Re: [PATCH] virtio-mmio: Devices parameter parsing

[Rusty Russell <rusty@rustcorp.com.au>]

On Mon, 21 Nov 2011 14:44:48 +0000, Pawel Moll <pawel.moll@arm.com> wrote:
> On Mon, 2011-11-21 at 03:32 +0000, Rusty Russell wrote:
> > 2) Doesn't leak mem on error paths.
> 
> Em, I don't think my code was leaking memory in any error case? (no
> offence meant or taken ;-)

You're right, I'm wrong.  In your original version, your innovative use
of free() then strdup() worked as intended.

> > 3) Handles error from platform_device_register_resndata().
> 
> As my code was ignoring wrong descriptions (all the continues) I ignored
> this result as well (the implementation complains on its own anyway),
> but I get your point.

I was referring to this:

	  return platform_device_register_resndata(&virtio_mmio_cmdline_parent,
			  "virtio-mmio", virtio_mmio_cmdline_id++,
			  resources, ARRAY_SIZE(resources), NULL, 0) != NULL;

The set function returns 0 or a -ve errno, not true/false.

> > 4) Uses shorter names for static functions/variables.
> 
> Ah, I get the hint :-) I'm trying to keep the naming convention in
> static symbols as well, as it makes cscope/ctags/grep usage easier...
> I'll just use the abbreviated "vm_" prefixes then.
> 
> > See what you think...
> 
> Funnily enough when I proposed some string parser few years ago (totally
> different story) I was flamed for using strchr() instead of strsep() ;-)
> But ok, I don't mind getting back to basics.
> 
> > +static int set_cmdline_device(const char *device, const struct kernel_param *kp)
> [...]
> > +       delim = strchr(device, '@');
> [...]
> > +	*delim = '\0';
> 
> Ah. I forgot that strchr() takes const char * but returns "non-const"
> char *... Cheating, that's what it is ;-), but will work. Probably what
> we really want is something like
> 	kstrtoull_delim(device, 0, &val, "@\0")
> I'll have a look and may try to propose something of that sort, but
> that's another story. Maybe I should just use simple_strtol() for the
> time being?

Or would it be simpler to enhance sscanf() with some weird format option
for suffixing?  I haven't looked for similar cases, but I'd suspect a
big win in general.

This would be neater than anything else we've got:
        if (sscanf(device, "%llu@%llu[KMG]:%u", ...) != 3
            && sscanf(device, "%llu@%llu[KMG]:%u:%u", ...) != 4)
                return -EINVAL;

Cheers,
Rusty.