kern/103602: drive gets wedged on READ CD CAPACITY if no disc is in

[Scott Long]

Thomas Quinot wrote:
> SCSI guys,
> 
> kern/103602 is a long-standing issue with a Sony DVD drive that has a
> strange behaviour (failing to attach and causing an interrupt storm).
> 
> One aspect of this PR was that an ATAPI command timing out was not
> correctly reported by the ATAPI/CAM SIM to the XPT layer. This has been
> fixed in atapi-cam.c rev. 1.49.
> 
> Now we have a second issue: if there is no disc in the drive, it
> times out a READ CD CAPACITY (and any subsequent retry, even if a disc
> is then inserted in the drive), instead of returning an appropriate
> error (the expected behaviour would be a NOT READY sense key and a
> MEDIUM NOT PRESENT ASC). A TEST UNIT READY command seems to behave
> properly (reporting the appropriate error if no disc is present,
> reporting no error and resetting the drive to a proper state if
> one is).
> 
> The timeout causes the unit to fail to attach at boot time if no disc is
> present. How should we deal with this situation? One possible option is
> to make sure that we first issue a TEST UNIT READY at attach time prior
> to issuing READ CD CAPACITY, and not issue the second command if the TUR
> returned a NOT READY status.
> 
> The traces below show how the drive behaves in response to TEST UNIT
> READY and READ CD CAPACITY in various cases. Complete boot messages are
> available in the PR history.
> 
> Thomas.
> 

I'm confused.  CAM actually sends a READ_CAPACITY command to the drive, 
not a READ_CD_CAPACITY command.  Are you saying that issuing both a
READ_CAPACITY and a READ_CD_CAPACITY to an empty Sony drive results in
the same bad behavior?

Re-arranging the probe process to issue a TUR first is tricky.  I'd 
rather not do it if possible.

Scott