mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
* [PATCH v2] serial: max310x: improve interrupt handling
@ 2025-09-05  8:54 Tapio Reijonen
  2025-09-05  9:04 ` Jiri Slaby
  0 siblings, 1 reply; 2+ messages in thread
From: Tapio Reijonen @ 2025-09-05  8:54 UTC (permalink / raw)
  To: Greg Kroah-Hartman, Jiri Slaby, Alexander Shiyan, Hugo Villeneuve
  Cc: linux-kernel, linux-serial, Tapio Reijonen

When there is a heavy load of receiving characters to all
four UART's, the warning 'Hardware RX FIFO overrun' is
sometimes detected.
The current implementation always service first the highest UART
until no more interrupt and then service another UART
(ex: UART3 will be serviced for as long as there are interrupts
for it, then UART2, etc).

This commit handle all individual interrupt sources before
reading the global IRQ register again.

This commit has also a nice side-effect of improving the efficiency
of the driver by reducing the number of reads of the global
IRQ register.

Signed-off-by: Tapio Reijonen <tapio.reijonen@vaisala.com>
---
Changes in v2:
- Improve content of the commit message
- Fix a line indention in prevoius patch
- According review comments, changed to use for_each_clear_bit
  to simplify serve all IRQs in a loop.
  NOTE: When a bit in IRQ[n] is set 0 the associated UART’s
  internal IRQ is generated.
- Link to v1: https://lore.kernel.org/r/20250903-master-max310x-improve-interrupt-handling-v1-1-bfb44829e760@vaisala.com
---
 drivers/tty/serial/max310x.c | 18 ++++++++++++++----
 1 file changed, 14 insertions(+), 4 deletions(-)

diff --git a/drivers/tty/serial/max310x.c b/drivers/tty/serial/max310x.c
index ce260e9949c3c268e706b2615d6fc01adc21e49b..464f2828bd304198d79046e79b2a8a0893e5cd77 100644
--- a/drivers/tty/serial/max310x.c
+++ b/drivers/tty/serial/max310x.c
@@ -825,14 +825,24 @@ static irqreturn_t max310x_ist(int irq, void *dev_id)
 	if (s->devtype->nr > 1) {
 		do {
 			unsigned int val = ~0;
+			unsigned long irq;
+			unsigned long port;
+			bool done = true;
 
 			WARN_ON_ONCE(regmap_read(s->regmap,
 						 MAX310X_GLOBALIRQ_REG, &val));
-			val = ((1 << s->devtype->nr) - 1) & ~val;
-			if (!val)
+
+			irq = val;
+
+			for_each_clear_bit(port, &irq, s->devtype->nr) {
+				done = false;
+
+				if (max310x_port_irq(s, port) == IRQ_HANDLED)
+					handled = true;
+			}
+
+			if (done)
 				break;
-			if (max310x_port_irq(s, fls(val) - 1) == IRQ_HANDLED)
-				handled = true;
 		} while (1);
 	} else {
 		if (max310x_port_irq(s, 0) == IRQ_HANDLED)

---
base-commit: c8bc81a52d5a2ac2e4b257ae123677cf94112755
change-id: 20250903-master-max310x-improve-interrupt-handling-aa22b7ba1c1d

Best regards,
-- 
Tapio Reijonen <tapio.reijonen@vaisala.com>


^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: [PATCH v2] serial: max310x: improve interrupt handling
  2025-09-05  8:54 [PATCH v2] serial: max310x: improve interrupt handling Tapio Reijonen
@ 2025-09-05  9:04 ` Jiri Slaby
  0 siblings, 0 replies; 2+ messages in thread
From: Jiri Slaby @ 2025-09-05  9:04 UTC (permalink / raw)
  To: Tapio Reijonen, Greg Kroah-Hartman, Alexander Shiyan, Hugo Villeneuve
  Cc: linux-kernel, linux-serial

On 05. 09. 25, 10:54, Tapio Reijonen wrote:
> When there is a heavy load of receiving characters to all
> four UART's, the warning 'Hardware RX FIFO overrun' is
> sometimes detected.
> The current implementation always service first the highest UART
> until no more interrupt and then service another UART
> (ex: UART3 will be serviced for as long as there are interrupts
> for it, then UART2, etc).
> 
> This commit handle all individual interrupt sources before
> reading the global IRQ register again.
> 
> This commit has also a nice side-effect of improving the efficiency
> of the driver by reducing the number of reads of the global
> IRQ register.
> 
> Signed-off-by: Tapio Reijonen <tapio.reijonen@vaisala.com>
> ---
> Changes in v2:
> - Improve content of the commit message
> - Fix a line indention in prevoius patch
> - According review comments, changed to use for_each_clear_bit
>    to simplify serve all IRQs in a loop.
>    NOTE: When a bit in IRQ[n] is set 0 the associated UART’s
>    internal IRQ is generated.
> - Link to v1: https://lore.kernel.org/r/20250903-master-max310x-improve-interrupt-handling-v1-1-bfb44829e760@vaisala.com
> ---
>   drivers/tty/serial/max310x.c | 18 ++++++++++++++----
>   1 file changed, 14 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/tty/serial/max310x.c b/drivers/tty/serial/max310x.c
> index ce260e9949c3c268e706b2615d6fc01adc21e49b..464f2828bd304198d79046e79b2a8a0893e5cd77 100644
> --- a/drivers/tty/serial/max310x.c
> +++ b/drivers/tty/serial/max310x.c
> @@ -825,14 +825,24 @@ static irqreturn_t max310x_ist(int irq, void *dev_id)
>   	if (s->devtype->nr > 1) {
>   		do {
>   			unsigned int val = ~0;
> +			unsigned long irq;
> +			unsigned long port;
> +			bool done = true;
>   
>   			WARN_ON_ONCE(regmap_read(s->regmap,
>   						 MAX310X_GLOBALIRQ_REG, &val));
> -			val = ((1 << s->devtype->nr) - 1) & ~val;
> -			if (!val)
> +
> +			irq = val;
> +
> +			for_each_clear_bit(port, &irq, s->devtype->nr) {

port can be uint, no need for ulong.

> +				done = false;
> +
> +				if (max310x_port_irq(s, port) == IRQ_HANDLED)
> +					handled = true;
> +			}

Hey, this looks to be the order of magnitude saner!

> +
> +			if (done)
>   				break;
> -			if (max310x_port_irq(s, fls(val) - 1) == IRQ_HANDLED)
> -				handled = true;
>   		} while (1);

So now, this should perhaps be "while (!done)"?

thanks,
-- 
js
suse labs

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2025-09-05  9:04 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2025-09-05  8:54 [PATCH v2] serial: max310x: improve interrupt handling Tapio Reijonen
2025-09-05  9:04 ` Jiri Slaby

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®