> Ok, here is the result of my tests:
>
> - in minix_read_super s_nzones wasn't set correctly for a V2 filesystem
> - remove silly and now obsolete restriction on filesystem size.
>
> Additionally i have included the following changes:
>
> - inodes are counted one-based, thus an inode with the number s_inodes is
> actually valid.
> - count_free didn't count correctly, the bitmaps are organized in 16 bit
> blocks in native byte order.
> - add some extra braces that gcc 2.8 warns about.
> - correctly handle file name aliases due to truncation
> - properly recognize a busy directory in minix_rmdir
> - V2 inodes have 16 bit link count.
Hi Andreas,
I looked through your patch and did have one reservation -- I'm not sure that
modifying the argument qstr in the dentry_op hash function is properly
supported. There's currently no precedent for doing this, and you might also
need to modify the name string for null termination. (Qstrs have a length but
also have null terminated strings.) So it might be safer to add a compare
function as well the hash.
Regards,
Bill