs after resuming the hardware and shortly before "Restarting Tasks". It does not hang hard, sysrq still works.
Created attachment 71591 [details] dmesg after boot
Created attachment 71592 [details] dmidecode
Did it work with earlier betas? Does it work with the latest KOTD? Andi's "enable APIC when BIOS is recent enough" patch would turn on APIC unconditionally, even when the BIOS had APIC support disabled. This should be fixed in the current KOTD.
It did work with 10.0 With Beta7, it still does not work, but with "nolapic" it does not boot anymore (hangs when discovering ide hda), "noapic" is needed now.
Handing over to Andi; IOAPIC trouble. Maybe the easiest thing is to disable the APIC based on the DMI info.
still in SLED10beta8.
Hmm, you said it booted with beta6 and didn't boot with beta7? Can you please narrow down which change broke it using kerneltest kernels? The noapic / nolapic confusion is weird. nolapic should include noapic. I think there is another bug for that open.
This is a UP+IOAPIC notebook. I suspect that previous releases didn't enable the IOAPIC. Now that IOAPIC is enabled by default, we notice that suspend/resume with IOAPIC enabled isn't reliable. So the question is with the old (working)( kernels booted with "apic" to force the use of the IOAPIC, does it fail the same way as the new kernel?
Unless Len has some good idea I will just blacklist it
>Without nolapic, it hangs after resuming the hardware and shortly before >"Restarting Tasks". It does not hang hard, sysrq still works Any clues to debug further? >Unless Len has some good idea I will just blacklist it Maybe it is just ok as a short term solution.
The (In reply to comment #8) > This is a UP+IOAPIC notebook. > > I suspect that previous releases didn't enable the IOAPIC. yes. > Now that IOAPIC is enabled by default, we notice that suspend/resume > with IOAPIC enabled isn't reliable. > > So the question is with the old (working)( kernels booted with "apic" > to force the use of the IOAPIC, does it fail the same way as the new kernel? yes. Should i further pursue the "nolapic" problem?
>Should i further pursue the "nolapic" problem? I think yes. as a long term solution, we do need to root-cause, and have a right fix either in kernel , or in bios.
Does adding a 'clear_IO_APIC' in 'ioapic_resume' change anything?
(In reply to comment #12) > >Should i further pursue the "nolapic" problem? > I think yes. as a long term solution, we do need to root-cause, and have a > right fix either in kernel , or in bios. Ok, i created bug #162098 for that since it is not suspend related.
Well I don't see how we can root cause anything if the reporter refuses to fix anything. Can you please try shaohua's suggestion from comment #13?
i tried it, but it does not help, it hangs just the same, before "restarting tasks..."
Please retest with latest kotd
without any options, it still hangs on resume, but both "noapic" and "nolapic" make it resume. Note that when the resume hangs, i can try to reboot with ctrl-alt-del, but i only get "stopping all md..." and then it does not reboot. It seems that some interrupt setup is wrong after resume if none of the no(l)apic options are present. Tested with kernel-default-2.6.16-20060412122502
Any progress here? BTW I am seeing a similar behaviour on my HP Compaq nx6125. In some cases the machine does not reboot, but hang at the end of the shutdown process.
No progress - probably you should pass your laptop to a kernel hacker and let them figure it out. I double checked the IO-APIC suspend/resume and as far as I can see every documented register is saved/restored. Or Luming do you have any ideas?
I'd think the machine is plain broken and needs to be blacklisted. It is a SONY after all, so there will be no BIOS update. Or would this be the only machine that's blacklisted wrt APIC?
Sure I can blacklist it. I suspect there will be more, but we didn't aggressively blacklist so far. If you can generate a list of laptops with dmidecode that need nolapic the might be useful.
Re: >It does not hang hard, sysrq still works. Could you post some information about this? Re: >probably you should pass your laptop to a kernel hacker Good idea. Is it possible to ship it to me. It makes some sense, especially when you want to just blacklist.
Luming, I assume you have a couple of test laptops already. Do they all suspend/resume with IO-APIC on?
Does this work with RC3/RC4?
SLED10B11 still needs "nolapic" or "noapic".
I am quite sure that i have seen this machine resume without problems on later openSUSE releases without noapic. There was a fix somewhen in late 2006 or 2007 that fixed APIC suspend / resume on a lot of machines and i think this fixed this machine, too. Since additionally this machine is totally unimportant for SLED, let's just wontfix this bug.