[PATCH] time: Fix signedness bug in sysfs_get_uname() and its callers

[PATCH] time: Fix signedness bug in sysfs_get_uname() and its callers

[Patrick Palka]

sysfs_get_uname() is erroneously declared as returning size_t even
though it may return a negative value, specifically -EINVAL.  Its
callers then check whether its return value is less than zero and indeed
that is never the case for size_t.

This patch changes sysfs_get_uname() to return ssize_t and makes sure
its callers use ssize_t accordingly.

Signed-off-by: Patrick Palka <patrick@parcs.ath.cx>
---
 kernel/time/clockevents.c   | 2 +-
 kernel/time/clocksource.c   | 6 +++---
 kernel/time/tick-internal.h | 2 +-
 3 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/kernel/time/clockevents.c b/kernel/time/clockevents.c
index 38959c8..30554b9 100644
--- a/kernel/time/clockevents.c
+++ b/kernel/time/clockevents.c
@@ -584,7 +584,7 @@ static ssize_t sysfs_unbind_tick_dev(struct device *dev,
 				     const char *buf, size_t count)
 {
 	char name[CS_NAME_LEN];
-	size_t ret = sysfs_get_uname(buf, name, count);
+	ssize_t ret = sysfs_get_uname(buf, name, count);
 	struct clock_event_device *ce;
 
 	if (ret < 0)
diff --git a/kernel/time/clocksource.c b/kernel/time/clocksource.c
index 50a8736..b286f42 100644
--- a/kernel/time/clocksource.c
+++ b/kernel/time/clocksource.c
@@ -893,7 +893,7 @@ sysfs_show_current_clocksources(struct device *dev,
 	return count;
 }
 
-size_t sysfs_get_uname(const char *buf, char *dst, size_t cnt)
+ssize_t sysfs_get_uname(const char *buf, char *dst, size_t cnt)
 {
 	size_t ret = cnt;
 
@@ -924,7 +924,7 @@ static ssize_t sysfs_override_clocksource(struct device *dev,
 					  struct device_attribute *attr,
 					  const char *buf, size_t count)
 {
-	size_t ret;
+	ssize_t ret;
 
 	mutex_lock(&clocksource_mutex);
 
@@ -952,7 +952,7 @@ static ssize_t sysfs_unbind_clocksource(struct device *dev,
 {
 	struct clocksource *cs;
 	char name[CS_NAME_LEN];
-	size_t ret;
+	ssize_t ret;
 
 	ret = sysfs_get_uname(buf, name, count);
 	if (ret < 0)
diff --git a/kernel/time/tick-internal.h b/kernel/time/tick-internal.h
index bc906ca..18e71f7 100644
--- a/kernel/time/tick-internal.h
+++ b/kernel/time/tick-internal.h
@@ -31,7 +31,7 @@ extern void tick_install_replacement(struct clock_event_device *dev);
 
 extern void clockevents_shutdown(struct clock_event_device *dev);
 
-extern size_t sysfs_get_uname(const char *buf, char *dst, size_t cnt);
+extern ssize_t sysfs_get_uname(const char *buf, char *dst, size_t cnt);
 
 /*
  * NO_HZ / high resolution timer shared code
-- 
1.8.4.rc3

Re: [PATCH] time: Fix signedness bug in sysfs_get_uname() and its callers

[John Stultz]

On 10/11/2013 10:11 AM, Patrick Palka wrote:
> sysfs_get_uname() is erroneously declared as returning size_t even
> though it may return a negative value, specifically -EINVAL.  Its
> callers then check whether its return value is less than zero and indeed
> that is never the case for size_t.
>
> This patch changes sysfs_get_uname() to return ssize_t and makes sure
> its callers use ssize_t accordingly.

So a similar fix has already been queued in tip/timers/core, but this
seems more complete, so I've resolved the collisions with the earlier
fix and queued it for 3.13.

Would you please take a look at the resulting commit and double check I
didn't flub the conflict resolution?

https://git.linaro.org/gitweb?p=people/jstultz/linux.git;a=commitdiff;h=891292a767c2453af0e5be9465e95b06b4b29ebe;hp=b7bc50e45111e59419474154736f419a555158d9

thanks
-john

Re: [PATCH] time: Fix signedness bug in sysfs_get_uname() and its callers

[Patrick Palka]

On Fri, Oct 18, 2013 at 8:16 PM, John Stultz <john.stultz@linaro.org> wrote:
> On 10/11/2013 10:11 AM, Patrick Palka wrote:
>> sysfs_get_uname() is erroneously declared as returning size_t even
>> though it may return a negative value, specifically -EINVAL.  Its
>> callers then check whether its return value is less than zero and indeed
>> that is never the case for size_t.
>>
>> This patch changes sysfs_get_uname() to return ssize_t and makes sure
>> its callers use ssize_t accordingly.
>
> So a similar fix has already been queued in tip/timers/core, but this
> seems more complete, so I've resolved the collisions with the earlier
> fix and queued it for 3.13.
>
> Would you please take a look at the resulting commit and double check I
> didn't flub the conflict resolution?
>
> https://git.linaro.org/gitweb?p=people/jstultz/linux.git;a=commitdiff;h=891292a767c2453af0e5be9465e95b06b4b29ebe;hp=b7bc50e45111e59419474154736f419a555158d9

Looks good.