Skip to contentSkip to conversation

perf: arch_perf_out_copy_user default

BUG New Peter Zijlstra 4 participants 9 messages Oct 30, 2013 – Nov 6, 2013

Subsystems (2)
lore
Share & tools
lore

Peter Zijlstra

author maintainer
#0 lore raw

Hi Frederic,

I just spotted:

#ifndef arch_perf_out_copy_user
#define arch_perf_out_copy_user __copy_from_user_inatomic #endif

vs:

arch/x86/include/asm/perf_event.h:#define arch_perf_out_copy_user copy_from_user_nmi

Now the problem is that copy_from_user_nmi() and __copy_from_user_inatomic() have different return semantics.

Furthermore, the macro you use them in DEFINE_OUTPUT_COPY() assumes the return value is the amount of memory copied; as also illustrated by memcpy_common().

Trouble is, __copy_from_user_inatomic() returns the number of bytes _NOT_ copied.

With this, my question to Will is, how did your ARM unwind support patches ever work? AFAICT they end up using the __copy_from_user_inatomic() thing.

---
 kernel/events/internal.h | 17 +++++++++++++++--
 1 file changed, 15 insertions(+), 2 deletions(-)
kernel/events/internal.h +15 -2
index ca6599723be5..d7a0f753e695 100644
@@ -110,7 +110,8 @@ func_name(struct perf_output_handle *handle,				\
 	return len;							\
 }
 
-static inline int memcpy_common(void *dst, const void *src, size_t n)
+static inline unsigned long
+memcpy_common(void *dst, const void *src, unsigned long n)
 {
 	memcpy(dst, src, n);
 	return n;
@@ -123,7 +124,19 @@ DEFINE_OUTPUT_COPY(__output_copy, memcpy_common)
 DEFINE_OUTPUT_COPY(__output_skip, MEMCPY_SKIP)
 
 #ifndef arch_perf_out_copy_user
-#define arch_perf_out_copy_user __copy_from_user_inatomic
+#define arch_perf_out_copy_user arch_perf_out_copy_user
+
+static inline unsigned long
+arch_perf_out_copy_user(void *dst, const void *src, unsigned long n)
+{
+	unsigned long ret;
+
+	pagefault_disable();
+	ret = __copy_from_user_inatomic(to, from, n);
+	pagefault_enable();
+
+	return n - ret;
+}
 #endif
 
 DEFINE_OUTPUT_COPY(__output_copy_user, arch_perf_out_copy_user)
author maintainer
#1 lore raw

PATCH 1/1 perf: Fix arch_perf_out_copy_user default implementation

Subject: perf: Fix arch_perf_out_copy_user default implementation

I noticed that the arch_perf_output_copy_user default of __copy_from_user_inatomic has different return semantics that all other memcpy functions we use -- notably also copy_from_user_nmi() for x86.

So provide a compatible default function.

Also change all the memcpy functions to use unsigned long.

---
 kernel/events/internal.h | 31 ++++++++++++++++++++++++-------
 1 file changed, 24 insertions(+), 7 deletions(-)
kernel/events/internal.h +24 -7
index ca6599723be5..325e85776406 100644
@@ -82,14 +82,14 @@ static inline unsigned long perf_data_size(struct ring_buffer *rb)
 }
 
 #define DEFINE_OUTPUT_COPY(func_name, memcpy_func)			\
-static inline unsigned int						\
+static inline unsigned long						\
 func_name(struct perf_output_handle *handle,				\
-	  const void *buf, unsigned int len)				\
+	  const void *buf, unsigned long len)				\
 {									\
 	unsigned long size, written;					\
 									\
 	do {								\
-		size = min_t(unsigned long, handle->size, len);		\
+		size = min(handle->size, len);				\
 									\
 		written = memcpy_func(handle->addr, buf, size);		\
 									\
@@ -110,7 +110,8 @@ func_name(struct perf_output_handle *handle,				\
 	return len;							\
 }
 
-static inline int memcpy_common(void *dst, const void *src, size_t n)
+static inline unsigned long
+memcpy_common(void *dst, const void *src, unsigned long n)
 {
 	memcpy(dst, src, n);
 	return n;
@@ -118,12 +119,28 @@ static inline int memcpy_common(void *dst, const void *src, size_t n)
 
 DEFINE_OUTPUT_COPY(__output_copy, memcpy_common)
 
-#define MEMCPY_SKIP(dst, src, n) (n)
+static inline unsigned long
+memcpy_skip(void *dst, const void *src, unsigned long n)
+{
+	return n;
+}
 
-DEFINE_OUTPUT_COPY(__output_skip, MEMCPY_SKIP)
+DEFINE_OUTPUT_COPY(__output_skip, memcpy_skip)
 
 #ifndef arch_perf_out_copy_user
-#define arch_perf_out_copy_user __copy_from_user_inatomic
+#define arch_perf_out_copy_user arch_perf_out_copy_user
+
+static inline unsigned long
+arch_perf_out_copy_user(void *dst, const void *src, unsigned long n)
+{
+	unsigned long ret;
+
+	pagefault_disable();
+	ret = __copy_from_user_inatomic(dst, src, n);
+	pagefault_enable();
+
+	return n - ret;
+}
 #endif
 
 DEFINE_OUTPUT_COPY(__output_copy_user, arch_perf_out_copy_user)
#2 lore raw
In reply to Peter Zijlstra
Quoted 1 line · Peter Zijlstra

Hi Frederic,

Hi Peter,

Quoted 18 lines

I just spotted:

#ifndef arch_perf_out_copy_user
#define arch_perf_out_copy_user __copy_from_user_inatomic #endif

vs:

arch/x86/include/asm/perf_event.h:#define arch_perf_out_copy_user copy_from_user_nmi

Now the problem is that copy_from_user_nmi() and __copy_from_user_inatomic() have different return semantics.

Furthermore, the macro you use them in DEFINE_OUTPUT_COPY() assumes the return value is the amount of memory copied; as also illustrated by memcpy_common().

Quoted 6 lines

Trouble is, __copy_from_user_inatomic() returns the number of bytes _NOT_ copied.

With this, my question to Will is, how did your ARM unwind support patches ever work? AFAICT they end up using the __copy_from_user_inatomic() thing.

Yeah, that's weird, because they *do* appear to work! In fact, looking at the code in kernel/events/core.c, it looks like __output_copy_user is expected to return the number of bytes not copied, so providing the __copy_from_user_inatomic succeeds first time around, the DEFINE_OUTPUT_COPY macro will return len (dump_size) and the perf_output_skip will deal with the buffer pointers for us. The issue then is that dynamic size will be 0, and the unwind code in perf will never be called (except I know that it *is* being called).

I'll go dig further...

Will

author maintainer
#3 lore raw
Quoted 3 lines · Will Deacon

Yeah, that's weird, because they *do* appear to work! In fact, looking at the code in kernel/events/core.c, it looks like __output_copy_user is expected to return the number of bytes not copied,

Oh, cute, double fail.. head hurts.

#4 lore raw
In reply to Peter Zijlstra
Quoted 15 lines · Peter Zijlstra

Hi Frederic,

I just spotted:

#ifndef arch_perf_out_copy_user
#define arch_perf_out_copy_user __copy_from_user_inatomic #endif

vs:

arch/x86/include/asm/perf_event.h:#define arch_perf_out_copy_user copy_from_user_nmi

Now the problem is that copy_from_user_nmi() and __copy_from_user_inatomic() have different return semantics.

Quoted 6 lines

Furthermore, the macro you use them in DEFINE_OUTPUT_COPY() assumes the return value is the amount of memory copied; as also illustrated by memcpy_common().

Trouble is, __copy_from_user_inatomic() returns the number of bytes _NOT_ copied.

Aie, sorry about that, I did a wrong assumption indeed.

Quoted 31 lines

With this, my question to Will is, how did your ARM unwind support patches ever work? AFAICT they end up using the __copy_from_user_inatomic() thing.

---
 kernel/events/internal.h | 17 +++++++++++++++--
 1 file changed, 15 insertions(+), 2 deletions(-)
kernel/events/internal.h +4 -2
index ca6599723be5..d7a0f753e695 100644
@@ -110,7 +110,8 @@ func_name(struct perf_output_handle *handle,				\
 	return len;							\
 }
 
-static inline int memcpy_common(void *dst, const void *src, size_t n)
+static inline unsigned long
+memcpy_common(void *dst, const void *src, unsigned long n)
 {
 	memcpy(dst, src, n);
 	return n;
@@ -123,7 +124,19 @@ DEFINE_OUTPUT_COPY(__output_copy, memcpy_common)
 DEFINE_OUTPUT_COPY(__output_skip, MEMCPY_SKIP)
 
 #ifndef arch_perf_out_copy_user
-#define arch_perf_out_copy_user __copy_from_user_inatomic
+#define arch_perf_out_copy_user arch_perf_out_copy_user
+

kernel/events/internal.h DEFINE_OUTPUT_COPY(__output_copy, memcpy_common) · line 138

Quoted 10 lines
kernel/events/internal.h +10 -0
+static inline unsigned long
+arch_perf_out_copy_user(void *dst, const void *src, unsigned long n)
+{
+	unsigned long ret;
+
+	pagefault_disable();
+	ret = __copy_from_user_inatomic(to, from, n);
+	pagefault_enable();
+
+	return n - ret;

Would it make sense to rather make copy_from_user_nmi() to use a return value pattern that is closer to those of the existing copy_from_user_*() ?

This way we avoid future mistakes of that kind.

Thanks.

Quoted 4 lines
kernel/events/internal.h +1 -0
+}
 #endif
 
 DEFINE_OUTPUT_COPY(__output_copy_user, arch_perf_out_copy_user)
author maintainer
#5 lore raw
Quoted 3 lines · Frederic Weisbecker

Would it make sense to rather make copy_from_user_nmi() to use a return value pattern that is closer to those of the existing copy_from_user_*() ?

Yeah we can do that I suppose; copy_form_user_nmi() actually uses __copy_from_user_inatomic() since about a week.

Something like so I suppose.. please check, I'm in fail mode.

It looks like DEFINE_OUTPUT_COPY() functions already returned the bytes not copied, and all its users appear to indeed expect that.

---
 arch/x86/kernel/cpu/perf_event.c           |  4 ++--
 arch/x86/kernel/cpu/perf_event_intel_ds.c  |  2 +-
 arch/x86/kernel/cpu/perf_event_intel_lbr.c |  2 +-
 arch/x86/lib/usercopy.c                    |  2 +-
 arch/x86/oprofile/backtrace.c              |  4 ++--
 kernel/events/internal.h                   | 35 ++++++++++++++++++++++--------
 6 files changed, 33 insertions(+), 16 deletions(-)
arch/x86/kernel/cpu/perf_event.c +2 -2
index 94bf3010bb39..0fa7dbe756e6 100644
@@ -1998,7 +1998,7 @@ perf_callchain_user32(struct pt_regs *regs, struct perf_callchain_entry *entry)
 		frame.return_address = 0;
 
 		bytes = copy_from_user_nmi(&frame, fp, sizeof(frame));
-		if (bytes != sizeof(frame))
+		if (bytes != 0)
 			break;
 
 		if (!valid_user_frame(fp, sizeof(frame)))
@@ -2050,7 +2050,7 @@ perf_callchain_user(struct perf_callchain_entry *entry, struct pt_regs *regs)
 		frame.return_address = 0;
 
 		bytes = copy_from_user_nmi(&frame, fp, sizeof(frame));
-		if (bytes != sizeof(frame))
+		if (bytes != 0)
 			break;
 
 		if (!valid_user_frame(fp, sizeof(frame)))
arch/x86/kernel/cpu/perf_event_intel_ds.c +1 -1
index c1760ff3c757..ae96cfa5eddd 100644
@@ -789,7 +789,7 @@ static int intel_pmu_pebs_fixup_ip(struct pt_regs *regs)
 
 		size = ip - to; /* Must fit our buffer, see above */
 		bytes = copy_from_user_nmi(buf, (void __user *)to, size);
-		if (bytes != size)
+		if (bytes != 0)
 			return 0;
 
 		kaddr = buf;
arch/x86/kernel/cpu/perf_event_intel_lbr.c +1 -1
index 90ee6c1d0542..d82d155aca8c 100644
@@ -491,7 +491,7 @@ static int branch_type(unsigned long from, unsigned long to, int abort)
 
 		/* may fail if text not present */
 		bytes = copy_from_user_nmi(buf, (void __user *)from, size);
-		if (bytes != size)
+		if (bytes != 0)
 			return X86_BR_NONE;
 
 		addr = buf;
arch/x86/lib/usercopy.c +1 -1
index 5465b8613944..ddf9ecb53cc3 100644
@@ -31,6 +31,6 @@ copy_from_user_nmi(void *to, const void __user *from, unsigned long n)
 	ret = __copy_from_user_inatomic(to, from, n);
 	pagefault_enable();
 
-	return n - ret;
+	return ret;
 }
 EXPORT_SYMBOL_GPL(copy_from_user_nmi);
arch/x86/oprofile/backtrace.c +2 -2
index d6aa6e8315d1..5d04be5efb64 100644
@@ -47,7 +47,7 @@ dump_user_backtrace_32(struct stack_frame_ia32 *head)
 	unsigned long bytes;
 
 	bytes = copy_from_user_nmi(bufhead, head, sizeof(bufhead));
-	if (bytes != sizeof(bufhead))
+	if (bytes != 0)
 		return NULL;
 
 	fp = (struct stack_frame_ia32 *) compat_ptr(bufhead[0].next_frame);
@@ -93,7 +93,7 @@ static struct stack_frame *dump_user_backtrace(struct stack_frame *head)
 	unsigned long bytes;
 
 	bytes = copy_from_user_nmi(bufhead, head, sizeof(bufhead));
-	if (bytes != sizeof(bufhead))
+	if (bytes != 0)
 		return NULL;
 
 	oprofile_add_trace(bufhead[0].return_address);
kernel/events/internal.h +26 -9
index ca6599723be5..569b218782ad 100644
@@ -82,16 +82,16 @@ static inline unsigned long perf_data_size(struct ring_buffer *rb)
 }
 
 #define DEFINE_OUTPUT_COPY(func_name, memcpy_func)			\
-static inline unsigned int						\
+static inline unsigned long						\
 func_name(struct perf_output_handle *handle,				\
-	  const void *buf, unsigned int len)				\
+	  const void *buf, unsigned long len)				\
 {									\
 	unsigned long size, written;					\
 									\
 	do {								\
-		size = min_t(unsigned long, handle->size, len);		\
-									\
+		size    = min(handle->size, len);			\
 		written = memcpy_func(handle->addr, buf, size);		\
+		written = size - written;				\
 									\
 		len -= written;						\
 		handle->addr += written;				\
@@ -110,20 +110,37 @@ func_name(struct perf_output_handle *handle,				\
 	return len;							\
 }
 
-static inline int memcpy_common(void *dst, const void *src, size_t n)
+static inline unsigned long
+memcpy_common(void *dst, const void *src, unsigned long n)
 {
 	memcpy(dst, src, n);
-	return n;
+	return 0;
 }
 
 DEFINE_OUTPUT_COPY(__output_copy, memcpy_common)
 
-#define MEMCPY_SKIP(dst, src, n) (n)
+static inline unsigned long
+memcpy_skip(void *dst, const void *src, unsigned long n)
+{
+	return 0;
+}
 
-DEFINE_OUTPUT_COPY(__output_skip, MEMCPY_SKIP)
+DEFINE_OUTPUT_COPY(__output_skip, memcpy_skip)
 
 #ifndef arch_perf_out_copy_user
-#define arch_perf_out_copy_user __copy_from_user_inatomic
+#define arch_perf_out_copy_user arch_perf_out_copy_user
+
+static inline unsigned long
+arch_perf_out_copy_user(void *dst, const void *src, unsigned long n)
+{
+	unsigned long ret;
+
+	pagefault_disable();
+	ret = __copy_from_user_inatomic(dst, src, n);
+	pagefault_enable();
+
+	return ret;
+}
 #endif
 
 DEFINE_OUTPUT_COPY(__output_copy_user, arch_perf_out_copy_user)
#6 lore raw
Quoted 7 lines · Peter Zijlstra
Quoted 3 lines · Frederic Weisbecker

Would it make sense to rather make copy_from_user_nmi() to use a return value pattern that is closer to those of the existing copy_from_user_*() ?

Yeah we can do that I suppose; copy_form_user_nmi() actually uses __copy_from_user_inatomic() since about a week.

Ah! Looks like we can even move copy_from_user_nmi() to asm-generic/uaccess.h then. Its fallback implementation after this patch is the same than the x86 one anyway.

Hmm but there is the range_not_ok()...

Quoted 109 lines

Something like so I suppose.. please check, I'm in fail mode.

It looks like DEFINE_OUTPUT_COPY() functions already returned the bytes not copied, and all its users appear to indeed expect that.

---
 arch/x86/kernel/cpu/perf_event.c           |  4 ++--
 arch/x86/kernel/cpu/perf_event_intel_ds.c  |  2 +-
 arch/x86/kernel/cpu/perf_event_intel_lbr.c |  2 +-
 arch/x86/lib/usercopy.c                    |  2 +-
 arch/x86/oprofile/backtrace.c              |  4 ++--
 kernel/events/internal.h                   | 35 ++++++++++++++++++++++--------
 6 files changed, 33 insertions(+), 16 deletions(-)
arch/x86/kernel/cpu/perf_event.c +2 -2
index 94bf3010bb39..0fa7dbe756e6 100644
@@ -1998,7 +1998,7 @@ perf_callchain_user32(struct pt_regs *regs, struct perf_callchain_entry *entry)
 		frame.return_address = 0;
 
 		bytes = copy_from_user_nmi(&frame, fp, sizeof(frame));
-		if (bytes != sizeof(frame))
+		if (bytes != 0)
 			break;
 
 		if (!valid_user_frame(fp, sizeof(frame)))
@@ -2050,7 +2050,7 @@ perf_callchain_user(struct perf_callchain_entry *entry, struct pt_regs *regs)
 		frame.return_address = 0;
 
 		bytes = copy_from_user_nmi(&frame, fp, sizeof(frame));
-		if (bytes != sizeof(frame))
+		if (bytes != 0)
 			break;
 
 		if (!valid_user_frame(fp, sizeof(frame)))
arch/x86/kernel/cpu/perf_event_intel_ds.c +1 -1
index c1760ff3c757..ae96cfa5eddd 100644
@@ -789,7 +789,7 @@ static int intel_pmu_pebs_fixup_ip(struct pt_regs *regs)
 
 		size = ip - to; /* Must fit our buffer, see above */
 		bytes = copy_from_user_nmi(buf, (void __user *)to, size);
-		if (bytes != size)
+		if (bytes != 0)
 			return 0;
 
 		kaddr = buf;
arch/x86/kernel/cpu/perf_event_intel_lbr.c +1 -1
index 90ee6c1d0542..d82d155aca8c 100644
@@ -491,7 +491,7 @@ static int branch_type(unsigned long from, unsigned long to, int abort)
 
 		/* may fail if text not present */
 		bytes = copy_from_user_nmi(buf, (void __user *)from, size);
-		if (bytes != size)
+		if (bytes != 0)
 			return X86_BR_NONE;
 
 		addr = buf;
arch/x86/lib/usercopy.c +1 -1
index 5465b8613944..ddf9ecb53cc3 100644
@@ -31,6 +31,6 @@ copy_from_user_nmi(void *to, const void __user *from, unsigned long n)
 	ret = __copy_from_user_inatomic(to, from, n);
 	pagefault_enable();
 
-	return n - ret;
+	return ret;
 }
 EXPORT_SYMBOL_GPL(copy_from_user_nmi);
arch/x86/oprofile/backtrace.c +2 -2
index d6aa6e8315d1..5d04be5efb64 100644
@@ -47,7 +47,7 @@ dump_user_backtrace_32(struct stack_frame_ia32 *head)
 	unsigned long bytes;
 
 	bytes = copy_from_user_nmi(bufhead, head, sizeof(bufhead));
-	if (bytes != sizeof(bufhead))
+	if (bytes != 0)
 		return NULL;
 
 	fp = (struct stack_frame_ia32 *) compat_ptr(bufhead[0].next_frame);
@@ -93,7 +93,7 @@ static struct stack_frame *dump_user_backtrace(struct stack_frame *head)
 	unsigned long bytes;
 
 	bytes = copy_from_user_nmi(bufhead, head, sizeof(bufhead));
-	if (bytes != sizeof(bufhead))
+	if (bytes != 0)
 		return NULL;
 
 	oprofile_add_trace(bufhead[0].return_address);
kernel/events/internal.h +1 -2
index ca6599723be5..569b218782ad 100644
@@ -82,16 +82,16 @@ static inline unsigned long perf_data_size(struct ring_buffer *rb)
 }
 
 #define DEFINE_OUTPUT_COPY(func_name, memcpy_func)			\
-static inline unsigned int						\
+static inline unsigned long						\
 func_name(struct perf_output_handle *handle,				\
-	  const void *buf, unsigned int len)				\

kernel/events/internal.h static inline unsigned long perf_data_size(struct ring_buffer *rb) · line 94

Quoted 10 lines
kernel/events/internal.h +3 -2
+	  const void *buf, unsigned long len)				\
 {									\
 	unsigned long size, written;					\
 									\
 	do {								\
-		size = min_t(unsigned long, handle->size, len);		\
-									\
+		size    = min(handle->size, len);			\
 		written = memcpy_func(handle->addr, buf, size);		\
+		written = size - written;				\

I assume copy_from_user_inatomic() shouldn't return negative value. I'm a bit confused by it's long return type but I haven't found a place where it returns anything else that the bytes copied. I only checked a very few archs though.

Quoted 34 lines
kernel/events/internal.h +13 -5
 									\
 		len -= written;						\
 		handle->addr += written;				\
@@ -110,20 +110,37 @@ func_name(struct perf_output_handle *handle,				\
 	return len;							\
 }
 
-static inline int memcpy_common(void *dst, const void *src, size_t n)
+static inline unsigned long
+memcpy_common(void *dst, const void *src, unsigned long n)
 {
 	memcpy(dst, src, n);
-	return n;
+	return 0;
 }
 
 DEFINE_OUTPUT_COPY(__output_copy, memcpy_common)
 
-#define MEMCPY_SKIP(dst, src, n) (n)
+static inline unsigned long
+memcpy_skip(void *dst, const void *src, unsigned long n)
+{
+	return 0;
+}
 
-DEFINE_OUTPUT_COPY(__output_skip, MEMCPY_SKIP)
+DEFINE_OUTPUT_COPY(__output_skip, memcpy_skip)
 
 #ifndef arch_perf_out_copy_user
-#define arch_perf_out_copy_user __copy_from_user_inatomic
+#define arch_perf_out_copy_user arch_perf_out_copy_user
+
+static inline unsigned long
+arch_perf_out_copy_user(void *dst, const void *src, unsigned long n)

kernel/events/internal.h func_name(struct perf_output_handle *handle, \ · line 144

Quoted 10 lines
kernel/events/internal.h +9 -0
+{
+	unsigned long ret;
+
+	pagefault_disable();
+	ret = __copy_from_user_inatomic(dst, src, n);
+	pagefault_enable();
+
+	return ret;
+}
 #endif

Ok, that all looks good!

Thanks!

Quoted 2 lines
kernel/events/internal.h +0 -0
 DEFINE_OUTPUT_COPY(__output_copy_user, arch_perf_out_copy_user)
Acked-by
#7 lore raw
In reply to Peter Zijlstra

Hi Peter,

Quoted 13 lines · Peter Zijlstra
Quoted 3 lines · Frederic Weisbecker

Would it make sense to rather make copy_from_user_nmi() to use a return value pattern that is closer to those of the existing copy_from_user_*() ?

Yeah we can do that I suppose; copy_form_user_nmi() actually uses __copy_from_user_inatomic() since about a week.

Something like so I suppose.. please check, I'm in fail mode.

It looks like DEFINE_OUTPUT_COPY() functions already returned the bytes not copied, and all its users appear to indeed expect that.

Quoted 8 lines
---
 arch/x86/kernel/cpu/perf_event.c           |  4 ++--
 arch/x86/kernel/cpu/perf_event_intel_ds.c  |  2 +-
 arch/x86/kernel/cpu/perf_event_intel_lbr.c |  2 +-
 arch/x86/lib/usercopy.c                    |  2 +-
 arch/x86/oprofile/backtrace.c              |  4 ++--
 kernel/events/internal.h                   | 35 ++++++++++++++++++++++--------
 6 files changed, 33 insertions(+), 16 deletions(-)

The core changes look good to me, and I've tested that this still produces callchain information for ARM. It's odd that this worked so well before...

Will

#8 lore raw

[tip:perf/core] perf: Fix arch_perf_out_copy_user default

In reply to Peter Zijlstra
perf: Fix arch_perf_out_copy_user default
Commit-ID:  0a196848ca365ec582c6d86659be456be6d4ed96
Gitweb:     http://git.kernel.org/tip/0a196848ca365ec582c6d86659be456be6d4ed96
Author:     Peter Zijlstra <[email protected]>
AuthorDate: Wed, 30 Oct 2013 21:16:22 +0100
Committer:  Ingo Molnar <[email protected]>
CommitDate: Wed, 6 Nov 2013 12:34:25 +0100

perf: Fix arch_perf_out_copy_user default

The arch_perf_output_copy_user() default of __copy_from_user_inatomic() returns bytes not copied, while all other argument functions given DEFINE_OUTPUT_COPY() return bytes copied.

Since copy_from_user_nmi() is the odd duck out by returning bytes copied where all other *copy_{to,from}* functions return bytes not copied, change it over and ammend DEFINE_OUTPUT_COPY() to expect bytes not copied.

Oddly enough DEFINE_OUTPUT_COPY() already returned bytes not copied while expecting its worker functions to return bytes copied.

---
 arch/x86/kernel/cpu/perf_event.c           |  4 ++--
 arch/x86/kernel/cpu/perf_event_intel_ds.c  |  2 +-
 arch/x86/kernel/cpu/perf_event_intel_lbr.c |  2 +-
 arch/x86/lib/usercopy.c                    |  2 +-
 arch/x86/oprofile/backtrace.c              |  4 ++--
 kernel/events/internal.h                   | 35 ++++++++++++++++++++++--------
 6 files changed, 33 insertions(+), 16 deletions(-)
arch/x86/kernel/cpu/perf_event.c +2 -2
index 8a87a32..8e13293 100644
@@ -1989,7 +1989,7 @@ perf_callchain_user32(struct pt_regs *regs, struct perf_callchain_entry *entry)
 		frame.return_address = 0;
 
 		bytes = copy_from_user_nmi(&frame, fp, sizeof(frame));
-		if (bytes != sizeof(frame))
+		if (bytes != 0)
 			break;
 
 		if (!valid_user_frame(fp, sizeof(frame)))
@@ -2041,7 +2041,7 @@ perf_callchain_user(struct perf_callchain_entry *entry, struct pt_regs *regs)
 		frame.return_address = 0;
 
 		bytes = copy_from_user_nmi(&frame, fp, sizeof(frame));
-		if (bytes != sizeof(frame))
+		if (bytes != 0)
 			break;
 
 		if (!valid_user_frame(fp, sizeof(frame)))
arch/x86/kernel/cpu/perf_event_intel_ds.c +1 -1
index c1760ff..ae96cfa 100644
@@ -789,7 +789,7 @@ static int intel_pmu_pebs_fixup_ip(struct pt_regs *regs)
 
 		size = ip - to; /* Must fit our buffer, see above */
 		bytes = copy_from_user_nmi(buf, (void __user *)to, size);
-		if (bytes != size)
+		if (bytes != 0)
 			return 0;
 
 		kaddr = buf;
arch/x86/kernel/cpu/perf_event_intel_lbr.c +1 -1
index 90ee6c1..d82d155 100644
@@ -491,7 +491,7 @@ static int branch_type(unsigned long from, unsigned long to, int abort)
 
 		/* may fail if text not present */
 		bytes = copy_from_user_nmi(buf, (void __user *)from, size);
-		if (bytes != size)
+		if (bytes != 0)
 			return X86_BR_NONE;
 
 		addr = buf;
arch/x86/lib/usercopy.c +1 -1
index 5465b86..ddf9ecb 100644
@@ -31,6 +31,6 @@ copy_from_user_nmi(void *to, const void __user *from, unsigned long n)
 	ret = __copy_from_user_inatomic(to, from, n);
 	pagefault_enable();
 
-	return n - ret;
+	return ret;
 }
 EXPORT_SYMBOL_GPL(copy_from_user_nmi);
arch/x86/oprofile/backtrace.c +2 -2
index d6aa6e8..5d04be5 100644
@@ -47,7 +47,7 @@ dump_user_backtrace_32(struct stack_frame_ia32 *head)
 	unsigned long bytes;
 
 	bytes = copy_from_user_nmi(bufhead, head, sizeof(bufhead));
-	if (bytes != sizeof(bufhead))
+	if (bytes != 0)
 		return NULL;
 
 	fp = (struct stack_frame_ia32 *) compat_ptr(bufhead[0].next_frame);
@@ -93,7 +93,7 @@ static struct stack_frame *dump_user_backtrace(struct stack_frame *head)
 	unsigned long bytes;
 
 	bytes = copy_from_user_nmi(bufhead, head, sizeof(bufhead));
-	if (bytes != sizeof(bufhead))
+	if (bytes != 0)
 		return NULL;
 
 	oprofile_add_trace(bufhead[0].return_address);
kernel/events/internal.h +26 -9
index ca65997..569b2187 100644
@@ -82,16 +82,16 @@ static inline unsigned long perf_data_size(struct ring_buffer *rb)
 }
 
 #define DEFINE_OUTPUT_COPY(func_name, memcpy_func)			\
-static inline unsigned int						\
+static inline unsigned long						\
 func_name(struct perf_output_handle *handle,				\
-	  const void *buf, unsigned int len)				\
+	  const void *buf, unsigned long len)				\
 {									\
 	unsigned long size, written;					\
 									\
 	do {								\
-		size = min_t(unsigned long, handle->size, len);		\
-									\
+		size    = min(handle->size, len);			\
 		written = memcpy_func(handle->addr, buf, size);		\
+		written = size - written;				\
 									\
 		len -= written;						\
 		handle->addr += written;				\
@@ -110,20 +110,37 @@ func_name(struct perf_output_handle *handle,				\
 	return len;							\
 }
 
-static inline int memcpy_common(void *dst, const void *src, size_t n)
+static inline unsigned long
+memcpy_common(void *dst, const void *src, unsigned long n)
 {
 	memcpy(dst, src, n);
-	return n;
+	return 0;
 }
 
 DEFINE_OUTPUT_COPY(__output_copy, memcpy_common)
 
-#define MEMCPY_SKIP(dst, src, n) (n)
+static inline unsigned long
+memcpy_skip(void *dst, const void *src, unsigned long n)
+{
+	return 0;
+}
 
-DEFINE_OUTPUT_COPY(__output_skip, MEMCPY_SKIP)
+DEFINE_OUTPUT_COPY(__output_skip, memcpy_skip)
 
 #ifndef arch_perf_out_copy_user
-#define arch_perf_out_copy_user __copy_from_user_inatomic
+#define arch_perf_out_copy_user arch_perf_out_copy_user
+
+static inline unsigned long
+arch_perf_out_copy_user(void *dst, const void *src, unsigned long n)
+{
+	unsigned long ret;
+
+	pagefault_disable();
+	ret = __copy_from_user_inatomic(dst, src, n);
+	pagefault_enable();
+
+	return ret;
+}
 #endif
 
 DEFINE_OUTPUT_COPY(__output_copy_user, arch_perf_out_copy_user)