linux-kernel - Re: [PATCH 02/11] vfs: Add better VFS support for page

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20090706103539.GA2611@infradead.org>
Date:	Mon, 6 Jul 2009 06:35:40 -0400
From:	Christoph Hellwig <hch@...radead.org>
To:	Nick Piggin <npiggin@...e.de>
Cc:	Christoph Hellwig <hch@...radead.org>, Jan Kara <jack@...e.cz>,
	LKML <linux-kernel@...r.kernel.org>, linux-mm@...ck.org,
	linux-fsdevel@...r.kernel.org
Subject: Re: [PATCH 02/11] vfs: Add better VFS support for page_mkwrite
	when blocksize < pagesize

On Mon, Jul 06, 2009 at 11:08:04AM +0200, Nick Piggin wrote:
> OK, hmm, but I wonder -- most of the time do_truncate will need to
> call notify_change anyway, so I wonder if avoiding the double
> indirection saves us anything? (It requires 2 indirect calls either
> way). And if we call ->setsize from ->setattr, then a filesystem
> which implements its own ->setattr could avoid one of those indirect
> calls. Not so if do_truncate has to call ->setattr then ->setsize.

I don't quite understand what you mean here. In the end there should
be one single indirect call, ->setsize (or whatever it's called by
then).

In the first round we'd split up a helper just for size updates from
notify_change, ala:

int vfs_truncate(struct dentry *dentry, loff_t size, int flags, file)
{
	int error;

	error = security_inode_truncate(dentry, size, flags, file);
	if (error)
		return error;

	if (inode->i_op->setsize) {
		inode->i_op->setsize(dentry, size, flags, file);

	} else {
		<... built up iattr here ...>

		if (inode->i_op->setattr) {	
			down_write(&dentry->d_inode->i_alloc_sem);
			error = inode->i_op->setattr(dentry, attr);
			up_write(&dentry->d_inode->i_alloc_sem);
		} else {
			down_write(&dentry->d_inode->i_alloc_sem);
			error = inode_setattr(inode, attr);
			up_write(&dentry->d_inode->i_alloc_sem);
		}
	}

	if (!error)
		fsnotify_truncate(dentry, size, flags);
	return error;
}

One all filesistem are converted to have a setsize method (either their
own or simple_setsize) the !inode->i_op->setsize case can go away.

Note that the above variant moves taking i_alloc_sem into ->setsize as
it's not required for most filesystems (I think only extN need for
O_DIRECT).

Also the above doesn't deal with killing the SUID/SGID bits yet, we'll
need some good way for that.

Actually it might be better to just pass the iattr to ->setsize to so
we can have the parsing for those arguments once, and that filesystems
can re-use parts of their ->setattr for ->setsize if it's complex enough
(timestamp updates and suid/sgid killing)

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/