x86/fpu, sched: Dynamically allocate 'struct fpu'

The FPU rewrite removed the dynamic allocations of 'struct fpu'.
But, this potentially wastes massive amounts of memory (2k per
task on systems that do not have AVX-512 for instance).

Instead of having a separate slab, this patch just appends the
space that we need to the 'task_struct' which we dynamically
allocate already.  This saves from doing an extra slab
allocation at fork().

The only real downside here is that we have to stick everything
and the end of the task_struct.  But, I think the
BUILD_BUG_ON()s I stuck in there should keep that from being too
fragile.

Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: Dave Hansen <dave@sr71.net>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/1437128892-9831-2-git-send-email-mingo@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>

4 files changed, 85 insertions, 38 deletions

diff --git a/arch/x86/include/asm/fpu/types.h b/arch/x86/include/asm/fpu/types.h
index 0637826292de..c49c5173158e 100644
--- a/arch/x86/include/asm/fpu/types.h
+++ b/arch/x86/include/asm/fpu/types.h
@@ -189,6 +189,7 @@ union fpregs_state {
 	struct fxregs_state		fxsave;
 	struct swregs_state		soft;
 	struct xregs_state		xsave;
+	u8 __padding[PAGE_SIZE];
 };
 
 /*
@@ -198,40 +199,6 @@ union fpregs_state {
  */
 struct fpu {
 	/*
-	 * @state:
-	 *
-	 * In-memory copy of all FPU registers that we save/restore
-	 * over context switches. If the task is using the FPU then
-	 * the registers in the FPU are more recent than this state
-	 * copy. If the task context-switches away then they get
-	 * saved here and represent the FPU state.
-	 *
-	 * After context switches there may be a (short) time period
-	 * during which the in-FPU hardware registers are unchanged
-	 * and still perfectly match this state, if the tasks
-	 * scheduled afterwards are not using the FPU.
-	 *
-	 * This is the 'lazy restore' window of optimization, which
-	 * we track though 'fpu_fpregs_owner_ctx' and 'fpu->last_cpu'.
-	 *
-	 * We detect whether a subsequent task uses the FPU via setting
-	 * CR0::TS to 1, which causes any FPU use to raise a #NM fault.
-	 *
-	 * During this window, if the task gets scheduled again, we
-	 * might be able to skip having to do a restore from this
-	 * memory buffer to the hardware registers - at the cost of
-	 * incurring the overhead of #NM fault traps.
-	 *
-	 * Note that on modern CPUs that support the XSAVEOPT (or other
-	 * optimized XSAVE instructions), we don't use #NM traps anymore,
-	 * as the hardware can track whether FPU registers need saving
-	 * or not. On such CPUs we activate the non-lazy ('eagerfpu')
-	 * logic, which unconditionally saves/restores all FPU state
-	 * across context switches. (if FPU state exists.)
-	 */
-	union fpregs_state		state;
-
-	/*
 	 * @last_cpu:
 	 *
 	 * Records the last CPU on which this context was loaded into
@@ -288,6 +255,43 @@ struct fpu {
 	 * deal with bursty apps that only use the FPU for a short time:
 	 */
 	unsigned char			counter;
+	/*
+	 * @state:
+	 *
+	 * In-memory copy of all FPU registers that we save/restore
+	 * over context switches. If the task is using the FPU then
+	 * the registers in the FPU are more recent than this state
+	 * copy. If the task context-switches away then they get
+	 * saved here and represent the FPU state.
+	 *
+	 * After context switches there may be a (short) time period
+	 * during which the in-FPU hardware registers are unchanged
+	 * and still perfectly match this state, if the tasks
+	 * scheduled afterwards are not using the FPU.
+	 *
+	 * This is the 'lazy restore' window of optimization, which
+	 * we track though 'fpu_fpregs_owner_ctx' and 'fpu->last_cpu'.
+	 *
+	 * We detect whether a subsequent task uses the FPU via setting
+	 * CR0::TS to 1, which causes any FPU use to raise a #NM fault.
+	 *
+	 * During this window, if the task gets scheduled again, we
+	 * might be able to skip having to do a restore from this
+	 * memory buffer to the hardware registers - at the cost of
+	 * incurring the overhead of #NM fault traps.
+	 *
+	 * Note that on modern CPUs that support the XSAVEOPT (or other
+	 * optimized XSAVE instructions), we don't use #NM traps anymore,
+	 * as the hardware can track whether FPU registers need saving
+	 * or not. On such CPUs we activate the non-lazy ('eagerfpu')
+	 * logic, which unconditionally saves/restores all FPU state
+	 * across context switches. (if FPU state exists.)
+	 */
+	union fpregs_state		state;
+	/*
+	 * WARNING: 'state' is dynamically-sized.  Do not put
+	 * anything after it here.
+	 */
 };
 
 #endif /* _ASM_X86_FPU_H */
diff --git a/arch/x86/include/asm/processor.h b/arch/x86/include/asm/processor.h
index 43e6519df0d5..944f1785ed0d 100644
--- a/arch/x86/include/asm/processor.h
+++ b/arch/x86/include/asm/processor.h
@@ -390,9 +390,6 @@ struct thread_struct {
 #endif
 	unsigned long		gs;
 
-	/* Floating point and extended processor state */
-	struct fpu		fpu;
-
 	/* Save middle states of ptrace breakpoints */
 	struct perf_event	*ptrace_bps[HBP_NUM];
 	/* Debug status used for traps, single steps, etc... */
@@ -418,6 +415,13 @@ struct thread_struct {
 	unsigned long		iopl;
 	/* Max allowed port in the bitmap, in bytes: */
 	unsigned		io_bitmap_max;
+
+	/* Floating point and extended processor state */
+	struct fpu		fpu;
+	/*
+	 * WARNING: 'fpu' is dynamically-sized.  It *MUST* be at
+	 * the end.
+	 */
 };
 
 /*
diff --git a/arch/x86/kernel/fpu/init.c b/arch/x86/kernel/fpu/init.c
index 32826791e675..deacbfa6b33e 100644
--- a/arch/x86/kernel/fpu/init.c
+++ b/arch/x86/kernel/fpu/init.c
@@ -136,6 +136,45 @@ static void __init fpu__init_system_generic(void)
 unsigned int xstate_size;
 EXPORT_SYMBOL_GPL(xstate_size);
 
+#define CHECK_MEMBER_AT_END_OF(TYPE, MEMBER)	\
+	BUILD_BUG_ON((sizeof(TYPE) -			\
+			offsetof(TYPE, MEMBER) -	\
+			sizeof(((TYPE *)0)->MEMBER)) > 	\
+			0)				\
+
+/*
+ * We append the 'struct fpu' to the task_struct.
+ */
+int __weak arch_task_struct_size(void)
+{
+	int task_size = sizeof(struct task_struct);
+
+	/*
+	 * Subtract off the static size of the register state.
+	 * It potentially has a bunch of padding.
+	 */
+	task_size -= sizeof(((struct task_struct *)0)->thread.fpu.state);
+
+	/*
+	 * Add back the dynamically-calculated register state
+	 * size.
+	 */
+	task_size += xstate_size;
+
+	/*
+	 * We dynamically size 'struct fpu', so we require that
+	 * it be at the end of 'thread_struct' and that
+	 * 'thread_struct' be at the end of 'task_struct'.  If
+	 * you hit a compile error here, check the structure to
+	 * see if something got added to the end.
+	 */
+	CHECK_MEMBER_AT_END_OF(struct fpu, state);
+	CHECK_MEMBER_AT_END_OF(struct thread_struct, fpu);
+	CHECK_MEMBER_AT_END_OF(struct task_struct, thread);
+
+	return task_size;
+}
+
 /*
  * Set up the xstate_size based on the legacy FPU context size.
  *
diff --git a/arch/x86/kernel/process.c b/arch/x86/kernel/process.c
index 9cad694ed7c4..975420eac105 100644
--- a/arch/x86/kernel/process.c
+++ b/arch/x86/kernel/process.c
@@ -81,7 +81,7 @@ EXPORT_SYMBOL_GPL(idle_notifier_unregister);
  */
 int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src)
 {
-	*dst = *src;
+	memcpy(dst, src, arch_task_struct_size());
 
 	return fpu__copy(&dst->thread.fpu, &src->thread.fpu);
 }