| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
| |
T-HEAD CPUs such as C906/C910/C920 support phy address based CMO, use
them so that we don't need to convert to virt address.
Signed-off-by: Jisheng Zhang <jszhang@kernel.org>
Reviewed-by: Guo Ren <guoren@kernel.org>
Link: https://lore.kernel.org/r/20231114143338.2406-3-jszhang@kernel.org
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Previously, we use alternative mechanism to dynamically patch
the CMO operations for THEAD C906/C910 during boot for performance
reason. But as pointed out by Arnd, "there is already a significant
cost in accessing the invalidated cache lines afterwards, which is
likely going to be much higher than the cost of an indirect branch".
And indeed, there's no performance difference with GMAC and EMMC per
my test on Sipeed Lichee Pi 4A board.
Use riscv_nonstd_cache_ops for THEAD C906/C910 CMO to simplify
the alternative code, and to acchieve Arnd's goal -- "I think
moving the THEAD ops at the same level as all nonstandard operations
makes sense, but I'd still leave CMO as an explicit fast path that
avoids the indirect branch. This seems like the right thing to do both
for readability and for platforms on which the indirect branch has a
noticeable overhead."
Signed-off-by: Jisheng Zhang <jszhang@kernel.org>
Tested-by: Emil Renner Berthing <emil.renner.berthing@canonical.com>
Reviewed-by: Conor Dooley <conor.dooley@microchip.com>
Link: https://lore.kernel.org/r/20231114143338.2406-2-jszhang@kernel.org
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Now that we're testing unaligned memory copy and making that
determination generically, there are no more users of the vendor
feature_probe_func(). While I think it's probably going to need to come
back, there are no users right now, so let's remove it until it's
needed.
Signed-off-by: Evan Green <evan@rivosinc.com>
Reviewed-by: Conor Dooley <conor.dooley@microchip.com>
Link: https://lore.kernel.org/r/20230818194136.4084400-3-evan@rivosinc.com
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
|
|
|
|
|
|
|
|
|
|
| |
The CBOM size and whether the HW is noncoherent is known and
determined during booting and won't change after that.
Signed-off-by: Jisheng Zhang <jszhang@kernel.org>
Reviewed-by: Conor Dooley <conor.dooley@microchip.com>
Link: https://lore.kernel.org/r/20230614165504.532-2-jszhang@kernel.org
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
When CONFIG_MODULES is set, __init_or_module becomes <empty>, but when
CONFIG_MODULES is not set, __init_or_module becomes __init.
In the latter case, it causes section mismatch warnings:
WARNING: modpost: vmlinux.o: section mismatch in reference: riscv_fill_cpu_mfr_info (section: .text) -> sifive_errata_patch_func (section: .init.text)
WARNING: modpost: vmlinux.o: section mismatch in reference: riscv_fill_cpu_mfr_info (section: .text) -> thead_errata_patch_func (section: .init.text)
Fixes: bb3f89487fd9 ("RISC-V: hwprobe: Remove __init on probe_vendor_features()")
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Reviewed-by: Evan Green <evan@rivosinc.com>
Reviewed-by: Conor Dooley <conor.dooley@microchip.com>
Link: https://lore.kernel.org/r/20230429155247.12131-1-rdunlap@infradead.org
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>
|
|\
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
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>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
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>
|
|/
|
|
|
|
|
|
|
|
|
|
| |
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>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Guenter reported a splat during boot, that Samuel pointed out was the
lockdep assertion failing in patch_insn_write():
WARNING: CPU: 0 PID: 0 at arch/riscv/kernel/patch.c:63 patch_insn_write+0x222/0x2f6
epc : patch_insn_write+0x222/0x2f6
ra : patch_insn_write+0x21e/0x2f6
epc : ffffffff800068c6 ra : ffffffff800068c2 sp : ffffffff81803df0
gp : ffffffff81a1ab78 tp : ffffffff81814f80 t0 : ffffffffffffe000
t1 : 0000000000000001 t2 : 4c45203a76637369 s0 : ffffffff81803e40
s1 : 0000000000000004 a0 : 0000000000000000 a1 : ffffffffffffffff
a2 : 0000000000000004 a3 : 0000000000000000 a4 : 0000000000000001
a5 : 0000000000000000 a6 : 0000000000000000 a7 : 0000000052464e43
s2 : ffffffff80b4889c s3 : 000000000000082c s4 : ffffffff80b48828
s5 : 0000000000000828 s6 : ffffffff8131a0a0 s7 : 0000000000000fff
s8 : 0000000008000200 s9 : ffffffff8131a520 s10: 0000000000000018
s11: 000000000000000b t3 : 0000000000000001 t4 : 000000000000000d
t5 : ffffffffd8180000 t6 : ffffffff81803bc8
status: 0000000200000100 badaddr: 0000000000000000 cause: 0000000000000003
[<ffffffff800068c6>] patch_insn_write+0x222/0x2f6
[<ffffffff80006a36>] patch_text_nosync+0xc/0x2a
[<ffffffff80003b86>] riscv_cpufeature_patch_func+0x52/0x98
[<ffffffff80003348>] _apply_alternatives+0x46/0x86
[<ffffffff80c02d36>] apply_boot_alternatives+0x3c/0xfa
[<ffffffff80c03ad8>] setup_arch+0x584/0x5b8
[<ffffffff80c0075a>] start_kernel+0xa2/0x8f8
This issue was exposed by 702e64550b12 ("riscv: fpu: switch has_fpu() to
riscv_has_extension_likely()"), as it is the patching in has_fpu() that
triggers the splats in Guenter's report.
Take the text_mutex before doing any code patching to satisfy lockdep.
Fixes: ff689fd21cb1 ("riscv: add RISC-V Svpbmt extension support")
Fixes: a35707c3d850 ("riscv: add memory-type errata for T-Head")
Fixes: 1a0e5dbd3723 ("riscv: sifive: Add SiFive alternative ports")
Reported-by: Guenter Roeck <linux@roeck-us.net>
Link: https://lore.kernel.org/all/20230212154333.GA3760469@roeck-us.net/
Signed-off-by: Conor Dooley <conor.dooley@microchip.com>
Reviewed-by: Samuel Holland <samuel@sholland.org>
Tested-by: Guenter Roeck <linux@roeck-us.net>
Link: https://lore.kernel.org/r/20230212194735.491785-1-conor@kernel.org
Cc: stable@vger.kernel.org
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Now that the text to patch is located using a relative offset from the
alternative entry, the text address should be computed without applying
the kernel mapping offset, both before and after VM setup.
Fixes: 8d23e94a4433 ("riscv: switch to relative alternative entries")
Signed-off-by: Samuel Holland <samuel@sholland.org>
Reviewed-by: Conor Dooley <conor.dooley@microchip.com>
Reviewed-by: Guo Ren <guoren@kernel.org>
Reviewed-by: Jisheng Zhang <jszhang@kernel.org>
Tested-by: Conor Dooley <conor.dooley@microchip.com>
Link: https://lore.kernel.org/r/20230212021534.59121-2-samuel@sholland.org
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Instead of using absolute addresses for both the old instrucions and
the alternative instructions, use offsets relative to the alt_entry
values. So this not only cuts the size of the alternative entry, but
also meets the prerequisite for patching alternatives in the vDSO,
since absolute alternative entries are subject to dynamic relocation,
which is incompatible with the vDSO building.
Signed-off-by: Jisheng Zhang <jszhang@kernel.org>
Reviewed-by: Andrew Jones <ajones@ventanamicro.com>
Reviewed-by: Conor Dooley <conor.dooley@microchip.com>
Link: https://lore.kernel.org/r/20230128172856.3814-10-jszhang@kernel.org
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
With the T-HEAD C9XX cores being designed before or during the ratification
to the SSCOFPMF extension, it implements functionality very similar but
not equal to it.
It implements overflow handling and also some privilege-mode filtering.
While SSCOFPMF supports this for all modes, the C9XX only implements the
filtering for M-mode and S-mode but not user-mode.
So add some adaptions to allow the C9XX to still handle
its PMU through the regular SBI PMU interface instead of defining new
interfaces or drivers.
To work properly, this requires a matching change in SBI, though the actual
interface between kernel and SBI does not change.
The main differences are a the overflow CSR and irq number.
As the reading of the overflow-csr is in the hot-path during irq handling,
use an errata and alternatives to not introduce new conditionals there.
Reviewed-by: Andrew Jones <ajones@ventanamicro.com>
Reviewed-by: Conor Dooley <conor.dooley@microchip.com>
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Link: https://lore.kernel.org/all/20221011231841.2951264-2-heiko@sntech.de/
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
|
|\
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Heiko Stuebner <heiko@sntech.de> says:
As noted by some people, some parts of the recently added extensions
(svpbmt, zicbom) + t-head errata could use some styling upgrades.
So this series provides these.
changes in v2:
- add patch also converting cpufeature probe to BIT()
- update commit message in patch1 (Conor)
Heiko Stuebner (5):
riscv: cleanup svpbmt cpufeature probing
riscv: drop some idefs from CMO initialization
riscv: use BIT() macros in t-head errata init
riscv: use BIT() marco for cpufeature probing
riscv: check for kernel config option in t-head memory types errata
arch/riscv/errata/thead/errata.c | 14 ++++++-----
arch/riscv/include/asm/cacheflush.h | 2 ++
arch/riscv/kernel/cpufeature.c | 39 ++++++++++++-----------------
3 files changed, 26 insertions(+), 29 deletions(-)
Link: https://lore.kernel.org/r/20220905111027.2463297-1-heiko@sntech.de
* b4-shazam-merge:
riscv: check for kernel config option in t-head memory types errata
riscv: use BIT() marco for cpufeature probing
riscv: use BIT() macros in t-head errata init
riscv: drop some idefs from CMO initialization
riscv: cleanup svpbmt cpufeature probing
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The t-head variant of page-based memory types should also check first
for the enabled kernel config option.
Fixes: a35707c3d850 ("riscv: add memory-type errata for T-Head")
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Reviewed-by: Conor Dooley <conor.dooley@microchip.com>
Reviewed-by: Andrew Jones <ajones@ventanamicro.com>
Reviewed-by: Guo Ren <guoren@kernel.org>
Link: https://lore.kernel.org/r/20220905111027.2463297-6-heiko@sntech.de
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Using the appropriate BIT macro makes the code better readable.
Suggested-by: Conor Dooley <conor.dooley@microchip.com>
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Reviewed-by: Guo Ren <guoren@kernel.org>
Reviewed-by: Conor Dooley <conor.dooley@microchip.com>
Reviewed-by: Andrew Jones <ajones@ventanamicro.com>
Link: https://lore.kernel.org/r/20220905111027.2463297-4-heiko@sntech.de
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Wrapping things in #ifdefs makes the code harder to read
while we also have IS_ENABLED() macros to do this in regular code
and the extension detection is not _that_ runtime critical.
So define a stub for riscv_noncoherent_supported() in the
non-CONFIG_RISCV_DMA_NONCOHERENT case and move the code to
us IS_ENABLED.
Suggested-by: Conor Dooley <conor.dooley@microchip.com>
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Reviewed-by: Guo Ren <guoren@kernel.org>
Reviewed-by: Conor Dooley <conor.dooley@microchip.com>
Reviewed-by: Andrew Jones <ajones@ventanamicro.com>
Link: https://lore.kernel.org/r/20220905111027.2463297-3-heiko@sntech.de
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
|
|/
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This fixes two issues: I truncated the warning's hart ID when porting to
the 64-bit hart ID code, and the original code's warning handling could
fire on an uninitialized hart ID.
The biggest change here is that riscv_cbom_block_size is no longer
initialized, as IMO the default isn't sane: there's nothing in the ISA
that mandates any specific cache block size, so falling back to one will
just silently produce the wrong answer on some systems. This also
changes the probing order so the cache block size is known before
enabling Zicbom support.
CC: stable@vger.kernel.org
CC: Andrew Jones <ajones@ventanamicro.com>
CC: Heiko Stuebner <heiko@sntech.de>
CC: Atish Patra <atishp@rivosinc.com>
Fixes: 3aefb2ee5bdd ("riscv: implement Zicbom-based CMO instructions + the t-head variant")
Fixes: 1631ba1259d6 ("riscv: Add support for non-coherent devices using zicbom extension")
Reported-by: kernel test robot <lkp@intel.com>
Reported-by: Conor Dooley <conor.dooley@microchip.com>
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
[Conor: fixed the redefinition errors]
Tested-by: Conor Dooley <conor.dooley@microchip.com>
Signed-off-by: Conor Dooley <conor.dooley@microchip.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20220912224800.998121-1-mail@conchuod.ie
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The T-Head C906 and C910 implement a scheme for handling
cache operations different from the generic Zicbom extension.
Add an errata for it next to the generic dma coherency ops.
Reviewed-by: Samuel Holland <samuel@sholland.org>
Tested-by: Samuel Holland <samuel@sholland.org>
Reviewed-by: Guo Ren <guoren@kernel.org>
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Link: https://lore.kernel.org/r/20220706231536.2041855-5-heiko@sntech.de
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Having a list of alternatives to check with a per-entry function pointer
to a check function is nice style-wise. But in case of early-alternatives
it can clash with the non-relocated kernel and the function pointer in
the list pointing to a completely wrong location.
This isn't an issue with one or two list entries, as in that case the
compiler seems to unroll the loop and even usage of the list structure
and then only does relative jumps into the check functions based on this.
When adding a third entry to either list though, the issue that was
hiding there from the beginning is triggered resulting a jump to a
memory address that isn't part of the kernel at all.
The list of features/erratas only contained an unused name and the
pointer to the check function, so an easy solution for the problem
is to just unroll the loop in code, dismantle the whole list structure
and just call the relevant check functions one by one ourself.
For the T-Head errata this includes moving the stage-check inside
the check functions.
The issue is only relevant for things that might be called for early-
alternatives (T-Head and possible future main extensions), so the
SiFive erratas were not affected from the beginning, as they got
an early return for early-alternatives in the original patchset.
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Tested-by: Samuel Holland <samuel@sholland.org>
Link: https://lore.kernel.org/r/20220526205646.258337-6-heiko@sntech.de
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
|
|
Some current cpus based on T-Head cores implement memory-types
way different than described in the svpbmt spec even going
so far as using PTE bits marked as reserved.
Add the T-Head vendor-id and necessary errata code to
replace the affected instructions.
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Tested-by: Samuel Holland <samuel@sholland.org>
Link: https://lore.kernel.org/r/20220511192921.2223629-13-heiko@sntech.de
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
|