added several syscalls for nacl mounts

added several syscalls for nacl mounts


BUG=
TEST=

Committed: https://git.chromium.org/gitweb?p=native_client/nacl-glibc.git;a=commit;h=87666a1

[Evgeniy Stepanov, 2012-03-30 11:04:55]

https://chromiumcodereview.appspot.com/9965011/diff/1/sysdeps/nacl/sysdep.h
File sysdeps/nacl/sysdep.h (right):

https://chromiumcodereview.appspot.com/9965011/diff/1/sysdeps/nacl/sysdep.h#n...
sysdeps/nacl/sysdep.h:593: // __nacl_irt_getcwd (buf, size); // TODO
TODO?

[vissi, 2012-03-30 11:28:12]

this patch is to expand nacl-mounts capabilities

https://chromiumcodereview.appspot.com/9965011/diff/1/sysdeps/nacl/sysdep.h
File sysdeps/nacl/sysdep.h (right):

https://chromiumcodereview.appspot.com/9965011/diff/1/sysdeps/nacl/sysdep.h#n...
sysdeps/nacl/sysdep.h:593: // __nacl_irt_getcwd (buf, size); // TODO
On 2012/03/30 11:04:55, Evgeniy Stepanov wrote:
> TODO?

Done in patch set 2.

[Evgeniy Stepanov, 2012-03-30 11:32:00]

https://chromiumcodereview.appspot.com/9965011/diff/10/sysdeps/nacl/sysdep.h
File sysdeps/nacl/sysdep.h (right):

https://chromiumcodereview.appspot.com/9965011/diff/10/sysdeps/nacl/sysdep.h#...
sysdeps/nacl/sysdep.h:591: *err = __nacl_irt_getcwd (buf, size);
This does not look right. Return value of __nacl_irt_getcwd is char* in
irt_syscalls.h

[vissi, 2012-03-30 11:43:27]

On 2012/03/30 11:32:00, Evgeniy Stepanov wrote:
> https://chromiumcodereview.appspot.com/9965011/diff/10/sysdeps/nacl/sysdep.h
> File sysdeps/nacl/sysdep.h (right):
> 
>
https://chromiumcodereview.appspot.com/9965011/diff/10/sysdeps/nacl/sysdep.h#...
> sysdeps/nacl/sysdep.h:591: *err = __nacl_irt_getcwd (buf, size);
> This does not look right. Return value of __nacl_irt_getcwd is char* in
> irt_syscalls.h

Now fixed

[vissi, 2012-03-30 12:04:10]

On 2012/03/30 11:43:27, vissi wrote:
> On 2012/03/30 11:32:00, Evgeniy Stepanov wrote:
> > https://chromiumcodereview.appspot.com/9965011/diff/10/sysdeps/nacl/sysdep.h
> > File sysdeps/nacl/sysdep.h (right):
> > 
> >
>
https://chromiumcodereview.appspot.com/9965011/diff/10/sysdeps/nacl/sysdep.h#...
> > sysdeps/nacl/sysdep.h:591: *err = __nacl_irt_getcwd (buf, size);
> > This does not look right. Return value of __nacl_irt_getcwd is char* in
> > irt_syscalls.h
> 
> Now fixed

If I need to stick to kernel getcwd, here it is, it returns error
http://lxr.free-electrons.com/source/fs/dcache.c?v=2.6.32#L2084
man getcwd says it returns char* pointer (not length, as the comment for
INTERNAL_SYSCALL_getcwd_2 says. So what convention to follow?

[vissi, 2012-03-30 12:32:52]

On 2012/03/30 12:04:10, vissi wrote:
> On 2012/03/30 11:43:27, vissi wrote:
> > On 2012/03/30 11:32:00, Evgeniy Stepanov wrote:
> > >
https://chromiumcodereview.appspot.com/9965011/diff/10/sysdeps/nacl/sysdep.h
> > > File sysdeps/nacl/sysdep.h (right):
> > > 
> > >
> >
>
https://chromiumcodereview.appspot.com/9965011/diff/10/sysdeps/nacl/sysdep.h#...
> > > sysdeps/nacl/sysdep.h:591: *err = __nacl_irt_getcwd (buf, size);
> > > This does not look right. Return value of __nacl_irt_getcwd is char* in
> > > irt_syscalls.h
> > 
> > Now fixed
> 
> If I need to stick to kernel getcwd, here it is, it returns error
> http://lxr.free-electrons.com/source/fs/dcache.c?v=2.6.32#L2084
> man getcwd says it returns char* pointer (not length, as the comment for
> INTERNAL_SYSCALL_getcwd_2 says. So what convention to follow?

Updated to match other syscalls.

[pasko-google - do not use, 2012-03-30 12:36:29]

https://chromiumcodereview.appspot.com/9965011/diff/5001/elf/Versions
File elf/Versions (right):

https://chromiumcodereview.appspot.com/9965011/diff/5001/elf/Versions#newcode87
elf/Versions:87: __nacl_irt_mkdir;
the list should be in alphabetical order

https://chromiumcodereview.appspot.com/9965011/diff/5001/sysdeps/nacl/sysdep.h
File sysdeps/nacl/sysdep.h (right):

https://chromiumcodereview.appspot.com/9965011/diff/5001/sysdeps/nacl/sysdep....
sysdeps/nacl/sysdep.h:592: return ret > 0 ? 0 : errno;
*err must be updated in case of error
and we should return the length because glibc will later attempt to realloc the
buffer to that length, see:
sysdeps/unix/sysv/linux/getcwd.c

[vissi, 2012-03-30 13:16:00]

On 2012/03/30 12:36:29, pasko wrote:
> https://chromiumcodereview.appspot.com/9965011/diff/5001/elf/Versions
> File elf/Versions (right):
> 
>
https://chromiumcodereview.appspot.com/9965011/diff/5001/elf/Versions#newcode87
> elf/Versions:87: __nacl_irt_mkdir;
> the list should be in alphabetical order
> 
> https://chromiumcodereview.appspot.com/9965011/diff/5001/sysdeps/nacl/sysdep.h
> File sysdeps/nacl/sysdep.h (right):
> 
>
https://chromiumcodereview.appspot.com/9965011/diff/5001/sysdeps/nacl/sysdep....
> sysdeps/nacl/sysdep.h:592: return ret > 0 ? 0 : errno;
> *err must be updated in case of error
> and we should return the length because glibc will later attempt to realloc
the
> buffer to that length, see:
> sysdeps/unix/sysv/linux/getcwd.c

PS 4 behaves as expected.

[pasko-google - do not use, 2012-03-30 13:24:58]

https://chromiumcodereview.appspot.com/9965011/diff/4014/sysdeps/nacl/sysdep.h
File sysdeps/nacl/sysdep.h (right):

https://chromiumcodereview.appspot.com/9965011/diff/4014/sysdeps/nacl/sysdep....
sysdeps/nacl/sysdep.h:592: if (buf == 0) {
GNU style code (ugly, huh?):
  if (buf == 0)
    {
      *err = EINVAL;
      return 0;
    }

[vissi, 2012-03-30 14:20:57]

On 2012/03/30 13:24:58, pasko wrote:
> https://chromiumcodereview.appspot.com/9965011/diff/4014/sysdeps/nacl/sysdep.h
> File sysdeps/nacl/sysdep.h (right):
> 
>
https://chromiumcodereview.appspot.com/9965011/diff/4014/sysdeps/nacl/sysdep....
> sysdeps/nacl/sysdep.h:592: if (buf == 0) {
> GNU style code (ugly, huh?):
>   if (buf == 0)
>     {
>       *err = EINVAL;
>       return 0;
>     }

removed incorrect returns, by @khim's suggestion

[khim, 2012-03-30 14:30:53]

LGTM