Message-ID: <f488382f0906161435w2559831cob8a5b8a78c22cf91@mail.gmail.com>
Date:	Tue, 16 Jun 2009 14:35:27 -0700
From:	Steven Noonan <steven@...inklabs.net>
To:	Greg KH <greg@...ah.com>
Cc:	Michal Januszewski <spock@...too.org>,
	Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: [BISECTED] uvesafb broken on 2.6.30 (v86d segfault)

I noticed that v86d segfaults on the final v2.6.30 release (breaking
uvesafb), but doesn't on the current master (v2.6.30-3392-g44b7532)
nor on v2.6.29.x. I originally tried to git-bisect v2.6.29 and
v2.6.30-rc1 (the first tag to contain the problem), but I ran into
multiple kernels that would either not compile or would simply panic
on boot.

So instead, I did a 'git bisect' on commits between tags 'v2.6.30' and
commit-ish '44b7532' to find the commit that _corrected_ the problem
(reversing the 'good'/'bad' logic of git-bisect to do the task).

First, here's the relevant dmesg output for broken kernels:

[    1.770858] general protection fault: 0000 [#1] SMP
[    1.771080] last sysfs file:
[    1.771163] Modules linked in:
[    1.771287]
[    1.771287] Pid: 531, comm: v86d Not tainted (2.6.30 #1) HP
Pavilion dv5000 (ES933AS#ABA)
[    1.771287] EIP: 0060:[<c10c1377>] EFLAGS: 00013286 CPU: 0
[    1.771287] EIP is at lockdep_sys_exit+0x7/0xa0
[    1.771287] EAX: 00000000 EBX: 00000000 ECX: 00000000 EDX: f73e0000
[    1.771287] ESI: c3d9e800 EDI: f73e1ed4 EBP: f73e1ec0 ESP: f73e1eac
[    1.771287]  DS: 007b ES: 007b FS: 00d8 GS: 00e0 SS: 0068
[    1.771287] Process v86d (pid: 531, ti=f73e0000 task=f73fe9c0
task.ti=f73e0000)
[    1.771287] Stack:
[    1.771287]  f73e1f00 674c2f57 f73e1fb4 f73fe9c0 00000000 f73e0000
c10334c4 f73e0000
[    1.771287]  00000000 00000000 00000000 00000000 00000000 00000000
00000000 00000000
[    1.771287]  00004f00 00000000 00000000 00000000 00000000 00000000
0000100a 0000c000
[    1.771287] Call Trace:
[    1.771287]  [<c10334c4>] ? resume_userspace+0x8/0x28
[    1.771287]  [<c1033644>] ? syscall_call+0x7/0xb
[    1.771287]  [<c109007b>] ? proc_taint+0xab/0x140
[    1.771287] Code: 04 24 00 4a b6 c1 e8 69 77 7e 00 85 f6 0f 8e 57
ff ff ff eb a5 eb 0d 90 90 90 90 90 90 90 90 90 90 90 90 90 55 89 e5
53 83 ec 10 <65> a1 14 00 00 00 89 45 f8 31 c0 64 8b 1d 20 20 ee c1 8b
83 b0
[    1.771287] EIP: [<c10c1377>] lockdep_sys_exit+0x7/0xa0 SS:ESP 0068:f73e1eac
[    1.779083] ---[ end trace e93713a9d40cd06c ]---
[    6.777823] uvesafb: Getting VBE info block failed (eax=0x4f00, err=1)
[    6.777924] uvesafb: vbe_init() failed with -22
[    6.778031] uvesafb: probe of uvesafb.0 failed with error -22



Michal (uvesafb creator) suggested that I try disabling lockdep to see
what happens. Doing so changes the dmesg output to this:

[    1.648494] general protection fault: 0000 [#1] SMP
[    1.648695] last sysfs file:
[    1.648771] Modules linked in:
[    1.648903]
[    1.648979] Pid: 529, comm: v86d Not tainted (2.6.30 #1) HP
Pavilion dv5000 (ES933AS#ABA)
[    1.649086] EIP: 0060:[<c110943b>] EFLAGS: 00013082 CPU: 0
[    1.649176] EIP is at trace_hardirqs_off_caller+0xb/0x110
[    1.649261] EAX: c1033242 EBX: c1033242 ECX: 00000000 EDX: f684a000
[    1.649346] ESI: c2f54640 EDI: f684bed4 EBP: f684bebc ESP: f684be98
[    1.649381]  DS: 007b ES: 007b FS: 00d8 GS: 00e0 SS: 0068
[    1.649381] Process v86d (pid: 529, ti=f684a000 task=f7103ed0
task.ti=f684a000)
[    1.649381] Stack:
[    1.649381]  f684bed4 f684bec0 c1057457 f684bec0 f7103ed0 f684bf00
00000000 f684bfb4
[    1.649381]  f7103ed0 f684a000 c148d5e4 f684a000 00000000 00000000
c1033242 00000000
[    1.649381]  00000000 00000000 00000000 00000000 00000000 00004f00
00000000 00000000
[    1.649381] Call Trace:
[    1.649381]  [<c1057457>] ? do_sys_vm86+0x1b7/0x390
[    1.649381]  [<c148d5e4>] ? trace_hardirqs_off_thunk+0xc/0x18
[    1.649381]  [<c1033242>] ? resume_userspace+0x6/0x1c
[    1.649381]  [<c10333b0>] ? syscall_call+0x7/0xb
[    1.649381]  [<c109007b>] ? do_sysctl+0x2db/0x3a0
[    1.649381] Code: c0 a3 f4 d1 de c1 31 c0 8b 55 fc 65 33 15 14 00
00 00 75 02 c9 c3 e8 05 8b f7 ff 90 8d 74 26 00 55 89 e5 83 ec 24 89
5d f4 89 c3 <65> a1 14 00 00 00 89 45 f0 31 c0 f6 05 f8 d1 de c1 02 89
75 f8
[    1.649381] EIP: [<c110943b>] trace_hardirqs_off_caller+0xb/0x110
SS:ESP 0068:f684be98
[    1.649381] ---[ end trace e93713a9d40cd06c ]---
[    6.655118] uvesafb: Getting VBE info block failed (eax=0x4f00, err=1)
[    6.655208] uvesafb: vbe_init() failed with -22
[    6.655299] uvesafb: probe of uvesafb.0 failed with error -22



Here's the bisection log (remember to mentally reverse 'bad' and 'good'):

# bad: [45e3e1935e2857c54783291107d33323b3ef33c8] Merge branch
'master' of git://git.kernel.org/pub/scm/linux/kernel/git/sam/kbuild-next
# good: [07a2039b8eb0af4ff464efd3dfd95de5c02648c6] Linux 2.6.30
git bisect start 'HEAD' 'v2.6.30'
# bad: [8a1ca8cedd108c8e76a6ab34079d0bbb4f244799] Merge branch
'perfcounters-for-linus' of
git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip
git bisect bad 8a1ca8cedd108c8e76a6ab34079d0bbb4f244799
# bad: [6cd8e300b49332eb9eeda45816c711c198d31505] Merge branch
'kvm-updates/2.6.31' of git://git.kernel.org/pub/scm/virt/kvm/kvm
git bisect bad 6cd8e300b49332eb9eeda45816c711c198d31505
# bad: [73fbad283cfbbcf02939bdbda31fc4a30e729cca] Merge branch 'next'
into for-linus
git bisect bad 73fbad283cfbbcf02939bdbda31fc4a30e729cca
# good: [bb7762961d3ce745688e9050e914c1d3f980268d] Merge branch
'x86-mm-for-linus' of
git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip
git bisect good bb7762961d3ce745688e9050e914c1d3f980268d
# bad: [75063600fd7b27fe447112c27997f100b9e2f99b] Merge branch
'futexes-for-linus' of
git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip
git bisect bad 75063600fd7b27fe447112c27997f100b9e2f99b
# good: [6b2e8523df148c15ea5abf13075026fb8bdb3f86] xen: reserve Xen
start_info rather than e820 reserving
git bisect good 6b2e8523df148c15ea5abf13075026fb8bdb3f86
# bad: [be15f9d63b97da0065187696962331de6cd9de9e] Merge branch
'x86-xen-for-linus' of
git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip
git bisect bad be15f9d63b97da0065187696962331de6cd9de9e
# good: [bec706838ec2f9c8c2b99e88a1270d7cba159b06] Merge branch
'x86-setup-for-linus' of
git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip
git bisect good bec706838ec2f9c8c2b99e88a1270d7cba159b06
# bad: [0b8c3d5ab000c22889af7f9409799a6cdc31a2b2] x86: Clear TS in
irq_ts_save() when in an atomic section
git bisect bad 0b8c3d5ab000c22889af7f9409799a6cdc31a2b2
# bad: [3aa6b186f86c5d06d6d92d14311ffed51f091f40] x86: Fix non-lazy GS
handling in sys_vm86()
git bisect bad 3aa6b186f86c5d06d6d92d14311ffed51f091f40
# good: [4a4aca641bc4598e77b866804f47c651ec4a764d] x86: Add quirk for
reboot stalls on a Dell Optiplex 360
git bisect good 4a4aca641bc4598e77b866804f47c651ec4a764d



And here's the commit that fixes the problem:

commit 3aa6b186f86c5d06d6d92d14311ffed51f091f40
Author: Lubomir Rintel <lkundrak@...sk>
Date:   Sun Jun 7 16:23:48 2009 +0200

    x86: Fix non-lazy GS handling in sys_vm86()

    This fixes a stack corruption panic or null dereference oops
    due to a bad GS in resume_userspace() when returning from
    sys_vm86() and calling lockdep_sys_exit().

    Only a problem when CONFIG_LOCKDEP and CONFIG_CC_STACKPROTECTOR
    enabled.

    Signed-off-by: Lubomir Rintel <lkundrak@...sk>
    Cc: H. Peter Anvin <hpa@...or.com>
    LKML-Reference: <1244384628.2323.4.camel@...bo>
    Signed-off-by: Ingo Molnar <mingo@...e.hu>


The cherry-pick of the bisected commit seems to work when applied to
vanilla v2.6.30.

Greg, for v2.6.30.1, could you please cherry-pick commit
3aa6b186f86c5d06d6d92d14311ffed51f091f40?


By the way, what happened to the versioning information that used to
be contained in the kernel binary filename and 'uname -r' output? I
used to see kernels that had version numbers similar to the output of
'git describe'. It was quite useful to have that kind of revision
information built into the kernel.

- Steven
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists