linux-2.6-microblaze.git
7 years agoRevert "Bluetooth: Add option for disabling legacy ioctl interfaces"
Linus Torvalds [Thu, 28 Sep 2017 20:20:32 +0000 (13:20 -0700)]
Revert "Bluetooth: Add option for disabling legacy ioctl interfaces"

This reverts commit dbbccdc4ced015cdd4051299bd87fbe0254ad351.

It turns out that the "legacy" users aren't so legacy at all, and that
turning off the legacy ioctl will break the current Qt bluetooth stack
for bluetooth LE devices that were released just a couple of months ago.

So it's simply not true that this was a legacy interface that hasn't
been needed and is only limited to old legacy BT devices.  Because I
actually read Kconfig help messages, and actively try to turn off
features that I don't need, I turned the option off.

Then I spent _way_ too much time debugging BLE issues until I realized
that it wasn't the Qt and subsurface development that had broken one of
my dive computer BLE downloads, but simply my broken kernel config.

Maybe in a decade it will be true that this is a legacy interface.  And
maybe with a better help-text and correct dependencies, this kind of
legacy removal might be acceptable.  But as things are right now both
the commit message and the Kconfig help text were misleading, and the
Kconfig option had the wrong dependenencies.

There's no reason to keep that broken Kconfig option in the tree.

Cc: Marcel Holtmann <marcel@holtmann.org>
Cc: Johan Hedberg <johan.hedberg@intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agoMerge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dledford/rdma
Linus Torvalds [Thu, 28 Sep 2017 19:12:51 +0000 (12:12 -0700)]
Merge tag 'for-linus' of git://git./linux/kernel/git/dledford/rdma

Pull rdma fixes from Doug Ledford:
 "Second -rc update for 4.14.

  Both Mellanox and Intel had a series of -rc fixes that landed this
  week. The Mellanox bunch is spread throughout the stack and not just
  in their driver, where as the Intel bunch was mostly in the hfi1
  driver. And, several of the fixes in the hfi1 driver were more than
  just simple 5 line fixes. As a result, the hfi1 driver fixes has a
  sizable LOC count.

  Everything else is as one would expect in an RC cycle in terms of LOC
  count. One item that might jump out and make you think "That's not an
  rc item" is the fix that corrects a typo. But, that change fixes a
  typo in a user visible API that was just added in this merge window,
  so if we fix it now, we can fix it. If we don't, the typo is in the
  API forever. Another that might not appear to be a fix at first glance
  is the Simplify mlx5_ib_cont_pages patch, but the simplification
  allows them to fix a bug in the existing function whenever the length
  of an SGE exceeded page size. We also had to revert one patch from the
  merge window that was wrong.

  Summary:

   - a few core fixes
   - a few ipoib fixes
   - a few mlx5 fixes
   - a 7-patch hfi1 related series"

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dledford/rdma:
  IB/hfi1: Unsuccessful PCIe caps tuning should not fail driver load
  IB/hfi1: On error, fix use after free during user context setup
  Revert "IB/ipoib: Update broadcast object if PKey value was changed in index 0"
  IB/hfi1: Return correct value in general interrupt handler
  IB/hfi1: Check eeprom config partition validity
  IB/hfi1: Only reset QSFP after link up and turn off AOC TX
  IB/hfi1: Turn off AOC TX after offline substates
  IB/mlx5: Fix NULL deference on mlx5_ib_update_xlt failure
  IB/mlx5: Simplify mlx5_ib_cont_pages
  IB/ipoib: Fix inconsistency with free_netdev and free_rdma_netdev
  IB/ipoib: Fix sysfs Pkey create<->remove possible deadlock
  IB: Correct MR length field to be 64-bit
  IB/core: Fix qp_sec use after free access
  IB/core: Fix typo in the name of the tag-matching cap struct

7 years agoMerge tag 'seccomp-v4.14-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/kees...
Linus Torvalds [Thu, 28 Sep 2017 18:20:52 +0000 (11:20 -0700)]
Merge tag 'seccomp-v4.14-rc3' of git://git./linux/kernel/git/kees/linux

Pull seccomp fix from Kees Cook:
 "Fix refcounting bug in CRIU interface, noticed by Chris Salls (Oleg &
  Tycho)"

* tag 'seccomp-v4.14-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/kees/linux:
  seccomp: fix the usage of get/put_seccomp_filter() in seccomp_get_filter()

7 years agoseccomp: fix the usage of get/put_seccomp_filter() in seccomp_get_filter()
Oleg Nesterov [Wed, 27 Sep 2017 15:25:30 +0000 (09:25 -0600)]
seccomp: fix the usage of get/put_seccomp_filter() in seccomp_get_filter()

As Chris explains, get_seccomp_filter() and put_seccomp_filter() can end
up using different filters. Once we drop ->siglock it is possible for
task->seccomp.filter to have been replaced by SECCOMP_FILTER_FLAG_TSYNC.

Fixes: f8e529ed941b ("seccomp, ptrace: add support for dumping seccomp filters")
Reported-by: Chris Salls <chrissalls5@gmail.com>
Cc: stable@vger.kernel.org # needs s/refcount_/atomic_/ for v4.12 and earlier
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
[tycho: add __get_seccomp_filter vs. open coding refcount_inc()]
Signed-off-by: Tycho Andersen <tycho@docker.com>
[kees: tweak commit log]
Signed-off-by: Kees Cook <keescook@chromium.org>
7 years agoMerge branch 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs
Linus Torvalds [Wed, 27 Sep 2017 19:22:12 +0000 (12:22 -0700)]
Merge branch 'for_linus' of git://git./linux/kernel/git/jack/linux-fs

Pull quota and isofs fixes from Jan Kara:
 "Two quota fixes (fallout of the quota locking changes) and an isofs
  build fix"

* 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs:
  quota: Fix quota corruption with generic/232 test
  isofs: fix build regression
  quota: add missing lock into __dquot_transfer()

7 years agoMerge tag 'linux-kselftest-4.14-rc3-fixes' of git://git.kernel.org/pub/scm/linux...
Linus Torvalds [Wed, 27 Sep 2017 17:51:08 +0000 (10:51 -0700)]
Merge tag 'linux-kselftest-4.14-rc3-fixes' of git://git./linux/kernel/git/shuah/linux-kselftest

Pull kselftest fixes from Shuah Khan:
 "This update consists of:

   - fixes to several existing tests

   - a test for regression introduced by b9470c27607b ("inet: kill
     smallest_size and smallest_port")

   - seccomp support for glibc 2.26 siginfo_t.h

   - fixes to kselftest framework and tests to run make O=dir use-case

   - fixes to silence unnecessary test output to de-clutter test results"

* tag 'linux-kselftest-4.14-rc3-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/shuah/linux-kselftest: (28 commits)
  selftests: timers: set-timer-lat: Fix hang when testing unsupported alarms
  selftests: timers: set-timer-lat: fix hang when std out/err are redirected
  selftests/memfd: correct run_tests.sh permission
  selftests/seccomp: Support glibc 2.26 siginfo_t.h
  selftests: futex: Makefile: fix for loops in targets to run silently
  selftests: Makefile: fix for loops in targets to run silently
  selftests: mqueue: Use full path to run tests from Makefile
  selftests: futex: copy sub-dir test scripts for make O=dir run
  selftests: lib.mk: copy test scripts and test files for make O=dir run
  selftests: sync: kselftest and kselftest-clean fail for make O=dir case
  selftests: sync: use TEST_CUSTOM_PROGS instead of TEST_PROGS
  selftests: lib.mk: add TEST_CUSTOM_PROGS to allow custom test run/install
  selftests: watchdog: fix to use TEST_GEN_PROGS and remove clean
  selftests: lib.mk: fix test executable status check to use full path
  selftests: Makefile: clear LDFLAGS for make O=dir use-case
  selftests: lib.mk: kselftest and kselftest-clean fail for make O=dir case
  Makefile: kselftest and kselftest-clean fail for make O=dir case
  selftests/net: msg_zerocopy enable build with older kernel headers
  selftests: actually run the various net selftests
  selftest: add a reuseaddr test
  ...

7 years agoMerge branch 'x86-fpu-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Wed, 27 Sep 2017 15:33:26 +0000 (08:33 -0700)]
Merge branch 'x86-fpu-for-linus' of git://git./linux/kernel/git/tip/tip

Pull x86 fpu fixes and cleanups from Ingo Molnar:
 "This is _way_ more cleanups than fixes, but the bugs were subtle and
  hard to hit, and the primary reason for them existing was the
  unnecessary historical complexity of some of the x86/fpu interfaces.

  The first bunch of commits clean up and simplify the xstate user copy
  handling functions, in reaction to the collective head-scratching
  about the xstate user-copy handling code that leads up to the fix for
  this SkyLake xstate handling bug:

     0852b374173b: x86/fpu: Add FPU state copying quirk to handle XRSTOR failure on Intel Skylake CPUs

  The cleanups don't change any functionality, they just (hopefully)
  make it all clearer, more consistent, more debuggable and more robust.

  Note that most of the linecount increase comes from these commits,
  where we better split the user/kernel copy logic by having more
  variants, instead repeated fragile patterns of:

               if (kbuf) {
                       memcpy(kbuf + pos, data, copy);
               } else {
                       if (__copy_to_user(ubuf + pos, data, copy))
                               return -EFAULT;
               }

  The next bunch of commits simplify the FPU state-machine to get rid of
  old lazy-FPU idiosyncrasies - a defensive simplification to make all
  the code easier to review and fix. No change in functionality.

  Then there's a couple of additional debugging tweaks: static checker
  warning fix and move an FPU related warning to under WARN_ON_FPU(),
  followed by another bunch of commits that represent a finegrained
  split-up of the fixes from Eric Biggers to handle weird xstate bits
  properly.

  I did this finegrained split-up because some of these fixes also
  impact the ABI for weird xstate handling, for which we'd like to have
  good bisection results, should they cause any problems. (We also had
  one regression with the more monolithic fixes, so splitting it all up
  sounded prudent for robustness reasons as well.)

  About the whole series: the commits up to 03eaec81ac09 have been in
  -next for months - but I've recently rebased them to remove a state
  machine clean-up commit that was objected to, and to make it more
  bisectable - so technically it's a new, rebased tree.

  Robustness history: this series had some regressions along the way,
  and all reported regressions have been fixed. All but one of the
  regressions manifested itself as easy to report warnings. The previous
  version of this latest series was also in linux-next, with one
  (warning-only) regression reported which is fixed in the latest
  version.

  Barring last minute brown paper bag bugs (and the commits are now
  older by a day which I'd hope helps paperbag reduction), I'm
  reasonably confident about its general robustness.

  Famous last words ..."

* 'x86-fpu-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (42 commits)
  x86/fpu: Use using_compacted_format() instead of open coded X86_FEATURE_XSAVES
  x86/fpu: Use validate_xstate_header() to validate the xstate_header in copy_user_to_xstate()
  x86/fpu: Eliminate the 'xfeatures' local variable in copy_user_to_xstate()
  x86/fpu: Copy the full header in copy_user_to_xstate()
  x86/fpu: Use validate_xstate_header() to validate the xstate_header in copy_kernel_to_xstate()
  x86/fpu: Eliminate the 'xfeatures' local variable in copy_kernel_to_xstate()
  x86/fpu: Copy the full state_header in copy_kernel_to_xstate()
  x86/fpu: Use validate_xstate_header() to validate the xstate_header in __fpu__restore_sig()
  x86/fpu: Use validate_xstate_header() to validate the xstate_header in xstateregs_set()
  x86/fpu: Introduce validate_xstate_header()
  x86/fpu: Rename fpu__activate_fpstate_read/write() to fpu__prepare_[read|write]()
  x86/fpu: Rename fpu__activate_curr() to fpu__initialize()
  x86/fpu: Simplify and speed up fpu__copy()
  x86/fpu: Fix stale comments about lazy FPU logic
  x86/fpu: Rename fpu::fpstate_active to fpu::initialized
  x86/fpu: Remove fpu__current_fpstate_write_begin/end()
  x86/fpu: Fix fpu__activate_fpstate_read() and update comments
  x86/fpu: Reinitialize FPU registers if restoring FPU state fails
  x86/fpu: Don't let userspace set bogus xcomp_bv
  x86/fpu: Turn WARN_ON() in context switch into WARN_ON_FPU()
  ...

7 years agoIB/hfi1: Unsuccessful PCIe caps tuning should not fail driver load
Harish Chegondi [Tue, 26 Sep 2017 13:06:34 +0000 (06:06 -0700)]
IB/hfi1: Unsuccessful PCIe caps tuning should not fail driver load

Failure to tune PCIe capabilities should not fail driver load. This can
cause the driver load to fail on systems with any of the following:
1. HFI's parent is not root. Example: HFI card is behind a PCIe bridge.
2. HFI's parent is not PCI Express capable.
In these situations, failure to tune PCIe capabilities should be logged
in the system message logs but not cause the driver load to fail.

This patch also ensures pcie capability word DevCtl is written only
after a successful read and the capability tuning process continues
even if read/write of the pcie capability word DevCtl fails.

Fixes: c53df62c7a9a ("IB/hfi1: Check return values from PCI config API calls")
Fixes: bf70a7757736 ("staging/rdma/hfi1: Enable WFR PCIe extended tags from the driver")
Reviewed-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Reviewed-by: Jakub Byczkowski <jakub.byczkowski@intel.com>
Signed-off-by: Harish Chegondi <harish.chegondi@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoIB/hfi1: On error, fix use after free during user context setup
Michael J. Ruhl [Tue, 26 Sep 2017 13:06:28 +0000 (06:06 -0700)]
IB/hfi1: On error, fix use after free during user context setup

During base context setup, if setup_base_ctxt() fails, the context is
deallocated. This is incorrect because the context is referenced on
return, to notify any waiting subcontext.  If there are no subcontexts
the pointer will be invalid.

Reorganize the error path so that deallocate_ctxt() is called after all
the possible subcontexts have been notified.

Reviewed-by: Ira Weiny <ira.weiny@intel.com>
Signed-off-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoRevert "IB/ipoib: Update broadcast object if PKey value was changed in index 0"
Alex Estrin [Tue, 26 Sep 2017 13:06:22 +0000 (06:06 -0700)]
Revert "IB/ipoib: Update broadcast object if PKey value was changed in index 0"

commit 9a9b8112699d will cause core to fail UD QP from being destroyed
on ipoib unload, therefore cause resources leakage.
On pkey change event above patch modifies mgid before calling underlying
driver to detach it from QP. Drivers' detach_mcast() will fail to find
modified mgid it was never given to attach in a first place.
Core qp->usecnt will never go down, so ib_destroy_qp() will fail.

IPoIB driver actually does take care of new broadcast mgid based on new
pkey by destroying an old mcast object in ipoib_mcast_dev_flush())
....
if (priv->broadcast) {
rb_erase(&priv->broadcast->rb_node, &priv->multicast_tree);
list_add_tail(&priv->broadcast->list, &remove_list);
priv->broadcast = NULL;
}
...

then in restarted ipoib_macst_join_task() creating a new broadcast mcast
object, sending join request and on completion tells the driver to attach
to reinitialized QP:
...
if (!priv->broadcast) {
...
broadcast = ipoib_mcast_alloc(dev, 0);
...
memcpy(broadcast->mcmember.mgid.raw, priv->dev->broadcast + 4,
       sizeof (union ib_gid));
priv->broadcast = broadcast;
...

Fixes: 9a9b8112699d ("IB/ipoib: Update broadcast object if PKey value was changed in index 0")
Cc: stable@vger.kernel.org
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Alex Estrin <alex.estrin@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Reviewed-by: Feras Daoud <ferasda@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoIB/hfi1: Return correct value in general interrupt handler
Kamenee Arumugam [Tue, 26 Sep 2017 13:06:15 +0000 (06:06 -0700)]
IB/hfi1: Return correct value in general interrupt handler

The general interrupt handler returns IRQ_HANDLED whether an IRQ
was handled or not.
Determine if an IRQ was handled and return the correct value.

Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Reviewed-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
Signed-off-by: Kamenee Arumugam <kamenee.arumugam@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoIB/hfi1: Check eeprom config partition validity
Jan Sokolowski [Tue, 26 Sep 2017 13:06:09 +0000 (06:06 -0700)]
IB/hfi1: Check eeprom config partition validity

Relying on a trailing magic value is incorrect. There are instances where
this is not present as trailing magic value has a specific purpose which is
not partition validation. Instead use the header magic value which is
present in all variants of the platform configuration and is intended for
validation. This is also used in other locations in the driver.

Fixes: bc5214ee2922 (IB/hfi1: Handle missing magic values in config file)
Reviewed-by: Jakub Byczkowski <jakub.byczkowski@intel.com>
Signed-off-by: Jan Sokolowski <jan.sokolowski@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoIB/hfi1: Only reset QSFP after link up and turn off AOC TX
Sebastian Sanchez [Tue, 26 Sep 2017 13:06:03 +0000 (06:06 -0700)]
IB/hfi1: Only reset QSFP after link up and turn off AOC TX

QSFP reset enables AOC transmitters by default. They should be off
before moving to high power mode to complete the setup. There is no
need to reset the QSFP during LNI failure as it was reset at link down.

Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Reviewed-by: Jakub Byczkowski <jakub.byczkowski@intel.com>
Signed-off-by: Sebastian Sanchez <sebastian.sanchez@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoIB/hfi1: Turn off AOC TX after offline substates
Sebastian Sanchez [Tue, 26 Sep 2017 13:05:57 +0000 (06:05 -0700)]
IB/hfi1: Turn off AOC TX after offline substates

Offline.quietDuration was added in the 8051 firmware, and the driver
only turns off the AOC transmitters when offline.quiet is reached.
However, the AOC transmitters need to be turned off at the new state.
Therefore, turn off the AOC transmitters at any offline substates
including offline.quiet and offline.quietDuration, then recheck we
reached offline.quiet to support backwards compatibility.

Reviewed-by: Jakub Byczkowski <jakub.byczkowski@intel.com>
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Sebastian Sanchez <sebastian.sanchez@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoquota: Fix quota corruption with generic/232 test
Jan Kara [Tue, 26 Sep 2017 08:36:05 +0000 (10:36 +0200)]
quota: Fix quota corruption with generic/232 test

Eric has reported that since commit d2faa415166b "quota: Do not acquire
dqio_sem for dquot overwrites in v2 format" test generic/232
occasionally fails due to quota information being incorrect. Indeed that
commit was too eager to remove dqio_sem completely from the path that
just overwrites quota structure with updated information. Although that
is innocent on its own, another process that inserts new quota structure
to the same block can perform read-modify-write cycle of that block thus
effectively discarding quota information update if they race in a wrong
way.

Fix the problem by acquiring dqio_sem for reading for overwrites of
quota structure. Note that it *is* possible to completely avoid taking
dqio_sem in the overwrite path however that will require modifying path
inserting / deleting quota structures to avoid RMW cycles of the full
block and for now it is not clear whether it is worth the hassle.

Fixes: d2faa415166b2883428efa92f451774ef44373ac
Reported-and-tested-by: Eric Whitney <enwlinux@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
7 years agoMerge tag 'mmc-v4.14-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/ulfh/mmc
Linus Torvalds [Tue, 26 Sep 2017 23:54:22 +0000 (16:54 -0700)]
Merge tag 'mmc-v4.14-rc1' of git://git./linux/kernel/git/ulfh/mmc

Pull MMC fixes from Ulf Hansson:

  - sdhci-pci: Fix voltage switch for some Intel host controllers

  - tmio: remove broken and noisy debug macro

* tag 'mmc-v4.14-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/ulfh/mmc:
  mmc: sdhci-pci: Fix voltage switch for some Intel host controllers
  mmc: tmio: remove broken and noisy debug macro

7 years agovfs: Return -ENXIO for negative SEEK_HOLE / SEEK_DATA offsets
Andreas Gruenbacher [Mon, 25 Sep 2017 10:23:03 +0000 (12:23 +0200)]
vfs: Return -ENXIO for negative SEEK_HOLE / SEEK_DATA offsets

In generic_file_llseek_size, return -ENXIO for negative offsets as well
as offsets beyond EOF.  This affects filesystems which don't implement
SEEK_HOLE / SEEK_DATA internally, possibly because they don't support
holes.

Fixes xfstest generic/448.

Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Cc: stable@vger.kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agoMerge branch 'WIP.x86/fpu' into x86/fpu, because it's ready
Ingo Molnar [Tue, 26 Sep 2017 08:17:43 +0000 (10:17 +0200)]
Merge branch 'WIP.x86/fpu' into x86/fpu, because it's ready

Signed-off-by: Ingo Molnar <mingo@kernel.org>
7 years agox86/fpu: Use using_compacted_format() instead of open coded X86_FEATURE_XSAVES
Eric Biggers [Sun, 24 Sep 2017 10:59:13 +0000 (12:59 +0200)]
x86/fpu: Use using_compacted_format() instead of open coded X86_FEATURE_XSAVES

This is the canonical method to use.

Signed-off-by: Eric Biggers <ebiggers@google.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Eric Biggers <ebiggers3@gmail.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Kevin Hao <haokexin@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Michael Halcrow <mhalcrow@google.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Wanpeng Li <wanpeng.li@hotmail.com>
Cc: Yu-cheng Yu <yu-cheng.yu@intel.com>
Cc: kernel-hardening@lists.openwall.com
Link: http://lkml.kernel.org/r/20170924105913.9157-11-mingo@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
7 years agox86/fpu: Use validate_xstate_header() to validate the xstate_header in copy_user_to_x...
Eric Biggers [Sun, 24 Sep 2017 10:59:12 +0000 (12:59 +0200)]
x86/fpu: Use validate_xstate_header() to validate the xstate_header in copy_user_to_xstate()

Tighten the checks in copy_user_to_xstate().

Signed-off-by: Eric Biggers <ebiggers@google.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Eric Biggers <ebiggers3@gmail.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Kevin Hao <haokexin@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Michael Halcrow <mhalcrow@google.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Wanpeng Li <wanpeng.li@hotmail.com>
Cc: Yu-cheng Yu <yu-cheng.yu@intel.com>
Cc: kernel-hardening@lists.openwall.com
Link: http://lkml.kernel.org/r/20170924105913.9157-10-mingo@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
7 years agox86/fpu: Eliminate the 'xfeatures' local variable in copy_user_to_xstate()
Eric Biggers [Sun, 24 Sep 2017 10:59:11 +0000 (12:59 +0200)]
x86/fpu: Eliminate the 'xfeatures' local variable in copy_user_to_xstate()

We now have this field in hdr.xfeatures.

Signed-off-by: Eric Biggers <ebiggers@google.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Eric Biggers <ebiggers3@gmail.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Kevin Hao <haokexin@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Michael Halcrow <mhalcrow@google.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Wanpeng Li <wanpeng.li@hotmail.com>
Cc: Yu-cheng Yu <yu-cheng.yu@intel.com>
Cc: kernel-hardening@lists.openwall.com
Link: http://lkml.kernel.org/r/20170924105913.9157-9-mingo@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
7 years agox86/fpu: Copy the full header in copy_user_to_xstate()
Eric Biggers [Sun, 24 Sep 2017 10:59:10 +0000 (12:59 +0200)]
x86/fpu: Copy the full header in copy_user_to_xstate()

This is in preparation to verify the full xstate header as supplied by user-space.

Signed-off-by: Eric Biggers <ebiggers@google.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Eric Biggers <ebiggers3@gmail.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Kevin Hao <haokexin@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Michael Halcrow <mhalcrow@google.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Wanpeng Li <wanpeng.li@hotmail.com>
Cc: Yu-cheng Yu <yu-cheng.yu@intel.com>
Cc: kernel-hardening@lists.openwall.com
Link: http://lkml.kernel.org/r/20170924105913.9157-8-mingo@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
7 years agox86/fpu: Use validate_xstate_header() to validate the xstate_header in copy_kernel_to...
Eric Biggers [Sun, 24 Sep 2017 10:59:09 +0000 (12:59 +0200)]
x86/fpu: Use validate_xstate_header() to validate the xstate_header in copy_kernel_to_xstate()

Tighten the checks in copy_kernel_to_xstate().

Signed-off-by: Eric Biggers <ebiggers@google.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Eric Biggers <ebiggers3@gmail.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Kevin Hao <haokexin@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Michael Halcrow <mhalcrow@google.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Wanpeng Li <wanpeng.li@hotmail.com>
Cc: Yu-cheng Yu <yu-cheng.yu@intel.com>
Cc: kernel-hardening@lists.openwall.com
Link: http://lkml.kernel.org/r/20170924105913.9157-7-mingo@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
7 years agox86/fpu: Eliminate the 'xfeatures' local variable in copy_kernel_to_xstate()
Eric Biggers [Sun, 24 Sep 2017 10:59:08 +0000 (12:59 +0200)]
x86/fpu: Eliminate the 'xfeatures' local variable in copy_kernel_to_xstate()

We have this information in the xstate_header.

Signed-off-by: Eric Biggers <ebiggers@google.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Eric Biggers <ebiggers3@gmail.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Kevin Hao <haokexin@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Michael Halcrow <mhalcrow@google.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Wanpeng Li <wanpeng.li@hotmail.com>
Cc: Yu-cheng Yu <yu-cheng.yu@intel.com>
Cc: kernel-hardening@lists.openwall.com
Link: http://lkml.kernel.org/r/20170924105913.9157-6-mingo@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
7 years agox86/fpu: Copy the full state_header in copy_kernel_to_xstate()
Eric Biggers [Sun, 24 Sep 2017 10:59:07 +0000 (12:59 +0200)]
x86/fpu: Copy the full state_header in copy_kernel_to_xstate()

This is in preparation to verify the full xstate header as supplied by user-space.

Signed-off-by: Eric Biggers <ebiggers@google.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Eric Biggers <ebiggers3@gmail.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Kevin Hao <haokexin@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Michael Halcrow <mhalcrow@google.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Wanpeng Li <wanpeng.li@hotmail.com>
Cc: Yu-cheng Yu <yu-cheng.yu@intel.com>
Cc: kernel-hardening@lists.openwall.com
Link: http://lkml.kernel.org/r/20170924105913.9157-5-mingo@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
7 years agox86/fpu: Use validate_xstate_header() to validate the xstate_header in __fpu__restore...
Eric Biggers [Sun, 24 Sep 2017 10:59:06 +0000 (12:59 +0200)]
x86/fpu: Use validate_xstate_header() to validate the xstate_header in __fpu__restore_sig()

Tighten the checks in __fpu__restore_sig() and update comments.

Signed-off-by: Eric Biggers <ebiggers@google.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Eric Biggers <ebiggers3@gmail.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Kevin Hao <haokexin@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Michael Halcrow <mhalcrow@google.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Wanpeng Li <wanpeng.li@hotmail.com>
Cc: Yu-cheng Yu <yu-cheng.yu@intel.com>
Cc: kernel-hardening@lists.openwall.com
Link: http://lkml.kernel.org/r/20170924105913.9157-4-mingo@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
7 years agox86/fpu: Use validate_xstate_header() to validate the xstate_header in xstateregs_set()
Eric Biggers [Sun, 24 Sep 2017 10:59:05 +0000 (12:59 +0200)]
x86/fpu: Use validate_xstate_header() to validate the xstate_header in xstateregs_set()

Tighten the checks in xstateregs_set().

Signed-off-by: Eric Biggers <ebiggers@google.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Eric Biggers <ebiggers3@gmail.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Kevin Hao <haokexin@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Michael Halcrow <mhalcrow@google.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Wanpeng Li <wanpeng.li@hotmail.com>
Cc: Yu-cheng Yu <yu-cheng.yu@intel.com>
Cc: kernel-hardening@lists.openwall.com
Link: http://lkml.kernel.org/r/20170924105913.9157-3-mingo@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
7 years agox86/fpu: Introduce validate_xstate_header()
Eric Biggers [Sun, 24 Sep 2017 10:59:04 +0000 (12:59 +0200)]
x86/fpu: Introduce validate_xstate_header()

Move validation of user-supplied xstate_header into a helper function,
in preparation of calling it from both the ptrace and sigreturn syscall
paths.

The new function also considers it to be an error if *any* reserved bits
are set, whereas before we were just clearing most of them silently.

This should reduce the chance of bugs that fail to correctly validate
user-supplied XSAVE areas.  It also will expose any broken userspace
programs that set the other reserved bits; this is desirable because
such programs will lose compatibility with future CPUs and kernels if
those bits are ever used for anything.  (There shouldn't be any such
programs, and in fact in the case where the compacted format is in use
we were already validating xfeatures.  But you never know...)

Signed-off-by: Eric Biggers <ebiggers@google.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Eric Biggers <ebiggers3@gmail.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Kevin Hao <haokexin@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Michael Halcrow <mhalcrow@google.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Wanpeng Li <wanpeng.li@hotmail.com>
Cc: Yu-cheng Yu <yu-cheng.yu@intel.com>
Cc: kernel-hardening@lists.openwall.com
Link: http://lkml.kernel.org/r/20170924105913.9157-2-mingo@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
7 years agox86/fpu: Rename fpu__activate_fpstate_read/write() to fpu__prepare_[read|write]()
Ingo Molnar [Sat, 23 Sep 2017 11:37:45 +0000 (13:37 +0200)]
x86/fpu: Rename fpu__activate_fpstate_read/write() to fpu__prepare_[read|write]()

As per the new nomenclature we don't 'activate' the FPU state
anymore, we initialize it. So drop the _activate_fpstate name
from these functions, which were a bit of a mouthful anyway,
and name them:

fpu__prepare_read()
fpu__prepare_write()

Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Eric Biggers <ebiggers3@gmail.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
7 years agox86/fpu: Rename fpu__activate_curr() to fpu__initialize()
Ingo Molnar [Sat, 23 Sep 2017 13:00:15 +0000 (15:00 +0200)]
x86/fpu: Rename fpu__activate_curr() to fpu__initialize()

Rename this function to better express that it's all about
initializing the FPU state of a task which goes hand in hand
with the fpu::initialized field.

Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Eric Biggers <ebiggers3@gmail.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Yu-cheng Yu <yu-cheng.yu@intel.com>
Link: http://lkml.kernel.org/r/20170923130016.21448-33-mingo@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
7 years agox86/fpu: Simplify and speed up fpu__copy()
Ingo Molnar [Sat, 23 Sep 2017 13:00:14 +0000 (15:00 +0200)]
x86/fpu: Simplify and speed up fpu__copy()

fpu__copy() has a preempt_disable()/enable() pair, which it had to do to
be able to atomically unlazy the current task when doing an FNSAVE.

But we don't unlazy tasks anymore, we always do direct saves/restores of
FPU context.

So remove both the unnecessary critical section, and update the comments.

Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Eric Biggers <ebiggers3@gmail.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Yu-cheng Yu <yu-cheng.yu@intel.com>
Link: http://lkml.kernel.org/r/20170923130016.21448-32-mingo@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
7 years agox86/fpu: Fix stale comments about lazy FPU logic
Ingo Molnar [Sat, 23 Sep 2017 13:00:13 +0000 (15:00 +0200)]
x86/fpu: Fix stale comments about lazy FPU logic

We don't do any lazy restore anymore, what we have are two pieces of optimization:

 - no-FPU tasks that don't save/restore the FPU context (kernel threads are such)

 - cached FPU registers maintained via the fpu->last_cpu field. This means that
   if an FPU task context switches to a non-FPU task then we can maintain the
   FPU registers as an in-FPU copies (cache), and skip the restoration of them
   once we switch back to the original FPU-using task.

Update all the comments that still referred to old 'lazy' and 'unlazy' concepts.

Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Eric Biggers <ebiggers3@gmail.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Yu-cheng Yu <yu-cheng.yu@intel.com>
Link: http://lkml.kernel.org/r/20170923130016.21448-31-mingo@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
7 years agox86/fpu: Rename fpu::fpstate_active to fpu::initialized
Ingo Molnar [Tue, 26 Sep 2017 07:43:36 +0000 (09:43 +0200)]
x86/fpu: Rename fpu::fpstate_active to fpu::initialized

The x86 FPU code used to have a complex state machine where both the FPU
registers and the FPU state context could be 'active' (or inactive)
independently of each other - which enabled features like lazy FPU restore.

Much of this complexity is gone in the current code: now we basically can
have FPU-less tasks (kernel threads) that don't use (and save/restore) FPU
state at all, plus full FPU users that save/restore directly with no laziness
whatsoever.

But the fpu::fpstate_active still carries bits of the old complexity - meanwhile
this flag has become a simple flag that shows whether the FPU context saving
area in the thread struct is initialized and used, or not.

Rename it to fpu::initialized to express this simplicity in the name as well.

Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Eric Biggers <ebiggers3@gmail.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Yu-cheng Yu <yu-cheng.yu@intel.com>
Link: http://lkml.kernel.org/r/20170923130016.21448-30-mingo@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
7 years agox86/fpu: Remove fpu__current_fpstate_write_begin/end()
Ingo Molnar [Sat, 23 Sep 2017 13:00:11 +0000 (15:00 +0200)]
x86/fpu: Remove fpu__current_fpstate_write_begin/end()

These functions are not used anymore, so remove them.

Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Bobby Powers <bobbypowers@gmail.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Eric Biggers <ebiggers3@gmail.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Yu-cheng Yu <yu-cheng.yu@intel.com>
Link: http://lkml.kernel.org/r/20170923130016.21448-29-mingo@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
7 years agox86/fpu: Fix fpu__activate_fpstate_read() and update comments
Ingo Molnar [Sat, 23 Sep 2017 13:00:10 +0000 (15:00 +0200)]
x86/fpu: Fix fpu__activate_fpstate_read() and update comments

fpu__activate_fpstate_read() can be called for the current task
when coredumping - or for stopped tasks when ptrace-ing.

Implement this properly in the code and update the comments.

This also fixes an incorrect (but harmless) warning introduced by
one of the earlier patches.

Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Eric Biggers <ebiggers3@gmail.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Yu-cheng Yu <yu-cheng.yu@intel.com>
Link: http://lkml.kernel.org/r/20170923130016.21448-28-mingo@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
7 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs
Linus Torvalds [Tue, 26 Sep 2017 01:24:14 +0000 (18:24 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/viro/vfs

Pull compat fix from Al Viro:
 "I really wish gcc warned about conversions from pointer to function
  into void *..."

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
  fix a typo in put_compat_shm_info()

7 years agofix a typo in put_compat_shm_info()
Al Viro [Tue, 26 Sep 2017 00:38:45 +0000 (20:38 -0400)]
fix a typo in put_compat_shm_info()

"uip" misspelled as "up"; unfortunately, the latter happens to be
a function and gcc is happy to convert it to void *...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
7 years agoMerge branch 'for-linus' of git://git.kernel.dk/linux-block
Linus Torvalds [Mon, 25 Sep 2017 22:46:04 +0000 (15:46 -0700)]
Merge branch 'for-linus' of git://git.kernel.dk/linux-block

Pull block fixes from Jens Axboe:

 - Two sets of NVMe pull requests from Christoph:
      - Fixes for the Fibre Channel host/target to fix spec compliance
      - Allow a zero keep alive timeout
      - Make the debug printk for broken SGLs work better
      - Fix queue zeroing during initialization
      - Set of RDMA and FC fixes
      - Target div-by-zero fix

 - bsg double-free fix.

 - ndb unknown ioctl fix from Josef.

 - Buffered vs O_DIRECT page cache inconsistency fix. Has been floating
   around for a long time, well reviewed. From Lukas.

 - brd overflow fix from Mikulas.

 - Fix for a loop regression in this merge window, where using a union
   for two members of the loop_cmd turned out to be a really bad idea.
   From Omar.

 - Fix for an iostat regression fix in this series, using the wrong API
   to get at the block queue. From Shaohua.

 - Fix for a potential blktrace delection deadlock. From Waiman.

* 'for-linus' of git://git.kernel.dk/linux-block: (30 commits)
  nvme-fcloop: fix port deletes and callbacks
  nvmet-fc: sync header templates with comments
  nvmet-fc: ensure target queue id within range.
  nvmet-fc: on port remove call put outside lock
  nvme-rdma: don't fully stop the controller in error recovery
  nvme-rdma: give up reconnect if state change fails
  nvme-core: Use nvme_wq to queue async events and fw activation
  nvme: fix sqhd reference when admin queue connect fails
  block: fix a crash caused by wrong API
  fs: Fix page cache inconsistency when mixing buffered and AIO DIO
  nvmet: implement valid sqhd values in completions
  nvme-fabrics: Allow 0 as KATO value
  nvme: allow timed-out ios to retry
  nvme: stop aer posting if controller state not live
  nvme-pci: Print invalid SGL only once
  nvme-pci: initialize queue memory before interrupts
  nvmet-fc: fix failing max io queue connections
  nvme-fc: use transport-specific sgl format
  nvme: add transport SGL definitions
  nvme.h: remove FC transport-specific error values
  ...

7 years agoMerge tag 'gfs2-for-linus-4.14-rc3' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Mon, 25 Sep 2017 22:41:56 +0000 (15:41 -0700)]
Merge tag 'gfs2-for-linus-4.14-rc3' of git://git./linux/kernel/git/gfs2/linux-gfs2

Pull gfs2 fix from Bob Peterson:
 "GFS2: Fix an old regression in GFS2's debugfs interface

 This fixes a regression introduced by commit 88ffbf3e037e ("GFS2: Use
 resizable hash table for glocks"). The regression caused the glock dump
 in debugfs to not report all the glocks, which makes debugging
 extremely difficult"

* tag 'gfs2-for-linus-4.14-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/gfs2/linux-gfs2:
  gfs2: Fix debugfs glocks dump

7 years agoMerge tag 'microblaze-4.14-rc3' of git://git.monstr.eu/linux-2.6-microblaze
Linus Torvalds [Mon, 25 Sep 2017 22:37:19 +0000 (15:37 -0700)]
Merge tag 'microblaze-4.14-rc3' of git://git.monstr.eu/linux-2.6-microblaze

Pull Microblaze fixes from Michal Simek:

 - Kbuild fix

 - use vma_pages

 - setup default little endians

* tag 'microblaze-4.14-rc3' of git://git.monstr.eu/linux-2.6-microblaze:
  arch: change default endian for microblaze
  microblaze: Cocci spatch "vma_pages"
  microblaze: Add missing kvm_para.h to Kbuild

7 years agoMerge tag 'trace-v4.14-rc1-2' of git://git.kernel.org/pub/scm/linux/kernel/git/rosted...
Linus Torvalds [Mon, 25 Sep 2017 22:22:31 +0000 (15:22 -0700)]
Merge tag 'trace-v4.14-rc1-2' of git://git./linux/kernel/git/rostedt/linux-trace

Pull tracing fixes from Steven Rostedt:
 "Stack tracing and RCU has been having issues with each other and
  lockdep has been pointing out constant problems.

  The changes have been going into the stack tracer, but it has been
  discovered that the problem isn't with the stack tracer itself, but it
  is with calling save_stack_trace() from within the internals of RCU.

  The stack tracer is the one that can trigger the issue the easiest,
  but examining the problem further, it could also happen from a WARN()
  in the wrong place, or even if an NMI happened in this area and it did
  an rcu_read_lock().

  The critical area is where RCU is not watching. Which can happen while
  going to and from idle, or bringing up or taking down a CPU.

  The final fix was to put the protection in kernel_text_address() as it
  is the one that requires RCU to be watching while doing the stack
  trace.

  To make this work properly, Paul had to allow rcu_irq_enter() happen
  after rcu_nmi_enter(). This should have been done anyway, since an NMI
  can page fault (reading vmalloc area), and a page fault triggers
  rcu_irq_enter().

  One patch is just a consolidation of code so that the fix only needed
  to be done in one location"

* tag 'trace-v4.14-rc1-2' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
  tracing: Remove RCU work arounds from stack tracer
  extable: Enable RCU if it is not watching in kernel_text_address()
  extable: Consolidate *kernel_text_address() functions
  rcu: Allow for page faults in NMI handlers

7 years agonvme-fcloop: fix port deletes and callbacks
James Smart [Tue, 19 Sep 2017 21:01:50 +0000 (14:01 -0700)]
nvme-fcloop: fix port deletes and callbacks

Now that there are potentially long delays between when a remoteport or
targetport delete calls is made and when the callback occurs (dev_loss_tmo
timeout), no longer block in the delete routines and move the final nport
puts to the callbacks.

Moved the fcloop_nport_get/put/free routines to avoid forward declarations.

Ensure port_info structs used in registrations are nulled in case fields
are not set (ex: devloss_tmo values).

Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
7 years agonvmet-fc: sync header templates with comments
James Smart [Wed, 20 Sep 2017 18:07:26 +0000 (11:07 -0700)]
nvmet-fc: sync header templates with comments

Comments were incorrect:
- defer_rcv was in host port template. moved to target port template
- Added Mandatory statements for target port template items

Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
7 years agonvmet-fc: ensure target queue id within range.
James Smart [Tue, 19 Sep 2017 23:33:56 +0000 (16:33 -0700)]
nvmet-fc: ensure target queue id within range.

When searching for queue id's ensure they are within the expected range.

Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
7 years agonvmet-fc: on port remove call put outside lock
James Smart [Tue, 19 Sep 2017 22:13:11 +0000 (15:13 -0700)]
nvmet-fc: on port remove call put outside lock

Avoid calling the put routine, as it may traverse to free routines while
holding the target lock.

Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
7 years agonvme-rdma: don't fully stop the controller in error recovery
Sagi Grimberg [Thu, 21 Sep 2017 14:01:38 +0000 (17:01 +0300)]
nvme-rdma: don't fully stop the controller in error recovery

By calling nvme_stop_ctrl on a already failed controller will wait for the
scan work to complete (only by identify timeout expiration which is 60
seconds). This is unnecessary when we already know that the controller has
failed.

Reported-by: Yi Zhang <yizhan@redhat.com>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
7 years agonvme-rdma: give up reconnect if state change fails
Sagi Grimberg [Thu, 21 Sep 2017 14:01:37 +0000 (17:01 +0300)]
nvme-rdma: give up reconnect if state change fails

If we failed to transition to state LIVE after a successful reconnect,
then controller deletion already started. In this case there is no
point moving forward with reconnect.

Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
7 years agonvme-core: Use nvme_wq to queue async events and fw activation
Sagi Grimberg [Thu, 21 Sep 2017 14:01:36 +0000 (17:01 +0300)]
nvme-core: Use nvme_wq to queue async events and fw activation

async_event_work might race as it is executed from two different
workqueues at the moment.

Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
7 years agonvme: fix sqhd reference when admin queue connect fails
James Smart [Thu, 21 Sep 2017 15:13:49 +0000 (08:13 -0700)]
nvme: fix sqhd reference when admin queue connect fails

Fix bug in sqhd patch.

It wasn't the sq that was at risk. In the case where the admin queue
connect command fails, the sq->size field is not set. Therefore, this
becomes a divide by zero error.

Add a quick check to bypass under this failure condition.

Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
7 years agogfs2: Fix debugfs glocks dump
Andreas Gruenbacher [Tue, 19 Sep 2017 12:15:35 +0000 (07:15 -0500)]
gfs2: Fix debugfs glocks dump

The switch to rhashtables (commit 88ffbf3e03) broke the debugfs glock
dump (/sys/kernel/debug/gfs2/<device>/glocks) for dumps bigger than a
single buffer: the right function for restarting an rhashtable iteration
from the beginning of the hash table is rhashtable_walk_enter;
rhashtable_walk_stop + rhashtable_walk_start will just resume from the
current position.

Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Cc: stable@vger.kernel.org # v4.3+
7 years agoselftests: timers: set-timer-lat: Fix hang when testing unsupported alarms
Shuah Khan [Thu, 21 Sep 2017 19:05:18 +0000 (13:05 -0600)]
selftests: timers: set-timer-lat: Fix hang when testing unsupported alarms

When timer_create() fails on a bootime or realtime clock, setup_timer()
returns 0 as if timer has been set. Callers wait forever for the timer
to expire.

This hang is seen on a system that doesn't have support for:

CLOCK_REALTIME_ALARM   ABSTIME missing CAP_WAKE_ALARM? : [UNSUPPORTED]

Test hangs waiting for a timer that hasn't been set to expire. Fix
setup_timer() to return 1, add handling in callers to detect the
unsupported case and return 0 without waiting to not fail the test.

Signed-off-by: Shuah Khan <shuahkh@osg.samsung.com>
7 years agoselftests: timers: set-timer-lat: fix hang when std out/err are redirected
Shuah Khan [Thu, 21 Sep 2017 19:46:01 +0000 (13:46 -0600)]
selftests: timers: set-timer-lat: fix hang when std out/err are redirected

do_timer_oneshot() uses select() as a timer with FD_SETSIZE and readfs
is cleared with FD_ZERO without FD_SET.

When stdout and stderr are redirected, the test hangs in select forever.
Fix the problem calling select() with readfds empty and nfds zero. This
is sufficient for using select() for timer.

With this fix "./set-timer-lat > /dev/null 2>&1" no longer hangs.

Signed-off-by: Shuah Khan <shuahkh@osg.samsung.com>
Acked-by: Greg Hackmann <ghackmann@google.com>
Signed-off-by: Shuah Khan <shuahkh@osg.samsung.com>
7 years agoselftests/memfd: correct run_tests.sh permission
Li Zhijian [Thu, 21 Sep 2017 09:13:27 +0000 (17:13 +0800)]
selftests/memfd: correct run_tests.sh permission

to fix the following issue:
------------------
TAP version 13
selftests: run_tests.sh
========================================
selftests: Warning: file run_tests.sh is not executable, correct this.
not ok 1..1 selftests: run_tests.sh [FAIL]
------------------

Signed-off-by: Li Zhijian <lizhijian@cn.fujitsu.com>
Signed-off-by: Shuah Khan <shuahkh@osg.samsung.com>
7 years agoselftests/seccomp: Support glibc 2.26 siginfo_t.h
Kees Cook [Thu, 7 Sep 2017 23:32:46 +0000 (16:32 -0700)]
selftests/seccomp: Support glibc 2.26 siginfo_t.h

The 2.26 release of glibc changed how siginfo_t is defined, and the earlier
work-around to using the kernel definition are no longer needed. The old
way needs to stay around for a while, though.

Reported-by: Seth Forshee <seth.forshee@canonical.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Will Drewry <wad@chromium.org>
Cc: Shuah Khan <shuah@kernel.org>
Cc: linux-kselftest@vger.kernel.org
Cc: stable@vger.kernel.org
Signed-off-by: Kees Cook <keescook@chromium.org>
Tested-by: Seth Forshee <seth.forshee@canonical.com>
Signed-off-by: Shuah Khan <shuahkh@osg.samsung.com>
7 years agoselftests: futex: Makefile: fix for loops in targets to run silently
Shuah Khan [Tue, 19 Sep 2017 00:46:23 +0000 (18:46 -0600)]
selftests: futex: Makefile: fix for loops in targets to run silently

Fix for loops in targets to run silently to avoid cluttering the test
results.

Suppresses the following from targets:

for DIR in functional; do               \
        BUILD_TARGET=./tools/testing/selftests/futex/$DIR; \
        mkdir $BUILD_TARGET  -p;        \
        make OUTPUT=$BUILD_TARGET -C $DIR all;\
done

./tools/testing/selftests/futex/run.sh

Signed-off-by: Shuah Khan <shuahkh@osg.samsung.com>
Reviewed-by: Darren Hart (VMware) <dvhart@infradead.org>
Signed-off-by: Shuah Khan <shuahkh@osg.samsung.com>
7 years agoselftests: Makefile: fix for loops in targets to run silently
Shuah Khan [Mon, 18 Sep 2017 23:30:50 +0000 (17:30 -0600)]
selftests: Makefile: fix for loops in targets to run silently

Fix for loops in targets to run silently to avoid cluttering the test
results.

Suppresses the following from targets: e.g run from breakpoints

for TARGET in breakpoints; do \
BUILD_TARGET=$BUILD/$TARGET; \
mkdir $BUILD_TARGET  -p; \
make OUTPUT=$BUILD_TARGET -C $TARGET;\
done;

Signed-off-by: Shuah Khan <shuahkh@osg.samsung.com>
7 years agoselftests: mqueue: Use full path to run tests from Makefile
Shuah Khan [Tue, 12 Sep 2017 14:52:13 +0000 (08:52 -0600)]
selftests: mqueue: Use full path to run tests from Makefile

Use full path including $(OUTPUT) to run tests from Makefile for
normal case when objects reside in the source tree as well as when
objects are relocated with make O=dir. In both cases $(OUTPUT) will
be set correctly by lib.mk.

Signed-off-by: Shuah Khan <shuahkh@osg.samsung.com>
7 years agoselftests: futex: copy sub-dir test scripts for make O=dir run
Shuah Khan [Tue, 12 Sep 2017 01:11:07 +0000 (19:11 -0600)]
selftests: futex: copy sub-dir test scripts for make O=dir run

For make O=dir run_tests to work, test scripts from sub-directories
need to be copied over to the object directory. Running tests from the
object directory is necessary to avoid making the source tree dirty.

Signed-off-by: Shuah Khan <shuahkh@osg.samsung.com>
Reviewed-by: Darren Hart (VMware) <dvhart@infradead.org>
Signed-off-by: Shuah Khan <shuahkh@osg.samsung.com>
7 years agoIB/mlx5: Fix NULL deference on mlx5_ib_update_xlt failure
Ilya Lesokhin [Sun, 24 Sep 2017 18:46:35 +0000 (21:46 +0300)]
IB/mlx5: Fix NULL deference on mlx5_ib_update_xlt failure

mlx5_ib_reg_user_mr called mlx5_ib_dereg_mr in case of MR population
failure. This resulted in a NULL dereference as ibmr->device wasn't
initialized yet.

We address this by adding an internal dereg_mr function that can handle
partially initialized MRs, and fixing clean_mr to work on partially
initialized MRs.

Fixes: ff740aefecb9 ("IB/mlx5: Decouple MR allocation and population flows")
Signed-off-by: Ilya Lesokhin <ilyal@mellanox.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoIB/mlx5: Simplify mlx5_ib_cont_pages
Ilya Lesokhin [Sun, 24 Sep 2017 18:46:34 +0000 (21:46 +0300)]
IB/mlx5: Simplify mlx5_ib_cont_pages

The patch simplifies mlx5_ib_cont_pages and fixes the following
issues in the original implementation:

First issues is related to alignment of the PFNs. After the check
base + p != PFN, the alignment of the PFN wasn't checked. So the PFN
sequence 0, 1, 1, 2 would result in a page_shift of 13 even though
the 3rd PFN is not 8KB aligned.

This wasn't actually a bug because it was supported by all the
existing mlx5 compatible device, but we don't want to require
this support in all future devices.

Another issue is because the inner loop didn't advance PFN so
the test "if (base + p != pfn)" always failed for SGE with
len > (1<<page_shift).

Fixes: e126ba97dba9 ("mlx5: Add driver for Mellanox Connect-IB adapters")
Signed-off-by: Ilya Lesokhin <ilyal@mellanox.com>
Reviewed-by: Eli Cohen <eli@mellanox.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoIB/ipoib: Fix inconsistency with free_netdev and free_rdma_netdev
Alex Vesker [Sun, 24 Sep 2017 18:46:33 +0000 (21:46 +0300)]
IB/ipoib: Fix inconsistency with free_netdev and free_rdma_netdev

Call free_rdma_netdev instead of free_netdev each time we want to
release a netdevice. This call is also relevant for future freeing
of offloaded child interfaces.

This patch also adds a missing call for free netdevice when releasing
a parent interface that has child interfaces using ipoib_remove_one.

Fixes: cd565b4b51e5 ('IB/IPoIB: Support acceleration options callbacks')
Signed-off-by: Alex Vesker <valex@mellanox.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoIB/ipoib: Fix sysfs Pkey create<->remove possible deadlock
Shalom Lagziel [Sun, 24 Sep 2017 18:46:32 +0000 (21:46 +0300)]
IB/ipoib: Fix sysfs Pkey create<->remove possible deadlock

A possible ABBA lock can happen with RTNL and vlan_rwsem.
For example:

Flow A: Device Flush
__ipoib_ib_dev_flush
down_read(vlan_rwsem)  // Lock A
ipoib_flush_ah
flush_workqueue(priv->wq)  // Wait for completion
A work on shared WQ (Mcast carrier)
ipoib_mcast_carrier_on_task
while (!rtnl_trylock())         // Wait for lock B

Flow B: Sysfs PKEY delete
ipoib_vlan_delete
lock(RTNL)  // Lock B
down_write(vlan_rwsem)  // Wait for lock A

This can happen with PKEY creates as well. The solution is to release
the RTNL lock in sysfs functions in case it is not possible to lock
VLAN RW semaphore and reset the SYS call.

Fixes: 69956d83267e ("IB/ipoib: Sync between remove_one to sysfs calls that use rtnl_lock")
Signed-off-by: Shalom Lagziel <shaloml@mellanox.com>
Signed-off-by: Alex Vesker <valex@mellanox.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoIB: Correct MR length field to be 64-bit
Parav Pandit [Sun, 24 Sep 2017 18:46:31 +0000 (21:46 +0300)]
IB: Correct MR length field to be 64-bit

The ib_mr->length represents the length of the MR in bytes as per
the IBTA spec 1.3 section 11.2.10.3 (REGISTER PHYSICAL MEMORY REGION).

Currently ib_mr->length field is defined as only 32-bits field.
This might result into truncation and failed WRs of consumers who
registers more than 4GB bytes memory regions and whose WRs accessing
such MRs.

This patch makes the length 64-bit to avoid such truncation.

Cc: Sagi Grimberg <sagi@grimberg.me>
Cc: Chuck Lever <chuck.lever@oracle.com>
Cc: Faisal Latif <faisal.latif@intel.com>
Fixes: 4c67e2bfc8b7 ("IB/core: Introduce new fast registration API")
Signed-off-by: Ilya Lesokhin <ilyal@mellanox.com>
Signed-off-by: Parav Pandit <parav@mellanox.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoIB/core: Fix qp_sec use after free access
Parav Pandit [Sun, 24 Sep 2017 18:46:30 +0000 (21:46 +0300)]
IB/core: Fix qp_sec use after free access

When security_ib_alloc_security fails, qp->qp_sec memory is freed.
However ib_destroy_qp still tries to access this memory which result
in kernel crash. So its initialized to NULL to avoid such access.

Fixes: d291f1a65232 ("IB/core: Enforce PKey security on QPs")
Signed-off-by: Parav Pandit <parav@mellanox.com>
Reviewed-by: Daniel Jurgens <danielj@mellanox.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoIB/core: Fix typo in the name of the tag-matching cap struct
Leon Romanovsky [Sun, 24 Sep 2017 18:46:29 +0000 (21:46 +0300)]
IB/core: Fix typo in the name of the tag-matching cap struct

The tag matching functionality is implemented by mlx5 driver
by extending XRQ, however this internal kernel information was
exposed to user space applications with *xrq* name instead of *tm*.

This patch renames *xrq* to *tm* to handle that.

Fixes: 8d50505ada72 ("IB/uverbs: Expose XRQ capabilities")
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Reviewed-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoblock: fix a crash caused by wrong API
Shaohua Li [Thu, 21 Sep 2017 19:17:16 +0000 (12:17 -0700)]
block: fix a crash caused by wrong API

part_stat_show takes a part device not a disk, so we should use
part_to_disk.

Fixes: d62e26b3ffd2("block: pass in queue to inflight accounting")
Cc: Bart Van Assche <bart.vanassche@wdc.com>
Cc: Omar Sandoval <osandov@fb.com>
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
7 years agofs: Fix page cache inconsistency when mixing buffered and AIO DIO
Lukas Czerner [Thu, 21 Sep 2017 14:16:29 +0000 (08:16 -0600)]
fs: Fix page cache inconsistency when mixing buffered and AIO DIO

Currently when mixing buffered reads and asynchronous direct writes it
is possible to end up with the situation where we have stale data in the
page cache while the new data is already written to disk. This is
permanent until the affected pages are flushed away. Despite the fact
that mixing buffered and direct IO is ill-advised it does pose a thread
for a data integrity, is unexpected and should be fixed.

Fix this by deferring completion of asynchronous direct writes to a
process context in the case that there are mapped pages to be found in
the inode. Later before the completion in dio_complete() invalidate
the pages in question. This ensures that after the completion the pages
in the written area are either unmapped, or populated with up-to-date
data. Also do the same for the iomap case which uses
iomap_dio_complete() instead.

This has a side effect of deferring the completion to a process context
for every AIO DIO that happens on inode that has pages mapped. However
since the consensus is that this is ill-advised practice the performance
implication should not be a problem.

This was based on proposal from Jeff Moyer, thanks!

Reviewed-by: Jan Kara <jack@suse.cz>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Jeff Moyer <jmoyer@redhat.com>
Signed-off-by: Lukas Czerner <lczerner@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
7 years agonvmet: implement valid sqhd values in completions
James Smart [Mon, 18 Sep 2017 16:08:29 +0000 (09:08 -0700)]
nvmet: implement valid sqhd values in completions

To support sqhd, for initiators that are following the spec and
paying attention to sqhd vs their sqtail values:

- add sqhd to struct nvmet_sq
- initialize sqhd to 0 in nvmet_sq_setup
- rather than propagate the 0's-based qsize value from the connect message
  which requires a +1 in every sqhd update, and as nothing else references
  it, convert to 1's-based value in nvmt_sq/cq_setup() calls.
- validate connect message sqsize being non-zero per spec.
- updated assign sqhd for every completion that goes back.

Also remove handling the NULL sq case in __nvmet_req_complete, as it can't
happen with the current code.

Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Max Gurtovoy <maxg@mellanox.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
7 years agonvme-fabrics: Allow 0 as KATO value
Guilherme G. Piccoli [Thu, 14 Sep 2017 16:59:28 +0000 (13:59 -0300)]
nvme-fabrics: Allow 0 as KATO value

Currently, driver code allows user to set 0 as KATO
(Keep Alive TimeOut), but this is not being respected.
This patch enforces the expected behavior.

Signed-off-by: Guilherme G. Piccoli <gpiccoli@linux.vnet.ibm.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
7 years agonvme: allow timed-out ios to retry
James Smart [Thu, 7 Sep 2017 20:18:04 +0000 (13:18 -0700)]
nvme: allow timed-out ios to retry

Currently the nvme_req_needs_retry() applies several checks to see if
a retry is allowed. On of those is whether the current time has exceeded
the start time of the io plus the timeout length. This check, if an io
times out, means there is never a retry allowed for the io. Which means
applications see the io failure.

Remove this check and allow the io to timeout, like it does on other
protocols, and retries to be made.

On the FC transport, a frame can be lost for an individual io, and there
may be no other errors that escalate for the connection/association.
The io will timeout, which causes the transport to escalate into creating
a new association, but the io that timed out, due to this retry logic, has
already failed back to the application and things are hosed.

Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Keith Busch <keith.busch@intel.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
7 years agonvme: stop aer posting if controller state not live
James Smart [Thu, 14 Sep 2017 18:03:09 +0000 (11:03 -0700)]
nvme: stop aer posting if controller state not live

If an nvme async_event command completes, in most cases, a new
async event is posted. However, if the controller enters a
resetting or reconnecting state, there is nothing to block the
scheduled work element from posting the async event again. Nor are
there calls from the transport to stop async events when an
association dies.

In the case of FC, where the association is torn down, the aer must
be aborted on the FC link and completes through the normal job
completion path. Thus the terminated async event ends up being
rescheduled even though the controller isn't in a valid state for
the aer, and the reposting gets the transport into a partially torn
down data structure.

It's possible to hit the scenario on rdma, although much less likely
due to an aer completing right as the association is terminated and
as the association teardown reclaims the blk requests via
nvme_cancel_request() so its immediate, not a link-related action
like on FC.

Fix by putting controller state checks in both the async event
completion routine where it schedules the async event and in the
async event work routine before it calls into the transport. It's
effectively a "stop_async_events()" behavior.  The transport, when
it creates a new association with the subsystem will transition
the state back to live and is already restarting the async event
posting.

Signed-off-by: James Smart <james.smart@broadcom.com>
[hch: remove taking a lock over reading the controller state]
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
7 years agonvme-pci: Print invalid SGL only once
Keith Busch [Fri, 15 Sep 2017 17:05:38 +0000 (13:05 -0400)]
nvme-pci: Print invalid SGL only once

The WARN_ONCE macro returns true if the condition is true, not if the
warn was raised, so we're printing the scatter list every time it's
invalid. This is excessive and makes debugging harder, so this patch
prints it just once.

Signed-off-by: Keith Busch <keith.busch@intel.com>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
7 years agonvme-pci: initialize queue memory before interrupts
Keith Busch [Thu, 14 Sep 2017 17:54:39 +0000 (13:54 -0400)]
nvme-pci: initialize queue memory before interrupts

A spurious interrupt before the nvme driver has initialized the completion
queue may inadvertently cause the driver to believe it has a completion
to process. This may result in a NULL dereference since the nvmeq's tags
are not set at this point.

The patch initializes the host's CQ memory so that a spurious interrupt
isn't mistaken for a real completion.

Signed-off-by: Keith Busch <keith.busch@intel.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
7 years agonvmet-fc: fix failing max io queue connections
James Smart [Mon, 11 Sep 2017 23:16:53 +0000 (16:16 -0700)]
nvmet-fc: fix failing max io queue connections

fc transport is treating NVMET_NR_QUEUES as maximum queue count, e.g.
admin queue plus NVMET_NR_QUEUES-1 io queues.  But NVMET_NR_QUEUES is
the number of io queues, so maximum queue count is really
NVMET_NR_QUEUES+1.

Fix the handling in the target fc transport

Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
7 years agonvme-fc: use transport-specific sgl format
James Smart [Thu, 7 Sep 2017 20:20:24 +0000 (13:20 -0700)]
nvme-fc: use transport-specific sgl format

Sync with NVM Express spec change and FC-NVME 1.18.

FC transport sets SGL type to Transport SGL Data Block Descriptor and
subtype to transport-specific value 0x0A.

Removed the warn-on's on the PRP fields. They are unneeded. They were
to check for values from the upper layer that weren't set right, and
for the most part were fine. But, with Async events, which reuse the
same structure and 2nd time issued the SGL overlay converted them to
the Transport SGL values - the warn-on's were errantly firing.

Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
7 years agonvme: add transport SGL definitions
James Smart [Thu, 7 Sep 2017 20:20:23 +0000 (13:20 -0700)]
nvme: add transport SGL definitions

Add transport SGL defintions from NVMe TP 4008, required for
the final NVMe-FC standard.

Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
7 years agonvme.h: remove FC transport-specific error values
James Smart [Thu, 7 Sep 2017 23:27:25 +0000 (16:27 -0700)]
nvme.h: remove FC transport-specific error values

The NVM express group recinded the reserved range for the transport.
Remove the FC-centric values that had been defined.

Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
7 years agoqla2xxx: remove use of FC-specific error codes
James Smart [Thu, 14 Sep 2017 18:30:15 +0000 (11:30 -0700)]
qla2xxx: remove use of FC-specific error codes

The qla2xxx driver uses the FC-specific error when it needed to return an
error to the FC-NVME transport.  Convert to use a generic value instead.

Signed-off-by: James Smart <james.smart@broadcom.com>
Acked-by: Himanshu Madhani <himanshu.madhani@cavium.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
7 years agolpfc: remove use of FC-specific error codes
James Smart [Thu, 7 Sep 2017 23:27:29 +0000 (16:27 -0700)]
lpfc: remove use of FC-specific error codes

The lpfc driver uses the FC-specific error when it needed to return an
error to the FC-NVME transport. Convert to use a generic value instead.

Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
7 years agonvmet-fcloop: remove use of FC-specific error codes
James Smart [Thu, 7 Sep 2017 23:27:28 +0000 (16:27 -0700)]
nvmet-fcloop: remove use of FC-specific error codes

The FC-NVME transport loopback test module used the FC-specific error
codes in cases where it emulated a transport abort case. Instead of
using the FC-specific values, now use a generic value (NVME_SC_INTERNAL).

Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
7 years agonvmet-fc: remove use of FC-specific error codes
James Smart [Thu, 7 Sep 2017 23:27:27 +0000 (16:27 -0700)]
nvmet-fc: remove use of FC-specific error codes

The FC-NVME target transport used the FC-specific error codes in
return codes when the transport or lldd failed. Instead of using the
FC-specific values, now use a generic value (NVME_SC_INTERNAL).

Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
7 years agonvme-fc: remove use of FC-specific error codes
James Smart [Thu, 7 Sep 2017 23:27:26 +0000 (16:27 -0700)]
nvme-fc: remove use of FC-specific error codes

The FC-NVME transport used the FC-specific error codes in cases where
it had to fabricate an error to go back up stack. Instead of using the
FC-specific values, now use a generic value (NVME_SC_INTERNAL).

Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
7 years agoloop: remove union of use_aio and ref in struct loop_cmd
Omar Sandoval [Wed, 20 Sep 2017 21:24:34 +0000 (14:24 -0700)]
loop: remove union of use_aio and ref in struct loop_cmd

When the request is completed, lo_complete_rq() checks cmd->use_aio.
However, if this is in fact an aio request, cmd->use_aio will have
already been reused as cmd->ref by lo_rw_aio*. Fix it by not using a
union. On x86_64, there's a hole after the union anyways, so this
doesn't make struct loop_cmd any bigger.

Fixes: 92d773324b7e ("block/loop: fix use after free")
Signed-off-by: Omar Sandoval <osandov@fb.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
7 years agoblktrace: Fix potential deadlock between delete & sysfs ops
Waiman Long [Wed, 20 Sep 2017 19:12:20 +0000 (13:12 -0600)]
blktrace: Fix potential deadlock between delete & sysfs ops

The lockdep code had reported the following unsafe locking scenario:

       CPU0                    CPU1
       ----                    ----
  lock(s_active#228);
                               lock(&bdev->bd_mutex/1);
                               lock(s_active#228);
  lock(&bdev->bd_mutex);

 *** DEADLOCK ***

The deadlock may happen when one task (CPU1) is trying to delete a
partition in a block device and another task (CPU0) is accessing
tracing sysfs file (e.g. /sys/block/dm-1/trace/act_mask) in that
partition.

The s_active isn't an actual lock. It is a reference count (kn->count)
on the sysfs (kernfs) file. Removal of a sysfs file, however, require
a wait until all the references are gone. The reference count is
treated like a rwsem using lockdep instrumentation code.

The fact that a thread is in the sysfs callback method or in the
ioctl call means there is a reference to the opended sysfs or device
file. That should prevent the underlying block structure from being
removed.

Instead of using bd_mutex in the block_device structure, a new
blk_trace_mutex is now added to the request_queue structure to protect
access to the blk_trace structure.

Suggested-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Waiman Long <longman@redhat.com>
Acked-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Fix typo in patch subject line, and prune a comment detailing how
the code used to work.

Signed-off-by: Jens Axboe <axboe@kernel.dk>
7 years agonbd: ignore non-nbd ioctl's
Josef Bacik [Sat, 6 May 2017 02:25:18 +0000 (22:25 -0400)]
nbd: ignore non-nbd ioctl's

In testing we noticed that nbd would spew if you ran a fio job against
the raw device itself.  This is because fio calls a block device
specific ioctl, however the block layer will first pass this back to the
driver ioctl handler in case the driver wants to do something special.
Since the device was setup using netlink this caused us to spew every
time fio called this ioctl.  Since we don't have special handling, just
error out for any non-nbd specific ioctl's that come in.  This fixes the
spew.

Signed-off-by: Josef Bacik <jbacik@fb.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
7 years agobsg-lib: don't free job in bsg_prepare_job
Christoph Hellwig [Thu, 7 Sep 2017 11:54:35 +0000 (13:54 +0200)]
bsg-lib: don't free job in bsg_prepare_job

The job structure is allocated as part of the request, so we should not
free it in the error path of bsg_prepare_job.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Cc: stable@vger.kernel.org
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
7 years agobrd: fix overflow in __brd_direct_access
Mikulas Patocka [Wed, 13 Sep 2017 13:17:57 +0000 (09:17 -0400)]
brd: fix overflow in __brd_direct_access

The code in __brd_direct_access multiplies the pgoff variable by page size
and divides it by 512. It can cause overflow on 32-bit architectures. The
overflow happens if we create ramdisk larger than 4G and use it as a
sparse device.

This patch replaces multiplication and division with multiplication by the
number of sectors per page.

Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Fixes: 1647b9b959c7 ("brd: add dax_operations support")
Cc: stable@vger.kernel.org # 4.12+
Signed-off-by: Jens Axboe <axboe@kernel.dk>
7 years agoarch: change default endian for microblaze microblaze-4.14-rc3
Babu Moger [Mon, 18 Sep 2017 16:53:29 +0000 (10:53 -0600)]
arch: change default endian for microblaze

Fix the default for microblaze. Michal Simek mentioned default for
microblaze should be CPU_LITTLE_ENDIAN.

Fixes : commit 206d3642d8ee ("arch/microblaze: add choice for endianness
and update Makefile")

Signed-off-by: Babu Moger <babu.moger@oracle.com>
Cc: Michal Simek <monstr@monstr.eu>
Signed-off-by: Michal Simek <michal.simek@xilinx.com>
7 years agomicroblaze: Cocci spatch "vma_pages"
Thomas Meyer [Wed, 20 Sep 2017 22:29:36 +0000 (00:29 +0200)]
microblaze: Cocci spatch "vma_pages"

Use vma_pages function on vma object instead of explicit computation.
Found by coccinelle spatch "api/vma_pages.cocci"

Signed-off-by: Thomas Meyer <thomas@m3y3r.de>
Signed-off-by: Michal Simek <michal.simek@xilinx.com>
7 years agomicroblaze: Add missing kvm_para.h to Kbuild
Michal Simek [Tue, 19 Sep 2017 14:40:21 +0000 (16:40 +0200)]
microblaze: Add missing kvm_para.h to Kbuild

Running make allmodconfig;make is throwing compilation error:
  CC      kernel/watchdog.o
In file included from ./include/linux/kvm_para.h:4:0,
                 from kernel/watchdog.c:29:
./include/uapi/linux/kvm_para.h:32:26: fatal error: asm/kvm_para.h: No
such file or directory
 #include <asm/kvm_para.h>
                          ^
compilation terminated.
make[1]: *** [kernel/watchdog.o] Error 1
make: *** [kernel/watchdog.o] Error 2

Reported-by: Michal Hocko <mhocko@kernel.org>
Suggested-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Fixes: 83f0124ad81e87b ("microblaze: remove asm-generic wrapper headers")
Reviewed-by: Tobias Klauser <tklauser@distanz.ch>
Tested-by: Michal Hocko <mhocko@suse.com>
7 years agox86/fpu: Reinitialize FPU registers if restoring FPU state fails
Eric Biggers [Sat, 23 Sep 2017 13:00:09 +0000 (15:00 +0200)]
x86/fpu: Reinitialize FPU registers if restoring FPU state fails

Userspace can change the FPU state of a task using the ptrace() or
rt_sigreturn() system calls.  Because reserved bits in the FPU state can
cause the XRSTOR instruction to fail, the kernel has to carefully
validate that no reserved bits or other invalid values are being set.

Unfortunately, there have been bugs in this validation code.  For
example, we were not checking that the 'xcomp_bv' field in the
xstate_header was 0.  As-is, such bugs are exploitable to read the FPU
registers of other processes on the system.  To do so, an attacker can
create a task, assign to it an invalid FPU state, then spin in a loop
and monitor the values of the FPU registers.  Because the task's FPU
registers are not being restored, sometimes the FPU registers will have
the values from another process.

This is likely to continue to be a problem in the future because the
validation done by the CPU instructions like XRSTOR is not immediately
visible to kernel developers.  Nor will invalid FPU states ever be
encountered during ordinary use --- they will only be seen during
fuzzing or exploits.  There can even be reserved bits outside the
xstate_header which are easy to forget about.  For example, the MXCSR
register contains reserved bits, which were not validated by the
KVM_SET_XSAVE ioctl until commit a575813bfe4b ("KVM: x86: Fix load
damaged SSEx MXCSR register").

Therefore, mitigate this class of vulnerability by restoring the FPU
registers from init_fpstate if restoring from the task's state fails.

We actually used to do this, but it was (perhaps unwisely) removed by
commit 9ccc27a5d297 ("x86/fpu: Remove error return values from
copy_kernel_to_*regs() functions").  This new patch is also a bit
different.  First, it only clears the registers, not also the bad
in-memory state; this is simpler and makes it easier to make the
mitigation cover all callers of __copy_kernel_to_fpregs().  Second, it
does the register clearing in an exception handler so that no extra
instructions are added to context switches.  In fact, we *remove*
instructions, since previously we were always zeroing the register
containing 'err' even if CONFIG_X86_DEBUG_FPU was disabled.

Signed-off-by: Eric Biggers <ebiggers@google.com>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Eric Biggers <ebiggers3@gmail.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: Kevin Hao <haokexin@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Michael Halcrow <mhalcrow@google.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Wanpeng Li <wanpeng.li@hotmail.com>
Cc: Yu-cheng Yu <yu-cheng.yu@intel.com>
Cc: kernel-hardening@lists.openwall.com
Link: http://lkml.kernel.org/r/20170922174156.16780-4-ebiggers3@gmail.com
Link: http://lkml.kernel.org/r/20170923130016.21448-27-mingo@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
7 years agox86/fpu: Don't let userspace set bogus xcomp_bv
Eric Biggers [Sat, 23 Sep 2017 13:00:07 +0000 (15:00 +0200)]
x86/fpu: Don't let userspace set bogus xcomp_bv

On x86, userspace can use the ptrace() or rt_sigreturn() system calls to
set a task's extended state (xstate) or "FPU" registers.  ptrace() can
set them for another task using the PTRACE_SETREGSET request with
NT_X86_XSTATE, while rt_sigreturn() can set them for the current task.
In either case, registers can be set to any value, but the kernel
assumes that the XSAVE area itself remains valid in the sense that the
CPU can restore it.

However, in the case where the kernel is using the uncompacted xstate
format (which it does whenever the XSAVES instruction is unavailable),
it was possible for userspace to set the xcomp_bv field in the
xstate_header to an arbitrary value.  However, all bits in that field
are reserved in the uncompacted case, so when switching to a task with
nonzero xcomp_bv, the XRSTOR instruction failed with a #GP fault.  This
caused the WARN_ON_FPU(err) in copy_kernel_to_xregs() to be hit.  In
addition, since the error is otherwise ignored, the FPU registers from
the task previously executing on the CPU were leaked.

Fix the bug by checking that the user-supplied value of xcomp_bv is 0 in
the uncompacted case, and returning an error otherwise.

The reason for validating xcomp_bv rather than simply overwriting it
with 0 is that we want userspace to see an error if it (incorrectly)
provides an XSAVE area in compacted format rather than in uncompacted
format.

Note that as before, in case of error we clear the task's FPU state.
This is perhaps non-ideal, especially for PTRACE_SETREGSET; it might be
better to return an error before changing anything.  But it seems the
"clear on error" behavior is fine for now, and it's a little tricky to
do otherwise because it would mean we couldn't simply copy the full
userspace state into kernel memory in one __copy_from_user().

This bug was found by syzkaller, which hit the above-mentioned
WARN_ON_FPU():

    WARNING: CPU: 1 PID: 0 at ./arch/x86/include/asm/fpu/internal.h:373 __switch_to+0x5b5/0x5d0
    CPU: 1 PID: 0 Comm: swapper/1 Not tainted 4.13.0 #453
    Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011
    task: ffff9ba2bc8e42c0 task.stack: ffffa78cc036c000
    RIP: 0010:__switch_to+0x5b5/0x5d0
    RSP: 0000:ffffa78cc08bbb88 EFLAGS: 00010082
    RAX: 00000000fffffffe RBX: ffff9ba2b8bf2180 RCX: 00000000c0000100
    RDX: 00000000ffffffff RSI: 000000005cb10700 RDI: ffff9ba2b8bf36c0
    RBP: ffffa78cc08bbbd0 R08: 00000000929fdf46 R09: 0000000000000001
    R10: 0000000000000000 R11: 0000000000000000 R12: ffff9ba2bc8e42c0
    R13: 0000000000000000 R14: ffff9ba2b8bf3680 R15: ffff9ba2bf5d7b40
    FS:  00007f7e5cb10700(0000) GS:ffff9ba2bf400000(0000) knlGS:0000000000000000
    CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
    CR2: 00000000004005cc CR3: 0000000079fd5000 CR4: 00000000001406e0
    Call Trace:
    Code: 84 00 00 00 00 00 e9 11 fd ff ff 0f ff 66 0f 1f 84 00 00 00 00 00 e9 e7 fa ff ff 0f ff 66 0f 1f 84 00 00 00 00 00 e9 c2 fa ff ff <0f> ff 66 0f 1f 84 00 00 00 00 00 e9 d4 fc ff ff 66 66 2e 0f 1f

Here is a C reproducer.  The expected behavior is that the program spin
forever with no output.  However, on a buggy kernel running on a
processor with the "xsave" feature but without the "xsaves" feature
(e.g. Sandy Bridge through Broadwell for Intel), within a second or two
the program reports that the xmm registers were corrupted, i.e. were not
restored correctly.  With CONFIG_X86_DEBUG_FPU=y it also hits the above
kernel warning.

    #define _GNU_SOURCE
    #include <stdbool.h>
    #include <inttypes.h>
    #include <linux/elf.h>
    #include <stdio.h>
    #include <sys/ptrace.h>
    #include <sys/uio.h>
    #include <sys/wait.h>
    #include <unistd.h>

    int main(void)
    {
        int pid = fork();
        uint64_t xstate[512];
        struct iovec iov = { .iov_base = xstate, .iov_len = sizeof(xstate) };

        if (pid == 0) {
            bool tracee = true;
            for (int i = 0; i < sysconf(_SC_NPROCESSORS_ONLN) && tracee; i++)
                tracee = (fork() != 0);
            uint32_t xmm0[4] = { [0 ... 3] = tracee ? 0x00000000 : 0xDEADBEEF };
            asm volatile("   movdqu %0, %%xmm0\n"
                         "   mov %0, %%rbx\n"
                         "1: movdqu %%xmm0, %0\n"
                         "   mov %0, %%rax\n"
                         "   cmp %%rax, %%rbx\n"
                         "   je 1b\n"
                         : "+m" (xmm0) : : "rax", "rbx", "xmm0");
            printf("BUG: xmm registers corrupted!  tracee=%d, xmm0=%08X%08X%08X%08X\n",
                   tracee, xmm0[0], xmm0[1], xmm0[2], xmm0[3]);
        } else {
            usleep(100000);
            ptrace(PTRACE_ATTACH, pid, 0, 0);
            wait(NULL);
            ptrace(PTRACE_GETREGSET, pid, NT_X86_XSTATE, &iov);
            xstate[65] = -1;
            ptrace(PTRACE_SETREGSET, pid, NT_X86_XSTATE, &iov);
            ptrace(PTRACE_CONT, pid, 0, 0);
            wait(NULL);
        }
        return 1;
    }

Note: the program only tests for the bug using the ptrace() system call.
The bug can also be reproduced using the rt_sigreturn() system call, but
only when called from a 32-bit program, since for 64-bit programs the
kernel restores the FPU state from the signal frame by doing XRSTOR
directly from userspace memory (with proper error checking).

Reported-by: Dmitry Vyukov <dvyukov@google.com>
Signed-off-by: Eric Biggers <ebiggers@google.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Rik van Riel <riel@redhat.com>
Acked-by: Dave Hansen <dave.hansen@linux.intel.com>
Cc: <stable@vger.kernel.org> [v3.17+]
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Eric Biggers <ebiggers3@gmail.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: Kevin Hao <haokexin@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Michael Halcrow <mhalcrow@google.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Wanpeng Li <wanpeng.li@hotmail.com>
Cc: Yu-cheng Yu <yu-cheng.yu@intel.com>
Cc: kernel-hardening@lists.openwall.com
Fixes: 0b29643a5843 ("x86/xsaves: Change compacted format xsave area header")
Link: http://lkml.kernel.org/r/20170922174156.16780-2-ebiggers3@gmail.com
Link: http://lkml.kernel.org/r/20170923130016.21448-25-mingo@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
7 years agoLinux 4.14-rc2
Linus Torvalds [Sun, 24 Sep 2017 23:38:56 +0000 (16:38 -0700)]
Linux 4.14-rc2

7 years agoMerge tag 'devicetree-fixes-for-4.14' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 24 Sep 2017 23:04:12 +0000 (16:04 -0700)]
Merge tag 'devicetree-fixes-for-4.14' of git://git./linux/kernel/git/robh/linux

Pull DeviceTree fixes from Rob Herring:

 - fix build for !OF providing empty of_find_device_by_node

 - fix Abracon vendor prefix

 - sync dtx_diff include paths (again)

 - a stm32h7 clock binding doc fix

* tag 'devicetree-fixes-for-4.14' of git://git.kernel.org/pub/scm/linux/kernel/git/robh/linux:
  dt-bindings: clk: stm32h7: fix clock-cell size
  scripts/dtc: dtx_diff - 2nd update of include dts paths to match build
  dt-bindings: fix vendor prefix for Abracon
  of: provide inline helper for of_find_device_by_node

7 years agoMerge branch 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 24 Sep 2017 19:33:58 +0000 (12:33 -0700)]
Merge branch 'x86-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull x86 fixes from Ingo Molnar:
 "Another round of CR3/PCID related fixes (I think this addresses all
  but one of the known problems with PCID support), an objtool fix plus
  a Clang fix that (finally) solves all Clang quirks to build a bootable
  x86 kernel as-is"

* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/asm: Fix inline asm call constraints for Clang
  objtool: Handle another GCC stack pointer adjustment bug
  x86/mm/32: Load a sane CR3 before cpu_init() on secondary CPUs
  x86/mm/32: Move setup_clear_cpu_cap(X86_FEATURE_PCID) earlier
  x86/mm/64: Stop using CR3.PCID == 0 in ASID-aware code
  x86/mm: Factor out CR3-building code

7 years agoMerge branch 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 24 Sep 2017 19:28:55 +0000 (12:28 -0700)]
Merge branch 'timers-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull timer fix from Ingo Molnar:
 "A clocksource driver section mismatch fix"

* 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  clocksource/integrator: Fix section mismatch warning

7 years agoMerge branch 'irq-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 24 Sep 2017 18:57:07 +0000 (11:57 -0700)]
Merge branch 'irq-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull irq fixes from Ingo Molnar:
 "Three irqchip driver fixes, and an affinity mask helper function bug
  fix affecting x86"

* 'irq-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  Revert "genirq: Restrict effective affinity to interrupts actually using it"
  irqchip.mips-gic: Fix shared interrupt mask writes
  irqchip/gic-v4: Fix building with ancient gcc
  irqchip/gic-v3: Iterate over possible CPUs by for_each_possible_cpu()

7 years agoMerge branch 'core-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 24 Sep 2017 18:53:13 +0000 (11:53 -0700)]
Merge branch 'core-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull address-limit checking fixes from Ingo Molnar:
 "This fixes a number of bugs in the address-limit (USER_DS) checks that
  got introduced in the merge window, (mostly) affecting the ARM and
  ARM64 platforms"

* 'core-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  arm64/syscalls: Move address limit check in loop
  arm/syscalls: Optimize address limit check
  Revert "arm/syscalls: Check address limit on user-mode return"
  syscalls: Use CHECK_DATA_CORRUPTION for addr_limit_user_check

7 years agoMerge branch 'next-general' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris...
Linus Torvalds [Sun, 24 Sep 2017 18:40:41 +0000 (11:40 -0700)]
Merge branch 'next-general' of git://git./linux/kernel/git/jmorris/linux-security

Pull misc security layer update from James Morris:
 "This is the remaining 'general' change in the security tree for v4.14,
  following the direct merging of SELinux (+ TOMOYO), AppArmor, and
  seccomp.

  That's everything now for the security tree except IMA, which will
  follow shortly (I've been traveling for the past week with patchy
  internet)"

* 'next-general' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris/linux-security:
  security: fix description of values returned by cap_inode_need_killpriv

7 years agoMerge branch 'next-tpm' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris...
Linus Torvalds [Sun, 24 Sep 2017 18:34:28 +0000 (11:34 -0700)]
Merge branch 'next-tpm' of git://git./linux/kernel/git/jmorris/linux-security

Pull TPM updates from James Morris:
 "Here are the TPM updates from Jarkko for v4.14, which I've placed in
  their own branch (next-tpm). I ended up cherry-picking them as other
  changes had been made in Jarkko's branch after he sent me his original
  pull request.

  I plan on maintaining a separate branch for TPM (and other security
  subsystems) from now on.

  From Jarkko: 'Not much this time except a few fixes'"

* 'next-tpm' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris/linux-security:
  tpm: ibmvtpm: simplify crq initialization and document crq format
  tpm: replace msleep() with  usleep_range() in TPM 1.2/2.0 generic drivers
  Documentation: tpm: add powered-while-suspended binding documentation
  tpm: tpm_crb: constify acpi_device_id.
  tpm: vtpm: constify vio_device_id