[Xenomai] [PATCH 1/1] posix skin: fix address passed to pse51_mutex_check_init in syscall
tbultel at free.fr
Sat Apr 30 11:41:33 CEST 2016
Le 29/04/2016 23:15, Gilles Chanteperdrix a écrit :
> On Fri, Apr 29, 2016 at 10:50:48PM +0200, Thierry Bultel wrote:
>> Le 29/04/2016 21:40, Gilles Chanteperdrix a écrit :
>>> On Thu, Apr 28, 2016 at 04:39:50PM +0200, Thierry Bultel wrote:
>>>> The address passed to pse51_mutex_check_init was the one of userland,
>>>> not the kernel one.
>>>> Signed-off-by: Thierry Bultel <tbultel at free.fr>
>>>> ksrc/skins/posix/syscall.c | 2 +-
>>>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>>> diff --git a/ksrc/skins/posix/syscall.c b/ksrc/skins/posix/syscall.c
>>>> index 87b95fa..31e2705 100644
>>>> --- a/ksrc/skins/posix/syscall.c
>>>> +++ b/ksrc/skins/posix/syscall.c
>>>> @@ -1202,7 +1202,7 @@ static int __pthread_mutex_check_init(struct pt_regs *regs)
>>>> } else
>>>> attr = NULL;
>>>> - return pse51_mutex_check_init(&umx->shadow_mutex, attr);
>>>> + return pse51_mutex_check_init(&mx.shadow_mutex, attr);
>>> Ah, what a coincidence, Jan reported this issue a long time ago:
>>> and I have merged a patch for it just a few days ago in my local
>>> branch (soon to be Xenomai 2.6.5).
>> As you probably read in my previous emails, I am currently achieving
>> to port ipipe to 4.5.
>> Whereas the existing version does not seem to make any problems
>> with kernel 3.18.12, it does a kernel crash with 4.5.
>> I did not investigate why this makes a difference.
>> With the patch, the 'leaks' non regression test successfully passes.
> Fine, but how is that relevant? The patch is already merged, and Jan
> probably explained the reason for the crash.
I was just wondering why this used to work with older kernels
and why it does not any longer with newer ones.
More information about the Xenomai