I'm starting to work on a kernel driver that will submit SCSI requests
directly to a device queue -- rather like the sg driver except that mine
will operate as a bottom half, not in process context. There won't be any
predetermined timeouts for the commands, but sometimes my driver will need
to abort a command and perform a device reset.
I see that scsi_reset_provider() is available to implement device reset
requests (although it needs to run in a process, no big deal). But it
won't abort a command that's already in progress.
So I need to know how to abort a queued request. That actually involves
two problems: how to abort a request that hasn't yet made it to the head
of the queue and begun processing, and how to abort a command that has
already gone to the low-level device driver. The first problem probably
doesn't matter much; I'm assuming that nobody else will be using the
device at the same time as my driver. But it would be nice to know how to
cancel _all_ the outstanding requests on the device queue.
The second problem is the real one I have to face. Is there some way to
abort an ongoing command, perhaps by following pointers to the scsi_cmnd
structure and invoking the timeout routine directly?
Alan Stern
-
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
|
