oprofile fixes for 3.2

GIT PULL Superseded Robert Richter 3 participants 9 messages Dec 19, 2011 – Dec 23, 2011

Subsystems (4)

Stable branch Arm (32-bit) s390 SuperH

lore

Share & tools

lore

Robert Richter

author

2011-12-19 13:47 UTC #0 lore raw

Ingo,

please pull this one fix for 3.2.

Thanks,

-Robert

The following changes since commit dc47ce90c3a822cd7c9e9339fe4d5f61dcb26b50:

Linux 3.2-rc5 (2011-12-09 15:09:32 -0800)

are available in the git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/rric/oprofile.git urgent

Robert Richter (1):
oprofile: Fix uninitialized memory access when writing to oprofilefs

 arch/s390/oprofile/init.c         |    3 +++
 drivers/oprofile/oprofile_files.c |    9 +++++++++
 drivers/oprofile/oprofilefs.c     |   15 ++++++++++++++-
 3 files changed, 26 insertions(+), 1 deletions(-)

Robert Richter

author

2011-12-19 13:46 UTC #1 lore raw

PATCH 1/1 oprofile: Fix uninitialized memory access when writing to oprofilefs

If oprofilefs_ulong_from_user() is called with count equals zero, *val must be initialized. Otherwise *val is later used uninitialized as no error is returned. Alternatively oprofilefs_ulong_from_user() may not be called if !count. This patch fixes usage of oprofilefs_ulong_from_ user().

This follows write syscall implementation when count is zero: "If count is zero ... [and if] no errors are detected, 0 will be returned without causing any other effect." (man 2 write)

Reported-By Mike Waychison <[email protected]>

Cc Andrew Morton <[email protected]>

Cc [email protected]

Signed-off-by Robert Richter <[email protected]>

---
 arch/s390/oprofile/init.c         |    3 +++
 drivers/oprofile/oprofile_files.c |    9 +++++++++
 drivers/oprofile/oprofilefs.c     |   15 ++++++++++++++-
 3 files changed, 26 insertions(+), 1 deletions(-)

arch/s390/oprofile/init.c +3 -0

index 6efc18b..5d605f1 100644
@@ -87,6 +87,9 @@ static ssize_t hwsampler_write(struct file *file, char const __user *buf,
 	if (*offset)
 		return -EINVAL;
 
+	if (!count)
+		return 0;
+
 	retval = oprofilefs_ulong_from_user(&val, buf, count);
 	if (retval)
 		return retval;

drivers/oprofile/oprofile_files.c +9 -0

index 89f6345..8265b41 100644
@@ -44,6 +44,9 @@ static ssize_t timeout_write(struct file *file, char const __user *buf,
 	if (*offset)
 		return -EINVAL;
 
+	if (!count)
+		return 0;
+
 	retval = oprofilefs_ulong_from_user(&val, buf, count);
 	if (retval)
 		return retval;

@@ -83,6 +86,9 @@ static ssize_t depth_write(struct file *file, char const __user *buf, size_t cou
 	if (!oprofile_ops.backtrace)
 		return -EINVAL;
 
+	if (!count)
+		return 0;
+
 	retval = oprofilefs_ulong_from_user(&val, buf, count);
 	if (retval)
 		return retval;

@@ -140,6 +146,9 @@ static ssize_t enable_write(struct file *file, char const __user *buf, size_t co
 	if (*offset)
 		return -EINVAL;
 
+	if (!count)
+		return 0;
+
 	retval = oprofilefs_ulong_from_user(&val, buf, count);
 	if (retval)
 		return retval;

drivers/oprofile/oprofilefs.c +14 -1

index d0de6cc..1caf1b0 100644
@@ -59,7 +59,17 @@ ssize_t oprofilefs_ulong_to_user(unsigned long val, char __user *buf, size_t cou
 	return simple_read_from_buffer(buf, count, offset, tmpbuf, maxlen);
 }
 
-
+/*
+ * Note: oprofilefs_ulong_from_user() must be called with *val
+ * initialized, otherwise *val is used uninitialized if !count. This
+ * follows write syscall implementation when count is zero: "If count
+ * is zero ... [and if] no errors are detected, 0 will be returned
+ * without causing any other effect." (man 2 write)
+ *
+ * In case *val is a temporary variable, oprofilefs_ulong_from_user()
+ * may not be called if !count. This causes race conditions due to
+ * missing locking of *var.
+ */
 int oprofilefs_ulong_from_user(unsigned long *val, char const __user *buf, size_t count)
 {
 	char tmpbuf[TMPBUFSIZE];

@@ -98,6 +108,9 @@ static ssize_t ulong_write_file(struct file *file, char const __user *buf, size_
 	if (*offset)
 		return -EINVAL;
 
+	if (!count)
+		return 0;
+
 	retval = oprofilefs_ulong_from_user(&value, buf, count);
 	if (retval)
 		return retval;

Signature · 1 line

1.7.7

Ingo Molnar

2011-12-19 14:02 UTC #2 lore raw

In reply to Robert Richter

Quoted 17 lines · Robert Richter

Ingo,
please pull this one fix for 3.2.
Thanks,
-Robert
The following changes since commit dc47ce90c3a822cd7c9e9339fe4d5f61dcb26b50:
Linux 3.2-rc5 (2011-12-09 15:09:32 -0800)
are available in the git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/rric/oprofile.git urgent

Quoted 7 lines

Robert Richter (1):
oprofile: Fix uninitialized memory access when writing to oprofilefs

 arch/s390/oprofile/init.c         |    3 +++
 drivers/oprofile/oprofile_files.c |    9 +++++++++
 drivers/oprofile/oprofilefs.c     |   15 ++++++++++++++-
 3 files changed, 26 insertions(+), 1 deletions(-)

Hm, i really don't like this:

+ if (!count)
+ return 0;
+
retval = oprofilefs_ulong_from_user(&val, buf, count); if (retval)
return retval;

+ if (!count)
+ return 0;
+
retval = oprofilefs_ulong_from_user(&value, buf, count); if (retval)
return retval;

See the ugly and fragile pattern?

This should *really* be solved via the oprofilefs_ulong_from_user() helper function, not by sprinkling the !count checks in half a dozen places ...

Thanks,

Ingo

Robert Richter

author

2011-12-19 14:17 UTC #3 lore raw

Quoted 7 lines · Ingo Molnar

(quoted diff) +3 -0

+	if (!count)
+		return 0;
+
 	retval = oprofilefs_ulong_from_user(&value, buf, count);
 	if (retval)
 		return retval;

Quoted 5 lines

See the ugly and fragile pattern?
This should *really* be solved via the oprofilefs_ulong_from_user() helper function, not by sprinkling the !count checks in half a dozen places ...

Hmm, I thought there was no way to leave the code path with count == 0 and retval. But thinking about it it would be possible with returning count or errors < 0. Will improve the patch.

Thanks,

-Robert

Signature · 2 lines

Advanced Micro Devices, Inc.
Operating System Research Center

Robert Richter

author

2011-12-19 15:39 UTC #4 lore raw

PATCH v2 1/1 oprofile: Fix uninitialized memory access when writing to

Quoted 11 lines · Robert Richter

Quoted 9 lines · Ingo Molnar
(quoted diff) +3 -0
+	if (!count)
+		return 0;
+
 	retval = oprofilefs_ulong_from_user(&value, buf, count);
 	if (retval)
 		return retval;
See the ugly and fragile pattern?

Quoted 7 lines

Quoted 3 lines
This should *really* be solved via the oprofilefs_ulong_from_user() helper function, not by sprinkling the !count checks in half a dozen places ...
Hmm, I thought there was no way to leave the code path with count == 0 and retval. But thinking about it it would be possible with returning count or errors < 0. Will improve the patch.

Ingo, see my updated version below. I am fine with appling it directly on tip/perf/urgent.

Thanks,

-Robert

Quoted 1 line

From f07214bc05560f63da8404d31aa83034a16c6229 Mon Sep 17 00:00:00 2001

From: Robert Richter <[email protected]>
Date: Fri, 16 Dec 2011 15:45:31 +0100
Subject: [PATCH] oprofile: Fix uninitialized memory access when writing to
oprofilefs

If oprofilefs_ulong_from_user() is called with count equals zero, *val remains unchanged. Depending on the implementation it might be uninitialized.

Change oprofilefs_ulong_from_user()'s interface to return count on success. Thus, we are able to return early if count equals zero which avoids using *val uninitialized. Fixing all users of oprofilefs_ulong_ from_user().

This follows write syscall implementation when count is zero: "If count is zero ... [and if] no errors are detected, 0 will be returned without causing any other effect." (man 2 write)

Reported-By Mike Waychison <[email protected]>

Cc Andrew Morton <[email protected]>

Cc [email protected]

Signed-off-by Robert Richter <[email protected]>

---
 arch/s390/oprofile/init.c         |    2 +-
 drivers/oprofile/oprofile_files.c |    7 ++++---
 drivers/oprofile/oprofilefs.c     |   11 +++++++++--
 3 files changed, 14 insertions(+), 6 deletions(-)

arch/s390/oprofile/init.c +1 -1

index 6efc18b..bd58b72 100644
@@ -88,7 +88,7 @@ static ssize_t hwsampler_write(struct file *file, char const __user *buf,
 		return -EINVAL;
 
 	retval = oprofilefs_ulong_from_user(&val, buf, count);
-	if (retval)
+	if (retval <= 0)
 		return retval;
 
 	if (oprofile_started)

drivers/oprofile/oprofile_files.c +4 -3

index 89f6345..84a208d 100644
@@ -45,7 +45,7 @@ static ssize_t timeout_write(struct file *file, char const __user *buf,
 		return -EINVAL;
 
 	retval = oprofilefs_ulong_from_user(&val, buf, count);
-	if (retval)
+	if (retval <= 0)
 		return retval;
 
 	retval = oprofile_set_timeout(val);

@@ -84,7 +84,7 @@ static ssize_t depth_write(struct file *file, char const __user *buf, size_t cou
 		return -EINVAL;
 
 	retval = oprofilefs_ulong_from_user(&val, buf, count);
-	if (retval)
+	if (retval <= 0)
 		return retval;
 
 	retval = oprofile_set_ulong(&oprofile_backtrace_depth, val);

@@ -141,9 +141,10 @@ static ssize_t enable_write(struct file *file, char const __user *buf, size_t co
 		return -EINVAL;
 
 	retval = oprofilefs_ulong_from_user(&val, buf, count);
-	if (retval)
+	if (retval <= 0)
 		return retval;
 
+	retval = 0;
 	if (val)
 		retval = oprofile_start();
 	else

drivers/oprofile/oprofilefs.c +9 -2

index d0de6cc..2f0aa0f 100644
@@ -60,6 +60,13 @@ ssize_t oprofilefs_ulong_to_user(unsigned long val, char __user *buf, size_t cou
 }
 
 
+/*
+ * Note: If oprofilefs_ulong_from_user() returns 0, then *val remains
+ * unchanged and might be uninitialized. This follows write syscall
+ * implementation when count is zero: "If count is zero ... [and if]
+ * no errors are detected, 0 will be returned without causing any
+ * other effect." (man 2 write)
+ */
 int oprofilefs_ulong_from_user(unsigned long *val, char const __user *buf, size_t count)
 {
 	char tmpbuf[TMPBUFSIZE];

@@ -79,7 +86,7 @@ int oprofilefs_ulong_from_user(unsigned long *val, char const __user *buf, size_
 	raw_spin_lock_irqsave(&oprofilefs_lock, flags);
 	*val = simple_strtoul(tmpbuf, NULL, 0);
 	raw_spin_unlock_irqrestore(&oprofilefs_lock, flags);
-	return 0;
+	return count;
 }

@@ -99,7 +106,7 @@ static ssize_t ulong_write_file(struct file *file, char const __user *buf, size_
 		return -EINVAL;
 
 	retval = oprofilefs_ulong_from_user(&value, buf, count);
-	if (retval)
+	if (retval <= 0)
 		return retval;
 
 	retval = oprofile_set_ulong(file->private_data, value);

Signature · 8 lines

1.7.7




-- 
Advanced Micro Devices, Inc.
Operating System Research Center

Ingo Molnar

Applied

2011-12-19 16:21 UTC #5 lore raw

Re: [PATCH v2] oprofile: Fix uninitialized memory access when writing to

Quoted 20 lines · Robert Richter

Quoted 11 lines · Robert Richter
Quoted 9 lines · Ingo Molnar
(quoted diff) +3 -0
+	if (!count)
+		return 0;
+
 	retval = oprofilefs_ulong_from_user(&value, buf, count);
 	if (retval)
 		return retval;
See the ugly and fragile pattern?
Quoted 7 lines
Quoted 3 lines
This should *really* be solved via the oprofilefs_ulong_from_user() helper function, not by sprinkling the !count checks in half a dozen places ...
Hmm, I thought there was no way to leave the code path with count == 0 and retval. But thinking about it it would be possible with returning count or errors < 0. Will improve the patch.

Quoted 1 line

Ingo, see my updated version below. [...]

Yeah, this looks a lot cleaner.

Quoted 1 line

[...] I am fine with appling it directly on tip/perf/urgent.

Applied,

Thanks,

Ingo

tip-bot for Robert Richter

author

2011-12-20 08:57 UTC #6 lore raw

[tip:perf/urgent] oprofile: Fix uninitialized memory access when writing to writing to oprofilefs

In reply to Robert Richter

Commit-ID:  913050b91eb94f194392dd797b1ff3779f606ac0
Gitweb:     http://git.kernel.org/tip/913050b91eb94f194392dd797b1ff3779f606ac0
Author:     Robert Richter <[email protected]>
AuthorDate: Mon, 19 Dec 2011 16:38:30 +0100
Committer:  Ingo Molnar <[email protected]>
CommitDate: Mon, 19 Dec 2011 17:18:43 +0100

oprofile: Fix uninitialized memory access when writing to writing to oprofilefs

If oprofilefs_ulong_from_user() is called with count equals zero, *val remains unchanged. Depending on the implementation it might be uninitialized.

This follows write syscall implementation when count is zero:
"If count is zero ... [and if] no errors are detected, 0 will be returned without causing any other effect." (man 2 write)

Reported-By Mike Waychison <[email protected]>

Signed-off-by Robert Richter <[email protected]>

Cc Andrew Morton <[email protected]>

Cc <[email protected]>

Cc oprofile-list <[email protected]>

Link http://lkml.kernel.org/r/[email protected] open here

Signed-off-by Ingo Molnar <[email protected]>

---
 arch/s390/oprofile/init.c         |    2 +-
 drivers/oprofile/oprofile_files.c |    7 ++++---
 drivers/oprofile/oprofilefs.c     |   11 +++++++++--
 3 files changed, 14 insertions(+), 6 deletions(-)

arch/s390/oprofile/init.c +1 -1

index 6efc18b..bd58b72 100644
@@ -88,7 +88,7 @@ static ssize_t hwsampler_write(struct file *file, char const __user *buf,
 		return -EINVAL;
 
 	retval = oprofilefs_ulong_from_user(&val, buf, count);
-	if (retval)
+	if (retval <= 0)
 		return retval;
 
 	if (oprofile_started)

drivers/oprofile/oprofile_files.c +4 -3

index 89f6345..84a208d 100644
@@ -45,7 +45,7 @@ static ssize_t timeout_write(struct file *file, char const __user *buf,
 		return -EINVAL;
 
 	retval = oprofilefs_ulong_from_user(&val, buf, count);
-	if (retval)
+	if (retval <= 0)
 		return retval;
 
 	retval = oprofile_set_timeout(val);

@@ -84,7 +84,7 @@ static ssize_t depth_write(struct file *file, char const __user *buf, size_t cou
 		return -EINVAL;
 
 	retval = oprofilefs_ulong_from_user(&val, buf, count);
-	if (retval)
+	if (retval <= 0)
 		return retval;
 
 	retval = oprofile_set_ulong(&oprofile_backtrace_depth, val);

@@ -141,9 +141,10 @@ static ssize_t enable_write(struct file *file, char const __user *buf, size_t co
 		return -EINVAL;
 
 	retval = oprofilefs_ulong_from_user(&val, buf, count);
-	if (retval)
+	if (retval <= 0)
 		return retval;
 
+	retval = 0;
 	if (val)
 		retval = oprofile_start();
 	else

drivers/oprofile/oprofilefs.c +9 -2

index d0de6cc..2f0aa0f 100644
@@ -60,6 +60,13 @@ ssize_t oprofilefs_ulong_to_user(unsigned long val, char __user *buf, size_t cou
 }
 
 
+/*
+ * Note: If oprofilefs_ulong_from_user() returns 0, then *val remains
+ * unchanged and might be uninitialized. This follows write syscall
+ * implementation when count is zero: "If count is zero ... [and if]
+ * no errors are detected, 0 will be returned without causing any
+ * other effect." (man 2 write)
+ */
 int oprofilefs_ulong_from_user(unsigned long *val, char const __user *buf, size_t count)
 {
 	char tmpbuf[TMPBUFSIZE];

@@ -79,7 +86,7 @@ int oprofilefs_ulong_from_user(unsigned long *val, char const __user *buf, size_
 	raw_spin_lock_irqsave(&oprofilefs_lock, flags);
 	*val = simple_strtoul(tmpbuf, NULL, 0);
 	raw_spin_unlock_irqrestore(&oprofilefs_lock, flags);
-	return 0;
+	return count;
 }

@@ -99,7 +106,7 @@ static ssize_t ulong_write_file(struct file *file, char const __user *buf, size_
 		return -EINVAL;
 
 	retval = oprofilefs_ulong_from_user(&value, buf, count);
-	if (retval)
+	if (retval <= 0)
 		return retval;
 
 	retval = oprofile_set_ulong(file->private_data, value);

Robert Richter

author

2011-12-22 15:15 UTC #7 lore raw

PATCH 1/1 oprofile, arm/sh: Fix oprofile_arch_exit() linkage issue

In reply to Robert Richter

Quoted 1 line · Robert Richter

please pull this one fix for 3.2.

Ingo, yet another fix. Please apply to tip/perf/urgent.

Thanks,

-Robert

Quoted 1 line

From a5ea1e51bb66c6081d081cd88d787f248e07d6af Mon Sep 17 00:00:00 2001

From: Vladimir Zapolskiy <[email protected]>
Date: Thu, 22 Dec 2011 15:36:56 +0200
Subject: [PATCH] oprofile, arm/sh: Fix oprofile_arch_exit() linkage issue

This change fixes a linking problem, which happens if oprofile is selected to be compiled as built-in:

oprofile_arch_exit' referenced in section .init.text' of arch/arm/oprofile/built-in.o: defined in discarded section `.exit.text' of arch/arm/oprofile/built-in.o

The problem is appeared after commit 87121ca504, which introduced oprofile_arch_exit() calls from __init function. Note that the aforementioned commit has been backported to stable branches, and the problem is known to be reproduced at least with 3.0.13 and 3.1.5 kernels.

Cc [email protected] # 3.0+

Cc Will Deacon <[email protected]>

Signed-off-by Vladimir Zapolskiy <[email protected]>

Signed-off-by Robert Richter <[email protected]>

---
 arch/arm/oprofile/common.c |    2 +-
 arch/sh/oprofile/common.c  |    4 ++--
 2 files changed, 3 insertions(+), 3 deletions(-)

arch/arm/oprofile/common.c +1 -1

index c074e66..4e0a371 100644
@@ -116,7 +116,7 @@ int __init oprofile_arch_init(struct oprofile_operations *ops)
 	return oprofile_perf_init(ops);
 }
 
-void __exit oprofile_arch_exit(void)
+void oprofile_arch_exit(void)
 {
 	oprofile_perf_exit();
 }

arch/sh/oprofile/common.c +2 -2

index b4c2d2b..e4dd5d5 100644
@@ -49,7 +49,7 @@ int __init oprofile_arch_init(struct oprofile_operations *ops)
 	return oprofile_perf_init(ops);
 }
 
-void __exit oprofile_arch_exit(void)
+void oprofile_arch_exit(void)
 {
 	oprofile_perf_exit();
 	kfree(sh_pmu_op_name);

@@ -60,5 +60,5 @@ int __init oprofile_arch_init(struct oprofile_operations *ops)
 	ops->backtrace = sh_backtrace;
 	return -ENODEV;
 }
-void __exit oprofile_arch_exit(void) {}
+void oprofile_arch_exit(void) {}
 #endif /* CONFIG_HW_PERF_EVENTS */

Signature · 7 lines

1.7.8



-- 
Advanced Micro Devices, Inc.
Operating System Research Center

tip-bot for Vladimir Zapolskiy

2011-12-23 13:19 UTC #8 lore raw

[tip:perf/urgent] oprofile, arm/sh: Fix oprofile_arch_exit() linkage issue

Commit-ID:  55205c916e179e09773d98d290334d319f45ac6b
Gitweb:     http://git.kernel.org/tip/55205c916e179e09773d98d290334d319f45ac6b
Author:     Vladimir Zapolskiy <[email protected]>
AuthorDate: Thu, 22 Dec 2011 16:15:40 +0100
Committer:  Ingo Molnar <[email protected]>
CommitDate: Fri, 23 Dec 2011 11:58:34 +0100

oprofile, arm/sh: Fix oprofile_arch_exit() linkage issue

This change fixes a linking problem, which happens if oprofile is selected to be compiled as built-in:

oprofile_arch_exit' referenced in section .init.text' of arch/arm/oprofile/built-in.o: defined in discarded section `.exit.text' of arch/arm/oprofile/built-in.o

Signed-off-by Vladimir Zapolskiy <[email protected]>

Signed-off-by Robert Richter <[email protected]>

Cc Will Deacon <[email protected]>

Cc oprofile-list <[email protected]>

Cc <[email protected]>

Link http://lkml.kernel.org/r/[email protected] open here

Signed-off-by Ingo Molnar <[email protected]>

---
 arch/arm/oprofile/common.c |    2 +-
 arch/sh/oprofile/common.c  |    4 ++--
 2 files changed, 3 insertions(+), 3 deletions(-)

arch/arm/oprofile/common.c +1 -1

index c074e66..4e0a371 100644
@@ -116,7 +116,7 @@ int __init oprofile_arch_init(struct oprofile_operations *ops)
 	return oprofile_perf_init(ops);
 }
 
-void __exit oprofile_arch_exit(void)
+void oprofile_arch_exit(void)
 {
 	oprofile_perf_exit();
 }

arch/sh/oprofile/common.c +2 -2

index b4c2d2b..e4dd5d5 100644
@@ -49,7 +49,7 @@ int __init oprofile_arch_init(struct oprofile_operations *ops)
 	return oprofile_perf_init(ops);
 }
 
-void __exit oprofile_arch_exit(void)
+void oprofile_arch_exit(void)
 {
 	oprofile_perf_exit();
 	kfree(sh_pmu_op_name);

@@ -60,5 +60,5 @@ int __init oprofile_arch_init(struct oprofile_operations *ops)
 	ops->backtrace = sh_backtrace;
 	return -ENODEV;
 }
-void __exit oprofile_arch_exit(void) {}
+void oprofile_arch_exit(void) {}
 #endif /* CONFIG_HW_PERF_EVENTS */