mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Steve French <smfltc@us.ibm.com>
To: joern@wohnheim.fh-wedel.de, linux-kernel@vger.kernel.org
Cc: linux-cifs-client@lists.samba.org, jra@samba.org
Subject: Re: [PATCH COW] sys_copyfile
Date: 27 Apr 2004 10:58:25 -0500	[thread overview]
Message-ID: <1083081505.12804.65.camel@stevef95.austin.ibm.com> (raw)

> With warm cache, copyfile() is about 10% faster

Over the network it would be a lot more than that.

With cifs vfs, I would expect that copyfile of files located on the same
server share would be much faster than that perhaps 5 to 10 times faster
due to the huge reduction in network bandwidth usage and latency - since
for the case of cifs the protocol already defines a single network
operation for copy file that could be then used (although it would be
preceeded by a couple of cheap cifs calls for the directory name lookup,
although the target mode complicates things for cifs - there would be a
minor change needed at the end to set the mode if the copy succeeds
which makes the copy error paths over the ).  For the CIFS Unix
Extensions version 2 would could add a trivial change to Samba to take
the mode as part of copy network parms.

in do_copyfile all I would need would be an op that looks a bit like 
rename (the cifs vfs part of the changes, to fs/cifs/cifssmb.c mostly,
would be trivial) e.g.:

 int do_copyfile(struct nameidata *old_nd, struct nameidata *new_nd,
                 struct dentry *new_dentry, umode_t mode)
{
-	int ret;
+       int ret = 0;

        if (!old_nd->dentry->d_inode)
                return -ENOENT;
        if (!S_ISREG(old_nd->dentry->d_inode->i_mode))
                return -EINVAL;
        /* FIXME: replace with proper permission check */
        if (new_dentry->d_inode)
                return -EEXIST;

+	if(old_nd->dentry->d_inode->i_op->copy) {
+		ret = old_dir->i_op->copy(old_nd->dentry, 
+			mode, new_dentry);
+	}

	if(!ret)
		return ret;
	else
		ret = vfs_create(new_nd->dentry->d_inode,
			 new_dentry, mode, new_nd);
	if (ret)
                return ret;

	ret = copy_data(old_nd->dentry, old_nd->mnt, new_dentry,
		new_nd->mnt);

        if (ret) {
                int error = vfs_unlink(new_nd->dentry->d_inode,
			new_dentry);
                BUG_ON(error);
               /* FIXME: not sure if there are return value we 
			should not BUG()
	               * on */
        }
        return ret;
}
	


             reply	other threads:[~2004-04-27 15:59 UTC|newest]

Thread overview: 7+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2004-04-27 15:58 Steve French [this message]
2004-04-27 16:42 ` Jörn Engel
2004-04-27 19:46   ` Steve French
2004-04-27 19:55     ` Jörn Engel
  -- strict thread matches above, loose matches on Subject: below --
2004-04-26  9:20 Jörn Engel
2004-04-27 11:30 ` Erik Mouw
2004-04-27 11:47   ` Jörn Engel

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=1083081505.12804.65.camel@stevef95.austin.ibm.com \
    --to=smfltc@us.ibm.com \
    --cc=joern@wohnheim.fh-wedel.de \
    --cc=jra@samba.org \
    --cc=linux-cifs-client@lists.samba.org \
    --cc=linux-kernel@vger.kernel.org \
    /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

all inboxes | Powered by JetHome®