Re: [PATCH/RFC] minix filesystem: Corrected patch

["Daniel Aragonés" <danarag@gmail.com>]

Hi Pekka!

On 1/22/06, you wrote:

 >+                                       offset = p - kaddr;
 >> +                                       over = filldir(dirent, de3->name, l,
 >> +                                       (n<<PAGE_CACHE_SHIFT) | offset,
 >> +                                       de3->inode, DT_UNKNOWN);
 >Hmm, strange formatting. Wouldn't it be better if you introduced a
 >name pointer and moved those filldir bits outside of the if-else
 >block? Less code duplication that way.
 >+                               if (namecompare(namelen,sbi->s_namelen,name,de3->name))
 >> +                                       goto found;
 >> +                       }
 >Same here.
 >+                                       goto out_unlock;
 >> +                               de = minix_next_entry(de, sbi);
 >> +                               de3 = minix_next_entry(de3, sbi);
 >Why do you do both here?

You are right, but I thought that duplication was the appropiate to be the most conservative with the preexistent code and also providing for the needed duplication of the strucutre minix_dir_entry.
The secondary structure (minix3_dir_entry) has to follow all the endeavours of its parent one, so both are here.

 >+               sbi->s_log_zone_size = *(__u16 *)(bh->b_data + 12);
 >> +               sbi->s_max_size = *(__u32 *)(bh->b_data + 16);
 >> +               sbi->s_nzones = *(__u32 *)(bh->b_data + 20);
 >You probably want to introduce a struct minix3_super_block for this.
 >It's much more readable that way.

Yes, but if I do, is closer to a rewrite of the preexistent code. And I think that it not deserves it. Minix is not so important (sorry if some one is listening).

 >+                               goto out_bad_hblock;
 >> +               }
 >You're now setting the block size twice for the V3 case.

You are right.

 >+#define MINIX2_INODES_PER_BLOCK(b) ((b)/(sizeof (struct minix2_inode)))
 >Maybe this should be called minix_inodes_per_block instead and be a
 >static inline function?

Just to follow the style found.