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

* for-next/perf: (25 commits)
  perf/marvell: Fix !CONFIG_OF build for CN10K DDR PMU driver
  drivers/perf: Add Apple icestorm/firestorm CPU PMU driver
  drivers/perf: arm_pmu: Handle 47 bit counters
  arm64: perf: Consistently make all event numbers as 16-bits
  arm64: perf: Expose some Armv9 common events under sysfs
  perf/marvell: cn10k DDR perf event core ownership
  perf/marvell: cn10k DDR perfmon event overflow handling
  perf/marvell: CN10k DDR performance monitor support
  dt-bindings: perf: marvell: cn10k ddr performance monitor
  perf/arm-cmn: Update watchpoint format
  perf/arm-cmn: Hide XP PUB events for CMN-600
  perf: replace bitmap_weight with bitmap_empty where appropriate
  perf: Replace acpi_bus_get_device()
  perf/marvell_cn10k: Fix unused variable warning when W=1 and CONFIG_OF=n
  perf/arm-cmn: Make arm_cmn_debugfs static
  perf: MARVELL_CN10K_TAD_PMU should depend on ARCH_THUNDER
  perf/arm-ccn: Use platform_get_irq() to get the interrupt
  irqchip/apple-aic: Move PMU-specific registers to their own include file
  arm64: dts: apple: Add t8303 PMU nodes
  arm64: dts: apple: Add t8103 PMU interrupt affinities
  ...

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

* for-next/pauth:
  arm64: Add support of PAuth QARMA3 architected algorithm
  arm64: cpufeature: Mark existing PAuth architected algorithm as QARMA5
  arm64: cpufeature: Account min_field_value when cheking secondaries for PAuth

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

* for-next/mte:
  docs: sysfs-devices-system-cpu: document "asymm" value for mte_tcf_preferred
  arm64/mte: Remove asymmetric mode from the prctl() interface
  kasan: fix a missing header include of static_keys.h
  arm64/mte: Add userspace interface for enabling asymmetric mode
  arm64/mte: Add hwcap for asymmetric mode
  arm64/mte: Add a little bit of documentation for mte_update_sctlr_user()
  arm64/mte: Document ABI for asymmetric mode
  arm64: mte: avoid clearing PSTATE.TCO on entry unless necessary
  kasan: split kasan_*enabled() functions into a separate header

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

* for-next/mm:
  Documentation: vmcoreinfo: Fix htmldocs warning
  arm64/mm: Drop use_1G_block()
  arm64: avoid flushing icache multiple times on contiguous HugeTLB
  arm64: crash_core: Export MODULES, VMALLOC, and VMEMMAP ranges
  arm64/hugetlb: Define __hugetlb_valid_size()
  arm64/mm: avoid fixmap race condition when create pud mapping
  arm64/mm: Consolidate TCR_EL1 fields

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

* for-next/misc:
  arm64: mm: Drop 'const' from conditional arm64_dma_phys_limit definition
  arm64: clean up tools Makefile
  arm64: drop unused includes of <linux/personality.h>
  arm64: Do not defer reserve_crashkernel() for platforms with no DMA memory zones
  arm64: prevent instrumentation of bp hardening callbacks
  arm64: cpufeature: Remove cpu_has_fwb() check
  arm64: atomics: remove redundant static branch
  arm64: entry: Save some nops when CONFIG_ARM64_PSEUDO_NMI is not set

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

* for-next/linkage:
  arm64: module: remove (NOLOAD) from linker script
  linkage: remove SYM_FUNC_{START,END}_ALIAS()
  x86: clean up symbol aliasing
  arm64: clean up symbol aliasing
  linkage: add SYM_FUNC_ALIAS{,_LOCAL,_WEAK}()

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

* for-next/kselftest:
  kselftest/arm64: Log the PIDs of the parent and child in sve-ptrace
  kselftest/arm64: signal: Allow tests to be incompatible with features
  kselftest/arm64: mte: user_mem: test a wider range of values
  kselftest/arm64: mte: user_mem: add more test types
  kselftest/arm64: mte: user_mem: add test type enum
  kselftest/arm64: mte: user_mem: check different offsets and sizes
  kselftest/arm64: mte: user_mem: rework error handling
  kselftest/arm64: mte: user_mem: introduce tag_offset and tag_len
  kselftest/arm64: Remove local definitions of MTE prctls
  kselftest/arm64: Remove local ARRAY_SIZE() definitions

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

* for-next/insn:
  arm64: insn: add encoders for atomic operations
  arm64: move AARCH64_BREAK_FAULT into insn-def.h
  arm64: insn: Generate 64 bit mask immediates correctly

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

* for-next/errata:
  arm64: Add cavium_erratum_23154_cpus missing sentinel
  irqchip/gic-v3: Workaround Marvell erratum 38545 when reading IAR

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

* for-next/docs:
  arm64/mte: Clarify mode reported by PR_GET_TAGGED_ADDR_CTRL
  arm64: booting.rst: Clarify on requiring non-secure EL2

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

* for-next/coredump:
  arm64: Change elfcore for_each_mte_vma() to use VMA iterator
  arm64: mte: Document the core dump file format
  arm64: mte: Dump the MTE tags in the core file
  arm64: mte: Define the number of bytes for storing the tags in a page
  elf: Introduce the ARM MTE ELF segment type
  elfcore: Replace CONFIG_{IA64, UML} checks with a new option

docs: sysfs-devices-system-cpu: document "asymm" value for mte_tcf_preferred

It was added in commit 766121ba5de3 ("arm64/mte: Add userspace interface
for enabling asymmetric mode").

Signed-off-by: Evgenii Stepanov <eugenis@google.com>
Reviewed-by: Mark Brown <broonie@kernel.org>
Link: https://lore.kernel.org/r/20220309215943.87831-1-eugenis@google.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64/mte: Remove asymmetric mode from the prctl() interface

As pointed out by Evgenii Stepanov one potential issue with the new ABI for
enabling asymmetric is that if there are multiple places where MTE is
configured in a process, some of which were compiled with the old prctl.h
and some of which were compiled with the new prctl.h, there may be problems
keeping track of which MTE modes are requested. For example some code may
disable only sync and async modes leaving asymmetric mode enabled when it
intended to fully disable MTE.

In order to avoid such mishaps remove asymmetric mode from the prctl(),
instead implicitly allowing it if both sync and async modes are requested.
This should not disrupt userspace since a process requesting both may
already see a mix of sync and async modes due to differing defaults between
CPUs or changes in default while the process is running but it does mean
that userspace is unable to explicitly request asymmetric mode without
changing the system default for CPUs.

Reported-by: Evgenii Stepanov <eugenis@google.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Reviewed-by: Evgenii Stepanov <eugenis@google.com>
Cc: Peter Collingbourne <pcc@google.com>
Cc: Joey Gouly <joey.gouly@arm.com>
Cc: Branislav Rankov <branislav.rankov@arm.com>
Link: https://lore.kernel.org/r/20220309131200.112637-1-broonie@kernel.org
Signed-off-by: Will Deacon <will@kernel.org>

arm64: Add cavium_erratum_23154_cpus missing sentinel

Qian Cai reported that playing with CPU hotplug resulted in a
out-of-bound access due to cavium_erratum_23154_cpus missing
a sentinel indicating the end of the array.

Add it in order to restore peace and harmony in the world
of broken HW.

Reported-by: Qian Cai <quic_qiancai@quicinc.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Fixes: 24a147bcef8c ("irqchip/gic-v3: Workaround Marvell erratum 38545 when reading IAR")
Link: https://lore.kernel.org/r/YijmkXp1VG7e8lDx@qian
Cc: Linu Cherian <lcherian@marvell.com>
Cc: Will Deacon <will@kernel.org>
Link: https://lore.kernel.org/r/20220309180600.3990874-1-maz@kernel.org
Signed-off-by: Will Deacon <will@kernel.org>

perf/marvell: Fix !CONFIG_OF build for CN10K DDR PMU driver

When compiling the Marvell CN10K DDR PMU driver with CONFIG_OF=n, the
build fails:

  | drivers/perf/marvell_cn10k_ddr_pmu.c:723:35: error: 'cn10k_ddr_pmu_of_match' undeclared here (not in a function); did you mean 'cn10k_ddr_pmu_driver'?

Use `of_match_ptr()` to avoid referencing the non-existent match table
in this configuration.

Link: https://lore.kernel.org/r/202203091424.Vfe8J4W9-lkp@intel.com
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Will Deacon <will@kernel.org>

arm64: mm: Drop 'const' from conditional arm64_dma_phys_limit definition

Commit 031495635b46 ("arm64: Do not defer reserve_crashkernel() for
platforms with no DMA memory zones") introduced different definitions
for 'arm64_dma_phys_limit' depending on CONFIG_ZONE_DMA{,32} based on
a late suggestion from Pasha. Sadly, this results in a build error when
passing W=1:

  | arch/arm64/mm/init.c:90:19: error: conflicting type qualifiers for 'arm64_dma_phys_limit'

Drop the 'const' for now and use '__ro_after_init' consistently.

Link: https://lore.kernel.org/r/202203090241.aj7paWeX-lkp@intel.com
Link: https://lore.kernel.org/r/CA+CK2bDbbx=8R=UthkMesWOST8eJMtOGJdfMRTFSwVmo0Vn0EA@mail.gmail.com
Fixes: 031495635b46 ("arm64: Do not defer reserve_crashkernel() for platforms with no DMA memory zones")
Signed-off-by: Will Deacon <will@kernel.org>

Documentation: vmcoreinfo: Fix htmldocs warning

Since commit 2369f171d5c5 ("arm64: crash_core: Export MODULES, VMALLOC,
and VMEMMAP ranges"), Stephen reports a warning when building htmldocs:

  | Documentation/admin-guide/kdump/vmcoreinfo.rst:498: WARNING: Title underline too short.

Extend the underline to squash the warning.

Fixes: 2369f171d5c5 ("arm64: crash_core: Export MODULES, VMALLOC, and VMEMMAP ranges")
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Will Deacon <will@kernel.org>

kasan: fix a missing header include of static_keys.h

The kasan-enabled.h header relies on static keys, so make sure
to include the header to avoid compilation errors (with JUMP_LABEL=n).

It fixes the following:
./include/linux/kasan-enabled.h:9:1: warning: data definition has no type or storage class
    9 | DECLARE_STATIC_KEY_FALSE(kasan_flag_enabled);
      | ^~~~~~~~~~~~~~~~~~~~~~~~
error: type defaults to 'int' in declaration of 'DECLARE_STATIC_KEY_FALSE' [-Werror=implicit-int]

Fixes: f9b5e46f4097eb29 ("kasan: split kasan_*enabled() functions into a separate header")
Cc: Peter Collingbourne <pcc@google.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will@kernel.org>
Acked-by: Andrey Konovalov <andreyknvl@gmail.com>
Signed-off-by: Joey Gouly <joey.gouly@arm.com>
Link: https://lore.kernel.org/r/20220301154518.19456-1-joey.gouly@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

Merge branch 'for-next/perf-m1' into for-next/perf

Support for the CPU PMUs on the Apple M1.

* for-next/perf-m1:
  drivers/perf: Add Apple icestorm/firestorm CPU PMU driver
  drivers/perf: arm_pmu: Handle 47 bit counters
  irqchip/apple-aic: Move PMU-specific registers to their own include file
  arm64: dts: apple: Add t8303 PMU nodes
  arm64: dts: apple: Add t8103 PMU interrupt affinities
  irqchip/apple-aic: Wire PMU interrupts
  irqchip/apple-aic: Parse FIQ affinities from device-tree
  dt-bindings: apple,aic: Add affinity description for per-cpu pseudo-interrupts
  dt-bindings: apple,aic: Add CPU PMU per-cpu pseudo-interrupts
  dt-bindings: arm-pmu: Document Apple PMU compatible strings

drivers/perf: Add Apple icestorm/firestorm CPU PMU driver

Add a new, weird and wonderful driver for the equally weird Apple
PMU HW. Although the PMU itself is functional, we don't know much
about the events yet, so this can be considered as yet another
random number generator...

Nonetheless, it can reliably count at least cycles and instructions
in the usually wonky big-little way. For anything else, it of course
supports raw event numbers.

Signed-off-by: Marc Zyngier <maz@kernel.org>
Signed-off-by: Will Deacon <will@kernel.org>

drivers/perf: arm_pmu: Handle 47 bit counters

The current ARM PMU framework can only deal with 32 or 64bit counters.
Teach it about a 47bit flavour.

Yes, this is odd.

Reviewed-by: Hector Martin <marcan@marcan.st>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Signed-off-by: Will Deacon <will@kernel.org>

Merge branch 'irq/aic-pmu' of git://git./linux/kernel/git/maz/arm-platforms into for-next/perf-m1

Pull in Apple AIC rework from Marc Zyngier to support PMU interrupts on
the M1 platform.

* 'irq/aic-pmu' of git://git.kernel.org/pub/scm/linux/kernel/git/maz/arm-platforms:
  irqchip/apple-aic: Move PMU-specific registers to their own include file
  arm64: dts: apple: Add t8303 PMU nodes
  arm64: dts: apple: Add t8103 PMU interrupt affinities
  irqchip/apple-aic: Wire PMU interrupts
  irqchip/apple-aic: Parse FIQ affinities from device-tree
  dt-bindings: apple,aic: Add affinity description for per-cpu pseudo-interrupts
  dt-bindings: apple,aic: Add CPU PMU per-cpu pseudo-interrupts
  dt-bindings: arm-pmu: Document Apple PMU compatible strings

arm64: perf: Consistently make all event numbers as 16-bits

Arm ARM documents PMU event numbers as 16-bits in the table and more 0x4XXX
events have been added in the header file, so use 16-bits for all event
numbers and make them consistent.

No functional change intended.

Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will@kernel.org>
Signed-off-by: Shaokun Zhang <zhangshaokun@hisilicon.com>
Link: https://lore.kernel.org/r/20220303100710.2238-1-zhangshaokun@hisilicon.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: perf: Expose some Armv9 common events under sysfs

Armv9[1] has introduced some common architectural events (0x400C-0x400F)
and common microarchitectural events (0x4010-0x401B), which can be detected
by PMCEID0_EL0 from bit44 to bit59, so expose these common events under
sysfs.

[1] https://developer.arm.com/documentation/ddi0608/ba

Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will@kernel.org>
Signed-off-by: Shaokun Zhang <zhangshaokun@hisilicon.com>
Link: https://lore.kernel.org/r/20220303085419.64085-1-zhangshaokun@hisilicon.com
Signed-off-by: Will Deacon <will@kernel.org>

perf/marvell: cn10k DDR perf event core ownership

As DDR perf event counters are not per core, so they should be accessed
only by one core at a time. Select new core when previously owning core
is going offline.

Signed-off-by: Bharat Bhushan <bbhushan2@marvell.com>
Reviewed-by: Bhaskara Budiredla <bbudiredla@marvell.com>
Link: https://lore.kernel.org/r/20220211045346.17894-5-bbhushan2@marvell.com
Signed-off-by: Will Deacon <will@kernel.org>

perf/marvell: cn10k DDR perfmon event overflow handling

CN10k DSS h/w perfmon does not support event overflow interrupt, so
periodic timer is being used. Each event counter is 48bit, which in worst
case scenario can increment at maximum 5.6 GT/s. At this rate it may take
many hours to overflow these counters. Therefore polling period for
overflow is set to 100 sec, which can be changed using sysfs parameter.

Two fixed event counters starts counting from zero on overflow, so
overflow condition is when new count less than previous count. While
eight programmable event counters freezes at maximum value. Also individual
counter cannot be restarted, so need to restart all eight counters.

Signed-off-by: Bharat Bhushan <bbhushan2@marvell.com>
Reviewed-by: Bhaskara Budiredla <bbudiredla@marvell.com>
Link: https://lore.kernel.org/r/20220211045346.17894-4-bbhushan2@marvell.com
Signed-off-by: Will Deacon <will@kernel.org>

perf/marvell: CN10k DDR performance monitor support

Marvell CN10k DRAM Subsystem (DSS) supports eight event counters for
monitoring performance and software can program each counter to monitor
any of the defined performance event. Performance events are for
interface between the DDR controller and the PHY, interface between the
DDR Controller and the CHI interconnect, or within the DDR Controller.
Additionally DSS also supports two fixed performance event counters, one
for number of ddr reads and other for ddr writes.

This patch add basic support for these performance monitoring events
on CN10k.

Signed-off-by: Bharat Bhushan <bbhushan2@marvell.com>
Reviewed-by: Bhaskara Budiredla <bbudiredla@marvell.com>
Link: https://lore.kernel.org/r/20220211045346.17894-3-bbhushan2@marvell.com
Signed-off-by: Will Deacon <will@kernel.org>

dt-bindings: perf: marvell: cn10k ddr performance monitor

Add binding documentation for the Marvell CN10k DDR
performance monitor unit.

Signed-off-by: Bharat Bhushan <bbhushan2@marvell.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Link: https://lore.kernel.org/r/20220211045346.17894-2-bbhushan2@marvell.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: clean up tools Makefile

Remove unused gen-y.

Remove redundant $(shell ...) because 'mkdir' is done in cmd_gen_cpucaps.

Replace $(filter-out $(PHONY), $^) with the $(real-prereqs) shorthand.

The '&&' in cmd_gen_cpucaps should be replaced with ';' because it is
run under 'set -e' environment.

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Link: https://lore.kernel.org/r/20220227085232.206529-1-masahiroy@kernel.org
Signed-off-by: Will Deacon <will@kernel.org>

perf/arm-cmn: Update watchpoint format

From CMN-650 onwards, some of the fields in the watchpoint config
registers moved subtly enough to easily overlook. Watchpoint events are
still only partially supported on newer IPs - which in itself deserves
noting - but were not intended to become any *less* functional than on
CMN-600.

Fixes: 60d1504070c2 ("perf/arm-cmn: Support new IP features")
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Link: https://lore.kernel.org/r/e1ce4c2f1e4f73ab1c60c3a85e4037cd62dd6352.1645727871.git.robin.murphy@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

perf/arm-cmn: Hide XP PUB events for CMN-600

CMN-600 doesn't have XP events for the PUB channel, but we missed
the appropriate check to avoid exposing them.

Fixes: 60d1504070c2 ("perf/arm-cmn: Support new IP features")
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Link: https://lore.kernel.org/r/4c108d39a0513def63acccf09ab52b328f242aeb.1645727871.git.robin.murphy@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: drop unused includes of <linux/personality.h>

Drop several includes of <linux/personality.h> which are not used.
git-blame indicates they were used at some point, but they're not needed
anymore.

Signed-off-by: Sagar Patel <sagarmp@cs.unc.edu>
Link: https://lore.kernel.org/r/20220307222412.146506-1-sagarmp@cs.unc.edu
Signed-off-by: Will Deacon <will@kernel.org>

arm64: Do not defer reserve_crashkernel() for platforms with no DMA memory zones

The following patches resulted in deferring crash kernel reservation to
mem_init(), mainly aimed at platforms with DMA memory zones (no IOMMU),
in particular Raspberry Pi 4.

commit 1a8e1cef7603 ("arm64: use both ZONE_DMA and ZONE_DMA32")
commit 8424ecdde7df ("arm64: mm: Set ZONE_DMA size based on devicetree's dma-ranges")
commit 0a30c53573b0 ("arm64: mm: Move reserve_crashkernel() into mem_init()")
commit 2687275a5843 ("arm64: Force NO_BLOCK_MAPPINGS if crashkernel reservation is required")

Above changes introduced boot slowdown due to linear map creation for
all the memory banks with NO_BLOCK_MAPPINGS, see discussion[1].  The proposed
changes restore crash kernel reservation to earlier behavior thus avoids
slow boot, particularly for platforms with IOMMU (no DMA memory zones).

Tested changes to confirm no ~150ms boot slowdown on our SoC with IOMMU
and 8GB memory.  Also tested with ZONE_DMA and/or ZONE_DMA32 configs to confirm
no regression to deferring scheme of crash kernel memory reservation.
In both cases successfully collected kernel crash dump.

[1] https://lore.kernel.org/all/9436d033-579b-55fa-9b00-6f4b661c2dd7@linux.microsoft.com/

Signed-off-by: Vijay Balakrishna <vijayb@linux.microsoft.com>
Cc: stable@vger.kernel.org
Reviewed-by: Pasha Tatashin <pasha.tatashin@soleen.com>
Link: https://lore.kernel.org/r/1646242689-20744-1-git-send-email-vijayb@linux.microsoft.com
[will: Add #ifdef CONFIG_KEXEC_CORE guards to fix 'crashk_res' references in allnoconfig build]
Signed-off-by: Will Deacon <will@kernel.org>

kselftest/arm64: Log the PIDs of the parent and child in sve-ptrace

If the test triggers a problem it may well result in a log message from
the kernel such as a WARN() or BUG(). If these include a PID it can help
with debugging to know if it was the parent or child process that triggered
the issue, since the test is just creating a new thread the process name
will be the same either way. Print the PIDs of the parent and child on
startup so users have this information to hand should it be needed.

Signed-off-by: Mark Brown <broonie@kernel.org>
Reviewed-by: Shuah Khan <skhan@linuxfoundation.org>
Link: https://lore.kernel.org/r/20220303192817.2732509-1-broonie@kernel.org
Signed-off-by: Will Deacon <will@kernel.org>

irqchip/gic-v3: Workaround Marvell erratum 38545 when reading IAR

When a IAR register read races with a GIC interrupt RELEASE event,
GIC-CPU interface could wrongly return a valid INTID to the CPU
for an interrupt that is already released(non activated) instead of 0x3ff.

As a side effect, an interrupt handler could run twice, once with
interrupt priority and then with idle priority.

As a workaround, gic_read_iar is updated so that it will return a
valid interrupt ID only if there is a change in the active priority list
after the IAR read on all the affected Silicons.

Since there are silicon variants where both 23154 and 38545 are applicable,
workaround for erratum 23154 has been extended to address both of them.

Signed-off-by: Linu Cherian <lcherian@marvell.com>
Reviewed-by: Marc Zyngier <maz@kernel.org>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/20220307143014.22758-1-lcherian@marvell.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64/mm: Drop use_1G_block()

pud_sect_supported() already checks for PUD level block mapping support i.e
on ARM64_4K_PAGES config. Hence pud_sect_supported(), along with some other
required alignment checks can help completely drop use_1G_block().

Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will@kernel.org>
Cc: linux-arm-kernel@lists.infradead.org
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Anshuman Khandual <anshuman.khandual@arm.com>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/1644988012-25455-1-git-send-email-anshuman.khandual@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: avoid flushing icache multiple times on contiguous HugeTLB

When a contiguous HugeTLB page is mapped, set_pte_at() will be called
CONT_PTES/CONT_PMDS times.  Therefore, __sync_icache_dcache() will
flush cache multiple times if the page is executable (to ensure
the I-D cache coherency).  However, the first flushing cache already
covers subsequent cache flush operations.  So only flusing cache
for the head page if it is a HugeTLB page to avoid redundant cache
flushing.  In the next patch, it is also depends on this change
since the tail vmemmap pages of HugeTLB is mapped with read-only
meanning only head page struct can be modified.

Signed-off-by: Muchun Song <songmuchun@bytedance.com>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/20220302084624.33340-1-songmuchun@bytedance.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: prevent instrumentation of bp hardening callbacks

We may call arm64_apply_bp_hardening() early during entry (e.g. in
el0_ia()) before it is safe to run instrumented code. Unfortunately this
may result in running instrumented code in two cases:

* The hardening callbacks called by arm64_apply_bp_hardening() are not
  marked as `noinstr`, and have been observed to be instrumented when
  compiled with either GCC or LLVM.

* Since arm64_apply_bp_hardening() itself is only marked as `inline`
  rather than `__always_inline`, it is possible that the compiler
  decides to place it out-of-line, whereupon it may be instrumented.

For example, with defconfig built with clang 13.0.0,
call_hvc_arch_workaround_1() is compiled as:

| <call_hvc_arch_workaround_1>:
|        d503233f        paciasp
|        f81f0ffe        str     x30, [sp, #-16]!
|        320183e0        mov     w0, #0x80008000
|        d503201f        nop
|        d4000002        hvc     #0x0
|        f84107fe        ldr     x30, [sp], #16
|        d50323bf        autiasp
|        d65f03c0        ret

... but when CONFIG_FTRACE=y and CONFIG_KCOV=y this is compiled as:

| <call_hvc_arch_workaround_1>:
|        d503245f        bti     c
|        d503201f        nop
|        d503201f        nop
|        d503233f        paciasp
|        a9bf7bfd        stp     x29, x30, [sp, #-16]!
|        910003fd        mov     x29, sp
|        94000000        bl      0 <__sanitizer_cov_trace_pc>
|        320183e0        mov     w0, #0x80008000
|        d503201f        nop
|        d4000002        hvc     #0x0
|        a8c17bfd        ldp     x29, x30, [sp], #16
|        d50323bf        autiasp
|        d65f03c0        ret

... with a patchable function entry registered with ftrace, and a direct
call to __sanitizer_cov_trace_pc(). Neither of these are safe early
during entry sequences.

This patch avoids the unsafe instrumentation by marking
arm64_apply_bp_hardening() as `__always_inline` and by marking the
hardening functions as `noinstr`. This avoids the potential for
instrumentation, and causes clang to consistently generate the function
as with the defconfig sample.

Note: in the defconfig compilation, when CONFIG_SVE=y, x30 is spilled to
the stack without being placed in a frame record, which will result in a
missing entry if call_hvc_arch_workaround_1() is backtraced. Similar is
true of qcom_link_stack_sanitisation(), where inline asm spills the LR
to a GPR prior to corrupting it. This is not a significant issue
presently as we will only backtrace here if an exception is taken, and
in such cases we may omit entries for other reasons today.

The relevant hardening functions were introduced in commits:

  ec82b567a74fbdff ("arm64: Implement branch predictor hardening for Falkor")
  b092201e00206141 ("arm64: Add ARM_SMCCC_ARCH_WORKAROUND_1 BP hardening support")

... and these were subsequently moved in commit:

  d4647f0a2ad71110 ("arm64: Rewrite Spectre-v2 mitigation code")

The arm64_apply_bp_hardening() function was introduced in commit:

  0f15adbb2861ce6f ("arm64: Add skeleton to harden the branch predictor against aliasing attacks")

... and was subsequently moved and reworked in commit:

  6279017e807708a0 ("KVM: arm64: Move BP hardening helpers into spectre.h")

Fixes: ec82b567a74fbdff ("arm64: Implement branch predictor hardening for Falkor")
Fixes: b092201e00206141 ("arm64: Add ARM_SMCCC_ARCH_WORKAROUND_1 BP hardening support")
Fixes: d4647f0a2ad71110 ("arm64: Rewrite Spectre-v2 mitigation code")
Fixes: 0f15adbb2861ce6f ("arm64: Add skeleton to harden the branch predictor against aliasing attacks")
Fixes: 6279017e807708a0 ("KVM: arm64: Move BP hardening helpers into spectre.h")
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Cc: Ard Biesheuvel <ardb@kernel.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: James Morse <james.morse@arm.com>
Cc: Marc Zyngier <maz@kernel.org>
Cc: Mark Brown <broonie@kernel.org>
Cc: Will Deacon <will@kernel.org>
Acked-by: Marc Zyngier <maz@kernel.org>
Reviewed-by: Mark Brown <broonie@kernel.org>
Link: https://lore.kernel.org/r/20220224181028.512873-1-mark.rutland@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: crash_core: Export MODULES, VMALLOC, and VMEMMAP ranges

The following interrelated ranges are needed by the kdump crash tool:
MODULES_VADDR ~ MODULES_END,
VMALLOC_START ~ VMALLOC_END,
VMEMMAP_START ~ VMEMMAP_END

Since these values change from time to time, it is preferable to export
them via vmcoreinfo than to change the crash's code frequently.

Signed-off-by: Huang Shijie <shijie@os.amperecomputing.com>
Link: https://lore.kernel.org/r/20220209092642.9181-1-shijie@os.amperecomputing.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64/mte: Add userspace interface for enabling asymmetric mode

The architecture provides an asymmetric mode for MTE where tag mismatches
are checked asynchronously for stores but synchronously for loads. Allow
userspace processes to select this and make it available as a default mode
via the existing per-CPU sysfs interface.

Since there PR_MTE_TCF_ values are a bitmask (allowing the kernel to choose
between the multiple modes) and there are no free bits adjacent to the
existing PR_MTE_TCF_ bits the set of bits used to specify the mode becomes
disjoint. Programs using the new interface should be aware of this and
programs that do not use it will not see any change in behaviour.

When userspace requests two possible modes but the system default for the
CPU is the third mode (eg, default is synchronous but userspace requests
either asynchronous or asymmetric) the preference order is:

   ASYMM > ASYNC > SYNC

This situation is not currently possible since there are only two modes and
it is mandatory to have a system default so there could be no ambiguity and
there is no ABI change. The chosen order is basically arbitrary as we do not
have a clear metric for what is better here.

If userspace requests specifically asymmetric mode via the prctl() and the
system does not support it then we will return an error, this mirrors
how we handle the case where userspace enables MTE on a system that does
not support MTE at all and the behaviour that will be seen if running on
an older kernel that does not support userspace use of asymmetric mode.

Attempts to set asymmetric mode as the default mode will result in an error
if the system does not support it.

Signed-off-by: Mark Brown <broonie@kernel.org>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Reviewed-by: Vincenzo Frascino <Vincenzo.Frascino@arm.com>
Tested-by: Branislav Rankov <branislav.rankov@arm.com>
Link: https://lore.kernel.org/r/20220216173224.2342152-5-broonie@kernel.org
Signed-off-by: Will Deacon <will@kernel.org>

arm64/mte: Add hwcap for asymmetric mode

Allow userspace to detect support for asymmetric mode by providing a hwcap
for it, using the official feature name FEAT_MTE3.

Signed-off-by: Mark Brown <broonie@kernel.org>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Reviewed-by: Vincenzo Frascino <Vincenzo.Frascino@arm.com>
Tested-by: Branislav Rankov <branislav.rankov@arm.com>
Link: https://lore.kernel.org/r/20220216173224.2342152-4-broonie@kernel.org
Signed-off-by: Will Deacon <will@kernel.org>

arm64/mte: Add a little bit of documentation for mte_update_sctlr_user()

The code isn't that obscure but it probably won't hurt to have a little
bit more documentation for anyone trying to find out where everything
actually takes effect.

Signed-off-by: Mark Brown <broonie@kernel.org>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Reviewed-by: Vincenzo Frascino <Vincenzo.Frascino@arm.com>
Tested-by: Branislav Rankov <branislav.rankov@arm.com>
Link: https://lore.kernel.org/r/20220216173224.2342152-3-broonie@kernel.org
Signed-off-by: Will Deacon <will@kernel.org>

arm64/mte: Document ABI for asymmetric mode

MTE3 adds a new mode which is synchronous for reads but asynchronous for
writes. Document the userspace ABI for this feature, we call the new
mode ASYMM and add a new prctl flag and mte_tcf_preferred value for it.

Signed-off-by: Mark Brown <broonie@kernel.org>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/20220216173224.2342152-2-broonie@kernel.org
Signed-off-by: Will Deacon <will@kernel.org>

kselftest/arm64: signal: Allow tests to be incompatible with features

Some features may invalidate some tests, for example by supporting an
operation which would trap otherwise. Allow tests to list features that
they are incompatible with so we can cover the case where a signal will
be generated without disruption on systems where that won't happen.

Signed-off-by: Mark Brown <broonie@kernel.org>
Reviewed-by: Shuah Khan <skhan@linuxfoundation.org>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/20220207152109.197566-6-broonie@kernel.org
Signed-off-by: Will Deacon <will@kernel.org>

arm64: module: remove (NOLOAD) from linker script

On ELF, (NOLOAD) sets the section type to SHT_NOBITS[1]. It is conceptually
inappropriate for .plt and .text.* sections which are always
SHT_PROGBITS.

In GNU ld, if PLT entries are needed, .plt will be SHT_PROGBITS anyway
and (NOLOAD) will be essentially ignored. In ld.lld, since
https://reviews.llvm.org/D118840 ("[ELF] Support (TYPE=<value>) to
customize the output section type"), ld.lld will report a `section type
mismatch` error. Just remove (NOLOAD) to fix the error.

[1] https://lld.llvm.org/ELF/linker_script.html As of today, "The
section should be marked as not loadable" on
https://sourceware.org/binutils/docs/ld/Output-Section-Type.html is
outdated for ELF.

Tested-by: Nathan Chancellor <nathan@kernel.org>
Reported-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Fangrui Song <maskray@google.com>
Acked-by: Ard Biesheuvel <ardb@kernel.org>
Link: https://lore.kernel.org/r/20220218081209.354383-1-maskray@google.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: cpufeature: Remove cpu_has_fwb() check

cpu_has_fwb() is supposed to warn user is following architectural
requirement is not valid:

LoUU, bits [29:27] - Level of Unification Uniprocessor for the cache
                     hierarchy.

  Note

    When FEAT_S2FWB is implemented, the architecture requires that
    this field is zero so that no levels of data cache need to be
    cleaned in order to manage coherency with instruction fetches.

LoUIS, bits [23:21] - Level of Unification Inner Shareable for the
                      cache hierarchy.

  Note

    When FEAT_S2FWB is implemented, the architecture requires that
    this field is zero so that no levels of data cache need to be
    cleaned in order to manage coherency with instruction fetches.

It is not really clear what user have to do if assertion fires. Having
assertions about the CPU design like this inspire even more assertions
to be added and the kernel definitely is not the right place for that,
so let's remove cpu_has_fwb() altogether.

Signed-off-by: Vladimir Murzin <vladimir.murzin@arm.com>
Link: https://lore.kernel.org/r/20220224164739.119168-1-vladimir.murzin@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: Add support of PAuth QARMA3 architected algorithm

QARMA3 is relaxed version of the QARMA5 algorithm which expected to
reduce the latency of calculation while still delivering a suitable
level of security.

Support for QARMA3 can be discovered via ID_AA64ISAR2_EL1

    APA3, bits [15:12] Indicates whether the QARMA3 algorithm is
                       implemented in the PE for address
                       authentication in AArch64 state.

    GPA3, bits [11:8]  Indicates whether the QARMA3 algorithm is
                       implemented in the PE for generic code
                       authentication in AArch64 state.

Signed-off-by: Vladimir Murzin <vladimir.murzin@arm.com>
Acked-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20220224124952.119612-4-vladimir.murzin@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: cpufeature: Mark existing PAuth architected algorithm as QARMA5

In preparation of supporting PAuth QARMA3 architected algorithm mark
existing one as QARMA5, so we can distingwish between two.

Signed-off-by: Vladimir Murzin <vladimir.murzin@arm.com>
Acked-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20220224124952.119612-3-vladimir.murzin@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: cpufeature: Account min_field_value when cheking secondaries for PAuth

In case, both boot_val and sec_val have value below min_field_value we
would wrongly report that address authentication is supported. It is
not a big issue because we enable address authentication based on boot
cpu (and check there is correct).

Signed-off-by: Vladimir Murzin <vladimir.murzin@arm.com>
Acked-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20220224124952.119612-2-vladimir.murzin@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: Change elfcore for_each_mte_vma() to use VMA iterator

Rework for_each_mte_vma() to use a VMA iterator instead of an explicit
linked-list. This will allow easy integration with the maple tree work
which removes the VMA list altogether.

Signed-off-by: Liam R. Howlett <Liam.Howlett@oracle.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/20220218023650.672072-1-Liam.Howlett@oracle.com
[will: Folded in fix from Catalin]
Link: https://lore.kernel.org/r/YhUcywqIhmHvX6dG@arm.com
Signed-off--by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Will Deacon <will@kernel.org>

arm64/hugetlb: Define __hugetlb_valid_size()

arch_hugetlb_valid_size() can be just factored out to create another helper
to be used in arch_hugetlb_migration_supported() as well. This just defines
__hugetlb_valid_size() for that purpose.

Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will@kernel.org>
Cc: linux-arm-kernel@lists.infradead.org
Cc: linux-kernel@vger.kernel.org
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Anshuman Khandual <anshuman.khandual@arm.com>
Link: https://lore.kernel.org/r/1645073557-6150-1-git-send-email-anshuman.khandual@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: mte: avoid clearing PSTATE.TCO on entry unless necessary

On some microarchitectures, clearing PSTATE.TCO is expensive. Clearing
TCO is only necessary if in-kernel MTE is enabled, or if MTE is
enabled in the userspace process in synchronous (or, soon, asymmetric)
mode, because we do not report uaccess faults to userspace in none
or asynchronous modes. Therefore, adjust the kernel entry code to
clear TCO only if necessary.

Because it is now possible to switch to a task in which TCO needs to
be clear from a task in which TCO is set, we also need to do the same
thing on task switch.

Signed-off-by: Peter Collingbourne <pcc@google.com>
Link: https://linux-review.googlesource.com/id/I52d82a580bd0500d420be501af2c35fa8c90729e
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/20220219012945.894950-2-pcc@google.com
Signed-off-by: Will Deacon <will@kernel.org>

kasan: split kasan_*enabled() functions into a separate header

In an upcoming commit we are going to need to call
kasan_hw_tags_enabled() from arch/arm64/include/asm/mte.h. This
would create a circular dependency between headers if KASAN_GENERIC
or KASAN_SW_TAGS is enabled: linux/kasan.h -> linux/pgtable.h ->
asm/pgtable.h -> asm/mte.h -> linux/kasan.h. Break the cycle
by introducing a new header linux/kasan-enabled.h with the
kasan_*enabled() functions that can be included from asm/mte.h.

Link: https://linux-review.googlesource.com/id/I5b0d96c6ed0026fc790899e14d42b2fac6ab568e
Signed-off-by: Peter Collingbourne <pcc@google.com>
Reviewed-by: Andrey Konovalov <andreyknvl@gmail.com>
Link: https://lore.kernel.org/r/20220219012945.894950-1-pcc@google.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: insn: add encoders for atomic operations

It is a preparation patch for eBPF atomic supports under arm64. eBPF
needs support atomic[64]_fetch_add, atomic[64]_[fetch_]{and,or,xor} and
atomic[64]_{xchg|cmpxchg}. The ordering semantics of eBPF atomics are
the same with the implementations in linux kernel.

Add three helpers to support LDCLR/LDEOR/LDSET/SWP, CAS and DMB
instructions. STADD/STCLR/STEOR/STSET are simply encoded as aliases for
LDADD/LDCLR/LDEOR/LDSET with XZR as the destination register, so no extra
helper is added. atomic_fetch_add() and other atomic ops needs support for
STLXR instruction, so extend enum aarch64_insn_ldst_type to do that.

LDADD/LDEOR/LDSET/SWP and CAS instructions are only available when LSE
atomics is enabled, so just return AARCH64_BREAK_FAULT directly in
these newly-added helpers if CONFIG_ARM64_LSE_ATOMICS is disabled.

Signed-off-by: Hou Tao <houtao1@huawei.com>
Link: https://lore.kernel.org/r/20220217072232.1186625-3-houtao1@huawei.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: move AARCH64_BREAK_FAULT into insn-def.h

If CONFIG_ARM64_LSE_ATOMICS is off, encoders for LSE-related instructions
can return AARCH64_BREAK_FAULT directly in insn.h. In order to access
AARCH64_BREAK_FAULT in insn.h, we can not include debug-monitors.h in
insn.h, because debug-monitors.h has already depends on insn.h, so just
move AARCH64_BREAK_FAULT into insn-def.h.

It will be used by the following patch to eliminate unnecessary LSE-related
encoders when CONFIG_ARM64_LSE_ATOMICS is off.

Signed-off-by: Hou Tao <houtao1@huawei.com>
Link: https://lore.kernel.org/r/20220217072232.1186625-2-houtao1@huawei.com
Signed-off-by: Will Deacon <will@kernel.org>

linkage: remove SYM_FUNC_{START,END}_ALIAS()

Now that all aliases are defined using SYM_FUNC_ALIAS(), remove the old
SYM_FUNC_{START,END}_ALIAS() macros.

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

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Acked-by: Ard Biesheuvel <ardb@kernel.org>
Acked-by: Josh Poimboeuf <jpoimboe@redhat.com>
Acked-by: Mark Brown <broonie@kernel.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Jiri Slaby <jslaby@suse.cz>
Cc: Peter Zijlstra <peterz@infradead.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20220216162229.1076788-5-mark.rutland@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

x86: clean up symbol aliasing

Now that we have SYM_FUNC_ALIAS() and SYM_FUNC_ALIAS_WEAK(), use those
to simplify the definition of function aliases across arch/x86.

For clarity, where there are multiple annotations such as
EXPORT_SYMBOL(), I've tried to keep annotations grouped by symbol. For
example, where a function has a name and an alias which are both
exported, this is organised as:

SYM_FUNC_START(func)
    ... asm insns ...
SYM_FUNC_END(func)
EXPORT_SYMBOL(func)

SYM_FUNC_ALIAS(alias, func)
EXPORT_SYMBOL(alias)

Where there are only aliases and no exports or other annotations, I have
not bothered with line spacing, e.g.

SYM_FUNC_START(func)
    ... asm insns ...
SYM_FUNC_END(func)
SYM_FUNC_ALIAS(alias, func)

The tools/perf/ copies of memset_64.S and memset_64.S are updated
likewise to avoid the build system complaining these are mismatched:

| Warning: Kernel ABI header at 'tools/arch/x86/lib/memcpy_64.S' differs from latest version at 'arch/x86/lib/memcpy_64.S'
| diff -u tools/arch/x86/lib/memcpy_64.S arch/x86/lib/memcpy_64.S
| Warning: Kernel ABI header at 'tools/arch/x86/lib/memset_64.S' differs from latest version at 'arch/x86/lib/memset_64.S'
| diff -u tools/arch/x86/lib/memset_64.S arch/x86/lib/memset_64.S

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

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Acked-by: Ard Biesheuvel <ardb@kernel.org>
Acked-by: Josh Poimboeuf <jpoimboe@redhat.com>
Acked-by: Mark Brown <broonie@kernel.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Slaby <jslaby@suse.cz>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20220216162229.1076788-4-mark.rutland@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: clean up symbol aliasing

Now that we have SYM_FUNC_ALIAS() and SYM_FUNC_ALIAS_WEAK(), use those
to simplify and more consistently define function aliases across
arch/arm64.

Aliases are now defined in terms of a canonical function name. For
position-independent functions I've made the __pi_<func> name the
canonical name, and defined other alises in terms of this.

The SYM_FUNC_{START,END}_PI(func) macros obscure the __pi_<func> name,
and make this hard to seatch for. The SYM_FUNC_START_WEAK_PI() macro
also obscures the fact that the __pi_<func> fymbol is global and the
<func> symbol is weak. For clarity, I have removed these macros and used
SYM_FUNC_{START,END}() directly with the __pi_<func> name.

For example:

SYM_FUNC_START_WEAK_PI(func)
... asm insns ...
SYM_FUNC_END_PI(func)
EXPORT_SYMBOL(func)

... becomes:

SYM_FUNC_START(__pi_func)
... asm insns ...
SYM_FUNC_END(__pi_func)

SYM_FUNC_ALIAS_WEAK(func, __pi_func)
EXPORT_SYMBOL(func)

For clarity, where there are multiple annotations such as
EXPORT_SYMBOL(), I've tried to keep annotations grouped by symbol. For
example, where a function has a name and an alias which are both
exported, this is organised as:

SYM_FUNC_START(func)
... asm insns ...
SYM_FUNC_END(func)
EXPORT_SYMBOL(func)

SYM_FUNC_ALIAS(alias, func)
EXPORT_SYMBOL(alias)

For consistency with the other string functions, I've defined strrchr as
a position-independent function, as it can safely be used as such even
though we have no users today.

As we no longer use SYM_FUNC_{START,END}_ALIAS(), our local copies are
removed. The common versions will be removed by a subsequent patch.

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

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Acked-by: Ard Biesheuvel <ardb@kernel.org>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Acked-by: Josh Poimboeuf <jpoimboe@redhat.com>
Acked-by: Mark Brown <broonie@kernel.org>
Cc: Joey Gouly <joey.gouly@arm.com>
Cc: Will Deacon <will@kernel.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20220216162229.1076788-3-mark.rutland@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

linkage: add SYM_FUNC_ALIAS{,_LOCAL,_WEAK}()

Currently aliasing an asm function requires adding START and END
annotations for each name, as per Documentation/asm-annotations.rst:

SYM_FUNC_START_ALIAS(__memset)
SYM_FUNC_START(memset)
    ... asm insns ...
SYM_FUNC_END(memset)
SYM_FUNC_END_ALIAS(__memset)

This is more painful than necessary to maintain, especially where a
function has many aliases, some of which we may wish to define
conditionally. For example, arm64's memcpy/memmove implementation (which
uses some arch-specific SYM_*() helpers) has:

SYM_FUNC_START_ALIAS(__memmove)
SYM_FUNC_START_ALIAS_WEAK_PI(memmove)
SYM_FUNC_START_ALIAS(__memcpy)
SYM_FUNC_START_WEAK_PI(memcpy)
    ... asm insns ...
SYM_FUNC_END_PI(memcpy)
EXPORT_SYMBOL(memcpy)
SYM_FUNC_END_ALIAS(__memcpy)
EXPORT_SYMBOL(__memcpy)
SYM_FUNC_END_ALIAS_PI(memmove)
EXPORT_SYMBOL(memmove)
SYM_FUNC_END_ALIAS(__memmove)
EXPORT_SYMBOL(__memmove)
SYM_FUNC_START(name)

It would be much nicer if we could define the aliases *after* the
standard function definition. This would avoid the need to specify each
symbol name twice, and would make it easier to spot the canonical
function definition.

This patch adds new macros to allow us to do so, which allows the above
example to be rewritten more succinctly as:

SYM_FUNC_START(__pi_memcpy)
    ... asm insns ...
SYM_FUNC_END(__pi_memcpy)

SYM_FUNC_ALIAS(__memcpy, __pi_memcpy)
EXPORT_SYMBOL(__memcpy)
SYM_FUNC_ALIAS_WEAK(memcpy, __memcpy)
EXPORT_SYMBOL(memcpy)

SYM_FUNC_ALIAS(__pi_memmove, __pi_memcpy)
SYM_FUNC_ALIAS(__memmove, __pi_memmove)
EXPORT_SYMBOL(__memmove)
SYM_FUNC_ALIAS_WEAK(memmove, __memmove)
EXPORT_SYMBOL(memmove)

The reduction in duplication will also make it possible to replace some
uses of WEAK with more accurate Kconfig guards, e.g.

#ifndef CONFIG_KASAN
SYM_FUNC_ALIAS(memmove, __memmove)
EXPORT_SYMBOL(memmove)
#endif

... which should make it easier to ensure that symbols are neither used
nor overidden unexpectedly.

The existing SYM_FUNC_START_ALIAS() and SYM_FUNC_START_LOCAL_ALIAS() are
marked as deprecated, and will be removed once existing users are moved
over to the new scheme.

The tools/perf/ copy of linkage.h is updated to match. A subsequent
patch will depend upon this when updating the x86 asm annotations.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Acked-by: Ard Biesheuvel <ardb@kernel.org>
Acked-by: Josh Poimboeuf <jpoimboe@redhat.com>
Acked-by: Mark Brown <broonie@kernel.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Jiri Slaby <jslaby@suse.cz>
Cc: Peter Zijlstra <peterz@infradead.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20220216162229.1076788-2-mark.rutland@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: mte: Document the core dump file format

Add the program header definition and data layout for the
PT_ARM_MEMTAG_MTE segments.

Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Acked-by: Luis Machado <luis.machado@linaro.org>
Link: https://lore.kernel.org/r/20220131165456.2160675-6-catalin.marinas@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: mte: Dump the MTE tags in the core file

For each vma mapped with PROT_MTE (the VM_MTE flag set), generate a
PT_ARM_MEMTAG_MTE segment in the core file and dump the corresponding
tags. The in-file size for such segments is 128 bytes per page.

For pages in a VM_MTE vma which are not present in the user page tables
or don't have the PG_mte_tagged flag set (e.g. execute-only), just write
zeros in the core file.

An example of program headers for two vmas, one 2-page, the other 4-page
long:

  Type           Offset   VirtAddr           PhysAddr           FileSiz  MemSiz   Flg Align
  ...
  LOAD           0x030000 0x0000ffff80034000 0x0000000000000000 0x000000 0x002000 RW  0x1000
  LOAD           0x030000 0x0000ffff80036000 0x0000000000000000 0x004000 0x004000 RW  0x1000
  ...
  LOPROC+0x1     0x05b000 0x0000ffff80034000 0x0000000000000000 0x000100 0x002000     0
  LOPROC+0x1     0x05b100 0x0000ffff80036000 0x0000000000000000 0x000200 0x004000     0

Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Acked-by: Luis Machado <luis.machado@linaro.org>
Reviewed-by: Mark Brown <broonie@kernel.org>
Link: https://lore.kernel.org/r/20220131165456.2160675-5-catalin.marinas@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: mte: Define the number of bytes for storing the tags in a page

Rather than explicitly calculating the number of bytes for a compact tag
storage format corresponding to a page, just add a MTE_PAGE_TAG_STORAGE
macro. With the current MTE implementation of 4 bits per tag, we store
2 tags in a byte.

Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Acked-by: Luis Machado <luis.machado@linaro.org>
Link: https://lore.kernel.org/r/20220131165456.2160675-4-catalin.marinas@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

elf: Introduce the ARM MTE ELF segment type

Memory tags will be dumped in the core file as segments with their own
type. Discussions with the binutils and the generic ABI community
settled on using new definitions in the PT_*PROC space (and to be
documented in the processor-specific ABIs).

Introduce PT_ARM_MEMTAG_MTE as (PT_LOPROC + 0x1). Not included in this
patch since there is no upstream support but the CHERI/BSD community
will also reserve:

  #define PT_ARM_MEMTAG_CHERI    (PT_LOPROC + 0x2)
  #define PT_RISCV_MEMTAG_CHERI  (PT_LOPROC + 0x3)

Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Acked-by: Luis Machado <luis.machado@linaro.org>
Link: https://lore.kernel.org/r/20220131165456.2160675-3-catalin.marinas@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

elfcore: Replace CONFIG_{IA64, UML} checks with a new option

As arm64 is about to introduce MTE-specific phdrs in the core dump, add
a common CONFIG_ARCH_BINFMT_ELF_EXTRA_PHDRS option currently selectable
by UML_X86 and IA64.

Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Cc: Eric Biederman <ebiederm@xmission.com>
Link: https://lore.kernel.org/r/20220131165456.2160675-2-catalin.marinas@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: atomics: remove redundant static branch

Due to a historical oversight, we emit a redundant static branch for
each atomic/atomic64 operation when CONFIG_ARM64_LSE_ATOMICS is
selected. We can safely remove this, making the kernel Image reasonably
smaller.

When CONFIG_ARM64_LSE_ATOMICS is selected, every LSE atomic operation
has two preceding static branches with the same target, e.g.

b f7c <kernel_init_freeable+0xa4>
b f7c <kernel_init_freeable+0xa4>
mov w0, #0x1                    // #1
ldadd w0, w0, [x19]

This is because the __lse_ll_sc_body() wrapper uses
system_uses_lse_atomics(), which checks both `arm64_const_caps_ready`
and `cpu_hwcap_keys[ARM64_HAS_LSE_ATOMICS]`, each of which emits a
static branch. This has been the case since commit:

  addfc38672c73efd ("arm64: atomics: avoid out-of-line ll/sc atomics")

However, there was never a need to check `arm64_const_caps_ready`, which
was itself introduced in commit:

  63a1e1c95e60e798 ("arm64/cpufeature: don't use mutex in bringup path")

... so that cpus_have_const_cap() could fall back to checking the
`cpu_hwcaps` bitmap prior to the static keys for individual caps
becoming enabled. As system_uses_lse_atomics() doesn't check
`cpu_hwcaps`, and doesn't need to as we can safely use the LL/SC atomics
prior to enabling the `ARM64_HAS_LSE_ATOMICS` static key, it doesn't
need to check `arm64_const_caps_ready`.

This patch removes the `arm64_const_caps_ready` check from
system_uses_lse_atomics(). As the arch_atomic_* routines are meant to be
safely usable in noinstr code, I've also marked
system_uses_lse_atomics() as __always_inline.

This results in one fewer static branch per atomic operation, with the
prior example becoming:

b f78 <kernel_init_freeable+0xa0>
mov w0, #0x1                    // #1
ldadd w0, w0, [x19]

Each static branch consists of the branch itself and an associated
__jump_table entry. Removing these has a reasonable impact on the Image
size, with a GCC 11.1.0 defconfig v5.17-rc2 Image being reduced by
128KiB:

| [mark@lakrids:~/src/linux]% ls -al Image*
| -rw-r--r-- 1 mark mark 34619904 Feb  3 18:24 Image.baseline
| -rw-r--r-- 1 mark mark 34488832 Feb  3 18:33 Image.onebranch

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Cc: Ard Biesheuvel <ardb@kernel.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Suzuki Poulose <suzuki.poulose@arm.com>
Cc: Will Deacon <will@kernel.org>
Link: https://lore.kernel.org/r/20220204104439.270567-1-mark.rutland@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

kselftest/arm64: mte: user_mem: test a wider range of values

Instead of hard coding a small amount of tests, generate a wider
range of tests to try catch any corner cases that could show up.

These new tests test different MTE tag lengths and offsets, which
previously would have caused infinite loops in the kernel. This was
fixed by 295cf156231c ("arm64: Avoid premature usercopy failure"),
so these are regressions tests for that corner case.

Signed-off-by: Joey Gouly <joey.gouly@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will@kernel.org>
Cc: Mark Brown <broonie@kernel.org>
Cc: Shuah Khan <shuah@kernel.org>
Reviewed-by: Mark Brown <broonie@kernel.org>
Tested-by: Mark Brown <broonie@kernel.org>
Reviewed-by: Shuah Khan <skhan@linuxfoundation.org>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/20220209152240.52788-7-joey.gouly@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

kselftest/arm64: mte: user_mem: add more test types

To expand the test coverage for MTE tags in userspace memory,
also perform the test with `write`, `readv` and `writev` syscalls.

Signed-off-by: Joey Gouly <joey.gouly@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will@kernel.org>
Cc: Mark Brown <broonie@kernel.org>
Cc: Shuah Khan <shuah@kernel.org>
Reviewed-by: Mark Brown <broonie@kernel.org>
Tested-by: Mark Brown <broonie@kernel.org>
Reviewed-by: Shuah Khan <skhan@linuxfoundation.org>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/20220209152240.52788-6-joey.gouly@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

kselftest/arm64: mte: user_mem: add test type enum

The test is currently hardcoded to use the `read` syscall, this commit adds
a test_type enum to support expanding the test coverage to other syscalls.

Signed-off-by: Joey Gouly <joey.gouly@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will@kernel.org>
Cc: Mark Brown <broonie@kernel.org>
Cc: Shuah Khan <shuah@kernel.org>
Reviewed-by: Mark Brown <broonie@kernel.org>
Tested-by: Mark Brown <broonie@kernel.org>
Reviewed-by: Shuah Khan <skhan@linuxfoundation.org>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/20220209152240.52788-5-joey.gouly@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

kselftest/arm64: mte: user_mem: check different offsets and sizes

To check there are no assumptions in the kernel about buffer sizes or alignments of
user space pointers, expand the test to cover different sizes and offsets.

Signed-off-by: Joey Gouly <joey.gouly@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will@kernel.org>
Cc: Mark Brown <broonie@kernel.org>
Cc: Shuah Khan <shuah@kernel.org>
Reviewed-by: Mark Brown <broonie@kernel.org>
Tested-by: Mark Brown <broonie@kernel.org>
Reviewed-by: Shuah Khan <skhan@linuxfoundation.org>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/20220209152240.52788-4-joey.gouly@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

kselftest/arm64: mte: user_mem: rework error handling

Future commits will have multiple iterations of tests in this function,
so make the error handling assume it will pass and then bail out if there
is an error.

Signed-off-by: Joey Gouly <joey.gouly@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will@kernel.org>
Cc: Mark Brown <broonie@kernel.org>
Cc: Shuah Khan <shuah@kernel.org>
Reviewed-by: Mark Brown <broonie@kernel.org>
Tested-by: Mark Brown <broonie@kernel.org>
Reviewed-by: Shuah Khan <skhan@linuxfoundation.org>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/20220209152240.52788-3-joey.gouly@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

kselftest/arm64: mte: user_mem: introduce tag_offset and tag_len

These can be used to place an MTE tag at an address that is not at a
page size boundary.

The kernel prior to 295cf156231c ("arm64: Avoid premature usercopy failure"),
would infinite loop if an MTE tag was placed not at a PAGE_SIZE boundary.
This is because the kernel checked if the pages were readable by checking the
first byte of each page, but would then fault in the middle of the page due
to the MTE tag.

Signed-off-by: Joey Gouly <joey.gouly@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will@kernel.org>
Cc: Mark Brown <broonie@kernel.org>
Cc: Shuah Khan <shuah@kernel.org>
Reviewed-by: Mark Brown <broonie@kernel.org>
Tested-by: Mark Brown <broonie@kernel.org>
Reviewed-by: Shuah Khan <skhan@linuxfoundation.org>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/20220209152240.52788-2-joey.gouly@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64/mte: Clarify mode reported by PR_GET_TAGGED_ADDR_CTRL

With the current wording readers might infer that PR_GET_TAGGED_ADDR_CTRL
will report the mode currently active in the thread however this is not the
actual behaviour, instead all modes currently selected by the process will
be reported with the mode used depending on the combination of the
requested modes and the default set for the current CPU. This has been the
case since 433c38f40f6a81 ("arm64: mte: change ASYNC and SYNC TCF settings
into bitfields"), before that we did not allow more than one mode to be
requested simultaneously.

Update the documentation to more clearly reflect current behaviour.

Signed-off-by: Mark Brown <broonie@kernel.org>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/20220127190324.660405-1-broonie@kernel.org
Signed-off-by: Will Deacon <will@kernel.org>

kselftest/arm64: Remove local definitions of MTE prctls

The GCR EL1 test unconditionally includes local definitions of the prctls
it tests. Since not only will the kselftest build infrastructure ensure
that the in tree uapi headers are available but the toolchain being used to
build kselftest may ensure that system uapi headers with MTE support are
available this causes the compiler to warn about duplicate definitions.
Remove these duplicate definitions.

Signed-off-by: Mark Brown <broonie@kernel.org>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Acked-by: Shuah Khan <skhan@linuxfoundation.org>
Link: https://lore.kernel.org/r/20220126174421.1712795-1-broonie@kernel.org
Signed-off-by: Will Deacon <will@kernel.org>

kselftest/arm64: Remove local ARRAY_SIZE() definitions

An ARRAY_SIZE() has been added to kselftest.h so remove the local versions
in some of the arm64 selftests.

Signed-off-by: Mark Brown <broonie@kernel.org>
Reviewed-by: Shuah Khan <skhan@linuxfoundation.org>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/20220124171748.2195875-1-broonie@kernel.org
Signed-off-by: Will Deacon <will@kernel.org>

arm64: insn: Generate 64 bit mask immediates correctly

When the insn framework is used to encode an AND/ORR/EOR instruction,
aarch64_encode_immediate() is used to pick the immr imms values.

If the immediate is a 64bit mask, with bit 63 set, and zeros in any
of the upper 32 bits, the immr value is incorrectly calculated meaning
the wrong mask is generated.
For example, 0x8000000000000001 should have an immr of 1, but 32 is used,
meaning the resulting mask is 0x0000000300000000.

It would appear eBPF is unable to hit these cases, as build_insn()'s
imm value is a s32, so when used with BPF_ALU64, the sign-extended
u64 immediate would always have all-1s or all-0s in the upper 32 bits.

KVM does not generate a va_mask with any of the top bits set as these
VA wouldn't be usable with TTBR0_EL2.

This happens because the rotation is calculated from fls(~imm), which
takes an unsigned int, but the immediate may be 64bit.

Use fls64() so the 64bit mask doesn't get truncated to a u32.

Signed-off-by: James Morse <james.morse@arm.com>
Brown-paper-bag-for: Marc Zyngier <maz@kernel.org>
Acked-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20220127162127.2391947-4-james.morse@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64/mm: avoid fixmap race condition when create pud mapping

The 'fixmap' is a global resource and is used recursively by
create pud mapping(), leading to a potential race condition in the
presence of a concurrent call to alloc_init_pud():

kernel_init thread                          virtio-mem workqueue thread
==================                          ===========================

  alloc_init_pud(...)                       alloc_init_pud(...)
  pudp = pud_set_fixmap_offset(...)         pudp = pud_set_fixmap_offset(...)
  READ_ONCE(*pudp)
  pud_clear_fixmap(...)
                                            READ_ONCE(*pudp) // CRASH!

As kernel may sleep during creating pud mapping, introduce a mutex lock to
serialise use of the fixmap entries by alloc_init_pud(). However, there is
no need for locking in early boot stage and it doesn't work well with
KASLR enabled when early boot. So, enable lock when system_state doesn't
equal to "SYSTEM_BOOTING".

Signed-off-by: Jianyong Wu <jianyong.wu@arm.com>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Fixes: f4710445458c ("arm64: mm: use fixmap when creating page tables")
Link: https://lore.kernel.org/r/20220201114400.56885-1-jianyong.wu@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64/mm: Consolidate TCR_EL1 fields

This renames and moves SYS_TCR_EL1_TCMA1 and SYS_TCR_EL1_TCMA0 definitions
into pgtable-hwdef.h thus consolidating all TCR fields in a single header.
This does not cause any functional change.

Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will@kernel.org>
Cc: linux-arm-kernel@lists.infradead.org
Cc: linux-kernel@vger.kernel.org
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Anshuman Khandual <anshuman.khandual@arm.com>
Link: https://lore.kernel.org/r/1643121513-21854-1-git-send-email-anshuman.khandual@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: entry: Save some nops when CONFIG_ARM64_PSEUDO_NMI is not set

Arm64 pseudo-NMI feature code brings some additional nops
when CONFIG_ARM64_PSEUDO_NMI is not set, which is not
necessary. So add necessary ifdeffery to avoid it.

Signed-off-by: He Ying <heying24@huawei.com>
Acked-by: Mark Rutland <mark.rutland@arm.com>
Link: https://lore.kernel.org/r/20220112032410.29231-1-heying24@huawei.com
Signed-off-by: Will Deacon <will@kernel.org>

arm64: booting.rst: Clarify on requiring non-secure EL2

The ARMv8.4 architecture revision introduced the EL2 exception level
to the secure world. Clarify the existing wording to make sure that
Linux relies on being executed in the non-secure state.

Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Reviewed-by: Mark Rutland <mark.rutland@arm.com>
Link: https://lore.kernel.org/r/20220107160056.322141-2-andre.przywara@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

perf: replace bitmap_weight with bitmap_empty where appropriate

In some places, drivers/perf code calls bitmap_weight() to check if any
bit of a given bitmap is set. It's better to use bitmap_empty() in that
case because bitmap_empty() stops traversing the bitmap as soon as it
finds first set bit, while bitmap_weight() counts all bits unconditionally.

Signed-off-by: Yury Norov <yury.norov@gmail.com>
Acked-by: Mark Rutland <mark.rutland@arm.com>
Link: https://lore.kernel.org/r/20220210224933.379149-13-yury.norov@gmail.com
Signed-off-by: Will Deacon <will@kernel.org>

perf: Replace acpi_bus_get_device()

Replace acpi_bus_get_device() that is going to be dropped with
acpi_fetch_acpi_dev().

No intentional functional impact.

Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-by: Will Deacon <will@kernel.org>
Link: https://lore.kernel.org/r/10025610.nUPlyArG6x@kreacher
Signed-off-by: Will Deacon <will@kernel.org>

perf/marvell_cn10k: Fix unused variable warning when W=1 and CONFIG_OF=n

The kbuild helpfully reports that the Marvell CN10K TAD PMU driver emits
a warning when building with W=1 and CONFIG_OF=n:

  | >> drivers/perf/marvell_cn10k_tad_pmu.c:371:34: warning: unused variable 'tad_pmu_of_match' [-Wunused-const-variable]
       static const struct of_device_id tad_pmu_of_match[] = {

Guard the match table with CONFIG_OF to squash the warning.

Link: https://lore.kernel.org/r/202201292349.zRQLcDDD-lkp@intel.com
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Will Deacon <will@kernel.org>

perf/arm-cmn: Make arm_cmn_debugfs static

Indeed our debugfs directory is driver-internal so should be static.

Link: https://lore.kernel.org/r/202202030812.II1K2ZXf-lkp@intel.com
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Reviewed-by: Anshuman Khandual <anshuman.khandual@arm.com>
Link: https://lore.kernel.org/r/ca9248caaae69b5134f69e085fe78905dfe74378.1643911278.git.robin.murphy@arm.com
Signed-off-by: Will Deacon <will@kernel.org>

perf: MARVELL_CN10K_TAD_PMU should depend on ARCH_THUNDER

The Marvell CN10K Last-Level cache Tag-and-data Units (LLC-TAD)
performance monitor is only present on Marvell CN10K SoCs.  Hence add a
dependency on ARCH_THUNDER, to prevent asking the user about this driver
when configuring a kernel without Cavium Thunder (incl. Marvell CN10K)
SoC support.

Fixes: 036a7584bede ("drivers: perf: Add LLC-TAD perf counter support")
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Link: https://lore.kernel.org/r/b4662a2c767d04cca19417e0c845edea2da262ad.1641995941.git.geert+renesas@glider.be
Signed-off-by: Will Deacon <will@kernel.org>

perf/arm-ccn: Use platform_get_irq() to get the interrupt

platform_get_resource(pdev, IORESOURCE_IRQ, ..) relies on static
allocation of IRQ resources in DT core code, this causes an issue
when using hierarchical interrupt domains using "interrupts" property
in the node as this bypasses the hierarchical setup and messes up the
irq chaining.

In preparation for removal of static setup of IRQ resource from DT core
code use platform_get_irq().

Link: https://lore.kernel.org/r/20211224161334.31123-1-prabhakar.mahadev-lad.rj@bp.renesas.com
Signed-off-by: Lad Prabhakar <prabhakar.mahadev-lad.rj@bp.renesas.com>
Signed-off-by: Will Deacon <will@kernel.org>

irqchip/apple-aic: Move PMU-specific registers to their own include file

As we are about to have a PMU driver, move the PMU bits from the AIC
driver into a common include file.

Reviewed-by: Hector Martin <marcan@marcan.st>
Signed-off-by: Marc Zyngier <maz@kernel.org>

arm64: dts: apple: Add t8303 PMU nodes

Advertise the two PMU nodes for the t8103 SoC.

Reviewed-by: Hector Martin <marcan@marcan.st>
Signed-off-by: Marc Zyngier <maz@kernel.org>

arm64: dts: apple: Add t8103 PMU interrupt affinities

The two PMU pseudo interrupts have specific affinities. One set
is affine to the small cores, and the other set affine to the
big ones.

Signed-off-by: Marc Zyngier <maz@kernel.org>

irqchip/apple-aic: Wire PMU interrupts

Add the necessary code to configure and P and E-core PMU interrupts
with their respective affinities. When such an interrupt fires, map
it onto the right pseudo-interrupt.

Reviewed-by: Hector Martin <marcan@marcan.st>
Signed-off-by: Marc Zyngier <maz@kernel.org>

irqchip/apple-aic: Parse FIQ affinities from device-tree

In order to be able to tell the core IRQ code about the affinity
of the PMU interrupt in later patches, parse the affinities kindly
provided in the device-tree.

Signed-off-by: Marc Zyngier <maz@kernel.org>

dt-bindings: apple,aic: Add affinity description for per-cpu pseudo-interrupts

Some of the FIQ per-cpu pseudo-interrupts are better described with
a specific affinity, the most obvious candidate being the CPU PMUs.

Augment the AIC binding to be able to specify that affinity in the
interrupt controller node.

Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: Marc Zyngier <maz@kernel.org>

dt-bindings: apple,aic: Add CPU PMU per-cpu pseudo-interrupts

Advertise the two pseudo-interrupts that tied to the two PMU
flavours present in the Apple M1 SoC.

We choose the expose two different pseudo-interrupts to the OS
as the e-core PMU is obviously different from the p-core one,
effectively presenting two different devices.

Acked-by: Rob Herring <robh@kernel.org>
Reviewed-by: Hector Martin <marcan@marcan.st>
Signed-off-by: Marc Zyngier <maz@kernel.org>

dt-bindings: arm-pmu: Document Apple PMU compatible strings

As we are about to add support fur the Apple PMUs, document the compatible
strings associated with the two micro-architectures present in the Apple M1.

Acked-by: Rob Herring <robh@kernel.org>
Reviewed-by: Hector Martin <marcan@marcan.st>
Signed-off-by: Marc Zyngier <maz@kernel.org>

Linux 5.17-rc3

Merge tag 'ext4_for_linus_stable' of git://git./linux/kernel/git/tytso/ext4

Pull ext4 fixes from Ted Ts'o:
 "Various bug fixes for ext4 fast commit and inline data handling.

  Also fix regression introduced as part of moving to the new mount API"

* tag 'ext4_for_linus_stable' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4:
  fs/ext4: fix comments mentioning i_mutex
  ext4: fix incorrect type issue during replay_del_range
  jbd2: fix kernel-doc descriptions for jbd2_journal_shrink_{scan,count}()
  ext4: fix potential NULL pointer dereference in ext4_fill_super()
  jbd2: refactor wait logic for transaction updates into a common function
  jbd2: cleanup unused functions declarations from jbd2.h
  ext4: fix error handling in ext4_fc_record_modified_inode()
  ext4: remove redundant max inline_size check in ext4_da_write_inline_data_begin()
  ext4: fix error handling in ext4_restore_inline_data()
  ext4: fast commit may miss file actions
  ext4: fast commit may not fallback for ineligible commit
  ext4: modify the logic of ext4_mb_new_blocks_simple
  ext4: prevent used blocks from being allocated during fast commit replay

Merge tag 'perf-tools-fixes-for-v5.17-2022-02-06' of git://git./linux/kernel/git/acme/linux

Pull perf tools fixes from Arnaldo Carvalho de Melo:

 - Fix display of grouped aliased events in 'perf stat'.

 - Add missing branch_sample_type to perf_event_attr__fprintf().

 - Apply correct label to user/kernel symbols in branch mode.

 - Fix 'perf ftrace' system_wide tracing, it has to be set before
   creating the maps.

 - Return error if procfs isn't mounted for PID namespaces when
   synthesizing records for pre-existing processes.

 - Set error stream of objdump process for 'perf annotate' TUI, to avoid
   garbling the screen.

 - Add missing arm64 support to perf_mmap__read_self(), the kernel part
   got into 5.17.

 - Check for NULL pointer before dereference writing debug info about a
   sample.

 - Update UAPI copies for asound, perf_event, prctl and kvm headers.

 - Fix a typo in bpf_counter_cgroup.c.

* tag 'perf-tools-fixes-for-v5.17-2022-02-06' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux:
  perf ftrace: system_wide collection is not effective by default
  libperf: Add arm64 support to perf_mmap__read_self()
  tools include UAPI: Sync sound/asound.h copy with the kernel sources
  perf stat: Fix display of grouped aliased events
  perf tools: Apply correct label to user/kernel symbols in branch mode
  perf bpf: Fix a typo in bpf_counter_cgroup.c
  perf synthetic-events: Return error if procfs isn't mounted for PID namespaces
  perf session: Check for NULL pointer before dereference
  perf annotate: Set error stream of objdump process for TUI
  perf tools: Add missing branch_sample_type to perf_event_attr__fprintf()
  tools headers UAPI: Sync linux/kvm.h with the kernel sources
  tools headers UAPI: Sync linux/prctl.h with the kernel sources
  perf beauty: Make the prctl arg regexp more strict to cope with PR_SET_VMA
  tools headers cpufeatures: Sync with the kernel sources
  tools headers UAPI: Sync linux/perf_event.h with the kernel sources
  tools include UAPI: Sync sound/asound.h copy with the kernel sources

Merge tag 'perf_urgent_for_v5.17_rc3' of git://git./linux/kernel/git/tip/tip

Pull perf fixes from Borislav Petkov:

 - Intel/PT: filters could crash the kernel

 - Intel: default disable the PMU for SMM, some new-ish EFI firmware has
   started using CPL3 and the PMU CPL filters don't discriminate against
   SMM, meaning that CPL3 (userspace only) events now also count EFI/SMM
   cycles.

 - Fixup for perf_event_attr::sig_data

* tag 'perf_urgent_for_v5.17_rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  perf/x86/intel/pt: Fix crash with stop filters in single-range mode
  perf: uapi: Document perf_event_attr::sig_data truncation on 32 bit architectures
  selftests/perf_events: Test modification of perf_event_attr::sig_data
  perf: Copy perf_event_attr::sig_data on modification
  x86/perf: Default set FREEZE_ON_SMI for all

Merge tag 'objtool_urgent_for_v5.17_rc3' of git://git./linux/kernel/git/tip/tip

Pull objtool fix from Borislav Petkov:
 "Fix a potential truncated string warning triggered by gcc12"

* tag 'objtool_urgent_for_v5.17_rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  objtool: Fix truncated string warning

Merge tag 'irq_urgent_for_v5.17_rc3' of git://git./linux/kernel/git/tip/tip

Pull irq fix from Borislav Petkov:
 "Remove a bogus warning introduced by the recent PCI MSI irq affinity
  overhaul"

* tag 'irq_urgent_for_v5.17_rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  PCI/MSI: Remove bogus warning in pci_irq_get_affinity()

Merge tag 'edac_urgent_for_v5.17_rc3' of git://git./linux/kernel/git/ras/ras

Pull EDAC fixes from Borislav Petkov:
 "Fix altera and xgene EDAC drivers to propagate the correct error code
  from platform_get_irq() so that deferred probing still works"

* tag 'edac_urgent_for_v5.17_rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/ras/ras:
  EDAC/xgene: Fix deferred probing
  EDAC/altera: Fix deferred probing