2.6.0-test9 breaks cdrecord w. ide-scsi device

2.6.0-test9 breaks cdrecord w. ide-scsi device

[P. Christeas]

Just rebooted my long-running machine and now strongly suspect that it is a 
kernel bug:
I try to burn (dummy mode) 20 tracks of audio on a 'PIONEER DVD-RW  DVR-106D' 
drive.
At -test8 it would work (tested 4 times now, several with even older kernels).
On -test 9 I always (> 15 times tested) get the messages quoted.
The drive is on ide-scsi [M],  sg is built-in, sr is a module. Ask me anything 
else you may want me to try..


cdrecord 2.01.xx :
Track 01:   31 of   38 MB written (fifo 100%) [buf  90%]  19.8x.cdrecord:
 Input/output error. write_g1: scsi sendcmd: no error CDB:  2A 00 00 00 36 F3
 00 00 1B 00
status: 0x2 (CHECK CONDITION)
Sense Bytes: 70 00 04 00 D3 10 FA 0E 26 01 30 FE 08 01 00 00
Sense Key: 0x4 Hardware Error, Segment 0
Sense Code: 0x08 Qual 0x01 (logical unit communication time-out) Fru 0x0
Sense flags: Blk 13832442 (not valid)
cmd finished after 2.004s timeout 200s

write track data: error after 33085584 bytes
cdrecord: A write error occured.
cdrecord: Please properly read the error message above.
Writing  time:   28.974s
Average write speed 154.5x.
Min drive buffer fill was 79%
Fixating...


cdrecord 1.10:
Track 04:   1 of  33 MB written (fifo  98%).cdrecord.old: Input/output error.
 write_g1: scsi sendcmd: no error CDB:  2A 00 00 00 BE 9D 00 00 1B 00
status: 0x2 (CHECK CONDITION)
Sense Bytes: 70 00 04 00 D3 10 FA 0E 26 01 30 FE 08 01 00 00
Sense Key: 0x4 Hardware Error, Segment 0
Sense Code: 0x08 Qual 0x01 (logical unit communication time-out) Fru 0x0
Sense flags: Blk 13832442 (not valid)
cmd finished after 2.004s timeout 200s

write track data: error after 1460592 bytes
Sense Bytes: 70 00 00 00 00 00 00 0E 00 00 00 00 00 00 00 00 00 00
Writing  time:   57.892s
Fixating...

Re: 2.6.0-test9 breaks cdrecord w. ide-scsi device

[P. Christeas]

> Hello
>
> I don't really know what the problem with ide-scsi is (I'm not much of a
> programmer) but is there any particular reason you're clinging to the old
> ide-scsi method for CD writing? This has been broken a few times now in the
> 2.5/2.6 tree. Even when it is working, it's still somewhat broken and they
> know it. I have found that the maximum speed I can get is 16x using
> ide-scsi in the 2.6 tree.
>
> There is now ide-cd writing support, and cdrecord 2 supports it. I build
> that stuff modular, so I just load the ide-cd and isofs modules (modprobe
> takes care of the rest). I achieve the fastest writing speeds I've ever
> had, using this driver. Far better than ide-scsi or even Sleazy CD Creator
> in Windows.
>
> In the cdrecord command, everything is the same except you use
> dev=/dev/hdxx instead of the dev=x,x,x scsi target notation. For example:
>
> cdrecord -v speed=48 dev=/dev/hdd -data filename.iso
>
> Just in case you didn't know this or have never tried this. I hope it helps
> you.
>
It's true that I hadn't tried the native ATAPI method lately.
My point, however, was that there is a regression from -test8 to -test9 about 
cdrecord-ing.

Re: 2.6.0-test9 breaks cdrecord w. ide-scsi device

[P. Christeas]

> Hello
>
> ...
>  
> There is now ide-cd writing support, and cdrecord 2 supports it. I build
> that stuff modular, so I just load the ide-cd and isofs modules (modprobe
> takes care of the rest). I achieve the fastest writing speeds I've ever
> had, using this driver. Far better than ide-scsi or even Sleazy CD Creator
> in Windows.
>
>...

Tried it. Now used /dev/hdd there..
This time it worked a few times and some other times it failed. 
It now seems that I got a buffer underrun, although I do have 'realtime 
priority' for the process.
Note that I'm not using 'burnfree'. That's why I have linux anyway :) .

Can anybody examine why is the buffer of cdrecord less than 100% with 
RR-scheduling? That would never happen in 2.4, cdrecord 1.10 ..


Track 10:   25 of   32 MB written (fifo  25%) [buf  94%]  15.8x.cdrecord: 
Success. write_g1: scsi sendcmd: no error
CDB:  2A 00 00 02 3C 76 00 00 1B 00
status: 0x2 (CHECK CONDITION)
Sense Bytes: F1 00 05 00 32 35 73 0E 00 00 00 00 21 02 00 00
Sense Key: 0x5 Illegal Request, deferred error, Segment 0
Sense Code: 0x21 Qual 0x02 (invalid address for write) Fru 0x0
Sense flags: Blk 3290483 (valid)
resid: 63504
cmd finished after 0.006s timeout 200s

write track data: error after 27179712 bytes
cdrecord: The current problem looks like a buffer underrun.
cdrecord: Try to use 'driveropts=burnfree'.
cdrecord: Make sure that you are root, enable DMA and check your HW/OS set up.
cdrecord: Success. test unit ready: scsi sendcmd: no error
CDB:  00 00 00 00 00 00
status: 0x2 (CHECK CONDITION)
Sense Bytes: 70 00 02 00 00 00 00 0E 00 00 00 00 30 00 00 00
Sense Key: 0x2 Not Ready, Segment 0
Sense Code: 0x30 Qual 0x00 (incompatible medium installed) Fru 0x0
Sense flags: Blk 0 (not valid)
cmd finished after 0.005s timeout 200s
Writing  time:  139.113s
Average write speed  32.0x.
Min drive buffer fill was 80%
Fixating...
WARNING: Some drives don't like fixation in dummy mode.
cdrecord: Success. flush cache: scsi sendcmd: no error
CDB:  35 00 00 00 00 00 00 00 00 00
status: 0x2 (CHECK CONDITION)
Sense Bytes: 70 00 02 00 00 00 00 0E 00 00 00 00 30 00 00 00
Sense Key: 0x2 Not Ready, Segment 0
Sense Code: 0x30 Qual 0x00 (incompatible medium installed) Fru 0x0
Sense flags: Blk 0 (not valid)
cmd finished after 0.005s timeout 200s
Trouble flushing the cache
cdrecord: Success. test unit ready: scsi sendcmd: no error
CDB:  00 00 00 00 00 00
status: 0x2 (CHECK CONDITION)
Sense Bytes: 70 00 02 00 00 00 00 0E 00 00 00 00 30 00 00 00
Sense Key: 0x2 Not Ready, Segment 0
Sense Code: 0x30 Qual 0x00 (incompatible medium installed) Fru 0x0
Sense flags: Blk 0 (not valid)
cmd finished after 0.006s timeout 200s
Fixating time:   26.117s


hdparm:
IO_support   =  3 (32-bit w/sync)
 unmaskirq    =  1 (on)
 using_dma    =  1 (on)
 keepsettings =  0 (off)
 readonly     =  1 (on)

Re: 2.6.0-test9 breaks cdrecord w. ide-scsi device

[bill davidsen]

In article <200311011219.21890.p_christ@hol.gr>,
P. Christeas <p_christ@hol.gr> wrote:
| > Hello
| >
| > ...
| >  
| > There is now ide-cd writing support, and cdrecord 2 supports it. I build
| > that stuff modular, so I just load the ide-cd and isofs modules (modprobe
| > takes care of the rest). I achieve the fastest writing speeds I've ever
| > had, using this driver. Far better than ide-scsi or even Sleazy CD Creator
| > in Windows.
| >
| >...
| 
| Tried it. Now used /dev/hdd there..
| This time it worked a few times and some other times it failed. 
| It now seems that I got a buffer underrun, although I do have 'realtime 
| priority' for the process.
| Note that I'm not using 'burnfree'. That's why I have linux anyway :) .
| 
| Can anybody examine why is the buffer of cdrecord less than 100% with 
| RR-scheduling? That would never happen in 2.4, cdrecord 1.10 ..

If you haven't solved this yet (firmware current?) you might try using a
large fifo, starting a burn, then looking at the system with vmstat to
see if there's a lot of system time. If so, I suspect the you are using
PIO in spite of the new method and selection of DMA for the device.

In any case it will provide a data point.
-- 
bill davidsen <davidsen@tmr.com>
  CTO, TMR Associates, Inc
Doing interesting things with little computers since 1979.