Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
From:   Dominique Martinet <asmadeus@codewreck.org>
To:     v9fs-developer@lists.sourceforge.net
Cc:     Leon Romanovsky <leon@kernel.org>, linux_oss@crudebyte.com,
        linux-kernel@vger.kernel.org, syzkaller-bugs@googlegroups.com,
        Dominique Martinet <asmadeus@codewreck.org>,
        syzbot+67d13108d855f451cafc@syzkaller.appspotmail.com
Subject: [PATCH 1/2 v2] 9p: client_create/destroy: only call trans_mod->close after create
Date:   Thu, 29 Sep 2022 06:58:00 +0900
Message-Id: <20220928215800.1749929-1-asmadeus@codewreck.org>
In-Reply-To: <20220928214417.1749609-1-asmadeus@codewreck.org>
References: <20220928214417.1749609-1-asmadeus@codewreck.org>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

destroy code would incorrectly call close() if trans_mod exists after some
hasty code cleanup: we need to make sure we only call close after create

Link: https://lkml.kernel.org/r/00000000000015ac7905e97ebaed@google.com
Reported-by: syzbot+67d13108d855f451cafc@syzkaller.appspotmail.com
Reported-by: Leon Romanovsky <leon@kernel.org>
Fixes: 3ff51294a055 ("9p: p9_client_create: use p9_client_destroy on failure")
Signed-off-by: Dominique Martinet <asmadeus@codewreck.org>
---

As pointed out in later mail, rdma actually does set trans->priv then
fails, so we also need to reset clnt->trans on create errors.

That's getting uglier than I wish it'd be, but the cleanup code I just
trashed away really isn't pretty either so I guess it'll have to do...
At least close() should now really never be called on create failures.

 net/9p/client.c | 11 +++++++++--
 1 file changed, 9 insertions(+), 2 deletions(-)

diff --git a/net/9p/client.c b/net/9p/client.c
index bfa80f01992e..8cf952f2de68 100644
--- a/net/9p/client.c
+++ b/net/9p/client.c
@@ -971,6 +971,7 @@ struct p9_client *p9_client_create(const char *dev_name, char *options)
 	spin_lock_init(&clnt->lock);
 	idr_init(&clnt->fids);
 	idr_init(&clnt->reqs);
+	clnt->trans = ERR_PTR(-EINVAL);
 
 	err = parse_opts(options, clnt);
 	if (err < 0)
@@ -990,8 +991,14 @@ struct p9_client *p9_client_create(const char *dev_name, char *options)
 		 clnt, clnt->trans_mod, clnt->msize, clnt->proto_version);
 
 	err = clnt->trans_mod->create(clnt, dev_name, options);
-	if (err)
+	// ensure clnt->trans is initialized to call close() on destroy
+	// if and only if create succeeded
+	if (err < 0) {
+		clnt->trans = PTR_ERR(err);
 		goto out;
+	}
+	if (IS_ERR(clnt->trans))
+		clnt->trans = NULL;
 
 	if (clnt->msize > clnt->trans_mod->maxsize) {
 		clnt->msize = clnt->trans_mod->maxsize;
@@ -1036,7 +1043,7 @@ void p9_client_destroy(struct p9_client *clnt)
 
 	p9_debug(P9_DEBUG_MUX, "clnt %p\n", clnt);
 
-	if (clnt->trans_mod)
+	if (clnt->trans_mod && !IS_ERR(client->trans))
 		clnt->trans_mod->close(clnt);
 
 	v9fs_put_trans(clnt->trans_mod);
-- 
2.35.1