kobject sample code

kobject sample code

[John de la Garza]

I'm looking at samples/kobject/kobject-example.c

at line 39 foo is read:

        sscanf(buf, "%du", &foo);

foo is an int so why is it read using %du not %d?

kobject sample code

[Paul Bolle]

On Mon, 2014-12-08 at 21:20 -0500, John de la Garza wrote:
> I'm looking at samples/kobject/kobject-example.c
> 
> at line 39 foo is read:
> 
>         sscanf(buf, "%du", &foo);
> 
> foo is an int so why is it read using %du not %d?

My reading of lib/vsprintf.c:vsscanf() is that a "%du" format expects
"buf" to contain an integer followed by a literal 'u' char. Is that your
reading too?


Paul Bolle

kobject sample code

[John de la Garza]

On Tue, Dec 09, 2014 at 10:26:03AM +0100, Paul Bolle wrote:
> On Mon, 2014-12-08 at 21:20 -0500, John de la Garza wrote:
> > I'm looking at samples/kobject/kobject-example.c
> > 
> > at line 39 foo is read:
> > 
> >         sscanf(buf, "%du", &foo);
> > 
> > foo is an int so why is it read using %du not %d?
> 
> My reading of lib/vsprintf.c:vsscanf() is that a "%du" format expects
> "buf" to contain an integer followed by a literal 'u' char. Is that your
> reading too?
> 
After reading the code and the man page for vsscanf, it seems like
it is not expecting the 'u', but rather ignores it.

vsscanf is a bit complicated for me, so I may be missing something.

kobject sample code

[Paul Bolle]

On Sat, 2014-12-13 at 13:01 -0500, John de la Garza wrote:
> On Tue, Dec 09, 2014 at 10:26:03AM +0100, Paul Bolle wrote:
> > On Mon, 2014-12-08 at 21:20 -0500, John de la Garza wrote:
> > > I'm looking at samples/kobject/kobject-example.c
> > > 
> > > at line 39 foo is read:
> > > 
> > >         sscanf(buf, "%du", &foo);
> > > 
> > > foo is an int so why is it read using %du not %d?
> > 
> > My reading of lib/vsprintf.c:vsscanf() is that a "%du" format expects
> > "buf" to contain an integer followed by a literal 'u' char. Is that your
> > reading too?
> > 
> After reading the code and the man page for vsscanf, it seems like
> it is not expecting the 'u', but rather ignores it.

The manpage _probably_ describes intended behavior. But I'm not sure
whether the kernel's vsscanf() tries to emulate it all that faithfully.
(And, besides, that manpage would gain quite a bit with a few examples.)
Anyhow, the net effect of "%du" appears to be to ignore any non digit
trailing input. So I think you're reading the code correctly.

> vsscanf is a bit complicated for me, so I may be missing something.

I wouldn't have bothered replying if I had not noticed
https://lkml.org/lkml/2014/12/17/386 . You might like to have a look at
that patch. My primary observation is that its commit explanation is not
very helpful. (That's why I CC-ed Ratislav.)


Paul Bolle

kobject sample code

[Rastislav Barlik]

On Wed, 17 Dec 2014 22:31:21 +0100
Paul Bolle <pebolle@tiscali.nl> wrote:

> On Sat, 2014-12-13 at 13:01 -0500, John de la Garza wrote:
> > On Tue, Dec 09, 2014 at 10:26:03AM +0100, Paul Bolle wrote:
> > > On Mon, 2014-12-08 at 21:20 -0500, John de la Garza wrote:
> > > > I'm looking at samples/kobject/kobject-example.c
> > > > 
> > > > at line 39 foo is read:
> > > > 
> > > >         sscanf(buf, "%du", &foo);
> > > > 
> > > > foo is an int so why is it read using %du not %d?
> > > 
> > > My reading of lib/vsprintf.c:vsscanf() is that a "%du" format
> > > expects "buf" to contain an integer followed by a literal 'u'
> > > char. Is that your reading too?
> > > 
> > After reading the code and the man page for vsscanf, it seems like
> > it is not expecting the 'u', but rather ignores it.
> 
> The manpage _probably_ describes intended behavior. But I'm not sure
> whether the kernel's vsscanf() tries to emulate it all that
> faithfully. (And, besides, that manpage would gain quite a bit with a
> few examples.) Anyhow, the net effect of "%du" appears to be to
> ignore any non digit trailing input. So I think you're reading the
> code correctly.
> 
> > vsscanf is a bit complicated for me, so I may be missing something.
> 
> I wouldn't have bothered replying if I had not noticed
> https://lkml.org/lkml/2014/12/17/386 . You might like to have a look
> at that patch. My primary observation is that its commit explanation
> is not very helpful. (That's why I CC-ed Ratislav.)
> 
> Paul Bolle

My understanding is that sscanf() would ignore any trailing input
anyway. So I think that "%du" is just a typo and should have been "%d"
instead. There should also be check for return value from sscanf().
I send a patch which uses kstrtoint() as it is preferred to single
variable sscanf().

-- 
Kind regards
Rastislav Barlik