Commit writes to media on system shutdown or pmem driver unload.
Signed-off-by: NDan Williams <dan.j.williams@intel.com>

Given that nvdimm_flush() has higher overhead than wmb_pmem() (pointer
chasing through nd_region), and that we otherwise assume a platform has
ADR capability when flush hints are not present, move nvdimm_flush() to
REQ_FLUSH context.

Note that we still arrange for nvdimm_flush() to be called even in the
ADR case. We need at least once wmb() fence to push buffered writes in
the cpu out to the ADR protected domain.

Cc: Toshi Kani <toshi.kani@hpe.com>
Cc: Ross Zwisler <ross.zwisler@linux.intel.com>
Signed-off-by: NDan Williams <dan.j.williams@intel.com>

nvdimm_flush() is a replacement for the x86 'pcommit' instruction.  It is
an optional write flushing mechanism that an nvdimm bus can provide for
the pmem driver to consume.  In the case of the NFIT nvdimm-bus-provider
nvdimm_flush() is implemented as a series of flush-hint-address [1]
writes to each dimm in the interleave set (region) that backs the
namespace.

The nvdimm_has_flush() routine relies on platform firmware to describe
the flushing capabilities of a platform.  It uses the heuristic of
whether an nvdimm bus provider provides flush address data to return a
ternary result:

      1: flush addresses defined
      0: dimm topology described without flush addresses (assume ADR)
 -errno: no topology information, unable to determine flush mechanism

The pmem driver is expected to take the following actions on this ternary
result:

      1: nvdimm_flush() in response to REQ_FUA / REQ_FLUSH and shutdown
      0: do not set, WC or FUA on the queue, take no further action
 -errno: warn and then operate as if nvdimm_has_flush() returned '0'

The caveat of this heuristic is that it can not distinguish the "dimm
does not have flush address" case from the "platform firmware is broken
and failed to describe a flush address".  Given we are already
explicitly trusting the NFIT there's not much more we can do beyond
blacklisting broken firmwares if they are ever encountered.

Cc: Ross Zwisler <ross.zwisler@linux.intel.com>
Signed-off-by: NDan Williams <dan.j.williams@intel.com>

Currently phys_to_pfn_t() is an exported symbol to allow nfit_test to
override it and indicate that nfit_test-pmem is not device-mapped.  Now,
we want to enable nfit_test to operate without DMA_CMA and the pmem it
provides will no longer be physically contiguous, i.e. won't be capable
of supporting direct_access requests larger than a page.  Make
pmem_direct_access() a weak symbol so that it can be replaced by the
tools/testing/nvdimm/ version, and move phys_to_pfn_t() to a static
inline now that it no longer needs to be overridden.
Acked-by: NJohannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: NDan Williams <dan.j.williams@intel.com>

Clean up needless calls to the action routine by letting
devm_add_action_or_reset() call it automatically.  This does cause the
disk to registered and immediately unregistered when a memory allocation
fails, but the block layer should be prepared for such an event.
Reported-by: NSudip Mukherjee <sudipm.mukherjee@gmail.com>
Signed-off-by: NDan Williams <dan.j.williams@intel.com>

For autodetecting a previously established dax configuration we need the
info block to indicate block-device vs device-dax mode, and we need to
have the default namespace probe hand-off the configuration to the
dax_pmem driver.
Signed-off-by: NDan Williams <dan.j.williams@intel.com>

1/ If a mapping overlaps a bad sector fail the request.

2/ Do not opportunistically report more dax-capable capacity than is
   requested when errors present.
Reviewed-by: NJeff Moyer <jmoyer@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NDan Williams <dan.j.williams@intel.com>
[vishal: fix a conflict with system RAM collision patches]
[vishal: add a 'size' parameter to ->direct_access]
[vishal: fix a conflict with DAX alignment check patches]
Signed-off-by: NVishal Verma <vishal.l.verma@intel.com>

I had relied on the kbuild robot for cross build coverage, however it
only builds alpha_defconfig.  Switch from HPAGE_SIZE to PMD_SIZE, which
is more widely defined.

Fixes: 658922e5 ("libnvdimm, pfn: fix memmap reservation sizing")
Cc: <stable@vger.kernel.org>
Reported-by: NGuenter Roeck <guenter@roeck-us.net>
Tested-by: NGuenter Roeck <guenter@roeck-us.net>
Signed-off-by: NDan Williams <dan.j.williams@intel.com>

When configuring a pfn-device instance to allocate the memmap array it
needs to account for the fact that vmemmap_populate_hugepages()
allocates struct page blocks in HPAGE_SIZE chunks.  We need to align the
reserved area size to 2MB otherwise arch_add_memory() runs out of memory
while establishing the memmap:

 WARNING: CPU: 0 PID: 496 at arch/x86/mm/init_64.c:704 arch_add_memory+0xe7/0xf0
 [..]
 Call Trace:
  [<ffffffff8148bdb3>] dump_stack+0x85/0xc2
  [<ffffffff810a749b>] __warn+0xcb/0xf0
  [<ffffffff810a75cd>] warn_slowpath_null+0x1d/0x20
  [<ffffffff8106a497>] arch_add_memory+0xe7/0xf0
  [<ffffffff811d2097>] devm_memremap_pages+0x287/0x450
  [<ffffffff811d1ffa>] ? devm_memremap_pages+0x1ea/0x450
  [<ffffffffa0000298>] __wrap_devm_memremap_pages+0x58/0x70 [nfit_test_iomap]
  [<ffffffffa0047a58>] pmem_attach_disk+0x318/0x420 [nd_pmem]
  [<ffffffffa0047bcf>] nd_pmem_probe+0x6f/0x90 [nd_pmem]
  [<ffffffffa0009469>] nvdimm_bus_probe+0x69/0x110 [libnvdimm]
 [..]
  ndbus0: nd_pmem.probe(pfn3.0) = -12
 nd_pmem: probe of pfn3.0 failed with error -12
libndctl: ndctl_pfn_enable: pfn3.0: failed to enable
Reported-by: NNamratha Kothapalli <namratha.n.kothapalli@intel.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: NDan Williams <dan.j.williams@intel.com>

The devm conversion obviates the need to continue to remember the queue
and disk locally in the driver.
Reviewed-by: NJohannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: NDan Williams <dan.j.williams@intel.com>

Now that pmem internals have been disentangled from pfn setup, that code
can move to the core.  This is in preparation for adding another user of
the pfn-device capabilities.
Reviewed-by: NJohannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: NDan Williams <dan.j.williams@intel.com>

In preparation for providing an alternative (to block device) access
mechanism to persistent memory, convert pmem_rw_bytes() to
nsio_rw_bytes().  This allows ->rw_bytes() functionality without
requiring a 'struct pmem_device' to be instantiated.

In other words, when ->rw_bytes() is in use i/o is driven through
'struct nd_namespace_io', otherwise it is driven through 'struct
pmem_device' and the block layer.  This consolidates the disjoint calls
to devm_exit_badblocks() and devm_memunmap() into a common
devm_nsio_disable() and cleans up the init path to use a unified
pmem_attach_disk() implementation.
Reviewed-by: NJohannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: NDan Williams <dan.j.williams@intel.com>

The leading '0x' in front of %pa is redundant, also we can just use %pR
to simplify the print statement.  The request parameters can be directly
taken from the resource as well.
Reviewed-by: NJohannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: NDan Williams <dan.j.williams@intel.com>

Register a callback to clean up the request_queue and put the gendisk at
driver disable time.

Cc: Ross Zwisler <ross.zwisler@linux.intel.com>
Reviewed-by: NJohannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: NDan Williams <dan.j.williams@intel.com>

Save a pointer chase by storing the driver private data in the
request_queue rather than the gendisk.
Reviewed-by: NJohannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: NDan Williams <dan.j.williams@intel.com>

Pass the device performing the probe so we can use a devm allocation for
the btt superblock.

Cc: Vishal Verma <vishal.l.verma@intel.com>
Reviewed-by: NJohannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: NDan Williams <dan.j.williams@intel.com>

Pass the device performing the probe so we can use a devm allocation for
the pfn superblock.

Cc: Ross Zwisler <ross.zwisler@linux.intel.com>
Reviewed-by: NJohannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: NDan Williams <dan.j.williams@intel.com>

We can derive the common namespace from other information.  We also do
not need to cache it because all the usages are in slow paths.
Reviewed-by: NJohannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: NDan Williams <dan.j.williams@intel.com>

The ACPI specification does not specify the state of data after a clear
poison operation.  Potential future libnvdimm bus implementations for
other architectures also might not specify or disagree on the state of
data after clear poison.  Clarify why we write twice.
Reported-by: NJeff Moyer <jmoyer@redhat.com>
Reported-by: NVishal Verma <vishal.l.verma@intel.com>
Signed-off-by: NDan Williams <dan.j.williams@intel.com>
Signed-off-by: NRoss Zwisler <ross.zwisler@linux.intel.com>
Reviewed-by: NJohannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: NJeff Moyer <jmoyer@redhat.com>
Reviewed-by: NVishal Verma <vishal.l.verma@intel.com>

When section alignment padding is in effect we need to shift / truncate
the range that is queried for poison by the 'start_pad' or 'end_trunc'
reservations.

It's easiest if we just pass in an adjusted resource range rather than
deriving it from the passed in namespace.  With the resource range
resolution pushed out to the caller we can also push the
namespace-to-region lookup to the caller and drop the implicit pmem-type
assumption about the passed in namespace object.

Cc: Vishal Verma <vishal.l.verma@intel.com>
Signed-off-by: NDan Williams <dan.j.williams@intel.com>

PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time
ago with promise that one day it will be possible to implement page
cache with bigger chunks than PAGE_SIZE.

This promise never materialized.  And unlikely will.

We have many places where PAGE_CACHE_SIZE assumed to be equal to
PAGE_SIZE.  And it's constant source of confusion on whether
PAGE_CACHE_* or PAGE_* constant should be used in a particular case,
especially on the border between fs and mm.

Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much
breakage to be doable.

Let's stop pretending that pages in page cache are special.  They are
not.

The changes are pretty straight-forward:

 - <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;

 - <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;

 - PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN};

 - page_cache_get() -> get_page();

 - page_cache_release() -> put_page();

This patch contains automated changes generated with coccinelle using
script below.  For some reason, coccinelle doesn't patch header files.
I've called spatch for them manually.

The only adjustment after coccinelle is revert of changes to
PAGE_CAHCE_ALIGN definition: we are going to drop it later.

There are few places in the code where coccinelle didn't reach.  I'll
fix them manually in a separate patch.  Comments and documentation also
will be addressed with the separate patch.

virtual patch

@@
expression E;
@@
- E << (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E

@@
expression E;
@@
- E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E

@@
@@
- PAGE_CACHE_SHIFT
+ PAGE_SHIFT

@@
@@
- PAGE_CACHE_SIZE
+ PAGE_SIZE

@@
@@
- PAGE_CACHE_MASK
+ PAGE_MASK

@@
expression E;
@@
- PAGE_CACHE_ALIGN(E)
+ PAGE_ALIGN(E)

@@
expression E;
@@
- page_cache_get(E)
+ get_page(E)

@@
expression E;
@@
- page_cache_release(E)
+ put_page(E)
Signed-off-by: NKirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: NMichal Hocko <mhocko@suse.com>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Update the definition of memcpy_from_pmem() to return 0 or a negative
error code.  Implement x86/arch_memcpy_from_pmem() with memcpy_mcsafe().

Cc: Borislav Petkov <bp@alien8.de>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Acked-by: NIngo Molnar <mingo@kernel.org>
Reviewed-by: NRoss Zwisler <ross.zwisler@linux.intel.com>
Signed-off-by: NDan Williams <dan.j.williams@intel.com>

If a write is directed at a known bad block perform the following:

1/ write the data

2/ send a clear poison command

3/ invalidate the poison out of the cache hierarchy

Cc: <x86@kernel.org>
Cc: Ross Zwisler <ross.zwisler@linux.intel.com>
Reviewed-by: NVishal Verma <vishal.l.verma@intel.com>
Signed-off-by: NDan Williams <dan.j.williams@intel.com>

When we enounter a bad block we need to kunmap_atomic() before
returning.

Cc: <stable@vger.kernel.org>
Cc: Ross Zwisler <ross.zwisler@linux.intel.com>
Reviewed-by: NVishal Verma <vishal.l.verma@intel.com>
Signed-off-by: NDan Williams <dan.j.williams@intel.com>

When alloc_disk(0) or alloc_disk-node(0, XX) is used, the ->major
number is completely ignored:  all devices are allocated with a
major of BLOCK_EXT_MAJOR.

So there is no point allocating pmem_major.
Signed-off-by: NNeilBrown <neilb@suse.com>
Signed-off-by: NDan Williams <dan.j.williams@intel.com>

   drivers/nvdimm/pmem.c: In function 'nvdimm_namespace_attach_pfn':
   drivers/nvdimm/pmem.c:367:3: error: implicit declaration of function
   	'__phys_to_pfn' [-Werror=implicit-function-declaration]
   .base_pfn = __phys_to_pfn(nsio->res.start),

ia64 does not provide __phys_to_pfn(), just use the PHYS_PFN() alias.

Cc: Guenter Roeck <linux@roeck-us.net>
Reported-by: Nkbuild test robot <fengguang.wu@intel.com>
Signed-off-by: NDan Williams <dan.j.williams@intel.com>

On a platform where 'Persistent Memory' and 'System RAM' are mixed
within a given sparsemem section, trim the namespace and notify about the
sub-optimal alignment.

Cc: Toshi Kani <toshi.kani@hpe.com>
Cc: Ross Zwisler <ross.zwisler@linux.intel.com>
Signed-off-by: NDan Williams <dan.j.williams@intel.com>

The altmap for a section-misaligned namespace needs to arrange for the
base_pfn to be section-aligned.  As a result the 'reserve' region (pfns
from base that do not have a struct page) must be increased.  Otherwise
we trip the altmap validation check in __add_pages:

	if (altmap->base_pfn != phys_start_pfn
			|| vmem_altmap_offset(altmap) > nr_pages) {
		pr_warn_once("memory add fail, invalid altmap\n");
		return -EINVAL;
	}
Signed-off-by: NDan Williams <dan.j.williams@intel.com>

In preparation for asynchronous address range scrub support add an
ability for the pmem driver to dynamically consume address range scrub
results.
Signed-off-by: NDan Williams <dan.j.williams@intel.com>

A recent bugfix changed pfn_t to always be 64-bit wide, but did not
change the code in pmem.c, which is now broken on 32-bit architectures
as reported by gcc:

In file included from ../drivers/nvdimm/pmem.c:28:0:
drivers/nvdimm/pmem.c: In function 'pmem_alloc':
include/linux/pfn_t.h:15:17: error: large integer implicitly truncated to unsigned type [-Werror=overflow]
 #define PFN_DEV (1ULL << (BITS_PER_LONG_LONG - 3))

This changes the intermediate pfn_flags in struct pmem_device to
be 64 bit wide as well, so they can store the flags correctly.
Signed-off-by: NArnd Bergmann <arnd@arndb.de>
Fixes: db78c222 ("mm: fix pfn_t vs highmem")
Signed-off-by: NDan Williams <dan.j.williams@intel.com>

get_dev_page() enables paths like get_user_pages() to pin a dynamically
mapped pfn-range (devm_memremap_pages()) while the resulting struct page
objects are in use.  Unlike get_page() it may fail if the device is, or
is in the process of being, disabled.  While the initial lookup of the
range may be an expensive list walk, the result is cached to speed up
subsequent lookups which are likely to be in the same mapped range.

devm_memremap_pages() now requires a reference counter to be specified
at init time.  For pmem this means moving request_queue allocation into
pmem_alloc() so the existing queue usage counter can track "device
pages".

ZONE_DEVICE pages always have an elevated count and will never be on an
lru reclaim list.  That space in 'struct page' can be redirected for
other uses, but for safety introduce a poison value that will always
trip __list_add() to assert.  This allows half of the struct list_head
storage to be reclaimed with some assurance to back up the assumption
that the page count never goes to zero and a list_add() is never
attempted.
Signed-off-by: NDan Williams <dan.j.williams@intel.com>
Tested-by: NLogan Gunthorpe <logang@deltatee.com>
Cc: Dave Hansen <dave@sr71.net>
Cc: Matthew Wilcox <willy@linux.intel.com>
Cc: Ross Zwisler <ross.zwisler@linux.intel.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Before the dynamically allocated struct pages from devm_memremap_pages()
can be put to use outside the driver, we need a mechanism to track
whether they are still in use at teardown.  Towards that goal reorder
the initialization sequence to allow the 'q_usage_counter' from the
request_queue to be used by the devm_memremap_pages() implementation (in
subsequent patches).
Signed-off-by: NDan Williams <dan.j.williams@intel.com>
Cc: Ross Zwisler <ross.zwisler@linux.intel.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Use the new vmem_altmap capability to enable the pmem driver to arrange
for a struct page memmap to be established in persistent memory.

[linux@roeck-us.net: mn10300: declare __pfn_to_phys() to fix build error]
Signed-off-by: NDan Williams <dan.j.williams@intel.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Ross Zwisler <ross.zwisler@linux.intel.com>
Signed-off-by: NGuenter Roeck <linux@roeck-us.net>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

In support of providing struct page for large persistent memory
capacities, use struct vmem_altmap to change the default policy for
allocating memory for the memmap array.  The default vmemmap_populate()
allocates page table storage area from the page allocator.  Given
persistent memory capacities relative to DRAM it may not be feasible to
store the memmap in 'System Memory'.  Instead vmem_altmap represents
pre-allocated "device pages" to satisfy vmemmap_alloc_block_buf()
requests.
Signed-off-by: NDan Williams <dan.j.williams@intel.com>
Reported-by: Nkbuild test robot <lkp@intel.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

There are several scenarios where we need to retrieve and update
metadata associated with a given devm_memremap_pages() mapping, and the
only lookup key available is a pfn in the range:

1/ We want to augment vmemmap_populate() (called via arch_add_memory())
   to allocate memmap storage from pre-allocated pages reserved by the
   device driver.  At vmemmap_alloc_block_buf() time it grabs device pages
   rather than page allocator pages.  This is in support of
   devm_memremap_pages() mappings where the memmap is too large to fit in
   main memory (i.e. large persistent memory devices).

2/ Taking a reference against the mapping when inserting device pages
   into the address_space radix of a given inode.  This facilitates
   unmap_mapping_range() and truncate_inode_pages() operations when the
   driver is tearing down the mapping.

3/ get_user_pages() operations on ZONE_DEVICE memory require taking a
   reference against the mapping so that the driver teardown path can
   revoke and drain usage of device pages.
Signed-off-by: NDan Williams <dan.j.williams@intel.com>
Tested-by: NLogan Gunthorpe <logang@deltatee.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Ross Zwisler <ross.zwisler@linux.intel.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

For the purpose of communicating the optional presence of a 'struct
page' for the pfn returned from ->direct_access(), introduce a type that
encapsulates a page-frame-number plus flags.  These flags contain the
historical "page_link" encoding for a scatterlist entry, but can also
denote "device memory".  Where "device memory" is a set of pfns that are
not part of the kernel's linear mapping by default, but are accessed via
the same memory controller as ram.

The motivation for this new type is large capacity persistent memory
that needs struct page entries in the 'memmap' to support 3rd party DMA
(i.e.  O_DIRECT I/O with a persistent memory source/target).  However,
we also need it in support of maintaining a list of mapped inodes which
need to be unmapped at driver teardown or freeze_bdev() time.
Signed-off-by: NDan Williams <dan.j.williams@intel.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Dave Hansen <dave@sr71.net>
Cc: Ross Zwisler <ross.zwisler@linux.intel.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Support badblock checking in all the pmem read paths that do not go
through the block layer.  This protects info block reads (btt or pfn) as
well as data reads to a pmem namespace via a btt instance.
Signed-off-by: NDan Williams <dan.j.williams@intel.com>

Longer term teach dax to punch "error" holes in mapping requests and
deliver SIGBUS to applications that consume a bad pmem page.  For now,
simply disable the dax performance optimization in the presence of known
errors.
Signed-off-by: NDan Williams <dan.j.williams@intel.com>

Check the sectors specified in a read bio to see if they hit a known bad
block, and return an error code pmem_do_bvec().

Note that the ->rw_page() is not in a position to return errors.  For
now, copy the same layering violation present in zram_rw_page() to avoid
crashes of the form:

 kernel BUG at mm/filemap.c:822!
 [..]
 Call Trace:
  [<ffffffff811c540e>] page_endio+0x1e/0x60
  [<ffffffff81290d29>] mpage_end_io+0x39/0x60
  [<ffffffff8141c4ef>] bio_endio+0x3f/0x60
  [<ffffffffa005c491>] pmem_make_request+0x111/0x230 [nd_pmem]

...i.e. unlock a page that was already unlocked via pmem_rw_page() =>
page_endio().
Reported-by: NVishal Verma <vishal.l.verma@intel.com>
Signed-off-by: NDan Williams <dan.j.williams@intel.com>

If a device will ever have badblocks it should always have a badblocks
instance available.  So, similar to md, embed a badblocks instance in
pmem_device.  This reduces pointer chasing in the i/o fast path, and
simplifies the init path.
Reported-by: NVishal Verma <vishal.l.verma@intel.com>
Signed-off-by: NDan Williams <dan.j.williams@intel.com>