Currently, the scsi error handler will issue a START_UNIT
command if the drive indicates it needs its motor started
and the allow_restart flag is set in the scsi_device. If,
after the scsi error handler invokes a host adapter reset
due to error recovery, a device is in a unit attention
state AND also needs a START_UNIT, that device will be placed
offline. The disk array devices on an ipr RAID adapter
will do exactly this when in a dual initiator configuration.
This patch adds a single retry to the EH initiated
START_UNIT.
Signed-off-by: NBrian King <brking@linux.vnet.ibm.com>

Patch modified and
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

It looks like megaraid_sas at least needs this to throttle its commands
as they begin to time out.  The code keeps the existing transport
template use of eh_timed_out (and allows the transport to override the
host if they both have this callback).
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

If an EH command times out today, the LLDD's abort handler
will be called to abort the command. It is assumed that this
completes successfully, which can result in the command getting
completed later resulting in an oops. Improve the current
implementation by escalating all the way to host reset if
necessary in order to clean up the EH command.
Signed-off-by: NBrian King <brking@linux.vnet.ibm.com>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

1) If the device reports an uncorrectable MEDIUM ERROR, such
as SK MEDIUM ERROR, ASC UNRECOVERED READ ERR, AMNF DATA
FIELD or RECORD NOT FOUND, then: In scsi_check_sense()
return SUCCESS so as to not retry -- the error is
uncorrectable -- this speeds up total processing time.
Signed-off-by: NLuben Tuikov <ltuikov@yahoo.com>

Extracted the MEDIUM ERROR piece and
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

Export a couple of functions from scsi_error that are needed to handle
failed SCSI commands from the SAS EH.
Signed-off-by: NDarrick J. Wong <djwong@us.ibm.com>

make exports GPL and
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

scsi_send_eh_cmnd is the last user of non-sg commands currently.
This patch switches it to a one-element SG list.  Also updates the
kerneldoc comment for scsi_send_eh_cmnd to reflect reality while we're
at it.

Test on my mptsas card, but this should get testing with as many
drivers as possible.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

Fixup for lockdep enabled kernels: Annotate an on-stack completion.
Signed-off-by: NStefan Richter <stefanr@s5r6.in-berlin.de>
Acked-by: NIngo Molnar <mingo@elte.hu>
Cc: James Bottomley <James.Bottomley@steeleye.com>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

The callers of scsi_send_eh_cmnd are setting the cmnd buffer,
and then scsi_send_eh_cmnd is copying that updated buffer to
the old_cmnd variable. Then after the command runs, we end up
copying that old_cmnd var which has the new cmnd to the scsi
command buffer. When this command gets recent, all types of fun
things happen like getting TUR or START_STOP commands with
data and scatterlists.

This patch made against scsi-rc-fixes, has the callers of
scsi_send_eh_cmnd pass in the command so scsi_send_eh_cmnd
can do the right thing. This should go into 2.6.18 since this
fixes a regression added when we removed some of the scsi_cmnd
fields and replaced them with local variables.
Signed-off-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

Currently struct scsi_cmnd has various fields that are used to backup
original data after the corresponding fields have been overridden for
EH commands.  This means drivers can easily get at it and misuse it.
Due to the old_ naming this doesn't happen for most of them, but two
that have different names have been used wrong a lot (see previous
patch).  Another downside is that they unessecarily bloat the scsi_cmnd
size.

This patch moves them onstack in scsi_send_eh_cmnd to fix those two
issues aswell as allowing future EH fixes like moving the EH command
submissions to use SG lists like everything else.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

The scsi midlayer portion of the patch
Signed-off-by: NJames Smart <James.Smart@emulex.com>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

The RQ_SCSI_* flags are a vestiage of a long past history.  The EH code
still sets them but we never make use of that information.  The other
users is pluto.c which never had a chance to work but needs to be kept
compiling to keep Davem happy, so copy over the definition there.

We could probably get rid of RQ_ACTIVE/RQ_INACTIVE aswell with some
work, there's only two more or less bogus looking uses in ubd and scsi.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

With Achim patch the last user (gdth) is switched away from scsi_request
so we an kill it now.  Also disables some code in i2o_scsi that was
broken since the sg driver stopped using scsi_requests.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

libata implemented a feature to schedule EH without an associated EH
by manipulating shost->host_eh_scheduled in ata_scsi_schedule_eh()
directly.  Move this function to scsi_error.c and rename it to
scsi_schedule_eh().  It is now an exported API for SCSI transports and
exported via new header file drivers/scsi/scsi_transport_api.h

This patch also de-export scsi_eh_wakeup() which was exported
specifically for ata_scsi_schedule_eh().
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

libata needs to invoke EH without scmd.  This patch adds
shost->host_eh_scheduled to implement such behavior.

Currently the only user of this feature is libata and no general
interface is defined.  This patch simply adds handling for
host_eh_scheduled where needed and exports scsi_eh_wakeup() to
modules.  The rest is upto libata.  This is the result of the
following discussion.

http://thread.gmane.org/gmane.linux.scsi/23853/focus=9760

In short, SCSI host is not supposed to know about exceptions unrelated
to specific device or command.  Such exceptions should be handled by
transport layer proper.  However, the distinction is not essential to
ATA and libata is planning to depart from SCSI, so, for the time
being, libata will be using SCSI EH to handle such exceptions.
Signed-off-by: NTejun Heo <htejun@gmail.com>

Overriding the whole EH code is a per-transport, not per-host thing.
Move ->eh_strategy_handler to the transport class, same as
->eh_timed_out.

Downside is that scsi_host_alloc can't check for the total lack of EH
anymore, but the transition period from old EH where we needed it is
long gone already.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

This moves the eh_timed_out functionality from the scsi_host_template
to the transport_template. Given that this is now a transport function,
the EH_RESET_TIMER case no longer caps the timer reschedulings. The
transport guarantees that this is not an infinite condition.
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

Fix up an off by one error in calculating retries for scsi
commands. This bug was discovered when an SG_IO request
was sent to scsi core with retries = 0, causing the overall
timeout check to go off in scsi_softirq_done.
Signed-off-by: NBrian King <brking@us.ibm.com>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

Export two SCSI EH command handling functions.  To be used by libata EH.
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jgarzik@pobox.com>

When the scsi_execute_async interface was added it ended up reducing
the flexibility of userspace to send arbitrary scsi commands through
sg using SG_IO. The SG_IO interface allows userspace to specify the
CDB length. This is now ignored in scsi_execute_async and it is
guessed using the COMMAND_SIZE macro, which is not always correct,
particularly for vendor specific commands. This patch adds a cmd_len
parameter to the scsi_execute_async interface to allow the caller
to specify the length of the CDB.
Signed-off-by: NBrian King <brking@us.ibm.com>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

Add scsi helpers to create really-large-requests and convert
scsi-ml to scsi_execute_async().

Per Jens's previous comments, I placed this function in scsi_lib.c.
I made it follow all the queue's limits - I think I did at least :), so
I removed the warning on the function header.

I think the scsi_execute_* functions should eventually take a request_queue
and be placed some place where the dm-multipath hw_handler can use them
if that failover code is going to stay in the kernel. That conversion
patch will be sent in another mail though.
Signed-off-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

The eh_action semaphore in scsi_eh_send_command is cleared after a
command timeout.  The command is subsequently aborted and the abort
will try to call scsi_done() on it.  Unfortunately, the scsi_eh_done()
routine unconditinally completes the semaphore (which is now null).
Fix this race by makiong the scsi_eh_done() routine check that the
semaphore is non null before completing it (mirroring the ordinary
command done/timeout logic).
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

scsi_send_eh_cmnd currently uses a semaphore and an overload of eh_timer
to either get a completion for a command for a timeout.
Switch to using a completion and wait_for_completion_timeout to simply
the code and not having to deal with the races ourselves.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

now that the abuse in qla2xxx is gone this field can be remove.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

adjust comments, remove a useless cast and remove a write-only variable.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

Wrap a highly common idiom.  Makes the code easier to read, helps pave
the way for sdev->{id,channel} removal, and adds a token that can easily
by grepped-for in the future.

There are a couple sdev_id() and scmd_printk() updates thrown in as well.

Rejections fixed up and
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

rejections fixed and
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

This should eliminate (at least in the mid layer) to make numeric
assumptions about any of the enumeration variables.  As a side effect,
it will also make all the messages consistent and line us up nicely for
the error logging strategy (if it ever shows itself again).
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

Found in the -rt patch set.  The scsi_error thread likely will be in the
TASK_INTERRUPTIBLE state upon exit.  This patch fixes this bug.
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

Found in the -rt patch set.  The scsi_error thread likely will be in the
TASK_INTERRUPTIBLE state upon exit.  This patch fixes this bug.
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

From: 	Alan Stern <stern@rowland.harvard.edu>

This patch (as561) fixes the error handler's thread-exit code.  The
kthread_stop call won't wake the thread from a down_interruptible, so
the patch gets rid of the semaphore and simply does

        set_current_state(TASK_INTERRUPTIBLE);
Signed-off-by: NAlan Stern <stern@rowland.harvard.edu>

Modified to simplify the termination loop and correct the sleep condition.
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

We fix the oops by enforcing the host state model.  There have also
been two extra states added: SHOST_CANCEL_RECOVERY and
SHOST_DEL_RECOVERY so we can take the model through host removal while
the recovery thread is active.
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

The problem lies in the way the error handler uses TEST UNIT READY to
tell whether error recovery has succeeded.  The scsi_eh_tur function
gives up after one round of retrying; after that it decides that more
error recovery is needed.

However TUR is liable to report sense data indicating a retry is needed
when in fact error recovery has succeeded.  A typical example might be
SK=2, ASC=4, ASCQ=1 (Logical unit in process of becoming ready).  The mere
fact that we were able to get a sensible reply to the TUR should indicate
that the device is working well enough to stop error recovery.

I ran across a case back in January where this happened.  A CD-ROM drive
timed out the INQUIRY command, and a device reset fixed the blockage.
But then the drive kept responding with 2/4/1 -- because it was spinning
up I suppose -- until the error handler gave up and placed it offline.
If the initial INQUIRY had received the 2/4/1 instead, everything would
have worked okay.  It doesn't seem reasonable for things to fail just
because the error handler had started running.
Signed-off-by: NAlan Stern <stern@rowland.harvard.edu>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

This one's slightly more difficult.  The transport class uses
REQ_FAILFAST, so another interface (scsi_execute) had to be invented to
take the extra flag.  Also, the sense functions are shifted around to
allow spi_execute to place data directly into a struct scsi_sense_hdr.
With this change, there's probably a lot of unnecessary sense buffer
allocation going on which we can fix later.
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

Migrate the current SCSI host state model to a model like SCSI
device is using.
Signed-off-by: NMike Anderson <andmike@us.ibm.com>

Rejections fixed up and
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

Just opencoded access to eh_eflags, it's much more readable anyway.
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

We never look at it except for the old megaraid driver that abuses it
for sending internal commands.  That usage can be fixed easily because
those internal commands are single-threaded by a mutex and we can easily
use a completion there.
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>