Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Mon, 4 Jun 2018 22:59:36 -0500
From:   "Serge E. Hallyn" <serge@hallyn.com>
To:     Tycho Andersen <tycho@tycho.ws>
Cc:     Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
        Jiri Slaby <jslaby@suse.com>, linux-serial@vger.kernel.org,
        linux-kernel@vger.kernel.org, "Serge E . Hallyn" <serge@hallyn.com>
Subject: Re: [PATCH] uart: fix race between uart_put_char() and
 uart_shutdown()
Message-ID: <20180605035936.GA19642@mail.hallyn.com>
References: <20180605000127.5495-1-tycho@tycho.ws>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: <20180605000127.5495-1-tycho@tycho.ws>
User-Agent: Mutt/1.5.21 (2010-09-15)
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

Quoting Tycho Andersen (tycho@tycho.ws):
> We have reports of the following crash:
> 
>     PID: 7 TASK: ffff88085c6d61c0 CPU: 1 COMMAND: "kworker/u25:0"
>     #0 [ffff88085c6db710] machine_kexec at ffffffff81046239
>     #1 [ffff88085c6db760] crash_kexec at ffffffff810fc248
>     #2 [ffff88085c6db830] oops_end at ffffffff81008ae7
>     #3 [ffff88085c6db860] no_context at ffffffff81050b8f
>     #4 [ffff88085c6db8b0] __bad_area_nosemaphore at ffffffff81050d75
>     #5 [ffff88085c6db900] bad_area_nosemaphore at ffffffff81050e83
>     #6 [ffff88085c6db910] __do_page_fault at ffffffff8105132e
>     #7 [ffff88085c6db9b0] do_page_fault at ffffffff8105152c
>     #8 [ffff88085c6db9c0] page_fault at ffffffff81a3f122
>     [exception RIP: uart_put_char+149]
>     RIP: ffffffff814b67b5 RSP: ffff88085c6dba78 RFLAGS: 00010006
>     RAX: 0000000000000292 RBX: ffffffff827c5120 RCX: 0000000000000081
>     RDX: 0000000000000000 RSI: 000000000000005f RDI: ffffffff827c5120
>     RBP: ffff88085c6dba98 R8: 000000000000012c R9: ffffffff822ea320
>     R10: ffff88085fe4db04 R11: 0000000000000001 R12: ffff881059f9c000
>     R13: 0000000000000001 R14: 000000000000005f R15: 0000000000000fba
>     ORIG_RAX: ffffffffffffffff CS: 0010 SS: 0018
>     #9 [ffff88085c6dbaa0] tty_put_char at ffffffff81497544
>     #10 [ffff88085c6dbac0] do_output_char at ffffffff8149c91c
>     #11 [ffff88085c6dbae0] __process_echoes at ffffffff8149cb8b
>     #12 [ffff88085c6dbb30] commit_echoes at ffffffff8149cdc2
>     #13 [ffff88085c6dbb60] n_tty_receive_buf_fast at ffffffff8149e49b
>     #14 [ffff88085c6dbbc0] __receive_buf at ffffffff8149ef5a
>     #15 [ffff88085c6dbc20] n_tty_receive_buf_common at ffffffff8149f016
>     #16 [ffff88085c6dbca0] n_tty_receive_buf2 at ffffffff8149f194
>     #17 [ffff88085c6dbcb0] flush_to_ldisc at ffffffff814a238a
>     #18 [ffff88085c6dbd50] process_one_work at ffffffff81090be2
>     #19 [ffff88085c6dbe20] worker_thread at ffffffff81091b4d
>     #20 [ffff88085c6dbeb0] kthread at ffffffff81096384
>     #21 [ffff88085c6dbf50] ret_from_fork at ffffffff81a3d69f​
> 
> after slogging through some dissasembly:
> 
> ffffffff814b6720 <uart_put_char>:
> ffffffff814b6720:	55                   	push   %rbp
> ffffffff814b6721:	48 89 e5             	mov    %rsp,%rbp
> ffffffff814b6724:	48 83 ec 20          	sub    $0x20,%rsp
> ffffffff814b6728:	48 89 1c 24          	mov    %rbx,(%rsp)
> ffffffff814b672c:	4c 89 64 24 08       	mov    %r12,0x8(%rsp)
> ffffffff814b6731:	4c 89 6c 24 10       	mov    %r13,0x10(%rsp)
> ffffffff814b6736:	4c 89 74 24 18       	mov    %r14,0x18(%rsp)
> ffffffff814b673b:	e8 b0 8e 58 00       	callq  ffffffff81a3f5f0 <mcount>
> ffffffff814b6740:	4c 8b a7 88 02 00 00 	mov    0x288(%rdi),%r12
> ffffffff814b6747:	45 31 ed             	xor    %r13d,%r13d
> ffffffff814b674a:	41 89 f6             	mov    %esi,%r14d
> ffffffff814b674d:	49 83 bc 24 70 01 00 	cmpq   $0x0,0x170(%r12)
> ffffffff814b6754:	00 00
> ffffffff814b6756:	49 8b 9c 24 80 01 00 	mov    0x180(%r12),%rbx
> ffffffff814b675d:	00
> ffffffff814b675e:	74 2f                	je     ffffffff814b678f <uart_put_char+0x6f>
> ffffffff814b6760:	48 89 df             	mov    %rbx,%rdi
> ffffffff814b6763:	e8 a8 67 58 00       	callq  ffffffff81a3cf10 <_raw_spin_lock_irqsave>
> ffffffff814b6768:	41 8b 8c 24 78 01 00 	mov    0x178(%r12),%ecx
> ffffffff814b676f:	00
> ffffffff814b6770:	89 ca                	mov    %ecx,%edx
> ffffffff814b6772:	f7 d2                	not    %edx
> ffffffff814b6774:	41 03 94 24 7c 01 00 	add    0x17c(%r12),%edx
> ffffffff814b677b:	00
> ffffffff814b677c:	81 e2 ff 0f 00 00    	and    $0xfff,%edx
> ffffffff814b6782:	75 23                	jne    ffffffff814b67a7 <uart_put_char+0x87>
> ffffffff814b6784:	48 89 c6             	mov    %rax,%rsi
> ffffffff814b6787:	48 89 df             	mov    %rbx,%rdi
> ffffffff814b678a:	e8 e1 64 58 00       	callq  ffffffff81a3cc70 <_raw_spin_unlock_irqrestore>
> ffffffff814b678f:	44 89 e8             	mov    %r13d,%eax
> ffffffff814b6792:	48 8b 1c 24          	mov    (%rsp),%rbx
> ffffffff814b6796:	4c 8b 64 24 08       	mov    0x8(%rsp),%r12
> ffffffff814b679b:	4c 8b 6c 24 10       	mov    0x10(%rsp),%r13
> ffffffff814b67a0:	4c 8b 74 24 18       	mov    0x18(%rsp),%r14
> ffffffff814b67a5:	c9                   	leaveq
> ffffffff814b67a6:	c3                   	retq
> ffffffff814b67a7:	49 8b 94 24 70 01 00 	mov    0x170(%r12),%rdx
> ffffffff814b67ae:	00
> ffffffff814b67af:	48 63 c9             	movslq %ecx,%rcx
> ffffffff814b67b2:	41 b5 01             	mov    $0x1,%r13b
> ffffffff814b67b5:	44 88 34 0a          	mov    %r14b,(%rdx,%rcx,1)
> ffffffff814b67b9:	41 8b 94 24 78 01 00 	mov    0x178(%r12),%edx
> ffffffff814b67c0:	00
> ffffffff814b67c1:	83 c2 01             	add    $0x1,%edx
> ffffffff814b67c4:	81 e2 ff 0f 00 00    	and    $0xfff,%edx
> ffffffff814b67ca:	41 89 94 24 78 01 00 	mov    %edx,0x178(%r12)
> ffffffff814b67d1:	00
> ffffffff814b67d2:	eb b0                	jmp    ffffffff814b6784 <uart_put_char+0x64>
> ffffffff814b67d4:	66 66 66 2e 0f 1f 84 	data32 data32 nopw %cs:0x0(%rax,%rax,1)
> ffffffff814b67db:	00 00 00 00 00
> 
> for our build, this is crashing at:
> 
>     circ->buf[circ->head] = c;
> 
> Looking in uart_port_startup(), it seems that circ->buf (state->xmit.buf)
> protected by the "per-port mutex", which based on uart_port_check() is
> state->port.mutex. Indeed, the lock acquired in uart_put_char() is
> uport->lock, i.e. not the same lock.
> 
> Anyway, since the lock is not acquired, if uart_shutdown() is called, the
> last chunk of that function may release state->xmit.buf before its assigned
> to null, and cause the race above.
> 
> To fix it, we simply also acquire state->port.mutex.
> 
> Unfortunately, I don't have any insightful thoughts about how to test this.
> Ideas are appreciated :)

I wonder whether there is something we can do with qemu -serial pipe: ?

> Signed-off-by: Tycho Andersen <tycho@tycho.ws>

Acked-by: Serge Hallyn <serge@hallyn.com>

> ---
>  drivers/tty/serial/serial_core.c | 11 ++++++++++-
>  1 file changed, 10 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/tty/serial/serial_core.c b/drivers/tty/serial/serial_core.c
> index 0466f9f08a91..883a8c15510c 100644
> --- a/drivers/tty/serial/serial_core.c
> +++ b/drivers/tty/serial/serial_core.c
> @@ -532,9 +532,15 @@ static int uart_put_char(struct tty_struct *tty, unsigned char c)
>  	unsigned long flags;
>  	int ret = 0;
>  
> +	/*
> +	 * state->xmit.buf is protected by state->port.mutex, see the note in
> +	 * uart_port_startup()
> +	 */
> +	mutex_lock(&state->port.mutex);
> +
>  	circ = &state->xmit;
>  	if (!circ->buf)
> -		return 0;
> +		goto out;
>  
>  	port = uart_port_lock(state, flags);
>  	if (port && uart_circ_chars_free(circ) != 0) {
> @@ -543,6 +549,9 @@ static int uart_put_char(struct tty_struct *tty, unsigned char c)
>  		ret = 1;
>  	}
>  	uart_port_unlock(port, flags);
> +
> +out:
> +	mutex_unlock(&state->port.mutex);
>  	return ret;
>  }
>  
> -- 
> 2.17.0