On Fri, Jan 19, 2024 at 12:29:13PM +0000, Mark Brown wrote:
> When we are in a syscall we will only save the FPSIMD subset even though
> the task still has access to the full register set, and on context switch
(Pedantic nit: "A even if B" (= "A applies even in that subset of cases
where B"), instead of "A even though B" (= "A applies notwithstanding
that it is always the case that B") (?) If the SVE trapping were
ripped out altogether, it would be a different and rather simpler
story...)
> we will only remove TIF_SVE when loading the register state. This means
> that the signal handling code should not assume that TIF_SVE means that
> the register state is stored in SVE format, it should instead check the
> format that was recorded during save.
>
> Fixes: 8c845e273104 ("arm64/sve: Leave SVE enabled on syscall if we don't context switch")
> Signed-off-by: Mark Brown <broonie@xxxxxxxxxx>
> Cc: <stable@xxxxxxxxxxxxxxx>
> ---
> arch/arm64/kernel/fpsimd.c | 2 +-
> arch/arm64/kernel/signal.c | 4 ++--
> 2 files changed, 3 insertions(+), 3 deletions(-)
>
> diff --git a/arch/arm64/kernel/fpsimd.c b/arch/arm64/kernel/fpsimd.c
> index 1559c706d32d..80133c190136 100644
> --- a/arch/arm64/kernel/fpsimd.c
> +++ b/arch/arm64/kernel/fpsimd.c
> @@ -1626,7 +1626,7 @@ void fpsimd_preserve_current_state(void)
> void fpsimd_signal_preserve_current_state(void)
> {
> fpsimd_preserve_current_state();
> - if (test_thread_flag(TIF_SVE))
> + if (current->thread.fp_type == FP_STATE_SVE)
> sve_to_fpsimd(current);
> }
>
> diff --git a/arch/arm64/kernel/signal.c b/arch/arm64/kernel/signal.c
> index 0e8beb3349ea..425b1bc17a3f 100644
> --- a/arch/arm64/kernel/signal.c
> +++ b/arch/arm64/kernel/signal.c
> @@ -242,7 +242,7 @@ static int preserve_sve_context(struct sve_context __user *ctx)
> vl = task_get_sme_vl(current);
> vq = sve_vq_from_vl(vl);
> flags |= SVE_SIG_FLAG_SM;
> - } else if (test_thread_flag(TIF_SVE)) {
> + } else if (current->thread.fp_type == FP_STATE_SVE) {
> vq = sve_vq_from_vl(vl);
> }
>
> @@ -878,7 +878,7 @@ static int setup_sigframe_layout(struct rt_sigframe_user_layout *user,
> if (system_supports_sve() || system_supports_sme()) {
> unsigned int vq = 0;
>
> - if (add_all || test_thread_flag(TIF_SVE) ||
> + if (add_all || current->thread.fp_type == FP_STATE_SVE ||
> thread_sm_enabled(¤t->thread)) {
> int vl = max(sve_max_vl(), sme_max_vl());
>
>
> ---
> base-commit: 0dd3ee31125508cd67f7e7172247f05b7fd1753a
> change-id: 20240118-arm64-sve-signal-regs-5711e0d10425
>
> Best regards,
> --
> Mark Brown <broonie@xxxxxxxxxx>
>
[...]
If the historical meanings of TIF_SVE have been split up (which seems a
good idea), does that resolve all of the "bare"
test_thread_flag(TIF_SVE) that were still there?
If there are any others remaining, they may need looking at if there is
any question over what condition they are trying to test for...
Cheers
---Dave
