[ARM] Eliminate tsk->used_math

Remove usage of tsk->used_math on ARM, moving the status to an array
of co-processor usage.  (ARM can have up to 15 co-processors
providing various extra facilities such as SIMD, VFP or FP.)

arch/arm/kernel/asm-offsets.c

@@ -45,7 +45,6 @@
 
 int main(void)
 {
-  DEFINE(TSK_USED_MATH,		offsetof(struct task_struct, used_math));
   DEFINE(TSK_ACTIVE_MM,		offsetof(struct task_struct, active_mm));
   BLANK();
   DEFINE(VMA_VM_MM,		offsetof(struct vm_area_struct, vm_mm));

arch/arm/kernel/entry-armv.S

@@ -963,23 +963,46 @@ __und_usr:	sub	sp, sp, #S_FRAME_SIZE		@ Allocate frame size in one go
  * co-processor instructions.  However, we have to watch out
  * for the ARM6/ARM7 SWI bug.
  *
- * Emulators may wish to make use of the instruction value we
- * prepared for them in r0.
+ * Emulators may wish to make use of the following registers:
+ *  r0  - instruction opcode.
+ *  r10 - this threads thread_info structure.
  */
 call_fpe:	enable_irq r10				@ Enable interrupts
 		tst	r0, #0x08000000			@ only CDP/CPRT/LDC/STC have bit 27
 #if defined(CONFIG_CPU_ARM610) || defined(CONFIG_CPU_ARM710)
-		and	r10, r0, #0x0f000000		@ mask out op-code bits
-		teqne	r10, #0x0f000000		@ SWI (ARM6/7 bug)?
+		and	r8, r0, #0x0f000000		@ mask out op-code bits
+		teqne	r8, #0x0f000000			@ SWI (ARM6/7 bug)?
 #endif
 		moveq	pc, lr
-do_fpe:		get_thread_info r10			@ get current thread
-		ldr	r4, [r10, #TI_TASK]		@ get current task
-		mov	r8, #1
-		strb	r8, [r4, #TSK_USED_MATH]	@ set current->used_math
-		ldr	r4, .LCfp
+		get_thread_info r10			@ get current thread
+		and	r8, r0, #0x00000f00		@ mask out CP number
+		mov	r7, #1
+		add	r6, r10, #TI_USED_CP
+		strb	r7, [r6, r8, lsr #8]		@ set appropriate used_cp[]
+		add	pc, pc, r8, lsr #6
+		mov	r0, r0
+
+		mov	pc, lr				@ CP#0
+		b	do_fpe				@ CP#1 (FPE)
+		b	do_fpe				@ CP#2 (FPE)
+		mov	pc, lr				@ CP#3
+		mov	pc, lr				@ CP#4
+		mov	pc, lr				@ CP#5
+		mov	pc, lr				@ CP#6
+		mov	pc, lr				@ CP#7
+		mov	pc, lr				@ CP#8
+		mov	pc, lr				@ CP#9
+		mov	pc, lr				@ CP#10 (VFP)
+		mov	pc, lr				@ CP#11 (VFP)
+		mov	pc, lr				@ CP#12
+		mov	pc, lr				@ CP#13
+		mov	pc, lr				@ CP#14 (Debug)
+		mov	pc, lr				@ CP#15 (Control)
+
+do_fpe:		ldr	r4, .LCfp
 		add	r10, r10, #TI_FPSTATE		@ r10 = workspace
 		ldr	pc, [r4]			@ Call FP module USR entry point
+
 /*
  * The FP module is called with these registers set:
  *  r0  = instruction
@@ -989,6 +1012,11 @@ do_fpe:		get_thread_info r10			@ get current thread
  *  lr  = unrecognised FP instruction return address
  */
 
+		.data
+ENTRY(fp_enter)
+		.word	fpe_not_present
+		.text
+
 fpundefinstr:	mov	r0, sp
 		adrsvc	al, lr, ret_from_exception
 		b	do_undefinstr
@@ -1016,10 +1044,6 @@ ENTRY(ret_from_exception)
 		mov	why, #0
 		b	ret_to_user
 
-		.data
-ENTRY(fp_enter)
-		.word	fpe_not_present
-		.text
 /*
  * Register switch for ARMv3 and ARMv4 processors
  * r0 = previous thread_info, r1 = next thread_info

arch/arm/kernel/process.c

@@ -307,8 +307,7 @@ void flush_thread(void)
 	struct thread_info *thread = current_thread_info();
 	struct task_struct *tsk = current;
 
-	tsk->used_math = 0;
-
+	memset(thread->used_cp, 0, sizeof(thread->used_cp));
 	memset(&tsk->thread.debug, 0, sizeof(struct debug_info));
 	fp_init(&thread->fpstate);
 }
@@ -344,12 +343,12 @@ copy_thread(int nr, unsigned long clone_flags, unsigned long stack_start,
 int dump_fpu (struct pt_regs *regs, struct user_fp *fp)
 {
 	struct thread_info *thread = current_thread_info();
-	int used_math = current->used_math;
+	int used_math = thread->used_cp[1] | thread->used_cp[2];
 
 	if (used_math)
 		memcpy(fp, &thread->fpstate.soft, sizeof (*fp));
 
-	return used_math;
+	return used_math != 0;
 }
 
 /*

arch/arm/kernel/ptrace.c

@@ -602,8 +602,9 @@ static int ptrace_getfpregs(struct task_struct *tsk, void *ufp)
  */
 static int ptrace_setfpregs(struct task_struct *tsk, void *ufp)
 {
-	tsk->used_math = 1;
-	return copy_from_user(&tsk->thread_info->fpstate, ufp,
+	struct thread_info *thread = tsk->thread_info;
+	thread->used_cp[1] = thread->used_cp[2] = 1;
+	return copy_from_user(&thread->fpstate, ufp,
 			      sizeof(struct user_fp)) ? -EFAULT : 0;
 }
 

include/asm-arm/thread_info.h

@@ -51,8 +51,9 @@ struct thread_info {
 	__u32			cpu;		/* cpu */
 	__u32			cpu_domain;	/* cpu domain */
 	struct cpu_context_save	cpu_context;	/* cpu context */
-	struct restart_block    restart_block;
+	__u8			used_cp[16];	/* thread used copro */
 	union fp_state		fpstate;
+	struct restart_block    restart_block;
 };
 
 #define INIT_THREAD_INFO(tsk)						\
@@ -107,7 +108,8 @@ extern void free_thread_info(struct thread_info *);
 #define TI_CPU		20
 #define TI_CPU_DOMAIN	24
 #define TI_CPU_SAVE	28
-#define TI_FPSTATE	76
+#define TI_USED_MATH	76
+#define TI_FPSTATE	(TI_USED_MATH+16)
 
 #endif