* [PATCH] bfs iget() abuses
@ 2005-10-04 16:48 Al Viro
2005-10-05 9:41 ` Tigran Aivazian
0 siblings, 1 reply; 2+ messages in thread
From: Al Viro @ 2005-10-04 16:48 UTC (permalink / raw)
To: Linus Torvalds; +Cc: Tigran A. Aivazian, linux-kernel
bfs_fill_super() walks the inode table to get the bitmap of
free inodes and collect stats. It has no business using iget() for
that - it's a lot of extra work, extra icache pollution and more
complex code. Switched to walking the damn thing directly.
Note: that also allows to kill ->i_dsk_ino in there - separate
patch if Tigran can confirm that this field can be zero only for deleted
inodes (i.e. something that could only be found during that scan and not
by normal lookups).
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
----
[this is a followup to Alexey's bfs annotations patch]
diff -urN RC14-rc3-git4-base/fs/bfs/inode.c current/fs/bfs/inode.c
--- RC14-rc3-git4-base/fs/bfs/inode.c 2005-10-04 12:30:06.000000000 -0400
+++ current/fs/bfs/inode.c 2005-10-04 05:57:46.000000000 -0400
@@ -362,23 +362,41 @@
info->si_lf_eblk = 0;
info->si_lf_sblk = 0;
info->si_lf_ioff = 0;
+ bh = NULL;
for (i=BFS_ROOT_INO; i<=info->si_lasti; i++) {
- inode = iget(s,i);
- if (BFS_I(inode)->i_dsk_ino == 0)
+ struct bfs_inode *di;
+ int block = (i - BFS_ROOT_INO)/BFS_INODES_PER_BLOCK + 1;
+ int off = (i - BFS_ROOT_INO) % BFS_INODES_PER_BLOCK;
+ unsigned long sblock, eblock;
+
+ if (!off) {
+ brelse(bh);
+ bh = sb_bread(s, block);
+ }
+
+ if (!bh)
+ continue;
+
+ di = (struct bfs_inode *)bh->b_data + off;
+
+ if (!di->i_ino) {
info->si_freei++;
- else {
- set_bit(i, info->si_imap);
- info->si_freeb -= inode->i_blocks;
- if (BFS_I(inode)->i_eblock > info->si_lf_eblk) {
- info->si_lf_eblk = BFS_I(inode)->i_eblock;
- info->si_lf_sblk = BFS_I(inode)->i_sblock;
- info->si_lf_ioff = BFS_INO2OFF(i);
- }
+ continue;
+ }
+ set_bit(i, info->si_imap);
+ info->si_freeb -= BFS_FILEBLOCKS(di);
+
+ sblock = le32_to_cpu(di->i_sblock);
+ eblock = le32_to_cpu(di->i_eblock);
+ if (eblock > info->si_lf_eblk) {
+ info->si_lf_eblk = eblock;
+ info->si_lf_sblk = sblock;
+ info->si_lf_ioff = BFS_INO2OFF(i);
}
- iput(inode);
}
+ brelse(bh);
if (!(s->s_flags & MS_RDONLY)) {
- mark_buffer_dirty(bh);
+ mark_buffer_dirty(info->si_sbh);
s->s_dirt = 1;
}
dump_imap("read_super", s);
^ permalink raw reply [flat|nested] 2+ messages in thread
* Re: [PATCH] bfs iget() abuses
2005-10-04 16:48 [PATCH] bfs iget() abuses Al Viro
@ 2005-10-05 9:41 ` Tigran Aivazian
0 siblings, 0 replies; 2+ messages in thread
From: Tigran Aivazian @ 2005-10-05 9:41 UTC (permalink / raw)
To: Al Viro; +Cc: Linus Torvalds, linux-kernel
Hi Al,
Yes, looks good to me and also I confirm that ->i_dsk_ino is only zero for
deletes inodes.
Kind regards
Tigran
On Tue, 4 Oct 2005, Al Viro wrote:
> bfs_fill_super() walks the inode table to get the bitmap of
> free inodes and collect stats. It has no business using iget() for
> that - it's a lot of extra work, extra icache pollution and more
> complex code. Switched to walking the damn thing directly.
>
> Note: that also allows to kill ->i_dsk_ino in there - separate
> patch if Tigran can confirm that this field can be zero only for deleted
> inodes (i.e. something that could only be found during that scan and not
> by normal lookups).
>
> Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
> ----
> [this is a followup to Alexey's bfs annotations patch]
>
> diff -urN RC14-rc3-git4-base/fs/bfs/inode.c current/fs/bfs/inode.c
> --- RC14-rc3-git4-base/fs/bfs/inode.c 2005-10-04 12:30:06.000000000 -0400
> +++ current/fs/bfs/inode.c 2005-10-04 05:57:46.000000000 -0400
> @@ -362,23 +362,41 @@
> info->si_lf_eblk = 0;
> info->si_lf_sblk = 0;
> info->si_lf_ioff = 0;
> + bh = NULL;
> for (i=BFS_ROOT_INO; i<=info->si_lasti; i++) {
> - inode = iget(s,i);
> - if (BFS_I(inode)->i_dsk_ino == 0)
> + struct bfs_inode *di;
> + int block = (i - BFS_ROOT_INO)/BFS_INODES_PER_BLOCK + 1;
> + int off = (i - BFS_ROOT_INO) % BFS_INODES_PER_BLOCK;
> + unsigned long sblock, eblock;
> +
> + if (!off) {
> + brelse(bh);
> + bh = sb_bread(s, block);
> + }
> +
> + if (!bh)
> + continue;
> +
> + di = (struct bfs_inode *)bh->b_data + off;
> +
> + if (!di->i_ino) {
> info->si_freei++;
> - else {
> - set_bit(i, info->si_imap);
> - info->si_freeb -= inode->i_blocks;
> - if (BFS_I(inode)->i_eblock > info->si_lf_eblk) {
> - info->si_lf_eblk = BFS_I(inode)->i_eblock;
> - info->si_lf_sblk = BFS_I(inode)->i_sblock;
> - info->si_lf_ioff = BFS_INO2OFF(i);
> - }
> + continue;
> + }
> + set_bit(i, info->si_imap);
> + info->si_freeb -= BFS_FILEBLOCKS(di);
> +
> + sblock = le32_to_cpu(di->i_sblock);
> + eblock = le32_to_cpu(di->i_eblock);
> + if (eblock > info->si_lf_eblk) {
> + info->si_lf_eblk = eblock;
> + info->si_lf_sblk = sblock;
> + info->si_lf_ioff = BFS_INO2OFF(i);
> }
> - iput(inode);
> }
> + brelse(bh);
> if (!(s->s_flags & MS_RDONLY)) {
> - mark_buffer_dirty(bh);
> + mark_buffer_dirty(info->si_sbh);
> s->s_dirt = 1;
> }
> dump_imap("read_super", s);
>
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2005-10-05 9:41 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2005-10-04 16:48 [PATCH] bfs iget() abuses Al Viro
2005-10-05 9:41 ` Tigran Aivazian
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.