Re: [PATCH] btree support in xfs driver

[Marco Gerards]

Bean <address@hidden> writes:

> On Feb 1, 2008 3:57 AM, Marco Gerards <address@hidden> wrote:
>> Bean <address@hidden> writes:
>>
>> Hi Bean!
>>
>> > I add btree support for the xfs driver, and fix a few bugs. Now i' m
>> > able to list a directory of over 5000 files, please test it.
>>
>> You rock!
>>
>> Can you load big files now, that are stored in BTrees.  Did you use
>> code from elsewhere?
>
> it should be, grub_xfs_read_block is used by both file and directory.
>
>>
>> > 2008-02-01  Bean  <address@hidden>
>> >
>> >       * fs/xfs.h (grub_xfs_sblock): New member log2_dirblk.
>> >       (grub_xfs_btree_node): New structure.
>> >       (grub_xfs_btree_root): New structure.
>> >       (grub_xfs_inode): New member nblocks, extsize, nextents and btree.
>>
>> members
>>
>> >       (GRUB_XFS_EXTENT_OFFSET): Use exts instead of inode->data.extents.
>> >       (GRUB_XFS_EXTENT_BLOCK): Likewise.
>> >       (GRUB_XFS_EXTENT_SIZE): Likewise.
>> >       (grub_xfs_read_block): Support btree format type.
>> >       (grub_xfs_iterate_dir): Use regparm(1) attribute in call_hook.
>> >       Use directory block as basic unit.
>>
>> This won't work on non-i386.  So I think this has to be fixed like
>> that other bug, for example with the autoconf check extension you
>> proposed.
>
> i take a look at the aclocal.m4, the place where NESTED_FUNC_ATTR is defined:
>
> AC_MSG_RESULT([$grub_cv_i386_check_nested_functions])
>
> if test "x$grub_cv_i386_check_nested_functions" = xyes; then
>   AC_DEFINE([NESTED_FUNC_ATTR],
>       [__attribute__ ((__regparm__ (2)))],
>       [Catch gcc bug])
> else
> dnl Unfortunately, the above test does not detect a bug in gcc-4.0.
> dnl So use regparm 2 until a better test is found.
>   AC_DEFINE([NESTED_FUNC_ATTR],
>       [__attribute__ ((__regparm__ (2)))],
>       [Catch gcc bug])
> fi
>
> is there a problem here ? i think NESTED_FUNC_ATTR is defined
> regardless of the test result.

Ehm, this is weird.  This wasn't here from the beginning, right?

But at least this is on i386 only...

--
Marco