linux-kernel - Re: [PATCH] dax: Allow block size > 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] [day] [month] [year] [list]

Message-ID: <672bcab0911a2_10bc62943f@dwillia2-xfh.jf.intel.com.notmuch>
Date: Wed, 6 Nov 2024 11:59:44 -0800
From: Dan Williams <dan.j.williams@...el.com>
To: Jan Kara <jack@...e.cz>, Asahi Lina <lina@...hilina.net>
CC: Dave Chinner <david@...morbit.com>, Jan Kara <jack@...e.cz>, Dan Williams
	<dan.j.williams@...el.com>, Matthew Wilcox <willy@...radead.org>, "Alexander
 Viro" <viro@...iv.linux.org.uk>, Christian Brauner <brauner@...nel.org>,
	Sergio Lopez Pascual <slp@...hat.com>, <linux-fsdevel@...r.kernel.org>,
	<nvdimm@...ts.linux.dev>, <linux-kernel@...r.kernel.org>,
	<asahi@...ts.linux.dev>
Subject: Re: [PATCH] dax: Allow block size > PAGE_SIZE

Jan Kara wrote:
[..]
> > This WARN still feels like the wrong thing, though. Right now it is the
> > only thing in DAX code complaining on a page size/block size mismatch
> > (at least for virtiofs). If this is so important, I feel like there
> > should be a higher level check elsewhere, like something happening at
> > mount time or on file open. It should actually cause the operations to
> > fail cleanly.
> 
> That's a fair point. Currently filesystems supporting DAX check for this in
> their mount code because there isn't really a DAX code that would get
> called during mount and would have enough information to perform the check.
> I'm not sure adding a new call just for this check makes a lot of sense.
> But if you have some good place in mind, please tell me.

Is not the reason that dax_writeback_mapping_range() the only thing
checking ->i_blkbits because 'struct writeback_control' does writeback
in terms of page-index ranges?

All other dax entry points are filesystem controlled that know the
block-to-pfn-to-mapping relationship.

Recall that dax_writeback_mapping_range() is historically for pmem
persistence guarantees to make sure that applications write through CPU
cache to media.

Presumably there are no cache coherency concerns with fuse and dax
writes from the guest side are not a risk of being stranded in CPU
cache. Host side filesystem writeback will take care of them when / if
the guest triggers a storage device cache flush, not a guest page cache
writeback.

So, I think the simplest fix here is potentially:

diff --git a/fs/fuse/dax.c b/fs/fuse/dax.c
index 12ef91d170bb..15cf7bb20b5e 100644
--- a/fs/fuse/dax.c
+++ b/fs/fuse/dax.c
@@ -777,11 +777,8 @@ ssize_t fuse_dax_write_iter(struct kiocb *iocb, struct iov_iter *from)
 static int fuse_dax_writepages(struct address_space *mapping,
 			       struct writeback_control *wbc)
 {
-
-	struct inode *inode = mapping->host;
-	struct fuse_conn *fc = get_fuse_conn(inode);
-
-	return dax_writeback_mapping_range(mapping, fc->dax->dev, wbc);
+	/* nothing to flush, fuse cache coherency is managed by the host */
+	return 0;
 }
 
 static vm_fault_t __fuse_dax_fault(struct vm_fault *vmf, unsigned int order,