Merge branch 'for-next/fixes' into for-next/core

Merge for-next/fixes to resolve conflicts in arm64_hugetlb_cma_reserve().

* for-next/fixes:
  acpi/arm64: fix next_platform_timer() section mismatch error
  arm64/hugetlb: fix CMA gigantic page order for non-4K PAGE_SIZE

Merge branch 'for-next/vdso' into for-next/core

* for-next/vdso:
  arm64: vdso32: require CROSS_COMPILE_COMPAT for gcc+bfd
  arm64: vdso32: suppress error message for 'make mrproper'
  arm64: vdso32: drop test for -march=armv8-a
  arm64: vdso32: drop the test for dmb ishld

Merge branch 'for-next/trbe-errata' into for-next/core

* for-next/trbe-errata:
  arm64: errata: Add detection for TRBE write to out-of-range
  arm64: errata: Add workaround for TSB flush failures
  arm64: errata: Add detection for TRBE overwrite in FILL mode
  arm64: Add Neoverse-N2, Cortex-A710 CPU part definition

Merge branch 'for-next/sve' into for-next/core

* for-next/sve:
  arm64/sve: Fix warnings when SVE is disabled
  arm64/sve: Add stub for sve_max_virtualisable_vl()
  arm64/sve: Track vector lengths for tasks in an array
  arm64/sve: Explicitly load vector length when restoring SVE state
  arm64/sve: Put system wide vector length information into structs
  arm64/sve: Use accessor functions for vector lengths in thread_struct
  arm64/sve: Rename find_supported_vector_length()
  arm64/sve: Make access to FFR optional
  arm64/sve: Make sve_state_size() static
  arm64/sve: Remove sve_load_from_fpsimd_state()
  arm64/fp: Reindent fpsimd_save()

Merge branch 'for-next/scs' into for-next/core

* for-next/scs:
  scs: Release kasan vmalloc poison in scs_free process

Merge branch 'for-next/pfn-valid' into for-next/core

* for-next/pfn-valid:
  arm64/mm: drop HAVE_ARCH_PFN_VALID
  dma-mapping: remove bogus test for pfn_valid from dma_map_resource

Merge branch 'for-next/perf' into for-next/core

* for-next/perf:
  drivers/perf: Improve build test coverage
  drivers/perf: thunderx2_pmu: Change data in size tx2_uncore_event_update()
  drivers/perf: hisi: Fix PA PMU counter offset

Merge branch 'for-next/mte' into for-next/core

* for-next/mte:
  kasan: Extend KASAN mode kernel parameter
  arm64: mte: Add asymmetric mode support
  arm64: mte: CPU feature detection for Asymm MTE
  arm64: mte: Bitfield definitions for Asymm MTE
  kasan: Remove duplicate of kasan_flag_async
  arm64: kasan: mte: move GCR_EL1 switch to task switch when KASAN disabled

Merge branch 'for-next/mm' into for-next/core

* for-next/mm:
  arm64: mm: update max_pfn after memory hotplug
  arm64/mm: Add pud_sect_supported()
  arm64: mm: Drop pointless call to set_max_mapnr()

Merge branch 'for-next/misc' into for-next/core

* for-next/misc:
  arm64: Select POSIX_CPU_TIMERS_TASK_WORK
  arm64: Document boot requirements for FEAT_SME_FA64
  arm64: ftrace: use function_nocfi for _mcount as well
  arm64: asm: setup.h: export common variables
  arm64/traps: Avoid unnecessary kernel/user pointer conversion

Merge branch 'for-next/kselftest' into for-next/core

* for-next/kselftest:
  selftests: arm64: Factor out utility functions for assembly FP tests
  selftests: arm64: Add coverage of ptrace flags for SVE VL inheritance
  selftests: arm64: Verify that all possible vector lengths are handled
  selftests: arm64: Fix and enable test for setting current VL in vec-syscfg
  selftests: arm64: Remove bogus error check on writing to files
  selftests: arm64: Fix printf() format mismatch in vec-syscfg
  selftests: arm64: Move FPSIMD in SVE ptrace test into a function
  selftests: arm64: More comprehensively test the SVE ptrace interface
  selftests: arm64: Verify interoperation of SVE and FPSIMD register sets
  selftests: arm64: Clarify output when verifying SVE register set
  selftests: arm64: Document what the SVE ptrace test is doing
  selftests: arm64: Remove extraneous register setting code
  selftests: arm64: Don't log child creation as a test in SVE ptrace test
  selftests: arm64: Use a define for the number of SVE ptrace tests to be run

Merge branch 'for-next/kexec' into for-next/core

* for-next/kexec:
  arm64: trans_pgd: remove trans_pgd_map_page()
  arm64: kexec: remove cpu-reset.h
  arm64: kexec: remove the pre-kexec PoC maintenance
  arm64: kexec: keep MMU enabled during kexec relocation
  arm64: kexec: install a copy of the linear-map
  arm64: kexec: use ld script for relocation function
  arm64: kexec: relocate in EL1 mode
  arm64: kexec: configure EL2 vectors for kexec
  arm64: kexec: pass kimage as the only argument to relocation function
  arm64: kexec: Use dcache ops macros instead of open-coding
  arm64: kexec: skip relocation code for inplace kexec
  arm64: kexec: flush image and lists during kexec load time
  arm64: hibernate: abstract ttrb0 setup function
  arm64: trans_pgd: hibernate: Add trans_pgd_copy_el2_vectors
  arm64: kernel: add helper for booted at EL2 and not VHE

Merge branch 'for-next/extable' into for-next/core

* for-next/extable:
  arm64: vmlinux.lds.S: remove `.fixup` section
  arm64: extable: add load_unaligned_zeropad() handler
  arm64: extable: add a dedicated uaccess handler
  arm64: extable: add `type` and `data` fields
  arm64: extable: use `ex` for `exception_table_entry`
  arm64: extable: make fixup_exception() return bool
  arm64: extable: consolidate definitions
  arm64: gpr-num: support W registers
  arm64: factor out GPR numbering helpers
  arm64: kvm: use kvm_exception_table_entry
  arm64: lib: __arch_copy_to_user(): fold fixups into body
  arm64: lib: __arch_copy_from_user(): fold fixups into body
  arm64: lib: __arch_clear_user(): fold fixups into body

Merge branch 'for-next/8.6-timers' into for-next/core

* for-next/8.6-timers:
  arm64: Add HWCAP for self-synchronising virtual counter
  arm64: Add handling of CNTVCTSS traps
  arm64: Add CNT{P,V}CTSS_EL0 alternatives to cnt{p,v}ct_el0
  arm64: Add a capability for FEAT_ECV
  clocksource/drivers/arch_arm_timer: Move workaround synchronisation around
  clocksource/drivers/arm_arch_timer: Fix masking for high freq counters
  clocksource/drivers/arm_arch_timer: Drop unnecessary ISB on CVAL programming
  clocksource/drivers/arm_arch_timer: Remove any trace of the TVAL programming interface
  clocksource/drivers/arm_arch_timer: Work around broken CVAL implementations
  clocksource/drivers/arm_arch_timer: Advertise 56bit timer to the core code
  clocksource/drivers/arm_arch_timer: Move MMIO timer programming over to CVAL
  clocksource/drivers/arm_arch_timer: Fix MMIO base address vs callback ordering issue
  clocksource/drivers/arm_arch_timer: Move drop _tval from erratum function names
  clocksource/drivers/arm_arch_timer: Move system register timer programming over to CVAL
  clocksource/drivers/arm_arch_timer: Extend write side of timer register accessors to u64
  clocksource/drivers/arm_arch_timer: Drop CNT*_TVAL read accessors
  clocksource/arm_arch_timer: Add build-time guards for unhandled register accesses

arm64: Select POSIX_CPU_TIMERS_TASK_WORK

With 6caa5812e2d1 ("KVM: arm64: Use generic KVM xfer to guest work
function") all arm64 exit paths are properly equipped to handle the
POSIX timers' task work.

Deferring timer callbacks to thread context, not only limits the amount
of time spent in hard interrupt context, but is a safer
implementation[1], and will allow PREEMPT_RT setups to use KVM[2].

So let's enable POSIX_CPU_TIMERS_TASK_WORK on arm64.

[1] https://lore.kernel.org/all/20200716201923.228696399@linutronix.de/
[2] https://lore.kernel.org/linux-rt-users/87v92bdnlx.ffs@tglx/

Signed-off-by: Nicolas Saenz Julienne <nsaenzju@redhat.com>
Acked-by: Mark Rutland <mark.rutland@arm.com>
Acked-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20211018144713.873464-1-nsaenzju@redhat.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: Document boot requirements for FEAT_SME_FA64

The EAC1 release of the SME specification adds the FA64 feature which
requires enablement at higher ELs before lower ELs can use it. Document
what we require from higher ELs in our boot requirements.

Signed-off-by: Mark Brown <broonie@kernel.org>
Link: https://lore.kernel.org/r/20211026111802.12853-1-broonie@kernel.org
Signed-off-by: Will Deacon <will@kernel.org>

arm64/sve: Fix warnings when SVE is disabled

In configurations where SVE is disabled we define but never reference the
functions for retrieving the default vector length, causing warnings. Fix
this by move the ifdef up, marking get_default_vl() inline since it is
referenced from code guarded by an IS_ENABLED() check, and do the same for
the other accessors for consistency.

Reported-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Link: https://lore.kernel.org/r/20211022141635.2360415-3-broonie@kernel.org
Signed-off-by: Will Deacon <will@kernel.org>

arm64/sve: Add stub for sve_max_virtualisable_vl()

Fixes build problems for configurations with KVM enabled but SVE disabled.

Reported-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Link: https://lore.kernel.org/r/20211022141635.2360415-2-broonie@kernel.org
Signed-off-by: Will Deacon <will@kernel.org>

arm64: errata: Add detection for TRBE write to out-of-range

Arm Neoverse-N2 and Cortex-A710 cores are affected by an erratum where
the trbe, under some circumstances, might write upto 64bytes to an
address after the Limit as programmed by the TRBLIMITR_EL1.LIMIT.
This might -
  - Corrupt a page in the ring buffer, which may corrupt trace from a
    previous session, consumed by userspace.
  - Hit the guard page at the end of the vmalloc area and raise a fault.

To keep the handling simpler, we always leave the last page from the
range, which TRBE is allowed to write. This can be achieved by ensuring
that we always have more than a PAGE worth space in the range, while
calculating the LIMIT for TRBE. And then the LIMIT pointer can be
adjusted to leave the PAGE (TRBLIMITR.LIMIT -= PAGE_SIZE), out of the
TRBE range while enabling it. This makes sure that the TRBE will only
write to an area within its allowed limit (i.e, [head-head+size]) and
we do not have to handle address faults within the driver.

Cc: Anshuman Khandual <anshuman.khandual@arm.com>
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: Will Deacon <will@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Reviewed-by: Anshuman Khandual <anshuman.khandual@arm.com>
Reviewed-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Link: https://lore.kernel.org/r/20211019163153.3692640-5-suzuki.poulose@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: errata: Add workaround for TSB flush failures

Arm Neoverse-N2 (#2067961) and Cortex-A710 (#2054223) suffers
from errata, where a TSB (trace synchronization barrier)
fails to flush the trace data completely, when executed from
a trace prohibited region. In Linux we always execute it
after we have moved the PE to trace prohibited region. So,
we can apply the workaround every time a TSB is executed.

The work around is to issue two TSB consecutively.

NOTE: This errata is defined as LOCAL_CPU_ERRATUM, implying
that a late CPU could be blocked from booting if it is the
first CPU that requires the workaround. This is because we
do not allow setting a cpu_hwcaps after the SMP boot. The
other alternative is to use "this_cpu_has_cap()" instead
of the faster system wide check, which may be a bit of an
overhead, given we may have to do this in nvhe KVM host
before a guest entry.

Cc: Will Deacon <will@kernel.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Anshuman Khandual <anshuman.khandual@arm.com>
Cc: Marc Zyngier <maz@kernel.org>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Reviewed-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Reviewed-by: Anshuman Khandual <anshuman.khandual@arm.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Link: https://lore.kernel.org/r/20211019163153.3692640-4-suzuki.poulose@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: errata: Add detection for TRBE overwrite in FILL mode

Arm Neoverse-N2 and the Cortex-A710 cores are affected
by a CPU erratum where the TRBE will overwrite the trace buffer
in FILL mode. The TRBE doesn't stop (as expected in FILL mode)
when it reaches the limit and wraps to the base to continue
writing upto 3 cache lines. This will overwrite any trace that
was written previously.

Add the Neoverse-N2 erratum(#2139208) and Cortex-A710 erratum
(#2119858) to the detection logic.

This will be used by the TRBE driver in later patches to work
around the issue. The detection has been kept with the core
arm64 errata framework list to make sure :
  - We don't duplicate the framework in TRBE driver
  - The errata detection is advertised like the rest
    of the CPU errata.

Note that the Kconfig entries are not fully active until the
TRBE driver implements the work around.

Cc: Will Deacon <will@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Anshuman Khandual <anshuman.khandual@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Mike Leach <mike.leach@linaro.org>
cc: Leo Yan <leo.yan@linaro.org>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Reviewed-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Reviewed-by: Anshuman Khandual <anshuman.khandual@arm.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Link: https://lore.kernel.org/r/20211019163153.3692640-3-suzuki.poulose@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: Add Neoverse-N2, Cortex-A710 CPU part definition

Add the CPU Partnumbers for the new Arm designs.

Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will@kernel.org>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Reviewed-by: Anshuman Khandual <anshuman.khandual@arm.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Link: https://lore.kernel.org/r/20211019163153.3692640-2-suzuki.poulose@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

selftests: arm64: Factor out utility functions for assembly FP tests

The various floating point test programs written in assembly have a bunch
of helper functions and macros which are cut'n'pasted between them. Factor
them out into a separate source file which is linked into all of them.

We don't include memcmp() since it isn't as generic as it should be and
directly branches to report an error in the programs.

Signed-off-by: Mark Brown <broonie@kernel.org>
Link: https://lore.kernel.org/r/20211019181851.3341232-1-broonie@kernel.org
Signed-off-by: Will Deacon <will@kernel.org>

arm64: vmlinux.lds.S: remove `.fixup` section

We no longer place anything into a `.fixup` section, so we no longer
need to place those sections into the `.text` section in the main kernel
Image.

Remove the use of `.fixup`.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Reviewed-by: Ard Biesheuvel <ardb@kernel.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: James Morse <james.morse@arm.com>
Cc: Robin Murphy <robin.murphy@arm.com>
Cc: Will Deacon <will@kernel.org>
Link: https://lore.kernel.org/r/20211019160219.5202-14-mark.rutland@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: extable: add load_unaligned_zeropad() handler

For inline assembly, we place exception fixups out-of-line in the
`.fixup` section such that these are out of the way of the fast path.
This has a few drawbacks:

* Since the fixup code is anonymous, backtraces will symbolize fixups as
  offsets from the nearest prior symbol, currently
  `__entry_tramp_text_end`. This is confusing, and painful to debug
  without access to the relevant vmlinux.

* Since the exception handler adjusts the PC to execute the fixup, and
  the fixup uses a direct branch back into the function it fixes,
  backtraces of fixups miss the original function. This is confusing,
  and violates requirements for RELIABLE_STACKTRACE (and therefore
  LIVEPATCH).

* Inline assembly and associated fixups are generated from templates,
  and we have many copies of logically identical fixups which only
  differ in which specific registers are written to and which address is
  branched to at the end of the fixup. This is potentially wasteful of
  I-cache resources, and makes it hard to add additional logic to fixups
  without significant bloat.

* In the case of load_unaligned_zeropad(), the logic in the fixup
  requires a temporary register that we must allocate even in the
  fast-path where it will not be used.

This patch address all four concerns for load_unaligned_zeropad() fixups
by adding a dedicated exception handler which performs the fixup logic
in exception context and subsequent returns back after the faulting
instruction. For the moment, the fixup logic is identical to the old
assembly fixup logic, but in future we could enhance this by taking the
ESR and FAR into account to constrain the faults we try to fix up, or to
specialize fixups for MTE tag check faults.

Other than backtracing, there should be no functional change as a result
of this patch.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Reviewed-by: Ard Biesheuvel <ardb@kernel.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: James Morse <james.morse@arm.com>
Cc: Robin Murphy <robin.murphy@arm.com>
Cc: Will Deacon <will@kernel.org>
Link: https://lore.kernel.org/r/20211019160219.5202-13-mark.rutland@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: extable: add a dedicated uaccess handler

For inline assembly, we place exception fixups out-of-line in the
`.fixup` section such that these are out of the way of the fast path.
This has a few drawbacks:

* Since the fixup code is anonymous, backtraces will symbolize fixups as
  offsets from the nearest prior symbol, currently
  `__entry_tramp_text_end`. This is confusing, and painful to debug
  without access to the relevant vmlinux.

* Since the exception handler adjusts the PC to execute the fixup, and
  the fixup uses a direct branch back into the function it fixes,
  backtraces of fixups miss the original function. This is confusing,
  and violates requirements for RELIABLE_STACKTRACE (and therefore
  LIVEPATCH).

* Inline assembly and associated fixups are generated from templates,
  and we have many copies of logically identical fixups which only
  differ in which specific registers are written to and which address is
  branched to at the end of the fixup. This is potentially wasteful of
  I-cache resources, and makes it hard to add additional logic to fixups
  without significant bloat.

This patch address all three concerns for inline uaccess fixups by
adding a dedicated exception handler which updates registers in
exception context and subsequent returns back into the function which
faulted, removing the need for fixups specialized to each faulting
instruction.

Other than backtracing, there should be no functional change as a result
of this patch.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Reviewed-by: Ard Biesheuvel <ardb@kernel.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: James Morse <james.morse@arm.com>
Cc: Robin Murphy <robin.murphy@arm.com>
Cc: Will Deacon <will@kernel.org>
Link: https://lore.kernel.org/r/20211019160219.5202-12-mark.rutland@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: extable: add `type` and `data` fields

Subsequent patches will add specialized handlers for fixups, in addition
to the simple PC fixup and BPF handlers we have today. In preparation,
this patch adds a new `type` field to struct exception_table_entry, and
uses this to distinguish the fixup and BPF cases. A `data` field is also
added so that subsequent patches can associate data specific to each
exception site (e.g. register numbers).

Handlers are named ex_handler_*() for consistency, following the exmaple
of x86. At the same time, get_ex_fixup() is split out into a helper so
that it can be used by other ex_handler_*() functions ins subsequent
patches.

This patch will increase the size of the exception tables, which will be
remedied by subsequent patches removing redundant fixup code. There
should be no functional change as a result of this patch.

Since each entry is now 12 bytes in size, we must reduce the alignment
of each entry from `.align 3` (i.e. 8 bytes) to `.align 2` (i.e. 4
bytes), which is the natrual alignment of the `insn` and `fixup` fields.
The current 8-byte alignment is a holdover from when the `insn` and
`fixup` fields was 8 bytes, and while not harmful has not been necessary
since commit:

  6c94f27ac847ff8e ("arm64: switch to relative exception tables")

Similarly, RO_EXCEPTION_TABLE_ALIGN is dropped to 4 bytes.

Concurrently with this patch, x86's exception table entry format is
being updated (similarly to a 12-byte format, with 32-bytes of absolute
data). Once both have been merged it should be possible to unify the
sorttable logic for the two.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Reviewed-by: Ard Biesheuvel <ardb@kernel.org>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Andrii Nakryiko <andrii@kernel.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: James Morse <james.morse@arm.com>
Cc: Jean-Philippe Brucker <jean-philippe@linaro.org>
Cc: Robin Murphy <robin.murphy@arm.com>
Cc: Will Deacon <will@kernel.org>
Link: https://lore.kernel.org/r/20211019160219.5202-11-mark.rutland@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: extable: use `ex` for `exception_table_entry`

Subsequent patches will extend `struct exception_table_entry` with more
fields, and the distinction between the entry and its `fixup` field will
become more important.

For clarity, let's consistently use `ex` to refer to refer to an entire
entry. In subsequent patches we'll use `fixup` to refer to the fixup
field specifically. This matches the naming convention used today in
arch/arm64/net/bpf_jit_comp.c.

There should be no functional change as a result of this patch.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Acked-by: Robin Murphy <robin.murphy@arm.com>
Reviewed-by: Ard Biesheuvel <ardb@kernel.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: James Morse <james.morse@arm.com>
Cc: Will Deacon <will@kernel.org>
Link: https://lore.kernel.org/r/20211019160219.5202-10-mark.rutland@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: extable: make fixup_exception() return bool

The return values of fixup_exception() and arm64_bpf_fixup_exception()
represent a boolean condition rather than an error code, so for clarity
it would be better to return `bool` rather than `int`.

This patch adjusts the code accordingly. While we're modifying the
prototype, we also remove the unnecessary `extern` keyword, so that this
won't look out of place when we make subsequent additions to the header.

There should be no functional change as a result of this patch.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Reviewed-by: Ard Biesheuvel <ardb@kernel.org>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Andrii Nakryiko <andrii@kernel.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: James Morse <james.morse@arm.com>
Cc: Jean-Philippe Brucker <jean-philippe@linaro.org>
Cc: Robin Murphy <robin.murphy@arm.com>
Cc: Will Deacon <will@kernel.org>
Link: https://lore.kernel.org/r/20211019160219.5202-9-mark.rutland@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: extable: consolidate definitions

In subsequent patches we'll alter the structure and usage of struct
exception_table_entry. For inline assembly, we create these using the
`_ASM_EXTABLE()` CPP macro defined in <asm/uaccess.h>, and for plain
assembly code we use the `_asm_extable()` GAS macro defined in
<asm/assembler.h>, which are largely identical save for different
escaping and stringification requirements.

This patch moves the common definitions to a new <asm/asm-extable.h>
header, so that it's easier to keep the two in-sync, and to remove the
implication that these are only used for uaccess helpers (as e.g.
load_unaligned_zeropad() is only used on kernel memory, and depends upon
`_ASM_EXTABLE()`.

At the same time, a few minor modifications are made for clarity and in
preparation for subsequent patches:

* The structure creation is factored out into an `__ASM_EXTABLE_RAW()`
  macro. This will make it easier to support different fixup variants in
  subsequent patches without needing to update all users of
  `_ASM_EXTABLE()`, and makes it easier to see tha the CPP and GAS
  variants of the macros are structurally identical.

  For the CPP macro, the stringification of fields is left to the
  wrapper macro, `_ASM_EXTABLE()`, as in subsequent patches it will be
  necessary to stringify fields in wrapper macros to safely concatenate
  strings which cannot be token-pasted together in CPP.

* The fields of the structure are created separately on their own lines.
  This will make it easier to add/remove/modify individual fields
  clearly.

* Additional parentheses are added around the use of macro arguments in
  field definitions to avoid any potential problems with evaluation due
  to operator precedence, and to make errors upon misuse clearer.

* USER() is moved into <asm/asm-uaccess.h>, as it is not required by all
  assembly code, and is already refered to by comments in that file.

There should be no functional change as a result of this patch.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Reviewed-by: Ard Biesheuvel <ardb@kernel.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: James Morse <james.morse@arm.com>
Cc: Robin Murphy <robin.murphy@arm.com>
Cc: Will Deacon <will@kernel.org>
Link: https://lore.kernel.org/r/20211019160219.5202-8-mark.rutland@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: gpr-num: support W registers

In subsequent patches we'll want to map W registers to their register
numbers. Update gpr-num.h so that we can do this.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Reviewed-by: Ard Biesheuvel <ardb@kernel.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: James Morse <james.morse@arm.com>
Cc: Robin Murphy <robin.murphy@arm.com>
Cc: Will Deacon <will@kernel.org>
Link: https://lore.kernel.org/r/20211019160219.5202-7-mark.rutland@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: factor out GPR numbering helpers

In <asm/sysreg.h> we have macros to convert the names of general purpose
registers (GPRs) into integer constants, which we use to manually build
the encoding for `MRS` and `MSR` instructions where we can't rely on the
assembler to do so for us.

In subsequent patches we'll need to map the same GPR names to integer
constants so that we can use this to build metadata for exception
fixups.

So that the we can use the mappings elsewhere, factor out the
definitions into a new <asm/gpr-num.h> header, renaming the definitions
to align with this "GPR num" naming for clarity.

There should be no functional change as a result of this patch.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Reviewed-by: Ard Biesheuvel <ardb@kernel.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: James Morse <james.morse@arm.com>
Cc: Robin Murphy <robin.murphy@arm.com>
Cc: Will Deacon <will@kernel.org>
Link: https://lore.kernel.org/r/20211019160219.5202-6-mark.rutland@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: kvm: use kvm_exception_table_entry

In subsequent patches we'll alter `struct exception_table_entry`, adding
fields that are not needed for KVM exception fixups.

In preparation for this, migrate KVM to its own `struct
kvm_exception_table_entry`, which is identical to the current format of
`struct exception_table_entry`. Comments are updated accordingly.

There should be no functional change as a result of this patch.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Reviewed-by: Ard Biesheuvel <ardb@kernel.org>
Cc: Alexandru Elisei <alexandru.elisei@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: James Morse <james.morse@arm.com>
Cc: Marc Zyngier <maz@kernel.org>
Cc: Robin Murphy <robin.murphy@arm.com>
Cc: Suzuki K Poulose <suzuki.poulose@arm.com>
Cc: Will Deacon <will@kernel.org>
Acked-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20211019160219.5202-5-mark.rutland@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: lib: __arch_copy_to_user(): fold fixups into body

Like other functions, __arch_copy_to_user() places its exception fixups
in the `.fixup` section without any clear association with
__arch_copy_to_user() itself. If we backtrace the fixup code, it will be
symbolized as an offset from the nearest prior symbol, which happens to
be `__entry_tramp_text_end`. Further, since the PC adjustment for the
fixup is akin to a direct branch rather than a function call,
__arch_copy_to_user() itself will be missing from the backtrace.

This is confusing and hinders debugging. In general this pattern will
also be problematic for CONFIG_LIVEPATCH, since fixups often return to
their associated function, but this isn't accurately captured in the
stacktrace.

To solve these issues for assembly functions, we must move fixups into
the body of the functions themselves, after the usual fast-path returns.
This patch does so for __arch_copy_to_user().

Inline assembly will be dealt with in subsequent patches.

Other than the improved backtracing, there should be no functional
change as a result of this patch.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Acked-by: Robin Murphy <robin.murphy@arm.com>
Reviewed-by: Ard Biesheuvel <ardb@kernel.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: James Morse <james.morse@arm.com>
Cc: Mark Brown <broonie@kernel.org>
Cc: Will Deacon <will@kernel.org>
Link: https://lore.kernel.org/r/20211019160219.5202-4-mark.rutland@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: lib: __arch_copy_from_user(): fold fixups into body

Like other functions, __arch_copy_from_user() places its exception
fixups in the `.fixup` section without any clear association with
__arch_copy_from_user() itself. If we backtrace the fixup code, it will
be symbolized as an offset from the nearest prior symbol, which happens
to be `__entry_tramp_text_end`. Further, since the PC adjustment for the
fixup is akin to a direct branch rather than a function call,
__arch_copy_from_user() itself will be missing from the backtrace.

This is confusing and hinders debugging. In general this pattern will
also be problematic for CONFIG_LIVEPATCH, since fixups often return to
their associated function, but this isn't accurately captured in the
stacktrace.

To solve these issues for assembly functions, we must move fixups into
the body of the functions themselves, after the usual fast-path returns.
This patch does so for __arch_copy_from_user().

Inline assembly will be dealt with in subsequent patches.

Other than the improved backtracing, there should be no functional
change as a result of this patch.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Acked-by: Robin Murphy <robin.murphy@arm.com>
Reviewed-by: Ard Biesheuvel <ardb@kernel.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: James Morse <james.morse@arm.com>
Cc: Mark Brown <broonie@kernel.org>
Cc: Will Deacon <will@kernel.org>
Link: https://lore.kernel.org/r/20211019160219.5202-3-mark.rutland@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: lib: __arch_clear_user(): fold fixups into body

Like other functions, __arch_clear_user() places its exception fixups in
the `.fixup` section without any clear association with
__arch_clear_user() itself. If we backtrace the fixup code, it will be
symbolized as an offset from the nearest prior symbol, which happens to
be `__entry_tramp_text_end`. Further, since the PC adjustment for the
fixup is akin to a direct branch rather than a function call,
__arch_clear_user() itself will be missing from the backtrace.

This is confusing and hinders debugging. In general this pattern will
also be problematic for CONFIG_LIVEPATCH, since fixups often return to
their associated function, but this isn't accurately captured in the
stacktrace.

To solve these issues for assembly functions, we must move fixups into
the body of the functions themselves, after the usual fast-path returns.
This patch does so for __arch_clear_user().

Inline assembly will be dealt with in subsequent patches.

Other than the improved backtracing, there should be no functional
change as a result of this patch.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Acked-by: Robin Murphy <robin.murphy@arm.com>
Reviewed-by: Ard Biesheuvel <ardb@kernel.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: James Morse <james.morse@arm.com>
Cc: Mark Brown <broonie@kernel.org>
Cc: Will Deacon <will@kernel.org>
Link: https://lore.kernel.org/r/20211019160219.5202-2-mark.rutland@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: vdso32: require CROSS_COMPILE_COMPAT for gcc+bfd

Similar to
commit 231ad7f409f1 ("Makefile: infer --target from ARCH for CC=clang")
There really is no point in setting --target based on
$CROSS_COMPILE_COMPAT for clang when the integrated assembler is being
used, since
commit ef94340583ee ("arm64: vdso32: drop -no-integrated-as flag").

Allows COMPAT_VDSO to be selected without setting $CROSS_COMPILE_COMPAT
when using clang and lld together.

Before:
$ ARCH=arm64 CROSS_COMPILE_COMPAT=arm-linux-gnueabi- make -j72 LLVM=1 defconfig
$ grep CONFIG_COMPAT_VDSO .config
CONFIG_COMPAT_VDSO=y
$ ARCH=arm64 make -j72 LLVM=1 defconfig
$ grep CONFIG_COMPAT_VDSO .config
$

After:
$ ARCH=arm64 CROSS_COMPILE_COMPAT=arm-linux-gnueabi- make -j72 LLVM=1 defconfig
$ grep CONFIG_COMPAT_VDSO .config
CONFIG_COMPAT_VDSO=y
$ ARCH=arm64 make -j72 LLVM=1 defconfig
$ grep CONFIG_COMPAT_VDSO .config
CONFIG_COMPAT_VDSO=y

Reviewed-by: Nathan Chancellor <nathan@kernel.org>
Suggested-by: Nathan Chancellor <nathan@kernel.org>
Tested-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Nick Desaulniers <ndesaulniers@google.com>
Reviewed-by: Vincenzo Frascino <vincenzo.frascino@arm.com>
Link: https://lore.kernel.org/r/20211019223646.1146945-5-ndesaulniers@google.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: vdso32: suppress error message for 'make mrproper'

When running the following command without arm-linux-gnueabi-gcc in
one's $PATH, the following warning is observed:

$ ARCH=arm64 CROSS_COMPILE_COMPAT=arm-linux-gnueabi- make -j72 LLVM=1 mrproper
make[1]: arm-linux-gnueabi-gcc: No such file or directory

This is because KCONFIG is not run for mrproper, so CONFIG_CC_IS_CLANG
is not set, and we end up eagerly evaluating various variables that try
to invoke CC_COMPAT.

This is a similar problem to what was observed in
commit dc960bfeedb0 ("h8300: suppress error messages for 'make clean'")

Reported-by: Lucas Henneman <henneman@google.com>
Suggested-by: Masahiro Yamada <masahiroy@kernel.org>
Signed-off-by: Nick Desaulniers <ndesaulniers@google.com>
Reviewed-by: Vincenzo Frascino <vincenzo.frascino@arm.com>
Reviewed-by: Nathan Chancellor <nathan@kernel.org>
Tested-by: Nathan Chancellor <nathan@kernel.org>
Link: https://lore.kernel.org/r/20211019223646.1146945-4-ndesaulniers@google.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: vdso32: drop test for -march=armv8-a

As Arnd points out:
  gcc-4.8 already supported -march=armv8, and we require gcc-5.1 now, so
  both this #if/#else construct and the corresponding
  "cc32-option,-march=armv8-a" check should be obsolete now.

Link: https://lore.kernel.org/lkml/CAK8P3a3UBEJ0Py2ycz=rHfgog8g3mCOeQOwO0Gmp-iz6Uxkapg@mail.gmail.com/
Suggested-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Nick Desaulniers <ndesaulniers@google.com>
Reviewed-by: Vincenzo Frascino <vincenzo.frascino@arm.com>
Reviewed-by: Nathan Chancellor <nathan@kernel.org>
Link: https://lore.kernel.org/r/20211019223646.1146945-3-ndesaulniers@google.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: vdso32: drop the test for dmb ishld

Binutils added support for this instruction in commit
e797f7e0b2bedc9328d4a9a0ebc63ca7a2dbbebc which shipped in 2.24 (just
missing the 2.23 release) but was cherry-picked into 2.23 in commit
27a50d6755bae906bc73b4ec1a8b448467f0bea1. Thanks to Christian and Simon
for helping me with the patch archaeology.

According to Documentation/process/changes.rst, the minimum supported
version of binutils is 2.23. Since all supported versions of GAS support
this instruction, drop the assembler invocation, preprocessor
flags/guards, and the cross assembler macro that's now unused.

This also avoids a recursive self reference in a follow up cleanup
patch.

Cc: Christian Biesinger <cbiesinger@google.com>
Cc: Simon Marchi <simon.marchi@polymtl.ca>
Signed-off-by: Nick Desaulniers <ndesaulniers@google.com>
Reviewed-by: Vincenzo Frascino <vincenzo.frascino@arm.com>
Reviewed-by: Nathan Chancellor <nathan@kernel.org>
Link: https://lore.kernel.org/r/20211019223646.1146945-2-ndesaulniers@google.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64/sve: Track vector lengths for tasks in an array

As for SVE we will track a per task SME vector length for tasks. Convert
the existing storage for the vector length into an array and update
fpsimd_flush_task() to initialise this in a function.

Signed-off-by: Mark Brown <broonie@kernel.org>
Link: https://lore.kernel.org/r/20211019172247.3045838-10-broonie@kernel.org
Signed-off-by: Will Deacon <will@kernel.org>

arm64/sve: Explicitly load vector length when restoring SVE state

Currently when restoring the SVE state we supply the SVE vector length
as an argument to sve_load_state() and the underlying macros. This becomes
inconvenient with the addition of SME since we may need to restore any
combination of SVE and SME vector lengths, and we already separately
restore the vector length in the KVM code. We don't need to know the vector
length during the actual register load since the SME load instructions can
index into the data array for us.

Refactor the interface so we explicitly set the vector length separately
to restoring the SVE registers in preparation for adding SME support, no
functional change should be involved.

Signed-off-by: Mark Brown <broonie@kernel.org>
Link: https://lore.kernel.org/r/20211019172247.3045838-9-broonie@kernel.org
Signed-off-by: Will Deacon <will@kernel.org>

arm64/sve: Put system wide vector length information into structs

With the introduction of SME we will have a second vector length in the
system, enumerated and configured in a very similar fashion to the
existing SVE vector length.  While there are a few differences in how
things are handled this is a relatively small portion of the overall
code so in order to avoid code duplication we factor out

We create two structs, one vl_info for the static hardware properties
and one vl_config for the runtime configuration, with an array
instantiated for each and update all the users to reference these. Some
accessor functions are provided where helpful for readability, and the
write to set the vector length is put into a function since the system
register being updated needs to be chosen at compile time.

This is a mostly mechanical replacement, further work will be required
to actually make things generic, ensuring that we handle those places
where there are differences properly.

Signed-off-by: Mark Brown <broonie@kernel.org>
Link: https://lore.kernel.org/r/20211019172247.3045838-8-broonie@kernel.org
Signed-off-by: Will Deacon <will@kernel.org>

arm64/sve: Use accessor functions for vector lengths in thread_struct

In a system with SME there are parallel vector length controls for SVE and
SME vectors which function in much the same way so it is desirable to
share the code for handling them as much as possible. In order to prepare
for doing this add a layer of accessor functions for the various VL related
operations on tasks.

Since almost all current interactions are actually via task->thread rather
than directly with the thread_info the accessors use that. Accessors are
provided for both generic and SVE specific usage, the generic accessors
should be used for cases where register state is being manipulated since
the registers are shared between streaming and regular SVE so we know that
when SME support is implemented we will always have to be in the appropriate
mode already and hence can generalise now.

Since we are using task_struct and we don't want to cause widespread
inclusion of sched.h the acessors are all out of line, it is hoped that
none of the uses are in a sufficiently critical path for this to be an
issue. Those that are most likely to present an issue are in the same
translation unit so hopefully the compiler may be able to inline anyway.

This is purely adding the layer of abstraction, additional work will be
needed to support tasks using SME.

Signed-off-by: Mark Brown <broonie@kernel.org>
Link: https://lore.kernel.org/r/20211019172247.3045838-7-broonie@kernel.org
Signed-off-by: Will Deacon <will@kernel.org>

arm64/sve: Rename find_supported_vector_length()

The function has SVE specific checks in it and it will be more trouble
to add conditional code for SME than it is to simply rename it to be SVE
specific.

Signed-off-by: Mark Brown <broonie@kernel.org>
Link: https://lore.kernel.org/r/20211019172247.3045838-6-broonie@kernel.org
Signed-off-by: Will Deacon <will@kernel.org>

arm64/sve: Make access to FFR optional

SME introduces streaming SVE mode in which FFR is not present and the
instructions for accessing it UNDEF. In preparation for handling this
update the low level SVE state access functions to take a flag specifying
if FFR should be handled. When saving the register state we store a zero
for FFR to guard against uninitialized data being read. No behaviour change
should be introduced by this patch.

Signed-off-by: Mark Brown <broonie@kernel.org>
Link: https://lore.kernel.org/r/20211019172247.3045838-5-broonie@kernel.org
Signed-off-by: Will Deacon <will@kernel.org>

arm64/sve: Make sve_state_size() static

There are no users outside fpsimd.c so make sve_state_size() static.
KVM open codes an equivalent.

Signed-off-by: Mark Brown <broonie@kernel.org>
Link: https://lore.kernel.org/r/20211019172247.3045838-4-broonie@kernel.org
Signed-off-by: Will Deacon <will@kernel.org>

arm64/sve: Remove sve_load_from_fpsimd_state()

Following optimisations of the SVE register handling we no longer load the
SVE state from a saved copy of the FPSIMD registers, we convert directly
in registers or from one saved state to another. Remove the function so we
don't need to update it during further refactoring.

Signed-off-by: Mark Brown <broonie@kernel.org>
Link: https://lore.kernel.org/r/20211019172247.3045838-3-broonie@kernel.org
Signed-off-by: Will Deacon <will@kernel.org>

arm64/fp: Reindent fpsimd_save()

Currently all the active code in fpsimd_save() is inside a check for
TIF_FOREIGN_FPSTATE. Reduce the indentation level by changing to return
from the function if TIF_FOREIGN_FPSTATE is set.

Signed-off-by: Mark Brown <broonie@kernel.org>
Link: https://lore.kernel.org/r/20211019172247.3045838-2-broonie@kernel.org
Signed-off-by: Will Deacon <will@kernel.org>

arm64: Add HWCAP for self-synchronising virtual counter

Since userspace can make use of the CNTVSS_EL0 instruction, expose
it via a HWCAP.

Suggested-by: Will Deacon <will@kernel.org>
Acked-by: Will Deacon <will@kernel.org>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20211017124225.3018098-18-maz@kernel.org
Signed-off-by: Will Deacon <will@kernel.org>

arm64: Add handling of CNTVCTSS traps

Since CNTVCTSS obey the same control bits as CNTVCT, add the necessary
decoding to the hook table. Note that there is no known user of
this at the moment.

Acked-by: Will Deacon <will@kernel.org>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20211017124225.3018098-17-maz@kernel.org
Signed-off-by: Will Deacon <will@kernel.org>

arm64: Add CNT{P,V}CTSS_EL0 alternatives to cnt{p,v}ct_el0

CNTPCTSS_EL0 and CNTVCTSS_EL0 are alternatives to the usual
CNTPCT_EL0 and CNTVCT_EL0 that do not require a previous ISB
to be synchronised (SS stands for Self-Synchronising).

Use the ARM64_HAS_ECV capability to control alternative sequences
that switch to these low(er)-cost primitives. Note that the
counter access in the VDSO is for now left alone until we decide
whether we want to allow this.

Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20211017124225.3018098-16-maz@kernel.org
Signed-off-by: Will Deacon <will@kernel.org>

arm64: Add a capability for FEAT_ECV

Add a new capability to detect the Enhanced Counter Virtualization
feature (FEAT_ECV).

Reviewed-by: Oliver Upton <oupton@google.com>
Acked-by: Will Deacon <will@kernel.org>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20211017124225.3018098-15-maz@kernel.org
Signed-off-by: Will Deacon <will@kernel.org>

Merge branch 'timers/drivers/armv8.6_arch_timer' of https://git.linaro.org/people/daniel.lezcano/linux into for-next/8.6-timers

Pull Arm architected timer driver rework from Marc (via Daniel) so that
we can add the Armv8.6 support on top.

Link: https://lore.kernel.org/r/d0c55386-2f7f-a940-45bb-d80ae5e0f378@linaro.org
* 'timers/drivers/armv8.6_arch_timer' of https://git.linaro.org/people/daniel.lezcano/linux:
  clocksource/drivers/arch_arm_timer: Move workaround synchronisation around
  clocksource/drivers/arm_arch_timer: Fix masking for high freq counters
  clocksource/drivers/arm_arch_timer: Drop unnecessary ISB on CVAL programming
  clocksource/drivers/arm_arch_timer: Remove any trace of the TVAL programming interface
  clocksource/drivers/arm_arch_timer: Work around broken CVAL implementations
  clocksource/drivers/arm_arch_timer: Advertise 56bit timer to the core code
  clocksource/drivers/arm_arch_timer: Move MMIO timer programming over to CVAL
  clocksource/drivers/arm_arch_timer: Fix MMIO base address vs callback ordering issue
  clocksource/drivers/arm_arch_timer: Move drop _tval from erratum function names
  clocksource/drivers/arm_arch_timer: Move system register timer programming over to CVAL
  clocksource/drivers/arm_arch_timer: Extend write side of timer register accessors to u64
  clocksource/drivers/arm_arch_timer: Drop CNT*_TVAL read accessors
  clocksource/arm_arch_timer: Add build-time guards for unhandled register accesses

clocksource/drivers/arch_arm_timer: Move workaround synchronisation around

We currently handle synchronisation when workarounds are enabled
by having an ISB in the __arch_counter_get_cnt?ct_stable() helpers.

While this works, this prevents us from relaxing this synchronisation.

Instead, move it closer to the point where the synchronisation is
actually needed. Further patches will subsequently relax this.

Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20211017124225.3018098-14-maz@kernel.org
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>

clocksource/drivers/arm_arch_timer: Fix masking for high freq counters

Unfortunately, the architecture provides no means to determine the bit
width of the system counter. However, we do know the following from the
specification:

 - the system counter is at least 56 bits wide
 - Roll-over time of not less than 40 years

To date, the arch timer driver has depended on the first property,
assuming any system counter to be 56 bits wide and masking off the rest.
However, combining a narrow clocksource mask with a high frequency
counter could result in prematurely wrapping the system counter by a
significant margin. For example, a 56 bit wide, 1GHz system counter
would wrap in a mere 2.28 years!

This is a problem for two reasons: v8.6+ implementations are required to
provide a 64 bit, 1GHz system counter. Furthermore, before v8.6,
implementers may select a counter frequency of their choosing.

Fix the issue by deriving a valid clock mask based on the second
property from above. Set the floor at 56 bits, since we know no system
counter is narrower than that.

[maz: fixed width computation not to lose the last bit, added
      max delta generation for the timer]

Suggested-by: Marc Zyngier <maz@kernel.org>
Signed-off-by: Oliver Upton <oupton@google.com>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20210807191428.3488948-1-oupton@google.com
Link: https://lore.kernel.org/r/20211017124225.3018098-13-maz@kernel.org
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>

clocksource/drivers/arm_arch_timer: Drop unnecessary ISB on CVAL programming

Switching from TVAL to CVAL has a small drawback: we need an ISB
before reading the counter. We cannot get rid of it, but we can
instead remove the one that comes just after writing to CVAL.

This reduces the number of ISBs from 3 to 2 when programming
the timer.

Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20211017124225.3018098-12-maz@kernel.org
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>

clocksource/drivers/arm_arch_timer: Remove any trace of the TVAL programming interface

TVAL usage is now long gone, get rid of the leftovers.

Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20211017124225.3018098-11-maz@kernel.org
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>

clocksource/drivers/arm_arch_timer: Work around broken CVAL implementations

The Applied Micro XGene-1 SoC has a busted implementation of the
CVAL register: it looks like it is based on TVAL instead of the
other way around. The net effect of this implementation blunder
is that the maximum deadline you can program in the timer is
32bit wide.

Use a MIDR check to notice the broken CPU, and reduce the width
of the timer to 32bit.

Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20211017124225.3018098-10-maz@kernel.org
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>

clocksource/drivers/arm_arch_timer: Advertise 56bit timer to the core code

Proudly tell the code code that we have a timer able to handle
56 bits deltas.

Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20211017124225.3018098-9-maz@kernel.org
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>

clocksource/drivers/arm_arch_timer: Move MMIO timer programming over to CVAL

Similarily to the sysreg-based timer, move the MMIO over to using
the CVAL registers instead of TVAL. Note that there is no warranty
that the 64bit MMIO access will be atomic, but the timer is always
disabled at the point where we program CVAL.

Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20211017124225.3018098-8-maz@kernel.org
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>

clocksource/drivers/arm_arch_timer: Fix MMIO base address vs callback ordering issue

The MMIO timer base address gets published after we have registered
the callbacks and the interrupt handler, which is... a bit dangerous.

Fix this by moving the base address publication to the point where
we register the timer, and expose a pointer to the timer structure
itself rather than a naked value.

Reviewed-by: Oliver Upton <oupton@google.com>
Reviewed-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20211017124225.3018098-7-maz@kernel.org
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>

clocksource/drivers/arm_arch_timer: Move drop _tval from erratum function names

The '_tval' name in the erratum handling function names doesn't
make much sense anymore (and they were using CVAL the first place).

Drop the _tval tag.

Reviewed-by: Oliver Upton <oupton@google.com>
Reviewed-by: Mark Rutland <mark.rutland@arm.com>
Tested-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20211017124225.3018098-6-maz@kernel.org
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>

clocksource/drivers/arm_arch_timer: Move system register timer programming over to CVAL

In order to cope better with high frequency counters, move the
programming of the timers from the countdown timer (TVAL) over
to the comparator (CVAL).

The programming model is slightly different, as we now need to
read the current counter value to have an absolute deadline
instead of a relative one.

There is a small overhead to this change, which we will address
in the following patches.

Reviewed-by: Oliver Upton <oupton@google.com>
Reviewed-by: Mark Rutland <mark.rutland@arm.com>
Tested-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20211017124225.3018098-5-maz@kernel.org
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>

clocksource/drivers/arm_arch_timer: Extend write side of timer register accessors to u64

The various accessors for the timer sysreg and MMIO registers are
currently hardwired to 32bit. However, we are about to introduce
the use of the CVAL registers, which require a 64bit access.

Upgrade the write side of the accessors to take a 64bit value
(the read side is left untouched as we don't plan to ever read
back any of these registers).

No functional change expected.

Reviewed-by: Oliver Upton <oupton@google.com>
Reviewed-by: Mark Rutland <mark.rutland@arm.com>
Tested-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20211017124225.3018098-4-maz@kernel.org
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>

clocksource/drivers/arm_arch_timer: Drop CNT*_TVAL read accessors

The arch timer driver never reads the various TVAL registers, only
writes to them. It is thus pointless to provide accessors
for them and to implement errata workarounds.

Drop these read-side accessors, and add a couple of BUG() statements
for the time being. These statements will be removed further down
the line.

Reviewed-by: Oliver Upton <oupton@google.com>
Reviewed-by: Mark Rutland <mark.rutland@arm.com>
Tested-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20211017124225.3018098-3-maz@kernel.org
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>

clocksource/arm_arch_timer: Add build-time guards for unhandled register accesses

As we are about to change the registers that are used by the driver,
start by adding build-time checks to ensure that we always handle
all registers and access modes.

Suggested-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20211017124225.3018098-2-maz@kernel.org
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>

acpi/arm64: fix next_platform_timer() section mismatch error

Fix modpost Section mismatch error in next_platform_timer().

  [...]
  WARNING: modpost: vmlinux.o(.text.unlikely+0x26e60): Section mismatch in reference from the function next_platform_timer() to the variable .init.data:acpi_gtdt_desc
  The function next_platform_timer() references
  the variable __initdata acpi_gtdt_desc.
  This is often because next_platform_timer lacks a __initdata
  annotation or the annotation of acpi_gtdt_desc is wrong.

  WARNING: modpost: vmlinux.o(.text.unlikely+0x26e64): Section mismatch in reference from the function next_platform_timer() to the variable .init.data:acpi_gtdt_desc
  The function next_platform_timer() references
  the variable __initdata acpi_gtdt_desc.
  This is often because next_platform_timer lacks a __initdata
  annotation or the annotation of acpi_gtdt_desc is wrong.

  ERROR: modpost: Section mismatches detected.
  Set CONFIG_SECTION_MISMATCH_WARN_ONLY=y to allow them.
  make[1]: *** [scripts/Makefile.modpost:59: vmlinux.symvers] Error 1
  make[1]: *** Deleting file 'vmlinux.symvers'
  make: *** [Makefile:1176: vmlinux] Error 2
  [...]

Fixes: a712c3ed9b8a ("acpi/arm64: Add memory-mapped timer support in GTDT driver")
Signed-off-by: Jackie Liu <liuyun01@kylinos.cn>
Acked-by: Hanjun Guo <guohanjun@huawei.com>
Link: https://lore.kernel.org/r/20210823092526.2407526-1-liu.yun@linux.dev
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>

arm64: ftrace: use function_nocfi for _mcount as well

Commit 800618f955a9 ("arm64: ftrace: use function_nocfi for ftrace_call")
only fixed address of ftrace_call but address of _mcount needs to be
fixed as well. Use function_nocfi() to get the actual address of _mcount
function as with CONFIG_CFI_CLANG, the compiler replaces function pointers
with jump table addresses which breaks dynamic ftrace as the address of
_mcount is replaced with the address of _mcount.cfi_jt.

With mainline, this won't be a problem since by default
CONFIG_DYNAMIC_FTRACE_WITH_REGS=y with Clang >= 10 as it supports
-fpatchable-function-entry and CFI requires Clang 12 but for consistency
we should add function_nocfi() for _mcount as well.

Signed-off-by: Sumit Garg <sumit.garg@linaro.org>
Acked-by: Mark Rutland <mark.rutland@arm.com>
Reviewed-by: Sami Tolvanen <samitolvanen@google.com>
Link: https://lore.kernel.org/r/20211011125059.3378646-1-sumit.garg@linaro.org
Signed-off-by: Will Deacon <will@kernel.org>

arm64: asm: setup.h: export common variables

When building the kernel with sparse enabled 'C=1' the following
warnings can be seen:

arch/arm64/kernel/setup.c:58:13: warning: symbol '__fdt_pointer' was not declared. Should it be static?
arch/arm64/kernel/setup.c:84:25: warning: symbol 'boot_args' was not declared. Should it be static?

Rework so the variables are exported, since these two variable are
created and used in setup.c, also used in head.S.

Signed-off-by: Anders Roxell <anders.roxell@linaro.org>
Link: https://lore.kernel.org/r/20211007195601.677474-1-anders.roxell@linaro.org
Signed-off-by: Will Deacon <will@kernel.org>

arm64/hugetlb: fix CMA gigantic page order for non-4K PAGE_SIZE

For non-4K PAGE_SIZE configs, the largest gigantic huge page size is
CONT_PMD_SHIFT order. On arm64 with 64K PAGE_SIZE, the gigantic page is
16G. Therefore, one should be able to specify 'hugetlb_cma=16G' on the
kernel command line so that one gigantic page can be allocated from CMA.
However, when adding such an option the following message is produced:

hugetlb_cma: cma area should be at least 8796093022208 MiB

This is because the calculation for non-4K gigantic page order is
incorrect in the arm64 specific routine arm64_hugetlb_cma_reserve().

Fixes: abb7962adc80 ("arm64/hugetlb: Reserve CMA areas for gigantic pages on 16K and 64K configs")
Cc: <stable@vger.kernel.org> # 5.9.x
Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: Anshuman Khandual <anshuman.khandual@arm.com>
Link: https://lore.kernel.org/r/20211005202529.213812-1-mike.kravetz@oracle.com
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>

kasan: Extend KASAN mode kernel parameter

Architectures supported by KASAN_HW_TAGS can provide an asymmetric mode
of execution. On an MTE enabled arm64 hw for example this can be
identified with the asymmetric tagging mode of execution. In particular,
when such a mode is present, the CPU triggers a fault on a tag mismatch
during a load operation and asynchronously updates a register when a tag
mismatch is detected during a store operation.

Extend the KASAN HW execution mode kernel command line parameter to
support asymmetric mode.

Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Andrey Konovalov <andreyknvl@gmail.com>
Signed-off-by: Vincenzo Frascino <vincenzo.frascino@arm.com>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Reviewed-by: Andrey Konovalov <andreyknvl@gmail.com>
Link: https://lore.kernel.org/r/20211006154751.4463-6-vincenzo.frascino@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: mte: Add asymmetric mode support

MTE provides an asymmetric mode for detecting tag exceptions. In
particular, when such a mode is present, the CPU triggers a fault
on a tag mismatch during a load operation and asynchronously updates
a register when a tag mismatch is detected during a store operation.

Add support for MTE asymmetric mode.

Note: If the CPU does not support MTE asymmetric mode the kernel falls
back on synchronous mode which is the default for kasan=on.

Cc: Will Deacon <will@kernel.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Andrey Konovalov <andreyknvl@gmail.com>
Signed-off-by: Vincenzo Frascino <vincenzo.frascino@arm.com>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Acked-by: Andrey Konovalov <andreyknvl@gmail.com>
Link: https://lore.kernel.org/r/20211006154751.4463-5-vincenzo.frascino@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: mte: CPU feature detection for Asymm MTE

Add the cpufeature entries to detect the presence of Asymmetric MTE.

Note: The tag checking mode is initialized via cpu_enable_mte() ->
kasan_init_hw_tags() hence to enable it we require asymmetric mode
to be at least on the boot CPU. If the boot CPU does not have it, it is
fine for late CPUs to have it as long as the feature is not enabled
(ARM64_CPUCAP_BOOT_CPU_FEATURE).

Cc: Will Deacon <will@kernel.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Suzuki K Poulose <Suzuki.Poulose@arm.com>
Signed-off-by: Vincenzo Frascino <vincenzo.frascino@arm.com>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Reviewed-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Link: https://lore.kernel.org/r/20211006154751.4463-4-vincenzo.frascino@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: mte: Bitfield definitions for Asymm MTE

Add Asymmetric Memory Tagging Extension bitfield definitions.

Cc: Will Deacon <will@kernel.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Vincenzo Frascino <vincenzo.frascino@arm.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/20211006154751.4463-3-vincenzo.frascino@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

kasan: Remove duplicate of kasan_flag_async

After merging async mode for KASAN_HW_TAGS a duplicate of the
kasan_flag_async flag was left erroneously inside the code.

Remove the duplicate.

Note: This change does not bring functional changes to the code
base.

Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Marco Elver <elver@google.com>
Cc: Evgenii Stepanov <eugenis@google.com>
Cc: Andrey Konovalov <andreyknvl@gmail.com>
Signed-off-by: Vincenzo Frascino <vincenzo.frascino@arm.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Reviewed-by: Andrey Konovalov <andreyknvl@gmail.com>
Link: https://lore.kernel.org/r/20211006154751.4463-2-vincenzo.frascino@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

selftests: arm64: Add coverage of ptrace flags for SVE VL inheritance

Add a test that covers enabling and disabling of SVE vector length
inheritance via the ptrace interface.

Signed-off-by: Mark Brown <broonie@kernel.org>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/20211005123537.976795-1-broonie@kernel.org
Signed-off-by: Will Deacon <will@kernel.org>

drivers/perf: Improve build test coverage

Improve build test cover by allowing some drivers to build under
COMPILE_TEST where possible.

Some notes:
- Mostly a dependency on CONFIG_ACPI is not really required for only
  building (but left untouched), but is required for TX2 which uses ACPI
  functions which have no stubs
- XGENE required 64b dependency as it relies on some unsigned long perf
  struct fields being 64b
- I don't see why TX2 requires NUMA to build, but left untouched
- Added an explicit dependency on GENERIC_MSI_IRQ_DOMAIN for
  ARM_SMMU_V3_PMU, which is required for platform MSI functions

Signed-off-by: John Garry <john.garry@huawei.com>
Link: https://lore.kernel.org/r/1633085326-156653-3-git-send-email-john.garry@huawei.com
Signed-off-by: Will Deacon <will@kernel.org>

drivers/perf: thunderx2_pmu: Change data in size tx2_uncore_event_update()

A LSL of 32 requires > 32b value to hold the result. However in
tx2_uncore_event_update(), 1UL << 32 currently only works as unsigned
long is 64b on a 64b system.

If we want to compile test for a 32b system, we need unsigned long long,
whose min size is 64b.

Signed-off-by: John Garry <john.garry@huawei.com>
Link: https://lore.kernel.org/r/1633085326-156653-2-git-send-email-john.garry@huawei.com
Signed-off-by: Will Deacon <will@kernel.org>

drivers/perf: hisi: Fix PA PMU counter offset

The PA PMU counter offset was correct in [1] and the driver has
already been verified. We want to keep the register offset using
lower case character in later version that is consistent with
the existed driver. Since there was no functional change, we
didn't do more test. However there is typo when modified the PA
PMU counter offset by mistake, so fix this bad mistake.

[1] https://www.spinics.net/lists/arm-kernel/msg865263.html

Cc: Will Deacon <will@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: John Garry <john.garry@huawei.com>
Cc: Qi Liu <liuqi115@huawei.com>
Signed-off-by: Shaokun Zhang <zhangshaokun@hisilicon.com>
Link: https://lore.kernel.org/r/20210928123022.23467-1-zhangshaokun@hisilicon.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64/mm: drop HAVE_ARCH_PFN_VALID

CONFIG_SPARSEMEM_VMEMMAP is now the only available memory model on arm64
platforms and free_unused_memmap() would just return without creating any
holes in the memmap mapping.  There is no need for any special handling in
pfn_valid() and HAVE_ARCH_PFN_VALID can just be dropped.  This also moves
the pfn upper bits sanity check into generic pfn_valid().

[rppt: rebased on v5.15-rc3]

Link: https://lkml.kernel.org/r/1621947349-25421-1-git-send-email-anshuman.khandual@arm.com
Signed-off-by: Anshuman Khandual <anshuman.khandual@arm.com>
Acked-by: David Hildenbrand <david@redhat.com>
Acked-by: Mike Rapoport <rppt@linux.ibm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will@kernel.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: Mike Rapoport <rppt@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Mike Rapoport <rppt@linux.ibm.com>
Link: https://lore.kernel.org/r/20210930013039.11260-3-rppt@kernel.org
Signed-off-by: Will Deacon <will@kernel.org>

dma-mapping: remove bogus test for pfn_valid from dma_map_resource

dma_map_resource() uses pfn_valid() to ensure the range is not RAM.
However, pfn_valid() only checks for availability of the memory map for a
PFN but it does not ensure that the PFN is actually backed by RAM.

As dma_map_resource() is the only method in DMA mapping APIs that has this
check, simply drop the pfn_valid() test from dma_map_resource().

Link: https://lore.kernel.org/all/20210824173741.GC623@arm.com/
Signed-off-by: Mike Rapoport <rppt@linux.ibm.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Acked-by: David Hildenbrand <david@redhat.com>
Link: https://lore.kernel.org/r/20210930013039.11260-2-rppt@kernel.org
Signed-off-by: Will Deacon <will@kernel.org>

arm64: trans_pgd: remove trans_pgd_map_page()

The intend of trans_pgd_map_page() was to map contiguous range of VA
memory to the memory that is getting relocated during kexec. However,
since we are now using linear map instead of contiguous range this
function is not needed

Suggested-by: Pingfan Liu <kernelfans@gmail.com>
Signed-off-by: Pasha Tatashin <pasha.tatashin@soleen.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/20210930143113.1502553-16-pasha.tatashin@soleen.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: kexec: remove cpu-reset.h

This header contains only cpu_soft_restart() which is never used directly
anymore. So, remove this header, and rename the helper to be
cpu_soft_restart().

Suggested-by: James Morse <james.morse@arm.com>
Signed-off-by: Pasha Tatashin <pasha.tatashin@soleen.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/20210930143113.1502553-15-pasha.tatashin@soleen.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: kexec: remove the pre-kexec PoC maintenance

Now that kexec does its relocations with the MMU enabled, we no longer
need to clean the relocation data to the PoC.

Suggested-by: James Morse <james.morse@arm.com>
Signed-off-by: Pasha Tatashin <pasha.tatashin@soleen.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/20210930143113.1502553-14-pasha.tatashin@soleen.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: kexec: keep MMU enabled during kexec relocation

Now, that we have linear map page tables configured, keep MMU enabled
to allow faster relocation of segments to final destination.

Cavium ThunderX2:
Kernel Image size: 38M Iniramfs size: 46M Total relocation size: 84M
MMU-disabled:
relocation 7.489539915s
MMU-enabled:
relocation 0.03946095s

Broadcom Stingray:
The performance data: for a moderate size kernel + initramfs: 25M the
relocation was taking 0.382s, with enabled MMU it now takes
0.019s only or x20 improvement.

The time is proportional to the size of relocation, therefore if initramfs
is larger, 100M it could take over a second.

Signed-off-by: Pasha Tatashin <pasha.tatashin@soleen.com>
Tested-by: Pingfan Liu <piliu@redhat.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/20210930143113.1502553-13-pasha.tatashin@soleen.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: kexec: install a copy of the linear-map

To perform the kexec relocation with the MMU enabled, we need a copy
of the linear map.

Create one, and install it from the relocation code. This has to be done
from the assembly code as it will be idmapped with TTBR0. The kernel
runs in TTRB1, so can't use the break-before-make sequence on the mapping
it is executing from.

The makes no difference yet as the relocation code runs with the MMU
disabled.

Suggested-by: James Morse <james.morse@arm.com>
Signed-off-by: Pasha Tatashin <pasha.tatashin@soleen.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/20210930143113.1502553-12-pasha.tatashin@soleen.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: kexec: use ld script for relocation function

Currently, relocation code declares start and end variables
which are used to compute its size.

The better way to do this is to use ld script, and put relocation
function in its own section.

Signed-off-by: Pasha Tatashin <pasha.tatashin@soleen.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/20210930143113.1502553-11-pasha.tatashin@soleen.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: kexec: relocate in EL1 mode

Since we are going to keep MMU enabled during relocation, we need to
keep EL1 mode throughout the relocation.

Keep EL1 enabled, and switch EL2 only before entering the new world.

Suggested-by: James Morse <james.morse@arm.com>
Signed-off-by: Pasha Tatashin <pasha.tatashin@soleen.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/20210930143113.1502553-10-pasha.tatashin@soleen.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: kexec: configure EL2 vectors for kexec

If we have a EL2 mode without VHE, the EL2 vectors are needed in order
to switch to EL2 and jump to new world with hypervisor privileges.

In preparation to MMU enabled relocation, configure our EL2 table now.

Kexec uses #HVC_SOFT_RESTART to branch to the new world, so extend
el1_sync vector that is provided by trans_pgd_copy_el2_vectors() to
support this case.

Signed-off-by: Pasha Tatashin <pasha.tatashin@soleen.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/20210930143113.1502553-9-pasha.tatashin@soleen.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: kexec: pass kimage as the only argument to relocation function

Currently, kexec relocation function (arm64_relocate_new_kernel) accepts
the following arguments:

head: start of array that contains relocation information.
entry: entry point for new kernel or purgatory.
dtb_mem: first and only argument to entry.

The number of arguments cannot be easily expended, because this
function is also called from HVC_SOFT_RESTART, which preserves only
three arguments. And, also arm64_relocate_new_kernel is written in
assembly but called without stack, thus no place to move extra arguments
to free registers.

Soon, we will need to pass more arguments: once we enable MMU we
will need to pass information about page tables.

Pass kimage to arm64_relocate_new_kernel, and teach it to get the
required fields from kimage.

Signed-off-by: Pasha Tatashin <pasha.tatashin@soleen.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/20210930143113.1502553-8-pasha.tatashin@soleen.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: kexec: Use dcache ops macros instead of open-coding

kexec does dcache maintenance when it re-writes all memory. Our
dcache_by_line_op macro depends on reading the sanitized DminLine
from memory. Kexec may have overwritten this, so open-codes the
sequence.

dcache_by_line_op is a whole set of macros, it uses dcache_line_size
which uses read_ctr for the sanitsed DminLine. Reading the DminLine
is the first thing the dcache_by_line_op does.

Rename dcache_by_line_op dcache_by_myline_op and take DminLine as
an argument. Kexec can now use the slightly smaller macro.

This makes up-coming changes to the dcache maintenance easier on
the eye.

Code generated by the existing callers is unchanged.

Suggested-by: James Morse <james.morse@arm.com>
Signed-off-by: Pasha Tatashin <pasha.tatashin@soleen.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/20210930143113.1502553-7-pasha.tatashin@soleen.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: kexec: skip relocation code for inplace kexec

In case of kdump or when segments are already in place the relocation
is not needed, therefore the setup of relocation function and call to
it can be skipped.

Signed-off-by: Pasha Tatashin <pasha.tatashin@soleen.com>
Suggested-by: James Morse <james.morse@arm.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/20210930143113.1502553-6-pasha.tatashin@soleen.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: kexec: flush image and lists during kexec load time

Currently, during kexec load we are copying relocation function and
flushing it. However, we can also flush kexec relocation buffers and
if new kernel image is already in place (i.e. crash kernel), we can
also flush the new kernel image itself.

Signed-off-by: Pasha Tatashin <pasha.tatashin@soleen.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/20210930143113.1502553-5-pasha.tatashin@soleen.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: hibernate: abstract ttrb0 setup function

Currently, only hibernate sets custom ttbr0 with safe idmaped function.
Kexec, is also going to be using this functionality when relocation code
is going to be idmapped.

Move the setup sequence to a dedicated cpu_install_ttbr0() for custom
ttbr0.

Suggested-by: James Morse <james.morse@arm.com>
Signed-off-by: Pasha Tatashin <pasha.tatashin@soleen.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/20210930143113.1502553-4-pasha.tatashin@soleen.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: trans_pgd: hibernate: Add trans_pgd_copy_el2_vectors

Users of trans_pgd may also need a copy of vector table because it is
also may be overwritten if a linear map can be overwritten.

Move setup of EL2 vectors from hibernate to trans_pgd, so it can be
later shared with kexec as well.

Signed-off-by: Pasha Tatashin <pasha.tatashin@soleen.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/20210930143113.1502553-3-pasha.tatashin@soleen.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: kernel: add helper for booted at EL2 and not VHE

Replace places that contain logic like this:
is_hyp_mode_available() && !is_kernel_in_hyp_mode()

With a dedicated boolean function  is_hyp_nvhe(). This will be needed
later in kexec in order to sooner switch back to EL2.

Suggested-by: James Morse <james.morse@arm.com>
Signed-off-by: Pasha Tatashin <pasha.tatashin@soleen.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/20210930143113.1502553-2-pasha.tatashin@soleen.com
Signed-off-by: Will Deacon <will@kernel.org>

scs: Release kasan vmalloc poison in scs_free process

Since scs allocation is moved to vmalloc region, the
shadow stack is protected by kasan_posion_vmalloc.
However, the vfree_atomic operation needs to access
its context for scs_free process and causes kasan error
as the dump info below.

This patch Adds kasan_unpoison_vmalloc() before vfree_atomic,
which aligns to the prior flow as using kmem_cache.
The vmalloc region will go back posioned in the following
vumap() operations.

 ==================================================================
 BUG: KASAN: vmalloc-out-of-bounds in llist_add_batch+0x60/0xd4
 Write of size 8 at addr ffff8000100b9000 by task kthreadd/2

 CPU: 0 PID: 2 Comm: kthreadd Not tainted 5.15.0-rc2-11681-g92477dd1faa6-dirty #1
 Hardware name: linux,dummy-virt (DT)
 Call trace:
  dump_backtrace+0x0/0x43c
  show_stack+0x1c/0x2c
  dump_stack_lvl+0x68/0x84
  print_address_description+0x80/0x394
  kasan_report+0x180/0x1dc
  __asan_report_store8_noabort+0x48/0x58
  llist_add_batch+0x60/0xd4
  vfree_atomic+0x60/0xe0
  scs_free+0x1dc/0x1fc
  scs_release+0xa4/0xd4
  free_task+0x30/0xe4
  __put_task_struct+0x1ec/0x2e0
  delayed_put_task_struct+0x5c/0xa0
  rcu_do_batch+0x62c/0x8a0
  rcu_core+0x60c/0xc14
  rcu_core_si+0x14/0x24
  __do_softirq+0x19c/0x68c
  irq_exit+0x118/0x2dc
  handle_domain_irq+0xcc/0x134
  gic_handle_irq+0x7c/0x1bc
  call_on_irq_stack+0x40/0x70
  do_interrupt_handler+0x78/0x9c
  el1_interrupt+0x34/0x60
  el1h_64_irq_handler+0x1c/0x2c
  el1h_64_irq+0x78/0x7c
  _raw_spin_unlock_irqrestore+0x40/0xcc
  sched_fork+0x4f0/0xb00
  copy_process+0xacc/0x3648
  kernel_clone+0x168/0x534
  kernel_thread+0x13c/0x1b0
  kthreadd+0x2bc/0x400
  ret_from_fork+0x10/0x20

 Memory state around the buggy address:
  ffff8000100b8f00: f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8
  ffff8000100b8f80: f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8
 >ffff8000100b9000: f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8
                    ^
  ffff8000100b9080: f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8
  ffff8000100b9100: f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8
 ==================================================================

Suggested-by: Kuan-Ying Lee <kuan-ying.lee@mediatek.com>
Acked-by: Will Deacon <will@kernel.org>
Tested-by: Will Deacon <will@kernel.org>
Reviewed-by: Sami Tolvanen <samitolvanen@google.com>
Signed-off-by: Yee Lee <yee.lee@mediatek.com>
Fixes: a2abe7cbd8fe ("scs: switch to vmapped shadow stacks")
Link: https://lore.kernel.org/r/20210930081619.30091-1-yee.lee@mediatek.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: kasan: mte: move GCR_EL1 switch to task switch when KASAN disabled

It is not necessary to write to GCR_EL1 on every kernel entry and
exit when HW tag-based KASAN is disabled because the kernel will not
execute any IRG instructions in that mode. Since accessing GCR_EL1
can be expensive on some microarchitectures, avoid doing so by moving
the access to task switch when HW tag-based KASAN is disabled.

Signed-off-by: Peter Collingbourne <pcc@google.com>
Acked-by: Andrey Konovalov <andreyknvl@gmail.com>
Link: https://linux-review.googlesource.com/id/I78e90d60612a94c24344526f476ac4ff216e10d2
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/20210924010655.2886918-1-pcc@google.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: mm: update max_pfn after memory hotplug

After new memory blocks have been hotplugged, max_pfn and max_low_pfn
needs updating to reflect on new PFNs being hot added to system.
Without this patch, debug-related functions that use max_pfn such as
get_max_dump_pfn() or read_page_owner() will not work with any page in
memory that is hot-added after boot.

Fixes: 4ab215061554 ("arm64: Add memory hotplug support")
Signed-off-by: Sudarshan Rajagopalan <quic_sudaraja@quicinc.com>
Signed-off-by: Chris Goldsworthy <quic_cgoldswo@quicinc.com>
Acked-by: David Hildenbrand <david@redhat.com>
Cc: Florian Fainelli <f.fainelli@gmail.com>
Cc: Georgi Djakov <quic_c_gdjako@quicinc.com>
Tested-by: Georgi Djakov <quic_c_gdjako@quicinc.com>
Link: https://lore.kernel.org/r/a51a27ee7be66024b5ce626310d673f24107bcb8.1632853776.git.quic_cgoldswo@quicinc.com
Signed-off-by: Will Deacon <will@kernel.org>