On Tuesday, 07/22/2008 at 01:57 EDT, "Schuh, Richard" <[EMAIL PROTECTED]>
wrote:
> TRACE TYPE IO, CPU 0005 TIME 16:28:05.323115
> TRACEID = T670, TRACESET = NULL, IODATA = 128
> USER = SYSTEM, I/O OLD PSW = 0704D001 80000000 00000000 002C0FBE
> DEVICE = 0670, SCSW = 01C04017 00AA4088 0E000001 ** I/O ERROR **
> ESW = 00800000
> I/O PRIORITIES: CHANNEL = 0, CURRENT = 255, ORIGINAL = 255
> OUT-PRIORITIZED COUNT = 0
> -> CCW(1) = 07600001 00000000, CCW ADDRESS = 00AA4080
> ********** INVALID DATA ADDRESS **********
> -> CCW(2) = 0F200001 00000000, CCW ADDRESS = 00AA4088
> ********** INVALID DATA ADDRESS **********
> The vendor says that the device has been modified to return Device End
for both
> the Rewind (CCW Op Code 07) and the Rewind and Unload (0F) commands. In
the
> penultimate entry, you can see that there is a Rewind chained to a
Rewind and
> Unload. Both of those CCWs are flagged in the trace as having invalid
data
> addresses. This is puzzling because the I/O was generated by CP. If
there are,
> indeed, invalid data addresses, sould that not cause a Channel Program
Check or
> some other error rather than Command Reject? Why is CP using invalid
channel
> programs in the first place?
Those are control operations. The address and length fields are not
[supposed to be] used by the device. The channel does not generate UNIT
CHECKs and does not manufacture SENSE data. The 'invalid data address' is
the result of a determination by CP, not the channel subsystem, at the
time the record was cut. I've seen this when CP puts bad addresses in
CCWs specifically to catch unexpected data transfer attempts.
> Here is the pertinent section from the OPERATOR console:
>
> 15:47:21 TAPE 0670 DETACHED RSCHUH 0670 BY
RSCHUH
> 15:47:21 HCPERP500I TAPE 0670 AN OPERATION WAS TERMINATED BECAUSE
A
> 15:47:21 HCPERP500I COMMAND REJECT ERROR
OCCURRED
> 15:47:21 HCPERP6300I SENSE DATA FORMAT = 02 MSG CODE =
00
> 15:47:21 HCPERP6301I CHANNEL COMMAND WORD COMMAND CODE =
07
> 15:47:21 HCPERP6303I SENSE = 80482427 00000020 00000000 00000000
00000000
> 15:47:21 HCPERP6303I 00000000 0F010000
00000000
> 15:47:21 HCPERP6304I IRB = 01C04017 00AA4088 0E000001
00800000
> 15:47:21 HCPERP6305I USERID =
SYSTEM
> 15:47:21 HCPERP2216I CHANNEL PATH ID =
52
> 15:47:21 HCPERP2220I PHYSICAL CHANNEL PATH ID =
0168
>
> According to the vendor, their internal trace shows that their machine
received
> the Rewind and responded with Device End. The next thing they saw was
the Sense.
The first 4 bytes of SENSE data (80482427) show that
- X'80': The device is rejecting the REWIND
- X'48': Drive is online and is at beginning-of-tape.
- X'24': Path 2 to the device is raising the error
: The autoloader has cartridges in it
- X'27': Error recovery action = (what you see in the HCPERP messages)
Byte 7 says it is format X'20' sense data, so starting at byte 24:
- X'0F': ESCON 4.5 MB/s channel
- X'01': Autoloader is installed
With no knowledge of the device, I'd say they rejected the REWIND because
it was already at BOT, but regardless, the evidence points to the device,
not to the channel or CP.
Alan Altmark
z/VM Development
IBM Endicott
