This consolidates the reads from each of the if/else branches
to one place making the code a lot nicer to look at.
Tested-by: NChristian Lamparter <chunkeey@googlemail.com>
Signed-off-by: NMans Rullgard <mans@mansr.com>
Signed-off-by: NTejun Heo <tj@kernel.org>

Rename the register access macros and use standard _relaxed()
ops instead of __raw variants with explicit byte swapping.

The original driver used the ppc-specific in/out_le32().  When it
was adapted to other systems, these were added to the driver
under ifdefs.  However, those names are not defined as macros on
ppc, so it ended up replacing them there as well with altered
semantics.  This patch restores the original semantics on ppc and
makes the accesses no less strict on other systems.

Also fixes too many sparse warnings to count.
Tested-by: NChristian Lamparter <chunkeey@googlemail.com>
Signed-off-by: NMans Rullgard <mans@mansr.com>
Signed-off-by: NTejun Heo <tj@kernel.org>

Convert dmaengine_terminate_all() calls to synchronous and asynchronous
versions where appropriate.
Tested-by: NChristian Lamparter <chunkeey@googlemail.com>
Signed-off-by: NAndy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: NTejun Heo <tj@kernel.org>

The pointer to the mmio register base is missing the __iomem
annotation.  Fix this.
Tested-by: NChristian Lamparter <chunkeey@googlemail.com>
Signed-off-by: NMans Rullgard <mans@mansr.com>
Signed-off-by: NTejun Heo <tj@kernel.org>

The (void *__iomem) cast is wrong.  Change the target type of the
"base" pointer to void __iomem instead and drop the cast.
Tested-by: NChristian Lamparter <chunkeey@googlemail.com>
Signed-off-by: NMans Rullgard <mans@mansr.com>
Signed-off-by: NTejun Heo <tj@kernel.org>

Casting a pointer to unsigned long only to immediately cast it back
to a pointer makes no sense.  Fix this.
Tested-by: NChristian Lamparter <chunkeey@googlemail.com>
Signed-off-by: NMans Rullgard <mans@mansr.com>
Signed-off-by: NTejun Heo <tj@kernel.org>

The sata_dwc_qc_prep() does nothing.  Use the default ata_noop_qc_prep
instead.
Tested-by: NChristian Lamparter <chunkeey@googlemail.com>
Signed-off-by: NMans Rullgard <mans@mansr.com>
Signed-off-by: NTejun Heo <tj@kernel.org>

Here we refactor HOSTDEV{P}_FROM_*() macros to fit one line and fix the
definition of HSDEV_FROM_HSDEVP() where wrong name of the parameter waas used.
Tested-by: NChristian Lamparter <chunkeey@googlemail.com>
Signed-off-by: NAndy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: NTejun Heo <tj@kernel.org>

This moves all global data into the driver private struct, thus
permitting multiple devices of this type to be used.

The core_scr_read/write() functions are replaced with equivalent
calls to the existing sata_dwc_scr_read/write().
Tested-by: NChristian Lamparter <chunkeey@googlemail.com>
Signed-off-by: NMans Rullgard <mans@mansr.com>
Signed-off-by: NTejun Heo <tj@kernel.org>

This adds support for powering on an optional PHY when activating the
device.
Tested-by: NChristian Lamparter <chunkeey@googlemail.com>
Signed-off-by: NMans Rullgard <mans@mansr.com>
Signed-off-by: NTejun Heo <tj@kernel.org>

Currently this driver only works with a DesignWare DMA engine which it
registers manually using the second "reg" address range and interrupt
number from the DT node.

This patch makes the driver instead use the "dmas" property if present,
otherwise optionally falling back on the old way so existing device
trees can continue to work.

With this change, there is no longer any reason to depend on the 460EX
machine type so drop that from Kconfig.
Tested-by: NChristian Lamparter <chunkeey@googlemail.com>
Signed-off-by: NMans Rullgard <mans@mansr.com>
Signed-off-by: NTejun Heo <tj@kernel.org>

ata_sff_qc_issue() can't handle DMA commands and thus we have to avoid it for
them. Do call ata_bmdma_qc_issue() instead for this case. Note that the former
one distinguishes PIO and DMA mode and behaves accordingly.
Suggested-by: NChristian Lamparter <chunkeey@googlemail.com>
Signed-off-by: NAndy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: NTejun Heo <tj@kernel.org>

Calling dmaengine_prep_slave_sg() for non-dma ATA commands is
unnecessary at best and could be harmful if the dma driver reacts
badly to this.  It also causes this driver to print a bogus error
message in these cases.

This patch changes sata_dwc_qc_issue() to only do the dma setup
for dma commands and also reports an error to libata if if fails.
Tested-by: NChristian Lamparter <chunkeey@googlemail.com>
Signed-off-by: NMans Rullgard <mans@mansr.com>
Signed-off-by: NTejun Heo <tj@kernel.org>

There is no need to have a platform driver compiled since the DMA driver is
used as a library.
Tested-by: NChristian Lamparter <chunkeey@googlemail.com>
Signed-off-by: NAndy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: NTejun Heo <tj@kernel.org>

In the original code the DMA is always a flow controller. Set this accordingly
in updated code.
Tested-by: NChristian Lamparter <chunkeey@googlemail.com>
Signed-off-by: NAndy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: NTejun Heo <tj@kernel.org>

The burst size as defined by DMAengine API is in items of address width. Derive
burst size from AHB_DMA_BRST_DFLT (64 bytes) by dividing it to
DMA_SLAVE_BUSWIDTH_4_BYTES (4 bytes) that gives us 16 items.
Tested-by: NChristian Lamparter <chunkeey@googlemail.com>
Signed-off-by: NAndy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: NTejun Heo <tj@kernel.org>

The original code states:

	Make sure a LLI block is not created that will span 8K max FIS
	boundary. If the block spans such a FIS boundary, there is a chance
	that a DMA burst will cross that boundary -- this results in an error
	in the host controller.

Since we have switched to generic DMAengine API we satisfy above by setting
dma_boundary value to 0x1fff.
Suggested-by: NMans Rullgard <mans@mansr.com>
Tested-by: NChristian Lamparter <chunkeey@googlemail.com>
Signed-off-by: NAndy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: NTejun Heo <tj@kernel.org>

This patch fixes Machine Check "Data Write PLB Error" which happens
when libata-sff's ata_sff_dev_select is trying to write into the
device_addr in order to select a drive. However, SATA has no master
or slave devices like the old ATA Bus, therefore selecting a
different drive is kind of pointless.

Data Write PLB Error
Oops: Machine check, sig: 7 [#1]
PowerPC 44x Platform
Modules linked in:
CPU: 0 PID: 508 Comm: scsi_eh_0 Not tainted 4.6.0-rc3-next-20160412+ #10
[...]
NIP [c027e820] ata_sff_dev_select+0x3c/0x44
LR [c027e810] ata_sff_dev_select+0x2c/0x44
Call Trace:
[cec31cd0] [c027da00] ata_sff_postreset+0x40/0xb4 (unreliable)
[cec31ce0] [c027a03c] ata_eh_reset+0x5cc/0x928
[cec31d60] [c027a840] ata_eh_recover+0x330/0x10bc
[cec31df0] [c027bae0] ata_do_eh+0x4c/0xa4
[...]
Signed-off-by: NChristian Lamparter <chunkeey@googlemail.com>
Signed-off-by: NTejun Heo <tj@kernel.org>

This lock is already taken in ata_scsi_queuecmd() a few levels up the
call stack so attempting to take it here is an error.  Moreover, it is
pointless in the first place since it only protects a single, atomic
assignment.

Enabling lock debugging gives the following output:

=============================================
[ INFO: possible recursive locking detected ]
4.4.0-rc5+ #189 Not tainted
---------------------------------------------
kworker/u2:3/37 is trying to acquire lock:
 (&(&host->lock)->rlock){-.-...}, at: [<90283294>] sata_dwc_exec_command_by_tag.constprop.14+0x44/0x8c

but task is already holding lock:
 (&(&host->lock)->rlock){-.-...}, at: [<902761ac>] ata_scsi_queuecmd+0x2c/0x330

other info that might help us debug this:
 Possible unsafe locking scenario:

       CPU0
       ----
  lock(&(&host->lock)->rlock);
  lock(&(&host->lock)->rlock);

 *** DEADLOCK ***
 May be due to missing lock nesting notation

4 locks held by kworker/u2:3/37:
 #0:  ("events_unbound"){.+.+.+}, at: [<9003a0a4>] process_one_work+0x12c/0x430
 #1:  ((&entry->work)){+.+.+.}, at: [<9003a0a4>] process_one_work+0x12c/0x430
 #2:  (&bdev->bd_mutex){+.+.+.}, at: [<9011fd54>] __blkdev_get+0x50/0x380
 #3:  (&(&host->lock)->rlock){-.-...}, at: [<902761ac>] ata_scsi_queuecmd+0x2c/0x330

stack backtrace:
CPU: 0 PID: 37 Comm: kworker/u2:3 Not tainted 4.4.0-rc5+ #189
Workqueue: events_unbound async_run_entry_fn
Stack : 90b38e30 00000021 00000003 9b2a6040 00000000 9005f3f0 904fc8dc 00000025
        906b96e4 00000000 90528648 9b3336c4 904fc8dc 9009bf18 00000002 00000004
        00000000 00000000 9b3336c4 9b3336e4 904fc8dc 9003d074 00000000 90500000
        9005e738 00000000 00000000 00000000 00000000 00000000 00000000 00000000
        6e657665 755f7374 756f626e 0000646e 00000000 00000000 9b00ca00 9b025000
          ...
Call Trace:
[<90009d6c>] show_stack+0x88/0xa4
[<90057744>] __lock_acquire+0x1ce8/0x2154
[<900583e4>] lock_acquire+0x64/0x8c
[<9045ff10>] _raw_spin_lock_irqsave+0x54/0x78
[<90283294>] sata_dwc_exec_command_by_tag.constprop.14+0x44/0x8c
[<90283484>] sata_dwc_qc_issue+0x1a8/0x24c
[<9026b39c>] ata_qc_issue+0x1f0/0x410
[<90273c6c>] ata_scsi_translate+0xb4/0x200
[<90276234>] ata_scsi_queuecmd+0xb4/0x330
[<9025800c>] scsi_dispatch_cmd+0xd0/0x128
[<90259934>] scsi_request_fn+0x58c/0x638
[<901a3e50>] __blk_run_queue+0x40/0x5c
[<901a83d4>] blk_queue_bio+0x27c/0x28c
[<901a5914>] generic_make_request+0xf0/0x188
[<901a5a54>] submit_bio+0xa8/0x194
[<9011adcc>] submit_bh_wbc.isra.23+0x15c/0x17c
[<9011c908>] block_read_full_page+0x3e4/0x428
[<9009e2e0>] do_read_cache_page+0xac/0x210
[<9009fd90>] read_cache_page+0x18/0x24
[<901bbd18>] read_dev_sector+0x38/0xb0
[<901bd174>] msdos_partition+0xb4/0x5c0
[<901bcb8c>] check_partition+0x140/0x274
[<901bba60>] rescan_partitions+0xa0/0x2b0
[<9011ff68>] __blkdev_get+0x264/0x380
[<901201ac>] blkdev_get+0x128/0x36c
[<901b9378>] add_disk+0x3c0/0x4bc
[<90268268>] sd_probe_async+0x100/0x224
[<90043a44>] async_run_entry_fn+0x50/0x124
[<9003a11c>] process_one_work+0x1a4/0x430
[<9003a4f4>] worker_thread+0x14c/0x4fc
[<900408f4>] kthread+0xd0/0xe8
[<90004338>] ret_from_kernel_thread+0x14/0x1c

Fixes: 62936009 ("[libata] Add 460EX on-chip SATA driver, sata_dwc_460ex")
Tested-by: NChristian Lamparter <chunkeey@googlemail.com>
Signed-off-by: NMans Rullgard <mans@mansr.com>
Signed-off-by: NTejun Heo <tj@kernel.org>

We pass struct dw_dma_chip to dw_dma_probe() anyway, thus we may use it to
pass a platform data as well.

While here, constify the source of the platform data.
Acked-by: NViresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: NAndy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: NVinod Koul <vinod.koul@intel.com>

Keep the entire platform data in the struct dw_dma.
It makes the driver a bit cleaner.
Acked-by: NViresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: NAndy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: NVinod Koul <vinod.koul@intel.com>

There several changes are done here:

- Convert the property to be in bytes

  Besides that this is a common practice for such property, the use of a value
  in bytes much more convenient than handling the encoded one.

- Rename data_width to data-width in the device tree bindings

  The change leaves the support for the old format as well just in case someone
  will use a newer kernel with an old device tree blob.

- While here, replace dwc_fast_ffs() by __ffs()
Signed-off-by: NAndy Shevchenko <andriy.shevchenko@linux.intel.com>
Acked-by: NViresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: NVinod Koul <vinod.koul@intel.com>

The value of nr_masters equal to 0 is invalid since this DMA controller has to
have at least one master.

Check this before we proceed with the rest of properties.
Acked-by: NViresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: NAndy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: NVinod Koul <vinod.koul@intel.com>

This patch changes the driver to allocate DMA descriptors when
needed. This stops memory resources to be wasted and letting
them sit idle in the free_list structure when the device doesn't
need it... This also solves the problem, that a driver has to
guess the number of how many descriptors it needs to allocate
in advance. Currently, the dma engine will just fail when put
under load by sata_dwc_460ex.
Signed-off-by: NChristian Lamparter <chunkeey@googlemail.com>
Signed-off-by: NVinod Koul <vinod.koul@intel.com>

To be sure we have the cyclic transfers already gone we set cdesc to NULL. It
will prevent the double free.
Signed-off-by: NAndy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: NVinod Koul <vinod.koul@intel.com>

Residue is a property of any active descriptor. So, any descriptor may be in
different state but residue is a feature of active descriptor. Check if the
asked descriptor is active and return proper residue value for it.
Signed-off-by: NAndy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: NVinod Koul <vinod.koul@intel.com>

We have already dedicated variable for flags, therefore no need to create an
additional storage for that. Covert dwc->initialized to use dwc->flags.
Signed-off-by: NAndy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: NVinod Koul <vinod.koul@intel.com>

We have already dedicated variable for flags, therefore no need to create an
additional storage for that. Convert dwc->paused to use dwc->flags.
Signed-off-by: NAndy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: NVinod Koul <vinod.koul@intel.com>

The code is fixed to satisfy a compiler otherwise we have

drivers/dma/dw/core.c: In function ‘dwc_handle_cyclic’:
drivers/dma/dw/core.c:568: warning: comparison between signed and unsigned
drivers/dma/dw/core.c: In function ‘dw_dma_tasklet’:
drivers/dma/dw/core.c:590: warning: comparison between signed and unsigned
drivers/dma/dw/core.c: In function ‘dw_dma_off’:
drivers/dma/dw/core.c:1103: warning: comparison between signed and unsigned
drivers/dma/dw/core.c: In function ‘dw_dma_cyclic_free’:
drivers/dma/dw/core.c:1469: warning: comparison between signed and unsigned
drivers/dma/dw/core.c: In function ‘dw_dma_probe’:
drivers/dma/dw/core.c:1574: warning: comparison between signed and unsigned

There is no functional change.
Signed-off-by: NAndy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: NVinod Koul <vinod.koul@intel.com>

Since struct dw_dma is allocated and regs member is assigned properly we can
use standard IO accessors to the DMA registers.
Signed-off-by: NAndy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: NVinod Koul <vinod.koul@intel.com>

The datasheet requires that the LLP_[SD]_EN bits be cleared whenever
LLP.LOC is zero, i.e. in the last descriptor of a multi-block chain.
Make the driver do this.
Signed-off-by: NMans Rullgard <mans@mansr.com>
Acked-by: NAndy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: NVinod Koul <vinod.koul@intel.com>

The LMS field indicates from which master the descriptor is to be
read.  This patch assumes this is always the same as the memory
side in a peripheral transfer which is true for all known systems.
Signed-off-by: NMans Rullgard <mans@mansr.com>
Acked-by: NAndy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: NVinod Koul <vinod.koul@intel.com>

If the DMA controller uses a different byte order than the host CPU,
the hardware linked list descriptor fields need to be byte-swapped.

This patch makes the driver write these fields using the same byte
order it uses for mmio accesses to the DMA engine. I do not know
if this is guaranteed to always be correct.
Signed-off-by: NMans Rullgard <mans@mansr.com>
Acked-by: NAndy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: NVinod Koul <vinod.koul@intel.com>

On some architectures the DMA controller can have two masters connected to
different buses and thus access to memory is possible only through one and
to peripheral through the other.

This patch changes the src and dst master setting to match the direction
of the transfer.
Signed-off-by: NMans Rullgard <mans@mansr.com>
Acked-by: NAndy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: NVinod Koul <vinod.koul@intel.com>

The source and destination masters are reflecting buses or their layers to
where the different devices can be connected. The patch changes the master
names to reflect which one is related to which independently on the transfer
direction.

The outcome of the change is that the memory data width is now always limited
by a data width of the master which is dedicated to communicate to memory.

The patch will not break anything since all current users have the same data
width for all masters. Though it would be nice to revisit avr32 platforms to
check what is the actual hardware topology in use there. It seems that it has
one bus and two masters on it as stated by Table 8-2, that's why everything
works independently on the master in use. The purpose of the sequential patch
is to fix the driver for configuration of more than one bus.

The change is done in the assumption that src_master and dst_master are
reflecting a connection to the memory and peripheral correspondently on avr32
and otherwise on the rest.
Acked-by: NHans-Christian Egtvedt <egtvedt@samfundet.no>
Acked-by: NMark Brown <broonie@kernel.org>
Signed-off-by: NAndy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: NVinod Koul <vinod.koul@intel.com>

The commit 89500520 ("dmaengine: dw: apply both HS interfaces and remove
slave_id usage") cleaned up the code to avoid usage of depricated slave_id
member of generic slave configuration.

Meanwhile it broke the master selection by removing important call to
dwc_set_masters() in ->device_alloc_chan_resources() which copied masters from
custom slave configuration to the internal channel structure.

Everything works until now since there is no customized connection of
DesignWare DMA IP to the bus, i.e. one bus and one or more masters are in use.
The configurations where 2 masters are connected to the different masters are
not working anymore. We are expecting one user of such configuration and need
to select masters properly. Besides that it is obviously a performance
regression since only one master is in use in multi-master configuration.

Select masters in accordance with what user asked for. Keep this patch in a form
more suitable for back porting.

We are safe to take necessary data in ->device_alloc_chan_resources() because
we don't support generic slave configuration embedded into custom one, and thus
the only way to provide such is to use the parameter to a filter function which
is called exactly before channel resource allocation.

While here, replase BUG_ON to less noisy dev_warn() and prevent channel
allocation in case of error.

Fixes: 89500520 ("dmaengine: dw: apply both HS interfaces and remove slave_id usage")
Cc: stable@vger.kernel.org
Signed-off-by: NAndy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: NVinod Koul <vinod.koul@intel.com>

Pull Ceph updates from Sage Weil:
 "There is quite a bit here, including some overdue refactoring and
  cleanup on the mon_client and osd_client code from Ilya, scattered
  writeback support for CephFS and a pile of bug fixes from Zheng, and a
  few random cleanups and fixes from others"

[ I already decided not to pull this because of it having been rebased
  recently, but ended up changing my mind after all.  Next time I'll
  really hold people to it.  Oh well.   - Linus ]

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client: (34 commits)
  libceph: use KMEM_CACHE macro
  ceph: use kmem_cache_zalloc
  rbd: use KMEM_CACHE macro
  ceph: use lookup request to revalidate dentry
  ceph: kill ceph_get_dentry_parent_inode()
  ceph: fix security xattr deadlock
  ceph: don't request vxattrs from MDS
  ceph: fix mounting same fs multiple times
  ceph: remove unnecessary NULL check
  ceph: avoid updating directory inode's i_size accidentally
  ceph: fix race during filling readdir cache
  libceph: use sizeof_footer() more
  ceph: kill ceph_empty_snapc
  ceph: fix a wrong comparison
  ceph: replace CURRENT_TIME by current_fs_time()
  ceph: scattered page writeback
  libceph: add helper that duplicates last extent operation
  libceph: enable large, variable-sized OSD requests
  libceph: osdc->req_mempool should be backed by a slab pool
  libceph: make r_request msg_size calculation clearer
  ...

Pull orangefs filesystem from Mike Marshall.

This finally merges the long-pending orangefs filesystem, which has been
much cleaned up with input from Al Viro over the last six months.  From
the documentation file:

 "OrangeFS is an LGPL userspace scale-out parallel storage system.  It
  is ideal for large storage problems faced by HPC, BigData, Streaming
  Video, Genomics, Bioinformatics.

  Orangefs, originally called PVFS, was first developed in 1993 by Walt
  Ligon and Eric Blumer as a parallel file system for Parallel Virtual
  Machine (PVM) as part of a NASA grant to study the I/O patterns of
  parallel programs.

  Orangefs features include:

    - Distributes file data among multiple file servers
    - Supports simultaneous access by multiple clients
    - Stores file data and metadata on servers using local file system
      and access methods
    - Userspace implementation is easy to install and maintain
    - Direct MPI support
    - Stateless"

see Documentation/filesystems/orangefs.txt for more in-depth details.

* tag 'ofs-pull-tag-1' of git://git.kernel.org/pub/scm/linux/kernel/git/hubcap/linux: (174 commits)
  orangefs: fix orangefs_superblock locking
  orangefs: fix do_readv_writev() handling of error halfway through
  orangefs: have ->kill_sb() evict the VFS side of things first
  orangefs: sanitize ->llseek()
  orangefs-bufmap.h: trim unused junk
  orangefs: saner calling conventions for getting a slot
  orangefs_copy_{to,from}_bufmap(): don't pass bufmap pointer
  orangefs: get rid of readdir_handle_s
  ornagefs: ensure that truncate has an up to date inode size
  orangefs: move code which sets i_link to orangefs_inode_getattr
  orangefs: remove needless wrapper around GFP_KERNEL
  orangefs: remove wrapper around mutex_lock(&inode->i_mutex)
  orangefs: refactor inode type or link_target change detection
  orangefs: use new getattr for revalidate and remove old getattr
  orangefs: use new getattr in inode getattr and permission
  orangefs: use new orangefs_inode_getattr to get size in write and llseek
  orangefs: use new orangefs_inode_getattr to create new inodes
  orangefs: rename orangefs_inode_getattr to orangefs_inode_old_getattr
  orangefs: remove inode->i_lock wrapper
  orangefs: put register_chrdev immediately before register_filesystem
  ...

Pull NTB bug fixes from Jon Mason:
 "NTB bug fixes for tasklet from spinning forever, link errors,
  translation window setup, NULL ptr dereference, and ntb-perf errors.

  Also, a modification to the driver API that makes _addr functions
  optional"

* tag 'ntb-4.6' of git://github.com/jonmason/ntb:
  NTB: Remove _addr functions from ntb_hw_amd
  NTB: Make _addr functions optional in the API
  NTB: Fix incorrect clean up routine in ntb_perf
  NTB: Fix incorrect return check in ntb_perf
  ntb: fix possible NULL dereference
  ntb: add missing setup of translation window
  ntb: stop link work when we do not have memory
  ntb: stop tasklet from spinning forever during shutdown.
  ntb: perf test: fix address space confusion