Re: [LTP] [PATCH] syscalls/sysfs: replace the TWARN to TCONF

[Li Wang <liwang@redhat.com>]

> 
> ----- Original Message -----
> > From: "Li Wang" <liwang@redhat.com>
> > To: "Jan Stancek" <jstancek@redhat.com>
> > Cc: ltp-list@lists.sourceforge.net
> > Sent: Monday, 27 July, 2015 7:54:26 AM
> > Subject: Re: [LTP] [PATCH] syscalls/sysfs: replace the TWARN to TCONF
> > 
> > Hi Jan,
> > 
> > ----- Original Message -----
> > > 
> > > 
> > > 
> > > 
> > > ----- Original Message -----
> > > > From: "Li Wang" <liwang@redhat.com>
> > > > To: ltp-list@lists.sourceforge.net
> > > > Sent: Friday, 24 July, 2015 9:22:00 AM
> > > > Subject: [LTP] [PATCH] syscalls/sysfs: replace the TWARN to TCONF
> > > > 
> > > > I run some of these cases on aarch64 today and get a series waring,
> > > > that's
> > > > because
> > > > sysfs syscall is not implemented on aarch64. So, maybe TCONF is better
> > > > than
> > > > TWARN
> > > > in the testcase.
> > > 
> > > Are you sure it's not supported? Or is it just disabled in your kernel
> > > config?
> > > I'm looking at Kconfig and nothing suggests you can't use it on aarch64:
> > 
> > Hmm, to be precise, it has supported sysfs_syscall but not define
> > '__NR_sysfs' on aarch64.
> 
> Hi,
> 
> Same question:
>   http://linux-arm-kernel.infradead.narkive.com/n9PCFfpR/arm64-dose-arm64-support-the-sysfs-system-call
> 
> So it looks like you can enable CONFIG_SYSFS_SYSCALL, but still you won't
> be able to use it (for 64bit on 64bit), since it doesn't appear to be in
> syscall table.

yes.

> 
> > 
> > My system config:
> > # grep -i sysfs_syscall /boot/config-4.1.0-0.12.el7.aarch64
> > CONFIG_SYSFS_SYSCALL=y
> > # grep -i EXPERT /boot/config-4.1.0-0.12.el7.aarch64
> > CONFIG_EXPERT is not set
> > 
> > (1)---->
> > If I change the code as:
> > 
> > diff --git a/testcases/kernel/syscalls/sysfs/sysfs01.c
> > b/testcases/kernel/syscalls/sysfs/sysfs01.c
> > index 11ebe43..770c9c9 100644
> > --- a/testcases/kernel/syscalls/sysfs/sysfs01.c
> > +++ b/testcases/kernel/syscalls/sysfs/sysfs01.c
> > @@ -67,6 +67,7 @@
> >   ******************************************************************************/
> >  
> >  #include "test.h"
> > +#include "linux_syscall_numbers.h"
> >  #include <errno.h>
> >  #include <unistd.h>
> >  #include <syscall.h>
> > @@ -85,14 +86,12 @@ int main(int ac, char **av)
> >  
> >         setup();
> >  
> > -#ifdef __NR_sysfs
> > -
> >         for (lc = 0; TEST_LOOPING(lc); lc++) {
> >  
> >                 tst_count = 0;
> >  
> >                 /* option 1, buf holds fs name */
> > -               TEST(syscall(__NR_sysfs, 1, "proc"));
> > +               TEST(ltp_syscall(__NR_sysfs, 1, "proc"));
> >  
> >                 /* check return code */
> >                 if (TEST_RETURN == -1) {
> > @@ -102,10 +101,6 @@ int main(int ac, char **av)
> >                         tst_resm(TPASS, "sysfs(2) Passed for " "option 1");
> >                 }
> >         }                       /*End of TEST_LOOPING */
> > -#else
> > -       tst_resm(TWARN,
> > -                "This test can only run on kernels that support the sysfs
> > system call");
> > -#endif
> >  
> >         /*Clean up and exit */
> >         cleanup();
> > 
> > 
> > It's show:
> > 
> > # ./sysfs01
> > sysfs01     1  TCONF  :  sysfs01.c:94: syscall __NR_sysfs not supported on
> > your arch
> > sysfs01     2  TCONF  :  sysfs01.c:94: Remaining cases not appropriate for
> > configuration
> 
> I think this is correct way to go.

Ok, I will post a patch V2.

> 
> > 
> > 
> > 
> > (2)---->
> > then I add the macro to "linux_syscall_numbers.h" by manual:
> 
> linux_syscall_numbers.h is generated from *.in files in the same directory.

ok, got it.

> 
> > 
> > # grep -e __aarch64 -e __NR_sysfs  linux_syscall_numbers.h -A 2 -B 2
> > 
> > #ifdef __aarch64__
> > ...
> > # ifndef __NR_sysfs
> > #  define __NR_sysfs 135
> > # endif
> 
> This appears to be taken from unistd32.h.
> 
> > 
> > 
> > It's failed as:
> > 
> > # ./sysfs01
> > sysfs01     1  TFAIL  :  sysfs01.c:99: sysfs(2) Failed for option 1 and set
> > errno to 22
> > 
> 
> Yeah, I think that's because the number you picked is used for
> __NR_rt_sigprocmask:

yes. 

Thanks,
Li Wang


> 
> $ grep -e __aarch64 -e __NR_rt_sigproc linux_syscall_numbers.h
> #ifdef __aarch64__
> # ifndef __NR_rt_sigprocmask
> #  define __NR_rt_sigprocmask 135
> 
> Regards,
> Jan
> 

------------------------------------------------------------------------------
_______________________________________________
Ltp-list mailing list
Ltp-list@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/ltp-list