Floating-point emulation code for systems lacking proper FPU (i.e. Octeon),

enabled by option FPUEMUL.

This is pretty straightforward, except for conditional branch on FPU condition
codes emulation (bc1f/bc1fl/bc1t/bc1tl instructions): unlike most
RISC-with-delay-slots designs (m88k, sparc), the branch pipeline is not exposed
to the kernel on Mips, therefore we can not resume a branch without losing the
delay slot instruction.

Some other operating systems work around this issue by emulating the delay
slot instruction, but this is error-prone (and requires the kernel code to
be aware of all supported instructions of the processor it is currently running
on), some use dedicated breakpoints to single-step through the delay slot and
then resume the branch as expected, but this causes a lot of copy-on-write
allocations.

This code chooses a third path, of copying the delay slot instructions to run toa special `magic' page, followed by a special trap instruction to give control
back to the kernel. This makes sure the instruction will actually be run by the
processor, and that no more than one page per process is wasted, regardless of
the number of branches to emulate.

Tested on octeon (big-endian) by syuu@ and on loongson (little-endian) by me.
Note that enabling option FPUEMUL in the kernel will completely disable the
hardware FPU, if there is one; there is currently no way to build a kernel
supporting both hardware and software FPU, and there is no reason to change
this until there is a strong need to support both.

3 files changed, 22 insertions, 4 deletions

diff --git a/sys/arch/mips64/include/_types.h b/sys/arch/mips64/include/_types.h
index bd59a97e6e8..e22ed5c01bf 100644
--- a/sys/arch/mips64/include/_types.h
+++ b/sys/arch/mips64/include/_types.h
@@ -1,4 +1,4 @@
-/*	$OpenBSD: _types.h,v 1.7 2009/11/04 19:14:10 kettenis Exp $	*/
+/*	$OpenBSD: _types.h,v 1.8 2010/11/24 21:16:26 miod Exp $	*/
 
 /*-
  * Copyright (c) 1990, 1993
@@ -135,5 +135,6 @@ typedef struct label_t {
 
 /* Feature test macros */
 #define __HAVE_TIMECOUNTER
+#define	__HAVE_EXEC_MD_MAP
 
 #endif	/* _MIPS64__TYPES_H_ */
diff --git a/sys/arch/mips64/include/cpu.h b/sys/arch/mips64/include/cpu.h
index a7843f21435..2ef12590d9d 100644
--- a/sys/arch/mips64/include/cpu.h
+++ b/sys/arch/mips64/include/cpu.h
@@ -1,4 +1,4 @@
-/*	$OpenBSD: cpu.h,v 1.68 2010/10/24 15:40:03 miod Exp $	*/
+/*	$OpenBSD: cpu.h,v 1.69 2010/11/24 21:16:26 miod Exp $	*/
 
 /*-
  * Copyright (c) 1992, 1993
@@ -279,11 +279,13 @@ extern vaddr_t uncached_base;
 #define	BREAK_BRKPT_VAL		514
 #define	BREAK_SOVER_VAL		515
 #define	BREAK_DDB_VAL		516
+#define	BREAK_FPUEMUL_VAL	517
 #define	BREAK_KDB	(BREAK_INSTR | (BREAK_KDB_VAL << BREAK_VAL_SHIFT))
 #define	BREAK_SSTEP	(BREAK_INSTR | (BREAK_SSTEP_VAL << BREAK_VAL_SHIFT))
 #define	BREAK_BRKPT	(BREAK_INSTR | (BREAK_BRKPT_VAL << BREAK_VAL_SHIFT))
 #define	BREAK_SOVER	(BREAK_INSTR | (BREAK_SOVER_VAL << BREAK_VAL_SHIFT))
 #define	BREAK_DDB	(BREAK_INSTR | (BREAK_DDB_VAL << BREAK_VAL_SHIFT))
+#define	BREAK_FPUEMUL	(BREAK_INSTR | (BREAK_FPUEMUL_VAL << BREAK_VAL_SHIFT))
 
 /*
  * The floating point version and status registers.
@@ -550,6 +552,7 @@ void cpu_startclock(struct cpu_info *);
 
 extern vaddr_t CpuCacheAliasMask;
 
+struct exec_package;
 struct tlb_entry;
 struct user;
 
@@ -604,10 +607,13 @@ int	tlb_update(vaddr_t, unsigned);
 void	tlb_read(int, struct tlb_entry *);
 
 void	build_trampoline(vaddr_t, vaddr_t);
+int	exec_md_map(struct proc *, struct exec_package *);
 void	savectx(struct user *, int);
 
 void	enable_fpu(struct proc *);
 void	save_fpu(void);
+int	fpe_branch_emulate(struct proc *, struct trap_frame *, uint32_t,
+	    vaddr_t);
 
 int	guarded_read_4(paddr_t, uint32_t *);
 int	guarded_write_4(paddr_t, uint32_t);
diff --git a/sys/arch/mips64/include/proc.h b/sys/arch/mips64/include/proc.h
index b78d8f109e8..f4d31e7201a 100644
--- a/sys/arch/mips64/include/proc.h
+++ b/sys/arch/mips64/include/proc.h
@@ -1,4 +1,4 @@
-/*	$OpenBSD: proc.h,v 1.6 2010/11/24 21:01:02 miod Exp $	*/
+/*	$OpenBSD: proc.h,v 1.7 2010/11/24 21:16:26 miod Exp $	*/
 
 /*
  * Copyright (c) 1992, 1993
@@ -50,6 +50,11 @@ struct mdproc {
 	vaddr_t	md_ss_addr;		/* single step address */
 	uint32_t md_ss_instr;		/* saved single step instruction */
 
+	/* fpu emulation fields */
+	vaddr_t	md_fppgva;		/* vaddr of the branch emulation page */
+	vaddr_t	md_fpbranchva;		/* vaddr of fp branch destination */
+	vaddr_t	md_fpslotva;		/* initial vaddr of delay slot */
+
 /* The following is RM7000 dependent, but kept in for compatibility */
 	int	md_pc_ctrl;		/* performance counter control */
 	int	md_pc_count;		/* performance counter */
@@ -59,7 +64,13 @@ struct mdproc {
 	int	md_watch_m;
 };
 
-/* md_flags */
+/*
+ * Values for md_flags.
+ * MDP_FPUSED has two meanings: if the floating point hardware (coprocessor
+ * #1) is available, it means it has been used; if there is no floating
+ * point hardware, it means the process is currently running a duplicated
+ * delay slot, created by the branch emulation logic.
+ */
 #define	MDP_FPUSED	0x00000001	/* floating point coprocessor used */
 #define	MDP_PERF	0x00010000	/* Performance counter used */
 #define	MDP_WATCH1	0x00020000	/* Watch register 1 used */