Skip to contentSkip to conversation

tty: serial: meson_uart: Init port lock early

PATCH In review Marc Zyngier 2 participants 5 messages Jul 5, 2020

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 tty: serial: meson_uart: Init port lock early

The meson UART driver triggers a lockdep splat at boot time, due to the new expectation that the driver has to initialize the per-port spinlock itself.

It remains unclear why a double initialization of the port spinlock is a desirable outcome, but in the meantime let's fix the splat.

Fixes a3cb39d258ef ("serial: core: Allow detach and attach serial device for console")
Cc Andy Shevchenko <[email protected]>
Cc Greg Kroah-Hartman <[email protected]>
---
 drivers/tty/serial/meson_uart.c | 3 +++
 1 file changed, 3 insertions(+)
drivers/tty/serial/meson_uart.c +3 -0
index d2c08b760f83..386e39c90628 100644
@@ -759,6 +759,9 @@ static int meson_uart_probe(struct platform_device *pdev)
 	if (ret)
 		return ret;
 
+	/* Init the spinlock early in case this is the console */
+	spin_lock_init(&port->lock);
+
 	port->iotype = UPIO_MEM;
 	port->mapbase = res_mem->start;
 	port->mapsize = resource_size(res_mem);
Signature · 1 line
2.26.2
#1 lore raw
Quoted 9 lines · Marc Zyngier

The meson UART driver triggers a lockdep splat at boot time, due to the new expectation that the driver has to initialize the per-port spinlock itself.

It remains unclear why a double initialization of the port spinlock is a desirable outcome, but in the meantime let's fix the splat.

Thanks!

Can you test patch from [1] if it helps and doesn't break anything in your case?

[1]: https://lore.kernel.org/linux-serial/[email protected]/T/#m9255e2a7474b160e66c7060fca5323ca3df49cfd

Quoted 25 lines
Fixes a3cb39d258ef ("serial: core: Allow detach and attach serial device for console")
Cc Andy Shevchenko <[email protected]>
Cc Greg Kroah-Hartman <[email protected]>
---
 drivers/tty/serial/meson_uart.c | 3 +++
 1 file changed, 3 insertions(+)
drivers/tty/serial/meson_uart.c +3 -1
index d2c08b760f83..386e39c90628 100644
@@ -759,6 +759,9 @@ static int meson_uart_probe(struct platform_device *pdev)
        if (ret)
                return ret;
 
+       /* Init the spinlock early in case this is the console */
+       spin_lock_init(&port->lock);
+
        port->iotype = UPIO_MEM;
        port->mapbase = res_mem->start;
        port->mapsize = resource_size(res_mem);
--

2.26.2

Signature · 2 lines
With Best Regards,
Andy Shevchenko
#2 lore raw
Quoted 11 lines · Andy Shevchenko
Quoted 9 lines · Marc Zyngier

The meson UART driver triggers a lockdep splat at boot time, due to the new expectation that the driver has to initialize the per-port spinlock itself.

It remains unclear why a double initialization of the port spinlock is a desirable outcome, but in the meantime let's fix the splat.

Quoted 6 lines

Thanks!

Can you test patch from [1] if it helps and doesn't break anything in your case?

[1]:
https://lore.kernel.org/linux-serial/[email protected]/T/#m9255e2a7474b160e66c7060fca5323ca3df49cfd

On its own, this patch doesn't seem to cure the issue (and it adds a compile-time warning due to unused flags).

Or did you mean to test it in complement of my patch?

        M.
Signature · 1 line
Jazz is not dead. It just smells funny...
#3 lore raw
Quoted 14 lines · Marc Zyngier
Quoted 13 lines · Andy Shevchenko
Quoted 9 lines · Marc Zyngier

The meson UART driver triggers a lockdep splat at boot time, due to the new expectation that the driver has to initialize the per-port spinlock itself.

It remains unclear why a double initialization of the port spinlock is a desirable outcome, but in the meantime let's fix the splat.

Thanks!

Quoted 8 lines
Quoted 5 lines

Can you test patch from [1] if it helps and doesn't break anything in your case?

[1]:
https://lore.kernel.org/linux-serial/[email protected]/T/#m9255e2a7474b160e66c7060fca5323ca3df49cfd

On its own, this patch doesn't seem to cure the issue (and it adds a compile-time warning due to unused flags).

Ah, sorry, I didn't compile it.
And after second though I think we simple need to initialise spin lock there. Can you try below (compile-tested only):

From ed4c882e7dc3fdfcea706ada0678c060c36163b3 Mon Sep 17 00:00:00 2001
From: Andy Shevchenko <[email protected]>
Date: Sat, 4 Jul 2020 19:30:39 +0300
Subject: [PATCH 1/1] serial: core: Initialise spin lock before use in
uart_configure_port()

In case of the port to be used as a console we must initialise a spin lock before use.

---
 drivers/tty/serial/serial_core.c | 7 +++++++
 1 file changed, 7 insertions(+)
drivers/tty/serial/serial_core.c +7 -0
index 3cc183acf7ba..a81b4900eb60 100644
@@ -2371,6 +2371,13 @@ uart_configure_port(struct uart_driver *drv, struct uart_state *state,
 		/* Power up port for set_mctrl() */
 		uart_change_pm(state, UART_PM_STATE_ON);
 
+		/*
+		 * If this driver supports console, and it hasn't been
+		 * successfully registered yet, initialise spin lock for it.
+		 */
+		if (port->cons && !(port->cons->flags & CON_ENABLED))
+			spin_lock_init(&port->lock);
+
 		/*
 		 * Ensure that the modem control lines are de-activated.
 		 * keep the DTR setting that is set in uart_set_options()
Signature · 1 line
2.27.0
Quoted 1 line

Or did you mean to test it in complement of my patch?

No, the idea to avoid "fixing" driver as you rightfully noticed a double init issue.

Signature · 2 lines
With Best Regards,
Andy Shevchenko
#4 lore raw
Quoted 66 lines · Andy Shevchenko
Quoted 14 lines · Marc Zyngier
Quoted 13 lines · Andy Shevchenko
Quoted 9 lines · Marc Zyngier

The meson UART driver triggers a lockdep splat at boot time, due to the new expectation that the driver has to initialize the per-port spinlock itself.

It remains unclear why a double initialization of the port spinlock is a desirable outcome, but in the meantime let's fix the splat.

Thanks!

Quoted 8 lines
Quoted 5 lines

Can you test patch from [1] if it helps and doesn't break anything in your case?

[1]:
https://lore.kernel.org/linux-serial/[email protected]/T/#m9255e2a7474b160e66c7060fca5323ca3df49cfd

On its own, this patch doesn't seem to cure the issue (and it adds a compile-time warning due to unused flags).

Ah, sorry, I didn't compile it.
And after second though I think we simple need to initialise spin lock there. Can you try below (compile-tested only):

From ed4c882e7dc3fdfcea706ada0678c060c36163b3 Mon Sep 17 00:00:00 2001
From: Andy Shevchenko <[email protected]>
Date: Sat, 4 Jul 2020 19:30:39 +0300
Subject: [PATCH 1/1] serial: core: Initialise spin lock before use in
uart_configure_port()

In case of the port to be used as a console we must initialise a spin lock before use.

---
 drivers/tty/serial/serial_core.c | 7 +++++++
 1 file changed, 7 insertions(+)
drivers/tty/serial/serial_core.c +7 -0
index 3cc183acf7ba..a81b4900eb60 100644
@@ -2371,6 +2371,13 @@ uart_configure_port(struct uart_driver *drv,
 struct uart_state *state,
 		/* Power up port for set_mctrl() */
 		uart_change_pm(state, UART_PM_STATE_ON);
 
+		/*
+		 * If this driver supports console, and it hasn't been
+		 * successfully registered yet, initialise spin lock for it.
+		 */
+		if (port->cons && !(port->cons->flags & CON_ENABLED))
+			spin_lock_init(&port->lock);
+
 		/*
 		 * Ensure that the modem control lines are de-activated.

* keep the DTR setting that is set in uart_set_options()
--
2.27.0

Quoted 4 lines
Quoted 1 line

Or did you mean to test it in complement of my patch?

No, the idea to avoid "fixing" driver as you rightfully noticed a double init issue.

This certainly keeps lockdep quiet on my system. You probably want to set the lockdep class whilst you're at it (keeping that code common with uart_port_spin_lock_init).

Thanks,

        M.
Signature · 1 line
Jazz is not dead. It just smells funny...