Skip to contentSkip to conversation

serial: mxc: Keep the original FIFO empty check

PATCH In review Fabio Estevam 4 participants 4 messages Oct 25, 2022 – Oct 26, 2022

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 serial: mxc: Keep the original FIFO empty check

From: Fabio Estevam <[email protected]>

Tim Harvey reported that since commit c7878a0483c5 ("serial: mxc:
have putc use the TXFIFO"), console messages put inside board_init() are no longer printed correctly.

This change added a check to handle the UART FIFO full condition and removed the UART FIFO empty check, which causes the problem.

To avoid console corruption, add back the original UART FIFO empty check so that when the FIFO is empty, mxc_serial_putc() returns -EAGAIN to the core serial-uclass.c.

This way the serial core can properly handle the UART empty condition by not doing character transmission when -EAGAIN is returned.

Fixes c7878a0483c5 ("serial: mxc: have putc use the TXFIFO")

---
Changes since v1:
- Add the check for FIFO full and FIFO empty.

 drivers/serial/serial_mxc.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/serial/serial_mxc.c +1 -1
index 4cf79c1ca24f..d3fbe76065ec 100644
@@ -311,7 +311,7 @@ static int mxc_serial_putc(struct udevice *dev, const char ch)
 	struct mxc_serial_plat *plat = dev_get_plat(dev);
 	struct mxc_uart *const uart = plat->reg;
 
-	if (readl(&uart->ts) & UTS_TXFULL)
+	if ((readl(&uart->ts) & UTS_TXFULL) || !(readl(&uart->ts) & UTS_TXEMPTY))
 		return -EAGAIN;
 
 	writel(ch, &uart->txd);
Signature · 1 line
2.25.1
#1 lore raw
Quoted 41 lines · Fabio Estevam

From: Fabio Estevam <[email protected]>

Tim Harvey reported that since commit c7878a0483c5 ("serial: mxc:
have putc use the TXFIFO"), console messages put inside board_init() are no longer printed correctly.

This change added a check to handle the UART FIFO full condition and removed the UART FIFO empty check, which causes the problem.

To avoid console corruption, add back the original UART FIFO empty check so that when the FIFO is empty, mxc_serial_putc() returns -EAGAIN to the core serial-uclass.c.

This way the serial core can properly handle the UART empty condition by not doing character transmission when -EAGAIN is returned.

Fixes c7878a0483c5 ("serial: mxc: have putc use the TXFIFO")

---
Changes since v1:
- Add the check for FIFO full and FIFO empty.

 drivers/serial/serial_mxc.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/serial/serial_mxc.c +1 -2
index 4cf79c1ca24f..d3fbe76065ec 100644
@@ -311,7 +311,7 @@ static int mxc_serial_putc(struct udevice *dev, const char ch)
        struct mxc_serial_plat *plat = dev_get_plat(dev);
        struct mxc_uart *const uart = plat->reg;
 
-       if (readl(&uart->ts) & UTS_TXFULL)
+       if ((readl(&uart->ts) & UTS_TXFULL) || !(readl(&uart->ts) & UTS_TXEMPTY))
                return -EAGAIN;
 
        writel(ch, &uart->txd);
--
Quoted 2 lines

2.25.1

Fabio,

This resolves the issue. Why would the kernel not suffer from this as well? We are essentially saying if the FIFO is not full but also not empty we need to wait.... so do we have a FIFO size of 1 or something here as opposed to the kernel?

Best Regards,

Tim

#2 lore raw

Hi Tim,

Quoted 6 lines · Tim Harvey

Fabio,

This resolves the issue. Why would the kernel not suffer from this as well? We are essentially saying if the FIFO is not full but also not empty we need to wait.... so do we have a FIFO size of 1 or something here as opposed to the kernel?

The FIFO has a fixed size of 32 chars.

Kernel uses TXTL, which "controls the threshold at which a maskable interrupt is generated by the TxFIFO".

We do not use interrupts in U-Boot. Prior to Johannes' patch, we only checked for the TXFIFO empty condition, which avoided sending data when nothing was available inside the FIFO (under-run condition).

Johannes' patch handles over-run condition.

My patch guarantees that both over-run and under-run conditions are handled.

When the number of chars in the FIFO is 1, 2, 3,...30, the transmission can safely happen. Special cases are FIFO empty and FIFO full, which returns -EAGAIN to the serial core.

Regards,

Fabio Estevam

Signature · 3 lines
DENX Software Engineering GmbH,      Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-60 Fax: (+49)-8142-66989-80 Email: [email protected]
#3 lore raw

Hi Fabio

Quoted 23 lines · Fabio Estevam

Hi Tim,

Quoted 6 lines · Tim Harvey

Fabio,

This resolves the issue. Why would the kernel not suffer from this as well? We are essentially saying if the FIFO is not full but also not empty we need to wait.... so do we have a FIFO size of 1 or something here as opposed to the kernel?

The FIFO has a fixed size of 32 chars.

Kernel uses TXTL, which "controls the threshold at which a maskable interrupt is generated by the
TxFIFO".

We do not use interrupts in U-Boot. Prior to Johannes' patch, we only checked for the TXFIFO empty condition, which avoided sending data when nothing was
available inside the FIFO (under-run condition).

Quoted 8 lines

Johannes' patch handles over-run condition.

My patch guarantees that both over-run and under-run conditions are handled.

When the number of chars in the FIFO is 1, 2, 3,...30, the transmission can safely happen. Special cases are FIFO empty and FIFO full, which returns -EAGAIN to the serial core.

Can you explain it better? If FIFO is empty there are no chars in the serial, and
the transmission can happen only when you have at least one byte in the fifo. The underrun condition in general is when the start of tx is enabled but you have no bytes to send
at the time that bytes are needed. Was the case to limit the change on the architecture that was tested?

Michael

Quoted 9 lines

Regards,

Fabio Estevam
--
DENX Software Engineering GmbH, Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-60 Fax: (+49)-8142-66989-80 Email:
[email protected]

Signature · 11 lines
Michael Nazzareno Trimarchi
Co-Founder & Chief Executive Officer
M. +39 347 913 2170
[email protected]
__________________________________

Amarula Solutions BV
Joop Geesinkweg 125, 1114 AB, Amsterdam, NL
T. +31 (0)85 111 9172
[email protected]
www.amarulasolutions.com