Re: [PATCH 05/05] x86: Rename paravirtualized TSC functions

[Ingo Molnar <mingo@elte.hu>]

* Yinghai Lu <yhlu.kernel@gmail.com> wrote:

> On Wed, Jul 9, 2008 at 10:30 AM, Alok Kataria <akataria@vmware.com> wrote:
> > On Wed, 2008-07-09 at 00:20 -0700, Yinghai Lu wrote:
> >> On Tue, Jul 8, 2008 at 11:13 PM, Ingo Molnar <mingo@elte.hu> wrote:
> >> >
> >> > * Alok Kataria <akataria@vmware.com> wrote:
> >> >
> >
> >> got
> >> calling  ixgb_init_module+0x0/0x76
> >> Intel(R) PRO/10GbE Network Driver - version 1.0.126-k4
> >> Copyright (c) 1999-2006 Intel Corporation.
> >> vendor=1022 device=7458
> >> ACPI: PCI Interrupt 0000:0c:01.0[A] -> GSI 56 (level, low) -> IRQ 56
> >> ixgb: eth18: ixgb_probe: Intel(R) PRO/10GbE Network Connection
> >> BUG: unable to handle kernel NULL pointer dereference at 0000000000000000
> >> IP: [<ffffffff80253b17>] hrtick_start_fair+0x89/0x173
> >> PGD 0
> >> Oops: 0000 [1] SMP
> >> CPU 4
> >> Modules linked in:
> >> Pid: 103, comm: events/4 Not tainted 2.6.26-rc9-tip-00026-geae1aa0-dirty #240
> >> RIP: 0010:[<ffffffff80253b17>]  [<ffffffff80253b17>]
> >> hrtick_start_fair+0x89/0x173
> >> RSP: 0018:ffff88082481fbd0  EFLAGS: 00010046
> >> RAX: 0000000000000000 RBX: ffff880824828000 RCX: 0000000000000006
> >> RDX: ffff88084423e700 RSI: ffff880824828000 RDI: ffff88084423e700
> >> RBP: ffff88082481fc00 R08: 0000000000000000 R09: 0000000000000000
> >> R10: 0000000000000000 R11: 0000000000000000 R12: ffff88084423e700
> >> R13: ffff880844239700 R14: ffff880824828000 R15: 0000000000000c1a
> >> FS:  0000000000000000(0000) GS:ffff881024c39600(0000) knlGS:0000000000000000
> >> CS:  0010 DS: 0018 ES: 0018 CR0: 000000008005003b
> >> CR2: 0000000000000000 CR3: 0000000000201000 CR4: 00000000000006e0
> >> DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
> >> DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
> >> Process events/4 (pid: 103, threadinfo ffff88082481e000, task ffff880824814230)
> >> Stack:  ffff880844248118 000000005c0ee9e3 ffff880844239700 ffff88082490d878
> >>  ffff88084423e700 0000000000000000 ffff88082481fc40 ffffffff80256591
> >>  ffff88082481fc30 ffffffff80516fc5 000000005c0ee9e3 ffff88082490d840
> >> Call Trace:
> >>  [<ffffffff80256591>] dequeue_task_fair+0x5f/0x7e
> >>  [<ffffffff80516fc5>] ? __first_cpu+0x26/0x49
> >>  [<ffffffff80252792>] dequeue_task+0xce/0xf0
> >>  [<ffffffff80252835>] deactivate_task+0x31/0x50
> >>  [<ffffffff80252b93>] pull_task+0x2c/0x78
> >>  [<ffffffff80254c04>] load_balance_fair+0x18d/0x277
> >>  [<ffffffff80a49f02>] schedule+0x3db/0x962
> >>  [<ffffffff802c0f0f>] ? vmstat_update+0x0/0x5e
> >>  [<ffffffff802787c9>] ? schedule_delayed_work+0x31/0x48
> >>  [<ffffffff802779ee>] worker_thread+0xbb/0x114
> >>  [<ffffffff8027c5ec>] ? autoremove_wake_function+0x0/0x63
> >>  [<ffffffff80277933>] ? worker_thread+0x0/0x114
> >>  [<ffffffff8027c147>] kthread+0x61/0xa4
> >>  [<ffffffff802614e5>] ? schedule_tail+0x36/0x81
> >>  [<ffffffff8022b509>] child_rip+0xa/0x11
> >>  [<ffffffff8027c0e6>] ? kthread+0x0/0xa4
> >>  [<ffffffff8022b4ff>] ? child_rip+0x0/0x11
> >>
> >>
> >> Code: c3 80 e8 86 04 01 00 80 3d 76 52 c0 00 00 0f 89 e1 00 00 00 41
> >> f6 84 24 70 08 00 00 04 0f 85 d2 00 00 00 49 8b 84 24 a8 08 00 00 <48>
> >> 8b 00 83 b8 c0 00 00 00 00 0f 84 ba 00 00 00 49 83 7d 10 01
> >> RIP  [<ffffffff80253b17>] hrtick_start_fair+0x89/0x173
> >>  RSP <ffff88082481fbd0>
> >> CR2: 0000000000000000
> >> ---[ end trace c05d5c1f5b126388 ]---
> >>
> >> yesterday tip/mater with tip/x86/modules
> >> tip-history-2008-07-08_16.08_Tue works well.
> >>
> >> others traps merge seems not to cause the problem..
> >>
> >
> > Hi Yinghai,
> >
> > Are we sure that these patches cause this null pointer dereference ?
> > The panic in scheduler seems to be totally disconnected to the changes
> > that these patches make. The only scheduler bit that we touch is the
> > sched_clock thingy....but that too looks harmless to me.
> >
> > Can you please bisect and see if the first patch in the series  is the
> > problem ?
> 
> tries last night,   it seems pgtable related patches cause that.

that would be the tip/xen64 stuff, right? Does this revert:

|    Revert "x86_64: there's no need to preallocate level1_fixmap_pgt"
|
|    This reverts commit 033786969d1d1b5af12a32a19d3a760314d05329.
|
|    Suresh Siddha reported that this broke booting on his 2GB testbox.

solve your problems, or are there other problems still?

	Ingo