dm thin metadata:  Avoid returning cmd->bm wild pointer on error

Maybe __create_persistent_data_objects() caller will use PTR_ERR as a
pointer, it will lead to some strange things.

Signed-off-by: Ye Bin <[email protected]>
Cc: [email protected]
Signed-off-by: Mike Snitzer <[email protected]>

dm cache metadata: Avoid returning cmd->bm wild pointer on error

Maybe __create_persistent_data_objects() caller will use PTR_ERR as a
pointer, it will lead to some strange things.

Signed-off-by: Ye Bin <[email protected]>
Cc: [email protected]
Signed-off-by: Mike Snitzer <[email protected]>

ALSA: hda: use consistent HDAudio spelling in comments/docs

We use HDaudio and HDAudio, pick one to make searches easier.
No functionality change

Also fix timestamping typo in documentation.

Reported-by: Guennadi Liakhovetski <[email protected]>
Signed-off-by: Pierre-Louis Bossart <[email protected]>
Reviewed-by: Bard Liao <[email protected]>
Reviewed-by: Guennadi Liakhovetski <[email protected]>
Signed-off-by: Kai Vehmanen <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Takashi Iwai <[email protected]>

libata: implement ATA_HORKAGE_MAX_TRIM_128M and apply to Sandisks

All three generations of Sandisk SSDs lock up hard intermittently.
Experiments showed that disabling NCQ lowered the failure rate significantly
and the kernel has been disabling NCQ for some models of SD7's and 8's,
which is obviously undesirable.

Karthik worked with Sandisk to root cause the hard lockups to trim commands
larger than 128M. This patch implements ATA_HORKAGE_MAX_TRIM_128M which
limits max trim size to 128M and applies it to all three generations of
Sandisk SSDs.

Signed-off-by: Tejun Heo <[email protected]>
Cc: Karthik Shivaram <[email protected]>
Cc: [email protected]
Signed-off-by: Jens Axboe <[email protected]>

ALSA: hda: add dev_dbg log when driver is not selected

On SKL+ Intel platforms, the driver selection is handled by the
snd_intel_dspcfg, and when the HDaudio legacy driver is not selected,
be it with the auto-selection or user preferences with a kernel
parameter, the probe aborts with no logs, only a -ENODEV return value.

Having no dmesg trace, even with dynamic debug enabled, makes support
more complicated than it needs to be, and even experienced users can
be fooled. A simple dev_dbg() trace solves this problem.

BugLink: https://github.com/thesofproject/linux/issues/2330
Signed-off-by: Pierre-Louis Bossart <[email protected]>
Reviewed-by: Bard Liao <[email protected]>
Reviewed-by: Guennadi Liakhovetski <[email protected]>
Signed-off-by: Kai Vehmanen <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Takashi Iwai <[email protected]>

ALSA: hda: fix a runtime pm issue in SOF when integrated GPU is disabled

In snd_hdac_device_init pm_runtime_set_active is called to
increase child_count in parent device. But when it is failed
to build connection with GPU for one case that integrated
graphic gpu is disabled, snd_hdac_ext_bus_device_exit will be
invoked to clean up a HD-audio extended codec base device. At
this time the child_count of parent is not decreased, which
makes parent device can't get suspended.

This patch calls pm_runtime_set_suspended to decrease child_count
in parent device in snd_hdac_device_exit to match with
snd_hdac_device_init. pm_runtime_set_suspended can make sure that
it will not decrease child_count if the device is already suspended.

Signed-off-by: Rander Wang <[email protected]>
Reviewed-by: Ranjani Sridharan <[email protected]>
Reviewed-by: Pierre-Louis Bossart <[email protected]>
Reviewed-by: Bard Liao <[email protected]>
Reviewed-by: Guennadi Liakhovetski <[email protected]>
Signed-off-by: Kai Vehmanen <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Takashi Iwai <[email protected]>

ALSA: hda: hdmi - add Rocketlake support

Add Rocketlake HDMI codec support. Rocketlake shares
the pin-to-port mapping table with Tigerlake.

Signed-off-by: Rander Wang <[email protected]>
Reviewed-by: Pierre-Louis Bossart <[email protected]>
Reviewed-by: Ranjani Sridharan <[email protected]>
Signed-off-by: Kai Vehmanen <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Takashi Iwai <[email protected]>

io_uring: no read/write-retry on -EAGAIN error and O_NONBLOCK marked file

Actually two things that need fixing up here:

- The io_rw_reissue() -EAGAIN retry is explicit to block devices and
  regular files, so don't ever attempt to do that on other types of
  files.

- If we hit -EAGAIN on a nonblock marked file, don't arm poll handler for
  it. It should just complete with -EAGAIN.

Cc: [email protected]
Reported-by: Norman Maurer <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>

fix regression in "epoll: Keep a reference on files added to the check list"

epoll_loop_check_proc() can run into a file already committed to destruction;
we can't grab a reference on those and don't need to add them to the set for
reverse path check anyway.

Tested-by: Marc Zyngier <[email protected]>
Fixes: a9ed4a6560b8 ("epoll: Keep a reference on files added to the check list")
Signed-off-by: Al Viro <[email protected]>

io_uring: set table->files[i] to NULL when io_sqe_file_register failed

While io_sqe_file_register() failed in __io_sqe_files_update(),
table->files[i] still point to the original file which may freed
soon, and that will trigger use-after-free problems.

Cc: [email protected]
Fixes: f3bd9dae3708 ("io_uring: fix memleak in __io_sqe_files_update()")
Signed-off-by: Jiufei Xue <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>

Merge branch 'topic/tasklet-convert' into for-linus

Pull tasklet API conversions.

Signed-off-by: Takashi Iwai <[email protected]>

ALSA: ua101: convert tasklets to use new tasklet_setup() API

In preparation for unconditionally passing the
struct tasklet_struct pointer to all tasklet
callbacks, switch to using the new tasklet_setup()
and from_tasklet() to pass the tasklet pointer explicitly.

Signed-off-by: Romain Perier <[email protected]>
Signed-off-by: Allen Pais <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Takashi Iwai <[email protected]>

ALSA: usb-audio: convert tasklets to use new tasklet_setup() API

In preparation for unconditionally passing the
struct tasklet_struct pointer to all tasklet
callbacks, switch to using the new tasklet_setup()
and from_tasklet() to pass the tasklet pointer explicitly.

Signed-off-by: Romain Perier <[email protected]>
Signed-off-by: Allen Pais <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Takashi Iwai <[email protected]>

ASoC: txx9: convert tasklets to use new tasklet_setup() API

In preparation for unconditionally passing the
struct tasklet_struct pointer to all tasklet
callbacks, switch to using the new tasklet_setup()
and from_tasklet() to pass the tasklet pointer explicitly.

Signed-off-by: Romain Perier <[email protected]>
Signed-off-by: Allen Pais <[email protected]>
Acked-by: Mark Brown <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Takashi Iwai <[email protected]>

ASoC: siu: convert tasklets to use new tasklet_setup() API

In preparation for unconditionally passing the
struct tasklet_struct pointer to all tasklet
callbacks, switch to using the new tasklet_setup()
and from_tasklet() to pass the tasklet pointer explicitly.

Signed-off-by: Romain Perier <[email protected]>
Signed-off-by: Allen Pais <[email protected]>
Acked-by: Mark Brown <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Takashi Iwai <[email protected]>

ASoC: fsl_esai: convert tasklets to use new tasklet_setup() API

In preparation for unconditionally passing the
struct tasklet_struct pointer to all tasklet
callbacks, switch to using the new tasklet_setup()
and from_tasklet() to pass the tasklet pointer explicitly.

Signed-off-by: Romain Perier <[email protected]>
Signed-off-by: Allen Pais <[email protected]>
Acked-by: Mark Brown <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Takashi Iwai <[email protected]>

ALSA: hdsp: convert tasklets to use new tasklet_setup() API

In preparation for unconditionally passing the
struct tasklet_struct pointer to all tasklet
callbacks, switch to using the new tasklet_setup()
and from_tasklet() to pass the tasklet pointer explicitly.

Signed-off-by: Romain Perier <[email protected]>
Signed-off-by: Allen Pais <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Takashi Iwai <[email protected]>

ALSA: riptide: convert tasklets to use new tasklet_setup() API

In preparation for unconditionally passing the
struct tasklet_struct pointer to all tasklet
callbacks, switch to using the new tasklet_setup()
and from_tasklet() to pass the tasklet pointer explicitly.

Signed-off-by: Romain Perier <[email protected]>
Signed-off-by: Allen Pais <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Takashi Iwai <[email protected]>

ALSA: pci/asihpi: convert tasklets to use new tasklet_setup() API

In preparation for unconditionally passing the
struct tasklet_struct pointer to all tasklet
callbacks, switch to using the new tasklet_setup()
and from_tasklet() to pass the tasklet pointer explicitly.

Signed-off-by: Romain Perier <[email protected]>
Signed-off-by: Allen Pais <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Takashi Iwai <[email protected]>

ALSA: firewire: convert tasklets to use new tasklet_setup() API

In preparation for unconditionally passing the
struct tasklet_struct pointer to all tasklet
callbacks, switch to using the new tasklet_setup()
and from_tasklet() to pass the tasklet pointer explicitly.

Signed-off-by: Romain Perier <[email protected]>
Signed-off-by: Allen Pais <[email protected]>
Acked-by: Takashi Sakamoto <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Takashi Iwai <[email protected]>

ALSA: core: convert tasklets to use new tasklet_setup() API

In preparation for unconditionally passing the
struct tasklet_struct pointer to all tasklet
callbacks, switch to using the new tasklet_setup()
and from_tasklet() to pass the tasklet pointer explicitly.

Signed-off-by: Romain Perier <[email protected]>
Signed-off-by: Allen Pais <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Takashi Iwai <[email protected]>

s390: update defconfigs

Signed-off-by: Heiko Carstens <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>

s390: fix GENERIC_LOCKBREAK dependency typo in Kconfig

Commit fa686453053b ("sched/rt, s390: Use CONFIG_PREEMPTION")
changed a bunch of uses of CONFIG_PREEMPT to _PREEMPTION.
Except in the Kconfig it used two T's. That's the only place
in the system where that spelling exists, so let's fix that.

Fixes: fa686453053b ("sched/rt, s390: Use CONFIG_PREEMPTION")
Cc: <[email protected]> # 5.6
Signed-off-by: Eric Farman <[email protected]>
Signed-off-by: Heiko Carstens <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>

drm/i915: Clear the repeater bit on HDCP disable

On HDCP disable, clear the repeater bit. This ensures if we connect a
non-repeater sink after a repeater, the bit is in the state we expect.

Fixes: ee5e5e7a5e0f ("drm/i915: Add HDCP framework + base implementation")
Cc: Chris Wilson <[email protected]>
Cc: Ramalingam C <[email protected]>
Cc: Daniel Vetter <[email protected]>
Cc: Sean Paul <[email protected]>
Cc: Jani Nikula <[email protected]>
Cc: Joonas Lahtinen <[email protected]>
Cc: Rodrigo Vivi <[email protected]>
Cc: [email protected]
Cc: <[email protected]> # v4.17+
Reviewed-by: Ramalingam C <[email protected]>
Signed-off-by: Sean Paul <[email protected]>
Signed-off-by: Ramalingam C <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]
(cherry picked from commit 2cc0c7b520bf8ea20ec42285d4e3d37b467eb7f9)
Signed-off-by: Jani Nikula <[email protected]>

drm/i915: Fix sha_text population code

This patch fixes a few bugs:

1- We weren't taking into account sha_leftovers when adding multiple
   ksvs to sha_text. As such, we were or'ing the end of ksv[j - 1] with
   the beginning of ksv[j]

2- In the sha_leftovers == 2 and sha_leftovers == 3 case, bstatus was
   being placed on the wrong half of sha_text, overlapping the leftover
   ksv value

3- In the sha_leftovers == 2 case, we need to manually terminate the
   byte stream with 0x80 since the hardware doesn't have enough room to
   add it after writing M0

The upside is that all of the HDCP supported HDMI repeaters I could
find on Amazon just strip HDCP anyways, so it turns out to be _really_
hard to hit any of these cases without an MST hub, which is not (yet)
supported. Oh, and the sha_leftovers == 1 case works perfectly!

Fixes: ee5e5e7a5e0f ("drm/i915: Add HDCP framework + base implementation")
Cc: Chris Wilson <[email protected]>
Cc: Ramalingam C <[email protected]>
Cc: Daniel Vetter <[email protected]>
Cc: Sean Paul <[email protected]>
Cc: Jani Nikula <[email protected]>
Cc: Joonas Lahtinen <[email protected]>
Cc: Rodrigo Vivi <[email protected]>
Cc: [email protected]
Cc: <[email protected]> # v4.17+
Reviewed-by: Ramalingam C <[email protected]>
Signed-off-by: Sean Paul <[email protected]>
Signed-off-by: Ramalingam C <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]
(cherry picked from commit 1f0882214fd0037b74f245d9be75c31516fed040)
Signed-off-by: Jani Nikula <[email protected]>

drm/i915/display: Ensure that ret is always initialized in icl_combo_phy_verify_state

Clang warns:

drivers/gpu/drm/i915/display/intel_combo_phy.c:268:3: warning: variable
'ret' is uninitialized when used here [-Wuninitialized]
                ret &= check_phy_reg(dev_priv, phy, ICL_PORT_TX_DW8_LN0(phy),
                ^~~
drivers/gpu/drm/i915/display/intel_combo_phy.c:261:10: note: initialize
the variable 'ret' to silence this warning
        bool ret;
                ^
                 = 0
1 warning generated.

In practice, the bug this warning appears to be concerned with would not
actually matter because ret gets initialized to the return value of
cnl_verify_procmon_ref_values. However, that does appear to be a bug
since it means the first hunk of the patch this fixes won't actually do
anything (since the values of check_phy_reg won't factor into the final
ret value). Initialize ret to true then make all of the assignments a
bitwise AND with itself so that the function always does what it should
do.

Fixes: 239bef676d8e ("drm/i915/display: Implement new combo phy initialization step")
Link: https://github.com/ClangBuiltLinux/linux/issues/1094
Signed-off-by: Nathan Chancellor <[email protected]>
Reviewed-by: Matt Roper <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]
Signed-off-by: José Roberto de Souza <[email protected]>
(cherry picked from commit 2034c2129bc4a91d471815d4dc7a2a69eaa5338d)
Signed-off-by: Jani Nikula <[email protected]>

arm64/module: set trampoline section flags regardless of CONFIG_DYNAMIC_FTRACE

In the arm64 module linker script, the section .text.ftrace_trampoline
is specified unconditionally regardless of whether CONFIG_DYNAMIC_FTRACE
is enabled (this is simply due to the limitation that module linker
scripts are not preprocessed like the vmlinux one).

Normally, for .plt and .text.ftrace_trampoline, the section flags
present in the module binary wouldn't matter since module_frob_arch_sections()
would assign them manually anyway. However, the arm64 module loader only
sets the section flags for .text.ftrace_trampoline when CONFIG_DYNAMIC_FTRACE=y.
That's only become problematic recently due to a recent change in
binutils-2.35, where the .text.ftrace_trampoline section (along with the
.plt section) is now marked writable and executable (WAX).

We no longer allow writable and executable sections to be loaded due to
commit 5c3a7db0c7ec ("module: Harden STRICT_MODULE_RWX"), so this is
causing all modules linked with binutils-2.35 to be rejected under arm64.
Drop the IS_ENABLED(CONFIG_DYNAMIC_FTRACE) check in module_frob_arch_sections()
so that the section flags for .text.ftrace_trampoline get properly set to
SHF_EXECINSTR|SHF_ALLOC, without SHF_WRITE.

Signed-off-by: Jessica Yu <[email protected]>
Acked-by: Will Deacon <[email protected]>
Acked-by: Ard Biesheuvel <[email protected]>
Link: http://lore.kernel.org/r/20200831094651.GA16385@linux-8ccs
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Catalin Marinas <[email protected]>

arm64: Remove exporting cpu_logical_map symbol

Commit eaecca9e7710 ("arm64: Fix __cpu_logical_map undefined issue")
exported cpu_logical_map in order to fix tegra194-cpufreq module build
failure.

As this might potentially cause problem while supporting physical CPU
hotplug, tegra194-cpufreq module was reworded to avoid use of
cpu_logical_map() via the commit 93d0c1ab2328 ("cpufreq: replace
cpu_logical_map() with read_cpuid_mpir()")

Since cpu_logical_map was exported to fix the module build temporarily,
let us remove the same before it gains any user again.

Signed-off-by: Sudeep Holla <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Catalin Marinas <[email protected]>

drm/virtio: fix unblank

When going through a disable/enable cycle without changing the
framebuffer the optimization added by commit 3954ff10e06e ("drm/virtio:
skip set_scanout if framebuffer didn't change") causes the screen stay
blank.  Add a bool to force an update to fix that.

v2: use drm_atomic_crtc_needs_modeset() (Daniel).

Cc: [email protected]
Fixes: 3954ff10e06e ("drm/virtio: skip set_scanout if framebuffer didn't change")
Signed-off-by: Gerd Hoffmann <[email protected]>
Tested-by: Jiri Slaby <[email protected]>
Tested-by: Diego Viola <[email protected]>
Acked-by: Daniel Vetter <[email protected]>
Link: http://patchwork.freedesktop.org/patch/msgid/[email protected]
(cherry picked from commit 1bc371cd0ec907bab870cacb6e898105f9c41dc8)

Merge tag 'perf-tools-fixes-for-v5.9-2020-09-01' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux

Pull perf tools fixes from Arnaldo Carvalho de Melo:

 - Fix infinite loop in the TUI for grouped events in 'perf top/record',
   eg when using "perf top -e '{cycles,instructions,cache-misses}'".

 - Fix segfault by skipping side-band event setup if HAVE_LIBBPF_SUPPORT
   is not set.

 - Fix synthesized branch stacks generated from CoreSight ETM trace and
   Intel PT hardware traces.

 - Fix error when synthesizing events from ARM SPE hardware trace.

 - The SNOOPX and REMOTE offsets in the data_src bitmask in perf records
   were were both 37, SNOOPX is 38, fix it.

 - Fix use of CPU list with summary option in 'perf sched timehist'.

 - Avoid an uninitialized read when using fake PMUs.

 - Set perf_event_attr.exclude_guest=1 for user-space counting.

 - Don't order events when doing a 'perf report -D' raw dump of
   perf.data records.

 - Set NULL sentinel in pmu_events table in "Parse and process metrics"
   'perf test'

 - Fix basic bpf filtering 'perf test' on s390x.

 - Fix out of bounds array access in the 'perf stat' print_counters()
   evlist method.

 - Add mwait_idle_with_hints.constprop.0 to the list of idle symbols.

 - Use %zd for size_t printf formats on 32-bit.

 - Correct the help info of "perf record --no-bpf-event" option.

 - Add entries for CoreSight and Arm SPE tooling to MAINTAINERS.

* tag 'perf-tools-fixes-for-v5.9-2020-09-01' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux:
  perf report: Disable ordered_events for raw dump
  perf tools: Correct SNOOPX field offset
  perf intel-pt: Fix corrupt data after perf inject from
  perf cs-etm: Fix corrupt data after perf inject from
  perf top/report: Fix infinite loop in the TUI for grouped events
  perf parse-events: Avoid an uninitialized read when using fake PMUs
  perf stat: Fix out of bounds array access in the print_counters() evlist method
  perf test: Set NULL sentinel in pmu_events table in "Parse and process metrics" test
  perf parse-events: Set exclude_guest=1 for user-space counting
  perf record: Correct the help info of option "--no-bpf-event"
  perf tools: Use %zd for size_t printf formats on 32-bit
  MAINTAINERS: Add entries for CoreSight and Arm SPE tooling
  perf: arm-spe: Fix check error when synthesizing events
  perf symbols: Add mwait_idle_with_hints.constprop.0 to the list of idle symbols
  perf top: Skip side-band event setup if HAVE_LIBBPF_SUPPORT is not set
  perf sched timehist: Fix use of CPU list with summary option
  perf test: Fix basic bpf filtering test

scsi: libsas: Set data_dir as DMA_NONE if libata marks qc as NODATA

It was discovered that sdparm will fail when attempting to disable write
cache on a SATA disk connected via libsas.

In the ATA command set the write cache state is controlled through the SET
FEATURES operation. This is roughly corresponds to MODE SELECT in SCSI and
the latter command is what is used in the SCSI-ATA translation layer. A
subtle difference is that a MODE SELECT carries data whereas SET FEATURES
is defined as a non-data command in ATA.

Set the DMA data direction to DMA_NONE if the requested ATA command is
identified as non-data.

[mkp: commit desc]

Fixes: fa1c1e8f1ece ("[SCSI] Add SATA support to libsas")
Link: https://lore.kernel.org/r/[email protected]
Reviewed-by: John Garry <[email protected]>
Reviewed-by: Jason Yan <[email protected]>
Signed-off-by: Luo Jiaxing <[email protected]>
Signed-off-by: Martin K. Petersen <[email protected]>

Merge tag 'for-5.9-rc3-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux

Pull btrfs fixes from David Sterba:
 "Two small fixes and a bunch of lockdep fixes for warnings that show up
  with an upcoming tree locking update but are valid with current locks
  as well"

* tag 'for-5.9-rc3-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux:
  btrfs: tree-checker: fix the error message for transid error
  btrfs: set the lockdep class for log tree extent buffers
  btrfs: set the correct lockdep class for new nodes
  btrfs: allocate scrub workqueues outside of locks
  btrfs: fix potential deadlock in the search ioctl
  btrfs: drop path before adding new uuid tree entry
  btrfs: block-group: fix free-space bitmap threshold

scsi: target: iscsi: Fix data digest calculation

Current code does not consider 'page_off' in data digest calculation. To
fix this, add a local variable 'first_sg' and set first_sg.offset to
sg->offset + page_off.

Link: https://lore.kernel.org/r/[email protected]
Fixes: e48354ce078c ("iscsi-target: Add iSCSI fabric support for target v4.1")
Cc: <[email protected]>
Reviewed-by: Mike Christie <[email protected]>
Signed-off-by: Varun Prakash <[email protected]>
Signed-off-by: Martin K. Petersen <[email protected]>

blk-stat: make q->stats->lock irqsafe

blk-iocost calls blk_stat_enable_accounting() while holding an irqsafe lock
which triggers a lockdep splat because q->stats->lock isn't irqsafe. Let's
make it irqsafe.

Signed-off-by: Tejun Heo <[email protected]>
Fixes: cd006509b0a9 ("blk-iocost: account for IO size when testing latencies")
Cc: [email protected] # v5.8+
Signed-off-by: Jens Axboe <[email protected]>

blk-iocost: ioc_pd_free() shouldn't assume irq disabled

ioc_pd_free() grabs irq-safe ioc->lock without ensuring that irq is disabled
when it can be called with irq disabled or enabled. This has a small chance
of causing A-A deadlocks and triggers lockdep splats. Use irqsave operations
instead.

Signed-off-by: Tejun Heo <[email protected]>
Fixes: 7caa47151ab2 ("blkcg: implement blk-iocost")
Cc: [email protected] # v5.4+
Signed-off-by: Jens Axboe <[email protected]>

net: usb: dm9601: Add USB ID of Keenetic Plus DSL

Keenetic Plus DSL is a xDSL modem that uses dm9620 as its USB interface.

Signed-off-by: Kamil Lorenc <[email protected]>
Signed-off-by: David S. Miller <[email protected]>

vhost: fix typo in error message

"enable" should be "disable" when the function name is
vhost_disable_notify(), which does the disabling work.

Signed-off-by: Yunsheng Lin <[email protected]>
Acked-by: Jason Wang <[email protected]>
Acked-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: David S. Miller <[email protected]>

Merge tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi

Pull SCSI fixes from James Bottomley:
 "Three minor fixes, all in drivers"

* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
  scsi: scsi_debug: Remove superfluous close zone in resp_open_zone()
  scsi: libcxgbi: Fix a use after free in cxgbi_conn_xmit_pdu()
  scsi: qedf: Fix null ptr reference in qedf_stag_change_work

dm integrity: fix error reporting in bitmap mode after creation

The dm-integrity target did not report errors in bitmap mode just after
creation. The reason is that the function integrity_recalc didn't clean up
ic->recalc_bitmap as it proceeded with recalculation.

Fix this by updating the bitmap accordingly -- the double shift serves
to rounddown.

Signed-off-by: Mikulas Patocka <[email protected]>
Fixes: 468dfca38b1a ("dm integrity: add a bitmap mode")
Cc: [email protected] # v5.2+
Signed-off-by: Mike Snitzer <[email protected]>

dm crypt: Initialize crypto wait structures

Use the DECLARE_CRYPTO_WAIT() macro to properly initialize the crypto
wait structures declared on stack before their use with
crypto_wait_req().

Fixes: 39d13a1ac41d ("dm crypt: reuse eboiv skcipher for IV generation")
Fixes: bbb1658461ac ("dm crypt: Implement Elephant diffuser for Bitlocker compatibility")
Cc: [email protected]
Signed-off-by: Damien Le Moal <[email protected]>
Signed-off-by: Mike Snitzer <[email protected]>

dm mpath: fix racey management of PG initialization

Commit 935fcc56abc3 ("dm mpath: only flush workqueue when needed")
changed flush_multipath_work() to avoid needless workqueue
flushing (of a multipath global workqueue). But that change didn't
realize the surrounding flush_multipath_work() code should also only
run if 'pg_init_in_progress' is set.

Fix this by only doing all of flush_multipath_work()'s PG init related
work if 'pg_init_in_progress' is set.

Otherwise multipath_wait_for_pg_init_completion() will run
unconditionally but the preceeding flush_workqueue(kmpath_handlerd)
may not. This could lead to deadlock (though only if kmpath_handlerd
never runs a corresponding work to decrement 'pg_init_in_progress').

It could also be, though highly unlikely, that the kmpath_handlerd
work that does PG init completes before 'pg_init_in_progress' is set,
and then an intervening DM table reload's multipath_postsuspend()
triggers flush_multipath_work().

Fixes: 935fcc56abc3 ("dm mpath: only flush workqueue when needed")
Cc: [email protected]
Reported-by: Ben Marzinski <[email protected]>
Signed-off-by: Mike Snitzer <[email protected]>

dm writecache: handle DAX to partitions on persistent memory correctly

The function dax_direct_access doesn't take partitions into account,
it always maps pages from the beginning of the device. Therefore,
persistent_memory_claim() must get the partition offset using
get_start_sect() and add it to the page offsets passed to
dax_direct_access().

Signed-off-by: Mikulas Patocka <[email protected]>
Fixes: 48debafe4f2f ("dm: add writecache target")
Cc: [email protected] # 4.18+
Signed-off-by: Mike Snitzer <[email protected]>

net: ethernet: mlx4: Fix memory allocation in mlx4_buddy_init()

On machines with much memory (> 2 TByte) and log_mtts_per_seg == 0, a
max_order of 31 will be passed to mlx_buddy_init(), which results in
s = BITS_TO_LONGS(1 << 31) becoming a negative value, leading to
kvmalloc_array() failure when it is converted to size_t.

  mlx4_core 0000:b1:00.0: Failed to initialize memory region table, aborting
  mlx4_core: probe of 0000:b1:00.0 failed with error -12

Fix this issue by changing the left shifting operand from a signed literal to
an unsigned one.

Fixes: 225c7b1feef1 ("IB/mlx4: Add a driver Mellanox ConnectX InfiniBand adapters")
Signed-off-by: Shung-Hsi Yu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>

cpufreq: intel_pstate: Fix intel_pstate_get_hwp_max() for turbo disabled

This fixes the behavior of the scaling_max_freq and scaling_min_freq
sysfs files in systems which had turbo disabled by the BIOS.

Caleb noticed that the HWP is programmed to operate in the wrong
P-state range on his system when the CPUFREQ policy min/max frequency
is set via sysfs.  This seems to be because in his system
intel_pstate_get_hwp_max() is returning the maximum turbo P-state even
though turbo was disabled by the BIOS, which causes intel_pstate to
scale kHz frequencies incorrectly e.g. setting the maximum turbo
frequency whenever the maximum guaranteed frequency is requested via
sysfs.

Tested-by: Caleb Callaway <[email protected]>
Signed-off-by: Francisco Jerez <[email protected]>
Acked-by: Srinivas Pandruvada <[email protected]>
[ rjw: Minor subject edits ]
Signed-off-by: Rafael J. Wysocki <[email protected]>

cpufreq: intel_pstate: Free memory only when turning off

When intel_pstate switches the operation mode from "active" to
"passive" or the other way around, freeing its data structures
representing CPUs and allocating them again from scratch is not
necessary and wasteful.  Moreover, if these data structures are
preserved, the cached HWP Request MSR value from there may be
written to the MSR to start with to reinitialize it and help to
restore the EPP value set previously (it is set to 0xFF when CPUs
go offline to allow their SMT siblings to use the full range of
EPP values and that also happens when the driver gets unregistered).

Accordingly, modify the driver to only do a full cleanup on driver
object registration errors and when its status is changed to "off"
via sysfs and to write the cached HWP Request MSR value back to
the MSR on CPU init if the data structure representing the given
CPU is still there.

Signed-off-by: Rafael J. Wysocki <[email protected]>
Acked-by: Srinivas Pandruvada <[email protected]>

cpufreq: intel_pstate: Add ->offline and ->online callbacks

Add ->offline and ->online driver callbacks to prepare for taking a
CPU offline and to restore its working configuration when it goes
back online, respectively, to avoid invoking the ->init callback on
every CPU online which is quite a bit of unnecessary overhead.

Define ->offline and ->online so that they can be used in the
passive mode as well as in the active mode and because ->offline
will do the majority of ->stop_cpu work, the passive mode does
not need that callback any more, so drop it from there.

Also modify the active mode ->suspend and ->resume callbacks to
prevent them from interfering with the new ->offline and ->online
ones in case the latter are invoked withing the system-wide suspend
and resume code flow and make the passive mode use them too.

Signed-off-by: Rafael J. Wysocki <[email protected]>
Acked-by: Srinivas Pandruvada <[email protected]>

cpufreq: intel_pstate: Tweak the EPP sysfs interface

Modify the EPP sysfs interface to reject attempts to change the EPP
to values different from 0 ("performance") in the active mode with
the "performance" policy (ie. scaling_governor set to "performance"),
to avoid situations in which the kernel appears to discard data
passed to it via the EPP sysfs attribute.

Signed-off-by: Rafael J. Wysocki <[email protected]>
Reviewed-by: Artem Bityutskiy <[email protected]>
Acked-by: Srinivas Pandruvada <[email protected]>

cpufreq: intel_pstate: Update cached EPP in the active mode

Make intel_pstate update the cached EPP value when setting the EPP
via sysfs in the active mode just like it is the case in the passive
mode, for consistency, but also for the benefit of subsequent
changes.

No intentional functional impact.

Signed-off-by: Rafael J. Wysocki <[email protected]>
Acked-by: Srinivas Pandruvada <[email protected]>

cpufreq: intel_pstate: Refuse to turn off with HWP enabled

After commit f6ebbcf08f37 ("cpufreq: intel_pstate: Implement passive
mode with HWP enabled") it is possible to change the driver status
to "off" via sysfs with HWP enabled, which effectively causes the
driver to unregister itself, but HWP remains active and it forces the
minimum performance, so even if another cpufreq driver is loaded,
it will not be able to control the CPU frequency.

For this reason, make the driver refuse to change the status to
"off" with HWP enabled.

Signed-off-by: Rafael J. Wysocki <[email protected]>
Acked-by: Srinivas Pandruvada <[email protected]>

ARC: [plat-hsdk]: Switch ethernet phy-mode to rgmii-id

HSDK board has Micrel KSZ9031, recent commit
bcf3440c6dd ("net: phy: micrel: add phy-mode support for the KSZ9031 PHY")
caused a breakdown of Ethernet.
Using 'phy-mode = "rgmii"' is not correct because accodring RGMII
specification it is necessary to have delay on RX (PHY to MAX)
which is not generated in case of "rgmii".
Using "rgmii-id" adds necessary delay and solves the issue.

Also adding name of PHY placed on HSDK board.

Signed-off-by: Evgeniy Didin <[email protected]>
Cc: Eugeniy Paltsev <[email protected]>
Cc: Alexey Brodkin <[email protected]>
Signed-off-by: Vineet Gupta <[email protected]>

pktgen: fix error message with wrong function name

Error on calling kthread_create_on_node prints wrong function name,
kernel_thread.

Fixes: 94dcf29a11b3 ("kthread: use kthread_create_on_node()")
Signed-off-by: Leesoo Ahn <[email protected]>
Acked-by: Gustavo A. R. Silva <[email protected]>
Signed-off-by: David S. Miller <[email protected]>

arc: fix memory initialization for systems with two memory banks

Rework of memory map initialization broke initialization of ARC systems
with two memory banks. Before these changes, memblock was not aware of
nodes configuration and the memory map was always allocated from the
"lowmem" bank. After the addition of node information to memblock, the core
mm attempts to allocate the memory map for the "highmem" bank from its
node. The access to this memory using __va() fails because it can be only
accessed using kmap.

Anther problem that was uncovered is that {min,max}_high_pfn are calculated
from u64 high_mem_start variable which prevents truncation to 32-bit
physical address and the PFN values are above the node and zone boundaries.

Use phys_addr_t type for high_mem_start and high_mem_size to ensure
correspondence between PFNs and highmem zone boundaries and reserve the
entire highmem bank until mem_init() to avoid accesses to it before highmem
is enabled.

To test this:
1. Enable HIGHMEM in ARC config
2. Enable 2 memory banks in haps_hs.dts (uncomment the 2nd bank)

Fixes: 51930df5801e ("mm: free_area_init: allow defining max_zone_pfn in descending order")
Cc: [email protected] [5.8]
Signed-off-by: Mike Rapoport <[email protected]>
Signed-off-by: Vineet Gupta <[email protected]>
[vgupta: added instructions to test highmem]

Merge branch 'opp/fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/vireshk/pm

Pull operating performance points (OPP) framework fixes for 5.9-rc4 from
Viresh Kumar:

"This fixes reference counting for OPP tables. Few patches are getting
 queued (for various subsystems) for 5.10 which depend on this to be
 fixed first."

* 'opp/fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/vireshk/pm:
  opp: Don't drop reference for an OPP table that was never parsed

ia64: fix min_low_pfn/max_low_pfn build errors

Fix min_low_pfn/max_low_pfn build errors for arch/ia64/: (e.g.)

 ERROR: "max_low_pfn" [drivers/rpmsg/virtio_rpmsg_bus.ko] undefined!
 ERROR: "min_low_pfn" [drivers/rpmsg/virtio_rpmsg_bus.ko] undefined!
 ERROR: "min_low_pfn" [drivers/hwtracing/intel_th/intel_th_msu.ko] undefined!
 ERROR: "max_low_pfn" [drivers/hwtracing/intel_th/intel_th_msu.ko] undefined!
 ERROR: "min_low_pfn" [drivers/crypto/cavium/nitrox/n5pf.ko] undefined!
 ERROR: "max_low_pfn" [drivers/crypto/cavium/nitrox/n5pf.ko] undefined!
 ERROR: "max_low_pfn" [drivers/md/dm-integrity.ko] undefined!
 ERROR: "min_low_pfn" [drivers/md/dm-integrity.ko] undefined!
 ERROR: "max_low_pfn" [crypto/tcrypt.ko] undefined!
 ERROR: "min_low_pfn" [crypto/tcrypt.ko] undefined!
 ERROR: "min_low_pfn" [security/keys/encrypted-keys/encrypted-keys.ko] undefined!
 ERROR: "max_low_pfn" [security/keys/encrypted-keys/encrypted-keys.ko] undefined!
 ERROR: "min_low_pfn" [arch/ia64/kernel/mca_recovery.ko] undefined!
 ERROR: "max_low_pfn" [arch/ia64/kernel/mca_recovery.ko] undefined!

David suggested just exporting min_low_pfn & max_low_pfn in
mm/memblock.c:
https://lore.kernel.org/lkml/alpine.DEB.2.22.394.2006291911220.1118534@chino.kir.corp.google.com/

Reported-by: kernel test robot <[email protected]>
Signed-off-by: Randy Dunlap <[email protected]>
Acked-by: David Rientjes <[email protected]>
Acked-by: Tony Luck <[email protected]>
Cc: [email protected]
Cc: Andrew Morton <[email protected]>
Cc: David Rientjes <[email protected]>
Cc: Mike Rapoport <[email protected]>
Cc: Tony Luck <[email protected]>
Cc: Fenghua Yu <[email protected]>
Cc: [email protected]
Signed-off-by: Mike Rapoport <[email protected]>

Merge tag 'misc-habanalabs-fixes-2020-08-31' of git://people.freedesktop.org/~gabbayo/linux into char-misc-nexxt

Oded writes:

This tag contains the two fixes:

- prevent user buffer overflow through a debugfs entry

- fix the report of the engines that create RAZWI error in GAUDI

* tag 'misc-habanalabs-fixes-2020-08-31' of git://people.freedesktop.org/~gabbayo/linux:
  habanalabs: fix report of RAZWI initiator coordinates
  habanalabs: prevent user buff overflow

perf report: Disable ordered_events for raw dump

Disable ordered_events for report raw dump, because for raw dump we want
to see events as they are stored in the perf.data file, not sorted by
time.

Signed-off-by: Jiri Olsa <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Michael Petlan <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Link: http://lore.kernel.org/lkml/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf tools: Correct SNOOPX field offset

perf_event.h has macros that define the field offsets in the data_src
bitmask in perf records. The SNOOPX and REMOTE offsets were both 37.

These are distinct fields, and the bitfield layout in perf_mem_data_src
confirms that SNOOPX should be at offset 38.

Committer notes:

This was extracted from a larger patch that also contained kernel
changes.

Fixes: 52839e653b5629bd ("perf tools: Add support for printing new mem_info encodings")
Signed-off-by: Al Grant <[email protected]>
Reviewed-by: Andi Kleen <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Link: http://lore.kernel.org/lkml/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf intel-pt: Fix corrupt data after perf inject from

Commit 42bbabed09ce6208 ("perf tools: Add hw_idx in struct branch_stack")
changed the format of branch stacks in perf samples. When samples use
this new format, a flag must be set in the corresponding event.

Synthesized branch stacks generated from Intel PT were using the new
format, but not setting the event attribute, leading to consumers
seeing corrupt data. This patch fixes the issue by setting the event
attribute to indicate use of the new format.

Fixes: 42bbabed09ce6208 ("perf tools: Add hw_idx in struct branch_stack")
Signed-off-by: Al Grant <[email protected]>
Acked-by: Adrian Hunter <[email protected]>
Reviewed-by: Mathieu Poirier <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Kan Liang <[email protected]>
Cc: Leo Yan <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Mike Leach <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Suzuki Poulouse <[email protected]>
Cc: [email protected]
Link: http://lore.kernel.org/lkml/[email protected]
Signed-off-by: Leo Yan <[email protected]>
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf cs-etm: Fix corrupt data after perf inject from

Commit 42bbabed09ce6208 ("perf tools: Add hw_idx in struct branch_stack")
changed the format of branch stacks in perf samples. When samples use
this new format, a flag must be set in the corresponding event.

Synthesized branch stacks generated from CoreSight ETM trace were using
the new format, but not setting the event attribute, leading to
consumers seeing corrupt data. This patch fixes the issue by setting the
event attribute to indicate use of the new format.

Fixes: 42bbabed09ce6208 ("perf tools: Add hw_idx in struct branch_stack")
Signed-off-by: Al Grant <[email protected]>
Reviewed-by: Andrea Brunato <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Kan Liang <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Mathieu Poirier <[email protected]>
Cc: Mike Leach <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Suzuki Poulouse <[email protected]>
Cc: [email protected]
Signed-off-by: Leo Yan <[email protected]>
Link: http://lore.kernel.org/lkml/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf top/report: Fix infinite loop in the TUI for grouped events

For a while we need to have a dummy event for doing things like
receiving PERF_RECORD_COMM, PERF_RECORD_EXEC, etc for threads being
created and dying while we synthesize the pre-existing ones at tool
start.

This 'dummy' event is needed for keeping track of thread lifetime events
early in the session but are uninteresting otherwise, i.e. no need to
have it in a initial events menu for the non-grouped case, i.e. for:

 # perf top -e cycles,instructions

or even for plain:

 # perf top

When 'cycles' and that 'dummy' event are in place.

The code to remove that 'dummy' event ended up creating an endless loop
for the grouped case, i.e.:

 # perf top -e '{cycles,instructions}'

Fix it.

Fixes: bee9ca1c8a237ca1 ("perf report TUI: Remove needless 'dummy' event from menu")
Cc: Adrian Hunter <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Namhyung Kim <[email protected]>
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf parse-events: Avoid an uninitialized read when using fake PMUs

With a fake_pmu the pmu_info isn't populated by perf_pmu__check_alias.
In this case, don't try to copy the uninitialized values to the evsel.

Signed-off-by: Ian Rogers <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Jin Yao <[email protected]>
Cc: Kajol Jain <[email protected]>
Cc: Kan Liang <[email protected]>
Cc: Leo Yan <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Stephane Eranian <[email protected]>
Link: http://lore.kernel.org/lkml/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf stat: Fix out of bounds array access in the print_counters() evlist method

Fix a compile error on F32 and gcc version 10.1 on s390 in file
utils/stat-display.c.  The error does not show up with make DEBUG=y.  In
fact the issue shows up when using both compiler options -O6 and
-D_FORTIFY_SOURCE=2 (which are omitted with DEBUG=Y).

This is the offending call chain:

print_counter_aggr()
  printout(config, -1, 0, ...)  with 2nd parm id set to -1
    aggr_printout(config, x, id --> -1, ...) which leads to this code:
case AGGR_NONE:
                if (evsel->percore && !config->percore_show_thread) {
                        ....
                } else {
                        fprintf(config->output, "CPU%*d%s",
                                config->csv_output ? 0 : -7,
                                evsel__cpus(evsel)->map[id],
                        ^^ id is -1 !!!!
                                config->csv_sep);
                }

This is a compiler inlining issue which is detected on s390 but not on
other plattforms.

Output before:

 # make util/stat-display.o
    .....

  util/stat-display.c: In function ‘perf_evlist__print_counters’:
  util/stat-display.c:121:4: error: array subscript -1 is below array
      bounds of ‘int[]’ [-Werror=array-bounds]
  121 |    fprintf(config->output, "CPU%*d%s",
      |    ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  122 |     config->csv_output ? 0 : -7,
      |     ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  123 |     evsel__cpus(evsel)->map[id],
      |     ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  124 |     config->csv_sep);
      |     ~~~~~~~~~~~~~~~~
  In file included from util/evsel.h:13,
                 from util/evlist.h:13,
                 from util/stat-display.c:9:
  /root/linux/tools/lib/perf/include/internal/cpumap.h:10:7:
  note: while referencing ‘map’
   10 |  int  map[];
      |       ^~~
  cc1: all warnings being treated as errors
  mv: cannot stat 'util/.stat-display.o.tmp': No such file or directory
  make[3]: *** [/root/linux/tools/build/Makefile.build:97: util/stat-display.o]
  Error 1
  make[2]: *** [Makefile.perf:716: util/stat-display.o] Error 2
  make[1]: *** [Makefile.perf:231: sub-make] Error 2
  make: *** [Makefile:110: util/stat-display.o] Error 2
  [root@t35lp46 perf]#

Output after:

  # make util/stat-display.o
    .....
  CC       util/stat-display.o
  [root@t35lp46 perf]#

Committer notes:

Removed the removal of {} enclosing the multiline else block, as pointed
out by Jiri Olsa.

Suggested-by: Jiri Olsa <[email protected]>
Signed-off-by: Thomas Richter <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Cc: Heiko Carstens <[email protected]>
Cc: Sumanth Korikkar <[email protected]>
Cc: Sven Schnelle <[email protected]>
Cc: Vasily Gorbik <[email protected]>
Link: http://lore.kernel.org/lkml/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf test: Set NULL sentinel in pmu_events table in "Parse and process metrics" test

Linux 5.9 introduced perf test case "Parse and process metrics" and
on s390 this test case always dumps core:

  [root@t35lp67 perf]# ./perf test -vvvv -F 67
  67: Parse and process metrics                             :
  --- start ---
  metric expr inst_retired.any / cpu_clk_unhalted.thread for IPC
  parsing metric: inst_retired.any / cpu_clk_unhalted.thread
  Segmentation fault (core dumped)
  [root@t35lp67 perf]#

I debugged this core dump and gdb shows this call chain:

  (gdb) where
   #0  0x000003ffabc3192a in __strnlen_c_1 () from /lib64/libc.so.6
   #1  0x000003ffabc293de in strcasestr () from /lib64/libc.so.6
   #2  0x0000000001102ba2 in match_metric(list=0x1e6ea20 "inst_retired.any",
            n=<optimized out>)
       at util/metricgroup.c:368
   #3  find_metric (map=<optimized out>, map=<optimized out>,
           metric=0x1e6ea20 "inst_retired.any")
      at util/metricgroup.c:765
   #4  __resolve_metric (ids=0x0, map=<optimized out>, metric_list=0x0,
           metric_no_group=<optimized out>, m=<optimized out>)
      at util/metricgroup.c:844
   #5  resolve_metric (ids=0x0, map=0x0, metric_list=0x0,
          metric_no_group=<optimized out>)
      at util/metricgroup.c:881
   #6  metricgroup__add_metric (metric=<optimized out>,
        metric_no_group=metric_no_group@entry=false, events=<optimized out>,
        events@entry=0x3ffd84fb878, metric_list=0x0,
        metric_list@entry=0x3ffd84fb868, map=0x0)
      at util/metricgroup.c:943
   #7  0x00000000011034ae in metricgroup__add_metric_list (map=0x13f9828 <map>,
        metric_list=0x3ffd84fb868, events=0x3ffd84fb878,
        metric_no_group=<optimized out>, list=<optimized out>)
      at util/metricgroup.c:988
   #8  parse_groups (perf_evlist=perf_evlist@entry=0x1e70260,
          str=str@entry=0x12f34b2 "IPC", metric_no_group=<optimized out>,
          metric_no_merge=<optimized out>,
          fake_pmu=fake_pmu@entry=0x1462f18 <perf_pmu.fake>,
          metric_events=0x3ffd84fba58, map=0x1)
      at util/metricgroup.c:1040
   #9  0x0000000001103eb2 in metricgroup__parse_groups_test(
   evlist=evlist@entry=0x1e70260, map=map@entry=0x13f9828 <map>,
   str=str@entry=0x12f34b2 "IPC",
   metric_no_group=metric_no_group@entry=false,
   metric_no_merge=metric_no_merge@entry=false,
   metric_events=0x3ffd84fba58)
      at util/metricgroup.c:1082
   #10 0x00000000010c84d8 in __compute_metric (ratio2=0x0, name2=0x0,
          ratio1=<synthetic pointer>, name1=0x12f34b2 "IPC",
   vals=0x3ffd84fbad8, name=0x12f34b2 "IPC")
      at tests/parse-metric.c:159
   #11 compute_metric (ratio=<synthetic pointer>, vals=0x3ffd84fbad8,
   name=0x12f34b2 "IPC")
      at tests/parse-metric.c:189
   #12 test_ipc () at tests/parse-metric.c:208
.....
..... omitted many more lines

This test case was added with
commit 218ca91df477 ("perf tests: Add parse metric test for frontend metric").

When I compile with make DEBUG=y it works fine and I do not get a core dump.

It turned out that the above listed function call chain worked on a struct
pmu_event array which requires a trailing element with zeroes which was
missing. The marco map_for_each_event() loops over that array tests for members
metric_expr/metric_name/metric_group being non-NULL. Adding this element fixes
the issue.

Output after:

  [root@t35lp46 perf]# ./perf test 67
  67: Parse and process metrics                             : Ok
  [root@t35lp46 perf]#

Committer notes:

As Ian remarks, this is not s390 specific:

<quote Ian>
  This also shows up with address sanitizer on all architectures
  (perhaps change the patch title) and perhaps add a "Fixes: <commit>"
  tag.

  =================================================================
  ==4718==ERROR: AddressSanitizer: global-buffer-overflow on address
  0x55c93b4d59e8 at pc 0x55c93a1541e2 bp 0x7ffd24327c60 sp
  0x7ffd24327c58
  READ of size 8 at 0x55c93b4d59e8 thread T0
      #0 0x55c93a1541e1 in find_metric tools/perf/util/metricgroup.c:764:2
      #1 0x55c93a153e6c in __resolve_metric tools/perf/util/metricgroup.c:844:9
      #2 0x55c93a152f18 in resolve_metric tools/perf/util/metricgroup.c:881:9
      #3 0x55c93a1528db in metricgroup__add_metric
  tools/perf/util/metricgroup.c:943:9
      #4 0x55c93a151996 in metricgroup__add_metric_list
  tools/perf/util/metricgroup.c:988:9
      #5 0x55c93a1511b9 in parse_groups tools/perf/util/metricgroup.c:1040:8
      #6 0x55c93a1513e1 in metricgroup__parse_groups_test
  tools/perf/util/metricgroup.c:1082:9
      #7 0x55c93a0108ae in __compute_metric tools/perf/tests/parse-metric.c:159:8
      #8 0x55c93a010744 in compute_metric tools/perf/tests/parse-metric.c:189:9
      #9 0x55c93a00f5ee in test_ipc tools/perf/tests/parse-metric.c:208:2
      #10 0x55c93a00f1e8 in test__parse_metric
  tools/perf/tests/parse-metric.c:345:2
      #11 0x55c939fd7202 in run_test tools/perf/tests/builtin-test.c:410:9
      #12 0x55c939fd6736 in test_and_print tools/perf/tests/builtin-test.c:440:9
      #13 0x55c939fd58c3 in __cmd_test tools/perf/tests/builtin-test.c:661:4
      #14 0x55c939fd4e02 in cmd_test tools/perf/tests/builtin-test.c:807:9
      #15 0x55c939e4763d in run_builtin tools/perf/perf.c:313:11
      #16 0x55c939e46475 in handle_internal_command tools/perf/perf.c:365:8
      #17 0x55c939e4737e in run_argv tools/perf/perf.c:409:2
      #18 0x55c939e45f7e in main tools/perf/perf.c:539:3

  0x55c93b4d59e8 is located 0 bytes to the right of global variable
  'pme_test' defined in 'tools/perf/tests/parse-metric.c:17:25'
  (0x55c93b4d54a0) of size 1352
  SUMMARY: AddressSanitizer: global-buffer-overflow
  tools/perf/util/metricgroup.c:764:2 in find_metric
  Shadow bytes around the buggy address:
    0x0ab9a7692ae0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
    0x0ab9a7692af0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
    0x0ab9a7692b00: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
    0x0ab9a7692b10: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
    0x0ab9a7692b20: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
  =>0x0ab9a7692b30: 00 00 00 00 00 00 00 00 00 00 00 00 00[f9]f9 f9
    0x0ab9a7692b40: f9 f9 f9 f9 f9 f9 f9 f9 f9 f9 f9 f9 f9 f9 f9 f9
    0x0ab9a7692b50: f9 f9 f9 f9 f9 f9 f9 f9 f9 f9 f9 f9 f9 f9 f9 f9
    0x0ab9a7692b60: f9 f9 f9 f9 f9 f9 f9 f9 00 00 00 00 00 00 00 00
    0x0ab9a7692b70: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
    0x0ab9a7692b80: f9 f9 f9 f9 f9 f9 f9 f9 f9 f9 f9 f9 f9 f9 f9 f9
  Shadow byte legend (one shadow byte represents 8 application bytes):
    Addressable:           00
    Partially addressable: 01 02 03 04 05 06 07
    Heap left redzone:    fa
    Freed heap region:    fd
    Stack left redzone:    f1
    Stack mid redzone:    f2
    Stack right redzone:     f3
    Stack after return:    f5
    Stack use after scope:   f8
    Global redzone:          f9
    Global init order:    f6
    Poisoned by user:        f7
    Container overflow:    fc
    Array cookie:            ac
    Intra object redzone:    bb
    ASan internal:           fe
    Left alloca redzone:     ca
    Right alloca redzone:    cb
    Shadow gap:              cc
</quote>

I'm also adding the missing "Fixes" tag and setting just .name to NULL,
as doing it that way is more compact (the compiler will zero out
everything else) and the table iterators look for .name being NULL as
the sentinel marking the end of the table.

Fixes: 0a507af9c681ac2a ("perf tests: Add parse metric test for ipc metric")
Signed-off-by: Thomas Richter <[email protected]>
Reviewed-by: Sumanth Korikkar <[email protected]>
Acked-by: Ian Rogers <[email protected]>
Cc: Heiko Carstens <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Sven Schnelle <[email protected]>
Cc: Vasily Gorbik <[email protected]>
Link: http://lore.kernel.org/lkml/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf parse-events: Set exclude_guest=1 for user-space counting

Currently if we run 'perf record -e cycles:u', exclude_guest=0.

But it doesn't make sense in most cases that we request for
user-space counting but we also get the guest report.

Of course, we also need to consider 'perf kvm' usage case that
authorized perf users on the host may only want to count guest user
space events. For example,

  # perf kvm --guest record -e cycles:u

When we have 'exclude_guest=1' for 'perf kvm' usage, we may get nothing
from guest events.

To keep perf semantics consistent and clear, this patch sets
exclude_guest=1 for user-space counting but except for 'perf kvm' usage.

Before:

  perf record -e cycles:u ./div
  perf evlist -v
  cycles:u: ..., exclude_kernel: 1, exclude_hv: 1, ...

After:
  perf record -e cycles:u ./div
  perf evlist -v
  cycles:u: ..., exclude_kernel: 1, exclude_hv: 1,  exclude_guest: 1, ...

Before:
  perf kvm --guest record -e cycles:u -vvv

perf_event_attr:

  size                             120
  { sample_period, sample_freq }   4000
  sample_type                      IP|TID|TIME|ID|CPU|PERIOD
  read_format                      ID
  disabled                         1
  inherit                          1
  exclude_kernel                   1
  exclude_hv                       1
  freq                             1
  sample_id_all                    1

After:

  perf kvm --guest record -e cycles:u -vvv

perf_event_attr:
  size                             120
  { sample_period, sample_freq }   4000
  sample_type                      IP|TID|TIME|ID|CPU|PERIOD
  read_format                      ID
  disabled                         1
  inherit                          1
  exclude_kernel                   1
  exclude_hv                       1
  freq                             1
  sample_id_all                    1

For Before/After, exclude_guest are both 0 for perf kvm usage.

perf test 6

 6: Parse event definition strings             : Ok

Signed-off-by: Jin Yao <[email protected]>
Tested-by: Like Xu <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Jin Yao <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Kan Liang <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Link: http://lore.kernel.org/lkml/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf record: Correct the help info of option "--no-bpf-event"

The help info of option "--no-bpf-event" is wrongly described as "record
bpf events", correct it.

Committer testing:

  $ perf record -h bpf

   Usage: perf record [<options>] [<command>]
      or: perf record [<options>] -- <command> [<options>]

          --clang-opt <clang options>
                            options passed to clang when compiling BPF scriptlets
          --clang-path <clang path>
                            clang binary to use for compiling BPF scriptlets
          --no-bpf-event    do not record bpf events

  $

Fixes: 71184c6ab7e6 ("perf record: Replace option --bpf-event with --no-bpf-event")
Signed-off-by: Wei Li <[email protected]>
Acked-by: Song Liu <[email protected]>
Tested-by: Arnaldo Carvalho de Melo <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Hanjun Guo <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Li Bin <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Namhyung Kim <[email protected]>
Link: http://lore.kernel.org/lkml/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf tools: Use %zd for size_t printf formats on 32-bit

A couple of trivial fixes for using %zd for size_t in the code
supporting the ZSTD compression library.

Signed-off-by: Chris Wilson <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Alexey Budankov <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Link: http://lore.kernel.org/lkml/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

block: fix locking in bdev_del_partition

We need to hold the whole device bd_mutex to protect against
other thread concurrently deleting out partition before we get
to it, and thus causing a use after free.

Fixes: cddae808aeb7 ("block: pass a hd_struct to delete_partition")
Reported-by: [email protected]
Signed-off-by: Christoph Hellwig <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>

block: release disk reference in hd_struct_free_work

Commit e8c7d14ac6c3 ("block: revert back to synchronous request_queue removal")
stops to release request queue from wq context because that commit
supposed all blk_put_queue() is called in context which is allowed
to sleep. However, this assumption isn't true because we release disk's
reference in partition's percpu_ref's ->release() which doesn't allow
to sleep, because the ->release() is run via call_rcu().

Fixes this issue by moving put disk reference into hd_struct_free_work()

Fixes: e8c7d14ac6c3 ("block: revert back to synchronous request_queue removal")
Reported-by: Ilya Dryomov <[email protected]>
Signed-off-by: Ming Lei <[email protected]>
Tested-by: Ilya Dryomov <[email protected]>
Reviewed-by: Christoph Hellwig <[email protected]>
Cc: Luis Chamberlain <[email protected]>
Cc: Christoph Hellwig <[email protected]>
Cc: Bart Van Assche <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>

iio: adc: mcp3422: fix locking on error path

Reading from the chip should be unlocked on error path else the lock
could never being released.

Fixes: 07914c84ba30 ("iio: adc: Add driver for Microchip MCP3422/3/4 high resolution ADC")
Fixes: 3f1093d83d71 ("iio: adc: mcp3422: fix locking scope")
Acked-by: Jonathan Cameron <[email protected]>
Signed-off-by: Angelo Compagnucci <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

io_uring: fix removing the wrong file in __io_sqe_files_update()

Index here is already the position of the file in fixed_file_table, we
should not use io_file_from_index() again to get it. Otherwise, the
wrong file which still in use may be released unexpectedly.

Cc: [email protected] # v5.6
Fixes: 05f3fb3c5397 ("io_uring: avoid ring quiesce for fixed file set unregister and update")
Signed-off-by: Jiufei Xue <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>

block: ensure bdi->io_pages is always initialized

If a driver leaves the limit settings as the defaults, then we don't
initialize bdi->io_pages. This means that file systems may need to
work around bdi->io_pages == 0, which is somewhat messy.

Initialize the default value just like we do for ->ra_pages.

Cc: [email protected]
Fixes: 9491ae4aade6 ("mm: don't cap request size based on read-ahead setting")
Reported-by: OGAWA Hirofumi <[email protected]>
Reviewed-by: Christoph Hellwig <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>

ALSA: pcm: oss: Remove superfluous WARN_ON() for mulaw sanity check

The PCM OSS mulaw plugin has a check of the format of the counter part
whether it's a linear format.  The check is with snd_BUG_ON() that
emits WARN_ON() when the debug config is set, and it confuses
syzkaller as if it were a serious issue.  Let's drop snd_BUG_ON() for
avoiding that.

While we're at it, correct the error code to a more suitable, EINVAL.

Reported-by: [email protected]
Cc: <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Takashi Iwai <[email protected]>

regulator: core: Fix slab-out-of-bounds in regulator_unlock_recursive()

The recent commit 7d8196641ee1 ("regulator: Remove pointer table
overallocation") changed the size of coupled_rdevs and now KASAN is able
to detect slab-out-of-bounds problem in regulator_unlock_recursive(),
which is a legit problem caused by a typo in the code. The recursive
unlock function uses n_coupled value of a parent regulator for unlocking
supply regulator, while supply's n_coupled should be used. In practice
problem may only affect platforms that use coupled regulators.

Cc: [email protected] # 5.0+
Fixes: f8702f9e4aa7 ("regulator: core: Use ww_mutex for regulators locking")
Signed-off-by: Dmitry Osipenko <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>

MIPS: perf: Fix wrong check condition of Loongson event IDs

According to the user's manual chapter 8.2.1 of Loongson 3A2000 CPU [1]
and 3A3000 CPU [2], we should take some event IDs such as 274, 358, 359
and 360 as valid in the check condition, otherwise they are recognized
as "not supported", fix it.

[1] http://www.loongson.cn/uploadfile/cpu/3A2000/Loongson3A2000_user2.pdf
[2] http://www.loongson.cn/uploadfile/cpu/3A3000/Loongson3A3000_3B3000user2.pdf

Fixes: e9dfbaaeef1c ("MIPS: perf: Add hardware perf events support for new Loongson-3")
Signed-off-by: Tiezhu Yang <[email protected]>
Acked-by: Huang Pei <[email protected]>
Signed-off-by: Thomas Bogendoerfer <[email protected]>

clang-format: Update with the latest for_each macro list

Re-run the shell fragment that generated the original list.

Signed-off-by: Miguel Ojeda <[email protected]>

HID: core: Sanitize event code and type when mapping input

When calling into hid_map_usage(), the passed event code is
blindly stored as is, even if it doesn't fit in the associated bitmap.

This event code can come from a variety of sources, including devices
masquerading as input devices, only a bit more "programmable".

Instead of taking the event code at face value, check that it actually
fits the corresponding bitmap, and if it doesn't:
- spit out a warning so that we know which device is acting up
- NULLify the bitmap pointer so that we catch unexpected uses

Code paths that can make use of untrusted inputs can now check
that the mapping was indeed correct and bail out if not.

Cc: [email protected]
Signed-off-by: Marc Zyngier <[email protected]>
Signed-off-by: Benjamin Tissoires <[email protected]>

HID: core: Correctly handle ReportSize being zero

It appears that a ReportSize value of zero is legal, even if a bit
non-sensical. Most of the HID code seems to handle that gracefully,
except when computing the total size in bytes. When fed as input to
memset, this leads to some funky outcomes.

Detect the corner case and correctly compute the size.

Cc: [email protected]
Signed-off-by: Marc Zyngier <[email protected]>
Signed-off-by: Benjamin Tissoires <[email protected]>

Documentation: fix dma-buf.rst underline length warning

/home/rdunlap/lnx/lnx-59-rc2/Documentation/driver-api/dma-buf.rst:182: WARNING: Title underline too short.
Indefinite DMA Fences
~~~~~~~~~~~~~~~~~~~~

Fixes: 72b6ede73623 ("dma-buf.rst: Document why indefinite fences are a bad idea")
Signed-off-by: Randy Dunlap <[email protected]>
Cc: Daniel Vetter <[email protected]>
Cc: Daniel Vetter <[email protected]>
Cc: Dave Airlie <[email protected]>
Signed-off-by: Daniel Vetter <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

misc: eeprom: at24: register nvmem only after eeprom is ready to use

During nvmem_register() the nvmem core sends notifications when:

    - cell added
    - nvmem added

and during these notifications some callback func may access the nvmem
device, which will fail in case of at24 eeprom because regulator and pm
are enabled after nvmem_register().

Fixes: cd5676db0574 ("misc: eeprom: at24: support pm_runtime control")
Fixes: b20eb4c1f026 ("eeprom: at24: drop unnecessary label")
Cc: [email protected]
Signed-off-by: Vadym Kochan <[email protected]>
Signed-off-by: Bartosz Golaszewski <[email protected]>

Merge tag 'thunderbolt-for-v5.9-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/westeri/thunderbolt into usb-linus

Mika writes:

thunderbolt: Fixes for v5.9-rc4

This includes two fixes, one that fixes a regression around reboot and
other that uses a correct link rate when USB3 bandwidth is reclaimed
when the link is not up.

Both have been in linux-next with no reported issues.

* tag 'thunderbolt-for-v5.9-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/westeri/thunderbolt:
  thunderbolt: Use maximum USB3 link rate when reclaiming if link is not up
  thunderbolt: Disable ports that are not implemented

HID: elan: Fix memleak in elan_input_configured

When input_mt_init_slots() fails, input should be freed
to prevent memleak. When input_register_device() fails,
we should call input_mt_destroy_slots() to free memory
allocated by input_mt_init_slots().

Signed-off-by: Dinghao Liu <[email protected]>
Signed-off-by: Jiri Kosina <[email protected]>

scsi: lpfc: Update lpfc version to 12.8.0.4

Update lpfc version to 12.8.0.4

Link: https://lore.kernel.org/r/[email protected]
Co-developed-by: Dick Kennedy <[email protected]>
Signed-off-by: Dick Kennedy <[email protected]>
Signed-off-by: James Smart <[email protected]>
Signed-off-by: Martin K. Petersen <[email protected]>

scsi: lpfc: Extend the RDF FPIN Registration descriptor for additional events

Currently the driver registers for Link Integrity events only.

This patch adds registration for the following FPIN types:

 - Delivery Notifications
 - Congestion Notification
 - Peer Congestion Notification

Link: https://lore.kernel.org/r/[email protected]
Co-developed-by: Dick Kennedy <[email protected]>
Signed-off-by: Dick Kennedy <[email protected]>
Signed-off-by: James Smart <[email protected]>
Signed-off-by: Martin K. Petersen <[email protected]>

scsi: lpfc: Fix FLOGI/PLOGI receive race condition in pt2pt discovery

The driver is unable to successfully login with remote device. During pt2pt
login, the driver completes its FLOGI request with the remote device having
WWN precedence.  The remote device issues its own (delayed) FLOGI after
accepting the driver's and, upon transmitting the FLOGI, immediately
recognizes it has already processed the driver's FLOGI thus it transitions
to sending a PLOGI before waiting for an ACC to its FLOGI.

In the driver, the FLOGI is received and an ACC sent, followed by the PLOGI
being received and an ACC sent. The issue is that the PLOGI reception
occurs before the response from the adapter from the FLOGI ACC is
received. Processing of the PLOGI sets state flags to perform the REG_RPI
mailbox command and proceed with the rest of discovery on the port. The
same completion routine used by both FLOGI and PLOGI is generic in
nature. One of the things it does is clear flags, and those flags happen to
drive the rest of discovery.  So what happened was the PLOGI processing set
the flags, the FLOGI ACC completion cleared them, thus when the PLOGI ACC
completes it doesn't see the flags and stops.

Fix by modifying the generic completion routine to not clear the rest of
discovery flag (NLP_ACC_REGLOGIN) unless the completion is also associated
with performing a mailbox command as part of its handling.  For things such
as FLOGI ACC, there isn't a subsequent action to perform with the adapter,
thus there is no mailbox cmd ptr. PLOGI ACC though will perform REG_RPI
upon completion, thus there is a mailbox cmd ptr.

Link: https://lore.kernel.org/r/[email protected]
Co-developed-by: Dick Kennedy <[email protected]>
Signed-off-by: Dick Kennedy <[email protected]>
Signed-off-by: James Smart <[email protected]>
Signed-off-by: Martin K. Petersen <[email protected]>

scsi: lpfc: Fix setting IRQ affinity with an empty CPU mask

Some systems are reporting the following log message during driver unload
or system shutdown:

  ics_rtas_set_affinity: No online cpus in the mask

A prior commit introduced the writing of an empty affinity mask in calls to
irq_set_affinity_hint() when disabling interrupts or when there are no
remaining online CPUs to service an eq interrupt. At least some ppc64
systems are checking whether affinity masks are empty or not.

Do not call irq_set_affinity_hint() with an empty CPU mask.

Fixes: dcaa21367938 ("scsi: lpfc: Change default IRQ model on AMD architectures")
Link: https://lore.kernel.org/r/[email protected]
Cc: <[email protected]> # v5.5+
Co-developed-by: Dick Kennedy <[email protected]>
Signed-off-by: Dick Kennedy <[email protected]>
Signed-off-by: James Smart <[email protected]>
Signed-off-by: Martin K. Petersen <[email protected]>

scsi: qla2xxx: Fix regression on sparc64

Commit 98aee70d19a7 ("qla2xxx: Add endianizer to max_payload_size
modifier.") in 2014 broke qla2xxx on sparc64, e.g. as in the Sun Blade 1000
/ 2000. Unbreak by partial revert to fix endianness in nvram firmware
default initialization. Also mark the second frame_payload_size in nvram_t
__le16 to avoid new sparse warnings.

Link: https://lore.kernel.org/r/[email protected]
Fixes: 98aee70d19a7 ("qla2xxx: Add endianizer to max_payload_size modifier.")
Reviewed-by: Himanshu Madhani <[email protected]>
Reviewed-by: Bart Van Assche <[email protected]>
Acked-by: Arun Easi <[email protected]>
Signed-off-by: René Rebe <[email protected]>
Signed-off-by: Martin K. Petersen <[email protected]>

scsi: libfc: Fix for double free()

Fix for '&fp->skb' double free.

Link:
https://lore.kernel.org/r/20200825093940[email protected]
Reported-by: Dan Carpenter <[email protected]>
Signed-off-by: Javed Hasan <[email protected]>
Signed-off-by: Martin K. Petersen <[email protected]>

scsi: pm8001: Fix memleak in pm8001_exec_internal_task_abort

When pm8001_tag_alloc() fails, task should be freed just like it is done in
the subsequent error paths.

Link: https://lore.kernel.org/r/[email protected]
Acked-by: Jack Wang <[email protected]>
Signed-off-by: Dinghao Liu <[email protected]>
Signed-off-by: Martin K. Petersen <[email protected]>

net: ethernet: ti: am65-cpsw: fix rmii 100Mbit link mode

In RMII link mode it's required to set bit 15 IFCTL_A in MAC_SL MAC_CONTROL
register to enable support for 100Mbit link speed.

Fixes: 93a76530316a ("net: ethernet: ti: introduce am65x/j721e gigabit eth subsystem driver")
Signed-off-by: Grygorii Strashko <[email protected]>
Signed-off-by: David S. Miller <[email protected]>

Merge tag 'docs-5.9-3' of git://git.lwn.net/linux

Pull documentation fixes from Jonathan Corbet:
 "A handful of documentation fixes for 5.9"

* tag 'docs-5.9-3' of git://git.lwn.net/linux:
  Documentation: laptops: thinkpad-acpi: fix underline length build warning
  Documentation: fix typo for abituguru documentation
  docs: Fix function name trailing double-()s
  devices.txt: fix typo of "ubd" as "udb"
  Documentation: add riscv entry in list of existing profiles
  MAINTAINERS: mention documentation maintainer entry profile
  Fpga: Documentation: Replace deprecated :c:func: Usage
  IIO: Documentation: Replace deprecated :c:func: Usage
  Documentation/locking/locktypes: fix local_locks documentation

cxgb4: fix thermal zone device registration

When multiple adapters are present in the system, pci hot-removing second
adapter leads to the following warning as both the adapters registered
thermal zone device with same thermal zone name/type.
Therefore, use unique thermal zone name during thermal zone device
initialization. Also mark thermal zone dev NULL once unregistered.

[  414.370143] ------------[ cut here ]------------
[  414.370944] sysfs group 'power' not found for kobject 'hwmon0'
[  414.371747] WARNING: CPU: 9 PID: 2661 at fs/sysfs/group.c:281
 sysfs_remove_group+0x76/0x80
[  414.382550] CPU: 9 PID: 2661 Comm: bash Not tainted 5.8.0-rc6+ #33
[  414.383593] Hardware name: Supermicro X10SRA-F/X10SRA-F, BIOS 2.0a 06/23/2016
[  414.384669] RIP: 0010:sysfs_remove_group+0x76/0x80
[  414.385738] Code: 48 89 df 5b 5d 41 5c e9 d8 b5 ff ff 48 89 df e8 60 b0 ff ff
 eb cb 49 8b 14 24 48 8b 75 00 48 c7 c7 90 ae 13 bb e8 6a 27 d0 ff <0f> 0b 5b 5d
 41 5c c3 0f 1f 00 0f 1f 44 00 00 48 85 f6 74 31 41 54
[  414.388404] RSP: 0018:ffffa22bc080fcb0 EFLAGS: 00010286
[  414.389638] RAX: 0000000000000000 RBX: 0000000000000000 RCX: 0000000000000000
[  414.390829] RDX: 0000000000000001 RSI: ffff8ee2de3e9510 RDI: ffff8ee2de3e9510
[  414.392064] RBP: ffffffffbaef2ee0 R08: 0000000000000000 R09: 0000000000000000
[  414.393224] R10: 0000000000000000 R11: 000000002b30006c R12: ffff8ee260720008
[  414.394388] R13: ffff8ee25e0a40e8 R14: ffffa22bc080ff08 R15: ffff8ee2c3be5020
[  414.395661] FS:  00007fd2a7171740(0000) GS:ffff8ee2de200000(0000)
 knlGS:0000000000000000
[  414.396825] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[  414.398011] CR2: 00007f178ffe5020 CR3: 000000084c5cc003 CR4: 00000000003606e0
[  414.399172] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[  414.400352] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[  414.401473] Call Trace:
[  414.402685]  device_del+0x89/0x400
[  414.403819]  device_unregister+0x16/0x60
[  414.405024]  hwmon_device_unregister+0x44/0xa0
[  414.406112]  thermal_remove_hwmon_sysfs+0x196/0x200
[  414.407256]  thermal_zone_device_unregister+0x1b5/0x1f0
[  414.408415]  cxgb4_thermal_remove+0x3c/0x4f [cxgb4]
[  414.409668]  remove_one+0x212/0x290 [cxgb4]
[  414.410875]  pci_device_remove+0x36/0xb0
[  414.412004]  device_release_driver_internal+0xe2/0x1c0
[  414.413276]  pci_stop_bus_device+0x64/0x90
[  414.414433]  pci_stop_and_remove_bus_device_locked+0x16/0x30
[  414.415609]  remove_store+0x75/0x90
[  414.416790]  kernfs_fop_write+0x114/0x1b0
[  414.417930]  vfs_write+0xcf/0x210
[  414.419059]  ksys_write+0xa7/0xe0
[  414.420120]  do_syscall_64+0x4c/0xa0
[  414.421278]  entry_SYSCALL_64_after_hwframe+0x44/0xa9
[  414.422335] RIP: 0033:0x7fd2a686afd0
[  414.423396] Code: Bad RIP value.
[  414.424549] RSP: 002b:00007fffc1446148 EFLAGS: 00000246 ORIG_RAX:
 0000000000000001
[  414.425638] RAX: ffffffffffffffda RBX: 0000000000000002 RCX: 00007fd2a686afd0
[  414.426830] RDX: 0000000000000002 RSI: 00007fd2a7196000 RDI: 0000000000000001
[  414.427927] RBP: 00007fd2a7196000 R08: 000000000000000a R09: 00007fd2a7171740
[  414.428923] R10: 00007fd2a7171740 R11: 0000000000000246 R12: 00007fd2a6b43400
[  414.430082] R13: 0000000000000002 R14: 0000000000000001 R15: 0000000000000000
[  414.431027] irq event stamp: 76300
[  414.435678] ---[ end trace 13865acb4d5ab00f ]---

Fixes: b18719157762 ("cxgb4: Add thermal zone support")
Signed-off-by: Potnuri Bharat Teja <[email protected]>
Signed-off-by: David S. Miller <[email protected]>

drivers/net/wan/hdlc_cisco: Add hard_header_len

This driver didn't set hard_header_len. This patch sets hard_header_len
for it according to its header_ops->create function.

This driver's header_ops->create function (cisco_hard_header) creates
a header of (struct hdlc_header), so hard_header_len should be set to
sizeof(struct hdlc_header).

Cc: Martin Schiller <[email protected]>
Signed-off-by: Xie He <[email protected]>
Acked-by: Krzysztof Halasa <[email protected]>
Signed-off-by: David S. Miller <[email protected]>

ionic: fix txrx work accounting

Take the tx accounting out of the work_done calculation to
prevent a possible duplicate napi_schedule call when under
high Tx stress but low Rx traffic.

Fixes: b14e4e95f9ec ("ionic: tx separate servicing")
Signed-off-by: Shannon Nelson <[email protected]>
Signed-off-by: David S. Miller <[email protected]>

Merge tag 'mmc-v5.9-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/ulfh/mmc

Pull MMC fixes from Ulf Hansson:

 - Fix HS400 tuning for ACPI ID AMDI0040

 - Fix reset of CQHCI for Intel GLK-based controllers

 - Use correct timeout clock for Tegra186/194/210

 - Fix eMMC mounting on mt7622/Bpi-64

* tag 'mmc-v5.9-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/ulfh/mmc:
  sdhci: tegra: Add missing TMCLK for data timeout
  arm64: tegra: Add missing timeout clock to Tegra194 SDMMC nodes
  arm64: tegra: Add missing timeout clock to Tegra186 SDMMC nodes
  arm64: tegra: Add missing timeout clock to Tegra210 SDMMC
  dt-bindings: mmc: tegra: Add tmclk for Tegra210 and later
  sdhci: tegra: Remove SDHCI_QUIRK_DATA_TIMEOUT_USES_SDCLK for Tegra186
  sdhci: tegra: Remove SDHCI_QUIRK_DATA_TIMEOUT_USES_SDCLK for Tegra210
  arm64: dts: mt7622: add reset node for mmc device
  dt-bindings: mmc: Add missing description for clk_in/out_sd1
  mmc: mediatek: add optional module reset property
  mmc: dt-bindings: Add resets/reset-names for Mediatek MMC bindings
  mmc: sdhci-pci: Fix SDHCI_RESET_ALL for CQHCI for Intel GLK-based controllers
  mmc: sdhci-acpi: Fix HS400 tuning for AMDI0040

Merge git://git.kernel.org/pub/scm/linux/kernel/git/pablo/nf

Pablo Neira Ayuso says:

====================
Netfilter fixes for net

The following patchset contains Netfilter fixes for net:

1) Do not delete clash entries on reply, let them expire instead,
   from Florian Westphal.

2) Do not report EAGAIN to nfnetlink, otherwise this enters a busy loop.
   Update nfnetlink_unicast() to translate EAGAIN to ENOBUFS.

3) Remove repeated words in code comments, from Randy Dunlap.

4) Several patches for the flowtable selftests, from Fabian Frederick.
====================

Signed-off-by: David S. Miller <[email protected]>

microblaze: fix min_low_pfn/max_low_pfn build errors

Fix min_low_pfn/max_low_pfn build errors for arch/microblaze/: (e.g.)

  ERROR: "min_low_pfn" [drivers/rpmsg/virtio_rpmsg_bus.ko] undefined!
  ERROR: "min_low_pfn" [drivers/hwtracing/intel_th/intel_th_msu_sink.ko] undefined!
  ERROR: "min_low_pfn" [drivers/hwtracing/intel_th/intel_th_msu.ko] undefined!
  ERROR: "min_low_pfn" [drivers/mmc/core/mmc_core.ko] undefined!
  ERROR: "min_low_pfn" [drivers/md/dm-crypt.ko] undefined!
  ERROR: "min_low_pfn" [drivers/net/wireless/ath/ath6kl/ath6kl_sdio.ko] undefined!
  ERROR: "min_low_pfn" [crypto/tcrypt.ko] undefined!
  ERROR: "min_low_pfn" [crypto/asymmetric_keys/asym_tpm.ko] undefined!

Mike had/has an alternate patch for Microblaze:
https://lore.kernel.org/lkml/20200630111519[email protected]/

David suggested just exporting min_low_pfn & max_low_pfn in
mm/memblock.c:
https://lore.kernel.org/lkml/alpine.DEB.2.22.394.2006291911220.1118534@chino.kir.corp.google.com/

Reported-by: kernel test robot <[email protected]>
Signed-off-by: Randy Dunlap <[email protected]>
Acked-by: Michal Simek <[email protected]>
Acked-by: David Rientjes <[email protected]>
Cc: [email protected]
Cc: Andrew Morton <[email protected]>
Cc: David Rientjes <[email protected]>
Cc: Mike Rapoport <[email protected]>
Cc: Michal Simek <[email protected]>
Cc: Michal Simek <[email protected]>
Signed-off-by: Mike Rapoport <[email protected]>

drm/sun4i: Fix dsi dcs long write function

It's writing too much data. regmap_bulk_write expects number of
register sized chunks to write, not a byte sized length of the
bounce buffer. Bounce buffer needs to be padded too, so that
regmap_bulk_write will not read past the end of the buffer.

Fixes: 133add5b5ad4 ("drm/sun4i: Add Allwinner A31 MIPI-DSI controller support")
Signed-off-by: Ondrej Jirman <[email protected]>
Signed-off-by: Maxime Ripard <[email protected]>
Reviewed-by: Jernej Skrabec <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

habanalabs: fix report of RAZWI initiator coordinates

All initiator coordinates received upon an 'MMU page fault RAZWI
event' should be the routers coordinates, the only exception is the
DMA initiators for which the reported coordinates correspond to
their actual location.

Signed-off-by: Ofir Bitton <[email protected]>
Reviewed-by: Oded Gabbay <[email protected]>
Signed-off-by: Oded Gabbay <[email protected]>

habanalabs: prevent user buff overflow

This commit fixes a potential debugfs issue that may occur when
reading the clock gating mask into the user buffer since the
user buffer size was not taken into consideration.

Signed-off-by: Moti Haimovski <[email protected]>
Reviewed-by: Oded Gabbay <[email protected]>
Signed-off-by: Oded Gabbay <[email protected]>

Documentation/llvm: Improve formatting of commands, variables, and arguments

While reviewing a separate patch, I noticed that the formatting of the
commands, variables, and arguments was not in a monospaced font like the
rest of the Kbuild documentation (see kbuild/kconfig.rst for an
example). This is due to a lack of "::" before indented command blocks
and single backticks instead of double backticks for inline formatting.

Add those so that the document looks nicer in an HTML format, while not
ruining the look in plain text.

As a result of this, we can remove the escaped backslashes in the last
code block and move them to single backslashes.

Signed-off-by: Nathan Chancellor <[email protected]>
Reviewed-by: Nick Desaulniers <[email protected]>
Signed-off-by: Masahiro Yamada <[email protected]>