While sending each RPC Reply, svc_rdma_sendto allocates and DMA-
maps a separate buffer where the RPC/RDMA transport header is
constructed. The buffer is unmapped and released in the Send
completion handler. This is significant per-RPC overhead,
especially for small RPCs.

Instead, allocate and DMA-map a buffer, and cache it in each
svc_rdma_send_ctxt. This buffer and its mapping can be re-used
for each RPC, saving the cost of memory allocation and DMA
mapping.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>

Clean up: No current caller of svc_rdma_send's passes in a chained
WR. The logic that counts the chain length can be replaced with a
constant (1).
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>

Clean up: Now that the send_wr is part of the svc_rdma_send_ctxt,
svc_rdma_post_send_wr is nearly empty.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>

Receive buffers are always the same size, but each Send WR has a
variable number of SGEs, based on the contents of the xdr_buf being
sent.

While assembling a Send WR, keep track of the number of SGEs so that
we don't exceed the device's maximum, or walk off the end of the
Send SGE array.

For now the Send path just fails if it exceeds the maximum.

The current logic in svc_rdma_accept bases the maximum number of
Send SGEs on the largest NFS request that can be sent or received.
In the transport layer, the limit is actually based on the
capabilities of the underlying device, not on properties of the
Upper Layer Protocol.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>

svc_rdma_op_ctxt's are pre-allocated and maintained on a per-xprt
free list. This eliminates the overhead of calling kmalloc / kfree,
both of which grab a globally shared lock that disables interrupts.
Introduce a replacement to svc_rdma_op_ctxt's that is built
especially for the svcrdma Send path.

Subsequent patches will take advantage of this new structure by
allocating real resources which are then cached in these objects.
The allocations are freed when the transport is torn down.

I've renamed the structure so that static type checking can be used
to ensure that uses of op_ctxt and send_ctxt are not confused. As an
additional clean up, structure fields are renamed to conform with
kernel coding conventions.

Additional clean ups:
- Handle svc_rdma_send_ctxt_get allocation failure at each call
  site, rather than pre-allocating and hoping we guessed correctly
- All send_ctxt_put call-sites request page freeing, so remove
  the @free_pages argument
- All send_ctxt_put call-sites unmap SGEs, so fold that into
  svc_rdma_send_ctxt_put
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>

Clean up: Since there's already a svc_rdma_op_ctxt being passed
around with the running count of mapped SGEs, drop unneeded
parameters to svc_rdma_post_send_wr().
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>

Clean up: svc_rdma_dma_map_buf does mostly the same thing as
svc_rdma_dma_map_page, so let's fold these together.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>

There is a significant latency penalty when processing an ingress
Receive if the Receive buffer resides in memory that is not on the
same NUMA node as the the CPU handling completions for a CQ.

The system administrator and the device driver determine which CPU
handles completions. This CPU does not change during life of the CQ.
Further the Upper Layer does not have any visibility of which CPU it
is.

Allocating Receive buffers in the Receive completion handler
guarantees that Receive buffers are allocated on the preferred NUMA
node for that CQ.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>

The current Receive path uses an array of pages which are allocated
and DMA mapped when each Receive WR is posted, and then handed off
to the upper layer in rqstp::rq_arg. The page flip releases unused
pages in the rq_pages pagelist. This mechanism introduces a
significant amount of overhead.

So instead, kmalloc the Receive buffer, and leave it DMA-mapped
while the transport remains connected. This confers a number of
benefits:

* Each Receive WR requires only one receive SGE, no matter how large
  the inline threshold is. This helps the server-side NFS/RDMA
  transport operate on less capable RDMA devices.

* The Receive buffer is left allocated and mapped all the time. This
  relieves svc_rdma_post_recv from the overhead of allocating and
  DMA-mapping a fresh buffer.

* svc_rdma_wc_receive no longer has to DMA unmap the Receive buffer.
  It has to DMA sync only the number of bytes that were received.

* svc_rdma_build_arg_xdr no longer has to free a page in rq_pages
  for each page in the Receive buffer, making it a constant-time
  function.

* The Receive buffer is now plugged directly into the rq_arg's
  head[0].iov_vec, and can be larger than a page without spilling
  over into rq_arg's page list. This enables simplification of
  the RDMA Read path in subsequent patches.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>

Rather than releasing the incoming svc_rdma_recv_ctxt at the end of
svc_rdma_recvfrom, hold onto it until svc_rdma_sendto.

This permits the contents of the Receive buffer to be preserved
through svc_process and then referenced directly in sendto as it
constructs Write and Reply chunks to return to the client.

The real changes will come in subsequent patches.

Note: I cannot use ->xpo_release_rqst for this purpose because that
is called _before_ ->xpo_sendto. svc_rdma_sendto uses information in
the received Call transport header to construct the Reply transport
header, which is preserved in the RPC's Receive buffer.

The historical comment in svc_send() isn't helpful: it is already
obvious that ->xpo_release_rqst is being called before ->xpo_sendto,
but there is no explanation for this ordering going back to the
beginning of the git era.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>

Currently svc_rdma_recv_ctxt_put's callers have to know whether they
want to free the ctxt's pages or not. This means the human
developers have to know when and why to set that free_pages
argument.

Instead, the ctxt should carry that information with it so that
svc_rdma_recv_ctxt_put does the right thing no matter who is
calling.

We want to keep track of the number of pages in the Receive buffer
separately from the number of pages pulled over by RDMA Read. This
is so that the correct number of pages can be freed properly and
that number is well-documented.

So now, rc_hdr_count is the number of pages consumed by head[0]
(ie., the page index where the Read chunk should start); and
rc_page_count is always the number of pages that need to be released
when the ctxt is put.

The @free_pages argument is no longer needed.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>

Clean up: No need to retain rq_depth in struct svcrdma_xprt, it is
used only in svc_rdma_accept().
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>

svc_rdma_op_ctxt's are pre-allocated and maintained on a per-xprt
free list. This eliminates the overhead of calling kmalloc / kfree,
both of which grab a globally shared lock that disables interrupts.
To reduce contention further, separate the use of these objects in
the Receive and Send paths in svcrdma.

Subsequent patches will take advantage of this separation by
allocating real resources which are then cached in these objects.
The allocations are freed when the transport is torn down.

I've renamed the structure so that static type checking can be used
to ensure that uses of op_ctxt and recv_ctxt are not confused. As an
additional clean up, structure fields are renamed to conform with
kernel coding conventions.

As a final clean up, helpers related to recv_ctxt are moved closer
to the functions that use them.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>

This includes:
  * Posting on the Send and Receive queues
  * Send, Receive, Read, and Write completion
  * Connect upcalls
  * QP errors
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>

This includes:
  * Transport accept and tear-down
  * Decisions about using Write and Reply chunks
  * Each RDMA segment that is handled
  * Whenever an RDMA_ERR is sent

As a clean-up, I've standardized the order of the includes, and
removed some now redundant dprintk call sites.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>

Clean up: Move #include <trace/events/rpcrdma.h> into source files,
similar to how it is done with trace/events/sunrpc.h.

Server-side trace points will be part of the rpcrdma subsystem,
just like the client-side trace points.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>

Ensure each RDMA listener and its children transports are created in
the same net namespace as the user that started the NFS service.
This is similar to how listener sockets are created in
svc_create_socket, required for enabling support for containers.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>

Currently the knfsd replay cache appears to try to refuse replying to
retries that come within 200ms of the cache entry being created. That
makes limited sense in today's world of high speed TCP.

After a TCP disconnection, a client can very easily reconnect and retry
an rpc in less than 200ms.  If this logic drops that retry, however, the
client may be quite slow to retry again.  This logic is original to the
first reply cache implementation in 2.1, and may have made more sense
for UDP clients that retried much more frequently.

After this patch we will still drop on finding the original request
still in progress.  We may want to fix that as well at some point,
though it's less likely.

Note that svc_check_conn_limits is often the cause of those
disconnections.  We may want to fix that some day.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
Acked-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>

nfsd4_scsi_identify_device() performs a single IDENTIFY command for the
device identification VPD page using a small buffer.  If the reply is
too large to fit in this buffer then the GETDEVICEINFO reply will not
contain any info for the SCSI volume aside from the registration key.
This can happen for example if the device has descriptors using long
SCSI name strings.

When the initial reply from the device indicates a larger buffer is
needed, retry once using the page length from that reply.
Signed-off-by: Scott Mayhew <smayhew@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>

nfsd4_readdir_rsize restricts rd_maxcount to svc_max_payload when
estimating the size of the readdir reply, but nfsd_encode_readdir
restricts it to INT_MAX when encoding the reply.  This can result in log
messages like "kernel: RPC request reserved 32896 but used 1049444".

Restrict rd_dircount similarly (no reason it should be larger than
svc_max_payload).
Signed-off-by: Scott Mayhew <smayhew@redhat.com>
Cc: stable@vger.kernel.org
Signed-off-by: J. Bruce Fields <bfields@redhat.com>

Pll KVM fixes from Radim Krčmář:
 "ARM:
   - Fix proxying of GICv2 CPU interface accesses
   - Fix crash when switching to BE
   - Track source vcpu git GICv2 SGIs
   - Fix an outdated bit of documentation

  x86:
   - Speed up injection of expired timers (for stable)"

* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
  KVM: x86: remove APIC Timer periodic/oneshot spikes
  arm64: vgic-v2: Fix proxying of cpuif access
  KVM: arm/arm64: vgic_init: Cleanup reference to process_maintenance
  KVM: arm64: Fix order of vcpu_write_sys_reg() arguments
  KVM: arm/arm64: vgic: Fix source vcpu issues for GICv2 SGI

Pull iommu fixes from Joerg Roedel:

 - fix a compile warning in the AMD IOMMU driver with irq remapping
   disabled

 - fix for VT-d interrupt remapping and invalidation size (caused a
   BUG_ON when trying to invalidate more than 4GB)

 - build fix and a regression fix for broken graphics with old DTS for
   the rockchip iommu driver

 - a revert in the PCI window reservation code which fixes a regression
   with VFIO.

* tag 'iommu-fixes-v4.17-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/joro/iommu:
  iommu: rockchip: fix building without CONFIG_OF
  iommu/vt-d: Use WARN_ON_ONCE instead of BUG_ON in qi_flush_dev_iotlb()
  iommu/vt-d: fix shift-out-of-bounds in bug checking
  iommu/dma: Move PCI window region reservation back into dma specific path.
  iommu/rockchip: Make clock handling optional
  iommu/amd: Hide unused iommu_table_lock
  iommu/vt-d: Fix usage of force parameter in intel_ir_reconfigure_irte()

Pull x86 fix from Thomas Gleixner:
 "Unbreak the CPUID CPUID_8000_0008_EBX reload which got dropped when
  the evaluation of physical and virtual bits which uses the same CPUID
  leaf was moved out of get_cpu_cap()"

* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/cpu: Restore CPUID_8000_0008_EBX reload

Pull clocksource fixes from Thomas Gleixner:
 "The recent addition of the early TSC clocksource breaks on machines
  which have an unstable TSC because in case that TSC is disabled, then
  the clocksource selection logic falls back to the early TSC which is
  obviously bogus.

  That also unearthed a few robustness issues in the clocksource
  derating code which are addressed as well"

* 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  clocksource: Rework stale comment
  clocksource: Consistent de-rate when marking unstable
  x86/tsc: Fix mark_tsc_unstable()
  clocksource: Initialize cs->wd_list
  clocksource: Allow clocksource_mark_unstable() on unregistered clocksources
  x86/tsc: Always unregister clocksource_tsc_early

Pull irq fix from Thomas Gleixner:
 "A single fix to prevent false positives in the spurious interrupt
  detector when more than a single demultiplex register is evaluated in
  the Qualcom irq combiner driver"

* 'irq-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  irqchip/qcom: Fix check for spurious interrupts

Pull x86 platform driver fixes from Darren Hart:

 - We missed a case in the Dell config dependencies resulting in a
   possible bad configuration, resolve it by giving up on trying to keep
   DELL_LAPTOP visible in the menu and make it depend on DELL_SMBIOS.

 - Fix a null pointer dereference at module unload for the asus-wireless
   driver.

* tag 'platform-drivers-x86-v4.17-2' of git://git.infradead.org/linux-platform-drivers-x86:
  platform/x86: Kconfig: Fix dell-laptop dependency chain.
  platform/x86: asus-wireless: Fix NULL pointer dereference

Pull USB fixes from Greg KH:
 "Here are some USB driver fixes for 4.17-rc4.

  The majority of them are some USB gadget fixes that missed my last
  pull request. The "largest" patch in here is a fix for the old visor
  driver that syzbot found 6 months or so ago and I finally remembered
  to fix it.

  All of these have been in linux-next with no reported issues"

* tag 'usb-4.17-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb:
  Revert "usb: host: ehci: Use dma_pool_zalloc()"
  usb: typec: tps6598x: handle block reads separately with plain-I2C adapters
  usb: typec: tcpm: Release the role mux when exiting
  USB: Accept bulk endpoints with 1024-byte maxpacket
  xhci: Fix use-after-free in xhci_free_virt_device
  USB: serial: visor: handle potential invalid device configuration
  USB: serial: option: adding support for ublox R410M
  usb: musb: trace: fix NULL pointer dereference in musb_g_tx()
  usb: musb: host: fix potential NULL pointer dereference
  usb: gadget: composite Allow for larger configuration descriptors
  usb: dwc3: gadget: Fix list_del corruption in dwc3_ep_dequeue
  usb: dwc3: gadget: dwc3_gadget_del_and_unmap_request() can be static
  usb: dwc2: pci: Fix error return code in dwc2_pci_probe()
  usb: dwc2: WA for Full speed ISOC IN in DDMA mode.
  usb: dwc2: dwc2_vbus_supply_init: fix error check
  usb: gadget: f_phonet: fix pn_net_xmit()'s return type

Since the commit "8003c9ae: add APIC Timer periodic/oneshot mode VMX
preemption timer support", a Windows 10 guest has some erratic timer
spikes.

Here the results on a 150000 times 1ms timer without any load:
	  Before 8003c9ae | After 8003c9ae
Max           1834us          |  86000us
Mean          1100us          |   1021us
Deviation       59us          |    149us
Here the results on a 150000 times 1ms timer with a cpu-z stress test:
	  Before 8003c9ae | After 8003c9ae
Max          32000us          | 140000us
Mean          1006us          |   1997us
Deviation      140us          |  11095us

The root cause of the problem is starting hrtimer with an expiry time
already in the past can take more than 20 milliseconds to trigger the
timer function.  It can be solved by forward such past timers
immediately, rather than submitting them to hrtimer_start().
In case the timer is periodic, update the target expiration and call
hrtimer_start with it.

v2: Check if the tsc deadline is already expired. Thank you Mika.
v3: Execute the past timers immediately rather than submitting them to
hrtimer_start().
v4: Rearm the periodic timer with advance_periodic_target_expiration() a
simpler version of set_target_expiration(). Thank you Paolo.

Cc: Mika Penttilä <mika.penttila@nextfour.com>
Cc: Wanpeng Li <kernellwp@gmail.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: stable@vger.kernel.org
Signed-off-by: Anthoine Bourgeois <anthoine.bourgeois@blade-group.com>
8003c9ae ("KVM: LAPIC: add APIC Timer periodic/oneshot mode VMX preemption timer support")
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>

KVM/arm fixes for 4.17, take #2

- Fix proxying of GICv2 CPU interface accesses
- Fix crash when switching to BE
- Track source vcpu git GICv2 SGIs
- Fix an outdated bit of documentation

Merge tag 'kbuild-fixes-v4.17' of git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild

Pull Kbuild fixes from Masahiro Yamada:

 - remove state comment in modpost

 - extend MAINTAINERS entry to cover modpost and more makefiles

 - fix missed building of SANCOV gcc-plugin

 - replace left-over 'bison' with $(YACC)

 - display short log when generating parer of genksyms

* tag 'kbuild-fixes-v4.17' of git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild:
  genksyms: fix typo in parse.tab.{c,h} generation rules
  kbuild: replace hardcoded bison in cmd_bison_h with $(YACC)
  gcc-plugins: fix build condition of SANCOV plugin
  MAINTAINERS: Update Kbuild entry with a few paths
  modpost: delete stale comment

Pull clk fixes froom Stephen Boyd:
 "A handful of fixes for the stm32mp1 clk driver came in during the
  merge window for the driver that got merged in the merge window.

  Plus a warning fix for unused PM ops and a couple fixes for the meson
  clk driver clk names that went unnoticed with the regmap rework.

  There's also another fix in here for the mux rounding flag which
  wasn't doing what it said it did, but now it does"

* tag 'clk-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/clk/linux:
  clk: meson: meson8b: fix meson8b_cpu_clk parent clock name
  clk: meson: meson8b: fix meson8b_fclk_div3_div clock name
  clk: meson: drop meson_aoclk_gate_regmap_ops
  clk: meson: honor CLK_MUX_ROUND_CLOSEST in clk_regmap
  clk: honor CLK_MUX_ROUND_CLOSEST in generic clk mux
  clk: cs2000: mark resume function as __maybe_unused
  clk: stm32mp1: remove ck_apb_dbg clock
  clk: stm32mp1: set stgen_k clock as critical
  clk: stm32mp1: add missing tzc2 clock
  clk: stm32mp1: fix SAI3 & SAI4 clocks
  clk: stm32mp1: remove unused dfsdm_src[] const
  clk: stm32mp1: add missing static

Pull remoteproc and rpmsg fixes from Bjorn Andersson:

 - fix screw-up when reversing boolean for rproc_stop()

 - add missing OF node refcounting dereferences

 - add missing MODULE_ALIAS in rpmsg_char

* tag 'rproc-v4.17-1' of git://github.com/andersson/remoteproc:
  rpmsg: added MODULE_ALIAS for rpmsg_char
  remoteproc: qcom: Fix potential device node leaks
  remoteproc: fix crashed parameter logic on stop call

Pull drm fixes from Dave Airlie:
 "vmwgfx, i915, vc4, vga dac fixes.

  This seems eerily quiet, so I expect it will explode next week or
  something.

  One i915 model firmware, two vmwgfx fixes, one vc4 fix and one bridge
  leak fix"

* tag 'drm-fixes-for-v4.17-rc4' of git://people.freedesktop.org/~airlied/linux:
  drm/bridge: vga-dac: Fix edid memory leak
  drm/vc4: Make sure vc4_bo_{inc,dec}_usecnt() calls are balanced
  drm/i915/glk: Add MODULE_FIRMWARE for Geminilake
  drm/vmwgfx: Fix a buffer object leak
  drm/vmwgfx: Clean up fbdev modeset locking

Pull tracing fixes from Steven Rostedt:
 "Some of the files in the tracing directory show file mode 0444 when
  they are writable by root. To fix the confusion, they should be 0644.
  Note, either case root can still write to them.

  Zhengyuan asked why I never applied that patch (the first one is from
  2014!). I simply forgot about it. /me lowers head in shame"

* tag 'trace-v4.17-rc1-3' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
  tracing: Fix the file mode of stack tracer
  ftrace: Have set_graph_* files have normal file modes

Pull rdma fixes from Doug Ledford:
 "This is our first pull request of the rc cycle. It's not that it's
  been overly quiet, we were just waiting on a few things before sending
  this off.

  For instance, the 6 patch series from Intel for the hfi1 driver had
  actually been pulled in on Tuesday for a Wednesday pull request, only
  to have Jason notice something I missed, so we held off for some
  testing, and then on Thursday had to respin the series because the
  very first patch needed a minor fix (unnecessary cast is all).

  There is a sizable hns patch series in here, as well as a reasonably
  largish hfi1 patch series, then all of the lines of uapi updates are
  just the change to the new official Linux-OpenIB SPDX tag (a bunch of
  our files had what amounts to a BSD-2-Clause + MIT Warranty statement
  as their license as a result of the initial code submission years ago,
  and the SPDX folks decided it was unique enough to warrant a unique
  tag), then the typical mlx4 and mlx5 updates, and finally some cxgb4
  and core/cache/cma updates to round out the bunch.

  None of it was overly large by itself, but in the 2 1/2 weeks we've
  been collecting patches, it has added up :-/.

  As best I can tell, it's been through 0day (I got a notice about my
  last for-next push, but not for my for-rc push, but Jason seems to
  think that failure messages are prioritized and success messages not
  so much). It's also been through linux-next. And yes, we did notice in
  the context portion of the CMA query gid fix patch that there is a
  dubious BUG_ON() in the code, and have plans to audit our BUG_ON usage
  and remove it anywhere we can.

  Summary:

   - Various build fixes (USER_ACCESS=m and ADDR_TRANS turned off)

   - SPDX license tag cleanups (new tag Linux-OpenIB)

   - RoCE GID fixes related to default GIDs

   - Various fixes to: cxgb4, uverbs, cma, iwpm, rxe, hns (big batch),
     mlx4, mlx5, and hfi1 (medium batch)"

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma: (52 commits)
  RDMA/cma: Do not query GID during QP state transition to RTR
  IB/mlx4: Fix integer overflow when calculating optimal MTT size
  IB/hfi1: Fix memory leak in exception path in get_irq_affinity()
  IB/{hfi1, rdmavt}: Fix memory leak in hfi1_alloc_devdata() upon failure
  IB/hfi1: Fix NULL pointer dereference when invalid num_vls is used
  IB/hfi1: Fix loss of BECN with AHG
  IB/hfi1 Use correct type for num_user_context
  IB/hfi1: Fix handling of FECN marked multicast packet
  IB/core: Make ib_mad_client_id atomic
  iw_cxgb4: Atomically flush per QP HW CQEs
  IB/uverbs: Fix kernel crash during MR deregistration flow
  IB/uverbs: Prevent reregistration of DM_MR to regular MR
  RDMA/mlx4: Add missed RSS hash inner header flag
  RDMA/hns: Fix a couple misspellings
  RDMA/hns: Submit bad wr
  RDMA/hns: Update assignment method for owner field of send wqe
  RDMA/hns: Adjust the order of cleanup hem table
  RDMA/hns: Only assign dqpn if IB_QP_PATH_DEST_QPN bit is set
  RDMA/hns: Remove some unnecessary attr_mask judgement
  RDMA/hns: Only assign mtu if IB_QP_PATH_MTU bit is set
  ...

Pull block fixes from Jens Axboe:
 "A collection of fixes that should to into this release. This contains:

   - Set of bcache fixes from Coly, fixing regression in patches that
     went into this series.

   - Set of NVMe fixes by way of Keith.

   - Set of bdi related fixes, one from Jan and two from Tetsuo Handa,
     fixing various issues around device addition/removal.

   - Two block inflight fixes from Omar, fixing issues around the
     transition to using tags for blk-mq inflight accounting that we
     did a few releases ago"

* tag 'for-linus-20180504' of git://git.kernel.dk/linux-block:
  bdi: Fix oops in wb_workfn()
  nvmet: switch loopback target state to connecting when resetting
  nvme/multipath: Fix multipath disabled naming collisions
  nvme/multipath: Disable runtime writable enabling parameter
  nvme: Set integrity flag for user passthrough commands
  nvme: fix potential memory leak in option parsing
  bdi: Fix use after free bug in debugfs_remove()
  bdi: wake up concurrent wb_shutdown() callers.
  bcache: use pr_info() to inform duplicated CACHE_SET_IO_DISABLE set
  bcache: set dc->io_disable to true in conditional_stop_bcache_device()
  bcache: add wait_for_kthread_stop() in bch_allocator_thread()
  bcache: count backing device I/O error for writeback I/O
  bcache: set CACHE_SET_IO_DISABLE in bch_cached_dev_error()
  bcache: store disk name in struct cache and struct cached_dev
  blk-mq: fix sysfs inflight counter
  blk-mq: count allocated but not started requests in iostats inflight

Pull xfs fixes from Darrick Wong:
 "I've got one more bug fix for xfs for 4.17-rc4, which caps the amount
  of data we try to handle in one dedupe request so that userspace can't
  livelock the kernel.

  This series has been run through a full xfstests run during the week
  and through a quick xfstests run against this morning's master, with
  no ajor failures reported.

  Summary:

  - Cap the maximum length of a deduplication request at MAX_RW_COUNT/2
    to avoid kernel livelock due to excessively large IO requests"

* tag 'xfs-4.17-fixes-2' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux:
  xfs: cap the length of deduplication requests

Pull btrfs fixes from David Sterba:
 "Two regression fixes and one fix for stable"

* tag 'for-4.17-rc3-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux:
  Btrfs: send, fix missing truncate for inode with prealloc extent past eof
  btrfs: Take trans lock before access running trans in check_delayed_ref
  btrfs: Fix wrong first_key parameter in replace_path