For existing adrenos, there is one or more ringbuffer, depending on
whether preemption is supported.  When preemption is supported, each
ringbuffer has it's own priority.  A submitqueue (which maps to a
gl context or vk queue in userspace) is mapped to a specific ring-
buffer at creation time, based on the submitqueue's priority.

Each ringbuffer has it's own drm_gpu_scheduler.  Each submitqueue
maps to a drm_sched_entity.  And each submit maps to a drm_sched_job.

Closes: https://gitlab.freedesktop.org/drm/msm/-/issues/4Signed-off-by: Rob Clark <robdclark@chromium.org>
Acked-by: Christian König <christian.koenig@amd.com>
Link: https://lore.kernel.org/r/20210728010632.2633470-10-robdclark@gmail.comSigned-off-by: Rob Clark <robdclark@chromium.org>

In the next patch, we start having more than a single potential failure
reason.
Signed-off-by: Rob Clark <robdclark@chromium.org>
Acked-by: Christian König <christian.koenig@amd.com>
Link: https://lore.kernel.org/r/20210728010632.2633470-9-robdclark@gmail.comSigned-off-by: Rob Clark <robdclark@chromium.org>

Previously the (non-fd) fence returned from submit ioctl was a raw
seqno, which is scoped to the ring.  But from UABI standpoint, the
ioctls related to seqno fences all specify a submitqueue.  We can
take advantage of that to replace the seqno fences with a cyclic idr
handle.

This is in preperation for moving to drm scheduler, at which point
the submit ioctl will return after queuing the submit job to the
scheduler, but before the submit is written into the ring (and
therefore before a ring seqno has been assigned).  Which means we
need to replace the dma_fence that userspace may need to wait on
with a scheduler fence.
Signed-off-by: Rob Clark <robdclark@chromium.org>
Acked-by: Christian König <christian.koenig@amd.com>
Link: https://lore.kernel.org/r/20210728010632.2633470-8-robdclark@gmail.comSigned-off-by: Rob Clark <robdclark@chromium.org>

Move all the locked/active/pinned state handling to msm_gem_submit.c.
In particular, for drm/scheduler, we'll need to do all this before
pushing the submit job to the scheduler.  But while we're at it we can
get rid of the dupicate pin and refcnt.
Signed-off-by: Rob Clark <robdclark@chromium.org>
Acked-by: Christian König <christian.koenig@amd.com>
Link: https://lore.kernel.org/r/20210728010632.2633470-7-robdclark@gmail.comSigned-off-by: Rob Clark <robdclark@chromium.org>

No need for this to be split in two parts.
Signed-off-by: Rob Clark <robdclark@chromium.org>
Acked-by: Christian König <christian.koenig@amd.com>
Link: https://lore.kernel.org/r/20210728010632.2633470-6-robdclark@gmail.comSigned-off-by: Rob Clark <robdclark@chromium.org>

Now that no one is using it, remove it.
Signed-off-by: Rob Clark <robdclark@chromium.org>
Acked-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://lore.kernel.org/r/20210728010632.2633470-5-robdclark@gmail.comSigned-off-by: Rob Clark <robdclark@chromium.org>

No idea why we were still using this.  It certainly hasn't been needed
for some time.  So drop the pointless twin codepaths.
Signed-off-by: Rob Clark <robdclark@chromium.org>
Acked-by: Christian König <christian.koenig@amd.com>
Link: https://lore.kernel.org/r/20210728010632.2633470-4-robdclark@gmail.comSigned-off-by: Rob Clark <robdclark@chromium.org>

If we don't have a gpu, there is no need to create a submitqueue, which
lets us simplify the error handling and submitqueue creation.
Signed-off-by: Rob Clark <robdclark@chromium.org>
Acked-by: Christian König <christian.koenig@amd.com>
Link: https://lore.kernel.org/r/20210728010632.2633470-3-robdclark@gmail.comSigned-off-by: Rob Clark <robdclark@chromium.org>

Fix a couple incorrect or misspelt comments, and add submitqueue doc
comment.
Signed-off-by: Rob Clark <robdclark@chromium.org>
Acked-by: Christian König <christian.koenig@amd.com>
Link: https://lore.kernel.org/r/20210728010632.2633470-2-robdclark@gmail.comSigned-off-by: Rob Clark <robdclark@chromium.org>

This adds a few things to try and make frequency scaling better match
the workload:

1) Longer polling interval to avoid whip-lashing between too-high and
   too-low frequencies in certain workloads, like mobile games which
   throttle themselves to 30fps.

   Previously our polling interval was short enough to let things
   ramp down to minimum freq in the "off" frame, but long enough to
   not react quickly enough when rendering started on the next frame,
   leading to uneven frame times.  (Ie. rather than a consistent 33ms
   it would alternate between 16/33/48ms.)

2) Awareness of when the GPU is active vs idle.  Since we know when
   the GPU is active vs idle, we can clamp the frequency down to the
   minimum while it is idle.  (If it is idle for long enough, then
   the autosuspend delay will eventually kick in and power down the
   GPU.)

   Since devfreq has no knowledge of powered-but-idle, this takes a
   small bit of trickery to maintain a "fake" frequency while idle.
   This, combined with the longer polling period allows devfreq to
   arrive at a reasonable "active" frequency, while still clamping
   to minimum freq when idle to reduce power draw.

3) Boost.  Because simple_ondemand needs to see a certain threshold
   of busyness to ramp up, we could end up needing multiple polling
   cycles before it reacts appropriately on interactive workloads
   (ex. scrolling a web page after reading for some time), on top
   of the already lengthened polling interval, when we see a idle
   to active transition after a period of idle time we boost the
   frequency that we return to.
Signed-off-by: Rob Clark <robdclark@chromium.org>
Link: https://lore.kernel.org/r/20210726144653.2180096-4-robdclark@gmail.comSigned-off-by: Rob Clark <robdclark@chromium.org>

In the next patch, it grows a bit more, so lets not duplicate the logic
in multiple places.
Signed-off-by: Rob Clark <robdclark@chromium.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Link: https://lore.kernel.org/r/20210726144653.2180096-3-robdclark@gmail.comSigned-off-by: Rob Clark <robdclark@chromium.org>

Before we start adding more cleverness, split it into it's own file.
Signed-off-by: Rob Clark <robdclark@chromium.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Link: https://lore.kernel.org/r/20210726144653.2180096-2-robdclark@gmail.comSigned-off-by: Rob Clark <robdclark@chromium.org>

Nothing we do to in update_fences() can't be done in an atomic context,
so move this into the GPU's irq context to reduce latency (and call
dma_fence_signal() so we aren't relying on dma_fence_is_signaled() which
would defeat the purpose).
Signed-off-by: Rob Clark <robdclark@chromium.org>
Link: https://lore.kernel.org/r/20210726144359.2179302-3-robdclark@gmail.comSigned-off-by: Rob Clark <robdclark@chromium.org>

Let dma_fence::signaled, etc, read directly from the address that the hw
is writing with updated completed fence seqno, so we can potentially
notice that the fence is signaled sooner.

Plus add some docs.
Signed-off-by: Rob Clark <robdclark@chromium.org>
Link: https://lore.kernel.org/r/20210726144359.2179302-2-robdclark@gmail.comSigned-off-by: Rob Clark <robdclark@chromium.org>

gcc doesn't care, but clang quite reasonably pointed out that the recent
commit e9ba16e6 ("smpboot: Mark idle_init() as __always_inlined to
work around aggressive compiler un-inlining") did some really odd
things:

    kernel/smpboot.c:50:20: warning: duplicate 'inline' declaration specifier [-Wduplicate-decl-specifier]
    static inline void __always_inline idle_init(unsigned int cpu)
                       ^

which not only has that duplicate inlining specifier, but the new
__always_inline was put in the wrong place of the function definition.

We put the storage class specifiers (ie things like "static" and
"extern") first, and the type information after that.  And while the
compiler may not care, we put the inline specifier before the types.

So it should be just

    static __always_inline void idle_init(unsigned int cpu)

instead.

Cc: Ingo Molnar <mingo@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Pull powerpc fixes from Michael Ellerman:

 - Fix guest to host memory corruption in H_RTAS due to missing nargs
   check.

 - Fix guest triggerable host crashes due to bad handling of nested
   guest TM state.

 - Fix possible crashes due to incorrect reference counting in
   kvm_arch_vcpu_ioctl().

 - Two commits fixing some regressions in KVM transactional memory
   handling introduced by the recent rework of the KVM code.

Thanks to Nicholas Piggin, Alexey Kardashevskiy, and Michael Neuling.

* tag 'powerpc-5.14-3' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
  KVM: PPC: Book3S HV Nested: Sanitise H_ENTER_NESTED TM state
  KVM: PPC: Book3S: Fix H_RTAS rets buffer overflow
  KVM: PPC: Fix kvm_arch_vcpu_ioctl vcpu_load leak
  KVM: PPC: Book3S: Fix CONFIG_TRANSACTIONAL_MEM=n crash
  KVM: PPC: Book3S HV P9: Fix guest TM support

Pull timer fixes from Thomas Gleixner:
 "A small set of timer related fixes:

   - Plug a race between rearm and process tick in the posix CPU timers
     code

   - Make the optimization to avoid recalculation of the next timer
     interrupt work correctly when there are no timers pending"

* tag 'timers-urgent-2021-07-25' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  timers: Fix get_next_timer_interrupt() with no timers pending
  posix-cpu-timers: Fix rearm racing against process tick

Pull x86 jump label fix from Thomas Gleixner:
 "A single fix for jump labels to prevent the compiler from agressive
  un-inlining which results in a section mismatch"

* tag 'locking-urgent-2021-07-25' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  jump_labels: Mark __jump_label_transform() as __always_inlined to work around aggressive compiler un-inlining

Pull EFI fixes from Thomas Gleixner:
 "A set of EFI fixes:

   - Prevent memblock and I/O reserved resources to get out of sync when
     EFI memreserve is in use.

   - Don't claim a non-existing table is invalid

   - Don't warn when firmware memory is already reserved correctly"

* tag 'efi-urgent-2021-07-25' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  efi/mokvar: Reserve the table only if it is in boot services data
  efi/libstub: Fix the efi_load_initrd function description
  firmware/efi: Tell memblock about EFI iomem reservations
  efi/tpm: Differentiate missing and invalid final event log table.

Pull core fix from Thomas Gleixner:
 "A single update for the boot code to prevent aggressive un-inlining
  which causes a section mismatch"

* tag 'core-urgent-2021-07-25' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  smpboot: Mark idle_init() as __always_inlined to work around aggressive compiler un-inlining

Pull dma-mapping fix from Christoph Hellwig:

 - handle vmalloc addresses in dma_common_{mmap,get_sgtable} (Roman
   Skakun)

* tag 'dma-mapping-5.14-1' of git://git.infradead.org/users/hch/dma-mapping:
  dma-mapping: handle vmalloc addresses in dma_common_{mmap,get_sgtable}

Pull cifs fixes from Steve French:
 "Five cifs/smb3 fixes, including a DFS failover fix, two fallocate
  fixes, and two trivial coverity cleanups"

* tag '5.14-rc2-smb3-fixes' of git://git.samba.org/sfrench/cifs-2.6:
  cifs: fix fallocate when trying to allocate a hole.
  CIFS: Clarify SMB1 code for POSIX delete file
  CIFS: Clarify SMB1 code for POSIX Create
  cifs: support share failover when remounting
  cifs: only write 64kb at a time when fallocating a small region of a file

Pull RISC-V fixes from Palmer Dabbelt:

 - properly set the memory size, which fixes 32-bit systems

 - allow initrd to load anywhere in memory, rather that restricting it
   to the first 256MiB

 - fix the 'mem=' parameter on 64-bit systems to properly account for
   the maximum supported memory now that the kernel is outside the
   linear map

 - avoid installing mappings into the last 4KiB of memory, which
   conflicts with error values

 - avoid the stack from being freed while it is being walked

 - a handful of fixes to the new copy to/from user routines

* tag 'riscv-for-linus-5.14-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux:
  riscv: __asm_copy_to-from_user: Fix: Typos in comments
  riscv: __asm_copy_to-from_user: Remove unnecessary size check
  riscv: __asm_copy_to-from_user: Fix: fail on RV32
  riscv: __asm_copy_to-from_user: Fix: overrun copy
  riscv: stacktrace: pin the task's stack in get_wchan
  riscv: Make sure the kernel mapping does not overlap with IS_ERR_VALUE
  riscv: Make sure the linear mapping does not use the kernel mapping
  riscv: Fix memory_limit for 64-bit kernel
  RISC-V: load initrd wherever it fits into memory
  riscv: Fix 32-bit RISC-V boot failure

Commit 71f64283 ("ACPI: utils: Fix reference counting in
for_each_acpi_dev_match()") started doing "acpi_dev_put()" on a pointer
that was possibly NULL.  That fails miserably, because that helper
inline function is not set up to handle that case.

Just make acpi_dev_put() silently accept a NULL pointer, rather than
calling down to put_device() with an invalid offset off that NULL
pointer.

Link: https://lore.kernel.org/lkml/a607c149-6bf6-0fd0-0e31-100378504da2@kernel.dk/Reported-and-tested-by: Jens Axboe <axboe@kernel.dk>
Tested-by: Daniel Scally <djrscally@gmail.com>
Cc: Andy Shevchenko <andy.shevchenko@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Pull SCSI fixes from James Bottomley:
 "Four fixes, all in drivers, all of which can lead to user visible
  problems in certain situations"

* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
  scsi: target: Fix NULL dereference on XCOPY completion
  scsi: mpt3sas: Transition IOC to Ready state during shutdown
  scsi: target: Fix protect handling in WRITE SAME(32)
  scsi: iscsi: Fix iface sysfs attr detection

Pull io_uring fixes from Jens Axboe:

 - Fix a memory leak due to a race condition in io_init_wq_offload
   (Yang)

 - Poll error handling fixes (Pavel)

 - Fix early fdput() regression (me)

 - Don't reissue iopoll requests off release path (me)

 - Add a safety check for io-wq queue off wrong path (me)

* tag 'io_uring-5.14-2021-07-24' of git://git.kernel.dk/linux-block:
  io_uring: explicitly catch any illegal async queue attempt
  io_uring: never attempt iopoll reissue from release path
  io_uring: fix early fdput() of file
  io_uring: fix memleak in io_init_wq_offload()
  io_uring: remove double poll entry on arm failure
  io_uring: explicitly count entries for poll reqs

Pull block fixes from Jens Axboe:

 - NVMe pull request (Christoph):
    - tracing fix (Keith Busch)
    - fix multipath head refcounting (Hannes Reinecke)
    - Write Zeroes vs PI fix (me)
    - drop a bogus WARN_ON (Zhihao Cheng)

 - Increase max blk-cgroup policy size, now that mq-deadline
   uses it too (Oleksandr)

* tag 'block-5.14-2021-07-24' of git://git.kernel.dk/linux-block:
  nvme: set the PRACT bit when using Write Zeroes with T10 PI
  nvme: fix nvme_setup_command metadata trace event
  nvme: fix refcounting imbalance when all paths are down
  nvme-pci: don't WARN_ON in nvme_reset_work if ctrl.state is not RESETTING
  block: increase BLKCG_MAX_POLS

Pull i2c fixes from Wolfram Sang:
 "Two bugfixes for the I2C subsystem"

* 'i2c/for-current' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux:
  i2c: mpc: Poll for MCF
  misc: eeprom: at24: Always append device id even if label property is set.

Merge misc mm fixes from Andrew Morton:
 "15 patches.

  VM subsystems affected by this patch series: userfaultfd, kfence,
  highmem, pagealloc, memblock, pagecache, secretmem, pagemap, and
  hugetlbfs"

* akpm:
  hugetlbfs: fix mount mode command line processing
  mm: fix the deadlock in finish_fault()
  mm: mmap_lock: fix disabling preemption directly
  mm/secretmem: wire up ->set_page_dirty
  writeback, cgroup: do not reparent dax inodes
  writeback, cgroup: remove wb from offline list before releasing refcnt
  memblock: make for_each_mem_range() traverse MEMBLOCK_HOTPLUG regions
  mm: page_alloc: fix page_poison=1 / INIT_ON_ALLOC_DEFAULT_ON interaction
  mm: use kmap_local_page in memzero_page
  mm: call flush_dcache_page() in memcpy_to_page() and memzero_page()
  kfence: skip all GFP_ZONEMASK allocations
  kfence: move the size check to the beginning of __kfence_alloc()
  kfence: defer kfence_test_init to ensure that kunit debugfs is created
  selftest: use mmap instead of posix_memalign to allocate memory
  userfaultfd: do not untag user pointers

Fixing typos and grammar mistakes and using more intuitive label
name.
Signed-off-by: Akira Tsukamoto <akira.tsukamoto@gmail.com>
Fixes: ca6eaaa2 ("riscv: __asm_copy_to-from_user: Optimize unaligned memory access and pipeline stall")
Signed-off-by: Palmer Dabbelt <palmerdabbelt@google.com>

Clean up:

The size of 0 will be evaluated in the next step. Not
required here.
Signed-off-by: Akira Tsukamoto <akira.tsukamoto@gmail.com>
Fixes: ca6eaaa2 ("riscv: __asm_copy_to-from_user: Optimize unaligned memory access and pipeline stall")
Signed-off-by: Palmer Dabbelt <palmerdabbelt@google.com>

Had a bug when converting bytes to bits when the cpu was rv32.

The a3 contains the number of bytes and multiple of 8
would be the bits. The LGREG is holding 2 for RV32 and 3 for
RV32, so to achieve multiple of 8 it must always be constant 3.
The 2 was mistakenly used for rv32.
Signed-off-by: Akira Tsukamoto <akira.tsukamoto@gmail.com>
Fixes: ca6eaaa2 ("riscv: __asm_copy_to-from_user: Optimize unaligned memory access and pipeline stall")
Signed-off-by: Palmer Dabbelt <palmerdabbelt@google.com>

There were two causes for the overrun memory access.

The threshold size was too small.
The aligning dst require one SZREG and unrolling word copy requires
8*SZREG, total have to be at least 9*SZREG.

Inside the unrolling copy, the subtracting -(8*SZREG-1) would make
iteration happening one extra loop. Proper value is -(8*SZREG).
Signed-off-by: Akira Tsukamoto <akira.tsukamoto@gmail.com>
Fixes: ca6eaaa2 ("riscv: __asm_copy_to-from_user: Optimize unaligned memory access and pipeline stall")
Signed-off-by: Palmer Dabbelt <palmerdabbelt@google.com>

In commit 32021982 ("hugetlbfs: Convert to fs_context") processing
of the mount mode string was changed from match_octal() to fsparam_u32.

This changed existing behavior as match_octal does not require octal
values to have a '0' prefix, but fsparam_u32 does.

Use fsparam_u32oct which provides the same behavior as match_octal.

Link: https://lkml.kernel.org/r/20210721183326.102716-1-mike.kravetz@oracle.com
Fixes: 32021982 ("hugetlbfs: Convert to fs_context")
Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>
Reported-by: Dennis Camera <bugs+kernel.org@dtnr.ch>
Reviewed-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: David Howells <dhowells@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Commit 63f3655f ("mm, memcg: fix reclaim deadlock with writeback")
fix the following ABBA deadlock by pre-allocating the pte page table
without holding the page lock.

	                                lock_page(A)
                                        SetPageWriteback(A)
                                        unlock_page(A)
  lock_page(B)
                                        lock_page(B)
  pte_alloc_one
    shrink_page_list
      wait_on_page_writeback(A)
                                        SetPageWriteback(B)
                                        unlock_page(B)

                                        # flush A, B to clear the writeback

Commit f9ce0be7 ("mm: Cleanup faultaround and finish_fault()
codepaths") reworked the relevant code but ignored this race.  This will
cause the deadlock above to appear again, so fix it.

Link: https://lkml.kernel.org/r/20210721074849.57004-1-zhengqi.arch@bytedance.com
Fixes: f9ce0be7 ("mm: Cleanup faultaround and finish_fault() codepaths")
Signed-off-by: Qi Zheng <zhengqi.arch@bytedance.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Vladimir Davydov <vdavydov.dev@gmail.com>
Cc: Muchun Song <songmuchun@bytedance.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Commit 832b5072 ("mm: mmap_lock: use local locks instead of
disabling preemption") fixed a bug by using local locks.

But commit d01079f3 ("mm/mmap_lock: remove dead code for
!CONFIG_TRACING configurations") changed those lines back to the
original version.

I guess it was introduced by fixing conflicts.

Link: https://lkml.kernel.org/r/20210720074228.76342-1-songmuchun@bytedance.com
Fixes: d01079f3 ("mm/mmap_lock: remove dead code for !CONFIG_TRACING configurations")
Signed-off-by: Muchun Song <songmuchun@bytedance.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Reviewed-by: Yang Shi <shy828301@gmail.com>
Reviewed-by: Pankaj Gupta <pankaj.gupta@ionos.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Make secretmem up to date with the changes done in commit 0af57378
("mm: require ->set_page_dirty to be explicitly wired up") so that
unconditional call to this method won't cause crashes.

Link: https://lkml.kernel.org/r/20210716063933.31633-1-rppt@kernel.org
Fixes: 0af57378 ("mm: require ->set_page_dirty to be explicitly wired up")
Signed-off-by: Mike Rapoport <rppt@linux.ibm.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

The inode switching code is not suited for dax inodes.  An attempt to
switch a dax inode to a parent writeback structure (as a part of a
writeback cleanup procedure) results in a panic like this:

  run fstests generic/270 at 2021-07-15 05:54:02
  XFS (pmem0p2): EXPERIMENTAL big timestamp feature in use.  Use at your own risk!
  XFS (pmem0p2): DAX enabled. Warning: EXPERIMENTAL, use at your own risk
  XFS (pmem0p2): EXPERIMENTAL inode btree counters feature in use. Use at your own risk!
  XFS (pmem0p2): Mounting V5 Filesystem
  XFS (pmem0p2): Ending clean mount
  XFS (pmem0p2): Quotacheck needed: Please wait.
  XFS (pmem0p2): Quotacheck: Done.
  XFS (pmem0p2): xlog_verify_grant_tail: space > BBTOB(tail_blocks)
  XFS (pmem0p2): xlog_verify_grant_tail: space > BBTOB(tail_blocks)
  XFS (pmem0p2): xlog_verify_grant_tail: space > BBTOB(tail_blocks)
  BUG: unable to handle page fault for address: 0000000005b0f669
  #PF: supervisor read access in kernel mode
  #PF: error_code(0x0000) - not-present page
  PGD 0 P4D 0
  Oops: 0000 [#1] SMP PTI
  CPU: 13 PID: 10479 Comm: kworker/13:16 Not tainted 5.14.0-rc1-master-8096acd7+ #8
  Hardware name: HP ProLiant DL360 Gen9/ProLiant DL360 Gen9, BIOS P89 09/13/2016
  Workqueue: inode_switch_wbs inode_switch_wbs_work_fn
  RIP: 0010:inode_do_switch_wbs+0xaf/0x470
  Code: 00 30 0f 85 c1 03 00 00 0f 1f 44 00 00 31 d2 48 c7 c6 ff ff ff ff 48 8d 7c 24 08 e8 eb 49 1a 00 48 85 c0 74 4a bb ff ff ff ff <48> 8b 50 08 48 8d 4a ff 83 e2 01 48 0f 45 c1 48 8b 00 a8 08 0f 85
  RSP: 0018:ffff9c66691abdc8 EFLAGS: 00010002
  RAX: 0000000005b0f661 RBX: 00000000ffffffff RCX: ffff89e6a21382b0
  RDX: 0000000000000001 RSI: ffff89e350230248 RDI: ffffffffffffffff
  RBP: ffff89e681d19400 R08: 0000000000000000 R09: 0000000000000228
  R10: ffffffffffffffff R11: ffffffffffffffc0 R12: ffff89e6a2138130
  R13: ffff89e316af7400 R14: ffff89e316af6e78 R15: ffff89e6a21382b0
  FS:  0000000000000000(0000) GS:ffff89ee5fb40000(0000) knlGS:0000000000000000
  CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
  CR2: 0000000005b0f669 CR3: 0000000cb2410004 CR4: 00000000001706e0
  Call Trace:
   inode_switch_wbs_work_fn+0xb6/0x2a0
   process_one_work+0x1e6/0x380
   worker_thread+0x53/0x3d0
   kthread+0x10f/0x130
   ret_from_fork+0x22/0x30
  Modules linked in: xt_CHECKSUM xt_MASQUERADE xt_conntrack ipt_REJECT nf_reject_ipv4 nft_compat nft_chain_nat nf_nat nf_conntrack nf_defrag_ipv6 nf_defrag_ipv4 nft_counter nf_tables nfnetlink bridge stp llc rfkill sunrpc intel_rapl_msr intel_rapl_common sb_edac x86_pkg_temp_thermal intel_powerclamp coretemp kvm_intel ipmi_ssif kvm mgag200 i2c_algo_bit iTCO_wdt irqbypass drm_kms_helper iTCO_vendor_support acpi_ipmi rapl syscopyarea sysfillrect intel_cstate ipmi_si sysimgblt ioatdma dax_pmem_compat fb_sys_fops ipmi_devintf device_dax i2c_i801 pcspkr intel_uncore hpilo nd_pmem cec dax_pmem_core dca i2c_smbus acpi_tad lpc_ich ipmi_msghandler acpi_power_meter drm fuse xfs libcrc32c sd_mod t10_pi crct10dif_pclmul crc32_pclmul crc32c_intel tg3 ghash_clmulni_intel serio_raw hpsa hpwdt scsi_transport_sas wmi dm_mirror dm_region_hash dm_log dm_mod
  CR2: 0000000005b0f669
  ---[ end trace ed2105faff8384f3 ]---
  RIP: 0010:inode_do_switch_wbs+0xaf/0x470
  Code: 00 30 0f 85 c1 03 00 00 0f 1f 44 00 00 31 d2 48 c7 c6 ff ff ff ff 48 8d 7c 24 08 e8 eb 49 1a 00 48 85 c0 74 4a bb ff ff ff ff <48> 8b 50 08 48 8d 4a ff 83 e2 01 48 0f 45 c1 48 8b 00 a8 08 0f 85
  RSP: 0018:ffff9c66691abdc8 EFLAGS: 00010002
  RAX: 0000000005b0f661 RBX: 00000000ffffffff RCX: ffff89e6a21382b0
  RDX: 0000000000000001 RSI: ffff89e350230248 RDI: ffffffffffffffff
  RBP: ffff89e681d19400 R08: 0000000000000000 R09: 0000000000000228
  R10: ffffffffffffffff R11: ffffffffffffffc0 R12: ffff89e6a2138130
  R13: ffff89e316af7400 R14: ffff89e316af6e78 R15: ffff89e6a21382b0
  FS:  0000000000000000(0000) GS:ffff89ee5fb40000(0000) knlGS:0000000000000000
  CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
  CR2: 0000000005b0f669 CR3: 0000000cb2410004 CR4: 00000000001706e0
  Kernel panic - not syncing: Fatal exception
  Kernel Offset: 0x15200000 from 0xffffffff81000000 (relocation range: 0xffffffff80000000-0xffffffffbfffffff)
  ---[ end Kernel panic - not syncing: Fatal exception ]---

The crash happens on an attempt to iterate over attached pagecache pages
and check the dirty flag: a dax inode's xarray contains pfn's instead of
generic struct page pointers.

This happens for DAX and not for other kinds of non-page entries in the
inodes because it's a tagged iteration, and shadow/swap entries are
never tagged; only DAX entries get tagged.

Fix the problem by bailing out (with the false return value) of
inode_prepare_sbs_switch() if a dax inode is passed.

[willy@infradead.org: changelog addition]

Link: https://lkml.kernel.org/r/20210719171350.3876830-1-guro@fb.com
Fixes: c22d70a1 ("writeback, cgroup: release dying cgwbs by switching attached inodes")
Signed-off-by: Roman Gushchin <guro@fb.com>
Reported-by: Murphy Zhou <jencce.kernel@gmail.com>
Reported-by: Darrick J. Wong <djwong@kernel.org>
Tested-by: Darrick J. Wong <djwong@kernel.org>
Tested-by: Murphy Zhou <jencce.kernel@gmail.com>
Acked-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Jan Kara <jack@suse.cz>
Cc: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Boyang reported that the commit c22d70a1 ("writeback, cgroup:
release dying cgwbs by switching attached inodes") causes the kernel to
crash while running xfstests generic/256 on ext4 on aarch64 and ppc64le.

  run fstests generic/256 at 2021-07-12 05:41:40
  EXT4-fs (vda3): mounted filesystem with ordered data mode. Opts: . Quota mode: none.
  Unable to handle kernel NULL pointer dereference at virtual address 0000000000000000
  Mem abort info:
     ESR = 0x96000005
     EC = 0x25: DABT (current EL), IL = 32 bits
     SET = 0, FnV = 0
     EA = 0, S1PTW = 0
     FSC = 0x05: level 1 translation fault
  Data abort info:
     ISV = 0, ISS = 0x00000005
     CM = 0, WnR = 0
  user pgtable: 64k pages, 48-bit VAs, pgdp=00000000b0502000
  [0000000000000000] pgd=0000000000000000, p4d=0000000000000000, pud=0000000000000000
  Internal error: Oops: 96000005 [#1] SMP
  Modules linked in: dm_flakey dm_snapshot dm_bufio dm_zero dm_mod loop tls rpcsec_gss_krb5 auth_rpcgss nfsv4 dns_resolver nfs lockd grace fscache netfs rfkill sunrpc ext4 vfat fat mbcache jbd2 drm fuse xfs libcrc32c crct10dif_ce ghash_ce sha2_ce sha256_arm64 sha1_ce virtio_blk virtio_net net_failover virtio_console failover virtio_mmio aes_neon_bs [last unloaded: scsi_debug]
  CPU: 0 PID: 408468 Comm: kworker/u8:5 Tainted: G X --------- ---  5.14.0-0.rc1.15.bx.el9.aarch64 #1
  Hardware name: QEMU KVM Virtual Machine, BIOS 0.0.0 02/06/2015
  Workqueue: events_unbound cleanup_offline_cgwbs_workfn
  pstate: 004000c5 (nzcv daIF +PAN -UAO -TCO BTYPE=--)
  pc : cleanup_offline_cgwbs_workfn+0x320/0x394
  lr : cleanup_offline_cgwbs_workfn+0xe0/0x394
  sp : ffff80001554fd10
  x29: ffff80001554fd10 x28: 0000000000000000 x27: 0000000000000001
  x26: 0000000000000000 x25: 00000000000000e0 x24: ffffd2a2fbe671a8
  x23: ffff80001554fd88 x22: ffffd2a2fbe67198 x21: ffffd2a2fc25a730
  x20: ffff210412bc3000 x19: ffff210412bc3280 x18: 0000000000000000
  x17: 0000000000000000 x16: 0000000000000000 x15: 0000000000000000
  x14: 0000000000000000 x13: 0000000000000030 x12: 0000000000000040
  x11: ffff210481572238 x10: ffff21048157223a x9 : ffffd2a2fa276c60
  x8 : ffff210484106b60 x7 : 0000000000000000 x6 : 000000000007d18a
  x5 : ffff210416a86400 x4 : ffff210412bc0280 x3 : 0000000000000000
  x2 : ffff80001554fd88 x1 : ffff210412bc0280 x0 : 0000000000000003
  Call trace:
     cleanup_offline_cgwbs_workfn+0x320/0x394
     process_one_work+0x1f4/0x4b0
     worker_thread+0x184/0x540
     kthread+0x114/0x120
     ret_from_fork+0x10/0x18
  Code: d63f0020 97f99963 17ffffa6 f8588263 (f9400061)
  ---[ end trace e250fe289272792a ]---
  Kernel panic - not syncing: Oops: Fatal exception
  SMP: stopping secondary CPUs
  SMP: failed to stop secondary CPUs 0-2
  Kernel Offset: 0x52a2e9fa0000 from 0xffff800010000000
  PHYS_OFFSET: 0xfff0defca0000000
  CPU features: 0x00200251,23200840
  Memory Limit: none
  ---[ end Kernel panic - not syncing: Oops: Fatal exception ]---

The problem happens when cgwb_release_workfn() races with
cleanup_offline_cgwbs_workfn(): wb_tryget() in
cleanup_offline_cgwbs_workfn() can be called after percpu_ref_exit() is
cgwb_release_workfn(), which is basically a use-after-free error.

Fix the problem by making removing the writeback structure from the
offline list before releasing the percpu reference counter.  It will
guarantee that cleanup_offline_cgwbs_workfn() will not see and not
access writeback structures which are about to be released.

Link: https://lkml.kernel.org/r/20210716201039.3762203-1-guro@fb.com
Fixes: c22d70a1 ("writeback, cgroup: release dying cgwbs by switching attached inodes")
Signed-off-by: Roman Gushchin <guro@fb.com>
Reported-by: Boyang Xue <bxue@redhat.com>
Suggested-by: Jan Kara <jack@suse.cz>
Tested-by: Darrick J. Wong <djwong@kernel.org>
Cc: Will Deacon <will@kernel.org>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Murphy Zhou <jencce.kernel@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>