riscv: Add vector ISA support
Subsystems (8)
v21 cover riscv: Add vector ISA support
This is the v21 patch series for adding Vector extension support in Linux. Please refer to [1] for the introduction of the patchset. The v21 patch series was aimed to solve build issues from v19, provide usage guideline for the prctl interface, and address review comments on v20.
Thank every one who has been reviewing, suggesting on the topic. Hope this get a step closer to the final merge.
Here points out where changes are located:
Updated patches: 10, 11, 21, 25
New patches: 20
Unchanged patches: 1, 2, 3, 4, 5, 6, 7, 8, 9, 12, 13, 14, 15, 16, 17, 18, 19, 22, 23, 24, 26, 27
Source tree:
https://github.com/sifive/riscv-linux/tree/riscv/for-next/vector-v21
Links:
- [1] https://lore.kernel.org/all/20230518161949.11203-1-andy.chiu at sifive.com/
---
Changelog V21
- Add usage guideline for the prctl interface (R?mi, patch 25)
- Properly define macros to prevent build fails (Bj?rn, patch 21)
- expose ELF_HWCAP as a function so we can set bits individually for processes. (patch 20)
- Turn off V in ELF_HWCAP when user is not allowed to use it. (R?mi, patch 21)
- Send SIGBUS to indicate the allocation for V context fails. (Darius, patch 11)
- Refine checks in riscv_v_first_use_handler(). (patch 11, 21)
- Refine location of callsite to disable Vector when forked (patch 10)
Changelog V20
- Add .gitignore into hwprobe (patch 26)
- Implement test for prctl and fix issues (patch 20, 25)
- Fix compile error with allmodconfig (patch 20).
- Check elf_hwcap in first-use trap (patch 11).
- Refine code (patch 11, 20, 21, 23, 24)
- Properly add V entry into hwprobe (patch 3).
- Fix typos (patch 3, 24)
- Use "_unlikely" to detect V since there is no public available hw that supports it (patch 2).
Changelog V19
- Rebase to the latest -next branch (at 6.4-rc1 ac9a786). Solve conflicts at patch 14, 15, and 19.
- Add a sysctl, and prctl intefaces for userspace Vector control, and a document for it. (patch 20, 21, 24)
- Add a Kconfig RISCV_V_DISABLE to set the default value of userspace Vector enablement status at compile-time. (patch 23)
- Allow hwprobe interface to probe Vector. (patch 3)
- Fix typos and commit msg at patch 6 and 8.
Changelog V18
- Rebase to the latest -next branch (at 9c2598d)
- patch 7: Detect inconsistent VLEN setup on an SMP system (Heiko).
- patch 10: Add blank lines (Heiko)
- patch 10: Return immediately in insn_is_vector() if an insn matches (Heiko)
- patch 11: Use sizeof(vstate->datap) instead of sizeof(void*) (Eike)
Changelog V17
- Rebase to the latest -next branch (at e45d6a5):
- Solve conflicts at 9 and 13 due to generic entry
- Use generic entry in do_trap_insn_illegal() trap handler
Changelog V16
- Rebase to the latest for-next (at 4b74077):
- Solve conflicts at 7, and 17
- Use as-instr to detect if assembler supports .option arch directive and remove dependency from GAS, for both ZBB and V.
- Cleanup code in KVM vector
- Address issue reported by sparse
- Refine code:
- Fix a mixed-use of space/tab
- Remove new lines at the end of file
Changelog V15
- Rebase to risc-v -next (v6.3-rc1)
- Make V depend on FD in Kconfig according to the spec and shut off v properly.
- Fix a syntax error for clang build. But mark RISCV_ISA_V GAS only due to https://reviews.llvm.org/D123515
- Use scratch reg in inline asm instead of t4.
- Refine code.
- Cleanup per-patch changelogs.
Changelog V14
- Rebase to risc-v -next (v6.2-rc7)
- Use TOOLCHAIN_HAS_V to detect if we can enable Vector. And refine KBUILD_CFLAGS to remove v from default compile option.
- Drop illegal instruction handling patch in kvm and leave it to a independent series[3]. The series has merged into 6.3-rc1
- Move KVM_RISCV_ISA_EXT_V to the end of enum to prevent potential ABI breaks.
- Use PT_SIZE_ON_STACK instead of PT_SIZE to fit alignment. Also, remove panic log from v13 (15/19) because it is no longer relevant.
- Rewrite insn_is_vector for better structuring (change if-else chain to a switch)
- Fix compilation error in the middle of the series
- Validate size of the alternative signal frame if V is enabled whenever:
- The user call sigaltstack to update altstack
- A signal is being delivered
- Rename __riscv_v_state to __riscv_v_ext_state.
- Add riscv_v_ prefix and rename rvv appropriately
- Organize riscv_v_vsize setup code into vector.c
- Address the issue mentioned by Heiko on !FPU case
- Honor orignal authors that got changed accidentally in v13 4,5,6
Changelog V13
- Rebase to latest risc-v next (v6.2-rc1)
- vineetg: Re-organize the series to comply with bisect-ability
- andy.chiu: Improve task switch with inline assembly
- Re-structure the signal frame to avoid user ABI break.
- Implemnt first-use trap and drop prctl for per-task V state enablement. Also, redirect this trap from hs to vs for kvm setup.
- Do not expose V context in ptrace/sigframe until the task start using V. But still reserve V context for size ofsigframe reported by auxv.
- Drop the kernel mode vector and leave it to another (future) series.
Changelog V12 (Chris)
- rebases to some point after v5.18-rc6
- add prctl to control per-process V state
Chnagelog V10
- Rebase to v5.18-rc6
- Merge several patches
- Refine codes
- Fix bugs
- Add kvm vector support
Changelog V9
- Rebase to v5.15
- Merge several patches
- Refine codes
- Fix a kernel panic issue
Changelog V8
- Rebase to v5.14
- Refine struct __riscv_v_ext_state with struct __riscv_ctx_hdr
- Refine has_vector into a static key
- Defined __reserved space in struct sigcontext for vector and future extensions
Changelog V7
- Add support for kernel mode vector
- Add vector extension XOR implementation
- Optimize task switch codes of vector
- Allocate space for vector registers in start_thread()
- Fix an illegal instruction exception when accessing vlenb
- Optimize vector registers initialization
- Initialize vector registers with proper vsetvli then it can work normally
- Refine ptrace porting due to generic API changed
- Code clean up
Changelog V6
- Replace vle.v/vse.v instructions with vle8.v/vse8.v based on 0.9 spec
- Add comments based on mailinglist feedback
- Fix rv32 build error
Changelog V5
- Using regset_size() correctly in generic ptrace
- Fix the ptrace porting
- Fix compile warning
Changelog V4
- Support dynamic vlen
- Fix bugs: lazy save/resotre, not saving vtype
- Update VS bit offset based on latest vector spec
- Add new vector csr based on latest vector spec
- Code refine and removed unused macros
Changelog V3
- Rebase linux-5.6-rc3 and tested with qemu
- Seperate patches with Anup's advice
- Give out a ABI puzzle with unlimited vlen
Changelog V2
- Fixup typo "vecotr, fstate_save->vstate_save".
- Fixup wrong saved registers' length in vector.S.
- Seperate unrelated patches from this one.
Andy Chiu (11):
riscv: hwprobe: Add support for probing V in
RISCV_HWPROBE_KEY_IMA_EXT_0
riscv: Allocate user's vector context in the first-use trap
riscv: signal: check fp-reserved words unconditionally
riscv: signal: validate altstack to reflect Vector
riscv: hwcap: change ELF_HWCAP to a function
riscv: Add prctl controls for userspace vector management
riscv: Add sysctl to set the default vector rule for new processes
riscv: detect assembler support for .option arch
riscv: Add documentation for Vector
selftests: Test RISC-V Vector prctl interface
selftests: add .gitignore file for RISC-V hwprobe
Greentime Hu (9):
riscv: Add new csr defines related to vector extension
riscv: Clear vector regfile on bootup
riscv: Introduce Vector enable/disable helpers
riscv: Introduce riscv_v_vsize to record size of Vector context
riscv: Introduce struct/helpers to save/restore per-task Vector state
riscv: Add task switch support for vector
riscv: Add ptrace vector support
riscv: signal: Add sigcontext save/restore for vector
riscv: prevent stack corruption by reserving task_pt_regs(p) early
Guo Ren (4):
riscv: Rename __switch_to_aux() -> fpu
riscv: Extending cpufeature.c to detect V-extension
riscv: Disable Vector Instructions for kernel itself
riscv: Enable Vector code to be built
Vincent Chen (3):
riscv: signal: Report signal frame size to userspace via auxv
riscv: kvm: Add V extension to KVM ISA
riscv: KVM: Add vector lazy save/restore support
Documentation/riscv/hwprobe.rst | 3 + Documentation/riscv/index.rst | 1 + Documentation/riscv/vector.rst | 132 +++++++++ arch/riscv/Kconfig | 39 ++- arch/riscv/Makefile | 6 +- arch/riscv/include/asm/csr.h | 18 +- arch/riscv/include/asm/elf.h | 11 +- arch/riscv/include/asm/hwcap.h | 3 + arch/riscv/include/asm/insn.h | 29 ++ arch/riscv/include/asm/kvm_host.h | 2 + arch/riscv/include/asm/kvm_vcpu_vector.h | 82 ++++++ arch/riscv/include/asm/processor.h | 13 + arch/riscv/include/asm/switch_to.h | 9 +- arch/riscv/include/asm/thread_info.h | 3 + arch/riscv/include/asm/vector.h | 184 ++++++++++++ arch/riscv/include/uapi/asm/auxvec.h | 1 + arch/riscv/include/uapi/asm/hwcap.h | 1 + arch/riscv/include/uapi/asm/hwprobe.h | 1 + arch/riscv/include/uapi/asm/kvm.h | 8 + arch/riscv/include/uapi/asm/ptrace.h | 39 +++ arch/riscv/include/uapi/asm/sigcontext.h | 16 +- arch/riscv/kernel/Makefile | 1 + arch/riscv/kernel/cpufeature.c | 25 ++ arch/riscv/kernel/entry.S | 6 +- arch/riscv/kernel/head.S | 41 ++- arch/riscv/kernel/process.c | 20 ++ arch/riscv/kernel/ptrace.c | 70 +++++ arch/riscv/kernel/setup.c | 3 + arch/riscv/kernel/signal.c | 220 ++++++++++++-- arch/riscv/kernel/smpboot.c | 7 + arch/riscv/kernel/sys_riscv.c | 4 + arch/riscv/kernel/traps.c | 26 +- arch/riscv/kernel/vector.c | 276 ++++++++++++++++++ arch/riscv/kvm/Makefile | 1 + arch/riscv/kvm/vcpu.c | 25 ++ arch/riscv/kvm/vcpu_vector.c | 186 ++++++++++++ include/uapi/linux/elf.h | 1 + include/uapi/linux/prctl.h | 11 + kernel/sys.c | 12 + tools/testing/selftests/riscv/Makefile | 2 +- .../selftests/riscv/hwprobe/.gitignore | 1 + .../testing/selftests/riscv/vector/.gitignore | 2 + tools/testing/selftests/riscv/vector/Makefile | 15 + .../riscv/vector/vstate_exec_nolibc.c | 111 +++++++ .../selftests/riscv/vector/vstate_prctl.c | 189 ++++++++++++ 45 files changed, 1805 insertions(+), 51 deletions(-) create mode 100644 Documentation/riscv/vector.rst create mode 100644 arch/riscv/include/asm/kvm_vcpu_vector.h create mode 100644 arch/riscv/include/asm/vector.h create mode 100644 arch/riscv/kernel/vector.c create mode 100644 arch/riscv/kvm/vcpu_vector.c create mode 100644 tools/testing/selftests/riscv/hwprobe/.gitignore create mode 100644 tools/testing/selftests/riscv/vector/.gitignore create mode 100644 tools/testing/selftests/riscv/vector/Makefile create mode 100644 tools/testing/selftests/riscv/vector/vstate_exec_nolibc.c create mode 100644 tools/testing/selftests/riscv/vector/vstate_prctl.c
Signature · 1 line
2.17.1
PATCH v21 1/27 riscv: Rename __switch_to_aux() -> fpu
From: Guo Ren <[email protected]>
The name of __switch_to_aux() is not clear and rename it with the determine function: __switch_to_fpu(). Next we could add other regs' switch.
--- arch/riscv/include/asm/switch_to.h | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
arch/riscv/include/asm/switch_to.h +3 -3
@@ -46,7 +46,7 @@ static inline void fstate_restore(struct task_struct *task, } } -static inline void __switch_to_aux (struct task_struct *prev, +static inline void __switch_to_fpu (struct task_struct *prev,struct task_struct *next) {struct pt_regs *regs;
@@ -66,7 +66,7 @@ static __always_inline bool has_fpu(void) static __always_inline bool has_fpu(void) { return false; }#define fstate_save (task, regs) do { } while (0)#define fstate_restore (task, regs) do { } while (0) -#define __switch_to_aux (__prev, __next) do { } while (0) +#define __switch_to_fpu (__prev, __next) do { } while (0)#endif extern struct task_struct *__switch_to(struct task_struct *,
@@ -77,7 +77,7 @@ do { \struct task_struct *__prev = (prev); \struct task_struct *__next = (next); \if (has_fpu ()) \ -__switch_to_aux (__prev, __next); \ +__switch_to_fpu (__prev, __next); \ ((last) =__switch_to (__prev, __next)); \ }while (0 )
Signature · 1 line
2.17.1
PATCH v21 2/27 riscv: Extending cpufeature.c to detect V-extension
In reply to Andy ChiuFrom: Guo Ren <[email protected]>
Add V-extension into riscv_isa_ext_keys array and detect it with isa string parsing.
---
Changelog v20:
s/riscv_has_extension_likely/riscv_has_extension_unlikely/ (Palmer)
--- arch/riscv/include/asm/hwcap.h | 1 + arch/riscv/include/asm/vector.h | 26 ++++++++++++++++++++++++++ arch/riscv/include/uapi/asm/hwcap.h | 1 + arch/riscv/kernel/cpufeature.c | 11 +++++++++++ 4 files changed, 39 insertions(+) create mode 100644 arch/riscv/include/asm/vector.h
arch/riscv/include/asm/hwcap.h +1 -0
@@ -22,6 +22,7 @@#define RISCV_ISA_EXT_m ('m' - 'a')#define RISCV_ISA_EXT_s ('s' - 'a')#define RISCV_ISA_EXT_u ('u' - 'a') +#define RISCV_ISA_EXT_v ('v' - 'a')/* * These macros represent the logical IDs of each multi-letter RISC-V ISA
arch/riscv/include/asm/vector.h new file +26 -0
@@ -0,0 +1,26 @@ +/* SPDX-License-Identifier: GPL-2.0-or-later */ +/* +* Copyright (C) 2020 SiFive +*/ + +#ifndef __ASM_RISCV_VECTOR_H +#define __ASM_RISCV_VECTOR_H + +#include <linux/types.h> + +#ifdef CONFIG_RISCV_ISA_V + +#include <asm/hwcap.h> + +static __always_inline bool has_vector(void) +{ +return riscv_has_extension_unlikely (RISCV_ISA_EXT_v); +} + +#else /* ! CONFIG_RISCV_ISA_V */ + +static __always_inline bool has_vector(void) { return false; } + +#endif /* CONFIG_RISCV_ISA_V */ + +#endif /* ! __ASM_RISCV_VECTOR_H */
arch/riscv/include/uapi/asm/hwcap.h +1 -0
@@ -21,5 +21,6 @@#define COMPAT_HWCAP_ISA_F (1 << ('F' - 'A'))#define COMPAT_HWCAP_ISA_D (1 << ('D' - 'A'))#define COMPAT_HWCAP_ISA_C (1 << ('C' - 'A')) +#define COMPAT_HWCAP_ISA_V (1 << ('V' - 'A'))#endif /* _UAPI_ASM_RISCV_HWCAP_H */
arch/riscv/kernel/cpufeature.c +11 -0
@@ -107,6 +107,7 @@ void __init riscv_fill_hwcap(void) isa2hwcap['f' -'a' ] =COMPAT_HWCAP_ISA_F ; isa2hwcap['d' -'a' ] =COMPAT_HWCAP_ISA_D ; isa2hwcap['c' -'a' ] =COMPAT_HWCAP_ISA_C ; + isa2hwcap['v' -'a' ] =COMPAT_HWCAP_ISA_V ; elf_hwcap =0 ;
@@ -267,6 +268,16 @@ void __init riscv_fill_hwcap(void) elf_hwcap &= ~COMPAT_HWCAP_ISA_F ; } +if (elf_hwcap &COMPAT_HWCAP_ISA_V ) { +/* +* ISA string in device tree might have 'v' flag, but +* CONFIG_RISCV_ISA_V is disabled in kernel. +* Clear V flag in elf_hwcap if CONFIG_RISCV_ISA_V is disabled. +*/ +if (!IS_ENABLED (CONFIG_RISCV_ISA_V )) + elf_hwcap &= ~COMPAT_HWCAP_ISA_V ; + } +memset (print_str,0 ,sizeof (print_str));for (i =0 , j =0 ; i <NUM_ALPHA_EXTS ; i++)if (riscv_isa[0 ] &BIT_MASK (i))
Signature · 1 line
2.17.1
PATCH v21 3/27 riscv: hwprobe: Add support for probing V in RISCV_HWPROBE_KEY_IMA_EXT_0
In reply to Andy ChiuProbing kernel support for Vector extension is available now. This only add detection for V only. Extenions like Zvfh, Zk are not in this scope.
---
Changelog v20:
- Fix a typo in document, and remove duplicated probes (Heiko)
- probe V extension in RISCV_HWPROBE_KEY_IMA_EXT_0 key only (Palmer, Evan)
--- Documentation/riscv/hwprobe.rst | 3 +++ arch/riscv/include/uapi/asm/hwprobe.h | 1 + arch/riscv/kernel/sys_riscv.c | 4 ++++ 3 files changed, 8 insertions(+)
Documentation/riscv/hwprobe.rst +3 -0
@@ -64,6 +64,9 @@ The following keys are defined: * :c:macro:`RISCV_HWPROBE_IMA_C`: The C extension is supported, as defined by version 2.2 of the RISC-V ISA manual. + * :c:macro:`RISCV_HWPROBE_IMA_V`: The V extension is supported, as defined by + version 1.0 of the RISC-V Vector extension manual. + * :c:macro:`RISCV_HWPROBE_KEY_CPUPERF_0`: A bitmask that contains performance information about the selected set of processors.
arch/riscv/include/uapi/asm/hwprobe.h +1 -0
@@ -25,6 +25,7 @@ struct riscv_hwprobe {#define RISCV_HWPROBE_KEY_IMA_EXT_0 4#define RISCV_HWPROBE_IMA_FD (1 << 0)#define RISCV_HWPROBE_IMA_C (1 << 1) +#define RISCV_HWPROBE_IMA_V (1 << 2)#define RISCV_HWPROBE_KEY_CPUPERF_0 5#define RISCV_HWPROBE_MISALIGNED_UNKNOWN (0 << 0)#define RISCV_HWPROBE_MISALIGNED_EMULATED (1 << 0)
arch/riscv/kernel/sys_riscv.c +4 -0
@@ -10,6 +10,7 @@#include <asm/cpufeature.h> #include <asm/hwprobe.h> #include <asm/sbi.h> +#include <asm/vector.h> #include <asm/switch_to.h> #include <asm/uaccess.h> #include <asm/unistd.h>
@@ -171,6 +172,9 @@ static void hwprobe_one_pair(struct riscv_hwprobe *pair,if (riscv_isa_extension_available (NULL , c)) pair->value |=RISCV_HWPROBE_IMA_C ; +if (has_vector ()) + pair->value |=RISCV_HWPROBE_IMA_V ; +break ;case RISCV_HWPROBE_KEY_CPUPERF_0 :
Signature · 1 line
2.17.1
Re: [PATCH -next v21 03/27] riscv: hwprobe: Add support for probing V in RISCV_HWPROBE_KEY_IMA_EXT_0
Am Montag, 5. Juni 2023, 13:07:00 CEST schrieb Andy Chiu:
Quoted 7 lines
Probing kernel support for Vector extension is available now. This only add detection for V only. Extenions like Zvfh, Zk are not in this scope.
Signed-off-by Andy Chiu <[email protected]>Reviewed-by Conor Dooley <[email protected]>Reviewed-by Evan Green <[email protected]>Reviewed-by Palmer Dabbelt <[email protected]>
Re: [PATCH -next v21 03/27] riscv: hwprobe: Add support for probing V in RISCV_HWPROBE_KEY_IMA_EXT_0
In reply to Andy ChiuQuoted 59 lines · Andy Chiu
Probing kernel support for Vector extension is available now. This only add detection for V only. Extenions like Zvfh, Zk are not in this scope.
Signed-off-by Andy Chiu <[email protected]>Reviewed-by Conor Dooley <[email protected]>Reviewed-by Evan Green <[email protected]>Reviewed-by Palmer Dabbelt <[email protected]>---
Changelog v20:
- Fix a typo in document, and remove duplicated probes (Heiko)
- probe V extension inRISCV_HWPROBE_KEY_IMA_EXT_0key only (Palmer, Evan)--- Documentation/riscv/hwprobe.rst | 3 +++ arch/riscv/include/uapi/asm/hwprobe.h | 1 + arch/riscv/kernel/sys_riscv.c | 4 ++++ 3 files changed, 8 insertions(+)Documentation/riscv/hwprobe.rst +3
-0@@ -64,6 +64,9 @@ The following keys are defined: * :c:macro:`RISCV_HWPROBE_IMA_C`: The C extension is supported, as defined by version 2.2 of the RISC-V ISA manual. + * :c:macro:`RISCV_HWPROBE_IMA_V`: The V extension is supported, as defined by + version 1.0 of the RISC-V Vector extension manual. + * :c:macro:`RISCV_HWPROBE_KEY_CPUPERF_0`: A bitmask that contains performance information about the selected set of processors.arch/riscv/include/uapi/asm/hwprobe.h +1
-0@@ -25,6 +25,7 @@ struct riscv_hwprobe {#define RISCV_HWPROBE_KEY_IMA_EXT_0 4#define RISCV_HWPROBE_IMA_FD (1 << 0)#define RISCV_HWPROBE_IMA_C (1 << 1) +#define RISCV_HWPROBE_IMA_V (1 << 2)#define RISCV_HWPROBE_KEY_CPUPERF_0 5#define RISCV_HWPROBE_MISALIGNED_UNKNOWN (0 << 0)#define RISCV_HWPROBE_MISALIGNED_EMULATED (1 << 0)arch/riscv/kernel/sys_riscv.c +1
-0@@ -10,6 +10,7 @@#include <asm/cpufeature.h> #include <asm/hwprobe.h> #include <asm/sbi.h> +#include <asm/vector.h> #include <asm/switch_to.h> #include <asm/uaccess.h> #include <asm/unistd.h>
arch/riscv/kernel/sys_riscv.c static void hwprobe_one_pair(struct riscv_hwprobe · line 179
Quoted 8 lines
arch/riscv/kernel/sys_riscv.c +3
-0@@ -171,6 +172,9 @@ static void hwprobe_one_pair(struct riscv_hwprobe *pair, if (riscv_isa_extension_available(NULL, c)) pair->value |= RISCV_HWPROBE_IMA_C; +if (has_vector ()) + pair->value |=RISCV_HWPROBE_IMA_V ; +break ;
I am concerned by the exception this is making. I believe the intention of riscv_hwprobe is to replace AT_HWCAP as the single point of truth for userspace to make instruction use decisions. Since this does not check riscv_v_vstate_ctrl_user_allowed, application code which wants to know if V instructions are usable must use AT_HWCAP instead, unlike all other extensions for which the relevant data is available within the hwprobe return.
Assuming this is intentional, what is the path forward for future extensions that cannot be used from userspace without additional conditions being met? For instance, if we add support in the future for the Zve* extensions, the V bit would not be set in HWCAP for them, which would require library code to use the prctl interface unless we define the hwcap bits to imply userspace usability.
-s
Quoted 9 lines
arch/riscv/kernel/sys_riscv.c +0
-0case RISCV_HWPROBE_KEY_CPUPERF_0 :Signature · 7 lines
2.17.1 _______________________________________________ linux-riscv mailing list linux-riscv at lists.infradead.org http://lists.infradead.org/mailman/listinfo/linux-riscv
Palmer Dabbelt
maintainerRe: [PATCH -next v21 03/27] riscv: hwprobe: Add support for probing V in RISCV_HWPROBE_KEY_IMA_EXT_0
Quoted 64 lines · sorear at fastmail.com
Quoted 63 lines · Andy Chiu
Probing kernel support for Vector extension is available now. This only add detection for V only. Extenions like Zvfh, Zk are not in this scope.
Signed-off-by Andy Chiu <[email protected]>Reviewed-by Conor Dooley <[email protected]>Reviewed-by Evan Green <[email protected]>Reviewed-by Palmer Dabbelt <[email protected]>---
Changelog v20:
- Fix a typo in document, and remove duplicated probes (Heiko)
- probe V extension inRISCV_HWPROBE_KEY_IMA_EXT_0key only (Palmer, Evan)--- Documentation/riscv/hwprobe.rst | 3 +++ arch/riscv/include/uapi/asm/hwprobe.h | 1 + arch/riscv/kernel/sys_riscv.c | 4 ++++ 3 files changed, 8 insertions(+)Documentation/riscv/hwprobe.rst +3
-0@@ -64,6 +64,9 @@ The following keys are defined: * :c:macro:`RISCV_HWPROBE_IMA_C`: The C extension is supported, as defined by version 2.2 of the RISC-V ISA manual. + * :c:macro:`RISCV_HWPROBE_IMA_V`: The V extension is supported, as defined by + version 1.0 of the RISC-V Vector extension manual. + * :c:macro:`RISCV_HWPROBE_KEY_CPUPERF_0`: A bitmask that contains performance information about the selected set of processors.arch/riscv/include/uapi/asm/hwprobe.h +1
-0@@ -25,6 +25,7 @@ struct riscv_hwprobe {#define RISCV_HWPROBE_KEY_IMA_EXT_0 4#define RISCV_HWPROBE_IMA_FD (1 << 0)#define RISCV_HWPROBE_IMA_C (1 << 1) +#define RISCV_HWPROBE_IMA_V (1 << 2)#define RISCV_HWPROBE_KEY_CPUPERF_0 5#define RISCV_HWPROBE_MISALIGNED_UNKNOWN (0 << 0)#define RISCV_HWPROBE_MISALIGNED_EMULATED (1 << 0)arch/riscv/kernel/sys_riscv.c +1
-0@@ -10,6 +10,7 @@#include <asm/cpufeature.h> #include <asm/hwprobe.h> #include <asm/sbi.h> +#include <asm/vector.h> #include <asm/switch_to.h> #include <asm/uaccess.h> #include <asm/unistd.h> @@ -171,6 +172,9 @@ static void hwprobe_one_pair(struct riscv_hwprobe *pair, if (riscv_isa_extension_available(NULL, c)) pair->value |= RISCV_HWPROBE_IMA_C;
Quoted 11 lines
arch/riscv/kernel/sys_riscv.c static void hwprobe_one_pair(struct riscv_hwprobe *pair, · line 178
Quoted 4 lines
arch/riscv/kernel/sys_riscv.c +3
-0+if (has_vector ()) + pair->value |=RISCV_HWPROBE_IMA_V ; +break ;I am concerned by the exception this is making. I believe the intention of
riscv_hwprobeis to replaceAT_HWCAPas the single point of truth for userspace to make instruction use decisions. Since this does not checkriscv_v_vstate_ctrl_user_allowed, application code which wants to know if V instructions are usable must useAT_HWCAPinstead, unlike all other extensions for which the relevant data is available within the hwprobe return.
I guess we were vague in the docs about what "supported" means, but IIRC the goal was for riscv_hwprobe() to indicate what's supported by both the HW and the kernel. In other words, hwprobe should indicate what's possible to enable -- even if there's some additional steps necessary to enable it.
We can at least make this a little more explicit with something like
diff --git a/Documentation/riscv/hwprobe.rst b/Documentation/riscv/hwprobe.rst index 19165ebd82ba..7f82a5385bc3 100644
--- a/Documentation/riscv/hwprobe.rst
+++ b/Documentation/riscv/hwprobe.rst
@@ -27,6 +27,13 @@ AND of the values for the specified CPUs. Usermode can supply NULL for cpus and
0 for cpu_count as a shortcut for all online CPUs. There are currently no flags, this value must be zero for future compatibility.
+Calls to sys_riscv_hwprobe() indicate the features supported by both the +kernel and the hardware that the system is running on. For example, if the +hardware supports the V extension and the kernel has V support enabled then +RISCV_HWPROBE_KEY_IMA_EXT_0/RISCV_HWPROBE_IMA_V will be set even if the V +extension is disabled via a userspace-controlled tunable such as +PR_RISCV_V_SET_CONTROL.
+
On success 0 is returned, on failure a negative error code is returned.
The following keys are defined:
@@ -65,7 +72,10 @@ The following keys are defined:
by version 2.2 of the RISC-V ISA manual.
* :c:macro:RISCV_HWPROBE_IMA_V: The V extension is supported, as defined by
- version 1.0 of the RISC-V Vector extension manual.
+ version 1.0 of the RISC-V Vector extension manual. For strict uABI
+ compatibility some systems may disable V by default even when the hardware
+ supports in, in which case users must call `prctl(PR_RISCV_V_SET_CONTROL,
+ ...` to explicitly allow V to be used.
* :c:macro:RISCV_HWPROBE_EXT_ZBA: The Zba address generation extension is supported, as defined in version 1.0 of the Bit-Manipulation ISA
IMO that's the better way to go that to require that userspace tries to enable V via the prctl() first, but we haven't released this yet so in theory we could still change it.
We'd have a similar discussion for some of the counters that need to feed through the perf interface, though those are still in flight...
Quoted 6 lines
Assuming this is intentional, what is the path forward for future extensions that cannot be used from userspace without additional conditions being met? For instance, if we add support in the future for the Zve* extensions, the V bit would not be set in HWCAP for them, which would require library code to use the prctl interface unless we define the hwcap bits to imply userspace usability.
In this case a system that supports some of the Zve extensions but not the full V extension would not be probably from userspace, as V would not be set anywhere. The way to support that would be to add new bits into hwprobe to indicate those extensions, it just wasn't clear that anyone was interested in building Linux-flavored systems that supported only some a strict subset of V.
Happy to see patches if you know of some hardware in the pipeline, though ;)
Quoted 12 lines
-s
Quoted 9 lines
arch/riscv/kernel/sys_riscv.c +0
-1case RISCV_HWPROBE_KEY_CPUPERF_0 : --2.17.1
_______________________________________________
linux-riscv mailing list
linux-riscv at lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
Re: [PATCH -next v21 03/27] riscv: hwprobe: Add support for probing V in RISCV_HWPROBE_KEY_IMA_EXT_0
Quoted 77 lines · Palmer Dabbelt
Quoted 64 lines · sorear at fastmail.com
Quoted 63 lines · Andy Chiu
Probing kernel support for Vector extension is available now. This only add detection for V only. Extenions like Zvfh, Zk are not in this scope.
Signed-off-by Andy Chiu <[email protected]>Reviewed-by Conor Dooley <[email protected]>Reviewed-by Evan Green <[email protected]>Reviewed-by Palmer Dabbelt <[email protected]>---
Changelog v20:
- Fix a typo in document, and remove duplicated probes (Heiko)
- probe V extension inRISCV_HWPROBE_KEY_IMA_EXT_0key only (Palmer, Evan)--- Documentation/riscv/hwprobe.rst | 3 +++ arch/riscv/include/uapi/asm/hwprobe.h | 1 + arch/riscv/kernel/sys_riscv.c | 4 ++++ 3 files changed, 8 insertions(+)Documentation/riscv/hwprobe.rst +3
-0@@ -64,6 +64,9 @@ The following keys are defined: * :c:macro:`RISCV_HWPROBE_IMA_C`: The C extension is supported, as defined by version 2.2 of the RISC-V ISA manual. + * :c:macro:`RISCV_HWPROBE_IMA_V`: The V extension is supported, as defined by + version 1.0 of the RISC-V Vector extension manual. + * :c:macro:`RISCV_HWPROBE_KEY_CPUPERF_0`: A bitmask that contains performance information about the selected set of processors.arch/riscv/include/uapi/asm/hwprobe.h +1
-0@@ -25,6 +25,7 @@ struct riscv_hwprobe {#define RISCV_HWPROBE_KEY_IMA_EXT_0 4#define RISCV_HWPROBE_IMA_FD (1 << 0)#define RISCV_HWPROBE_IMA_C (1 << 1) +#define RISCV_HWPROBE_IMA_V (1 << 2)#define RISCV_HWPROBE_KEY_CPUPERF_0 5#define RISCV_HWPROBE_MISALIGNED_UNKNOWN (0 << 0)#define RISCV_HWPROBE_MISALIGNED_EMULATED (1 << 0)arch/riscv/kernel/sys_riscv.c +1
-0@@ -10,6 +10,7 @@#include <asm/cpufeature.h> #include <asm/hwprobe.h> #include <asm/sbi.h> +#include <asm/vector.h> #include <asm/switch_to.h> #include <asm/uaccess.h> #include <asm/unistd.h> @@ -171,6 +172,9 @@ static void hwprobe_one_pair(struct riscv_hwprobe *pair, if (riscv_isa_extension_available(NULL, c)) pair->value |= RISCV_HWPROBE_IMA_C;Quoted 11 lines
arch/riscv/kernel/sys_riscv.c static void hwprobe_one_pair(struct riscv_hwprobe *pair, · line 178
Quoted 4 lines
arch/riscv/kernel/sys_riscv.c +3
-0+if (has_vector ()) + pair->value |=RISCV_HWPROBE_IMA_V ; +break ;I am concerned by the exception this is making. I believe the intention of
riscv_hwprobeis to replaceAT_HWCAPas the single point of truth for userspace to make instruction use decisions. Since this does not checkriscv_v_vstate_ctrl_user_allowed, application code which wants to know if V instructions are usable must useAT_HWCAPinstead, unlike all other extensions for which the relevant data is available within the hwprobe return.
Quoted 3 lines
I guess we were vague in the docs about what "supported" means, but IIRC the goal was for
riscv_hwprobe()to indicate what's supported by both the HW and the kernel. In other words, hwprobe should indicate what's
Should this be "the HW, the firmware, and the kernel" in the cases where it matters, or are you considering the firmware part of the kernel's view of the hardware?
Quoted 45 lines
possible to enable -- even if there's some additional steps necessary to enable it.
We can at least make this a little more explicit with something like
diff --git a/Documentation/riscv/hwprobe.rst b/Documentation/riscv/hwprobe.rst index 19165ebd82ba..7f82a5385bc3 100644 --- a/Documentation/riscv/hwprobe.rst +++ b/Documentation/riscv/hwprobe.rst @@ -27,6 +27,13 @@ AND of the values for the specified CPUs.Usermode can supply NULL for cpus and
0 forcpu_countas a shortcut for all online CPUs. There are
currently no flags,
this value must be zero for future compatibility.+Calls to
sys_riscv_hwprobe()indicate the features supported by
both the
+kernel and the hardware that the system is running on. For
example, if the
+hardware supports the V extension and the kernel has V support
enabled then
+RISCV_HWPROBE_KEY_IMA_EXT_0/RISCV_HWPROBE_IMA_Vwill be set
even if the V
+extension is disabled via a userspace-controlled tunable such as +PR_RISCV_V_SET_CONTROL.
+
On success 0 is returned, on failure a negative error code is
returned.The following keys are defined:
@@ -65,7 +72,10 @@ The following keys are defined:
by version 2.2 of the RISC-V ISA manual.* :c:macro:
RISCV_HWPROBE_IMA_V: The V extension is supported,
as defined by
- version 1.0 of the RISC-V Vector extension manual.
+ version 1.0 of the RISC-V Vector extension manual. For strict
uABI
+ compatibility some systems may disable V by default even when
the hardware
+ supports in, in which case users must call
`prctl(PR_RISCV_V_SET_CONTROL,
+ ...` to explicitly allow V to be used.
Quoted 8 lines
* :c:macro:
RISCV_HWPROBE_EXT_ZBA: The Zba address generation
extension is
supported, as defined in version 1.0 of the
Bit-Manipulation ISAIMO that's the better way to go that to require that userspace tries to enable V via the
prctl()first, but we haven't released this yet so in theory we could still change it.
It's certainly a more precise definition but I'm arguing it's not a useful one.
The description of the prctl() in Documentation/riscv/vector.rst is fairly clear that it is intended only for use by init systems, and not by libraries.
Would you agree that "V is supported by the hardware and kernel, and could have been enabled by the init system but wasn't" is not actionable information for most applications and libraries?
https://sourceware.org/pipermail/libc-alpha/2023-April/147062.html proposes to use hwprobe to "do things like dynamically choose a memcpy implementation". Would you agree that hwprobe as currently defined in for-next is not suitable for the purpose described in that message, since it describes features that could be enabled, not features that are enabled?
Quoted 2 lines
We'd have a similar discussion for some of the counters that need to feed through the perf interface, though those are still in flight...
The documented intent of the vector prctl is to enable or disable vector use as a policy for a tree of processes. If I understand them correctly the perf counter user access patches require _individual processes_ to enable perf counters for their own use, which makes it a very different story from the perspective of the hwprobe API consumers.
-s
Quoted 33 lines
Quoted 6 lines
Assuming this is intentional, what is the path forward for future extensions that cannot be used from userspace without additional conditions being met? For instance, if we add support in the future for the Zve* extensions, the V bit would not be set in HWCAP for them, which would require library code to use the prctl interface unless we define the hwcap bits to imply userspace usability.
In this case a system that supports some of the Zve extensions but not the full V extension would not be probably from userspace, as V would not be set anywhere. The way to support that would be to add new bits into hwprobe to indicate those extensions, it just wasn't clear that anyone was interested in building Linux-flavored systems that supported only some a strict subset of V.
Happy to see patches if you know of some hardware in the pipeline, though ;)
Quoted 12 lines
-s
Quoted 9 lines
arch/riscv/kernel/sys_riscv.c +0
-1case RISCV_HWPROBE_KEY_CPUPERF_0 : --2.17.1
_______________________________________________
linux-riscv mailing list
linux-riscv at lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv_______________________________________________
linux-riscv mailing list
linux-riscv at lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
PATCH v21 4/27 riscv: Add new csr defines related to vector extension
In reply to Andy ChiuFrom: Greentime Hu <[email protected]>
Follow the riscv vector spec to add new csr numbers.
--- arch/riscv/include/asm/csr.h | 18 ++++++++++++++++-- 1 file changed, 16 insertions(+), 2 deletions(-)
arch/riscv/include/asm/csr.h +16 -2
@@ -24,16 +24,24 @@#define SR_FS_CLEAN _AC(0x00004000, UL)#define SR_FS_DIRTY _AC(0x00006000, UL) +#define SR_VS _AC(0x00000600, UL)/* Vector Status */ +#define SR_VS_OFF _AC(0x00000000, UL) +#define SR_VS_INITIAL _AC(0x00000200, UL) +#define SR_VS_CLEAN _AC(0x00000400, UL) +#define SR_VS_DIRTY _AC(0x00000600, UL) +#define SR_XS _AC(0x00018000, UL)/* Extension Status */ #define SR_XS_OFF _AC(0x00000000, UL)#define SR_XS_INITIAL _AC(0x00008000, UL)#define SR_XS_CLEAN _AC(0x00010000, UL)#define SR_XS_DIRTY _AC(0x00018000, UL) +#define SR_FS_VS (SR_FS | SR_VS)/* Vector and Floating-Point Unit */ +#ifndef CONFIG_64BIT -#define SR_SD _AC(0x80000000, UL)/* FS/XS dirty */ +#define SR_SD _AC(0x80000000, UL)/* FS/VS/XS dirty */ #else -#define SR_SD _AC(0x8000000000000000, UL)/* FS/XS dirty */ +#define SR_SD _AC(0x8000000000000000, UL)/* FS/VS/XS dirty */ #endif #ifdef CONFIG_64BIT
@@ -375,6 +383,12 @@#define CSR_MVIPH 0x319#define CSR_MIPH 0x354 +#define CSR_VSTART 0x8 +#define CSR_VCSR 0xf +#define CSR_VL 0xc20 +#define CSR_VTYPE 0xc21 +#define CSR_VLENB 0xc22 +#ifdef CONFIG_RISCV_M_MODE # define CSR_STATUS CSR_MSTATUS# define CSR_IE CSR_MIE
Signature · 1 line
2.17.1
PATCH v21 5/27 riscv: Clear vector regfile on bootup
In reply to Andy ChiuFrom: Greentime Hu <[email protected]>
clear vector registers on boot if kernel supports V.
--- arch/riscv/kernel/head.S | 27 +++++++++++++++++++++++++-- 1 file changed, 25 insertions(+), 2 deletions(-)
arch/riscv/kernel/head.S +25 -2
@@ -392,7 +392,7 @@ ENTRY(reset_regs) #ifdef CONFIG_FPU csrr t0, CSR_MISA andi t0, t0, (COMPAT_HWCAP_ISA_F | COMPAT_HWCAP_ISA_D) - beqz t0, .Lreset_regs_done + beqz t0, .Lreset_regs_done_fpu li t1, SR_FS csrs CSR_STATUS, t1
@@ -430,8 +430,31 @@ ENTRY(reset_regs) fmv.s.x f31, zero csrw fcsr, 0 /* note that the caller must clear SR_FS */ +.Lreset_regs_done_fpu: #endif /* CONFIG_FPU */ -.Lreset_regs_done: + +#ifdef CONFIG_RISCV_ISA_V + csrr t0, CSR_MISA + li t1, COMPAT_HWCAP_ISA_V + and t0, t0, t1 + beqz t0, .Lreset_regs_done_vector + + /* + * Clear vector registers and reset vcsr + * VLMAX has a defined value, VLEN is a constant, + * and this form of vsetvli is defined to set vl to VLMAX. + */ + li t1, SR_VS + csrs CSR_STATUS, t1 + csrs CSR_VCSR, x0 + vsetvli t1, x0, e8, m8, ta, ma + vmv.v.i v0, 0 + vmv.v.i v8, 0 + vmv.v.i v16, 0 + vmv.v.i v24, 0 + /* note that the caller must clear SR_VS */ +.Lreset_regs_done_vector: +#endif /* CONFIG_RISCV_ISA_V */ ret END(reset_regs) #endif /* CONFIG_RISCV_M_MODE */
Signature · 1 line
2.17.1
PATCH v21 6/27 riscv: Disable Vector Instructions for kernel itself
In reply to Andy ChiuFrom: Guo Ren <[email protected]>
Disable vector instructions execution for kernel mode at its entrances. This helps find illegal uses of vector in the kernel space, which is similar to the fpu.
---
Changelog V19:
- Add description in commit msg (Heiko's suggestion on v17)
--- arch/riscv/kernel/entry.S | 6 +++--- arch/riscv/kernel/head.S | 12 ++++++------ 2 files changed, 9 insertions(+), 9 deletions(-)
arch/riscv/kernel/entry.S +3 -3
@@ -48,10 +48,10 @@ _save_context: * Disable user-mode memory access as it should only be set in the * actual user copy routines. * - * Disable the FPU to detect illegal usage of floating point in kernel - * space. + * Disable the FPU/Vector to detect illegal usage of floating point + * or vector in kernel space. */ - li t0, SR_SUM | SR_FS + li t0, SR_SUM | SR_FS_VS REG_L s0, TASK_TI_USER_SP(tp) csrrc s1, CSR_STATUS, t0
arch/riscv/kernel/head.S +6 -6
@@ -140,10 +140,10 @@ secondary_start_sbi: .option pop /* - * Disable FPU to detect illegal usage of - * floating point in kernel space + * Disable FPU & VECTOR to detect illegal usage of + * floating point or vector in kernel space */ - li t0, SR_FS + li t0, SR_FS_VS csrc CSR_STATUS, t0 /* Set trap vector to spin forever to help debug */
@@ -234,10 +234,10 @@ pmp_done: .option pop /* - * Disable FPU to detect illegal usage of - * floating point in kernel space + * Disable FPU & VECTOR to detect illegal usage of + * floating point or vector in kernel space */ - li t0, SR_FS + li t0, SR_FS_VS csrc CSR_STATUS, t0 #ifdef CONFIG_RISCV_BOOT_SPINWAIT
Signature · 1 line
2.17.1
PATCH v21 7/27 riscv: Introduce Vector enable/disable helpers
In reply to Andy ChiuFrom: Greentime Hu <[email protected]>
These are small and likely to be frequently called so implement as inline routines (vs. function call).
--- arch/riscv/include/asm/vector.h | 11 +++++++++++ 1 file changed, 11 insertions(+)
arch/riscv/include/asm/vector.h +11 -0
@@ -11,12 +11,23 @@#ifdef CONFIG_RISCV_ISA_V #include <asm/hwcap.h> +#include <asm/csr.h> static __always_inline bool has_vector(void) {return riscv_has_extension_unlikely (RISCV_ISA_EXT_v); } +static __always_inline void riscv_v_enable(void) +{ +csr_set (CSR_SSTATUS ,SR_VS ); +} + +static __always_inline void riscv_v_disable(void) +{ +csr_clear (CSR_SSTATUS ,SR_VS ); +} +#else /* ! CONFIG_RISCV_ISA_V */ static __always_inline bool has_vector(void) { return false; }
Signature · 1 line
2.17.1
PATCH v21 8/27 riscv: Introduce riscv_v_vsize to record size of Vector context
In reply to Andy ChiuFrom: Greentime Hu <[email protected]>
This patch is used to detect the size of CPU vector registers and use riscv_v_vsize to save the size of all the vector registers. It assumes all harts has the same capabilities in a SMP system. If a core detects VLENB that is different from the boot core, then it warns and turns off V support for user space.
---
Changelog V19:
- Fix grammar in WARN() (Conor)
Changelog V18:
- Detect inconsistent VLEN setup on an SMP system (Heiko).
--- arch/riscv/include/asm/vector.h | 8 ++++++++ arch/riscv/kernel/Makefile | 1 + arch/riscv/kernel/cpufeature.c | 2 ++ arch/riscv/kernel/smpboot.c | 7 +++++++ arch/riscv/kernel/vector.c | 36 +++++++++++++++++++++++++++++++++ 5 files changed, 54 insertions(+) create mode 100644 arch/riscv/kernel/vector.c
arch/riscv/include/asm/vector.h +8 -0
@@ -7,12 +7,16 @@#define __ASM_RISCV_VECTOR_H#include <linux/types.h> +#include <uapi/asm-generic/errno.h> #ifdef CONFIG_RISCV_ISA_V #include <asm/hwcap.h> #include <asm/csr.h> +extern unsigned long riscv_v_vsize; +int riscv_v_setup_vsize (void ); + static __always_inline bool has_vector(void) {return riscv_has_extension_unlikely (RISCV_ISA_EXT_v);
@@ -30,7 +34,11 @@ static __always_inline void riscv_v_disable(void)#else /* ! CONFIG_RISCV_ISA_V */ +struct pt_regs ; + +static inline int riscv_v_setup_vsize (void ) {return -EOPNOTSUPP ; } static __always_inline bool has_vector(void) { return false; } +#define riscv_v_vsize (0)#endif /* CONFIG_RISCV_ISA_V */
arch/riscv/kernel/Makefile +1 -0
@@ -56,6 +56,7 @@ obj-$(CONFIG_MMU) += vdso.o vdso/ obj-$(CONFIG_RISCV_M_MODE) += traps_misaligned.o obj-$(CONFIG_FPU) += fpu.o +obj-$(CONFIG_RISCV_ISA_V) += vector.o obj-$(CONFIG_SMP) += smpboot.o obj-$(CONFIG_SMP) += smp.o obj-$(CONFIG_SMP) += cpu_ops.o
arch/riscv/kernel/cpufeature.c +2 -0
@@ -18,6 +18,7 @@#include <asm/hwcap.h> #include <asm/patch.h> #include <asm/processor.h> +#include <asm/vector.h> #define NUM_ALPHA_EXTS ('z' - 'a' + 1)
@@ -269,6 +270,7 @@ void __init riscv_fill_hwcap(void) }if (elf_hwcap &COMPAT_HWCAP_ISA_V ) { +riscv_v_setup_vsize ();/* * ISA string in device tree might have 'v' flag, but * CONFIG_RISCV_ISA_V is disabled in kernel.
arch/riscv/kernel/smpboot.c +7 -0
@@ -31,6 +31,8 @@#include <asm/tlbflush.h> #include <asm/sections.h> #include <asm/smp.h> +#include <uapi/asm/hwcap.h> +#include <asm/vector.h> #include "head.h"
@@ -169,6 +171,11 @@ asmlinkage __visible void smp_callin(void)set_cpu_online (curr_cpuid,1 );probe_vendor_features (curr_cpuid); +if (has_vector ()) { +if (riscv_v_setup_vsize ()) + elf_hwcap &= ~COMPAT_HWCAP_ISA_V ; + } +/* * Remote TLB flushes are ignored while the CPU is offline, so emit * a local TLB flush right now just in case.
arch/riscv/kernel/vector.c new file +36 -0
@@ -0,0 +1,36 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +/* +* Copyright (C) 2023 SiFive +* Author: Andy Chiu <[email protected]> +*/ +#include <linux/export.h> + +#include <asm/vector.h> +#include <asm/csr.h> +#include <asm/elf.h> +#include <asm/bug.h> + +unsigned long riscv_v_vsize__read_mostly ; +EXPORT_SYMBOL_GPL (riscv_v_vsize); + +int riscv_v_setup_vsize (void ) +{ +unsigned long this_vsize; + +/* There are 32 vector registers with vlenb length. */ +riscv_v_enable (); + this_vsize =csr_read (CSR_VLENB ) *32 ; +riscv_v_disable (); + +if (!riscv_v_vsize) { + riscv_v_vsize = this_vsize; +return 0 ; + } + +if (riscv_v_vsize != this_vsize) { +WARN (1 ,"RISCV_ISA_V only supports one vlenb on SMP systems" ); +return -EOPNOTSUPP ; + } + +return 0 ; +}
Signature · 1 line
2.17.1
PATCH v21 9/27 riscv: Introduce struct/helpers to save/restore per-task Vector state
In reply to Andy ChiuFrom: Greentime Hu <[email protected]>
Add vector state context struct to be added later in thread_struct. And prepare low-level helper functions to save/restore vector contexts.
This include Vector Regfile and CSRs holding dynamic configuration state (vstart, vl, vtype, vcsr). The Vec Register width could be implementation defined, but same for all processes, so that is saved separately.
This is not yet wired into final thread_struct - will be done when __switch_to actually starts doing this in later patches.
Given the variable (and potentially large) size of regfile, they are saved in dynamically allocated memory, pointed to by datap pointer in __riscv_v_ext_state.
--- arch/riscv/include/asm/vector.h | 95 ++++++++++++++++++++++++++++ arch/riscv/include/uapi/asm/ptrace.h | 17 +++++ 2 files changed, 112 insertions(+)
arch/riscv/include/asm/vector.h +95 -0
@@ -11,8 +11,10 @@#ifdef CONFIG_RISCV_ISA_V +#include <linux/stringify.h> #include <asm/hwcap.h> #include <asm/csr.h> +#include <asm/asm.h> extern unsigned long riscv_v_vsize;int riscv_v_setup_vsize (void );
@@ -22,6 +24,26 @@ static __always_inline bool has_vector(void)return riscv_has_extension_unlikely (RISCV_ISA_EXT_v); } +static inline void __riscv_v_vstate_clean (struct pt_regs *regs) +{ + regs->status = (regs->status & ~SR_VS ) |SR_VS_CLEAN ; +} + +static inline void riscv_v_vstate_off (struct pt_regs *regs) +{ + regs->status = (regs->status & ~SR_VS ) |SR_VS_OFF ; +} + +static inline void riscv_v_vstate_on (struct pt_regs *regs) +{ + regs->status = (regs->status & ~SR_VS ) |SR_VS_INITIAL ; +} + +static inline bool riscv_v_vstate_query (struct pt_regs *regs) +{ +return (regs->status &SR_VS ) !=0 ; +} + static __always_inline void riscv_v_enable(void) {csr_set (CSR_SSTATUS ,SR_VS );
@@ -32,13 +54,86 @@ static __always_inline void riscv_v_disable(void)csr_clear (CSR_SSTATUS ,SR_VS ); } +static __always_inline void __vstate_csr_save(struct __riscv_v_ext_state *dest) +{ + asmvolatile ( + "csrr %0, " __stringify(CSR_VSTART) "\n\t" + "csrr %1, " __stringify(CSR_VTYPE) "\n\t" + "csrr %2, " __stringify(CSR_VL) "\n\t" + "csrr %3, " __stringify(CSR_VCSR) "\n\t" + : "=r" (dest->vstart), "=r" (dest->vtype), "=r" (dest->vl), + "=r" (dest->vcsr) : :); +} + +static __always_inline void __vstate_csr_restore(struct __riscv_v_ext_state *src) +{ + asmvolatile ( +".option push\n\t" +".option arch, +v\n\t" +"vsetvl x0, %2, %1\n\t" +".option pop\n\t" + "csrw " __stringify(CSR_VSTART) ", %0\n\t" + "csrw " __stringify(CSR_VCSR) ", %3\n\t" + : : "r" (src->vstart), "r" (src->vtype), "r" (src->vl), + "r" (src->vcsr) :); +} + +static inline void __riscv_v_vstate_save (struct __riscv_v_ext_state *save_to, +void *datap) +{ +unsigned long vl; + +riscv_v_enable (); +__vstate_csr_save (save_to); + asmvolatile ( +".option push\n\t" +".option arch, +v\n\t" +"vsetvli %0, x0, e8, m8, ta, ma\n\t" +"vse8.v v0, (%1)\n\t" +"add %1, %1, %0\n\t" +"vse8.v v8, (%1)\n\t" +"add %1, %1, %0\n\t" +"vse8.v v16, (%1)\n\t" +"add %1, %1, %0\n\t" +"vse8.v v24, (%1)\n\t" +".option pop\n\t" + :"=&r" (vl) :"r" (datap) :"memory" ); +riscv_v_disable (); +} + +static inline void __riscv_v_vstate_restore (struct __riscv_v_ext_state *restore_from, +void *datap) +{ +unsigned long vl; + +riscv_v_enable (); + asmvolatile ( +".option push\n\t" +".option arch, +v\n\t" +"vsetvli %0, x0, e8, m8, ta, ma\n\t" +"vle8.v v0, (%1)\n\t" +"add %1, %1, %0\n\t" +"vle8.v v8, (%1)\n\t" +"add %1, %1, %0\n\t" +"vle8.v v16, (%1)\n\t" +"add %1, %1, %0\n\t" +"vle8.v v24, (%1)\n\t" +".option pop\n\t" + :"=&r" (vl) :"r" (datap) :"memory" ); +__vstate_csr_restore (restore_from); +riscv_v_disable (); +} +#else /* ! CONFIG_RISCV_ISA_V */ struct pt_regs ;static inline int riscv_v_setup_vsize (void ) {return -EOPNOTSUPP ; } static __always_inline bool has_vector(void) { return false; } +static inline bool riscv_v_vstate_query (struct pt_regs *regs) {return false; }#define riscv_v_vsize (0) +#define riscv_v_vstate_off (regs) do {} while (0) +#define riscv_v_vstate_on (regs) do {} while (0)#endif /* CONFIG_RISCV_ISA_V */
arch/riscv/include/uapi/asm/ptrace.h +17 -0
@@ -77,6 +77,23 @@ union __riscv_fp_state {struct __riscv_q_ext_state q; }; +struct __riscv_v_ext_state { +unsigned long vstart; +unsigned long vl; +unsigned long vtype; +unsigned long vcsr; +void *datap; +/* +* In signal handler, datap will be set a correct user stack offset +* and vector registers will be copied to the address of datap +* pointer. +* +* In ptrace syscall, datap will be set to zero and the vector +* registers will be copied to the address right after this +* structure. +*/ +}; +#endif /* __ASSEMBLY__ */ #endif /* _UAPI_ASM_RISCV_PTRACE_H */
Signature · 1 line
2.17.1
Re: [PATCH -next v21 09/27] riscv: Introduce struct/helpers to save/restore per-task Vector state
Le maanantaina 5. kes?kuuta 2023, 14.07.06 EEST Andy Chiu a ?crit :
Quoted 41 lines
arch/riscv/include/asm/vector.h +33
-0@@ -32,13 +54,86 @@ static __always_inline void riscv_v_disable(void)csr_clear (CSR_SSTATUS ,SR_VS ); } +static __always_inline void __vstate_csr_save(struct __riscv_v_ext_state *dest) +{ + asmvolatile ( + "csrr %0, " __stringify(CSR_VSTART) "\n\t" + "csrr %1, " __stringify(CSR_VTYPE) "\n\t" + "csrr %2, " __stringify(CSR_VL) "\n\t" + "csrr %3, " __stringify(CSR_VCSR) "\n\t" + : "=r" (dest->vstart), "=r" (dest->vtype), "=r" (dest- vl), + "=r" (dest->vcsr) : :); +} + +static __always_inline void __vstate_csr_restore(struct __riscv_v_ext_state *src) +{ + asmvolatile ( +".option push\n\t" +".option arch, +v\n\t" +"vsetvl x0, %2, %1\n\t" +".option pop\n\t" + "csrw " __stringify(CSR_VSTART) ", %0\n\t" + "csrw " __stringify(CSR_VCSR) ", %3\n\t" + : : "r" (src->vstart), "r" (src->vtype), "r" (src->vl), + "r" (src->vcsr) :); +} + +static inline void __riscv_v_vstate_save (struct __riscv_v_ext_state *save_to, + void *datap) +{ +unsigned long vl; + +riscv_v_enable (); +__vstate_csr_save (save_to); + asmvolatile ( +".option push\n\t" +".option arch, +v\n\t" +"vsetvli %0, x0, e8, m8, ta, ma\n\t" + "vse8.v v0, (%1)\n\t"
arch/riscv/include/asm/vector.h static __always_inline void riscv_v_disable(void) · line 103
Quoted 10 lines
arch/riscv/include/asm/vector.h +10
-0+"add %1, %1, %0\n\t" +"vse8.v v8, (%1)\n\t" +"add %1, %1, %0\n\t" +"vse8.v v16, (%1)\n\t" +"add %1, %1, %0\n\t" +"vse8.v v24, (%1)\n\t" +".option pop\n\t" + : "=&r" (vl) : "r" (datap) : "memory"); + riscv_v_disable(); +}
Shouldn't this use vs8r.v rather than vse8.v, and do away with vsetvli? This seems like a textbook use case for the whole-register store instruction, no?
Review static __always_inline void riscv_v_disable(void) · line 106
Quoted 3 lines
(quoted diff) +2
-0+ +static inline void __riscv_v_vstate_restore(struct __riscv_v_ext_state *restore_from, + void
*datap)
Quoted 11 lines
arch/riscv/include/asm/vector.h +11
-0+{ +unsigned long vl; + +riscv_v_enable (); + asmvolatile ( +".option push\n\t" +".option arch, +v\n\t" +"vsetvli %0, x0, e8, m8, ta, ma\n\t" +"vle8.v v0, (%1)\n\t" +"add %1, %1, %0\n\t" + "vle8.v v8, (%1)\n\t"
arch/riscv/include/asm/vector.h static __always_inline void riscv_v_disable(void) · line 126
Quoted 10 lines
arch/riscv/include/asm/vector.h +10
-0+"add %1, %1, %0\n\t" +"vle8.v v16, (%1)\n\t" +"add %1, %1, %0\n\t" +"vle8.v v24, (%1)\n\t" +".option pop\n\t" + : "=&r" (vl) : "r" (datap) : "memory"); + __vstate_csr_restore(restore_from); +riscv_v_disable (); +} +
Ditto but vl8r.v.
Quoted 13 lines
arch/riscv/include/asm/vector.h +1
-0#else /* ! CONFIG_RISCV_ISA_V */ struct pt_regs ;static inline int riscv_v_setup_vsize (void ) {return -EOPNOTSUPP ; } static __always_inline bool has_vector(void) { return false; } +static inline bool riscv_v_vstate_query (struct pt_regs *regs) {return false; } #define
riscv_v_vsize(0)arch/riscv/include/asm/vector.h +2
-0+#define riscv_v_vstate_off (regs) do {} while (0) +#define riscv_v_vstate_on (regs) do {} while (0)#endif /* CONFIG_RISCV_ISA_V */
Signature · 2 lines
???? ????-?????? http://www.remlab.net/
Re: [PATCH -next v21 09/27] riscv: Introduce struct/helpers to save/restore per-task Vector state
Quoted 49 lines · On Mon, Jun 12, 2023 at 10:36?PM R?mi Denis-Courmont <[email protected]> wrote:
Le maanantaina 5. kes?kuuta 2023, 14.07.06 EEST Andy Chiu a ?crit :
Quoted 47 lines
arch/riscv/include/asm/vector.h +39
-0@@ -32,13 +54,86 @@ static __always_inline void riscv_v_disable(void)csr_clear (CSR_SSTATUS ,SR_VS ); } +static __always_inline void __vstate_csr_save(struct __riscv_v_ext_state *dest) +{ + asmvolatile ( + "csrr %0, " __stringify(CSR_VSTART) "\n\t" + "csrr %1, " __stringify(CSR_VTYPE) "\n\t" + "csrr %2, " __stringify(CSR_VL) "\n\t" + "csrr %3, " __stringify(CSR_VCSR) "\n\t" + : "=r" (dest->vstart), "=r" (dest->vtype), "=r" (dest- vl), + "=r" (dest->vcsr) : :); +} + +static __always_inline void __vstate_csr_restore(struct __riscv_v_ext_state *src) +{ + asmvolatile ( +".option push\n\t" +".option arch, +v\n\t" +"vsetvl x0, %2, %1\n\t" +".option pop\n\t" + "csrw " __stringify(CSR_VSTART) ", %0\n\t" + "csrw " __stringify(CSR_VCSR) ", %3\n\t" + : : "r" (src->vstart), "r" (src->vtype), "r" (src->vl), + "r" (src->vcsr) :); +} + +static inline void __riscv_v_vstate_save (struct __riscv_v_ext_state *save_to, + void *datap) +{ +unsigned long vl; + +riscv_v_enable (); +__vstate_csr_save (save_to); + asmvolatile ( +".option push\n\t" +".option arch, +v\n\t" +"vsetvli %0, x0, e8, m8, ta, ma\n\t" +"vse8.v v0, (%1)\n\t" +"add %1, %1, %0\n\t" +"vse8.v v8, (%1)\n\t" +"add %1, %1, %0\n\t" +"vse8.v v16, (%1)\n\t" +"add %1, %1, %0\n\t" + "vse8.v v24, (%1)\n\t"
Quoted 8 lines
arch/riscv/include/asm/vector.h static __always_inline void riscv_v_disable(void) · line 102
Quoted 4 lines
arch/riscv/include/asm/vector.h +4
-0+".option pop\n\t" + : "=&r" (vl) : "r" (datap) : "memory"); + riscv_v_disable(); +}Shouldn't this use
vs8r.vrather thanvse8.v, and do away withvsetvli? This seems like a textbook use case for the whole-register store instruction, no?
Yes, I think it is worth changing to whole-register load/store instruction. Let me form a follow-up patch to improve it a bit.
Quoted 40 lines
Review static __always_inline void riscv_v_disable(void) · line 106
Quoted 4 lines
(quoted diff) +2
-0+ +static inline void __riscv_v_vstate_restore(struct __riscv_v_ext_state *restore_from, + void*datap)
Quoted 11 lines
arch/riscv/include/asm/vector.h +11
-0+{ +unsigned long vl; + +riscv_v_enable (); + asmvolatile ( +".option push\n\t" +".option arch, +v\n\t" +"vsetvli %0, x0, e8, m8, ta, ma\n\t" +"vle8.v v0, (%1)\n\t" +"add %1, %1, %0\n\t" + "vle8.v v8, (%1)\n\t"arch/riscv/include/asm/vector.h static __always_inline void riscv_v_disable(void) · line 126
Quoted 10 lines
arch/riscv/include/asm/vector.h +10
-0+"add %1, %1, %0\n\t" +"vle8.v v16, (%1)\n\t" +"add %1, %1, %0\n\t" +"vle8.v v24, (%1)\n\t" +".option pop\n\t" + : "=&r" (vl) : "r" (datap) : "memory"); + __vstate_csr_restore(restore_from); +riscv_v_disable (); +} +Ditto but
vl8r.v.Quoted 11 lines
arch/riscv/include/asm/vector.h +1
-0#else /* ! CONFIG_RISCV_ISA_V */ struct pt_regs ;static inline int riscv_v_setup_vsize (void ) {return -EOPNOTSUPP ; } static __always_inline bool has_vector(void) { return false; } +static inline bool riscv_v_vstate_query (struct pt_regs *regs) {return false; } #define
riscv_v_vsize(0)arch/riscv/include/asm/vector.h +2
-0+#define riscv_v_vstate_off (regs) do {} while (0) +#define riscv_v_vstate_on (regs) do {} while (0)
Quoted 10 lines
Quoted 2 lines
#endif /*
CONFIG_RISCV_ISA_V*/--
???? ????-??????
http://www.remlab.net/
Thanks,
Andy
PATCH v21 10/27 riscv: Add task switch support for vector
In reply to Andy ChiuFrom: Greentime Hu <[email protected]>
This patch adds task switch support for vector. It also supports all lengths of vlen.
---
Changelog v21:
- move riscv_v_vstate_off(childregs); to a proper place.
--- arch/riscv/include/asm/processor.h | 1 + arch/riscv/include/asm/switch_to.h | 3 +++ arch/riscv/include/asm/thread_info.h | 3 +++ arch/riscv/include/asm/vector.h | 38 ++++++++++++++++++++++++++++ arch/riscv/kernel/process.c | 19 ++++++++++++++ 5 files changed, 64 insertions(+)
arch/riscv/include/asm/processor.h +1 -0
@@ -39,6 +39,7 @@ struct thread_struct {unsigned long s[12 ];/* s[0]: frame pointer */ struct __riscv_d_ext_state fstate;unsigned long bad_cause; +struct __riscv_v_ext_state vstate; };/* Whitelist the fstate from the task_struct for hardened usercopy */
arch/riscv/include/asm/switch_to.h +3 -0
@@ -8,6 +8,7 @@#include <linux/jump_label.h> #include <linux/sched/task_stack.h> +#include <asm/vector.h> #include <asm/hwcap.h> #include <asm/processor.h> #include <asm/ptrace.h>
@@ -78,6 +79,8 @@ do { \struct task_struct *__next = (next); \if (has_fpu ()) \__switch_to_fpu (__prev, __next); \ +if (has_vector ()) \ +__switch_to_vector (__prev, __next); \ ((last) =__switch_to (__prev, __next)); \ }while (0 )
arch/riscv/include/asm/thread_info.h +3 -0
@@ -81,6 +81,9 @@ struct thread_info { .preempt_count = INIT_PREEMPT_COUNT, \ } +void arch_release_task_struct (struct task_struct *tsk); +int arch_dup_task_struct (struct task_struct *dst,struct task_struct *src); +#endif /* !__ASSEMBLY__ */ /*
arch/riscv/include/asm/vector.h +38 -0
@@ -12,6 +12,9 @@#ifdef CONFIG_RISCV_ISA_V #include <linux/stringify.h> +#include <linux/sched.h> +#include <linux/sched/task_stack.h> +#include <asm/ptrace.h> #include <asm/hwcap.h> #include <asm/csr.h> #include <asm/asm.h>
@@ -124,6 +127,38 @@ static inline void __riscv_v_vstate_restore(struct __riscv_v_ext_state *restore_riscv_v_disable (); } +static inline void riscv_v_vstate_save (struct task_struct *task, +struct pt_regs *regs) +{ +if ((regs->status &SR_VS ) ==SR_VS_DIRTY ) { +struct __riscv_v_ext_state *vstate = &task->thread.vstate; + +__riscv_v_vstate_save (vstate, vstate->datap); +__riscv_v_vstate_clean (regs); + } +} + +static inline void riscv_v_vstate_restore (struct task_struct *task, +struct pt_regs *regs) +{ +if ((regs->status &SR_VS ) !=SR_VS_OFF ) { +struct __riscv_v_ext_state *vstate = &task->thread.vstate; + +__riscv_v_vstate_restore (vstate, vstate->datap); +__riscv_v_vstate_clean (regs); + } +} + +static inline void __switch_to_vector (struct task_struct *prev, +struct task_struct *next) +{ +struct pt_regs *regs; + + regs =task_pt_regs (prev); +riscv_v_vstate_save (prev, regs); +riscv_v_vstate_restore (next,task_pt_regs (next)); +} +#else /* ! CONFIG_RISCV_ISA_V */ struct pt_regs ;
@@ -132,6 +167,9 @@ static inline int riscv_v_setup_vsize(void) { return -EOPNOTSUPP; } static __always_inline bool has_vector(void) { return false; }static inline bool riscv_v_vstate_query (struct pt_regs *regs) {return false; }#define riscv_v_vsize (0) +#define riscv_v_vstate_save (task, regs) do {} while (0) +#define riscv_v_vstate_restore (task, regs) do {} while (0) +#define __switch_to_vector (__prev, __next) do {} while (0)#define riscv_v_vstate_off (regs) do {} while (0)#define riscv_v_vstate_on (regs) do {} while (0)
arch/riscv/kernel/process.c +19 -0
@@ -24,6 +24,7 @@#include <asm/switch_to.h> #include <asm/thread_info.h> #include <asm/cpuidle.h> +#include <asm/vector.h> registerunsigned long gp_in_global __asm__("gp" );
@@ -146,12 +147,28 @@ void flush_thread(void)fstate_off (current,task_pt_regs (current));memset (¤t->thread.fstate,0 ,sizeof (current->thread.fstate));#endif +#ifdef CONFIG_RISCV_ISA_V +/* Reset vector state */ +riscv_v_vstate_off (task_pt_regs (current)); +kfree (current->thread.vstate.datap); +memset (¤t->thread.vstate,0 ,sizeof (struct __riscv_v_ext_state )); +#endif +} + +void arch_release_task_struct (struct task_struct *tsk) +{ +/* Free the vector context of datap. */ +if (has_vector ()) +kfree (tsk->thread.vstate.datap); }int arch_dup_task_struct (struct task_struct *dst,struct task_struct *src) {fstate_save (src,task_pt_regs (src)); *dst = *src; +/* clear entire V context, including datap for a new task */ +memset (&dst->thread.vstate,0 ,sizeof (struct __riscv_v_ext_state )); +return 0 ; }
@@ -176,6 +193,8 @@ int copy_thread(struct task_struct *p, const struct kernel_clone_args *args) p->thread.s[1 ] = (unsigned long )args->fn_arg; }else { *childregs = *(current_pt_regs ()); +/* Turn off status.VS */ +riscv_v_vstate_off (childregs);if (usp)/* User fork */ childregs->sp = usp;if (clone_flags &CLONE_SETTLS )
Signature · 1 line
2.17.1
PATCH v21 11/27 riscv: Allocate user's vector context in the first-use trap
In reply to Andy ChiuVector unit is disabled by default for all user processes. Thus, a process will take a trap (illegal instruction) into kernel at the first time when it uses Vector. Only after then, the kernel allocates V context and starts take care of the context for that user process.
---
Changelog v21:
- Remove has_vector() and use ELF_HWCAP instead. The V bit in the hwcap, if preesents, implies has_vector() returning true. So we can drop has_vector() test here safely, because we also don't handle !V but has_vector() cases.
- Change SIGKILL to SIGBUS when allocation of V-context fails. (Darius)
Changelog v20:
- move has_vector() into vector.c for better code readibility
- check elf_hwcap in the first-use trap because it might get turned off if cores have different VLENs.
Changelog v18:
- Add blank lines (Heiko)
- Return immediately in insn_is_vector() if an insn matches (Heiko)
--- arch/riscv/include/asm/insn.h | 29 ++++++++++ arch/riscv/include/asm/vector.h | 2 + arch/riscv/kernel/traps.c | 26 ++++++++- arch/riscv/kernel/vector.c | 95 +++++++++++++++++++++++++++++++++ 4 files changed, 150 insertions(+), 2 deletions(-)
arch/riscv/include/asm/insn.h +29 -0
@@ -137,6 +137,26 @@#define RVG_OPCODE_JALR 0x67#define RVG_OPCODE_JAL 0x6f#define RVG_OPCODE_SYSTEM 0x73 +#define RVG_SYSTEM_CSR_OFF 20 +#define RVG_SYSTEM_CSR_MASK GENMASK(12, 0) + +/* parts of opcode for RVF, RVD and RVQ */ +#define RVFDQ_FL_FS_WIDTH_OFF 12 +#define RVFDQ_FL_FS_WIDTH_MASK GENMASK(3, 0) +#define RVFDQ_FL_FS_WIDTH_W 2 +#define RVFDQ_FL_FS_WIDTH_D 3 +#define RVFDQ_LS_FS_WIDTH_Q 4 +#define RVFDQ_OPCODE_FL 0x07 +#define RVFDQ_OPCODE_FS 0x27 + +/* parts of opcode for RVV */ +#define RVV_OPCODE_VECTOR 0x57 +#define RVV_VL_VS_WIDTH_8 0 +#define RVV_VL_VS_WIDTH_16 5 +#define RVV_VL_VS_WIDTH_32 6 +#define RVV_VL_VS_WIDTH_64 7 +#define RVV_OPCODE_VL RVFDQ_OPCODE_FL +#define RVV_OPCODE_VS RVFDQ_OPCODE_FS/* parts of opcode for RVC*/ #define RVC_OPCODE_C0 0x0
@@ -304,6 +324,15 @@ static __always_inline bool riscv_insn_is_branch(u32 code) (RVC_X (x_,RVC_B_IMM_7_6_OPOFF ,RVC_B_IMM_7_6_MASK ) <<RVC_B_IMM_7_6_OFF ) | \ (RVC_IMM_SIGN(x_) << RVC_B_IMM_SIGN_OFF); }) +#define RVG_EXTRACT_SYSTEM_CSR (x) \ + ({typeof(x) x_ = (x); RV_X(x_, RVG_SYSTEM_CSR_OFF, RVG_SYSTEM_CSR_MASK); }) + +#define RVFDQ_EXTRACT_FL_FS_WIDTH (x) \ + ({typeof(x) x_ = (x); RV_X(x_, RVFDQ_FL_FS_WIDTH_OFF, \ + RVFDQ_FL_FS_WIDTH_MASK); }) + +#define RVV_EXRACT_VL_VS_WIDTH (x) RVFDQ_EXTRACT_FL_FS_WIDTH(x) +/* * Get the immediate from a J-type instruction. *
arch/riscv/include/asm/vector.h +2 -0
@@ -21,6 +21,7 @@extern unsigned long riscv_v_vsize;int riscv_v_setup_vsize (void ); +bool riscv_v_first_use_handler (struct pt_regs *regs); static __always_inline bool has_vector(void) {
@@ -165,6 +166,7 @@ struct pt_regs;static inline int riscv_v_setup_vsize (void ) {return -EOPNOTSUPP ; } static __always_inline bool has_vector(void) { return false; } +static inline bool riscv_v_first_use_handler (struct pt_regs *regs) {return false; }static inline bool riscv_v_vstate_query (struct pt_regs *regs) {return false; }#define riscv_v_vsize (0)#define riscv_v_vstate_save (task, regs) do {} while (0)
arch/riscv/kernel/traps.c +24 -2
@@ -26,6 +26,7 @@#include <asm/ptrace.h> #include <asm/syscall.h> #include <asm/thread_info.h> +#include <asm/vector.h> int show_unhandled_signals =1 ;
@@ -145,8 +146,29 @@ DO_ERROR_INFO(do_trap_insn_misaligned, SIGBUS, BUS_ADRALN, "instruction address misaligned");DO_ERROR_INFO (do_trap_insn_fault,SIGSEGV ,SEGV_ACCERR ,"instruction access fault" ); -DO_ERROR_INFO (do_trap_insn_illegal, -SIGILL ,ILL_ILLOPC ,"illegal instruction" ); + +asmlinkage __visible __trap_section void do_trap_insn_illegal(struct pt_regs *regs) +{ +if (user_mode (regs)) { +irqentry_enter_from_user_mode (regs); + +local_irq_enable (); + +if (!riscv_v_first_use_handler (regs)) +do_trap_error (regs,SIGILL ,ILL_ILLOPC , regs->epc, +"Oops - illegal instruction" ); + +irqentry_exit_to_user_mode (regs); + }else { +irqentry_state_t state =irqentry_nmi_enter (regs); + +do_trap_error (regs,SIGILL ,ILL_ILLOPC , regs->epc, +"Oops - illegal instruction" ); + +irqentry_nmi_exit (regs, state); + } +} +DO_ERROR_INFO (do_trap_load_fault,SIGSEGV ,SEGV_ACCERR ,"load access fault" );#ifndef CONFIG_RISCV_M_MODE
arch/riscv/kernel/vector.c +95 -0
108 lines not on this page · raw message
Signature · 1 line
2.17.1
Re: [PATCH -next v21 11/27] riscv: Allocate user's vector context in the first-use trap
Quoted 8 lines · Andy Chiu
Vector unit is disabled by default for all user processes. Thus, a process will take a trap (illegal instruction) into kernel at the first time when it uses Vector. Only after then, the kernel allocates V context and starts take care of the context for that user process.
Suggested-by Richard Henderson <[email protected]>Link https://lore.kernel.org/r/3923eeee-e4dc-0911-40bf-84c34aee962d at linaro.orgSigned-off-by Andy Chiu <[email protected]>
Quoted 6 lines
---
Changelog v21:
- Removehas_vector()and useELF_HWCAPinstead. The V bit in the hwcap, if preesents, implieshas_vector()returning true. So we can drophas_vector()test here safely, because we also don't handle !V buthas_vector()cases.
I think you could probably have carried my R-b from v20, but I realise I did not reply to you - sorry about that.
The removal of the has_vector() check seems to tally with the conversation we had on v20, so
Cheers,
Conor.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 228 bytes
Desc: not available
URL: <http://lists.infradead.org/pipermail/kvm-riscv/attachments/20230605/daa8c1a8/attachment-0001.sig>
Re: [PATCH -next v21 11/27] riscv: Allocate user's vector context in the first-use trap
In reply to Andy ChiuAm Montag, 5. Juni 2023, 13:07:08 CEST schrieb Andy Chiu:
Quoted 8 lines
Vector unit is disabled by default for all user processes. Thus, a process will take a trap (illegal instruction) into kernel at the first time when it uses Vector. Only after then, the kernel allocates V context and starts take care of the context for that user process.
Suggested-by Richard Henderson <[email protected]>Link https://lore.kernel.org/r/3923eeee-e4dc-0911-40bf-84c34aee962d at linaro.orgSigned-off-by Andy Chiu <[email protected]>
PATCH v21 12/27 riscv: Add ptrace vector support
In reply to Andy ChiuFrom: Greentime Hu <[email protected]>
This patch adds ptrace support for riscv vector. The vector registers will be saved in datap pointer of __riscv_v_ext_state. This pointer will be set right after the __riscv_v_ext_state data structure then it will be put in ubuf for ptrace system call to get or set. It will check if the datap got from ubuf is set to the correct address or not when the ptrace system call is trying to set the vector registers.
---
Changelog V18:
- Use sizeof(vstate->datap) instead of sizeof(void*) (Eike)
--- arch/riscv/include/uapi/asm/ptrace.h | 7 +++ arch/riscv/kernel/ptrace.c | 70 ++++++++++++++++++++++++++++ include/uapi/linux/elf.h | 1 + 3 files changed, 78 insertions(+)
arch/riscv/include/uapi/asm/ptrace.h +7 -0
@@ -94,6 +94,13 @@ struct __riscv_v_ext_state {*/ }; +/* +* According to spec: The number of bits in a single vector register, +* VLEN >= ELEN, which must be a power of 2, and must be no greater than +* 2^16 = 65536bits = 8192bytes +*/ +#define RISCV_MAX_VLENB (8192) +#endif /* __ASSEMBLY__ */ #endif /* _UAPI_ASM_RISCV_PTRACE_H */
arch/riscv/kernel/ptrace.c +70 -0
94 lines not on this page · raw message
include/uapi/linux/elf.h +1 -0
@@ -440,6 +440,7 @@ typedef struct elf64_shdr {#define NT_MIPS_DSP 0x800/* MIPS DSP ASE registers */ #define NT_MIPS_FP_MODE 0x801/* MIPS floating-point mode */ #define NT_MIPS_MSA 0x802/* MIPS SIMD registers */ +#define NT_RISCV_VECTOR 0x900/* RISC-V vector registers */ #define NT_LOONGARCH_CPUCFG 0xa00/* LoongArch CPU config registers */ #define NT_LOONGARCH_CSR 0xa01/* LoongArch control and status registers */ #define NT_LOONGARCH_LSX 0xa02/* LoongArch Loongson SIMD Extension registers */
Signature · 1 line
2.17.1
PATCH v21 13/27 riscv: signal: check fp-reserved words unconditionally
In reply to Andy ChiuIn order to let kernel/user locate and identify an extension context on the existing sigframe, we are going to utilize reserved space of fp and encode the information there. And since the sigcontext has already preserved a space for fp context w or w/o CONFIG_FPU, we move those reserved words checking/setting routine back into generic code.
This commit also undone an additional logical change carried by the refactor commit 007f5c3589578
("Refactor FPU code in signal setup/return procedures"). Originally we did not restore fp context if restoring of gpr have failed. And it was fine on the other side. In such way the kernel could keep the regfiles intact, and potentially react at the failing point of restore.
--- arch/riscv/kernel/signal.c | 55 +++++++++++++++++++------------------- 1 file changed, 28 insertions(+), 27 deletions(-)
arch/riscv/kernel/signal.c +28 -27
@@ -40,26 +40,13 @@ static long restore_fp_state(struct pt_regs *regs, {long err;struct __riscv_d_ext_state __user *state = &sc_fpregs->d; -size_t i; err =__copy_from_user (¤t->thread.fstate, state,sizeof (*state));if (unlikely (err))return err;fstate_restore (current, regs); - -/* We support no other extension state at this time. */ -for (i =0 ; i <ARRAY_SIZE (sc_fpregs->q.reserved); i++) { -u32 value; - - err =__get_user (value, &sc_fpregs->q.reserved[i]); -if (unlikely (err)) -break ; -if (value !=0 ) -return -EINVAL ; - } - -return err; +return 0 ; } static long save_fp_state(struct pt_regs *regs,
@@ -67,20 +54,9 @@ static long save_fp_state(struct pt_regs *regs, {long err;struct __riscv_d_ext_state __user *state = &sc_fpregs->d; -size_t i;fstate_save (current, regs); err =__copy_to_user (state, ¤t->thread.fstate,sizeof (*state)); -if (unlikely (err)) -return err; - -/* We support no other extension state at this time. */ -for (i =0 ; i <ARRAY_SIZE (sc_fpregs->q.reserved); i++) { - err =__put_user (0 , &sc_fpregs->q.reserved[i]); -if (unlikely (err)) -break ; - } -return err; }#else
@@ -92,11 +68,30 @@ static long restore_sigcontext(struct pt_regs *regs, struct sigcontext __user *sc) {long err; +size_t i; +/* sc_regs is structured the same as the start of pt_regs */ err =__copy_from_user (regs, &sc->sc_regs,sizeof (sc->sc_regs)); +if (unlikely (err)) +return err; +/* Restore the floating-point state. */ -if (has_fpu ()) - err |=restore_fp_state (regs, &sc->sc_fpregs); +if (has_fpu ()) { + err =restore_fp_state (regs, &sc->sc_fpregs); +if (unlikely (err)) +return err; + } + +/* We support no other extension state at this time. */ +for (i =0 ; i <ARRAY_SIZE (sc->sc_fpregs.q.reserved); i++) { +u32 value; + + err =__get_user (value, &sc->sc_fpregs.q.reserved[i]); +if (unlikely (err)) +break ; +if (value !=0 ) +return -EINVAL ; + }return err; }
@@ -147,11 +142,17 @@ static long setup_sigcontext(struct rt_sigframe __user *frame, {struct sigcontext __user *sc = &frame->uc.uc_mcontext;long err; +size_t i; +/* sc_regs is structured the same as the start of pt_regs */ err =__copy_to_user (&sc->sc_regs, regs,sizeof (sc->sc_regs));/* Save the floating-point state. */ if (has_fpu ()) err |=save_fp_state (regs, &sc->sc_fpregs); +/* We support no other extension state at this time. */ +for (i =0 ; i <ARRAY_SIZE (sc->sc_fpregs.q.reserved); i++) + err |=__put_user (0 , &sc->sc_fpregs.q.reserved[i]); +return err; }
Signature · 1 line
2.17.1
PATCH v21 14/27 riscv: signal: Add sigcontext save/restore for vector
In reply to Andy ChiuFrom: Greentime Hu <[email protected]>
This patch facilitates the existing fp-reserved words for placement of the first extension's context header on the user's sigframe. A context header consists of a distinct magic word and the size, including the header itself, of an extension on the stack. Then, the frame is followed by the context of that extension, and then a header + context body for another extension if exists. If there is no more extension to come, then the frame must be ended with a null context header. A special case is rv64gc, where the kernel support no extensions requiring to expose additional regfile to the user. In such case the kernel would place the null context header right after the first reserved word of __riscv_q_ext_state when saving sigframe. And the kernel would check if all reserved words are zeros when a signal handler returns.
__riscv_q_ext_state---->| |<-__riscv_extra_ext_header
~ ~ .reserved[0]--->|0 |<- .reserved <-------|magic |<- .hdr | |size |_______ end of sc_fpregs | |ext-bdy| | ~ ~ +)size ------->|magic |<- another context header |size | |ext-bdy| ~ ~ |magic:0|<- null context header |size:0 |
The vector registers will be saved in datap pointer. The datap pointer will be allocated dynamically when the task needs in kernel space. On the other hand, datap pointer on the sigframe will be set right after the __riscv_v_ext_state data structure.
---
Changelog V19:
- Fix a conflict in signal.c due to commit 8d736482749f ("riscv: add icache flush for nommu sigreturn trampoline")
--- arch/riscv/include/uapi/asm/ptrace.h | 15 ++ arch/riscv/include/uapi/asm/sigcontext.h | 16 ++- arch/riscv/kernel/setup.c | 3 + arch/riscv/kernel/signal.c | 174 +++++++++++++++++++++-- 4 files changed, 193 insertions(+), 15 deletions(-)
arch/riscv/include/uapi/asm/ptrace.h +15 -0
@@ -71,6 +71,21 @@ struct __riscv_q_ext_state {__u32 reserved[3 ]; }; +struct __riscv_ctx_hdr { +__u32 magic; +__u32 size; +}; + +struct __riscv_extra_ext_header { +__u32 __padding[129 ] __attribute__((aligned (16 ))); +/* +* Reserved for expansion of sigcontext structure. Currently zeroed +* upon signal, and must be zero upon sigreturn. +*/ +__u32 reserved; +struct __riscv_ctx_hdr hdr; +}; +union __riscv_fp_state {struct __riscv_f_ext_state f;struct __riscv_d_ext_state d;
arch/riscv/include/uapi/asm/sigcontext.h +15 -1
@@ -8,6 +8,17 @@#include <asm/ptrace.h> +/* The Magic number for signal context frame header. */ +#define RISCV_V_MAGIC 0x53465457 +#define END_MAGIC 0x0 + +/* The size of END signal context header. */ +#define END_HDR_SIZE 0x0 + +struct __sc_riscv_v_state { +struct __riscv_v_ext_state v_state; +} __attribute__((aligned (16 ))); +/* * Signal context structure *
@@ -16,7 +27,10 @@*/ struct sigcontext {struct user_regs_struct sc_regs; -union __riscv_fp_state sc_fpregs; +union { +union __riscv_fp_state sc_fpregs; +struct __riscv_extra_ext_header sc_extdesc; + }; };#endif /* _UAPI_ASM_RISCV_SIGCONTEXT_H */
arch/riscv/kernel/setup.c +3 -0
@@ -262,6 +262,8 @@ static void __init parse_dtb(void)#endif } +extern void __init init_rt_signal_env (void ); +void __init setup_arch (char **cmdline_p) {parse_dtb ();
@@ -295,6 +297,7 @@ void __init setup_arch(char **cmdline_p)riscv_init_cbo_blocksizes ();riscv_fill_hwcap (); +init_rt_signal_env ();apply_boot_alternatives ();if (IS_ENABLED (CONFIG_RISCV_ISA_ZICBOM ) &&riscv_isa_extension_available (NULL ,ZICBOM ))
arch/riscv/kernel/signal.c +160 -14
249 lines not on this page · raw message
Signature · 1 line
2.17.1
Re: [PATCH -next v21 14/27] riscv: signal: Add sigcontext save/restore for vector
Hi,
Quoted 29 lines · Andy Chiu
From: Greentime Hu <[email protected]>
This patch facilitates the existing fp-reserved words for placement of the first extension's context header on the user's sigframe. A context header consists of a distinct magic word and the size, including the header itself, of an extension on the stack. Then, the frame is followed by the context of that extension, and then a header + context body for another extension if exists. If there is no more extension to come, then the frame must be ended with a null context header. A special case is rv64gc, where the kernel support no extensions requiring to expose additional regfile to the user. In such case the kernel would place the null context header right after the first reserved word of
__riscv_q_ext_statewhen saving sigframe. And the kernel would check if all reserved words are zeros when a signal handler returns.__riscv_q_ext_state---->| |<-__riscv_extra_ext_header
~ ~ .reserved[0]--->|0 |<- .reserved <-------|magic |<- .hdr | |size |_______ end of sc_fpregs | |ext-bdy| | ~ ~ +)size ------->|magic |<- another context header |size | |ext-bdy| ~ ~ |magic:0|<- null context header |size:0 |
Quoted 4 lines
The vector registers will be saved in datap pointer. The datap pointer will be allocated dynamically when the task needs in kernel space. On the other hand, datap pointer on the sigframe will be set right after the
__riscv_v_ext_statedata structure.
It appears that this patch somehow breaks userland, at least the rust compiler. This can be observed for instance by building the rust-lsd package in Debian, but many other rust packages are also affected:
* Failed build with kernel 6.5.3:
https://buildd.debian.org/status/fetch.php?pkg=rust-lsd&arch=riscv64&ver=0.23.1-7%2Bb1&stamp=1696475386&raw=0
* Successful build with kernel 6.4.13:
https://buildd.debian.org/status/fetch.php?pkg=rust-lsd&arch=riscv64&ver=0.23.1-7%2Bb1&stamp=1696491025&raw=0
It happens on hardware which does not have the V extension (in the above case on a Hifive Unmatched board). This can also be reproduced in a QEMU VM. Unfortunately disabling CONFIG_RISCV_ISA_V does not workaround the issue.
It is not clear to me if it is a kernel issue or a wrong assumption on the rust side. Any hint on how to continue investigating?
Regards
Aurelien
Signature · 2 lines
Aurelien Jarno GPG: 4096R/1DDD8C9B aurelien at aurel32.net http://aurel32.net
Re: [PATCH -next v21 14/27] riscv: signal: Add sigcontext save/restore for vector
Quoted 33 lines · On Sun, Oct 8, 2023 at 6:19?PM Aurelien Jarno <[email protected]> wrote:
Hi,
Quoted 29 lines · Andy Chiu
From: Greentime Hu <[email protected]>
This patch facilitates the existing fp-reserved words for placement of the first extension's context header on the user's sigframe. A context header consists of a distinct magic word and the size, including the header itself, of an extension on the stack. Then, the frame is followed by the context of that extension, and then a header + context body for another extension if exists. If there is no more extension to come, then the frame must be ended with a null context header. A special case is rv64gc, where the kernel support no extensions requiring to expose additional regfile to the user. In such case the kernel would place the null context header right after the first reserved word of
__riscv_q_ext_statewhen saving sigframe. And the kernel would check if all reserved words are zeros when a signal handler returns.__riscv_q_ext_state---->| |<-__riscv_extra_ext_header ~ ~ .reserved[0]--->|0 |<- .reserved <-------|magic |<- .hdr | |size |_______ end of sc_fpregs | |ext-bdy| | ~ ~ +)size ------->|magic |<- another context header |size | |ext-bdy| ~ ~ |magic:0|<- null context header |size:0 |
Quoted 8 lines
Quoted 4 lines
The vector registers will be saved in datap pointer. The datap pointer will be allocated dynamically when the task needs in kernel space. On the other hand, datap pointer on the sigframe will be set right after the
__riscv_v_ext_statedata structure.It appears that this patch somehow breaks userland, at least the rust compiler. This can be observed for instance by building the rust-lsd package in Debian, but many other rust packages are also affected:
Sorry for the time spent on pinpointing the issue. Yes, this is a bug and we had a fix [1]. This fix was accidently not getting into the -fixes branch, but it will. And it should be going into linux stable as well, though I am not certain about the timing. Otherwise, this bug may potentially break any processes which allocate a sigaltstack at an address higher than their stack.
Quoted 15 lines
* Failed build with kernel 6.5.3:
https://buildd.debian.org/status/fetch.php?pkg=rust-lsd&arch=riscv64&ver=0.23.1-7%2Bb1&stamp=1696475386&raw=0* Successful build with kernel 6.4.13:
https://buildd.debian.org/status/fetch.php?pkg=rust-lsd&arch=riscv64&ver=0.23.1-7%2Bb1&stamp=1696491025&raw=0It happens on hardware which does not have the V extension (in the above case on a Hifive Unmatched board). This can also be reproduced in a QEMU VM. Unfortunately disabling
CONFIG_RISCV_ISA_Vdoes not workaround the issue.It is not clear to me if it is a kernel issue or a wrong assumption on the rust side. Any hint on how to continue investigating?
Quoted 6 lines
Regards
Aurelien--
Aurelien Jarno GPG: 4096R/1DDD8C9B
aurelien at aurel32.net http://aurel32.net
[1]: https://yhbt.net/lore/all/mhng-7799d3a1-c12a-48e9-bb5f-e0a596892d78 at palmer-ri-x1c9/
Re: [PATCH -next v21 14/27] riscv: signal: Add sigcontext save/restore for vector
Hi,
Quoted 43 lines · Andy Chiu
Quoted 33 lines · On Sun, Oct 8, 2023 at 6:19?PM Aurelien Jarno <[email protected]> wrote:
Hi,
Quoted 29 lines · Andy Chiu
From: Greentime Hu <[email protected]>
This patch facilitates the existing fp-reserved words for placement of the first extension's context header on the user's sigframe. A context header consists of a distinct magic word and the size, including the header itself, of an extension on the stack. Then, the frame is followed by the context of that extension, and then a header + context body for another extension if exists. If there is no more extension to come, then the frame must be ended with a null context header. A special case is rv64gc, where the kernel support no extensions requiring to expose additional regfile to the user. In such case the kernel would place the null context header right after the first reserved word of
__riscv_q_ext_statewhen saving sigframe. And the kernel would check if all reserved words are zeros when a signal handler returns.__riscv_q_ext_state---->| |<-__riscv_extra_ext_header ~ ~ .reserved[0]--->|0 |<- .reserved <-------|magic |<- .hdr | |size |_______ end of sc_fpregs | |ext-bdy| | ~ ~ +)size ------->|magic |<- another context header |size | |ext-bdy| ~ ~ |magic:0|<- null context header |size:0 |Quoted 8 lines
Quoted 4 lines
The vector registers will be saved in datap pointer. The datap pointer will be allocated dynamically when the task needs in kernel space. On the other hand, datap pointer on the sigframe will be set right after the
__riscv_v_ext_statedata structure.It appears that this patch somehow breaks userland, at least the rust compiler. This can be observed for instance by building the rust-lsd package in Debian, but many other rust packages are also affected:
Quoted 6 lines
Sorry for the time spent on pinpointing the issue. Yes, this is a bug and we had a fix [1]. This fix was accidently not getting into the -fixes branch, but it will. And it should be going into linux stable as well, though I am not certain about the timing. Otherwise, this bug may potentially break any processes which allocate a sigaltstack at an address higher than their stack.
Thanks for the pointer, I somehow missed that patch when looking for a potential fix on the mailing list. I have just tried it, and I confirm it fixes the issue. Let's hope it can get merged soon.
Regards
Aurelien
Signature · 2 lines
Aurelien Jarno GPG: 4096R/1DDD8C9B aurelien at aurel32.net http://aurel32.net
PATCH v21 15/27 riscv: signal: Report signal frame size to userspace via auxv
In reply to Andy ChiuFrom: Vincent Chen <[email protected]>
The vector register belongs to the signal context. They need to be stored and restored as entering and leaving the signal handler. According to the V-extension specification, the maximum length of the vector registers can be 2^16. Hence, if userspace refers to the MINSIGSTKSZ to create a sigframe, it may not be enough. To resolve this problem, this patch refers to the commit 94b07c1f8c39c
("arm64: signal: Report signal frame size to userspace via auxv") to enable userspace to know the minimum required sigframe size through the auxiliary vector and use it to allocate enough memory for signal context.
Note that auxv always reports size of the sigframe as if V exists for all starting processes, whenever the kernel has CONFIG_RISCV_ISA_V. The reason is that users usually reference this value to allocate an alternative signal stack, and the user may use V anytime. So the user must reserve a space for V-context in sigframe in case that the signal handler invokes after the kernel allocating V.
---
Changelog V19:
- Fix a conflict in signal.c due to commit 8d736482749f ("riscv: add icache flush for nommu sigreturn trampoline")
--- arch/riscv/include/asm/elf.h | 9 +++++++++ arch/riscv/include/asm/processor.h | 2 ++ arch/riscv/include/uapi/asm/auxvec.h | 1 + arch/riscv/kernel/signal.c | 20 +++++++++++++++----- 4 files changed, 27 insertions(+), 5 deletions(-)
arch/riscv/include/asm/elf.h +9 -0
@@ -105,6 +105,15 @@ do { \ get_cache_size(3, CACHE_TYPE_UNIFIED)); \NEW_AUX_ENT (AT_L3_CACHEGEOMETRY , \get_cache_geometry (3 ,CACHE_TYPE_UNIFIED )); \ +/* \ +* Should always be nonzero unless there's a kernel bug. \ +* If we haven't determined a sensible value to give to \ +* userspace, omit the entry: \ +*/ \ +if (likely (signal_minsigstksz)) \ +NEW_AUX_ENT (AT_MINSIGSTKSZ , signal_minsigstksz); \ +else \ +NEW_AUX_ENT (AT_IGNORE ,0 ); \ } while (0)#define ARCH_HAS_SETUP_ADDITIONAL_PAGES struct linux_binprm ;
arch/riscv/include/asm/processor.h +2 -0
@@ -7,6 +7,7 @@#define _ASM_RISCV_PROCESSOR_H#include <linux/const.h> +#include <linux/cache.h> #include <vdso/processor.h>
@@ -81,6 +82,7 @@ int riscv_of_parent_hartid(struct device_node *node, unsigned long *hartid);extern void riscv_fill_hwcap (void );extern int arch_dup_task_struct (struct task_struct *dst,struct task_struct *src); +extern unsigned long signal_minsigstksz__ro_after_init ;#endif /* __ASSEMBLY__ */ #endif /* _ASM_RISCV_PROCESSOR_H */
arch/riscv/include/uapi/asm/auxvec.h +1 -0
@@ -35,5 +35,6 @@/* entries in ARCH_DLINFO */ #define AT_VECTOR_SIZE_ARCH 9 +#define AT_MINSIGSTKSZ 51#endif /* _UAPI_ASM_RISCV_AUXVEC_H */
arch/riscv/kernel/signal.c +15 -5
@@ -23,6 +23,8 @@#include <asm/csr.h> #include <asm/cacheflush.h> +unsigned long signal_minsigstksz__ro_after_init ; +extern u32 __user_rt_sigreturn[2 ];static size_t riscv_v_sc_size__ro_after_init ;
@@ -197,7 +199,7 @@ static long restore_sigcontext(struct pt_regs *regs,return err; } -static size_t get_rt_frame_size (void ) +static size_t get_rt_frame_size (bool cal_all) {struct rt_sigframe __user *frame;size_t frame_size;
@@ -205,8 +207,10 @@ static size_t get_rt_frame_size(void) frame_size =sizeof (*frame); -if (has_vector () &&riscv_v_vstate_query (task_pt_regs (current))) - total_context_size += riscv_v_sc_size; +if (has_vector ()) { +if (cal_all ||riscv_v_vstate_query (task_pt_regs (current))) + total_context_size += riscv_v_sc_size; + }/* * Preserved a __riscv_ctx_hdr for END signal context header if an * extension uses __riscv_extra_ext_header
@@ -226,7 +230,7 @@ SYSCALL_DEFINE0(rt_sigreturn)struct rt_sigframe __user *frame;struct task_struct *task;sigset_t set; -size_t frame_size =get_rt_frame_size (); +size_t frame_size =get_rt_frame_size (false);/* Always make any pending restarted system calls return -EINTR */ current->restart_block.fn = do_no_restart_syscall;
@@ -323,7 +327,7 @@ static int setup_rt_frame(struct ksignal *ksig, sigset_t *set,struct rt_sigframe __user *frame;long err =0 ;unsigned long __maybe_unused addr; -size_t frame_size =get_rt_frame_size (); +size_t frame_size =get_rt_frame_size (false); frame =get_sigframe (ksig, regs, frame_size);if (!access_ok (frame, frame_size))
@@ -465,4 +469,10 @@ void __init init_rt_signal_env(void) { riscv_v_sc_size =sizeof (struct __riscv_ctx_hdr ) +sizeof (struct __sc_riscv_v_state ) + riscv_v_vsize; +/* +* Determine the stack space required for guaranteed signal delivery. +* The signal_minsigstksz will be populated into the AT_MINSIGSTKSZ entry +* in the auxiliary array at process startup. +*/ + signal_minsigstksz =get_rt_frame_size (true); }
Signature · 1 line
2.17.1
PATCH v21 16/27 riscv: signal: validate altstack to reflect Vector
In reply to Andy ChiuSome extensions, such as Vector, dynamically change footprint on a signal frame, so MINSIGSTKSZ is no longer accurate. For example, an RV64V implementation with vlen = 512 may occupy 2K + 40 + 12 Bytes of a signal frame with the upcoming support. And processes that do not execute any vector instructions do not need to reserve the extra sigframe. So we need a way to guard the allocation size of the sigframe at process runtime according to current status of V.
Thus, provide the function sigaltstack_size_valid() to validate its size based on current allocation status of supported extensions.
--- arch/riscv/kernel/signal.c | 7 +++++++ 1 file changed, 7 insertions(+)
arch/riscv/kernel/signal.c +7 -0
@@ -476,3 +476,10 @@ void __init init_rt_signal_env(void)*/ signal_minsigstksz =get_rt_frame_size (true); } + +#ifdef CONFIG_DYNAMIC_SIGFRAME +bool sigaltstack_size_valid (size_t ss_size) +{ +return ss_size >get_rt_frame_size (false); +} +#endif /* CONFIG_DYNAMIC_SIGFRAME */
Signature · 1 line
2.17.1
PATCH v21 17/27 riscv: prevent stack corruption by reserving task_pt_regs(p) early
In reply to Andy ChiuFrom: Greentime Hu <[email protected]>
Early function calls, such as setup_vm(), relocate_enable_mmu(), soc_early_init() etc, are free to operate on stack. However, PT_SIZE_ON_STACK bytes at the head of the kernel stack are purposedly reserved for the placement of per-task register context pointed by task_pt_regs(p). Those functions may corrupt task_pt_regs if we overlap the $sp with it. In fact, we had accidentally corrupted sstatus.VS in some tests, treating the kernel to save V context before V was actually allocated, resulting in a kernel panic.
Thus, we should skip PT_SIZE_ON_STACK for $sp before making C function calls from the top-level assembly.
--- arch/riscv/kernel/head.S | 2 ++ 1 file changed, 2 insertions(+)
arch/riscv/kernel/head.S +2 -0
@@ -301,6 +301,7 @@ clear_bss_done: la tp, init_task la sp, init_thread_union + THREAD_SIZE XIP_FIXUP_OFFSET sp + addi sp, sp, -PT_SIZE_ON_STACK #ifdef CONFIG_BUILTIN_DTB la a0, __dtb_start XIP_FIXUP_OFFSET a0
@@ -318,6 +319,7 @@ clear_bss_done: /* Restore C environment */ la tp, init_task la sp, init_thread_union + THREAD_SIZE + addi sp, sp, -PT_SIZE_ON_STACK #ifdef CONFIG_KASAN call kasan_early_init
Signature · 1 line
2.17.1
PATCH v21 18/27 riscv: kvm: Add V extension to KVM ISA
In reply to Andy ChiuFrom: Vincent Chen <[email protected]>
Add V extension to KVM isa extension list to enable supporting of V extension on VCPUs.
--- arch/riscv/include/uapi/asm/kvm.h | 1 + arch/riscv/kvm/vcpu.c | 1 + 2 files changed, 2 insertions(+)
arch/riscv/include/uapi/asm/kvm.h +1 -0
@@ -121,6 +121,7 @@ enum KVM_RISCV_ISA_EXT_ID {KVM_RISCV_ISA_EXT_ZICBOZ ,KVM_RISCV_ISA_EXT_ZBB ,KVM_RISCV_ISA_EXT_SSAIA , +KVM_RISCV_ISA_EXT_V , KVM_RISCV_ISA_EXT_MAX, };
arch/riscv/kvm/vcpu.c +1 -0
@@ -57,6 +57,7 @@ static const unsigned long kvm_isa_ext_arr[] = { [KVM_RISCV_ISA_EXT_H] = RISCV_ISA_EXT_h, [KVM_RISCV_ISA_EXT_I] = RISCV_ISA_EXT_i, [KVM_RISCV_ISA_EXT_M] = RISCV_ISA_EXT_m, + [KVM_RISCV_ISA_EXT_V] = RISCV_ISA_EXT_v,KVM_ISA_EXT_ARR (SSAIA ), KVM_ISA_EXT_ARR(SSTC),
Signature · 1 line
2.17.1
PATCH v21 19/27 riscv: KVM: Add vector lazy save/restore support
In reply to Andy ChiuFrom: Vincent Chen <[email protected]>
This patch adds vector context save/restore for guest VCPUs. To reduce the impact on KVM performance, the implementation imitates the FP context switch mechanism to lazily store and restore the vector context only when the kernel enters/exits the in-kernel run loop and not during the KVM world switch.
---
Changelog V19:
- remap V extension registers as type 9 in uapi/asm/kvm.h
--- arch/riscv/include/asm/kvm_host.h | 2 + arch/riscv/include/asm/kvm_vcpu_vector.h | 82 ++++++++++ arch/riscv/include/uapi/asm/kvm.h | 7 + arch/riscv/kvm/Makefile | 1 + arch/riscv/kvm/vcpu.c | 22 +++ arch/riscv/kvm/vcpu_vector.c | 186 +++++++++++++++++++++++ 6 files changed, 300 insertions(+) create mode 100644 arch/riscv/include/asm/kvm_vcpu_vector.h create mode 100644 arch/riscv/kvm/vcpu_vector.c
arch/riscv/include/asm/kvm_host.h +2 -0
@@ -15,6 +15,7 @@#include <linux/spinlock.h> #include <asm/hwcap.h> #include <asm/kvm_aia.h> +#include <asm/ptrace.h> #include <asm/kvm_vcpu_fp.h> #include <asm/kvm_vcpu_insn.h> #include <asm/kvm_vcpu_sbi.h>
@@ -145,6 +146,7 @@ struct kvm_cpu_context {unsigned long sstatus;unsigned long hstatus;union __riscv_fp_state fp; +struct __riscv_v_ext_state vector; };struct kvm_vcpu_csr {
arch/riscv/include/asm/kvm_vcpu_vector.h new file +82 -0
82 lines not on this page · raw message
arch/riscv/include/uapi/asm/kvm.h +7 -0
@@ -204,6 +204,13 @@ enum KVM_RISCV_SBI_EXT_ID {#define KVM_REG_RISCV_SBI_MULTI_REG_LAST \ KVM_REG_RISCV_SBI_MULTI_REG(KVM_RISCV_SBI_EXT_MAX - 1) +/* V extension registers are mapped as type 9 */ +#define KVM_REG_RISCV_VECTOR (0x09 << KVM_REG_RISCV_TYPE_SHIFT) +#define KVM_REG_RISCV_VECTOR_CSR_REG (name) \ + (offsetof(struct __riscv_v_ext_state, name) / sizeof(unsigned long)) +#define KVM_REG_RISCV_VECTOR_REG (n) \ + ((n) + sizeof(struct __riscv_v_ext_state) / sizeof(unsigned long)) +#endif #endif /* __LINUX_KVM_RISCV_H */
arch/riscv/kvm/Makefile +1 -0
@@ -17,6 +17,7 @@ kvm-y += mmu.o kvm-y += vcpu.o kvm-y += vcpu_exit.o kvm-y += vcpu_fp.o +kvm-y += vcpu_vector.o kvm-y += vcpu_insn.o kvm-y += vcpu_switch.o kvm-y += vcpu_sbi.o
arch/riscv/kvm/vcpu.c +22 -0
@@ -22,6 +22,8 @@#include <asm/cacheflush.h> #include <asm/hwcap.h> #include <asm/sbi.h> +#include <asm/vector.h> +#include <asm/kvm_vcpu_vector.h> const struct _kvm_stats_desc kvm_vcpu_stats_desc[] = {KVM_GENERIC_VCPU_STATS (),
@@ -139,6 +141,8 @@ static void kvm_riscv_reset_vcpu(struct kvm_vcpu *vcpu)kvm_riscv_vcpu_fp_reset (vcpu); +kvm_riscv_vcpu_vector_reset (vcpu); +kvm_riscv_vcpu_timer_reset (vcpu);kvm_riscv_vcpu_aia_reset (vcpu);
@@ -199,6 +203,9 @@ int kvm_arch_vcpu_create(struct kvm_vcpu *vcpu) cntx->hstatus |=HSTATUS_SPVP ; cntx->hstatus |=HSTATUS_SPV ; +if (kvm_riscv_vcpu_alloc_vector_context (vcpu, cntx)) +return -ENOMEM ; +/* By default, make CY, TM, and IR counters accessible in VU mode */ reset_csr->scounteren =0x7 ;
@@ -242,6 +249,9 @@ void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu)/* Free unused pages pre-allocated for G-stage page table mappings */ kvm_mmu_free_memory_cache (&vcpu->arch.mmu_page_cache); + +/* Free vector context space for host and guest kernel */ +kvm_riscv_vcpu_free_vector_context (vcpu); }int kvm_cpu_has_pending_timer (struct kvm_vcpu *vcpu)
@@ -680,6 +690,9 @@ static int kvm_riscv_vcpu_set_reg(struct kvm_vcpu *vcpu,return kvm_riscv_vcpu_set_reg_isa_ext (vcpu, reg);case KVM_REG_RISCV_SBI_EXT :return kvm_riscv_vcpu_set_reg_sbi_ext (vcpu, reg); +case KVM_REG_RISCV_VECTOR : +return kvm_riscv_vcpu_set_reg_vector (vcpu, reg, +KVM_REG_RISCV_VECTOR );default :break ; }
@@ -709,6 +722,9 @@ static int kvm_riscv_vcpu_get_reg(struct kvm_vcpu *vcpu,return kvm_riscv_vcpu_get_reg_isa_ext (vcpu, reg);case KVM_REG_RISCV_SBI_EXT :return kvm_riscv_vcpu_get_reg_sbi_ext (vcpu, reg); +case KVM_REG_RISCV_VECTOR : +return kvm_riscv_vcpu_get_reg_vector (vcpu, reg, +KVM_REG_RISCV_VECTOR );default :break ; }
@@ -1003,6 +1019,9 @@ void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)kvm_riscv_vcpu_host_fp_save (&vcpu->arch.host_context);kvm_riscv_vcpu_guest_fp_restore (&vcpu->arch.guest_context, vcpu->arch.isa); +kvm_riscv_vcpu_host_vector_save (&vcpu->arch.host_context); +kvm_riscv_vcpu_guest_vector_restore (&vcpu->arch.guest_context, + vcpu->arch.isa);kvm_riscv_vcpu_aia_load (vcpu, cpu);
@@ -1022,6 +1041,9 @@ void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)kvm_riscv_vcpu_host_fp_restore (&vcpu->arch.host_context);kvm_riscv_vcpu_timer_save (vcpu); +kvm_riscv_vcpu_guest_vector_save (&vcpu->arch.guest_context, + vcpu->arch.isa); +kvm_riscv_vcpu_host_vector_restore (&vcpu->arch.host_context); csr->vsstatus =csr_read (CSR_VSSTATUS ); csr->vsie =csr_read (CSR_VSIE );
arch/riscv/kvm/vcpu_vector.c new file +186 -0
186 lines not on this page · raw message
Signature · 1 line
2.17.1
PATCH v21 20/27 riscv: hwcap: change ELF_HWCAP to a function
In reply to Andy ChiuUsing a function is flexible to represent ELF_HWCAP. So the kernel may encode hwcap reflecting supported hardware features just at the moment of the start of each program.
This will be helpful when we introduce prctl/sysctl interface to control per-process availability of Vector extension in following patches. Programs started with V disabled should see V masked off in theirs ELF_HWCAP.
--- arch/riscv/include/asm/elf.h | 2 +- arch/riscv/include/asm/hwcap.h | 2 ++ arch/riscv/kernel/cpufeature.c | 5 +++++ 3 files changed, 8 insertions(+), 1 deletion(-)
arch/riscv/include/asm/elf.h +1 -1
@@ -66,7 +66,7 @@ extern bool compat_elf_check_arch(Elf32_Ehdr *hdr);* via a bitmap that coorespends to each single-letter ISA extension. This is * essentially defunct, but will remain for compatibility with userspace. */ -#define ELF_HWCAP (elf_hwcap & ((1UL << RISCV_ISA_EXT_BASE) - 1)) +#define ELF_HWCAP riscv_get_elf_hwcap()extern unsigned long elf_hwcap;/*
arch/riscv/include/asm/hwcap.h +2 -0
@@ -61,6 +61,8 @@#include <linux/jump_label.h> +unsigned long riscv_get_elf_hwcap (void ); +struct riscv_isa_ext_data {/* Name of the extension displayed to userspace via /proc/cpuinfo */ char uprop[RISCV_ISA_EXT_NAME_LEN_MAX ];
arch/riscv/kernel/cpufeature.c +5 -0
@@ -293,6 +293,11 @@ void __init riscv_fill_hwcap(void)pr_info ("riscv: ELF capabilities %s\n" , print_str); } +unsigned long riscv_get_elf_hwcap (void ) +{ +return (elf_hwcap & ((1UL <<RISCV_ISA_EXT_BASE ) -1 )); +} +#ifdef CONFIG_RISCV_ALTERNATIVE /* * Alternative patch sites consider 48 bits when determining when to patch
Signature · 1 line
2.17.1
Re: [PATCH -next v21 20/27] riscv: hwcap: change ELF_HWCAP to a function
Quoted 8 lines · Andy Chiu
Using a function is flexible to represent
ELF_HWCAP. So the kernel may encode hwcap reflecting supported hardware features just at the moment of the start of each program.This will be helpful when we introduce prctl/sysctl interface to control per-process availability of Vector extension in following patches. Programs started with V disabled should see V masked off in theirs
ELF_HWCAP.
For the uninformed, like myself, this needs to be a function, rather than open coding the masking of the V bit at the one user of ELF_HWCAP this series adds, because the binfmt stuff needs to get the value in create_elf_fdpic_tables() & co?
If that's your purpose,
One minor comment below.
Quoted 39 lines
Signed-off-by Andy Chiu <[email protected]>--- arch/riscv/include/asm/elf.h | 2 +- arch/riscv/include/asm/hwcap.h | 2 ++ arch/riscv/kernel/cpufeature.c | 5 +++++ 3 files changed, 8 insertions(+), 1 deletion(-)arch/riscv/include/asm/elf.h +1
-1@@ -66,7 +66,7 @@ extern bool compat_elf_check_arch(Elf32_Ehdr *hdr);* via a bitmap that coorespends to each single-letter ISA extension. This is * essentially defunct, but will remain for compatibility with userspace. */ -#define ELF_HWCAP (elf_hwcap & ((1UL << RISCV_ISA_EXT_BASE) - 1)) +#define ELF_HWCAP riscv_get_elf_hwcap()extern unsigned long elf_hwcap;/* arch/riscv/include/asm/hwcap.h +2
-0@@ -61,6 +61,8 @@#include <linux/jump_label.h> +unsigned long riscv_get_elf_hwcap (void ); +struct riscv_isa_ext_data {/* Name of the extension displayed to userspace via /proc/cpuinfo */ char uprop[RISCV_ISA_EXT_NAME_LEN_MAX ];arch/riscv/kernel/cpufeature.c +0
-0
arch/riscv/kernel/cpufeature.c void __init riscv_fill_hwcap(void) · line 298
Quoted 6 lines
arch/riscv/kernel/cpufeature.c +3
-0@@ -293,6 +293,11 @@ void __init riscv_fill_hwcap(void)pr_info ("riscv: ELF capabilities %s\n" , print_str); } +unsigned long riscv_get_elf_hwcap (void ) +{ +return (elf_hwcap & ((1UL <<RISCV_ISA_EXT_BASE ) -1 ));
If you respin for some other reason, could you drop the open coded mask creation as part of these changes?
Cheers,
Conor.
Quoted 13 lines
arch/riscv/kernel/cpufeature.c +2
-0+} +#ifdef CONFIG_RISCV_ALTERNATIVE /* * Alternative patch sites consider 48 bits when determining when to patch Signature · 7 lines
2.17.1 _______________________________________________ linux-riscv mailing list linux-riscv at lists.infradead.org http://lists.infradead.org/mailman/listinfo/linux-riscv
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 228 bytes
Desc: not available
URL: <http://lists.infradead.org/pipermail/kvm-riscv/attachments/20230605/9dd4e51c/attachment-0001.sig>
Re: [PATCH -next v21 20/27] riscv: hwcap: change ELF_HWCAP to a function
Quoted 10 lines · Le maanantaina 5. kesäkuuta 2023, 19.24.53 EEST Conor Dooley a écrit :
Quoted 8 lines · Andy Chiu
Using a function is flexible to represent
ELF_HWCAP. So the kernel may encode hwcap reflecting supported hardware features just at the moment of the start of each program.This will be helpful when we introduce prctl/sysctl interface to control per-process availability of Vector extension in following patches. Programs started with V disabled should see V masked off in theirs
ELF_HWCAP.
Quoted 4 lines
For the uninformed, like myself, this needs to be a function, rather than open coding the masking of the V bit at the one user of
ELF_HWCAPthis series adds, because the binfmt stuff needs to get the value increate_elf_fdpic_tables()& co?
No, this is to deal with existing userspace that expects vectors to "just work" when the V bit is set (e.g. FFmpeg 6.0).
Signature · 10 lines
レミ・デニ-クールモン http://www.remlab.net/ _______________________________________________ linux-riscv mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-riscv
Re: [PATCH -next v21 20/27] riscv: hwcap: change ELF_HWCAP to a function
Quoted 11 lines · Rémi Denis-Courmont
Quoted 10 lines · Le maanantaina 5. kesäkuuta 2023, 19.24.53 EEST Conor Dooley a écrit :
Quoted 8 lines · Andy Chiu
Using a function is flexible to represent
ELF_HWCAP. So the kernel may encode hwcap reflecting supported hardware features just at the moment of the start of each program.This will be helpful when we introduce prctl/sysctl interface to control per-process availability of Vector extension in following patches. Programs started with V disabled should see V masked off in theirs
ELF_HWCAP.
Quoted 7 lines
Quoted 4 lines
For the uninformed, like myself, this needs to be a function, rather than open coding the masking of the V bit at the one user of
ELF_HWCAPthis series adds, because the binfmt stuff needs to get the value increate_elf_fdpic_tables()& co?No, this is to deal with existing userspace that expects vectors to "just work" when the V bit is set (e.g. FFmpeg 6.0).
Is that not what create_elf_fdpic_tables() & its ilk do? Inform userspace of whether the V bit is set?
Cheers,
Conor.
_______________________________________________
linux-riscv mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-riscv
PATCH v21 21/27 riscv: Add prctl controls for userspace vector management
In reply to Andy ChiuThis patch add two riscv-specific prctls, to allow usespace control the use of vector unit:
* PR_RISCV_V_SET_CONTROL: control the permission to use Vector at next, or all following execve for a thread. Turning off a thread's Vector live is not possible since libraries may have registered ifunc that may execute Vector instructions.
* PR_RISCV_V_GET_CONTROL: get the same permission setting for the current thread, and the setting for following execve(s).
---
Changelog v21:
- Remove the check in riscv_v_first_use_handler() because it is checked in ELF_HWCAP.
- Mask off COMPAT_HWCAP_ISA_V in ELF_HWCAP when the user process is not allowed to use it.
- properly define RISCV_V_{GET,SET}_CONTROL and remove unecessary #else clause. (Bj?rn)
Changelog v20:
- address build issue when KVM is compile as a module (Heiko)
- s/RISCV_V_DISABLE/RISCV_ISA_V_DEFAULT_ENABLE/ (Conor)
- change function names to have better scoping
- check has_vector() before accessing vstate_ctrl
- use proper return type for prctl calls (long instead of uint)
--- arch/riscv/include/asm/processor.h | 10 +++ arch/riscv/include/asm/vector.h | 4 + arch/riscv/kernel/cpufeature.c | 9 ++- arch/riscv/kernel/process.c | 1 + arch/riscv/kernel/vector.c | 114 +++++++++++++++++++++++++++++ arch/riscv/kvm/vcpu.c | 2 + include/uapi/linux/prctl.h | 11 +++ kernel/sys.c | 12 +++ 8 files changed, 162 insertions(+), 1 deletion(-)
arch/riscv/include/asm/processor.h +10 -0
@@ -40,6 +40,7 @@ struct thread_struct {unsigned long s[12 ];/* s[0]: frame pointer */ struct __riscv_d_ext_state fstate;unsigned long bad_cause; +unsigned long vstate_ctrl;struct __riscv_v_ext_state vstate; };
@@ -83,6 +84,15 @@ extern void riscv_fill_hwcap(void);extern int arch_dup_task_struct (struct task_struct *dst,struct task_struct *src);extern unsigned long signal_minsigstksz__ro_after_init ; + +#ifdef CONFIG_RISCV_ISA_V +/* Userspace interface for PR_RISCV_V_{SET,GET}_VS prctl()s: */ +#define RISCV_V_SET_CONTROL (arg) riscv_v_vstate_ctrl_set_current(arg) +#define RISCV_V_GET_CONTROL () riscv_v_vstate_ctrl_get_current() +extern long riscv_v_vstate_ctrl_set_current (unsigned long arg); +extern long riscv_v_vstate_ctrl_get_current (void ); +#endif /* CONFIG_RISCV_ISA_V */ +#endif /* __ASSEMBLY__ */ #endif /* _ASM_RISCV_PROCESSOR_H */
arch/riscv/include/asm/vector.h +4 -0
@@ -160,6 +160,9 @@ static inline void __switch_to_vector(struct task_struct *prev, riscv_v_vstate_restore(next, task_pt_regs(next)); } +void riscv_v_vstate_ctrl_init (struct task_struct *tsk); +bool riscv_v_vstate_ctrl_user_allowed (void ); +#else /* ! CONFIG_RISCV_ISA_V */ struct pt_regs ;
@@ -168,6 +171,7 @@ static inline int riscv_v_setup_vsize(void) { return -EOPNOTSUPP; } static __always_inline bool has_vector(void) { return false; }static inline bool riscv_v_first_use_handler (struct pt_regs *regs) {return false; }static inline bool riscv_v_vstate_query (struct pt_regs *regs) {return false; } +static inline bool riscv_v_vstate_ctrl_user_allowed (void ) {return false; }#define riscv_v_vsize (0)#define riscv_v_vstate_save (task, regs) do {} while (0)#define riscv_v_vstate_restore (task, regs) do {} while (0)
arch/riscv/kernel/cpufeature.c +8 -1
@@ -295,7 +295,14 @@ void __init riscv_fill_hwcap(void)unsigned long riscv_get_elf_hwcap (void ) { -return (elf_hwcap & ((1UL <<RISCV_ISA_EXT_BASE ) -1 )); +unsigned long hwcap; + + hwcap = (elf_hwcap & ((1UL <<RISCV_ISA_EXT_BASE ) -1 )); + +if (!riscv_v_vstate_ctrl_user_allowed ()) + hwcap &= ~COMPAT_HWCAP_ISA_V ; + +return hwcap; }#ifdef CONFIG_RISCV_ALTERNATIVE
arch/riscv/kernel/process.c +1 -0
@@ -149,6 +149,7 @@ void flush_thread(void)#endif #ifdef CONFIG_RISCV_ISA_V /* Reset vector state */ +riscv_v_vstate_ctrl_init (current);riscv_v_vstate_off (task_pt_regs (current));kfree (current->thread.vstate.datap);memset (¤t->thread.vstate,0 ,sizeof (struct __riscv_v_ext_state ));
arch/riscv/kernel/vector.c +114 -0
135 lines not on this page · raw message
arch/riscv/kvm/vcpu.c +2 -0
@@ -88,6 +88,8 @@ static bool kvm_riscv_vcpu_isa_enable_allowed(unsigned long ext)switch (ext) {case KVM_RISCV_ISA_EXT_H :return false; +case KVM_RISCV_ISA_EXT_V : +return riscv_v_vstate_ctrl_user_allowed ();default :break ; }
include/uapi/linux/prctl.h +11 -0
@@ -294,4 +294,15 @@ struct prctl_mm_map {#define PR_SET_MEMORY_MERGE 67#define PR_GET_MEMORY_MERGE 68 + +#define PR_RISCV_V_SET_CONTROL 69 +#define PR_RISCV_V_GET_CONTROL 70 +# define PR_RISCV_V_VSTATE_CTRL_DEFAULT 0 +# define PR_RISCV_V_VSTATE_CTRL_OFF 1 +# define PR_RISCV_V_VSTATE_CTRL_ON 2 +# define PR_RISCV_V_VSTATE_CTRL_INHERIT (1 << 4) +# define PR_RISCV_V_VSTATE_CTRL_CUR_MASK 0x3 +# define PR_RISCV_V_VSTATE_CTRL_NEXT_MASK 0xc +# define PR_RISCV_V_VSTATE_CTRL_MASK 0x1f +#endif /* _LINUX_PRCTL_H */
kernel/sys.c +12 -0
@@ -140,6 +140,12 @@#ifndef GET_TAGGED_ADDR_CTRL # define GET_TAGGED_ADDR_CTRL () (-EINVAL)#endif +#ifndef RISCV_V_SET_CONTROL +# define RISCV_V_SET_CONTROL @ (-EINVAL) +#endif +#ifndef RISCV_V_GET_CONTROL +# define RISCV_V_GET_CONTROL () (-EINVAL) +#endif /* * this is where the system-wide overflow UID and GID are defined, for
@@ -2708,6 +2714,12 @@ SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3, error = !!test_bit (MMF_VM_MERGE_ANY , &me->mm->flags);break ;#endif +case PR_RISCV_V_SET_CONTROL : + error =RISCV_V_SET_CONTROL (arg2); +break ; +case PR_RISCV_V_GET_CONTROL : + error =RISCV_V_GET_CONTROL (); +break ;default : error = -EINVAL ;break ;
Signature · 1 line
2.17.1
PATCH v21 22/27 riscv: Add sysctl to set the default vector rule for new processes
In reply to Andy ChiuTo support Vector extension, the series exports variable-length vector registers on the signal frame. However, this potentially breaks abi if processing vector registers is required in the signal handler for old binaries. For example, there is such need if user-level context switch is triggerred via signals[1].
For this reason, it is best to leave a decision to distro maintainers, where the enablement of userspace Vector for new launching programs can be controlled. Developers may also need the switch to experiment with. The parameter is configurable through sysctl interface so a distro may turn off Vector early at init script if the break really happens in the wild.
The switch will only take effects on new execve() calls once set. This will not effect existing processes that do not call execve(), nor processes which has been set with a non-default vstate_ctrl by making explicit PR_RISCV_V_SET_CONTROL prctl() calls.
---
Changelog v20:
- Use READ_ONCE to access riscv_v_implicit_uacc (Bj?rn)
--- arch/riscv/kernel/vector.c | 33 ++++++++++++++++++++++++++++++++- 1 file changed, 32 insertions(+), 1 deletion(-)
arch/riscv/kernel/vector.c +32 -1
@@ -180,7 +180,7 @@ void riscv_v_vstate_ctrl_init(struct task_struct *tsk) next =riscv_v_ctrl_get_next (tsk);if (!next) { -if (riscv_v_implicit_uacc) +if (READ_ONCE (riscv_v_implicit_uacc)) cur =PR_RISCV_V_VSTATE_CTRL_ON ;else cur =PR_RISCV_V_VSTATE_CTRL_OFF ;
@@ -243,3 +243,34 @@ long riscv_v_vstate_ctrl_set_current(unsigned long arg)return -EINVAL ; } + +#ifdef CONFIG_SYSCTL + +static struct ctl_table riscv_v_default_vstate_table[] = { + { + .procname ="riscv_v_default_allow" , + .data = &riscv_v_implicit_uacc, + .maxlen =sizeof (riscv_v_implicit_uacc), + .mode =0644 , + .proc_handler = proc_dobool, + }, + { } +}; + +static int __init riscv_v_sysctl_init (void ) +{ +if (has_vector ()) +if (!register_sysctl ("abi" , riscv_v_default_vstate_table)) +return -EINVAL ; +return 0 ; +} + +#else /* ! CONFIG_SYSCTL */ +static int __init riscv_v_sysctl_init (void ) {return 0 ; } +#endif /* ! CONFIG_SYSCTL */ + +static int riscv_v_init (void ) +{ +return riscv_v_sysctl_init (); +} +core_initcall (riscv_v_init);
Signature · 1 line
2.17.1
PATCH v21 23/27 riscv: detect assembler support for .option arch
In reply to Andy ChiuSome extensions use .option arch directive to selectively enable certain extensions in parts of its assembly code. For example, Zbb uses it to inform assmebler to emit bit manipulation instructions. However, supporting of this directive only exist on GNU assembler and has not landed on clang at the moment, making TOOLCHAIN_HAS_ZBB depend on AS_IS_GNU.
While it is still under review at https://reviews.llvm.org/D123515, the upcoming Vector patch also requires this feature in assembler. Thus, provide Kconfig AS_HAS_OPTION_ARCH to detect such feature. Then TOOLCHAIN_HAS_XXX will be turned on automatically when the feature land.
--- arch/riscv/Kconfig | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-)
arch/riscv/Kconfig +7 -1
@@ -262,6 +262,12 @@ config RISCV_DMA_NONCOHERENT config AS_HAS_INSN def_bool $(as-instr,.insn r 51$(comma) 0$(comma) 0$(comma) t0$(comma) t0$(comma) zero) +config AS_HAS_OPTION_ARCH + # https://reviews.llvm.org/D123515 + def_bool y + depends on $(as-instr, .option arch$(comma) +m) + depends on !$(as-instr, .option arch$(comma) -i) + source "arch/riscv/Kconfig.socs" source "arch/riscv/Kconfig.errata"
@@ -466,7 +472,7 @@ config TOOLCHAIN_HAS_ZBB depends on !64BIT || $(cc-option,-mabi=lp64 -march=rv64ima_zbb) depends on !32BIT || $(cc-option,-mabi=ilp32 -march=rv32ima_zbb) depends on LLD_VERSION >= 150000 || LD_VERSION >= 23900 - depends on AS_IS_GNU + depends on AS_HAS_OPTION_ARCH config RISCV_ISA_ZBB bool "Zbb extension support for bit manipulation instructions"
Signature · 1 line
2.17.1
Nathan Chancellor
maintainerRe: [PATCH -next v21 23/27] riscv: detect assembler support for .option arch
Hi Andy,
Quoted 11 lines · Andy Chiu
Some extensions use .option arch directive to selectively enable certain extensions in parts of its assembly code. For example, Zbb uses it to inform assmebler to emit bit manipulation instructions. However, supporting of this directive only exist on GNU assembler and has not landed on clang at the moment, making
TOOLCHAIN_HAS_ZBBdepend onAS_IS_GNU.While it is still under review at https://reviews.llvm.org/D123515, the upcoming Vector patch also requires this feature in assembler. Thus, provide
KconfigAS_HAS_OPTION_ARCHto detect such feature. ThenTOOLCHAIN_HAS_XXXwill be turned on automatically when the feature land.
Just an FYI, this change has landed in LLVM main, so it should be in LLVM 17 in a few months:
https://github.com/llvm/llvm-project/commit/9e8ed3403c191ab9c4903e8eeb8f732ff8a43cb4
If you have to spin another revision for some reason, consider updating the Phabricator link to that one, as I expect that link to remain more stable in the long run over the Phabricator one, as LLVM is planning to eventually move away from Phabricator to GitHub pull requests. I don't think this is worth respinning on its own (obviously, heh).
The rest of the change still looks good to me, thanks again for taking this up.
Quoted 39 lines
Suggested-by Nathan Chancellor <[email protected]>Signed-off-by Andy Chiu <[email protected]>Reviewed-by Conor Dooley <[email protected]>Reviewed-by Nathan Chancellor <[email protected]>Reviewed-by Heiko Stuebner <[email protected]>Tested-by Heiko Stuebner <[email protected]>--- arch/riscv/Kconfig | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-)arch/riscv/Kconfig +7
-1@@ -262,6 +262,12 @@ config RISCV_DMA_NONCOHERENT config AS_HAS_INSN def_bool $(as-instr,.insn r 51$(comma) 0$(comma) 0$(comma) t0$(comma) t0$(comma) zero) +config AS_HAS_OPTION_ARCH + # https://reviews.llvm.org/D123515 + def_bool y + depends on $(as-instr, .option arch$(comma) +m) + depends on !$(as-instr, .option arch$(comma) -i) + source "arch/riscv/Kconfig.socs" source "arch/riscv/Kconfig.errata"@@ -466,7 +472,7 @@ config TOOLCHAIN_HAS_ZBB depends on !64BIT || $(cc-option,-mabi=lp64 -march=rv64ima_zbb) depends on !32BIT || $(cc-option,-mabi=ilp32 -march=rv32ima_zbb) depends on LLD_VERSION >= 150000 || LD_VERSION >= 23900 - depends on AS_IS_GNU + depends on AS_HAS_OPTION_ARCH config RISCV_ISA_ZBB bool "Zbb extension support for bit manipulation instructions"Signature · 1 line
2.17.1
Re: [PATCH -next v21 23/27] riscv: detect assembler support for .option arch
Quoted 2 lines · Nathan Chancellor
Just an FYI, this change has landed in LLVM main, so it should be in LLVM 17 in a few months:
Great! :)
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 228 bytes
Desc: not available
URL: <http://lists.infradead.org/pipermail/kvm-riscv/attachments/20230605/a932daa4/attachment-0001.sig>
Re: [PATCH -next v21 23/27] riscv: detect assembler support for .option arch
In reply to Andy ChiuHi Andy,
arch/riscv/Kconfig config RISCV_DMA_NONCOHERENT · line 269
Quoted 5 lines · Andy Chiu
arch/riscv/Kconfig +5
-0+config AS_HAS_OPTION_ARCH + # https://reviews.llvm.org/D123515 + def_bool y + depends on $(as-instr, .option arch$(comma) +m) + depends on !$(as-instr, .option arch$(comma) -i)
With tip-of-tree clang (llvm-project commit 85a8e5c3e0586e85), I'm seeing AS_HAS_OPTION_ARCH be set to n. It's the second "depends on" that makes it be set to n, so apparently clang started accepting ".option arch -i". What was your intent here for checking that ".option arch -i" is not supported? I'd think that just the first "depends on" would be sufficient.
- Eric
Palmer Dabbelt
maintainerRe: [PATCH -next v21 23/27] riscv: detect assembler support for .option arch
Quoted 14 lines · ebiggers at kernel.org
Hi Andy,
arch/riscv/Kconfig config RISCV_DMA_NONCOHERENT · line 269
Quoted 5 lines · Andy Chiu
arch/riscv/Kconfig +5
-0+config AS_HAS_OPTION_ARCH + # https://reviews.llvm.org/D123515 + def_bool y + depends on $(as-instr, .option arch$(comma) +m) + depends on !$(as-instr, .option arch$(comma) -i)With tip-of-tree clang (llvm-project commit
85a8e5c3e0586e85), I'm seeingAS_HAS_OPTION_ARCHbe set to n. It's the second "depends on" that makes it be set to n, so apparently clang started accepting ".option arch -i". What was your intent here for checking that ".option arch -i" is not supported? I'd think that just the first "depends on" would be sufficient.
I'm not sure what Andy's rationale was, but de3a913df6e ("RISC-V:
Clarify the behavior of .option arch directive.") in binutils-gdb stopped accepting .option arch, -i along with fixing a handful of other oddities in our .option arch handling.
If that's all this is testing for then we should probably add some sort of version check for old binutils (or maybe just ignore it, looks like it was a bugfix and the old version was never released).
+Nelson, as he probably knows better than I do.
That said: what does LLVM do if you ask it to turn the "I" base ISA off? I'd argue there's no instructions left at that point...
Re: [PATCH -next v21 23/27] riscv: detect assembler support for .option arch
Quoted 11 lines · On Sun, Jan 21, 2024 at 10:55?AM Palmer Dabbelt <[email protected]> wrote:
Quoted 9 lines · ebiggers at kernel.org
Hi Andy,
Quoted 5 lines · Andy Chiu
arch/riscv/Kconfig +5
-0+config AS_HAS_OPTION_ARCH + # https://reviews.llvm.org/D123515 + def_bool y + depends on $(as-instr, .option arch$(comma) +m) + depends on !$(as-instr, .option arch$(comma) -i)
Quoted 5 lines
Quoted 5 lines
With tip-of-tree clang (llvm-project commit
85a8e5c3e0586e85), I'm seeingAS_HAS_OPTION_ARCHbe set to n. It's the second "depends on" that makes it be set to n, so apparently clang started accepting ".option arch -i". What was your intent here for checking that ".option arch -i" is not supported? I'd think that just the first "depends on" would be sufficient.
The reason why I added the second check is because clang and gcc only return an assembler warning when it does not support ".option arch" at all. No errors were reported. So, it ended up passing the first condition check for old toolchains that shouldn't be passing.
Quoted 6 lines
I'm not sure what Andy's rationale was, but
de3a913df6e("RISC-V:
Clarify the behavior of .option arch directive.") in binutils-gdb stopped accepting.option arch, -ialong with fixing a handful of other oddities in our.option archhandling.
Quoted 8 lines
If that's all this is testing for then we should probably add some sort of version check for old binutils (or maybe just ignore it, looks like it was a bugfix and the old version was never released).
+Nelson, as he probably knows better than I do.
That said: what does LLVM do if you ask it to turn the "I" base ISA off? I'd argue there's no instructions left at that point...
Maybe what we really should do is to upgrade the condition check to a one liner shell script and grep if "Warning" is being printed. Sadly this warning is not failing the compilation with -Werror.
I can try forming a patch on this if it feels alright to people.
Thanks,
Andy
Re: [PATCH -next v21 23/27] riscv: detect assembler support for .option arch
Quoted 6 lines · Andy Chiu
Maybe what we really should do is to upgrade the condition check to a one liner shell script and grep if "Warning" is being printed. Sadly this warning is not failing the compilation with -Werror.
I can try forming a patch on this if it feels alright to people.
What about -Wa,--fatal-warnings ?
- Eric
Nathan Chancellor
maintainerRe: [PATCH -next v21 23/27] riscv: detect assembler support for .option arch
Quoted 9 lines · Eric Biggers
Quoted 6 lines · Andy Chiu
Maybe what we really should do is to upgrade the condition check to a one liner shell script and grep if "Warning" is being printed. Sadly this warning is not failing the compilation with -Werror.
I can try forming a patch on this if it feels alright to people.
What about -Wa,--fatal-warnings ?
I suspect that would work, the following diff appears to work for me with a version of clang that does and does not support '.option arch', (although I am not sure if adding -Wa,--fatal-warnings will have any other consequences):
arch/riscv/Kconfig +0 -1
@@ -315,7 +315,6 @@ config AS_HAS_OPTION_ARCH # https://reviews.llvm.org/D123515 def_bool y depends on $(as-instr, .option arch$(comma) +m) - depends on !$(as-instr, .option arch$(comma) -i) source "arch/riscv/Kconfig.socs" source "arch/riscv/Kconfig.errata"
scripts/Kconfig.include +1 -1
@@ -33,7 +33,7 @@ ld-option = $(success,$(LD) -v $(1)) # $(as-instr,<instr>) # Return y if the assembler supports <instr>, n otherwise -as-instr = $(success,printf "%b\n" "$(1)" | $(CC) $(CLANG_FLAGS) -c -x assembler-with-cpp -o /dev/null -) +as-instr = $(success,printf "%b\n" "$(1)" | $(CC) $(CLANG_FLAGS) -Wa$(comma)--fatal-warnings -c -x assembler-with-cpp -o /dev/null -) # check if $(CC) and $(LD) exist $(error-if,$(failure,command -v $(CC)),C compiler '$(CC)' not found)
Re: [PATCH -next v21 23/27] riscv: detect assembler support for .option arch
Quoted 33 lines · Nathan Chancellor
Quoted 9 lines · Eric Biggers
Quoted 6 lines · Andy Chiu
Maybe what we really should do is to upgrade the condition check to a one liner shell script and grep if "Warning" is being printed. Sadly this warning is not failing the compilation with -Werror.
I can try forming a patch on this if it feels alright to people.
What about -Wa,--fatal-warnings ?
I suspect that would work, the following diff appears to work for me with a version of clang that does and does not support '.option arch', (although I am not sure if adding -Wa,--fatal-warnings will have any other consequences):
arch/riscv/Kconfig +0
-1@@ -315,7 +315,6 @@ config AS_HAS_OPTION_ARCH # https://reviews.llvm.org/D123515 def_bool y depends on $(as-instr, .option arch$(comma) +m) - depends on !$(as-instr, .option arch$(comma) -i) source "arch/riscv/Kconfig.socs" source "arch/riscv/Kconfig.errata"scripts/Kconfig.include +0
-0
scripts/Kconfig.include ld-option = $(success,$(LD) -v $(1)) · line 39
Quoted 8 lines
scripts/Kconfig.include +1
-1@@ -33,7 +33,7 @@ ld-option = $(success,$(LD) -v $(1)) # $(as-instr,<instr>) # Return y if the assembler supports <instr>, n otherwise -as-instr = $(success,printf "%b\n" "$(1)" | $(CC) $(CLANG_FLAGS) -c -x assembler-with-cpp -o /dev/null -) +as-instr = $(success,printf "%b\n" "$(1)" | $(CC) $(CLANG_FLAGS) -Wa$(comma)--fatal-warnings -c -x assembler-with-cpp -o /dev/null -) # check if $(CC) and $(LD) exist $(error-if,$(failure,command -v $(CC)),C compiler '$(CC)' not found)
It looks good to me as long as none of the as-instr users turn out to have any expected warnings.
Maybe send the change to scripts/Kconfig.include as a separate patch?
- Eric
PATCH v21 24/27 riscv: Enable Vector code to be built
In reply to Andy ChiuFrom: Guo Ren <[email protected]>
This patch adds configs for building Vector code. First it detects the reqired toolchain support for building the code. Then it provides an option setting whether Vector is implicitly enabled to userspace.
---
Changelog v20:
- s/RISCV_V_DISABLE/RISCV_ISA_V_DEFAULT_ENABLE/ for better understanding (Conor)
- Update commit message (Conor)
Changelog V19:
- Add RISCV_V_DISABLE to set compile-time default.
--- arch/riscv/Kconfig | 31 +++++++++++++++++++++++++++++++ arch/riscv/Makefile | 6 +++++- 2 files changed, 36 insertions(+), 1 deletion(-)
arch/riscv/Kconfig +31 -0
@@ -466,6 +466,37 @@ config RISCV_ISA_SVPBMT If you don't know what to do here, say Y. +config TOOLCHAIN_HAS_V + bool + default y + depends on !64BIT || $(cc-option,-mabi=lp64 -march=rv64iv) + depends on !32BIT || $(cc-option,-mabi=ilp32 -march=rv32iv) + depends on LLD_VERSION >= 140000 || LD_VERSION >= 23800 + depends on AS_HAS_OPTION_ARCH + +config RISCV_ISA_V + bool "VECTOR extension support" + depends on TOOLCHAIN_HAS_V + depends on FPU + select DYNAMIC_SIGFRAME + default y + help + Say N here if you want to disable all vector related procedure + in the kernel. + + If you don't know what to do here, say Y. + +config RISCV_ISA_V_DEFAULT_ENABLE + bool "Enable userspace Vector by default" + depends on RISCV_ISA_V + default y + help + Say Y here if you want to enable Vector in userspace by default. + Otherwise, userspace has to make explicit prctl() call to enable + Vector, or enable it via the sysctl interface. + + If you don't know what to do here, say Y. + config TOOLCHAIN_HAS_ZBB bool default y
arch/riscv/Makefile +5 -1
@@ -60,6 +60,7 @@ riscv-march-$(CONFIG_ARCH_RV32I) := rv32ima riscv-march-$(CONFIG_ARCH_RV64I) := rv64ima riscv-march-$(CONFIG_FPU) := $(riscv-march-y)fd riscv-march-$(CONFIG_RISCV_ISA_C) := $(riscv-march-y)c +riscv-march-$(CONFIG_RISCV_ISA_V) := $(riscv-march-y)v ifdef CONFIG_TOOLCHAIN_NEEDS_OLD_ISA_SPEC KBUILD_CFLAGS += -Wa,-misa-spec=2.2
@@ -71,7 +72,10 @@ endif # Check if the toolchain supports Zihintpause extension riscv-march-$(CONFIG_TOOLCHAIN_HAS_ZIHINTPAUSE) := $(riscv-march-y)_zihintpause -KBUILD_CFLAGS += -march=$(subst fd,,$(riscv-march-y)) +# Remove F,D,V from isa string for all. Keep extensions between "fd" and "v" by +# matching non-v and non-multi-letter extensions out with the filter ([^v_]*) +KBUILD_CFLAGS += -march=$(shell echo $(riscv-march-y) | sed -E 's/(rv32ima|rv64ima)fd([^v_]*)v?/\1\2/') + KBUILD_AFLAGS += -march=$(riscv-march-y) KBUILD_CFLAGS += -mno-save-restore
Signature · 1 line
2.17.1
PATCH v21 25/27 riscv: Add documentation for Vector
In reply to Andy ChiuThis patch add a brief documentation of the userspace interface in regard to the RISC-V Vector extension.
---
Changelog v21:
- add usage guideline into doc (R?mi) Changelog v20:
- Drop bit-field repressentation and typos (Bj?rn)
- Fix document styling (Bagas)
--- Documentation/riscv/index.rst | 1 + Documentation/riscv/vector.rst | 132 +++++++++++++++++++++++++++++++++ 2 files changed, 133 insertions(+) create mode 100644 Documentation/riscv/vector.rst
Documentation/riscv/index.rst +1 -0
@@ -10,6 +10,7 @@ RISC-V architecture hwprobe patch-acceptance uabi + vector features
Documentation/riscv/vector.rst new file +132 -0
132 lines not on this page · raw message
Signature · 1 line
2.17.1
Re: [PATCH -next v21 25/27] riscv: Add documentation for Vector
Nits...
Quoted 53 lines · Le maanantaina 5. kesäkuuta 2023, 14.07.22 EEST Andy Chiu a écrit :
This patch add a brief documentation of the userspace interface in regard to the RISC-V Vector extension.
Signed-off-by Andy Chiu <[email protected]>Reviewed-by Greentime Hu <[email protected]>Reviewed-by Vincent Chen <[email protected]>Co-developed-by Bagas Sanjaya <[email protected]>Signed-off-by Bagas Sanjaya <[email protected]>---
Changelog v21:
- add usage guideline into doc (Rémi) Changelog v20:
- Drop bit-field repressentation and typos (Björn)
- Fix document styling (Bagas)--- Documentation/riscv/index.rst | 1 + Documentation/riscv/vector.rst | 132 +++++++++++++++++++++++++++++++++ 2 files changed, 133 insertions(+) create mode 100644 Documentation/riscv/vector.rstDocumentation/riscv/index.rst +1
-0@@ -10,6 +10,7 @@ RISC-V architecture hwprobe patch-acceptance uabi + vector featuresDocumentation/riscv/vector.rst new file +7
-0@@ -0,0 +1,132 @@ +.. SPDX-License-Identifier: GPL-2.0 + +========================================= +Vector Extension Support for RISC-V Linux +========================================= + +This document briefly outlines the interface provided to userspace by Linuxin +order to support the use of the RISC-V Vector Extension.
Documentation/riscv/vector.rst new file +7
-0+ +1. prctl() Interface +--------------------- + +Two new prctl() calls are added to allow programs to manage the enablement +status for the use of Vector in userspace. The intended usage guideline for +these interfaces is to give init systems a way to modify the availability
Quoted 1 line
of V +for processes running under its domain.
Nit: their domain
Quoted 1 line
Calling thess interfaces is
those interfaces
Quoted 2 lines
not +recommended in libraries routines because libraries should not override policies +configured from the parant process. Also, users must
parent
Quoted 1 line
noted that these interfaces
note
Quoted 2 lines
+are not portable to non-Linux, nor non-RISC-V environments, so it is discourage +to use in a portable code.
"so their use in portable code is discouraged."
Quoted 23 lines
To get the
availability of V in an ELF program, +please read :c:macro:COMPAT_HWCAP_ISA_Vbit of :c:macro:ELF_HWCAPin the +auxiliary vector.Documentation/riscv/vector.rst new file +4
-0+ +* prctl(PR_RISCV_V_SET_CONTROL, unsigned long arg) + + Sets the Vector enablement status of the calling thread, where thecontrol + argument consists of two 2-bit enablement statuses and a bit
for inheritance + mode. Other threads of the calling process are
unaffected.Documentation/riscv/vector.rst new file +5
-0+ + Enablement status is a tri-state value each occupying 2-bit of space in + the control argument: + + * :c:macro:`PR_RISCV_V_VSTATE_CTRL_DEFAULT`: Use the system-widedefault + enablement status on
execve(). The system-wide default
setting can be + controlled via sysctl interface (see sysctl section
below). +Documentation/riscv/vector.rst new file +4
-0+ * :c:macro:`PR_RISCV_V_VSTATE_CTRL_ON`: Allow Vector to be run for the + thread. + + * :c:macro:`PR_RISCV_V_VSTATE_CTRL_OFF`: Disallow Vector. Executing
Quoted 2 lines
Vector + instructions under such condition will trap and casuse the
termination of the thread.
cause
Quoted 94 lines
Documentation/riscv/vector.rst new file +8
-0+ + arg: The control argument is a 5-bit value consisting of 3 parts, and + accessed by 3 masks respectively. + + The 3 masks, PR_RISCV_V_VSTATE_CTRL_CUR_MASK, + PR_RISCV_V_VSTATE_CTRL_NEXT_MASK, and PR_RISCV_V_VSTATE_CTRL_INHERIT + represents bit[1:0], bit[3:2], and bit[4]. bit[1:0] accounts for the + enablement status of current thread, and the setting at bit[3:2] takesplace + at next
execve(). bit[4] defines the inheritance mode of the
setting in + bit[3:2].Documentation/riscv/vector.rst new file +2
-0+ + * :c:macro:`PR_RISCV_V_VSTATE_CTRL_CUR_MASK`: bit[1:0]: Account forthe + Vector enablement status for the calling thread. The calling
thread is + not able to turn off Vector once it has been enabled.
Theprctl()call + fails with EPERM if the value in this mask isPR_RISCV_V_VSTATE_CTRL_OFF+ but the current enablement status is
not off. Setting
+PR_RISCV_V_VSTATE_CTRL_DEFAULThere takes no effect but to set
back + the original enablement status.Documentation/riscv/vector.rst new file +2
-0+ + * :c:macro:`PR_RISCV_V_VSTATE_CTRL_NEXT_MASK`: bit[3:2]: Accountfor the + Vector enablement setting for the calling thread at the
nextexecve()+ system call. IfPR_RISCV_V_VSTATE_CTRL_DEFAULTis
used in this mask, + then the enablement status will be decided by
the system-wide + enablement status whenexecve()happen.Documentation/riscv/vector.rst new file +2
-0+ + * :c:macro:`PR_RISCV_V_VSTATE_CTRL_INHERIT`: bit[4]: theinheritance + mode for the setting at
PR_RISCV_V_VSTATE_CTRL_NEXT_MASK. If the bit + is set then the
followingexecve()will not clear the setting in both +PR_RISCV_V_VSTATE_CTRL_NEXT_MASKandPR_RISCV_V_VSTATE_CTRL_INHERIT. +
This setting persists across changes in the system-wide default value.Documentation/riscv/vector.rst new file +4
-0+ + Return value: + * 0 on success; + * EINVAL: Vector not supported, invalid enablement status forcurrent or + next mask;
+ * EPERM: Turning off Vector inPR_RISCV_V_VSTATE_CTRL_CUR_MASKif
Vector + was enabled for the calling thread.Documentation/riscv/vector.rst new file +5
-0+ + On success: + * A valid setting for PR_RISCV_V_VSTATE_CTRL_CUR_MASK takes place + immediately. The enablement status specified in + PR_RISCV_V_VSTATE_CTRL_NEXT_MASK happens at the next execve()call, or + all following
execve()calls ifPR_RISCV_V_VSTATE_CTRL_INHERITbit is + set.
+ * Every successful call overwrites a previous setting for the
calling + thread.Documentation/riscv/vector.rst new file +4
-0+ +* prctl(PR_RISCV_V_GET_CONTROL) + + Gets the same Vector enablement status for the calling thread. Settingfor + next
execve()call and the inheritance bit are all OR-ed together.Documentation/riscv/vector.rst new file +2
-0+ + Note that ELF programs are able to get the availability of V for itselfby + reading :c:macro:
COMPAT_HWCAP_ISA_Vbit of :c:macro:ELF_HWCAP
in the + auxiliary vector.Documentation/riscv/vector.rst new file +17
-0+ + Return value: + * a nonnegative value on success; + * EINVAL: Vector not supported. + +2. System runtime configuration (sysctl) +----------------------------------------- + +To mitigate the ABI impact of expansion of the signal stack, a +policy mechanism is provided to the administrators, distro maintainers, and +developers to control the default Vector enablement status for userspace +processes in form of sysctl knob: + +* /proc/sys/abi/riscv_v_default_allow + + Writing the text representation of 0 or 1 to this file sets the default + system enablement status for new starting userspace programs. Validvalues + are:
Documentation/riscv/vector.rst new file +2
-0+ + * 0: Do not allow Vector code to be executed as the default for newprocesses. + * 1: Allow Vector code to be executed as the default for
new processes. +Documentation/riscv/vector.rst new file +3
-0+ Reading this file returns the current system default enablement status. + + At every execve() call, a new enablement status of the new process isset to + the system default, unless:
Documentation/riscv/vector.rst new file +2
-0+ + * PR_RISCV_V_VSTATE_CTRL_INHERIT is set for the calling process, andthe + setting in
PR_RISCV_V_VSTATE_CTRL_NEXT_MASKis notDocumentation/riscv/vector.rst new file +6
-0+ PR_RISCV_V_VSTATE_CTRL_DEFAULT. Or, + + * The setting in PR_RISCV_V_VSTATE_CTRL_NEXT_MASK is not + PR_RISCV_V_VSTATE_CTRL_DEFAULT. + + Modifying the system default enablement status does not affect theenablement + status of any existing process of thread that do not make
anexecve()call.
Signature · 10 lines
レミ・デニ-クールモン http://www.remlab.net/ _______________________________________________ linux-riscv mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-riscv
PATCH v21 26/27 selftests: Test RISC-V Vector prctl interface
In reply to Andy ChiuThis add a test for prctl interface that controls the use of userspace Vector.
--- tools/testing/selftests/riscv/Makefile | 2 +- .../testing/selftests/riscv/vector/.gitignore | 2 + tools/testing/selftests/riscv/vector/Makefile | 15 ++ .../riscv/vector/vstate_exec_nolibc.c | 111 ++++++++++ .../selftests/riscv/vector/vstate_prctl.c | 189 ++++++++++++++++++ 5 files changed, 318 insertions(+), 1 deletion(-) create mode 100644 tools/testing/selftests/riscv/vector/.gitignore create mode 100644 tools/testing/selftests/riscv/vector/Makefile create mode 100644 tools/testing/selftests/riscv/vector/vstate_exec_nolibc.c create mode 100644 tools/testing/selftests/riscv/vector/vstate_prctl.c
tools/testing/selftests/riscv/Makefile +1 -1
@@ -5,7 +5,7 @@ ARCH ?= $(shell uname -m 2>/dev/null || echo not) ifneq (,$(filter $(ARCH),riscv)) -RISCV_SUBTARGETS ?= hwprobe +RISCV_SUBTARGETS ?= hwprobe vector else RISCV_SUBTARGETS := endif
tools/testing/selftests/riscv/vector/.gitignore new file +2 -0
@@ -0,0 +1,2 @@ +vstate_exec_nolibc +vstate_prctl
tools/testing/selftests/riscv/vector/Makefile new file +15 -0
@@ -0,0 +1,15 @@ +# SPDX-License-Identifier: GPL-2.0 +# Copyright (C) 2021 ARM Limited +# Originally tools/testing/arm64/abi/Makefile + +TEST_GEN_PROGS := vstate_prctl +TEST_GEN_PROGS_EXTENDED := vstate_exec_nolibc + +include ../../lib.mk + +$(OUTPUT)/vstate_prctl: vstate_prctl.c ../hwprobe/sys_hwprobe.S + $(CC) -static -o$@ $(CFLAGS) $(LDFLAGS) $^ + +$(OUTPUT)/vstate_exec_nolibc: vstate_exec_nolibc.c + $(CC) -nostdlib -static -include ../../../../include/nolibc/nolibc.h \ + -Wall $(CFLAGS) $(LDFLAGS) $^ -o $@ -lgcc
tools/testing/selftests/riscv/vector/vstate_exec_nolibc.c new file +111 -0
111 lines not on this page · raw message
tools/testing/selftests/riscv/vector/vstate_prctl.c new file +189 -0
189 lines not on this page · raw message
Signature · 1 line
2.17.1
PATCH v21 27/27 selftests: add .gitignore file for RISC-V hwprobe
In reply to Andy ChiuThe executable file "hwprobe" should be ignored by git, adding it to fix that.
--- tools/testing/selftests/riscv/hwprobe/.gitignore | 1 + 1 file changed, 1 insertion(+) create mode 100644 tools/testing/selftests/riscv/hwprobe/.gitignore
tools/testing/selftests/riscv/hwprobe/.gitignore new file +1 -0
@@ -0,0 +1 @@ +hwprobe
Signature · 1 line
2.17.1
Palmer Dabbelt
AppliedQuoted 9 lines · Andy Chiu
This is the v21 patch series for adding Vector extension support in Linux. Please refer to [1] for the introduction of the patchset. The v21 patch series was aimed to solve build issues from v19, provide usage guideline for the prctl interface, and address review comments on v20.
Thank every one who has been reviewing, suggesting on the topic. Hope this get a step closer to the final merge.
[...]
Applied, thanks!
[01/27] riscv: Rename __switch_to_aux() -> fpu
https://git.kernel.org/palmer/c/419d5d38ac5d [02/27] riscv: Extending cpufeature.c to detect V-extension https://git.kernel.org/palmer/c/dc6667a4e7e3 [03/27] riscv: hwprobe: Add support for probing V in RISCV_HWPROBE_KEY_IMA_EXT_0 https://git.kernel.org/palmer/c/162e4df137c1 [04/27] riscv: Add new csr defines related to vector extension https://git.kernel.org/palmer/c/b5665d2a9432 [05/27] riscv: Clear vector regfile on bootup https://git.kernel.org/palmer/c/6b533828726a [06/27] riscv: Disable Vector Instructions for kernel itself https://git.kernel.org/palmer/c/74abe5a39d3a [07/27] riscv: Introduce Vector enable/disable helpers https://git.kernel.org/palmer/c/0a3381a01dcc [08/27] riscv: Introduce riscv_v_vsize to record size of Vector context https://git.kernel.org/palmer/c/7017858eb2d7 [09/27] riscv: Introduce struct/helpers to save/restore per-task Vector state https://git.kernel.org/palmer/c/03c3fcd9941a [10/27] riscv: Add task switch support for vector https://git.kernel.org/palmer/c/3a2df6323def [11/27] riscv: Allocate user's vector context in the first-use trap https://git.kernel.org/palmer/c/cd054837243b [12/27] riscv: Add ptrace vector support https://git.kernel.org/palmer/c/0c59922c769a [13/27] riscv: signal: check fp-reserved words unconditionally https://git.kernel.org/palmer/c/a45cedaa1ac0 [14/27] riscv: signal: Add sigcontext save/restore for vector https://git.kernel.org/palmer/c/8ee0b41898fa [15/27] riscv: signal: Report signal frame size to userspace via auxv https://git.kernel.org/palmer/c/e92f469b0771 [16/27] riscv: signal: validate altstack to reflect Vector https://git.kernel.org/palmer/c/76e22fdc2c26 [17/27] riscv: prevent stack corruption by reserving task_pt_regs(p) early https://git.kernel.org/palmer/c/c7cdd96eca28 [18/27] riscv: kvm: Add V extension to KVM ISA https://git.kernel.org/palmer/c/bf78f1ea6e51 [19/27] riscv: KVM: Add vector lazy save/restore support https://git.kernel.org/palmer/c/0f4b82579716 [20/27] riscv: hwcap: change ELF_HWCAP to a function https://git.kernel.org/palmer/c/50724efcb370 [21/27] riscv: Add prctl controls for userspace vector management https://git.kernel.org/palmer/c/1fd96a3e9d5d [22/27] riscv: Add sysctl to set the default vector rule for new processes https://git.kernel.org/palmer/c/7ca7a7b9b635 [23/27] riscv: detect assembler support for .option arch https://git.kernel.org/palmer/c/e4bb020f3dbb [24/27] riscv: Enable Vector code to be built https://git.kernel.org/palmer/c/fa8e7cce55da [25/27] riscv: Add documentation for Vector https://git.kernel.org/palmer/c/04a4722eeede [26/27] selftests: Test RISC-V Vector prctl interface https://git.kernel.org/palmer/c/7cf6198ce22d [27/27] selftests: add .gitignore file for RISC-V hwprobe https://git.kernel.org/palmer/c/1e72695137ef
Best regards,
Signature · 1 line
Palmer Dabbelt <[email protected]>
This series was applied to riscv/linux.git (for-next) by Palmer Dabbelt <[email protected]>:
Hello:
This series was applied to riscv/linux.git (for-next) by Palmer Dabbelt <[email protected]>:
Quoted 9 lines · On Mon, 5 Jun 2023 11:06:57 +0000 you wrote:
This is the v21 patch series for adding Vector extension support in Linux. Please refer to [1] for the introduction of the patchset. The v21 patch series was aimed to solve build issues from v19, provide usage guideline for the prctl interface, and address review comments on v20.
Thank every one who has been reviewing, suggesting on the topic. Hope this get a step closer to the final merge.
[...]
Here is the summary with links:
- [-next,v21,01/27] riscv: Rename __switch_to_aux() -> fpu
https://git.kernel.org/riscv/c/419d5d38ac5d - [-next,v21,02/27] riscv: Extending cpufeature.c to detect V-extension https://git.kernel.org/riscv/c/dc6667a4e7e3 - [-next,v21,03/27] riscv: hwprobe: Add support for probing V in RISCV_HWPROBE_KEY_IMA_EXT_0 https://git.kernel.org/riscv/c/162e4df137c1 - [-next,v21,04/27] riscv: Add new csr defines related to vector extension https://git.kernel.org/riscv/c/b5665d2a9432 - [-next,v21,05/27] riscv: Clear vector regfile on bootup https://git.kernel.org/riscv/c/6b533828726a - [-next,v21,06/27] riscv: Disable Vector Instructions for kernel itself https://git.kernel.org/riscv/c/74abe5a39d3a - [-next,v21,07/27] riscv: Introduce Vector enable/disable helpers https://git.kernel.org/riscv/c/0a3381a01dcc - [-next,v21,08/27] riscv: Introduce riscv_v_vsize to record size of Vector context https://git.kernel.org/riscv/c/7017858eb2d7 - [-next,v21,09/27] riscv: Introduce struct/helpers to save/restore per-task Vector state https://git.kernel.org/riscv/c/03c3fcd9941a - [-next,v21,10/27] riscv: Add task switch support for vector https://git.kernel.org/riscv/c/3a2df6323def - [-next,v21,11/27] riscv: Allocate user's vector context in the first-use trap https://git.kernel.org/riscv/c/cd054837243b - [-next,v21,12/27] riscv: Add ptrace vector support https://git.kernel.org/riscv/c/0c59922c769a - [-next,v21,13/27] riscv: signal: check fp-reserved words unconditionally https://git.kernel.org/riscv/c/a45cedaa1ac0 - [-next,v21,14/27] riscv: signal: Add sigcontext save/restore for vector https://git.kernel.org/riscv/c/8ee0b41898fa - [-next,v21,15/27] riscv: signal: Report signal frame size to userspace via auxv https://git.kernel.org/riscv/c/e92f469b0771 - [-next,v21,16/27] riscv: signal: validate altstack to reflect Vector https://git.kernel.org/riscv/c/76e22fdc2c26 - [-next,v21,17/27] riscv: prevent stack corruption by reserving task_pt_regs(p) early https://git.kernel.org/riscv/c/c7cdd96eca28 - [-next,v21,18/27] riscv: kvm: Add V extension to KVM ISA https://git.kernel.org/riscv/c/bf78f1ea6e51 - [-next,v21,19/27] riscv: KVM: Add vector lazy save/restore support https://git.kernel.org/riscv/c/0f4b82579716 - [-next,v21,20/27] riscv: hwcap: change ELF_HWCAP to a function https://git.kernel.org/riscv/c/50724efcb370 - [-next,v21,21/27] riscv: Add prctl controls for userspace vector management https://git.kernel.org/riscv/c/1fd96a3e9d5d - [-next,v21,22/27] riscv: Add sysctl to set the default vector rule for new processes https://git.kernel.org/riscv/c/7ca7a7b9b635 - [-next,v21,23/27] riscv: detect assembler support for .option arch https://git.kernel.org/riscv/c/e4bb020f3dbb - [-next,v21,24/27] riscv: Enable Vector code to be built https://git.kernel.org/riscv/c/fa8e7cce55da - [-next,v21,25/27] riscv: Add documentation for Vector https://git.kernel.org/riscv/c/04a4722eeede - [-next,v21,26/27] selftests: Test RISC-V Vector prctl interface https://git.kernel.org/riscv/c/7cf6198ce22d - [-next,v21,27/27] selftests: add .gitignore file for RISC-V hwprobe https://git.kernel.org/riscv/c/1e72695137ef
You are awesome, thank you!
Signature · 2 lines
Deet-doot-dot, I am a bot. https://korg.docs.kernel.org/patchwork/pwbot.html