Currently the sve-ptrace test for setting and reading FPSIMD data assumes that the child will start off in FPSIMD only mode and that it can use this to read some FPSIMD mode SVE ptrace data, skipping the test if it can't. This isn't an assumption guaranteed by the ABI and also limits how we can use this testcase within the program. Instead skip the initial read and just generate a FPSIMD format buffer for the write part of the test, making the coverage more robust in the face of future kernel and test program changes.

Signed-off-by: Mark Brown broonie@kernel.org Reviewed-by: Shuah Khan skhan@linuxfoundation.org --- tools/testing/selftests/arm64/fp/sve-ptrace.c | 39 ++++++++----------- 1 file changed, 16 insertions(+), 23 deletions(-)

diff --git a/tools/testing/selftests/arm64/fp/sve-ptrace.c b/tools/testing/selftests/arm64/fp/sve-ptrace.c index 7682798adbba..8f6146d89ca4 100644 --- a/tools/testing/selftests/arm64/fp/sve-ptrace.c +++ b/tools/testing/selftests/arm64/fp/sve-ptrace.c @@ -46,7 +46,7 @@ static const struct vec_type vec_types[] = {

#define VL_TESTS (((SVE_VQ_MAX - SVE_VQ_MIN) + 1) * 3) #define FLAG_TESTS 2 -#define FPSIMD_TESTS 3 +#define FPSIMD_TESTS 2

#define EXPECTED_TESTS ((VL_TESTS + FLAG_TESTS + FPSIMD_TESTS) * ARRAY_SIZE(vec_types))

@@ -240,28 +240,24 @@ static void check_u32(unsigned int vl, const char *reg, /* Access the FPSIMD registers via the SVE regset */ static void ptrace_sve_fpsimd(pid_t child, const struct vec_type *type) { - void *svebuf = NULL; - size_t svebufsz = 0; + void *svebuf; struct user_sve_header *sve; struct user_fpsimd_state *fpsimd, new_fpsimd; unsigned int i, j; unsigned char *p; + int ret;

- /* New process should start with FPSIMD registers only */ - sve = get_sve(child, type, &svebuf, &svebufsz); - if (!sve) { - ksft_test_result_fail("get_sve(%s): %s\n", - type->name, strerror(errno)); - + svebuf = malloc(SVE_PT_SIZE(0, SVE_PT_REGS_FPSIMD)); + if (!svebuf) { + ksft_test_result_fail("Failed to allocate FPSIMD buffer\n"); return; - } else { - ksft_test_result_pass("get_sve(%s FPSIMD)\n", type->name); }

- ksft_test_result((sve->flags & SVE_PT_REGS_MASK) == SVE_PT_REGS_FPSIMD, - "Got FPSIMD registers via %s\n", type->name); - if ((sve->flags & SVE_PT_REGS_MASK) != SVE_PT_REGS_FPSIMD) - goto out; + memset(svebuf, 0, SVE_PT_SIZE(0, SVE_PT_REGS_FPSIMD)); + sve = svebuf; + sve->flags = SVE_PT_REGS_FPSIMD; + sve->size = SVE_PT_SIZE(0, SVE_PT_REGS_FPSIMD); + sve->vl = 16; /* We don't care what the VL is */

/* Try to set a known FPSIMD state via PT_REGS_SVE */ fpsimd = (struct user_fpsimd_state *)((char *)sve + @@ -273,12 +269,11 @@ static void ptrace_sve_fpsimd(pid_t child, const struct vec_type *type) p[j] = j; }

- if (set_sve(child, type, sve)) { - ksft_test_result_fail("set_sve(%s FPSIMD): %s\n", - type->name, strerror(errno)); - + ret = set_sve(child, type, sve); + ksft_test_result(ret == 0, "%s FPSIMD set via SVE: %d\n", + type->name, ret); + if (ret) goto out; - }

/* Verify via the FPSIMD regset */ if (get_fpsimd(child, &new_fpsimd)) { @@ -548,11 +543,9 @@ static int do_parent(pid_t child) if (getauxval(vec_types[i].hwcap_type) & vec_types[i].hwcap) { ptrace_sve_fpsimd(child, &vec_types[i]); } else { - ksft_test_result_skip("%s FPSIMD get via SVE\n", - vec_types[i].name); ksft_test_result_skip("%s FPSIMD set via SVE\n", vec_types[i].name); - ksft_test_result_skip("%s set read via FPSIMD\n", + ksft_test_result_skip("%s FPSIMD read\n", vec_types[i].name); }

Provide ABI documentation for SME similar to that for SVE. Due to the very large overlap around streaming SVE mode in both implementation and interfaces documentation for streaming mode SVE is added to the SVE document rather than the SME one.

Signed-off-by: Mark Brown broonie@kernel.org Reviewed-by: Catalin Marinas catalin.marinas@arm.com Reviewed-by: Szabolcs Nagy szabolcs.nagy@arm.com --- Documentation/arm64/index.rst | 1 + Documentation/arm64/sme.rst | 428 ++++++++++++++++++++++++++++++++++ Documentation/arm64/sve.rst | 70 +++++- 3 files changed, 489 insertions(+), 10 deletions(-) create mode 100644 Documentation/arm64/sme.rst

diff --git a/Documentation/arm64/index.rst b/Documentation/arm64/index.rst index 4f840bac083e..ae21f8118830 100644 --- a/Documentation/arm64/index.rst +++ b/Documentation/arm64/index.rst @@ -21,6 +21,7 @@ ARM64 Architecture perf pointer-authentication silicon-errata + sme sve tagged-address-abi tagged-pointers diff --git a/Documentation/arm64/sme.rst b/Documentation/arm64/sme.rst new file mode 100644 index 000000000000..8ba677b87e90 --- /dev/null +++ b/Documentation/arm64/sme.rst @@ -0,0 +1,428 @@ +=================================================== +Scalable Matrix Extension support for AArch64 Linux +=================================================== + +This document outlines briefly the interface provided to userspace by Linux in +order to support use of the ARM Scalable Matrix Extension (SME). + +This is an outline of the most important features and issues only and not +intended to be exhaustive. It should be read in conjunction with the SVE +documentation in sve.rst which provides details on the Streaming SVE mode +included in SME. + +This document does not aim to describe the SME architecture or programmer's +model. To aid understanding, a minimal description of relevant programmer's +model features for SME is included in Appendix A. + + +1. General +----------- + +* PSTATE.SM, PSTATE.ZA, the streaming mode vector length, the ZA + register state and TPIDR2_EL0 are tracked per thread. + +* The presence of SME is reported to userspace via HWCAP2_SME in the aux vector + AT_HWCAP2 entry. Presence of this flag implies the presence of the SME + instructions and registers, and the Linux-specific system interfaces + described in this document. SME is reported in /proc/cpuinfo as "sme". + +* Support for the execution of SME instructions in userspace can also be + detected by reading the CPU ID register ID_AA64PFR1_EL1 using an MRS + instruction, and checking that the value of the SME field is nonzero. [3] + + It does not guarantee the presence of the system interfaces described in the + following sections: software that needs to verify that those interfaces are + present must check for HWCAP2_SME instead. + +* There are a number of optional SME features, presence of these is reported + through AT_HWCAP2 through: + + HWCAP2_SME_I16I64 + HWCAP2_SME_F64F64 + HWCAP2_SME_I8I32 + HWCAP2_SME_F16F32 + HWCAP2_SME_B16F32 + HWCAP2_SME_F32F32 + HWCAP2_SME_FA64 + + This list may be extended over time as the SME architecture evolves. + + These extensions are also reported via the CPU ID register ID_AA64SMFR0_EL1, + which userspace can read using an MRS instruction. See elf_hwcaps.txt and + cpu-feature-registers.txt for details. + +* Debuggers should restrict themselves to interacting with the target via the + NT_ARM_SVE, NT_ARM_SSVE and NT_ARM_ZA regsets. The recommended way + of detecting support for these regsets is to connect to a target process + first and then attempt a + + ptrace(PTRACE_GETREGSET, pid, NT_ARM_<regset>, &iov). + +* Whenever ZA register values are exchanged in memory between userspace and + the kernel, the register value is encoded in memory as a series of horizontal + vectors from 0 to VL/8-1 stored in the same endianness invariant format as is + used for SVE vectors. + +* On thread creation TPIDR2_EL0 is preserved unless CLONE_SETTLS is specified, + in which case it is set to 0. + +2. Vector lengths +------------------ + +SME defines a second vector length similar to the SVE vector length which is +controls the size of the streaming mode SVE vectors and the ZA matrix array. +The ZA matrix is square with each side having as many bytes as a streaming +mode SVE vector. + + +3. Sharing of streaming and non-streaming mode SVE state +--------------------------------------------------------- + +It is implementation defined which if any parts of the SVE state are shared +between streaming and non-streaming modes. When switching between modes +via software interfaces such as ptrace if no register content is provided as +part of switching no state will be assumed to be shared and everything will +be zeroed. + + +4. System call behaviour +------------------------- + +* On syscall PSTATE.ZA is preserved, if PSTATE.ZA==1 then the contents of the + ZA matrix are preserved. + +* On syscall PSTATE.SM will be cleared and the SVE registers will be handled + as per the standard SVE ABI. + +* Neither the SVE registers nor ZA are used to pass arguments to or receive + results from any syscall. + +* On process creation (eg, clone()) the newly created process will have + PSTATE.SM cleared. + +* All other SME state of a thread, including the currently configured vector + length, the state of the PR_SME_VL_INHERIT flag, and the deferred vector + length (if any), is preserved across all syscalls, subject to the specific + exceptions for execve() described in section 6. + + +5. Signal handling +------------------- + +* Signal handlers are invoked with streaming mode and ZA disabled. + +* A new signal frame record za_context encodes the ZA register contents on + signal delivery. [1] + +* The signal frame record for ZA always contains basic metadata, in particular + the thread's vector length (in za_context.vl). + +* The ZA matrix may or may not be included in the record, depending on + the value of PSTATE.ZA. The registers are present if and only if: + za_context.head.size >= ZA_SIG_CONTEXT_SIZE(sve_vq_from_vl(za_context.vl)) + in which case PSTATE.ZA == 1. + +* If matrix data is present, the remainder of the record has a vl-dependent + size and layout. Macros ZA_SIG_* are defined [1] to facilitate access to + them. + +* The matrix is stored as a series of horizontal vectors in the same format as + is used for SVE vectors. + +* If the ZA context is too big to fit in sigcontext.__reserved[], then extra + space is allocated on the stack, an extra_context record is written in + __reserved[] referencing this space. za_context is then written in the + extra space. Refer to [1] for further details about this mechanism. + + +5. Signal return +----------------- + +When returning from a signal handler: + +* If there is no za_context record in the signal frame, or if the record is + present but contains no register data as described in the previous section, + then ZA is disabled. + +* If za_context is present in the signal frame and contains matrix data then + PSTATE.ZA is set to 1 and ZA is populated with the specified data. + +* The vector length cannot be changed via signal return. If za_context.vl in + the signal frame does not match the current vector length, the signal return + attempt is treated as illegal, resulting in a forced SIGSEGV. + + +6. prctl extensions +-------------------- + +Some new prctl() calls are added to allow programs to manage the SME vector +length: + +prctl(PR_SME_SET_VL, unsigned long arg) + + Sets the vector length of the calling thread and related flags, where + arg == vl | flags. Other threads of the calling process are unaffected. + + vl is the desired vector length, where sve_vl_valid(vl) must be true. + + flags: + + PR_SME_VL_INHERIT + + Inherit the current vector length across execve(). Otherwise, the + vector length is reset to the system default at execve(). (See + Section 9.) + + PR_SME_SET_VL_ONEXEC + + Defer the requested vector length change until the next execve() + performed by this thread. + + The effect is equivalent to implicit execution of the following + call immediately after the next execve() (if any) by the thread: + + prctl(PR_SME_SET_VL, arg & ~PR_SME_SET_VL_ONEXEC) + + This allows launching of a new program with a different vector + length, while avoiding runtime side effects in the caller. + + Without PR_SME_SET_VL_ONEXEC, the requested change takes effect + immediately. + + + Return value: a nonnegative on success, or a negative value on error: + EINVAL: SME not supported, invalid vector length requested, or + invalid flags. + + + On success: + + * Either the calling thread's vector length or the deferred vector length + to be applied at the next execve() by the thread (dependent on whether + PR_SME_SET_VL_ONEXEC is present in arg), is set to the largest value + supported by the system that is less than or equal to vl. If vl == + SVE_VL_MAX, the value set will be the largest value supported by the + system. + + * Any previously outstanding deferred vector length change in the calling + thread is cancelled. + + * The returned value describes the resulting configuration, encoded as for + PR_SME_GET_VL. The vector length reported in this value is the new + current vector length for this thread if PR_SME_SET_VL_ONEXEC was not + present in arg; otherwise, the reported vector length is the deferred + vector length that will be applied at the next execve() by the calling + thread. + + * Changing the vector length causes all of ZA, P0..P15, FFR and all bits of + Z0..Z31 except for Z0 bits [127:0] .. Z31 bits [127:0] to become + unspecified, including both streaming and non-streaming SVE state. + Calling PR_SME_SET_VL with vl equal to the thread's current vector + length, or calling PR_SME_SET_VL with the PR_SVE_SET_VL_ONEXEC flag, + does not constitute a change to the vector length for this purpose. + + * Changing the vector length causes PSTATE.ZA and PSTATE.SM to be cleared. + Calling PR_SME_SET_VL with vl equal to the thread's current vector + length, or calling PR_SME_SET_VL with the PR_SVE_SET_VL_ONEXEC flag, + does not constitute a change to the vector length for this purpose. + + +prctl(PR_SME_GET_VL) + + Gets the vector length of the calling thread. + + The following flag may be OR-ed into the result: + + PR_SME_VL_INHERIT + + Vector length will be inherited across execve(). + + There is no way to determine whether there is an outstanding deferred + vector length change (which would only normally be the case between a + fork() or vfork() and the corresponding execve() in typical use). + + To extract the vector length from the result, bitwise and it with + PR_SME_VL_LEN_MASK. + + Return value: a nonnegative value on success, or a negative value on error: + EINVAL: SME not supported. + + +7. ptrace extensions +--------------------- + +* A new regset NT_ARM_SSVE is defined for access to streaming mode SVE + state via PTRACE_GETREGSET and PTRACE_SETREGSET, this is documented in + sve.rst. + +* A new regset NT_ARM_ZA is defined for ZA state for access to ZA state via + PTRACE_GETREGSET and PTRACE_SETREGSET. + + Refer to [2] for definitions. + +The regset data starts with struct user_za_header, containing: + + size + + Size of the complete regset, in bytes. + This depends on vl and possibly on other things in the future. + + If a call to PTRACE_GETREGSET requests less data than the value of + size, the caller can allocate a larger buffer and retry in order to + read the complete regset. + + max_size + + Maximum size in bytes that the regset can grow to for the target + thread. The regset won't grow bigger than this even if the target + thread changes its vector length etc. + + vl + + Target thread's current streaming vector length, in bytes. + + max_vl + + Maximum possible streaming vector length for the target thread. + + flags + + Zero or more of the following flags, which have the same + meaning and behaviour as the corresponding PR_SET_VL_* flags: + + SME_PT_VL_INHERIT + + SME_PT_VL_ONEXEC (SETREGSET only). + +* The effects of changing the vector length and/or flags are equivalent to + those documented for PR_SME_SET_VL. + + The caller must make a further GETREGSET call if it needs to know what VL is + actually set by SETREGSET, unless is it known in advance that the requested + VL is supported. + +* The size and layout of the payload depends on the header fields. The + SME_PT_ZA_*() macros are provided to facilitate access to the data. + +* In either case, for SETREGSET it is permissible to omit the payload, in which + case the vector length and flags are changed and PSTATE.ZA is set to 0 + (along with any consequences of those changes). If a payload is provided + then PSTATE.ZA will be set to 1. + +* For SETREGSET, if the requested VL is not supported, the effect will be the + same as if the payload were omitted, except that an EIO error is reported. + No attempt is made to translate the payload data to the correct layout + for the vector length actually set. It is up to the caller to translate the + payload layout for the actual VL and retry. + +* The effect of writing a partial, incomplete payload is unspecified. + + +8. ELF coredump extensions +--------------------------- + +* NT_ARM_SSVE notes will be added to each coredump for + each thread of the dumped process. The contents will be equivalent to the + data that would have been read if a PTRACE_GETREGSET of the corresponding + type were executed for each thread when the coredump was generated. + +* A NT_ARM_ZA note will be added to each coredump for each thread of the + dumped process. The contents will be equivalent to the data that would have + been read if a PTRACE_GETREGSET of NT_ARM_ZA were executed for each thread + when the coredump was generated. + + +9. System runtime configuration +-------------------------------- + +* To mitigate the ABI impact of expansion of the signal frame, a policy + mechanism is provided for administrators, distro maintainers and developers + to set the default vector length for userspace processes: + +/proc/sys/abi/sme_default_vector_length + + Writing the text representation of an integer to this file sets the system + default vector length to the specified value, unless the value is greater + than the maximum vector length supported by the system in which case the + default vector length is set to that maximum. + + The result can be determined by reopening the file and reading its + contents. + + At boot, the default vector length is initially set to 32 or the maximum + supported vector length, whichever is smaller and supported. This + determines the initial vector length of the init process (PID 1). + + Reading this file returns the current system default vector length. + +* At every execve() call, the new vector length of the new process is set to + the system default vector length, unless + + * PR_SME_VL_INHERIT (or equivalently SME_PT_VL_INHERIT) is set for the + calling thread, or + + * a deferred vector length change is pending, established via the + PR_SME_SET_VL_ONEXEC flag (or SME_PT_VL_ONEXEC). + +* Modifying the system default vector length does not affect the vector length + of any existing process or thread that does not make an execve() call. + + +Appendix A. SME programmer's model (informative) +================================================= + +This section provides a minimal description of the additions made by SVE to the +ARMv8-A programmer's model that are relevant to this document. + +Note: This section is for information only and not intended to be complete or +to replace any architectural specification. + +A.1. Registers +--------------- + +In A64 state, SME adds the following: + +* A new mode, streaming mode, in which a subset of the normal FPSIMD and SVE + features are available. When supported EL0 software may enter and leave + streaming mode at any time. + + For best system performance it is strongly encouraged for software to enable + streaming mode only when it is actively being used. + +* A new vector length controlling the size of ZA and the Z registers when in + streaming mode, separately to the vector length used for SVE when not in + streaming mode. There is no requirement that either the currently selected + vector length or the set of vector lengths supported for the two modes in + a given system have any relationship. The streaming mode vector length + is referred to as SVL. + +* A new ZA matrix register. This is a square matrix of SVLxSVL bits. Most + operations on ZA require that streaming mode be enabled but ZA can be + enabled without streaming mode in order to load, save and retain data. + + For best system performance it is strongly encouraged for software to enable + ZA only when it is actively being used. + +* Two new 1 bit fields in PSTATE which may be controlled via the SMSTART and + SMSTOP instructions or by access to the SVCR system register: + + * PSTATE.ZA, if this is 1 then the ZA matrix is accessible and has valid + data while if it is 0 then ZA can not be accessed. When PSTATE.ZA is + changed from 0 to 1 all bits in ZA are cleared. + + * PSTATE.SM, if this is 1 then the PE is in streaming mode. When the value + of PSTATE.SM is changed then it is implementation defined if the subset + of the floating point register bits valid in both modes may be retained. + Any other bits will be cleared. + + +References +========== + +[1] arch/arm64/include/uapi/asm/sigcontext.h + AArch64 Linux signal ABI definitions + +[2] arch/arm64/include/uapi/asm/ptrace.h + AArch64 Linux ptrace ABI definitions + +[3] Documentation/arm64/cpu-feature-registers.rst diff --git a/Documentation/arm64/sve.rst b/Documentation/arm64/sve.rst index 9d9a4de5bc34..93c2c2990584 100644 --- a/Documentation/arm64/sve.rst +++ b/Documentation/arm64/sve.rst @@ -7,7 +7,9 @@ Author: Dave Martin Dave.Martin@arm.com Date: 4 August 2017

This document outlines briefly the interface provided to userspace by Linux in -order to support use of the ARM Scalable Vector Extension (SVE). +order to support use of the ARM Scalable Vector Extension (SVE), including +interactions with Streaming SVE mode added by the Scalable Matrix Extension +(SME).

This is an outline of the most important features and issues only and not intended to be exhaustive. @@ -23,6 +25,10 @@ model features for SVE is included in Appendix A. * SVE registers Z0..Z31, P0..P15 and FFR and the current vector length VL, are tracked per-thread.

+* In streaming mode FFR is not accessible unless HWCAP2_SME_FA64 is present + in the system, when it is not supported and these interfaces are used to + access streaming mode FFR is read and written as zero. + * The presence of SVE is reported to userspace via HWCAP_SVE in the aux vector AT_HWCAP entry. Presence of this flag implies the presence of the SVE instructions and registers, and the Linux-specific system interfaces @@ -53,10 +59,19 @@ model features for SVE is included in Appendix A. which userspace can read using an MRS instruction. See elf_hwcaps.txt and cpu-feature-registers.txt for details.

+* On hardware that supports the SME extensions, HWCAP2_SME will also be + reported in the AT_HWCAP2 aux vector entry. Among other things SME adds + streaming mode which provides a subset of the SVE feature set using a + separate SME vector length and the same Z/V registers. See sme.rst + for more details. + * Debuggers should restrict themselves to interacting with the target via the NT_ARM_SVE regset. The recommended way of detecting support for this regset is to connect to a target process first and then attempt a - ptrace(PTRACE_GETREGSET, pid, NT_ARM_SVE, &iov). + ptrace(PTRACE_GETREGSET, pid, NT_ARM_SVE, &iov). Note that when SME is + present and streaming SVE mode is in use the FPSIMD subset of registers + will be read via NT_ARM_SVE and NT_ARM_SVE writes will exit streaming mode + in the target.

* Whenever SVE scalable register values (Zn, Pn, FFR) are exchanged in memory between userspace and the kernel, the register value is encoded in memory in @@ -126,6 +141,11 @@ the SVE instruction set architecture. are only present in fpsimd_context. For convenience, the content of V0..V31 is duplicated between sve_context and fpsimd_context.

+* The record contains a flag field which includes a flag SVE_SIG_FLAG_SM which + if set indicates that the thread is in streaming mode and the vector length + and register data (if present) describe the streaming SVE data and vector + length. + * The signal frame record for SVE always contains basic metadata, in particular the thread's vector length (in sve_context.vl).

@@ -170,6 +190,11 @@ When returning from a signal handler: the signal frame does not match the current vector length, the signal return attempt is treated as illegal, resulting in a forced SIGSEGV.

+* It is permitted to enter or leave streaming mode by setting or clearing + the SVE_SIG_FLAG_SM flag but applications should take care to ensure that + when doing so sve_context.vl and any register data are appropriate for the + vector length in the new mode. +

6. prctl extensions -------------------- @@ -265,8 +290,14 @@ prctl(PR_SVE_GET_VL) 7. ptrace extensions ---------------------

-* A new regset NT_ARM_SVE is defined for use with PTRACE_GETREGSET and - PTRACE_SETREGSET. +* New regsets NT_ARM_SVE and NT_ARM_SSVE are defined for use with + PTRACE_GETREGSET and PTRACE_SETREGSET. NT_ARM_SSVE describes the + streaming mode SVE registers and NT_ARM_SVE describes the + non-streaming mode SVE registers. + + In this description a register set is referred to as being "live" when + the target is in the appropriate streaming or non-streaming mode and is + using data beyond the subset shared with the FPSIMD Vn registers.

Refer to [2] for definitions.

@@ -297,7 +328,7 @@ The regset data starts with struct user_sve_header, containing:

flags

- either + at most one of

SVE_PT_REGS_FPSIMD

@@ -331,6 +362,10 @@ The regset data starts with struct user_sve_header, containing:

SVE_PT_VL_ONEXEC (SETREGSET only).

+ If neither FPSIMD nor SVE flags are provided then no register + payload is available, this is only possible when SME is implemented. + + * The effects of changing the vector length and/or flags are equivalent to those documented for PR_SVE_SET_VL.

@@ -346,6 +381,13 @@ The regset data starts with struct user_sve_header, containing: case only the vector length and flags are changed (along with any consequences of those changes).

+* In systems supporting SME when in streaming mode a GETREGSET for + NT_REG_SVE will return only the user_sve_header with no register data, + similarly a GETREGSET for NT_REG_SSVE will not return any register data + when not in streaming mode. + +* A GETREGSET for NT_ARM_SSVE will never return SVE_PT_REGS_FPSIMD. + * For SETREGSET, if an SVE_PT_REGS_SVE payload is present and the requested VL is not supported, the effect will be the same as if the payload were omitted, except that an EIO error is reported. No @@ -355,17 +397,25 @@ The regset data starts with struct user_sve_header, containing: unspecified. It is up to the caller to translate the payload layout for the actual VL and retry.

+* Where SME is implemented it is not possible to GETREGSET the register + state for normal SVE when in streaming mode, nor the streaming mode + register state when in normal mode, regardless of the implementation defined + behaviour of the hardware for sharing data between the two modes. + +* Any SETREGSET of NT_ARM_SVE will exit streaming mode if the target was in + streaming mode and any SETREGSET of NT_ARM_SSVE will enter streaming mode + if the target was not in streaming mode. + * The effect of writing a partial, incomplete payload is unspecified.

8. ELF coredump extensions ---------------------------

-* A NT_ARM_SVE note will be added to each coredump for each thread of the - dumped process. The contents will be equivalent to the data that would have - been read if a PTRACE_GETREGSET of NT_ARM_SVE were executed for each thread - when the coredump was generated. - +* NT_ARM_SVE and NT_ARM_SSVE notes will be added to each coredump for + each thread of the dumped process. The contents will be equivalent to the + data that would have been read if a PTRACE_GETREGSET of the corresponding + type were executed for each thread when the coredump was generated.

9. System runtime configuration --------------------------------

As with SVE rather than impose ambitious toolchain requirements for SME we manually encode the few instructions which we require in order to perform the work the kernel needs to do. The instructions used to save and restore context are provided as assembler macros while those for entering and leaving streaming mode are done in asm volatile blocks since they are expected to be used from C.

We could do the SMSTART and SMSTOP operations with read/modify/write cycles on SVCR but using the aliases provided for individual field accesses should be slightly faster. These instructions are aliases for MSR but since our minimum toolchain requirements are old enough to mean that we can't use the sX_X_cX_cX_X form and they always use xzr rather than taking a value like write_sysreg_s() wants we just use .inst.

Signed-off-by: Mark Brown broonie@kernel.org Reviewed-by: Catalin Marinas catalin.marinas@arm.com --- arch/arm64/include/asm/fpsimd.h | 25 +++++++++++++ arch/arm64/include/asm/fpsimdmacros.h | 54 +++++++++++++++++++++++++++ 2 files changed, 79 insertions(+)

diff --git a/arch/arm64/include/asm/fpsimd.h b/arch/arm64/include/asm/fpsimd.h index cb24385e3632..6e2dc9dcbf49 100644 --- a/arch/arm64/include/asm/fpsimd.h +++ b/arch/arm64/include/asm/fpsimd.h @@ -249,6 +249,31 @@ static inline void sve_setup(void) { }

#endif /* ! CONFIG_ARM64_SVE */

+#ifdef CONFIG_ARM64_SME + +static inline void sme_smstart_sm(void) +{ + asm volatile(__msr_s(SYS_SVCR_SMSTART_SM_EL0, "xzr")); +} + +static inline void sme_smstop_sm(void) +{ + asm volatile(__msr_s(SYS_SVCR_SMSTOP_SM_EL0, "xzr")); +} + +static inline void sme_smstop(void) +{ + asm volatile(__msr_s(SYS_SVCR_SMSTOP_SMZA_EL0, "xzr")); +} + +#else + +static inline void sme_smstart_sm(void) { } +static inline void sme_smstop_sm(void) { } +static inline void sme_smstop(void) { } + +#endif /* ! CONFIG_ARM64_SME */ + /* For use by EFI runtime services calls only */ extern void __efi_fpsimd_begin(void); extern void __efi_fpsimd_end(void); diff --git a/arch/arm64/include/asm/fpsimdmacros.h b/arch/arm64/include/asm/fpsimdmacros.h index 2509d7dde55a..2e9a33155081 100644 --- a/arch/arm64/include/asm/fpsimdmacros.h +++ b/arch/arm64/include/asm/fpsimdmacros.h @@ -93,6 +93,12 @@ .endif .endm

+.macro _sme_check_wv v + .if (\v) < 12 || (\v) > 15 + .error "Bad vector select register \v." + .endif +.endm + /* SVE instruction encodings for non-SVE-capable assemblers */ /* (pre binutils 2.28, all kernel capable clang versions support SVE) */

@@ -174,6 +180,54 @@ | (\np) .endm

+/* SME instruction encodings for non-SME-capable assemblers */ +/* (pre binutils 2.38/LLVM 13) */ + +/* RDSVL X\nx, #\imm */ +.macro _sme_rdsvl nx, imm + _check_general_reg \nx + _check_num (\imm), -0x20, 0x1f + .inst 0x04bf5800 \ + | (\nx) \ + | (((\imm) & 0x3f) << 5) +.endm + +/* + * STR (vector from ZA array): + * STR ZA[\nw, #\offset], [X\nxbase, #\offset, MUL VL] + */ +.macro _sme_str_zav nw, nxbase, offset=0 + _sme_check_wv \nw + _check_general_reg \nxbase + _check_num (\offset), -0x100, 0xff + .inst 0xe1200000 \ + | (((\nw) & 3) << 13) \ + | ((\nxbase) << 5) \ + | ((\offset) & 7) +.endm + +/* + * LDR (vector to ZA array): + * LDR ZA[\nw, #\offset], [X\nxbase, #\offset, MUL VL] + */ +.macro _sme_ldr_zav nw, nxbase, offset=0 + _sme_check_wv \nw + _check_general_reg \nxbase + _check_num (\offset), -0x100, 0xff + .inst 0xe1000000 \ + | (((\nw) & 3) << 13) \ + | ((\nxbase) << 5) \ + | ((\offset) & 7) +.endm + +/* + * Zero the entire ZA array + * ZERO ZA + */ +.macro zero_za + .inst 0xc00800ff +.endm + .macro __for from:req, to:req .if (\from) == (\to) _for__body %\from

This patch introduces basic cpufeature support for discovering the presence of the Scalable Matrix Extension.

Signed-off-by: Mark Brown broonie@kernel.org Reviewed-by: Catalin Marinas catalin.marinas@arm.com --- Documentation/arm64/elf_hwcaps.rst | 33 ++++++++++++++++ arch/arm64/include/asm/cpu.h | 1 + arch/arm64/include/asm/cpufeature.h | 12 ++++++ arch/arm64/include/asm/fpsimd.h | 2 + arch/arm64/include/asm/hwcap.h | 8 ++++ arch/arm64/include/uapi/asm/hwcap.h | 8 ++++ arch/arm64/kernel/cpufeature.c | 59 +++++++++++++++++++++++++++++ arch/arm64/kernel/cpuinfo.c | 9 +++++ arch/arm64/kernel/fpsimd.c | 30 +++++++++++++++ arch/arm64/tools/cpucaps | 2 + 10 files changed, 164 insertions(+)

diff --git a/Documentation/arm64/elf_hwcaps.rst b/Documentation/arm64/elf_hwcaps.rst index a8f30963e550..f8d818eaaff5 100644 --- a/Documentation/arm64/elf_hwcaps.rst +++ b/Documentation/arm64/elf_hwcaps.rst @@ -264,6 +264,39 @@ HWCAP2_MTE3 Functionality implied by ID_AA64PFR1_EL1.MTE == 0b0011, as described by Documentation/arm64/memory-tagging-extension.rst.

+HWCAP2_SME + + Functionality implied by ID_AA64PFR1_EL1.SME == 0b0001, as described + by Documentation/arm64/sme.rst. + +HWCAP2_SME_I16I64 + + Functionality implied by ID_AA64SMFR0_EL1.I16I64 == 0b1111. + +HWCAP2_SME_F64F64 + + Functionality implied by ID_AA64SMFR0_EL1.F64F64 == 0b1. + +HWCAP2_SME_I8I32 + + Functionality implied by ID_AA64SMFR0_EL1.I8I32 == 0b1111. + +HWCAP2_SME_F16F32 + + Functionality implied by ID_AA64SMFR0_EL1.F16F32 == 0b1. + +HWCAP2_SME_B16F32 + + Functionality implied by ID_AA64SMFR0_EL1.B16F32 == 0b1. + +HWCAP2_SME_F32F32 + + Functionality implied by ID_AA64SMFR0_EL1.F32F32 == 0b1. + +HWCAP2_SME_FA64 + + Functionality implied by ID_AA64SMFR0_EL1.FA64 == 0b1. + 4. Unused AT_HWCAP bits -----------------------

diff --git a/arch/arm64/include/asm/cpu.h b/arch/arm64/include/asm/cpu.h index a58e366f0b07..d08062bcb9c1 100644 --- a/arch/arm64/include/asm/cpu.h +++ b/arch/arm64/include/asm/cpu.h @@ -58,6 +58,7 @@ struct cpuinfo_arm64 { u64 reg_id_aa64pfr0; u64 reg_id_aa64pfr1; u64 reg_id_aa64zfr0; + u64 reg_id_aa64smfr0;

struct cpuinfo_32bit aarch32;

diff --git a/arch/arm64/include/asm/cpufeature.h b/arch/arm64/include/asm/cpufeature.h index c62e7e5e2f0c..8ac12e4094aa 100644 --- a/arch/arm64/include/asm/cpufeature.h +++ b/arch/arm64/include/asm/cpufeature.h @@ -759,6 +759,18 @@ static __always_inline bool system_supports_sve(void) cpus_have_const_cap(ARM64_SVE); }

+static __always_inline bool system_supports_sme(void) +{ + return IS_ENABLED(CONFIG_ARM64_SME) && + cpus_have_const_cap(ARM64_SME); +} + +static __always_inline bool system_supports_fa64(void) +{ + return IS_ENABLED(CONFIG_ARM64_SME) && + cpus_have_const_cap(ARM64_SME_FA64); +} + static __always_inline bool system_supports_cnp(void) { return IS_ENABLED(CONFIG_ARM64_CNP) && diff --git a/arch/arm64/include/asm/fpsimd.h b/arch/arm64/include/asm/fpsimd.h index 6e2dc9dcbf49..2e8ef00e7520 100644 --- a/arch/arm64/include/asm/fpsimd.h +++ b/arch/arm64/include/asm/fpsimd.h @@ -74,6 +74,8 @@ extern void sve_set_vq(unsigned long vq_minus_1);

struct arm64_cpu_capabilities; extern void sve_kernel_enable(const struct arm64_cpu_capabilities *__unused); +extern void sme_kernel_enable(const struct arm64_cpu_capabilities *__unused); +extern void fa64_kernel_enable(const struct arm64_cpu_capabilities *__unused);

extern u64 read_zcr_features(void);

diff --git a/arch/arm64/include/asm/hwcap.h b/arch/arm64/include/asm/hwcap.h index 8db5ec0089db..9f0ce004fdbc 100644 --- a/arch/arm64/include/asm/hwcap.h +++ b/arch/arm64/include/asm/hwcap.h @@ -109,6 +109,14 @@ #define KERNEL_HWCAP_AFP __khwcap2_feature(AFP) #define KERNEL_HWCAP_RPRES __khwcap2_feature(RPRES) #define KERNEL_HWCAP_MTE3 __khwcap2_feature(MTE3) +#define KERNEL_HWCAP_SME __khwcap2_feature(SME) +#define KERNEL_HWCAP_SME_I16I64 __khwcap2_feature(SME_I16I64) +#define KERNEL_HWCAP_SME_F64F64 __khwcap2_feature(SME_F64F64) +#define KERNEL_HWCAP_SME_I8I32 __khwcap2_feature(SME_I8I32) +#define KERNEL_HWCAP_SME_F16F32 __khwcap2_feature(SME_F16F32) +#define KERNEL_HWCAP_SME_B16F32 __khwcap2_feature(SME_B16F32) +#define KERNEL_HWCAP_SME_F32F32 __khwcap2_feature(SME_F32F32) +#define KERNEL_HWCAP_SME_FA64 __khwcap2_feature(SME_FA64)

/* * This yields a mask that user programs can use to figure out what diff --git a/arch/arm64/include/uapi/asm/hwcap.h b/arch/arm64/include/uapi/asm/hwcap.h index 99cb5d383048..b0256cec63b5 100644 --- a/arch/arm64/include/uapi/asm/hwcap.h +++ b/arch/arm64/include/uapi/asm/hwcap.h @@ -79,5 +79,13 @@ #define HWCAP2_AFP (1 << 20) #define HWCAP2_RPRES (1 << 21) #define HWCAP2_MTE3 (1 << 22) +#define HWCAP2_SME (1 << 23) +#define HWCAP2_SME_I16I64 (1 << 24) +#define HWCAP2_SME_F64F64 (1 << 25) +#define HWCAP2_SME_I8I32 (1 << 26) +#define HWCAP2_SME_F16F32 (1 << 27) +#define HWCAP2_SME_B16F32 (1 << 28) +#define HWCAP2_SME_F32F32 (1 << 29) +#define HWCAP2_SME_FA64 (1 << 30)

#endif /* _UAPI__ASM_HWCAP_H */ diff --git a/arch/arm64/kernel/cpufeature.c b/arch/arm64/kernel/cpufeature.c index d72c4b4d389c..0f2d7ddd69ae 100644 --- a/arch/arm64/kernel/cpufeature.c +++ b/arch/arm64/kernel/cpufeature.c @@ -261,6 +261,8 @@ static const struct arm64_ftr_bits ftr_id_aa64pfr0[] = { };

static const struct arm64_ftr_bits ftr_id_aa64pfr1[] = { + ARM64_FTR_BITS(FTR_VISIBLE_IF_IS_ENABLED(CONFIG_ARM64_SME), + FTR_STRICT, FTR_LOWER_SAFE, ID_AA64PFR1_SME_SHIFT, 4, 0), ARM64_FTR_BITS(FTR_HIDDEN, FTR_STRICT, FTR_LOWER_SAFE, ID_AA64PFR1_MPAMFRAC_SHIFT, 4, 0), ARM64_FTR_BITS(FTR_HIDDEN, FTR_STRICT, FTR_LOWER_SAFE, ID_AA64PFR1_RASFRAC_SHIFT, 4, 0), ARM64_FTR_BITS(FTR_VISIBLE_IF_IS_ENABLED(CONFIG_ARM64_MTE), @@ -293,6 +295,24 @@ static const struct arm64_ftr_bits ftr_id_aa64zfr0[] = { ARM64_FTR_END, };

+static const struct arm64_ftr_bits ftr_id_aa64smfr0[] = { + ARM64_FTR_BITS(FTR_VISIBLE_IF_IS_ENABLED(CONFIG_ARM64_SME), + FTR_STRICT, FTR_EXACT, ID_AA64SMFR0_FA64_SHIFT, 1, 0), + ARM64_FTR_BITS(FTR_VISIBLE_IF_IS_ENABLED(CONFIG_ARM64_SME), + FTR_STRICT, FTR_EXACT, ID_AA64SMFR0_I16I64_SHIFT, 4, 0), + ARM64_FTR_BITS(FTR_VISIBLE_IF_IS_ENABLED(CONFIG_ARM64_SME), + FTR_STRICT, FTR_EXACT, ID_AA64SMFR0_F64F64_SHIFT, 1, 0), + ARM64_FTR_BITS(FTR_VISIBLE_IF_IS_ENABLED(CONFIG_ARM64_SME), + FTR_STRICT, FTR_EXACT, ID_AA64SMFR0_I8I32_SHIFT, 4, 0), + ARM64_FTR_BITS(FTR_VISIBLE_IF_IS_ENABLED(CONFIG_ARM64_SME), + FTR_STRICT, FTR_EXACT, ID_AA64SMFR0_F16F32_SHIFT, 1, 0), + ARM64_FTR_BITS(FTR_VISIBLE_IF_IS_ENABLED(CONFIG_ARM64_SME), + FTR_STRICT, FTR_EXACT, ID_AA64SMFR0_B16F32_SHIFT, 1, 0), + ARM64_FTR_BITS(FTR_VISIBLE_IF_IS_ENABLED(CONFIG_ARM64_SME), + FTR_STRICT, FTR_EXACT, ID_AA64SMFR0_F32F32_SHIFT, 1, 0), + ARM64_FTR_END, +}; + static const struct arm64_ftr_bits ftr_id_aa64mmfr0[] = { ARM64_FTR_BITS(FTR_VISIBLE, FTR_STRICT, FTR_LOWER_SAFE, ID_AA64MMFR0_ECV_SHIFT, 4, 0), ARM64_FTR_BITS(FTR_HIDDEN, FTR_STRICT, FTR_LOWER_SAFE, ID_AA64MMFR0_FGT_SHIFT, 4, 0), @@ -645,6 +665,7 @@ static const struct __ftr_reg_entry { ARM64_FTR_REG_OVERRIDE(SYS_ID_AA64PFR1_EL1, ftr_id_aa64pfr1, &id_aa64pfr1_override), ARM64_FTR_REG(SYS_ID_AA64ZFR0_EL1, ftr_id_aa64zfr0), + ARM64_FTR_REG(SYS_ID_AA64SMFR0_EL1, ftr_id_aa64smfr0),

/* Op1 = 0, CRn = 0, CRm = 5 */ ARM64_FTR_REG(SYS_ID_AA64DFR0_EL1, ftr_id_aa64dfr0), @@ -960,6 +981,7 @@ void __init init_cpu_features(struct cpuinfo_arm64 *info) init_cpu_ftr_reg(SYS_ID_AA64PFR0_EL1, info->reg_id_aa64pfr0); init_cpu_ftr_reg(SYS_ID_AA64PFR1_EL1, info->reg_id_aa64pfr1); init_cpu_ftr_reg(SYS_ID_AA64ZFR0_EL1, info->reg_id_aa64zfr0); + init_cpu_ftr_reg(SYS_ID_AA64SMFR0_EL1, info->reg_id_aa64smfr0);

if (id_aa64pfr0_32bit_el0(info->reg_id_aa64pfr0)) init_32bit_cpu_features(&info->aarch32); @@ -2442,6 +2464,33 @@ static const struct arm64_cpu_capabilities arm64_features[] = { .matches = has_cpuid_feature, .min_field_value = 1, }, +#ifdef CONFIG_ARM64_SME + { + .desc = "Scalable Matrix Extension", + .type = ARM64_CPUCAP_SYSTEM_FEATURE, + .capability = ARM64_SME, + .sys_reg = SYS_ID_AA64PFR1_EL1, + .sign = FTR_UNSIGNED, + .field_pos = ID_AA64PFR1_SME_SHIFT, + .field_width = 4, + .min_field_value = ID_AA64PFR1_SME, + .matches = has_cpuid_feature, + .cpu_enable = sme_kernel_enable, + }, + /* FA64 should be sorted after the base SME capability */ + { + .desc = "FA64", + .type = ARM64_CPUCAP_SYSTEM_FEATURE, + .capability = ARM64_SME_FA64, + .sys_reg = SYS_ID_AA64SMFR0_EL1, + .sign = FTR_UNSIGNED, + .field_pos = ID_AA64SMFR0_FA64_SHIFT, + .field_width = 1, + .min_field_value = ID_AA64SMFR0_FA64, + .matches = has_cpuid_feature, + .cpu_enable = fa64_kernel_enable, + }, +#endif /* CONFIG_ARM64_SME */ {}, };

@@ -2575,6 +2624,16 @@ static const struct arm64_cpu_capabilities arm64_elf_hwcaps[] = { HWCAP_CAP(SYS_ID_AA64MMFR0_EL1, ID_AA64MMFR0_ECV_SHIFT, 4, FTR_UNSIGNED, 1, CAP_HWCAP, KERNEL_HWCAP_ECV), HWCAP_CAP(SYS_ID_AA64MMFR1_EL1, ID_AA64MMFR1_AFP_SHIFT, 4, FTR_UNSIGNED, 1, CAP_HWCAP, KERNEL_HWCAP_AFP), HWCAP_CAP(SYS_ID_AA64ISAR2_EL1, ID_AA64ISAR2_RPRES_SHIFT, 4, FTR_UNSIGNED, 1, CAP_HWCAP, KERNEL_HWCAP_RPRES), +#ifdef CONFIG_ARM64_SME + HWCAP_CAP(SYS_ID_AA64PFR1_EL1, ID_AA64PFR1_SME_SHIFT, 4, FTR_UNSIGNED, ID_AA64PFR1_SME, CAP_HWCAP, KERNEL_HWCAP_SME), + HWCAP_CAP(SYS_ID_AA64SMFR0_EL1, ID_AA64SMFR0_FA64_SHIFT, 1, FTR_UNSIGNED, ID_AA64SMFR0_FA64, CAP_HWCAP, KERNEL_HWCAP_SME_FA64), + HWCAP_CAP(SYS_ID_AA64SMFR0_EL1, ID_AA64SMFR0_I16I64_SHIFT, 4, FTR_UNSIGNED, ID_AA64SMFR0_I16I64, CAP_HWCAP, KERNEL_HWCAP_SME_I16I64), + HWCAP_CAP(SYS_ID_AA64SMFR0_EL1, ID_AA64SMFR0_F64F64_SHIFT, 1, FTR_UNSIGNED, ID_AA64SMFR0_F64F64, CAP_HWCAP, KERNEL_HWCAP_SME_F64F64), + HWCAP_CAP(SYS_ID_AA64SMFR0_EL1, ID_AA64SMFR0_I8I32_SHIFT, 4, FTR_UNSIGNED, ID_AA64SMFR0_I8I32, CAP_HWCAP, KERNEL_HWCAP_SME_I8I32), + HWCAP_CAP(SYS_ID_AA64SMFR0_EL1, ID_AA64SMFR0_F16F32_SHIFT, 1, FTR_UNSIGNED, ID_AA64SMFR0_F16F32, CAP_HWCAP, KERNEL_HWCAP_SME_F16F32), + HWCAP_CAP(SYS_ID_AA64SMFR0_EL1, ID_AA64SMFR0_B16F32_SHIFT, 1, FTR_UNSIGNED, ID_AA64SMFR0_B16F32, CAP_HWCAP, KERNEL_HWCAP_SME_B16F32), + HWCAP_CAP(SYS_ID_AA64SMFR0_EL1, ID_AA64SMFR0_F32F32_SHIFT, 1, FTR_UNSIGNED, ID_AA64SMFR0_F32F32, CAP_HWCAP, KERNEL_HWCAP_SME_F32F32), +#endif /* CONFIG_ARM64_SME */ {}, };

diff --git a/arch/arm64/kernel/cpuinfo.c b/arch/arm64/kernel/cpuinfo.c index 330b92ea863a..a73fe2888b7e 100644 --- a/arch/arm64/kernel/cpuinfo.c +++ b/arch/arm64/kernel/cpuinfo.c @@ -98,6 +98,14 @@ static const char *const hwcap_str[] = { [KERNEL_HWCAP_AFP] = "afp", [KERNEL_HWCAP_RPRES] = "rpres", [KERNEL_HWCAP_MTE3] = "mte3", + [KERNEL_HWCAP_SME] = "sme", + [KERNEL_HWCAP_SME_I16I64] = "smei16i64", + [KERNEL_HWCAP_SME_F64F64] = "smef64f64", + [KERNEL_HWCAP_SME_I8I32] = "smei8i32", + [KERNEL_HWCAP_SME_F16F32] = "smef16f32", + [KERNEL_HWCAP_SME_B16F32] = "smeb16f32", + [KERNEL_HWCAP_SME_F32F32] = "smef32f32", + [KERNEL_HWCAP_SME_FA64] = "smefa64", };

#ifdef CONFIG_COMPAT @@ -401,6 +409,7 @@ static void __cpuinfo_store_cpu(struct cpuinfo_arm64 *info) info->reg_id_aa64pfr0 = read_cpuid(ID_AA64PFR0_EL1); info->reg_id_aa64pfr1 = read_cpuid(ID_AA64PFR1_EL1); info->reg_id_aa64zfr0 = read_cpuid(ID_AA64ZFR0_EL1); + info->reg_id_aa64smfr0 = read_cpuid(ID_AA64SMFR0_EL1);

if (id_aa64pfr1_mte(info->reg_id_aa64pfr1)) info->reg_gmid = read_cpuid(GMID_EL1); diff --git a/arch/arm64/kernel/fpsimd.c b/arch/arm64/kernel/fpsimd.c index 47af76e53221..e4fba0bfb55e 100644 --- a/arch/arm64/kernel/fpsimd.c +++ b/arch/arm64/kernel/fpsimd.c @@ -993,6 +993,32 @@ void fpsimd_release_task(struct task_struct *dead_task)

#endif /* CONFIG_ARM64_SVE */

+#ifdef CONFIG_ARM64_SME + +void sme_kernel_enable(const struct arm64_cpu_capabilities *__always_unused p) +{ + /* Set priority for all PEs to architecturally defined minimum */ + write_sysreg_s(read_sysreg_s(SYS_SMPRI_EL1) & ~SMPRI_EL1_PRIORITY_MASK, + SYS_SMPRI_EL1); + + /* Allow SME in kernel */ + write_sysreg(read_sysreg(CPACR_EL1) | CPACR_EL1_SMEN_EL1EN, CPACR_EL1); + isb(); +} + +/* + * This must be called after sme_kernel_enable(), we rely on the + * feature table being sorted to ensure this. + */ +void fa64_kernel_enable(const struct arm64_cpu_capabilities *__always_unused p) +{ + /* Allow use of FA64 */ + write_sysreg_s(read_sysreg_s(SYS_SMCR_EL1) | SMCR_ELx_FA64_MASK, + SYS_SMCR_EL1); +} + +#endif /* CONFIG_ARM64_SVE */ + /* * Trapped SVE access * @@ -1538,6 +1564,10 @@ static int __init fpsimd_init(void) if (!cpu_have_named_feature(ASIMD)) pr_notice("Advanced SIMD is not implemented\n");

+ + if (cpu_have_named_feature(SME) && !cpu_have_named_feature(SVE)) + pr_notice("SME is implemented but not SVE\n"); + return sve_sysctl_init(); } core_initcall(fpsimd_init); diff --git a/arch/arm64/tools/cpucaps b/arch/arm64/tools/cpucaps index 3ed418f70e3b..e52b289a27c2 100644 --- a/arch/arm64/tools/cpucaps +++ b/arch/arm64/tools/cpucaps @@ -43,6 +43,8 @@ KVM_PROTECTED_MODE MISMATCHED_CACHE_TYPE MTE MTE_ASYMM +SME +SME_FA64 SPECTRE_V2 SPECTRE_V3A SPECTRE_V4

As for SVE provide a sysctl which allows the default SME vector length to be configured.

Signed-off-by: Mark Brown broonie@kernel.org Reviewed-by: Catalin Marinas catalin.marinas@arm.com --- arch/arm64/kernel/fpsimd.c | 29 ++++++++++++++++++++++++++++- 1 file changed, 28 insertions(+), 1 deletion(-)

diff --git a/arch/arm64/kernel/fpsimd.c b/arch/arm64/kernel/fpsimd.c index 5e5fbd9cba75..754a96563f6f 100644 --- a/arch/arm64/kernel/fpsimd.c +++ b/arch/arm64/kernel/fpsimd.c @@ -489,6 +489,30 @@ static int __init sve_sysctl_init(void) static int __init sve_sysctl_init(void) { return 0; } #endif /* ! (CONFIG_ARM64_SVE && CONFIG_SYSCTL) */

+#if defined(CONFIG_ARM64_SME) && defined(CONFIG_SYSCTL) +static struct ctl_table sme_default_vl_table[] = { + { + .procname = "sme_default_vector_length", + .mode = 0644, + .proc_handler = vec_proc_do_default_vl, + .extra1 = &vl_info[ARM64_VEC_SME], + }, + { } +}; + +static int __init sme_sysctl_init(void) +{ + if (system_supports_sme()) + if (!register_sysctl("abi", sme_default_vl_table)) + return -EINVAL; + + return 0; +} + +#else /* ! (CONFIG_ARM64_SME && CONFIG_SYSCTL) */ +static int __init sme_sysctl_init(void) { return 0; } +#endif /* ! (CONFIG_ARM64_SME && CONFIG_SYSCTL) */ + #define ZREG(sve_state, vq, n) ((char *)(sve_state) + \ (SVE_SIG_ZREG_OFFSET(vq, n) - SVE_SIG_REGS_OFFSET))

@@ -1687,6 +1711,9 @@ static int __init fpsimd_init(void) if (cpu_have_named_feature(SME) && !cpu_have_named_feature(SVE)) pr_notice("SME is implemented but not SVE\n");

- return sve_sysctl_init(); + sve_sysctl_init(); + sme_sysctl_init(); + + return 0; } core_initcall(fpsimd_init);

The Scalable Matrix Extension introduces support for a new thread specific data register TPIDR2 intended for use by libc. The kernel must save the value of TPIDR2 on context switch and should ensure that all new threads start off with a default value of 0. Add a field to the thread_struct to store TPIDR2 and context switch it with the other thread specific data.

In case there are future extensions which also use TPIDR2 we introduce system_supports_tpidr2() and use that rather than system_supports_sme() for TPIDR2 handling.

Signed-off-by: Mark Brown broonie@kernel.org Reviewed-by: Catalin Marinas catalin.marinas@arm.com --- arch/arm64/include/asm/cpufeature.h | 5 +++++ arch/arm64/include/asm/processor.h | 1 + arch/arm64/kernel/fpsimd.c | 4 ++++ arch/arm64/kernel/process.c | 14 ++++++++++++-- 4 files changed, 22 insertions(+), 2 deletions(-)

diff --git a/arch/arm64/include/asm/cpufeature.h b/arch/arm64/include/asm/cpufeature.h index 5ddfae233ea5..14a8f3d93add 100644 --- a/arch/arm64/include/asm/cpufeature.h +++ b/arch/arm64/include/asm/cpufeature.h @@ -778,6 +778,11 @@ static __always_inline bool system_supports_fa64(void) cpus_have_const_cap(ARM64_SME_FA64); }

+static __always_inline bool system_supports_tpidr2(void) +{ + return system_supports_sme(); +} + static __always_inline bool system_supports_cnp(void) { return IS_ENABLED(CONFIG_ARM64_CNP) && diff --git a/arch/arm64/include/asm/processor.h b/arch/arm64/include/asm/processor.h index 7a57cbff8a03..849e97d418a8 100644 --- a/arch/arm64/include/asm/processor.h +++ b/arch/arm64/include/asm/processor.h @@ -169,6 +169,7 @@ struct thread_struct { u64 mte_ctrl; #endif u64 sctlr_user; + u64 tpidr2_el0; };

static inline unsigned int thread_get_vl(struct thread_struct *thread, diff --git a/arch/arm64/kernel/fpsimd.c b/arch/arm64/kernel/fpsimd.c index 39f44fcb9b99..231f2d85b65e 100644 --- a/arch/arm64/kernel/fpsimd.c +++ b/arch/arm64/kernel/fpsimd.c @@ -1098,6 +1098,10 @@ void sme_kernel_enable(const struct arm64_cpu_capabilities *__always_unused p) /* Allow SME in kernel */ write_sysreg(read_sysreg(CPACR_EL1) | CPACR_EL1_SMEN_EL1EN, CPACR_EL1); isb(); + + /* Allow EL0 to access TPIDR2 */ + write_sysreg(read_sysreg(SCTLR_EL1) | SCTLR_ELx_ENTP2, SCTLR_EL1); + isb(); }

/* diff --git a/arch/arm64/kernel/process.c b/arch/arm64/kernel/process.c index 7fa97df55e3a..e20571f19718 100644 --- a/arch/arm64/kernel/process.c +++ b/arch/arm64/kernel/process.c @@ -250,6 +250,8 @@ void show_regs(struct pt_regs *regs) static void tls_thread_flush(void) { write_sysreg(0, tpidr_el0); + if (system_supports_tpidr2()) + write_sysreg_s(0, SYS_TPIDR2_EL0);

if (is_compat_task()) { current->thread.uw.tp_value = 0; @@ -343,6 +345,8 @@ int copy_thread(unsigned long clone_flags, unsigned long stack_start, * out-of-sync with the saved value. */ *task_user_tls(p) = read_sysreg(tpidr_el0); + if (system_supports_tpidr2()) + p->thread.tpidr2_el0 = read_sysreg_s(SYS_TPIDR2_EL0);

if (stack_start) { if (is_compat_thread(task_thread_info(p))) @@ -353,10 +357,12 @@ int copy_thread(unsigned long clone_flags, unsigned long stack_start,

/* * If a TLS pointer was passed to clone, use it for the new - * thread. + * thread. We also reset TPIDR2 if it's in use. */ - if (clone_flags & CLONE_SETTLS) + if (clone_flags & CLONE_SETTLS) { p->thread.uw.tp_value = tls; + p->thread.tpidr2_el0 = 0; + } } else { /* * A kthread has no context to ERET to, so ensure any buggy @@ -387,6 +393,8 @@ int copy_thread(unsigned long clone_flags, unsigned long stack_start, void tls_preserve_current_state(void) { *task_user_tls(current) = read_sysreg(tpidr_el0); + if (system_supports_tpidr2() && !is_compat_task()) + current->thread.tpidr2_el0 = read_sysreg_s(SYS_TPIDR2_EL0); }

static void tls_thread_switch(struct task_struct *next) @@ -399,6 +407,8 @@ static void tls_thread_switch(struct task_struct *next) write_sysreg(0, tpidrro_el0);

write_sysreg(*task_user_tls(next), tpidr_el0); + if (system_supports_tpidr2()) + write_sysreg_s(next->thread.tpidr2_el0, SYS_TPIDR2_EL0); }

/*

When in streaming mode we need to save and restore the streaming mode SVE register state rather than the regular SVE register state. This uses the streaming mode vector length and omits FFR but is otherwise identical, if TIF_SVE is enabled when we are in streaming mode then streaming mode takes precedence.

This does not handle use of streaming SVE state with KVM, ptrace or signals. This will be updated in further patches.

Signed-off-by: Mark Brown broonie@kernel.org Reviewed-by: Catalin Marinas catalin.marinas@arm.com --- arch/arm64/include/asm/fpsimd.h | 22 +++++- arch/arm64/include/asm/fpsimdmacros.h | 11 +++ arch/arm64/include/asm/processor.h | 10 +++ arch/arm64/kernel/entry-fpsimd.S | 5 ++ arch/arm64/kernel/fpsimd.c | 109 +++++++++++++++++++++----- arch/arm64/kvm/fpsimd.c | 2 +- 6 files changed, 136 insertions(+), 23 deletions(-)

diff --git a/arch/arm64/include/asm/fpsimd.h b/arch/arm64/include/asm/fpsimd.h index 821d270980da..cd94f5c5b516 100644 --- a/arch/arm64/include/asm/fpsimd.h +++ b/arch/arm64/include/asm/fpsimd.h @@ -47,11 +47,21 @@ extern void fpsimd_update_current_state(struct user_fpsimd_state const *state);

extern void fpsimd_bind_state_to_cpu(struct user_fpsimd_state *state, void *sve_state, unsigned int sve_vl, - u64 *svcr); + unsigned int sme_vl, u64 *svcr);

extern void fpsimd_flush_task_state(struct task_struct *target); extern void fpsimd_save_and_flush_cpu_state(void);

+static inline bool thread_sm_enabled(struct thread_struct *thread) +{ + return system_supports_sme() && (thread->svcr & SYS_SVCR_EL0_SM_MASK); +} + +static inline bool thread_za_enabled(struct thread_struct *thread) +{ + return system_supports_sme() && (thread->svcr & SYS_SVCR_EL0_ZA_MASK); +} + /* Maximum VL that SVE/SME VL-agnostic software can transparently support */ #define VL_ARCH_MAX 0x100

@@ -63,7 +73,14 @@ static inline size_t sve_ffr_offset(int vl)

static inline void *sve_pffr(struct thread_struct *thread) { - return (char *)thread->sve_state + sve_ffr_offset(thread_get_sve_vl(thread)); + unsigned int vl; + + if (system_supports_sme() && thread_sm_enabled(thread)) + vl = thread_get_sme_vl(thread); + else + vl = thread_get_sve_vl(thread); + + return (char *)thread->sve_state + sve_ffr_offset(vl); }

extern void sve_save_state(void *state, u32 *pfpsr, int save_ffr); @@ -72,6 +89,7 @@ extern void sve_load_state(void const *state, u32 const *pfpsr, extern void sve_flush_live(bool flush_ffr, unsigned long vq_minus_1); extern unsigned int sve_get_vl(void); extern void sve_set_vq(unsigned long vq_minus_1); +extern void sme_set_vq(unsigned long vq_minus_1);

struct arm64_cpu_capabilities; extern void sve_kernel_enable(const struct arm64_cpu_capabilities *__unused); diff --git a/arch/arm64/include/asm/fpsimdmacros.h b/arch/arm64/include/asm/fpsimdmacros.h index 2e9a33155081..f6ab36e0cd8d 100644 --- a/arch/arm64/include/asm/fpsimdmacros.h +++ b/arch/arm64/include/asm/fpsimdmacros.h @@ -262,6 +262,17 @@ 921: .endm

+/* Update SMCR_EL1.LEN with the new VQ */ +.macro sme_load_vq xvqminus1, xtmp, xtmp2 + mrs_s \xtmp, SYS_SMCR_EL1 + bic \xtmp2, \xtmp, SMCR_ELx_LEN_MASK + orr \xtmp2, \xtmp2, \xvqminus1 + cmp \xtmp2, \xtmp + b.eq 921f + msr_s SYS_SMCR_EL1, \xtmp2 //self-synchronising +921: +.endm + /* Preserve the first 128-bits of Znz and zero the rest. */ .macro _sve_flush_z nz _sve_check_zreg \nz diff --git a/arch/arm64/include/asm/processor.h b/arch/arm64/include/asm/processor.h index 22cd11e86854..7542310b4e6b 100644 --- a/arch/arm64/include/asm/processor.h +++ b/arch/arm64/include/asm/processor.h @@ -184,6 +184,11 @@ static inline unsigned int thread_get_sve_vl(struct thread_struct *thread) return thread_get_vl(thread, ARM64_VEC_SVE); }

+static inline unsigned int thread_get_sme_vl(struct thread_struct *thread) +{ + return thread_get_vl(thread, ARM64_VEC_SME); +} + unsigned int task_get_vl(const struct task_struct *task, enum vec_type type); void task_set_vl(struct task_struct *task, enum vec_type type, unsigned long vl); @@ -197,6 +202,11 @@ static inline unsigned int task_get_sve_vl(const struct task_struct *task) return task_get_vl(task, ARM64_VEC_SVE); }

+static inline unsigned int task_get_sme_vl(const struct task_struct *task) +{ + return task_get_vl(task, ARM64_VEC_SME); +} + static inline void task_set_sve_vl(struct task_struct *task, unsigned long vl) { task_set_vl(task, ARM64_VEC_SVE, vl); diff --git a/arch/arm64/kernel/entry-fpsimd.S b/arch/arm64/kernel/entry-fpsimd.S index deee5f01462e..6f88c0f86d50 100644 --- a/arch/arm64/kernel/entry-fpsimd.S +++ b/arch/arm64/kernel/entry-fpsimd.S @@ -94,4 +94,9 @@ SYM_FUNC_START(sme_get_vl) ret SYM_FUNC_END(sme_get_vl)

+SYM_FUNC_START(sme_set_vq) + sme_load_vq x0, x1, x2 + ret +SYM_FUNC_END(sme_set_vq) + #endif /* CONFIG_ARM64_SME */ diff --git a/arch/arm64/kernel/fpsimd.c b/arch/arm64/kernel/fpsimd.c index 1c113349f6cc..f8506a875eb2 100644 --- a/arch/arm64/kernel/fpsimd.c +++ b/arch/arm64/kernel/fpsimd.c @@ -123,6 +123,7 @@ struct fpsimd_last_state_struct { void *sve_state; u64 *svcr; unsigned int sve_vl; + unsigned int sme_vl; };

static DEFINE_PER_CPU(struct fpsimd_last_state_struct, fpsimd_last_state); @@ -301,17 +302,28 @@ void task_set_vl_onexec(struct task_struct *task, enum vec_type type, task->thread.vl_onexec[type] = vl; }

+/* + * TIF_SME controls whether a task can use SME without trapping while + * in userspace, when TIF_SME is set then we must have storage + * alocated in sve_state and za_state to store the contents of both ZA + * and the SVE registers for both streaming and non-streaming modes. + * + * If both SVCR.ZA and SVCR.SM are disabled then at any point we + * may disable TIF_SME and reenable traps. + */ + + /* * TIF_SVE controls whether a task can use SVE without trapping while - * in userspace, and also the way a task's FPSIMD/SVE state is stored - * in thread_struct. + * in userspace, and also (together with TIF_SME) the way a task's + * FPSIMD/SVE state is stored in thread_struct. * * The kernel uses this flag to track whether a user task is actively * using SVE, and therefore whether full SVE register state needs to * be tracked. If not, the cheaper FPSIMD context handling code can * be used instead of the more costly SVE equivalents. * - * * TIF_SVE set: + * * TIF_SVE or SVCR.SM set: * * The task can execute SVE instructions while in userspace without * trapping to the kernel. @@ -319,7 +331,8 @@ void task_set_vl_onexec(struct task_struct *task, enum vec_type type, * When stored, Z0-Z31 (incorporating Vn in bits[127:0] or the * corresponding Zn), P0-P15 and FFR are encoded in in * task->thread.sve_state, formatted appropriately for vector - * length task->thread.sve_vl. + * length task->thread.sve_vl or, if SVCR.SM is set, + * task->thread.sme_vl. * * task->thread.sve_state must point to a valid buffer at least * sve_state_size(task) bytes in size. @@ -357,19 +370,40 @@ void task_set_vl_onexec(struct task_struct *task, enum vec_type type, */ static void task_fpsimd_load(void) { + bool restore_sve_regs = false; + bool restore_ffr; + WARN_ON(!system_supports_fpsimd()); WARN_ON(!have_cpu_fpsimd_context());

- if (IS_ENABLED(CONFIG_ARM64_SME) && test_thread_flag(TIF_SME)) - write_sysreg_s(current->thread.svcr, SYS_SVCR_EL0); - + /* Check if we should restore SVE first */ if (IS_ENABLED(CONFIG_ARM64_SVE) && test_thread_flag(TIF_SVE)) { sve_set_vq(sve_vq_from_vl(task_get_sve_vl(current)) - 1); + restore_sve_regs = true; + restore_ffr = true; + } + + /* Restore SME, override SVE register configuration if needed */ + if (system_supports_sme()) { + unsigned long sme_vl = task_get_sme_vl(current); + + if (test_thread_flag(TIF_SME)) + sme_set_vq(sve_vq_from_vl(sme_vl) - 1); + + write_sysreg_s(current->thread.svcr, SYS_SVCR_EL0); + + if (thread_sm_enabled(&current->thread)) { + restore_sve_regs = true; + restore_ffr = system_supports_fa64(); + } + } + + if (restore_sve_regs) sve_load_state(sve_pffr(&current->thread), - &current->thread.uw.fpsimd_state.fpsr, true); - } else { + &current->thread.uw.fpsimd_state.fpsr, + restore_ffr); + else fpsimd_load_state(&current->thread.uw.fpsimd_state); - } }

/* @@ -387,6 +421,9 @@ static void fpsimd_save(void) struct fpsimd_last_state_struct const *last = this_cpu_ptr(&fpsimd_last_state); /* set by fpsimd_bind_task_to_cpu() or fpsimd_bind_state_to_cpu() */ + bool save_sve_regs = false; + bool save_ffr; + unsigned int vl;

WARN_ON(!system_supports_fpsimd()); WARN_ON(!have_cpu_fpsimd_context()); @@ -394,15 +431,33 @@ static void fpsimd_save(void) if (test_thread_flag(TIF_FOREIGN_FPSTATE)) return;

- if (IS_ENABLED(CONFIG_ARM64_SME) && - test_thread_flag(TIF_SME)) { + if (test_thread_flag(TIF_SVE)) { + save_sve_regs = true; + save_ffr = true; + vl = last->sve_vl; + } + + if (system_supports_sme()) { u64 *svcr = last->svcr; *svcr = read_sysreg_s(SYS_SVCR_EL0); + + if (thread_za_enabled(&current->thread)) { + /* ZA state managment is not implemented yet */ + force_signal_inject(SIGKILL, SI_KERNEL, 0, 0); + return; + } + + /* If we are in streaming mode override regular SVE. */ + if (*svcr & SYS_SVCR_EL0_SM_MASK) { + save_sve_regs = true; + save_ffr = system_supports_fa64(); + vl = last->sme_vl; + } }

- if (IS_ENABLED(CONFIG_ARM64_SVE) && - test_thread_flag(TIF_SVE)) { - if (WARN_ON(sve_get_vl() != last->sve_vl)) { + if (IS_ENABLED(CONFIG_ARM64_SVE) && save_sve_regs) { + /* Get the configured VL from RDVL, will account for SM */ + if (WARN_ON(sve_get_vl() != vl)) { /* * Can't save the user regs, so current would * re-enter user with corrupt state. @@ -413,8 +468,8 @@ static void fpsimd_save(void) }

sve_save_state((char *)last->sve_state + - sve_ffr_offset(last->sve_vl), - &last->st->fpsr, true); + sve_ffr_offset(vl), + &last->st->fpsr, save_ffr); } else { fpsimd_save_state(last->st); } @@ -619,7 +674,14 @@ static void sve_to_fpsimd(struct task_struct *task) */ static size_t sve_state_size(struct task_struct const *task) { - return SVE_SIG_REGS_SIZE(sve_vq_from_vl(task_get_sve_vl(task))); + unsigned int vl = 0; + + if (system_supports_sve()) + vl = task_get_sve_vl(task); + if (system_supports_sme()) + vl = max(vl, task_get_sme_vl(task)); + + return SVE_SIG_REGS_SIZE(sve_vq_from_vl(vl)); }

/* @@ -748,7 +810,8 @@ int vec_set_vector_length(struct task_struct *task, enum vec_type type, }

fpsimd_flush_task_state(task); - if (test_and_clear_tsk_thread_flag(task, TIF_SVE)) + if (test_and_clear_tsk_thread_flag(task, TIF_SVE) || + thread_sm_enabled(&task->thread)) sve_to_fpsimd(task);

if (system_supports_sme() && type == ARM64_VEC_SME) @@ -1375,6 +1438,9 @@ void fpsimd_flush_thread(void) fpsimd_flush_thread_vl(ARM64_VEC_SVE); }

+ if (system_supports_sme()) + fpsimd_flush_thread_vl(ARM64_VEC_SME); + put_cpu_fpsimd_context(); }

@@ -1418,6 +1484,7 @@ static void fpsimd_bind_task_to_cpu(void) last->st = &current->thread.uw.fpsimd_state; last->sve_state = current->thread.sve_state; last->sve_vl = task_get_sve_vl(current); + last->sme_vl = task_get_sme_vl(current); last->svcr = &current->thread.svcr; current->thread.fpsimd_cpu = smp_processor_id();

@@ -1433,7 +1500,8 @@ static void fpsimd_bind_task_to_cpu(void) }

void fpsimd_bind_state_to_cpu(struct user_fpsimd_state *st, void *sve_state, - unsigned int sve_vl, u64 *svcr) + unsigned int sve_vl, unsigned int sme_vl, + u64 *svcr) { struct fpsimd_last_state_struct *last = this_cpu_ptr(&fpsimd_last_state); @@ -1445,6 +1513,7 @@ void fpsimd_bind_state_to_cpu(struct user_fpsimd_state *st, void *sve_state, last->svcr = svcr; last->sve_state = sve_state; last->sve_vl = sve_vl; + last->sme_vl = sme_vl; }

/* diff --git a/arch/arm64/kvm/fpsimd.c b/arch/arm64/kvm/fpsimd.c index ac09f1f682ff..394e583bb73e 100644 --- a/arch/arm64/kvm/fpsimd.c +++ b/arch/arm64/kvm/fpsimd.c @@ -116,7 +116,7 @@ void kvm_arch_vcpu_ctxsync_fp(struct kvm_vcpu *vcpu) fpsimd_bind_state_to_cpu(&vcpu->arch.ctxt.fp_regs, vcpu->arch.sve_state, vcpu->arch.sve_max_vl, - NULL); + 0, NULL);

clear_thread_flag(TIF_FOREIGN_FPSTATE); update_thread_flag(TIF_SVE, vcpu_has_sve(vcpu));

By default all SME operations in userspace will trap. When this happens we allocate storage space for the SME register state, set up the SVE registers and disable traps. We do not need to initialize ZA since the architecture guarantees that it will be zeroed when enabled and when we trap ZA is disabled.

On syscall we exit streaming mode if we were previously in it and ensure that all but the lower 128 bits of the registers are zeroed while preserving the state of ZA. This follows the aarch64 PCS for SME, ZA state is preserved over a function call and streaming mode is exited. Since the traps for SME do not distinguish between streaming mode SVE and ZA usage if ZA is in use rather than reenabling traps we instead zero the parts of the SVE registers not shared with FPSIMD and leave SME enabled, this simplifies handling SME traps. If ZA is not in use then we reenable SME traps and fall through to normal handling of SVE.

Signed-off-by: Mark Brown broonie@kernel.org --- arch/arm64/include/asm/esr.h | 1 + arch/arm64/include/asm/exception.h | 1 + arch/arm64/include/asm/fpsimd.h | 39 +++++++ arch/arm64/kernel/entry-common.c | 11 ++ arch/arm64/kernel/fpsimd.c | 167 +++++++++++++++++++++++++---- arch/arm64/kernel/process.c | 30 +++++- arch/arm64/kernel/syscall.c | 29 ++++- 7 files changed, 255 insertions(+), 23 deletions(-)

diff --git a/arch/arm64/include/asm/esr.h b/arch/arm64/include/asm/esr.h index 43872e0cfd1e..0467837fd66b 100644 --- a/arch/arm64/include/asm/esr.h +++ b/arch/arm64/include/asm/esr.h @@ -76,6 +76,7 @@ #define ESR_ELx_IL_SHIFT (25) #define ESR_ELx_IL (UL(1) << ESR_ELx_IL_SHIFT) #define ESR_ELx_ISS_MASK (ESR_ELx_IL - 1) +#define ESR_ELx_ISS(esr) ((esr) & ESR_ELx_ISS_MASK)

/* ISS field definitions shared by different classes */ #define ESR_ELx_WNR_SHIFT (6) diff --git a/arch/arm64/include/asm/exception.h b/arch/arm64/include/asm/exception.h index 339477dca551..2add7f33b7c2 100644 --- a/arch/arm64/include/asm/exception.h +++ b/arch/arm64/include/asm/exception.h @@ -64,6 +64,7 @@ void do_debug_exception(unsigned long addr_if_watchpoint, unsigned int esr, struct pt_regs *regs); void do_fpsimd_acc(unsigned int esr, struct pt_regs *regs); void do_sve_acc(unsigned int esr, struct pt_regs *regs); +void do_sme_acc(unsigned int esr, struct pt_regs *regs); void do_fpsimd_exc(unsigned int esr, struct pt_regs *regs); void do_sysinstr(unsigned int esr, struct pt_regs *regs); void do_sp_pc_abort(unsigned long addr, unsigned int esr, struct pt_regs *regs); diff --git a/arch/arm64/include/asm/fpsimd.h b/arch/arm64/include/asm/fpsimd.h index 1a709c03bb6c..6c33bc832ed4 100644 --- a/arch/arm64/include/asm/fpsimd.h +++ b/arch/arm64/include/asm/fpsimd.h @@ -239,6 +239,8 @@ static inline bool sve_vq_available(unsigned int vq) return vq_available(ARM64_VEC_SVE, vq); }

+size_t sve_state_size(struct task_struct const *task); + #else /* ! CONFIG_ARM64_SVE */

static inline void sve_alloc(struct task_struct *task) { } @@ -278,10 +280,25 @@ static inline void vec_update_vq_map(enum vec_type t) { } static inline int vec_verify_vq_map(enum vec_type t) { return 0; } static inline void sve_setup(void) { }

+static inline size_t sve_state_size(struct task_struct const *task) +{ + return 0; +} + #endif /* ! CONFIG_ARM64_SVE */

#ifdef CONFIG_ARM64_SME

+static inline void sme_user_disable(void) +{ + sysreg_clear_set(cpacr_el1, CPACR_EL1_SMEN_EL0EN, 0); +} + +static inline void sme_user_enable(void) +{ + sysreg_clear_set(cpacr_el1, 0, CPACR_EL1_SMEN_EL0EN); +} + static inline void sme_smstart_sm(void) { asm volatile(__msr_s(SYS_SVCR_SMSTART_SM_EL0, "xzr")); @@ -309,16 +326,33 @@ static inline int sme_max_virtualisable_vl(void) return vec_max_virtualisable_vl(ARM64_VEC_SME); }

+extern void sme_alloc(struct task_struct *task); extern unsigned int sme_get_vl(void); extern int sme_set_current_vl(unsigned long arg); extern int sme_get_current_vl(void);

+/* + * Return how many bytes of memory are required to store the full SME + * specific state (currently just ZA) for task, given task's currently + * configured vector length. + */ +static inline size_t za_state_size(struct task_struct const *task) +{ + unsigned int vl = task_get_sme_vl(task); + + return ZA_SIG_REGS_SIZE(sve_vq_from_vl(vl)); +} + #else

+static inline void sme_user_disable(void) { BUILD_BUG(); } +static inline void sme_user_enable(void) { BUILD_BUG(); } + static inline void sme_smstart_sm(void) { } static inline void sme_smstop_sm(void) { } static inline void sme_smstop(void) { }

+static inline void sme_alloc(struct task_struct *task) { } static inline void sme_setup(void) { } static inline unsigned int sme_get_vl(void) { return 0; } static inline int sme_max_vl(void) { return 0; } @@ -326,6 +360,11 @@ static inline int sme_max_virtualisable_vl(void) { return 0; } static inline int sme_set_current_vl(unsigned long arg) { return -EINVAL; } static inline int sme_get_current_vl(void) { return -EINVAL; }

+static inline size_t za_state_size(struct task_struct const *task) +{ + return 0; +} + #endif /* ! CONFIG_ARM64_SME */

/* For use by EFI runtime services calls only */ diff --git a/arch/arm64/kernel/entry-common.c b/arch/arm64/kernel/entry-common.c index 878c65aa7206..29139e9a1517 100644 --- a/arch/arm64/kernel/entry-common.c +++ b/arch/arm64/kernel/entry-common.c @@ -537,6 +537,14 @@ static void noinstr el0_sve_acc(struct pt_regs *regs, unsigned long esr) exit_to_user_mode(regs); }

+static void noinstr el0_sme_acc(struct pt_regs *regs, unsigned long esr) +{ + enter_from_user_mode(regs); + local_daif_restore(DAIF_PROCCTX); + do_sme_acc(esr, regs); + exit_to_user_mode(regs); +} + static void noinstr el0_fpsimd_exc(struct pt_regs *regs, unsigned long esr) { enter_from_user_mode(regs); @@ -645,6 +653,9 @@ asmlinkage void noinstr el0t_64_sync_handler(struct pt_regs *regs) case ESR_ELx_EC_SVE: el0_sve_acc(regs, esr); break; + case ESR_ELx_EC_SME: + el0_sme_acc(regs, esr); + break; case ESR_ELx_EC_FP_EXC64: el0_fpsimd_exc(regs, esr); break; diff --git a/arch/arm64/kernel/fpsimd.c b/arch/arm64/kernel/fpsimd.c index dc38f3f2a28a..00a0cbd01ce5 100644 --- a/arch/arm64/kernel/fpsimd.c +++ b/arch/arm64/kernel/fpsimd.c @@ -209,6 +209,12 @@ static void set_sme_default_vl(int val) set_default_vl(ARM64_VEC_SME, val); }

+static void sme_free(struct task_struct *); + +#else + +static inline void sme_free(struct task_struct *t) { } + #endif

DEFINE_PER_CPU(bool, fpsimd_context_busy); @@ -676,7 +682,7 @@ static void sve_to_fpsimd(struct task_struct *task) * Return how many bytes of memory are required to store the full SVE * state for task, given task's currently configured vector length. */ -static size_t sve_state_size(struct task_struct const *task) +size_t sve_state_size(struct task_struct const *task) { unsigned int vl = 0;

@@ -818,18 +824,22 @@ int vec_set_vector_length(struct task_struct *task, enum vec_type type, thread_sm_enabled(&task->thread)) sve_to_fpsimd(task);

- if (system_supports_sme() && type == ARM64_VEC_SME) + if (system_supports_sme() && type == ARM64_VEC_SME) { task->thread.svcr &= ~(SYS_SVCR_EL0_SM_MASK | SYS_SVCR_EL0_ZA_MASK); + clear_thread_flag(TIF_SME); + }

if (task == current) put_cpu_fpsimd_context();

/* - * Force reallocation of task SVE state to the correct size - * on next use: + * Force reallocation of task SVE and SME state to the correct + * size on next use: */ sve_free(task); + if (system_supports_sme() && type == ARM64_VEC_SME) + sme_free(task);

task_set_vl(task, type, vl);

@@ -1164,12 +1174,43 @@ void __init sve_setup(void) void fpsimd_release_task(struct task_struct *dead_task) { __sve_free(dead_task); + sme_free(dead_task); }

#endif /* CONFIG_ARM64_SVE */

#ifdef CONFIG_ARM64_SME

+/* This will move to uapi/asm/sigcontext.h when signals are implemented */ +#define ZA_SIG_REGS_SIZE(vq) ((vq * __SVE_VQ_BYTES) * (vq * __SVE_VQ_BYTES)) + +/* + * Ensure that task->thread.za_state is allocated and sufficiently large. + * + * This function should be used only in preparation for replacing + * task->thread.za_state with new data. The memory is always zeroed + * here to prevent stale data from showing through: this is done in + * the interest of testability and predictability, the architecture + * guarantees that when ZA is enabled it will be zeroed. + */ +void sme_alloc(struct task_struct *task) +{ + if (task->thread.za_state) { + memset(task->thread.za_state, 0, za_state_size(task)); + return; + } + + /* This could potentially be up to 64K. */ + task->thread.za_state = + kzalloc(za_state_size(task), GFP_KERNEL); +} + +static void sme_free(struct task_struct *task) +{ + kfree(task->thread.za_state); + task->thread.za_state = NULL; +} + void sme_kernel_enable(const struct arm64_cpu_capabilities *__always_unused p) { /* Set priority for all PEs to architecturally defined minimum */ @@ -1279,6 +1320,29 @@ void __init sme_setup(void)

#endif /* CONFIG_ARM64_SME */

+static void sve_init_regs(void) +{ + /* + * Convert the FPSIMD state to SVE, zeroing all the state that + * is not shared with FPSIMD. If (as is likely) the current + * state is live in the registers then do this there and + * update our metadata for the current task including + * disabling the trap, otherwise update our in-memory copy. + * We are guaranteed to not be in streaming mode, we can only + * take a SVE trap when not in streaming mode and we can't be + * in streaming mode when taking a SME trap. + */ + if (!test_thread_flag(TIF_FOREIGN_FPSTATE)) { + unsigned long vq_minus_one = + sve_vq_from_vl(task_get_sve_vl(current)) - 1; + sve_set_vq(vq_minus_one); + sve_flush_live(true, vq_minus_one); + fpsimd_bind_task_to_cpu(); + } else { + fpsimd_to_sve(current); + } +} + /* * Trapped SVE access * @@ -1310,22 +1374,77 @@ void do_sve_acc(unsigned int esr, struct pt_regs *regs) WARN_ON(1); /* SVE access shouldn't have trapped */

/* - * Convert the FPSIMD state to SVE, zeroing all the state that - * is not shared with FPSIMD. If (as is likely) the current - * state is live in the registers then do this there and - * update our metadata for the current task including - * disabling the trap, otherwise update our in-memory copy. + * Even if the task can have used streaming mode we can only + * generate SVE access traps in normal SVE mode and + * transitioning out of streaming mode may discard any + * streaming mode state. Always clear the high bits to avoid + * any potential errors tracking what is properly initialised. + */ + sve_init_regs(); + + put_cpu_fpsimd_context(); +} + +/* + * Trapped SME access + * + * Storage is allocated for the full SVE and SME state, the current + * FPSIMD register contents are migrated to SVE if SVE is not already + * active, and the access trap is disabled. + * + * TIF_SME should be clear on entry: otherwise, fpsimd_restore_current_state() + * would have disabled the SME access trap for userspace during + * ret_to_user, making an SVE access trap impossible in that case. + */ +void do_sme_acc(unsigned int esr, struct pt_regs *regs) +{ + /* Even if we chose not to use SME, the hardware could still trap: */ + if (unlikely(!system_supports_sme()) || WARN_ON(is_compat_task())) { + force_signal_inject(SIGILL, ILL_ILLOPC, regs->pc, 0); + return; + } + + /* + * If this not a trap due to SME being disabled then something + * is being used in the wrong mode, report as SIGILL. */ + if (ESR_ELx_ISS(esr) != ESR_ELx_SME_ISS_SME_DISABLED) { + force_signal_inject(SIGILL, ILL_ILLOPC, regs->pc, 0); + return; + } + + sve_alloc(current); + sme_alloc(current); + if (!current->thread.sve_state || !current->thread.za_state) { + force_sig(SIGKILL); + return; + } + + get_cpu_fpsimd_context(); + + /* With TIF_SME userspace shouldn't generate any traps */ + if (test_and_set_thread_flag(TIF_SME)) + WARN_ON(1); + if (!test_thread_flag(TIF_FOREIGN_FPSTATE)) { unsigned long vq_minus_one = - sve_vq_from_vl(task_get_sve_vl(current)) - 1; - sve_set_vq(vq_minus_one); - sve_flush_live(true, vq_minus_one); + sve_vq_from_vl(task_get_sme_vl(current)) - 1; + sme_set_vq(vq_minus_one); + fpsimd_bind_task_to_cpu(); - } else { - fpsimd_to_sve(current); }

+ /* + * If SVE was not already active initialise the SVE registers, + * any non-shared state between the streaming and regular SVE + * registers is architecturally guaranteed to be zeroed when + * we enter streaming mode. We do not need to initialize ZA + * since ZA must be disabled at this point and enabling ZA is + * architecturally defined to zero ZA. + */ + if (system_supports_sve() && !test_thread_flag(TIF_SVE)) + sve_init_regs(); + put_cpu_fpsimd_context(); }

@@ -1442,8 +1561,12 @@ void fpsimd_flush_thread(void) fpsimd_flush_thread_vl(ARM64_VEC_SVE); }

- if (system_supports_sme()) + if (system_supports_sme()) { + clear_thread_flag(TIF_SME); + sme_free(current); fpsimd_flush_thread_vl(ARM64_VEC_SME); + current->thread.svcr = 0; + }

put_cpu_fpsimd_context(); } @@ -1493,14 +1616,22 @@ static void fpsimd_bind_task_to_cpu(void) last->svcr = &current->thread.svcr; current->thread.fpsimd_cpu = smp_processor_id();

+ /* + * Toggle SVE and SME trapping for userspace if needed, these + * are serialsied by ret_to_user(). + */ + if (system_supports_sme()) { + if (test_thread_flag(TIF_SME)) + sme_user_enable(); + else + sme_user_disable(); + } + if (system_supports_sve()) { - /* Toggle SVE trapping for userspace if needed */ if (test_thread_flag(TIF_SVE)) sve_user_enable(); else sve_user_disable(); - - /* Serialised by exception return to user */ } }

diff --git a/arch/arm64/kernel/process.c b/arch/arm64/kernel/process.c index 07f235b46cf5..99c293513817 100644 --- a/arch/arm64/kernel/process.c +++ b/arch/arm64/kernel/process.c @@ -300,17 +300,41 @@ int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src)

/* * Detach src's sve_state (if any) from dst so that it does not - * get erroneously used or freed prematurely. dst's sve_state + * get erroneously used or freed prematurely. dst's copies * will be allocated on demand later on if dst uses SVE. * For consistency, also clear TIF_SVE here: this could be done * later in copy_process(), but to avoid tripping up future - * maintainers it is best not to leave TIF_SVE and sve_state in + * maintainers it is best not to leave TIF flags and buffers in * an inconsistent state, even temporarily. */ dst->thread.sve_state = NULL; clear_tsk_thread_flag(dst, TIF_SVE);

- dst->thread.svcr = 0; + /* + * In the unlikely event that we create a new thread with ZA + * enabled we should retain the ZA state so duplicate it here. + * This may be shortly freed if we exec() or if CLONE_SETTLS + * but it's simpler to do it here. To avoid confusing the rest + * of the code ensure that we have a sve_state allocated + * whenever za_state is allocated. + */ + if (thread_za_enabled(&src->thread)) { + dst->thread.sve_state = kzalloc(sve_state_size(src), + GFP_KERNEL); + if (!dst->thread.za_state) + return -ENOMEM; + dst->thread.za_state = kmemdup(src->thread.za_state, + za_state_size(src), + GFP_KERNEL); + if (!dst->thread.za_state) { + kfree(dst->thread.sve_state); + dst->thread.sve_state = NULL; + return -ENOMEM; + } + } else { + dst->thread.za_state = NULL; + clear_tsk_thread_flag(dst, TIF_SME); + }

/* clear any pending asynchronous tag fault raised by the parent */ clear_tsk_thread_flag(dst, TIF_MTE_ASYNC_FAULT); diff --git a/arch/arm64/kernel/syscall.c b/arch/arm64/kernel/syscall.c index c938603b3ba0..92c69e5ac269 100644 --- a/arch/arm64/kernel/syscall.c +++ b/arch/arm64/kernel/syscall.c @@ -158,11 +158,36 @@ static void el0_svc_common(struct pt_regs *regs, int scno, int sc_nr, syscall_trace_exit(regs); }

-static inline void sve_user_discard(void) +/* + * As per the ABI exit SME streaming mode and clear the SVE state not + * shared with FPSIMD on syscall entry. + */ +static inline void fp_user_discard(void) { + /* + * If SME is active then exit streaming mode. If ZA is active + * then flush the SVE registers but leave userspace access to + * both SVE and SME enabled, otherwise disable SME for the + * task and fall through to disabling SVE too. This means + * that after a syscall we never have any streaming mode + * register state to track, if this changes the KVM code will + * need updating. + */ + if (system_supports_sme() && test_thread_flag(TIF_SME)) { + u64 svcr = read_sysreg_s(SYS_SVCR_EL0); + + if (svcr & SYS_SVCR_EL0_SM_MASK) + sme_smstop_sm(); + } + if (!system_supports_sve()) return;

+ /* + * If SME is not active then disable SVE, the registers will + * be cleared when userspace next attempts to access them and + * we do not need to track the SVE register state until then. + */ clear_thread_flag(TIF_SVE);

/* @@ -177,7 +202,7 @@ static inline void sve_user_discard(void)

void do_el0_svc(struct pt_regs *regs) { - sve_user_discard(); + fp_user_discard(); el0_svc_common(regs, regs->regs[8], __NR_syscalls, sys_call_table); }

When in streaming mode we have the same set of SVE registers as we do in regular SVE mode with the exception of FFR and the use of the SME vector length. Provide signal handling for these registers by taking one of the reserved words in the SVE signal context as a flags field and defining a flag which is set for streaming mode. When the flag is set the vector length is set to the streaming mode vector length and we save and restore streaming mode data. We support entering or leaving streaming mode based on the value of the flag but do not support changing the vector length, this is not currently supported SVE signal handling.

We could instead allocate a separate record in the signal frame for the streaming mode SVE context but this inflates the size of the maximal signal frame required and adds complication when validating signal frames from userspace, especially given the current structure of the code.

Any implementation of support for streaming mode vectors in signals will have some potential for causing issues for applications that attempt to handle SVE vectors in signals, use streaming mode but do not understand streaming mode in their signal handling code, it is hard to identify a case that is clearly better than any other - they all have cases where they could cause unexpected register corruption or faults.

Signed-off-by: Mark Brown broonie@kernel.org Reviewed-by: Catalin Marinas catalin.marinas@arm.com --- arch/arm64/include/asm/processor.h | 8 +++++ arch/arm64/include/uapi/asm/sigcontext.h | 16 +++++++-- arch/arm64/kernel/signal.c | 42 ++++++++++++++++++------ 3 files changed, 53 insertions(+), 13 deletions(-)

diff --git a/arch/arm64/include/asm/processor.h b/arch/arm64/include/asm/processor.h index 6a3a6c3dec90..1d2ca4870b84 100644 --- a/arch/arm64/include/asm/processor.h +++ b/arch/arm64/include/asm/processor.h @@ -190,6 +190,14 @@ static inline unsigned int thread_get_sme_vl(struct thread_struct *thread) return thread_get_vl(thread, ARM64_VEC_SME); }

+static inline unsigned int thread_get_cur_vl(struct thread_struct *thread) +{ + if (system_supports_sme() && (thread->svcr & SYS_SVCR_EL0_SM_MASK)) + return thread_get_sme_vl(thread); + else + return thread_get_sve_vl(thread); +} + unsigned int task_get_vl(const struct task_struct *task, enum vec_type type); void task_set_vl(struct task_struct *task, enum vec_type type, unsigned long vl); diff --git a/arch/arm64/include/uapi/asm/sigcontext.h b/arch/arm64/include/uapi/asm/sigcontext.h index 0c796c795dbe..57e9f8c3ee9e 100644 --- a/arch/arm64/include/uapi/asm/sigcontext.h +++ b/arch/arm64/include/uapi/asm/sigcontext.h @@ -134,9 +134,12 @@ struct extra_context { struct sve_context { struct _aarch64_ctx head; __u16 vl; - __u16 __reserved[3]; + __u16 flags; + __u16 __reserved[2]; };

+#define SVE_SIG_FLAG_SM 0x1 /* Context describes streaming mode */ + #endif /* !__ASSEMBLY__ */

#include <asm/sve_context.h> @@ -186,9 +189,16 @@ struct sve_context { * sve_context.vl must equal the thread's current vector length when * doing a sigreturn. * + * On systems with support for SME the SVE register state may reflect either + * streaming or non-streaming mode. In streaming mode the streaming mode + * vector length will be used and the flag SVE_SIG_FLAG_SM will be set in + * the flags field. It is permitted to enter or leave streaming mode in + * a signal return, applications should take care to ensure that any difference + * in vector length between the two modes is handled, including any resizing + * and movement of context blocks. * - * Note: for all these macros, the "vq" argument denotes the SVE - * vector length in quadwords (i.e., units of 128 bits). + * Note: for all these macros, the "vq" argument denotes the vector length + * in quadwords (i.e., units of 128 bits). * * The correct way to obtain vq is to use sve_vq_from_vl(vl). The * result is valid if and only if sve_vl_valid(vl) is true. This is diff --git a/arch/arm64/kernel/signal.c b/arch/arm64/kernel/signal.c index 42efa464e46e..0ddce6afd2a3 100644 --- a/arch/arm64/kernel/signal.c +++ b/arch/arm64/kernel/signal.c @@ -226,11 +226,17 @@ static int preserve_sve_context(struct sve_context __user *ctx) { int err = 0; u16 reserved[ARRAY_SIZE(ctx->__reserved)]; + u16 flags = 0; unsigned int vl = task_get_sve_vl(current); unsigned int vq = 0;

- if (test_thread_flag(TIF_SVE)) + if (thread_sm_enabled(&current->thread)) { + vl = task_get_sme_vl(current); vq = sve_vq_from_vl(vl); + flags |= SVE_SIG_FLAG_SM; + } else if (test_thread_flag(TIF_SVE)) { + vq = sve_vq_from_vl(vl); + }

memset(reserved, 0, sizeof(reserved));

@@ -238,6 +244,7 @@ static int preserve_sve_context(struct sve_context __user *ctx) __put_user_error(round_up(SVE_SIG_CONTEXT_SIZE(vq), 16), &ctx->head.size, err); __put_user_error(vl, &ctx->vl, err); + __put_user_error(flags, &ctx->flags, err); BUILD_BUG_ON(sizeof(ctx->__reserved) != sizeof(reserved)); err |= __copy_to_user(&ctx->__reserved, reserved, sizeof(reserved));

@@ -258,18 +265,28 @@ static int preserve_sve_context(struct sve_context __user *ctx) static int restore_sve_fpsimd_context(struct user_ctxs *user) { int err; - unsigned int vq; + unsigned int vl, vq; struct user_fpsimd_state fpsimd; struct sve_context sve;

if (__copy_from_user(&sve, user->sve, sizeof(sve))) return -EFAULT;

- if (sve.vl != task_get_sve_vl(current)) + if (sve.flags & SVE_SIG_FLAG_SM) { + if (!system_supports_sme()) + return -EINVAL; + + vl = task_get_sme_vl(current); + } else { + vl = task_get_sve_vl(current); + } + + if (sve.vl != vl) return -EINVAL;

if (sve.head.size <= sizeof(*user->sve)) { clear_thread_flag(TIF_SVE); + current->thread.svcr &= ~SYS_SVCR_EL0_SM_MASK; goto fpsimd_only; }

@@ -301,7 +318,10 @@ static int restore_sve_fpsimd_context(struct user_ctxs *user) if (err) return -EFAULT;

- set_thread_flag(TIF_SVE); + if (sve.flags & SVE_SIG_FLAG_SM) + current->thread.svcr |= SYS_SVCR_EL0_SM_MASK; + else + set_thread_flag(TIF_SVE);

fpsimd_only: /* copy the FP and status/control registers */ @@ -393,7 +413,7 @@ static int parse_user_sigframe(struct user_ctxs *user, break;

case SVE_MAGIC: - if (!system_supports_sve()) + if (!system_supports_sve() && !system_supports_sme()) goto invalid;

if (user->sve) @@ -594,11 +614,12 @@ static int setup_sigframe_layout(struct rt_sigframe_user_layout *user, if (system_supports_sve()) { unsigned int vq = 0;

- if (add_all || test_thread_flag(TIF_SVE)) { - int vl = sve_max_vl(); + if (add_all || test_thread_flag(TIF_SVE) || + thread_sm_enabled(&current->thread)) { + int vl = max(sve_max_vl(), sme_max_vl());

if (!add_all) - vl = task_get_sve_vl(current); + vl = thread_get_cur_vl(&current->thread);

vq = sve_vq_from_vl(vl); } @@ -649,8 +670,9 @@ static int setup_sigframe(struct rt_sigframe_user_layout *user, __put_user_error(current->thread.fault_code, &esr_ctx->esr, err); }

- /* Scalable Vector Extension state, if present */ - if (system_supports_sve() && err == 0 && user->sve_offset) { + /* Scalable Vector Extension state (including streaming), if present */ + if ((system_supports_sve() || system_supports_sme()) && + err == 0 && user->sve_offset) { struct sve_context __user *sve_ctx = apply_user_offset(user, user->sve_offset); err |= preserve_sve_context(sve_ctx);

The streaming mode SVE registers are represented using the same data structures as for SVE but since the vector lengths supported and in use may not be the same as SVE we represent them with a new type NT_ARM_SSVE. Unfortunately we only have a single 16 bit reserved field available in the header so there is no space to fit the current and maximum vector length for both standard and streaming SVE mode without redefining the structure in a way the creates a complicatd and fragile ABI. Since FFR is not present in streaming mode it is read and written as zero.

Setting NT_ARM_SSVE registers will put the task into streaming mode, similarly setting NT_ARM_SVE registers will exit it. Reads that do not correspond to the current mode of the task will return the header with no register data. For compatibility reasons on write setting no flag for the register type will be interpreted as setting SVE registers, though users can provide no register data as an alternative mechanism for doing so.

Signed-off-by: Mark Brown broonie@kernel.org Reviewed-by: Catalin Marinas catalin.marinas@arm.com --- arch/arm64/include/asm/fpsimd.h | 1 + arch/arm64/include/uapi/asm/ptrace.h | 13 +- arch/arm64/kernel/fpsimd.c | 31 +++- arch/arm64/kernel/ptrace.c | 214 +++++++++++++++++++++------ include/uapi/linux/elf.h | 1 + 5 files changed, 201 insertions(+), 59 deletions(-)

diff --git a/arch/arm64/include/asm/fpsimd.h b/arch/arm64/include/asm/fpsimd.h index 6c33bc832ed4..5afcd0709aae 100644 --- a/arch/arm64/include/asm/fpsimd.h +++ b/arch/arm64/include/asm/fpsimd.h @@ -144,6 +144,7 @@ struct vl_info { extern void sve_alloc(struct task_struct *task); extern void fpsimd_release_task(struct task_struct *task); extern void fpsimd_sync_to_sve(struct task_struct *task); +extern void fpsimd_force_sync_to_sve(struct task_struct *task); extern void sve_sync_to_fpsimd(struct task_struct *task); extern void sve_sync_from_fpsimd_zeropad(struct task_struct *task);

diff --git a/arch/arm64/include/uapi/asm/ptrace.h b/arch/arm64/include/uapi/asm/ptrace.h index 758ae984ff97..522b925a78c1 100644 --- a/arch/arm64/include/uapi/asm/ptrace.h +++ b/arch/arm64/include/uapi/asm/ptrace.h @@ -109,7 +109,7 @@ struct user_hwdebug_state { } dbg_regs[16]; };

-/* SVE/FP/SIMD state (NT_ARM_SVE) */ +/* SVE/FP/SIMD state (NT_ARM_SVE & NT_ARM_SSVE) */

struct user_sve_header { __u32 size; /* total meaningful regset content in bytes */ @@ -220,6 +220,7 @@ struct user_sve_header { (SVE_PT_SVE_PREG_OFFSET(vq, __SVE_NUM_PREGS) - \ SVE_PT_SVE_PREGS_OFFSET(vq))

+/* For streaming mode SVE (SSVE) FFR must be read and written as zero */ #define SVE_PT_SVE_FFR_OFFSET(vq) \ (SVE_PT_REGS_OFFSET + __SVE_FFR_OFFSET(vq))

@@ -240,10 +241,12 @@ struct user_sve_header { - SVE_PT_SVE_OFFSET + (__SVE_VQ_BYTES - 1)) \ / __SVE_VQ_BYTES * __SVE_VQ_BYTES)

-#define SVE_PT_SIZE(vq, flags) \ - (((flags) & SVE_PT_REGS_MASK) == SVE_PT_REGS_SVE ? \ - SVE_PT_SVE_OFFSET + SVE_PT_SVE_SIZE(vq, flags) \ - : SVE_PT_FPSIMD_OFFSET + SVE_PT_FPSIMD_SIZE(vq, flags)) +#define SVE_PT_SIZE(vq, flags) \ + (((flags) & SVE_PT_REGS_MASK) == SVE_PT_REGS_SVE ? \ + SVE_PT_SVE_OFFSET + SVE_PT_SVE_SIZE(vq, flags) \ + : ((((flags) & SVE_PT_REGS_MASK) == SVE_PT_REGS_FPSIMD ? \ + SVE_PT_FPSIMD_OFFSET + SVE_PT_FPSIMD_SIZE(vq, flags) \ + : SVE_PT_REGS_OFFSET)))

/* pointer authentication masks (NT_ARM_PAC_MASK) */

diff --git a/arch/arm64/kernel/fpsimd.c b/arch/arm64/kernel/fpsimd.c index 80f7ca12f855..94f06e9d37cf 100644 --- a/arch/arm64/kernel/fpsimd.c +++ b/arch/arm64/kernel/fpsimd.c @@ -643,7 +643,7 @@ static void fpsimd_to_sve(struct task_struct *task) if (!system_supports_sve()) return;

- vq = sve_vq_from_vl(task_get_sve_vl(task)); + vq = sve_vq_from_vl(thread_get_cur_vl(&task->thread)); __fpsimd_to_sve(sst, fst, vq); }

@@ -660,7 +660,7 @@ static void fpsimd_to_sve(struct task_struct *task) */ static void sve_to_fpsimd(struct task_struct *task) { - unsigned int vq; + unsigned int vq, vl; void const *sst = task->thread.sve_state; struct user_fpsimd_state *fst = &task->thread.uw.fpsimd_state; unsigned int i; @@ -669,7 +669,8 @@ static void sve_to_fpsimd(struct task_struct *task) if (!system_supports_sve()) return;

- vq = sve_vq_from_vl(task_get_sve_vl(task)); + vl = thread_get_cur_vl(&task->thread); + vq = sve_vq_from_vl(vl); for (i = 0; i < SVE_NUM_ZREGS; ++i) { p = (__uint128_t const *)ZREG(sst, vq, i); fst->vregs[i] = arm64_le128_to_cpu(*p); @@ -717,6 +718,19 @@ void sve_alloc(struct task_struct *task) }

+/* + * Force the FPSIMD state shared with SVE to be updated in the SVE state + * even if the SVE state is the current active state. + * + * This should only be called by ptrace. task must be non-runnable. + * task->thread.sve_state must point to at least sve_state_size(task) + * bytes of allocated kernel memory. + */ +void fpsimd_force_sync_to_sve(struct task_struct *task) +{ + fpsimd_to_sve(task); +} + /* * Ensure that task->thread.sve_state is up to date with respect to * the user task, irrespective of when SVE is in use or not. @@ -727,7 +741,8 @@ void sve_alloc(struct task_struct *task) */ void fpsimd_sync_to_sve(struct task_struct *task) { - if (!test_tsk_thread_flag(task, TIF_SVE)) + if (!test_tsk_thread_flag(task, TIF_SVE) && + !thread_sm_enabled(&task->thread)) fpsimd_to_sve(task); }

@@ -741,7 +756,8 @@ void fpsimd_sync_to_sve(struct task_struct *task) */ void sve_sync_to_fpsimd(struct task_struct *task) { - if (test_tsk_thread_flag(task, TIF_SVE)) + if (test_tsk_thread_flag(task, TIF_SVE) || + thread_sm_enabled(&task->thread)) sve_to_fpsimd(task); }

@@ -766,7 +782,7 @@ void sve_sync_from_fpsimd_zeropad(struct task_struct *task) if (!test_tsk_thread_flag(task, TIF_SVE)) return;

- vq = sve_vq_from_vl(task_get_sve_vl(task)); + vq = sve_vq_from_vl(thread_get_cur_vl(&task->thread));

memset(sst, 0, SVE_SIG_REGS_SIZE(vq)); __fpsimd_to_sve(sst, fst, vq); @@ -810,8 +826,7 @@ int vec_set_vector_length(struct task_struct *task, enum vec_type type, /* * To ensure the FPSIMD bits of the SVE vector registers are preserved, * write any live register state back to task_struct, and convert to a - * regular FPSIMD thread. Since the vector length can only be changed - * with a syscall we can't be in streaming mode while reconfiguring. + * regular FPSIMD thread. */ if (task == current) { get_cpu_fpsimd_context(); diff --git a/arch/arm64/kernel/ptrace.c b/arch/arm64/kernel/ptrace.c index 230a47b9189e..60185c27b394 100644 --- a/arch/arm64/kernel/ptrace.c +++ b/arch/arm64/kernel/ptrace.c @@ -713,21 +713,51 @@ static int system_call_set(struct task_struct *target, #ifdef CONFIG_ARM64_SVE

static void sve_init_header_from_task(struct user_sve_header *header, - struct task_struct *target) + struct task_struct *target, + enum vec_type type) { unsigned int vq; + bool active; + bool fpsimd_only; + enum vec_type task_type;

memset(header, 0, sizeof(*header));

- header->flags = test_tsk_thread_flag(target, TIF_SVE) ? - SVE_PT_REGS_SVE : SVE_PT_REGS_FPSIMD; - if (test_tsk_thread_flag(target, TIF_SVE_VL_INHERIT)) - header->flags |= SVE_PT_VL_INHERIT; + /* Check if the requested registers are active for the task */ + if (thread_sm_enabled(&target->thread)) + task_type = ARM64_VEC_SME; + else + task_type = ARM64_VEC_SVE; + active = (task_type == type); + + switch (type) { + case ARM64_VEC_SVE: + if (test_tsk_thread_flag(target, TIF_SVE_VL_INHERIT)) + header->flags |= SVE_PT_VL_INHERIT; + fpsimd_only = !test_tsk_thread_flag(target, TIF_SVE); + break; + case ARM64_VEC_SME: + if (test_tsk_thread_flag(target, TIF_SME_VL_INHERIT)) + header->flags |= SVE_PT_VL_INHERIT; + fpsimd_only = false; + break; + default: + WARN_ON_ONCE(1); + return; + }

- header->vl = task_get_sve_vl(target); + if (active) { + if (fpsimd_only) { + header->flags |= SVE_PT_REGS_FPSIMD; + } else { + header->flags |= SVE_PT_REGS_SVE; + } + } + + header->vl = task_get_vl(target, type); vq = sve_vq_from_vl(header->vl);

- header->max_vl = sve_max_vl(); + header->max_vl = vec_max_vl(type); header->size = SVE_PT_SIZE(vq, header->flags); header->max_size = SVE_PT_SIZE(sve_vq_from_vl(header->max_vl), SVE_PT_REGS_SVE); @@ -738,19 +768,17 @@ static unsigned int sve_size_from_header(struct user_sve_header const *header) return ALIGN(header->size, SVE_VQ_BYTES); }

-static int sve_get(struct task_struct *target, - const struct user_regset *regset, - struct membuf to) +static int sve_get_common(struct task_struct *target, + const struct user_regset *regset, + struct membuf to, + enum vec_type type) { struct user_sve_header header; unsigned int vq; unsigned long start, end;

- if (!system_supports_sve()) - return -EINVAL; - /* Header */ - sve_init_header_from_task(&header, target); + sve_init_header_from_task(&header, target, type); vq = sve_vq_from_vl(header.vl);

membuf_write(&to, &header, sizeof(header)); @@ -758,49 +786,61 @@ static int sve_get(struct task_struct *target, if (target == current) fpsimd_preserve_current_state();

- /* Registers: FPSIMD-only case */ - BUILD_BUG_ON(SVE_PT_FPSIMD_OFFSET != sizeof(header)); - if ((header.flags & SVE_PT_REGS_MASK) == SVE_PT_REGS_FPSIMD) + BUILD_BUG_ON(SVE_PT_SVE_OFFSET != sizeof(header)); + + switch ((header.flags & SVE_PT_REGS_MASK)) { + case SVE_PT_REGS_FPSIMD: return __fpr_get(target, regset, to);

- /* Otherwise: full SVE case */ + case SVE_PT_REGS_SVE: + start = SVE_PT_SVE_OFFSET; + end = SVE_PT_SVE_FFR_OFFSET(vq) + SVE_PT_SVE_FFR_SIZE(vq); + membuf_write(&to, target->thread.sve_state, end - start);

- BUILD_BUG_ON(SVE_PT_SVE_OFFSET != sizeof(header)); - start = SVE_PT_SVE_OFFSET; - end = SVE_PT_SVE_FFR_OFFSET(vq) + SVE_PT_SVE_FFR_SIZE(vq); - membuf_write(&to, target->thread.sve_state, end - start); + start = end; + end = SVE_PT_SVE_FPSR_OFFSET(vq); + membuf_zero(&to, end - start);

- start = end; - end = SVE_PT_SVE_FPSR_OFFSET(vq); - membuf_zero(&to, end - start); + /* + * Copy fpsr, and fpcr which must follow contiguously in + * struct fpsimd_state: + */ + start = end; + end = SVE_PT_SVE_FPCR_OFFSET(vq) + SVE_PT_SVE_FPCR_SIZE; + membuf_write(&to, &target->thread.uw.fpsimd_state.fpsr, + end - start);

- /* - * Copy fpsr, and fpcr which must follow contiguously in - * struct fpsimd_state: - */ - start = end; - end = SVE_PT_SVE_FPCR_OFFSET(vq) + SVE_PT_SVE_FPCR_SIZE; - membuf_write(&to, &target->thread.uw.fpsimd_state.fpsr, end - start); + start = end; + end = sve_size_from_header(&header); + return membuf_zero(&to, end - start);

- start = end; - end = sve_size_from_header(&header); - return membuf_zero(&to, end - start); + default: + return 0; + } }

-static int sve_set(struct task_struct *target, +static int sve_get(struct task_struct *target, const struct user_regset *regset, - unsigned int pos, unsigned int count, - const void *kbuf, const void __user *ubuf) + struct membuf to) +{ + if (!system_supports_sve()) + return -EINVAL; + + return sve_get_common(target, regset, to, ARM64_VEC_SVE); +} + +static int sve_set_common(struct task_struct *target, + const struct user_regset *regset, + unsigned int pos, unsigned int count, + const void *kbuf, const void __user *ubuf, + enum vec_type type) { int ret; struct user_sve_header header; unsigned int vq; unsigned long start, end;

- if (!system_supports_sve()) - return -EINVAL; - /* Header */ if (count < sizeof(header)) return -EINVAL; @@ -813,13 +853,37 @@ static int sve_set(struct task_struct *target, * Apart from SVE_PT_REGS_MASK, all SVE_PT_* flags are consumed by * vec_set_vector_length(), which will also validate them for us: */ - ret = vec_set_vector_length(target, ARM64_VEC_SVE, header.vl, + ret = vec_set_vector_length(target, type, header.vl, ((unsigned long)header.flags & ~SVE_PT_REGS_MASK) << 16); if (ret) goto out;

/* Actual VL set may be less than the user asked for: */ - vq = sve_vq_from_vl(task_get_sve_vl(target)); + vq = sve_vq_from_vl(task_get_vl(target, type)); + + /* Enter/exit streaming mode */ + if (system_supports_sme()) { + u64 old_svcr = target->thread.svcr; + + switch (type) { + case ARM64_VEC_SVE: + target->thread.svcr &= ~SYS_SVCR_EL0_SM_MASK; + break; + case ARM64_VEC_SME: + target->thread.svcr |= SYS_SVCR_EL0_SM_MASK; + break; + default: + WARN_ON_ONCE(1); + return -EINVAL; + } + + /* + * If we switched then invalidate any existing SVE + * state and ensure there's storage. + */ + if (target->thread.svcr != old_svcr) + sve_alloc(target); + }

/* Registers: FPSIMD-only case */

@@ -828,10 +892,15 @@ static int sve_set(struct task_struct *target, ret = __fpr_set(target, regset, pos, count, kbuf, ubuf, SVE_PT_FPSIMD_OFFSET); clear_tsk_thread_flag(target, TIF_SVE); + if (type == ARM64_VEC_SME) + fpsimd_force_sync_to_sve(target); goto out; }

- /* Otherwise: full SVE case */ + /* + * Otherwise: no registers or full SVE case. For backwards + * compatibility reasons we treat empty flags as SVE registers. + */

/* * If setting a different VL from the requested VL and there is @@ -852,8 +921,9 @@ static int sve_set(struct task_struct *target,

/* * Ensure target->thread.sve_state is up to date with target's - * FPSIMD regs, so that a short copyin leaves trailing registers - * unmodified. + * FPSIMD regs, so that a short copyin leaves trailing + * registers unmodified. Always enable SVE even if going into + * streaming mode. */ fpsimd_sync_to_sve(target); set_tsk_thread_flag(target, TIF_SVE); @@ -889,8 +959,46 @@ static int sve_set(struct task_struct *target, return ret; }

+static int sve_set(struct task_struct *target, + const struct user_regset *regset, + unsigned int pos, unsigned int count, + const void *kbuf, const void __user *ubuf) +{ + if (!system_supports_sve()) + return -EINVAL; + + return sve_set_common(target, regset, pos, count, kbuf, ubuf, + ARM64_VEC_SVE); +} + #endif /* CONFIG_ARM64_SVE */

+#ifdef CONFIG_ARM64_SME + +static int ssve_get(struct task_struct *target, + const struct user_regset *regset, + struct membuf to) +{ + if (!system_supports_sme()) + return -EINVAL; + + return sve_get_common(target, regset, to, ARM64_VEC_SME); +} + +static int ssve_set(struct task_struct *target, + const struct user_regset *regset, + unsigned int pos, unsigned int count, + const void *kbuf, const void __user *ubuf) +{ + if (!system_supports_sme()) + return -EINVAL; + + return sve_set_common(target, regset, pos, count, kbuf, ubuf, + ARM64_VEC_SME); +} + +#endif /* CONFIG_ARM64_SME */ + #ifdef CONFIG_ARM64_PTR_AUTH static int pac_mask_get(struct task_struct *target, const struct user_regset *regset, @@ -1108,6 +1216,9 @@ enum aarch64_regset { #ifdef CONFIG_ARM64_SVE REGSET_SVE, #endif +#ifdef CONFIG_ARM64_SVE + REGSET_SSVE, +#endif #ifdef CONFIG_ARM64_PTR_AUTH REGSET_PAC_MASK, REGSET_PAC_ENABLED_KEYS, @@ -1188,6 +1299,17 @@ static const struct user_regset aarch64_regsets[] = { .set = sve_set, }, #endif +#ifdef CONFIG_ARM64_SME + [REGSET_SSVE] = { /* Streaming mode SVE */ + .core_note_type = NT_ARM_SSVE, + .n = DIV_ROUND_UP(SVE_PT_SIZE(SVE_VQ_MAX, SVE_PT_REGS_SVE), + SVE_VQ_BYTES), + .size = SVE_VQ_BYTES, + .align = SVE_VQ_BYTES, + .regset_get = ssve_get, + .set = ssve_set, + }, +#endif #ifdef CONFIG_ARM64_PTR_AUTH [REGSET_PAC_MASK] = { .core_note_type = NT_ARM_PAC_MASK, diff --git a/include/uapi/linux/elf.h b/include/uapi/linux/elf.h index 787c657bfae8..a8dc688e1826 100644 --- a/include/uapi/linux/elf.h +++ b/include/uapi/linux/elf.h @@ -431,6 +431,7 @@ typedef struct elf64_shdr { #define NT_ARM_PACG_KEYS 0x408 /* ARM pointer authentication generic key */ #define NT_ARM_TAGGED_ADDR_CTRL 0x409 /* arm64 tagged address control (prctl()) */ #define NT_ARM_PAC_ENABLED_KEYS 0x40a /* arm64 ptr auth enabled keys (prctl()) */ +#define NT_ARM_SSVE 0x40b /* ARM Streaming SVE registers */ #define NT_ARC_V2 0x600 /* ARCv2 accumulator/extra registers */ #define NT_VMCOREDD 0x700 /* Vmcore Device Dump Note */ #define NT_MIPS_DSP 0x800 /* MIPS DSP ASE registers */

Both streaming mode and ZA may increase power consumption when they are enabled and streaming mode makes many FPSIMD and SVE instructions undefined which will cause problems for any kernel mode floating point so disable both when we flush the CPU state. This covers both kernel_neon_begin() and idle and after flushing the state a reload is always required anyway.

Signed-off-by: Mark Brown broonie@kernel.org Reviewed-by: Catalin Marinas catalin.marinas@arm.com --- arch/arm64/kernel/fpsimd.c | 9 +++++++++ 1 file changed, 9 insertions(+)

diff --git a/arch/arm64/kernel/fpsimd.c b/arch/arm64/kernel/fpsimd.c index 94f06e9d37cf..9592cdd7d635 100644 --- a/arch/arm64/kernel/fpsimd.c +++ b/arch/arm64/kernel/fpsimd.c @@ -1759,6 +1759,15 @@ static void fpsimd_flush_cpu_state(void) { WARN_ON(!system_supports_fpsimd()); __this_cpu_write(fpsimd_last_state.st, NULL); + + /* + * Leaving streaming mode enabled will cause issues for any kernel + * NEON and leaving streaming mode or ZA enabled may increase power + * consumption. + */ + if (system_supports_sme()) + sme_smstop(); + set_thread_flag(TIF_FOREIGN_FPSTATE); }

For the time being we do not support use of SME by KVM guests, support for this will be enabled in future. In order to prevent any side effects or side channels via the new system registers, including the EL0 read/write register TPIDR2, explicitly undefine all the system registers added by SME and mask out the SME bitfield in SYS_ID_AA64PFR1.

Signed-off-by: Mark Brown broonie@kernel.org Reviewed-by: Catalin Marinas catalin.marinas@arm.com --- arch/arm64/kvm/sys_regs.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-)

diff --git a/arch/arm64/kvm/sys_regs.c b/arch/arm64/kvm/sys_regs.c index 7b45c040cc27..689e53dd4cb1 100644 --- a/arch/arm64/kvm/sys_regs.c +++ b/arch/arm64/kvm/sys_regs.c @@ -1132,6 +1132,8 @@ static u64 read_id_reg(const struct kvm_vcpu *vcpu, case SYS_ID_AA64PFR1_EL1: if (!kvm_has_mte(vcpu->kvm)) val &= ~ARM64_FEATURE_MASK(ID_AA64PFR1_MTE); + + val &= ~ARM64_FEATURE_MASK(ID_AA64PFR1_SME); break; case SYS_ID_AA64ISAR1_EL1: if (!vcpu_has_ptrauth(vcpu)) @@ -1553,7 +1555,7 @@ static const struct sys_reg_desc sys_reg_descs[] = { ID_UNALLOCATED(4,2), ID_UNALLOCATED(4,3), ID_SANITISED(ID_AA64ZFR0_EL1), - ID_UNALLOCATED(4,5), + ID_HIDDEN(ID_AA64SMFR0_EL1), ID_UNALLOCATED(4,6), ID_UNALLOCATED(4,7),

@@ -1596,6 +1598,8 @@ static const struct sys_reg_desc sys_reg_descs[] = {

{ SYS_DESC(SYS_ZCR_EL1), NULL, reset_val, ZCR_EL1, 0, .visibility = sve_visibility }, { SYS_DESC(SYS_TRFCR_EL1), undef_access }, + { SYS_DESC(SYS_SMPRI_EL1), undef_access }, + { SYS_DESC(SYS_SMCR_EL1), undef_access }, { SYS_DESC(SYS_TTBR0_EL1), access_vm_reg, reset_unknown, TTBR0_EL1 }, { SYS_DESC(SYS_TTBR1_EL1), access_vm_reg, reset_unknown, TTBR1_EL1 }, { SYS_DESC(SYS_TCR_EL1), access_vm_reg, reset_val, TCR_EL1, 0 }, @@ -1678,8 +1682,10 @@ static const struct sys_reg_desc sys_reg_descs[] = {

{ SYS_DESC(SYS_CCSIDR_EL1), access_ccsidr }, { SYS_DESC(SYS_CLIDR_EL1), access_clidr }, + { SYS_DESC(SYS_SMIDR_EL1), undef_access }, { SYS_DESC(SYS_CSSELR_EL1), access_csselr, reset_unknown, CSSELR_EL1 }, { SYS_DESC(SYS_CTR_EL0), access_ctr }, + { SYS_DESC(SYS_SVCR_EL0), undef_access },

{ PMU_SYS_REG(SYS_PMCR_EL0), .access = access_pmcr, .reset = reset_pmcr, .reg = PMCR_EL0 }, @@ -1719,6 +1725,7 @@ static const struct sys_reg_desc sys_reg_descs[] = {

{ SYS_DESC(SYS_TPIDR_EL0), NULL, reset_unknown, TPIDR_EL0 }, { SYS_DESC(SYS_TPIDRRO_EL0), NULL, reset_unknown, TPIDRRO_EL0 }, + { SYS_DESC(SYS_TPIDR2_EL0), undef_access },

{ SYS_DESC(SYS_SCXTNUM_EL0), undef_access },

While we don't currently support SME in guests we do currently support it for the host system so we need to take care of SME's impact, including the floating point register state, when running guests. Simiarly to SVE we need to manage the traps in CPACR_RL1, what is new is the handling of streaming mode and ZA.

Normally we defer any handling of the floating point register state until the guest first uses it however if the system is in streaming mode FPSIMD and SVE operations may generate SME traps which we would need to distinguish from actual attempts by the guest to use SME. Rather than do this for the time being if we are in streaming mode when entering the guest we force the floating point state to be saved immediately and exit streaming mode, meaning that the guest won't generate SME traps for supported operations.

We could handle ZA in the access trap similarly to the FPSIMD/SVE state without the disruption caused by streaming mode but for simplicity handle it the same way as streaming mode for now.

This will be revisited when we support SME for guests (hopefully before SME hardware becomes available), for now it will only incur additional cost on systems with SME and even there only if streaming mode or ZA are enabled.

Signed-off-by: Mark Brown broonie@kernel.org Reviewed-by: Catalin Marinas catalin.marinas@arm.com --- arch/arm64/include/asm/kvm_host.h | 1 + arch/arm64/kvm/fpsimd.c | 36 +++++++++++++++++++++++++++++++ 2 files changed, 37 insertions(+)

diff --git a/arch/arm64/include/asm/kvm_host.h b/arch/arm64/include/asm/kvm_host.h index 8a7c442d5b57..f8f0d30dd1a2 100644 --- a/arch/arm64/include/asm/kvm_host.h +++ b/arch/arm64/include/asm/kvm_host.h @@ -454,6 +454,7 @@ struct kvm_vcpu_arch { #define KVM_ARM64_DEBUG_STATE_SAVE_TRBE (1 << 13) /* Save TRBE context if active */ #define KVM_ARM64_FP_FOREIGN_FPSTATE (1 << 14) #define KVM_ARM64_ON_UNSUPPORTED_CPU (1 << 15) /* Physical CPU not in supported_cpus */ +#define KVM_ARM64_HOST_SME_ENABLED (1 << 16) /* SME enabled for EL0 */

#define KVM_GUESTDBG_VALID_MASK (KVM_GUESTDBG_ENABLE | \ KVM_GUESTDBG_USE_SW_BP | \ diff --git a/arch/arm64/kvm/fpsimd.c b/arch/arm64/kvm/fpsimd.c index 57d7ac3cfa0c..441edb9c398c 100644 --- a/arch/arm64/kvm/fpsimd.c +++ b/arch/arm64/kvm/fpsimd.c @@ -82,6 +82,26 @@ void kvm_arch_vcpu_load_fp(struct kvm_vcpu *vcpu)

if (read_sysreg(cpacr_el1) & CPACR_EL1_ZEN_EL0EN) vcpu->arch.flags |= KVM_ARM64_HOST_SVE_ENABLED; + + /* + * We don't currently support SME guests but if we leave + * things in streaming mode then when the guest starts running + * FPSIMD or SVE code it may generate SME traps so as a + * special case if we are in streaming mode we force the host + * state to be saved now and exit streaming mode so that we + * don't have to handle any SME traps for valid guest + * operations. Do this for ZA as well for now for simplicity. + */ + if (system_supports_sme()) { + if (read_sysreg(cpacr_el1) & CPACR_EL1_SMEN_EL0EN) + vcpu->arch.flags |= KVM_ARM64_HOST_SME_ENABLED; + + if (read_sysreg_s(SYS_SVCR_EL0) & + (SYS_SVCR_EL0_SM_MASK | SYS_SVCR_EL0_ZA_MASK)) { + vcpu->arch.flags &= ~KVM_ARM64_FP_HOST; + fpsimd_save_and_flush_cpu_state(); + } + } }

/* @@ -135,6 +155,22 @@ void kvm_arch_vcpu_put_fp(struct kvm_vcpu *vcpu)

local_irq_save(flags);

+ /* + * If we have VHE then the Hyp code will reset CPACR_EL1 to + * CPACR_EL1_DEFAULT and we need to reenable SME. + */ + if (has_vhe() && system_supports_sme()) { + /* Also restore EL0 state seen on entry */ + if (vcpu->arch.flags & KVM_ARM64_HOST_SME_ENABLED) + sysreg_clear_set(CPACR_EL1, 0, + CPACR_EL1_SMEN_EL0EN | + CPACR_EL1_SMEN_EL1EN); + else + sysreg_clear_set(CPACR_EL1, + CPACR_EL1_SMEN_EL0EN, + CPACR_EL1_SMEN_EL1EN); + } + if (vcpu->arch.flags & KVM_ARM64_FP_ENABLED) { if (vcpu_has_sve(vcpu)) { __vcpu_sys_reg(vcpu, ZCR_EL1) = read_sysreg_el1(SYS_ZCR);

As for the kernel so that we don't have ambitious toolchain requirements to build the tests manually encode some of the SVE instructions.

Signed-off-by: Mark Brown broonie@kernel.org Reviewed-by: Shuah Khan skhan@linuxfoundation.org Acked-by: Catalin Marinas catalin.marinas@arm.com --- tools/testing/selftests/arm64/fp/sme-inst.h | 51 +++++++++++++++++++++ 1 file changed, 51 insertions(+) create mode 100644 tools/testing/selftests/arm64/fp/sme-inst.h

diff --git a/tools/testing/selftests/arm64/fp/sme-inst.h b/tools/testing/selftests/arm64/fp/sme-inst.h new file mode 100644 index 000000000000..7191e53ca1c0 --- /dev/null +++ b/tools/testing/selftests/arm64/fp/sme-inst.h @@ -0,0 +1,51 @@ +// SPDX-License-Identifier: GPL-2.0-only +// Copyright (C) 2021-2 ARM Limited. +// Original author: Mark Brown broonie@kernel.org + +#ifndef SME_INST_H +#define SME_INST_H + +/* + * RDSVL X\nx, #\imm + */ +.macro rdsvl nx, imm + .inst 0x4bf5800 \ + | (\imm << 5) \ + | (\nx) +.endm + +.macro smstop + msr S0_3_C4_C6_3, xzr +.endm + +.macro smstart_za + msr S0_3_C4_C5_3, xzr +.endm + +.macro smstart_sm + msr S0_3_C4_C3_3, xzr +.endm + +/* + * LDR (vector to ZA array): + * LDR ZA[\nw, #\offset], [X\nxbase, #\offset, MUL VL] + */ +.macro _ldr_za nw, nxbase, offset=0 + .inst 0xe1000000 \ + | (((\nw) & 3) << 13) \ + | ((\nxbase) << 5) \ + | ((\offset) & 7) +.endm + +/* + * STR (vector from ZA array): + * STR ZA[\nw, #\offset], [X\nxbase, #\offset, MUL VL] + */ +.macro _str_za nw, nxbase, offset=0 + .inst 0xe1200000 \ + | (((\nw) & 3) << 13) \ + | ((\nxbase) << 5) \ + | ((\offset) & 7) +.endm + +#endif

The Scalable Matrix Extension adds a new system register TPIDR2 intended to be used by libc for its own thread specific use, add some kselftests which exercise the ABI for it.

Since this test should with some adjustment work for TPIDR and any other similar registers added in future add tests for it in a separate directory rather than placing it with the other floating point tests, nothing existing looked suitable so I created a new test directory called "abi".

Since this feature is intended to be used by libc the test is built as freestanding code using nolibc so we don't end up with the test program and libc both trying to manage the register simultaneously and distrupting each other. As a result of being written using nolibc rather than using hwcaps to identify if SME is available in the system we check for the default SME vector length configuration in proc, adding hwcap support to nolibc seems like disproportionate effort and didn't feel entirely idiomatic for what nolibc is trying to do.

Signed-off-by: Mark Brown broonie@kernel.org Reviewed-by: Shuah Khan skhan@linuxfoundation.org Acked-by: Catalin Marinas catalin.marinas@arm.com --- tools/testing/selftests/arm64/abi/.gitignore | 1 + tools/testing/selftests/arm64/abi/Makefile | 9 +- tools/testing/selftests/arm64/abi/tpidr2.c | 298 +++++++++++++++++++ 3 files changed, 307 insertions(+), 1 deletion(-) create mode 100644 tools/testing/selftests/arm64/abi/tpidr2.c

diff --git a/tools/testing/selftests/arm64/abi/.gitignore b/tools/testing/selftests/arm64/abi/.gitignore index b79cf5814c23..b9e54417250d 100644 --- a/tools/testing/selftests/arm64/abi/.gitignore +++ b/tools/testing/selftests/arm64/abi/.gitignore @@ -1 +1,2 @@ syscall-abi +tpidr2 diff --git a/tools/testing/selftests/arm64/abi/Makefile b/tools/testing/selftests/arm64/abi/Makefile index 96eba974ac8d..c8d7f2495eb2 100644 --- a/tools/testing/selftests/arm64/abi/Makefile +++ b/tools/testing/selftests/arm64/abi/Makefile @@ -1,8 +1,15 @@ # SPDX-License-Identifier: GPL-2.0 # Copyright (C) 2021 ARM Limited

-TEST_GEN_PROGS := syscall-abi +TEST_GEN_PROGS := syscall-abi tpidr2

include ../../lib.mk

$(OUTPUT)/syscall-abi: syscall-abi.c syscall-abi-asm.S + +# Build with nolibc since TPIDR2 is intended to be actively managed by +# libc and we're trying to test the functionality that it depends on here. +$(OUTPUT)/tpidr2: tpidr2.c + $(CC) -fno-asynchronous-unwind-tables -fno-ident -s -Os -nostdlib \ + -static -include ../../../../include/nolibc/nolibc.h \ + -ffreestanding -Wall $^ -o $@ -lgcc diff --git a/tools/testing/selftests/arm64/abi/tpidr2.c b/tools/testing/selftests/arm64/abi/tpidr2.c new file mode 100644 index 000000000000..351a098b503a --- /dev/null +++ b/tools/testing/selftests/arm64/abi/tpidr2.c @@ -0,0 +1,298 @@ +// SPDX-License-Identifier: GPL-2.0-only + +#include <linux/sched.h> +#include <linux/wait.h> + +#define SYS_TPIDR2 "S3_3_C13_C0_5" + +#define EXPECTED_TESTS 5 + +static void putstr(const char *str) +{ + write(1, str, strlen(str)); +} + +static void putnum(unsigned int num) +{ + char c; + + if (num / 10) + putnum(num / 10); + + c = '0' + (num % 10); + write(1, &c, 1); +} + +static int tests_run; +static int tests_passed; +static int tests_failed; +static int tests_skipped; + +static void set_tpidr2(uint64_t val) +{ + asm volatile ( + "msr " SYS_TPIDR2 ", %0\n" + : + : "r"(val) + : "cc"); +} + +static uint64_t get_tpidr2(void) +{ + uint64_t val; + + asm volatile ( + "mrs %0, " SYS_TPIDR2 "\n" + : "=r"(val) + : + : "cc"); + + return val; +} + +static void print_summary(void) +{ + if (tests_passed + tests_failed + tests_skipped != EXPECTED_TESTS) + putstr("# UNEXPECTED TEST COUNT: "); + + putstr("# Totals: pass:"); + putnum(tests_passed); + putstr(" fail:"); + putnum(tests_failed); + putstr(" xfail:0 xpass:0 skip:"); + putnum(tests_skipped); + putstr(" error:0\n"); +} + +/* Processes should start with TPIDR2 == 0 */ +static int default_value(void) +{ + return get_tpidr2() == 0; +} + +/* If we set TPIDR2 we should read that value */ +static int write_read(void) +{ + set_tpidr2(getpid()); + + return getpid() == get_tpidr2(); +} + +/* If we set a value we should read the same value after scheduling out */ +static int write_sleep_read(void) +{ + set_tpidr2(getpid()); + + msleep(100); + + return getpid() == get_tpidr2(); +} + +/* + * If we fork the value in the parent should be unchanged and the + * child should start with the same value and be able to set its own + * value. + */ +static int write_fork_read(void) +{ + pid_t newpid, waiting, oldpid; + int status; + + set_tpidr2(getpid()); + + oldpid = getpid(); + newpid = fork(); + if (newpid == 0) { + /* In child */ + if (get_tpidr2() != oldpid) { + putstr("# TPIDR2 changed in child: "); + putnum(get_tpidr2()); + putstr("\n"); + exit(0); + } + + set_tpidr2(getpid()); + if (get_tpidr2() == getpid()) { + exit(1); + } else { + putstr("# Failed to set TPIDR2 in child\n"); + exit(0); + } + } + if (newpid < 0) { + putstr("# fork() failed: -"); + putnum(-newpid); + putstr("\n"); + return 0; + } + + for (;;) { + waiting = waitpid(newpid, &status, 0); + + if (waiting < 0) { + if (errno == EINTR) + continue; + putstr("# waitpid() failed: "); + putnum(errno); + putstr("\n"); + return 0; + } + if (waiting != newpid) { + putstr("# waitpid() returned wrong PID\n"); + return 0; + } + + if (!WIFEXITED(status)) { + putstr("# child did not exit\n"); + return 0; + } + + if (getpid() != get_tpidr2()) { + putstr("# TPIDR2 corrupted in parent\n"); + return 0; + } + + return WEXITSTATUS(status); + } +} + +/* + * sys_clone() has a lot of per architecture variation so just define + * it here rather than adding it to nolibc, plus the raw API is a + * little more convenient for this test. + */ +static int sys_clone(unsigned long clone_flags, unsigned long newsp, + int *parent_tidptr, unsigned long tls, + int *child_tidptr) +{ + return my_syscall5(__NR_clone, clone_flags, newsp, parent_tidptr, tls, + child_tidptr); +} + +/* + * If we clone with CLONE_SETTLS then the value in the parent should + * be unchanged and the child should start with zero and be able to + * set its own value. + */ +static int write_clone_read(void) +{ + int parent_tid, child_tid; + pid_t parent, waiting; + int ret, status; + + parent = getpid(); + set_tpidr2(parent); + + ret = sys_clone(CLONE_SETTLS, 0, &parent_tid, 0, &child_tid); + if (ret == -1) { + putstr("# clone() failed\n"); + putnum(errno); + putstr("\n"); + return 0; + } + + if (ret == 0) { + /* In child */ + if (get_tpidr2() != 0) { + putstr("# TPIDR2 non-zero in child: "); + putnum(get_tpidr2()); + putstr("\n"); + exit(0); + } + + if (gettid() == 0) + putstr("# Child TID==0\n"); + set_tpidr2(gettid()); + if (get_tpidr2() == gettid()) { + exit(1); + } else { + putstr("# Failed to set TPIDR2 in child\n"); + exit(0); + } + } + + for (;;) { + waiting = wait4(ret, &status, __WCLONE, NULL); + + if (waiting < 0) { + if (errno == EINTR) + continue; + putstr("# wait4() failed: "); + putnum(errno); + putstr("\n"); + return 0; + } + if (waiting != ret) { + putstr("# wait4() returned wrong PID "); + putnum(waiting); + putstr("\n"); + return 0; + } + + if (!WIFEXITED(status)) { + putstr("# child did not exit\n"); + return 0; + } + + if (parent != get_tpidr2()) { + putstr("# TPIDR2 corrupted in parent\n"); + return 0; + } + + return WEXITSTATUS(status); + } +} + +#define run_test(name) \ + if (name()) { \ + tests_passed++; \ + } else { \ + tests_failed++; \ + putstr("not "); \ + } \ + putstr("ok "); \ + putnum(++tests_run); \ + putstr(" " #name "\n"); + +int main(int argc, char **argv) +{ + int ret, i; + + putstr("TAP version 13\n"); + putstr("1.."); + putnum(EXPECTED_TESTS); + putstr("\n"); + + putstr("# PID: "); + putnum(getpid()); + putstr("\n"); + + /* + * This test is run with nolibc which doesn't support hwcap and + * it's probably disproportionate to implement so instead check + * for the default vector length configuration in /proc. + */ + ret = open("/proc/sys/abi/sme_default_vector_length", O_RDONLY, 0); + if (ret >= 0) { + run_test(default_value); + run_test(write_read); + run_test(write_sleep_read); + run_test(write_fork_read); + run_test(write_clone_read); + + } else { + putstr("# SME support not present\n"); + + for (i = 0; i < EXPECTED_TESTS; i++) { + putstr("ok "); + putnum(i); + putstr(" skipped, TPIDR2 not supported\n"); + } + + tests_skipped += EXPECTED_TESTS; + } + + print_summary(); + + return 0; +}

One of the features of SME is the addition of streaming mode, in which we have access to a set of streaming mode SVE registers at the SME vector length. Since these are accessed using the SVE instructions let's reuse the existing SVE stress test for testing with a compile time option for controlling the few small differences needed:

- Enter streaming mode immediately on starting the program. - In streaming mode FFR is removed so skip reading and writing FFR.

Signed-off-by: Mark Brown broonie@kernel.org Reviewed-by: Shuah Khan skhan@linuxfoundation.org Acked-by: Catalin Marinas catalin.marinas@arm.com --- tools/testing/selftests/arm64/fp/.gitignore | 1 + tools/testing/selftests/arm64/fp/Makefile | 3 + tools/testing/selftests/arm64/fp/ssve-stress | 59 ++++++++++++++++++++ tools/testing/selftests/arm64/fp/sve-test.S | 20 +++++++ 4 files changed, 83 insertions(+) create mode 100644 tools/testing/selftests/arm64/fp/ssve-stress

diff --git a/tools/testing/selftests/arm64/fp/.gitignore b/tools/testing/selftests/arm64/fp/.gitignore index 6e9a610c5e5d..5729a5b1adfc 100644 --- a/tools/testing/selftests/arm64/fp/.gitignore +++ b/tools/testing/selftests/arm64/fp/.gitignore @@ -5,5 +5,6 @@ rdvl-sve sve-probe-vls sve-ptrace sve-test +ssve-test vec-syscfg vlset diff --git a/tools/testing/selftests/arm64/fp/Makefile b/tools/testing/selftests/arm64/fp/Makefile index a224fff8082b..e6643c9b0474 100644 --- a/tools/testing/selftests/arm64/fp/Makefile +++ b/tools/testing/selftests/arm64/fp/Makefile @@ -5,6 +5,7 @@ TEST_GEN_PROGS := sve-ptrace sve-probe-vls vec-syscfg TEST_PROGS_EXTENDED := fp-pidbench fpsimd-test fpsimd-stress \ rdvl-sme rdvl-sve \ sve-test sve-stress \ + ssve-test ssve-stress \ vlset

all: $(TEST_GEN_PROGS) $(TEST_PROGS_EXTENDED) @@ -19,6 +20,8 @@ sve-ptrace: sve-ptrace.o sve-probe-vls: sve-probe-vls.o rdvl.o sve-test: sve-test.o asm-utils.o $(CC) -nostdlib $^ -o $@ +ssve-test: sve-test.S asm-utils.o + $(CC) -DSSVE -nostdlib $^ -o $@ vec-syscfg: vec-syscfg.o rdvl.o vlset: vlset.o

diff --git a/tools/testing/selftests/arm64/fp/ssve-stress b/tools/testing/selftests/arm64/fp/ssve-stress new file mode 100644 index 000000000000..e2bd2cc184ad --- /dev/null +++ b/tools/testing/selftests/arm64/fp/ssve-stress @@ -0,0 +1,59 @@ +#!/bin/bash +# SPDX-License-Identifier: GPL-2.0-only +# Copyright (C) 2015-2019 ARM Limited. +# Original author: Dave Martin Dave.Martin@arm.com + +set -ue + +NR_CPUS=`nproc` + +pids= +logs= + +cleanup () { + trap - INT TERM CHLD + set +e + + if [ -n "$pids" ]; then + kill $pids + wait $pids + pids= + fi + + if [ -n "$logs" ]; then + cat $logs + rm $logs + logs= + fi +} + +interrupt () { + cleanup + exit 0 +} + +child_died () { + cleanup + exit 1 +} + +trap interrupt INT TERM EXIT + +for x in `seq 0 $((NR_CPUS * 4))`; do + log=`mktemp` + logs=$logs\ $log + ./ssve-test >$log & + pids=$pids\ $! +done + +# Wait for all child processes to be created: +sleep 10 + +while :; do + kill -USR1 $pids +done & +pids=$pids\ $! + +wait + +exit 1 diff --git a/tools/testing/selftests/arm64/fp/sve-test.S b/tools/testing/selftests/arm64/fp/sve-test.S index f5b1b48ffff2..589264231a2d 100644 --- a/tools/testing/selftests/arm64/fp/sve-test.S +++ b/tools/testing/selftests/arm64/fp/sve-test.S @@ -13,6 +13,7 @@ #include <asm/unistd.h> #include "assembler.h" #include "asm-offsets.h" +#include "sme-inst.h"

#define NZR 32 #define NPR 16 @@ -156,6 +157,7 @@ endfunction // We fill the upper lanes of FFR with zeros. // Beware: corrupts P0. function setup_ffr +#ifndef SSVE mov x4, x30

and w0, w0, #0x3 @@ -178,6 +180,9 @@ function setup_ffr wrffr p0.b

ret x4 +#else + ret +#endif endfunction

// Trivial memory compare: compare x2 bytes starting at address x0 with @@ -260,6 +265,7 @@ endfunction // Beware -- corrupts P0. // Clobbers x0-x5. function check_ffr +#ifndef SSVE mov x3, x30

ldr x4, =scratch @@ -280,6 +286,9 @@ function check_ffr mov x2, x5 mov x30, x3 b memcmp +#else + ret +#endif endfunction

// Any SVE register modified here can cause corruption in the main @@ -295,10 +304,12 @@ function irritator_handler movi v0.8b, #1 movi v9.16b, #2 movi v31.8b, #3 +#ifndef SSVE // And P0 rdffr p0.b // And FFR wrffr p15.b +#endif

ret endfunction @@ -359,6 +370,11 @@ endfunction .globl _start function _start _start: +#ifdef SSVE + puts "Streaming mode " + smstart_sm +#endif + // Sanity-check and report the vector length

rdvl x19, #8 @@ -407,6 +423,10 @@ _start: orr w2, w2, #SA_NODEFER bl setsignal

+#ifdef SSVE + smstart_sm // syscalls will have exited streaming mode +#endif + mov x22, #0 // generation number, increments per iteration .Ltest_loop: rdvl x0, #8

Add a stress test for context switching of the ZA register state based on the similar tests Dave Martin wrote for FPSIMD and SVE registers. The test loops indefinitely writing a data pattern to ZA then reading it back and verifying that it's what was expected.

Unlike the other tests we manually assemble the SME instructions since at present no released toolchain has SME support integrated.

Signed-off-by: Mark Brown broonie@kernel.org Reviewed-by: Shuah Khan skhan@linuxfoundation.org Acked-by: Catalin Marinas catalin.marinas@arm.com --- tools/testing/selftests/arm64/fp/.gitignore | 1 + tools/testing/selftests/arm64/fp/Makefile | 3 + tools/testing/selftests/arm64/fp/za-stress | 59 +++ tools/testing/selftests/arm64/fp/za-test.S | 388 ++++++++++++++++++++ 4 files changed, 451 insertions(+) create mode 100644 tools/testing/selftests/arm64/fp/za-stress create mode 100644 tools/testing/selftests/arm64/fp/za-test.S

diff --git a/tools/testing/selftests/arm64/fp/.gitignore b/tools/testing/selftests/arm64/fp/.gitignore index 5729a5b1adfc..ead3197e720b 100644 --- a/tools/testing/selftests/arm64/fp/.gitignore +++ b/tools/testing/selftests/arm64/fp/.gitignore @@ -8,3 +8,4 @@ sve-test ssve-test vec-syscfg vlset +za-test diff --git a/tools/testing/selftests/arm64/fp/Makefile b/tools/testing/selftests/arm64/fp/Makefile index e6643c9b0474..38d2d0d5a0eb 100644 --- a/tools/testing/selftests/arm64/fp/Makefile +++ b/tools/testing/selftests/arm64/fp/Makefile @@ -6,6 +6,7 @@ TEST_PROGS_EXTENDED := fp-pidbench fpsimd-test fpsimd-stress \ rdvl-sme rdvl-sve \ sve-test sve-stress \ ssve-test ssve-stress \ + za-test za-stress \ vlset

all: $(TEST_GEN_PROGS) $(TEST_PROGS_EXTENDED) @@ -24,5 +25,7 @@ ssve-test: sve-test.S asm-utils.o $(CC) -DSSVE -nostdlib $^ -o $@ vec-syscfg: vec-syscfg.o rdvl.o vlset: vlset.o +za-test: za-test.o asm-utils.o + $(CC) -nostdlib $^ -o $@

include ../../lib.mk diff --git a/tools/testing/selftests/arm64/fp/za-stress b/tools/testing/selftests/arm64/fp/za-stress new file mode 100644 index 000000000000..5ac386b55b95 --- /dev/null +++ b/tools/testing/selftests/arm64/fp/za-stress @@ -0,0 +1,59 @@ +#!/bin/bash +# SPDX-License-Identifier: GPL-2.0-only +# Copyright (C) 2015-2019 ARM Limited. +# Original author: Dave Martin Dave.Martin@arm.com + +set -ue + +NR_CPUS=`nproc` + +pids= +logs= + +cleanup () { + trap - INT TERM CHLD + set +e + + if [ -n "$pids" ]; then + kill $pids + wait $pids + pids= + fi + + if [ -n "$logs" ]; then + cat $logs + rm $logs + logs= + fi +} + +interrupt () { + cleanup + exit 0 +} + +child_died () { + cleanup + exit 1 +} + +trap interrupt INT TERM EXIT + +for x in `seq 0 $((NR_CPUS * 4))`; do + log=`mktemp` + logs=$logs\ $log + ./za-test >$log & + pids=$pids\ $! +done + +# Wait for all child processes to be created: +sleep 10 + +while :; do + kill -USR1 $pids +done & +pids=$pids\ $! + +wait + +exit 1 diff --git a/tools/testing/selftests/arm64/fp/za-test.S b/tools/testing/selftests/arm64/fp/za-test.S new file mode 100644 index 000000000000..9ab6f9cd9623 --- /dev/null +++ b/tools/testing/selftests/arm64/fp/za-test.S @@ -0,0 +1,388 @@ +// SPDX-License-Identifier: GPL-2.0-only +// Copyright (C) 2021 ARM Limited. +// Original author: Mark Brown broonie@kernel.org +// +// Scalable Matrix Extension ZA context switch test +// Repeatedly writes unique test patterns into each ZA tile +// and reads them back to verify integrity. +// +// for x in `seq 1 NR_CPUS`; do sve-test & pids=$pids\ $! ; done +// (leave it running for as long as you want...) +// kill $pids + +#include <asm/unistd.h> +#include "assembler.h" +#include "asm-offsets.h" +#include "sme-inst.h" + +.arch_extension sve + +#define MAXVL 2048 +#define MAXVL_B (MAXVL / 8) + +// Declare some storage space to shadow ZA register contents and a +// scratch buffer for a vector. +.pushsection .text +.data +.align 4 +zaref: + .space MAXVL_B * MAXVL_B +scratch: + .space MAXVL_B +.popsection + +// Trivial memory copy: copy x2 bytes, starting at address x1, to address x0. +// Clobbers x0-x3 +function memcpy + cmp x2, #0 + b.eq 1f +0: ldrb w3, [x1], #1 + strb w3, [x0], #1 + subs x2, x2, #1 + b.ne 0b +1: ret +endfunction + +// Generate a test pattern for storage in ZA +// x0: pid +// x1: row in ZA +// x2: generation + +// These values are used to constuct a 32-bit pattern that is repeated in the +// scratch buffer as many times as will fit: +// bits 31:28 generation number (increments once per test_loop) +// bits 27:16 pid +// bits 15: 8 row number +// bits 7: 0 32-bit lane index + +function pattern + mov w3, wzr + bfi w3, w0, #16, #12 // PID + bfi w3, w1, #8, #8 // Row + bfi w3, w2, #28, #4 // Generation + + ldr x0, =scratch + mov w1, #MAXVL_B / 4 + +0: str w3, [x0], #4 + add w3, w3, #1 // Lane + subs w1, w1, #1 + b.ne 0b + + ret +endfunction + +// Get the address of shadow data for ZA horizontal vector xn +.macro _adrza xd, xn, nrtmp + ldr \xd, =zaref + rdsvl \nrtmp, 1 + madd \xd, x\nrtmp, \xn, \xd +.endm + +// Set up test pattern in a ZA horizontal vector +// x0: pid +// x1: row number +// x2: generation +function setup_za + mov x4, x30 + mov x12, x1 // Use x12 for vector select + + bl pattern // Get pattern in scratch buffer + _adrza x0, x12, 2 // Shadow buffer pointer to x0 and x5 + mov x5, x0 + ldr x1, =scratch + bl memcpy // length set up in x2 by _adrza + + _ldr_za 12, 5 // load vector w12 from pointer x5 + + ret x4 +endfunction + +// Trivial memory compare: compare x2 bytes starting at address x0 with +// bytes starting at address x1. +// Returns only if all bytes match; otherwise, the program is aborted. +// Clobbers x0-x5. +function memcmp + cbz x2, 2f + + stp x0, x1, [sp, #-0x20]! + str x2, [sp, #0x10] + + mov x5, #0 +0: ldrb w3, [x0, x5] + ldrb w4, [x1, x5] + add x5, x5, #1 + cmp w3, w4 + b.ne 1f + subs x2, x2, #1 + b.ne 0b + +1: ldr x2, [sp, #0x10] + ldp x0, x1, [sp], #0x20 + b.ne barf + +2: ret +endfunction + +// Verify that a ZA vector matches its shadow in memory, else abort +// x0: row number +// Clobbers x0-x7 and x12. +function check_za + mov x3, x30 + + mov x12, x0 + _adrza x5, x0, 6 // pointer to expected value in x5 + mov x4, x0 + ldr x7, =scratch // x7 is scratch + + mov x0, x7 // Poison scratch + mov x1, x6 + bl memfill_ae + + _str_za 12, 7 // save vector w12 to pointer x7 + + mov x0, x5 + mov x1, x7 + mov x2, x6 + mov x30, x3 + b memcmp +endfunction + +// Any SME register modified here can cause corruption in the main +// thread -- but *only* the locations modified here. +function irritator_handler + // Increment the irritation signal count (x23): + ldr x0, [x2, #ucontext_regs + 8 * 23] + add x0, x0, #1 + str x0, [x2, #ucontext_regs + 8 * 23] + + // Corrupt some random ZA data +#if 0 + adr x0, .text + (irritator_handler - .text) / 16 * 16 + movi v0.8b, #1 + movi v9.16b, #2 + movi v31.8b, #3 +#endif + + ret +endfunction + +function terminate_handler + mov w21, w0 + mov x20, x2 + + puts "Terminated by signal " + mov w0, w21 + bl putdec + puts ", no error, iterations=" + ldr x0, [x20, #ucontext_regs + 8 * 22] + bl putdec + puts ", signals=" + ldr x0, [x20, #ucontext_regs + 8 * 23] + bl putdecn + + mov x0, #0 + mov x8, #__NR_exit + svc #0 +endfunction + +// w0: signal number +// x1: sa_action +// w2: sa_flags +// Clobbers x0-x6,x8 +function setsignal + str x30, [sp, #-((sa_sz + 15) / 16 * 16 + 16)]! + + mov w4, w0 + mov x5, x1 + mov w6, w2 + + add x0, sp, #16 + mov x1, #sa_sz + bl memclr + + mov w0, w4 + add x1, sp, #16 + str w6, [x1, #sa_flags] + str x5, [x1, #sa_handler] + mov x2, #0 + mov x3, #sa_mask_sz + mov x8, #__NR_rt_sigaction + svc #0 + + cbz w0, 1f + + puts "sigaction failure\n" + b .Labort + +1: ldr x30, [sp], #((sa_sz + 15) / 16 * 16 + 16) + ret +endfunction + +// Main program entry point +.globl _start +function _start +_start: + puts "Streaming mode " + smstart_za + + // Sanity-check and report the vector length + + rdsvl 19, 8 + cmp x19, #128 + b.lo 1f + cmp x19, #2048 + b.hi 1f + tst x19, #(8 - 1) + b.eq 2f + +1: puts "bad vector length: " + mov x0, x19 + bl putdecn + b .Labort + +2: puts "vector length:\t" + mov x0, x19 + bl putdec + puts " bits\n" + + // Obtain our PID, to ensure test pattern uniqueness between processes + mov x8, #__NR_getpid + svc #0 + mov x20, x0 + + puts "PID:\t" + mov x0, x20 + bl putdecn + + mov x23, #0 // Irritation signal count + + mov w0, #SIGINT + adr x1, terminate_handler + mov w2, #SA_SIGINFO + bl setsignal + + mov w0, #SIGTERM + adr x1, terminate_handler + mov w2, #SA_SIGINFO + bl setsignal + + mov w0, #SIGUSR1 + adr x1, irritator_handler + mov w2, #SA_SIGINFO + orr w2, w2, #SA_NODEFER + bl setsignal + + mov x22, #0 // generation number, increments per iteration +.Ltest_loop: + rdsvl 0, 8 + cmp x0, x19 + b.ne vl_barf + + rdsvl 21, 1 // Set up ZA & shadow with test pattern +0: mov x0, x20 + sub x1, x21, #1 + mov x2, x22 + bl setup_za + subs x21, x21, #1 + b.ne 0b + + and x8, x22, #127 // Every 128 interations... + cbz x8, 0f + mov x8, #__NR_getpid // (otherwise minimal syscall) + b 1f +0: + mov x8, #__NR_sched_yield // ...encourage preemption +1: + svc #0 + + mrs x0, S3_3_C4_C2_2 // SVCR should have ZA=1,SM=0 + and x1, x0, #3 + cmp x1, #2 + b.ne svcr_barf + + rdsvl 21, 1 // Verify that the data made it through + rdsvl 24, 1 // Verify that the data made it through +0: sub x0, x24, x21 + bl check_za + subs x21, x21, #1 + bne 0b + + add x22, x22, #1 // Everything still working + b .Ltest_loop + +.Labort: + mov x0, #0 + mov x1, #SIGABRT + mov x8, #__NR_kill + svc #0 +endfunction + +function barf +// fpsimd.c acitivty log dump hack +// ldr w0, =0xdeadc0de +// mov w8, #__NR_exit +// svc #0 +// end hack + smstop + mov x10, x0 // expected data + mov x11, x1 // actual data + mov x12, x2 // data size + + puts "Mismatch: PID=" + mov x0, x20 + bl putdec + puts ", iteration=" + mov x0, x22 + bl putdec + puts ", row=" + mov x0, x21 + bl putdecn + puts "\tExpected [" + mov x0, x10 + mov x1, x12 + bl dumphex + puts "]\n\tGot [" + mov x0, x11 + mov x1, x12 + bl dumphex + puts "]\n" + + mov x8, #__NR_getpid + svc #0 +// fpsimd.c acitivty log dump hack +// ldr w0, =0xdeadc0de +// mov w8, #__NR_exit +// svc #0 +// ^ end of hack + mov x1, #SIGABRT + mov x8, #__NR_kill + svc #0 +// mov x8, #__NR_exit +// mov x1, #1 +// svc #0 +endfunction + +function vl_barf + mov x10, x0 + + puts "Bad active VL: " + mov x0, x10 + bl putdecn + + mov x8, #__NR_exit + mov x1, #1 + svc #0 +endfunction + +function svcr_barf + mov x10, x0 + + puts "Bad SVCR: " + mov x0, x10 + bl putdecn + + mov x8, #__NR_exit + mov x1, #1 + svc #0 +endfunction

In order to allow ptrace of streaming mode SVE registers we have added a new regset for streaming mode which in isolation offers the same ABI as regular SVE with a different vector type. Add this to the array of regsets we handle, together with additional tests for the interoperation of the two regsets.

Signed-off-by: Mark Brown broonie@kernel.org Reviewed-by: Shuah Khan skhan@linuxfoundation.org Acked-by: Catalin Marinas catalin.marinas@arm.com --- tools/testing/selftests/arm64/fp/sve-ptrace.c | 11 +++++++++++ 1 file changed, 11 insertions(+)

diff --git a/tools/testing/selftests/arm64/fp/sve-ptrace.c b/tools/testing/selftests/arm64/fp/sve-ptrace.c index c2564480b750..df45e4f9465c 100644 --- a/tools/testing/selftests/arm64/fp/sve-ptrace.c +++ b/tools/testing/selftests/arm64/fp/sve-ptrace.c @@ -26,6 +26,10 @@ #define NT_ARM_SVE 0x405 #endif

+#ifndef NT_ARM_SSVE +#define NT_ARM_SSVE 0x40b +#endif + struct vec_type { const char *name; unsigned long hwcap_type; @@ -42,6 +46,13 @@ static const struct vec_type vec_types[] = { .regset = NT_ARM_SVE, .prctl_set = PR_SVE_SET_VL, }, + { + .name = "Streaming SVE", + .hwcap_type = AT_HWCAP2, + .hwcap = HWCAP2_SME, + .regset = NT_ARM_SSVE, + .prctl_set = PR_SME_SET_VL, + }, };

#define VL_TESTS (((SVE_VQ_MAX - SVE_VQ_MIN) + 1) * 4)

For every possible combination of SVE and SME vector length verify that for each possible value of SVCR after a syscall we leave streaming mode and ZA is preserved. We don't need to take account of any streaming/non streaming SVE vector length changes in the assembler code since the store instructions will handle the vector length for us. We log if the system supports FA64 and only try to set FFR in streaming mode if it does.

Signed-off-by: Mark Brown broonie@kernel.org Reviewed-by: Shuah Khan skhan@linuxfoundation.org Acked-by: Catalin Marinas catalin.marinas@arm.com --- .../selftests/arm64/abi/syscall-abi-asm.S | 79 ++++++- .../testing/selftests/arm64/abi/syscall-abi.c | 204 ++++++++++++++++-- .../testing/selftests/arm64/abi/syscall-abi.h | 15 ++ 3 files changed, 275 insertions(+), 23 deletions(-) create mode 100644 tools/testing/selftests/arm64/abi/syscall-abi.h

diff --git a/tools/testing/selftests/arm64/abi/syscall-abi-asm.S b/tools/testing/selftests/arm64/abi/syscall-abi-asm.S index 983467cfcee0..b523c21c2278 100644 --- a/tools/testing/selftests/arm64/abi/syscall-abi-asm.S +++ b/tools/testing/selftests/arm64/abi/syscall-abi-asm.S @@ -9,15 +9,42 @@ // invoked is configured in x8 of the input GPR data. // // x0: SVE VL, 0 for FP only +// x1: SME VL // // GPRs: gpr_in, gpr_out // FPRs: fpr_in, fpr_out // Zn: z_in, z_out // Pn: p_in, p_out // FFR: ffr_in, ffr_out +// ZA: za_in, za_out +// SVCR: svcr_in, svcr_out + +#include "syscall-abi.h"

.arch_extension sve

+/* + * LDR (vector to ZA array): + * LDR ZA[\nw, #\offset], [X\nxbase, #\offset, MUL VL] + */ +.macro _ldr_za nw, nxbase, offset=0 + .inst 0xe1000000 \ + | (((\nw) & 3) << 13) \ + | ((\nxbase) << 5) \ + | ((\offset) & 7) +.endm + +/* + * STR (vector from ZA array): + * STR ZA[\nw, #\offset], [X\nxbase, #\offset, MUL VL] + */ +.macro _str_za nw, nxbase, offset=0 + .inst 0xe1200000 \ + | (((\nw) & 3) << 13) \ + | ((\nxbase) << 5) \ + | ((\offset) & 7) +.endm + .globl do_syscall do_syscall: // Store callee saved registers x19-x29 (80 bytes) plus x0 and x1 @@ -30,6 +57,24 @@ do_syscall: stp x25, x26, [sp, #80] stp x27, x28, [sp, #96]

+ // Set SVCR if we're doing SME + cbz x1, 1f + adrp x2, svcr_in + ldr x2, [x2, :lo12:svcr_in] + msr S3_3_C4_C2_2, x2 +1: + + // Load ZA if it's enabled - uses x12 as scratch due to SME LDR + tbz x2, #SVCR_ZA_SHIFT, 1f + mov w12, #0 + ldr x2, =za_in +2: _ldr_za 12, 2 + add x2, x2, x1 + add x12, x12, #1 + cmp x1, x12 + bne 2b +1: + // Load GPRs x8-x28, and save our SP/FP for later comparison ldr x2, =gpr_in add x2, x2, #64 @@ -68,7 +113,7 @@ do_syscall: ldp q30, q31, [x2, #16 * 30] 1:

- // Load the SVE registers if we're doing SVE + // Load the SVE registers if we're doing SVE/SME cbz x0, 1f

ldr x2, =z_in @@ -105,9 +150,14 @@ do_syscall: ldr z30, [x2, #30, MUL VL] ldr z31, [x2, #31, MUL VL]

+ // Only set a non-zero FFR, test patterns must be zero since the + // syscall should clear it - this lets us handle FA64. ldr x2, =ffr_in ldr p0, [x2, #0] + ldr x2, [x2, #0] + cbz x2, 2f wrffr p0.b +2:

ldr x2, =p_in ldr p0, [x2, #0, MUL VL] @@ -169,6 +219,24 @@ do_syscall: stp q28, q29, [x2, #16 * 28] stp q30, q31, [x2, #16 * 30]

+ // Save SVCR if we're doing SME + cbz x1, 1f + mrs x2, S3_3_C4_C2_2 + adrp x3, svcr_out + str x2, [x3, :lo12:svcr_out] +1: + + // Save ZA if it's enabled - uses x12 as scratch due to SME STR + tbz x2, #SVCR_ZA_SHIFT, 1f + mov w12, #0 + ldr x2, =za_out +2: _str_za 12, 2 + add x2, x2, x1 + add x12, x12, #1 + cmp x1, x12 + bne 2b +1: + // Save the SVE state if we have some cbz x0, 1f

@@ -224,6 +292,10 @@ do_syscall: str p14, [x2, #14, MUL VL] str p15, [x2, #15, MUL VL]

+ // Only save FFR if we wrote a value for SME + ldr x2, =ffr_in + ldr x2, [x2, #0] + cbz x2, 1f ldr x2, =ffr_out rdffr p0.b str p0, [x2, #0] @@ -237,4 +309,9 @@ do_syscall: ldp x27, x28, [sp, #96] ldp x29, x30, [sp], #112

+ // Clear SVCR if we were doing SME so future tests don't have ZA + cbz x1, 1f + msr S3_3_C4_C2_2, xzr +1: + ret diff --git a/tools/testing/selftests/arm64/abi/syscall-abi.c b/tools/testing/selftests/arm64/abi/syscall-abi.c index 1e13b7523918..b632bfe9e022 100644 --- a/tools/testing/selftests/arm64/abi/syscall-abi.c +++ b/tools/testing/selftests/arm64/abi/syscall-abi.c @@ -18,9 +18,13 @@

#include "../../kselftest.h"

+#include "syscall-abi.h" + #define NUM_VL ((SVE_VQ_MAX - SVE_VQ_MIN) + 1)

-extern void do_syscall(int sve_vl); +static int default_sme_vl; + +extern void do_syscall(int sve_vl, int sme_vl);

static void fill_random(void *buf, size_t size) { @@ -48,14 +52,15 @@ static struct syscall_cfg { uint64_t gpr_in[NUM_GPR]; uint64_t gpr_out[NUM_GPR];

-static void setup_gpr(struct syscall_cfg *cfg, int sve_vl) +static void setup_gpr(struct syscall_cfg *cfg, int sve_vl, int sme_vl, + uint64_t svcr) { fill_random(gpr_in, sizeof(gpr_in)); gpr_in[8] = cfg->syscall_nr; memset(gpr_out, 0, sizeof(gpr_out)); }

-static int check_gpr(struct syscall_cfg *cfg, int sve_vl) +static int check_gpr(struct syscall_cfg *cfg, int sve_vl, int sme_vl, uint64_t svcr) { int errors = 0; int i; @@ -79,13 +84,15 @@ static int check_gpr(struct syscall_cfg *cfg, int sve_vl) uint64_t fpr_in[NUM_FPR * 2]; uint64_t fpr_out[NUM_FPR * 2];

-static void setup_fpr(struct syscall_cfg *cfg, int sve_vl) +static void setup_fpr(struct syscall_cfg *cfg, int sve_vl, int sme_vl, + uint64_t svcr) { fill_random(fpr_in, sizeof(fpr_in)); memset(fpr_out, 0, sizeof(fpr_out)); }

-static int check_fpr(struct syscall_cfg *cfg, int sve_vl) +static int check_fpr(struct syscall_cfg *cfg, int sve_vl, int sme_vl, + uint64_t svcr) { int errors = 0; int i; @@ -109,13 +116,15 @@ static uint8_t z_zero[__SVE_ZREG_SIZE(SVE_VQ_MAX)]; uint8_t z_in[SVE_NUM_PREGS * __SVE_ZREG_SIZE(SVE_VQ_MAX)]; uint8_t z_out[SVE_NUM_PREGS * __SVE_ZREG_SIZE(SVE_VQ_MAX)];

-static void setup_z(struct syscall_cfg *cfg, int sve_vl) +static void setup_z(struct syscall_cfg *cfg, int sve_vl, int sme_vl, + uint64_t svcr) { fill_random(z_in, sizeof(z_in)); fill_random(z_out, sizeof(z_out)); }

-static int check_z(struct syscall_cfg *cfg, int sve_vl) +static int check_z(struct syscall_cfg *cfg, int sve_vl, int sme_vl, + uint64_t svcr) { size_t reg_size = sve_vl; int errors = 0; @@ -126,13 +135,17 @@ static int check_z(struct syscall_cfg *cfg, int sve_vl)

/* * After a syscall the low 128 bits of the Z registers should - * be preserved and the rest be zeroed or preserved. + * be preserved and the rest be zeroed or preserved, except if + * we were in streaming mode in which case the low 128 bits may + * also be cleared by the transition out of streaming mode. */ for (i = 0; i < SVE_NUM_ZREGS; i++) { void *in = &z_in[reg_size * i]; void *out = &z_out[reg_size * i];

- if (memcmp(in, out, SVE_VQ_BYTES) != 0) { + if ((memcmp(in, out, SVE_VQ_BYTES) != 0) && + !((svcr & SVCR_SM_MASK) && + memcmp(z_zero, out, SVE_VQ_BYTES) == 0)) { ksft_print_msg("%s SVE VL %d Z%d low 128 bits changed\n", cfg->name, sve_vl, i); errors++; @@ -145,13 +158,15 @@ static int check_z(struct syscall_cfg *cfg, int sve_vl) uint8_t p_in[SVE_NUM_PREGS * __SVE_PREG_SIZE(SVE_VQ_MAX)]; uint8_t p_out[SVE_NUM_PREGS * __SVE_PREG_SIZE(SVE_VQ_MAX)];

-static void setup_p(struct syscall_cfg *cfg, int sve_vl) +static void setup_p(struct syscall_cfg *cfg, int sve_vl, int sme_vl, + uint64_t svcr) { fill_random(p_in, sizeof(p_in)); fill_random(p_out, sizeof(p_out)); }

-static int check_p(struct syscall_cfg *cfg, int sve_vl) +static int check_p(struct syscall_cfg *cfg, int sve_vl, int sme_vl, + uint64_t svcr) { size_t reg_size = sve_vq_from_vl(sve_vl) * 2; /* 1 bit per VL byte */

@@ -175,8 +190,19 @@ static int check_p(struct syscall_cfg *cfg, int sve_vl) uint8_t ffr_in[__SVE_PREG_SIZE(SVE_VQ_MAX)]; uint8_t ffr_out[__SVE_PREG_SIZE(SVE_VQ_MAX)];

-static void setup_ffr(struct syscall_cfg *cfg, int sve_vl) +static void setup_ffr(struct syscall_cfg *cfg, int sve_vl, int sme_vl, + uint64_t svcr) { + /* + * If we are in streaming mode and do not have FA64 then FFR + * is unavailable. + */ + if ((svcr & SVCR_SM_MASK) && + !(getauxval(AT_HWCAP2) & HWCAP2_SME_FA64)) { + memset(&ffr_in, 0, sizeof(ffr_in)); + return; + } + /* * It is only valid to set a contiguous set of bits starting * at 0. For now since we're expecting this to be cleared by @@ -186,7 +212,8 @@ static void setup_ffr(struct syscall_cfg *cfg, int sve_vl) fill_random(ffr_out, sizeof(ffr_out)); }

-static int check_ffr(struct syscall_cfg *cfg, int sve_vl) +static int check_ffr(struct syscall_cfg *cfg, int sve_vl, int sme_vl, + uint64_t svcr) { size_t reg_size = sve_vq_from_vl(sve_vl) * 2; /* 1 bit per VL byte */ int errors = 0; @@ -195,6 +222,10 @@ static int check_ffr(struct syscall_cfg *cfg, int sve_vl) if (!sve_vl) return 0;

+ if ((svcr & SVCR_SM_MASK) && + !(getauxval(AT_HWCAP2) & HWCAP2_SME_FA64)) + return 0; + /* After a syscall the P registers should be preserved or zeroed */ for (i = 0; i < reg_size; i++) if (ffr_out[i] && (ffr_in[i] != ffr_out[i])) @@ -206,8 +237,65 @@ static int check_ffr(struct syscall_cfg *cfg, int sve_vl) return errors; }

-typedef void (*setup_fn)(struct syscall_cfg *cfg, int sve_vl); -typedef int (*check_fn)(struct syscall_cfg *cfg, int sve_vl); +uint64_t svcr_in, svcr_out; + +static void setup_svcr(struct syscall_cfg *cfg, int sve_vl, int sme_vl, + uint64_t svcr) +{ + svcr_in = svcr; +} + +static int check_svcr(struct syscall_cfg *cfg, int sve_vl, int sme_vl, + uint64_t svcr) +{ + int errors = 0; + + if (svcr_out & SVCR_SM_MASK) { + ksft_print_msg("%s Still in SM, SVCR %llx\n", + cfg->name, svcr_out); + errors++; + } + + if ((svcr_in & SVCR_ZA_MASK) != (svcr_out & SVCR_ZA_MASK)) { + ksft_print_msg("%s PSTATE.ZA changed, SVCR %llx != %llx\n", + cfg->name, svcr_in, svcr_out); + errors++; + } + + return errors; +} + +uint8_t za_in[SVE_NUM_PREGS * __SVE_ZREG_SIZE(SVE_VQ_MAX)]; +uint8_t za_out[SVE_NUM_PREGS * __SVE_ZREG_SIZE(SVE_VQ_MAX)]; + +static void setup_za(struct syscall_cfg *cfg, int sve_vl, int sme_vl, + uint64_t svcr) +{ + fill_random(za_in, sizeof(za_in)); + memset(za_out, 0, sizeof(za_out)); +} + +static int check_za(struct syscall_cfg *cfg, int sve_vl, int sme_vl, + uint64_t svcr) +{ + size_t reg_size = sme_vl * sme_vl; + int errors = 0; + + if (!(svcr & SVCR_ZA_MASK)) + return 0; + + if (memcmp(za_in, za_out, reg_size) != 0) { + ksft_print_msg("SME VL %d ZA does not match\n", sme_vl); + errors++; + } + + return errors; +} + +typedef void (*setup_fn)(struct syscall_cfg *cfg, int sve_vl, int sme_vl, + uint64_t svcr); +typedef int (*check_fn)(struct syscall_cfg *cfg, int sve_vl, int sme_vl, + uint64_t svcr);

/* * Each set of registers has a setup function which is called before @@ -225,20 +313,23 @@ static struct { { setup_z, check_z }, { setup_p, check_p }, { setup_ffr, check_ffr }, + { setup_svcr, check_svcr }, + { setup_za, check_za }, };

-static bool do_test(struct syscall_cfg *cfg, int sve_vl) +static bool do_test(struct syscall_cfg *cfg, int sve_vl, int sme_vl, + uint64_t svcr) { int errors = 0; int i;

for (i = 0; i < ARRAY_SIZE(regset); i++) - regset[i].setup(cfg, sve_vl); + regset[i].setup(cfg, sve_vl, sme_vl, svcr);

- do_syscall(sve_vl); + do_syscall(sve_vl, sme_vl);

for (i = 0; i < ARRAY_SIZE(regset); i++) - errors += regset[i].check(cfg, sve_vl); + errors += regset[i].check(cfg, sve_vl, sme_vl, svcr);

return errors == 0; } @@ -246,9 +337,10 @@ static bool do_test(struct syscall_cfg *cfg, int sve_vl) static void test_one_syscall(struct syscall_cfg *cfg) { int sve_vq, sve_vl; + int sme_vq, sme_vl;

/* FPSIMD only case */ - ksft_test_result(do_test(cfg, 0), + ksft_test_result(do_test(cfg, 0, default_sme_vl, 0), "%s FPSIMD\n", cfg->name);

if (!(getauxval(AT_HWCAP) & HWCAP_SVE)) @@ -265,8 +357,36 @@ static void test_one_syscall(struct syscall_cfg *cfg) if (sve_vq != sve_vq_from_vl(sve_vl)) sve_vq = sve_vq_from_vl(sve_vl);

- ksft_test_result(do_test(cfg, sve_vl), + ksft_test_result(do_test(cfg, sve_vl, default_sme_vl, 0), "%s SVE VL %d\n", cfg->name, sve_vl); + + if (!(getauxval(AT_HWCAP2) & HWCAP2_SME)) + continue; + + for (sme_vq = SVE_VQ_MAX; sme_vq > 0; --sme_vq) { + sme_vl = prctl(PR_SME_SET_VL, sme_vq * 16); + if (sme_vl == -1) + ksft_exit_fail_msg("PR_SME_SET_VL failed: %s (%d)\n", + strerror(errno), errno); + + sme_vl &= PR_SME_VL_LEN_MASK; + + if (sme_vq != sve_vq_from_vl(sme_vl)) + sme_vq = sve_vq_from_vl(sme_vl); + + ksft_test_result(do_test(cfg, sve_vl, sme_vl, + SVCR_ZA_MASK | SVCR_SM_MASK), + "%s SVE VL %d/SME VL %d SM+ZA\n", + cfg->name, sve_vl, sme_vl); + ksft_test_result(do_test(cfg, sve_vl, sme_vl, + SVCR_SM_MASK), + "%s SVE VL %d/SME VL %d SM\n", + cfg->name, sve_vl, sme_vl); + ksft_test_result(do_test(cfg, sve_vl, sme_vl, + SVCR_ZA_MASK), + "%s SVE VL %d/SME VL %d ZA\n", + cfg->name, sve_vl, sme_vl); + } } }

@@ -299,14 +419,54 @@ int sve_count_vls(void) return vl_count; }

+int sme_count_vls(void) +{ + unsigned int vq; + int vl_count = 0; + int vl; + + if (!(getauxval(AT_HWCAP2) & HWCAP2_SME)) + return 0; + + /* Ensure we configure a SME VL, used to flag if SVCR is set */ + default_sme_vl = 16; + + /* + * Enumerate up to SVE_VQ_MAX vector lengths + */ + for (vq = SVE_VQ_MAX; vq > 0; --vq) { + vl = prctl(PR_SME_SET_VL, vq * 16); + if (vl == -1) + ksft_exit_fail_msg("PR_SME_SET_VL failed: %s (%d)\n", + strerror(errno), errno); + + vl &= PR_SME_VL_LEN_MASK; + + if (vq != sve_vq_from_vl(vl)) + vq = sve_vq_from_vl(vl); + + vl_count++; + } + + return vl_count; +} + int main(void) { int i; + int tests = 1; /* FPSIMD */

srandom(getpid());

ksft_print_header(); - ksft_set_plan(ARRAY_SIZE(syscalls) * (sve_count_vls() + 1)); + tests += sve_count_vls(); + tests += (sve_count_vls() * sme_count_vls()) * 3; + ksft_set_plan(ARRAY_SIZE(syscalls) * tests); + + if (getauxval(AT_HWCAP2) & HWCAP2_SME_FA64) + ksft_print_msg("SME with FA64\n"); + else if (getauxval(AT_HWCAP2) & HWCAP2_SME) + ksft_print_msg("SME without FA64\n");

for (i = 0; i < ARRAY_SIZE(syscalls); i++) test_one_syscall(&syscalls[i]); diff --git a/tools/testing/selftests/arm64/abi/syscall-abi.h b/tools/testing/selftests/arm64/abi/syscall-abi.h new file mode 100644 index 000000000000..bda5a87ad381 --- /dev/null +++ b/tools/testing/selftests/arm64/abi/syscall-abi.h @@ -0,0 +1,15 @@ +/* SPDX-License-Identifier: GPL-2.0-only */ +/* + * Copyright (C) 2021 ARM Limited. + */ + +#ifndef SYSCALL_ABI_H +#define SYSCALL_ABI_H + +#define SVCR_ZA_MASK 2 +#define SVCR_SM_MASK 1 + +#define SVCR_ZA_SHIFT 1 +#define SVCR_SM_SHIFT 0 + +#endif