Re: "aborting sbp2 command" expected in kernel.org FireWire maybe

Mark K:

> hdparm -tT /dev/scd0
> under both 2.4.26 and 2.6.6.

Whole `script` logs I quote inline below for 2.6.6 and 2.4.27, past all
my English and signature. Consistent summary results now are:

"aborting sbp2 command" in dd of 2.6 of kernel.org
"aborting sbp2 command" in hdparm -tT or dd of 2.6.7 of Knoppix 3.6

"10.80 MB/sec" in hdparm -tT of 2.6.6 of Knoppix 3.4 (and 11.4 MB/s in
dd)
"24.10 MB/sec" in hdparm -tT of 2.4.27 of Knoppix 3.6 (and 24.6 MB/s in
dd)

25.3 MB/s in dd when SCSI over something other than Linux FireWire.

You can see, in Knoppix, my runlevel was 4, /dev/scd0 was my boot CD,
/dev/scd1 was my FireWire drive, bash time told me elapsed real time,
and Knoppix dd calculated B/s for me.

> > Knoppix 3.6 ... 3.4 ...

When writing before, I had only Knoppix 3.4, now I also have the 3.6
that a friend found for me.

>> sudo blockdev --flushbufs /dev/scd0
>> sudo dd if=/dev/scd0 bs=1M skip=0 count=1 >/dev/null
> ...
> I'm guessing that 'count=1' tells you the number of bytes dd moves?

Me, I say dd if=$if bs=$bs skip=$skip count=$count because I think it
means,

Open $if, seek to ($skip * $bs) bytes, and then loop $count times,
reading $bs bytes in each iteration of the loop.

All that happens way way up high, above the /dev/scd0 level of the
kernel. The consequent bus traces vary greatly, courtesy cache. If I
ever I find the source for dd, I'll post patches for it e.g. let me
seek to a hex LBA even when $bs != bytes per LBA. Meanwhile, I rolled
my own pldd to give us results seen when doing exactly this but
bypassing the cache via ioctl SG_IO, ioctl CDROM_SEND_PACKET, etc. Do
you want to see those results?

> kernel.org is what the 1394 developers write.

Good to hear. False Google rumour was that experiments with
linux1394.org sources get more attention.

> it may be that some distro's test and modify 1394 to somehow improve it

People tell me Debian 1394 sources differ. Why I do not yet know.

> once we agree on and understand the data.

Thank you for your patient interest. Better now?

Pat LaVarre

--- 3.4 boot: knoppix26 4
--- then e-mailed via: init 5

$ cat /etc/knoppix-version
3.4 2004-05-17
$ uname -msr
Linux 2.6.6 i686
$ sudo modprobe sbp2
$ cat /proc/scsi/scsi
Attached devices:
Host: scsi0 Channel: 00 Id: 00 Lun: 00
   Vendor: LITE-ON Model: COMBO LTC-48161H Rev: KH0K
   Type: CD-ROM ANSI SCSI revision: 02
Host: scsi1 Channel: 00 Id: 00 Lun: 00
   Vendor: Iomega Model: RRD Rev: 83.B
   Type: CD-ROM ANSI SCSI revision: 02
$
$ sudo hdparm -tT /dev/scd1

/dev/scd1:
  Timing buffer-cache reads: 2380 MB in 2.00 seconds = 1188.40 MB/sec
  Timing buffered disk reads: 34 MB in 3.15 seconds = 10.80 MB/sec
$
$ sudo hdparm -tT /dev/scd1

/dev/scd1:
  Timing buffer-cache reads: 2224 MB in 2.00 seconds = 1111.61 MB/sec
  Timing buffered disk reads: 34 MB in 3.15 seconds = 10.80 MB/sec
$
$ sudo hdparm -tT /dev/scd1

/dev/scd1:
  Timing buffer-cache reads: 2324 MB in 2.00 seconds = 1161.60 MB/sec
  Timing buffered disk reads: 34 MB in 3.15 seconds = 10.81 MB/sec
$
$ sudo blockdev --flushbufs /dev/scd1
$ time sudo dd if=/dev/scd1 bs=1M skip=0 count=200 >/dev/null
200+0 records in
200+0 records out
209715200 bytes transferred in 18.475686 seconds (11350875 bytes/sec)

real 0m18.490s
user 0m0.002s
sys 0m1.087s
$
$ # show: no dmesg emitted
$ dmesg | tail -8
sbp2: $Rev: 1205 $ Ben Collins <bcollins@debian.org>
scsi1 : SCSI emulation for IEEE-1394 SBP-2 Devices
ieee1394: sbp2: Logged into SBP-2 device
ieee1394: Node 0-00:1023: Max speed [S400] - Max payload [2048]
   Vendor: Iomega Model: RRD Rev: 83.B
   Type: CD-ROM ANSI SCSI revision: 02
sr1: scsi3-mmc drive: 125x/125x caddy
Attached scsi CD-ROM sr1 at scsi1, channel 0, id 0, lun 0
$

--- 3.6 boot: knoppix 4
--- then e-mailed via: init 5

$ uname -msr
Linux 2.4.27 i686
$
$ cat /etc/knoppix-version
3.6 2004-08-16
$ uname -msr
Linux 2.4.27 i686
$ # no explicit modprobe required
$ cat /proc/scsi/scsi
Attached devices:
Host: scsi0 Channel: 00 Id: 00 Lun: 00
   Vendor: LITE-ON Model: COMBO LTC-48161H Rev: KH0K
   Type: CD-ROM ANSI SCSI revision: 02
Host: scsi2 Channel: 00 Id: 00 Lun: 00
   Vendor: Iomega Model: RRD Rev: 83.B
   Type: CD-ROM ANSI SCSI revision: 02
$
$ # part 1 of 2 to show: no dmesg emitted
$ echo h | sudo dd of=/proc/sysrq-trigger
0+1 records in
0+1 records out
2 bytes transferred in 0.005006 seconds (400 bytes/sec)
$ dmesg | tail -1
SysRq : HELP : loglevel0-8 reBoot tErm kIll saK showMem Off showPc
unRaw Sync showTasks Unmount
$
$ sudo hdparm -tT /dev/scd1

/dev/scd1:
  Timing buffer-cache reads: 2448 MB in 2.00 seconds = 1224.00 MB/sec
  Timing buffered disk reads: 74 MB in 3.07 seconds = 24.10 MB/sec
$
$ sudo hdparm -tT /dev/scd1

/dev/scd1:
  Timing buffer-cache reads: 2248 MB in 2.00 seconds = 1124.00 MB/sec
  Timing buffered disk reads: 74 MB in 3.07 seconds = 24.10 MB/sec
$
$ sudo hdparm -tT /dev/scd1

/dev/scd1:
  Timing buffer-cache reads: 2240 MB in 2.00 seconds = 1120.00 MB/sec
  Timing buffered disk reads: 74 MB in 3.07 seconds = 24.10 MB/sec
$
$ sudo blockdev --flushbufs /dev/scd1
$ time sudo dd if=/dev/scd1 bs=1M skip=0 count=200 >/dev/null
200+0 records in
200+0 records out
209715200 bytes transferred in 8.512928 seconds (24634908 bytes/sec)

real 0m8.536s
user 0m0.000s
sys 0m0.530s
$
$ # part 2 of 2 to show: no dmesg emitted
$ dmesg | tail -1
SysRq : HELP : loglevel0-8 reBoot tErm kIll saK showMem Off showPc
unRaw Sync showTasks Unmount
$

--- 3.6 boot: knoppix26 4
--- then e-mailed via: init 5

$ cat /etc/knoppix-version
3.6 2004-08-16
$ uname -msr
Linux 2.6.7 i686
$ runlevel
N 4
$

---
-------------------------------------------------------
This SF.Net email is sponsored by BEA Weblogic Workshop
FREE Java Enterprise J2EE developer tools!
Get your free copy of BEA WebLogic Workshop 8.1 today.
http://ads.osdn.com/?ad_id=5047&alloc_id=10808&op=click
_______________________________________________
mailing list Linux1394-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux1394-user