Received-SPF: pass (google.com: domain of linux-nfs+bounces-3121-linux.lists.archive=gmail.com@vger.kernel.org designates 139.178.88.99 as permitted sender) client-ip=139.178.88.99;
Organization: Red Hat UK Ltd. Registered Address: Red Hat UK Ltd, Amberley
	Place, 107-111 Peascod Street, Windsor, Berkshire, SI4 1TE, United
	Kingdom.
	Registered in England and Wales under Company Registration No. 3798903
From: David Howells <dhowells@redhat.com>
In-Reply-To: <20240430140056.261997-15-dhowells@redhat.com>
References: <20240430140056.261997-15-dhowells@redhat.com> <20240430140056.261997-1-dhowells@redhat.com>
To: Christian Brauner <christian@brauner.io>
Cc: dhowells@redhat.com, Jeff Layton <jlayton@kernel.org>,
    Gao Xiang <hsiangkao@linux.alibaba.com>,
    Dominique Martinet <asmadeus@codewreck.org>,
    Matthew Wilcox <willy@infradead.org>,
    Steve French <smfrench@gmail.com>,
    Marc Dionne <marc.dionne@auristor.com>,
    Paulo Alcantara <pc@manguebit.com>,
    Shyam Prasad N <sprasad@microsoft.com>, Tom Talpey <tom@talpey.com>,
    Eric Van Hensbergen <ericvh@kernel.org>,
    Ilya Dryomov <idryomov@gmail.com>, netfs@lists.linux.dev,
    linux-cachefs@redhat.com, linux-afs@lists.infradead.org,
    linux-cifs@vger.kernel.org, linux-nfs@vger.kernel.org,
    ceph-devel@vger.kernel.org, v9fs@lists.linux.dev,
    linux-erofs@lists.ozlabs.org, linux-fsdevel@vger.kernel.org,
    linux-mm@kvack.org, netdev@vger.kernel.org,
    linux-kernel@vger.kernel.org, Latchesar Ionkov <lucho@ionkov.net>,
    Christian Schoenebeck <linux_oss@crudebyte.com>
Subject: Re: [PATCH v2 14/22] netfs: New writeback implementation
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <458059.1714582859.1@warthog.procyon.org.uk>
Content-Transfer-Encoding: quoted-printable
Date: Wed, 01 May 2024 18:00:59 +0100
Message-ID: <458060.1714582859@warthog.procyon.org.uk>

This needs the attached change.  It needs to allow for netfs_perform_write=
()
changing i_size whilst we're doing writeback.  The issue is that i_size is
cached in the netfs_io_request struct (as that's what we're going to tell =
the
server the new i_size should be), but we're not updating this properly if
i_size moves between us creating the request and us deciding to write out =
the
folio in which i_size was when we created the request.

This can lead to the folio_zero_segment() that can be seen in the patch be=
low
clearing the wrong amount of the final page - assuming it's still the fina=
l
page.

David
---
diff --git a/fs/netfs/write_issue.c b/fs/netfs/write_issue.c
index 69c50f4cbf41..e190043bc0da 100644
--- a/fs/netfs/write_issue.c
+++ b/fs/netfs/write_issue.c
@@ -315,13 +315,19 @@ static int netfs_write_folio(struct netfs_io_request=
 *wreq,
 	struct netfs_group *fgroup; /* TODO: Use this with ceph */
 	struct netfs_folio *finfo;
 	size_t fsize =3D folio_size(folio), flen =3D fsize, foff =3D 0;
-	loff_t fpos =3D folio_pos(folio);
+	loff_t fpos =3D folio_pos(folio), i_size;
 	bool to_eof =3D false, streamw =3D false;
 	bool debug =3D false;
 =

 	_enter("");
 =

-	if (fpos >=3D wreq->i_size) {
+	/* netfs_perform_write() may shift i_size around the page or from out
+	 * of the page to beyond it, but cannot move i_size into or through the
+	 * page since we have it locked.
+	 */
+	i_size =3D i_size_read(wreq->inode);
+
+	if (fpos >=3D i_size) {
 		/* mmap beyond eof. */
 		_debug("beyond eof");
 		folio_start_writeback(folio);
@@ -332,6 +338,9 @@ static int netfs_write_folio(struct netfs_io_request *=
wreq,
 		return 0;
 	}
 =

+	if (fpos + fsize > wreq->i_size)
+		wreq->i_size =3D i_size;
+
 	fgroup =3D netfs_folio_group(folio);
 	finfo =3D netfs_folio_info(folio);
 	if (finfo) {
@@ -342,14 +351,14 @@ static int netfs_write_folio(struct netfs_io_request=
 *wreq,
 =

 	if (wreq->origin =3D=3D NETFS_WRITETHROUGH) {
 		to_eof =3D false;
-		if (flen > wreq->i_size - fpos)
-			flen =3D wreq->i_size - fpos;
-	} else if (flen > wreq->i_size - fpos) {
-		flen =3D wreq->i_size - fpos;
+		if (flen > i_size - fpos)
+			flen =3D i_size - fpos;
+	} else if (flen > i_size - fpos) {
+		flen =3D i_size - fpos;
 		if (!streamw)
 			folio_zero_segment(folio, flen, fsize);
 		to_eof =3D true;
-	} else if (flen =3D=3D wreq->i_size - fpos) {
+	} else if (flen =3D=3D i_size - fpos) {
 		to_eof =3D true;
 	}
 	flen -=3D foff;