lkml.org 
[lkml]   [2012]   [Feb]   [20]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
Date
From
Subject[PATCH v2 3/3] i387: support lazy restore of FPU state

From: Linus Torvalds <torvalds@linux-foundation.org>
Date: Sun, 19 Feb 2012 13:27:00 -0800
Subject: [PATCH v2 3/3] i387: support lazy restore of FPU state

This makes us recognize when we try to restore FPU state that matches
what we already have in the FPU on this CPU, and avoids the restore
entirely if so.

To do this, we add two new data fields:

- a percpu 'fpu_owner_task' variable that gets written any time we
update the "has_fpu" field, and thus acts as a kind of back-pointer
to the task that owns the CPU. The exception is when we save the FPU
state as part of a context switch - if the save can keep the FPU
state around, we leave the 'fpu_owner_task' variable pointing at the
task whose FP state still remains on the CPU.

- a per-thread 'last_cpu' field, that indicates which CPU that thread
used its FPU on last. We update this on every context switch
(writing an invalid CPU number if the last context switch didn't
leave the FPU in a lazily usable state), so we know that *that*
thread has done nothing else with the FPU since.

These two fields together can be used when next switching back to the
task to see if the CPU still matches: if 'fpu_owner_task' matches the
task we are switching to, we know that no other task (or kernel FPU
usage) touched the FPU on this CPU in the meantime, and if the current
CPU number matches the 'last_cpu' field, we know that this thread did no
other FP work on any other CPU, so the FPU state on the CPU must match
what was saved on last context switch.

In that case, we can avoid the 'f[x]rstor' entirely, and just clear the
CR0.TS bit.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
---

No changes since last time, although the fpu_counter changes nearby
means that the patch itself isn't quite identical. And there's a bit
more commentary on why it's being so anal on task switch time in the
commit log.

Kernel profiles and just timing things both agree: this works. That
said, I don't know how much it matters in real life, but it seems to be
the RightThing(tm) to do, and it really falls out quite nicely without a
lot of complexity.

arch/x86/include/asm/i387.h | 35 +++++++++++++++++++++++------------
arch/x86/include/asm/processor.h | 3 ++-
arch/x86/kernel/cpu/common.c | 2 ++
arch/x86/kernel/process_32.c | 2 +-
arch/x86/kernel/process_64.c | 2 +-
5 files changed, 29 insertions(+), 15 deletions(-)

diff --git a/arch/x86/include/asm/i387.h b/arch/x86/include/asm/i387.h
index 74c607b37e87..247904945d3f 100644
--- a/arch/x86/include/asm/i387.h
+++ b/arch/x86/include/asm/i387.h
@@ -32,6 +32,8 @@ extern int init_fpu(struct task_struct *child);
extern void math_state_restore(void);
extern int dump_fpu(struct pt_regs *, struct user_i387_struct *);

+DECLARE_PER_CPU(struct task_struct *, fpu_owner_task);
+
extern user_regset_active_fn fpregs_active, xfpregs_active;
extern user_regset_get_fn fpregs_get, xfpregs_get, fpregs_soft_get,
xstateregs_get;
@@ -276,7 +278,7 @@ static inline int restore_fpu_checking(struct task_struct *tsk)
"emms\n\t" /* clear stack tags */
"fildl %P[addr]", /* set F?P to defined value */
X86_FEATURE_FXSAVE_LEAK,
- [addr] "m" (tsk->thread.has_fpu));
+ [addr] "m" (tsk->thread.fpu.has_fpu));

return fpu_restore_checking(&tsk->thread.fpu);
}
@@ -288,19 +290,21 @@ static inline int restore_fpu_checking(struct task_struct *tsk)
*/
static inline int __thread_has_fpu(struct task_struct *tsk)
{
- return tsk->thread.has_fpu;
+ return tsk->thread.fpu.has_fpu;
}

/* Must be paired with an 'stts' after! */
static inline void __thread_clear_has_fpu(struct task_struct *tsk)
{
- tsk->thread.has_fpu = 0;
+ tsk->thread.fpu.has_fpu = 0;
+ percpu_write(fpu_owner_task, NULL);
}

/* Must be paired with a 'clts' before! */
static inline void __thread_set_has_fpu(struct task_struct *tsk)
{
- tsk->thread.has_fpu = 1;
+ tsk->thread.fpu.has_fpu = 1;
+ percpu_write(fpu_owner_task, tsk);
}

/*
@@ -345,18 +349,22 @@ typedef struct { int preload; } fpu_switch_t;
* We don't do that yet, so "fpu_lazy_restore()" always returns
* false, but some day..
*/
-#define fpu_lazy_restore(tsk) (0)
-#define fpu_lazy_state_intact(tsk) do { } while (0)
+static inline int fpu_lazy_restore(struct task_struct *new, unsigned int cpu)
+{
+ return new == percpu_read_stable(fpu_owner_task) &&
+ cpu == new->thread.fpu.last_cpu;
+}

-static inline fpu_switch_t switch_fpu_prepare(struct task_struct *old, struct task_struct *new)
+static inline fpu_switch_t switch_fpu_prepare(struct task_struct *old, struct task_struct *new, int cpu)
{
fpu_switch_t fpu;

fpu.preload = tsk_used_math(new) && new->fpu_counter > 5;
if (__thread_has_fpu(old)) {
- if (__save_init_fpu(old))
- fpu_lazy_state_intact(old);
- __thread_clear_has_fpu(old);
+ if (!__save_init_fpu(old))
+ cpu = ~0;
+ old->thread.fpu.last_cpu = cpu;
+ old->thread.fpu.has_fpu = 0; /* But leave fpu_owner_task! */

/* Don't change CR0.TS if we just switch! */
if (fpu.preload) {
@@ -367,9 +375,10 @@ static inline fpu_switch_t switch_fpu_prepare(struct task_struct *old, struct ta
stts();
} else {
old->fpu_counter = 0;
+ old->thread.fpu.last_cpu = ~0;
if (fpu.preload) {
new->fpu_counter++;
- if (fpu_lazy_restore(new))
+ if (fpu_lazy_restore(new, cpu))
fpu.preload = 0;
else
prefetch(new->thread.fpu.state);
@@ -463,8 +472,10 @@ static inline void kernel_fpu_begin(void)
__save_init_fpu(me);
__thread_clear_has_fpu(me);
/* We do 'stts()' in kernel_fpu_end() */
- } else
+ } else {
+ percpu_write(fpu_owner_task, NULL);
clts();
+ }
}

static inline void kernel_fpu_end(void)
diff --git a/arch/x86/include/asm/processor.h b/arch/x86/include/asm/processor.h
index f7c89e231c6c..58545c97d071 100644
--- a/arch/x86/include/asm/processor.h
+++ b/arch/x86/include/asm/processor.h
@@ -374,6 +374,8 @@ union thread_xstate {
};

struct fpu {
+ unsigned int last_cpu;
+ unsigned int has_fpu;
union thread_xstate *state;
};

@@ -454,7 +456,6 @@ struct thread_struct {
unsigned long trap_no;
unsigned long error_code;
/* floating point and extended processor state */
- unsigned long has_fpu;
struct fpu fpu;
#ifdef CONFIG_X86_32
/* Virtual 86 mode info */
diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c
index d43cad74f166..b667148dfad7 100644
--- a/arch/x86/kernel/cpu/common.c
+++ b/arch/x86/kernel/cpu/common.c
@@ -1044,6 +1044,8 @@ DEFINE_PER_CPU(char *, irq_stack_ptr) =

DEFINE_PER_CPU(unsigned int, irq_count) = -1;

+DEFINE_PER_CPU(struct task_struct *, fpu_owner_task);
+
/*
* Special IST stacks which the CPU switches to when it calls
* an IST-marked descriptor entry. Up to 7 stacks (hardware
diff --git a/arch/x86/kernel/process_32.c b/arch/x86/kernel/process_32.c
index bc32761bc27a..c08d1ff12b7c 100644
--- a/arch/x86/kernel/process_32.c
+++ b/arch/x86/kernel/process_32.c
@@ -304,7 +304,7 @@ __switch_to(struct task_struct *prev_p, struct task_struct *next_p)

/* never put a printk in __switch_to... printk() calls wake_up*() indirectly */

- fpu = switch_fpu_prepare(prev_p, next_p);
+ fpu = switch_fpu_prepare(prev_p, next_p, cpu);

/*
* Reload esp0.
diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c
index 8ad880b3bc1c..cfa5c90c01db 100644
--- a/arch/x86/kernel/process_64.c
+++ b/arch/x86/kernel/process_64.c
@@ -389,7 +389,7 @@ __switch_to(struct task_struct *prev_p, struct task_struct *next_p)
unsigned fsindex, gsindex;
fpu_switch_t fpu;

- fpu = switch_fpu_prepare(prev_p, next_p);
+ fpu = switch_fpu_prepare(prev_p, next_p, cpu);

/*
* Reload esp0, LDT and the page table pointer:
--
1.7.9.188.g12766.dirty


\
 
 \ /
  Last update: 2012-02-20 20:51    [W:0.119 / U:67.672 seconds]
©2003-2017 Jasper Spaans. hosted at Digital OceanAdvertise on this site