Re: [BUG] in 2.6.5 at msr.c and cpuid.c

Re: [BUG] in 2.6.5 at msr.c and cpuid.c

[H. Peter Anvin]

Bjoern Schmidt wrote:
> Hello H. Peter,
> 
> sorry for mailing to you directly, but sending this bugreport
> to linux-kernel@vger.kernel.org failed and i don't know why...
>
> The server was at an uptime of ~8 days until this bug appeared.
> At 15:33 the smbd was killed by Signal 7.
> 

Why is your smbd touching /dev/cpu/*/msr?  Something is very odd about
that... assuming you're not exporting /dev through Samba I would guess
this is a preemption bug.

Could you please send me a) your System.map and b) your kernel
configuration?

Could you also try disabling preemption and see if that helps?

	-hpa

Re: [BUG] in 2.6.5 at msr.c and cpuid.c

[H. Peter Anvin]

This is pretty interesting... whomever changed cpuid.c/msr.c to the new
cpu_online() macro missed the fact that cpu_online() doesn't handle
out-of-range numbers (on SMP it'll return garbage, on UP it will BUG);
it still doesn't change the fact that it shouldn't have been there in
the first place...

Bjoern Schmidt wrote:
> 
> Apr 18 15:05:51 kilobyte kernel: ------------[ cut here ]------------
> Apr 18 15:05:51 kilobyte kernel: kernel BUG at arch/i386/kernel/msr.c:244!
> Apr 18 15:05:51 kilobyte kernel: invalid operand: 0000 [#1]
> Apr 18 15:05:51 kilobyte kernel: PREEMPT
> Apr 18 15:05:51 kilobyte kernel: CPU:    0
> Apr 18 15:05:51 kilobyte kernel: EIP:    0060:[<c0115482>]    Not tainted
> Apr 18 15:05:51 kilobyte kernel: EFLAGS: 00010202   (2.6.5)
> Apr 18 15:05:51 kilobyte kernel: EIP is at msr_open+0x32/0x40
> Apr 18 15:05:51 kilobyte kernel: eax: c035fae0   ebx: c7d64000   ecx:
> c9dfbc9c   edx: 00000001
> Apr 18 15:05:51 kilobyte kernel: esi: c9dfbc60   edi: 00000000   ebp:
> c0115450   esp: c7d65f18
> Apr 18 15:05:51 kilobyte kernel: ds: 007b   es: 007b   ss: 0068
> Apr 18 15:05:51 kilobyte kernel: Process smbd (pid: 17921,
> threadinfo=c7d64000 task=c8a8a6c0)
> Apr 18 15:05:51 kilobyte kernel: Stack: c01595a2 c3866ad4 c326c0c0
> c7d65f28 00000001 c326c0c0 c3866ad4 00000001
> Apr 18 15:05:51 kilobyte kernel: c9fec1e0 c014f680 c3866ad4 c326c0c0
> 404c88d7 00008000 00008000 c4d40000
> Apr 18 15:05:51 kilobyte kernel: c7d64000 c014f528 c269a160 c9fec1e0
> 00008000 c7d65f70 c269a160 c9fec1e0
> Apr 18 15:05:51 kilobyte kernel: Call Trace:
> Apr 18 15:05:51 kilobyte kernel: [<c01595a2>] chrdev_open+0xf2/0x220
> Apr 18 15:05:51 kilobyte kernel: [<c014f680>] dentry_open+0x150/0x210
> Apr 18 15:05:51 kilobyte kernel: [<c014f528>] filp_open+0x68/0x70
> Apr 18 15:05:51 kilobyte kernel: [<c014f9cb>] sys_open+0x5b/0x90
> Apr 18 15:05:51 kilobyte kernel: [<c0107319>] sysenter_past_esp+0x52/0x71
> Apr 18 15:05:51 kilobyte kernel:
> Apr 18 15:05:51 kilobyte kernel: Code: 0f 0b f4 00 4a 32 32 c0 eb e6 90
> 90 90 90 55 b8 00 e0 ff ff