All drivers must implement this hook, otherwise ATA commands would go
nowhere (and a lot of other oopsen would appear as well).
Signed-off-by: NJeff Garzik <jeff@garzik.org>

Separate sata_vsc interrupt handling into a normal (per-port) path and an
error path with the addition of vsc_port_intr and vsc_error_intr
respectively.  The error path handles interrupt based
hotplug events which requires the definition of vsc_freeze and vsc_thaw.

Note: vsc_port_intr has a workaround for unexpected interrupts that occur
during polled commands.  This fixes a regression between 2.6.19 and 2.6.20.

Changes in take2:
* removed definition of invalid fis bit
* let standard ata-error-handling handle the serror register
* clear all unhandled interrupts
* revert changes to vsc_intr_mask_update (vsc_thaw enables all interrupts)
* use unlikely() for the pci-abort and not-our-interrupt cases in vsc_sata_interrupt

Changes in take3:
* Unify the "add" + "hook-up" patches into this single patch

[htejun@gmail.com: clean up comments and suggestions]
Cc: Jeremy Higdon <jeremy@sgi.com>
Signed-off-by: NDan Williams <dan.j.williams@intel.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

sata_sil used to trigger HSM error if IRQ occurs during polling
command.  This didn't matter because polling wasn't used in sata_sil.
However, as of 2.6.20, all IDENTIFYs are performed by polling and
device detection sometimes fails due to spurious IRQ.  This patch
makes sata_sil ignore and clear spurious IRQ while executing commands
by polling.
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

intx should be turned on when pci_enable_msi() fails not when it
succeeds.  Fix it.
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

80c test mask is at bits 18 and 19 of EIDE Controller Configuration
not 22 and 23.  Fix it.
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

Duplicate ids noticed by Kay Sievers <kay.sievers@suse.de>

Although 100% different, this is based on a patch by Greg Kroah-Hartman
<gregkh@suse.de>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

Recently Tejun wrote a patch to ahci.c to make it raise a HSM violation
if the drive attempted to complete a tag that wasn't outstanding. We could
run into the same problem with sata_nv ADMA. This adds code to raise a HSM
violation error if the controller gives us a notifier tag that isn't
outstanding, since the drive may be issuing spurious completions.
Signed-off-by: NRobert Hancock <hancockr@shaw.ca>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

sata_nv implemented its own copies of the BMDMA helper functions for ADMA,
since the ADMA BMDMA status registers are PIO while the other registers
are MMIO, and this was the only way to handle this previously. Now that
we have iomap support, the standard routines should just work, so use them.
The only thing we need to override as far as ADMA and BMDMA is the
post_internal_cmd callback, where we should only call ata_post_internal_cmd
if we are in port-register mode.
Signed-off-by: NRobert Hancock <hancockr@shaw.ca>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

Merge order left libata-acpi and pata_scc with remainling usage of
ap->id.  Kill superflous id printing and substitute the remaining ones
with ap->print_id.
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

SDB FIS containing spurious NCQ completions is a clear protocol
violation.  Currently, only some Maxtors with early firmware revisions
are showing this problem.  Those firmwares have other NCQ related
problems including buggy NCQ error reporting and occasional lock up
after NCQ errors.

Consider spurious NCQ completions HSM violation and freeze the port
after it.  EH will turn off NCQ after this happens several times.
Eventually drives which show this behavior should be blacklisted for
NCQ.
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

Remove the DMA setup function. As pointed out by Sergey we set the actual
DMA clock timing in set_dmamode so we don't actually need to do anything
with it at set up time, but just leave the PIO timings loaded.
Signed-off-by: NAlan Cox <alan@redhat.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

Correct iordy handling and DMA bit flag handling.
Signed-off-by: NAlan Cox <alan@redhat.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

Keeps the behaviour consistent and easier to understand.
Signed-off-by: NAlan Cox <alan@redhat.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

Signed-off-by: NAlan Cox <alan@redhat.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

The CS5520 isn't just an ATA controller and we must not
pci_disable_device it as it turns into pci_disable_computer.
Signed-off-by: NAlan Cox <alan@redhat.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

We can't specify which mode in the cases below but we can at least say
PIO and look consistent with the default.
Signed-off-by: NAlan Cox <alan@redhat.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

ata_port has two different id fields - id and port_no.  id is
system-wide 1-based unique id for the port while port_no is 0-based
host-wide port number.  The former is primarily used to identify the
ATA port to the user in printk messages while the latter is used in
various places in libata core and LLDs to index the port inside the
host.

The two fields feel quite similar and sometimes ap->id is used in
place of ap->port_no, which is very difficult to spot.  This patch
renames ap->id to ap->print_id to reduce the possibility of such bugs.

Some printk messages are adjusted such that id string (ata%u[.%u])
isn't printed twice and/or to use ata_*_printk() instead of hardcoded
id format.
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

We already have code that handles hotplug interrupt indications in ADMA
mode, this turns on the control flag that actually enables these interrupts.
Also fixes some cases in the same functions where a 16-bit register was read
using a readl instead of a readw.
Signed-off-by: NRobert Hancock <hancockr@shaw.ca>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

The hardware provides us a notifier register that indicates what command
tags have completed. Use this to determine which CPBs to check, rather
than blindly checking all active CPBs. This should provide a minor
performance win, since if the controller has touched some of these
incomplete CPBs, accessing them will likely result in a cache miss.
Signed-off-by: NRobert Hancock <hancockr@shaw.ca>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

This edits the taskfile setup to more closely match the way that libata
sends the taskfile for other controllers. This avoids putting taskfile writes
into the CPB buffer that are not needed according to the taskfile flags.
Signed-off-by: NRobert Hancock <hancockr@shaw.ca>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

Clean up the initialization of the CPB and APRD structures so that we
strictly follow the rules for ordering of writes to the CPB flags and
response flags, and prevent duplicate initialization.
Signed-off-by: NRobert Hancock <hancockr@shaw.ca>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

When error handling occurs with pending commands, output the contents
of the next CPB count and next CPB index registers as well as the others,
since these may be useful for debugging.
Signed-off-by: NRobert Hancock <hancockr@shaw.ca>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

libata: Remove duplicate dma blacklist entry

The exact same entry is already present.
Signed-off-by: NMagnus Damm <damm@opensource.se>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

pata_pcmcia: Update device table

Add CFA devices from I-O Data, Mitsubishi and Viking. Add SanDisk comment.
Signed-off-by: NMagnus Damm <damm@opensource.se>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

Conversion to resource-managed iomap was buggy causing init failures
on both vt6420 and 6421 - BAR5 wasn't mapped for both controllers
while on vt6420 sata_via tried to map BAR0-4 twice.  Fix it.
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

Fix ata_scsi_change_queue_depth() such that...

* NCQ on/off is exactly determined using the same logic as the issue path.

* queue depth is adjusted to 1 if NCQ is not enabled.

* -EINVAL is returned if requested action is ignored due to limitations.

This fixes the bug which allows queue depth to be increased on
blacklisted NCQ hosts/devices.
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

Fix ata_scmd_need_defer() such that...

* whether NCQ is used or not is exactly determined using the same
  criteria as the issue path.

* defer-check is performed in all cases.

This fixes race condition where turning off NCQ on the fly causes
non-NCQ commands sneak into NCQ phase.
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

Separate out ata_ncq_enabled().
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

libata used disable pdev only on PM_EVENT_SUSPEND while re-enable pdev
unconditionally.  This was okay before ref-counted pdev enable update
but it now makes the pdev pinned after swsusp cycle (enabled twice but
disabled only once) and devres sanity check whines about it.

Fix it by unconditionally disabling pdev on all suspend events.
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

This patch appears to solve some problems with commands timing out in
cases where an NCQ command is immediately followed by a non-NCQ command
(or possibly vice versa). This is a rather ugly solution, but until we
know more about why this is needed, this is about all we can do.
Signed-off-by: NRobert Hancock <hancockr@shaw.ca>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

The driver requires in_be32(), and so should not be built on many PCI
platforms.
Signed-off-by: NJeff Garzik <jeff@garzik.org>

ata_probe_ent_alloc() had a temporary hack such that devm_kzalloc()
was used for allocation if devres had been previously initialized on
the device; otherwise, plain kzalloc() was used.  This was to make the
code useable from both the old and devres-aware libata drivers during
transition.  This hack made ata_sas_port_alloc() unable to determine
how the probe_ent is allocated, causing double free in some cases.

Remove the now-unneeded hack and make ata_sas_port_alloc() use
devm_kfree().
Signed-off-by: NTejun Heo <htejun@gmail.com>
Cc: James Bottomley <James.Bottomley@SteelEye.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

Fix sparse warnings in SATA:
drivers/ata/sata_sil.c:342:9: warning: Using plain integer as NULL pointer
drivers/ata/sata_mv.c:2056:55: warning: Using plain integer as NULL pointer
Signed-off-by: NRandy Dunlap <randy.dunlap@oracle.com>
Cc: Jeff Garzik <jeff@garzik.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

Some debug output in the ADMA error_handler function was removed recently,
but it may be useful in certain cases, like NCQ commands timing out. Add it
back in, but make it a bit more intelligent so that it only prints if
command(s) are active and only prints the CPBs for those commands.
That way it won't spew at inappropriate times like suspend/resume.
Signed-off-by: NRobert Hancock <hancockr@shaw.ca>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

Somehow the sis_info133 external definition ended up in libata.h and that
was included by both drivers.  However libata.h contains libata-* specific
internals and clashing defines like DRV_NAME so this makes a mess.  Move
the extern into the C file and remove the warnings

[akpm@linux-foundation.org: create sis.h to avoid extern-decl-in-C]
Signed-off-by: NAlan Cox <alan@redhat.com>
Cc: Jeff Garzik <jeff@garzik.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

I was trying to use HDIO_DRIVE_TASK for something today,
and discovered that the libata implementation does not copy
over the upper four LBA bits from args[6].

This is serious, as any tools using this ioctl would have their
commands applied to the wrong sectors on the drive, possibly resulting
in disk corruption.

Ideally, newer apps should use SG_IO/ATA_16 directly,
avoiding this bug.  But with libata poised to displace drivers/ide,
better compatibility here is a must.

This patch fixes libata to use the upper four LBA bits passed
in from the ioctl.

The original drivers/ide implementation copies over all bits
except for the master/slave select bit.  With this patch,
libata will copy only the four high-order LBA bits,
just in case there are assumptions elsewhere in libata (?).
Signed-Off-By: NMark Lord <mlord@pobox.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

ATA_DNXFER_ANY isn't used anymore.  Kill it.
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

The current EH speed down code is more of a proof that the EH
framework is capable of adjusting transfer speed in response to error.
This patch puts some intelligence into EH speed down sequence.  The
rules are..

* If there have been more than three timeout, HSM violation or
  unclassified DEV errors for known supported commands during last 10
  mins, NCQ is turned off.

* If there have been more than three timeout or HSM violation for known
  supported command, transfer mode is slowed down.  If DMA is active,
  it is first slowered by one grade (e.g. UDMA133->100).  If that
  doesn't help, it's slowered to 40c limit (UDMA33).  If PIO is
  active, it's slowered by one grade first.  If that doesn't help,
  PIO0 is forced.  Note that this rule does not change transfer mode.
  DMA is never degraded into PIO by this rule.

* If there have been more than ten ATA bus, timeout, HSM violation or
  unclassified device errors for known supported commands && speeding
  down DMA mode didn't help, the device is forced into PIO mode.  Note
  that this rule is considered only for PATA devices and is pretty
  difficult to trigger.

One error can only trigger one rule at a time.  After a rule is
triggered, error history is cleared such that the next speed down
happens only after some number of errors are accumulated.  This makes
sense because now speed down is done in bigger stride.
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

* Move forcing device to PIO0 on device disable into
  ata_dev_disable().  This makes both old and new EHs act the same
  way.

* Speed down only PIO mode on probe failure.  All commands used during
  probing are PIO commands.  There's no point in speeding down DMA.

* Retry at least once after -ENODEV.  Some devices report garbled
  IDENTIFY data after certain events.  This shouldn't cause device
  detach and re-attach.

* Rearrange EH failure path for simplicity.
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

Make ata_down_xfermask_limit() accept @sel instead of @force_pio0.
@sel selects how the xfermask limit will be adjusted.  The following
selectors are defined.

* ATA_DNXFER_PIO	: only speed down PIO
* ATA_DNXFER_DMA	: only speed down DMA, don't cause transfer mode change
* ATA_DNXFER_40C	: apply 40c cable limit
* ATA_DNXFER_FORCE_PIO	: force PIO
* ATA_DNXFER_FORCE_PIO0	: force PIO0 (same as original with @force_pio0 == 1)
* ATA_DNXFER_ANY	: same as original with @force_pio0 == 0

Currently, only ANY and FORCE_PIO0 are used to maintain the original
behavior.  Other selectors will be used later to improve EH speed down
sequence.
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>