[ Upstream commit 0e68b5517d3767562889f1d83fdb828c26adb24f ]
Running a rt-kernel base on 6.2.0-rc3-rt1 on an Ampere Altra outputs
the following:
BUG: sleeping function called from invalid context at kernel/locking/spinlock_rt.c:46
in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 9, name: kworker/u320:0
preempt_count: 2, expected: 0
RCU nest depth: 0, expected: 0
3 locks held by kworker/u320:0/9:
#0: ffff3fff8c27d128 ((wq_completion)efi_rts_wq){+.+.}-{0:0}, at: process_one_work (./include/linux/atomic/atomic-long.h:41)
#1: ffff80000861bdd0 ((work_completion)(&efi_rts_work.work)){+.+.}-{0:0}, at: process_one_work (./include/linux/atomic/atomic-long.h:41)
#2: ffffdf7e1ed3e460 (efi_rt_lock){+.+.}-{3:3}, at: efi_call_rts (drivers/firmware/efi/runtime-wrappers.c:101)
Preemption disabled at:
efi_virtmap_load (./arch/arm64/include/asm/mmu_context.h:248)
CPU: 0 PID: 9 Comm: kworker/u320:0 Tainted: G W 6.2.0-rc3-rt1
Hardware name: WIWYNN Mt.Jade Server System B81.03001.0005/Mt.Jade Motherboard, BIOS 1.08.20220218 (SCP: 1.08.20220218) 2022/02/18
Workqueue: efi_rts_wq efi_call_rts
Call trace:
dump_backtrace (arch/arm64/kernel/stacktrace.c:158)
show_stack (arch/arm64/kernel/stacktrace.c:165)
dump_stack_lvl (lib/dump_stack.c:107 (discriminator 4))
dump_stack (lib/dump_stack.c:114)
__might_resched (kernel/sched/core.c:10134)
rt_spin_lock (kernel/locking/rtmutex.c:1769 (discriminator 4))
efi_call_rts (drivers/firmware/efi/runtime-wrappers.c:101)
[...]
This seems to come from commit ff7a167961d1 ("arm64: efi: Execute
runtime services from a dedicated stack") which adds a spinlock. This
spinlock is taken through:
efi_call_rts()
\-efi_call_virt()
\-efi_call_virt_pointer()
\-arch_efi_call_virt_setup()
Make 'efi_rt_lock' a raw_spinlock to avoid being preempted.
[ardb: The EFI runtime services are called with a different set of
translation tables, and are permitted to use the SIMD registers.
The context switch code preserves/restores neither, and so EFI
calls must be made with preemption disabled, rather than only
disabling migration.]
Fixes: ff7a167961d1 ("arm64: efi: Execute runtime services from a dedicated stack")
Signed-off-by: Pierre Gondois <pierre.gondois@arm.com>
Cc: <stable@vger.kernel.org> # v6.1+
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit ff7a167961d1b97e0e205f245f806e564d3505e7 upstream. With the introduction of PRMT in the ACPI subsystem, the EFI rts workqueue is no longer the only caller of efi_call_virt_pointer() in the kernel. This means the EFI runtime services lock is no longer sufficient to manage concurrent calls into firmware, but also that firmware calls may occur that are not marshalled via the workqueue mechanism, but originate directly from the caller context. For added robustness, and to ensure that the runtime services have 8 KiB of stack space available as per the EFI spec, introduce a spinlock protected EFI runtime stack of 8 KiB, where the spinlock also ensures serialization between the EFI rts workqueue (which itself serializes EFI runtime calls) and other callers of efi_call_virt_pointer(). While at it, use the stack pivot to avoid reloading the shadow call stack pointer from the ordinary stack, as doing so could produce a gadget to defeat it. Signed-off-by: Ard Biesheuvel <ardb@kernel.org> Cc: Lee Jones <lee@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 031af50045ea97ed4386eb3751ca2c134d0fc911 ]
The inline assembly for arm64's cmpxchg_double*() implementations use a
+Q constraint to hazard against other accesses to the memory location
being exchanged. However, the pointer passed to the constraint is a
pointer to unsigned long, and thus the hazard only applies to the first
8 bytes of the location.
GCC can take advantage of this, assuming that other portions of the
location are unchanged, leading to a number of potential problems.
This is similar to what we fixed back in commit:
fee960bed5e857eb ("arm64: xchg: hazard against entire exchange variable")
... but we forgot to adjust cmpxchg_double*() similarly at the same
time.
The same problem applies, as demonstrated with the following test:
| struct big {
| u64 lo, hi;
| } __aligned(128);
|
| unsigned long foo(struct big *b)
| {
| u64 hi_old, hi_new;
|
| hi_old = b->hi;
| cmpxchg_double_local(&b->lo, &b->hi, 0x12, 0x34, 0x56, 0x78);
| hi_new = b->hi;
|
| return hi_old ^ hi_new;
| }
... which GCC 12.1.0 compiles as:
| 0000000000000000 <foo>:
| 0: d503233f paciasp
| 4: aa0003e4 mov x4, x0
| 8: 1400000e b 40 <foo+0x40>
| c: d2800240 mov x0, #0x12 // #18
| 10: d2800681 mov x1, #0x34 // #52
| 14: aa0003e5 mov x5, x0
| 18: aa0103e6 mov x6, x1
| 1c: d2800ac2 mov x2, #0x56 // #86
| 20: d2800f03 mov x3, #0x78 // #120
| 24: 48207c82 casp x0, x1, x2, x3, [x4]
| 28: ca050000 eor x0, x0, x5
| 2c: ca060021 eor x1, x1, x6
| 30: aa010000 orr x0, x0, x1
| 34: d2800000 mov x0, #0x0 // #0 <--- BANG
| 38: d50323bf autiasp
| 3c: d65f03c0 ret
| 40: d2800240 mov x0, #0x12 // #18
| 44: d2800681 mov x1, #0x34 // #52
| 48: d2800ac2 mov x2, #0x56 // #86
| 4c: d2800f03 mov x3, #0x78 // #120
| 50: f9800091 prfm pstl1strm, [x4]
| 54: c87f1885 ldxp x5, x6, [x4]
| 58: ca0000a5 eor x5, x5, x0
| 5c: ca0100c6 eor x6, x6, x1
| 60: aa0600a6 orr x6, x5, x6
| 64: b5000066 cbnz x6, 70 <foo+0x70>
| 68: c8250c82 stxp w5, x2, x3, [x4]
| 6c: 35ffff45 cbnz w5, 54 <foo+0x54>
| 70: d2800000 mov x0, #0x0 // #0 <--- BANG
| 74: d50323bf autiasp
| 78: d65f03c0 ret
Notice that at the lines with "BANG" comments, GCC has assumed that the
higher 8 bytes are unchanged by the cmpxchg_double() call, and that
`hi_old ^ hi_new` can be reduced to a constant zero, for both LSE and
LL/SC versions of cmpxchg_double().
This patch fixes the issue by passing a pointer to __uint128_t into the
+Q constraint, ensuring that the compiler hazards against the entire 16
bytes being modified.
With this change, GCC 12.1.0 compiles the above test as:
| 0000000000000000 <foo>:
| 0: f9400407 ldr x7, [x0, #8]
| 4: d503233f paciasp
| 8: aa0003e4 mov x4, x0
| c: 1400000f b 48 <foo+0x48>
| 10: d2800240 mov x0, #0x12 // #18
| 14: d2800681 mov x1, #0x34 // #52
| 18: aa0003e5 mov x5, x0
| 1c: aa0103e6 mov x6, x1
| 20: d2800ac2 mov x2, #0x56 // #86
| 24: d2800f03 mov x3, #0x78 // #120
| 28: 48207c82 casp x0, x1, x2, x3, [x4]
| 2c: ca050000 eor x0, x0, x5
| 30: ca060021 eor x1, x1, x6
| 34: aa010000 orr x0, x0, x1
| 38: f9400480 ldr x0, [x4, #8]
| 3c: d50323bf autiasp
| 40: ca0000e0 eor x0, x7, x0
| 44: d65f03c0 ret
| 48: d2800240 mov x0, #0x12 // #18
| 4c: d2800681 mov x1, #0x34 // #52
| 50: d2800ac2 mov x2, #0x56 // #86
| 54: d2800f03 mov x3, #0x78 // #120
| 58: f9800091 prfm pstl1strm, [x4]
| 5c: c87f1885 ldxp x5, x6, [x4]
| 60: ca0000a5 eor x5, x5, x0
| 64: ca0100c6 eor x6, x6, x1
| 68: aa0600a6 orr x6, x5, x6
| 6c: b5000066 cbnz x6, 78 <foo+0x78>
| 70: c8250c82 stxp w5, x2, x3, [x4]
| 74: 35ffff45 cbnz w5, 5c <foo+0x5c>
| 78: f9400480 ldr x0, [x4, #8]
| 7c: d50323bf autiasp
| 80: ca0000e0 eor x0, x7, x0
| 84: d65f03c0 ret
... sampling the high 8 bytes before and after the cmpxchg, and
performing an EOR, as we'd expect.
For backporting, I've tested this atop linux-4.9.y with GCC 5.5.0. Note
that linux-4.9.y is oldest currently supported stable release, and
mandates GCC 5.1+. Unfortunately I couldn't get a GCC 5.1 binary to run
on my machines due to library incompatibilities.
I've also used a standalone test to check that we can use a __uint128_t
pointer in a +Q constraint at least as far back as GCC 4.8.5 and LLVM
3.9.1.
Fixes: 5284e1b4bc8a ("arm64: xchg: Implement cmpxchg_double")
Fixes: e9a4b795652f ("arm64: cmpxchg_dbl: patch in lse instructions when supported by the CPU")
Reported-by: Boqun Feng <boqun.feng@gmail.com>
Link: https://lore.kernel.org/lkml/Y6DEfQXymYVgL3oJ@boqun-archlinux/
Reported-by: Peter Zijlstra <peterz@infradead.org>
Link: https://lore.kernel.org/lkml/Y6GXoO4qmH9OIZ5Q@hirez.programming.kicks-ass.net/
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Cc: stable@vger.kernel.org
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Steve Capper <steve.capper@arm.com>
Cc: Will Deacon <will@kernel.org>
Link: https://lore.kernel.org/r/20230104151626.3262137-1-mark.rutland@arm.com
Signed-off-by: Will Deacon <will@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit b2c3ccbd0011bb3b51d0fec24cb3a5812b1ec8ea ]
When CONFIG_ARM64_LSE_ATOMICS=y, each use of an LL/SC atomic results in
a fragment of code being generated in a subsection without a clear
association with its caller. A trampoline in the caller branches to the
LL/SC atomic with with a direct branch, and the atomic directly branches
back into its trampoline.
This breaks backtracing, as any PC within the out-of-line fragment will
be symbolized as an offset from the nearest prior symbol (which may not
be the function using the atomic), and since the atomic returns with a
direct branch, the caller's PC may be missing from the backtrace.
For example, with secondary_start_kernel() hacked to contain
atomic_inc(NULL), the resulting exception can be reported as being taken
from cpus_are_stuck_in_kernel():
| Unable to handle kernel NULL pointer dereference at virtual address 0000000000000000
| Mem abort info:
| ESR = 0x0000000096000004
| EC = 0x25: DABT (current EL), IL = 32 bits
| SET = 0, FnV = 0
| EA = 0, S1PTW = 0
| FSC = 0x04: level 0 translation fault
| Data abort info:
| ISV = 0, ISS = 0x00000004
| CM = 0, WnR = 0
| [0000000000000000] user address but active_mm is swapper
| Internal error: Oops: 96000004 [#1] PREEMPT SMP
| Modules linked in:
| CPU: 1 PID: 0 Comm: swapper/1 Not tainted 5.19.0-11219-geb555cb5b794-dirty #3
| Hardware name: linux,dummy-virt (DT)
| pstate: 60000005 (nZCv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--)
| pc : cpus_are_stuck_in_kernel+0xa4/0x120
| lr : secondary_start_kernel+0x164/0x170
| sp : ffff80000a4cbe90
| x29: ffff80000a4cbe90 x28: 0000000000000000 x27: 0000000000000000
| x26: 0000000000000000 x25: 0000000000000000 x24: 0000000000000000
| x23: 0000000000000000 x22: 0000000000000000 x21: 0000000000000000
| x20: 0000000000000001 x19: 0000000000000001 x18: 0000000000000008
| x17: 3030383832343030 x16: 3030303030307830 x15: ffff80000a4cbab0
| x14: 0000000000000001 x13: 5d31666130663133 x12: 3478305b20313030
| x11: 3030303030303078 x10: 3020726f73736563 x9 : 726f737365636f72
| x8 : ffff800009ff2ef0 x7 : 0000000000000003 x6 : 0000000000000000
| x5 : 0000000000000000 x4 : 0000000000000000 x3 : 0000000000000100
| x2 : 0000000000000000 x1 : ffff0000029bd880 x0 : 0000000000000000
| Call trace:
| cpus_are_stuck_in_kernel+0xa4/0x120
| __secondary_switched+0xb0/0xb4
| Code: 35ffffa3 17fffc6c d53cd040 f9800011 (885f7c01)
| ---[ end trace 0000000000000000 ]---
This is confusing and hinders debugging, and will be problematic for
CONFIG_LIVEPATCH as these cases cannot be unwound reliably.
This is very similar to recent issues with out-of-line exception fixups,
which were removed in commits:
35d67794b8828333 ("arm64: lib: __arch_clear_user(): fold fixups into body")
4012e0e22739eef9 ("arm64: lib: __arch_copy_from_user(): fold fixups into body")
139f9ab73d60cf76 ("arm64: lib: __arch_copy_to_user(): fold fixups into body")
When the trampolines were introduced in commit:
addfc38672c73efd ("arm64: atomics: avoid out-of-line ll/sc atomics")
The rationale was to improve icache performance by grouping the LL/SC
atomics together. This has never been measured, and this theoretical
benefit is outweighed by other factors:
* As the subsections are collapsed into sections at object file
granularity, these are spread out throughout the kernel and can share
cachelines with unrelated code regardless.
* GCC 12.1.0 has been observed to place the trampoline out-of-line in
specialised __ll_sc_*() functions, introducing more branching than was
intended.
* Removing the trampolines has been observed to shrink a defconfig
kernel Image by 64KiB when building with GCC 12.1.0.
This patch removes the LL/SC trampolines, meaning that the LL/SC atomics
will be inlined into their callers (or placed in out-of line functions
using regular BL/RET pairs). When CONFIG_ARM64_LSE_ATOMICS=y, the LL/SC
atomics are always called in an unlikely branch, and will be placed in a
cold portion of the function, so this should have minimal impact to the
hot paths.
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>
Cc: Will Deacon <will@kernel.org>
Link: https://lore.kernel.org/r/20220817155914.3975112-2-mark.rutland@arm.com
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Stable-dep-of: 031af50045ea ("arm64: cmpxchg_double*: hazard against entire exchange variable")
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 8e6082e94aac6d0338883b5953631b662a5a9188 ]
The code for the atomic ops is formatted inconsistently, and while this
is not a functional problem it is rather distracting when working on
them.
Some have ops have consistent indentation, e.g.
| #define ATOMIC_OP_ADD_RETURN(name, mb, cl...) \
| static inline int __lse_atomic_add_return##name(int i, atomic_t *v) \
| { \
| u32 tmp; \
| \
| asm volatile( \
| __LSE_PREAMBLE \
| " ldadd" #mb " %w[i], %w[tmp], %[v]\n" \
| " add %w[i], %w[i], %w[tmp]" \
| : [i] "+r" (i), [v] "+Q" (v->counter), [tmp] "=&r" (tmp) \
| : "r" (v) \
| : cl); \
| \
| return i; \
| }
While others have negative indentation for some lines, and/or have
misaligned trailing backslashes, e.g.
| static inline void __lse_atomic_##op(int i, atomic_t *v) \
| { \
| asm volatile( \
| __LSE_PREAMBLE \
| " " #asm_op " %w[i], %[v]\n" \
| : [i] "+r" (i), [v] "+Q" (v->counter) \
| : "r" (v)); \
| }
This patch makes the indentation consistent and also aligns the trailing
backslashes. This makes the code easier to read for those (like myself)
who are easily distracted by these inconsistencies.
This is intended as a cleanup.
There should be no functional change as a result of this patch.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Cc: Boqun Feng <boqun.feng@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Will Deacon <will@kernel.org>
Acked-by: Will Deacon <will@kernel.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20211210151410.2782645-2-mark.rutland@arm.com
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Stable-dep-of: 031af50045ea ("arm64: cmpxchg_double*: hazard against entire exchange variable")
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit 406504c7b0405d74d74c15a667cd4c4620c3e7a9 upstream.
A recent development on the EFI front has resulted in guests having
their page tables baked in the firmware binary, and mapped into the
IPA space as part of a read-only memslot. Not only is this legitimate,
but it also results in added security, so thumbs up.
It is possible to take an S1PTW translation fault if the S1 PTs are
unmapped at stage-2. However, KVM unconditionally treats S1PTW as a
write to correctly handle hardware AF/DB updates to the S1 PTs.
Furthermore, KVM injects an exception into the guest for S1PTW writes.
In the aforementioned case this results in the guest taking an abort
it won't recover from, as the S1 PTs mapping the vectors suffer from
the same problem.
So clearly our handling is... wrong.
Instead, switch to a two-pronged approach:
- On S1PTW translation fault, handle the fault as a read
- On S1PTW permission fault, handle the fault as a write
This is of no consequence to SW that *writes* to its PTs (the write
will trigger a non-S1PTW fault), and SW that uses RO PTs will not
use HW-assisted AF/DB anyway, as that'd be wrong.
Only in the case described in c4ad98e4b72c ("KVM: arm64: Assume write
fault on S1PTW permission fault on instruction fetch") do we end-up
with two back-to-back faults (page being evicted and faulted back).
I don't think this is a case worth optimising for.
Fixes: c4ad98e4b72c ("KVM: arm64: Assume write fault on S1PTW permission fault on instruction fetch")
Reviewed-by: Oliver Upton <oliver.upton@linux.dev>
Reviewed-by: Ard Biesheuvel <ardb@kernel.org>
Regression-tested-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit d8c1d798a2e5091128c391c6dadcc9be334af3f5 ] We use is_ttbr0_addr() in noinstr code, but as it's only marked as inline, it's theoretically possible for the compiler to place it out-of-line and instrument it, which would be problematic. Mark is_ttbr0_addr() as __always_inline such that that can safely be used from noinstr code. For consistency, do the same to is_ttbr1_addr(). Note that while is_ttbr1_addr() calls arch_kasan_reset_tag(), this is a macro (and its callees are either macros or __always_inline), so there is not a risk of transient instrumentation. Signed-off-by: Mark Rutland <mark.rutland@arm.com> Cc: Catalin Marinas <catalin.marinas@arm.com> Cc: Will Deacon <will@kernel.org> Link: https://lore.kernel.org/r/20221114144042.3001140-1-mark.rutland@arm.com Signed-off-by: Will Deacon <will@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 8d56e5c5a99ce1d17d39ce5a8260e42c2a2d7682 ]
In the initial release of the ARM Architecture Reference Manual for
ARMv8-A, the ESR_ELx registers were defined as 32-bit registers. This
changed in 2018 with version D.a (ARM DDI 0487D.a) of the architecture,
when they became 64-bit registers, with bits [63:32] defined as RES0. In
version G.a, a new field was added to ESR_ELx, ISS2, which covers bits
[36:32]. This field is used when the Armv8.7 extension FEAT_LS64 is
implemented.
As a result of the evolution of the register width, Linux stores it as
both a 64-bit value and a 32-bit value, which hasn't affected correctness
so far as Linux only uses the lower 32 bits of the register.
Make the register type consistent and always treat it as 64-bit wide. The
register is redefined as an "unsigned long", which is an unsigned
double-word (64-bit quantity) for the LP64 machine (aapcs64 [1], Table 1,
page 14). The type was chosen because "unsigned int" is the most frequent
type for ESR_ELx and because FAR_ELx, which is used together with ESR_ELx
in exception handling, is also declared as "unsigned long". The 64-bit type
also makes adding support for architectural features that use fields above
bit 31 easier in the future.
The KVM hypervisor will receive a similar update in a subsequent patch.
[1] https://github.com/ARM-software/abi-aa/releases/download/2021Q3/aapcs64.pdf
Signed-off-by: Alexandru Elisei <alexandru.elisei@arm.com>
Reviewed-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20220425114444.368693-4-alexandru.elisei@arm.com
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Stable-dep-of: 0bb1fbffc631 ("arm64: mm: kfence: only handle translation faults")
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit acfc35cfcee5df419391671ef1a631f43feee4e3 ]
Add the same change for ARM64 as done in the commit 9440c4294160
("x86/syscall: Include asm/ptrace.h in syscall_wrapper header") to
make sure all syscalls see 'struct pt_regs' definition and resulted
BTF for '__arm64_sys_*(struct pt_regs *regs)' functions point to
actual struct.
Without this patch, the BPF verifier refuses to load a tracing prog
which accesses pt_regs.
bpf(BPF_PROG_LOAD, {prog_type=0x1a, ...}, 128) = -1 EACCES
With this patch, we can see the correct error, which saves us time
in debugging the prog.
bpf(BPF_PROG_LOAD, {prog_type=0x1a, ...}, 128) = 4
bpf(BPF_RAW_TRACEPOINT_OPEN, {raw_tracepoint={name=NULL, prog_fd=4}}, 128) = -1 ENOTSUPP
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/r/20221031215728.50389-1-kuniyu@amazon.com
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 8ec8490a1950efeccb00967698cf7cb2fcd25ca7 ]
CONFIG_UBSAN_SHIFT with gcc-5 complains that the shifting of
ARM_CPU_IMP_AMPERE (0xC0) into bits [31:24] by MIDR_CPU_MODEL() is
undefined behavior. Well, sort of, it actually spells the error as:
arch/arm64/kernel/proton-pack.c: In function 'spectre_bhb_loop_affected':
arch/arm64/include/asm/cputype.h:44:2: error: initializer element is not constant
(((imp) << MIDR_IMPLEMENTOR_SHIFT) | \
^
This isn't an issue for other Implementor codes, as all the other codes
have zero in the top bit and so are representable as a signed int.
Cast the implementor code to unsigned in MIDR_CPU_MODEL to remove the
undefined behavior.
Fixes: 0e5d5ae837c8 ("arm64: Add AMPERE1 to the Spectre-BHB affected list")
Reported-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: D Scott Phillips <scott@os.amperecomputing.com>
Link: https://lore.kernel.org/r/20221102160106.1096948-1-scott@os.amperecomputing.com
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>