Add qc->result_tf and ATA_QCFLAG_RESULT_TF.  This moves the
responsibility of loading result TF from post-compltion path to qc
execution path.  qc->result_tf is loaded if explicitly requested or
the qc failsa.  This allows more efficient completion implementation
and correct handling of result TF for controllers which don't have
global TF representation such as sil3124/32.
Signed-off-by: NTejun Heo <htejun@gmail.com>

Don't directly access &qc->tf when tf == &qc->tf.
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>

Convert the ATAPI_ENABLE_DMADIR compile time option needed
by some SATA-PATA bridge to runtime module parameter.
Signed-off-by: NAlbert Lee <albertcc@tw.ibm.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

A lot of EH codes are about to be added to libata.  Separate out
libata-eh.c.  ata_scsi_timed_out(), ata_scsi_error(),
ata_qc_timeout(), ata_eng_timeout(), ata_eh_qc_complete() and
ata_eh_qc_retry() are moved.  No code is changed by this patch.
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

qcs might get retried because of unrelated failure.  e.g. NCQ command
failure causes the whole command set to be aborted.  Decrement
scmd->retries for such retrials to avoid unnecessarily failing
commands.  Note that scmd->retries will be incremented the first time.
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

TF register might not be directly accessible depending on errors.
e.g. TF of failed NCQ command is in log page 10h.  Make reading TF
responsibility of error handlers.  For the current EH, simply push TF
reading into qc completion functions as they are practically part of
EH.  New EH will fill qc->tf with status registers before complting
qcs.
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

Current sense generation code does not generate sense error if status
register value doesn't indicate error condition.  However, LLDD's may
indicate errors which 't show up in status register.  Completing such
qc's without generating sense results in successful completion of
failed commands.

Invoke ata_to_sense_error() regardless of status register if
qc->err_mask is not zero such that ata_to_sense_error() generates
default sense error.
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

Add a new qc flag ATA_QCFLAG_IO.  This flag gets set for normal IO
commands originating from SCSI midlayer.  This information will be
used by EH to determine transfer speed reconfiguration.
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

Rename ATA_FLAG_PORT_DISABLED to ATA_FLAG_DISABLED for consistency.
(ATA_FLAG_* are always about ports).
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

Kill unused flag ATA_FLAG_IN_EH.
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

This patch renames ata_dev_present() to ata_dev_enabled() and adds
ata_dev_disabled().  This is to discern the state where a device is
present but disabled from not-present state.  This disctinction is
necessary when configuring transfer mode because device selection
timing must not be violated even if a device fails to configure.
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

There is no reason for the issuer to diddle with a failed qc as the
issuer has complete control over when a qc gets freed (usually in
->complete_fn).  Make ata_qc_issue() responsible for completing qcs
which failed to issue.
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

In preparation for SAS, kill some unnecessary code in ata_scsi_ioctl
to find the ATA port and device given the scsi_device. Neither local
is used in the function.
Signed-off-by: NBrian King <brking@us.ibm.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

Encapsulate part of ata_scsi_queuecmd so that it can be
reused by future SAS patches.
Signed-off-by: NBrian King <brking@us.ibm.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

Mostly, trim trailing whitespace.

Also:
* export ata_dev_pair
* move ata_dev_classify export closer to ata_dev_pair export

At the moment libata doesn't pass pm_message_t down ata_device_suspend.
This causes drives to be powered down when we just want a freeze,
causing unnecessary wear and tear. This patch gets pm_message_t passed
down so that it can be used to determine whether to power down the
drive.
Signed-off-by: NNigel Cunningham <nigel@suspend2.net>

 drivers/scsi/libata-core.c |    5 +++--
 drivers/scsi/libata-scsi.c |    4 ++--
 drivers/scsi/scsi_sysfs.c  |    2 +-
 include/linux/libata.h     |    4 ++--
 include/scsi/scsi_host.h   |    2 +-
 5 files changed, 9 insertions(+), 8 deletions(-)
Signed-off-by: NJeff Garzik <jeff@garzik.org>

Signed-off-by: NJeff Garzik <jeff@garzik.org>

Encapsulate some of ata_scsi_slave_config so that parts
can be reused in future SAS patches.
Signed-off-by: NBrian King <brking@us.ibm.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

Remove the ATA_PROT_PIO_MULT protocol.
Signed-off-by: NAlbert Lee <albertcc@tw.ibm.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

Original patch (and description) by Douglas Gilbert, with minor fixes
and API updates from me.

Changelog:
  - make existing libata VPD device identification page (0x83)
    supply the ATA serial number in the libata "vendor
    specific" designator (from Chris Paulson-Ellis)
  - add a "t10 vendor id based" designator as defined in
    SAT rev 08 (section 10.3.4.2.3) that supplies ATA
    model and serial numbers
  - make the libata VPD page 0x83 more extensible (for
    adding more designators in the future).
  - rename EVPD to VPD in various places. Enable Vital
    Product Data (EVPD) is a bit in the INQUIRY cdb.
Signed-off-by: NJeff Garzik <jeff@garzik.org>

SCSI midlayer has moved hostt->eh_timed_out to transport template.  As
libata doesn't need full-blown transport support yet, implement
minimal transport for libata.  No transport class or whatsoever, just
empty transport template with ->eh_timed_out hook.
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

Implement port_task.  LLDD's can schedule a function to be executed
with context after specified delay.  libata core takes care of
synchronization against EH.  This is generalized form of pio_task and
packet_task which are tied to PIO hsm implementation.
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

Until problems are sorted.

This patch renames ata_dev_id_[c_]string() to ata_id_[c_]string().
All other functions which read data from ATA ID data start with ata_id
and those two function names were getting too long.
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jgarzik@pobox.com>

If a low level driver wants to control max_sectors, it had to adjust
ap->host->max_sectors and set ATA_DFLAG_LOCK_SECTORS to tell
ata_scsi_slave_config not to override the limit.  This is not only
cumbersome but also incorrect for hosts which support more than one
devices per port.

This patch adds per-device ->max_sectors.  If the field is unset
(zero), libata core layer will adjust ->max_sectors according to
default rules.  If the field is set, libata honors the setting.
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jgarzik@pobox.com>

cdb_len is per-device property.  Sharing cdb_len on ap results in
inaccurate configuration on revalidation and hotplugging.  This patch
makes cdb_len per-device.
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jgarzik@pobox.com>

Use ata_dev_id_c_string()
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jgarzik@pobox.com>

In an effort to kill libata-specific assert() and use generic
WARN_ON(), this patch converts all assert(X)'s in libata core layer to
WARN_ON(!X)'s.  Most conversions are straight-forward logical negation
exception for the followings.

* In libata-core.c:ata_fill_sg(),
  assert(qc->n_elem > 0) is converted to WARN_ON(qc->n_elem == 0) because
  qc->n_elem is unsigned and unsigned <= 0 is weird.

* In libata-scsi.c:ata_gen_ata_desc/fixed_sense(),
  assert(NULL != qc->ap->ops->tf_read) is converted to
  WARN_ON(qc->ap->ops->tf_read == NULL), as there are no other users of
  'constant cond var' style in libata.
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jgarzik@pobox.com>

->eng_timeout cannot be invoked with NULL qc anymore.  Add an
assertion in ata_scsi_error() and kill NULL qc handling from all
->eng_timeout callbacks.
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jgarzik@pobox.com>

Implement ata_scsi_timed_out(), to be used as
scsi_host_template->eh_timed_out callback for all libata drivers.
Without this function, the following race exists.

If a qc completes after SCSI timer expires but before libata EH kicks
in, the qc gets completed but the scsicmd still gets passed to libata
EH resulting in ->eng_timeout invocation with NULL qc, which none is
handling properly.

This patch makes sure that scmd and qc share the same lifetime.
Original idea from Jeff Garzik <jgarzik@pobox.com>.
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jgarzik@pobox.com>

Add ATA_QCFLAG_EH_SCHEDULED.  If this flag is set, the qc is owned by
EH and normal completion path is not allowed to finish it.  This patch
doesn't actually use this flag.
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jgarzik@pobox.com>

ATA_FLAG_IN_EH flag is set on entry to EH and cleared on completion.
This patch just sets and clears the flag.  Following patches will
build normal qc execution / EH synchronization aroung this flag.
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jgarzik@pobox.com>

It looks like they are either discarding or corrupting data when the FUA
command is used, bad.
Signed-off-by: NJens Axboe <axboe@suse.de>

This is a merger of libata docs + cleanups from
Martin Waitz <tali@admingilde.org> and me.

From: Randy Dunlap <rdunlap@xenotime.net>
From: Martin Waitz <tali@admingilde.org>

Fix libata kernel-doc comments to match code.
Add some function parameters to kernel-doc.
Fix some typos/spellos.
Put comments in <= 80 columns.
Make one DPRINTK string unique.
Fix sparse cast warnings.
Signed-off-by: NRandy Dunlap <rdunlap@xenotime.net>
Signed-off-by: NJeff Garzik <jgarzik@pobox.com>

Implement ata_eh_qc_complete/retry() using scsi_eh_finish_cmd() and
scsi_eh_flush_done_q().  This removes all eh scsicmd finish hacks from
low level drivers.

This change was first suggested by Jeff Garzik.
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jgarzik@pobox.com>

Return AC_ERR_* mask from issue fuctions instead of 0/-1.  This
enables things like failing a qc with AC_ERR_HSM when the device
doesn't set DRDY when the qc is about to be issued.
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jgarzik@pobox.com>

When ata_qc_issue() fails, the qc might have been dma mapped or not.
So, performing only ata_qc_free() results in dma map leak.  This patch
makes ata_qc_issue() mark dma map flags correctly on failure and calls
ata_qc_complete() after ata_qc_issue() fails.
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jgarzik@pobox.com>