Re: [Kernel Bug] WARNING in vfat_rmdir

[Zhiyu Zhang]

Hi Matthew,

Thanks for your reply!

> Ooh, -1 is a real errno, so -1 or errno is a bad API.  I'd suggest just
> returning -ENOENT directly instead of translating it in the caller.

I agree with this advice. I can let fat_get_entry() return -ENOENT like:

static int fat__get_entry(struct inode *dir, loff_t *pos,
        *bh = NULL;
        iblock = *pos >> sb->s_blocksize_bits;
        err = fat_bmap(dir, iblock, &phys, &mapped_blocks, 0, false);
-       if (err || !phys)
-               return -1;      /* beyond EOF or error */
+       if (err)
+               return err;     /* real error (e.g., -EIO)*/
+       if (!phys)
+               return -ENOENT;      /* EOF */

Then I also need to re-handle the callers of fat_get_entry() as some
of them are still waiting for a "-1".

I'm willing to submit a patch to standardize the errno (though it
seems to unrelated to this bug), as long as it is approved :)

Best,
Zhiyu Zhang