lkml.org 
[lkml]   [2022]   [Jul]   [8]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH 1/2] lib/string_helpers: Introduce strsplit_u32()
On Fri, Jul 08, 2022 at 02:13:14PM +0200, Cezary Rojewski wrote:
> On 2022-07-08 1:46 PM, Andy Shevchenko wrote:
> > On Fri, Jul 8, 2022 at 1:33 PM Cezary Rojewski
> > <cezary.rojewski@intel.com> wrote:

...

> > > When I'd written the very first version of this function many months
> > > ago, get_options() looked as it does not fulfill our needs. It seems to
> > > be true even today: caller needs to know the number of elements in an
> > > array upfront.
> >
> > Have you read a kernel doc for it? It does return the number of
> > elements at the first pass.
>
> Yes, I've checked several parts of it. Perhaps I did miss something but
> simple_strtoull() doc reads: use kstrtox() instead.

Doc was fixed to make clearer that in some cases it's okay to use
simple_strtox(). And this was exactly due to obfuscation code with this
recommendation. Yes, in general one supposed to use kstrtox(), but it's
not 100% obligatory.

> Thus the strsplit_u32()
> makes use of kstrtox().

Yeah...

> > > Also, kstrtox() takes into account '0x' and modifies the
> > > base accordingly if that's the case. simple_strtoull() looks as not
> > > capable of doing the same thing.
> >
> > How come?! It does parse all known prefixes: 0x, 0, +, -.
>
> Hmm.. doc says that it stops at the first non-digit character. Will
> re-check.

Yes, but under non-digit implies the standard prefixes of digits.
simple_strtox() and kstrotox() use the very same function for prefixes.

> > > The goal is to be able to parse input such as:
> > >
> > > 0x1000003,0,0,0x1000004,0,0
> > >
> > > into a sequence of 6 uints, filling the *tkns and *num_tkns for the caller.
> >
> > Yes. Have you checked the test cases for get_options()?

(1)

...

> > > avs-driver, which is also part of the ASoC framework has very similar
> > > debug-interface. I believe there's no need to duplicate the functions -
> > > move them to common code instead.
> >
> > Taking the above into account, please try to use get_options() and
> > then tell me what's not working with it. If so, we will add test cases
> > to get_options() and fix it.
>
> There is a difference:
>
> // get_options
> int ints[5];
>
> s = get_options(str, ARRAY_SIZE(ints), ints);
>
> // strsplit_u32()
> u32 *tkns, num_tkns;
>
> ret = strsplit_u32(str, delim, &tkns, &num_tkns);
>
> Nothing has been told upfront for in the second case.

It seems you are missing the (1). The code has checks for the case where you
can do get number upfront, it would just require two passes, but it's nothing
in comparison of heave realloc().

unsigned int *tokens;
char *p;
int num;

p = get_options(str, 0, &num);
if (num == 0)
// No numbers in the string!

tokens = kcalloc(num + 1, sizeof(*tokens), GFP_KERNEL);
if (!tokens)
return -ENOMEM;

p = get_oprions(str, num, &tokens);
if (*p)
// String was parsed only partially!
// assuming it's not a fatal error

return tokens;

--
With Best Regards,
Andy Shevchenko


\
 
 \ /
  Last update: 2022-07-08 14:31    [W:0.051 / U:0.428 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site