IF YOU WOULD LIKE TO GET AN ACCOUNT, please write an
email to Administrator. User accounts are meant only to access repo
and report issues and/or generate pull requests.
This is a purpose-specific Git hosting for
BaseALT
projects. Thank you for your understanding!
Только зарегистрированные пользователи имеют доступ к сервису!
Для получения аккаунта, обратитесь к администратору.
Only capture the first cpu_id in order for the comparison
below to be of any use.
Fixes: ea3de9ce8aa2 ("RISC-V: Add a syscall for HW probing")
Signed-off-by: Andrew Jones <ajones@ventanamicro.com>
Reviewed-by: Conor Dooley <conor.dooley@microchip.com>
Reviewed-by: Evan Green <evan@rivosinc.com>
Link: https://lore.kernel.org/r/20230426141333.10063-2-ajones@ventanamicro.com
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
Add 2 early command line parameters that allow to downgrade satp mode
(using the same naming as x86):
- "no5lvl": use a 4-level page table (down from sv57 to sv48)
- "no4lvl": use a 3-level page table (down from sv57/sv48 to sv39)
Note that going through the device tree to get the kernel command line
works with ACPI too since the efi stub creates a device tree anyway with
the command line.
In KASAN kernels, we can't use the libfdt that early in the boot process
since we are not ready to execute instrumented functions. So instead of
using the "generic" libfdt, we compile our own versions of those functions
that are not instrumented and that are prefixed so that they do not
conflict with the generic ones. We also need the non-instrumented versions
of the string functions and the prefixed versions of memcpy/memmove.
This is largely inspired by commit aacd149b6238 ("arm64: head: avoid
relocating the kernel twice for KASLR") from which I removed compilation
flags that were not relevant to RISC-V at the moment (LTO, SCS). Also
note that we have to link with -z norelro to avoid ld.lld to throw a
warning with the new .got sections, like in commit 311bea3cb9ee ("arm64:
link with -z norelro for LLD or aarch64-elf").
Signed-off-by: Alexandre Ghiti <alexghiti@rivosinc.com>
Tested-by: Björn Töpel <bjorn@rivosinc.com>
Reviewed-by: Björn Töpel <bjorn@rivosinc.com>
Link: https://lore.kernel.org/r/20230424092313.178699-2-alexghiti@rivosinc.com
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
probe_vendor_features() is now called from smp_callin(), which is not
__init code and runs during cpu hotplug events. Remove the
__init_or_module decoration from it and the functions it calls to avoid
walking into outer space.
Fixes: 62a31d6e38bd ("RISC-V: hwprobe: Support probing of misaligned access performance")
Signed-off-by: Evan Green <evan@rivosinc.com>
Reviewed-by: Conor Dooley <conor.dooley@microchip.com>
Link: https://lore.kernel.org/r/20230420194934.1871356-1-evan@rivosinc.com
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
- Improve the VDSO build time checks to cover all dynamic relocations
VDSO does not allow dynamic relcations, but the build time check is
incomplete and fragile.
It's based on architectures specifying the relocation types to search
for and does not handle R_*_NONE relocation entries correctly.
R_*_NONE relocations are injected by some GNU ld variants if they fail
to determine the exact .rel[a]/dyn_size to cover trailing zeros.
R_*_NONE relocations must be ignored by dynamic loaders, so they
should be ignored in the build time check too.
Remove the architecture specific relocation types to check for and
validate strictly that no other relocations than R_*_NONE end up
in the VSDO .so file.
- Prefer signal delivery to the current thread for
CLOCK_PROCESS_CPUTIME_ID based posix-timers
Such timers prefer to deliver the signal to the main thread of a
process even if the context in which the timer expires is the current
task. This has the downside that it might wake up an idle thread.
As there is no requirement or guarantee that the signal has to be
delivered to the main thread, avoid this by preferring the current
task if it is part of the thread group which shares sighand.
This not only avoids waking idle threads, it also distributes the
signal delivery in case of multiple timers firing in the context
of different threads close to each other better.
- Align the tick period properly (again)
For a long time the tick was starting at CLOCK_MONOTONIC zero, which
allowed users space applications to either align with the tick or to
place a periodic computation so that it does not interfere with the
tick. The alignement of the tick period was more by chance than by
intention as the tick is set up before a high resolution clocksource is
installed, i.e. timekeeping is still tick based and the tick period
advances from there.
The early enablement of sched_clock() broke this alignement as the time
accumulated by sched_clock() is taken into account when timekeeping is
initialized. So the base value now(CLOCK_MONOTONIC) is not longer a
multiple of tick periods, which breaks applications which relied on
that behaviour.
Cure this by aligning the tick starting point to the next multiple of
tick periods, i.e 1000ms/CONFIG_HZ.
- A set of NOHZ fixes and enhancements
- Cure the concurrent writer race for idle and IO sleeptime statistics
The statitic values which are exposed via /proc/stat are updated from
the CPU local idle exit and remotely by cpufreq, but that happens
without any form of serialization. As a consequence sleeptimes can be
accounted twice or worse.
Prevent this by restricting the accumulation writeback to the CPU
local idle exit and let the remote access compute the accumulated
value.
- Protect idle/iowait sleep time with a sequence count
Reading idle/iowait sleep time, e.g. from /proc/stat, can race with
idle exit updates. As a consequence the readout may result in random
and potentially going backwards values.
Protect this by a sequence count, which fixes the idle time
statistics issue, but cannot fix the iowait time problem because
iowait time accounting races with remote wake ups decrementing the
remote runqueues nr_iowait counter. The latter is impossible to fix,
so the only way to deal with that is to document it properly and to
remove the assertion in the selftest which triggers occasionally due
to that.
- Restructure struct tick_sched for better cache layout
- Some small cleanups and a better cache layout for struct tick_sched
- Implement the missing timer_wait_running() callback for POSIX CPU timers
For unknown reason the introduction of the timer_wait_running() callback
missed to fixup posix CPU timers, which went unnoticed for almost four
years.
While initially only targeted to prevent livelocks between a timer
deletion and the timer expiry function on PREEMPT_RT enabled kernels, it
turned out that fixing this for mainline is not as trivial as just
implementing a stub similar to the hrtimer/timer callbacks.
The reason is that for CONFIG_POSIX_CPU_TIMERS_TASK_WORK enabled systems
there is a livelock issue independent of RT.
CONFIG_POSIX_CPU_TIMERS_TASK_WORK=y moves the expiry of POSIX CPU timers
out from hard interrupt context to task work, which is handled before
returning to user space or to a VM. The expiry mechanism moves the
expired timers to a stack local list head with sighand lock held. Once
sighand is dropped the task can be preempted and a task which wants to
delete a timer will spin-wait until the expiry task is scheduled back
in. In the worst case this will end up in a livelock when the preempting
task and the expiry task are pinned on the same CPU.
The timer wheel has a timer_wait_running() mechanism for RT, which uses
a per CPU timer-base expiry lock which is held by the expiry code and the
task waiting for the timer function to complete blocks on that lock.
This does not work in the same way for posix CPU timers as there is no
timer base and expiry for process wide timers can run on any task
belonging to that process, but the concept of waiting on an expiry lock
can be used too in a slightly different way.
Add a per task mutex to struct posix_cputimers_work, let the expiry task
hold it accross the expiry function and let the deleting task which
waits for the expiry to complete block on the mutex.
In the non-contended case this results in an extra mutex_lock()/unlock()
pair on both sides.
This avoids spin-waiting on a task which is scheduled out, prevents the
livelock and cures the problem for RT and !RT systems.
-----BEGIN PGP SIGNATURE-----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=SVnj
-----END PGP SIGNATURE-----
Merge tag 'timers-core-2023-04-24' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull timers and timekeeping updates from Thomas Gleixner:
- Improve the VDSO build time checks to cover all dynamic relocations
VDSO does not allow dynamic relocations, but the build time check is
incomplete and fragile.
It's based on architectures specifying the relocation types to search
for and does not handle R_*_NONE relocation entries correctly.
R_*_NONE relocations are injected by some GNU ld variants if they
fail to determine the exact .rel[a]/dyn_size to cover trailing zeros.
R_*_NONE relocations must be ignored by dynamic loaders, so they
should be ignored in the build time check too.
Remove the architecture specific relocation types to check for and
validate strictly that no other relocations than R_*_NONE end up in
the VSDO .so file.
- Prefer signal delivery to the current thread for
CLOCK_PROCESS_CPUTIME_ID based posix-timers
Such timers prefer to deliver the signal to the main thread of a
process even if the context in which the timer expires is the current
task. This has the downside that it might wake up an idle thread.
As there is no requirement or guarantee that the signal has to be
delivered to the main thread, avoid this by preferring the current
task if it is part of the thread group which shares sighand.
This not only avoids waking idle threads, it also distributes the
signal delivery in case of multiple timers firing in the context of
different threads close to each other better.
- Align the tick period properly (again)
For a long time the tick was starting at CLOCK_MONOTONIC zero, which
allowed users space applications to either align with the tick or to
place a periodic computation so that it does not interfere with the
tick. The alignement of the tick period was more by chance than by
intention as the tick is set up before a high resolution clocksource
is installed, i.e. timekeeping is still tick based and the tick
period advances from there.
The early enablement of sched_clock() broke this alignement as the
time accumulated by sched_clock() is taken into account when
timekeeping is initialized. So the base value now(CLOCK_MONOTONIC) is
not longer a multiple of tick periods, which breaks applications
which relied on that behaviour.
Cure this by aligning the tick starting point to the next multiple of
tick periods, i.e 1000ms/CONFIG_HZ.
- A set of NOHZ fixes and enhancements:
* Cure the concurrent writer race for idle and IO sleeptime
statistics
The statitic values which are exposed via /proc/stat are updated
from the CPU local idle exit and remotely by cpufreq, but that
happens without any form of serialization. As a consequence
sleeptimes can be accounted twice or worse.
Prevent this by restricting the accumulation writeback to the CPU
local idle exit and let the remote access compute the accumulated
value.
* Protect idle/iowait sleep time with a sequence count
Reading idle/iowait sleep time, e.g. from /proc/stat, can race
with idle exit updates. As a consequence the readout may result
in random and potentially going backwards values.
Protect this by a sequence count, which fixes the idle time
statistics issue, but cannot fix the iowait time problem because
iowait time accounting races with remote wake ups decrementing
the remote runqueues nr_iowait counter. The latter is impossible
to fix, so the only way to deal with that is to document it
properly and to remove the assertion in the selftest which
triggers occasionally due to that.
* Restructure struct tick_sched for better cache layout
* Some small cleanups and a better cache layout for struct
tick_sched
- Implement the missing timer_wait_running() callback for POSIX CPU
timers
For unknown reason the introduction of the timer_wait_running()
callback missed to fixup posix CPU timers, which went unnoticed for
almost four years.
While initially only targeted to prevent livelocks between a timer
deletion and the timer expiry function on PREEMPT_RT enabled kernels,
it turned out that fixing this for mainline is not as trivial as just
implementing a stub similar to the hrtimer/timer callbacks.
The reason is that for CONFIG_POSIX_CPU_TIMERS_TASK_WORK enabled
systems there is a livelock issue independent of RT.
CONFIG_POSIX_CPU_TIMERS_TASK_WORK=y moves the expiry of POSIX CPU
timers out from hard interrupt context to task work, which is handled
before returning to user space or to a VM. The expiry mechanism moves
the expired timers to a stack local list head with sighand lock held.
Once sighand is dropped the task can be preempted and a task which
wants to delete a timer will spin-wait until the expiry task is
scheduled back in. In the worst case this will end up in a livelock
when the preempting task and the expiry task are pinned on the same
CPU.
The timer wheel has a timer_wait_running() mechanism for RT, which
uses a per CPU timer-base expiry lock which is held by the expiry
code and the task waiting for the timer function to complete blocks
on that lock.
This does not work in the same way for posix CPU timers as there is
no timer base and expiry for process wide timers can run on any task
belonging to that process, but the concept of waiting on an expiry
lock can be used too in a slightly different way.
Add a per task mutex to struct posix_cputimers_work, let the expiry
task hold it accross the expiry function and let the deleting task
which waits for the expiry to complete block on the mutex.
In the non-contended case this results in an extra
mutex_lock()/unlock() pair on both sides.
This avoids spin-waiting on a task which is scheduled out, prevents
the livelock and cures the problem for RT and !RT systems
* tag 'timers-core-2023-04-24' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
posix-cpu-timers: Implement the missing timer_wait_running callback
selftests/proc: Assert clock_gettime(CLOCK_BOOTTIME) VS /proc/uptime monotonicity
selftests/proc: Remove idle time monotonicity assertions
MAINTAINERS: Remove stale email address
timers/nohz: Remove middle-function __tick_nohz_idle_stop_tick()
timers/nohz: Add a comment about broken iowait counter update race
timers/nohz: Protect idle/iowait sleep time under seqcount
timers/nohz: Only ever update sleeptime from idle exit
timers/nohz: Restructure and reshuffle struct tick_sched
tick/common: Align tick period with the HZ tick.
selftests/timers/posix_timers: Test delivery of signals across threads
posix-timers: Prefer delivery of signals to the current thread
vdso: Improve cmd_vdso_check to check all dynamic relocations
- Core:
- Add tracepoints for tasklet callbacks which makes it possible to
analyze individual tasklet functions instead of guess working
from the overall duration of tasklet processing
- Ensure that secondary interrupt threads have their affinity adjusted
correctly.
- Drivers:
- A large rework of the RISC-V IPI management to prepare for a new
RISC-V interrupt architecture
- Small fixes and enhancements all over the place
- Removal of support for various obsolete hardware platforms and the
related code
-----BEGIN PGP SIGNATURE-----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=EWF0
-----END PGP SIGNATURE-----
Merge tag 'irq-core-2023-04-24' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull interrupt updates from Thomas Gleixner:
"Core:
- Add tracepoints for tasklet callbacks which makes it possible to
analyze individual tasklet functions instead of guess working from
the overall duration of tasklet processing
- Ensure that secondary interrupt threads have their affinity
adjusted correctly
Drivers:
- A large rework of the RISC-V IPI management to prepare for a new
RISC-V interrupt architecture
- Small fixes and enhancements all over the place
- Removal of support for various obsolete hardware platforms and the
related code"
* tag 'irq-core-2023-04-24' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (21 commits)
irqchip/st: Remove stih415/stih416 and stid127 platforms support
irqchip/gic-v3: Add Rockchip 3588001 erratum workaround
genirq: Update affinity of secondary threads
softirq: Add trace points for tasklet entry/exit
irqchip/loongson-pch-pic: Fix pch_pic_acpi_init calling
irqchip/loongson-pch-pic: Fix registration of syscore_ops
irqchip/loongson-eiointc: Fix registration of syscore_ops
irqchip/loongson-eiointc: Fix incorrect use of acpi_get_vec_parent
irqchip/loongson-eiointc: Fix returned value on parsing MADT
irqchip/riscv-intc: Add empty irq_eoi() for chained irq handlers
RISC-V: Use IPIs for remote icache flush when possible
RISC-V: Use IPIs for remote TLB flush when possible
RISC-V: Allow marking IPIs as suitable for remote FENCEs
RISC-V: Treat IPIs as normal Linux IRQs
irqchip/riscv-intc: Allow drivers to directly discover INTC hwnode
RISC-V: Clear SIP bit only when using SBI IPI operations
irqchip/irq-sifive-plic: Add syscore callbacks for hibernation
irqchip: Use of_property_read_bool() for boolean properties
irqchip/bcm-6345-l1: Request memory region
irqchip/gicv3: Workaround for NVIDIA erratum T241-FABRIC-4
...
We have two extension names for AIA ISA support: Smaia (M-mode AIA CSRs)
and Ssaia (S-mode AIA CSRs).
We extend the ISA string parsing to detect Smaia and Ssaia extensions.
Signed-off-by: Anup Patel <apatel@ventanamicro.com>
Reviewed-by: Andrew Jones <ajones@ventanamicro.com>
Reviewed-by: Atish Patra <atishp@rivosinc.com>
Reviewed-by: Conor Dooley <conor.dooley@microchip.com>
Signed-off-by: Anup Patel <anup@brainfault.org>
Acked-by: Palmer Dabbelt <palmer@rivosinc.com>
Alexandre Ghiti <alexghiti@rivosinc.com> says:
After multiple attempts, this patchset is now based on the fact that the
64b kernel mapping was moved outside the linear mapping.
The first patch allows to build relocatable kernels but is not selected
by default. That patch is a requirement for KASLR.
The second and third patches take advantage of an already existing powerpc
script that checks relocations at compile-time, and uses it for riscv.
* b4-shazam-merge:
riscv: Use --emit-relocs in order to move .rela.dyn in init
riscv: Check relocations at compile time
powerpc: Move script to check relocations at compile time in scripts/
riscv: Introduce CONFIG_RELOCATABLE
riscv: Move .rela.dyn outside of init to avoid empty relocations
riscv: Prepare EFI header for relocatable kernels
Link: https://lore.kernel.org/r/20230329045329.64565-1-alexghiti@rivosinc.com
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
This config allows to compile 64b kernel as PIE and to relocate it at
any virtual address at runtime: this paves the way to KASLR.
Runtime relocation is possible since relocation metadata are embedded into
the kernel.
Note that relocating at runtime introduces an overhead even if the
kernel is loaded at the same address it was linked at and that the compiler
options are those used in arm64 which uses the same RELA relocation
format.
Signed-off-by: Alexandre Ghiti <alexghiti@rivosinc.com>
Link: https://lore.kernel.org/r/20230329045329.64565-4-alexghiti@rivosinc.com
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
Alexandre Ghiti <alexghiti@rivosinc.com> says:
As described in patch 2, our current kasan implementation is intricate,
so I tried to simplify the implementation and mimic what arm64/x86 are
doing.
In addition it fixes UEFI bootflow with a kasan kernel and kasan inline
instrumentation: all kasan configurations were tested on a large ubuntu
kernel with success with KASAN_KUNIT_TEST and KASAN_MODULE_TEST.
inline ubuntu config + uefi:
sv39: OK
sv48: OK
sv57: OK
outline ubuntu config + uefi:
sv39: OK
sv48: OK
sv57: OK
Actually 1 test always fails with KASAN_KUNIT_TEST that I have to check:
KASAN failure expected in "set_bit(nr, addr)", but none occurrred
Note that Palmer recently proposed to remove COMMAND_LINE_SIZE from the
userspace abi
https://lore.kernel.org/lkml/20221211061358.28035-1-palmer@rivosinc.com/T/
so that we can finally increase the command line to fit all kasan kernel
parameters.
All of this should hopefully fix the syzkaller riscv build that has been
failing for a few months now, any test is appreciated and if I can help
in any way, please ask.
* b4-shazam-merge:
riscv: Unconditionnally select KASAN_VMALLOC if KASAN
riscv: Fix ptdump when KASAN is enabled
riscv: Fix EFI stub usage of KASAN instrumented strcmp function
riscv: Move DTB_EARLY_BASE_VA to the kernel address space
riscv: Rework kasan population functions
riscv: Split early and final KASAN population functions
Link: https://lore.kernel.org/r/20230203075232.274282-1-alexghiti@rivosinc.com
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
The EFI stub must not use any KASAN instrumented code as the kernel
proper did not initialize the thread pointer and the mapping for the
KASAN shadow region.
Avoid using the generic strcmp function, instead use the one in
drivers/firmware/efi/libstub/string.c.
Signed-off-by: Alexandre Ghiti <alexghiti@rivosinc.com>
Acked-by: Ard Biesheuvel <ardb@kernel.org>
Reviewed-by: Atish Patra <atishp@rivosinc.com>
Link: https://lore.kernel.org/r/20230203075232.274282-5-alexghiti@rivosinc.com
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
Evan Green <evan@rivosinc.com> says:
There's been a bunch of off-list discussions about this, including at
Plumbers. The original plan was to do something involving providing an
ISA string to userspace, but ISA strings just aren't sufficient for a
stable ABI any more: in order to parse an ISA string users need the
version of the specifications that the string is written to, the version
of each extension (sometimes at a finer granularity than the RISC-V
releases/versions encode), and the expected use case for the ISA string
(ie, is it a U-mode or M-mode string). That's a lot of complexity to
try and keep ABI compatible and it's probably going to continue to grow,
as even if there's no more complexity in the specifications we'll have
to deal with the various ISA string parsing oddities that end up all
over userspace.
Instead this patch set takes a very different approach and provides a set
of key/value pairs that encode various bits about the system. The big
advantage here is that we can clearly define what these mean so we can
ensure ABI stability, but it also allows us to encode information that's
unlikely to ever appear in an ISA string (see the misaligned access
performance, for example). The resulting interface looks a lot like
what arm64 and x86 do, and will hopefully fit well into something like
ACPI in the future.
The actual user interface is a syscall, with a vDSO function in front of
it. The vDSO function can answer some queries without a syscall at all,
and falls back to the syscall for cases it doesn't have answers to.
Currently we prepopulate it with an array of answers for all keys and
a CPU set of "all CPUs". This can be adjusted as necessary to provide
fast answers to the most common queries.
An example series in glibc exposing this syscall and using it in an
ifunc selector for memcpy can be found at [1].
I was asked about the performance delta between this and something like
sysfs. I created a small test program and ran it on a Nezha D1
Allwinner board. Doing each operation 100000 times and dividing, these
operations take the following amount of time:
- open()+read()+close() of /sys/kernel/cpu_byteorder: 3.8us
- access("/sys/kernel/cpu_byteorder", R_OK): 1.3us
- riscv_hwprobe() vDSO and syscall: .0094us
- riscv_hwprobe() vDSO with no syscall: 0.0091us
These numbers get farther apart if we query multiple keys, as sysfs will
scale linearly with the number of keys, where the dedicated syscall
stays the same. To frame these numbers, I also did a tight
fork/exec/wait loop, which I measured as 4.8ms. So doing 4
open/read/close operations is a delta of about 0.3%, versus a single vDSO
call is a delta of essentially zero.
[1] https://patchwork.ozlabs.org/project/glibc/list/?series=343050
* b4-shazam-merge:
RISC-V: Add hwprobe vDSO function and data
selftests: Test the new RISC-V hwprobe interface
RISC-V: hwprobe: Support probing of misaligned access performance
RISC-V: hwprobe: Add support for RISCV_HWPROBE_BASE_BEHAVIOR_IMA
RISC-V: Add a syscall for HW probing
RISC-V: Move struct riscv_cpuinfo to new header
Link: https://lore.kernel.org/r/20230407231103.2622178-1-evan@rivosinc.com
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
Add a vDSO function __vdso_riscv_hwprobe, which can sit in front of the
riscv_hwprobe syscall and answer common queries. We stash a copy of
static answers for the "all CPUs" case in the vDSO data page. This data
is private to the vDSO, so we can decide later to change what's stored
there or under what conditions we defer to the syscall. Currently all
data can be discovered at boot, so the vDSO function answers all queries
when the cpumask is set to the "all CPUs" hint.
There's also a boolean in the data that lets the vDSO function know that
all CPUs are the same. In that case, the vDSO will also answer queries
for arbitrary CPU masks in addition to the "all CPUs" hint.
Signed-off-by: Evan Green <evan@rivosinc.com>
Link: https://lore.kernel.org/r/20230407231103.2622178-7-evan@rivosinc.com
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
This allows userspace to select various routines to use based on the
performance of misaligned access on the target hardware.
Rather than adding DT bindings, this change taps into the alternatives
mechanism used to probe CPU errata. Add a new function pointer alongside
the vendor-specific errata_patch_func() that probes for desirable errata
(otherwise known as "features"). Unlike the errata_patch_func(), this
function is called on each CPU as it comes up, so it can save
feature information per-CPU.
The T-head C906 has fast unaligned access, both as defined by GCC [1],
and in performing a basic benchmark, which determined that byte copies
are >50% slower than a misaligned word copy of the same data size (source
for this test at [2]):
bytecopy size f000 count 50000 offset 0 took 31664899 us
wordcopy size f000 count 50000 offset 0 took 5180919 us
wordcopy size f000 count 50000 offset 1 took 13416949 us
[1] https://github.com/gcc-mirror/gcc/blob/master/gcc/config/riscv/riscv.cc#L353
[2] https://pastebin.com/EPXvDHSW
Co-developed-by: Palmer Dabbelt <palmer@rivosinc.com>
Signed-off-by: Evan Green <evan@rivosinc.com>
Reviewed-by: Heiko Stuebner <heiko.stuebner@vrull.eu>
Tested-by: Heiko Stuebner <heiko.stuebner@vrull.eu>
Reviewed-by: Conor Dooley <conor.dooley@microchip.com>
Reviewed-by: Paul Walmsley <paul.walmsley@sifive.com>
Link: https://lore.kernel.org/r/20230407231103.2622178-5-evan@rivosinc.com
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
We have an implicit set of base behaviors that userspace depends on,
which are mostly defined in various ISA specifications.
Co-developed-by: Palmer Dabbelt <palmer@rivosinc.com>
Signed-off-by: Evan Green <evan@rivosinc.com>
Reviewed-by: Conor Dooley <conor.dooley@microchip.com>
Reviewed-by: Heiko Stuebner <heiko.stuebner@vrull.eu>
Tested-by: Heiko Stuebner <heiko.stuebner@vrull.eu>
Reviewed-by: Paul Walmsley <paul.walmsley@sifive.com>
Link: https://lore.kernel.org/r/20230407231103.2622178-4-evan@rivosinc.com
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
We don't have enough space for these all in ELF_HWCAP{,2} and there's no
system call that quite does this, so let's just provide an arch-specific
one to probe for hardware capabilities. This currently just provides
m{arch,imp,vendor}id, but with the key-value pairs we can pass more in
the future.
Co-developed-by: Palmer Dabbelt <palmer@rivosinc.com>
Signed-off-by: Evan Green <evan@rivosinc.com>
Reviewed-by: Conor Dooley <conor.dooley@microchip.com>
Reviewed-by: Heiko Stuebner <heiko.stuebner@vrull.eu>
Tested-by: Heiko Stuebner <heiko.stuebner@vrull.eu>
Reviewed-by: Paul Walmsley <paul.walmsley@sifive.com>
Link: https://lore.kernel.org/r/20230407231103.2622178-3-evan@rivosinc.com
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
In preparation for tracking and exposing microarchitectural details to
userspace (like whether or not unaligned accesses are fast), move the
riscv_cpuinfo struct out to its own new cpufeatures.h header. It will
need to be used by more than just cpu.c.
Signed-off-by: Evan Green <evan@rivosinc.com>
Reviewed-by: Conor Dooley <conor.dooley@microchip.com>
Reviewed-by: Heiko Stuebner <heiko.stuebner@vrull.eu>
Tested-by: Heiko Stuebner <heiko.stuebner@vrull.eu>
Reviewed-by: Paul Walmsley <paul.walmsley@sifive.com>
Link: https://lore.kernel.org/r/20230407231103.2622178-2-evan@rivosinc.com
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
early_init_dt_verify() is already called in parse_dtb() and since the dtb
address does not change anymore (it is now in the fixmap region), no need
to reset initial_boot_params by calling early_init_dt_verify() again.
Signed-off-by: Alexandre Ghiti <alexghiti@rivosinc.com>
Link: https://lore.kernel.org/r/20230329081932.79831-3-alexghiti@rivosinc.com
Cc: stable@vger.kernel.org
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
riscv establishes 2 virtual mappings:
- early_pg_dir maps the kernel which allows to discover the system
memory
- swapper_pg_dir installs the final mapping (linear mapping included)
We used to map the dtb in early_pg_dir using DTB_EARLY_BASE_VA, and this
mapping was not carried over in swapper_pg_dir. It happens that
early_init_fdt_scan_reserved_mem() must be called before swapper_pg_dir is
setup otherwise we could allocate reserved memory defined in the dtb.
And this function initializes reserved_mem variable with addresses that
lie in the early_pg_dir dtb mapping: when those addresses are reused
with swapper_pg_dir, this mapping does not exist and then we trap.
The previous "fix" was incorrect as early_init_fdt_scan_reserved_mem()
must be called before swapper_pg_dir is set up otherwise we could
allocate in reserved memory defined in the dtb.
So move the dtb mapping in the fixmap region which is established in
early_pg_dir and handed over to swapper_pg_dir.
Fixes: 922b0375fc93 ("riscv: Fix memblock reservation for device tree blob")
Fixes: 8f3a2b4a96dc ("RISC-V: Move DT mapping outof fixmap")
Fixes: 50e63dd8ed92 ("riscv: fix reserved memory setup")
Reported-by: Conor Dooley <conor.dooley@microchip.com>
Link: https://lore.kernel.org/all/f8e67f82-103d-156c-deb0-d6d6e2756f5e@microchip.com/
Signed-off-by: Alexandre Ghiti <alexghiti@rivosinc.com>
Reviewed-by: Conor Dooley <conor.dooley@microchip.com>
Tested-by: Conor Dooley <conor.dooley@microchip.com>
Link: https://lore.kernel.org/r/20230329081932.79831-2-alexghiti@rivosinc.com
Cc: stable@vger.kernel.org
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
Now that of_cpu_device_node_get() is defined in of.h, of_device.h is just
implicitly including other includes, and is no longer needed. Adjust the
include files with what was implicitly included by of_device.h (cpu.h and
of.h) and drop including of_device.h.
Link: https://lore.kernel.org/r/20230329-dt-cpu-header-cleanups-v1-9-581e2605fe47@kernel.org
Signed-off-by: Rob Herring <robh@kernel.org>
Removing the include of cpu.h from of_device.h (included by
of_platform.h) causes an error in setup.c:
arch/riscv/kernel/setup.c:313:22: error: arithmetic on a pointer to an incomplete type 'typeof(struct cpu)' (aka 'struct cpu')
The of_platform.h header is not necessary either, so it can be dropped.
Link: https://lore.kernel.org/r/20230329-dt-cpu-header-cleanups-v1-8-581e2605fe47@kernel.org
Signed-off-by: Rob Herring <robh@kernel.org>
This reverts commit baf7cbd94b5688f167443a2cc3dcea3300132099.
There are some duplicate cache attributes populations executed
in both ci_leaf_init() and later cache_setup_properties().
Revert the commit baf7cbd94b56 ("riscv: Set more data to cacheinfo")
to setup only the level and type attributes at this early place.
Signed-off-by: Song Shuai <suagrfillet@gmail.com>
Acked-by: Sudeep Holla <sudeep.holla@arm.com>
Acked-by: Conor Dooley <conor.dooley@microchip.com>
Link: https://lore.kernel.org/r/20230308064734.512457-1-suagrfillet@gmail.com
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
In a NOMMU kernel, sigreturn trampolines are generated on the user
stack by setup_rt_frame. Currently, these trampolines are not instruction
fenced, thus their visibility to ifetch is not guaranteed.
This patch adds a flush_icache_range in setup_rt_frame to fix this
problem.
Signed-off-by: Mathis Salmen <mathis.salmen@matsal.de>
Fixes: 6bd33e1ece52 ("riscv: add nommu support")
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20230406101130.82304-1-mathis.salmen@matsal.de
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
The RISC-V calling convention passes the first argument, and the
return value in the a0 register. For this reason, the a0 register
needs some extra care; When handling syscalls, the a0 register is
saved into regs->orig_a0, so a0 can be properly restored for,
e.g. interrupted syscalls.
This functionality was broken with the introduction of the generic
entry patches. Here, a0 was saved into orig_a0 after calling
syscall_enter_from_user_mode(), which can change regs->a0 for some
paths, incorrectly restoring a0.
This is resolved, by saving a0 prior doing the
syscall_enter_from_user_mode() call.
Fixes: f0bddf50586d ("riscv: entry: Convert to generic entry")
Reviewed-by: Heiko Stuebner <heiko.stuebner@vrull.eu>
Tested-by: Heiko Stuebner <heiko.stuebner@vrull.eu>
Signed-off-by: Björn Töpel <bjorn@rivosinc.com>
Reported-by: Conor Dooley <conor.dooley@microchip.com>
Reviewed-by: Conor Dooley <conor.dooley@microchip.com>
Tested-by: Conor Dooley <conor.dooley@microchip.com>
Tested-by: Geert Uytterhoeven <geert+renesas@glider.be>
Tested-by: Andy Chiu <andy.chiu@sifive.com>
Link: https://lore.kernel.org/r/20230403065207.1070974-1-bjorn@kernel.org
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
To do remote FENCEs (i.e. remote TLB flushes) using IPI calls on the
RISC-V kernel, we need hardware mechanism to directly inject IPI from
the supervisor mode (i.e. RISC-V kernel) instead of using SBI calls.
The upcoming AIA IMSIC devices allow direct IPI injection from the
supervisor mode (i.e. RISC-V kernel). To support this, we extend the
riscv_ipi_set_virq_range() function so that IPI provider (i.e. irqchip
drivers can mark IPIs as suitable for remote FENCEs.
Signed-off-by: Anup Patel <apatel@ventanamicro.com>
Reviewed-by: Atish Patra <atishp@rivosinc.com>
Acked-by: Palmer Dabbelt <palmer@rivosinc.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20230328035223.1480939-5-apatel@ventanamicro.com
Currently, the RISC-V kernel provides arch specific hooks (i.e.
struct riscv_ipi_ops) to register IPI handling methods. The stats
gathering of IPIs is also arch specific in the RISC-V kernel.
Other architectures (such as ARM, ARM64, and MIPS) have moved away
from custom arch specific IPI handling methods. Currently, these
architectures have Linux irqchip drivers providing a range of Linux
IRQ numbers to be used as IPIs and IPI triggering is done using
generic IPI APIs. This approach allows architectures to treat IPIs
as normal Linux IRQs and IPI stats gathering is done by the generic
Linux IRQ subsystem.
We extend the RISC-V IPI handling as-per above approach so that arch
specific IPI handling methods (struct riscv_ipi_ops) can be removed
and the IPI handling is done through the Linux IRQ subsystem.
Signed-off-by: Anup Patel <apatel@ventanamicro.com>
Acked-by: Palmer Dabbelt <palmer@rivosinc.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20230328035223.1480939-4-apatel@ventanamicro.com
Various RISC-V drivers (such as SBI IPI, SBI Timer, SBI PMU, and
KVM RISC-V) don't have associated DT node but these drivers need
standard per-CPU (local) interrupts defined by the RISC-V privileged
specification.
We add riscv_get_intc_hwnode() in arch/riscv which allows RISC-V
drivers not having DT node to discover INTC hwnode which in-turn
helps these drivers to map per-CPU (local) interrupts provided
by the INTC driver.
Signed-off-by: Anup Patel <apatel@ventanamicro.com>
Reviewed-by: Atish Patra <atishp@rivosinc.com>
Acked-by: Palmer Dabbelt <palmer@rivosinc.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20230328035223.1480939-3-apatel@ventanamicro.com
The software interrupt pending (i.e. [M|S]SIP) bit is writeable for
S-mode but read-only for M-mode so we clear this bit only when using
SBI IPI operations.
Signed-off-by: Anup Patel <apatel@ventanamicro.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Atish Patra <atishp@rivosinc.com>
Acked-by: Palmer Dabbelt <palmer@rivosinc.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20230328035223.1480939-2-apatel@ventanamicro.com
guoren@kernel.org <guoren@kernel.org> says:
From: Guo Ren <guoren@linux.alibaba.com>
The patches convert riscv to use the generic entry infrastructure from
kernel/entry/*. Some optimization for entry.S with new .macro and merge
ret_from_kernel_thread into ret_from_fork.
* b4-shazam-merge:
riscv: entry: Consolidate general regs saving/restoring
riscv: entry: Consolidate ret_from_kernel_thread into ret_from_fork
riscv: entry: Remove extra level wrappers of trace_hardirqs_{on,off}
riscv: entry: Convert to generic entry
riscv: entry: Add noinstr to prevent instrumentation inserted
riscv: ptrace: Remove duplicate operation
Link: https://lore.kernel.org/r/20230222033021.983168-1-guoren@kernel.org
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
To be able to trace invocations of smp_send_reschedule(), rename the
arch-specific definitions of it to arch_smp_send_reschedule() and wrap it
into an smp_send_reschedule() that contains a tracepoint.
Changes to include the declaration of the tracepoint were driven by the
following coccinelle script:
@func_use@
@@
smp_send_reschedule(...);
@include@
@@
#include <trace/events/ipi.h>
@no_include depends on func_use && !include@
@@
#include <...>
+
+ #include <trace/events/ipi.h>
[csky bits]
[riscv bits]
Signed-off-by: Valentin Schneider <vschneid@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Guo Ren <guoren@kernel.org>
Acked-by: Palmer Dabbelt <palmer@rivosinc.com>
Link: https://lore.kernel.org/r/20230307143558.294354-6-vschneid@redhat.com
The ret_from_kernel_thread() behaves similarly with ret_from_fork(),
the only difference is whether call the fn(arg) or not, this can be
achieved by testing fn is NULL or not, I.E s0 is 0 or not. Many
architectures have done the same thing, it makes entry.S more clean.
Signed-off-by: Jisheng Zhang <jszhang@kernel.org>
Reviewed-by: Björn Töpel <bjorn@rivosinc.com>
Reviewed-by: Guo Ren <guoren@kernel.org>
Tested-by: Guo Ren <guoren@kernel.org>
Signed-off-by: Guo Ren <guoren@kernel.org>
Link: https://lore.kernel.org/r/20230222033021.983168-7-guoren@kernel.org
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
Since riscv is converted to generic entry, there's no need for the
extra wrappers of trace_hardirqs_{on,off}.
Signed-off-by: Jisheng Zhang <jszhang@kernel.org>
Reviewed-by: Guo Ren <guoren@kernel.org>
Reviewed-by: Björn Töpel <bjorn@rivosinc.com>
Tested-by: Guo Ren <guoren@kernel.org>
Signed-off-by: Guo Ren <guoren@kernel.org>
Link: https://lore.kernel.org/r/20230222033021.983168-6-guoren@kernel.org
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
This patch converts riscv to use the generic entry infrastructure from
kernel/entry/*. The generic entry makes maintainers' work easier and
codes more elegant. Here are the changes:
- More clear entry.S with handle_exception and ret_from_exception
- Get rid of complex custom signal implementation
- Move syscall procedure from assembly to C, which is much more
readable.
- Connect ret_from_fork & ret_from_kernel_thread to generic entry.
- Wrap with irqentry_enter/exit and syscall_enter/exit_from_user_mode
- Use the standard preemption code instead of custom
Suggested-by: Huacai Chen <chenhuacai@kernel.org>
Reviewed-by: Björn Töpel <bjorn@rivosinc.com>
Tested-by: Yipeng Zou <zouyipeng@huawei.com>
Tested-by: Jisheng Zhang <jszhang@kernel.org>
Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
Signed-off-by: Guo Ren <guoren@kernel.org>
Cc: Ben Hutchings <ben@decadent.org.uk>
Link: https://lore.kernel.org/r/20230222033021.983168-5-guoren@kernel.org
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
Without noinstr the compiler is free to insert instrumentation (think
all the k*SAN, KCov, GCov, ftrace etc..) which can call code we're not
yet ready to run this early in the entry path, for instance it could
rely on RCU which isn't on yet, or expect lockdep state. (by peterz)
Link: https://lore.kernel.org/linux-riscv/YxcQ6NoPf3AH0EXe@hirez.programming.kicks-ass.net/
Reviewed-by: Björn Töpel <bjorn@rivosinc.com>
Suggested-by: Peter Zijlstra <peterz@infradead.org>
Tested-by: Jisheng Zhang <jszhang@kernel.org>
Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
Signed-off-by: Guo Ren <guoren@kernel.org>
Link: https://lore.kernel.org/r/20230222033021.983168-4-guoren@kernel.org
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
The TIF_SYSCALL_TRACE is controlled by a common code, see
kernel/ptrace.c and include/linux/thread_info.h.
clear_task_syscall_work(child, SYSCALL_TRACE);
Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
Signed-off-by: Guo Ren <guoren@kernel.org>
Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Reviewed-by: Björn Töpel <bjorn@rivosinc.com>
Link: https://lore.kernel.org/r/20230222033021.983168-3-guoren@kernel.org
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
The actual intention is that no dynamic relocation exists in the VDSO. For
this the VDSO build validates that the resulting .so file does not have any
relocations which are specified via $(ARCH_REL_TYPE_ABS) per architecture,
which is fragile as e.g. ARM64 lacks an entry for R_AARCH64_RELATIVE. Aside
of that ARCH_REL_TYPE_ABS is a misnomer as it checks for relative
relocations too.
However, some GNU ld ports produce unneeded R_*_NONE relocation entries. If
a port fails to determine the exact .rel[a].dyn size, the trailing zeros
become R_*_NONE relocations. E.g. ld's powerpc port recently fixed
https://sourceware.org/bugzilla/show_bug.cgi?id=29540). R_*_NONE are
generally a no-op in the dynamic loaders. So just ignore them.
Remove the ARCH_REL_TYPE_ABS defines and just validate that the resulting
.so file does not contain any R_* relocation entries except R_*_NONE.
Signed-off-by: Fangrui Song <maskray@google.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Vincenzo Frascino <vincenzo.frascino@arm.com> # for aarch64
Reviewed-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Reviewed-by: Vincenzo Frascino <vincenzo.frascino@arm.com> # for vDSO, aarch64
Acked-by: Michael Ellerman <mpe@ellerman.id.au> (powerpc)
Link: https://lore.kernel.org/r/20230310190750.3323802-1-maskray@google.com
Andrew Jones <ajones@ventanamicro.com> says:
When the Zicboz extension is available we can more rapidly zero naturally
aligned Zicboz block sized chunks of memory. As pages are always page
aligned and are larger than any Zicboz block size will be, then
clear_page() appears to be a good candidate for the extension. While cycle
count and energy consumption should also be considered, we can be pretty
certain that implementing clear_page() with the Zicboz extension is a win
by comparing the new dynamic instruction count with its current count[1].
Doing so we see that the new count is just over a quarter of the old count
(see patch6's commit message for more details).
For those of you who reviewed v1[2], you may be looking for the memset()
patches. As pointed out in v1, and a couple follow-up emails, it's not
clear that patching memset() is a win yet. When I get a chance to test
on real hardware with a comprehensive benchmark collection then I can
post the memset() patches separately (assuming the benchmarks show it's
worthwhile).
* b4-shazam-merge:
RISC-V: KVM: Expose Zicboz to the guest
RISC-V: KVM: Provide UAPI for Zicboz block size
RISC-V: Use Zicboz in clear_page when available
RISC-V: cpufeatures: Put the upper 16 bits of patch ID to work
RISC-V: Add Zicboz detection and block size parsing
dt-bindings: riscv: Document cboz-block-size
RISC-V: Factor out body of riscv_init_cbom_blocksize loop
RISC-V: alternatives: Support patching multiple insns in assembly
Link: https://lore.kernel.org/r/20230224162631.405473-1-ajones@ventanamicro.com
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
Using memset() to zero a 4K page takes 563 total instructions, where
20 are branches. clear_page(), with Zicboz and a 64 byte block size,
takes 169 total instructions, where 4 are branches and 33 are nops.
Even though the block size is a variable, thanks to alternatives, we
can still implement a Duff device without having to do any preliminary
calculations. This is achieved by using the alternatives' cpufeature
value (the upper 16 bits of patch_id). The value used is the maximum
zicboz block size order accepted at the patch site. This enables us
to stop patching / unrolling when 4K bytes have been zeroed (we would
loop and continue after 4K if the page size would be larger)
For 4K pages, unrolling 16 times allows block sizes of 64 and 128 to
only loop a few times and larger block sizes to not loop at all. Since
cbo.zero doesn't take an offset, we also need an 'add' after each
instruction, making the loop body 112 to 160 bytes. Hopefully this
is small enough to not cause icache misses.
Signed-off-by: Andrew Jones <ajones@ventanamicro.com>
Acked-by: Conor Dooley <conor.dooley@microchip.com>
Link: https://lore.kernel.org/r/20230224162631.405473-7-ajones@ventanamicro.com
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
cpufeature IDs are consecutive integers starting at 26, so a 32-bit
patch ID allows an aircraft carrier load of feature IDs. Repurposing
the upper 16 bits still leaves a boat load of feature IDs and gains
16 bits which may be used to control patching on a per patch-site
basis.
This will be initially used in Zicboz's application to clear_page(),
as Zicboz's block size must also be considered. In that case, the
upper 16-bit value's role will be to convey the maximum block size
which the Zicboz clear_page() implementation supports.
cpufeature patch sites which need to check for the existence or
absence of other cpufeatures may also be able to make use of this.
Signed-off-by: Andrew Jones <ajones@ventanamicro.com>
Reviewed-by: Conor Dooley <conor.dooley@microchip.com>
Link: https://lore.kernel.org/r/20230224162631.405473-6-ajones@ventanamicro.com
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
Parse "riscv,cboz-block-size" from the DT by piggybacking on Zicbom's
riscv_init_cbom_blocksize(). Additionally check the DT for the presence
of the "zicboz" extension and, when it's present, validate the parsed
cboz block size as we do Zicbom's cbom block size with
riscv_isa_extension_check().
Signed-off-by: Andrew Jones <ajones@ventanamicro.com>
Reviewed-by: Heiko Stuebner <heiko@sntech.de>
Reviewed-by: Conor Dooley <conor.dooley@microchip.com>
Link: https://lore.kernel.org/r/20230224162631.405473-5-ajones@ventanamicro.com
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
Andrew Jones <ajones@ventanamicro.com> says:
This series has no intended functional change. These cleanups were
found while renaming errata_id to patch_id in order to better
convey that its purpose is larger than errata (it's also for
cpufeatures).
* b4-shazam-merge:
riscv: cpufeature: Drop errata_list.h and other unused includes
riscv: lib: Include hwcap.h directly
riscv: alternatives: Rename errata_id to patch_id
riscv: alternatives: Remove unnecessary define and unused struct
riscv: Rename Kconfig.erratas to Kconfig.errata
riscv: Clarify RISCV_ALTERNATIVE help text
Link: https://lore.kernel.org/r/20230224154601.88163-1-ajones@ventanamicro.com
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
Drop errata_list.h, since cpufeature.c includes hwcap.h directly to
get cpufeature IDs. And, while there, prune the rest of the unused
includes too.
Signed-off-by: Andrew Jones <ajones@ventanamicro.com>
Reviewed-by: Conor Dooley <conor.dooley@microchip.com>
Reviewed-by: Heiko Stuebner <heiko.stuebner@vrull.eu>
Link: https://lore.kernel.org/r/20230224154601.88163-7-ajones@ventanamicro.com
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
Alternatives are used for both errata and cpufeatures. Use a more
generic name, 'patch_id', as in "ID of code patching site", to
avoid confusion when alternatives are used for cpufeatures.
Signed-off-by: Andrew Jones <ajones@ventanamicro.com>
Reviewed-by: Conor Dooley <conor.dooley@microchip.com>
Reviewed-by: Heiko Stuebner <heiko.stuebner@vrull.eu>
Link: https://lore.kernel.org/r/20230224154601.88163-5-ajones@ventanamicro.com
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
Qinglin Pan <panqinglin00@gmail.com> says:
Svnapot is a RISC-V extension for marking contiguous 4K pages as a non-4K
page. This patch set is for using Svnapot in hugetlb fs and huge vmap.
This patchset adds a Kconfig item for using Svnapot in
"Platform type"->"SVNAPOT extension support". Its default value is on,
and people can set it off if they don't allow kernel to detect Svnapot
hardware support and leverage it.
Tested on:
- qemu rv64 with "Svnapot support" off and svnapot=true.
- qemu rv64 with "Svnapot support" on and svnapot=true.
- qemu rv64 with "Svnapot support" off and svnapot=false.
- qemu rv64 with "Svnapot support" on and svnapot=false.
* b4-shazam-merge:
riscv: mm: support Svnapot in huge vmap
riscv: mm: support Svnapot in hugetlb page
riscv: mm: modify pte format for Svnapot
Link: https://lore.kernel.org/r/20230209131647.17245-1-panqinglin00@gmail.com
[Palmer: fix up the feature ordering in the merge]
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
We're currently using stop_machine() to update ftrace & kprobes, which
means that the thread that takes text_mutex during may not be the same
as the thread that eventually patches the code. This isn't actually a
race because the lock is still held (preventing any other concurrent
accesses) and there is only one thread running during stop_machine(),
but it does trigger a lockdep failure.
This patch just elides the lockdep check during stop_machine.
Fixes: c15ac4fd60d5 ("riscv/ftrace: Add dynamic function tracer support")
Suggested-by: Steven Rostedt <rostedt@goodmis.org>
Reported-by: Changbin Du <changbin.du@gmail.com>
Signed-off-by: Palmer Dabbelt <palmerdabbelt@google.com>
Signed-off-by: Conor Dooley <conor.dooley@microchip.com>
Link: https://lore.kernel.org/r/20230303143754.4005217-1-conor.dooley@microchip.com
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
Before commit 076cbf5d2163 ("x86/xen: don't let xen_pv_play_dead()
return"), in Xen, when a previously offlined CPU was brought back
online, it unexpectedly resumed execution where it left off in the
middle of the idle loop.
There were some hacks to make that work, but the behavior was surprising
as do_idle() doesn't expect an offlined CPU to return from the dead (in
arch_cpu_idle_dead()).
Now that Xen has been fixed, and the arch-specific implementations of
arch_cpu_idle_dead() also don't return, give it a __noreturn attribute.
This will cause the compiler to complain if an arch-specific
implementation might return. It also improves code generation for both
caller and callee.
Also fixes the following warning:
vmlinux.o: warning: objtool: do_idle+0x25f: unreachable instruction
Reported-by: Paul E. McKenney <paulmck@kernel.org>
Tested-by: Paul E. McKenney <paulmck@kernel.org>
Link: https://lore.kernel.org/r/60d527353da8c99d4cf13b6473131d46719ed16d.1676358308.git.jpoimboe@kernel.org
Signed-off-by: Josh Poimboeuf <jpoimboe@kernel.org>
Add one alternative to enable/disable svnapot support, enable this static
key when "svnapot" is in the "riscv,isa" field of fdt and SVNAPOT compile
option is set. It will influence the behavior of has_svnapot. All code
dependent on svnapot should make sure that has_svnapot return true firstly.
Modify PTE definition for Svnapot, and creates some functions in pgtable.h
to mark a PTE as napot and check if it is a Svnapot PTE. Until now, only
64KB napot size is supported in spec, so some macros has only 64KB version.
Signed-off-by: Qinglin Pan <panqinglin00@gmail.com>
Reviewed-by: Andrew Jones <ajones@ventanamicro.com>
Link: https://lore.kernel.org/r/20230209131647.17245-2-panqinglin00@gmail.com
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
