From: Miklos Szeredi <mszeredi@redhat.com>
To: Al Viro <viro@zeniv.linux.org.uk>,
Linus Torvalds <torvalds@linux-foundation.org>
Cc: linux-unionfs@vger.kernel.org, linux-fsdevel@vger.kernel.org,
linux-kernel@vger.kernel.org,
Andreas Gruenbacher <agruenba@redhat.com>,
garyhuang <zjh.20052005@163.com>
Subject: [PATCH v2 2/2] ovl: enable RCU'd ->get_acl()
Date: Wed, 18 Aug 2021 15:34:00 +0200 [thread overview]
Message-ID: <20210818133400.830078-3-mszeredi@redhat.com> (raw)
In-Reply-To: <20210818133400.830078-1-mszeredi@redhat.com>
Overlayfs does not cache ACL's (to avoid double caching). Instead it just
calls the underlying filesystem's i_op->get_acl(), which will return the
cached value, if possible.
In rcu path walk, however, get_cached_acl_rcu() is employed to get the
value from the cache, which will fail on overlayfs resulting in dropping
out of rcu walk mode. This can result in a big performance hit in certain
situations.
Fix by calling ->get_acl() with LOOKUP_RCU flag in case of ACL_DONT_CACHE
(which indicates pass-through)
Reported-by: garyhuang <zjh.20052005@163.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
---
fs/overlayfs/inode.c | 7 ++++---
fs/posix_acl.c | 8 +++++++-
include/linux/fs.h | 5 +++++
include/linux/posix_acl.h | 3 ++-
4 files changed, 18 insertions(+), 5 deletions(-)
diff --git a/fs/overlayfs/inode.c b/fs/overlayfs/inode.c
index 727154a1d3ce..6a55231b262a 100644
--- a/fs/overlayfs/inode.c
+++ b/fs/overlayfs/inode.c
@@ -13,6 +13,7 @@
#include <linux/fiemap.h>
#include <linux/fileattr.h>
#include <linux/security.h>
+#include <linux/namei.h>
#include "overlayfs.h"
@@ -454,12 +455,12 @@ struct posix_acl *ovl_get_acl(struct inode *inode, int type, int flags)
const struct cred *old_cred;
struct posix_acl *acl;
- if (flags)
- return ERR_PTR(-EINVAL);
-
if (!IS_ENABLED(CONFIG_FS_POSIX_ACL) || !IS_POSIXACL(realinode))
return NULL;
+ if (flags & LOOKUP_RCU)
+ return get_cached_acl_rcu(realinode, type);
+
old_cred = ovl_override_creds(inode->i_sb);
acl = get_acl(realinode, type);
revert_creds(old_cred);
diff --git a/fs/posix_acl.c b/fs/posix_acl.c
index 6b7f793e2b6f..4d1c6c266cf0 100644
--- a/fs/posix_acl.c
+++ b/fs/posix_acl.c
@@ -22,6 +22,7 @@
#include <linux/xattr.h>
#include <linux/export.h>
#include <linux/user_namespace.h>
+#include <linux/namei.h>
static struct posix_acl **acl_by_type(struct inode *inode, int type)
{
@@ -56,7 +57,12 @@ EXPORT_SYMBOL(get_cached_acl);
struct posix_acl *get_cached_acl_rcu(struct inode *inode, int type)
{
- return rcu_dereference(*acl_by_type(inode, type));
+ struct posix_acl *acl = rcu_dereference(*acl_by_type(inode, type));
+
+ if (acl == ACL_DONT_CACHE)
+ acl = inode->i_op->get_acl(inode, type, LOOKUP_RCU);
+
+ return acl;
}
EXPORT_SYMBOL(get_cached_acl_rcu);
diff --git a/include/linux/fs.h b/include/linux/fs.h
index 1c56d4fc4efe..20b7db2d0a85 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -581,6 +581,11 @@ static inline void mapping_allow_writable(struct address_space *mapping)
struct posix_acl;
#define ACL_NOT_CACHED ((void *)(-1))
+/*
+ * ACL_DONT_CACHE is for stacked filesystems, that rely on underlying fs to
+ * cache the ACL. This also means that ->get_acl() can be called in RCU mode
+ * with the LOOKUP_RCU flag.
+ */
#define ACL_DONT_CACHE ((void *)(-3))
static inline struct posix_acl *
diff --git a/include/linux/posix_acl.h b/include/linux/posix_acl.h
index 307094ebb88c..b65c877d92b8 100644
--- a/include/linux/posix_acl.h
+++ b/include/linux/posix_acl.h
@@ -72,6 +72,8 @@ extern struct posix_acl *get_posix_acl(struct inode *, int);
extern int set_posix_acl(struct user_namespace *, struct inode *, int,
struct posix_acl *);
+struct posix_acl *get_cached_acl_rcu(struct inode *inode, int type);
+
#ifdef CONFIG_FS_POSIX_ACL
int posix_acl_chmod(struct user_namespace *, struct inode *, umode_t);
extern int posix_acl_create(struct inode *, umode_t *, struct posix_acl **,
@@ -84,7 +86,6 @@ extern int simple_set_acl(struct user_namespace *, struct inode *,
extern int simple_acl_create(struct inode *, struct inode *);
struct posix_acl *get_cached_acl(struct inode *inode, int type);
-struct posix_acl *get_cached_acl_rcu(struct inode *inode, int type);
void set_cached_acl(struct inode *inode, int type, struct posix_acl *acl);
void forget_cached_acl(struct inode *inode, int type);
void forget_all_cached_acls(struct inode *inode);
--
2.31.1
next prev parent reply other threads:[~2021-08-18 13:34 UTC|newest]
Thread overview: 8+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-08-18 13:33 [PATCH v2 0/2] allow overlayfs to do RCU lookups Miklos Szeredi
2021-08-18 13:33 ` [PATCH v2 1/2] vfs: add flags argument to ->get_acl() callback Miklos Szeredi
2021-08-18 13:49 ` Christian Brauner
2021-08-18 13:34 ` Miklos Szeredi [this message]
2021-08-18 18:34 ` [PATCH v2 2/2] ovl: enable RCU'd ->get_acl() Linus Torvalds
2021-08-18 18:53 ` Miklos Szeredi
2021-08-18 14:26 ` [PATCH v2 0/2] allow overlayfs to do RCU lookups Al Viro
2021-08-18 15:03 ` Miklos Szeredi
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=20210818133400.830078-3-mszeredi@redhat.com \
--to=mszeredi@redhat.com \
--cc=agruenba@redhat.com \
--cc=linux-fsdevel@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-unionfs@vger.kernel.org \
--cc=torvalds@linux-foundation.org \
--cc=viro@zeniv.linux.org.uk \
--cc=zjh.20052005@163.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;
as well as URLs for NNTP newsgroup(s).