Several mutexes are taken while setting up console serial ports. In
particular, the tty_port->mutex and @console_mutex are taken:
serial_pnp_probe
serial8250_register_8250_port
uart_add_one_port (locks tty_port->mutex)
uart_configure_port
register_console (locks @console_mutex)
In order to synchronize kgdb's tty_find_polling_driver() with
register_console(), commit 6193bc90849a ("tty: serial: kgdboc:
synchronize tty_find_polling_driver() and register_console()") takes
the @console_mutex. However, this leads to the following call chain
(with locking):
platform_probe
kgdboc_probe
configure_kgdboc (locks @console_mutex)
tty_find_polling_driver
uart_poll_init (locks tty_port->mutex)
uart_set_options
This is clearly deadlock potential due to the reverse lock ordering.
Since uart_set_options() requires holding @console_mutex in order to
serialize early initialization of the serial-console lock, take the
@console_mutex in uart_poll_init() instead of configure_kgdboc().
Since configure_kgdboc() was using @console_mutex for safe traversal
of the console list, change it to use the SRCU iterator instead.
Add comments to uart_set_options() kerneldoc mentioning that it
requires holding @console_mutex (aka the console_list_lock).
Fixes: 6193bc90849a ("tty: serial: kgdboc: synchronize tty_find_polling_driver() and register_console()")
Signed-off-by: John Ogness <[email protected]>
---
drivers/tty/serial/kgdboc.c | 16 ++++------------
drivers/tty/serial/serial_core.c | 5 +++++
2 files changed, 9 insertions(+), 12 deletions(-)
diff --git a/drivers/tty/serial/kgdboc.c b/drivers/tty/serial/kgdboc.c
index a3ed9b34e2ab..f7df46cc1890 100644
--- a/drivers/tty/serial/kgdboc.c
+++ b/drivers/tty/serial/kgdboc.c
@@ -171,6 +171,7 @@ static int configure_kgdboc(void)
int err = -ENODEV;
char *cptr = config;
struct console *cons;
+ int cookie;
if (!strlen(config) || isspace(config[0])) {
err = 0;
@@ -189,15 +190,6 @@ static int configure_kgdboc(void)
if (kgdboc_register_kbd(&cptr))
goto do_register;
- /*
- * tty_find_polling_driver() can call uart_set_options()
- * (via poll_init) to configure the uart. Take the console_list_lock
- * in order to synchronize against register_console(), which can also
- * configure the uart via uart_set_options(). This also allows safe
- * traversal of the console list.
- */
- console_list_lock();
-
p = tty_find_polling_driver(cptr, &tty_line);
if (!p) {
console_list_unlock();
@@ -211,7 +203,8 @@ static int configure_kgdboc(void)
*/
console_lock();
- for_each_console(cons) {
+ cookie = console_srcu_read_lock();
+ for_each_console_srcu(cons) {
int idx;
if (cons->device && cons->device(cons, &idx) == p &&
idx == tty_line) {
@@ -219,11 +212,10 @@ static int configure_kgdboc(void)
break;
}
}
+ console_srcu_read_unlock(cookie);
console_unlock();
- console_list_unlock();
-
kgdb_tty_driver = p;
kgdb_tty_line = tty_line;
diff --git a/drivers/tty/serial/serial_core.c b/drivers/tty/serial/serial_core.c
index b9fbbee598b8..ec874f3a567c 100644
--- a/drivers/tty/serial/serial_core.c
+++ b/drivers/tty/serial/serial_core.c
@@ -2212,6 +2212,9 @@ EXPORT_SYMBOL_GPL(uart_parse_options);
* @parity: parity character - 'n' (none), 'o' (odd), 'e' (even)
* @bits: number of data bits
* @flow: flow control character - 'r' (rts)
+ *
+ * Locking: Caller must hold console_list_lock in order to serialize
+ * early initialization of the serial-console lock.
*/
int
uart_set_options(struct uart_port *port, struct console *co,
@@ -2619,7 +2622,9 @@ static int uart_poll_init(struct tty_driver *driver, int line, char *options)
if (!ret && options) {
uart_parse_options(options, &baud, &parity, &bits, &flow);
+ console_list_lock();
ret = uart_set_options(port, NULL, baud, parity, bits, flow);
+ console_list_unlock();
}
out:
mutex_unlock(&tport->mutex);
base-commit: b7bfaa761d760e72a969d116517eaa12e404c262
--
2.30.2
On Wed 2023-01-11 15:57:10, John Ogness wrote:
> Several mutexes are taken while setting up console serial ports. In
> particular, the tty_port->mutex and @console_mutex are taken:
>
> serial_pnp_probe
> serial8250_register_8250_port
> uart_add_one_port (locks tty_port->mutex)
> uart_configure_port
> register_console (locks @console_mutex)
>
> In order to synchronize kgdb's tty_find_polling_driver() with
> register_console(), commit 6193bc90849a ("tty: serial: kgdboc:
> synchronize tty_find_polling_driver() and register_console()") takes
> the @console_mutex. However, this leads to the following call chain
> (with locking):
>
> platform_probe
> kgdboc_probe
> configure_kgdboc (locks @console_mutex)
> tty_find_polling_driver
> uart_poll_init (locks tty_port->mutex)
> uart_set_options
>
> This is clearly deadlock potential due to the reverse lock ordering.
Great catch!
> Since uart_set_options() requires holding @console_mutex in order to
> serialize early initialization of the serial-console lock, take the
> @console_mutex in uart_poll_init() instead of configure_kgdboc().
>
> Since configure_kgdboc() was using @console_mutex for safe traversal
> of the console list, change it to use the SRCU iterator instead.
>
> Add comments to uart_set_options() kerneldoc mentioning that it
> requires holding @console_mutex (aka the console_list_lock).
>
> Fixes: 6193bc90849a ("tty: serial: kgdboc: synchronize tty_find_polling_driver() and register_console()")
> Signed-off-by: John Ogness <[email protected]>
> --- a/drivers/tty/serial/kgdboc.c
> +++ b/drivers/tty/serial/kgdboc.c
> @@ -189,15 +190,6 @@ static int configure_kgdboc(void)
> if (kgdboc_register_kbd(&cptr))
> goto do_register;
>
> - /*
> - * tty_find_polling_driver() can call uart_set_options()
> - * (via poll_init) to configure the uart. Take the console_list_lock
> - * in order to synchronize against register_console(), which can also
> - * configure the uart via uart_set_options(). This also allows safe
> - * traversal of the console list.
> - */
> - console_list_lock();
> -
> p = tty_find_polling_driver(cptr, &tty_line);
> if (!p) {
> console_list_unlock();
This console_list_unlock() should be removed here as well.
Best Regards,
Petr
PS: I still have to double check all the dependencies. I think that I need
to do it with a fresh head tomorrow.
On Wed 2023-01-11 17:23:55, Petr Mladek wrote:
> On Wed 2023-01-11 15:57:10, John Ogness wrote:
> > Several mutexes are taken while setting up console serial ports. In
> > particular, the tty_port->mutex and @console_mutex are taken:
> >
> > serial_pnp_probe
> > serial8250_register_8250_port
> > uart_add_one_port (locks tty_port->mutex)
> > uart_configure_port
> > register_console (locks @console_mutex)
> >
> > In order to synchronize kgdb's tty_find_polling_driver() with
> > register_console(), commit 6193bc90849a ("tty: serial: kgdboc:
> > synchronize tty_find_polling_driver() and register_console()") takes
> > the @console_mutex. However, this leads to the following call chain
> > (with locking):
> >
> > platform_probe
> > kgdboc_probe
> > configure_kgdboc (locks @console_mutex)
> > tty_find_polling_driver
> > uart_poll_init (locks tty_port->mutex)
> > uart_set_options
> >
> > This is clearly deadlock potential due to the reverse lock ordering.
>
> Great catch!
>
> > Since uart_set_options() requires holding @console_mutex in order to
> > serialize early initialization of the serial-console lock, take the
> > @console_mutex in uart_poll_init() instead of configure_kgdboc().
> >
> > Since configure_kgdboc() was using @console_mutex for safe traversal
> > of the console list, change it to use the SRCU iterator instead.
> >
> > Add comments to uart_set_options() kerneldoc mentioning that it
> > requires holding @console_mutex (aka the console_list_lock).
> >
> > Fixes: 6193bc90849a ("tty: serial: kgdboc: synchronize tty_find_polling_driver() and register_console()")
> > Signed-off-by: John Ogness <[email protected]>
>
> > --- a/drivers/tty/serial/kgdboc.c
> > +++ b/drivers/tty/serial/kgdboc.c
> > @@ -189,15 +190,6 @@ static int configure_kgdboc(void)
> > if (kgdboc_register_kbd(&cptr))
> > goto do_register;
> >
> > - /*
> > - * tty_find_polling_driver() can call uart_set_options()
> > - * (via poll_init) to configure the uart. Take the console_list_lock
> > - * in order to synchronize against register_console(), which can also
> > - * configure the uart via uart_set_options(). This also allows safe
> > - * traversal of the console list.
> > - */
> > - console_list_lock();
> > -
> > p = tty_find_polling_driver(cptr, &tty_line);
> > if (!p) {
> > console_list_unlock();
>
> This console_list_unlock() should be removed here as well.
This seems to be the only problem. Otherwise, the patch looks good to
me.
After removing the superfluous console_list_unlock():
Reviewed-by: Petr Mladek <[email protected]>
John, would you mind to send v2, please?
Best Regards,
Petr
On (23/01/12 16:40), Petr Mladek wrote:
> On Wed 2023-01-11 17:23:55, Petr Mladek wrote:
> > On Wed 2023-01-11 15:57:10, John Ogness wrote:
> > > Several mutexes are taken while setting up console serial ports. In
> > > particular, the tty_port->mutex and @console_mutex are taken:
> > >
> > > serial_pnp_probe
> > > serial8250_register_8250_port
> > > uart_add_one_port (locks tty_port->mutex)
> > > uart_configure_port
> > > register_console (locks @console_mutex)
> > >
> > > In order to synchronize kgdb's tty_find_polling_driver() with
> > > register_console(), commit 6193bc90849a ("tty: serial: kgdboc:
> > > synchronize tty_find_polling_driver() and register_console()") takes
> > > the @console_mutex. However, this leads to the following call chain
> > > (with locking):
> > >
> > > platform_probe
> > > kgdboc_probe
> > > configure_kgdboc (locks @console_mutex)
> > > tty_find_polling_driver
> > > uart_poll_init (locks tty_port->mutex)
> > > uart_set_options
> > >
> > > This is clearly deadlock potential due to the reverse lock ordering.
[..]
> > > - console_list_lock();
> > > -
> > > p = tty_find_polling_driver(cptr, &tty_line);
> > > if (!p) {
> > > console_list_unlock();
> >
> > This console_list_unlock() should be removed here as well.
>
> This seems to be the only problem. Otherwise, the patch looks good to
> me.
Good catch.
Reviewed-by: Sergey Senozhatsky <[email protected]>