linux-kernel - Re: [PATCH 4/6] usbip: fix stub_dev usbip_sockfd

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <5df3d221-9e78-4cbe-826b-81cbfc4d5888@i-love.sakura.ne.jp>
Date:   Mon, 8 Mar 2021 19:10:42 +0900
From:   Tetsuo Handa <penguin-kernel@...ove.sakura.ne.jp>
To:     Shuah Khan <skhan@...uxfoundation.org>, shuah@...nel.org,
        valentina.manea.m@...il.com, gregkh@...uxfoundation.org
Cc:     linux-usb@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH 4/6] usbip: fix stub_dev usbip_sockfd_store() races
 leading to gpf

On 2021/03/08 16:35, Tetsuo Handa wrote:
> On 2021/03/08 12:53, Shuah Khan wrote:
>> Fix the above problems:
>> - Stop using kthread_get_run() macro to create/start threads.
>> - Create threads and get task struct reference.
>> - Add kthread_create() failure handling and bail out.
>> - Hold usbip_device lock to update local and shared states after
>>   creating rx and tx threads.
>> - Update usbip_device status to SDEV_ST_USED.
>> - Update usbip_device tcp_socket, sockfd, tcp_rx, and tcp_tx
>> - Start threads after usbip_device (tcp_socket, sockfd, tcp_rx, tcp_tx,
>>   and status) is complete.
> 
> No, the whole usbip_sockfd_store() etc. should be serialized using a mutex,
> for two different threads can open same file and write the same content at
> the same moment. This results in seeing SDEV_ST_AVAILABLE and creating two
> threads and overwiting global variables and setting SDEV_ST_USED and starting
> two threads by each of two thread, which will later fail to call kthread_stop()
> on one of two thread because global variables are overwritten.
> 
> kthread_crate() (which involves GFP_KERNEL allocation) can take long time
> enough to hit
> 
>   usbip_sockfd_store() must perform
> 
>       if (sdev->ud.status != SDEV_ST_AVAILABLE) {

Oops. This is

	if (sdev->ud.status == SDEV_ST_AVAILABLE) {

of course.

>         /* misc assignments for attach operation */
>         sdev->ud.status = SDEV_ST_USED;
>       }
> 
>   under a lock, or multiple ud->tcp_{tx,rx} are created (which will later
>   cause a crash like [1]) and refcount on ud->tcp_socket is leaked when
>   usbip_sockfd_store() is concurrently called.
> 
> problem. That's why my patch introduced usbip_event_mutex lock.
> 

And I think that same serialization is required between "rh_port_connect() from attach_store()" and
"rh_port_disconnect() from vhci_shutdown_connection() via usbip_event_add(&vdev->ud, VDEV_EVENT_DOWN)
 from vhci_port_disconnect() from detach_store()", for both vhci_rx_pdu() from vhci_rx_loop() and
vhci_port_disconnect() from detach_store() can queue VDEV_EVENT_DOWN event which can be processed
without waiting for attach_store() to complete.