[PATCH 1/2] math: Introduce div64_long

[PATCH 1/2] math: Introduce div64_long

[Sasha Levin]

Add a div64_long macro which is used to devide a 64bit number by a long (which
can be 4 bytes on 32bit systems and 8 bytes on 64bit systems).

Suggested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Sasha Levin <levinsasha928@gmail.com>
---
 include/linux/math64.h |    4 ++++
 1 files changed, 4 insertions(+), 0 deletions(-)

diff --git a/include/linux/math64.h b/include/linux/math64.h
index 23fcdfc..b8ba855 100644
--- a/include/linux/math64.h
+++ b/include/linux/math64.h
@@ -6,6 +6,8 @@
 
 #if BITS_PER_LONG == 64
 
+#define div64_long(x,y) div64_s64((x),(y))
+
 /**
  * div_u64_rem - unsigned 64bit divide with 32bit divisor with remainder
  *
@@ -45,6 +47,8 @@ static inline s64 div64_s64(s64 dividend, s64 divisor)
 
 #elif BITS_PER_LONG == 32
 
+#define div64_long(x,y) div_s64((x),(y))
+
 #ifndef div_u64_rem
 static inline u64 div_u64_rem(u64 dividend, u32 divisor, u32 *remainder)
 {
-- 
1.7.8.4

[PATCH 2/2] ntp: Fix integer overflow when setting time

[Sasha Levin]

'long secs' was being passed as divisor to div_s64, which accepts a 32bit
divisor. On 64bit machines that value would be trimmed back from 8 bytes
back to 4, allowing a divide by zero when the number is bigger than
(1 << 32) - 1 and all 32 lower bits are 0.

Signed-off-by: Sasha Levin <levinsasha928@gmail.com>
---
 kernel/time/ntp.c |    6 +++++-
 1 files changed, 5 insertions(+), 1 deletions(-)

diff --git a/kernel/time/ntp.c b/kernel/time/ntp.c
index 17fb1b9..c83c228 100644
--- a/kernel/time/ntp.c
+++ b/kernel/time/ntp.c
@@ -289,7 +289,11 @@ static inline s64 ntp_update_offset_fll(s64 offset64, long secs)
 
 	time_status |= STA_MODE;
 
-	return div_s64(offset64 << (NTP_SCALE_SHIFT - SHIFT_FLL), secs);
+	/* 
+	 * secs is 8 bytes on 64bit systems, which means that it can
+	 * wrap around when dividing.
+	 */
+	return div64_long(offset64 << (NTP_SCALE_SHIFT - SHIFT_FLL), secs);
 }
 
 static void ntp_update_offset(long offset)
-- 
1.7.8.4

[tip:timers/core] ntp: Fix integer overflow when setting time

[tip-bot for Sasha Levin]

Commit-ID:  a078c6d0e6288fad6d83fb6d5edd91ddb7b6ab33
Gitweb:     http://git.kernel.org/tip/a078c6d0e6288fad6d83fb6d5edd91ddb7b6ab33
Author:     Sasha Levin <levinsasha928@gmail.com>
AuthorDate: Thu, 15 Mar 2012 12:36:14 -0400
Committer:  Thomas Gleixner <tglx@linutronix.de>
CommitDate: Thu, 15 Mar 2012 21:41:34 +0100

ntp: Fix integer overflow when setting time

'long secs' is passed as divisor to div_s64, which accepts a 32bit
divisor. On 64bit machines that value is trimmed back from 8 bytes
back to 4, causing a divide by zero when the number is bigger than
(1 << 32) - 1 and all 32 lower bits are 0.

Use div64_long() instead.

Signed-off-by: Sasha Levin <levinsasha928@gmail.com>
Cc: johnstul@us.ibm.com
Link: http://lkml.kernel.org/r/1331829374-31543-2-git-send-email-levinsasha928@gmail.com
Cc: stable@vger.kernel.org
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
---
 kernel/time/ntp.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/kernel/time/ntp.c b/kernel/time/ntp.c
index 17fb1b9..6e039b1 100644
--- a/kernel/time/ntp.c
+++ b/kernel/time/ntp.c
@@ -289,7 +289,7 @@ static inline s64 ntp_update_offset_fll(s64 offset64, long secs)
 
 	time_status |= STA_MODE;
 
-	return div_s64(offset64 << (NTP_SCALE_SHIFT - SHIFT_FLL), secs);
+	return div64_long(offset64 << (NTP_SCALE_SHIFT - SHIFT_FLL), secs);
 }
 
 static void ntp_update_offset(long offset)

Re: [tip:timers/core] ntp: Fix integer overflow when setting time

[H. Peter Anvin]

On 03/15/2012 02:53 PM, tip-bot for Sasha Levin wrote:
> Commit-ID:  a078c6d0e6288fad6d83fb6d5edd91ddb7b6ab33
> Gitweb:     http://git.kernel.org/tip/a078c6d0e6288fad6d83fb6d5edd91ddb7b6ab33
> Author:     Sasha Levin <levinsasha928@gmail.com>
> AuthorDate: Thu, 15 Mar 2012 12:36:14 -0400
> Committer:  Thomas Gleixner <tglx@linutronix.de>
> CommitDate: Thu, 15 Mar 2012 21:41:34 +0100
> 
> ntp: Fix integer overflow when setting time
> 
> 'long secs' is passed as divisor to div_s64, which accepts a 32bit
> divisor. On 64bit machines that value is trimmed back from 8 bytes
> back to 4, causing a divide by zero when the number is bigger than
> (1 << 32) - 1 and all 32 lower bits are 0.
> 
> Use div64_long() instead.
> 

Perhaps we should use the same kind of multiply-and-shift tricks we're
doing for jiffies conversion?  If nothing else it ought to perform better.

	-hpa

-- 
H. Peter Anvin, Intel Open Source Technology Center
I work for Intel.  I don't speak on their behalf.

Re: [tip:timers/core] ntp: Fix integer overflow when setting time

[Thomas Gleixner]

On Thu, 15 Mar 2012, H. Peter Anvin wrote:

> On 03/15/2012 02:53 PM, tip-bot for Sasha Levin wrote:
> > Commit-ID:  a078c6d0e6288fad6d83fb6d5edd91ddb7b6ab33
> > Gitweb:     http://git.kernel.org/tip/a078c6d0e6288fad6d83fb6d5edd91ddb7b6ab33
> > Author:     Sasha Levin <levinsasha928@gmail.com>
> > AuthorDate: Thu, 15 Mar 2012 12:36:14 -0400
> > Committer:  Thomas Gleixner <tglx@linutronix.de>
> > CommitDate: Thu, 15 Mar 2012 21:41:34 +0100
> > 
> > ntp: Fix integer overflow when setting time
> > 
> > 'long secs' is passed as divisor to div_s64, which accepts a 32bit
> > divisor. On 64bit machines that value is trimmed back from 8 bytes
> > back to 4, causing a divide by zero when the number is bigger than
> > (1 << 32) - 1 and all 32 lower bits are 0.
> > 
> > Use div64_long() instead.
> > 
> 
> Perhaps we should use the same kind of multiply-and-shift tricks we're
> doing for jiffies conversion?  If nothing else it ought to perform better.

It's not a hotpath where performance matters, but divide by zero does :)

Re: [tip:timers/core] ntp: Fix integer overflow when setting time

[H. Peter Anvin]

On 03/16/2012 02:48 AM, Thomas Gleixner wrote:
>>
>> Perhaps we should use the same kind of multiply-and-shift tricks we're
>> doing for jiffies conversion?  If nothing else it ought to perform better.
> 
> It's not a hotpath where performance matters, but divide by zero does :)

Doing the inverse multiply wouldn't have a divide by zero problem.  What
perhaps is more important (and the reason we went to shift-multiply) is
that in addition to being faster, it also avoids unnnecessary overflows.
 I would really like to figure out how to do proper shift-multiply for
the 64-bit jiffy conversions too.

	-hpa

-- 
H. Peter Anvin, Intel Open Source Technology Center
I work for Intel.  I don't speak on their behalf.

[tip:timers/core] math: Introduce div64_long

[tip-bot for Sasha Levin]

Commit-ID:  f910381a55cdaa097030291f272f6e6e4380c39a
Gitweb:     http://git.kernel.org/tip/f910381a55cdaa097030291f272f6e6e4380c39a
Author:     Sasha Levin <levinsasha928@gmail.com>
AuthorDate: Thu, 15 Mar 2012 12:36:13 -0400
Committer:  Thomas Gleixner <tglx@linutronix.de>
CommitDate: Thu, 15 Mar 2012 21:41:34 +0100

math: Introduce div64_long

Add a div64_long macro which is used to devide a 64bit number by a long (which
can be 4 bytes on 32bit systems and 8 bytes on 64bit systems).

Suggested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Sasha Levin <levinsasha928@gmail.com>
Cc: johnstul@us.ibm.com
Link: http://lkml.kernel.org/r/1331829374-31543-1-git-send-email-levinsasha928@gmail.com
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
---
 include/linux/math64.h |    4 ++++
 1 files changed, 4 insertions(+), 0 deletions(-)

diff --git a/include/linux/math64.h b/include/linux/math64.h
index 23fcdfc..b8ba855 100644
--- a/include/linux/math64.h
+++ b/include/linux/math64.h
@@ -6,6 +6,8 @@
 
 #if BITS_PER_LONG == 64
 
+#define div64_long(x,y) div64_s64((x),(y))
+
 /**
  * div_u64_rem - unsigned 64bit divide with 32bit divisor with remainder
  *
@@ -45,6 +47,8 @@ static inline s64 div64_s64(s64 dividend, s64 divisor)
 
 #elif BITS_PER_LONG == 32
 
+#define div64_long(x,y) div_s64((x),(y))
+
 #ifndef div_u64_rem
 static inline u64 div_u64_rem(u64 dividend, u32 divisor, u32 *remainder)
 {