powerpc: ftrace updates to previous patch series

PATCH In review Steven Rostedt 4 participants 13 messages Nov 26, 2008 – Dec 19, 2008

Subsystems (2)

PowerPC ftrace

lore

Share & tools

lore

Steven Rostedt

author maintainer

2008-11-26 22:04 UTC #0 lore raw

cover powerpc: ftrace updates to previous patch series

Paul,

This patch series addresses the issues you brought up as well as adds some more enhancements and fixes. This series is added on top of the previous patch series.

The new patches are: (and are posted now)
5987225... powerpc/ppc32: static ftrace fixes for PPC32 382d6db... powerpc: ftrace, use create_branch
d7d0ab8... powerpc: ftrace, added missing icache flush 10ec622... powerpc: ftrace, fix cast aliasing and add code verification d208ca5... powerpc: ftrace, do nothing in mcount call for dyn ftrace

The patches you already commented on:
5c4f5d7... powerpc/ppc32: ftrace, dynamic ftrace to handle modules a73af3e... powerpc/ppc64: ftrace, handle module trampolines for dyn ftrace 009104f... powerpc: ftrace, use probe_kernel API to modify code a352036... powerpc: ftrace, convert to new dynamic ftrace arch API 6d07bb4... powerpc: ftrace, do not latency trace idle

Note, I also fixed the spelling of your name in the change log of commit a352036... powerpc: ftrace, convert to new dynamic ftrace arch API

I'm only posting the new patches. I've tested this on both my PPC64 and my PPC32 boxes. You can test this as well by checking out my tip/ppc branch which is based off of an older version of tip and has the ppc dynamic ftrace code enabled. That branch is not to pull from, but is there for you to try out this code if you like.

The previous patch series, which have not changed except for the spelling fix of Paul's last name, can be found here:

http://lkml.org/lkml/2008/11/20/356

The following patches are in:

git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-2.6-trace.git

branch: ppc/ftrace

Steven Rostedt (5):
powerpc: ftrace, do nothing in mcount call for dyn ftrace
powerpc: ftrace, fix cast aliasing and add code verification
powerpc: ftrace, added missing icache flush
powerpc: ftrace, use create_branch
powerpc/ppc32: static ftrace fixes for PPC32

----

 arch/powerpc/kernel/Makefile   |    1 +
 arch/powerpc/kernel/entry_32.S |   40 ++-------
 arch/powerpc/kernel/entry_64.S |   12 ---
 arch/powerpc/kernel/ftrace.c   |  182 +++++++++++++++++++---------------------
 arch/powerpc/lib/Makefile      |    3 +
 5 files changed, 98 insertions(+), 140 deletions(-)

Steven Rostedt

author maintainer

2008-11-26 22:04 UTC #1 lore raw

PATCH 1/5 powerpc: ftrace, do nothing in mcount call for dyn ftrace

From: Steven Rostedt <[email protected]>

Impact: quicken mcount calls that are not replaced by dyn ftrace

Dynamic ftrace no longer does on the fly recording of mcount locations. The mcount locations are now found at compile time. The mcount function no longer needs to store registers and call a stub function. It can now just simply return.

Since there are some functions that do not get converted to a nop (.init sections and other code that may disappear), this patch should help speed up that code.

Also, the stub for mcount on PowerPC 32 can not be a simple branch link register like it is on PowerPC 64. According to the ABI specification:

"The _mcount routine is required to restore the link register from
the stack so that the profiling code can be inserted transparently, whether or not the profiled function saves the link register itself."

This means that we must restore the link register that was used to make the call to mcount. The minimal mcount function for PPC32 ends up being:

mcount:

        mflr    r0
        mtctr   r0
        lwz     r0, 4(r1)
        mtlr    r0

bctr

Where we move the link register used to call mcount into the ctr register, and then restore the link register from the stack. Then we use the ctr register to jump back to the mcount caller. The r0 register is free for us to use.

Signed-off-by Steven Rostedt <[email protected]>

---
 arch/powerpc/kernel/entry_32.S |   40 +++++++++-------------------------------
 arch/powerpc/kernel/entry_64.S |   12 ------------
 2 files changed, 9 insertions(+), 43 deletions(-)

arch/powerpc/kernel/entry_32.S +9 ~~-31~~

index 7ecc0d1..6f7eb7e 100644
@@ -1162,39 +1162,17 @@ machine_check_in_rtas:
 #ifdef CONFIG_DYNAMIC_FTRACE
 _GLOBAL(mcount)
 _GLOBAL(_mcount)
-	stwu	r1,-48(r1)
-	stw	r3, 12(r1)
-	stw	r4, 16(r1)
-	stw	r5, 20(r1)
-	stw	r6, 24(r1)
-	mflr	r3
-	stw	r7, 28(r1)
-	mfcr	r5
-	stw	r8, 32(r1)
-	stw	r9, 36(r1)
-	stw	r10,40(r1)
-	stw	r3, 44(r1)
-	stw	r5, 8(r1)
-	subi	r3, r3, MCOUNT_INSN_SIZE
-	.globl mcount_call
-mcount_call:
-	bl	ftrace_stub
-	nop
-	lwz	r6, 8(r1)
-	lwz	r0, 44(r1)
-	lwz	r3, 12(r1)
+	/*
+	 * It is required that _mcount on PPC32 must preserve the
+	 * link register. But we have r0 to play with. We use r0
+	 * to push the return address back to the caller of mcount
+	 * into the ctr register, restore the link register and
+	 * then jump back using the ctr register.
+	 */
+	mflr	r0
 	mtctr	r0
-	lwz	r4, 16(r1)
-	mtcr	r6
-	lwz	r5, 20(r1)
-	lwz	r6, 24(r1)
-	lwz	r0, 52(r1)
-	lwz	r7, 28(r1)
-	lwz	r8, 32(r1)
+	lwz	r0, 4(r1)
 	mtlr	r0
-	lwz	r9, 36(r1)
-	lwz	r10,40(r1)
-	addi	r1, r1, 48
 	bctr
 
 _GLOBAL(ftrace_caller)

arch/powerpc/kernel/entry_64.S +0 ~~-12~~

index e6d5284..b00982e 100644
@@ -888,18 +888,6 @@ _GLOBAL(enter_prom)
 #ifdef CONFIG_DYNAMIC_FTRACE
 _GLOBAL(mcount)
 _GLOBAL(_mcount)
-	/* Taken from output of objdump from lib64/glibc */
-	mflr	r3
-	stdu	r1, -112(r1)
-	std	r3, 128(r1)
-	subi	r3, r3, MCOUNT_INSN_SIZE
-	.globl mcount_call
-mcount_call:
-	bl	ftrace_stub
-	nop
-	ld	r0, 128(r1)
-	mtlr	r0
-	addi	r1, r1, 112
 	blr
 
 _GLOBAL(ftrace_caller)

Signature · 3 lines

1.5.6.5

--

Steven Rostedt

author maintainer

2008-11-26 22:04 UTC #2 lore raw

PATCH 2/5 powerpc: ftrace, fix cast aliasing and add code verification

In reply to Steven Rostedt

From: Steven Rostedt <[email protected]>

Impact: clean up and robustness addition

This patch addresses the comments made by Paul Mackerras. It removes the type casting between unsigned int and unsigned char pointers, and replaces them with a use of all unsigned int.

Verification that the jump is indeed made to a trampoline has also been added.

Signed-off-by Steven Rostedt <[email protected]>

---
 arch/powerpc/kernel/ftrace.c |  121 ++++++++++++++++++++++-------------------
 1 files changed, 65 insertions(+), 56 deletions(-)

arch/powerpc/kernel/ftrace.c +65 ~~-56~~

index 3271cd6..ea454a0 100644
@@ -162,26 +162,25 @@ static int
 __ftrace_make_nop(struct module *mod,
 		  struct dyn_ftrace *rec, unsigned long addr)
 {
-	unsigned char replaced[MCOUNT_INSN_SIZE * 2];
-	unsigned int *op = (unsigned *)&replaced;
-	unsigned char jmp[8];
-	unsigned long *ptr = (unsigned long *)&jmp;
+	unsigned int op;
+	unsigned int jmp[5];
+	unsigned long ptr;
 	unsigned long ip = rec->ip;
 	unsigned long tramp;
 	int offset;
 
 	/* read where this goes */
-	if (probe_kernel_read(replaced, (void *)ip, MCOUNT_INSN_SIZE))
+	if (probe_kernel_read(&op, (void *)ip, sizeof(int)))
 		return -EFAULT;
 
 	/* Make sure that that this is still a 24bit jump */
-	if (!is_bl_op(*op)) {
-		printk(KERN_ERR "Not expected bl: opcode is %x\n", *op);
+	if (!is_bl_op(op)) {
+		printk(KERN_ERR "Not expected bl: opcode is %x\n", op);
 		return -EINVAL;
 	}
 
 	/* lets find where the pointer goes */
-	tramp = find_bl_target(ip, *op);
+	tramp = find_bl_target(ip, op);
 
 	/*
 	 * On PPC64 the trampoline looks like:

@@ -200,19 +199,25 @@ __ftrace_make_nop(struct module *mod,
 	DEBUGP("ip:%lx jumps to %lx r2: %lx", ip, tramp, mod->arch.toc);
 
 	/* Find where the trampoline jumps to */
-	if (probe_kernel_read(jmp, (void *)tramp, 8)) {
+	if (probe_kernel_read(jmp, (void *)tramp, sizeof(jmp))) {
 		printk(KERN_ERR "Failed to read %lx\n", tramp);
 		return -EFAULT;
 	}
 
-	DEBUGP(" %08x %08x",
-	       (unsigned)(*ptr >> 32),
-	       (unsigned)*ptr);
+	DEBUGP(" %08x %08x", jmp[0], jmp[1]);
+
+	/* verify that this is what we expect it to be */
+	if (((jmp[0] & 0xffff0000) != 0x3d820000) ||
+	    ((jmp[1] & 0xffff0000) != 0x398c0000) ||
+	    (jmp[2] != 0xf8410028) ||
+	    (jmp[3] != 0xe96c0020) ||
+	    (jmp[4] != 0xe84c0028)) {
+		printk(KERN_ERR "Not a trampoline\n");
+		return -EINVAL;
+	}
 
-	offset = (unsigned)jmp[2] << 24 |
-		(unsigned)jmp[3] << 16 |
-		(unsigned)jmp[6] << 8 |
-		(unsigned)jmp[7];
+	offset = (unsigned)((unsigned short)jmp[0]) << 16 |
+		(unsigned)((unsigned short)jmp[1]);
 
 	DEBUGP(" %x ", offset);

@@ -225,13 +230,13 @@ __ftrace_make_nop(struct module *mod,
 		return -EFAULT;
 	}
 
-	DEBUGP(" %08x %08x\n",
-	       (unsigned)(*ptr >> 32),
-	       (unsigned)*ptr);
+	DEBUGP(" %08x %08x\n", jmp[0], jmp[1]);
+
+	ptr = ((unsigned long)jmp[0] << 32) + jmp[1];
 
 	/* This should match what was called */
-	if (*ptr != GET_ADDR(addr)) {
-		printk(KERN_ERR "addr does not match %lx\n", *ptr);
+	if (ptr != GET_ADDR(addr)) {
+		printk(KERN_ERR "addr does not match %lx\n", ptr);
 		return -EINVAL;
 	}

@@ -240,11 +245,11 @@ __ftrace_make_nop(struct module *mod,
 	 *  0xe8, 0x41, 0x00, 0x28   ld r2,40(r1)
 	 * This needs to be turned to a nop too.
 	 */
-	if (probe_kernel_read(replaced, (void *)(ip+4), MCOUNT_INSN_SIZE))
+	if (probe_kernel_read(&op, (void *)(ip+4), MCOUNT_INSN_SIZE))
 		return -EFAULT;
 
-	if (*op != 0xe8410028) {
-		printk(KERN_ERR "Next line is not ld! (%08x)\n", *op);
+	if (op != 0xe8410028) {
+		printk(KERN_ERR "Next line is not ld! (%08x)\n", op);
 		return -EINVAL;
 	}

@@ -261,9 +266,9 @@ __ftrace_make_nop(struct module *mod,
 	 *   ld r2,40(r1)
 	 *  1:
 	 */
-	op[0] = 0x48000008;	/* b +8 */
+	op = 0x48000008;	/* b +8 */
 
-	if (probe_kernel_write((void *)ip, replaced, MCOUNT_INSN_SIZE))
+	if (probe_kernel_write((void *)ip, &op, MCOUNT_INSN_SIZE))
 		return -EPERM;
 
 	return 0;

@@ -274,46 +279,52 @@ static int
 __ftrace_make_nop(struct module *mod,
 		  struct dyn_ftrace *rec, unsigned long addr)
 {
-	unsigned char replaced[MCOUNT_INSN_SIZE];
-	unsigned int *op = (unsigned *)&replaced;
-	unsigned char jmp[8];
-	unsigned int *ptr = (unsigned int *)&jmp;
+	unsigned int op;
+	unsigned int jmp[4];
 	unsigned long ip = rec->ip;
 	unsigned long tramp;
-	int offset;
 
-	if (probe_kernel_read(replaced, (void *)ip, MCOUNT_INSN_SIZE))
+	if (probe_kernel_read(&op, (void *)ip, MCOUNT_INSN_SIZE))
 		return -EFAULT;
 
 	/* Make sure that that this is still a 24bit jump */
-	if (!is_bl_op(*op)) {
-		printk(KERN_ERR "Not expected bl: opcode is %x\n", *op);
+	if (!is_bl_op(op)) {
+		printk(KERN_ERR "Not expected bl: opcode is %x\n", op);
 		return -EINVAL;
 	}
 
 	/* lets find where the pointer goes */
-	tramp = find_bl_target(ip, *op);
+	tramp = find_bl_target(ip, op);
 
 	/*
 	 * On PPC32 the trampoline looks like:
-	 * lis r11,sym@ha
-	 * addi r11,r11,sym@l
-	 * mtctr r11
-	 * bctr
+	 *  0x3d, 0x60, 0x00, 0x00  lis r11,sym@ha
+	 *  0x39, 0x6b, 0x00, 0x00  addi r11,r11,sym@l
+	 *  0x7d, 0x69, 0x03, 0xa6  mtctr r11
+	 *  0x4e, 0x80, 0x04, 0x20  bctr
 	 */
 
 	DEBUGP("ip:%lx jumps to %lx", ip, tramp);
 
 	/* Find where the trampoline jumps to */
-	if (probe_kernel_read(jmp, (void *)tramp, 8)) {
+	if (probe_kernel_read(jmp, (void *)tramp, sizeof(jmp))) {
 		printk(KERN_ERR "Failed to read %lx\n", tramp);
 		return -EFAULT;
 	}
 
-	DEBUGP(" %08x %08x ", ptr[0], ptr[1]);
+	DEBUGP(" %08x %08x ", jmp[0], jmp[1]);
+
+	/* verify that this is what we expect it to be */
+	if (((jmp[0] & 0xffff0000) != 0x3d600000) ||
+	    ((jmp[1] & 0xffff0000) != 0x396b0000) ||
+	    (jmp[2] != 0x7d6903a6) ||
+	    (jmp[3] != 0x4e800420)) {
+		printk(KERN_ERR "Not a trampoline\n");
+		return -EINVAL;
+	}
 
-	tramp = (ptr[1] & 0xffff) |
-		((ptr[0] & 0xffff) << 16);
+	tramp = (jmp[1] & 0xffff) |
+		((jmp[0] & 0xffff) << 16);
 	if (tramp & 0x8000)
 		tramp -= 0x10000;

@@ -326,9 +337,9 @@ __ftrace_make_nop(struct module *mod,
 		return -EINVAL;
 	}
 
-	op[0] = PPC_NOP_INSTR;
+	op = PPC_NOP_INSTR;
 
-	if (probe_kernel_write((void *)ip, replaced, MCOUNT_INSN_SIZE))
+	if (probe_kernel_write((void *)ip, &op, MCOUNT_INSN_SIZE))
 		return -EPERM;
 
 	return 0;

@@ -384,13 +395,12 @@ int ftrace_make_nop(struct module *mod,
 static int
 __ftrace_make_call(struct dyn_ftrace *rec, unsigned long addr)
 {
-	unsigned char replaced[MCOUNT_INSN_SIZE * 2];
-	unsigned int *op = (unsigned *)&replaced;
+	unsigned int op[2];
 	unsigned long ip = rec->ip;
 	unsigned long offset;
 
 	/* read where this goes */
-	if (probe_kernel_read(replaced, (void *)ip, MCOUNT_INSN_SIZE * 2))
+	if (probe_kernel_read(op, (void *)ip, MCOUNT_INSN_SIZE * 2))
 		return -EFAULT;
 
 	/*

@@ -425,7 +435,7 @@ __ftrace_make_call(struct dyn_ftrace *rec, unsigned long addr)
 
 	DEBUGP("write to %lx\n", rec->ip);
 
-	if (probe_kernel_write((void *)ip, replaced, MCOUNT_INSN_SIZE * 2))
+	if (probe_kernel_write((void *)ip, op, MCOUNT_INSN_SIZE * 2))
 		return -EPERM;
 
 	return 0;

@@ -434,18 +444,17 @@ __ftrace_make_call(struct dyn_ftrace *rec, unsigned long addr)
 static int
 __ftrace_make_call(struct dyn_ftrace *rec, unsigned long addr)
 {
-	unsigned char replaced[MCOUNT_INSN_SIZE];
-	unsigned int *op = (unsigned *)&replaced;
+	unsigned int op;
 	unsigned long ip = rec->ip;
 	unsigned long offset;
 
 	/* read where this goes */
-	if (probe_kernel_read(replaced, (void *)ip, MCOUNT_INSN_SIZE))
+	if (probe_kernel_read(&op, (void *)ip, MCOUNT_INSN_SIZE))
 		return -EFAULT;
 
 	/* It should be pointing to a nop */
-	if (op[0] != PPC_NOP_INSTR) {
-		printk(KERN_ERR "Expected NOP but have %x\n", op[0]);
+	if (op != PPC_NOP_INSTR) {
+		printk(KERN_ERR "Expected NOP but have %x\n", op);
 		return -EINVAL;
 	}

@@ -465,11 +474,11 @@ __ftrace_make_call(struct dyn_ftrace *rec, unsigned long addr)
 	}
 
 	/* Set to "bl addr" */
-	op[0] = branch_offset(offset);
+	op = branch_offset(offset);
 
 	DEBUGP("write to %lx\n", rec->ip);
 
-	if (probe_kernel_write((void *)ip, replaced, MCOUNT_INSN_SIZE))
+	if (probe_kernel_write((void *)ip, &op, MCOUNT_INSN_SIZE))
 		return -EPERM;
 
 	return 0;

Signature · 3 lines

1.5.6.5

--

Steven Rostedt

author maintainer

2008-11-26 22:04 UTC #3 lore raw

PATCH 3/5 powerpc: ftrace, added missing icache flush

In reply to Steven Rostedt

From: Steven Rostedt <[email protected]>

Impact: fix to PowerPC code modification

After modifying code it is essential to flush the icache. This patch adds the missing flush.

Reported-by Paul Mackerras <[email protected]>

Signed-off-by Steven Rostedt <[email protected]>

---
 arch/powerpc/kernel/ftrace.c |    9 +++++++++
 1 files changed, 9 insertions(+), 0 deletions(-)

arch/powerpc/kernel/ftrace.c +9 -0

index ea454a0..a4640e4 100644
@@ -271,6 +271,9 @@ __ftrace_make_nop(struct module *mod,
 	if (probe_kernel_write((void *)ip, &op, MCOUNT_INSN_SIZE))
 		return -EPERM;
 
+
+	flush_icache_range(ip, ip + 8);
+
 	return 0;
 }

@@ -342,6 +345,8 @@ __ftrace_make_nop(struct module *mod,
 	if (probe_kernel_write((void *)ip, &op, MCOUNT_INSN_SIZE))
 		return -EPERM;
 
+	flush_icache_range(ip, ip + 8);
+
 	return 0;
 }
 #endif /* PPC64 */

@@ -438,6 +443,8 @@ __ftrace_make_call(struct dyn_ftrace *rec, unsigned long addr)
 	if (probe_kernel_write((void *)ip, op, MCOUNT_INSN_SIZE * 2))
 		return -EPERM;
 
+	flush_icache_range(ip, ip + 8);
+
 	return 0;
 }
 #else

@@ -481,6 +488,8 @@ __ftrace_make_call(struct dyn_ftrace *rec, unsigned long addr)
 	if (probe_kernel_write((void *)ip, &op, MCOUNT_INSN_SIZE))
 		return -EPERM;
 
+	flush_icache_range(ip, ip + 8);
+
 	return 0;
 }
 #endif /* CONFIG_PPC64 */

Signature · 3 lines

1.5.6.5

--

Steven Rostedt

author maintainer

2008-11-26 22:04 UTC #4 lore raw

PATCH 4/5 powerpc: ftrace, use create_branch

In reply to Steven Rostedt

From: Steven Rostedt <[email protected]>

Impact: clean up

Paul Mackerras pointed out that the code to determine if the branch can reach the destination is incorrect. Michael Ellerman suggested to pull out the code from create_branch and use that.

Simply using create_branch is probably the best.

Reported-by Michael Ellerman <[email protected]>

Reported-by Paul Mackerras <[email protected]>

Signed-off-by Steven Rostedt <[email protected]>

---
 arch/powerpc/kernel/ftrace.c |   54 +++++++++--------------------------------
 1 files changed, 12 insertions(+), 42 deletions(-)

arch/powerpc/kernel/ftrace.c +12 ~~-42~~

index a4640e4..5355244 100644
@@ -114,19 +114,9 @@ ftrace_modify_code(unsigned long ip, unsigned char *old_code,
  */
 static int test_24bit_addr(unsigned long ip, unsigned long addr)
 {
-	long diff;
 
-	/*
-	 * Can we get to addr from ip in 24 bits?
-	 *  (26 really, since we mulitply by 4 for 4 byte alignment)
-	 */
-	diff = addr - ip;
-
-	/*
-	 * Return true if diff is less than 1 << 25
-	 *  and greater than -1 << 26.
-	 */
-	return (diff < (1 << 25)) && (diff > (-1 << 26));
+	/* use the create_branch to verify that this offset can be branched */
+	return create_branch((unsigned int *)ip, addr, 0);
 }
 
 static int is_bl_op(unsigned int op)

@@ -134,11 +124,6 @@ static int is_bl_op(unsigned int op)
 	return (op & 0xfc000003) == 0x48000001;
 }
 
-static int test_offset(unsigned long offset)
-{
-	return (offset + 0x2000000 > 0x3ffffff) || ((offset & 3) != 0);
-}
-
 static unsigned long find_bl_target(unsigned long ip, unsigned int op)
 {
 	static int offset;

@@ -151,12 +136,6 @@ static unsigned long find_bl_target(unsigned long ip, unsigned int op)
 	return ip + (long)offset;
 }
 
-static unsigned int branch_offset(unsigned long offset)
-{
-	/* return "bl ip+offset" */
-	return 0x48000001 | (offset & 0x03fffffc);
-}
-
 #ifdef CONFIG_PPC64
 static int
 __ftrace_make_nop(struct module *mod,

@@ -402,7 +381,6 @@ __ftrace_make_call(struct dyn_ftrace *rec, unsigned long addr)
 {
 	unsigned int op[2];
 	unsigned long ip = rec->ip;
-	unsigned long offset;
 
 	/* read where this goes */
 	if (probe_kernel_read(op, (void *)ip, MCOUNT_INSN_SIZE * 2))

@@ -424,17 +402,14 @@ __ftrace_make_call(struct dyn_ftrace *rec, unsigned long addr)
 		return -EINVAL;
 	}
 
-	/* now calculate a jump to the ftrace caller trampoline */
-	offset = rec->arch.mod->arch.tramp - ip;
-
-	if (test_offset(offset)) {
-		printk(KERN_ERR "REL24 %li out of range!\n",
-		       (long int)offset);
+	/* create the branch to the trampoline */
+	op[0] = create_branch((unsigned int *)ip,
+			      rec->arch.mod->arch.tramp, BRANCH_SET_LINK);
+	if (!op[0]) {
+		printk(KERN_ERR "REL24 out of range!\n");
 		return -EINVAL;
 	}
 
-	/* Set to "bl addr" */
-	op[0] = branch_offset(offset);
 	/* ld r2,40(r1) */
 	op[1] = 0xe8410028;

@@ -453,7 +428,6 @@ __ftrace_make_call(struct dyn_ftrace *rec, unsigned long addr)
 {
 	unsigned int op;
 	unsigned long ip = rec->ip;
-	unsigned long offset;
 
 	/* read where this goes */
 	if (probe_kernel_read(&op, (void *)ip, MCOUNT_INSN_SIZE))

@@ -471,18 +445,14 @@ __ftrace_make_call(struct dyn_ftrace *rec, unsigned long addr)
 		return -EINVAL;
 	}
 
-	/* now calculate a jump to the ftrace caller trampoline */
-	offset = rec->arch.mod->arch.tramp - ip;
-
-	if (test_offset(offset)) {
-		printk(KERN_ERR "REL24 %li out of range!\n",
-		       (long int)offset);
+	/* create the branch to the trampoline */
+	op = create_branch((unsigned int *)ip,
+			   rec->arch.mod->arch.tramp, BRANCH_SET_LINK);
+	if (!op) {
+		printk(KERN_ERR "REL24 out of range!\n");
 		return -EINVAL;
 	}
 
-	/* Set to "bl addr" */
-	op = branch_offset(offset);
-
 	DEBUGP("write to %lx\n", rec->ip);
 
 	if (probe_kernel_write((void *)ip, &op, MCOUNT_INSN_SIZE))

Signature · 3 lines

1.5.6.5

--

Michael Ellerman

Acked-by on v1 4/5

2008-11-27 22:06 UTC #5 lore raw

Re: [PATCH 4/5] powerpc: ftrace, use create_branch

Quoted 11 lines · Steven Rostedt

plain text document attachment
(0004-powerpc-ftrace-use-create_branch.patch)
From: Steven Rostedt <[email protected]>
Impact: clean up
Paul Mackerras pointed out that the code to determine if the branch can reach the destination is incorrect. Michael Ellerman suggested to pull out the code from create_branch and use that.
Simply using create_branch is probably the best.

You're right, that's a lot simpler :)

Acked-by Michael Ellerman <[email protected]>

cheers

Signature · 8 lines

Michael Ellerman
OzLabs, IBM Australia Development Lab

wwweb: http://michael.ellerman.id.au
phone: +61 2 6212 1183 (tie line 70 21183)

We do not inherit the earth from our ancestors,
we borrow it from our children. - S.M.A.R.T Person

Steven Rostedt

author maintainer

2008-11-26 22:04 UTC #6 lore raw

PATCH 5/5 powerpc/ppc32: static ftrace fixes for PPC32

In reply to Steven Rostedt

From: Steven Rostedt <rostedt@gollum.(none)>

Impact: fix for PowerPC 32 code

There were some early init code that was not safe for static ftrace to boot on my PowerBook. This code must only use relative addressing, and static mcount performs a compare of the ftrace_trace_function pointer, and gets that with an absolute address. In the early init boot up code, this will cause a fault.

This patch removes tracing from the files containing the offending functions.

Signed-off-by Steven Rostedt <rostedt@gollum.(none)>

---
 arch/powerpc/kernel/Makefile |    1 +
 arch/powerpc/lib/Makefile    |    3 +++
 2 files changed, 4 insertions(+), 0 deletions(-)

arch/powerpc/kernel/Makefile +1 -0

index 92673b4..d17edb4 100644
@@ -17,6 +17,7 @@ ifdef CONFIG_FUNCTION_TRACER
 CFLAGS_REMOVE_cputable.o = -pg -mno-sched-epilog
 CFLAGS_REMOVE_prom_init.o = -pg -mno-sched-epilog
 CFLAGS_REMOVE_btext.o = -pg -mno-sched-epilog
+CFLAGS_REMOVE_prom.o = -pg -mno-sched-epilog
 
 ifdef CONFIG_DYNAMIC_FTRACE
 # dynamic ftrace setup.

arch/powerpc/lib/Makefile +3 -0

index d69912c..8db3527 100644
@@ -6,6 +6,9 @@ ifeq ($(CONFIG_PPC64),y)
 EXTRA_CFLAGS		+= -mno-minimal-toc
 endif
 
+CFLAGS_REMOVE_code-patching.o = -pg
+CFLAGS_REMOVE_feature-fixups.o = -pg
+
 obj-y			:= string.o alloc.o \
 			   checksum_$(CONFIG_WORD_SIZE).o
 obj-$(CONFIG_PPC32)	+= div64.o copy_32.o crtsavres.o

Signature · 3 lines

1.5.6.5

--

Steven Rostedt

author maintainer

2008-11-26 22:41 UTC #7 lore raw

Re: [PATCH 5/5] powerpc/ppc32: static ftrace fixes for PPC32

Quoted 1 line · Steven Rostedt

From: Steven Rostedt <rostedt@gollum.(none)>

Ah, I need to add GIT_AUTHOR on my PowerBook. And yes, it's name is gollum ;-)

I'll fix this and push it again.

-- Steve

Quoted 50 lines

Impact: fix for PowerPC 32 code
There were some early init code that was not safe for static ftrace to boot on my PowerBook. This code must only use relative addressing, and static mcount performs a compare of the ftrace_trace_function pointer, and gets that with an absolute address. In the early init boot up code, this will cause a fault.
This patch removes tracing from the files containing the offending functions.
Signed-off-by Steven Rostedt <rostedt@gollum.(none)>
---
 arch/powerpc/kernel/Makefile |    1 +
 arch/powerpc/lib/Makefile    |    3 +++
 2 files changed, 4 insertions(+), 0 deletions(-)
arch/powerpc/kernel/Makefile +1 -0
index 92673b4..d17edb4 100644
@@ -17,6 +17,7 @@ ifdef CONFIG_FUNCTION_TRACER
 CFLAGS_REMOVE_cputable.o = -pg -mno-sched-epilog
 CFLAGS_REMOVE_prom_init.o = -pg -mno-sched-epilog
 CFLAGS_REMOVE_btext.o = -pg -mno-sched-epilog
+CFLAGS_REMOVE_prom.o = -pg -mno-sched-epilog
 
 ifdef CONFIG_DYNAMIC_FTRACE
 # dynamic ftrace setup.
arch/powerpc/lib/Makefile +3 -0
index d69912c..8db3527 100644
@@ -6,6 +6,9 @@ ifeq ($(CONFIG_PPC64),y)
 EXTRA_CFLAGS		+= -mno-minimal-toc
 endif
 
+CFLAGS_REMOVE_code-patching.o = -pg
+CFLAGS_REMOVE_feature-fixups.o = -pg
+
 obj-y			:= string.o alloc.o \
 			   checksum_$(CONFIG_WORD_SIZE).o
 obj-$(CONFIG_PPC32)	+= div64.o copy_32.o crtsavres.o
Signature · 3 lines
1.5.6.5

-- 

Steven Rostedt

author maintainer

2008-11-26 22:50 UTC #8 lore raw

In reply to Steven Rostedt

Quoted 8 lines · Steven Rostedt

Paul,
This patch series addresses the issues you brought up as well as adds some more enhancements and fixes. This series is added on top of the previous patch series.
The new patches are: (and are posted now) 5987225... powerpc/ppc32: static ftrace fixes for PPC32

The new commit id with the fix in Author and SoB is:
fd90db4bdc1ccf71927365cd634188cec4d7b151
powerpc/ppc32: static ftrace fixes for

-- Steve

Quoted 12 lines

382d6db... powerpc: ftrace, use create_branch
d7d0ab8... powerpc: ftrace, added missing icache flush 10ec622... powerpc: ftrace, fix cast aliasing and add code verification d208ca5... powerpc: ftrace, do nothing in mcount call for dyn ftrace
The patches you already commented on:
5c4f5d7... powerpc/ppc32: ftrace, dynamic ftrace to handle modules a73af3e... powerpc/ppc64: ftrace, handle module trampolines for dyn ftrace 009104f... powerpc: ftrace, use probe_kernel API to modify code a352036... powerpc: ftrace, convert to new dynamic ftrace arch API 6d07bb4... powerpc: ftrace, do not latency trace idle

Ingo Molnar

2008-11-28 13:10 UTC #9 lore raw

Quoted 8 lines · Steven Rostedt

Quoted 6 lines · Steven Rostedt
Paul,
This patch series addresses the issues you brought up as well as adds some more enhancements and fixes. This series is added on top of the previous patch series.

Quoted 6 lines

Quoted 2 lines
The new patches are: (and are posted now) 5987225... powerpc/ppc32: static ftrace fixes for PPC32
The new commit id with the fix in Author and SoB is:
fd90db4bdc1ccf71927365cd634188cec4d7b151
powerpc/ppc32: static ftrace fixes for

Steve, you rebased the previous commits, and that's a no-no. I sorted it all out, find below the stable URI for stuff that we'd like Paul to pull shortly before the merge window.

Thanks,

Ingo

-------------------------->
Please pull the latest tracing/powerpc git tree from:

git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip.git tracing/powerpc

Thanks,

Ingo

------------------>
Steven Rostedt (10):
powerpc: ftrace, do not latency trace idle
powerpc: ftrace, convert to new dynamic ftrace arch API
powerpc: ftrace, use probe_kernel API to modify code powerpc/ppc64: ftrace, handle module trampolines for dyn ftrace powerpc/ppc32: ftrace, dynamic ftrace to handle modules
powerpc: ftrace, do nothing in mcount call for dyn ftrace
powerpc: ftrace, fix cast aliasing and add code verification
powerpc: ftrace, added missing icache flush
powerpc: ftrace, use create_branch
powerpc/ppc32: static ftrace fixes for PPC32

 arch/powerpc/include/asm/ftrace.h |   14 +-
 arch/powerpc/include/asm/module.h |   16 ++-
 arch/powerpc/kernel/Makefile      |    1 +
 arch/powerpc/kernel/entry_32.S    |   40 +---
 arch/powerpc/kernel/entry_64.S    |   12 -
 arch/powerpc/kernel/ftrace.c      |  461 +++++++++++++++++++++++++++++++++----
 arch/powerpc/kernel/idle.c        |    5 +
 arch/powerpc/kernel/module_32.c   |   10 +
 arch/powerpc/kernel/module_64.c   |   13 +
 arch/powerpc/lib/Makefile         |    3 +
 10 files changed, 490 insertions(+), 85 deletions(-)

arch/powerpc/include/asm/ftrace.h +13 -1

index b298f7a..e5f2ae8 100644
@@ -7,7 +7,19 @@
 
 #ifndef __ASSEMBLY__
 extern void _mcount(void);
-#endif
+
+#ifdef CONFIG_DYNAMIC_FTRACE
+static inline unsigned long ftrace_call_adjust(unsigned long addr)
+{
+       /* reloction of mcount call site is the same as the address */
+       return addr;
+}
+
+struct dyn_arch_ftrace {
+	struct module *mod;
+};
+#endif /*  CONFIG_DYNAMIC_FTRACE */
+#endif /* __ASSEMBLY__ */
 
 #endif

arch/powerpc/include/asm/module.h +15 -1

index e5f14b1..0845488 100644
@@ -34,11 +34,19 @@ struct mod_arch_specific {
 #ifdef __powerpc64__
 	unsigned int stubs_section;	/* Index of stubs section in module */
 	unsigned int toc_section;	/* What section is the TOC? */
-#else
+#ifdef CONFIG_DYNAMIC_FTRACE
+	unsigned long toc;
+	unsigned long tramp;
+#endif
+
+#else /* powerpc64 */
 	/* Indices of PLT sections within module. */
 	unsigned int core_plt_section;
 	unsigned int init_plt_section;
+#ifdef CONFIG_DYNAMIC_FTRACE
+	unsigned long tramp;
 #endif
+#endif /* powerpc64 */
 
 	/* List of BUG addresses, source line numbers and filenames */
 	struct list_head bug_list;

@@ -68,6 +76,12 @@ struct mod_arch_specific {
 #    endif	/* MODULE */
 #endif
 
+#ifdef CONFIG_DYNAMIC_FTRACE
+#    ifdef MODULE
+	asm(".section .ftrace.tramp,\"ax\",@nobits; .align 3; .previous");
+#    endif	/* MODULE */
+#endif
+
 
 struct exception_table_entry;
 void sort_ex_table(struct exception_table_entry *start,

arch/powerpc/kernel/Makefile +1 -0

index 92673b4..d17edb4 100644
@@ -17,6 +17,7 @@ ifdef CONFIG_FUNCTION_TRACER
 CFLAGS_REMOVE_cputable.o = -pg -mno-sched-epilog
 CFLAGS_REMOVE_prom_init.o = -pg -mno-sched-epilog
 CFLAGS_REMOVE_btext.o = -pg -mno-sched-epilog
+CFLAGS_REMOVE_prom.o = -pg -mno-sched-epilog
 
 ifdef CONFIG_DYNAMIC_FTRACE
 # dynamic ftrace setup.

arch/powerpc/kernel/entry_32.S +9 ~~-31~~

index 7ecc0d1..6f7eb7e 100644
@@ -1162,39 +1162,17 @@ machine_check_in_rtas:
 #ifdef CONFIG_DYNAMIC_FTRACE
 _GLOBAL(mcount)
 _GLOBAL(_mcount)
-	stwu	r1,-48(r1)
-	stw	r3, 12(r1)
-	stw	r4, 16(r1)
-	stw	r5, 20(r1)
-	stw	r6, 24(r1)
-	mflr	r3
-	stw	r7, 28(r1)
-	mfcr	r5
-	stw	r8, 32(r1)
-	stw	r9, 36(r1)
-	stw	r10,40(r1)
-	stw	r3, 44(r1)
-	stw	r5, 8(r1)
-	subi	r3, r3, MCOUNT_INSN_SIZE
-	.globl mcount_call
-mcount_call:
-	bl	ftrace_stub
-	nop
-	lwz	r6, 8(r1)
-	lwz	r0, 44(r1)
-	lwz	r3, 12(r1)
+	/*
+	 * It is required that _mcount on PPC32 must preserve the
+	 * link register. But we have r0 to play with. We use r0
+	 * to push the return address back to the caller of mcount
+	 * into the ctr register, restore the link register and
+	 * then jump back using the ctr register.
+	 */
+	mflr	r0
 	mtctr	r0
-	lwz	r4, 16(r1)
-	mtcr	r6
-	lwz	r5, 20(r1)
-	lwz	r6, 24(r1)
-	lwz	r0, 52(r1)
-	lwz	r7, 28(r1)
-	lwz	r8, 32(r1)
+	lwz	r0, 4(r1)
 	mtlr	r0
-	lwz	r9, 36(r1)
-	lwz	r10,40(r1)
-	addi	r1, r1, 48
 	bctr
 
 _GLOBAL(ftrace_caller)

arch/powerpc/kernel/entry_64.S +0 ~~-12~~

index e6d5284..b00982e 100644
@@ -888,18 +888,6 @@ _GLOBAL(enter_prom)
 #ifdef CONFIG_DYNAMIC_FTRACE
 _GLOBAL(mcount)
 _GLOBAL(_mcount)
-	/* Taken from output of objdump from lib64/glibc */
-	mflr	r3
-	stdu	r1, -112(r1)
-	std	r3, 128(r1)
-	subi	r3, r3, MCOUNT_INSN_SIZE
-	.globl mcount_call
-mcount_call:
-	bl	ftrace_stub
-	nop
-	ld	r0, 128(r1)
-	mtlr	r0
-	addi	r1, r1, 112
 	blr
 
 _GLOBAL(ftrace_caller)

arch/powerpc/kernel/ftrace.c +421 ~~-40~~

514 lines not on this page · raw message

arch/powerpc/kernel/idle.c +5 -0

index 31982d0..88d9c1d 100644
@@ -69,10 +69,15 @@ void cpu_idle(void)
 				smp_mb();
 				local_irq_disable();
 
+				/* Don't trace irqs off for idle */
+				stop_critical_timings();
+
 				/* check again after disabling irqs */
 				if (!need_resched() && !cpu_should_die())
 					ppc_md.power_save();
 
+				start_critical_timings();
+
 				local_irq_enable();
 				set_thread_flag(TIF_POLLING_NRFLAG);

arch/powerpc/kernel/module_32.c +10 -0

index 2df91a0..f832773 100644
@@ -22,6 +22,7 @@
 #include <linux/fs.h>
 #include <linux/string.h>
 #include <linux/kernel.h>
+#include <linux/ftrace.h>
 #include <linux/cache.h>
 #include <linux/bug.h>
 #include <linux/sort.h>

@@ -53,6 +54,9 @@ static unsigned int count_relocs(const Elf32_Rela *rela, unsigned int num)
 			r_addend = rela[i].r_addend;
 		}
 
+#ifdef CONFIG_DYNAMIC_FTRACE
+	_count_relocs++;	/* add one for ftrace_caller */
+#endif
 	return _count_relocs;
 }

@@ -306,5 +310,11 @@ int apply_relocate_add(Elf32_Shdr *sechdrs,
 			return -ENOEXEC;
 		}
 	}
+#ifdef CONFIG_DYNAMIC_FTRACE
+	module->arch.tramp =
+		do_plt_call(module->module_core,
+			    (unsigned long)ftrace_caller,
+			    sechdrs, module);
+#endif
 	return 0;
 }

arch/powerpc/kernel/module_64.c +13 -0

index 1af2377..8992b03 100644
@@ -20,6 +20,7 @@
 #include <linux/moduleloader.h>
 #include <linux/err.h>
 #include <linux/vmalloc.h>
+#include <linux/ftrace.h>
 #include <linux/bug.h>
 #include <asm/module.h>
 #include <asm/firmware.h>

@@ -163,6 +164,11 @@ static unsigned long get_stubs_size(const Elf64_Ehdr *hdr,
 		}
 	}
 
+#ifdef CONFIG_DYNAMIC_FTRACE
+	/* make the trampoline to the ftrace_caller */
+	relocs++;
+#endif
+
 	DEBUGP("Looks like a total of %lu stubs, max\n", relocs);
 	return relocs * sizeof(struct ppc64_stub_entry);
 }

@@ -441,5 +447,12 @@ int apply_relocate_add(Elf64_Shdr *sechdrs,
 		}
 	}
 
+#ifdef CONFIG_DYNAMIC_FTRACE
+	me->arch.toc = my_r2(sechdrs, me);
+	me->arch.tramp = stub_for_addr(sechdrs,
+				       (unsigned long)ftrace_caller,
+				       me);
+#endif
+
 	return 0;
 }

arch/powerpc/lib/Makefile +3 -0

index d69912c..8db3527 100644
@@ -6,6 +6,9 @@ ifeq ($(CONFIG_PPC64),y)
 EXTRA_CFLAGS		+= -mno-minimal-toc
 endif
 
+CFLAGS_REMOVE_code-patching.o = -pg
+CFLAGS_REMOVE_feature-fixups.o = -pg
+
 obj-y			:= string.o alloc.o \
 			   checksum_$(CONFIG_WORD_SIZE).o
 obj-$(CONFIG_PPC32)	+= div64.o copy_32.o crtsavres.o

Steven Rostedt

author maintainer

2008-12-01 18:15 UTC #10 lore raw

Quoted 14 lines · Ingo Molnar

Quoted 11 lines · Steven Rostedt
Quoted 8 lines · Steven Rostedt
Paul,
This patch series addresses the issues you brought up as well as adds some more enhancements and fixes. This series is added on top of the previous patch series.
The new patches are: (and are posted now) 5987225... powerpc/ppc32: static ftrace fixes for PPC32

Quoted 7 lines

Quoted 3 lines
The new commit id with the fix in Author and SoB is:
fd90db4bdc1ccf71927365cd634188cec4d7b151
powerpc/ppc32: static ftrace fixes for
Steve, you rebased the previous commits, and that's a no-no. I sorted it all out, find below the stable URI for stuff that we'd like Paul to pull shortly before the merge window.

Sorry, I rebased to edit the change log where I mispelled Paul's name.

-- Steve

Paul Mackerras

2008-12-19 05:01 UTC #11 lore raw

Quoted 1 line · Steven Rostedt

Sorry, I rebased to edit the change log where I mispelled Paul's name.

So, what commits should I be pulling into powerpc.git, or what patches should I be committing, for dynamic ftrace on powerpc?

Paul.

Steven Rostedt

author maintainer

2008-12-19 05:42 UTC #12 lore raw

Quoted 6 lines · Paul Mackerras

Quoted 1 line · Steven Rostedt
Sorry, I rebased to edit the change log where I mispelled Paul's name.
So, what commits should I be pulling into powerpc.git, or what patches should I be committing, for dynamic ftrace on powerpc?

Ingo did not like the rebase and kept the original commits (the one with the change log with the incorrect spelling).

From: git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip.git
tracing/powerpc
====================================
commit 8fd6e5a8c81e2e9b912ea33c8425a10729db469b

Author: Steven Rostedt <[email protected]>
Date:   Fri Nov 14 16:21:19 2008 -0800

powerpc: ftrace, convert to new dynamic ftrace arch API

Impact: update to PowerPC ftrace arch API

This patch converts PowerPC to use the new dynamic ftrace arch API.

Thanks to Paul Mackennas for pointing out the mistakes of my original test_24bit_addr function.

Signed-off-by Steven Rostedt <[email protected]>

====================================

The branch in my tree 'ppc/ftrace' has not been changed since. The only reason the commits in my tree are different than the ones in tip was because I rebased to fixed the spelling in the change log.

From: git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-2.6-trace.git
ppc/ftrace
====================================
commit a3520364f9376943e4d6c76931bf4df8569fa555

Author: Steven Rostedt <[email protected]>
Date:   Fri Nov 14 16:21:19 2008 -0800

powerpc: ftrace, convert to new dynamic ftrace arch API

Impact: update to PowerPC ftrace arch API

This patch converts PowerPC to use the new dynamic ftrace arch API.

Thanks to Paul Mackerras for pointing out the mistakes of my original test_24bit_addr function.

Signed-off-by Steven Rostedt <[email protected]>

====================================

Take your pick for which to pull from.

Thanks,

-- Steve