2020-06-13 12:55:48

by Xiyu Yang

[permalink] [raw]
Subject: [PATCH] tty: serial_core: Fix uart_state refcnt leak when the port startup

uart_port_startup() invokes uart_port_lock(), which returns a reference
of the uart_port object if increases the refcount of the uart_state
object successfully or returns NULL if fails.

However, uart_port_startup() don't take the return value of
uart_port_lock() as the new uart_port object to "uport" and use the old
"uport" instead to balance refcount in uart_port_unlock(), which may
cause a redundant decrement of refcount occurred when the new "uport"
equals to NULL and then cause a potential memory leak.

Fix this issue by update the "uport" object to the return value of
uart_port_lock() when invoking uart_port_lock().

Signed-off-by: Xiyu Yang <[email protected]>
Signed-off-by: Xin Tan <[email protected]>
---
drivers/tty/serial/serial_core.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/tty/serial/serial_core.c b/drivers/tty/serial/serial_core.c
index 57840cf90388..968fd619aec0 100644
--- a/drivers/tty/serial/serial_core.c
+++ b/drivers/tty/serial/serial_core.c
@@ -205,7 +205,7 @@ static int uart_port_startup(struct tty_struct *tty, struct uart_state *state,
if (!page)
return -ENOMEM;

- uart_port_lock(state, flags);
+ uport = uart_port_lock(state, flags);
if (!state->xmit.buf) {
state->xmit.buf = (unsigned char *) page;
uart_circ_clear(&state->xmit);
--
2.7.4


2020-06-24 09:35:47

by Greg Kroah-Hartman

[permalink] [raw]
Subject: Re: [PATCH] tty: serial_core: Fix uart_state refcnt leak when the port startup

On Sat, Jun 13, 2020 at 08:52:18PM +0800, Xiyu Yang wrote:
> uart_port_startup() invokes uart_port_lock(), which returns a reference
> of the uart_port object if increases the refcount of the uart_state
> object successfully or returns NULL if fails.
>
> However, uart_port_startup() don't take the return value of
> uart_port_lock() as the new uart_port object to "uport" and use the old
> "uport" instead to balance refcount in uart_port_unlock(), which may
> cause a redundant decrement of refcount occurred when the new "uport"
> equals to NULL and then cause a potential memory leak.
>
> Fix this issue by update the "uport" object to the return value of
> uart_port_lock() when invoking uart_port_lock().
>
> Signed-off-by: Xiyu Yang <[email protected]>
> Signed-off-by: Xin Tan <[email protected]>
> ---
> drivers/tty/serial/serial_core.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/tty/serial/serial_core.c b/drivers/tty/serial/serial_core.c
> index 57840cf90388..968fd619aec0 100644
> --- a/drivers/tty/serial/serial_core.c
> +++ b/drivers/tty/serial/serial_core.c
> @@ -205,7 +205,7 @@ static int uart_port_startup(struct tty_struct *tty, struct uart_state *state,
> if (!page)
> return -ENOMEM;
>
> - uart_port_lock(state, flags);
> + uport = uart_port_lock(state, flags);

How is this a different pointer than you originally had?

And if it is a different pointer, shouldn't you be calling this function
and using the pointer much earlier in the function instead of just here?

Can you trigger a problem that this patch solves? If so, how?

thanks,

greg k-h

2020-06-24 09:48:34

by Jiri Slaby

[permalink] [raw]
Subject: Re: [PATCH] tty: serial_core: Fix uart_state refcnt leak when the port startup

On 24. 06. 20, 11:34, Greg Kroah-Hartman wrote:
> On Sat, Jun 13, 2020 at 08:52:18PM +0800, Xiyu Yang wrote:
>> uart_port_startup() invokes uart_port_lock(), which returns a reference
>> of the uart_port object if increases the refcount of the uart_state
>> object successfully or returns NULL if fails.
>>
>> However, uart_port_startup() don't take the return value of
>> uart_port_lock() as the new uart_port object to "uport" and use the old
>> "uport" instead to balance refcount in uart_port_unlock(), which may
>> cause a redundant decrement of refcount occurred when the new "uport"
>> equals to NULL and then cause a potential memory leak.
>>
>> Fix this issue by update the "uport" object to the return value of
>> uart_port_lock() when invoking uart_port_lock().
>>
>> Signed-off-by: Xiyu Yang <[email protected]>
>> Signed-off-by: Xin Tan <[email protected]>
>> ---
>> drivers/tty/serial/serial_core.c | 2 +-
>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/drivers/tty/serial/serial_core.c b/drivers/tty/serial/serial_core.c
>> index 57840cf90388..968fd619aec0 100644
>> --- a/drivers/tty/serial/serial_core.c
>> +++ b/drivers/tty/serial/serial_core.c
>> @@ -205,7 +205,7 @@ static int uart_port_startup(struct tty_struct *tty, struct uart_state *state,
>> if (!page)
>> return -ENOMEM;
>>
>> - uart_port_lock(state, flags);
>> + uport = uart_port_lock(state, flags);
>
> How is this a different pointer than you originally had?

Was this patch sent twice? As I had very same questions on the other
one, but never received a feedback:
https://lore.kernel.org/linux-serial/[email protected]/


Oh, wait: this is uart_port_startup, I commented on the uart_shutdown
one. But whatever, I would scratch both of them.

> And if it is a different pointer, shouldn't you be calling this function
> and using the pointer much earlier in the function instead of just here?
>
> Can you trigger a problem that this patch solves? If so, how?

thanks,
--
js
suse labs

2020-06-27 11:46:52

by Greg Kroah-Hartman

[permalink] [raw]
Subject: Re: [PATCH] tty: serial_core: Fix uart_state refcnt leak when the port startup

On Wed, Jun 24, 2020 at 11:42:59AM +0200, Jiri Slaby wrote:
> On 24. 06. 20, 11:34, Greg Kroah-Hartman wrote:
> > On Sat, Jun 13, 2020 at 08:52:18PM +0800, Xiyu Yang wrote:
> >> uart_port_startup() invokes uart_port_lock(), which returns a reference
> >> of the uart_port object if increases the refcount of the uart_state
> >> object successfully or returns NULL if fails.
> >>
> >> However, uart_port_startup() don't take the return value of
> >> uart_port_lock() as the new uart_port object to "uport" and use the old
> >> "uport" instead to balance refcount in uart_port_unlock(), which may
> >> cause a redundant decrement of refcount occurred when the new "uport"
> >> equals to NULL and then cause a potential memory leak.
> >>
> >> Fix this issue by update the "uport" object to the return value of
> >> uart_port_lock() when invoking uart_port_lock().
> >>
> >> Signed-off-by: Xiyu Yang <[email protected]>
> >> Signed-off-by: Xin Tan <[email protected]>
> >> ---
> >> drivers/tty/serial/serial_core.c | 2 +-
> >> 1 file changed, 1 insertion(+), 1 deletion(-)
> >>
> >> diff --git a/drivers/tty/serial/serial_core.c b/drivers/tty/serial/serial_core.c
> >> index 57840cf90388..968fd619aec0 100644
> >> --- a/drivers/tty/serial/serial_core.c
> >> +++ b/drivers/tty/serial/serial_core.c
> >> @@ -205,7 +205,7 @@ static int uart_port_startup(struct tty_struct *tty, struct uart_state *state,
> >> if (!page)
> >> return -ENOMEM;
> >>
> >> - uart_port_lock(state, flags);
> >> + uport = uart_port_lock(state, flags);
> >
> > How is this a different pointer than you originally had?
>
> Was this patch sent twice? As I had very same questions on the other
> one, but never received a feedback:
> https://lore.kernel.org/linux-serial/[email protected]/
>
>
> Oh, wait: this is uart_port_startup, I commented on the uart_shutdown
> one. But whatever, I would scratch both of them.

Yeah, you are right, dropping them both now, thanks.

greg k-h