Pull fs hole punching vs cache filling race fixes from Jan Kara:
 "Fix races leading to possible data corruption or stale data exposure
  in multiple filesystems when hole punching races with operations such
  as readahead.

  This is the series I was sending for the last merge window but with
  your objection fixed - now filemap_fault() has been modified to take
  invalidate_lock only when we need to create new page in the page cache
  and / or bring it uptodate"

* tag 'hole_punch_for_v5.15-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs:
  filesystems/locking: fix Malformed table warning
  cifs: Fix race between hole punch and page fault
  ceph: Fix race between hole punch and page fault
  fuse: Convert to using invalidate_lock
  f2fs: Convert to using invalidate_lock
  zonefs: Convert to using invalidate_lock
  xfs: Convert double locking of MMAPLOCK to use VFS helpers
  xfs: Convert to use invalidate_lock
  xfs: Refactor xfs_isilocked()
  ext2: Convert to using invalidate_lock
  ext4: Convert to use mapping->invalidate_lock
  mm: Add functions to lock invalidate_lock for two mappings
  mm: Protect operations adding pages to page cache with invalidate_lock
  documentation: Sync file_operations members with reality
  mm: Fix comments mentioning i_mutex

Pull UDF and isofs updates from Jan Kara:
 "Several smaller fixes and cleanups in UDF and isofs"

* tag 'fs_for_v5.15-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs:
  udf_get_extendedattr() had no boundary checks.
  isofs: joliet: Fix iocharset=utf8 mount option
  udf: Fix iocharset=utf8 mount option
  udf: Get rid of 0-length arrays in struct fileIdentDesc
  udf: Get rid of 0-length arrays
  udf: Remove unused declaration
  udf: Check LVID earlier

Pull FIEMAP cleanups from Jan Kara:
 "FIEMAP cleanups from Christoph transitioning all remaining filesystems
  supporting FIEMAP (ext2, hpfs) to iomap API and removing the old
  helper"

* tag 'fiemap_for_v5.15-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs:
  fs: remove generic_block_fiemap
  hpfs: use iomap_fiemap to implement ->fiemap
  ext2: use iomap_fiemap to implement ->fiemap
  ext2: make ext2_iomap_ops available unconditionally

Pull fsnotify updates from Jan Kara:
 "fsnotify speedups when notification actually isn't used and support
  for identifying processes which caused fanotify events through pidfd
  instead of normal pid"

* tag 'fsnotify_for_v5.15-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs:
  fsnotify: optimize the case of no marks of any type
  fsnotify: count all objects with attached connectors
  fsnotify: count s_fsnotify_inode_refs for attached connectors
  fsnotify: replace igrab() with ihold() on attach connector
  fanotify: add pidfd support to the fanotify API
  fanotify: introduce a generic info record copying helper
  fanotify: minor cosmetic adjustments to fid labels
  kernel/pid.c: implement additional checks upon pidfd_create() parameters
  kernel/pid.c: remove static qualifier from pidfd_create()

As per the long-suffering comment.

Reported-by: Minh Yuan <yuanmingbuaa@gmail.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Jiri Slaby <jirislaby@kernel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

The Linux console's VT102 implementation already consumes OSC
("Operating System Command") sequences, probably because that's how
palette changes are transmitted.

In addition to OSC, there are three other major clases of ANSI control
strings: APC ("Application Program Command"), PM ("Privacy Message"),
and DCS ("Device Control String").  They are handled similarly to OSC in
terms of termination.

Source: vt100.net

Add three new enumerated states, one for each of these types.  All three
are handled the same way right now--they simply consume input until
terminated.  I hope to expand upon this firmament in the future.  Add
new predicate ansi_control_string(), returning true for any of these
states.  Replace explicit checks against ESosc with calls to this
function.  Transition to these states appropriately from the escape
initiation (ESesc) state.

This was motivated by the following Notcurses bugs:

 https://github.com/dankamongmen/notcurses/issues/2050
 https://github.com/dankamongmen/notcurses/issues/1828
 https://github.com/dankamongmen/notcurses/issues/2069

where standard VT sequences are not consumed by the Linux console.  It's
not necessary that the Linux console *support* these sequences, but it
ought *consume* these well-specified classes of sequences.

Tested by sending a variety of escape sequences to the console, and
verifying that they still worked, or were now properly consumed.
Verified that the escapes were properly terminated at a generic level.
Verified that the Notcurses tools continued to show expected output on
the Linux console, except now without escape bleedthrough.

Link: https://lore.kernel.org/lkml/YSydL0q8iaUfkphg@schwarzgerat.orthanc/


Signed-off-by: nick black <dankamongmen@gmail.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Jiri Slaby <jirislaby@kernel.org>
Cc: Tetsuo Handa <penguin-kernel@i-love.sakura.ne.jp>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

This just does the "if the architecture does efficient unaligned
handling, start the memcmp using 'unsigned long' accesses", since
Nikolay Borisov found a load that cares.

This is basically the minimal patch, and limited to architectures that
are known to not have slow unaligned handling.  We've had the stupid
byte-at-a-time version forever, and nobody has ever even noticed before,
so let's keep the fix minimal.

A potential further improvement would be to align one of the sources in
order to at least minimize unaligned cases, but the only real case of
bigger memcmp() users seems to be the FIDEDUPERANGE ioctl().  As David
Sterba says, the dedupe ioctl is typically called on ranges spanning
many pages so the common case will all be page-aligned anyway.

All the relevant architectures select HAVE_EFFICIENT_UNALIGNED_ACCESS,
so I'm not going to worry about the combination of a very rare use-case
and a rare architecture until somebody actually hits it.  Particularly
since Nikolay also tested the more complex patch with extra alignment
handling code, and it only added overhead.

Link: https://lore.kernel.org/lkml/20210721135926.602840-1-nborisov@suse.com/


Reported-by: Nikolay Borisov <nborisov@suse.com>
Cc: David Sterba <dsterba@suse.cz>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Pull clk fix from Stephen Boyd:
 "One hotfix for a NULL pointer deref in the Renesas usb clk driver"

* tag 'clk-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/clk/linux:
  clk: renesas: rcar-usb2-clock-sel: Fix kernel NULL pointer dereference

Pull scheduler fixes from Borislav Petkov:

 - Have get_push_task() check whether current has migration disabled and
   thus avoid useless invocations of the migration thread

 - Rework initialization flow so that all rq->core's are initialized,
   even of CPUs which have not been onlined yet, so that iterating over
   them all works as expected

* tag 'sched_urgent_for_v5.14' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  sched: Fix get_push_task() vs migrate_disable()
  sched: Fix Core-wide rq->lock for uninitialized CPUs

Pull irq fix from Borislav Petkov:

 - Have msix_mask_all() check a global control which says whether MSI-X
   masking should be done and thus make it usable on Xen-PV too

* tag 'irq_urgent_for_v5.14' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  PCI/MSI: Skip masking MSI-X on Xen PV

Pull perf fixes from Borislav Petkov:

 - Prevent the amd/power module from being removed while in use

 - Mark AMD IBS as not supporting content exclusion

 - Add a workaround for AMD erratum #1197 where IBS registers might not
   be restored properly after exiting CC6 state

 - Fix a potential truncation of a 32-bit variable due to shifting

 - Read the correct bits describing the number of configurable address
   ranges on Intel PT

* tag 'perf_urgent_for_v5.14' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  perf/x86/amd/power: Assign pmu.module
  perf/x86/amd/ibs: Extend PERF_PMU_CAP_NO_EXCLUDE to IBS Op
  perf/x86/amd/ibs: Work around erratum #1197
  perf/x86/intel/uncore: Fix integer overflow on 23 bit left shift of a u32
  perf/x86/intel/pt: Fix mask of num_address_ranges

Pull x86 fixes from Borislav Petkov:

 - Fix build error on RHEL where -Werror=maybe-uninitialized is set.

 - Restore the firmware's IDT when calling EFI boot services and before
   ExitBootServices() has been called. This fixes a boot failure on what
   appears to be a tablet with 32-bit UEFI running a 64-bit kernel.

* tag 'x86_urgent_for_v5.14' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/resctrl: Fix a maybe-uninitialized build warning treated as error
  x86/efi: Restore Firmware IDT before calling ExitBootServices()

This reverts commit 83af58f8

.

It turns out that at least the assembly implementation for strncpy() was
buggy.  Revert the whole commit and return back to the default coding.

Signed-off-by: Helge Deller <deller@gmx.de>
Cc: <stable@vger.kernel.org> # v5.4+
Cc: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

The probe was manually passing NULL instead of dev to devm_clk_hw_register.
This caused a Unable to handle kernel NULL pointer dereference error.
Fix this by passing 'dev'.

Signed-off-by: Adam Ford <aford173@gmail.com>
Fixes: a20a40a8

 ("clk: renesas: rcar-usb2-clock-sel: Fix error handling in .probe()")
Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: Stephen Boyd <sboyd@kernel.org>

Pull SCSI fix from James Bottomley:
 "A single fix for a race introduced by a fix that went into 5.14-rc5"

* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
  scsi: core: Fix hang of freezing queue between blocking and running device

Pull USB fixes from Greg KH:
 "Here are a few tiny USB fixes for reported issues with some USB
  drivers.

  These fixes include:

   - gadget driver fixes for regressions

   - tcpm driver fix

   - dwc3 driver fixes

   - xhci renesas firmware loading fix, again.

   - usb serial option driver device id addition

   - usb serial ch341 revert for regression

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

* tag 'usb-5.14' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb:
  usb: gadget: u_audio: fix race condition on endpoint stop
  usb: gadget: f_uac2: fixup feedback endpoint stop
  usb: typec: tcpm: Raise vdm_sm_running flag only when VDM SM is running
  usb: renesas-xhci: Prefer firmware loading on unknown ROM state
  usb: dwc3: gadget: Stop EP0 transfers during pullup disable
  usb: dwc3: gadget: Fix dwc3_calc_trbs_left()
  Revert "USB: serial: ch341: fix character loss at high transfer rates"
  USB: serial: option: add new VID/PID to support Fibocom FG150

Pull powerpc fixes from Michael Ellerman:

 - Fix scv implicit soft-mask table for relocated (eg. kdump) kernels

 - Re-enable ARCH_ENABLE_SPLIT_PMD_PTLOCK, which was disabled due to a
   typo

Thanks to Lukas Bulwahn, Nicholas Piggin, and Daniel Axtens.

* tag 'powerpc-5.14-7' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
  powerpc/64s: Fix scv implicit soft-mask table for relocated kernels
  powerpc: Re-enable ARCH_ENABLE_SPLIT_PMD_PTLOCK

Pull block fixes from Jens Axboe:

 - Revert the mq-deadline priority handling, it's causing serious
   performance regressions. While experimental patches exists to fix
   this up, it's too late to do so now. Revert it and re-do it properly
   for 5.15 instead.

 - Fix a NULL vs IS_ERR() regression in this release (Dan)

 - Fix a mq-deadline accounting regression in this release (Bart)

 - Mark cryptoloop as deprecated. It's broken and dm-crypt fully
   supports it, and it's actively intefering with loop. Plan on removal
   for 5.16 (Christoph)

* tag 'block-5.14-2021-08-27' of git://git.kernel.dk/linux-block:
  cryptoloop: add a deprecation warning
  pd: fix a NULL vs IS_ERR() check
  Revert "block/mq-deadline: Prioritize high-priority requests"
  mq-deadline: Fix request accounting

Pull ARM SoC fixes from Arnd Bergmann:
 "Just two trivial fixes from the reset driver tree, nothing else came
  up since the last soc fixes"

* tag 'soc-fixes-5.14-4' of git://git.kernel.org/pub/scm/linux/kernel/git/soc/soc:
  reset: reset-zynqmp: Fixed the argument data type
  reset: RESET_MCHP_SPARX5 should depend on ARCH_SPARX5

Pull ACPI fix from Rafael Wysocki:
 "Fix a regression introduced during this cycle that has been partially
  addressed by an earlier commit (Andy Shevchenko)"

* tag 'acpi-5.14-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  media: ipu3-cio2: Drop reference on error path in cio2_bridge_connect_sensor()

Pull power management fixes from Rafael Wysocki:
 "These fix two issues introduced during this cycle, one of which is a
  regression and the other one affects new code.

  Specifics:

   - Prevent the operating performance points (OPP) code from crashing
     when some entries in the table of required OPPs are set to error
     pointer values (Marijn Suijten)

   - Prevent the generic power domains (genpd) framework from
     incorrectly overriding the performance state of a device set by its
     driver while it is runtime-suspended or when runtime PM of it is
     disabled (Dmitry Osipenko)"

* tag 'pm-5.14-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  PM: domains: Improve runtime PM performance state handling
  opp: core: Check for pending links before reading required_opp pointers

virtio_mem_set_fake_offline() might sleep now, and we call it under
rcu_read_lock(). To fix it, simply move the rcu_read_unlock() further
up, as we're done with the device.

Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Fixes: 6cc26d77

: "virtio-mem: use page_offline_(start|end) when setting PageOffline()
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Jason Wang <jasowang@redhat.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: virtualization@lists.linux-foundation.org
Signed-off-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

* pm-opp:
  opp: core: Check for pending links before reading required_opp pointers

Pull RISC-V fixes from Palmer Dabbelt:

 - device tree updates for the Microsemi Polarfire development kit that
   fix some mismatches between the u-boot and Linux enternet entries

 - ensure that the F register state is correctly reflected in core dumps

* tag 'riscv-for-linus-5.14-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux:
  riscv: dts: microchip: Add ethernet0 to the aliases node
  riscv: dts: microchip: Use 'local-mac-address' for emac1
  riscv: Ensure the value of FP registers in the core dump file is up to date

Pull MMC host fix from Ulf Hansson:

 - sdhci-iproc: Fix clock error for ACPI rpi's

* tag 'mmc-v5.14-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/ulfh/mmc:
  Revert "mmc: sdhci-iproc: Set SDHCI_QUIRK_CAP_CLOCK_BASE_BROKEN on BCM2711"

Support for cryptoloop has been officially marked broken and deprecated
in favor of dm-crypt (which supports the same broken algorithms if
needed) in Linux 2.6.4 (released in March 2004), and support for it has
been entirely removed from losetup in util-linux 2.23 (released in April
2013).  Add a warning and a deprecation schedule.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20210827163250.255325-1-hch@lst.de


Signed-off-by: Jens Axboe <axboe@kernel.dk>

Pull ARM fix from Russell King:
 "Resolve a Keystone 2 kernel mapping regression"

* tag 'for-linus' of git://git.armlinux.org.uk/~rmk/linux-arm:
  ARM: 9104/2: Fix Keystone 2 kernel mapping regression

This reverts commit 419dd626

.

It turned out that the change from the reverted commit breaks the ACPI
based rpi's because it causes the 100Mhz max clock to be overridden to the
return from sdhci_iproc_get_max_clock(), which is 0 because there isn't a
OF/DT based clock device.

Reported-by: Jeremy Linton <jeremy.linton@arm.com>
Fixes: 419dd626

 ("mmc: sdhci-iproc: Set SDHCI_QUIRK_CAP_CLOCK_BASE_BROKEN on BCM2711")
Acked-by: Stefan Wahren <stefan.wahren@i2se.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

If the endpoint completion callback is call right after the ep_enabled flag
is cleared and before usb_ep_dequeue() is call, we could do a double free
on the request and the associated buffer.

Fix this by clearing ep_enabled after all the endpoint requests have been
dequeued.

Fixes: 7de8681b

 ("usb: gadget: u_audio: Free requests only after callback")
Cc: stable <stable@vger.kernel.org>
Reported-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Signed-off-by: Jerome Brunet <jbrunet@baylibre.com>
Link: https://lore.kernel.org/r/20210827092927.366482-1-jbrunet@baylibre.com


Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

When the uac2 function is stopped, there seems to be an issue reported on
some platforms (Intel Merrifield at least)

BUG: kernel NULL pointer dereference, address: 0000000000000008
...
RIP: 0010:dwc3_gadget_del_and_unmap_request+0x19/0xe0
...
Call Trace:
 dwc3_remove_requests.constprop.0+0x12f/0x170
 __dwc3_gadget_ep_disable+0x7a/0x160
 dwc3_gadget_ep_disable+0x3d/0xd0
 usb_ep_disable+0x1c/0x70
 u_audio_stop_capture+0x79/0x120 [u_audio]
 afunc_set_alt+0x73/0x80 [usb_f_uac2]
 composite_setup+0x224/0x1b90 [libcomposite]

The issue happens only when the gadget is using the sync type "async", not
"adaptive". This indicates that problem is coming from the feedback
endpoint, which is only used with async synchronization mode.

The problem is that request is freed regardless of usb_ep_dequeue(), which
ends up badly if the request is not actually dequeued yet.

Update the feedback endpoint free function to release the endpoint the same
way it is done for the data endpoint, which takes care of the problem.

Fixes: 24f779da

 ("usb: gadget: f_uac2/u_audio: add feedback endpoint support")
Reported-by: Ferry Toth <ftoth@exalondelft.nl>
Tested-by: Ferry Toth <ftoth@exalondelft.nl>
Acked-by: Felipe Balbi <balbi@kernel.org>
Signed-off-by: Jerome Brunet <jbrunet@baylibre.com>
Link: https://lore.kernel.org/r/20210827075853.266912-1-jbrunet@baylibre.com


Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

blk_mq_alloc_disk() returns error pointers, it doesn't return NULL
so correct the check.

Fixes: 262d431f

 ("pd: use blk_mq_alloc_disk and blk_cleanup_disk")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Link: https://lore.kernel.org/r/20210827100023.GB9449@kili


Signed-off-by: Jens Axboe <axboe@kernel.dk>

Pull drm fixes from Dave Airlie:
 "Last set of fixes for 5.14, nothing major a couple of i915, couple of
  imx and a few amdgpu. All pretty small.

  i915:
   - Fix syncmap memory leak
   - Drop redundant display port debug print

  amdgpu:
   - Fix for pinning display buffers multiple times
   - Fix delayed work handling for GFXOFF
   - Fix build when CONFIG_SUSPEND is not set

  imx:
   - fix planar offset calculations
   - fix accidental partial revert"

* tag 'drm-fixes-2021-08-27' of git://anongit.freedesktop.org/drm/drm:
  drm/i915/dp: Drop redundant debug print
  drm/i915: Fix syncmap memory leak
  drm/amdgpu: Fix build with missing pm_suspend_target_state module export
  drm/amdgpu: Cancel delayed work when GFXOFF is disabled
  drm/amdgpu: use the preferred pin domain after the check
  drm/imx: ipuv3-plane: fix accidental partial revert of 8 pixel alignment fix
  gpu: ipu-v3: Fix i.MX IPU-v3 offset calculations for (semi)planar U/V formats

drm/imx: imx-drm alignment and plane offset fixes

Fix an accidental partial revert of commit 94dfec48

 ("drm/imx: Add 8
pixel alignment fix") and plane offset calculations for capture of
non-aligned resolutions.

Signed-off-by: Dave Airlie <airlied@redhat.com>

From: Philipp Zabel <p.zabel@pengutronix.de>
Link: https://patchwork.freedesktop.org/patch/msgid/85a41af99beb2c9e7d6020435a135bf9f205a5ff.camel@pengutronix.de

Merge tag 'amd-drm-fixes-5.14-2021-08-25' of https://gitlab.freedesktop.org/agd5f/linux

 into drm-fixes

amd-drm-fixes-5.14-2021-08-25:

amdgpu:
- Fix for pinning display buffers multiple times
- Fix delayed work handling for GFXOFF
- Fix build when CONFIG_SUSPEND is not set

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Alex Deucher <alexander.deucher@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210826032658.4068-1-alexander.deucher@amd.com

Merge tag 'drm-intel-fixes-2021-08-26' of git://anongit.freedesktop.org/drm/drm-intel

 into drm-fixes

- Fix syncmap memory leak
- Drop redundant display port debug print

Signed-off-by: Dave Airlie <airlied@redhat.com>

From: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/YSfSeHbyS5wBZtNJ@intel.com

When running as Xen PV guest, masking MSI-X is a responsibility of the
hypervisor. The guest has no write access to the relevant BAR at all - when
it tries to, it results in a crash like this:

    BUG: unable to handle page fault for address: ffffc9004069100c
    #PF: supervisor write access in kernel mode
    #PF: error_code(0x0003) - permissions violation
    RIP: e030:__pci_enable_msix_range.part.0+0x26b/0x5f0
     e1000e_set_interrupt_capability+0xbf/0xd0 [e1000e]
     e1000_probe+0x41f/0xdb0 [e1000e]
     local_pci_probe+0x42/0x80
    (...)

The recently introduced function msix_mask_all() does not check the global
variable pci_msi_ignore_mask which is set by XEN PV to bypass the masking
of MSI[-X] interrupts.

Add the check to make this function XEN PV compatible.

Fixes: 7d5ec3d3

 ("PCI/MSI: Mask all unused MSI-X entries")
Signed-off-by: Marek Marczykowski-Górecki <marmarek@invisiblethingslab.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20210826170342.135172-1-marmarek@invisiblethingslab.com

Pull nfsd fix from Bruce Fields:
 "This is a one-liner fix for a serious bug that can cause the server to
  become unresponsive to a client, so I think it's worth the last-minute
  inclusion for 5.14"

* tag 'nfsd-5.14-1' of git://linux-nfs.org/~bfields/linux:
  SUNRPC: Fix XPT_BUSY flag leakage in svc_handle_xprt()...

Pull networking fixes from Jakub Kicinski:
 "Networking fixes, including fixes from can and bpf.

  Closing three hw-dependent regressions. Any fixes of note are in the
  'old code' category. Nothing blocking release from our perspective.

  Current release - regressions:

   - stmmac: revert "stmmac: align RX buffers"

   - usb: asix: ax88772: move embedded PHY detection as early as
     possible

   - usb: asix: do not call phy_disconnect() for ax88178

   - Revert "net: really fix the build...", from Kalle to fix QCA6390

  Current release - new code bugs:

   - phy: mediatek: add the missing suspend/resume callbacks

  Previous releases - regressions:

   - qrtr: fix another OOB Read in qrtr_endpoint_post

   - stmmac: dwmac-rk: fix unbalanced pm_runtime_enable warnings

  Previous releases - always broken:

   - inet: use siphash in exception handling

   - ip_gre: add validation for csum_start

   - bpf: fix ringbuf helper function compatibility

   - rtnetlink: return correct error on changing device netns

   - e1000e: do not try to recover the NVM checksum on Tiger Lake"

* tag 'net-5.14-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (43 commits)
  Revert "net: really fix the build..."
  net: hns3: fix get wrong pfc_en when query PFC configuration
  net: hns3: fix GRO configuration error after reset
  net: hns3: change the method of getting cmd index in debugfs
  net: hns3: fix duplicate node in VLAN list
  net: hns3: fix speed unknown issue in bond 4
  net: hns3: add waiting time before cmdq memory is released
  net: hns3: clear hardware resource when loading driver
  net: fix NULL pointer reference in cipso_v4_doi_free
  rtnetlink: Return correct error on changing device netns
  net: dsa: hellcreek: Adjust schedule look ahead window
  net: dsa: hellcreek: Fix incorrect setting of GCL
  cxgb4: dont touch blocked freelist bitmap after free
  ipv4: use siphash instead of Jenkins in fnhe_hashfun()
  ipv6: use siphash in rt6_exception_hash()
  can: usb: esd_usb2: esd_usb2_rx_event(): fix the interchange of the CAN RX and TX error counters
  net: usb: asix: ax88772: fix boolconv.cocci warnings
  net/sched: ets: fix crash when flipping from 'strict' to 'quantum'
  qede: Fix memset corruption
  net: stmmac: fix kernel panic due to NULL pointer dereference of buf->xdp
  ...

This reverts commit fb926032.

Zhen reports that this commit slows down mq-deadline on a 128 thread
box, going from 258K IOPS to 170-180K. My testing shows that Optane
gen2 IOPS goes from 2.3M IOPS to 1.2M IOPS on a 64 thread box.

Looking in detail at the code, the main culprit here is needing to sum
percpu counters in the dispatch hot path, leading to very high CPU
utilization there. To make matters worse, the code currently needs to
sum 2 percpu counters, and it does so in the most naive way of iterating
possible CPUs _twice_.

Since we're close to release, revert this commit and we can re-do it
with regular per-priority counters instead for the 5.15 kernel.

Link: https://lore.kernel.org/linux-block/20210826144039.2143-1-thunder.leizhen@huawei.com/


Reported-by: Zhen Lei <thunder.leizhen@huawei.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>