Skip to contentSkip to conversation

serial: 8250: Set fifo timeout with uart_fifo_timeout()

PATCH RESEND Superseded Michael Pratt 4 participants 6 messages Jan 11, 2024 TTY layer and serial drivers

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

PATCH RESEND 1/1 serial: 8250: Set fifo timeout with uart_fifo_timeout()

Commit 8f3631f0f6eb ("serial/8250: Use fifo in 8250 console driver") reworked functions for basic 8250 and 16550 type serial devices in order to enable and use the internal FIFO device for buffering, however the default timeout of 10 ms remained, which is proving to be insufficient for low baud rates like 9600, causing data overrun.

Unforunately, that commit was written and accepted just before commit 31f6bd7fad3b ("serial: Store character timing information to uart_port") which introduced the frame_time member of the uart_port struct in order to store the amount of time it takes to send one UART frame relative to the baud rate and other serial port configuration, and commit f9008285bb69 ("serial: Drop timeout from uart_port") which established function uart_fifo_timeout() in order to calculate a reasonable timeout to wait for all frames
in the FIFO device to flush before writing data again
using the now stored frame_time value and size of the buffer.

Fix this by using the new function to calculate the timeout whenever the buffer is larger than 1 byte (unknown port default).

Tested on a MIPS device (ar934x) at baud rates 625, 9600, 115200.

---
v1 thread: https://lore.kernel.org/linux-serial/[email protected]/

 drivers/tty/serial/8250/8250_port.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
drivers/tty/serial/8250/8250_port.c +4 -1
index 8ca061d3bbb9..777b61a79c5e 100644
@@ -2076,7 +2076,10 @@ static void wait_for_lsr(struct uart_8250_port *up, int bits)
 {
 	unsigned int status, tmout = 10000;
 
-	/* Wait up to 10ms for the character(s) to be sent. */
+	/* Wait for a time relative to buffer size and baud */
+	if (up->port.fifosize > 1)
+		tmout = jiffies_to_usecs(uart_fifo_timeout(&up->port));
+
 	for (;;) {
 		status = serial_lsr_in(up);
 
Signature · 1 line
2.30.2
#1 lore raw
Quoted 17 lines · Michael Pratt

Commit 8f3631f0f6eb ("serial/8250: Use fifo in 8250 console driver") reworked functions for basic 8250 and 16550 type serial devices in order to enable and use the internal FIFO device for buffering, however the default timeout of 10 ms remained, which is proving to be insufficient for low baud rates like 9600, causing data overrun.

Unforunately, that commit was written and accepted just before commit 31f6bd7fad3b ("serial: Store character timing information to uart_port") which introduced the frame_time member of the uart_port struct in order to store the amount of time it takes to send one UART frame relative to the baud rate and other serial port configuration, and commit f9008285bb69 ("serial: Drop timeout from uart_port") which established function uart_fifo_timeout() in order to calculate a reasonable timeout to wait for all frames
in the FIFO device to flush before writing data again
using the now stored frame_time value and size of the buffer.

Quoted 8 lines

Fix this by using the new function to calculate the timeout whenever the buffer is larger than 1 byte (unknown port default).

Tested on a MIPS device (ar934x) at baud rates 625, 9600, 115200.

---
v1 thread: https://lore.kernel.org/linux-serial/[email protected]/

What commit id does this fix?

thanks,

greg k-h

#2 lore raw
Quoted 3 lines · Greg Kroah-Hartman
Quoted 1 line · Michael Pratt

Commit 8f3631f0f6eb ("serial/8250: Use fifo in 8250 console driver")

This is the commit that made the issue present itself.

I'm not sure whether it's right to say that this "fixes" that commit since it only caused the issue indirectly, and the diff for this patch doesn't touch any lines that the other commit touched, and the method I'm using to fix the issue was not available at the time, and also that for high baud rates like 115200 everything is still fine... (the 10 ms timeout is as old as the tree)

If that's enough for a "Fixes" tag then go ahead (or tell me to add it), but maybe a "Ref" tag would be enough?

You can see the other thread linked for more discussion on that point if you like...

Quoted 26 lines
Quoted 16 lines

reworked functions for basic 8250 and 16550 type serial devices in order to enable and use the internal FIFO device for buffering, however the default timeout of 10 ms remained, which is proving to be insufficient for low baud rates like 9600, causing data overrun.

Unforunately, that commit was written and accepted just before commit 31f6bd7fad3b ("serial: Store character timing information to uart_port") which introduced the frame_time member of the uart_port struct in order to store the amount of time it takes to send one UART frame relative to the baud rate and other serial port configuration, and commit f9008285bb69 ("serial: Drop timeout from uart_port") which established function uart_fifo_timeout() in order to calculate a reasonable timeout to wait for all frames
in the FIFO device to flush before writing data again
using the now stored frame_time value and size of the buffer.

Quoted 8 lines

Fix this by using the new function to calculate the timeout whenever the buffer is larger than 1 byte (unknown port default).

Tested on a MIPS device (ar934x) at baud rates 625, 9600, 115200.

---
v1 thread: https://lore.kernel.org/linux-serial/[email protected]/

Quoted 5 lines

What commit id does this fix?

thanks,

greg k-h

--
MCP

#3 lore raw
Quoted 17 lines · Michael Pratt
Quoted 3 lines · Greg Kroah-Hartman
Quoted 1 line · Michael Pratt

Commit 8f3631f0f6eb ("serial/8250: Use fifo in 8250 console driver")

This is the commit that made the issue present itself.

I'm not sure whether it's right to say that this "fixes" that commit since it only caused the issue indirectly, and the diff for this patch doesn't touch any lines that the other commit touched, and the method I'm using to fix the issue was not available at the time, and also that for high baud rates like 115200 everything is still fine... (the 10 ms timeout is as old as the tree)

Quoted 4 lines

If that's enough for a "Fixes" tag then go ahead (or tell me to add it), but maybe a "Ref" tag would be enough?

You can see the other thread linked for more discussion on that point if you like...

Besides Fixes tags, you can add dependency information for stable folks with Cc tag using the format described in Documentation/process/stable-kernel-rules.rst

Signature · 1 line
 i.
Quoted 35 lines
Quoted 26 lines
Quoted 16 lines

reworked functions for basic 8250 and 16550 type serial devices in order to enable and use the internal FIFO device for buffering, however the default timeout of 10 ms remained, which is proving to be insufficient for low baud rates like 9600, causing data overrun.

Unforunately, that commit was written and accepted just before commit 31f6bd7fad3b ("serial: Store character timing information to uart_port") which introduced the frame_time member of the uart_port struct in order to store the amount of time it takes to send one UART frame relative to the baud rate and other serial port configuration, and commit f9008285bb69 ("serial: Drop timeout from uart_port") which established function uart_fifo_timeout() in order to calculate a reasonable timeout to wait for all frames
in the FIFO device to flush before writing data again
using the now stored frame_time value and size of the buffer.

Quoted 8 lines

Fix this by using the new function to calculate the timeout whenever the buffer is larger than 1 byte (unknown port default).

Tested on a MIPS device (ar934x) at baud rates 625, 9600, 115200.

---
v1 thread: https://lore.kernel.org/linux-serial/[email protected]/

Quoted 5 lines

What commit id does this fix?

thanks,

greg k-h

--
MCP

#4 lore raw
In reply to Michael Pratt
Quoted 35 lines · Michael Pratt

Commit 8f3631f0f6eb ("serial/8250: Use fifo in 8250 console driver") reworked functions for basic 8250 and 16550 type serial devices in order to enable and use the internal FIFO device for buffering, however the default timeout of 10 ms remained, which is proving to be insufficient for low baud rates like 9600, causing data overrun.

Unforunately, that commit was written and accepted just before commit 31f6bd7fad3b ("serial: Store character timing information to uart_port") which introduced the frame_time member of the uart_port struct in order to store the amount of time it takes to send one UART frame relative to the baud rate and other serial port configuration, and commit f9008285bb69 ("serial: Drop timeout from uart_port") which established function uart_fifo_timeout() in order to calculate a reasonable timeout to wait for all frames
in the FIFO device to flush before writing data again
using the now stored frame_time value and size of the buffer.

Fix this by using the new function to calculate the timeout whenever the buffer is larger than 1 byte (unknown port default).

Tested on a MIPS device (ar934x) at baud rates 625, 9600, 115200.

---
v1 thread: https://lore.kernel.org/linux-serial/[email protected]/

 drivers/tty/serial/8250/8250_port.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
drivers/tty/serial/8250/8250_port.c +0 -0
index 8ca061d3bbb9..777b61a79c5e 100644

drivers/tty/serial/8250/8250_port.c static void wait_for_lsr(struct uart_8250_port *up, int bits) · line 2082

Quoted 8 lines
drivers/tty/serial/8250/8250_port.c +4 -1
@@ -2076,7 +2076,10 @@ static void wait_for_lsr(struct uart_8250_port *up, int bits)
 {
        unsigned int status, tmout = 10000;
 
-       /* Wait up to 10ms for the character(s) to be sent. */
+       /* Wait for a time relative to buffer size and baud */
+       if (up->port.fifosize > 1)
+               tmout = jiffies_to_usecs(uart_fifo_timeout(&up->port));
+

Michael, I have copied you onto a patch addressing the similar issue, where it was considered to avoid the computation of timeout on call to wait_for_lsr.

Quoted 8 lines
drivers/tty/serial/8250/8250_port.c +0 -1
        for (;;) {
                status = serial_lsr_in(up);
 
--

2.30.2

#5 lore raw
In reply to Michael Pratt
Quoted 7 lines · Michael Pratt

Commit 8f3631f0f6eb ("serial/8250: Use fifo in 8250 console driver") reworked functions for basic 8250 and 16550 type serial devices in order to enable and use the internal FIFO device for buffering, however the default timeout of 10 ms remained, which is proving to be insufficient for low baud rates like 9600, causing data overrun.

Unforunately, that commit was written and accepted just before commit

Unfortunately,

Signature · 1 line
 i.
Quoted 38 lines

31f6bd7fad3b ("serial: Store character timing information to uart_port") which introduced the frame_time member of the uart_port struct in order to store the amount of time it takes to send one UART frame relative to the baud rate and other serial port configuration, and commit f9008285bb69 ("serial: Drop timeout from uart_port") which established function uart_fifo_timeout() in order to calculate a reasonable timeout to wait for all frames
in the FIFO device to flush before writing data again
using the now stored frame_time value and size of the buffer.

Fix this by using the new function to calculate the timeout whenever the buffer is larger than 1 byte (unknown port default).

Tested on a MIPS device (ar934x) at baud rates 625, 9600, 115200.

---
v1 thread: https://lore.kernel.org/linux-serial/[email protected]/

 drivers/tty/serial/8250/8250_port.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
drivers/tty/serial/8250/8250_port.c +4 -1
index 8ca061d3bbb9..777b61a79c5e 100644
@@ -2076,7 +2076,10 @@ static void wait_for_lsr(struct uart_8250_port *up, int bits)
 {
 	unsigned int status, tmout = 10000;
 
-	/* Wait up to 10ms for the character(s) to be sent. */
+	/* Wait for a time relative to buffer size and baud */
+	if (up->port.fifosize > 1)
+		tmout = jiffies_to_usecs(uart_fifo_timeout(&up->port));
+
 	for (;;) {
 		status = serial_lsr_in(up);