From: Luis Henriques <luis@igalia.com>
To: Miklos Szeredi <miklos@szeredi.hu>
Cc: linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org,
Bernd Schubert <bernd@bsbernd.com>,
Luis Henriques <luis@igalia.com>, Teng Qin <tqin@jumptrading.com>
Subject: [RFC PATCH v2] fuse: fix race in fuse_notify_store()
Date: Thu, 30 Jan 2025 10:16:07 +0000 [thread overview]
Message-ID: <20250130101607.21756-1-luis@igalia.com> (raw)
Userspace filesystems can push data for a specific inode without it being
explicitly requested. This can be accomplished by using NOTIFY_STORE.
However, this may race against another process performing different
operations on the same inode.
If, for example, there is a process reading from it, it may happen that it
will block waiting for data to be available (locking the folio), while the
FUSE server will also block trying to lock the same folio to update it with
the inode data.
The easiest solution, as suggested by Miklos, is to allow the userspace
filesystem to skip locked folios.
Link: https://lore.kernel.org/CH2PR14MB41040692ABC50334F500789ED6C89@CH2PR14MB4104.namprd14.prod.outlook.com
Reported-by: Teng Qin <tqin@jumptrading.com>
Originally-by: Miklos Szeredi <miklos@szeredi.hu>
Signed-off-by: Luis Henriques <luis@igalia.com>
---
Hi!
Here's v2. Other than fixing the bug pointed out by Bernd (thanks!), I've
also added an explanation to the 'XXX' comment. As a matter of fact, I've
took another look at that code, and I felt compelled to remove that comment,
as using PAGE_SIZE seems to be the right thing.
Anyway, I'm still thinking that probably NOTIFY_STORE should *always* have
this behaviour, without the need for userspace to explicitly setting a flag.
Changes since v1:
- Only skip if __filemap_get_folio() returns -EAGAIN (Bernd)
fs/fuse/dev.c | 30 +++++++++++++++++++++++-------
include/uapi/linux/fuse.h | 8 +++++++-
2 files changed, 30 insertions(+), 8 deletions(-)
diff --git a/fs/fuse/dev.c b/fs/fuse/dev.c
index 27ccae63495d..309651f82ca4 100644
--- a/fs/fuse/dev.c
+++ b/fs/fuse/dev.c
@@ -1630,6 +1630,7 @@ static int fuse_notify_store(struct fuse_conn *fc, unsigned int size,
unsigned int num;
loff_t file_size;
loff_t end;
+ int fgp_flags = FGP_LOCK | FGP_ACCESSED | FGP_CREAT;
err = -EINVAL;
if (size < sizeof(outarg))
@@ -1645,6 +1646,9 @@ static int fuse_notify_store(struct fuse_conn *fc, unsigned int size,
nodeid = outarg.nodeid;
+ if (outarg.flags & FUSE_NOTIFY_STORE_NOWAIT)
+ fgp_flags |= FGP_NOWAIT;
+
down_read(&fc->killsb);
err = -ENOENT;
@@ -1668,14 +1672,26 @@ static int fuse_notify_store(struct fuse_conn *fc, unsigned int size,
struct page *page;
unsigned int this_num;
- folio = filemap_grab_folio(mapping, index);
- err = PTR_ERR(folio);
- if (IS_ERR(folio))
- goto out_iput;
+ folio = __filemap_get_folio(mapping, index, fgp_flags,
+ mapping_gfp_mask(mapping));
+ err = PTR_ERR_OR_ZERO(folio);
+ if (err) {
+ if (!(outarg.flags & FUSE_NOTIFY_STORE_NOWAIT) ||
+ (err != -EAGAIN))
+ goto out_iput;
+ page = NULL;
+ /* XXX is it OK to use PAGE_SIZE here? */
+ this_num = min_t(unsigned int, num, PAGE_SIZE - offset);
+ } else {
+ page = &folio->page;
+ this_num = min_t(unsigned int, num,
+ folio_size(folio) - offset);
+ }
- page = &folio->page;
- this_num = min_t(unsigned, num, folio_size(folio) - offset);
err = fuse_copy_page(cs, &page, offset, this_num, 0);
+ if (!page)
+ goto skip;
+
if (!folio_test_uptodate(folio) && !err && offset == 0 &&
(this_num == folio_size(folio) || file_size == end)) {
folio_zero_segment(folio, this_num, folio_size(folio));
@@ -1683,7 +1699,7 @@ static int fuse_notify_store(struct fuse_conn *fc, unsigned int size,
}
folio_unlock(folio);
folio_put(folio);
-
+skip:
if (err)
goto out_iput;
diff --git a/include/uapi/linux/fuse.h b/include/uapi/linux/fuse.h
index e9e78292d107..59725f89340e 100644
--- a/include/uapi/linux/fuse.h
+++ b/include/uapi/linux/fuse.h
@@ -576,6 +576,12 @@ struct fuse_file_lock {
*/
#define FUSE_EXPIRE_ONLY (1 << 0)
+/**
+ * notify_store flags
+ * FUSE_NOTIFY_STORE_NOWAIT: skip locked pages
+ */
+#define FUSE_NOTIFY_STORE_NOWAIT (1 << 0)
+
/**
* extension type
* FUSE_MAX_NR_SECCTX: maximum value of &fuse_secctx_header.nr_secctx
@@ -1075,7 +1081,7 @@ struct fuse_notify_store_out {
uint64_t nodeid;
uint64_t offset;
uint32_t size;
- uint32_t padding;
+ uint32_t flags;
};
struct fuse_notify_retrieve_out {
next reply other threads:[~2025-01-30 10:16 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-01-30 10:16 Luis Henriques [this message]
2025-02-21 17:40 ` Luis Henriques
2025-02-24 13:36 ` Miklos Szeredi
2025-02-24 14:30 ` Luis Henriques
2025-02-24 14:39 ` Miklos Szeredi
2025-02-25 10:37 ` Luis Henriques
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=20250130101607.21756-1-luis@igalia.com \
--to=luis@igalia.com \
--cc=bernd@bsbernd.com \
--cc=linux-fsdevel@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=miklos@szeredi.hu \
--cc=tqin@jumptrading.com \
/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®