[SCSI] improved eh timeout handler
authorHannes Reinecke <hare@suse.de>
Mon, 11 Nov 2013 12:44:54 +0000 (13:44 +0100)
committerJames Bottomley <JBottomley@Parallels.com>
Thu, 19 Dec 2013 15:39:02 +0000 (07:39 -0800)
commite494f6a728394ab0df194342549ee20e6f0752df
tree2cbae9eea944540b2777e40f1787a90b78123334
parent2451079bc2ae1334058be8babd44be03ecfa7041
[SCSI] improved eh timeout handler

When a command runs into a timeout we need to send an 'ABORT TASK'
TMF. This is typically done by the 'eh_abort_handler' LLDD callback.

Conceptually, however, this function is a normal SCSI command, so
there is no need to enter the error handler.

This patch implements a new scsi_abort_command() function which
invokes an asynchronous function scsi_eh_abort_handler() to
abort the commands via the usual 'eh_abort_handler'.

If abort succeeds the command is either retried or terminated,
depending on the number of allowed retries. However, 'eh_eflags'
records the abort, so if the retry would fail again the
command is pushed onto the error handler without trying to
abort it (again); it'll be cleared up from SCSI EH.

[hare: smatch detected stray switch fixed]
Signed-off-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
drivers/scsi/hosts.c
drivers/scsi/scsi.c
drivers/scsi/scsi_error.c
drivers/scsi/scsi_priv.h
include/scsi/scsi_cmnd.h
include/scsi/scsi_host.h