[PATCH v2 0/3] Early 8250 fixlets

[PATCH v2 0/3] Early 8250 fixlets

[Vineet Gupta]

Hi,

This series addresses some of the issues seen on early 8250 uart when
working with ARC SDP platform.


Changes since v1 [1]
	-Rebased to linux-next of 20150105
	-Patch 3/3 SNAFU fixed (arch over-ride was not correct due to mix up
	 of orig work based off 3.13 vs. 3.19+ based linux next submission
	 (different core earlycon code)
Thx,
-Vineet

[1] http://www.spinics.net/lists/kernel/msg1893299.html

Vineet Gupta (3):
  serial: 8250_early: optimize early 8250 uart
  serial: 8250_early: prepare for dynamic BASE_BAUD
  ARC: runtime determine BASE_BAUD per machine

 arch/arc/include/asm/serial.h        | 24 ++++--------------------
 arch/arc/kernel/devtree.c            | 24 ++++++++++++++++++++++++
 drivers/tty/serial/8250/8250_early.c |  7 +++++--
 drivers/tty/serial/earlycon.c        | 13 +++++++++++--
 4 files changed, 44 insertions(+), 24 deletions(-)

-- 
1.9.1

[PATCH v2 1/3] serial: 8250_early: optimize early 8250 uart

[Vineet Gupta]

In early 8250, IER is already zero so no point in writing this - twice
per char.

This helped improve the SystemC model based ARC OSCI platform

Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Jiri Slaby <jslaby@suse.cz>
Cc: linux-serial@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Cc: Rob Herring <robh@kernel.org>
---
 drivers/tty/serial/8250/8250_early.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/drivers/tty/serial/8250/8250_early.c b/drivers/tty/serial/8250/8250_early.c
index 4858b8a99d3b..ce2a8abd1f54 100644
--- a/drivers/tty/serial/8250/8250_early.c
+++ b/drivers/tty/serial/8250/8250_early.c
@@ -95,13 +95,16 @@ static void __init early_serial8250_write(struct console *console,
 
 	/* Save the IER and disable interrupts */
 	ier = serial8250_early_in(port, UART_IER);
-	serial8250_early_out(port, UART_IER, 0);
+	if (ier)
+		serial8250_early_out(port, UART_IER, 0);
 
 	uart_console_write(port, s, count, serial_putc);
 
 	/* Wait for transmitter to become empty and restore the IER */
 	wait_for_xmitr(port);
-	serial8250_early_out(port, UART_IER, ier);
+
+	if (ier)
+		serial8250_early_out(port, UART_IER, ier);
 }
 
 static unsigned int __init probe_baud(struct uart_port *port)
-- 
1.9.1

[PATCH v2 2/3] serial: 8250_early: prepare for dynamic BASE_BAUD

[Vineet Gupta]

BASE_BAUD is pain in neck for multi-platform support as it hard codes at
build time the clk value for early uart.

Mitigate this by allowing arches/platforms to provide their own version
which can do dynamic setup based on DT values etc (see next patch for
usage)

This was needed for ARC SDP platforms based on 2 different FPGA flows, each
with a different UART clk value.

Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Jiri Slaby <jslaby@suse.cz>
Cc: linux-serial@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Cc: Rob Herring <robh@kernel.org>
Cc: Arnd Bergmann <arnd@arndb.de>
---
 drivers/tty/serial/earlycon.c | 13 +++++++++++--
 1 file changed, 11 insertions(+), 2 deletions(-)

diff --git a/drivers/tty/serial/earlycon.c b/drivers/tty/serial/earlycon.c
index 64fe25a4285c..bfd3537739e0 100644
--- a/drivers/tty/serial/earlycon.c
+++ b/drivers/tty/serial/earlycon.c
@@ -54,6 +54,15 @@ static void __iomem * __init earlycon_map(unsigned long paddr, size_t size)
 	return base;
 }
 
+unsigned int __weak __init earlycon_base_baud(char *options)
+{
+#ifdef BASE_BAUD
+	return BASE_BAUD;
+#else
+	return 1843200/16;	/* x86 early console */
+#endif
+}
+
 static int __init parse_options(struct earlycon_device *device,
 				char *options)
 {
@@ -87,7 +96,7 @@ static int __init parse_options(struct earlycon_device *device,
 		return -EINVAL;
 	}
 
-	port->uartclk = BASE_BAUD * 16;
+	port->uartclk = earlycon_base_baud(options) * 16;
 
 	options = strchr(options, ',');
 	if (options) {
@@ -156,7 +165,7 @@ int __init of_setup_earlycon(unsigned long addr,
 
 	port->iotype = UPIO_MEM;
 	port->mapbase = addr;
-	port->uartclk = BASE_BAUD * 16;
+	port->uartclk = earlycon_base_baud(NULL) * 16;
 	port->membase = earlycon_map(addr, SZ_4K);
 
 	early_console_dev.con->data = &early_console_dev;
-- 
1.9.1

Re: [PATCH v2 2/3] serial: 8250_early: prepare for dynamic BASE_BAUD

[Peter Hurley]

Hi Vineet,

On 01/05/2015 06:30 AM, Vineet Gupta wrote:
> BASE_BAUD is pain in neck for multi-platform support as it hard codes at
> build time the clk value for early uart.
> 
> Mitigate this by allowing arches/platforms to provide their own version
> which can do dynamic setup based on DT values etc (see next patch for
> usage)
> 
> This was needed for ARC SDP platforms based on 2 different FPGA flows, each
> with a different UART clk value.
> 
> Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
> Cc: Jiri Slaby <jslaby@suse.cz>
> Cc: linux-serial@vger.kernel.org
> Cc: linux-kernel@vger.kernel.org
> Cc: Rob Herring <robh@kernel.org>
> Cc: Arnd Bergmann <arnd@arndb.de>
> ---
>  drivers/tty/serial/earlycon.c | 13 +++++++++++--
>  1 file changed, 11 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/tty/serial/earlycon.c b/drivers/tty/serial/earlycon.c
> index 64fe25a4285c..bfd3537739e0 100644
> --- a/drivers/tty/serial/earlycon.c
> +++ b/drivers/tty/serial/earlycon.c
> @@ -54,6 +54,15 @@ static void __iomem * __init earlycon_map(unsigned long paddr, size_t size)
>  	return base;
>  }
>  
> +unsigned int __weak __init earlycon_base_baud(char *options)
> +{
> +#ifdef BASE_BAUD
> +	return BASE_BAUD;
> +#else
> +	return 1843200/16;	/* x86 early console */
> +#endif
> +}
> +
>  static int __init parse_options(struct earlycon_device *device,
>  				char *options)
>  {
> @@ -87,7 +96,7 @@ static int __init parse_options(struct earlycon_device *device,
>  		return -EINVAL;
>  	}
>  
> -	port->uartclk = BASE_BAUD * 16;
> +	port->uartclk = earlycon_base_baud(options) * 16;

I don't understand why BASE_BAUD can't be defined by arc as arc_earlycon_base_baud().
Why is the weak binding necessary?

I'm pushing back on this because port initialization (and especially 8250)
is already a ridiculous spiderweb.

Regards,
Peter Hurley

>  	options = strchr(options, ',');
>  	if (options) {
> @@ -156,7 +165,7 @@ int __init of_setup_earlycon(unsigned long addr,
>  
>  	port->iotype = UPIO_MEM;
>  	port->mapbase = addr;
> -	port->uartclk = BASE_BAUD * 16;
> +	port->uartclk = earlycon_base_baud(NULL) * 16;
>  	port->membase = earlycon_map(addr, SZ_4K);
>  
>  	early_console_dev.con->data = &early_console_dev;
> 

Re: [PATCH v2 2/3] serial: 8250_early: prepare for dynamic BASE_BAUD

[Vineet Gupta]

Hi Peter,

On Monday 05 January 2015 06:31 PM, Peter Hurley wrote:
> Hi Vineet,
>
> On 01/05/2015 06:30 AM, Vineet Gupta wrote:
>> -	port->uartclk = BASE_BAUD * 16;
>> +	port->uartclk = earlycon_base_baud(options) * 16;
> I don't understand why BASE_BAUD can't be defined by arc as arc_earlycon_base_baud().
> Why is the weak binding necessary?

Indeed weak semantics are not really needed - even now we have what u r suggesting
- the missing part was existing conditional definition which will simply go away
if I implement the fdt parsing in that singly defined function. Thx for the tip :-)

In that case this requires no change to driver and I can fixup the platform -
perhaps ur Ack on next version will be nice.
Also what do u think of the early IER optimization ?

>
> I'm pushing back on this because port initialization (and especially 8250)
> is already a ridiculous spiderweb.

I totally agree.

Thx,
-Vineet

>
> Regards,
> Peter Hurley
>
>>  	options = strchr(options, ',');
>>  	if (options) {
>> @@ -156,7 +165,7 @@ int __init of_setup_earlycon(unsigned long addr,
>>  
>>  	port->iotype = UPIO_MEM;
>>  	port->mapbase = addr;
>> -	port->uartclk = BASE_BAUD * 16;
>> +	port->uartclk = earlycon_base_baud(NULL) * 16;
>>  	port->membase = earlycon_map(addr, SZ_4K);
>>  
>>  	early_console_dev.con->data = &early_console_dev;
>>
>

Re: [PATCH v2 2/3] serial: 8250_early: prepare for dynamic BASE_BAUD

[Peter Hurley]

On 01/05/2015 08:21 AM, Vineet Gupta wrote:
> Hi Peter,
> 
> On Monday 05 January 2015 06:31 PM, Peter Hurley wrote:
>> Hi Vineet,
>>
>> On 01/05/2015 06:30 AM, Vineet Gupta wrote:
>>> -	port->uartclk = BASE_BAUD * 16;
>>> +	port->uartclk = earlycon_base_baud(options) * 16;
>> I don't understand why BASE_BAUD can't be defined by arc as arc_earlycon_base_baud().
>> Why is the weak binding necessary?
> 
> Indeed weak semantics are not really needed - even now we have what u r suggesting
> - the missing part was existing conditional definition which will simply go away
> if I implement the fdt parsing in that singly defined function. Thx for the tip :-)
> 
> In that case this requires no change to driver and I can fixup the platform -
> perhaps ur Ack on next version will be nice.

Will do.

> Also what do u think of the early IER optimization ?

Seems ok except the commit changelog is wrong: IER is disabled twice per _line_.
How did you profile it? It's ok if the answer is "seemed faster"; I'm just
curious.

Regards,
Peter Hurley

>> I'm pushing back on this because port initialization (and especially 8250)
>> is already a ridiculous spiderweb.
> 
> I totally agree.
> 
> Thx,
> -Vineet
> 
>>
>> Regards,
>> Peter Hurley
>>
>>>  	options = strchr(options, ',');
>>>  	if (options) {
>>> @@ -156,7 +165,7 @@ int __init of_setup_earlycon(unsigned long addr,
>>>  
>>>  	port->iotype = UPIO_MEM;
>>>  	port->mapbase = addr;
>>> -	port->uartclk = BASE_BAUD * 16;
>>> +	port->uartclk = earlycon_base_baud(NULL) * 16;
>>>  	port->membase = earlycon_map(addr, SZ_4K);
>>>  
>>>  	early_console_dev.con->data = &early_console_dev;
>>>
>>
> 

Re: [PATCH v2 2/3] serial: 8250_early: prepare for dynamic BASE_BAUD

[Vineet Gupta]

On Monday 05 January 2015 07:04 PM, Peter Hurley wrote:
>> Also what do u think of the early IER optimization ?
> Seems ok except the commit changelog is wrong: IER is disabled twice per _line_.
> How did you profile it? It's ok if the answer is "seemed faster"; I'm just
> curious.

Looking at the code yes this would be per line - but I vaguely remember seeing it
for every char - this was when debugging a systemC model for 8250. Its been a while...
I'd added prints in the model whenever IER would wiggle and it was just being
written with 0's again and again when in the early polling based printing mode.

I'll send a revised version with fixed changelog.

Thx,
-Vineet

Re: [PATCH v2 2/3] serial: 8250_early: prepare for dynamic BASE_BAUD

[Peter Hurley]

On 01/05/2015 08:52 AM, Vineet Gupta wrote:
> On Monday 05 January 2015 07:04 PM, Peter Hurley wrote:
>>> Also what do u think of the early IER optimization ?
>> Seems ok except the commit changelog is wrong: IER is disabled twice per _line_.
>> How did you profile it? It's ok if the answer is "seemed faster"; I'm just
>> curious.
> 
> Looking at the code yes this would be per line - but I vaguely remember seeing it
> for every char - this was when debugging a systemC model for 8250. Its been a while...
> I'd added prints in the model whenever IER would wiggle and it was just being
> written with 0's again and again when in the early polling based printing mode.

Maybe that was while running kgdb through serial8250_put_poll_char(), which is
per-char.

Regards,
Peter Hurley

[PATCH v2 3/3] ARC: runtime determine BASE_BAUD per machine

[Vineet Gupta]

Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Jiri Slaby <jslaby@suse.cz>
Cc: linux-serial@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Cc: Rob Herring <robh@kernel.org>
Cc: Arnd Bergmann <arnd@arndb.de>
---
 arch/arc/include/asm/serial.h | 24 ++++--------------------
 arch/arc/kernel/devtree.c     | 24 ++++++++++++++++++++++++
 2 files changed, 28 insertions(+), 20 deletions(-)

diff --git a/arch/arc/include/asm/serial.h b/arch/arc/include/asm/serial.h
index 602b0970a764..a8c0b8743709 100644
--- a/arch/arc/include/asm/serial.h
+++ b/arch/arc/include/asm/serial.h
@@ -10,26 +10,10 @@
 #define _ASM_ARC_SERIAL_H
 
 /*
- * early-8250 requires BASE_BAUD to be defined and includes this header.
- * We put in a typical value:
- * 	(core clk / 16) - i.e. UART samples 16 times per sec.
- * Athough in multi-platform-image this might not work, specially if the
- * clk driving the UART is different.
- * We can't use DeviceTree as this is typically for early serial.
+ * 8250_early (now earlycon) include this hdr and expect BASE_BAUD definition
+ * for setting up uartclk.
+ * Now that is done via arch provided earlycon_base_baud() which can use FDT
+ * provided info as well to provide uartclk based on DT in use.
  */
 
-#include <asm/clk.h>
-
-#define BASE_BAUD	(arc_get_core_freq() / 16)
-
-/*
- * This is definitely going to break early 8250 consoles on multi-platform
- * images but hey, it won't add any code complexity for a debug feature of
- * one broken driver.
- */
-#ifdef CONFIG_ARC_PLAT_TB10X
-#undef BASE_BAUD
-#define BASE_BAUD	(arc_get_core_freq() / 16 / 3)
-#endif
-
 #endif /* _ASM_ARC_SERIAL_H */
diff --git a/arch/arc/kernel/devtree.c b/arch/arc/kernel/devtree.c
index fffdb5e41b20..4530f5a7eeb1 100644
--- a/arch/arc/kernel/devtree.c
+++ b/arch/arc/kernel/devtree.c
@@ -17,6 +17,28 @@
 #include <asm/clk.h>
 #include <asm/mach_desc.h>
 
+#ifdef CONFIG_SERIAL_8250_CONSOLE
+
+static unsigned int arc_base_baud = 33333333;
+
+unsigned int __init earlycon_base_baud(char *options)
+{
+	return arc_base_baud/16;
+}
+
+static void __init arc_set_early_base_baud(unsigned long dt_root)
+{
+	unsigned int core_clk = arc_get_core_freq();
+
+	if (of_flat_dt_is_compatible(dt_root, "abilis,arc-tb10x"))
+		arc_base_baud = core_clk/3;
+	else
+		arc_base_baud = core_clk;
+}
+#else
+#define arc_set_early_base_baud(dt_root)
+#endif
+
 static const void * __init arch_get_next_mach(const char *const **match)
 {
 	static const struct machine_desc *mdesc = __arch_info_begin;
@@ -56,5 +78,7 @@ const struct machine_desc * __init setup_machine_fdt(void *dt)
 	if (clk)
 		arc_set_core_freq(of_read_ulong(clk, len/4));
 
+	arc_set_early_base_baud(dt_root);
+
 	return mdesc;
 }
-- 
1.9.1