netfs, 9p: Fix race between umount and async request completion

There's a problem in 9p's interaction with netfslib whereby a crash occurs
because the 9p_fid structs get forcibly destroyed during client teardown
(without paying attention to their refcounts) before netfslib has finished
with them.  However, it's not a simple case of deferring the clunking that
p9_fid_put() does as that requires the p9_client record to still be
present.

The problem is that netfslib has to unlock pages and clear the IN_PROGRESS
flag before destroying the objects involved - including the fid - and, in
any case, nothing checks to see if writeback completed barring looking at
the page flags.

Fix this by keeping a count of outstanding I/O requests (of any type) and
waiting for it to quiesce during inode eviction.

Reported-by: syzbot+df038d463cca332e8414@syzkaller.appspotmail.com
Link: https://lore.kernel.org/all/0000000000005be0aa061846f8d6@google.com/
Reported-by: syzbot+d7c7a495a5e466c031b6@syzkaller.appspotmail.com
Link: https://lore.kernel.org/all/000000000000b86c5e06130da9c6@google.com/
Reported-by: syzbot+1527696d41a634cc1819@syzkaller.appspotmail.com
Link: https://lore.kernel.org/all/000000000000041f960618206d7e@google.com/
Signed-off-by: David Howells <dhowells@redhat.com>
Link: https://lore.kernel.org/r/755891.1716560771@warthog.procyon.org.uk
Tested-by: syzbot+d7c7a495a5e466c031b6@syzkaller.appspotmail.com
Reviewed-by: Dominique Martinet <asmadeus@codewreck.org>
cc: Eric Van Hensbergen <ericvh@kernel.org>
cc: Latchesar Ionkov <lucho@ionkov.net>
cc: Christian Schoenebeck <linux_oss@crudebyte.com>
cc: Jeff Layton <jlayton@kernel.org>
cc: Steve French <sfrench@samba.org>
cc: Hillf Danton <hdanton@sina.com>
cc: v9fs@lists.linux.dev
cc: linux-afs@lists.infradead.org
cc: linux-cifs@vger.kernel.org
cc: netfs@lists.linux.dev
cc: linux-fsdevel@vger.kernel.org
Reported-and-tested-by: syzbot+d7c7a495a5e466c031b6@syzkaller.appspotmail.com
Signed-off-by: Christian Brauner <brauner@kernel.org>

diff --git a/fs/9p/vfs_inode.c b/fs/9p/vfs_inode.c
index 7a3308d776060e2e2565af09d358f2cf33416b6b..fd72fc38c8f5b1ff971414388f6fab16ce53ac4c 100644 (file)
--- a/fs/9p/vfs_inode.c
+++ b/fs/9p/vfs_inode.c
@@ -348,6 +348,7 @@ void v9fs_evict_inode(struct inode *inode)
        __le32 __maybe_unused version;
 
        if (!is_bad_inode(inode)) {
+               netfs_wait_for_outstanding_io(inode);
                truncate_inode_pages_final(&inode->i_data);
 
                version = cpu_to_le32(v9inode->qid.version);

diff --git a/fs/afs/inode.c b/fs/afs/inode.c
index 94fc049aff584f43e622d164a13fc30962dd04f1..15bb7989c387ae59a4e1e4fd49c9b9f99e4f0ecf 100644 (file)
--- a/fs/afs/inode.c
+++ b/fs/afs/inode.c
@@ -648,6 +648,7 @@ void afs_evict_inode(struct inode *inode)
 
        ASSERTCMP(inode->i_ino, ==, vnode->fid.vnode);
 
+       netfs_wait_for_outstanding_io(inode);
        truncate_inode_pages_final(&inode->i_data);
 
        afs_set_cache_aux(vnode, &aux);

diff --git a/fs/netfs/objects.c b/fs/netfs/objects.c
index c90d482b16505d319dede502bcf23c7bed38e254..f4a642727479252de0212f80fe32ef23468e0ebf 100644 (file)
--- a/fs/netfs/objects.c
+++ b/fs/netfs/objects.c
@@ -72,6 +72,7 @@ struct netfs_io_request *netfs_alloc_request(struct address_space *mapping,
                }
        }
 
+       atomic_inc(&ctx->io_count);
        trace_netfs_rreq_ref(rreq->debug_id, 1, netfs_rreq_trace_new);
        netfs_proc_add_rreq(rreq);
        netfs_stat(&netfs_n_rh_rreq);
@@ -124,6 +125,7 @@ static void netfs_free_request(struct work_struct *work)
 {
        struct netfs_io_request *rreq =
                container_of(work, struct netfs_io_request, work);
+       struct netfs_inode *ictx = netfs_inode(rreq->inode);
        unsigned int i;
 
        trace_netfs_rreq(rreq, netfs_rreq_trace_free);
@@ -142,6 +144,9 @@ static void netfs_free_request(struct work_struct *work)
                }
                kvfree(rreq->direct_bv);
        }
+
+       if (atomic_dec_and_test(&ictx->io_count))
+               wake_up_var(&ictx->io_count);
        call_rcu(&rreq->rcu, netfs_free_request_rcu);
 }
 

diff --git a/fs/smb/client/cifsfs.c b/fs/smb/client/cifsfs.c
index ec5b639f421a27e46c894ff977237eb4c994942f..14810ffd15c89a98965128936d5600846ac7c5d1 100644 (file)
--- a/fs/smb/client/cifsfs.c
+++ b/fs/smb/client/cifsfs.c
@@ -431,6 +431,7 @@ cifs_free_inode(struct inode *inode)
 static void
 cifs_evict_inode(struct inode *inode)
 {
+       netfs_wait_for_outstanding_io(inode);
        truncate_inode_pages_final(&inode->i_data);
        if (inode->i_state & I_PINNING_NETFS_WB)
                cifs_fscache_unuse_inode_cookie(inode, true);

diff --git a/include/linux/netfs.h b/include/linux/netfs.h
index ca56a4428043d58392954fbfbedd43178cde7a74..3b22ce0d064cc4c43f006e5497cec1e2fa365dd5 100644 (file)
--- a/include/linux/netfs.h
+++ b/include/linux/netfs.h
@@ -68,6 +68,7 @@ struct netfs_inode {
        loff_t                  remote_i_size;  /* Size of the remote file */
        loff_t                  zero_point;     /* Size after which we assume there's no data
                                                 * on the server */
+       atomic_t                io_count;       /* Number of outstanding reqs */
        unsigned long           flags;
 #define NETFS_ICTX_ODIRECT     0               /* The file has DIO in progress */
 #define NETFS_ICTX_UNBUFFERED  1               /* I/O should not use the pagecache */
@@ -472,6 +473,7 @@ static inline void netfs_inode_init(struct netfs_inode *ctx,
        ctx->remote_i_size = i_size_read(&ctx->inode);
        ctx->zero_point = LLONG_MAX;
        ctx->flags = 0;
+       atomic_set(&ctx->io_count, 0);
 #if IS_ENABLED(CONFIG_FSCACHE)
        ctx->cache = NULL;
 #endif
@@ -515,4 +517,20 @@ static inline struct fscache_cookie *netfs_i_cookie(struct netfs_inode *ctx)
 #endif
 }
 
+/**
+ * netfs_wait_for_outstanding_io - Wait for outstanding I/O to complete
+ * @ctx: The netfs inode to wait on
+ *
+ * Wait for outstanding I/O requests of any type to complete.  This is intended
+ * to be called from inode eviction routines.  This makes sure that any
+ * resources held by those requests are cleaned up before we let the inode get
+ * cleaned up.
+ */
+static inline void netfs_wait_for_outstanding_io(struct inode *inode)
+{
+       struct netfs_inode *ictx = netfs_inode(inode);
+
+       wait_var_event(&ictx->io_count, atomic_read(&ictx->io_count) == 0);
+}
+
 #endif /* _LINUX_NETFS_H */