Patch series "huge vmalloc mappings", v13.
The kernel virtual mapping layer grew support for mapping memory with >
PAGE_SIZE ptes with commit 0ddab1d2ed ("lib/ioremap.c: add huge I/O
map capability interfaces"), and implemented support for using those
huge page mappings with ioremap.
According to the submission, the use-case is mapping very large
non-volatile memory devices, which could be GB or TB:
https://lore.kernel.org/lkml/1425404664-19675-1-git-send-email-toshi.kani@hp.com/
The benefit is said to be in the overhead of maintaining the mapping,
perhaps both in memory overhead and setup / teardown time. Memory
overhead for the mapping with a 4kB page and 8 byte page table is 2GB
per TB of mapping, down to 4MB / TB with 2MB pages.
The same huge page vmap infrastructure can be quite easily adapted and
used for mapping vmalloc memory pages without more complexity for arch
or core vmap code. However unlike ioremap, vmalloc page table overhead
is not a real problem, so the advantage to justify this is performance.
Several of the most structures in the kernel (e.g., vfs and network hash
tables) are allocated with vmalloc on NUMA machines, in order to
distribute access bandwidth over the machine. Mapping these with larger
pages can improve TLB usage significantly, for example this reduces TLB
misses by nearly 30x on a `git diff` workload on a 2-node POWER9 (59,800
-> 2,100) and reduces CPU cycles by 0.54%, due to vfs hashes being
allocated with 2MB pages.
[ Other numbers?
- The difference is even larger in a guest due to more costly TLB
misses.
- Eric Dumazet was keen on the network hash performance possibilities.
- Other archs? Ding was doing x86 testing. ]
The kernel module allocator also uses vmalloc to map module images even on
non-NUMA, which can result in high iTLB pressure on highly modular distro
type of kernels. This series does not implement huge mappings for modules
yet, but it's a step along the way. Rick Edgecombe was looking at that
IIRC.
The per-cpu allocator similarly might be able to take advantage of this.
Also on the todo list.
The disadvantages of this I can see are:
* Memory fragmentation can waste some physical memory because it will
attempt to allocate larger pages to fit the required size, rounding up
(once the requested size is >= 2MB).
- I don't see it being a big problem in practice unless some user
crops up that allocates thousands of 2.5MB ranges. We can tewak
heuristics a bit there if needed to reduce peak waste.
* Less granular mappings can make the NUMA distribution less balanced.
- Similar to the above.
- Could also allocate all major system hashes with one allocation
up-front and spread them all across the one block, which should help
overall NUMA distribution and reduce fragmentation waste.
* Callers might expect something about the underlying allocated pages.
- Tried to keep the apperance of base PAGE_SIZE pages throughout the
APIs and exposed data structures.
- Added a VM_NO_HUGE_VMAP flag to hammer troublesome cases with.
- Finally, added a nohugevmalloc boot option to turn it off (independent
of nohugeiomap).
This patch (of 14):
ARM uses its own PMD folding scheme which is missing pud_page which should
just pass through to pmd_page. Move this from the 3-level page table to
common header.
Link: https://lkml.kernel.org/r/20210317062402.533919-2-npiggin@gmail.com
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Cc: Russell King <linux@armlinux.org.uk>
Cc: Ding Tianhong <dingtianhong@huawei.com>
Cc: Uladzislau Rezki (Sony) <urezki@gmail.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Will Deacon <will@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Pull locking updates from Ingo Molnar:
- rtmutex cleanup & spring cleaning pass that removes ~400 lines of
code
- Futex simplifications & cleanups
- Add debugging to the CSD code, to help track down a tenacious race
(or hw problem)
- Add lockdep_assert_not_held(), to allow code to require a lock to not
be held, and propagate this into the ath10k driver
- Misc LKMM documentation updates
- Misc KCSAN updates: cleanups & documentation updates
- Misc fixes and cleanups
- Fix locktorture bugs with ww_mutexes
* tag 'locking-core-2021-04-28' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (44 commits)
kcsan: Fix printk format string
static_call: Relax static_call_update() function argument type
static_call: Fix unused variable warn w/o MODULE
locking/rtmutex: Clean up signal handling in __rt_mutex_slowlock()
locking/rtmutex: Restrict the trylock WARN_ON() to debug
locking/rtmutex: Fix misleading comment in rt_mutex_postunlock()
locking/rtmutex: Consolidate the fast/slowpath invocation
locking/rtmutex: Make text section and inlining consistent
locking/rtmutex: Move debug functions as inlines into common header
locking/rtmutex: Decrapify __rt_mutex_init()
locking/rtmutex: Remove pointless CONFIG_RT_MUTEXES=n stubs
locking/rtmutex: Inline chainwalk depth check
locking/rtmutex: Move rt_mutex_debug_task_free() to rtmutex.c
locking/rtmutex: Remove empty and unused debug stubs
locking/rtmutex: Consolidate rt_mutex_init()
locking/rtmutex: Remove output from deadlock detector
locking/rtmutex: Remove rtmutex deadlock tester leftovers
locking/rtmutex: Remove rt_mutex_timed_lock()
MAINTAINERS: Add myself as futex reviewer
locking/mutex: Remove repeated declaration
...
Pull xen updates from Juergen Gross:
- remove some PV ACPI cpu/memory hotplug code which has been broken for
a long time
- support direct mapped guests (other than dom0) on Arm
- several small fixes and cleanups
* tag 'for-linus-5.13-rc1-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip:
xen/arm: introduce XENFEAT_direct_mapped and XENFEAT_not_direct_mapped
xen-pciback: simplify vpci's find hook
xen-blkfront: Fix 'physical' typos
xen-blkback: fix compatibility bug with single page rings
xen: Remove support for PV ACPI cpu/memory hotplug
xen/pciback: Fix incorrect type warnings
Newer Xen versions expose two Xen feature flags to tell us if the domain
is directly mapped or not. Only when a domain is directly mapped it
makes sense to enable swiotlb-xen on ARM.
Introduce a function on ARM to check the new Xen feature flags and also
to deal with the legacy case. Call the function xen_swiotlb_detect.
Signed-off-by: Stefano Stabellini <stefano.stabellini@xilinx.com>
Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Link: https://lore.kernel.org/r/20210319200140.12512-1-sstabellini@kernel.org
Signed-off-by: Juergen Gross <jgross@suse.com>
The time pvops functions are the only ones left which might be
used in 32-bit mode and which return a 64-bit value.
Switch them to use the static_call() mechanism instead of pvops, as
this allows quite some simplification of the pvops implementation.
Signed-off-by: Juergen Gross <jgross@suse.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20210311142319.4723-5-jgross@suse.com
Pull ARM updates from Russell King:
- Generalise byte swapping assembly
- Update debug addresses for STI
- Validate start of physical memory with DTB
- Do not clear SCTLR.nTLSMD in decompressor
- amba/locomo/sa1111 devices remove method return type is void
- address markers for KASAN in page table dump
* tag 'for-linus' of git://git.armlinux.org.uk/~rmk/linux-arm:
ARM: 9065/1: OABI compat: fix build when EPOLL is not enabled
ARM: 9055/1: mailbox: arm_mhuv2: make remove callback return void
amba: Make use of bus_type functions
amba: Make the remove callback return void
vfio: platform: simplify device removal
amba: reorder functions
amba: Fix resource leak for drivers without .remove
ARM: 9054/1: arch/arm/mm/mmu.c: Remove duplicate header
ARM: 9053/1: arm/mm/ptdump:Add address markers for KASAN regions
ARM: 9051/1: vdso: remove unneded extra-y addition
ARM: 9050/1: Kconfig: Select ARCH_HAVE_NMI_SAFE_CMPXCHG where possible
ARM: 9049/1: locomo: make locomo bus's remove callback return void
ARM: 9048/1: sa1111: make sa1111 bus's remove callback return void
ARM: 9047/1: smp: remove unused variable
ARM: 9046/1: decompressor: Do not clear SCTLR.nTLSMD for ARMv7+ cores
ARM: 9045/1: uncompress: Validate start of physical memory against passed DTB
ARM: 9042/1: debug: no uncompress debugging while semihosting
ARM: 9041/1: sti LL_UART: add STiH418 SBC UART0 support
ARM: 9040/1: use DEBUG_UART_PHYS and DEBUG_UART_VIRT for sti LL_UART
ARM: 9039/1: assembler: generalize byte swapping macro into rev_l
Pull arm64 updates from Will Deacon:
- vDSO build improvements including support for building with BSD.
- Cleanup to the AMU support code and initialisation rework to support
cpufreq drivers built as modules.
- Removal of synthetic frame record from exception stack when entering
the kernel from EL0.
- Add support for the TRNG firmware call introduced by Arm spec
DEN0098.
- Cleanup and refactoring across the board.
- Avoid calling arch_get_random_seed_long() from
add_interrupt_randomness()
- Perf and PMU updates including support for Cortex-A78 and the v8.3
SPE extensions.
- Significant steps along the road to leaving the MMU enabled during
kexec relocation.
- Faultaround changes to initialise prefaulted PTEs as 'old' when
hardware access-flag updates are supported, which drastically
improves vmscan performance.
- CPU errata updates for Cortex-A76 (#1463225) and Cortex-A55
(#1024718)
- Preparatory work for yielding the vector unit at a finer granularity
in the crypto code, which in turn will one day allow us to defer
softirq processing when it is in use.
- Support for overriding CPU ID register fields on the command-line.
* tag 'arm64-upstream' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux: (85 commits)
drivers/perf: Replace spin_lock_irqsave to spin_lock
mm: filemap: Fix microblaze build failure with 'mmu_defconfig'
arm64: Make CPU_BIG_ENDIAN depend on ld.bfd or ld.lld 13.0.0+
arm64: cpufeatures: Allow disabling of Pointer Auth from the command-line
arm64: Defer enabling pointer authentication on boot core
arm64: cpufeatures: Allow disabling of BTI from the command-line
arm64: Move "nokaslr" over to the early cpufeature infrastructure
KVM: arm64: Document HVC_VHE_RESTART stub hypercall
arm64: Make kvm-arm.mode={nvhe, protected} an alias of id_aa64mmfr1.vh=0
arm64: Add an aliasing facility for the idreg override
arm64: Honor VHE being disabled from the command-line
arm64: Allow ID_AA64MMFR1_EL1.VH to be overridden from the command line
arm64: cpufeature: Add an early command-line cpufeature override facility
arm64: Extract early FDT mapping from kaslr_early_init()
arm64: cpufeature: Use IDreg override in __read_sysreg_by_encoding()
arm64: cpufeature: Add global feature override facility
arm64: Move SCTLR_EL1 initialisation to EL-agnostic code
arm64: Simplify init_el2_state to be non-VHE only
arm64: Move VHE-specific SPE setup to mutate_to_vhe()
arm64: Drop early setting of MDSCR_EL2.TPMS
...
Pull ARM fixes from Russell King:
- Fix latent bug with DC21285 (Footbridge PCI bridge) configuration
accessors that affects GCC >= 4.9.2
- Fix misplaced tegra_uart_config in decompressor
- Ensure signal page contents are initialised
- Fix kexec oops
* tag 'for-linus' of git://git.armlinux.org.uk/~rmk/linux-arm:
ARM: kexec: fix oops after TLB are invalidated
ARM: ensure the signal page contains defined contents
ARM: 9043/1: tegra: Fix misplaced tegra_uart_config in decompressor
ARM: footbridge: fix dc21285 PCI configuration accessors
Giancarlo Ferrari reports the following oops while trying to use kexec:
Unable to handle kernel paging request at virtual address 80112f38
pgd = fd7ef03e
[80112f38] *pgd=0001141e(bad)
Internal error: Oops: 80d [#1] PREEMPT SMP ARM
...
This is caused by machine_kexec() trying to set the kernel text to be
read/write, so it can poke values into the relocation code before
copying it - and an interrupt occuring which changes the page tables.
The subsequent writes then hit read-only sections that trigger a
data abort resulting in the above oops.
Fix this by copying the relocation code, and then writing the variables
into the destination, thereby avoiding the need to make the kernel text
read/write.
Reported-by: Giancarlo Ferrari <giancarlo.ferrari89@gmail.com>
Tested-by: Giancarlo Ferrari <giancarlo.ferrari89@gmail.com>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Take the 4 instruction byte swapping sequence from the decompressor's
head.S, and turn it into a rev_l GAS macro for general use. While
at it, make it use the 'rev' instruction when compiling for v6 or
later.
Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be>
Tested-by: Geert Uytterhoeven <geert+renesas@glider.be>
Reviewed-by: Nicolas Pitre <nico@fluxnic.net>
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
The ARM DEN0098 document describe an SMCCC based firmware service to
deliver hardware generated random numbers. Its existence is advertised
according to the SMCCC v1.1 specification.
Add a (dummy) call to probe functions implemented in each architecture
(ARM and arm64), to determine the existence of this interface.
For now this return false, but this will be overwritten by each
architecture's support patch.
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Reviewed-by: Sudeep Holla <sudeep.holla@arm.com>
Signed-off-by: Will Deacon <will@kernel.org>
Pull EFI updates from Borislav Petkov:
"These got delayed due to a last minute ia64 build issue which got
fixed in the meantime.
EFI updates collected by Ard Biesheuvel:
- Don't move BSS section around pointlessly in the x86 decompressor
- Refactor helper for discovering the EFI secure boot mode
- Wire up EFI secure boot to IMA for arm64
- Some fixes for the capsule loader
- Expose the RT_PROP table via the EFI test module
- Relax DT and kernel placement restrictions on ARM
with a few followup fixes:
- fix the build breakage on IA64 caused by recent capsule loader
changes
- suppress a type mismatch build warning in the expansion of
EFI_PHYS_ALIGN on ARM"
* tag 'efi_updates_for_v5.11' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
efi: arm: force use of unsigned type for EFI_PHYS_ALIGN
efi: ia64: disable the capsule loader
efi: stub: get rid of efi_get_max_fdt_addr()
efi/efi_test: read RuntimeServicesSupported
efi: arm: reduce minimum alignment of uncompressed kernel
efi: capsule: clean scatter-gather entries from the D-cache
efi: capsule: use atomic kmap for transient sglist mappings
efi: x86/xen: switch to efi_get_secureboot_mode helper
arm64/ima: add ima_arch support
ima: generalize x86/EFI arch glue for other EFI architectures
efi: generalize efi_get_secureboot
efi/libstub: EFI_GENERIC_STUB_INITRD_CMDLINE_LOADER should not default to yes
efi/x86: Only copy the compressed kernel image in efi_relocate_kernel()
efi/libstub/x86: simplify efi_is_native()
Pull ARM updates from Russell King:
- Rework phys/virt translation
- Add KASan support
- Move DT out of linear map region
- Use more PC-relative addressing in assembly
- Remove FP emulation handling while in kernel mode
- Link with '-z norelro'
- remove old check for GCC <= 4.2 in ARM unwinder code
- disable big endian if using clang's linker
* tag 'for-linus' of git://git.armlinux.org.uk/~rmk/linux-arm: (46 commits)
ARM: 9027/1: head.S: explicitly map DT even if it lives in the first physical section
ARM: 9038/1: Link with '-z norelro'
ARM: 9037/1: uncompress: Add OF_DT_MAGIC macro
ARM: 9036/1: uncompress: Fix dbgadtb size parameter name
ARM: 9035/1: uncompress: Add be32tocpu macro
ARM: 9033/1: arm/smp: Drop the macro S(x,s)
ARM: 9032/1: arm/mm: Convert PUD level pgtable helper macros into functions
ARM: 9031/1: hyp-stub: remove unused .L__boot_cpu_mode_offset symbol
ARM: 9044/1: vfp: use undef hook for VFP support detection
ARM: 9034/1: __div64_32(): straighten up inline asm constraints
ARM: 9030/1: entry: omit FP emulation for UND exceptions taken in kernel mode
ARM: 9029/1: Make iwmmxt.S support Clang's integrated assembler
ARM: 9028/1: disable KASAN in call stack capturing routines
ARM: 9026/1: unwind: remove old check for GCC <= 4.2
ARM: 9025/1: Kconfig: CPU_BIG_ENDIAN depends on !LD_IS_LLD
ARM: 9024/1: Drop useless cast of "u64" to "long long"
ARM: 9023/1: Spelling s/mmeory/memory/
ARM: 9022/1: Change arch/arm/lib/mem*.S to use WEAK instead of .weak
ARM: kvm: replace open coded VA->PA calculations with adr_l call
ARM: head.S: use PC relative insn sequence to calculate PHYS_OFFSET
...
Pull RISC-V updates from Palmer Dabbelt:
"We have a handful of new kernel features for 5.11:
- Support for the contiguous memory allocator.
- Support for IRQ Time Accounting
- Support for stack tracing
- Support for strict /dev/mem
- Support for kernel section protection
I'm being a bit conservative on the cutoff for this round due to the
timing, so this is all the new development I'm going to take for this
cycle (even if some of it probably normally would have been OK). There
are, however, some fixes on the list that I will likely be sending
along either later this week or early next week.
There is one issue in here: one of my test configurations
(PREEMPT{,_DEBUG}=y) fails to boot on QEMU 5.0.0 (from April) as of
the .text.init alignment patch.
With any luck we'll sort out the issue, but given how many bugs get
fixed all over the place and how unrelated those features seem my
guess is that we're just running into something that's been lurking
for a while and has already been fixed in the newer QEMU (though I
wouldn't be surprised if it's one of these implicit assumptions we
have in the boot flow). If it was hardware I'd be strongly inclined to
look more closely, but given that users can upgrade their simulators
I'm less worried about it"
* tag 'riscv-for-linus-5.11-mw0' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux:
arm64: Use the generic devmem_is_allowed()
arm: Use the generic devmem_is_allowed()
RISC-V: Use the new generic devmem_is_allowed()
lib: Add a generic version of devmem_is_allowed()
riscv: Fixed kernel test robot warning
riscv: kernel: Drop unused clean rule
riscv: provide memmove implementation
RISC-V: Move dynamic relocation section under __init
RISC-V: Protect all kernel sections including init early
RISC-V: Align the .init.text section
RISC-V: Initialize SBI early
riscv: Enable ARCH_STACKWALK
riscv: Make stack walk callback consistent with generic code
riscv: Cleanup stacktrace
riscv: Add HAVE_IRQ_TIME_ACCOUNTING
riscv: Enable CMA support
riscv: Ignore Image.* and loader.bin
riscv: Clean up boot dir
riscv: Fix compressed Image formats build
RISC-V: Add kernel image sections to the resource tree
Pull TIF_NOTIFY_SIGNAL updates from Jens Axboe:
"This sits on top of of the core entry/exit and x86 entry branch from
the tip tree, which contains the generic and x86 parts of this work.
Here we convert the rest of the archs to support TIF_NOTIFY_SIGNAL.
With that done, we can get rid of JOBCTL_TASK_WORK from task_work and
signal.c, and also remove a deadlock work-around in io_uring around
knowing that signal based task_work waking is invoked with the sighand
wait queue head lock.
The motivation for this work is to decouple signal notify based
task_work, of which io_uring is a heavy user of, from sighand. The
sighand lock becomes a huge contention point, particularly for
threaded workloads where it's shared between threads. Even outside of
threaded applications it's slower than it needs to be.
Roman Gershman <romger@amazon.com> reported that his networked
workload dropped from 1.6M QPS at 80% CPU to 1.0M QPS at 100% CPU
after io_uring was changed to use TIF_NOTIFY_SIGNAL. The time was all
spent hammering on the sighand lock, showing 57% of the CPU time there
[1].
There are further cleanups possible on top of this. One example is
TIF_PATCH_PENDING, where a patch already exists to use
TIF_NOTIFY_SIGNAL instead. Hopefully this will also lead to more
consolidation, but the work stands on its own as well"
[1] https://github.com/axboe/liburing/issues/215
* tag 'tif-task_work.arch-2020-12-14' of git://git.kernel.dk/linux-block: (28 commits)
io_uring: remove 'twa_signal_ok' deadlock work-around
kernel: remove checking for TIF_NOTIFY_SIGNAL
signal: kill JOBCTL_TASK_WORK
io_uring: JOBCTL_TASK_WORK is no longer used by task_work
task_work: remove legacy TWA_SIGNAL path
sparc: add support for TIF_NOTIFY_SIGNAL
riscv: add support for TIF_NOTIFY_SIGNAL
nds32: add support for TIF_NOTIFY_SIGNAL
ia64: add support for TIF_NOTIFY_SIGNAL
h8300: add support for TIF_NOTIFY_SIGNAL
c6x: add support for TIF_NOTIFY_SIGNAL
alpha: add support for TIF_NOTIFY_SIGNAL
xtensa: add support for TIF_NOTIFY_SIGNAL
arm: add support for TIF_NOTIFY_SIGNAL
microblaze: add support for TIF_NOTIFY_SIGNAL
hexagon: add support for TIF_NOTIFY_SIGNAL
csky: add support for TIF_NOTIFY_SIGNAL
openrisc: add support for TIF_NOTIFY_SIGNAL
sh: add support for TIF_NOTIFY_SIGNAL
um: add support for TIF_NOTIFY_SIGNAL
...
Pull asm-generic cross-architecture timer cleanup from Arnd Bergmann:
"This cleans up two ancient timer features that were never completed in
the past, CONFIG_GENERIC_CLOCKEVENTS and CONFIG_ARCH_USES_GETTIMEOFFSET.
There was only one user left for the ARCH_USES_GETTIMEOFFSET variant
of clocksource implementations, the ARM EBSA110 platform. Rather than
changing to use modern timekeeping, we remove the platform entirely as
Russell no longer uses his machine and nobody else seems to have one
any more.
The conditional code for using arch_gettimeoffset() is removed as a
result.
For CONFIG_GENERIC_CLOCKEVENTS, there are still a couple of platforms
not using clockevent drivers: parisc, ia64, most of m68k, and one Arm
platform. These all do timer ticks slighly differently, and this gets
cleaned up to the point they at least all call the same helper
function.
Instead of most platforms using 'select GENERIC_CLOCKEVENTS' in
Kconfig, the polarity is now reversed, with the few remaining ones
selecting LEGACY_TIMER_TICK instead"
* tag 'asm-generic-timers-5.11' of git://git.kernel.org/pub/scm/linux/kernel/git/arnd/asm-generic:
timekeeping: default GENERIC_CLOCKEVENTS to enabled
timekeeping: remove xtime_update
m68k: remove timer_interrupt() function
m68k: change remaining timers to legacy_timer_tick
m68k: m68328: use legacy_timer_tick()
m68k: sun3/sun3c: use legacy_timer_tick
m68k: split heartbeat out of timer function
m68k: coldfire: use legacy_timer_tick()
parisc: use legacy_timer_tick
ARM: rpc: use legacy_timer_tick
ia64: convert to legacy_timer_tick
timekeeping: add CONFIG_LEGACY_TIMER_TICK
timekeeping: remove arch_gettimeoffset
net: remove am79c961a driver
ARM: remove ebsa110 platform
Pull asm-generic mmu-context cleanup from Arnd Bergmann:
"This is a cleanup series from Nicholas Piggin, preparing for later
changes. The asm/mmu_context.h header are generalized and common code
moved to asm-gneneric/mmu_context.h.
This saves a bit of code and makes it easier to change in the future"
* tag 'asm-generic-mmu-context-5.11' of git://git.kernel.org/pub/scm/linux/kernel/git/arnd/asm-generic: (25 commits)
h8300: Fix generic mmu_context build
m68k: mmu_context: Fix Sun-3 build
xtensa: use asm-generic/mmu_context.h for no-op implementations
x86: use asm-generic/mmu_context.h for no-op implementations
um: use asm-generic/mmu_context.h for no-op implementations
sparc: use asm-generic/mmu_context.h for no-op implementations
sh: use asm-generic/mmu_context.h for no-op implementations
s390: use asm-generic/mmu_context.h for no-op implementations
riscv: use asm-generic/mmu_context.h for no-op implementations
powerpc: use asm-generic/mmu_context.h for no-op implementations
parisc: use asm-generic/mmu_context.h for no-op implementations
openrisc: use asm-generic/mmu_context.h for no-op implementations
nios2: use asm-generic/mmu_context.h for no-op implementations
nds32: use asm-generic/mmu_context.h for no-op implementations
mips: use asm-generic/mmu_context.h for no-op implementations
microblaze: use asm-generic/mmu_context.h for no-op implementations
m68k: use asm-generic/mmu_context.h for no-op implementations
ia64: use asm-generic/mmu_context.h for no-op implementations
hexagon: use asm-generic/mmu_context.h for no-op implementations
csky: use asm-generic/mmu_context.h for no-op implementations
...
Pull regset updates from Al Viro:
"Dead code removal, mostly.
The only exception is a bit of cleanups on itanic (getting rid of
redundant stack unwinds - each access_uarea() call does it and we call
that 7 times in a row in ptrace_[sg]etregs(), *after* having done it
ourselves in the caller; location where the user registers have been
spilled won't change under us, and we can bloody well just call
access_elf_reg() directly, giving it the unw_frame_info we'd
calculated for our own purposes)"
* 'regset.followup' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
c6x: kill ELF_CORE_COPY_FPREGS
whack-a-mole: USE_ELF_CORE_DUMP
[ia64] ptrace_[sg]etregs(): use access_elf_reg() instead of access_uarea()
[ia64] missed cleanups from switch to regset coredumps
arm: kill dump_task_regs()
Pull irq updates from Thomas Gleixner:
"Generic interrupt and irqchips subsystem updates. Unusually, there is
not a single completely new irq chip driver, just new DT bindings and
extensions of existing drivers to accomodate new variants!
Core:
- Consolidation and robustness changes for irq time accounting
- Cleanup and consolidation of irq stats
- Remove the fasteoi IPI flow which has been proved useless
- Provide an interface for converting legacy interrupt mechanism into
irqdomains
Drivers:
- Preliminary support for managed interrupts on platform devices
- Correctly identify allocation of MSIs proxyied by another device
- Generalise the Ocelot support to new SoCs
- Improve GICv4.1 vcpu entry, matching the corresponding KVM
optimisation
- Work around spurious interrupts on Qualcomm PDC
- Random fixes and cleanups"
* tag 'irq-core-2020-12-15' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (54 commits)
irqchip/qcom-pdc: Fix phantom irq when changing between rising/falling
driver core: platform: Add devm_platform_get_irqs_affinity()
ACPI: Drop acpi_dev_irqresource_disabled()
resource: Add irqresource_disabled()
genirq/affinity: Add irq_update_affinity_desc()
irqchip/gic-v3-its: Flag device allocation as proxied if behind a PCI bridge
irqchip/gic-v3-its: Tag ITS device as shared if allocating for a proxy device
platform-msi: Track shared domain allocation
irqchip/ti-sci-intr: Fix freeing of irqs
irqchip/ti-sci-inta: Fix printing of inta id on probe success
drivers/irqchip: Remove EZChip NPS interrupt controller
Revert "genirq: Add fasteoi IPI flow"
irqchip/hip04: Make IPIs use handle_percpu_devid_irq()
irqchip/bcm2836: Make IPIs use handle_percpu_devid_irq()
irqchip/armada-370-xp: Make IPIs use handle_percpu_devid_irq()
irqchip/gic, gic-v3: Make SGIs use handle_percpu_devid_irq()
irqchip/ocelot: Add support for Jaguar2 platforms
irqchip/ocelot: Add support for Serval platforms
irqchip/ocelot: Add support for Luton platforms
irqchip/ocelot: prepare to support more SoC
...
Pull irqchip updates for 5.11 from Marc Zyngier:
- Preliminary support for managed interrupts on platform devices
- Correctly identify allocation of MSIs proxyied by another device
- Remove the fasteoi IPI flow which has been proved useless
- Generalise the Ocelot support to new SoCs
- Improve GICv4.1 vcpu entry, matching the corresponding KVM optimisation
- Work around spurious interrupts on Qualcomm PDC
- Random fixes and cleanups
Link: https://lore.kernel.org/r/20201212135626.1479884-1-maz@kernel.org
Pull kmap updates from Thomas Gleixner:
"The new preemtible kmap_local() implementation:
- Consolidate all kmap_atomic() internals into a generic
implementation which builds the base for the kmap_local() API and
make the kmap_atomic() interface wrappers which handle the
disabling/enabling of preemption and pagefaults.
- Switch the storage from per-CPU to per task and provide scheduler
support for clearing mapping when scheduling out and restoring them
when scheduling back in.
- Merge the migrate_disable/enable() code, which is also part of the
scheduler pull request. This was required to make the kmap_local()
interface available which does not disable preemption when a
mapping is established. It has to disable migration instead to
guarantee that the virtual address of the mapped slot is the same
across preemption.
- Provide better debug facilities: guard pages and enforced
utilization of the mapping mechanics on 64bit systems when the
architecture allows it.
- Provide the new kmap_local() API which can now be used to cleanup
the kmap_atomic() usage sites all over the place. Most of the usage
sites do not require the implicit disabling of preemption and
pagefaults so the penalty on 64bit and 32bit non-highmem systems is
removed and quite some of the code can be simplified. A wholesale
conversion is not possible because some usage depends on the
implicit side effects and some need to be cleaned up because they
work around these side effects.
The migrate disable side effect is only effective on highmem
systems and when enforced debugging is enabled. On 64bit and 32bit
non-highmem systems the overhead is completely avoided"
* tag 'core-mm-2020-12-14' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (33 commits)
ARM: highmem: Fix cache_is_vivt() reference
x86/crashdump/32: Simplify copy_oldmem_page()
io-mapping: Provide iomap_local variant
mm/highmem: Provide kmap_local*
sched: highmem: Store local kmaps in task struct
x86: Support kmap_local() forced debugging
mm/highmem: Provide CONFIG_DEBUG_KMAP_LOCAL_FORCE_MAP
mm/highmem: Provide and use CONFIG_DEBUG_KMAP_LOCAL
microblaze/mm/highmem: Add dropped #ifdef back
xtensa/mm/highmem: Make generic kmap_atomic() work correctly
mm/highmem: Take kmap_high_get() properly into account
highmem: High implementation details and document API
Documentation/io-mapping: Remove outdated blurb
io-mapping: Cleanup atomic iomap
mm/highmem: Remove the old kmap_atomic cruft
highmem: Get rid of kmap_types.h
xtensa/mm/highmem: Switch to generic kmap atomic
sparc/mm/highmem: Switch to generic kmap atomic
powerpc/mm/highmem: Switch to generic kmap atomic
nds32/mm/highmem: Switch to generic kmap atomic
...
Pull arm64 updates from Catalin Marinas:
- Expose tag address bits in siginfo. The original arm64 ABI did not
expose any of the bits 63:56 of a tagged address in siginfo. In the
presence of user ASAN or MTE, this information may be useful. The
implementation is generic to other architectures supporting tags
(like SPARC ADI, subject to wiring up the arch code). The user will
have to opt in via sigaction(SA_EXPOSE_TAGBITS) so that the extra
bits, if available, become visible in si_addr.
- Default to 32-bit wide ZONE_DMA. Previously, ZONE_DMA was set to the
lowest 1GB to cope with the Raspberry Pi 4 limitations, to the
detriment of other platforms. With these changes, the kernel scans
the Device Tree dma-ranges and the ACPI IORT information before
deciding on a smaller ZONE_DMA.
- Strengthen READ_ONCE() to acquire when CONFIG_LTO=y. When building
with LTO, there is an increased risk of the compiler converting an
address dependency headed by a READ_ONCE() invocation into a control
dependency and consequently allowing for harmful reordering by the
CPU.
- Add CPPC FFH support using arm64 AMU counters.
- set_fs() removal on arm64. This renders the User Access Override
(UAO) ARMv8 feature unnecessary.
- Perf updates: PMU driver for the ARM DMC-620 memory controller, sysfs
identifier file for SMMUv3, stop event counters support for i.MX8MP,
enable the perf events-based hard lockup detector.
- Reorganise the kernel VA space slightly so that 52-bit VA
configurations can use more virtual address space.
- Improve the robustness of the arm64 memory offline event notifier.
- Pad the Image header to 64K following the EFI header definition
updated recently to increase the section alignment to 64K.
- Support CONFIG_CMDLINE_EXTEND on arm64.
- Do not use tagged PC in the kernel (TCR_EL1.TBID1==1), freeing up 8
bits for PtrAuth.
- Switch to vmapped shadow call stacks.
- Miscellaneous clean-ups.
* tag 'arm64-upstream' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux: (78 commits)
perf/imx_ddr: Add system PMU identifier for userspace
bindings: perf: imx-ddr: add compatible string
arm64: Fix build failure when HARDLOCKUP_DETECTOR_PERF is enabled
arm64: mte: fix prctl(PR_GET_TAGGED_ADDR_CTRL) if TCF0=NONE
arm64: mark __system_matches_cap as __maybe_unused
arm64: uaccess: remove vestigal UAO support
arm64: uaccess: remove redundant PAN toggling
arm64: uaccess: remove addr_limit_user_check()
arm64: uaccess: remove set_fs()
arm64: uaccess cleanup macro naming
arm64: uaccess: split user/kernel routines
arm64: uaccess: refactor __{get,put}_user
arm64: uaccess: simplify __copy_user_flushcache()
arm64: uaccess: rename privileged uaccess routines
arm64: sdei: explicitly simulate PAN/UAO entry
arm64: sdei: move uaccess logic to arch/arm64/
arm64: head.S: always initialize PSTATE
arm64: head.S: cleanup SCTLR_ELx initialization
arm64: head.S: rename el2_setup -> init_kernel_el
arm64: add C wrappers for SET_PSTATE_*()
...
As part of adding STRICT_DEVMEM support to the RISC-V port, Zong provided an
implementation of devmem_is_allowed() that's exactly the same as the version in
a handful of other ports. Rather than duplicate code, I've put a generic
version of this in lib/ and used it for the RISC-V port.
* palmer/generic-devmem:
arm64: Use the generic devmem_is_allowed()
arm: Use the generic devmem_is_allowed()
RISC-V: Use the new generic devmem_is_allowed()
lib: Add a generic version of devmem_is_allowed()
This is exactly the same as the arm64 version, which I recently copied
into lib/ for use by the RISC-V port.
Reviewed-by: Luis Chamberlain <mcgrof@kernel.org>
Signed-off-by: Palmer Dabbelt <palmerdabbelt@google.com>
Now that ARM started following the example of arm64 and RISC-V, and
no longer imposes any restrictions on the placement of the FDT in
memory at boot, we no longer need per-arch implementations of
efi_get_max_fdt_addr() to factor out the differences. So get rid of
it.
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Reviewed-by: Atish Patra <atish.patra@wdc.com>
Link: https://lore.kernel.org/r/20201029134901.9773-1-ardb@kernel.org
Now that we reduced the minimum relative alignment between PHYS_OFFSET
and PAGE_OFFSET to 2 MiB, we can take this into account when allocating
memory for the decompressed kernel when booting via EFI. This minimizes
the amount of unusable memory we may end up with due to the base of DRAM
being occupied by firmware.
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Scatter-gather lists passed to UpdateCapsule() should be cleaned
from the D-cache to ensure that they are visible to the CPU after a
warm reboot before the MMU is enabled. On ARM and arm64 systems, this
implies a D-cache clean by virtual address to the point of coherency.
However, due to the fact that the firmware itself is not able to map
physical addresses back to virtual addresses when running under the OS,
this must be done by the caller.
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
The ARM version of __div64_32() encapsulates a call to __do_div64 with
non-standard argument passing. In particular, __n is a 64-bit input
argument assigned to r0-r1 and __rem is an output argument sharing half
of that r0-r1 register pair.
With __n being an input argument, the compiler is in its right to
presume that r0-r1 would still hold the value of __n past the inline
assembly statement. Normally, the compiler would have assigned non
overlapping registers to __n and __rem if the value for __n is needed
again.
However, here we enforce our own register assignment and gcc fails to
notice the conflict. In practice this doesn't cause any problem as __n
is considered dead after the asm statement and *n is overwritten.
However this is not always guaranteed and clang rightfully complains.
Let's fix it properly by making __n into an input-output variable. This
makes it clear that those registers representing __n have been modified.
Then we can extract __rem as the high part of __n with plain C code.
This asm constraint "abuse" was likely relied upon back when gcc didn't
handle 64-bit values optimally. Turns out that gcc is now able to
optimize things and produces the same code with this patch applied.
Reported-by: Antony Yu <swpenim@gmail.com>
Signed-off-by: Nicolas Pitre <nico@fluxnic.net>
Reviewed-by: Ard Biesheuvel <ardb@kernel.org>
Tested-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Pull asm-generic fix from Arnd Bergmann:
"Add correct MAX_POSSIBLE_PHYSMEM_BITS setting to asm-generic.
This is a single bugfix for a bug that Stefan Agner found on 32-bit
Arm, but that exists on several other architectures"
* tag 'asm-generic-fixes-5.10-2' of git://git.kernel.org/pub/scm/linux/kernel/git/arnd/asm-generic:
arch: pgtable: define MAX_POSSIBLE_PHYSMEM_BITS where needed
Previously we were not clearing non-uapi flag bits in
sigaction.sa_flags when storing the userspace-provided sa_flags or
when returning them via oldact. Start doing so.
This allows userspace to detect missing support for flag bits and
allows the kernel to use non-uapi bits internally, as we are already
doing in arch/x86 for two flag bits. Now that this change is in
place, we no longer need the code in arch/x86 that was hiding these
bits from userspace, so remove it.
This is technically a userspace-visible behavior change for sigaction, as
the unknown bits returned via oldact.sa_flags are no longer set. However,
we are free to define the behavior for unknown bits exactly because
their behavior is currently undefined, so for now we can define the
meaning of each of them to be "clear the bit in oldact.sa_flags unless
the bit becomes known in the future". Furthermore, this behavior is
consistent with OpenBSD [1], illumos [2] and XNU [3] (FreeBSD [4] and
NetBSD [5] fail the syscall if unknown bits are set). So there is some
precedent for this behavior in other kernels, and in particular in XNU,
which is probably the most popular kernel among those that I looked at,
which means that this change is less likely to be a compatibility issue.
Link: [1] f634a6a4b5/sys/kern/kern_sig.c (L278)
Link: [2] 76f19f5fdc/usr/src/uts/common/syscall/sigaction.c (L86)
Link: [3] a449c6a3b8/bsd/kern/kern_sig.c (L480)
Link: [4] eded70c370/sys/kern/kern_sig.c (L699)
Link: [5] 3365779bec/sys/kern/sys_sig.c (L473)
Signed-off-by: Peter Collingbourne <pcc@google.com>
Reviewed-by: Dave Martin <Dave.Martin@arm.com>
Acked-by: "Eric W. Biederman" <ebiederm@xmission.com>
Link: https://linux-review.googlesource.com/id/I35aab6f5be932505d90f3b3450c083b4db1eca86
Link: https://lkml.kernel.org/r/878dbcb5f47bc9b11881c81f745c0bef5c23f97f.1605235762.git.pcc@google.com
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
To enable seccomp constant action bitmaps, we need to have a static
mapping to the audit architecture and system call table size. Add these
for arm.
Signed-off-by: Kees Cook <keescook@chromium.org>
Stefan Agner reported a bug when using zsram on 32-bit Arm machines
with RAM above the 4GB address boundary:
Unable to handle kernel NULL pointer dereference at virtual address 00000000
pgd = a27bd01c
[00000000] *pgd=236a0003, *pmd=1ffa64003
Internal error: Oops: 207 [#1] SMP ARM
Modules linked in: mdio_bcm_unimac(+) brcmfmac cfg80211 brcmutil raspberrypi_hwmon hci_uart crc32_arm_ce bcm2711_thermal phy_generic genet
CPU: 0 PID: 123 Comm: mkfs.ext4 Not tainted 5.9.6 #1
Hardware name: BCM2711
PC is at zs_map_object+0x94/0x338
LR is at zram_bvec_rw.constprop.0+0x330/0xa64
pc : [<c0602b38>] lr : [<c0bda6a0>] psr: 60000013
sp : e376bbe0 ip : 00000000 fp : c1e2921c
r10: 00000002 r9 : c1dda730 r8 : 00000000
r7 : e8ff7a00 r6 : 00000000 r5 : 02f9ffa0 r4 : e3710000
r3 : 000fdffe r2 : c1e0ce80 r1 : ebf979a0 r0 : 00000000
Flags: nZCv IRQs on FIQs on Mode SVC_32 ISA ARM Segment user
Control: 30c5383d Table: 235c2a80 DAC: fffffffd
Process mkfs.ext4 (pid: 123, stack limit = 0x495a22e6)
Stack: (0xe376bbe0 to 0xe376c000)
As it turns out, zsram needs to know the maximum memory size, which
is defined in MAX_PHYSMEM_BITS when CONFIG_SPARSEMEM is set, or in
MAX_POSSIBLE_PHYSMEM_BITS on the x86 architecture.
The same problem will be hit on all 32-bit architectures that have a
physical address space larger than 4GB and happen to not enable sparsemem
and include asm/sparsemem.h from asm/pgtable.h.
After the initial discussion, I suggested just always defining
MAX_POSSIBLE_PHYSMEM_BITS whenever CONFIG_PHYS_ADDR_T_64BIT is
set, or provoking a build error otherwise. This addresses all
configurations that can currently have this runtime bug, but
leaves all other configurations unchanged.
I looked up the possible number of bits in source code and
datasheets, here is what I found:
- on ARC, CONFIG_ARC_HAS_PAE40 controls whether 32 or 40 bits are used
- on ARM, CONFIG_LPAE enables 40 bit addressing, without it we never
support more than 32 bits, even though supersections in theory allow
up to 40 bits as well.
- on MIPS, some MIPS32r1 or later chips support 36 bits, and MIPS32r5
XPA supports up to 60 bits in theory, but 40 bits are more than
anyone will ever ship
- On PowerPC, there are three different implementations of 36 bit
addressing, but 32-bit is used without CONFIG_PTE_64BIT
- On RISC-V, the normal page table format can support 34 bit
addressing. There is no highmem support on RISC-V, so anything
above 2GB is unused, but it might be useful to eventually support
CONFIG_ZRAM for high pages.
Fixes: 61989a80fb ("staging: zsmalloc: zsmalloc memory allocation library")
Fixes: 02390b87a9 ("mm/zsmalloc: Prepare to variable MAX_PHYSMEM_BITS")
Acked-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Reviewed-by: Stefan Agner <stefan@agner.ch>
Tested-by: Stefan Agner <stefan@agner.ch>
Acked-by: Mike Rapoport <rppt@linux.ibm.com>
Link: https://lore.kernel.org/linux-mm/bdfa44bf1c570b05d6c70898e2bbb0acf234ecdf.1604762181.git.stefan@agner.ch/
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
rpc is the only user of the timer_tick() function now, and can
just call the newly added generic version instead.
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Currently, the .alt.smp.init section contains the virtual addresses
of the patch sites. Since patching may occur both before and after
switching into virtual mode, this requires some manual handling of
the address when applying the UP alternative.
Let's simplify this by using relative offsets in the table entries:
this allows us to simply add each entry's address to its contents,
regardless of whether we are running in virtual mode or not.
Reviewed-by: Nicolas Pitre <nico@fluxnic.net>
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
The ARM kernel's linear map starts at PAGE_OFFSET, which maps to a
physical address (PHYS_OFFSET) that is platform specific, and is
discovered at boot. Since we don't want to slow down translations
between physical and virtual addresses by keeping the offset in a
variable in memory, we implement this by patching the code performing
the translation, and putting the offset between PAGE_OFFSET and the
start of physical RAM directly into the instruction opcodes.
As we only patch up to 8 bits of offset, yielding 4 GiB >> 8 == 16 MiB
of granularity, we have to round up PHYS_OFFSET to the next multiple if
the start of physical RAM is not a multiple of 16 MiB. This wastes some
physical RAM, since the memory that was skipped will now live below
PAGE_OFFSET, making it inaccessible to the kernel.
We can improve this by changing the patchable sequences and the patching
logic to carry more bits of offset: 11 bits gives us 4 GiB >> 11 == 2 MiB
of granularity, and so we will never waste more than that amount by
rounding up the physical start of DRAM to the next multiple of 2 MiB.
(Note that 2 MiB granularity guarantees that the linear mapping can be
created efficiently, whereas less than 2 MiB may result in the linear
mapping needing another level of page tables)
This helps Zhen Lei's scenario, where the start of DRAM is known to be
occupied. It also helps EFI boot, which relies on the firmware's page
allocator to allocate space for the decompressed kernel as low as
possible. And if the KASLR patches ever land for 32-bit, it will give
us 3 more bits of randomization of the placement of the kernel inside
the linear region.
For the ARM code path, it simply comes down to using two add/sub
instructions instead of one for the carryless version, and patching
each of them with the correct immediate depending on the rotation
field. For the LPAE calculation, which has to deal with a carry, it
patches the MOVW instruction with up to 12 bits of offset (but we only
need 11 bits anyway)
For the Thumb2 code path, patching more than 11 bits of displacement
would be somewhat cumbersome, but the 11 bits we need fit nicely into
the second word of the u16[2] opcode, so we simply update the immediate
assignment and the left shift to create an addend of the right magnitude.
Suggested-by: Zhen Lei <thunder.leizhen@huawei.com>
Acked-by: Nicolas Pitre <nico@fluxnic.net>
Acked-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
In preparation for reducing the phys-to-virt minimum relative alignment
from 16 MiB to 2 MiB, switch to patchable sequences involving MOVW
instructions that can more easily be manipulated to carry a 12-bit
immediate. Note that the non-LPAE ARM sequence is not updated: MOVW
may not be supported on non-LPAE platforms, and the sequence itself
can be updated more easily to apply the 12 bits of displacement.
For Thumb2, which has many more versions of opcodes, switch to a sequence
that can be patched by the same patching code for both versions. Note
that the Thumb2 opcodes for MOVW and MVN are unambiguous, and have no
rotation bits in their immediate fields, so there is no need to use
placeholder constants in the asm blocks.
While at it, drop the 'volatile' qualifiers from the asm blocks: the
code does not have any side effects that are invisible to the compiler,
so it is free to omit these sequences if the outputs are not used.
Suggested-by: Russell King <linux@armlinux.org.uk>
Acked-by: Nicolas Pitre <nico@fluxnic.net>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Free up a register in the p2v patching code by switching to relative
references, which don't require keeping the phys-to-virt displacement
live in a register.
Acked-by: Nicolas Pitre <nico@fluxnic.net>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
