From: libaokun@huaweicloud.com
To: netfs@lists.linux.dev, dhowells@redhat.com, jlayton@kernel.org
Cc: libaokun@huaweicloud.com, yangerkun@huawei.com,
linux-kernel@vger.kernel.org, linux-fsdevel@vger.kernel.org,
hsiangkao@linux.alibaba.com, linux-erofs@lists.ozlabs.org,
yukuai3@huawei.com
Subject: [PATCH v2 07/12] cachefiles: add spin_lock for cachefiles_ondemand_info
Date: Wed, 15 May 2024 16:45:56 +0800 [thread overview]
Message-ID: <20240515084601.3240503-8-libaokun@huaweicloud.com> (raw)
In-Reply-To: <20240515084601.3240503-1-libaokun@huaweicloud.com>
From: Baokun Li <libaokun1@huawei.com>
The following concurrency may cause a read request to fail to be completed
and result in a hung:
t1 | t2
---------------------------------------------------------
cachefiles_ondemand_copen
req = xa_erase(&cache->reqs, id)
// Anon fd is maliciously closed.
cachefiles_ondemand_fd_release
xa_lock(&cache->reqs)
cachefiles_ondemand_set_object_close(object)
xa_unlock(&cache->reqs)
cachefiles_ondemand_set_object_open
// No one will ever close it again.
cachefiles_ondemand_daemon_read
cachefiles_ondemand_select_req
// Get a read req but its fd is already closed.
// The daemon can't issue a cread ioctl with an closed fd, then hung.
So add spin_lock for cachefiles_ondemand_info to protect ondemand_id and
state, thus we can avoid the above problem in cachefiles_ondemand_copen()
by using ondemand_id to determine if fd has been closed.
Fixes: c8383054506c ("cachefiles: notify the user daemon when looking up cookie")
Signed-off-by: Baokun Li <libaokun1@huawei.com>
---
fs/cachefiles/internal.h | 1 +
fs/cachefiles/ondemand.c | 35 ++++++++++++++++++++++++++++++++++-
2 files changed, 35 insertions(+), 1 deletion(-)
diff --git a/fs/cachefiles/internal.h b/fs/cachefiles/internal.h
index 7745b8abc3aa..45c8bed60538 100644
--- a/fs/cachefiles/internal.h
+++ b/fs/cachefiles/internal.h
@@ -55,6 +55,7 @@ struct cachefiles_ondemand_info {
int ondemand_id;
enum cachefiles_object_state state;
struct cachefiles_object *object;
+ spinlock_t lock;
};
/*
diff --git a/fs/cachefiles/ondemand.c b/fs/cachefiles/ondemand.c
index 898fab68332b..d04ddc6576e3 100644
--- a/fs/cachefiles/ondemand.c
+++ b/fs/cachefiles/ondemand.c
@@ -16,13 +16,16 @@ static int cachefiles_ondemand_fd_release(struct inode *inode,
struct cachefiles_object *object = file->private_data;
struct cachefiles_cache *cache = object->volume->cache;
struct cachefiles_ondemand_info *info = object->ondemand;
- int object_id = info->ondemand_id;
+ int object_id;
struct cachefiles_req *req;
XA_STATE(xas, &cache->reqs, 0);
xa_lock(&cache->reqs);
+ spin_lock(&info->lock);
+ object_id = info->ondemand_id;
info->ondemand_id = CACHEFILES_ONDEMAND_ID_CLOSED;
cachefiles_ondemand_set_object_close(object);
+ spin_unlock(&info->lock);
/* Only flush CACHEFILES_REQ_NEW marked req to avoid race with daemon_read */
xas_for_each_marked(&xas, req, ULONG_MAX, CACHEFILES_REQ_NEW) {
@@ -127,6 +130,7 @@ int cachefiles_ondemand_copen(struct cachefiles_cache *cache, char *args)
{
struct cachefiles_req *req;
struct fscache_cookie *cookie;
+ struct cachefiles_ondemand_info *info;
char *pid, *psize;
unsigned long id;
long size;
@@ -185,6 +189,33 @@ int cachefiles_ondemand_copen(struct cachefiles_cache *cache, char *args)
goto out;
}
+ info = req->object->ondemand;
+ spin_lock(&info->lock);
+ /*
+ * The anonymous fd was closed before copen ? Fail the request.
+ *
+ * t1 | t2
+ * ---------------------------------------------------------
+ * cachefiles_ondemand_copen
+ * req = xa_erase(&cache->reqs, id)
+ * // Anon fd is maliciously closed.
+ * cachefiles_ondemand_fd_release
+ * xa_lock(&cache->reqs)
+ * cachefiles_ondemand_set_object_close(object)
+ * xa_unlock(&cache->reqs)
+ * cachefiles_ondemand_set_object_open
+ * // No one will ever close it again.
+ * cachefiles_ondemand_daemon_read
+ * cachefiles_ondemand_select_req
+ *
+ * Get a read req but its fd is already closed. The daemon can't
+ * issue a cread ioctl with an closed fd, then hung.
+ */
+ if (info->ondemand_id == CACHEFILES_ONDEMAND_ID_CLOSED) {
+ spin_unlock(&info->lock);
+ req->error = -EBADFD;
+ goto out;
+ }
cookie = req->object->cookie;
cookie->object_size = size;
if (size)
@@ -194,6 +225,7 @@ int cachefiles_ondemand_copen(struct cachefiles_cache *cache, char *args)
trace_cachefiles_ondemand_copen(req->object, id, size);
cachefiles_ondemand_set_object_open(req->object);
+ spin_unlock(&info->lock);
wake_up_all(&cache->daemon_pollwq);
out:
@@ -596,6 +628,7 @@ int cachefiles_ondemand_init_obj_info(struct cachefiles_object *object,
return -ENOMEM;
object->ondemand->object = object;
+ spin_lock_init(&object->ondemand->lock);
INIT_WORK(&object->ondemand->ondemand_work, ondemand_object_worker);
return 0;
}
--
2.39.2
next prev parent reply other threads:[~2024-05-15 8:56 UTC|newest]
Thread overview: 38+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-05-15 8:45 [PATCH v2 00/12] cachefiles: some bugfixes and cleanups for ondemand requests libaokun
2024-05-15 8:45 ` [PATCH v2 01/12] cachefiles: remove request from xarry during flush requests libaokun
2024-05-20 2:20 ` Gao Xiang
2024-05-20 4:11 ` Baokun Li
2024-05-20 7:09 ` Jingbo Xu
2024-05-15 8:45 ` [PATCH v2 02/12] cachefiles: remove err_put_fd tag in cachefiles_ondemand_daemon_read() libaokun
2024-05-20 2:23 ` Gao Xiang
2024-05-20 4:15 ` Baokun Li
2024-05-15 8:45 ` [PATCH v2 03/12] cachefiles: fix slab-use-after-free in cachefiles_ondemand_get_fd() libaokun
2024-05-20 7:24 ` Jingbo Xu
2024-05-20 8:38 ` Baokun Li
2024-05-20 8:45 ` Gao Xiang
2024-05-20 9:10 ` Jingbo Xu
2024-05-20 9:19 ` Baokun Li
2024-05-20 12:22 ` Baokun Li
2024-05-20 8:06 ` Jingbo Xu
2024-05-20 9:10 ` Baokun Li
2024-05-15 8:45 ` [PATCH v2 04/12] cachefiles: fix slab-use-after-free in cachefiles_ondemand_daemon_read() libaokun
2024-05-20 7:36 ` Jingbo Xu
2024-05-20 8:56 ` Baokun Li
2024-05-15 8:45 ` [PATCH v2 05/12] cachefiles: add output string to cachefiles_obj_[get|put]_ondemand_fd libaokun
2024-05-20 7:40 ` Jingbo Xu
2024-05-20 9:02 ` Baokun Li
2024-05-15 8:45 ` [PATCH v2 06/12] cachefiles: add consistency check for copen/cread libaokun
2024-05-15 8:45 ` libaokun [this message]
2024-05-15 8:45 ` [PATCH v2 08/12] cachefiles: never get a new anonymous fd if ondemand_id is valid libaokun
2024-05-20 8:43 ` Jingbo Xu
2024-05-20 9:07 ` Baokun Li
2024-05-20 9:24 ` Jingbo Xu
2024-05-20 11:14 ` Baokun Li
2024-05-20 11:24 ` Gao Xiang
2024-05-15 8:45 ` [PATCH v2 09/12] cachefiles: defer exposing anon_fd until after copy_to_user() succeeds libaokun
2024-05-20 9:39 ` Jingbo Xu
2024-05-20 11:36 ` Baokun Li
2024-05-15 8:45 ` [PATCH v2 10/12] cachefiles: Set object to close if ondemand_id < 0 in copen libaokun
2024-05-15 8:46 ` [PATCH v2 11/12] cachefiles: flush all requests after setting CACHEFILES_DEAD libaokun
2024-05-15 8:46 ` [PATCH v2 12/12] cachefiles: make on-demand read killable libaokun
2024-05-19 10:56 ` [PATCH v2 00/12] cachefiles: some bugfixes and cleanups for ondemand requests Jeff Layton
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=20240515084601.3240503-8-libaokun@huaweicloud.com \
--to=libaokun@huaweicloud.com \
--cc=dhowells@redhat.com \
--cc=hsiangkao@linux.alibaba.com \
--cc=jlayton@kernel.org \
--cc=linux-erofs@lists.ozlabs.org \
--cc=linux-fsdevel@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=netfs@lists.linux.dev \
--cc=yangerkun@huawei.com \
--cc=yukuai3@huawei.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 read-only IMAP folder(s) and NNTP newsgroup(s).