Re: [PATCH 00 of 31] x86: unification and xen updates

[Jeremy Fitzhardinge]

Ingo Molnar wrote:
> * Jeremy Fitzhardinge <jeremy@xxxxxxxx> wrote:
>
>   
> > > config attached. randconfig generated, so it can have random 
> > > surprises enabled :)
> > >       
> > Thanks.  What was the workload?  Did it just happen, or were you doing 
> > something specific?
> >     
>
> simple bootup triggered it.
>   

Haven't managed to repro it yet, but DEBUG_PAGE_ALLOC turned up another 
case of doing tlb flushes with preempt on:

BUG: using smp_processor_id() in preemptible [00000000] code: init/1
caller is xen_flush_tlb+0xe/0xa2
Pid: 1, comm: init Not tainted 2.6.25-rc6-x86-latest.git #281
[<c0234eed>] debug_smp_processor_id+0x99/0xb0
[<c0103ca0>] xen_flush_tlb+0xe/0xa2
[<c011d46b>] kernel_map_pages+0x104/0x125
[<c0154f45>] get_page_from_freelist+0x32f/0x433
[<c01444c4>] ? lock_acquire+0x90/0x9d
[<c01550bd>] __alloc_pages+0x5e/0x2a5
[<c015c5b3>] do_wp_page+0x2c9/0x653
[<c017953f>] ? do_lookup+0x4f/0x146
[<c0103e11>] ? xen_restore_fl+0x2e/0x52
[<c01444c4>] ? lock_acquire+0x90/0x9d
[<c015f720>] ? handle_mm_fault+0xa37/0xbb8
[<c015f7a4>] handle_mm_fault+0xabb/0xbb8
[<c0179332>] ? path_put+0x20/0x23
[<c0103e11>] ? xen_restore_fl+0x2e/0x52
[<c0103e11>] ? xen_restore_fl+0x2e/0x52
[<c045699d>] ? do_page_fault+0x460/0x952
[<c013b80d>] ? down_read_trylock+0x37/0x41
[<c0456a51>] do_page_fault+0x514/0x952
[<c0232116>] ? copy_to_user+0x2a/0x34
[<c017e3ce>] ? sys_select+0x13e/0x164
[<c045653d>] ? do_page_fault+0x0/0x952
[<c0454eaa>] error_code+0x72/0x78


This code seems a bit fast and loose anyway, since it only does a local 
TLB flush, which means that other CPUs can still accessed free memory if 
they still have a stale TLB for it.  Having the tlb flush happen on a 
random CPU doesn't really affect things much.

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