Re: [PATCH] bpf: avoid warning for wrong pointer cast

[Alexei Starovoitov]

On 4/19/16 3:09 AM, Philip Li wrote:

On Tue, Apr 19, 2016 at 10:33:34AM +0800, Fengguang Wu wrote:

Fengguang, any idea why build-bot sometimes silent?


Sorry I went off for some time.. Philip, would you help have a check?

Hi Alexei, i have done some investigation for this. Fengguang, pls correct me 
if my understanding is wrong.

0day has caught the warning for f1ff54 commit around 2016-3-22, which considers 
the same warning in future is not new one, thus neglect it.

rli9@inn /kbuild-tests/build-error$ cat 
kernel-trace-bpf_trace.c:warning:cast-to-pointer-from-integer-of-different-size
kernel/trace/bpf_trace.c:345:13: warning: cast to pointer from integer 
of different size [-Wint-to-pointer-cast]

2016-03-22 17:12:38 xian 
bpf:master:321e70e79c94e5e6394a882d567baac949b74000 i386-randconfig-x009-201612 
f1ff543dbdd3eff53c8328cfb582f18e6c3d56ec

And for f1ff54's warning, 0day actually bisects to it, but then it checks the 
head of bpf/master at that time which is 321e70e, find the error is not 
existed, so
it ignores the interim warning to not send email report.

2016-03-22_17:15:49 O: R: /kbuild/src/consumer/kernel/trace/bpf_trace.c:345:13: 
warning: cast to pointer from integer of different size [-Wint-to-pointer-cast]
2016-03-22_17:15:49 O: grep -h -r . 
/tmp/kernel/i386-randconfig-x009-201612/gcc-5/f1ff543dbdd3eff53c8328cfb582f18e6c3d56ec/build-error
 # 0 errors
2016-03-22_17:15:49 O: grep -h -r . 
/tmp/kernel/i386-randconfig-x009-201612/gcc-5/f1ff543dbdd3eff53c8328cfb582f18e6c3d56ec/build-error/
 # 0 errors
2016-03-22_17:15:49 O: /c/kernel-tests/list-head 
bpf:master:321e70e79c94e5e6394a882d567baac949b74000 i386-randconfig-x009-201612
2016-03-22_17:16:03 O: .
2016-03-22_17:16:03 O: ..
2016-03-22_17:16:03 O: 
kernel-trace-bpf_trace.c:warning:cast-to-pointer-from-integer-of-different-size
2016-03-22_17:16:03 O: 2016-03-22 17:16:03 don't email interim warnings
2016-03-22_17:16:03 O: 2016-03-22 17:16:03 error no longer exist in head 
bpf/master 321e70e79c94e5e6394a882d567baac949b74000, check
/tmp/kernel/i386-randconfig-x009-201612/gcc-5/321e70e79c94e5e6394a882d567baac949b74000,
 errno 14
2016-03-22_17:16:03 O: 2016-03-22 17:16:03 Don't email interim warnings!


what really happens to 321e70e is it has build error (i rerun it to confirm), 
which is reported. So this is
a possible change to be done in 0day to report the first bad commit when HEAD 
is build error.


excellent sleuthing! Sounds like the adjustment to 0day is forthcoming.
Thank you for great tool!

Re: [PATCH] bpf: avoid warning for wrong pointer cast

[Philip Li]

On Tue, Apr 19, 2016 at 10:33:34AM +0800, Fengguang Wu wrote:
> Hi Alexei,
> 
> On Sat, Apr 16, 2016 at 05:47:42PM -0700, Alexei Starovoitov wrote:
> > On Sat, Apr 16, 2016 at 10:29:33PM +0200, Arnd Bergmann wrote:
> > > Two new functions in bpf contain a cast from a 'u64' to a
> > > pointer. This works on 64-bit architectures but causes a warning
> > > on all 32-bit architectures:
> > > 
> > > kernel/trace/bpf_trace.c: In function 'bpf_perf_event_output_tp':
> > > kernel/trace/bpf_trace.c:350:13: error: cast to pointer from integer of 
> > > different size [-Werror=int-to-pointer-cast]
> > >   u64 ctx = *(long *)r1;
> > > 
> > > This changes the cast to first convert the u64 argument into a uintptr_t,
> > > which is guaranteed to be the same size as a pointer.
> > > 
> > > Signed-off-by: Arnd Bergmann 
> > > Fixes: 9940d67c93b5 ("bpf: support bpf_get_stackid() and 
> > > bpf_perf_event_output() in tracepoint programs")
> > 
> > Thanks.
> > Acked-by: Alexei Starovoitov 
> > 
> > I guess I started to rely on 0-day build-bot too much.
> > This patch has been in my tree for 2+ weeks and then in net-next and
> > I didn't receive a single email from build-bot about this warning,
> > though I do receive them for my other work-in-progress stuff. Odd.
> > Fengguang, any idea why build-bot sometimes silent?
> 
> Sorry I went off for some time.. Philip, would you help have a check?
Hi Alexei, i have done some investigation for this. Fengguang, pls correct me 
if my understanding is wrong.

0day has caught the warning for f1ff54 commit around 2016-3-22, which considers 
the same warning in future is not new one, thus neglect it.

rli9@inn /kbuild-tests/build-error$ cat 
kernel-trace-bpf_trace.c:warning:cast-to-pointer-from-integer-of-different-size
kernel/trace/bpf_trace.c:345:13: warning: cast to pointer from integer 
of different size [-Wint-to-pointer-cast]

2016-03-22 17:12:38 xian 
bpf:master:321e70e79c94e5e6394a882d567baac949b74000 i386-randconfig-x009-201612 
f1ff543dbdd3eff53c8328cfb582f18e6c3d56ec

And for f1ff54's warning, 0day actually bisects to it, but then it checks the 
head of bpf/master at that time which is 321e70e, find the error is not 
existed, so
it ignores the interim warning to not send email report.

2016-03-22_17:15:49 O: R: /kbuild/src/consumer/kernel/trace/bpf_trace.c:345:13: 
warning: cast to pointer from integer of different size [-Wint-to-pointer-cast]
2016-03-22_17:15:49 O: grep -h -r . 
/tmp/kernel/i386-randconfig-x009-201612/gcc-5/f1ff543dbdd3eff53c8328cfb582f18e6c3d56ec/build-error
 # 0 errors 
2016-03-22_17:15:49 O: grep -h -r . 
/tmp/kernel/i386-randconfig-x009-201612/gcc-5/f1ff543dbdd3eff53c8328cfb582f18e6c3d56ec/build-error/
 # 0 errors 
2016-03-22_17:15:49 O: /c/kernel-tests/list-head 
bpf:master:321e70e79c94e5e6394a882d567baac949b74000 i386-randconfig-x009-201612 
2016-03-22_17:16:03 O: .
2016-03-22_17:16:03 O: ..
2016-03-22_17:16:03 O: 
kernel-trace-bpf_trace.c:warning:cast-to-pointer-from-integer-of-different-size
2016-03-22_17:16:03 O: 2016-03-22 17:16:03 don't email interim warnings
2016-03-22_17:16:03 O: 2016-03-22 17:16:03 error no longer exist in head 
bpf/master 321e70e79c94e5e6394a882d567baac949b74000, check 
/tmp/kernel/i386-randconfig-x009-201612/gcc-5/321e70e79c94e5e6394a882d567baac949b74000,
 errno 14
2016-03-22_17:16:03 O: 2016-03-22 17:16:03 Don't email interim warnings!


what really happens to 321e70e is it has build error (i rerun it to confirm), 
which is reported at 
https://lists.01.org/pipermail/kbuild-all/2016-March/018625.html. So this is
a possible change to be done in 0day to report the first bad commit when HEAD 
is build error.

2016-04-19_15:00:39 O: 2016-04-19 15:00  Running 
/kbuild-tests/build-queue/lkp-nex06-consumer/i386-randconfig-x009-201612-gcc-5-321e70e79c94e5e6394a882d567baac949b74000
2016-04-19_15:00:39 O: create new KBUILD_OUTPUT dir 
/kbuild/obj/consumer/i386-randconfig-x009-201612
2016-04-19_15:00:39 O: git checkout -B build-queue 
321e70e79c94e5e6394a882d567baac949b74000
2016-04-19_15:07:08 E: 330 real  4732 user  629 sys  1620.50% cpu   
i386-randconfig-x009-201612
2016-04-19_15:07:13 O: status: FAIL: build error
2016-04-19_15:07:13 O: ERROR: "tcp_sendpage" 
[drivers/staging/lustre/lnet/klnds/socklnd/ksocklnd.ko] undefined!

> 
> Thanks,
> Fengguang

Re: [PATCH] bpf: avoid warning for wrong pointer cast

[Fengguang Wu]

Hi Alexei,

On Sat, Apr 16, 2016 at 05:47:42PM -0700, Alexei Starovoitov wrote:
> On Sat, Apr 16, 2016 at 10:29:33PM +0200, Arnd Bergmann wrote:
> > Two new functions in bpf contain a cast from a 'u64' to a
> > pointer. This works on 64-bit architectures but causes a warning
> > on all 32-bit architectures:
> > 
> > kernel/trace/bpf_trace.c: In function 'bpf_perf_event_output_tp':
> > kernel/trace/bpf_trace.c:350:13: error: cast to pointer from integer of 
> > different size [-Werror=int-to-pointer-cast]
> >   u64 ctx = *(long *)r1;
> > 
> > This changes the cast to first convert the u64 argument into a uintptr_t,
> > which is guaranteed to be the same size as a pointer.
> > 
> > Signed-off-by: Arnd Bergmann 
> > Fixes: 9940d67c93b5 ("bpf: support bpf_get_stackid() and 
> > bpf_perf_event_output() in tracepoint programs")
> 
> Thanks.
> Acked-by: Alexei Starovoitov 
> 
> I guess I started to rely on 0-day build-bot too much.
> This patch has been in my tree for 2+ weeks and then in net-next and
> I didn't receive a single email from build-bot about this warning,
> though I do receive them for my other work-in-progress stuff. Odd.
> Fengguang, any idea why build-bot sometimes silent?

Sorry I went off for some time.. Philip, would you help have a check?

Thanks,
Fengguang

Re: [PATCH] bpf: avoid warning for wrong pointer cast

[David Miller]

From: Arnd Bergmann 
Date: Sat, 16 Apr 2016 22:29:33 +0200

> Two new functions in bpf contain a cast from a 'u64' to a
> pointer. This works on 64-bit architectures but causes a warning
> on all 32-bit architectures:
> 
> kernel/trace/bpf_trace.c: In function 'bpf_perf_event_output_tp':
> kernel/trace/bpf_trace.c:350:13: error: cast to pointer from integer of 
> different size [-Werror=int-to-pointer-cast]
>   u64 ctx = *(long *)r1;
> 
> This changes the cast to first convert the u64 argument into a uintptr_t,
> which is guaranteed to be the same size as a pointer.
> 
> Signed-off-by: Arnd Bergmann 
> Fixes: 9940d67c93b5 ("bpf: support bpf_get_stackid() and 
> bpf_perf_event_output() in tracepoint programs")

Applied.

Re: [PATCH] bpf: avoid warning for wrong pointer cast

[Alexei Starovoitov]

On Sat, Apr 16, 2016 at 10:29:33PM +0200, Arnd Bergmann wrote:
> Two new functions in bpf contain a cast from a 'u64' to a
> pointer. This works on 64-bit architectures but causes a warning
> on all 32-bit architectures:
> 
> kernel/trace/bpf_trace.c: In function 'bpf_perf_event_output_tp':
> kernel/trace/bpf_trace.c:350:13: error: cast to pointer from integer of 
> different size [-Werror=int-to-pointer-cast]
>   u64 ctx = *(long *)r1;
> 
> This changes the cast to first convert the u64 argument into a uintptr_t,
> which is guaranteed to be the same size as a pointer.
> 
> Signed-off-by: Arnd Bergmann 
> Fixes: 9940d67c93b5 ("bpf: support bpf_get_stackid() and 
> bpf_perf_event_output() in tracepoint programs")

Thanks.
Acked-by: Alexei Starovoitov 

I guess I started to rely on 0-day build-bot too much.
This patch has been in my tree for 2+ weeks and then in net-next and
I didn't receive a single email from build-bot about this warning,
though I do receive them for my other work-in-progress stuff. Odd.
Fengguang, any idea why build-bot sometimes silent?

[PATCH] bpf: avoid warning for wrong pointer cast

[Arnd Bergmann]

Two new functions in bpf contain a cast from a 'u64' to a
pointer. This works on 64-bit architectures but causes a warning
on all 32-bit architectures:

kernel/trace/bpf_trace.c: In function 'bpf_perf_event_output_tp':
kernel/trace/bpf_trace.c:350:13: error: cast to pointer from integer of 
different size [-Werror=int-to-pointer-cast]
  u64 ctx = *(long *)r1;

This changes the cast to first convert the u64 argument into a uintptr_t,
which is guaranteed to be the same size as a pointer.

Signed-off-by: Arnd Bergmann 
Fixes: 9940d67c93b5 ("bpf: support bpf_get_stackid() and 
bpf_perf_event_output() in tracepoint programs")
---
 kernel/trace/bpf_trace.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/kernel/trace/bpf_trace.c b/kernel/trace/bpf_trace.c
index 413ec5614180..c082313a523a 100644
--- a/kernel/trace/bpf_trace.c
+++ b/kernel/trace/bpf_trace.c
@@ -347,7 +347,7 @@ static u64 bpf_perf_event_output_tp(u64 r1, u64 r2, u64 
index, u64 r4, u64 size)
 * from bpf program and contain a pointer to 'struct pt_regs'. Fetch it
 * from there and call the same bpf_perf_event_output() helper
 */
-   u64 ctx = *(long *)r1;
+   u64 ctx = *(long *)(uintptr_t)r1;
 
return bpf_perf_event_output(ctx, r2, index, r4, size);
 }
@@ -365,7 +365,7 @@ static const struct bpf_func_proto 
bpf_perf_event_output_proto_tp = {
 
 static u64 bpf_get_stackid_tp(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5)
 {
-   u64 ctx = *(long *)r1;
+   u64 ctx = *(long *)(uintptr_t)r1;
 
return bpf_get_stackid(ctx, r2, r3, r4, r5);
 }
-- 
2.7.0