Bugzilla – Bug 298944
kernel-bigsmp tears down laptop support almost to zero
Last modified: 2007-08-24 12:47:41 UTC
YaST installs kernel-bigsmp instead of kernel-default. My notebook is a Thinkpad T60 with 2GB of memory. I saw this on a fresh installed Acer notebook (512MB memory), too. AFAIK we should install kernel-default on all machines with less than 4GB memory.
Since this also mean there is no battery module (missing all battery infos on laptops) on bigsmp this bug is CRITICAL
This was discussed on the kernel mailing list, I suggest you follow up on opensuse-kernel if you disagree. We will install bigsmp on all machines since it enables PAE which gives us some security benefits. I consider this part of the bug report invalid. Why is there no battery module? That's a bug and it should be fixed -> on to kernel team.
http://mailman.suse.de/mlarch/SuSE/kernel/2006/kernel.2006.01/msg00030.html The suspend issues we discussed then are assumed to be fixed now. So please just add the necessary ACPI modules that are missing.
we have been working on this for quite some time, and also discussed this in the respective fate entry and AJ and coolo. What is the problem with just adding the missing ACPI / battery modules?
(In reply to comment #6 from Marcus Meissner) > we have been working on this for quite some time, and also discussed this > in the respective fate entry and AJ and coolo. Respective fate entry which did not reach the relevant people. As you can see we now get some new kernel which invalidates all previous testing with regard to STR and STD for mobile devices. > What is the problem with just adding the missing ACPI / battery modules? No problem at all, but getting it _that_ late without prior notice is just a bad joke.
> Respective fate entry which did not reach the relevant people. As you can see > we now get some new kernel which invalidates all previous testing with regard > to STR and STD for mobile devices. At least all of the relevant people are on vacation ;-)
Can we please get a bigsmp kernel with the missing modules soon? If we wait for Beta 2 we're going to lose two important weeks of testing.
Created attachment 156749 [details] diff (default vs. bigsmp) The diff shows that it is much worse than just the battery module. Oh my goodness...
Adjusting summary & component, raising severity.
STD does not work at all with bigsmp on a T60p. $ cat /sys/power/state mem Very well done.
The laptop related configs should be merged into bigsmp (and I hope all can). We can then release a kernel update for the beta customers (+kmps).
Really, I think that the only difference between default and bigsmp should be CONFIG_HIGHMEM64G=y (with all symbols that depend on it) and CONFIG_NR_CPUS=128. Are there any known issues with the ACPI modules with PAE? I'd expect not, but before enabling them I'd like to be certain it's just a simple oversight and not intentional. I guess that's what the beta cycle is for, though. If there are bugs lurking, they should be shaken out now. I'll commit a new config for bigsmp and submit the kernel to STABLE.
Andi, is there a reason X86_MCE_NONFATAL is disabled on bigsmp? Is it another oversight, or does some big hardware spew messages when this is enabled?
It must be an oversight. I can't think of any reason for it. BTW in autoboot I use a macro based approach to generate the different configurations: always use the same base config and then just change specific options for sub kernels. Perhaps it would make sense to switch the i386 and x86-64 kernels to a similar scheme. Make one config the base and auto generate the others. It's not that difficult to do.
I was thinking something along those lines as well. That would certainly eliminate divergences like this in the future.
please install the testupdate: zypper sa ftp://you.suse.de/pub/suse/update/10.3 YOU zypper refresh zypper up this should update the kernel + kmps. (it will unfortunately pull in xen,kernel-xenpae due to the iwlwifi patch, ignore this problem). STD works for me (and restore from disk too) ;)
The update looks good at a first glance.
kernel + kmp patches approved.
The kernel update works for me, too.
Sidenote: with bigsmp as default it seems to be critical, to detect if the target cpu is a pentium m with banias core.. as bigsmp (HIGMEM64G=y) does _not_ work there at all (kernel is immediatelly panicing with "CPU too old for this kernel."). I thought this was a urban myth, but i just checked it here with a centrino notebook (asus m6000) which really doesn't boot at all with a bigsmp kernel (from sl103 beta2).
Frank, I suggest to file a separate bug for #24. For these systems we should not use kernel-bigsmp.