Messages in this thread | | | Date | Tue, 14 Apr 2009 18:11:54 +0100 (BST) | From | Hugh Dickins <> | Subject | Re: [2.6.30-rc1] RCU detected CPU 1 stall |
| |
On Mon, 13 Apr 2009, Tetsuo Handa wrote: > Paul E. McKenney wrote: > > Is this reproducible? > Not always, but it is reproducible. > > Al Viro wrote: > > I'd really love to see results of repeated alt-sysrq-p/alt-sysrq-l, just > > to see where was it actually spinning. > Below is sysrq message. > Maybe something related to khelper's current->mm == NULL warning problem.
Maybe, up to a point, and I'll post separately on those warnings.
> Full log is at http://I-love.SAKURA.ne.jp/tmp/dmesg-2.6.30-rc1-200904130930.txt . > > [ 47.412519] SysRq : Show Regs > [ 47.413986] > [ 47.414584] Pid: 3655, comm: khelper Tainted: G W (2.6.30-rc1 #1) VMware Virtual Platform > [ 47.415804] EIP: 0060:[<c0379c3d>] EFLAGS: 00000293 CPU: 0 > [ 47.415804] EIP is at __get_user_4+0x11/0x17 > [ 47.415804] EAX: f7150003 EBX: f7150000 ECX: 00000000 EDX: f6744000 > [ 47.415804] ESI: 000007b8 EDI: 7fffffff EBP: f6744f20 ESP: f6744f10 > [ 47.415804] DS: 007b ES: 007b FS: 00d8 GS: 0000 SS: 0068 > [ 47.415804] CR0: 8005003b CR2: f7150000 CR3: 3599a000 CR4: 000006d0 > [ 47.415804] DR0: 00000000 DR1: 00000000 DR2: 00000000 DR3: 00000000 > [ 47.415804] DR6: ffff0ff0 DR7: 00000400 > [ 47.415804] Call Trace: > [ 47.415804] [<c0225f4e>] ? count+0x3e/0xb0 > [ 47.415804] [<c0228581>] do_execve+0x621/0x890 > [ 47.415804] [<c022bd8b>] ? getname+0x6b/0xa0 > [ 47.415804] [<c010237e>] sys_execve+0x5e/0xb0 > [ 47.415804] [<c0103d19>] syscall_call+0x7/0xb > [ 47.415804] [<c010aee4>] ? kernel_execve+0x24/0x30 > [ 47.415804] [<c0172b6f>] ? ____call_usermodehelper+0xff/0x170 > [ 47.415804] [<c0172a70>] ? ____call_usermodehelper+0x0/0x170 > [ 47.415804] [<c0104707>] ? kernel_thread_helper+0x7/0x10
I'm now thinking that this really is hitting in count(), despite the ? on that in the backtrace, and is entirely unrelated to the recent check_unsafe_exec() changes. Stuck in a loop scanning the the kernelspace exec args without an mm.
But my compiler on your config gives quite different function sizes: please would you post to the list or send me privately the output of "objdump -trd fs/exec.o", so we can check that.
Thanks, Hugh
| |