Summary: | relocatable x86 kernel panics on boot when built with sys-devel/binutils-2.21 | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | Graham Murray <gmurray> |
Component: | [OLD] Core system | Assignee: | Gentoo Toolchain Maintainers <toolchain> |
Status: | RESOLVED FIXED | ||
Severity: | critical | CC: | contact, joakim, martin.dummer, non7top, sascha_lucas, vanya |
Priority: | High | ||
Version: | 10.0 | ||
Hardware: | x86 | ||
OS: | Linux | ||
Whiteboard: | linux-2.6.36.2-regression | ||
Package list: | Runtime testing required: | --- | |
Bug Depends on: | 378485 | ||
Bug Blocks: | |||
Attachments: | Kernel Configuration |
Description
Graham Murray
2010-12-15 12:18:36 UTC
Created attachment 257220 [details]
Kernel Configuration
is this the first kernel with this problem? What is the latest working kernel? (In reply to comment #2) > is this the first kernel with this problem? What is the latest working kernel? > This is the first kernel with the problem. gentoo-sources-2.6.36-r4 is the latest working kernel. Same thing, same symptoms here. 2.6.36-r4 is working but 2.6.36-r5 is not. The config had not changed in between. I installed also git-sources-2.6.37_rc5-r4 with the same problem. I tried to boot without splash, did not change and with "noapic" in the kernel parameters. With "noapic" the kernel panic is not coming, but booting does halt in the middle. For the time being I will stick with 2.6.36-r4. Thanks for your help Dirk Can you test with vanilla-sources-2.6.36.2 and let me know if it fails in the same manner? It looks as though this is NOT a kernel problem, but a binutils problem. While trying to identify the failing commit using git bisect, I found that even previously working kernel versions would exhibit the failure. So, I tried reverting to using gcc-4.4.5 but this still failed. So I reverted back to sys-devel/binutils-2.20.1-r1 and rebuilt gentoo-sources-2.6.36-r5 and it boots and runs sucessfully. Adding toolchain for their thoughts http://sourceware.org/bugzilla/show_bug.cgi?id=12327 maybe? Relevant thread is http://thread.gmane.org/gmane.comp.gnu.binutils/48959, but no one posted the actual panic message so I don't know if it's the same issue. 2.21.51.0.4 has a fix for that bug if you want to try it. i dont think it's that big of a deal. the relocatable option is rarely useful to the vast majority of people. once upstream has merged a fix, we'll pull it in, but i dont see us doing anything beyond waiting. a change merged into the "linux binutils" isnt something i'd like added to the "gnu binutils" either. *** Bug 354339 has been marked as a duplicate of this bug. *** Please unmask a version of binutils which is not affected by this bug (I've tested sucessfully with 2.21.51.0.4) or either mask 2.21 (In reply to comment #6) > So I reverted back to > sys-devel/binutils-2.20.1-r1 and rebuilt gentoo-sources-2.6.36-r5 and it boots > and runs sucessfully. > I have exactly the same symptom and after lots of attempts to fix I found this bug and can confirm that downgrading binutils to 2.20.1-r1 fixes the problem! i'm not masking binutils-2.21, and it's ridiculous to even suggest it if someone wants to verify this fixes things for them with binutils-2.21, i can add it to the tree fairly easily: http://sourceware.org/git/?p=binutils.git;a=commit;h=5daebc6a6606a30e60716f5bdee3d2018b560e8e The bug is in main tree for two months already. I see nothing good in making users fight with known bugs, especially as this bug is rather hard to track. if you want to help, try the patch i already posted and post your results. otherwise, take your complaints to the forums. they have no business here. I encountered exactly the same problem, but in another setup: binutils-2.21 collaborated just fine with gcc-4.4.4-r2 into buiding a relocatable kernel. The issue showed up only after upgrading to the (stable now) gcc-4.4.5... I lost tons of time chasing the guilty package, because I only related the fault to the compiler upgrade (and you know the implications, i.e. emerge -e system and such). All my thanks to the original reporter! seems the upstream 2.21 branch merged a diff patch than was merged into CVS head, so extracting just the relevant pieces might be troublesome. perhaps just wait for the 2.21.1 release. we're stabilizing 2.21.1 |