Skip to contentSkip to conversation

serial: 8250_early: Add earlycon for BCM2835 aux uart

PATCH In review [email protected] 4 participants 10 messages Jan 26, 2020 – Feb 12, 2020

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 serial: 8250_early: Add earlycon for BCM2835 aux uart

From: Matthias Brugger <[email protected]>

Define the OF early console for BCM2835 aux UART, which can be enabled by passing "earlycon" on the boot command line. This UART is found on BCM283x and BCM27xx SoCs, a.k.a. Raspberry Pi in its variants.

---

 drivers/tty/serial/8250/8250_bcm2835aux.c | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)
drivers/tty/serial/8250/8250_bcm2835aux.c +18 -0
index 8ce700c1a7fc..6769cea2964a 100644
@@ -135,6 +135,24 @@ static struct platform_driver bcm2835aux_serial_driver = {
 };
 module_platform_driver(bcm2835aux_serial_driver);
 
+#ifdef CONFIG_SERIAL_8250_CONSOLE
+
+static int __init early_bcm2835aux_setup(struct earlycon_device *device,
+					const char *options)
+{
+	if (!device->port.membase)
+		return -ENODEV;
+
+	device->port.iotype = UPIO_MEM32;
+	device->port.regshift = 2;
+
+	return early_serial8250_setup(device, NULL);
+}
+
+OF_EARLYCON_DECLARE(bcm2835aux, "brcm,bcm2835-aux-uart",
+		    early_bcm2835aux_setup);
+#endif
+
 MODULE_DESCRIPTION("BCM2835 auxiliar UART driver");
 MODULE_AUTHOR("Martin Sperl <[email protected]>");
 MODULE_LICENSE("GPL v2");
Signature · 1 line
2.24.1
#1 lore raw
Quoted 7 lines · On Sun, Jan 26, 2020 at 01:33:14PM +0100, [email protected] wrote:
drivers/tty/serial/8250/8250_bcm2835aux.c +7 -0
+#ifdef CONFIG_SERIAL_8250_CONSOLE
+
+static int __init early_bcm2835aux_setup(struct earlycon_device *device,
+					const char *options)
+{
+	if (!device->port.membase)
+		return -ENODEV;

drivers/tty/serial/8250/8250_bcm2835aux.c static struct platform_driver bcm2835aux_serial_driver = { · line 154

Quoted 10 lines
drivers/tty/serial/8250/8250_bcm2835aux.c +10 -0
+
+	device->port.iotype = UPIO_MEM32;
+	device->port.regshift = 2;
+
+	return early_serial8250_setup(device, NULL);
+}
+
+OF_EARLYCON_DECLARE(bcm2835aux, "brcm,bcm2835-aux-uart",
+		    early_bcm2835aux_setup);
+#endif

Does this really work? I also tried to get it working recently and the system just hung on boot. Looking at it with a JTAG debugger showed that the bcm2835aux registers were inaccessible because the mini UART wasn't enabled in the AUXENB register.

Maybe if you use OF_EARLYCON_DECLARE, the firmware recognizes that serial1 is set as stdout-path and performs enablement of the mini UART? Or are you using U-Boot which perhaps does the enablement?

I also saw in the JTAG debugger that the uartclk member contained an incorrect value, so I'd expect that it has to be set as well in early_bcm2835aux_setup().

I'll see to it that I give this patch a whirl when I return to the office next week.

Thanks,

Lukas

#2 lore raw
Quoted 24 lines · Lukas Wunner
Quoted 7 lines · On Sun, Jan 26, 2020 at 01:33:14PM +0100, [email protected] wrote:
drivers/tty/serial/8250/8250_bcm2835aux.c +7 -0
+#ifdef CONFIG_SERIAL_8250_CONSOLE
+
+static int __init early_bcm2835aux_setup(struct earlycon_device *device,
+					const char *options)
+{
+	if (!device->port.membase)
+		return -ENODEV;

drivers/tty/serial/8250/8250_bcm2835aux.c static struct platform_driver bcm2835aux_serial_driver = { · line 154

Quoted 10 lines
drivers/tty/serial/8250/8250_bcm2835aux.c +10 -0
+
+	device->port.iotype = UPIO_MEM32;
+	device->port.regshift = 2;
+
+	return early_serial8250_setup(device, NULL);
+}
+
+OF_EARLYCON_DECLARE(bcm2835aux, "brcm,bcm2835-aux-uart",
+		    early_bcm2835aux_setup);
+#endif

Does this really work? I also tried to get it working recently and the system just hung on boot. Looking at it with a JTAG debugger showed that the bcm2835aux registers were inaccessible because the mini UART wasn't enabled in the AUXENB register.

Quoted 3 lines

Maybe if you use OF_EARLYCON_DECLARE, the firmware recognizes that serial1 is set as stdout-path and performs enablement of the mini UART? Or are you using U-Boot which perhaps does the enablement?

Yes I'm using U-Boot which enables the console for me. My understanding is that the early console is thought as a re-use of the console the boot FW used for logging. AFAIK for example it does not enable any needed clocks but expects these to be enabled already.

Looking on the source code of U-Boot [1] I don't see that the AUXENB is written somewhere, so I suppose that the FW should already has enabled the aux-uart.

I any case if it's just to set one bit, I think we can do that in early_bcm2835aux_setup().

[1]
https://gitlab.denx.de/u-boot/u-boot/blob/master/drivers/serial/serial_bcm283x_mu.c

Quoted 4 lines

I also saw in the JTAG debugger that the uartclk member contained an incorrect value, so I'd expect that it has to be set as well in early_bcm2835aux_setup().

In my case the clock was set by U-Boot already.

Regards,
Matthias

Quoted 8 lines

I'll see to it that I give this patch a whirl when I return to the office next week.

Thanks,

Lukas

#3 lore raw
Quoted 51 lines · Matthias Brugger
Quoted 24 lines · Lukas Wunner
Quoted 7 lines · On Sun, Jan 26, 2020 at 01:33:14PM +0100, [email protected] wrote:
drivers/tty/serial/8250/8250_bcm2835aux.c +7 -0
+#ifdef CONFIG_SERIAL_8250_CONSOLE
+
+static int __init early_bcm2835aux_setup(struct earlycon_device *device,
+					const char *options)
+{
+	if (!device->port.membase)
+		return -ENODEV;

drivers/tty/serial/8250/8250_bcm2835aux.c static struct platform_driver bcm2835aux_serial_driver = { · line 154

Quoted 10 lines
drivers/tty/serial/8250/8250_bcm2835aux.c +10 -0
+
+	device->port.iotype = UPIO_MEM32;
+	device->port.regshift = 2;
+
+	return early_serial8250_setup(device, NULL);
+}
+
+OF_EARLYCON_DECLARE(bcm2835aux, "brcm,bcm2835-aux-uart",
+		    early_bcm2835aux_setup);
+#endif

Does this really work? I also tried to get it working recently and the system just hung on boot. Looking at it with a JTAG debugger showed that the bcm2835aux registers were inaccessible because the mini UART wasn't enabled in the AUXENB register.

Quoted 3 lines

Maybe if you use OF_EARLYCON_DECLARE, the firmware recognizes that serial1 is set as stdout-path and performs enablement of the mini UART? Or are you using U-Boot which perhaps does the enablement?

Yes I'm using U-Boot which enables the console for me. My understanding is that
the early console is thought as a re-use of the console the boot FW used for
logging. AFAIK for example it does not enable any needed clocks but expects
these to be enabled already.

Looking on the source code of U-Boot [1] I don't see that the AUXENB is written
somewhere, so I suppose that the FW should already has enabled the aux-uart.

I any case if it's just to set one bit, I think we can do that in early_bcm2835aux_setup().

[1]
https://gitlab.denx.de/u-boot/u-boot/blob/master/drivers/serial/serial_bcm283x_mu.c

Quoted 1 line
Quoted 5 lines
Quoted 3 lines

I also saw in the JTAG debugger that the uartclk member contained an incorrect value, so I'd expect that it has to be set as well in early_bcm2835aux_setup().

In my case the clock was set by U-Boot already.

I'm testing this by booting directly from RPi4's bootloader. And it works as long as I add this to config.txt:

enable_uart=1
gpu_freq=500

Which AFAIK blocks frequency scalin on the GPU and fixes the clock to a point where the serial is set at 115200 bauds.

Ideally it'd be nice to be able to query the clock frequency, and recalculate the divisors based on that. But I don't know if it's feasible at that point in the boot process.

We could maybe leave a comment explaining what is expected from the bootloader, given the different options around.

Regards,
Nicolas

#4 lore raw
Quoted 6 lines · Nicolas Saenz Julienne

I'm testing this by booting directly from RPi4's bootloader. And it works as long as I add this to config.txt:

enable_uart=1
gpu_freq=500

Quoted 6 lines

Which AFAIK blocks frequency scalin on the GPU and fixes the clock to a point where the serial is set at 115200 bauds.

Ideally it'd be nice to be able to query the clock frequency, and recalculate the divisors based on that. But I don't know if it's feasible at that point in the boot process.

Well, we don't even support adjusting the baudrate *after* the kernel has booted.

The problem is that in mainline, bcm2835_defconfig contains:
CONFIG_CPU_FREQ_DEFAULT_GOV_CONSERVATIVE=y

Likewise in the Foundation's downstream tree, bcmrpi_defconfig as well as bcm2711_defconfig and bcm2709_defconfig contain:
CONFIG_CPU_FREQ_DEFAULT_GOV_POWERSAVE=y

In contrast to this, we set the following on Revolution Pi devices:
CONFIG_CPU_FREQ_DEFAULT_GOV_PERFORMANCE=y

Downclocking influences not only the uart1 baud rate but also the spi0 clock. We attach Ethernet chips to spi0, throughput was significantly worse with the ondemand governor (which is what we used previously). We felt that maximum Ethernet performance outweighs the relatively small powersaving gains.

Thanks,

Lukas

#5 lore raw
Quoted 10 lines · Lukas Wunner
Quoted 9 lines · Nicolas Saenz Julienne

I'm testing this by booting directly from RPi4's bootloader. And it works as long as I add this to config.txt:

enable_uart=1
gpu_freq=500

Which AFAIK blocks frequency scalin on the GPU and fixes the clock to a point where the serial is set at 115200 bauds.

Quoted 6 lines
Quoted 3 lines

Ideally it'd be nice to be able to query the clock frequency, and recalculate the divisors based on that. But I don't know if it's feasible at that point in the boot process.

Well, we don't even support adjusting the baudrate *after* the kernel has booted.

Don't 8250 uarts have a clk divisor? I've seen other 8250 earlycon drivers do this (see 8250_ingenic.c), that said I think it's a lost cause for us.

BTW did you had the oportunity to have a go at the patch?

Quoted 7 lines

The problem is that in mainline, bcm2835_defconfig contains:
CONFIG_CPU_FREQ_DEFAULT_GOV_CONSERVATIVE=y

Likewise in the Foundation's downstream tree, bcmrpi_defconfig as well as bcm2711_defconfig and bcm2709_defconfig contain:
CONFIG_CPU_FREQ_DEFAULT_GOV_POWERSAVE=y

Quoted 8 lines

In contrast to this, we set the following on Revolution Pi devices:
CONFIG_CPU_FREQ_DEFAULT_GOV_PERFORMANCE=y

Downclocking influences not only the uart1 baud rate but also the spi0 clock. We attach Ethernet chips to spi0, throughput was significantly worse with the ondemand governor (which is what we used previously). We felt that maximum Ethernet performance outweighs the relatively small powersaving gains.

In that regard I suggest you use the upstream cpufreq driver which behaves properly in that regard. It disables GPU freq scaling, so as to change CPU frequencies without SPI/I2C/UART issues.

Regards,
Nicolas

#6 lore raw
Quoted 1 line · Nicolas Saenz Julienne

BTW did you had the oportunity to have a go at the patch?

I've just performed a quick test and it doesn't work for me. If I add stdout-path = "serial1:115200n8"; to the chosen node, I only get a regular console with this patch, not an earlycon.

Quoted 16 lines
Quoted 7 lines

The problem is that in mainline, bcm2835_defconfig contains:
CONFIG_CPU_FREQ_DEFAULT_GOV_CONSERVATIVE=y

Likewise in the Foundation's downstream tree, bcmrpi_defconfig as well as bcm2711_defconfig and bcm2709_defconfig contain:
CONFIG_CPU_FREQ_DEFAULT_GOV_POWERSAVE=y

Quoted 8 lines

In contrast to this, we set the following on Revolution Pi devices:
CONFIG_CPU_FREQ_DEFAULT_GOV_PERFORMANCE=y

Downclocking influences not only the uart1 baud rate but also the spi0 clock. We attach Ethernet chips to spi0, throughput was significantly worse with the ondemand governor (which is what we used previously). We felt that maximum Ethernet performance outweighs the relatively small powersaving gains.

Quoted 3 lines

In that regard I suggest you use the upstream cpufreq driver which behaves properly in that regard. It disables GPU freq scaling, so as to change CPU frequencies without SPI/I2C/UART issues.

Okay, I'll take a look. But the uart1 baudrate will still be wrong if the firmware downclocks because of overheating, right?

Thanks,

Lukas

#7 lore raw
Quoted 6 lines · Lukas Wunner
Quoted 1 line · Nicolas Saenz Julienne

BTW did you had the oportunity to have a go at the patch?

I've just performed a quick test and it doesn't work for me. If I add stdout-path = "serial1:115200n8"; to the chosen node, I only get a regular console with this patch, not an earlycon.

That's surprising, you're using u-boot isn't it? and the upstream device tree?

Quoted 16 lines
Quoted 16 lines
Quoted 15 lines

The problem is that in mainline, bcm2835_defconfig contains:
CONFIG_CPU_FREQ_DEFAULT_GOV_CONSERVATIVE=y

Likewise in the Foundation's downstream tree, bcmrpi_defconfig as well as bcm2711_defconfig and bcm2709_defconfig contain:
CONFIG_CPU_FREQ_DEFAULT_GOV_POWERSAVE=y

In contrast to this, we set the following on Revolution Pi devices:
CONFIG_CPU_FREQ_DEFAULT_GOV_PERFORMANCE=y

Downclocking influences not only the uart1 baud rate but also the spi0 clock. We attach Ethernet chips to spi0, throughput was significantly worse with the ondemand governor (which is what we used previously). We felt that maximum Ethernet performance outweighs the relatively small powersaving gains.

Quoted 6 lines
Quoted 3 lines

In that regard I suggest you use the upstream cpufreq driver which behaves properly in that regard. It disables GPU freq scaling, so as to change CPU frequencies without SPI/I2C/UART issues.

Okay, I'll take a look. But the uart1 baudrate will still be wrong if the firmware downclocks because of overheating, right?

You're right, it might happen. The way I understand it you're bound to leave the GPU at it's lower frequency if you want to use those peripherals and for them to be reliable. You could technically try to empirically fine tune the CPU thermal trip point so as to make sure the upstream kernel cpufreq downclock always kicks in before videocore's one. I'd actually like to see someone try that. However short of using an RT kernel It think you'll never be 100% sure it can never happen.

Regards,
Nicolas

#8 lore raw
Quoted 9 lines · Nicolas Saenz Julienne
Quoted 6 lines · Lukas Wunner
Quoted 1 line · Nicolas Saenz Julienne

BTW did you had the oportunity to have a go at the patch?

I've just performed a quick test and it doesn't work for me. If I add stdout-path = "serial1:115200n8"; to the chosen node, I only get a regular console with this patch, not an earlycon.

That's surprising, you're using u-boot isn't it? and the upstream device tree?

My apologies for the delay.

We boot the kernel directly from the Foundation's bootloader without U-Boot as intermediary, and we use the downstream DT and kernel (with RT & custom patches).

As far as I could see, the 8250_bcm2835aux UART wasn't even found when the kernel searched for available earlycons. The compatible string in the DT does match the one in Matthias' patch.

I notice that Greg has already applied the patch to tty/next. I'm currently stretched a little thin but I'll debug this further once the dust has settled here.

Thanks,

Lukas

#9 lore raw
Quoted 13 lines · Lukas Wunner
Quoted 9 lines · Nicolas Saenz Julienne
Quoted 6 lines · Lukas Wunner
Quoted 1 line · Nicolas Saenz Julienne

BTW did you had the oportunity to have a go at the patch?

I've just performed a quick test and it doesn't work for me. If I add stdout-path = "serial1:115200n8"; to the chosen node, I only get a regular console with this patch, not an earlycon.

That's surprising, you're using u-boot isn't it? and the upstream device tree?

My apologies for the delay.

Quoted 8 lines

We boot the kernel directly from the Foundation's bootloader without U-Boot as intermediary, and we use the downstream DT and kernel (with RT & custom patches).

As far as I could see, the 8250_bcm2835aux UART wasn't even found when the kernel searched for available earlycons. The compatible string in the DT does match the one in Matthias' patch.

Can you provide your device tree please. I suspect that we find the culprit there.

Regards,
Matthias

Quoted 8 lines

I notice that Greg has already applied the patch to tty/next. I'm currently stretched a little thin but I'll debug this further once the dust has settled here.

Thanks,

Lukas