1 files changed, 58 insertions, 43 deletions
diff --git a/arch/powerpc/platforms/pseries/hvCall.S b/arch/powerpc/platforms/pseries/hvCall.S
index c1427b3634ec..01e95ab18d35 100644
--- a/arch/powerpc/platforms/pseries/hvCall.S
+++ b/arch/powerpc/platforms/pseries/hvCall.S
@@ -14,20 +14,54 @@
        
 #define STK_PARM(i)     (48 + ((i)-3)*8)
-#ifdef CONFIG_HCALL_STATS
+#ifdef CONFIG_TRACEPOINTS
+        .section        ".toc","aw"
+        .globl hcall_tracepoint_refcount
+hcall_tracepoint_refcount:
+        .llong  0
+        .section        ".text"
 /*
 * precall must preserve all registers.  use unused STK_PARM()
- * areas to save snapshots and opcode.
+ * areas to save snapshots and opcode. We branch around this
+ * in early init (eg when populating the MMU hashtable) by using an
+ * unconditional cpu feature.
 */
 #define HCALL_INST_PRECALL                                      \
-        std     r3,STK_PARM(r3)(r1);    /* save opcode */       \
-        mftb    r0;                     /* get timebase and */  \
-        std     r0,STK_PARM(r5)(r1);    /* save for later */    \
 BEGIN_FTR_SECTION;                                              \
-        mfspr   r0,SPRN_PURR;           /* get PURR and */      \
+        b       1f;                                             \
-        std     r0,STK_PARM(r6)(r1);    /* save for later */    \
+END_FTR_SECTION(0, 1);                                          \
-END_FTR_SECTION_IFSET(CPU_FTR_PURR);
+        ld      r12,hcall_tracepoint_refcount@toc(r2);          \
-        
+        cmpdi   r12,0;                                          \
+        beq+    1f;                                             \
+        mflr    r0;                                             \
+        std     r3,STK_PARM(r3)(r1);                            \
+        std     r4,STK_PARM(r4)(r1);                            \
+        std     r5,STK_PARM(r5)(r1);                            \
+        std     r6,STK_PARM(r6)(r1);                            \
+        std     r7,STK_PARM(r7)(r1);                            \
+        std     r8,STK_PARM(r8)(r1);                            \
+        std     r9,STK_PARM(r9)(r1);                            \
+        std     r10,STK_PARM(r10)(r1);                          \
+        std     r0,16(r1);                                      \
+        stdu    r1,-STACK_FRAME_OVERHEAD(r1);                   \
+        bl      .__trace_hcall_entry;                           \
+        addi    r1,r1,STACK_FRAME_OVERHEAD;                     \
+        ld      r0,16(r1);                                      \
+        ld      r3,STK_PARM(r3)(r1);                            \
+        ld      r4,STK_PARM(r4)(r1);                            \
+        ld      r5,STK_PARM(r5)(r1);                            \
+        ld      r6,STK_PARM(r6)(r1);                            \
+        ld      r7,STK_PARM(r7)(r1);                            \
+        ld      r8,STK_PARM(r8)(r1);                            \
+        ld      r9,STK_PARM(r9)(r1);                            \
+        ld      r10,STK_PARM(r10)(r1);                          \
+        mtlr    r0;                                             \
+1:
 /*
 * postcall is performed immediately before function return which
 * allows liberal use of volatile registers.  We branch around this
@@ -38,40 +72,21 @@ END_FTR_SECTION_IFSET(CPU_FTR_PURR);
 BEGIN_FTR_SECTION;                                              \
        b       1f;                                             \
 END_FTR_SECTION(0, 1);                                          \
-        ld      r4,STK_PARM(r3)(r1);    /* validate opcode */   \
+        ld      r12,hcall_tracepoint_refcount@toc(r2);          \
-        cmpldi  cr7,r4,MAX_HCALL_OPCODE;                        \
+        cmpdi   r12,0;                                          \
-        bgt-    cr7,1f;                                         \
+        beq+    1f;                                             \
-                                                                \
+        mflr    r0;                                             \
-        /* get time and PURR snapshots after hcall */           \
+        ld      r6,STK_PARM(r3)(r1);                            \
-        mftb    r7;                     /* timebase after */    \
+        std     r3,STK_PARM(r3)(r1);                            \
-BEGIN_FTR_SECTION;                                              \
+        mr      r4,r3;                                          \
-        mfspr   r8,SPRN_PURR;           /* PURR after */        \
+        mr      r3,r6;                                          \
-        ld      r6,STK_PARM(r6)(r1);    /* PURR before */       \
+        std     r0,16(r1);                                      \
-        subf    r6,r6,r8;               /* delta */             \
+        stdu    r1,-STACK_FRAME_OVERHEAD(r1);                   \
-END_FTR_SECTION_IFSET(CPU_FTR_PURR);                            \
+        bl      .__trace_hcall_exit;                            \
-        ld      r5,STK_PARM(r5)(r1);    /* timebase before */   \
+        addi    r1,r1,STACK_FRAME_OVERHEAD;                     \
-        subf    r5,r5,r7;               /* time delta */        \
+        ld      r0,16(r1);                                      \
-                                                                \
+        ld      r3,STK_PARM(r3)(r1);                            \
-        /* calculate address of stat structure r4 = opcode */   \
+        mtlr    r0;                                             \
-        srdi    r4,r4,2;                /* index into array */  \
-        mulli   r4,r4,HCALL_STAT_SIZE;                          \
-        LOAD_REG_ADDR(r7, per_cpu__hcall_stats);                \
-        add     r4,r4,r7;                                       \
-        ld      r7,PACA_DATA_OFFSET(r13); /* per cpu offset */  \
-        add     r4,r4,r7;                                       \
-                                                                \
-        /* update stats */                                      \
-        ld      r7,HCALL_STAT_CALLS(r4); /* count */            \
-        addi    r7,r7,1;                                        \
-        std     r7,HCALL_STAT_CALLS(r4);                        \
-        ld      r7,HCALL_STAT_TB(r4);   /* timebase */          \
-        add     r7,r7,r5;                                       \
-        std     r7,HCALL_STAT_TB(r4);                           \
-BEGIN_FTR_SECTION;                                              \
-        ld      r7,HCALL_STAT_PURR(r4); /* PURR */              \
-        add     r7,r7,r6;                                       \
-        std     r7,HCALL_STAT_PURR(r4);                         \
-END_FTR_SECTION_IFSET(CPU_FTR_PURR);                            \
 1:
 #else
 #define HCALL_INST_PRECALL

diff --git a/arch/powerpc/platforms/pseries/hvCall.S b/arch/powerpc/platforms/pseries/hvCall.S index c1427b3634ec..01e95ab18d35 100644 --- a/arch/powerpc/platforms/pseries/hvCall.S +++ b/arch/powerpc/platforms/pseries/hvCall.S
@@ -14,20 +14,54 @@
14		14
15	#define STK_PARM(i) (48 + ((i)-3)*8)	15	#define STK_PARM(i) (48 + ((i)-3)*8)
16		16
17	#ifdef CONFIG_HCALL_STATS	17	#ifdef CONFIG_TRACEPOINTS
		18
		19	.section ".toc","aw"
		20
		21	.globl hcall_tracepoint_refcount
		22	hcall_tracepoint_refcount:
		23	.llong 0
		24
		25	.section ".text"
		26
18	/*	27	/*
19	* precall must preserve all registers. use unused STK_PARM()	28	* precall must preserve all registers. use unused STK_PARM()
20	* areas to save snapshots and opcode.	29	* areas to save snapshots and opcode. We branch around this
		30	* in early init (eg when populating the MMU hashtable) by using an
		31	* unconditional cpu feature.
21	*/	32	*/
22	#define HCALL_INST_PRECALL \	33	#define HCALL_INST_PRECALL \
23	std r3,STK_PARM(r3)(r1); /* save opcode */ \
24	mftb r0; /* get timebase and */ \
25	std r0,STK_PARM(r5)(r1); /* save for later */ \
26	BEGIN_FTR_SECTION; \	34	BEGIN_FTR_SECTION; \
27	mfspr r0,SPRN_PURR; /* get PURR and */ \	35	b 1f; \
28	std r0,STK_PARM(r6)(r1); /* save for later */ \	36	END_FTR_SECTION(0, 1); \
29	END_FTR_SECTION_IFSET(CPU_FTR_PURR);	37	ld r12,hcall_tracepoint_refcount@toc(r2); \
30		38	cmpdi r12,0; \
		39	beq+ 1f; \
		40	mflr r0; \
		41	std r3,STK_PARM(r3)(r1); \
		42	std r4,STK_PARM(r4)(r1); \
		43	std r5,STK_PARM(r5)(r1); \
		44	std r6,STK_PARM(r6)(r1); \
		45	std r7,STK_PARM(r7)(r1); \
		46	std r8,STK_PARM(r8)(r1); \
		47	std r9,STK_PARM(r9)(r1); \
		48	std r10,STK_PARM(r10)(r1); \
		49	std r0,16(r1); \
		50	stdu r1,-STACK_FRAME_OVERHEAD(r1); \
		51	bl .__trace_hcall_entry; \
		52	addi r1,r1,STACK_FRAME_OVERHEAD; \
		53	ld r0,16(r1); \
		54	ld r3,STK_PARM(r3)(r1); \
		55	ld r4,STK_PARM(r4)(r1); \
		56	ld r5,STK_PARM(r5)(r1); \
		57	ld r6,STK_PARM(r6)(r1); \
		58	ld r7,STK_PARM(r7)(r1); \
		59	ld r8,STK_PARM(r8)(r1); \
		60	ld r9,STK_PARM(r9)(r1); \
		61	ld r10,STK_PARM(r10)(r1); \
		62	mtlr r0; \
		63	1:
		64
31	/*	65	/*
32	* postcall is performed immediately before function return which	66	* postcall is performed immediately before function return which
33	* allows liberal use of volatile registers. We branch around this	67	* allows liberal use of volatile registers. We branch around this
@@ -38,40 +72,21 @@ END_FTR_SECTION_IFSET(CPU_FTR_PURR);
38	BEGIN_FTR_SECTION; \	72	BEGIN_FTR_SECTION; \
39	b 1f; \	73	b 1f; \
40	END_FTR_SECTION(0, 1); \	74	END_FTR_SECTION(0, 1); \
41	ld r4,STK_PARM(r3)(r1); /* validate opcode */ \	75	ld r12,hcall_tracepoint_refcount@toc(r2); \
42	cmpldi cr7,r4,MAX_HCALL_OPCODE; \	76	cmpdi r12,0; \
43	bgt- cr7,1f; \	77	beq+ 1f; \
44	\	78	mflr r0; \
45	/* get time and PURR snapshots after hcall */ \	79	ld r6,STK_PARM(r3)(r1); \
46	mftb r7; /* timebase after */ \	80	std r3,STK_PARM(r3)(r1); \
47	BEGIN_FTR_SECTION; \	81	mr r4,r3; \
48	mfspr r8,SPRN_PURR; /* PURR after */ \	82	mr r3,r6; \
49	ld r6,STK_PARM(r6)(r1); /* PURR before */ \	83	std r0,16(r1); \
50	subf r6,r6,r8; /* delta */ \	84	stdu r1,-STACK_FRAME_OVERHEAD(r1); \
51	END_FTR_SECTION_IFSET(CPU_FTR_PURR); \	85	bl .__trace_hcall_exit; \
52	ld r5,STK_PARM(r5)(r1); /* timebase before */ \	86	addi r1,r1,STACK_FRAME_OVERHEAD; \
53	subf r5,r5,r7; /* time delta */ \	87	ld r0,16(r1); \
54	\	88	ld r3,STK_PARM(r3)(r1); \
55	/* calculate address of stat structure r4 = opcode */ \	89	mtlr r0; \
56	srdi r4,r4,2; /* index into array */ \
57	mulli r4,r4,HCALL_STAT_SIZE; \
58	LOAD_REG_ADDR(r7, per_cpu__hcall_stats); \
59	add r4,r4,r7; \
60	ld r7,PACA_DATA_OFFSET(r13); /* per cpu offset */ \
61	add r4,r4,r7; \
62	\
63	/* update stats */ \
64	ld r7,HCALL_STAT_CALLS(r4); /* count */ \
65	addi r7,r7,1; \
66	std r7,HCALL_STAT_CALLS(r4); \
67	ld r7,HCALL_STAT_TB(r4); /* timebase */ \
68	add r7,r7,r5; \
69	std r7,HCALL_STAT_TB(r4); \
70	BEGIN_FTR_SECTION; \
71	ld r7,HCALL_STAT_PURR(r4); /* PURR */ \
72	add r7,r7,r6; \
73	std r7,HCALL_STAT_PURR(r4); \
74	END_FTR_SECTION_IFSET(CPU_FTR_PURR); \
75	1:	90	1:
76	#else	91	#else
77	#define HCALL_INST_PRECALL	92	#define HCALL_INST_PRECALL