In v4.16-RT I noticed a number of warnings from task_fpsimd_load(). The
code disables BH and expects that it is not preemptible. On -RT the
task remains preemptible but remains the same CPU. This may corrupt the
content of the SIMD registers if the task is preempted during
saving/restoring those registers.
Add a locallock around this process. This avoids that the any function
within the locallock block is invoked more than once on the same CPU.
The kernel_neon_begin() can't be kept preemptible. If the task-switch notices
TIF_FOREIGN_FPSTATE then it would restore task's SIMD state and we lose the
state of registers used for in-kernel-work. We would require additional storage
for the in-kernel copy of the registers. But then the NEON-crypto checks for
the need-resched flag so it shouldn't that bad.
The preempt_disable() avoids the context switch while the kernel uses the SIMD
registers. Unfortunately we have to balance out the migrate_disable() counter
because local_lock_bh() is invoked in different context compared to its unlock
counterpart.
__efi_fpsimd_begin() should not use kernel_fpu_begin() due to its
preempt_disable() context and instead save the registers always in its
extra spot on RT.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@xxxxxxxxxxxxx>
---
This seems to make work (crypto chacha20-neon + cyclictest). I have no
EFI so I have no clue if saving SIMD while calling to EFI works.
arch/arm64/kernel/fpsimd.c | 47 ++++++++++++++++++++++++++-------------------
1 file changed, 28 insertions(+), 19 deletions(-)
--- a/arch/arm64/kernel/fpsimd.c
+++ b/arch/arm64/kernel/fpsimd.c
@@ -38,6 +38,7 @@
#include <linux/signal.h>
#include <linux/slab.h>
#include <linux/sysctl.h>
+#include <linux/locallock.h>
#include <asm/fpsimd.h>
#include <asm/cputype.h>
@@ -235,7 +236,7 @@ static void sve_user_enable(void)
* whether TIF_SVE is clear or set, since these are not vector length
* dependent.
*/
-
+static DEFINE_LOCAL_IRQ_LOCK(fpsimd_lock);
/*
* Update current's FPSIMD/SVE registers from thread_struct.
*
@@ -594,7 +595,7 @@ int sve_set_vector_length(struct task_st
* non-SVE thread.
*/
if (task == current) {
- local_bh_disable();
+ local_lock_bh(fpsimd_lock);
task_fpsimd_save();
set_thread_flag(TIF_FOREIGN_FPSTATE);
@@ -605,7 +606,7 @@ int sve_set_vector_length(struct task_st
sve_to_fpsimd(task);
if (task == current)
- local_bh_enable();
+ local_unlock_bh(fpsimd_lock);
/*
* Force reallocation of task SVE state to the correct size
@@ -837,7 +838,7 @@ asmlinkage void do_sve_acc(unsigned int
sve_alloc(current);
- local_bh_disable();
+ local_lock_bh(fpsimd_lock);
task_fpsimd_save();
fpsimd_to_sve(current);
@@ -849,7 +850,7 @@ asmlinkage void do_sve_acc(unsigned int
if (test_and_set_thread_flag(TIF_SVE))
WARN_ON(1); /* SVE access shouldn't have trapped */
- local_bh_enable();
+ local_unlock_bh(fpsimd_lock);
}
/*
@@ -925,7 +926,7 @@ void fpsimd_flush_thread(void)
if (!system_supports_fpsimd())
return;
- local_bh_disable();
+ local_lock_bh(fpsimd_lock);
memset(¤t->thread.fpsimd_state, 0, sizeof(struct fpsimd_state));
fpsimd_flush_task_state(current);
@@ -967,7 +968,7 @@ void fpsimd_flush_thread(void)
set_thread_flag(TIF_FOREIGN_FPSTATE);
- local_bh_enable();
+ local_unlock_bh(fpsimd_lock);
}
/*
@@ -979,9 +980,9 @@ void fpsimd_preserve_current_state(void)
if (!system_supports_fpsimd())
return;
- local_bh_disable();
+ local_lock_bh(fpsimd_lock);
task_fpsimd_save();
- local_bh_enable();
+ local_unlock_bh(fpsimd_lock);
}
/*
@@ -1021,14 +1022,14 @@ void fpsimd_restore_current_state(void)
if (!system_supports_fpsimd())
return;
- local_bh_disable();
+ local_lock_bh(fpsimd_lock);
if (test_and_clear_thread_flag(TIF_FOREIGN_FPSTATE)) {
task_fpsimd_load();
fpsimd_bind_to_cpu();
}
- local_bh_enable();
+ local_unlock_bh(fpsimd_lock);
}
/*
@@ -1041,7 +1042,7 @@ void fpsimd_update_current_state(struct
if (!system_supports_fpsimd())
return;
- local_bh_disable();
+ local_lock_bh(fpsimd_lock);
current->thread.fpsimd_state.user_fpsimd = *state;
if (system_supports_sve() && test_thread_flag(TIF_SVE))
@@ -1052,7 +1053,7 @@ void fpsimd_update_current_state(struct
if (test_and_clear_thread_flag(TIF_FOREIGN_FPSTATE))
fpsimd_bind_to_cpu();
- local_bh_enable();
+ local_unlock_bh(fpsimd_lock);
}
/*
@@ -1115,7 +1116,7 @@ void kernel_neon_begin(void)
BUG_ON(!may_use_simd());
- local_bh_disable();
+ local_lock_bh(fpsimd_lock);
__this_cpu_write(kernel_neon_busy, true);
@@ -1129,8 +1130,14 @@ void kernel_neon_begin(void)
fpsimd_flush_cpu_state();
preempt_disable();
-
- local_bh_enable();
+ /*
+ * ballance atomic vs !atomic context of migrate_disable().
+ * local_lock_bh = get_local_var() + spin_lock_bh (2x migrate_disable)
+ */
+ migrate_disable();
+ migrate_disable();
+ migrate_disable();
+ local_unlock_bh(fpsimd_lock);
}
EXPORT_SYMBOL(kernel_neon_begin);
@@ -1154,6 +1161,10 @@ void kernel_neon_end(void)
WARN_ON(!busy); /* No matching kernel_neon_begin()? */
preempt_enable();
+ /* balance migrate_disable(). See kernel_neon_begin() */
+ migrate_enable();
+ migrate_enable();
+ migrate_enable();
}
EXPORT_SYMBOL(kernel_neon_end);
@@ -1185,9 +1196,7 @@ void __efi_fpsimd_begin(void)
if (!system_supports_fpsimd())
return;
- WARN_ON(preemptible());
-
- if (may_use_simd()) {
+ if (!IS_ENABLED(CONFIG_PREEMPT_RT_BASE) && may_use_simd()) {
kernel_neon_begin();
} else {
/*
--
To unsubscribe from this list: send the line "unsubscribe linux-rt-users" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
