RDMA/core: Fix null-ptr-deref in ib_core_cleanup()
authorChen Zhongjin <chenzhongjin@huawei.com>
Tue, 25 Oct 2022 02:41:46 +0000 (10:41 +0800)
committerJason Gunthorpe <jgg@nvidia.com>
Fri, 28 Oct 2022 15:59:40 +0000 (12:59 -0300)
KASAN reported a null-ptr-deref error:

  KASAN: null-ptr-deref in range [0x0000000000000118-0x000000000000011f]
  CPU: 1 PID: 379
  Hardware name: QEMU Standard PC (i440FX + PIIX, 1996)
  RIP: 0010:destroy_workqueue+0x2f/0x740
  RSP: 0018:ffff888016137df8 EFLAGS: 00000202
  ...
  Call Trace:
   ib_core_cleanup+0xa/0xa1 [ib_core]
   __do_sys_delete_module.constprop.0+0x34f/0x5b0
   do_syscall_64+0x3a/0x90
   entry_SYSCALL_64_after_hwframe+0x63/0xcd
  RIP: 0033:0x7fa1a0d221b7
  ...

It is because the fail of roce_gid_mgmt_init() is ignored:

 ib_core_init()
   roce_gid_mgmt_init()
     gid_cache_wq = alloc_ordered_workqueue # fail
 ...
 ib_core_cleanup()
   roce_gid_mgmt_cleanup()
     destroy_workqueue(gid_cache_wq)
     # destroy an unallocated wq

Fix this by catching the fail of roce_gid_mgmt_init() in ib_core_init().

Fixes: 03db3a2d81e6 ("IB/core: Add RoCE GID table management")
Signed-off-by: Chen Zhongjin <chenzhongjin@huawei.com>
Link: https://lore.kernel.org/r/20221025024146.109137-1-chenzhongjin@huawei.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
drivers/infiniband/core/device.c
drivers/infiniband/core/nldev.c

index ae60c73..b69e2c4 100644 (file)
@@ -2815,10 +2815,18 @@ static int __init ib_core_init(void)
 
        nldev_init();
        rdma_nl_register(RDMA_NL_LS, ibnl_ls_cb_table);
-       roce_gid_mgmt_init();
+       ret = roce_gid_mgmt_init();
+       if (ret) {
+               pr_warn("Couldn't init RoCE GID management\n");
+               goto err_parent;
+       }
 
        return 0;
 
+err_parent:
+       rdma_nl_unregister(RDMA_NL_LS);
+       nldev_exit();
+       unregister_pernet_device(&rdma_dev_net_ops);
 err_compat:
        unregister_blocking_lsm_notifier(&ibdev_lsm_nb);
 err_sa:
index b92358f..12dc970 100644 (file)
@@ -2537,7 +2537,7 @@ void __init nldev_init(void)
        rdma_nl_register(RDMA_NL_NLDEV, nldev_cb_table);
 }
 
-void __exit nldev_exit(void)
+void nldev_exit(void)
 {
        rdma_nl_unregister(RDMA_NL_NLDEV);
 }