We allocate the MTT table with alloc_pages() and then do pci_map_sg(),
so we must call pci_dma_sync_sg() after the CPU writes to the MTT
table.  This works since the device will never write MTTs on mem-free
HCAs, once we get rid of the use of the WRITE_MTT firmware command.
This change is needed to make that work, and is an improvement for
now, since it gives FMRs a chance at working.

For MPTs, both the device and CPU might write there, so we must
allocate DMA coherent memory for these.
Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

MTTs are allocated in non-cache-coherent memory, so we must give
reserved MTTs their own cache line, to prevent both device and
CPU from writing into the same cache line at the same time.
Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

The reserved_mtts field has different meaning in Tavor and Arbel, so
we are wasting mtt entries on memfree. Fix the Arbel case to match
Tavor semantics.
Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

Add an RDMA/iWARP driver for the Chelsio T3 1GbE and 10GbE adapters.
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

Signed-off-by: Roland Dreier <rolandd@cisco.com>

Randomize the starting port number and avoid re-using port values
immediately after they are closed.  Instead keep track of the last
port value used and increment it every time a new port number is
assigned, to better replicate other port spaces.
Signed-off-by: Roland Dreier <rolandd@cisco.com>

Percpu data is not freed on module unloading.

Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Christoph Raisch <raisch@de.ibm.com>
Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Hoang-Nam Nguyen <hnguyen@de.ibm.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

For some reason gcc-3.4.5 on sparc64 does:

 WARNING: "____ilog2_NaN" [drivers/infiniband/hw/mthca/ib_mthca.ko] undefined!

Points to note:

 (1) The asm volatile flush/flushw are just markers for viewing what comes out
     in the assembly; removing them has no effect on the result.

 (2) Changing almost anything else in dwh__mthca_arbel_init_srq_context() or
     dwh__mthca_alloc_srq() causes the problem to go away.

The compiler command line issued by the kernel build is:

/opt/crosstool/gcc-3.4.5-glibc-2.3.6/sparc64-unknown-linux-gnu/bin/sparc64-unknown-linux-gnu-gcc -fno-strict-aliasing -fno-common -Os -m64 -mno-fpu -mcpu=ultrasparc -mcmodel=medlow -ffixed-g4 -ffixed-g5 -fcall-used-g7 -Wa,--undeclared-regs -pg -fno-omit-frame-pointer -fno-optimize-sibling-calls -fasynchronous-unwind-tables -g  -c -o drivers/infiniband/hw/mthca/.tmp_mthca_srq.o drivers/infiniband/hw/mthca/mthca_srq.c

This can be reduced to this whilst still retaining the problem:

/opt/crosstool/gcc-3.4.5-glibc-2.3.6/sparc64-unknown-linux-gnu/bin/sparc64-unknown-linux-gnu-gcc -m64 -c -o drivers/infiniband/hw/mthca/mthca_srq.o drivers/infiniband/hw/mthca/mthca_srq.c -Os

Removing -Os or changing it to -O or -O0 thru -O6 gets rid of the problem.

This patch to the kernel code fixes the problem:

Cc: "David S. Miller" <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

The following patch adds experimental support for IPoIB connected
mode, as defined by the draft from the IETF ipoib working group.  The
idea is to increase performance by increasing the MTU from the maximum
of 2K (theoretically 4K) supported by IPoIB on top of UD.  With this
code, I'm able to get 800MByte/sec or more with netperf without
options on a Mellanox 4x back-to-back DDR system.

Some notes on code:
1. SRQ is used for scalability to large cluster sizes
2. Only RC connections are used (UC does not support SRQ now)
3. Retry count is set to 0 since spec draft warns against retries
4. Each connection is used for data transfers in only 1 direction, so
   each connection is either active(TX) or passive (RX).  2 sides that
   want to communicate create 2 connections.
5. Each active (TX) connection has a separate CQ for send completions -
   this keeps the code simple without CQ resize and other tricks
6. To detect stale passive side connections (where the remote side is
   down), we keep an LRU list of passive connections (updated once per
   second per connection) and destroy a connection after it has been
   unused for several seconds. The LRU rule makes it possible to avoid
   scanning connections that have recently been active.
Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

Use ARRAY_SIZE() macro already defined in kernel.h instead of open
coding equivalent code.
Signed-off-by: Ahmed S. Darwish <darwish.07@gmail.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

When clearing the ib_ah_attr parameter in to_ib_ah_attr(), use sizeof
*ib_ah_attr instead of sizeof *path.

Pointed out by Jack Morgenstein <jackm@mellanox.co.il>.
Signed-off-by: Roland Dreier <rolandd@cisco.com>

The ATA_ENABLE_PATA define was never meant to be permanent, and in
recent kernels, it's already been unconditionally enabled.  Remove.
Signed-off-by: Jeff Garzik <jeff@garzik.org>

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

If we are doing a PIO setup for a CFA card and it blows up with a device
error then assume it is an older CFA card which doesn't support this
rather than failing the device out of existance.

Stands seperate to the quieting patch but that is obviously useful with
this change.
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

ata_pci_device_do_resume can fail if the PCI device couldn't be re-enabled.
 Update sata_nv to propagate the return value from this call and to not try
to do any other resume activities if it fails.  Fixes a compile warning.
Signed-off-by: Robert Hancock <hancockr@shaw.ca>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

Update sata_nv to wait for the controller to indicate via the status
register that it has entered the requested state when switching between
ADMA mode and register mode.  This issue came up recently when debugging
some problems with cache flush command timeouts and while it didn't appear
to fix that problem, this is something we should likely be doing in any
case.
Signed-off-by: Robert Hancock <hancockr@shaw.ca>
Cc: Tejun Heo <htejun@gmail.com>
Cc: Jeff Garzik <jeff@garzik.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

Some problems showed up recently with cache flush commands timing out on
sata_nv.  Previously these commands were always handled by transitioning to
legacy mode from ADMA mode first.  The timeout problem was worked around
already by a change to the interrupt handling code for legacy mode, but for
non-data commands like these it appears we can handle them in ADMA mode, so
the switch to legacy mode is not needed.

This patch changes the behavior so that we use ADMA mode to submit
interrupt-driven commands with ATA_PROT_NODATA protocol.  In addition to
avoiding the problem mentioned above entirely, this avoids the overhead of
switching to legacy mode and back to ADMA mode for handling cache flushes.
When handling non-DMA-mapped commands, we leave the APRD blank and clear
the NV_CPB_CTL_APRD_VALID field in the CPB so the controller does not
attempt to read it.
Signed-off-by: Robert Hancock <hancockr@shaw.ca>
Cc: Jeff Garzik <jeff@garzik.org>
Cc: Tejun Heo <htejun@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

This cleans up a few issues with the error handling in sata_nv in ADMA mode
to make it more consistent with other NCQ-capable drivers like ahci and
sata_sil24:

- When a command failed, we would effectively set AC_ERR_DEV on the
  queued command always.  In the case of NCQ commands this prevents libata
  from doing a log page query to determine the details of the failed
  command, since it thinks we've already analyzed.  Just set flags in the
  port ehi->err_mask, then freeze or abort and let libata figure out what
  went wrong.

- The code handled NV_ADMA_STAT_CPBERR as a "really bad error" which
  caused it to set error flags on every queued command.  I don't know
  exactly what this flag means (no docs, grr!) but from what I can guess
  from the standard ADMA spec, it just means that one or more of the CPBs
  had an error, so we just need to go through and do our normal checks in
  this case.

- In the error_handler function the code would always dump the state of
  all the CPBs.  This output seems redundant at this point since libata
  already dumps the state of all active commands on errors (and it also
  triggers at times when it shouldn't, like when suspending).  Take this
  out.

[akpm@osdl.org: many coding-style fixes]
Signed-off-by: Robert Hancock <hancockr@shaw.ca>
Cc: Jeff Garzik <jeff@garzik.org>
Cc: Tejun Heo <htejun@gmail.com>
Cc: Allen Martin <AMartin@nvidia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

MPIIX has only single channel IDE which can be configured for either primary or
secondary legacy I/O ports and IRQ.  So, get rid of the unneeded second probe
entry in mpiix_init_one() and of the invalid (but unused anyway) enable bits in
mpiix_pre_reset().

Warning: this cleanup has only been compile-tested...
Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

Fix clearing/setting the wrong TIME/IE/PPE bits for a slave drive caused by a
wrong shift count.
Fix the PIO mode 1 being overclocked by wrongly selecting the fast timing bank.
Also, fix/rephrase some comments while at it.
Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

Fix the PIO mode 2 using mode 0 timings -- this driver should enable the
fast timing bank starting with PIO2, just like the ata_piix driver does.
Also, fix/rephrase some comments while at it.
Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

IORDY and IORDY enable/disable flags.
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

People are getting confused about which drivers to enable for PATA PIIX
type devices. Change the ATA_PIIX line and help to make it clearer.
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

* Hardreset must not exit without actually performing reset regardless
  of link status.  We're resetting the link after all.

* Minor message update.

* 150ms delay is meaningful iff link is online after reset is
  complete.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

Follow the old SRST rule and delay 150ms between completion of
hardreset and status checking.  Debouncing delay should usually cover
this but debounce duration could be shorter than 150ms under certain
circumstances.

Usefulness depends on host controller implementation but it can't hurt
and serves as a reminder that 2s delay for GoVault should also be
added here.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

Per Jeff's suggestion, this patch rearranges the info printed for ATA
drives into dmesg to add the full ATA firmware revision and model
information, while keeping the output to 2 lines.
Signed-off-by: Eric D. Mudama <edmudama@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

Fix the wrong "compatible" PIO mode choices: MWDMA0 has 480 ns cycle while PIO1
only has 383 ns cycle, and MWDMA2 timings matchs those of PIO4 exactly.
Signed-off-by: Jeff Garzik <jeff@garzik.org>

This patch is against each libata driver.

Two IRQ calls are added in ata_port_operations.
- irq_on() is used to enable interrupts.
- irq_ack() is used to acknowledge a device interrupt.

In most drivers, ata_irq_on() and ata_irq_ack() are used for
irq_on and irq_ack respectively.

In some drivers (ex: ahci, sata_sil24) which cannot use them
as is, ata_dummy_irq_on() and ata_dummy_irq_ack() are used.
Signed-off-by: Kou Ishizaki <kou.ishizaki@toshiba.co.jp>
Signed-off-by: Akira Iguchi <akira2.iguchi@toshiba.co.jp>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

This patch is against the libata core and headers.

Two IRQ calls are added in ata_port_operations.
- irq_on() is used to enable interrupts.
- irq_ack() is used to acknowledge a device interrupt.

In most drivers, ata_irq_on() and ata_irq_ack() are used for
irq_on and irq_ack respectively.

In some drivers (ex: ahci, sata_sil24) which cannot use them
as is, ata_dummy_irq_on() and ata_dummy_irq_ack() are used.
Signed-off-by: Kou Ishizaki <kou.ishizaki@toshiba.co.jp>
Signed-off-by: Akira Iguchi <akira2.iguchi@toshiba.co.jp>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

In file included from drivers/infiniband/hw/ipath/ipath_diag.c:44:
include/linux/io.h:35: warning: 'struct device' declared inside parameter list
include/linux/io.h:35: warning: its scope is only this definition or declaration

Cc: Jeff Garzik <jeff@garzik.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

Convert libata core layer and LLDs to use iomap.

* managed iomap is used.  Pointer to pcim_iomap_table() is cached at
  host->iomap and used through out LLDs.  This basically replaces
  host->mmio_base.

* if possible, pcim_iomap_regions() is used

Most iomap operation conversions are taken from Jeff Garzik
<jgarzik@pobox.com>'s iomap branch.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

devres updates for pata_platform were dropped while merging devres
patches due to merge conflict.  This is the updated version.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

devres change moved iomap.o from obj-$(CONFIG_GENERIC_IOMAP) to lib-y
making it not linked if no in-kernel driver uses it.  Fix it.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

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

Implement pcim_iomap_regions().  This function takes mask of BARs to
request and iomap.  No BAR should have length of zero.  BARs are
iomapped using pcim_iomap_table().
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

Now that all LLDs are converted to use devres, default stop callbacks
are unused.  Remove them.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

Update libata LLDs to use devres.  Core layer is already converted to
support managed LLDs.  This patch simplifies initialization and fixes
many resource related bugs in init failure and detach path.  For
example, all converted drivers now handle ata_device_add() failure
gracefully without excessive resource rollback code.

As most resources are released automatically on driver detach, many
drivers don't need or can do with much simpler ->{port|host}_stop().
In general, stop callbacks are need iff port or host needs to be given
commands to shut it down.  Note that freezing is enough in many cases
and ports are automatically frozen before being detached.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

Update libata core layer to use devres.

* ata_device_add() acquires all resources in managed mode.

* ata_host is allocated as devres associated with ata_host_release.

* Port attached status is handled as devres associated with
  ata_host_attach_release().

* Initialization failure and host removal is handedl by releasing
  devres group.

* Except for ata_scsi_release() removal, LLD interface remains the
  same.  Some functions use hacky is_managed test to support both
  managed and unmanaged devices.  These will go away once all LLDs are
  updated to use devres.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

Implement ata_host_detach() which calls ata_port_detach() for each
port in the host and export it.  ata_port_detach() is now internal and
thus un-exported.  ata_host_detach() will be used as the 'deregister
from libata layer' function after devres conversion.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

Implement device resource management, in short, devres.  A device
driver can allocate arbirary size of devres data which is associated
with a release function.  On driver detach, release function is
invoked on the devres data, then, devres data is freed.

devreses are typed by associated release functions.  Some devreses are
better represented by single instance of the type while others need
multiple instances sharing the same release function.  Both usages are
supported.

devreses can be grouped using devres group such that a device driver
can easily release acquired resources halfway through initialization
or selectively release resources (e.g. resources for port 1 out of 4
ports).

This patch adds devres core including documentation and the following
managed interfaces.

* alloc/free	: devm_kzalloc(), devm_kzfree()
* IO region	: devm_request_region(), devm_release_region()
* IRQ		: devm_request_irq(), devm_free_irq()
* DMA		: dmam_alloc_coherent(), dmam_free_coherent(),
		  dmam_declare_coherent_memory(), dmam_pool_create(),
		  dmam_pool_destroy()
* PCI		: pcim_enable_device(), pcim_pin_device(), pci_is_managed()
* iomap		: devm_ioport_map(), devm_ioport_unmap(), devm_ioremap(),
		  devm_ioremap_nocache(), devm_iounmap(), pcim_iomap_table(),
		  pcim_iomap(), pcim_iounmap()
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>