linux-kernel - Re: [RFC PATCH v4 3/8] iomap: pass blocksize to iomap_truncate

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]

Date: Sun, 2 Jun 2024 07:16:53 -0400
From: Brian Foster <bfoster@...hat.com>
To: Christoph Hellwig <hch@...radead.org>
Cc: Zhang Yi <yi.zhang@...weicloud.com>, linux-xfs@...r.kernel.org,
	linux-fsdevel@...r.kernel.org, linux-kernel@...r.kernel.org,
	djwong@...nel.org, brauner@...nel.org, david@...morbit.com,
	chandanbabu@...nel.org, jack@...e.cz, willy@...radead.org,
	yi.zhang@...wei.com, chengzhihao1@...wei.com, yukuai3@...wei.com
Subject: Re: [RFC PATCH v4 3/8] iomap: pass blocksize to iomap_truncate_page()

On Fri, May 31, 2024 at 05:39:10AM -0700, Christoph Hellwig wrote:
> > -		const struct iomap_ops *ops)
> > +iomap_truncate_page(struct inode *inode, loff_t pos, unsigned int blocksize,
> > +		bool *did_zero, const struct iomap_ops *ops)
> >  {
> > -	unsigned int blocksize = i_blocksize(inode);
> > -	unsigned int off = pos & (blocksize - 1);
> > +	unsigned int off = rem_u64(pos, blocksize);
> >  
> >  	/* Block boundary? Nothing to do */
> >  	if (!off)
> 
> Instad of passing yet another argument here, can we just kill
> iomap_truncate_page?
> 
> I.e. just open code the rem_u64 and 0 offset check in the only caller
> and call iomap_zero_range.  Same for the DAX variant and it's two
> callers.
> 
> 

Hey Christoph,

I've wondered the same about killing off iomap_truncate_page(), but JFYI
one of the several prototypes I have around of other potential ways to
address this problem slightly splits off truncate page from being a
straight zero range wrapper. A quick diff [2] of that is inlined below
for reference (only lightly tested, may be busted, etc.).

The idea is that IIRC truncate_page() was really the only zero range
user that might actually encounter dirty cache over unwritten mappings,
so given that and the typically constrained range size of truncate page,
just let it be more aggressive about bypassing the unwritten mapping
optimization in iomap_zero_iter(). Just something else to consider, and
this is definitely not something you'd want to do for zero range proper.

Brian

P.S., I think the last patches I actually posted around this were here
[1], but I also have multiple versions of that selective flush approach
ported to iomap instead of being XFS specific as well.

[1] https://lore.kernel.org/linux-xfs/20221104182358.2007475-1-bfoster@redhat.com/
    https://lore.kernel.org/linux-xfs/20221128173945.3953659-1-bfoster@redhat.com/
[2] truncate page POC:

--- 8< ---

diff --git a/fs/iomap/buffered-io.c b/fs/iomap/buffered-io.c
index c5802a459334..a261e732ea05 100644
--- a/fs/iomap/buffered-io.c
+++ b/fs/iomap/buffered-io.c
@@ -1380,7 +1380,8 @@ iomap_file_unshare(struct inode *inode, loff_t pos, loff_t len,
 }
 EXPORT_SYMBOL_GPL(iomap_file_unshare);
 
-static loff_t iomap_zero_iter(struct iomap_iter *iter, bool *did_zero)
+static loff_t iomap_zero_iter(struct iomap_iter *iter, bool *did_zero,
+			      bool dirty_cache)
 {
 	const struct iomap *srcmap = iomap_iter_srcmap(iter);
 	loff_t pos = iter->pos;
@@ -1388,7 +1389,8 @@ static loff_t iomap_zero_iter(struct iomap_iter *iter, bool *did_zero)
 	loff_t written = 0;
 
 	/* already zeroed?  we're done. */
-	if (srcmap->type == IOMAP_HOLE || srcmap->type == IOMAP_UNWRITTEN)
+	if (srcmap->type == IOMAP_HOLE ||
+	    (!dirty_cache && srcmap->type == IOMAP_UNWRITTEN))
 		return length;
 
 	do {
@@ -1439,7 +1441,7 @@ iomap_zero_range(struct inode *inode, loff_t pos, loff_t len, bool *did_zero,
 	int ret;
 
 	while ((ret = iomap_iter(&iter, ops)) > 0)
-		iter.processed = iomap_zero_iter(&iter, did_zero);
+		iter.processed = iomap_zero_iter(&iter, did_zero, false);
 	return ret;
 }
 EXPORT_SYMBOL_GPL(iomap_zero_range);
@@ -1450,11 +1452,29 @@ iomap_truncate_page(struct inode *inode, loff_t pos, bool *did_zero,
 {
 	unsigned int blocksize = i_blocksize(inode);
 	unsigned int off = pos & (blocksize - 1);
+	struct iomap_iter iter = {
+		.inode		= inode,
+		.pos		= pos,
+		.len		= blocksize - off,
+		.flags		= IOMAP_ZERO,
+	};
+	loff_t end;
+	int ret;
+	bool dirty_cache = false;
 
 	/* Block boundary? Nothing to do */
 	if (!off)
 		return 0;
-	return iomap_zero_range(inode, pos, blocksize - off, did_zero, ops);
+
+	/* overwrite unwritten ranges if any part of the range is dirty for
+	 * truncate page */
+	end = iter.pos + iter.len - 1;
+	if (filemap_range_needs_writeback(inode->i_mapping, iter.pos, end))
+		dirty_cache = true;
+
+	while ((ret = iomap_iter(&iter, ops)) > 0)
+		iter.processed = iomap_zero_iter(&iter, did_zero, dirty_cache);
+	return ret;
 }
 EXPORT_SYMBOL_GPL(iomap_truncate_page);