When last_task_used_math gets the cpu, ST0_CU1 is set during context

switch. All other FP programs have ST0_CU1 cleared in their thread_struct,
so that they incur faults on first touching the FPU, and are made the
FPU owner, aka last_task_used_math. used_math determines whether a
thread has used the FPU before; if not, the FPU needs to be initialized
before it can own the FPU. Note: a FP program that has already been
using the fpu (not neccesarily the owner), must reinit the fpu if
it uses the fpu in a signal handler; a sigreturn restores the
original context, discarding the context of the handler.

3 files changed, 90 insertions, 23 deletions

diff --git a/arch/mips64/kernel/process.c b/arch/mips64/kernel/process.c
index c2c077e35..d6d8b7435 100644
--- a/arch/mips64/kernel/process.c
+++ b/arch/mips64/kernel/process.c
@@ -78,7 +78,6 @@ int copy_thread(int nr, unsigned long clone_flags, unsigned long usp,
 	childksp = (unsigned long)p + KERNEL_STACK_SIZE - 32;
 
 	if (last_task_used_math == current) {
-		set_cp0_status(ST0_CU1, ST0_CU1);
 		save_fp(p);
 	}
 	/* set up new TSS. */
diff --git a/arch/mips64/kernel/signal.c b/arch/mips64/kernel/signal.c
index 7823efb3b..40a727831 100644
--- a/arch/mips64/kernel/signal.c
+++ b/arch/mips64/kernel/signal.c
@@ -37,6 +37,35 @@ extern asmlinkage int do_signal(sigset_t *oldset, struct pt_regs *regs);
 extern asmlinkage int save_fp_context(struct sigcontext *sc);
 extern asmlinkage int restore_fp_context(struct sigcontext *sc);
 
+static inline int store_fp_context(struct sigcontext *sc)
+{
+	unsigned int fcr0;
+	int err = 0;
+
+	err |= __copy_to_user(&sc->sc_fpregs[0], 
+		&current->thread.fpu.hard.fp_regs[0], NUM_FPU_REGS * 
+						sizeof(unsigned long));
+	err |= __copy_to_user(&sc->sc_fpc_csr, &current->thread.fpu.hard.control,
+						sizeof(unsigned int));
+	__asm__ __volatile__("cfc1 %0, $0\n\t" : "=r" (fcr0));
+	err |= __copy_to_user(&sc->sc_fpc_eir, &fcr0, sizeof(unsigned int));
+
+	return err;
+}
+
+static inline int refill_fp_context(struct sigcontext *sc)
+{
+	int err = 0;
+
+	if (verify_area(VERIFY_READ, sc, sizeof(*sc)))
+		return -EFAULT;
+	err |= __copy_from_user(&current->thread.fpu.hard.fp_regs[0], 
+			&sc->sc_fpregs[0], NUM_FPU_REGS * sizeof(unsigned long));
+	err |= __copy_from_user(&current->thread.fpu.hard.control, &sc->sc_fpc_csr,
+							sizeof(unsigned int));
+	return err;
+}
+
 /*
  * Atomically swap in the new signal mask, and wait for a signal.
  */
@@ -180,8 +209,12 @@ restore_sigcontext(struct pt_regs *regs, struct sigcontext *sc)
 
 	err |= __get_user(owned_fp, &sc->sc_ownedfp);
 	if (owned_fp) {
-		err |= restore_fp_context(sc);
-		last_task_used_math = current;
+		if (current == last_task_used_math) {
+			last_task_used_math = 0;
+			regs->cp0_status &= ~ST0_CU1;
+		}
+		current->used_math = 1;
+		err |= refill_fp_context(sc);
 	}
 
 	return err;
@@ -281,11 +314,9 @@ badframe:
 static int inline
 setup_sigcontext(struct pt_regs *regs, struct sigcontext *sc)
 {
-	int owned_fp;
 	int err = 0;
 
 	err |= __put_user(regs->cp0_epc, &sc->sc_pc);
-	err |= __put_user(regs->cp0_status, &sc->sc_status);
 
 #define save_gp_reg(i) {						\
 	err |= __put_user(regs->regs[i], &sc->sc_regs[i]);		\
@@ -306,16 +337,19 @@ setup_sigcontext(struct pt_regs *regs, struct sigcontext *sc)
 	err |= __put_user(regs->cp0_cause, &sc->sc_cause);
 	err |= __put_user(regs->cp0_badvaddr, &sc->sc_badvaddr);
 
-	owned_fp = (current == last_task_used_math);
-	err |= __put_user(owned_fp, &sc->sc_ownedfp);
-
 	if (current->used_math) {	/* fp is active.  */
-		set_cp0_status(ST0_CU1, ST0_CU1);
-		err |= save_fp_context(sc);
-		last_task_used_math = NULL;
-		regs->cp0_status &= ~ST0_CU1;
+		if (current == last_task_used_math) {
+			lazy_fpu_switch(current, 0);
+			last_task_used_math = NULL;
+			regs->cp0_status &= ~ST0_CU1;
+		}
+		err |= __put_user(1, &sc->sc_ownedfp);
+		err |= store_fp_context(sc);
 		current->used_math = 0;
+	} else {
+		err |= __put_user(0, &sc->sc_ownedfp);
 	}
+	err |= __put_user(regs->cp0_status, &sc->sc_status);
 
 	return err;
 }
diff --git a/arch/mips64/kernel/signal32.c b/arch/mips64/kernel/signal32.c
index 5a1ee45b0..4addc707c 100644
--- a/arch/mips64/kernel/signal32.c
+++ b/arch/mips64/kernel/signal32.c
@@ -64,6 +64,35 @@ typedef struct sigaltstack32 {
 } stack32_t;
 
 
+static inline int store_fp_context(struct sigcontext *sc)
+{
+	unsigned int fcr0;
+	int err = 0;
+
+	err |= __copy_to_user(&sc->sc_fpregs[0], 
+		&current->thread.fpu.hard.fp_regs[0], NUM_FPU_REGS * 
+						sizeof(unsigned long));
+	err |= __copy_to_user(&sc->sc_fpc_csr, &current->thread.fpu.hard.control,
+						sizeof(unsigned int));
+	__asm__ __volatile__("cfc1 %0, $0\n\t" : "=r" (fcr0));
+	err |= __copy_to_user(&sc->sc_fpc_eir, &fcr0, sizeof(unsigned int));
+
+	return err;
+}
+
+static inline int refill_fp_context(struct sigcontext *sc)
+{
+	int err = 0;
+
+	if (verify_area(VERIFY_READ, sc, sizeof(*sc)))
+		return -EFAULT;
+	err |= __copy_from_user(&current->thread.fpu.hard.fp_regs[0], 
+			&sc->sc_fpregs[0], NUM_FPU_REGS * sizeof(unsigned long));
+	err |= __copy_from_user(&current->thread.fpu.hard.control, &sc->sc_fpc_csr,
+							sizeof(unsigned int));
+	return err;
+}
+
 /*
  * Atomically swap in the new signal mask, and wait for a signal.
  */
@@ -210,8 +239,12 @@ restore_sigcontext(struct pt_regs *regs, struct sigcontext *sc)
 
 	err |= __get_user(owned_fp, &sc->sc_ownedfp);
 	if (owned_fp) {
-		err |= restore_fp_context(sc);
-		last_task_used_math = current;
+		if (current == last_task_used_math) {
+			last_task_used_math = 0;
+			regs->cp0_status &= ~ST0_CU1;
+		}
+		current->used_math = 1;
+		err |= refill_fp_context(sc);
 	}
 
 	return err;
@@ -316,11 +349,9 @@ badframe:
 static int inline
 setup_sigcontext(struct pt_regs *regs, struct sigcontext *sc)
 {
-	int owned_fp;
 	int err = 0;
 
 	err |= __put_user(regs->cp0_epc, &sc->sc_pc);
-	err |= __put_user(regs->cp0_status, &sc->sc_status);
 
 #define save_gp_reg(i) {						\
 	err |= __put_user(regs->regs[i], &sc->sc_regs[i]);		\
@@ -341,16 +372,19 @@ setup_sigcontext(struct pt_regs *regs, struct sigcontext *sc)
 	err |= __put_user(regs->cp0_cause, &sc->sc_cause);
 	err |= __put_user(regs->cp0_badvaddr, &sc->sc_badvaddr);
 
-	owned_fp = (current == last_task_used_math);
-	err |= __put_user(owned_fp, &sc->sc_ownedfp);
-
 	if (current->used_math) {	/* fp is active.  */
-		set_cp0_status(ST0_CU1, ST0_CU1);
-		err |= save_fp_context(sc);
-		last_task_used_math = NULL;
-		regs->cp0_status &= ~ST0_CU1;
+		if (current == last_task_used_math) {
+			lazy_fpu_switch(current, 0);
+			last_task_used_math = NULL;
+			regs->cp0_status &= ~ST0_CU1;
+		}
+		err |= __put_user(1, &sc->sc_ownedfp);
+		err |= store_fp_context(sc);
 		current->used_math = 0;
+	} else {
+		err |= __put_user(0, &sc->sc_ownedfp);
 	}
+	err |= __put_user(regs->cp0_status, &sc->sc_status);
 
 	return err;
 }