public inbox for linux-kernel@vger.kernel.org
 help / color / mirror / Atom feed
From: Eric Sandeen <sandeen@redhat.com>
To: Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
	linux-fsdevel <linux-fsdevel@vger.kernel.org>
Cc: Theodore Tso <tytso@mit.edu>
Subject: [RFC/PATCH] - revert generic_fillattr stat->blksize to PAGE_CACHE_SIZE
Date: Mon, 06 Nov 2006 15:50:02 -0600	[thread overview]
Message-ID: <454FAE0A.3070409@redhat.com> (raw)

The inode diet patches first did this to generic_fillattr():

-	stat->blksize = inode->i_blksize;
+	stat->blksize = PAGE_CACHE_SIZE;

but by 2.6.19-rc3 this was changed again:

-	stat->blksize = PAGE_CACHE_SIZE;
+	stat->blksize = (1 << inode->i_blkbits);

I can't find for sure why this was done; perhaps because it exposed a bug
in cifs[1]

However, if we are going to pick a default for generic_fillattr, it should probably
be what most filesystems were using before, and let filesystems which need something
else re-set it to according to their needs.  As it stands today, doing readdirs and
the like in block-sized chunks rather than page sized will probably not be the
best thing for performance.

A bit of quick parsing of the original inode diet patch shows that by far most
filesystems were using the page size:

(count) (line which sets i_blksize)
      1 cifs_inode->vfs_inode.i_blksize = CIFS_MAX_MSGSIZE;
      1 i->i_blksize = 512;
      1 inode->i_blksize = attr->va_blocksize;
      1 inode->i_blksize = befs_sb->block_size;
      1 inode->i_blksize = HFSPLUS_SB(inode->i_sb).alloc_blksz;
      1 inode->i_blksize = HFSPLUS_SB(sb).alloc_blksz;
      1 inode->i_blksize = HFS_SB(inode->i_sb)->alloc_blksz;
      1 inode->i_blksize = HFS_SB(sb)->alloc_blksz;
      1 inode->i_blksize = HPAGE_SIZE;
      1 inode->i_blksize = NCP_BLOCK_SIZE;
      1 inode->i_blksize = QNX4_DIR_ENTRY_SIZE;
      1 inode->i_blksize = (u32)osb->s_clustersize;
      1 inode->i_blksize = xfs_preferred_iosize(mp);
      1 ino->i_blksize = i_blksize;
      1 ino->i_blksize = proc_ino->i_blksize;
      1 ip->i_blksize = ip->i_sb->s_blocksize;
      1 ip->i_blksize = PAGE_SIZE;
      2 inode->i_blksize = fattr->du.nfs2.blocksize;
      2 inode->i_blksize = inode->i_sb->s_blocksize;
      2 inode->i_blksize = reiserfs_default_io_size;
      2 inode->i_blksize = sbi->cluster_size;
      2 vi->i_blksize = PAGE_CACHE_SIZE;
      3 inode->i_blksize = sb->s_blocksize;
      3 ret->i_blksize = PAGE_CACHE_SIZE;
      5 inode->i_blksize = 1024;
      7 inode->i_blksize = 0;
     12 inode->i_blksize = PAGE_SIZE;
     22 inode->i_blksize = PAGE_CACHE_SIZE;

so I would propose the following patch to make PAGE_CACHE_SIZE the default (again), 
and let filesystems which need something -else- do that on their own.

[1] http://lists.samba.org/archive/linux-cifs-client/2006-September/001481.html

Thanks,
-Eric

Signed-off-by: Eric Sandeen <sandeen@redhat.com>

--- linux.orig/fs/stat.c	2006-11-05 23:27:04.962482569 -0600
+++ linux/fs/stat.c	2006-11-05 23:27:29.394396050 -0600
@@ -33,7 +33,7 @@
  	stat->ctime = inode->i_ctime;
  	stat->size = i_size_read(inode);
  	stat->blocks = inode->i_blocks;
-	stat->blksize = (1 << inode->i_blkbits);
+	stat->blksize = PAGE_CACHE_SIZE;
  }

  EXPORT_SYMBOL(generic_fillattr);


             reply	other threads:[~2006-11-06 21:50 UTC|newest]

Thread overview: 18+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2006-11-06 21:50 Eric Sandeen [this message]
2006-11-06 22:27 ` [RFC/PATCH] - revert generic_fillattr stat->blksize to PAGE_CACHE_SIZE Dave Kleikamp
2006-11-06 22:37   ` Eric Sandeen
2006-11-07  0:26   ` Steve French
2006-11-07 13:40     ` Dave Kleikamp
2006-11-07 15:34       ` Dave Kleikamp
2006-11-07 15:50         ` Steve French
2006-11-06 23:05 ` Christoph Hellwig
2006-11-06 23:15   ` Eric Sandeen
2006-11-06 23:17     ` Christoph Hellwig
2006-11-06 23:28       ` Eric Sandeen
2006-11-07  0:08     ` Andreas Dilger
2006-11-07  0:13       ` Hua Zhong
2006-11-07  1:39       ` Eric Sandeen
2006-11-07  3:43     ` Theodore Tso
2006-11-07  4:02       ` Eric Sandeen
  -- strict thread matches above, loose matches on Subject: below --
2006-12-03 13:21 col-pepper
2006-12-03 13:48 ` Adrian Bunk

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=454FAE0A.3070409@redhat.com \
    --to=sandeen@redhat.com \
    --cc=linux-fsdevel@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=tytso@mit.edu \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox