Make IOMMU_DEFAULT_LAZY default for when AMD_IOMMU config is set, which
matches current behaviour.

For "fullflush" param, just call iommu_set_dma_strict(true) directly.

Since we get a strict vs lazy mode print already in iommu_subsys_init(),
and maintain a deprecation print when "fullflush" param is passed, drop the
prints in amd_iommu_init_dma_ops().

Finally drop global flag amd_iommu_unmap_flush, as it has no longer has any
purpose.

[jpg: Rebase for relocated file and drop amd_iommu_unmap_flush]
Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com>
Signed-off-by: John Garry <john.garry@huawei.com>
Link: https://lore.kernel.org/r/1626088340-5838-6-git-send-email-john.garry@huawei.comSigned-off-by: Joerg Roedel <jroedel@suse.de>

Make IOMMU_DEFAULT_LAZY default for when INTEL_IOMMU config is set,
as is current behaviour.

Also delete global flag intel_iommu_strict:
- In intel_iommu_setup(), call iommu_set_dma_strict(true) directly. Also
  remove the print, as iommu_subsys_init() prints the mode and we have
  already marked this param as deprecated.

- For cap_caching_mode() check in intel_iommu_setup(), call
  iommu_set_dma_strict(true) directly; also reword the accompanying print
  with a level downgrade and also add the missing '\n'.

- For Ironlake GPU, again call iommu_set_dma_strict(true) directly and
  keep the accompanying print.

[jpg: Remove intel_iommu_strict]
Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com>
Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: https://lore.kernel.org/r/1626088340-5838-5-git-send-email-john.garry@huawei.comSigned-off-by: Joerg Roedel <jroedel@suse.de>

First, add build options IOMMU_DEFAULT_{LAZY|STRICT}, so that we have the
opportunity to set {lazy|strict} mode as default at build time. Then put
the two config options in an choice, as they are mutually exclusive.

[jpg: Make choice between strict and lazy only (and not passthrough)]
Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com>
Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Robin Murphy <robin.murphy@arm.com>
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: https://lore.kernel.org/r/1626088340-5838-4-git-send-email-john.garry@huawei.comSigned-off-by: Joerg Roedel <jroedel@suse.de>

As well as the default domain type, it's useful to know whether strict
or lazy for DMA domains, so add this info in a separate print.

The (stict/lazy) mode may be also set via iommu.strict earlyparm, but
this will be processed prior to iommu_subsys_init(), so that print will be
accurate for drivers which don't set the mode via custom means.

For the drivers which set the mode via custom means - AMD and Intel drivers
- they maintain prints to inform a change in policy or that custom cmdline
methods to change policy are deprecated.
Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Robin Murphy <robin.murphy@arm.com>
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: https://lore.kernel.org/r/1626088340-5838-3-git-send-email-john.garry@huawei.comSigned-off-by: Joerg Roedel <jroedel@suse.de>

Now that the x86 drivers support iommu.strict, deprecate the custom
methods.
Signed-off-by: John Garry <john.garry@huawei.com>
Acked-by: Robin Murphy <robin.murphy@arm.com>
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: https://lore.kernel.org/r/1626088340-5838-2-git-send-email-john.garry@huawei.comSigned-off-by: Joerg Roedel <jroedel@suse.de>

Implement the map_pages() callback for the ARM SMMU driver
to allow calls from iommu_map to map multiple pages of
the same size in one call. Also, remove the map() callback
for the ARM SMMU driver, as it will no longer be used.
Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
Suggested-by: Will Deacon <will@kernel.org>
Signed-off-by: Georgi Djakov <quic_c_gdjako@quicinc.com>
Link: https://lore.kernel.org/r/1623850736-389584-16-git-send-email-quic_c_gdjako@quicinc.comSigned-off-by: Joerg Roedel <jroedel@suse.de>

Implement the unmap_pages() callback for the ARM SMMU driver
to allow calls from iommu_unmap to unmap multiple pages of
the same size in one call. Also, remove the unmap() callback
for the SMMU driver, as it will no longer be used.
Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
Suggested-by: Will Deacon <will@kernel.org>
Signed-off-by: Georgi Djakov <quic_c_gdjako@quicinc.com>
Link: https://lore.kernel.org/r/1623850736-389584-15-git-send-email-quic_c_gdjako@quicinc.comSigned-off-by: Joerg Roedel <jroedel@suse.de>

Implement the map_pages() callback for the ARM v7s io-pgtable
format.
Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
Signed-off-by: Georgi Djakov <quic_c_gdjako@quicinc.com>
Link: https://lore.kernel.org/r/1623850736-389584-14-git-send-email-quic_c_gdjako@quicinc.comSigned-off-by: Joerg Roedel <jroedel@suse.de>

Implement the unmap_pages() callback for the ARM v7s io-pgtable
format.
Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
Signed-off-by: Georgi Djakov <quic_c_gdjako@quicinc.com>
Link: https://lore.kernel.org/r/1623850736-389584-13-git-send-email-quic_c_gdjako@quicinc.comSigned-off-by: Joerg Roedel <jroedel@suse.de>

Implement the map_pages() callback for the ARM LPAE io-pgtable
format.
Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
Signed-off-by: Georgi Djakov <quic_c_gdjako@quicinc.com>
Link: https://lore.kernel.org/r/1623850736-389584-12-git-send-email-quic_c_gdjako@quicinc.comSigned-off-by: Joerg Roedel <jroedel@suse.de>

Implement the unmap_pages() callback for the ARM LPAE io-pgtable
format.
Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
Suggested-by: Will Deacon <will@kernel.org>
Signed-off-by: Georgi Djakov <quic_c_gdjako@quicinc.com>
Link: https://lore.kernel.org/r/1623850736-389584-11-git-send-email-quic_c_gdjako@quicinc.comSigned-off-by: Joerg Roedel <jroedel@suse.de>

The PTE methods currently operate on a single entry. In preparation
for manipulating multiple PTEs in one map or unmap call, allow them
to handle multiple PTEs.
Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
Suggested-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Georgi Djakov <quic_c_gdjako@quicinc.com>
Link: https://lore.kernel.org/r/1623850736-389584-10-git-send-email-quic_c_gdjako@quicinc.comSigned-off-by: Joerg Roedel <jroedel@suse.de>

Since iommu_pgsize can calculate how many pages of the
same size can be mapped/unmapped before the next largest
page size boundary, add support for invoking an IOMMU
driver's map_pages() callback, if it provides one.
Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
Suggested-by: Will Deacon <will@kernel.org>
Signed-off-by: Georgi Djakov <quic_c_gdjako@quicinc.com>
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: https://lore.kernel.org/r/1623850736-389584-9-git-send-email-quic_c_gdjako@quicinc.comSigned-off-by: Joerg Roedel <jroedel@suse.de>

Extend iommu_pgsize() to populate an optional 'count' parameter so that
we can direct unmapping operation to the ->unmap_pages callback if it
has been provided by the driver.
Signed-off-by: Will Deacon <will@kernel.org>
Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
Signed-off-by: Georgi Djakov <quic_c_gdjako@quicinc.com>
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: https://lore.kernel.org/r/1623850736-389584-8-git-send-email-quic_c_gdjako@quicinc.comSigned-off-by: Joerg Roedel <jroedel@suse.de>

The 'addr_merge' parameter to iommu_pgsize() is a fabricated address
intended to describe the alignment requirements to consider when
choosing an appropriate page size. On the iommu_map() path, this address
is the logical OR of the virtual and physical addresses.

Subsequent improvements to iommu_pgsize() will need to check the
alignment of the virtual and physical components of 'addr_merge'
independently, so pass them in as separate parameters and reconstruct
'addr_merge' locally.

No functional change.
Signed-off-by: Will Deacon <will@kernel.org>
Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
Signed-off-by: Georgi Djakov <quic_c_gdjako@quicinc.com>
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: https://lore.kernel.org/r/1623850736-389584-7-git-send-email-quic_c_gdjako@quicinc.comSigned-off-by: Joerg Roedel <jroedel@suse.de>

Avoid the potential for shifting values by amounts greater than the
width of their type by using a bitmap to compute page size in
iommu_pgsize().
Signed-off-by: Will Deacon <will@kernel.org>
Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
Signed-off-by: Georgi Djakov <quic_c_gdjako@quicinc.com>
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: https://lore.kernel.org/r/1623850736-389584-6-git-send-email-quic_c_gdjako@quicinc.comSigned-off-by: Joerg Roedel <jroedel@suse.de>

Add a callback for IOMMU drivers to provide a path for the
IOMMU framework to call into an IOMMU driver, which can
call into the io-pgtable code, to map a physically contiguous
rnage of pages of the same size.

For IOMMU drivers that do not specify a map_pages() callback,
the existing logic of mapping memory one page block at a time
will be used.
Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
Suggested-by: Will Deacon <will@kernel.org>
Acked-by: Lu Baolu <baolu.lu@linux.intel.com>
Signed-off-by: Georgi Djakov <quic_c_gdjako@quicinc.com>
Link: https://lore.kernel.org/r/1623850736-389584-5-git-send-email-quic_c_gdjako@quicinc.comSigned-off-by: Joerg Roedel <jroedel@suse.de>

Mapping memory into io-pgtables follows the same semantics
that unmapping memory used to follow (i.e. a buffer will be
mapped one page block per call to the io-pgtable code). This
means that it can be optimized in the same way that unmapping
memory was, so add a map_pages() callback to the io-pgtable
ops structure, so that a range of pages of the same size
can be mapped within the same call.
Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
Suggested-by: Will Deacon <will@kernel.org>
Signed-off-by: Georgi Djakov <quic_c_gdjako@quicinc.com>
Link: https://lore.kernel.org/r/1623850736-389584-4-git-send-email-quic_c_gdjako@quicinc.comSigned-off-by: Joerg Roedel <jroedel@suse.de>

Add a callback for IOMMU drivers to provide a path for the
IOMMU framework to call into an IOMMU driver, which can call
into the io-pgtable code, to unmap a virtually contiguous
range of pages of the same size.

For IOMMU drivers that do not specify an unmap_pages() callback,
the existing logic of unmapping memory one page block at a time
will be used.
Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
Suggested-by: Will Deacon <will@kernel.org>
Signed-off-by: Will Deacon <will@kernel.org>
Acked-by: Lu Baolu <baolu.lu@linux.intel.com>
Signed-off-by: Georgi Djakov <quic_c_gdjako@quicinc.com>
Link: https://lore.kernel.org/r/1623850736-389584-3-git-send-email-quic_c_gdjako@quicinc.comSigned-off-by: Joerg Roedel <jroedel@suse.de>

The io-pgtable code expects to operate on a single block or
granule of memory that is supported by the IOMMU hardware when
unmapping memory.

This means that when a large buffer that consists of multiple
such blocks is unmapped, the io-pgtable code will walk the page
tables to the correct level to unmap each block, even for blocks
that are virtually contiguous and at the same level, which can
incur an overhead in performance.

Introduce the unmap_pages() page table op to express to the
io-pgtable code that it should unmap a number of blocks of
the same size, instead of a single block. Doing so allows
multiple blocks to be unmapped in one call to the io-pgtable
code, reducing the number of page table walks, and indirect
calls.
Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
Suggested-by: Will Deacon <will@kernel.org>
Signed-off-by: Will Deacon <will@kernel.org>
Signed-off-by: Georgi Djakov <quic_c_gdjako@quicinc.com>
Link: https://lore.kernel.org/r/1623850736-389584-2-git-send-email-quic_c_gdjako@quicinc.comSigned-off-by: Joerg Roedel <jroedel@suse.de>

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>