Re: [PATCH][FAT] FAT dirent scan with hin take #3

[Machida, Hiroyuki]

Hi,

Pekka Enberg wrote:
> Hi,
>
> On 8/31/05, Machida, Hiroyuki <machida@xxxxxxxxxxxxx> wrote:
>
> > +inline
> > +static int hint_allocate(struct inode *dir)
> > +{
> > +       loff_t *hints;
> > +       int err = 0;
> > +
> > +       if (!MSDOS_I(dir)->scan_hints) {
> > +               hints = kcalloc(FAT_SCAN_NWAY, sizeof(loff_t), GFP_KERNEL);
> > +               if (!hints)
> > +                       err = -ENOMEM;
>
>
> Better to bail out here as...
>
>
> > +
> > +               down(&MSDOS_I(dir)->scan_lock);
> > +               if (MSDOS_I(dir)->scan_hints)
> > +                       err = -EINVAL;
>
>
> ...you might overwrite -ENOMEM here masking the real problem.

It's ok. If MSDOS_I(dir)->scan_hints isn't NULL, it's not error case.
We need just kfree(hints) and return 0.(Assuming  kfree() accepts NULL)
I think EINVAL confuse you.
> > +               if (!err)
> > +                       MSDOS_I(dir)->scan_hints = hints;
> > +               up(&MSDOS_I(dir)->scan_lock);
> > +               if (err == -EINVAL) {
>
>
> Gotos would make error handling much cleaner.
How about this ?

	if (!MSDOS_I(dir)->scan_hints) {
		hints  = kcalllo(....);

		down
		if (MSDOS_I(dir)->scan_hints) {
			up
			goto already_allocated;
		}
		if (hints)
			MSDOS_I(dir)->scan_hints = hints;
		up
	}
	return (hints == 0) ? -ENOMEM : 0;

already_allocated:
	kfree(hints); /* kfree accepts NULL */
	return 0;
		


> > +inline
> > +static int hint_index_body(const unsigned char *name, int name_len, int check_null)
>
>
> Please consider calling this __hint_index() instead as per normal
> naming conventions.

Agree.

> > +{
> > +       int i;
> > +       int val = 0;
> > +       unsigned char *p = (unsigned char *) name;
> > +       int id = current->pid;
> > +
> > +       for (i=0; i<name_len; i++) {
> > +               if (check_null && !*p) break;
>
>
> Please put break on separate line. You still have quite a few of these.

Agree

> > +               val = ((val << 1) & 0xfe) | ((val & 0x80) ? 1 : 0);
> > +               val ^= *p;
> > +               p ++;
> > +       }
> > +       id = ((id >> 8) & 0xf) ^ (id & 0xf);
> > +       val = (val << 1) | (id & 1);
> > +       return val & (FAT_SCAN_NWAY-1);
> > +}
>
>
>                               Pekka


--
Hiroyuki Machida		machida@xxxxxxxxxxxxx		
SSW Dept. HENC, Sony Corp.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/