Only allocate the FPU area when the application actually uses FPU, i.e., in the
first lazy FPU trap. This could save memory for non-fpu using apps.
for example: on my system after boot, there are around 300 processes, with
only 17 using FPU.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Cc: Arjan van de Ven <arjan@linux.intel.com>
---
v5: Processor device not available exception automatically disables interrupts.
Handle interrupts properly while doing blocking calls. And use SIGKILL
instead of SIGSEGV during OOM failures(consistent with other out of memory
handling cases we have today. for ex: oom handling in page fault handler).
v4: Handles the kmem_cache_alloc() failures.
v3: Fixed the non-atomic calling sequence in atomic context.
v2: Ported to x86.git#testing with some name changes.
---
Index: linux-2.6-x86/arch/x86/kernel/i387.c
===================================================================
--- linux-2.6-x86.orig/arch/x86/kernel/i387.c 2008-03-10 14:26:35.000000000 -0700
+++ linux-2.6-x86/arch/x86/kernel/i387.c 2008-03-10 14:40:53.000000000 -0700
@@ -8,7 +8,6 @@
#include <linux/module.h>
#include <linux/regset.h>
#include <linux/sched.h>
-#include <linux/bootmem.h>
#include <asm/sigcontext.h>
#include <asm/processor.h>
@@ -63,7 +62,6 @@
else
xstate_size = sizeof(struct i387_fsave_struct);
#endif
- init_task.thread.xstate = alloc_bootmem(xstate_size);
}
#ifdef CONFIG_X86_64
@@ -93,12 +91,22 @@
* value at reset if we support XMM instructions and then
* remeber the current task has used the FPU.
*/
-void init_fpu(struct task_struct *tsk)
+int init_fpu(struct task_struct *tsk)
{
if (tsk_used_math(tsk)) {
if (tsk == current)
unlazy_fpu(tsk);
- return;
+ return 0;
+ }
+
+ /*
+ * Memory allocation at the first usage of the FPU and other state.
+ */
+ if (!tsk->thread.xstate) {
+ tsk->thread.xstate = kmem_cache_alloc(task_xstate_cachep,
+ GFP_KERNEL);
+ if (!tsk->thread.xstate)
+ return -ENOMEM;
}
if (cpu_has_fxsr) {
@@ -120,6 +128,7 @@
* Only the device not available exception or ptrace can call init_fpu.
*/
set_stopped_child_used_math(tsk);
+ return 0;
}
int fpregs_active(struct task_struct *target, const struct user_regset *regset)
@@ -136,10 +145,14 @@
unsigned int pos, unsigned int count,
void *kbuf, void __user *ubuf)
{
+ int ret;
+
if (!cpu_has_fxsr)
return -ENODEV;
- init_fpu(target);
+ ret = init_fpu(target);
+ if (ret)
+ return ret;
return user_regset_copyout(&pos, &count, &kbuf, &ubuf,
&target->thread.xstate->fxsave, 0, -1);
@@ -154,7 +167,10 @@
if (!cpu_has_fxsr)
return -ENODEV;
- init_fpu(target);
+ ret = init_fpu(target);
+ if (ret)
+ return ret;
+
set_stopped_child_used_math(target);
ret = user_regset_copyin(&pos, &count, &kbuf, &ubuf,
@@ -312,11 +328,14 @@
void *kbuf, void __user *ubuf)
{
struct user_i387_ia32_struct env;
+ int ret;
if (!HAVE_HWFP)
return fpregs_soft_get(target, regset, pos, count, kbuf, ubuf);
- init_fpu(target);
+ ret = init_fpu(target);
+ if (ret)
+ return ret;
if (!cpu_has_fxsr) {
return user_regset_copyout(&pos, &count, &kbuf, &ubuf,
@@ -344,7 +363,10 @@
if (!HAVE_HWFP)
return fpregs_soft_set(target, regset, pos, count, kbuf, ubuf);
- init_fpu(target);
+ ret = init_fpu(target);
+ if (ret)
+ return ret;
+
set_stopped_child_used_math(target);
if (!cpu_has_fxsr) {
Index: linux-2.6-x86/arch/x86/kernel/process.c
===================================================================
--- linux-2.6-x86.orig/arch/x86/kernel/process.c 2008-03-10 14:26:35.000000000 -0700
+++ linux-2.6-x86/arch/x86/kernel/process.c 2008-03-10 14:26:36.000000000 -0700
@@ -5,24 +5,34 @@
#include <linux/slab.h>
#include <linux/sched.h>
-static struct kmem_cache *task_xstate_cachep;
+struct kmem_cache *task_xstate_cachep;
int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src)
{
*dst = *src;
- dst->thread.xstate = kmem_cache_alloc(task_xstate_cachep, GFP_KERNEL);
- if (!dst->thread.xstate)
- return -ENOMEM;
- WARN_ON((unsigned long)dst->thread.xstate & 15);
- memcpy(dst->thread.xstate, src->thread.xstate, xstate_size);
+ if (src->thread.xstate) {
+ dst->thread.xstate = kmem_cache_alloc(task_xstate_cachep,
+ GFP_KERNEL);
+ if (!dst->thread.xstate)
+ return -ENOMEM;
+ WARN_ON((unsigned long)dst->thread.xstate & 15);
+ memcpy(dst->thread.xstate, src->thread.xstate, xstate_size);
+ }
return 0;
}
-void free_thread_info(struct thread_info *ti)
+void free_thread_xstate(struct task_struct *tsk)
{
- kmem_cache_free(task_xstate_cachep, ti->task->thread.xstate);
- ti->task->thread.xstate = NULL;
+ if (tsk->thread.xstate) {
+ kmem_cache_free(task_xstate_cachep, tsk->thread.xstate);
+ tsk->thread.xstate = NULL;
+ }
+}
+
+void free_thread_info(struct thread_info *ti)
+{
+ free_thread_xstate(ti->task);
free_pages((unsigned long)(ti), get_order(THREAD_SIZE));
}
Index: linux-2.6-x86/include/asm-x86/processor.h
===================================================================
--- linux-2.6-x86.orig/include/asm-x86/processor.h 2008-03-10 14:26:35.000000000 -0700
+++ linux-2.6-x86/include/asm-x86/processor.h 2008-03-10 14:26:36.000000000 -0700
@@ -367,6 +367,8 @@
extern void print_cpu_info(struct cpuinfo_x86 *);
extern unsigned int xstate_size;
+extern void free_thread_xstate(struct task_struct *);
+extern struct kmem_cache *task_xstate_cachep;
extern void init_scattered_cpuid_features(struct cpuinfo_x86 *c);
extern unsigned int init_intel_cacheinfo(struct cpuinfo_x86 *c);
extern unsigned short num_cache_leaves;
Index: linux-2.6-x86/arch/x86/kernel/process_32.c
===================================================================
--- linux-2.6-x86.orig/arch/x86/kernel/process_32.c 2008-03-10 14:26:35.000000000 -0700
+++ linux-2.6-x86/arch/x86/kernel/process_32.c 2008-03-10 14:26:36.000000000 -0700
@@ -524,6 +524,10 @@
regs->cs = __USER_CS;
regs->ip = new_ip;
regs->sp = new_sp;
+ /*
+ * Free the old FP and other extended state
+ */
+ free_thread_xstate(current);
}
EXPORT_SYMBOL_GPL(start_thread);
Index: linux-2.6-x86/arch/x86/kernel/process_64.c
===================================================================
--- linux-2.6-x86.orig/arch/x86/kernel/process_64.c 2008-03-10 14:26:35.000000000 -0700
+++ linux-2.6-x86/arch/x86/kernel/process_64.c 2008-03-10 14:26:36.000000000 -0700
@@ -552,6 +552,10 @@
regs->ss = __USER_DS;
regs->flags = 0x200;
set_fs(USER_DS);
+ /*
+ * Free the old FP and other extended state
+ */
+ free_thread_xstate(current);
}
EXPORT_SYMBOL_GPL(start_thread);
Index: linux-2.6-x86/arch/x86/kernel/traps_32.c
===================================================================
--- linux-2.6-x86.orig/arch/x86/kernel/traps_32.c 2008-03-10 14:26:35.000000000 -0700
+++ linux-2.6-x86/arch/x86/kernel/traps_32.c 2008-03-10 15:16:50.000000000 -0700
@@ -1168,9 +1168,22 @@
struct thread_info *thread = current_thread_info();
struct task_struct *tsk = thread->task;
+ if (!tsk_used_math(tsk)) {
+ local_irq_enable();
+ /*
+ * does a slab alloc which can sleep
+ */
+ if (init_fpu(tsk)) {
+ /*
+ * ran out of memory!
+ */
+ do_group_exit(SIGKILL);
+ return;
+ }
+ local_irq_disable();
+ }
+
clts(); /* Allow maths ops (or we recurse) */
- if (!tsk_used_math(tsk))
- init_fpu(tsk);
restore_fpu(tsk);
thread->status |= TS_USEDFPU; /* So we fnsave on switch_to() */
tsk->fpu_counter++;
Index: linux-2.6-x86/arch/x86/kernel/traps_64.c
===================================================================
--- linux-2.6-x86.orig/arch/x86/kernel/traps_64.c 2008-03-10 14:26:35.000000000 -0700
+++ linux-2.6-x86/arch/x86/kernel/traps_64.c 2008-03-10 15:17:03.000000000 -0700
@@ -1117,10 +1117,23 @@
asmlinkage void math_state_restore(void)
{
struct task_struct *me = current;
- clts(); /* Allow maths ops (or we recurse) */
- if (!used_math())
- init_fpu(me);
+ if (!used_math()) {
+ local_irq_enable();
+ /*
+ * does a slab alloc which can sleep
+ */
+ if (init_fpu(me)) {
+ /*
+ * ran out of memory!
+ */
+ do_group_exit(SIGKILL);
+ return;
+ }
+ local_irq_disable();
+ }
+
+ clts(); /* Allow maths ops (or we recurse) */
restore_fpu_checking(&me->thread.xstate->fxsave);
task_thread_info(me)->status |= TS_USEDFPU;
me->fpu_counter++;
Index: linux-2.6-x86/include/asm-x86/i387.h
===================================================================
--- linux-2.6-x86.orig/include/asm-x86/i387.h 2008-03-10 14:26:35.000000000 -0700
+++ linux-2.6-x86/include/asm-x86/i387.h 2008-03-10 14:26:36.000000000 -0700
@@ -21,7 +21,7 @@
extern void fpu_init(void);
extern void mxcsr_feature_mask_init(void);
-extern void init_fpu(struct task_struct *child);
+extern int init_fpu(struct task_struct *child);
extern asmlinkage void math_state_restore(void);
extern void init_thread_xstate(void);
--
--
