Bug 30486

On 6-Current accessing a floppy drive in a DS10
causes a kernel crash. Sadly things seem to have
deteriorated.

I'll update the pr with a proper backtrace this evening.

Would it make sense to combine this with PR 30487 ?
I can't imagine there is much differen between ds10
and ds20 in the floppy drive area.

-- 
Sten Spans

"There is a crack in everything, that's how the light gets in."
Leonard Cohen - Anthem

On Tue, May 24, 2005 at 09:18:51AM +0200, Sten Spans wrote..
> 
> On 6-Current accessing a floppy drive in a DS10
> causes a kernel crash. Sadly things seem to have
> deteriorated.
> 
> I'll update the pr with a proper backtrace this evening.
> 
> Would it make sense to combine this with PR 30487 ?
> I can't imagine there is much differen between ds10
> and ds20 in the floppy drive area.

Probably correct.  And (long shot) ES40 might have the same issue.
All Tsunami based machines.

-- 
Wilko Bulte				wilko@FreeBSD.org

On Tue, 24 May 2005, Wilko Bulte wrote:

> On Tue, May 24, 2005 at 09:18:51AM +0200, Sten Spans wrote..
>>
>> On 6-Current accessing a floppy drive in a DS10
>> causes a kernel crash. Sadly things seem to have
>> deteriorated.
>>
>> I'll update the pr with a proper backtrace this evening.
>>

fatal kernel trap:

     trap entry     = 0x2 (memory management fault)
     faulting va    = 0x0
     type           = access violation
     cause          = store instruction
     pc             = 0xfffffc00005e82c0
     ra             = 0xfffffc0000600634
     sp             = 0xfffffe001a5a5b50
     curthread      = 0xfffffc003eae2a80
         pid = 7, comm = fdc0

[thread pid 7 tid 100026 ]
Stopped at      bus_dmamap_load+0x80:   stq     t0,0(t1) <0x0>  <t0=0x800000,t1=0x0>
db> trace
Tracing pid 7 tid 100026 td 0xfffffc003eae2a80
bus_dmamap_load() at bus_dmamap_load+0x80
isa_dmastart() at isa_dmastart+0x124
fdc_worker() at fdc_worker+0xc80
fdc_thread() at fdc_thread+0x88
fork_exit() at fork_exit+0xf0
exception_return() at exception_return
--- root of call graph ---
db>


-- 
Sten Spans

"There is a crack in everything, that's how the light gets in."
Leonard Cohen - Anthem

disabling dma (FDC_NODMA) recreates the RELENG_4
behaviour :

deepthought# dd if=/dev/zero of=/dev/fd0 bs=1440k count=1
dd: /dev/fd0: Input/output error
1+0 records in
0+0 records out
0 bytes transferred in 1.928038 secs (0 bytes/sec)
deepthought# fdformat /dev/fd0
Format 1440K floppy `/dev/fd0'? (y/n): y
Processing fdformat: ioctl(FD_FORM): Input/output error

So it's either kernel crash in isadma or broken pio,
gotta love free choice :}

-- 
Sten Spans

"There is a crack in everything, that's how the light gets in."
Leonard Cohen - Anthem

On Tue, May 24, 2005 at 10:19:01PM +0200, Sten Spans wrote..
> 
> disabling dma (FDC_NODMA) recreates the RELENG_4
> behaviour :
> 
> deepthought# dd if=/dev/zero of=/dev/fd0 bs=1440k count=1
> dd: /dev/fd0: Input/output error
> 1+0 records in
> 0+0 records out
> 0 bytes transferred in 1.928038 secs (0 bytes/sec)
> deepthought# fdformat /dev/fd0
> Format 1440K floppy `/dev/fd0'? (y/n): y
> Processing fdformat: ioctl(FD_FORM): Input/output error
> 
> So it's either kernel crash in isadma or broken pio,
> gotta love free choice :}

A long time ago Andrew Gallatin (gallatin@) and myself tried
to narrow it down.  ISTR Drew thought it was isadma getting something
wrong.  

-- 
Wilko Bulte				wilko@FreeBSD.org

Comment 6 Marcel Moolenaar 2005-05-25 08:27:57 UTC

State Changed
From-To: open->feedback

Please try with rev. 1.51 of src/sys/alpha/alpha/busdma_machdep.c

Comment 7 Kris Kennaway 2005-07-30 01:41:28 UTC

State Changed
From-To: feedback->closed

Feedback timeout