All of lore.kernel.org
 help / color / mirror / Atom feed
* [OpenRISC] [PATCH v2] time: Fix overflow itimer tests on 32-bit systems
@ 2021-08-06  9:42 Stafford Horne
       [not found] ` <0f577bc8-bef8-6c06-aaa9-57bf16d8443b@linaro.org>
  0 siblings, 1 reply; 6+ messages in thread
From: Stafford Horne @ 2021-08-06  9:42 UTC (permalink / raw)
  To: openrisc

On the port of OpenRISC I am working on and it appears the rv32 port
we have sets __TIMESIZE == 64 && __WORDSIZE == 32.  This causes the
size of time_t to be 8 bytes, but the tv_sec in the kernel is still 32-bit
causing truncation.

The truncations are unavoidable on these systems so skip the
testing/failures by guarding with __KERNEL_OLD_TIMEVAL_MATCHES_TIMEVAL64.

Also, futher in the tests and in other parts of code checking for time_t
overflow does not work on 32-bit systems when time_t is 64-bit.  As
suggested by Adhemerval, update the in_time_t_range function to assume
32-bits by using int32_t.

This also brings in the header for stdint.h so we can update other
usages of __int32_t to int32_t as suggested by Adhemerval.
---

Hello,

Sorry for the delay to get this out I have been busy on the hardware side of
openrisc the last month so I haven't been able to spend time on getting this
out.

The patch ends up doing a test fix and some lib code fixes, I can split it to
separate small patches.  But since as a whole it's small I feel leaving it
together is best.

-Stafford

 include/time.h    | 10 ++++++----
 time/tst-itimer.c |  4 ++--
 2 files changed, 8 insertions(+), 6 deletions(-)

diff --git a/include/time.h b/include/time.h
index 4372bfbd96..ba3c5116cf 100644
--- a/include/time.h
+++ b/include/time.h
@@ -11,6 +11,7 @@
 # include <sys/time.h>
 # include <time-clockid.h>
 # include <sys/time.h>
+# include <stdint.h>
 
 extern __typeof (strftime_l) __strftime_l;
 libc_hidden_proto (__strftime_l)
@@ -334,11 +335,12 @@ libc_hidden_proto (__time64)
    actual clock ID.  */
 #define CLOCK_IDFIELD_SIZE	3
 
-/* Check whether T fits in time_t.  */
+/* Check whether T fits in int32_t, assume all usages are for
+   sizeof(time_t) == 32.  */
 static inline bool
 in_time_t_range (__time64_t t)
 {
-  time_t s = t;
+  int32_t s = t;
   return s == t;
 }
 
@@ -445,8 +447,8 @@ timespec64_to_timeval64 (const struct __timespec64 ts64)
    and suseconds_t.  */
 struct __timeval32
 {
-  __int32_t tv_sec;         /* Seconds.  */
-  __int32_t tv_usec;        /* Microseconds.  */
+  int32_t tv_sec;         /* Seconds.  */
+  int32_t tv_usec;        /* Microseconds.  */
 };
 
 /* Conversion functions for converting to/from __timeval32  */
diff --git a/time/tst-itimer.c b/time/tst-itimer.c
index 929c2b74c7..bd7d7afe83 100644
--- a/time/tst-itimer.c
+++ b/time/tst-itimer.c
@@ -100,7 +100,7 @@ do_test (void)
 
       /* Linux does not provide 64 bit time_t support for getitimer and
 	 setitimer on architectures with 32 bit time_t support.  */
-      if (sizeof (__time_t) == 8)
+      if (__KERNEL_OLD_TIMEVAL_MATCHES_TIMEVAL64)
 	{
 	  TEST_COMPARE (setitimer (timers[i], &it, NULL), 0);
 	  TEST_COMPARE (setitimer (timers[i], &(struct itimerval) { 0 },
@@ -131,7 +131,7 @@ do_test (void)
       it.it_interval.tv_usec = 20;
       it.it_value.tv_sec = 30;
       it.it_value.tv_usec = 40;
-      if (sizeof (__time_t) == 8)
+      if (__KERNEL_OLD_TIMEVAL_MATCHES_TIMEVAL64)
 	{
 	  TEST_COMPARE (setitimer (timers[i], &it, NULL), 0);
 
-- 
2.31.1


^ permalink raw reply related	[flat|nested] 6+ messages in thread

* [OpenRISC] [PATCH v2] time: Fix overflow itimer tests on 32-bit systems
       [not found] ` <0f577bc8-bef8-6c06-aaa9-57bf16d8443b@linaro.org>
@ 2021-08-14 22:24   ` Stafford Horne
  2021-08-16 17:12     ` Joseph Myers
  0 siblings, 1 reply; 6+ messages in thread
From: Stafford Horne @ 2021-08-14 22:24 UTC (permalink / raw)
  To: openrisc

On Wed, Aug 11, 2021 at 05:30:40PM -0300, Adhemerval Zanella wrote:
> 
> 
> On 06/08/2021 06:42, Stafford Horne wrote:
> > On the port of OpenRISC I am working on and it appears the rv32 port
> > we have sets __TIMESIZE == 64 && __WORDSIZE == 32.  This causes the
> > size of time_t to be 8 bytes, but the tv_sec in the kernel is still 32-bit
> > causing truncation.
> > 
> > The truncations are unavoidable on these systems so skip the
> > testing/failures by guarding with __KERNEL_OLD_TIMEVAL_MATCHES_TIMEVAL64.
> > 
> > Also, futher in the tests and in other parts of code checking for time_t
> > overflow does not work on 32-bit systems when time_t is 64-bit.  As
> > suggested by Adhemerval, update the in_time_t_range function to assume
> > 32-bits by using int32_t.
> > 
> > This also brings in the header for stdint.h so we can update other
> > usages of __int32_t to int32_t as suggested by Adhemerval.
> > ---
> > 
> > Hello,
> > 
> > Sorry for the delay to get this out I have been busy on the hardware side of
> > openrisc the last month so I haven't been able to spend time on getting this
> > out.
> > 
> > The patch ends up doing a test fix and some lib code fixes, I can split it to
> > separate small patches.  But since as a whole it's small I feel leaving it
> > together is best.
> 
> LGTM, thanks.
> 
> Reviewed-by: Adhemerval Zanella  <adhemerval.zanella@linaro.org>

Thanks for the review.

> > 
> > -Stafford
> > 
> >  include/time.h    | 10 ++++++----
> >  time/tst-itimer.c |  4 ++--
> >  2 files changed, 8 insertions(+), 6 deletions(-)
> > 
> > diff --git a/include/time.h b/include/time.h
> > index 4372bfbd96..ba3c5116cf 100644
> > --- a/include/time.h
> > +++ b/include/time.h
> > @@ -11,6 +11,7 @@
> >  # include <sys/time.h>
> >  # include <time-clockid.h>
> >  # include <sys/time.h>
> > +# include <stdint.h>
> >  
> >  extern __typeof (strftime_l) __strftime_l;
> >  libc_hidden_proto (__strftime_l)
> > @@ -334,11 +335,12 @@ libc_hidden_proto (__time64)
> >     actual clock ID.  */
> >  #define CLOCK_IDFIELD_SIZE	3
> >  
> > -/* Check whether T fits in time_t.  */
> > +/* Check whether T fits in int32_t, assume all usages are for
> > +   sizeof(time_t) == 32.  */
> >  static inline bool
> >  in_time_t_range (__time64_t t)
> >  {
> > -  time_t s = t;
> > +  int32_t s = t;
> >    return s == t;
> >  }
> >  
> > @@ -445,8 +447,8 @@ timespec64_to_timeval64 (const struct __timespec64 ts64)
> >     and suseconds_t.  */
> >  struct __timeval32
> >  {
> > -  __int32_t tv_sec;         /* Seconds.  */
> > -  __int32_t tv_usec;        /* Microseconds.  */
> > +  int32_t tv_sec;         /* Seconds.  */
> > +  int32_t tv_usec;        /* Microseconds.  */
> >  };
> >  
> >  /* Conversion functions for converting to/from __timeval32  */
> > diff --git a/time/tst-itimer.c b/time/tst-itimer.c
> > index 929c2b74c7..bd7d7afe83 100644
> > --- a/time/tst-itimer.c
> > +++ b/time/tst-itimer.c
> > @@ -100,7 +100,7 @@ do_test (void)
> >  
> >        /* Linux does not provide 64 bit time_t support for getitimer and
> >  	 setitimer on architectures with 32 bit time_t support.  */
> > -      if (sizeof (__time_t) == 8)
> > +      if (__KERNEL_OLD_TIMEVAL_MATCHES_TIMEVAL64)
> >  	{
> >  	  TEST_COMPARE (setitimer (timers[i], &it, NULL), 0);
> >  	  TEST_COMPARE (setitimer (timers[i], &(struct itimerval) { 0 },
> > @@ -131,7 +131,7 @@ do_test (void)
> >        it.it_interval.tv_usec = 20;
> >        it.it_value.tv_sec = 30;
> >        it.it_value.tv_usec = 40;
> > -      if (sizeof (__time_t) == 8)
> > +      if (__KERNEL_OLD_TIMEVAL_MATCHES_TIMEVAL64)
> >  	{
> >  	  TEST_COMPARE (setitimer (timers[i], &it, NULL), 0);
> >  
> > 

^ permalink raw reply	[flat|nested] 6+ messages in thread

* [OpenRISC] [PATCH v2] time: Fix overflow itimer tests on 32-bit systems
  2021-08-14 22:24   ` Stafford Horne
@ 2021-08-16 17:12     ` Joseph Myers
  2021-08-16 21:54       ` Stafford Horne
  0 siblings, 1 reply; 6+ messages in thread
From: Joseph Myers @ 2021-08-16 17:12 UTC (permalink / raw)
  To: openrisc

I'm seeing a build failure in the glibc testsuite for i686-gnu:

tst-itimer.c: In function 'do_test':
tst-itimer.c:103:11: error: '__KERNEL_OLD_TIMEVAL_MATCHES_TIMEVAL64' undeclared (first use in this function)
  103 |       if (__KERNEL_OLD_TIMEVAL_MATCHES_TIMEVAL64)
      |           ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
tst-itimer.c:103:11: note: each undeclared identifier is reported only once for each function it appears in

https://sourceware.org/pipermail/libc-testresults/2021q3/008412.html

-- 
Joseph S. Myers
joseph at codesourcery.com

^ permalink raw reply	[flat|nested] 6+ messages in thread

* [OpenRISC] [PATCH v2] time: Fix overflow itimer tests on 32-bit systems
  2021-08-16 17:12     ` Joseph Myers
@ 2021-08-16 21:54       ` Stafford Horne
  2021-08-16 22:32         ` Joseph Myers
  0 siblings, 1 reply; 6+ messages in thread
From: Stafford Horne @ 2021-08-16 21:54 UTC (permalink / raw)
  To: openrisc

On Mon, Aug 16, 2021 at 05:12:04PM +0000, Joseph Myers wrote:
> I'm seeing a build failure in the glibc testsuite for i686-gnu:
> 
> tst-itimer.c: In function 'do_test':
> tst-itimer.c:103:11: error: '__KERNEL_OLD_TIMEVAL_MATCHES_TIMEVAL64' undeclared (first use in this function)
>   103 |       if (__KERNEL_OLD_TIMEVAL_MATCHES_TIMEVAL64)
>       |           ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> tst-itimer.c:103:11: note: each undeclared identifier is reported only once for each function it appears in
> 
> https://sourceware.org/pipermail/libc-testresults/2021q3/008412.html

Right sorry about that, so __KERNEL_OLD_TIMEVAL_MATCHES_TIMEVAL64 only exists in
linux.

So it looks like for non linux the new timer test changes break.

Should we provide __KERNEL_OLD_TIMEVAL_MATCHES_TIMEVAL64 for not linux builds,
or remove __KERNEL_OLD_TIMEVAL_MATCHES_TIMEVAL64 from the itimer test again?
The reason for using __KERNEL_OLD_TIMEVAL_MATCHES_TIMEVAL64 is to pick up the
timeval size which is different on each architecture.

Maybe the easiest is adding something like:

#ifndef __KERNEL_OLD_TIMEVAL_MATCHES_TIMEVAL64
#define __KERNEL_OLD_TIMEVAL_MATCHES_TIMEVAL64 (sizeof (__time_t) == 8)
#endif

To the top of tst-itimer.c.  Sorry I don't have time right now to send a patch
or test this.

-Stafford 

^ permalink raw reply	[flat|nested] 6+ messages in thread

* [OpenRISC] [PATCH v2] time: Fix overflow itimer tests on 32-bit systems
  2021-08-16 21:54       ` Stafford Horne
@ 2021-08-16 22:32         ` Joseph Myers
  2021-08-17 23:33           ` Stafford Horne
  0 siblings, 1 reply; 6+ messages in thread
From: Joseph Myers @ 2021-08-16 22:32 UTC (permalink / raw)
  To: openrisc

On Tue, 17 Aug 2021, Stafford Horne via Libc-alpha wrote:

> Should we provide __KERNEL_OLD_TIMEVAL_MATCHES_TIMEVAL64 for not linux builds,
> or remove __KERNEL_OLD_TIMEVAL_MATCHES_TIMEVAL64 from the itimer test again?
> The reason for using __KERNEL_OLD_TIMEVAL_MATCHES_TIMEVAL64 is to pick up the
> timeval size which is different on each architecture.

I'd suggest having a macro that doesn't refer to either "kernel" or "old 
timeval" (and that is defined for both Linux and Hurd).  As far as I 
understand, the logical concept that's relevant for this test isn't either 
one of those, it's more like "setitimer supports times that cannot be 
represented in 32 bits".

-- 
Joseph S. Myers
joseph at codesourcery.com

^ permalink raw reply	[flat|nested] 6+ messages in thread

* [OpenRISC] [PATCH v2] time: Fix overflow itimer tests on 32-bit systems
  2021-08-16 22:32         ` Joseph Myers
@ 2021-08-17 23:33           ` Stafford Horne
  0 siblings, 0 replies; 6+ messages in thread
From: Stafford Horne @ 2021-08-17 23:33 UTC (permalink / raw)
  To: openrisc

On Tue, Aug 17, 2021, 7:33 AM Joseph Myers <joseph@codesourcery.com> wrote:

> On Tue, 17 Aug 2021, Stafford Horne via Libc-alpha wrote:
>
> > Should we provide __KERNEL_OLD_TIMEVAL_MATCHES_TIMEVAL64 for not linux
> builds,
> > or remove __KERNEL_OLD_TIMEVAL_MATCHES_TIMEVAL64 from the itimer test
> again?
> > The reason for using __KERNEL_OLD_TIMEVAL_MATCHES_TIMEVAL64 is to pick
> up the
> > timeval size which is different on each architecture.
>
> I'd suggest having a macro that doesn't refer to either "kernel" or "old
> timeval" (and that is defined for both Linux and Hurd).  As far as I
> understand, the logical concept that's relevant for this test isn't either
> one of those, it's more like "setitimer supports times that cannot be
> represented in 32 bits".
>

Hello,

That makes sense, currently with the hurd build being broken how urgent is
this?  I worked on reproducing the build issue with build-many but didn't
get it working yet, probably about 80% there before I ran out of time.

I'll try to get it fixed in a few days as my top priority, but I only have
an hour or two a day to look at it. If we need to revert or add a temporary
patch please feel free.

-stafford

>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.librecores.org/pipermail/openrisc/attachments/20210818/ed52d64b/attachment.htm>

^ permalink raw reply	[flat|nested] 6+ messages in thread

end of thread, other threads:[~2021-08-17 23:33 UTC | newest]

Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-08-06  9:42 [OpenRISC] [PATCH v2] time: Fix overflow itimer tests on 32-bit systems Stafford Horne
     [not found] ` <0f577bc8-bef8-6c06-aaa9-57bf16d8443b@linaro.org>
2021-08-14 22:24   ` Stafford Horne
2021-08-16 17:12     ` Joseph Myers
2021-08-16 21:54       ` Stafford Horne
2021-08-16 22:32         ` Joseph Myers
2021-08-17 23:33           ` Stafford Horne

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.