From: Mark Brown broonie@kernel.org
mainline inclusion from mainline-v5.13-rc1 commit cccb78ce89c45a4414db712be4986edfb92434bd category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/I5ITJT CVE: NA
Reference: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=...
-------------------------------------------------
When we enable SVE usage in userspace after taking a SVE access trap we need to ensure that the portions of the register state that are not shared with the FPSIMD registers are zeroed. Currently we do this by forcing the FPSIMD registers to be saved to the task struct and converting them there. This is wasteful in the common case where the task state is loaded into the registers and we will immediately return to userspace since we can initialise the SVE state directly in registers instead of accessing multiple copies of the register state in memory.
Instead in that common case do the conversion in the registers and update the task metadata so that we can return to userspace without spilling the register state to memory unless there is some other reason to do so.
Signed-off-by: Mark Brown broonie@kernel.org Link: https://lore.kernel.org/r/20210312190313.24598-1-broonie@kernel.org Signed-off-by: Catalin Marinas catalin.marinas@arm.com Signed-off-by: Wang ShaoBo bobo.shaobowang@huawei.com --- arch/arm64/include/asm/fpsimd.h | 1 + arch/arm64/kernel/entry-fpsimd.S | 5 +++++ arch/arm64/kernel/fpsimd.c | 26 +++++++++++++++++--------- 3 files changed, 23 insertions(+), 9 deletions(-)
diff --git a/arch/arm64/include/asm/fpsimd.h b/arch/arm64/include/asm/fpsimd.h index 34fb33a2a110..7de0b34476f3 100644 --- a/arch/arm64/include/asm/fpsimd.h +++ b/arch/arm64/include/asm/fpsimd.h @@ -73,6 +73,7 @@ extern void sve_flush_live(void); extern void sve_load_from_fpsimd_state(struct user_fpsimd_state const *state, unsigned long vq_minus_1); extern unsigned int sve_get_vl(void); +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); diff --git a/arch/arm64/kernel/entry-fpsimd.S b/arch/arm64/kernel/entry-fpsimd.S index 2ca395c25448..3ecec60d3295 100644 --- a/arch/arm64/kernel/entry-fpsimd.S +++ b/arch/arm64/kernel/entry-fpsimd.S @@ -48,6 +48,11 @@ SYM_FUNC_START(sve_get_vl) ret SYM_FUNC_END(sve_get_vl)
+SYM_FUNC_START(sve_set_vq) + sve_load_vq x0, x1, x2 + ret +SYM_FUNC_END(sve_set_vq) + /* * Load SVE state from FPSIMD state. * diff --git a/arch/arm64/kernel/fpsimd.c b/arch/arm64/kernel/fpsimd.c index 901096d74b7e..9c3edb0899aa 100644 --- a/arch/arm64/kernel/fpsimd.c +++ b/arch/arm64/kernel/fpsimd.c @@ -927,9 +927,8 @@ void fpsimd_release_task(struct task_struct *dead_task) * Trapped SVE access * * Storage is allocated for the full SVE state, the current FPSIMD - * register contents are migrated across, and TIF_SVE is set so that - * the SVE access trap will be disabled the next time this task - * reaches ret_to_user. + * register contents are migrated across, and the access trap is + * disabled. * * TIF_SVE should be clear on entry: otherwise, fpsimd_restore_current_state() * would have disabled the SVE access trap for userspace during @@ -947,15 +946,24 @@ void do_sve_acc(unsigned int esr, struct pt_regs *regs)
get_cpu_fpsimd_context();
- fpsimd_save(); - - /* Force ret_to_user to reload the registers: */ - fpsimd_flush_task_state(current); - - fpsimd_to_sve(current); if (test_and_set_thread_flag(TIF_SVE)) 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. + */ + if (!test_thread_flag(TIF_FOREIGN_FPSTATE)) { + sve_set_vq(sve_vq_from_vl(current->thread.sve_vl) - 1); + sve_flush_live(); + fpsimd_bind_task_to_cpu(); + } else { + fpsimd_to_sve(current); + } + put_cpu_fpsimd_context(); }