Peng Tao points out that the call to pnfs_mark_matching_lsegs_return()
could race with pnfs_put_lseg(), in which case the layout segment is
cleared, but no layoutreturn will be sent.
Fix is to replace the call to pnfs_mark_matching_lsegs_invalid().
Reported-by: Peng Tao <tao.peng@primarydata.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

Fix a bug whereby if all the layout segments could be immediately freed,
the call to pnfs_error_mark_layout_for_return() would never result in
a layoutreturn.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

If pnfs_mark_matching_lsegs_return() needs to mark a layout segment for
return, then it must also set the return iomode.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

A stateid is a structure, pass it as a pointer.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

If two processes share the same credentials and NFSv4 open stateid, then
allow them both to dirty the same page, even if their nfs_open_context
differs.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

If the layout segment is invalid, then we should not be adding more
write requests to the commit list. Instead, those writes should be
replayed after requesting a new layout.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

Allow synchronous RPC calls to wait for pending RPC calls to finish,
but also allow asynchronous ones to just fire off another commit.

With this patch, the xfstests generic/074 test completes in 226s
instead of 242s
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

The flexfiles layout in particular, seems to want to poke around in the
O_DIRECT flags when retransmitting.
This patch sets up an interface to allow it to call back into O_DIRECT
to handle retransmission correctly. It also fixes a potential bug whereby
we could change the behaviour of O_DIRECT if an error is already pending.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

The stateid is extremely valuable when debugging.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

If the client is promising to return the layout ASAP, then there is no
need to return DELAY and have the server retry. Instead default to the
normal procedure described in RFC5661.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

The RFC requires us to check if the server is recalling a stateid that we
haven't yet received. If so, tell it to wait.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

If the client needs to delay the layout callback, then speed up the recall
process by marking the remaining layout segments to be actively returned
by the client.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

This ensures that we don't reuse the stateid if a layout return or
implied layout return means that we've returned all layout segments
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

If we're unable to perform the layoutget due to an invalid open stateid
or a bulk recall, ensure that we return the error so that the caller
can decide on an appropriate action.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

Instead of displaying a layout segment pointer in these tracepoints,
let's use the layout stateid, now that Olga gave us a set of tools for
displaying them.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

pnfs_update_layout is really the "nexus" of layout handling. If it
returns NULL then we end up going through the MDS. This patch adds
some tracepoints to that function that allow us to determine the
cause when we end up going through the MDS unexpectedly.
Signed-off-by: Jeff Layton <jeff.layton@primarydata.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

Signed-off-by: Olga Kornievskaia <kolga@netapp.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

Operations to which stateid information is added:
close, delegreturn, open, read, setattr, layoutget, layoutcommit, test_stateid,
write, lock, locku, lockt

Format is "stateid=<seqid>:<crc32 hash stateid.other>", also "openstateid=",
"layoutstateid=", and "lockstateid=" for open_file, layoutget, set_lock
tracepoints.

New function is added to internal.h, nfs_stateid_hash(), to compute the hash

trace_nfs4_setattr() is moved from nfs4_do_setattr() to _nfs4_do_setattr()
to get access to stateid.

trace_nfs4_setattr and trace_nfs4_delegreturn are changed from INODE_EVENT
to new event type, INODE_STATEID_EVENT which is same as INODE_EVENT but adds
stateid information

for locking tracepoints, moved trace_nfs4_set_lock() into _nfs4_do_setlk()
to get access to stateid information, and removed trace_nfs4_lock_reclaim(),
trace_nfs4_lock_expired() as they call into _nfs4_do_setlk() and both were
previously same LOCK_EVENT type.
Signed-off-by: Olga Kornievskaia <kolga@netapp.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

Pull MIPS fixes from Ralf Baechle:

 - Fix bitrot in __get_user_unaligned()
 - EVA userspace accessor bug fixes.
 - Fix for build issues with certain toolchains.
 - Fix build error for VDSO with particular toolchain versions.
 - Fix build error due to a variable that should have been removed by an
   earlier patch

* 'upstream' of git://git.linux-mips.org/pub/scm/ralf/upstream-linus:
  MIPS: Fix bitrot in __get_user_unaligned()
  MIPS: Fix build error due to unused variables.
  MIPS: VDSO: Fix build error
  MIPS: CPS: drop .set mips64r2 directives
  MIPS: uaccess: Take EVA into account in [__]clear_user
  MIPS: uaccess: Take EVA into account in __copy_from_user()
  MIPS: uaccess: Fix strlen_user with EVA

Pull ARM SoC fixes from Olof Johansson:
 "A smallish set of fixes that we've been sitting on for a while now,
  flushing the queue here so they go in.  Summary:

  A handful of fixes for OMAP, i.MX, Allwinner and Tegra:

   - A clock rate and a PHY setup fix for i.MX6Q/DL
   - A couple of fixes for the reduced serial bus (sunxi-rsb) on
     Allwinner
   - UART wakeirq fix for an OMAP4 board, timer config fixes for AM43XX.
   - Suspend fix for Tegra124 Chromebooks
   - Fix for missing implicit include that's different between
     ARM/ARM64"

* tag 'armsoc-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc:
  ARM: tegra: Fix suspend hang on Tegra124 Chromebooks
  bus: sunxi-rsb: Fix peripheral IC mapping runtime address
  bus: sunxi-rsb: Fix primary PMIC mapping hardware address
  ARM: dts: Fix UART wakeirq for omap4 duovero parlor
  ARM: OMAP2+: AM43xx: select ARM TWD timer
  ARM: OMAP2+: am43xx: enable GENERIC_CLOCKEVENTS_BROADCAST
  fsl-ifc: add missing include on ARM64
  ARM: dts: imx6: Fix Ethernet PHY mode on Ventana boards
  ARM: dts: imx: Fix the assigned-clock mismatch issue on imx6q/dl
  bus: sunxi-rsb: unlock on error in sunxi_rsb_read()
  ARM: dts: sunxi: sun6i-a31s-primo81.dts: add touchscreen axis swapping property

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Pull power management and ACPI fixes from Rafael Wysocki:
 "These fix an ACPI processor driver regression introduced during the
  4.3 cycle and a mistake in the recently added SCPI support in the
  arm_big_little cpufreq driver.

  Specifics:

   - Fix a thermal management issue introduced by an ACPI processor
     driver change made during the 4.3 development cycle that failed to
     return 0 from a function on success which triggered an error
     cleanup path every time it had been called that deleted useful data
     structures created previously (Srinivas Pandruvada).

   - Fix a variable data type issue in the arm_big_little cpufreq
     driver's SCPI support code added recently that prevents error
     handling in there from working correctly (Dan Carpenter)"

* tag 'pm+acpi-4.4-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  cpufreq: scpi-cpufreq: signedness bug in scpi_get_dvfs_info()
  ACPI / processor: Fix thermal cooling device regression

Pull md bugfix from Neil Brown:
 "One more md fix for 4.4-rc

  Fix a regression which causes reshape to not start properly sometimes"

* tag 'md/4.4-rc6-fix' of git://neil.brown.name/md:
  md: remove check for MD_RECOVERY_NEEDED in action_store.

Pull UBI bug fixes from Richard Weinberger:
 "This contains four bug fixes for UBI"

* tag 'upstream-4.4-rc7' of git://git.infradead.org/linux-ubifs:
  mtd: ubi: don't leak e if schedule_erase() fails
  mtd: ubi: fixup error correction in do_sync_erase()
  UBI: fix use of "VID" vs. "EC" in header self-check
  UBI: fix return error code

Pull ftrace/recordmcount fix from Steven Rostedt:
 "Russell King was reporting lots of warnings when he compiled his
  kernel with ftrace enabled.  With some investigation it was discovered
  that it was his compile setup.  He was using ccache with hard links,
  which allowed recordmcount to process the same .o twice.  When this
  happens, recordmcount will detect that it was already done and give a
  warning about it.

  Russell fixed this by having recordmcount detect that the object file
  has more than one hard link, and if it does, it unlinks the object
  file after it maps it and processes then.  This appears to fix the
  issue.

  As you did not like the fact that recordmcount modified the file in
  place and thought that it should do the modifications in memory and
  then write it out to disk and move it over the old file to prevent
  other more subtle issues like the one above, a second patch is added
  on top of Russell's to do just that.  Luckily the original code had
  write and lseek wrappers that I was able to modify to not do inplace
  writes, but simply keep track of the changes made in memory.  When a
  write is made, a "update" flag is set, and at the end of processing,
  if the update is set, then it writes the file with changes out to a
  new file, and then renames it over the original one.

  The file descriptor is still passed to the write and lseek wrappers
  because removing that would cause the change to be more intrusive.
  That can be removed in a follow up cleanup patch that can wait till
  the next merge window"

* tag 'trace-v4.4-rc4-2' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
  ftrace/scripts: Have recordmcount copy the object file
  scripts: recordmcount: break hardlinks

Pull ARC fixes from Vineet Gupta:
 "Sorry for this late pull request, but these are all important fixes
  for code introduced/updated in this release which we will otherwise
  end up back porting.

   - Unwinder rework (A revert followed by better fix)
   - Build errors: MMUv2, modules with -Os
   - highmem section mismatch build splat"

* tag 'arc-4.4-rc7-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/vgupta/arc:
  ARC: dw2 unwind: Catch Dwarf SNAFUs early
  ARC: dw2 unwind: Don't bail for CIE.version != 1
  Revert "ARC: dw2 unwind: Ignore CIE version !=1 gracefully instead of bailing"
  ARC: Fix linking errors with CONFIG_MODULE + CONFIG_CC_OPTIMIZE_FOR_SIZE
  ARC: mm: fix building for MMU v2
  ARC: mm: HIGHMEM: Fix section mismatch splat

* acpi-processor:
  ACPI / processor: Fix thermal cooling device regression

* pm-cpufreq:
  cpufreq: scpi-cpufreq: signedness bug in scpi_get_dvfs_info()

Pull parisc system call restart fix from Helge Deller:
 "The architectural design of parisc always uses two instructions to
  call kernel syscalls (delayed branch feature).  This means that the
  instruction following the branch (located in the delay slot of the
  branch instruction) is executed before control passes to the branch
  destination.

  Depending on which assembler instruction and how it is used in
  usersapce in the delay slot, this sometimes made restarted syscalls
  like futex() and poll() failing with -ENOSYS"

* 'parisc-4.4-4' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/parisc-linux:
  parisc: Fix syscall restarts

Pull sparc fixes from David Miller:

 1) Finally make perf stack backtraces stable on sparc, several problems
    (mostly due to the context in which the user copies from the stack
    are done) contributed to this.

    From Rob Gardner.

 2) Export ADI capability if the cpu supports it.

 3) Hook up userfaultfd system call.

 4) When faults happen during user copies we really have to clean up and
    restore the FPU state fully.  Also from Rob Gardner

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc:
  tty/serial: Skip 'NULL' char after console break when sysrq enabled
  sparc64: fix FP corruption in user copy functions
  sparc64: Perf should save/restore fault info
  sparc64: Ensure perf can access user stacks
  sparc64: Don't set %pil in rtrap_nmi too early
  sparc64: Add ADI capability to cpu capabilities
  tty: serial: constify sunhv_ops structs
  sparc: Hook up userfaultfd system call

When sysrq is triggered from console, serial driver for SUN hypervisor
console receives a console break and enables the sysrq. It expects a valid
sysrq char following with break. Meanwhile if driver receives 'NULL'
ASCII char then it disables sysrq and sysrq handler will never be invoked.

This fix skips calling uart sysrq handler when 'NULL' is received while
sysrq is enabled.
Signed-off-by: Vijay Kumar <vijay.ac.kumar@oracle.com>
Acked-by: Karl Volz <karl.volz@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Short story: Exception handlers used by some copy_to_user() and
copy_from_user() functions do not diligently clean up floating point
register usage, and this can result in a user process seeing invalid
values in floating point registers. This sometimes makes the process
fail.

Long story: Several cpu-specific (NG4, NG2, U1, U3) memcpy functions
use floating point registers and VIS alignaddr/faligndata to
accelerate data copying when source and dest addresses don't align
well. Linux uses a lazy scheme for saving floating point registers; It
is not done upon entering the kernel since it's a very expensive
operation. Rather, it is done only when needed. If the kernel ends up
not using FP regs during the course of some trap or system call, then
it can return to user space without saving or restoring them.

The various memcpy functions begin their FP code with VISEntry (or a
variation thereof), which saves the FP regs. They conclude their FP
code with VISExit (or a variation) which essentially marks the FP regs
"clean", ie, they contain no unsaved values. fprs.FPRS_FEF is turned
off so that a lazy restore will be triggered when/if the user process
accesses floating point regs again.

The bug is that the user copy variants of memcpy, copy_from_user() and
copy_to_user(), employ an exception handling mechanism to detect faults
when accessing user space addresses, and when this handler is invoked,
an immediate return from the function is forced, and VISExit is not
executed, thus leaving the fprs register in an indeterminate state,
but often with fprs.FPRS_FEF set and one or more dirty bits. This
results in a return to user space with invalid values in the FP regs,
and since fprs.FPRS_FEF is on, no lazy restore occurs.

This bug affects copy_to_user() and copy_from_user() for NG4, NG2,
U3, and U1. All are fixed by using a new exception handler for those
loads and stores that are done during the time between VISEnter and
VISExit.

n.b. In NG4memcpy, the problematic code can be triggered by a copy
size greater than 128 bytes and an unaligned source address.  This bug
is known to be the cause of random user process memory corruptions
while perf is running with the callgraph option (ie, perf record -g).
This occurs because perf uses copy_from_user() to read user stacks,
and may fault when it follows a stack frame pointer off to an
invalid page. Validation checks on the stack address just obscure
the underlying problem.
Signed-off-by: Rob Gardner <rob.gardner@oracle.com>
Signed-off-by: Dave Aldridge <david.j.aldridge@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

There have been several reports of random processes being killed with
a bus error or segfault during userspace stack walking in perf.  One
of the root causes of this problem is an asynchronous modification to
thread_info fault_address and fault_code, which stems from a perf
counter interrupt arriving during kernel processing of a "benign"
fault, such as a TSB miss. Since perf_callchain_user() invokes
copy_from_user() to read user stacks, a fault is not only possible,
but probable. Validity checks on the stack address merely cover up the
problem and reduce its frequency.

The solution here is to save and restore fault_address and fault_code
in perf_callchain_user() so that the benign fault handler is not
disturbed by a perf interrupt.
Signed-off-by: Rob Gardner <rob.gardner@oracle.com>
Signed-off-by: Dave Aldridge <david.j.aldridge@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

When an interrupt (such as a perf counter interrupt) is delivered
while executing in user space, the trap entry code puts ASI_AIUS in
%asi so that copy_from_user() and copy_to_user() will access the
correct memory. But if a perf counter interrupt is delivered while the
cpu is already executing in kernel space, then the trap entry code
will put ASI_P in %asi, and this will prevent copy_from_user() from
reading any useful stack data in either of the perf_callchain_user_X
functions, and thus no user callgraph data will be collected for this
sample period. An additional problem is that a fault is guaranteed
to occur, and though it will be silently covered up, it wastes time
and could perturb state.

In perf_callchain_user(), we ensure that %asi contains ASI_AIUS
because we know for a fact that the subsequent calls to
copy_from_user() are intended to read the user's stack.

[ Use get_fs()/set_fs() -DaveM ]
Signed-off-by: Rob Gardner <rob.gardner@oracle.com>
Signed-off-by: Dave Aldridge <david.j.aldridge@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Commit 28a1f533 delays setting %pil to avoid potential
hardirq stack overflow in the common rtrap_irq path.
Setting %pil also needs to be delayed in the rtrap_nmi
path for the same reason.
Signed-off-by: Rob Gardner <rob.gardner@oracle.com>
Signed-off-by: Dave Aldridge <david.j.aldridge@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add ADI (Application Data Integrity) capability to cpu capabilities list.
ADI capability allows virtual addresses to be encoded with a tag in
bits 63-60. This tag serves as an access control key for the regions
of virtual address with ADI enabled and a key set on them. Hypervisor
encodes this capability as "adp" in "hwcap-list" property in machine
description.
Signed-off-by: Khalid Aziz <khalid.aziz@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Constifies sunhv_ops structures in tty's serial
driver since they are not modified after their
initialization.

Detected and found using Coccinelle.
Suggested-by: Julia Lawall <Julia.Lawall@lip6.fr>
Signed-off-by: Aya Mahfouz <mahfouz.saif.elyazal@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The "domain" variable needs to be signed for the error handling to work.

Fixes: 8def3103 (cpufreq: arm_big_little: add SCPI interface driver)
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Acked-by: Sudeep Holla <sudeep.holla@arm.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>