Dave Martin <dave.mar...@arm.com> writes:
> Currently the FPSIMD handling code uses the condition task->mm ==
> NULL as a hint that task has no FPSIMD register context.
>
> The ->mm check is only there to filter out tasks that cannot
> possibly have FPSIMD context loaded, for optimisation purposes.
> Also, TIF_FOREIGN_FPSTATE must always be checked anyway before
> saving FPSIMD context back to memory. For these reasons, the ->mm
> checks are not useful, providing that TIF_FOREIGN_FPSTATE is
> maintained in a consistent way for all threads.
>
> The context switch logic is already deliberately optimised to defer
> reloads of the regs until ret_to_user (or sigreturn as a special
> case), and save them only if they have been previously loaded.
> These paths are the only places where the wrong_task and wrong_cpu
> conditions can be made false, by calling fpsimd_bind_task_to_cpu().
> Kernel threads by definition never reach these paths. As a result,
> the wrong_task and wrong_cpu tests in fpsimd_thread_switch() will
> always yield true for kernel threads.
>
> This patch removes the redundant checks and special-case code,
> ensuring that TIF_FOREIGN_FPSTATE is set whenever a kernel thread
> is scheduled in, and ensures that this flag is set for the init
> task. The fpsimd_flush_task_state() call already present in
> copy_thread() ensures the same for any new task.
>
> With TIF_FOREIGN_FPSTATE always set for kernel threads, this patch
> ensures that no extra context save work is added for kernel
> threads, and eliminates the redundant context saving that may
> currently occur for kernel threads that have acquired an mm via
> use_mm().
>
> Signed-off-by: Dave Martin <dave.mar...@arm.com>
> Reviewed-by: Catalin Marinas <catalin.mari...@arm.com>
> Reviewed-by: Alex Bennée <alex.ben...@linaro.org>
Still good (although obviously without the ws damage in the commit).
> Cc: Catalin Marinas <catalin.mari...@arm.com>
> Cc: Will Deacon <will.dea...@arm.com>
> Cc: Ard Biesheuvel <ard.biesheu...@linaro.org>
>
> ---
>
> Changes since v10:
>
> * The INIT_THREAD flag change is split out into the prior
> patch, since it is in principle a fix rather than simply a
> tidy-up.
>
> Requested by Christoffer Dall / Catalin Marinas:
>
> * Reworded commit message to explain the change more clearly,
> and remove confusing claims about things being true by
> construction.
>
> * Added a comment to the code explaining that wrong_cpu and
> wrong_task will always be true for kernel threads.
>
> * Ensure .fpsimd_cpu = NR_CPUS for the init task.
>
> This does not seem to be a bug, because the wrong_task check in
> fpsimd_thread_switch() should still always be true for the init
> task; but it is nonetheless an inconsistency compared with what
> copy_thread() does.
>
> So fix it to avoid future surprises.
> ---
> arch/arm64/include/asm/processor.h | 4 +++-
> arch/arm64/kernel/fpsimd.c | 40
> +++++++++++++++-----------------------
> 2 files changed, 19 insertions(+), 25 deletions(-)
>
> diff --git a/arch/arm64/include/asm/processor.h
> b/arch/arm64/include/asm/processor.h
> index 7675989..36d64f8 100644
> --- a/arch/arm64/include/asm/processor.h
> +++ b/arch/arm64/include/asm/processor.h
> @@ -156,7 +156,9 @@ static inline void arch_thread_struct_whitelist(unsigned
> long *offset,
> /* Sync TPIDR_EL0 back to thread_struct for current */
> void tls_preserve_current_state(void);
>
> -#define INIT_THREAD { }
> +#define INIT_THREAD { \
> + .fpsimd_cpu = NR_CPUS, \
> +}
>
> static inline void start_thread_common(struct pt_regs *regs, unsigned long
> pc)
> {
> diff --git a/arch/arm64/kernel/fpsimd.c b/arch/arm64/kernel/fpsimd.c
> index 2d9a9e8..d736b6c 100644
> --- a/arch/arm64/kernel/fpsimd.c
> +++ b/arch/arm64/kernel/fpsimd.c
> @@ -892,31 +892,25 @@ asmlinkage void do_fpsimd_exc(unsigned int esr, struct
> pt_regs *regs)
>
> void fpsimd_thread_switch(struct task_struct *next)
> {
> + bool wrong_task, wrong_cpu;
> +
> if (!system_supports_fpsimd())
> return;
> +
> + /* Save unsaved fpsimd state, if any: */
> + fpsimd_save();
> +
> /*
> - * Save the current FPSIMD state to memory, but only if whatever is in
> - * the registers is in fact the most recent userland FPSIMD state of
> - * 'current'.
> + * Fix up TIF_FOREIGN_FPSTATE to correctly describe next's
> + * state. For kernel threads, FPSIMD registers are never loaded
> + * and wrong_task and wrong_cpu will always be true.
> */
> - if (current->mm)
> - fpsimd_save();
> -
> - if (next->mm) {
> - /*
> - * If we are switching to a task whose most recent userland
> - * FPSIMD state is already in the registers of *this* cpu,
> - * we can skip loading the state from memory. Otherwise, set
> - * the TIF_FOREIGN_FPSTATE flag so the state will be loaded
> - * upon the next return to userland.
> - */
> - bool wrong_task = __this_cpu_read(fpsimd_last_state.st) !=
> + wrong_task = __this_cpu_read(fpsimd_last_state.st) !=
> &next->thread.uw.fpsimd_state;
> - bool wrong_cpu = next->thread.fpsimd_cpu != smp_processor_id();
> + wrong_cpu = next->thread.fpsimd_cpu != smp_processor_id();
>
> - update_tsk_thread_flag(next, TIF_FOREIGN_FPSTATE,
> - wrong_task || wrong_cpu);
> - }
> + update_tsk_thread_flag(next, TIF_FOREIGN_FPSTATE,
> + wrong_task || wrong_cpu);
> }
>
> void fpsimd_flush_thread(void)
> @@ -1121,9 +1115,8 @@ void kernel_neon_begin(void)
>
> __this_cpu_write(kernel_neon_busy, true);
>
> - /* Save unsaved task fpsimd state, if any: */
> - if (current->mm)
> - fpsimd_save();
> + /* Save unsaved fpsimd state, if any: */
> + fpsimd_save();
>
> /* Invalidate any task state remaining in the fpsimd regs: */
> fpsimd_flush_cpu_state();
> @@ -1245,8 +1238,7 @@ static int fpsimd_cpu_pm_notifier(struct notifier_block
> *self,
> {
> switch (cmd) {
> case CPU_PM_ENTER:
> - if (current->mm)
> - fpsimd_save();
> + fpsimd_save();
> fpsimd_flush_cpu_state();
> break;
> case CPU_PM_EXIT:
--
Alex Bennée
_______________________________________________
kvmarm mailing list
kvmarm@lists.cs.columbia.edu
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
