Re: Race condition in interrupt handling

Here is another run triggering this bug:

Assertion irq_tcb->get_state().is_halted() failed in file
/home/mai4/work/dOS/src/dOS/l4ka/kernel/src/api/v4/interrupt.cc, line 213
(fn=ffffffffc0607664)
--- "KD# assert" ---
--------------------------------- (eip=ffffffffc0618ba5,
esp=fffffffe80040f10) ---
> showtcb
tcb/tid/name [current]: IRQ_11

=== TCB: fffffffe8000b000 === ID: 0000000b00000001 =
ffffffffffffffc0/ffffffffc0cd2400 === PRIO: 0xff === CPU: 0 ===
UIP: fffffffe8000b000   queues: rSwl              wait : NIL_THRD
 :NIL_THRD           space: 0000000000000000
USP: 0000000000000000   tstate: POLLING           ready: NIL_THRD
 :NIL_THRD           pdir : 0000000000cb3000
KSP: fffffffe8000bf98   sndhd : NIL_THRD          send :
IRQ_000000000011:IRQ_000000000011   pager: NIL_THRD
total quant:                   0us, ts length  :                  10000us,
curr ts:            10000us
abs timeout:                   0us, rel timeout:                      0us
sens prio: 255, delay: max=0us, curr=0us
resources: 0000000000000000 []   flags: 0000000000000000 [t]
partner: 0000004000000001, saved partner: NIL_THRD, saved state: ABORTED ,
scheduler: 0000004000000001
> showtcb
tcb/tid/name [current]: 0000004000000001
=== TCB: fffffffe80040000 === ID: 0000004000000001 =
0000000080000200/ffffffffc0cb5000 === PRIO: 0x64 === CPU: 0 ===
UIP: 00000000020001df   queues: Rswl              wait :
0000004100000001:0000004100000001   space: ffffffffc0cb2000
USP: 00007fff7ffffcf8   tstate: RUNNING           ready:
0000004400000001:0000003e00000001   pdir : 0000000000cb3000
KSP: fffffffe80040e10   sndhd : IRQ_000000000011  send : NIL_THRD
 :NIL_THRD           pager: ROOTTASK
total quant:                   0us, ts length  :                  10000us,
curr ts:             3087us
abs timeout:            50661820us, rel timeout:       -        15554645us
sens prio: 100, delay: max=0us, curr=0us
resources: 0000000000000000 []   flags: 0000000000000000 [t]
partner: NIL_THRD, saved partner: NIL_THRD, saved state: ABORTED ,
scheduler: ROOTTASK

Haohui

On Fri, Apr 23, 2010 at 3:15 PM, Mai, Haohui <[email protected]> wrote:

> It's pretty difficult to reproduce this problem since it only happens once
> a while, Here is some information when the kernel goes wild:
>
> Assertion irq_tcb->get_state().is_halted() failed in file
> /home/mai4/work/dOS/src/dOS/l4ka/kernel/src/api/v4/interrupt.cc, line 213
> (fn=ffffffffc0607664)
> --- "KD# assert" ---
> --------------------------------- (eip=ffffffffc0618ba5,
> esp=fffffffe8003ef10) ---
> > showqueue
>
> [255]: (SIGMA0:0) (ROOTTASK:0) (IRQ_12:0) (IRQ_11:0)
> [100]: (0000003b00000001:0) (0000003c00000001:0) (0000003d00000001:0)
> 0000003e00000001:0 (0000003f00000001:0) 0000004000000001:0
> 0000004100000001:0 0000004200000001:0 (0000004300000001:0)
> 0000004400000001:0
> [000]: (0000001800000001:0)
> idle : IDLETHRD
>
> > showtcb
> tcb/tid/name [current]: IRQ_11
>
> === TCB: fffffffe8000b000 === ID: 0000000b00000001 =
> ffffffffffffffc0/ffffffffc0cd2400 === PRIO: 0xff === CPU: 0 ===
> UIP: fffffffe8000b000   queues: rswl              wait : NIL_THRD
>  :NIL_THRD           space: 0000000000000000
> USP: 0000000000000000   tstate: WAIT_FE           ready: NIL_THRD
>  :NIL_THRD           pdir : 0000000000cb3000
> KSP: fffffffe8000bf98   sndhd : NIL_THRD          send : NIL_THRD
>  :NIL_THRD           pager: NIL_THRD
> total quant:                   0us, ts length  :                  10000us,
> curr ts:            10000us
> abs timeout:                   0us, rel timeout:                      0us
> sens prio: 255, delay: max=0us, curr=0us
> resources: 0000000000000000 []   flags: 0000000000000000 [t]
> partner: 0000004000000001, saved partner: NIL_THRD, saved state: ABORTED ,
> scheduler: 0000004000000001
>
> > showtcb
> tcb/tid/name [current]: 4000000001
> === TCB: fffffffe80040000 === ID: 0000004000000001 =
> 0000000080000200/ffffffffc0cb5000 === PRIO: 0x64 === CPU: 0 ===
> UIP: 00000000020001df   queues: Rswl              wait :
> 0000004100000001:0000004100000001   space: ffffffffc0cb2000
> USP: 00007fff7ffffd08   tstate: RUNNING           ready:
> 0000003e00000001:0000004200000001   pdir : 0000000000cb3000
> KSP: fffffffe80040ee8   sndhd : NIL_THRD          send : NIL_THRD
>  :NIL_THRD           pager: ROOTTASK
> total quant:                   0us, ts length  :                  10000us,
> curr ts:             8476us
> abs timeout:            50581747us, rel timeout:       -         9969065us
> sens prio: 100, delay: max=0us, curr=0us
> resources: 0000000000000000 []   flags: 0000000000000000 [t]
> partner: IRQ_11, saved partner: NIL_THRD, saved state: ABORTED , scheduler:
> ROOTTASK
> >
>
> I'm wondering why the IRQ thread is in WAIT_FE state.. Do you have an idea?
>
> Haohui
>
> On Fri, Apr 16, 2010 at 2:15 AM, Jan Stoess <[email protected]> wrote:
>
>> > Actually I'm hitting this bug once a while under qemu. It seems to me
>> that
>> > handle_interrupt and irq_thread() are executed on different CPUs.
>> >
>> > What do you need me to do to clarify the problem?
>>
>> Can you dump the tracebuffer output when the assert hits and send the dump
>> over here?
>>
>> --
>> Jan Stoess
>> KIT/UKa System Architecture Group
>> Phone: +49 (721) 608 4056
>> Fax: +49 (721) 608 7664
>> http://os.ibds.kit.edu/stoess
>>
>>
>