serial: fix UART_BUG_TXEN test
PATCH 1/1 serial: fix UART_BUG_TXEN test
Hi,
There is a bug in the UART_BUG_TXEN test: It gives false positives in case the UART_IER_THRI bit is set. Fixed by explicitly clearing the UART_IER register first.
It may trigger with an active serial console as serial console writes set the UART_IER_THRI bit.
cheers,
Gerd
drivers/serial/8250.c renamed from drivers/serial/8250.c.serial +1 -0
@@ -1712,6 +1712,7 @@* Do a quick test to see if we receive an * interrupt when we enable the TX irq. */ +serial_outp (up,UART_IER ,0 );serial_outp (up,UART_IER ,UART_IER_THRI ); lsr =serial_in (up,UART_LSR ); iir =serial_in (up,UART_IIR );
Quoted 6 lines · Gerd Hoffmann
There is a bug in the
UART_BUG_TXENtest: It gives false positives in case theUART_IER_THRIbit is set. Fixed by explicitly clearing theUART_IERregister first.It may trigger with an active serial console as serial console writes set the
UART_IER_THRIbit.
Actually, I think Alan's (f91a3715db2bb44fcf08cec642e68f919b70f7f4) idea of setting UART_IER_THRI after serial console writes is buggy. If the serial port being used as a console is sharing its interrupt line with other devices, then there's the very real possibility for causing spurious interrupts.
I think that's what needs to be fixed rather than working around this potentially buggy behaviour.
Maybe we have no option but to take the spinlock in the serial console code, and suck it if we oops with that spinlock held.
Quoted 11 lines
Signed-off-by Gerd Hoffmann <[email protected]>drivers/serial/8250.c renamed from drivers/serial/8250.c.serial +1
-0@@ -1712,6 +1712,7 @@* Do a quick test to see if we receive an * interrupt when we enable the TX irq. */ +serial_outp (up,UART_IER ,0 );serial_outp (up,UART_IER ,UART_IER_THRI ); lsr =serial_in (up,UART_LSR ); iir =serial_in (up,UART_IIR );
Signature · 3 lines
Russell King Linux kernel 2.6 ARM Linux - http://www.arm.linux.org.uk/ maintainer of: 2.6 Serial core
Quoted 14 lines · Russell King
Quoted 6 lines · Gerd Hoffmann
There is a bug in the
UART_BUG_TXENtest: It gives false positives in case theUART_IER_THRIbit is set. Fixed by explicitly clearing theUART_IERregister first.It may trigger with an active serial console as serial console writes set the
UART_IER_THRIbit.Actually, I think Alan's (
f91a3715db2bb44fcf08cec642e68f919b70f7f4) idea of settingUART_IER_THRIafter serial console writes is buggy. If the serial port being used as a console is sharing its interrupt line with other devices, then there's the very real possibility for causing spurious interrupts.
Quoted 5 lines
I think that's what needs to be fixed rather than working around this potentially buggy behaviour.
Maybe we have no option but to take the spinlock in the serial console code, and suck it if we oops with that spinlock held.
So here's the patch. I've been waiting for feedback from Alan before sending this upstream... Nevertheless, could you (Gerd) test it please?
Thanks.
drivers/serial/8250.c +11 -2
@@ -2200,10 +2200,17 @@ static void serial8250_console_write(struct console *co, const char *s, unsigned int count) { struct uart_8250_port *up = &serial8250_ports[co->index]; +unsigned long flags;unsigned int ier; +int locked =1 ;touch_nmi_watchdog (); +if (oops_in_progress) { + locked =spin_trylock_irqsave (&up->port.lock, flags); + }else +spin_lock_irqsave (&up->port.lock, flags); +/* * First save the IER then disable the interrupts */
@@ -2221,8 +2228,10 @@ serial8250_console_write(struct console* and restore the IER */ wait_for_xmitr (up,BOTH_EMPTY ); - up->ier |=UART_IER_THRI ; -serial_out (up,UART_IER , ier |UART_IER_THRI ); +serial_out (up,UART_IER , ier); + +if (locked) +spin_unlock_irqrestore (&up->port.lock, flags); }static int serial8250_console_setup (struct console *co,char *options)
Signature · 3 lines
Russell King Linux kernel 2.6 ARM Linux - http://www.arm.linux.org.uk/ maintainer of: 2.6 Serial core