]> git.proxmox.com Git - mirror_ubuntu-eoan-kernel.git/log
mirror_ubuntu-eoan-kernel.git
6 years agoMerge branch 'work.dcache' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs
Linus Torvalds [Mon, 4 Jun 2018 15:57:36 +0000 (08:57 -0700)]
Merge branch 'work.dcache' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs

Pull dcache updates from Al Viro:
 "This is the first part of dealing with livelocks etc around
  shrink_dcache_parent()."

* 'work.dcache' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
  restore cond_resched() in shrink_dcache_parent()
  dput(): turn into explicit while() loop
  dcache: move cond_resched() into the end of __dentry_kill()
  d_walk(): kill 'finish' callback
  d_invalidate(): unhash immediately

6 years agoMerge tag 'for-4.18/block-20180603' of git://git.kernel.dk/linux-block
Linus Torvalds [Mon, 4 Jun 2018 14:58:06 +0000 (07:58 -0700)]
Merge tag 'for-4.18/block-20180603' of git://git.kernel.dk/linux-block

Pull block updates from Jens Axboe:

 - clean up how we pass around gfp_t and
   blk_mq_req_flags_t (Christoph)

 - prepare us to defer scheduler attach (Christoph)

 - clean up drivers handling of bounce buffers (Christoph)

 - fix timeout handling corner cases (Christoph/Bart/Keith)

 - bcache fixes (Coly)

 - prep work for bcachefs and some block layer optimizations (Kent).

 - convert users of bio_sets to using embedded structs (Kent).

 - fixes for the BFQ io scheduler (Paolo/Davide/Filippo)

 - lightnvm fixes and improvements (Matias, with contributions from Hans
   and Javier)

 - adding discard throttling to blk-wbt (me)

 - sbitmap blk-mq-tag handling (me/Omar/Ming).

 - remove the sparc jsflash block driver, acked by DaveM.

 - Kyber scheduler improvement from Jianchao, making it more friendly
   wrt merging.

 - conversion of symbolic proc permissions to octal, from Joe Perches.
   Previously the block parts were a mix of both.

 - nbd fixes (Josef and Kevin Vigor)

 - unify how we handle the various kinds of timestamps that the block
   core and utility code uses (Omar)

 - three NVMe pull requests from Keith and Christoph, bringing AEN to
   feature completeness, file backed namespaces, cq/sq lock split, and
   various fixes

 - various little fixes and improvements all over the map

* tag 'for-4.18/block-20180603' of git://git.kernel.dk/linux-block: (196 commits)
  blk-mq: update nr_requests when switching to 'none' scheduler
  block: don't use blocking queue entered for recursive bio submits
  dm-crypt: fix warning in shutdown path
  lightnvm: pblk: take bitmap alloc. out of critical section
  lightnvm: pblk: kick writer on new flush points
  lightnvm: pblk: only try to recover lines with written smeta
  lightnvm: pblk: remove unnecessary bio_get/put
  lightnvm: pblk: add possibility to set write buffer size manually
  lightnvm: fix partial read error path
  lightnvm: proper error handling for pblk_bio_add_pages
  lightnvm: pblk: fix smeta write error path
  lightnvm: pblk: garbage collect lines with failed writes
  lightnvm: pblk: rework write error recovery path
  lightnvm: pblk: remove dead function
  lightnvm: pass flag on graceful teardown to targets
  lightnvm: pblk: check for chunk size before allocating it
  lightnvm: pblk: remove unnecessary argument
  lightnvm: pblk: remove unnecessary indirection
  lightnvm: pblk: return NVM_ error on failed submission
  lightnvm: pblk: warn in case of corrupted write buffer
  ...

6 years agoLinux 4.17
Linus Torvalds [Sun, 3 Jun 2018 21:15:21 +0000 (14:15 -0700)]
Linux 4.17

6 years agoMerge branch 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs
Linus Torvalds [Sun, 3 Jun 2018 18:01:28 +0000 (11:01 -0700)]
Merge branch 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs

Pull vfs fixes from Al Viro.

 - fix io_destroy()/aio_complete() race

 - the vfs_open() change to get rid of open_check_o_direct() boilerplate
   was nice, but buggy. Al has a patch avoiding a revert, but that's
   definitely not a last-day fodder, so for now revert it is...

* 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
  Revert "fs: fold open_check_o_direct into do_dentry_open"
  fix io_destroy()/aio_complete() race

6 years agoRevert "fs: fold open_check_o_direct into do_dentry_open"
Al Viro [Sat, 2 Jun 2018 05:31:02 +0000 (01:31 -0400)]
Revert "fs: fold open_check_o_direct into do_dentry_open"

This reverts commit cab64df194667dc5d9d786f0a895f647f5501c0d.

Having vfs_open() in some cases drop the reference to
struct file combined with

error = vfs_open(path, f, cred);
if (error) {
put_filp(f);
return ERR_PTR(error);
}
return f;

is flat-out wrong.  It used to be

error = vfs_open(path, f, cred);
if (!error) {
/* from now on we need fput() to dispose of f */
error = open_check_o_direct(f);
if (error) {
fput(f);
f = ERR_PTR(error);
}
} else {
put_filp(f);
f = ERR_PTR(error);
}

and sure, having that open_check_o_direct() boilerplate gotten rid of is
nice, but not that way...

Worse, another call chain (via finish_open()) is FUBAR now wrt
FILE_OPENED handling - in that case we get error returned, with file
already hit by fput() *AND* FILE_OPENED not set.  Guess what happens in
path_openat(), when it hits

if (!(opened & FILE_OPENED)) {
BUG_ON(!error);
put_filp(file);
}

The root cause of all that crap is that the callers of do_dentry_open()
have no way to tell which way did it fail; while that could be fixed up
(by passing something like int *opened to do_dentry_open() and have it
marked if we'd called ->open()), it's probably much too late in the
cycle to do so right now.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agoMerge branch 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 3 Jun 2018 16:01:41 +0000 (09:01 -0700)]
Merge branch 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull scheduler fixes from Thomas Gleixner:

 - two patches addressing the problem that the scheduler allows under
   certain conditions user space tasks to be scheduled on CPUs which are
   not yet fully booted which causes a few subtle and hard to debug
   issue

 - add a missing runqueue clock update in the deadline scheduler which
   triggers a warning under certain circumstances

 - fix a silly typo in the scheduler header file

* 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  sched/headers: Fix typo
  sched/deadline: Fix missing clock update
  sched/core: Require cpu_active() in select_task_rq(), for user tasks
  sched/core: Fix rules for running on online && !active CPUs

6 years agoMerge branch 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 3 Jun 2018 15:58:59 +0000 (08:58 -0700)]
Merge branch 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull perf tooling fixes from Thomas Gleixner:

 - fix 'perf test Session topology' segfault on s390 (Thomas Richter)

 - fix NULL return handling in bpf__prepare_load() (YueHaibing)

 - fix indexing on Coresight ETM packet queue decoder (Mathieu Poirier)

 - fix perf.data format description of NRCPUS header (Arnaldo Carvalho
   de Melo)

 - update perf.data documentation section on cpu topology

 - handle uncore event aliases in small groups properly (Kan Liang)

 - add missing perf_sample.addr into python sample dictionary (Leo Yan)

* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  perf tools: Fix perf.data format description of NRCPUS header
  perf script python: Add addr into perf sample dict
  perf data: Update documentation section on cpu topology
  perf cs-etm: Fix indexing for decoder packet queue
  perf bpf: Fix NULL return handling in bpf__prepare_load()
  perf test: "Session topology" dumps core on s390
  perf parse-events: Handle uncore event aliases in small groups properly

6 years agoblk-mq: update nr_requests when switching to 'none' scheduler
Ming Lei [Sat, 2 Jun 2018 07:18:09 +0000 (15:18 +0800)]
blk-mq: update nr_requests when switching to 'none' scheduler

Now we setup q->nr_requests when switching to one new scheduler,
but not do it for 'none', then q->nr_requests may not be correct
for 'none'.

This patch fixes this issue by always updating 'nr_requests' when
switching to 'none'.

Cc: Marco Patalano <mpatalan@redhat.com>
Cc: "Ewan D. Milne" <emilne@redhat.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoblock: don't use blocking queue entered for recursive bio submits
Jens Axboe [Sat, 2 Jun 2018 20:04:07 +0000 (14:04 -0600)]
block: don't use blocking queue entered for recursive bio submits

If we end up splitting a bio and the queue goes away between
the initial submission and the later split submission, then we
can block forever in blk_queue_enter() waiting for the reference
to drop to zero. This will never happen, since we already hold
a reference.

Mark a split bio as already having entered the queue, so we can
just use the live non-blocking queue enter variant.

Thanks to Tetsuo Handa for the analysis.

Reported-by: syzbot+c4f9cebf9d651f6e54de@syzkaller.appspotmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agodm-crypt: fix warning in shutdown path
Kent Overstreet [Sat, 2 Jun 2018 17:45:04 +0000 (13:45 -0400)]
dm-crypt: fix warning in shutdown path

The counter for the number of allocated pages includes pages in the
mempool's reserve, so checking that the number of allocated pages is 0
needs to happen after we exit the mempool.

Fixes: 6f1c819c219f ("dm: convert to bioset_init()/mempool_init()")
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Reported-by: Krzysztof Kozlowski <krzk@kernel.org>
Acked-by: Mike Snitzer <snitzer@redhat.com>
Fixed to always just use percpu_counter_sum()

Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
Linus Torvalds [Sun, 3 Jun 2018 00:35:53 +0000 (17:35 -0700)]
Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net

Pull networking fixes from David Miller:

 1) Infinite loop in _decode_session6(), from Eric Dumazet.

 2) Pass correct argument to nla_strlcpy() in netfilter, also from Eric
    Dumazet.

 3) Out of bounds memory access in ipv6 srh code, from Mathieu Xhonneux.

 4) NULL deref in XDP_REDIRECT handling of tun driver, from Toshiaki
    Makita.

 5) Incorrect idr release in cls_flower, from Paul Blakey.

 6) Probe error handling fix in davinci_emac, from Dan Carpenter.

 7) Memory leak in XPS configuration, from Alexander Duyck.

 8) Use after free with cloned sockets in kcm, from Kirill Tkhai.

 9) MTU handling fixes fo ip_tunnel and ip6_tunnel, from Nicolas
    Dichtel.

10) Fix UAPI hole in bpf data structure for 32-bit compat applications,
    from Daniel Borkmann.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (33 commits)
  bpf: fix uapi hole for 32 bit compat applications
  net: usb: cdc_mbim: add flag FLAG_SEND_ZLP
  ip6_tunnel: remove magic mtu value 0xFFF8
  ip_tunnel: restore binding to ifaces with a large mtu
  net: dsa: b53: Add BCM5389 support
  kcm: Fix use-after-free caused by clonned sockets
  net-sysfs: Fix memory leak in XPS configuration
  ixgbe: fix parsing of TC actions for HW offload
  net: ethernet: davinci_emac: fix error handling in probe()
  net/ncsi: Fix array size in dumpit handler
  cls_flower: Fix incorrect idr release when failing to modify rule
  net/sonic: Use dma_mapping_error()
  xfrm Fix potential error pointer dereference in xfrm_bundle_create.
  vhost_net: flush batched heads before trying to busy polling
  tun: Fix NULL pointer dereference in XDP redirect
  be2net: Fix error detection logic for BE3
  net: qmi_wwan: Add Netgear Aircard 779S
  mlxsw: spectrum: Forbid creation of VLAN 1 over port/LAG
  atm: zatm: fix memcmp casting
  iwlwifi: pcie: compare with number of IRQs requested for, not number of CPUs
  ...

6 years agoMerge tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi
Linus Torvalds [Sat, 2 Jun 2018 22:54:49 +0000 (15:54 -0700)]
Merge tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi

Pull SCSI fix from James Bottomley:
 "Eve of merge window fix: The original code was so bogus as to be
  casting the wrong generic device to an rport and proceeding to take
  actions based on the bogus values it found.

  Fortunately it seems the location that is dereferenced always exists,
  so the code hasn't oopsed yet, but it certainly annoys the memory
  checkers"

* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
  scsi: scsi_transport_srp: Fix shost to rport translation

6 years agoMerge tag 'drm-fixes-for-v4.17-rc8' of git://people.freedesktop.org/~airlied/linux
Linus Torvalds [Sat, 2 Jun 2018 22:24:45 +0000 (15:24 -0700)]
Merge tag 'drm-fixes-for-v4.17-rc8' of git://people.freedesktop.org/~airlied/linux

Pull drm fixes from Dave Airlie:
 "A few final fixes:

  i915:
   - fix for potential Spectre vector in the new query uAPI
   - fix NULL pointer deref (FDO #106559)
   - DMI fix to hide LVDS for Radiant P845 (FDO #105468)

  amdgpu:
   - suspend/resume DC regression fix
   - underscan flicker fix on fiji
   - gamma setting fix after dpms

  omap:
   - fix oops regression

  core:
   - fix PSR timing

  dw-hdmi:
   - fix oops regression"

* tag 'drm-fixes-for-v4.17-rc8' of git://people.freedesktop.org/~airlied/linux:
  drm/amd/display: Update color props when modeset is required
  drm/amd/display: Make atomic-check validate underscan changes
  drm/bridge/synopsys: dw-hdmi: fix dw_hdmi_setup_rx_sense
  drm/amd/display: Fix BUG_ON during CRTC atomic check update
  drm/i915/query: nospec expects no more than an unsigned long
  drm/i915/query: Protect tainted function pointer lookup
  drm/i915/lvds: Move acpi lid notification registration to registration phase
  drm/i915: Disable LVDS on Radiant P845
  drm/omap: fix NULL deref crash with SDI displays
  drm/psr: Fix missed entry in PSR setup time table.

6 years agoMerge branch 'drm-fixes-4.17' of git://people.freedesktop.org/~agd5f/linux into drm...
Dave Airlie [Sat, 2 Jun 2018 20:13:57 +0000 (06:13 +1000)]
Merge branch 'drm-fixes-4.17' of git://people.freedesktop.org/~agd5f/linux into drm-fixes

Two last minute DC fixes for 4.17.  A fix for underscan on fiji and
a fix for gamma settings getting after dpms.

* 'drm-fixes-4.17' of git://people.freedesktop.org/~agd5f/linux:
  drm/amd/display: Update color props when modeset is required
  drm/amd/display: Make atomic-check validate underscan changes

6 years agoMerge tag 'mips_fixes_4.17_3' of git://git.kernel.org/pub/scm/linux/kernel/git/mips...
Linus Torvalds [Sat, 2 Jun 2018 17:12:23 +0000 (10:12 -0700)]
Merge tag 'mips_fixes_4.17_3' of git://git.kernel.org/pub/scm/linux/kernel/git/mips/linux

Pull MIPS fixes from James Hogan:
 "A final few MIPS fixes for 4.17:

   - drop Lantiq gphy reboot/remove reset (4.14)

   - prctl(PR_SET_FP_MODE): Disallow PRE without FR (4.0)

   - ptrace(PTRACE_PEEKUSR): Fix 64-bit FGRs (3.15)"

* tag 'mips_fixes_4.17_3' of git://git.kernel.org/pub/scm/linux/kernel/git/mips/linux:
  MIPS: ptrace: Fix PTRACE_PEEKUSR requests for 64-bit FGRs
  MIPS: prctl: Disallow FRE without FR with PR_SET_FP_MODE requests
  MIPS: lantiq: gphy: Drop reboot/remove reset asserts

6 years agoMerge tag 'vfio-v4.17' of git://github.com/awilliam/linux-vfio
Linus Torvalds [Sat, 2 Jun 2018 17:08:45 +0000 (10:08 -0700)]
Merge tag 'vfio-v4.17' of git://github.com/awilliam/linux-vfio

Pull VFIO fix from Alex Williamson:
 "Revert a pfn page mapping optimization identified as introducing a bad
  page state regression (Alex Williamson)"

* tag 'vfio-v4.17' of git://github.com/awilliam/linux-vfio:
  Revert "vfio/type1: Improve memory pinning process for raw PFN mapping"

6 years agoMerge tag 'char-misc-4.17-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/gregk...
Linus Torvalds [Sat, 2 Jun 2018 17:05:45 +0000 (10:05 -0700)]
Merge tag 'char-misc-4.17-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc

Pull char/misc driver fixes from Greg KH:
 "Here are four small bugfixes for some char/misc drivers. Well, really
  three fixes and one fix for one of those fixes due to problems found
  by 0-day.

  This resolves some reported issues with the hwtracing drivers, and a
  reported regression for the thunderbolt subsystem. All of these have
  been in linux-next for a while now with no reported problems"

* tag 'char-misc-4.17-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc:
  hwtracing: stm: fix build error on some arches
  intel_th: Use correct device when freeing buffers
  stm class: Use vmalloc for the master map
  thunderbolt: Handle NULL boot ACL entries properly

6 years agoMerge tag 'staging-4.17-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh...
Linus Torvalds [Sat, 2 Jun 2018 17:02:14 +0000 (10:02 -0700)]
Merge tag 'staging-4.17-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging

Pull IIO driver fixes from Greg KH:
 "Here are some old IIO driver fixes that were sitting in my tree for a
  few weeks. Sorry about not getting them to you sooner. They fix a
  number of small IIO driver issues that have been reported.

  All of these have been in linux-next for a while with no reported
  problems"

* tag 'staging-4.17-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging:
  iio: adc: select buffer for at91-sama5d2_adc
  iio: hid-sensor-trigger: Fix sometimes not powering up the sensor after resume
  iio: adc: at91-sama5d2_adc: fix channel configuration for differential channels
  iio:kfifo_buf: check for uint overflow
  iio:buffer: make length types match kfifo types
  iio: adc: stm32-dfsdm: fix sample rate for div2 spi clock
  iio: adc: stm32-dfsdm: fix successive oversampling settings
  iio: ad7793: implement IIO_CHAN_INFO_SAMP_FREQ

6 years agoMerge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma
Linus Torvalds [Sat, 2 Jun 2018 16:55:44 +0000 (09:55 -0700)]
Merge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma

Pull rdma fixes from Jason Gunthorpe:
 "Just three small last minute regressions that were found in the last
  week. The Broadcom fix is a bit big for rc7, but since it is fixing
  driver crash regressions that were merged via netdev into rc1, I am
  sending it.

   - bnxt netdev changes merged this cycle caused the bnxt RDMA driver
     to crash under certain situations

   - Arnd found (several, unfortunately) kconfig problems with the
     patches adding INFINIBAND_ADDR_TRANS. Reverting this last part,
     will fix it more fully outside -rc.

   - Subtle change in error code for a uapi function caused breakage in
     userspace. This was bug was subtly introduced cycle"

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma:
  IB/core: Fix error code for invalid GID entry
  IB: Revert "remove redundant INFINIBAND kconfig dependencies"
  RDMA/bnxt_re: Fix broken RoCE driver due to recent L2 driver changes

6 years agoMerge branch 'i2c/for-current' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa...
Linus Torvalds [Sat, 2 Jun 2018 16:52:22 +0000 (09:52 -0700)]
Merge branch 'i2c/for-current' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux

Pull i2c fixes from Wolfram Sang:
 "A documentation bugfix and a MAINTAINERS addition"

* 'i2c/for-current' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux:
  i2c: ocores: update HDL sources URL
  i2c: xlp9xx: Add MAINTAINERS entry

6 years agoMerge branch 'akpm' (patches from Andrew)
Linus Torvalds [Sat, 2 Jun 2018 16:44:15 +0000 (09:44 -0700)]
Merge branch 'akpm' (patches from Andrew)

Merge two fixes from Andrew Morton.

* emailed patches from Andrew Morton <akpm@linux-foundation.org>:
  mm: fix the NULL mapping case in __isolate_lru_page()
  mm/huge_memory.c: __split_huge_page() use atomic ClearPageDirty()

6 years agomm: fix the NULL mapping case in __isolate_lru_page()
Hugh Dickins [Fri, 1 Jun 2018 23:50:50 +0000 (16:50 -0700)]
mm: fix the NULL mapping case in __isolate_lru_page()

George Boole would have noticed a slight error in 4.16 commit
69d763fc6d3a ("mm: pin address_space before dereferencing it while
isolating an LRU page").  Fix it, to match both the comment above it,
and the original behaviour.

Although anonymous pages are not marked PageDirty at first, we have an
old habit of calling SetPageDirty when a page is removed from swap
cache: so there's a category of ex-swap pages that are easily
migratable, but were inadvertently excluded from compaction's async
migration in 4.16.

Link: http://lkml.kernel.org/r/alpine.LSU.2.11.1805302014001.12558@eggly.anvils
Fixes: 69d763fc6d3a ("mm: pin address_space before dereferencing it while isolating an LRU page")
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Minchan Kim <minchan@kernel.org>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Reported-by: Ivan Kalvachev <ikalvachev@gmail.com>
Cc: "Huang, Ying" <ying.huang@intel.com>
Cc: Jan Kara <jack@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agomm/huge_memory.c: __split_huge_page() use atomic ClearPageDirty()
Hugh Dickins [Fri, 1 Jun 2018 23:50:45 +0000 (16:50 -0700)]
mm/huge_memory.c: __split_huge_page() use atomic ClearPageDirty()

Swapping load on huge=always tmpfs (with khugepaged tuned up to be very
eager, but I'm not sure that is relevant) soon hung uninterruptibly,
waiting for page lock in shmem_getpage_gfp()'s find_lock_entry(), most
often when "cp -a" was trying to write to a smallish file.  Debug showed
that the page in question was not locked, and page->mapping NULL by now,
but page->index consistent with having been in a huge page before.

Reproduced in minutes on a 4.15 kernel, even with 4.17's 605ca5ede764
("mm/huge_memory.c: reorder operations in __split_huge_page_tail()") added
in; but took hours to reproduce on a 4.17 kernel (no idea why).

The culprit proved to be the __ClearPageDirty() on tails beyond i_size in
__split_huge_page(): the non-atomic __bitoperation may have been safe when
4.8's baa355fd3314 ("thp: file pages support for split_huge_page()")
introduced it, but liable to erase PageWaiters after 4.10's 62906027091f
("mm: add PageWaiters indicating tasks are waiting for a page bit").

Link: http://lkml.kernel.org/r/alpine.LSU.2.11.1805291841070.3197@eggly.anvils
Fixes: 62906027091f ("mm: add PageWaiters indicating tasks are waiting for a page bit")
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Cc: Nicholas Piggin <npiggin@gmail.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agoRevert "vfio/type1: Improve memory pinning process for raw PFN mapping"
Alex Williamson [Sat, 2 Jun 2018 14:41:44 +0000 (08:41 -0600)]
Revert "vfio/type1: Improve memory pinning process for raw PFN mapping"

Bisection by Amadeusz Sławiński implicates this commit leading to bad
page state issues after VM shutdown, likely due to unbalanced page
references.  The original commit was intended only as a performance
improvement, therefore revert for offline rework.

Link: https://lkml.org/lkml/2018/6/2/97
Fixes: 356e88ebe447 ("vfio/type1: Improve memory pinning process for raw PFN mapping")
Cc: Jason Cai (Xiang Feng) <jason.cai@linux.alibaba.com>
Reported-by: Amadeusz Sławiński <amade@asmblr.net>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
6 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf
David S. Miller [Sat, 2 Jun 2018 12:07:52 +0000 (08:07 -0400)]
Merge git://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf

Daniel Borkmann says:

====================
pull-request: bpf 2018-06-02

The following pull-request contains BPF updates for your *net* tree.

The main changes are:

1) BPF uapi fix in struct bpf_prog_info and struct bpf_map_info in
   order to fix offsets on 32 bit archs.

This will have a minor merge conflict with net-next which has the
__u32 gpl_compatible:1 bitfield in struct bpf_prog_info at this
location. Resolution is to use the gpl_compatible member.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agobpf: fix uapi hole for 32 bit compat applications
Daniel Borkmann [Sat, 2 Jun 2018 03:21:59 +0000 (05:21 +0200)]
bpf: fix uapi hole for 32 bit compat applications

In 64 bit, we have a 4 byte hole between ifindex and netns_dev in the
case of struct bpf_map_info but also struct bpf_prog_info. In net-next
commit b85fab0e67b ("bpf: Add gpl_compatible flag to struct bpf_prog_info")
added a bitfield into it to expose some flags related to programs. Thus,
add an unnamed __u32 bitfield for both so that alignment keeps the same
in both 32 and 64 bit cases, and can be naturally extended from there
as in b85fab0e67b.

Before:

  # file test.o
  test.o: ELF 32-bit LSB relocatable, Intel 80386, version 1 (SYSV), not stripped
  # pahole test.o
  struct bpf_map_info {
__u32                      type;                 /*     0     4 */
__u32                      id;                   /*     4     4 */
__u32                      key_size;             /*     8     4 */
__u32                      value_size;           /*    12     4 */
__u32                      max_entries;          /*    16     4 */
__u32                      map_flags;            /*    20     4 */
char                       name[16];             /*    24    16 */
__u32                      ifindex;              /*    40     4 */
__u64                      netns_dev;            /*    44     8 */
__u64                      netns_ino;            /*    52     8 */

/* size: 64, cachelines: 1, members: 10 */
/* padding: 4 */
  };

After (same as on 64 bit):

  # file test.o
  test.o: ELF 32-bit LSB relocatable, Intel 80386, version 1 (SYSV), not stripped
  # pahole test.o
  struct bpf_map_info {
__u32                      type;                 /*     0     4 */
__u32                      id;                   /*     4     4 */
__u32                      key_size;             /*     8     4 */
__u32                      value_size;           /*    12     4 */
__u32                      max_entries;          /*    16     4 */
__u32                      map_flags;            /*    20     4 */
char                       name[16];             /*    24    16 */
__u32                      ifindex;              /*    40     4 */

/* XXX 4 bytes hole, try to pack */

__u64                      netns_dev;            /*    48     8 */
__u64                      netns_ino;            /*    56     8 */
/* --- cacheline 1 boundary (64 bytes) --- */

/* size: 64, cachelines: 1, members: 10 */
/* sum members: 60, holes: 1, sum holes: 4 */
  };

Reported-by: Dmitry V. Levin <ldv@altlinux.org>
Reported-by: Eugene Syromiatnikov <esyr@redhat.com>
Fixes: 52775b33bb507 ("bpf: offload: report device information about offloaded maps")
Fixes: 675fc275a3a2d ("bpf: offload: report device information for offloaded programs")
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
6 years agonet: usb: cdc_mbim: add flag FLAG_SEND_ZLP
Daniele Palmas [Thu, 31 May 2018 09:18:29 +0000 (11:18 +0200)]
net: usb: cdc_mbim: add flag FLAG_SEND_ZLP

Testing Telit LM940 with ICMP packets > 14552 bytes revealed that
the modem needs FLAG_SEND_ZLP to properly work, otherwise the cdc
mbim data interface won't be anymore responsive.

Signed-off-by: Daniele Palmas <dnlplm@gmail.com>
Acked-by: Bjørn Mork <bjorn@mork.no>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'tunnel-mtus'
David S. Miller [Fri, 1 Jun 2018 17:56:31 +0000 (13:56 -0400)]
Merge branch 'tunnel-mtus'

Nicolas Dichtel says:

====================
ip[6] tunnels: fix mtu calculations

The first patch restores the possibility to bind an ip4 tunnel to an
interface whith a large mtu.
The second patch was spotted after the first fix. I also target it to net
because it fixes the max mtu value that can be used for ipv6 tunnels.

v2: remove the 0xfff8 in ip_tunnel_newlink()
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoip6_tunnel: remove magic mtu value 0xFFF8
Nicolas Dichtel [Thu, 31 May 2018 08:59:33 +0000 (10:59 +0200)]
ip6_tunnel: remove magic mtu value 0xFFF8

I don't know where this value comes from (probably a copy and paste and
paste and paste ...).
Let's use standard values which are a bit greater.

Link: https://git.kernel.org/pub/scm/linux/kernel/git/davem/netdev-vger-cvs.git/commit/?id=e5afd356a411a
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoip_tunnel: restore binding to ifaces with a large mtu
Nicolas Dichtel [Thu, 31 May 2018 08:59:32 +0000 (10:59 +0200)]
ip_tunnel: restore binding to ifaces with a large mtu

After commit f6cc9c054e77, the following conf is broken (note that the
default loopback mtu is 65536, ie IP_MAX_MTU + 1):

$ ip tunnel add gre1 mode gre local 10.125.0.1 remote 10.125.0.2 dev lo
add tunnel "gre0" failed: Invalid argument
$ ip l a type dummy
$ ip l s dummy1 up
$ ip l s dummy1 mtu 65535
$ ip tunnel add gre1 mode gre local 10.125.0.1 remote 10.125.0.2 dev dummy1
add tunnel "gre0" failed: Invalid argument

dev_set_mtu() doesn't allow to set a mtu which is too large.
First, let's cap the mtu returned by ip_tunnel_bind_dev(). Second, remove
the magic value 0xFFF8 and use IP_MAX_MTU instead.
0xFFF8 seems to be there for ages, I don't know why this value was used.

With a recent kernel, it's also possible to set a mtu > IP_MAX_MTU:
$ ip l s dummy1 mtu 66000
After that patch, it's also possible to bind an ip tunnel on that kind of
interface.

CC: Petr Machata <petrm@mellanox.com>
CC: Ido Schimmel <idosch@mellanox.com>
Link: https://git.kernel.org/pub/scm/linux/kernel/git/davem/netdev-vger-cvs.git/commit/?id=e5afd356a411a
Fixes: f6cc9c054e77 ("ip_tunnel: Emit events for post-register MTU changes")
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/klassert/ipsec
David S. Miller [Fri, 1 Jun 2018 17:25:41 +0000 (13:25 -0400)]
Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/klassert/ipsec

Steffen Klassert says:

====================
pull request (net): ipsec 2018-05-31

1) Avoid possible overflow of the offset variable
   in  _decode_session6(), this fixes an infinite
   lookp there. From Eric Dumazet.

2) We may use an error pointer in the error path of
   xfrm_bundle_create(). Fix this by returning this
   pointer directly to the caller.

Please pull or let me know if there are problems.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: dsa: b53: Add BCM5389 support
Damien Thébault [Thu, 31 May 2018 07:04:01 +0000 (07:04 +0000)]
net: dsa: b53: Add BCM5389 support

This patch adds support for the BCM5389 switch connected through MDIO.

Signed-off-by: Damien Thébault <damien.thebault@vitec.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agolightnvm: pblk: take bitmap alloc. out of critical section
Javier González [Fri, 1 Jun 2018 14:41:14 +0000 (16:41 +0200)]
lightnvm: pblk: take bitmap alloc. out of critical section

pblk allocates line bitmaps within the line lock unnecessarily. In order
to take pressure out of the fast patch, allocate line bitmaps outside
of this lock and refactor accordingly.

Signed-off-by: Javier González <javier@cnexlabs.com>
Signed-off-by: Matias Bjørling <mb@lightnvm.io>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agolightnvm: pblk: kick writer on new flush points
Hans Holmberg [Fri, 1 Jun 2018 14:41:13 +0000 (16:41 +0200)]
lightnvm: pblk: kick writer on new flush points

Unless we kick the writer directly when setting a new flush point, the
user risks having to wait for up to one second (the default timeout for
the write thread to be kicked) for the IO to complete.

Signed-off-by: Hans Holmberg <hans.holmberg@cnexlabs.com>
Signed-off-by: Matias Bjørling <mb@lightnvm.io>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agolightnvm: pblk: only try to recover lines with written smeta
Hans Holmberg [Fri, 1 Jun 2018 14:41:12 +0000 (16:41 +0200)]
lightnvm: pblk: only try to recover lines with written smeta

When switching between different lun configurations, there is no
guarantee that all lines that contain closed/open chunks have some
valid data to recover.

Check that the smeta chunk has been written to instead. Also
skip bad lines (that does not have enough good chunks).

Signed-off-by: Hans Holmberg <hans.holmberg@cnexlabs.com>
Signed-off-by: Matias Bjørling <mb@lightnvm.io>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agolightnvm: pblk: remove unnecessary bio_get/put
Javier González [Fri, 1 Jun 2018 14:41:11 +0000 (16:41 +0200)]
lightnvm: pblk: remove unnecessary bio_get/put

In the read path, pblk gets a reference to the incoming bio and puts it
after ending the bio. Though this behavior is correct, it is unnecessary
since pblk is the one putting the bio, therefore, it cannot disappear
underneath it.

Removing this reference, allows to clean up rqd->bio and avoids pointer
bouncing for the different read paths. Now, the incoming bio always
resides in the read context and pblk's internal bios (if any) reside in
rqd->bio.

Signed-off-by: Javier González <javier@cnexlabs.com>
Signed-off-by: Matias Bjørling <mb@lightnvm.io>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agolightnvm: pblk: add possibility to set write buffer size manually
Marcin Dziegielewski [Fri, 1 Jun 2018 14:41:10 +0000 (16:41 +0200)]
lightnvm: pblk: add possibility to set write buffer size manually

In some cases, users can want set write buffer size manually, e.g. to
adjust it to specific workload. This patch provides the possibility
to set write buffer size via module parameter feature.

Signed-off-by: Marcin Dziegielewski <marcin.dziegielewski@intel.com>
Signed-off-by: Igor Konopko <igor.j.konopko@intel.com>
Signed-off-by: Matias Bjørling <mb@lightnvm.io>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agolightnvm: fix partial read error path
Igor Konopko [Fri, 1 Jun 2018 14:41:09 +0000 (16:41 +0200)]
lightnvm: fix partial read error path

When error occurs during bio_add_page on partial read path, pblk
tries to free pages twice.

Signed-off-by: Igor Konopko <igor.j.konopko@intel.com>
Signed-off-by: Marcin Dziegielewski <marcin.dziegielewski@intel.com>
Signed-off-by: Matias Bjørling <mb@lightnvm.io>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agolightnvm: proper error handling for pblk_bio_add_pages
Igor Konopko [Fri, 1 Jun 2018 14:41:08 +0000 (16:41 +0200)]
lightnvm: proper error handling for pblk_bio_add_pages

Currently in case of error caused by bio_pc_add_page in
pblk_bio_add_pages two issues occur when calling from
pblk_rb_read_to_bio(). First one is in pblk_bio_free_pages, since we
are trying to free pages not allocated from our mempool. Second one
is the warn from dma_pool_free, that we are trying to free NULL
pointer dma.

This commit fix both issues.

Signed-off-by: Igor Konopko <igor.j.konopko@intel.com>
Signed-off-by: Marcin Dziegielewski <marcin.dziegielewski@intel.com>
Signed-off-by: Matias Bjørling <mb@lightnvm.io>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agolightnvm: pblk: fix smeta write error path
Hans Holmberg [Fri, 1 Jun 2018 14:41:07 +0000 (16:41 +0200)]
lightnvm: pblk: fix smeta write error path

Smeta write errors were previously ignored. Skip these
lines instead and throw them back on the free
list, so the chunks will go through a reset cycle
before we attempt to use the line again.

Signed-off-by: Hans Holmberg <hans.holmberg@cnexlabs.com>
Reviewed-by: Javier González <javier@cnexlabs.com>
Signed-off-by: Matias Bjørling <mb@lightnvm.io>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agolightnvm: pblk: garbage collect lines with failed writes
Hans Holmberg [Fri, 1 Jun 2018 14:41:06 +0000 (16:41 +0200)]
lightnvm: pblk: garbage collect lines with failed writes

Write failures should not happen under normal circumstances,
so in order to bring the chunk back into a known state as soon
as possible, evacuate all the valid data out of the line and let the
fw judge if the block can be written to in the next reset cycle.

Do this by introducing a new gc list for lines with failed writes,
and ensure that the rate limiter allocates a small portion of
the write bandwidth to get the job done.

The lba list is saved in memory for use during gc as we
cannot gurantee that the emeta data is readable if a write
error occurred.

Signed-off-by: Hans Holmberg <hans.holmberg@cnexlabs.com>
Reviewed-by: Javier González <javier@cnexlabs.com>
Signed-off-by: Matias Bjørling <mb@lightnvm.io>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agolightnvm: pblk: rework write error recovery path
Hans Holmberg [Fri, 1 Jun 2018 14:41:05 +0000 (16:41 +0200)]
lightnvm: pblk: rework write error recovery path

The write error recovery path is incomplete, so rework
the write error recovery handling to do resubmits directly
from the write buffer.

When a write error occurs, the remaining sectors in the chunk are
mapped out and invalidated and the request inserted in a resubmit list.

The writer thread checks if there are any requests to resubmit,
scans and invalidates any lbas that have been overwritten by later
writes and resubmits the failed entries.

Signed-off-by: Hans Holmberg <hans.holmberg@cnexlabs.com>
Reviewed-by: Javier González <javier@cnexlabs.com>
Signed-off-by: Matias Bjørling <mb@lightnvm.io>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agokcm: Fix use-after-free caused by clonned sockets
Kirill Tkhai [Fri, 1 Jun 2018 11:30:38 +0000 (14:30 +0300)]
kcm: Fix use-after-free caused by clonned sockets

(resend for properly queueing in patchwork)

kcm_clone() creates kernel socket, which does not take net counter.
Thus, the net may die before the socket is completely destructed,
i.e. kcm_exit_net() is executed before kcm_done().

Reported-by: syzbot+5f1a04e374a635efc426@syzkaller.appspotmail.com
Signed-off-by: Kirill Tkhai <ktkhai@virtuozzo.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agolightnvm: pblk: remove dead function
Javier González [Fri, 1 Jun 2018 13:04:25 +0000 (15:04 +0200)]
lightnvm: pblk: remove dead function

Remove dead function for manual sync. I/O

Signed-off-by: Javier González <javier@cnexlabs.com>
Signed-off-by: Matias Bjørling <mb@lightnvm.io>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agolightnvm: pass flag on graceful teardown to targets
Javier González [Fri, 1 Jun 2018 13:04:24 +0000 (15:04 +0200)]
lightnvm: pass flag on graceful teardown to targets

If the namespace is unregistered before the LightNVM target is removed
(e.g., on hot unplug) it is too late for the target to store any metadata
on the device - any attempt to write to the device will fail. In this
case, pass on a "gracefull teardown" flag to the target to let it know
when this happens.

In the case of pblk, we pad the open line (close all open chunks) to
improve data retention. In the event of an ungraceful shutdown, avoid
this part and just clean up.

Signed-off-by: Javier González <javier@cnexlabs.com>
Signed-off-by: Matias Bjørling <mb@lightnvm.io>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agolightnvm: pblk: check for chunk size before allocating it
Javier González [Fri, 1 Jun 2018 13:04:23 +0000 (15:04 +0200)]
lightnvm: pblk: check for chunk size before allocating it

Do the check for the chunk state after making sure that the chunk type
is supported.

Fixes: 32ef9412c114 ("lightnvm: pblk: implement get log report chunk")
Signed-off-by: Javier González <javier@cnexlabs.com>
Signed-off-by: Matias Bjørling <mb@lightnvm.io>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agolightnvm: pblk: remove unnecessary argument
Javier González [Fri, 1 Jun 2018 13:04:22 +0000 (15:04 +0200)]
lightnvm: pblk: remove unnecessary argument

Remove unnecessary argument on pblk_line_free()

Signed-off-by: Javier González <javier@cnexlabs.com>
Signed-off-by: Matias Bjørling <mb@lightnvm.io>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agolightnvm: pblk: remove unnecessary indirection
Javier González [Fri, 1 Jun 2018 13:04:21 +0000 (15:04 +0200)]
lightnvm: pblk: remove unnecessary indirection

Call nvm_submit_io directly and remove an unnecessary indirection on the
read path.

Signed-off-by: Javier González <javier@cnexlabs.com>
Signed-off-by: Matias Bjørling <mb@lightnvm.io>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agolightnvm: pblk: return NVM_ error on failed submission
Javier González [Fri, 1 Jun 2018 13:04:20 +0000 (15:04 +0200)]
lightnvm: pblk: return NVM_ error on failed submission

Return a meaningful error when the sanity vector I/O check fails.

Signed-off-by: Javier González <javier@cnexlabs.com>
Signed-off-by: Matias Bjørling <mb@lightnvm.io>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agolightnvm: pblk: warn in case of corrupted write buffer
Javier González [Fri, 1 Jun 2018 13:04:19 +0000 (15:04 +0200)]
lightnvm: pblk: warn in case of corrupted write buffer

When cleaning up buffer entries as we wrap up, their state should be
"completed". If any of the entries is in "submitted" state, it means
that something bad has happened. Trigger a warning immediately instead of
waiting for the state flag to eventually be updated, thus hiding the
issue.

Signed-off-by: Javier González <javier@cnexlabs.com>
Signed-off-by: Matias Bjørling <mb@lightnvm.io>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agolightnvm: pblk: improve error msg on corrupted LBAs
Javier González [Fri, 1 Jun 2018 13:04:18 +0000 (15:04 +0200)]
lightnvm: pblk: improve error msg on corrupted LBAs

In the event of a mismatch between the read LBA and the metadata pointer
reported by the device, improve the error message to be able to detect
the offending physical address (PPA) mapped to the corrupted LBA.

Signed-off-by: Javier González <javier@cnexlabs.com>
Signed-off-by: Matias Bjørling <mb@lightnvm.io>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agolightnvm: pblk: check read lba on gc path
Javier González [Fri, 1 Jun 2018 13:04:17 +0000 (15:04 +0200)]
lightnvm: pblk: check read lba on gc path

Check that the lba stored in the LBA metadata is correct in the GC path
too. This requires a new helper function to check random reads in the
vector read.

Signed-off-by: Javier González <javier@cnexlabs.com>
Signed-off-by: Matias Bjørling <mb@lightnvm.io>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agolightnvm: pblk: recheck for bad lines at runtime
Javier González [Fri, 1 Jun 2018 13:04:16 +0000 (15:04 +0200)]
lightnvm: pblk: recheck for bad lines at runtime

Bad blocks can grow at runtime. Check that the number of valid blocks in
a line are within the sanity threshold before allocating the line for
new writes.

Signed-off-by: Javier González <javier@cnexlabs.com>
Signed-off-by: Matias Bjørling <mb@lightnvm.io>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agolightnvm: pblk: fail gracefully on line alloc. failure
Javier González [Fri, 1 Jun 2018 13:04:15 +0000 (15:04 +0200)]
lightnvm: pblk: fail gracefully on line alloc. failure

In the event of a line failing to allocate, fail gracefully and stop the
pipeline to avoid more write failing in the same place.

Signed-off-by: Javier González <javier@cnexlabs.com>
Signed-off-by: Matias Bjørling <mb@lightnvm.io>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoMerge branch 'nvme-4.18' of git://git.infradead.org/nvme into for-4.18/block
Jens Axboe [Fri, 1 Jun 2018 13:39:48 +0000 (07:39 -0600)]
Merge branch 'nvme-4.18' of git://git.infradead.org/nvme into for-4.18/block

Pull NVMe changes from Christoph:

"Below is another set of NVMe updates for 4.18.  Besides the usual bug
 fixes this includes more feature completness in terms of AEN and log
 page handling on the target."

* 'nvme-4.18' of git://git.infradead.org/nvme:
  nvme: use the changed namespaces list log to clear ns data changed AENs
  nvme: mark nvme_queue_scan static
  nvme: submit AEN event configuration on startup
  nvmet: mask pending AENs
  nvmet: add AEN configuration support
  nvmet: implement the changed namespaces log
  nvmet: split log page implementation
  nvmet: add a new nvmet_zero_sgl helper
  nvme.h: add AEN configuration symbols
  nvme.h: add the changed namespace list log
  nvme.h: untangle AEN notice definitions
  nvmet: fix error return code in nvmet_file_ns_enable()
  nvmet: fix a typo in nvmet_file_ns_enable()
  nvme-fabrics: allow internal passthrough command on deleting controllers
  nvme-loop: add support for multiple ports
  nvme-pci: simplify __nvme_submit_cmd
  nvme-pci: Rate limit the nvme timeout warnings
  nvme: allow duplicate controller if prior controller being deleted

6 years agoblock: split the blk-mq case from elevator_init
Christoph Hellwig [Thu, 31 May 2018 17:11:40 +0000 (19:11 +0200)]
block: split the blk-mq case from elevator_init

There is almost no shared logic, which leads to a very confusing code
flow.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Damien Le Moal <damien.lemoal@wdc.com>
Tested-by: Damien Le Moal <damien.lemoal@wdc.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoblock: move sysfs_lock into elevator_init
Christoph Hellwig [Thu, 31 May 2018 17:11:39 +0000 (19:11 +0200)]
block: move sysfs_lock into elevator_init

Both callers take just around so function call, so move it in.
Also remove the now pointless blk_mq_sched_init wrapper.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Damien Le Moal <damien.lemoal@wdc.com>
Tested-by: Damien Le Moal <damien.lemoal@wdc.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoblock: remove the always unused name argument to elevator_init
Christoph Hellwig [Thu, 31 May 2018 17:11:38 +0000 (19:11 +0200)]
block: remove the always unused name argument to elevator_init

Reported-by: Damien Le Moal <Damien.LeMoal@wdc.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Damien Le Moal <damien.lemoal@wdc.com>
Tested-by: Damien Le Moal <damien.lemoal@wdc.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoblock: unexport elevator_init/exit
Christoph Hellwig [Thu, 31 May 2018 17:11:37 +0000 (19:11 +0200)]
block: unexport elevator_init/exit

These are only used by the block core.  Also move the declarations to
block/blk.h.

Reported-by: Damien Le Moal <Damien.LeMoal@wdc.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Damien Le Moal <damien.lemoal@wdc.com>
Tested-by: Damien Le Moal <damien.lemoal@wdc.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoblock: move initialization of elevator-related fields to blk_alloc_queue_node
Christoph Hellwig [Thu, 31 May 2018 17:11:36 +0000 (19:11 +0200)]
block: move initialization of elevator-related fields to blk_alloc_queue_node

No point in doing this in elevator_init.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reported-by: Damien Le Moal <Damien.LeMoal@wdc.com>
Reviewed-by: Damien Le Moal <damien.lemoal@wdc.com>
Tested-by: Damien Le Moal <damien.lemoal@wdc.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agonvme: use the changed namespaces list log to clear ns data changed AENs
Christoph Hellwig [Fri, 25 May 2018 16:17:41 +0000 (18:17 +0200)]
nvme: use the changed namespaces list log to clear ns data changed AENs

Per section 5.2 we need to issue the corresponding log page to clear an
AEN, so for a namespace data changed AEN we need to read the changed
namespace list log.  And once we read that log anyway we might as well
use it to optimize the rescan.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
6 years agonvme: mark nvme_queue_scan static
Christoph Hellwig [Fri, 25 May 2018 16:15:47 +0000 (18:15 +0200)]
nvme: mark nvme_queue_scan static

And move it toward the top of the file to avoid a forward declaration.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
6 years agonvme: submit AEN event configuration on startup
Hannes Reinecke [Tue, 22 May 2018 09:09:55 +0000 (11:09 +0200)]
nvme: submit AEN event configuration on startup

We should register for AEN events; some law-abiding targets might
not be sending us AENs otherwise.

Signed-off-by: Hannes Reinecke <hare@suse.com>
[hch: slight cleanups]
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
6 years agonvmet: mask pending AENs
Christoph Hellwig [Wed, 30 May 2018 13:05:09 +0000 (15:05 +0200)]
nvmet: mask pending AENs

Per section 5.2 of the NVMe 1.3 spec:

  "When the controller posts a completion queue entry for an outstanding
  Asynchronous Event Request command and thus reports an asynchronous
  event, subsequent events of that event type are automatically masked by
  the controller until the host clears that event. An event is cleared by
  reading the log page associated with that event using the Get Log Page
  command (see section 5.14)."

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
6 years agonvmet: add AEN configuration support
Christoph Hellwig [Wed, 30 May 2018 13:04:47 +0000 (15:04 +0200)]
nvmet: add AEN configuration support

AEN configuration via the 'Get Features' and 'Set Features' admin
command is mandatory, so we should be implemeting handling for it.

Signed-off-by: Hannes Reinecke <hare@suse.com>
[hch: use WRITE_ONCE, check for invalid values]
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
6 years agonvmet: implement the changed namespaces log
Christoph Hellwig [Fri, 25 May 2018 15:16:09 +0000 (17:16 +0200)]
nvmet: implement the changed namespaces log

Just keep a per-controller buffer of changed namespaces and copy it out
in the get log page implementation.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
6 years agonvmet: split log page implementation
Christoph Hellwig [Tue, 22 May 2018 09:10:03 +0000 (11:10 +0200)]
nvmet: split log page implementation

Remove the common code to allocate a buffer and copy it into the SGL.
Instead the two no-op implementations just zero the SGL directly, and
the smart log allocates a buffer on its own.  This prepares for the
more elaborate ANA log page.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
6 years agonvmet: add a new nvmet_zero_sgl helper
Christoph Hellwig [Tue, 22 May 2018 09:10:02 +0000 (11:10 +0200)]
nvmet: add a new nvmet_zero_sgl helper

Zeroes the SGL in the payload.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
6 years agonvme.h: add AEN configuration symbols
Hannes Reinecke [Fri, 25 May 2018 15:34:00 +0000 (17:34 +0200)]
nvme.h: add AEN configuration symbols

Signed-off-by: Hannes Reinecke <hare@suse.com>
[hch: split from a larger patch]
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
6 years agonet-sysfs: Fix memory leak in XPS configuration
Alexander Duyck [Thu, 31 May 2018 19:59:46 +0000 (15:59 -0400)]
net-sysfs: Fix memory leak in XPS configuration

This patch reorders the error cases in showing the XPS configuration so
that we hold off on memory allocation until after we have verified that we
can support XPS on a given ring.

Fixes: 184c449f91fe ("net: Add support for XPS with QoS via traffic classes")
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoixgbe: fix parsing of TC actions for HW offload
Ondřej Hlavatý [Thu, 31 May 2018 21:21:04 +0000 (23:21 +0200)]
ixgbe: fix parsing of TC actions for HW offload

The previous code was optimistic, accepting the offload of whole action
chain when there was a single known action (drop/redirect). This results
in offloading a rule which should not be offloaded, because its behavior
cannot be reproduced in the hardware.

For example:

$ tc filter add dev eno1 parent ffff: protocol ip \
    u32 ht 800: order 1 match tcp src 42 FFFF \
    action mirred egress mirror dev enp1s16 pipe \
    drop

The controller is unable to mirror the packet to a VF, but still
offloads the rule by dropping the packet.

Change the approach of the function to a pessimistic one, rejecting the
chain when an unknown action is found. This is better suited for future
extensions.

Note that both recognized actions always return TC_ACT_SHOT, therefore
it is safe to ignore actions behind them.

Signed-off-by: Ondřej Hlavatý <ohlavaty@redhat.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge tag 'xfs-4.17-fixes-3' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux
Linus Torvalds [Thu, 31 May 2018 21:23:07 +0000 (16:23 -0500)]
Merge tag 'xfs-4.17-fixes-3' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux

Pull xfs fix from Darrick Wong:
 "Clear out i_mapping error state when we're reinitializing inodes.

  This last minute fix prevents writeback error state from persisting
  past the end of the in-core inode lifecycle and causing EIO errors to
  be reported to userspace when no error has occurred.

  This fix for the behavioral regression has been soaking in for-next
  for a while, but various fs developers persuaded me to try to get it
  upstream for 4.17 because the patch that broke things was introduced
  in 4.17-rc4"

* tag 'xfs-4.17-fixes-3' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux:
  fs: clear writeback errors in inode_init_always

6 years agonet: ethernet: davinci_emac: fix error handling in probe()
Dan Carpenter [Thu, 31 May 2018 06:44:49 +0000 (09:44 +0300)]
net: ethernet: davinci_emac: fix error handling in probe()

The current error handling code has an issue where it does:

if (priv->txchan)
cpdma_chan_destroy(priv->txchan);

The problem is that ->txchan is either valid or an error pointer (which
would lead to an Oops).  I've changed it to use multiple error labels so
that the test can be removed.

Also there were some missing calls to netif_napi_del().

Fixes: 3ef0fdb2342c ("net: davinci_emac: switch to new cpdma layer")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet/ncsi: Fix array size in dumpit handler
Samuel Mendoza-Jonas [Thu, 31 May 2018 04:10:04 +0000 (14:10 +1000)]
net/ncsi: Fix array size in dumpit handler

With CONFIG_CC_STACKPROTECTOR enabled the kernel panics as below when
parsing a NCSI_CMD_PKG_INFO command:

[  150.149711] Kernel panic - not syncing: stack-protector: Kernel stack is corrupted in: 805cff08
[  150.149711]
[  150.159919] CPU: 0 PID: 1301 Comm: ncsi-netlink Not tainted 4.13.16-468cbec6d2c91239332cb91b1f0a73aafcb6f0c6 #1
[  150.170004] Hardware name: Generic DT based system
[  150.174852] [<80109930>] (unwind_backtrace) from [<80106bc4>] (show_stack+0x20/0x24)
[  150.182641] [<80106bc4>] (show_stack) from [<805d36e4>] (dump_stack+0x20/0x28)
[  150.189888] [<805d36e4>] (dump_stack) from [<801163ac>] (panic+0xdc/0x278)
[  150.196780] [<801163ac>] (panic) from [<801162cc>] (__stack_chk_fail+0x20/0x24)
[  150.204111] [<801162cc>] (__stack_chk_fail) from [<805cff08>] (ncsi_pkg_info_all_nl+0x244/0x258)
[  150.212912] [<805cff08>] (ncsi_pkg_info_all_nl) from [<804f939c>] (genl_lock_dumpit+0x3c/0x54)
[  150.221535] [<804f939c>] (genl_lock_dumpit) from [<804f873c>] (netlink_dump+0xf8/0x284)
[  150.229550] [<804f873c>] (netlink_dump) from [<804f8d44>] (__netlink_dump_start+0x124/0x17c)
[  150.237992] [<804f8d44>] (__netlink_dump_start) from [<804f9880>] (genl_rcv_msg+0x1c8/0x3d4)
[  150.246440] [<804f9880>] (genl_rcv_msg) from [<804f9174>] (netlink_rcv_skb+0xd8/0x134)
[  150.254361] [<804f9174>] (netlink_rcv_skb) from [<804f96a4>] (genl_rcv+0x30/0x44)
[  150.261850] [<804f96a4>] (genl_rcv) from [<804f7790>] (netlink_unicast+0x198/0x234)
[  150.269511] [<804f7790>] (netlink_unicast) from [<804f7ffc>] (netlink_sendmsg+0x368/0x3b0)
[  150.277783] [<804f7ffc>] (netlink_sendmsg) from [<804abea4>] (sock_sendmsg+0x24/0x34)
[  150.285625] [<804abea4>] (sock_sendmsg) from [<804ac1dc>] (___sys_sendmsg+0x244/0x260)
[  150.293556] [<804ac1dc>] (___sys_sendmsg) from [<804ad98c>] (__sys_sendmsg+0x5c/0x9c)
[  150.301400] [<804ad98c>] (__sys_sendmsg) from [<804ad9e4>] (SyS_sendmsg+0x18/0x1c)
[  150.308984] [<804ad9e4>] (SyS_sendmsg) from [<80102640>] (ret_fast_syscall+0x0/0x3c)
[  150.316743] ---[ end Kernel panic - not syncing: stack-protector: Kernel stack is corrupted in: 805cff08

This turns out to be because the attrs array in ncsi_pkg_info_all_nl()
is initialised to a length of NCSI_ATTR_MAX which is the maximum
attribute number, not the number of attributes.

Fixes: 955dc68cb9b2 ("net/ncsi: Add generic netlink family")
Signed-off-by: Samuel Mendoza-Jonas <sam@mendozajonas.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge tag 'wireless-drivers-for-davem-2018-05-30' of git://git.kernel.org/pub/scm...
David S. Miller [Thu, 31 May 2018 19:27:39 +0000 (15:27 -0400)]
Merge tag 'wireless-drivers-for-davem-2018-05-30' of git://git.kernel.org/pub/scm/linux/kernel/git/kvalo/wireless-drivers

Kalle Valo says:

====================
wireless-drivers fixes for 4.17

Two last minute fixes, hopefully they make it to 4.17 still.

rt2x00

* revert a fix which caused even more problems

iwlwifi

* fix a crash when there are 16 or more logical CPUs
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agocls_flower: Fix incorrect idr release when failing to modify rule
Paul Blakey [Wed, 30 May 2018 08:29:15 +0000 (11:29 +0300)]
cls_flower: Fix incorrect idr release when failing to modify rule

When we fail to modify a rule, we incorrectly release the idr handle
of the unmodified old rule.

Fix that by checking if we need to release it.

Fixes: fe2502e49b58 ("net_sched: remove cls_flower idr on failure")
Reported-by: Vlad Buslov <vladbu@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Paul Blakey <paulb@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet/sonic: Use dma_mapping_error()
Finn Thain [Wed, 30 May 2018 03:03:51 +0000 (13:03 +1000)]
net/sonic: Use dma_mapping_error()

With CONFIG_DMA_API_DEBUG=y, calling sonic_open() produces the
message, "DMA-API: device driver failed to check map error".
Add the missing dma_mapping_error() call.

Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Signed-off-by: Finn Thain <fthain@telegraphics.com.au>
Acked-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agodrm/amd/display: Update color props when modeset is required
Leo (Sunpeng) Li [Thu, 31 May 2018 14:23:37 +0000 (10:23 -0400)]
drm/amd/display: Update color props when modeset is required

This fixes issues where color management properties don't persist
over DPMS on/off, or when the CRTC is moved across connectors.

Signed-off-by: Leo (Sunpeng) Li <sunpeng.li@amd.com>
Reviewed-by: Harry Wentland <Harry.Wentland@amd.com>
Cc: stable@vger.kernel.org
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
6 years agodrm/amd/display: Make atomic-check validate underscan changes
David Francis [Thu, 31 May 2018 17:48:31 +0000 (13:48 -0400)]
drm/amd/display: Make atomic-check validate underscan changes

When the underscan state was changed, atomic-check was triggering a
validation but passing the old underscan values.  This change adds a
somewhat hacky check in dm_update_crtcs_state that will update the
stream if old and newunderscan values are different.
This was causing 4k on Fiji to allow underscan when it wasn't permitted.

Signed-off-by: David Francis <David.Francis@amd.com>
Reviewed-by: David Francis <David.Francis@amd.com>
Acked-by: Harry Wentland <harry.wentland@amd.com>
Cc: stable@vger.kernel.org
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
6 years agonvme.h: add the changed namespace list log
Christoph Hellwig [Fri, 25 May 2018 15:18:33 +0000 (17:18 +0200)]
nvme.h: add the changed namespace list log

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
6 years agonvme.h: untangle AEN notice definitions
Christoph Hellwig [Tue, 22 May 2018 09:09:54 +0000 (11:09 +0200)]
nvme.h: untangle AEN notice definitions

Stop including the event type in the definitions for the notice type.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
6 years agonvmet: fix error return code in nvmet_file_ns_enable()
Wei Yongjun [Thu, 31 May 2018 11:41:30 +0000 (11:41 +0000)]
nvmet: fix error return code in nvmet_file_ns_enable()

Fix to return error code -ENOMEM from the memory alloc fail error
handling case instead of 0, as done elsewhere in this function.

Fixes: d5eff33ee6f8 ("nvmet: add simple file backed ns support")
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.e>
Signed-off-by: Christoph Hellwig <hch@lst.de>
6 years agonvmet: fix a typo in nvmet_file_ns_enable()
Wei Yongjun [Thu, 31 May 2018 01:55:58 +0000 (01:55 +0000)]
nvmet: fix a typo in nvmet_file_ns_enable()

Fix a typo in nvmet_file_ns_enable().

Fixes: d5eff33ee6f8 ("nvmet: add simple file backed ns support")
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.e>
Signed-off-by: Christoph Hellwig <hch@lst.de>
6 years agonvme-fabrics: allow internal passthrough command on deleting controllers
Christoph Hellwig [Fri, 25 May 2018 13:41:54 +0000 (15:41 +0200)]
nvme-fabrics: allow internal passthrough command on deleting controllers

Without this we can't cleanly shut down.

Based on analysis an an earlier patch from Hannes Reinecke.

Fixes: bb06ec31452f ("nvme: expand nvmf_check_if_ready checks")
Reported-by: Hannes Reinecke <hare@suse.de>
Tested-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: James Smart <james.smart@broadcom.com>
6 years agoblock, bfq: prevent soft_rt_next_start from being stuck at infinity
Davide Sapienza [Thu, 31 May 2018 14:45:08 +0000 (16:45 +0200)]
block, bfq: prevent soft_rt_next_start from being stuck at infinity

BFQ can deem a bfq_queue as soft real-time only if the queue
- periodically becomes completely idle, i.e., empty and with
  no still-outstanding I/O request;
- after becoming idle, gets new I/O only after a special reference
  time soft_rt_next_start.

In this respect, after commit "block, bfq: consider also past I/O in
soft real-time detection", the value of soft_rt_next_start can never
decrease. This causes a problem with the following special updating
case for soft_rt_next_start: to prevent queues that are not completely
idle to be wrongly detected as soft real-time (when they become
non-empty again), soft_rt_next_start is temporarily set to infinity
for empty queues with still outstanding I/O requests. But, if such an
update is actually performed, then, because of the above commit,
soft_rt_next_start will be stuck at infinity forever, and the queue
will have no more chance to be considered soft real-time.

On slow systems, this problem does cause actual soft real-time
applications to be occasionally not detected as such.

This commit addresses this issue by eliminating the pushing of
soft_rt_next_start to infinity, and by changing the way non-empty
queues are prevented from being wrongly detected as soft
real-time. Simply, a queue that becomes non-empty again can now be
detected as soft real-time only if it has no outstanding I/O request.

Signed-off-by: Davide Sapienza <sapienza.dav@gmail.com>
Signed-off-by: Paolo Valente <paolo.valente@linaro.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoblock, bfq: increase weight-raising duration for interactive apps
Davide Sapienza [Thu, 31 May 2018 14:45:07 +0000 (16:45 +0200)]
block, bfq: increase weight-raising duration for interactive apps

The maximum possible duration of the weight-raising period for
interactive applications is limited to 13 seconds, as this is the time
needed to load the largest application that we considered when tuning
weight raising. Unfortunately, in such an evaluation, we did not
consider the case of very slow virtual machines.

For example, on a QEMU/KVM virtual machine
- running in a slow PC;
- with a virtual disk stacked on a slow low-end 5400rpm HDD;
- serving a heavy I/O workload, such as the sequential reading of
several files;
mplayer takes 23 seconds to start, if constantly weight-raised.

To address this issue, this commit conservatively sets the upper limit
for weight-raising duration to 25 seconds.

Signed-off-by: Davide Sapienza <sapienza.dav@gmail.com>
Signed-off-by: Paolo Valente <paolo.valente@linaro.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoblock, bfq: remove slow-system class
Paolo Valente [Thu, 31 May 2018 14:45:06 +0000 (16:45 +0200)]
block, bfq: remove slow-system class

BFQ computes the duration of weight raising for interactive
applications automatically, using some reference parameters. In
particular, BFQ uses the best durations (see comments in the code for
how these durations have been assessed) for two classes of systems:
slow and fast ones. Examples of slow systems are old phones or systems
using micro HDDs. Fast systems are all the remaining ones. Using these
parameters, BFQ computes the actual duration of the weight raising,
for the system at hand, as a function of the relative speed of the
system w.r.t. the speed of a reference system, belonging to the same
class of systems as the system at hand.

This slow vs fast differentiation proved to be useful in the past, but
happens to have little meaning with current hardware. Even worse, it
does cause problems in virtual systems, where the speed of the system
can vary frequently, and so widely to just confuse the class-detection
mechanism, and, as we have verified experimentally, to cause BFQ to
compute non-sensical weight-raising durations.

This commit addresses this issue by removing the slow class and the
class-detection mechanism.

Signed-off-by: Paolo Valente <paolo.valente@linaro.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoblock, bfq: add description of weight-raising heuristics
Paolo Valente [Thu, 31 May 2018 14:45:05 +0000 (16:45 +0200)]
block, bfq: add description of weight-raising heuristics

A description of how weight raising works is missing in BFQ
sources. In addition, the code for handling weight raising is
scattered across a few functions. This makes it rather hard to
understand the mechanism and its rationale. This commits adds such a
description at the beginning of the main source file.

Signed-off-by: Paolo Valente <paolo.valente@linaro.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoblock, bfq: remove the removal of 'next' rq in bfq_requests_merged
Filippo Muzzini [Thu, 31 May 2018 13:23:13 +0000 (15:23 +0200)]
block, bfq: remove the removal of 'next' rq in bfq_requests_merged

Since bfq_finish_request() is always called on the request 'next',
after bfq_requests_merged() is finished, and bfq_finish_request()
removes 'next' from its bfq_queue if needed, it isn't necessary to do
such a removal in advance in bfq_merged_requests().

This commit removes such a useless 'next' removal.

Signed-off-by: Filippo Muzzini <filippo.muzzini@outlook.it>
Signed-off-by: Paolo Valente <paolo.valente@linaro.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoblock, bfq: remove wrong check in bfq_requests_merged
Paolo Valente [Thu, 31 May 2018 14:48:05 +0000 (08:48 -0600)]
block, bfq: remove wrong check in bfq_requests_merged

The request rq passed to the function bfq_requests_merged is always in
a bfq_queue, so the check !RB_EMPTY_NODE(&rq->rb_node) at the
beginning of bfq_requests_merged always succeeds, and the control
flow systematically skips to the end of the function.  This implies
that the body of the function is never executed, i.e., the
repositioning of rq is never performed.

On the opposite end, a control is missing in the body of the function:
'next' must be removed only if it is inside a bfq_queue.

This commit removes the wrong check on rq, and adds the missing check
on 'next'. In addition, this commit adds comments on
bfq_requests_merged.

Signed-off-by: Filippo Muzzini <filippo.muzzini@outlook.it>
Signed-off-by: Paolo Valente <paolo.valente@linaro.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoblock, bfq: remove wrong lock in bfq_requests_merged
Filippo Muzzini [Thu, 31 May 2018 13:23:11 +0000 (15:23 +0200)]
block, bfq: remove wrong lock in bfq_requests_merged

In bfq_requests_merged(), there is a deadlock because the lock on
bfqq->bfqd->lock is held by the calling function, but the code of
this function tries to grab the lock again.

This deadlock is currently hidden by another bug (fixed by next commit
for this source file), which causes the body of bfq_requests_merged()
to be never executed.

This commit removes the deadlock by removing the lock/unlock pair.

Signed-off-by: Filippo Muzzini <filippo.muzzini@outlook.it>
Signed-off-by: Paolo Valente <paolo.valente@linaro.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
6 years agoMerge tag 'platform-drivers-x86-v4.17-4' of git://git.infradead.org/linux-platform...
Linus Torvalds [Thu, 31 May 2018 14:39:57 +0000 (09:39 -0500)]
Merge tag 'platform-drivers-x86-v4.17-4' of git://git.infradead.org/linux-platform-drivers-x86

Pull x86 platform driver fix from Andy Shevchenko:
 "Fix NULL pointer dereference in asus-wmi on rfkill cleanup.

  The effective change is just one new condition - two lines of code.
  But it required moving one static helper function, which is why the
  diff looks a bit bigger"

* tag 'platform-drivers-x86-v4.17-4' of git://git.infradead.org/linux-platform-drivers-x86:
  platform/x86: asus-wmi: Fix NULL pointer dereference

6 years agoplatform/x86: asus-wmi: Fix NULL pointer dereference
João Paulo Rechi Vita [Tue, 22 May 2018 21:30:15 +0000 (14:30 -0700)]
platform/x86: asus-wmi: Fix NULL pointer dereference

Do not perform the rfkill cleanup routine when
(asus->driver->wlan_ctrl_by_user && ashs_present()) is true, since
nothing is registered with the rfkill subsystem in that case. Doing so
leads to the following kernel NULL pointer dereference:

  BUG: unable to handle kernel NULL pointer dereference at           (null)
  IP: [<ffffffff816c7348>] __mutex_lock_slowpath+0x98/0x120
  PGD 1a3aa8067
  PUD 1a3b3d067
  PMD 0

  Oops: 0002 [#1] PREEMPT SMP
  Modules linked in: bnep ccm binfmt_misc uvcvideo videobuf2_vmalloc videobuf2_memops videobuf2_v4l2 videobuf2_core hid_a4tech videodev x86_pkg_temp_thermal intel_powerclamp coretemp ath3k btusb btrtl btintel bluetooth kvm_intel snd_hda_codec_hdmi kvm snd_hda_codec_realtek snd_hda_codec_generic irqbypass crc32c_intel arc4 i915 snd_hda_intel snd_hda_codec ath9k ath9k_common ath9k_hw ath i2c_algo_bit snd_hwdep mac80211 ghash_clmulni_intel snd_hda_core snd_pcm snd_timer cfg80211 ehci_pci xhci_pci drm_kms_helper syscopyarea sysfillrect sysimgblt fb_sys_fops drm xhci_hcd ehci_hcd asus_nb_wmi(-) asus_wmi sparse_keymap r8169 rfkill mxm_wmi serio_raw snd mii mei_me lpc_ich i2c_i801 video soundcore mei i2c_smbus wmi i2c_core mfd_core
  CPU: 3 PID: 3275 Comm: modprobe Not tainted 4.9.34-gentoo #34
  Hardware name: ASUSTeK COMPUTER INC. K56CM/K56CM, BIOS K56CM.206 08/21/2012
  task: ffff8801a639ba00 task.stack: ffffc900014cc000
  RIP: 0010:[<ffffffff816c7348>]  [<ffffffff816c7348>] __mutex_lock_slowpath+0x98/0x120
  RSP: 0018:ffffc900014cfce0  EFLAGS: 00010282
  RAX: 0000000000000000 RBX: ffff8801a54315b0 RCX: 00000000c0000100
  RDX: 0000000000000001 RSI: 0000000000000000 RDI: ffff8801a54315b4
  RBP: ffffc900014cfd30 R08: 0000000000000000 R09: 0000000000000002
  R10: 0000000000000000 R11: 0000000000000000 R12: ffff8801a54315b4
  R13: ffff8801a639ba00 R14: 00000000ffffffff R15: ffff8801a54315b8
  FS:  00007faa254fb700(0000) GS:ffff8801aef80000(0000) knlGS:0000000000000000
  CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
  CR2: 0000000000000000 CR3: 00000001a3b1b000 CR4: 00000000001406e0
  Stack:
   ffff8801a54315b8 0000000000000000 ffffffff814733ae ffffc900014cfd28
   ffffffff8146a28c ffff8801a54315b0 0000000000000000 ffff8801a54315b0
   ffff8801a66f3820 0000000000000000 ffffc900014cfd48 ffffffff816c73e7
  Call Trace:
   [<ffffffff814733ae>] ? acpi_ut_release_mutex+0x5d/0x61
   [<ffffffff8146a28c>] ? acpi_ns_get_node+0x49/0x52
   [<ffffffff816c73e7>] mutex_lock+0x17/0x30
   [<ffffffffa00a3bb4>] asus_rfkill_hotplug+0x24/0x1a0 [asus_wmi]
   [<ffffffffa00a4421>] asus_wmi_rfkill_exit+0x61/0x150 [asus_wmi]
   [<ffffffffa00a49f1>] asus_wmi_remove+0x61/0xb0 [asus_wmi]
   [<ffffffff814a5128>] platform_drv_remove+0x28/0x40
   [<ffffffff814a2901>] __device_release_driver+0xa1/0x160
   [<ffffffff814a29e3>] device_release_driver+0x23/0x30
   [<ffffffff814a1ffd>] bus_remove_device+0xfd/0x170
   [<ffffffff8149e5a9>] device_del+0x139/0x270
   [<ffffffff814a5028>] platform_device_del+0x28/0x90
   [<ffffffff814a50a2>] platform_device_unregister+0x12/0x30
   [<ffffffffa00a4209>] asus_wmi_unregister_driver+0x19/0x30 [asus_wmi]
   [<ffffffffa00da0ea>] asus_nb_wmi_exit+0x10/0xf26 [asus_nb_wmi]
   [<ffffffff8110c692>] SyS_delete_module+0x192/0x270
   [<ffffffff810022b2>] ? exit_to_usermode_loop+0x92/0xa0
   [<ffffffff816ca560>] entry_SYSCALL_64_fastpath+0x13/0x94
  Code: e8 5e 30 00 00 8b 03 83 f8 01 0f 84 93 00 00 00 48 8b 43 10 4c 8d 7b 08 48 89 63 10 41 be ff ff ff ff 4c 89 3c 24 48 89 44 24 08 <48> 89 20 4c 89 6c 24 10 eb 1d 4c 89 e7 49 c7 45 08 02 00 00 00
  RIP  [<ffffffff816c7348>] __mutex_lock_slowpath+0x98/0x120
   RSP <ffffc900014cfce0>
  CR2: 0000000000000000
  ---[ end trace 8d484233fa7cb512 ]---
  note: modprobe[3275] exited with preempt_count 2

https://bugzilla.kernel.org/show_bug.cgi?id=196467

Reported-by: red.f0xyz@gmail.com
Signed-off-by: João Paulo Rechi Vita <jprvita@endlessm.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
6 years agoMerge tag 'perf-urgent-for-mingo-4.17-20180531' of git://git.kernel.org/pub/scm/linux...
Ingo Molnar [Thu, 31 May 2018 10:37:07 +0000 (12:37 +0200)]
Merge tag 'perf-urgent-for-mingo-4.17-20180531' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux into perf/urgent

Pull perf/urgent fixes from Arnaldo Carvalho de Melo:

- Fix 'perf test Session topology' segfault on s390 (Thomas Richter)

- Fix NULL return handling in bpf__prepare_load() (YueHaibing)

- Fix indexing on Coresight ETM packet queue decoder (Mathieu Poirier)

- Fix perf.data format description of NRCPUS header (Arnaldo Carvalho de Melo)

- Update perf.data documentation section on cpu topology

- Handle uncore event aliases in small groups properly (Kan Liang)

- Add missing perf_sample.addr into python sample dictionary (Leo Yan)

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
6 years agosched/headers: Fix typo
Davidlohr Bueso [Wed, 30 May 2018 22:49:40 +0000 (15:49 -0700)]
sched/headers: Fix typo

I cannot spell 'throttling'.

Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20180530224940.17839-1-dave@stgolabs.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
6 years agosched/deadline: Fix missing clock update
Juri Lelli [Wed, 30 May 2018 16:08:09 +0000 (18:08 +0200)]
sched/deadline: Fix missing clock update

A missing clock update is causing the following warning:

 rq->clock_update_flags < RQCF_ACT_SKIP
 WARNING: CPU: 10 PID: 0 at kernel/sched/sched.h:963 inactive_task_timer+0x5d6/0x720
 Call Trace:
  <IRQ>
  __hrtimer_run_queues+0x10f/0x530
  hrtimer_interrupt+0xe5/0x240
  smp_apic_timer_interrupt+0x79/0x2b0
  apic_timer_interrupt+0xf/0x20
  </IRQ>
  do_idle+0x203/0x280
  cpu_startup_entry+0x6f/0x80
  start_secondary+0x1b0/0x200
  secondary_startup_64+0xa5/0xb0
 hardirqs last  enabled at (793919): [<ffffffffa27c5f6e>] cpuidle_enter_state+0x9e/0x360
 hardirqs last disabled at (793920): [<ffffffffa2a0096e>] interrupt_entry+0xce/0xe0
 softirqs last  enabled at (793922): [<ffffffffa20bef78>] irq_enter+0x68/0x70
 softirqs last disabled at (793921): [<ffffffffa20bef5d>] irq_enter+0x4d/0x70

This happens because inactive_task_timer() calls sub_running_bw() (if
TASK_DEAD and non_contending) that might trigger a schedutil update,
which might access the clock. Clock is however currently updated only
later in inactive_task_timer() function.

Fix the problem by updating the clock right after task_rq_lock().

Reported-by: kernel test robot <xiaolong.ye@intel.com>
Signed-off-by: Juri Lelli <juri.lelli@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Claudio Scordino <claudio@evidence.eu.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Luca Abeni <luca.abeni@santannapisa.it>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20180530160809.9074-1-juri.lelli@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
6 years agosched/core: Require cpu_active() in select_task_rq(), for user tasks
Paul Burton [Sat, 26 May 2018 15:46:47 +0000 (08:46 -0700)]
sched/core: Require cpu_active() in select_task_rq(), for user tasks

select_task_rq() is used in a few paths to select the CPU upon which a
thread should be run - for example it is used by try_to_wake_up() & by
fork or exec balancing. As-is it allows use of any online CPU that is
present in the task's cpus_allowed mask.

This presents a problem because there is a period whilst CPUs are
brought online where a CPU is marked online, but is not yet fully
initialized - ie. the period where CPUHP_AP_ONLINE_IDLE <= state <
CPUHP_ONLINE. Usually we don't run any user tasks during this window,
but there are corner cases where this can happen. An example observed
is:

  - Some user task A, running on CPU X, forks to create task B.

  - sched_fork() calls __set_task_cpu() with cpu=X, setting task B's
    task_struct::cpu field to X.

  - CPU X is offlined.

  - Task A, currently somewhere between the __set_task_cpu() in
    copy_process() and the call to wake_up_new_task(), is migrated to
    CPU Y by migrate_tasks() when CPU X is offlined.

  - CPU X is onlined, but still in the CPUHP_AP_ONLINE_IDLE state. The
    scheduler is now active on CPU X, but there are no user tasks on
    the runqueue.

  - Task A runs on CPU Y & reaches wake_up_new_task(). This calls
    select_task_rq() with cpu=X, taken from task B's task_struct,
    and select_task_rq() allows CPU X to be returned.

  - Task A enqueues task B on CPU X's runqueue, via activate_task() &
    enqueue_task().

  - CPU X now has a user task on its runqueue before it has reached the
    CPUHP_ONLINE state.

In most cases, the user tasks that schedule on the newly onlined CPU
have no idea that anything went wrong, but one case observed to be
problematic is if the task goes on to invoke the sched_setaffinity
syscall. The newly onlined CPU reaches the CPUHP_AP_ONLINE_IDLE state
before the CPU that brought it online calls stop_machine_unpark(). This
means that for a portion of the window of time between
CPUHP_AP_ONLINE_IDLE & CPUHP_ONLINE the newly onlined CPU's struct
cpu_stopper has its enabled field set to false. If a user thread is
executed on the CPU during this window and it invokes sched_setaffinity
with a CPU mask that does not include the CPU it's running on, then when
__set_cpus_allowed_ptr() calls stop_one_cpu() intending to invoke
migration_cpu_stop() and perform the actual migration away from the CPU
it will simply return -ENOENT rather than calling migration_cpu_stop().
We then return from the sched_setaffinity syscall back to the user task
that is now running on a CPU which it just asked not to run on, and
which is not present in its cpus_allowed mask.

This patch resolves the problem by having select_task_rq() enforce that
user tasks run on CPUs that are active - the same requirement that
select_fallback_rq() already enforces. This should ensure that newly
onlined CPUs reach the CPUHP_AP_ACTIVE state before being able to
schedule user tasks, and also implies that bringup_wait_for_ap() will
have called stop_machine_unpark() which resolves the sched_setaffinity
issue above.

I haven't yet investigated them, but it may be of interest to review
whether any of the actions performed by hotplug states between
CPUHP_AP_ONLINE_IDLE & CPUHP_AP_ACTIVE could have similar unintended
effects on user tasks that might schedule before they are reached, which
might widen the scope of the problem from just affecting the behaviour
of sched_setaffinity.

Signed-off-by: Paul Burton <paul.burton@mips.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20180526154648.11635-2-paul.burton@mips.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
6 years agosched/core: Fix rules for running on online && !active CPUs
Peter Zijlstra [Tue, 25 Jul 2017 16:58:21 +0000 (18:58 +0200)]
sched/core: Fix rules for running on online && !active CPUs

As already enforced by the WARN() in __set_cpus_allowed_ptr(), the rules
for running on an online && !active CPU are stricter than just being a
kthread, you need to be a per-cpu kthread.

If you're not strictly per-CPU, you have better CPUs to run on and
don't need the partially booted one to get your work done.

The exception is to allow smpboot threads to bootstrap the CPU itself
and get kernel 'services' initialized before we allow userspace on it.

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Tejun Heo <tj@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Fixes: 955dbdf4ce87 ("sched: Allow migrating kthreads into online but inactive CPUs")
Link: http://lkml.kernel.org/r/20170725165821.cejhb7v2s3kecems@hirez.programming.kicks-ass.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
6 years agoxfrm Fix potential error pointer dereference in xfrm_bundle_create.
Steffen Klassert [Thu, 31 May 2018 07:45:18 +0000 (09:45 +0200)]
xfrm Fix potential error pointer dereference in xfrm_bundle_create.

We may derference an invalid pointer in the error path of
xfrm_bundle_create(). Fix this by returning this error
pointer directly instead of assigning it to xdst0.

Fixes: 45b018beddb6 ("ipsec: Create and use new helpers for dst child access.")
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
6 years agofs: clear writeback errors in inode_init_always
Darrick J. Wong [Thu, 31 May 2018 02:43:53 +0000 (19:43 -0700)]
fs: clear writeback errors in inode_init_always

In inode_init_always(), we clear the inode mapping flags, which clears
any retained error (AS_EIO, AS_ENOSPC) bits.  Unfortunately, we do not
also clear wb_err, which means that old mapping errors can leak through
to new inodes.

This is crucial for the XFS inode allocation path because we recycle old
in-core inodes and we do not want error state from an old file to leak
into the new file.  This bug was discovered by running generic/036 and
generic/047 in a loop and noticing that the EIOs generated by the
collision of direct and buffered writes in generic/036 would survive the
remount between 036 and 047, and get reported to the fsyncs (on
different files!) in generic/047.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: Brian Foster <bfoster@redhat.com>