ARCv2: fpu: preserve userspace fpu state

Signed-off-by: Vineet Gupta <[email protected]>

Author: vineetgarc

arch/arc/Kconfig

@@ -351,24 +351,19 @@ config NODES_SHIFT
 	  Accessing memory beyond 1GB (with or w/o PAE) requires 2 memory
 	  zones.
 
-if ISA_ARCOMPACT
-
 config ARC_COMPACT_IRQ_LEVELS
+	depends on ISA_ARCOMPACT
 	bool "Setup Timer IRQ as high Priority"
 	# if SMP, LV2 enabled ONLY if ARC implementation has LV2 re-entrancy
 	depends on !SMP
 
 config ARC_FPU_SAVE_RESTORE
 	bool "Enable FPU state persistence across context switch"
 	help
-	  Double Precision Floating Point unit had dedicated regs which
-	  need to be saved/restored across context-switch.
-	  Note that ARC FPU is overly simplistic, unlike say x86, which has
-	  hardware pieces to allow software to conditionally save/restore,
-	  based on actual usage of FPU by a task. Thus our implemn does
-	  this for all tasks in system.
-
-endif #ISA_ARCOMPACT
+	  ARCompact FPU has internal registers to assist with Double precision
+	  Floating Point operations. There are control and stauts registers
+	  for floating point exceptions and rounding modes. These are
+	  preserved across task context switch when enabled.
 
 config ARC_CANT_LLSC
 	def_bool n

arch/arc/include/asm/arcregs.h

@@ -39,6 +39,8 @@
 #define ARC_REG_CLUSTER_BCR	0xcf
 #define ARC_REG_AUX_ICCM	0x208	/* ICCM Base Addr (ARCv2) */
 #define ARC_REG_LPB_CTRL	0x488	/* ARCv2 Loop Buffer control */
+#define ARC_REG_FPU_CTRL	0x300
+#define ARC_REG_FPU_STATUS	0x301
 
 /* Common for ARCompact and ARCv2 status register */
 #define ARC_REG_STATUS32	0x0A

arch/arc/include/asm/fpu.h

@@ -11,18 +11,44 @@
 
 #include <asm/ptrace.h>
 
+#ifdef CONFIG_ISA_ARCOMPACT
+
 /* These DPFP regs need to be saved/restored across ctx-sw */
 struct arc_fpu {
 	struct {
 		unsigned int l, h;
 	} aux_dpfp[2];
 };
 
-extern void fpu_save_restore(struct task_struct *p, struct task_struct *n);
+#define fpu_init_task(regs)
 
 #else
 
+/*
+ * ARCv2 FPU Control aux register
+ *   - bits to enable Traps on Exceptions
+ *   - Rounding mode
+ *
+ * ARCv2 FPU Status aux register
+ *   - FPU exceptions flags (Inv, Div-by-Zero, overflow, underflow, inexact)
+ *   - Flag Write Enable to clear flags explicitly (vs. by fpu instructions
+ *     only
+ */
+
+struct arc_fpu {
+	unsigned int ctrl, status;
+};
+
+extern void fpu_init_task(struct pt_regs *regs);
+
+#endif	/* !CONFIG_ISA_ARCOMPACT */
+
+extern void fpu_save_restore(struct task_struct *p, struct task_struct *n);
+
+#else	/* !CONFIG_ARC_FPU_SAVE_RESTORE */
+
 #define fpu_save_restore(p, n)
+#define fpu_init_task(regs)
 
 #endif	/* CONFIG_ARC_FPU_SAVE_RESTORE */
 

arch/arc/kernel/Makefile

@@ -23,7 +23,9 @@ obj-$(CONFIG_PERF_EVENTS)		+= perf_event.o
 obj-$(CONFIG_JUMP_LABEL)		+= jump_label.o
 
 obj-$(CONFIG_ARC_FPU_SAVE_RESTORE)	+= fpu.o
+ifdef CONFIG_ISA_ARCOMPACT
 CFLAGS_fpu.o   += -mdpfp
+endif
 
 ifdef CONFIG_ARC_DW2_UNWIND
 CFLAGS_ctx_sw.o += -fno-omit-frame-pointer

arch/arc/kernel/fpu.c

@@ -8,6 +8,8 @@
 #include <linux/sched.h>
 #include <asm/fpu.h>
 
+#ifdef CONFIG_ISA_ARCOMPACT
+
 /*
  * To save/restore FPU regs, simplest scheme would use LR/SR insns.
  * However since SR serializes the pipeline, an alternate "hack" can be used
@@ -50,3 +52,28 @@ void fpu_save_restore(struct task_struct *prev, struct task_struct *next)
 		: "r" (zero), "r" (*(readfrom + 3)), "r" (*(readfrom + 2))
 	);
 }
+
+#else
+
+void fpu_init_task(struct pt_regs *regs)
+{
+	/* default rounding mode */
+	write_aux_reg(ARC_REG_FPU_CTRL, 0x100);
+
+	/* set "Write enable" to allow explicit write to exception flags */
+	write_aux_reg(ARC_REG_FPU_STATUS, 0x80000000);
+}
+
+void fpu_save_restore(struct task_struct *prev, struct task_struct *next)
+{
+	struct arc_fpu *save = &prev->thread.fpu;
+	struct arc_fpu *restore = &next->thread.fpu;
+
+	save->ctrl = read_aux_reg(ARC_REG_FPU_CTRL);
+	save->status = read_aux_reg(ARC_REG_FPU_STATUS);
+
+	write_aux_reg(ARC_REG_FPU_CTRL, restore->ctrl);
+	write_aux_reg(ARC_REG_FPU_STATUS, restore->status);
+}
+
+#endif

arch/arc/kernel/process.c

@@ -20,6 +20,8 @@
 #include <linux/elf.h>
 #include <linux/tick.h>
 
+#include <asm/fpu.h>
+
 SYSCALL_DEFINE1(arc_settls, void *, user_tls_data_ptr)
 {
 	task_thread_info(current)->thr_ptr = (unsigned int)user_tls_data_ptr;
@@ -263,7 +265,7 @@ int copy_thread_tls(unsigned long clone_flags, unsigned long usp,
 /*
  * Do necessary setup to start up a new user task
  */
-void start_thread(struct pt_regs * regs, unsigned long pc, unsigned long usp)
+void start_thread(struct pt_regs *regs, unsigned long pc, unsigned long usp)
 {
 	regs->sp = usp;
 	regs->ret = pc;
@@ -279,6 +281,8 @@ void start_thread(struct pt_regs * regs, unsigned long pc, unsigned long usp)
 	regs->eflags = 0;
 #endif
 
+	fpu_init_task(regs);
+
 	/* bogus seed values for debugging */
 	regs->lp_start = 0x10;
 	regs->lp_end = 0x80;