[PATCH v2 00/15] serial: add new I/O type for SPI and I2C bus devices

[PATCH v2 00/15] serial: add new I/O type for SPI and I2C bus devices

[Hugo Villeneuve]

Hello,
this patch series add a new I/O type for serial devices on a SPI/I2C bus,
and a few related cleanups/improvements.

These changes are based on a suggestion [1] made by Ilpo Järvinen during
past sc16is7xx driver patches review.

The first patches make use of uart_iotype_*() functions to simplify and
cleanup the UART report functions.

This cleanup sets the stage to avoid displaying irrelevant MMIO or legacy
I/O information for serial devices on I2C or SPI busses when we add the new
UPIO_BUS type.

The patch "serial: uniformize serial port I/O infos display" is more of an
RFC as an attempt to uniformize the display of I/O informations. If deemed
usefull/appropriate, it can be reworked and integrasted with previous
patches.

Tested on a imx6 board with two SC16is752 using SPI mode, and a dummy
device-tree entry for a MAX3100:
  dmesg -t | grep "base_baud"
    2020000.serial: ttymxc0 at MMIO 0x2020000 (irq = 197, base_baud = 5000000) is a IMX
    21e8000.serial: ttymxc1 at MMIO 0x21e8000 (irq = 198, base_baud = 5000000) is a IMX
    spi1.0: ttySC0 (irq = 165, base_baud = 1500000) is a SC16IS752
    spi1.0: ttySC1 (irq = 165, base_baud = 1500000) is a SC16IS752
    spi3.0: ttySC2 (irq = 37, base_baud = 1500000) is a SC16IS752
    spi3.0: ttySC3 (irq = 37, base_baud = 1500000) is a SC16IS752
    spi3.1: ttyMAX0 (irq = 0, base_baud = 0) is a MAX3100
Note that before these patches, max3100 silently failed in uart_configure_port()
because membase/iobase/mapbase were zero.

For max310x and SC16is7xx in i2c mode, tested only that driver is properly
registered by using i2c-stub.

Also tested on a custom board with a Renesas RZ/G2L cpu (sh-sci driver) to
confirm there is no regression (also with earlycon):
  dmesg -t | grep "MMIO"
    earlycon: scif0 at MMIO 0x000000001004b800 (options '115200n8')
    1004b800.serial: ttySC0 at MMIO 0x000000001004b800 (irq = 35, base_baud = 0) is a scif
    1004bc00.serial: ttySC1 at MMIO 0x000000001004bc00 (irq = 40, base_baud = 0) is a scif

Maybe some of the patches could be merged together, but for now I decided
to keep them separate to help the review process.

Thank you.

[1] https://lore.kernel.org/lkml/2936e18f-44ea-faed-9fa0-2ddefe7c3194@linux.intel.com/raw

Link: [v1] https://lore.kernel.org/all/20260423-tty-upio-v1-0-baf82d3b86d1@dimonoff.com/

Changes for v2:
- Replace snprintf with scnprintf (Ilpo)
- Fix alignment of : to ? (Ilpo)
- Move hub6 match port to 8250_hub6.c (Ilpo)
- earlycon: use uart_iotype_*()
- 8250_rsa: use uart_iotype_*()
- 8250_port: use uart_iotype_*() in serial_port_out_sync()
- simplified uart_line_info() even more and remove intermediate mmio
  variable
- Split and rename patch serial: core: prevent irrelevant I/O infos display
  for UPIO_BUS to serial: core: use uart_iotype_*() to simplify
  uart_report_port() and moved before UPIO_BUS patch.
- Add a new patch to implement a function to uniformize the display
  of I/O infos for uart_report_port(), uart_line_info() and
  earlycon_print_info(). If accepted could be better integrated/merged with
  the next series submission. If not, can simply be dropped.
- Add patch for uart_regiowidth_to_iotype()

---
Hugo Villeneuve (15):
      serial: 8250_hub6: add hub6_match_port()
      serial: core: add uart_iotype_mmio/legacy_io helper functions
      serial: core: use uart_iotype_*() to simplify uart_match_port()
      serial: core: use uart_iotype_*() to simplify uart_line_info()
      serial: core: replace snprintf with more robust scnprintf
      serial: core: fix indentation/alignment
      serial: core: use uart_iotype_*() to simplify uart_report_port()
      serial: earlycon: use uart_iotype_*() to simplify code
      serial: 8250: use uart_iotype_*() to simplify code
      serial: 8250_rsa: use uart_iotype_*() to simplify code
      serial: core: add new I/O type for SPI and I2C bus devices
      serial: sc16is7xx: use new UPIO_BUS as iotype
      serial: max310x: use new UPIO_BUS as iotype
      serial: max3100: use new UPIO_BUS as iotype
      serial: uniformize serial port I/O infos display

 drivers/tty/serial/8250/8250.h      |   7 ++
 drivers/tty/serial/8250/8250_hub6.c |   6 ++
 drivers/tty/serial/8250/8250_port.c |  43 ++---------
 drivers/tty/serial/8250/8250_rsa.c  |  40 +++++------
 drivers/tty/serial/earlycon.c       |  17 ++---
 drivers/tty/serial/max3100.c        |   1 +
 drivers/tty/serial/max310x.c        |   9 +--
 drivers/tty/serial/sc16is7xx.c      |   9 +--
 drivers/tty/serial/serial_core.c    | 137 ++++++++++++++++++++++--------------
 include/linux/serial_core.h         |   6 ++
 include/uapi/linux/serial.h         |   1 +
 11 files changed, 137 insertions(+), 139 deletions(-)
---
base-commit: 254f49634ee16a731174d2ae34bc50bd5f45e731
change-id: 20260428-tty-upio-93a350a4d69a

Best regards,
-- 
Hugo Villeneuve <hvilleneuve@dimonoff.com>

[PATCH v2 01/15] serial: 8250_hub6: add hub6_match_port()

[Hugo Villeneuve]

From: Hugo Villeneuve <hvilleneuve@dimonoff.com>

Move the entire hub6 related match port check into its own function in
8250_hub6.c and add a stub for the case when hub6 code is not even built
into kernel.

Suggested-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Signed-off-by: Hugo Villeneuve <hvilleneuve@dimonoff.com>
---
 drivers/tty/serial/8250/8250.h      | 7 +++++++
 drivers/tty/serial/8250/8250_hub6.c | 6 ++++++
 drivers/tty/serial/serial_core.c    | 4 ++--
 3 files changed, 15 insertions(+), 2 deletions(-)

diff --git a/drivers/tty/serial/8250/8250.h b/drivers/tty/serial/8250/8250.h
index 77fe0588fd6bd5c5baac0fa03b7c466742aa9432..9d1068d0489dc4285b1d30489926bd676bcbe60e 100644
--- a/drivers/tty/serial/8250/8250.h
+++ b/drivers/tty/serial/8250/8250.h
@@ -334,6 +334,13 @@ int fintek_8250_probe(struct uart_8250_port *uart);
 static inline int fintek_8250_probe(struct uart_8250_port *uart) { return 0; }
 #endif
 
+#ifdef CONFIG_SERIAL_8250_HUB6
+bool hub6_match_port(const struct uart_port *port1, const struct uart_port *port2);
+#else
+static inline bool hub6_match_port(const struct uart_port *port1, const struct uart_port *port2)
+{ return false; }
+#endif
+
 #ifdef CONFIG_ARCH_OMAP1
 #include <linux/soc/ti/omap1-soc.h>
 static inline int is_omap1_8250(struct uart_8250_port *pt)
diff --git a/drivers/tty/serial/8250/8250_hub6.c b/drivers/tty/serial/8250/8250_hub6.c
index 273f59b9bca555e9fb035005d3df2089c5b0a039..eae32c924e29bcaf608d66988aca461b9a53dc91 100644
--- a/drivers/tty/serial/8250/8250_hub6.c
+++ b/drivers/tty/serial/8250/8250_hub6.c
@@ -41,6 +41,12 @@ static struct platform_device hub6_device = {
 	},
 };
 
+bool hub6_match_port(const struct uart_port *port1, const struct uart_port *port2)
+{
+	return port1->iobase == port2->iobase && port1->hub6 == port2->hub6;
+}
+EXPORT_SYMBOL_GPL(hub6_match_port);
+
 static int __init hub6_init(void)
 {
 	return platform_device_register(&hub6_device);
diff --git a/drivers/tty/serial/serial_core.c b/drivers/tty/serial/serial_core.c
index 89cebdd278410a5a97f3f5ee1b9c171715145067..7354f8af4853d34e5e819c2887f287970c653b69 100644
--- a/drivers/tty/serial/serial_core.c
+++ b/drivers/tty/serial/serial_core.c
@@ -33,6 +33,7 @@
 #include <linux/uaccess.h>
 
 #include "serial_base.h"
+#include "8250/8250.h" /* For hub6_match_port() */
 
 /*
  * This is used to lock changes in serial line configuration.
@@ -3198,8 +3199,7 @@ bool uart_match_port(const struct uart_port *port1,
 	case UPIO_PORT:
 		return port1->iobase == port2->iobase;
 	case UPIO_HUB6:
-		return port1->iobase == port2->iobase &&
-		       port1->hub6   == port2->hub6;
+		return hub6_match_port(port1, port2);
 	case UPIO_MEM:
 	case UPIO_MEM16:
 	case UPIO_MEM32:

-- 
2.47.3

[PATCH v2 02/15] serial: core: add uart_iotype_mmio/legacy_io helper functions

[Hugo Villeneuve]

From: Hugo Villeneuve <hvilleneuve@dimonoff.com>

To help simplify code that check on the io type mode of the port.

Signed-off-by: Hugo Villeneuve <hvilleneuve@dimonoff.com>
---
Will also be used by future commit that add UPIO_BUS.
---
 drivers/tty/serial/serial_core.c | 28 ++++++++++++++++++++++++++++
 include/linux/serial_core.h      |  4 ++++
 2 files changed, 32 insertions(+)

diff --git a/drivers/tty/serial/serial_core.c b/drivers/tty/serial/serial_core.c
index 7354f8af4853d34e5e819c2887f287970c653b69..c177054f9515e1cb5a1e557c0dac096cb1cbd8cf 100644
--- a/drivers/tty/serial/serial_core.c
+++ b/drivers/tty/serial/serial_core.c
@@ -1967,6 +1967,34 @@ static const char *uart_type(struct uart_port *port)
 	return str;
 }
 
+bool uart_iotype_mmio(enum uart_iotype iotype)
+{
+	switch (iotype) {
+	case UPIO_MEM:
+	case UPIO_MEM32:
+	case UPIO_AU:
+	case UPIO_TSI:
+	case UPIO_MEM32BE:
+	case UPIO_MEM16:
+		return true;
+	default:
+		return false;
+	}
+}
+EXPORT_SYMBOL_GPL(uart_iotype_mmio);
+
+bool uart_iotype_legacy_io(enum uart_iotype iotype)
+{
+	switch (iotype) {
+	case UPIO_PORT:
+	case UPIO_HUB6:
+		return true;
+	default:
+		return false;
+	}
+}
+EXPORT_SYMBOL_GPL(uart_iotype_legacy_io);
+
 #ifdef CONFIG_PROC_FS
 
 static void uart_line_info(struct seq_file *m, struct uart_state *state)
diff --git a/include/linux/serial_core.h b/include/linux/serial_core.h
index 666430b4789977ae19641442b369eb1c773233f4..fc3f5ebe389658c197ffc105ce4ac11cacef59bb 100644
--- a/include/linux/serial_core.h
+++ b/include/linux/serial_core.h
@@ -1306,4 +1306,8 @@ static inline int uart_handle_break(struct uart_port *port)
 					 !((cflag) & CLOCAL))
 
 int uart_get_rs485_mode(struct uart_port *port);
+
+bool uart_iotype_mmio(enum uart_iotype iotype);
+bool uart_iotype_legacy_io(enum uart_iotype iotype);
+
 #endif /* LINUX_SERIAL_CORE_H */

-- 
2.47.3

Re: [PATCH v2 02/15] serial: core: add uart_iotype_mmio/legacy_io helper functions

[Andy Shevchenko]

On Tue, Apr 28, 2026 at 01:53:48PM -0400, Hugo Villeneuve wrote:

> To help simplify code that check on the io type mode of the port.

...

> +bool uart_iotype_legacy_io(enum uart_iotype iotype)

Why do we use 'legacy'? Still in use in modern CPUs...

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v2 02/15] serial: core: add uart_iotype_mmio/legacy_io helper functions

[Hugo Villeneuve]

On Thu, 30 Apr 2026 17:08:25 +0200
Andy Shevchenko <andriy.shevchenko@intel.com> wrote:

> On Tue, Apr 28, 2026 at 01:53:48PM -0400, Hugo Villeneuve wrote:
> 
> > To help simplify code that check on the io type mode of the port.
> 
> ...
> 
> > +bool uart_iotype_legacy_io(enum uart_iotype iotype)
> 
> Why do we use 'legacy'? Still in use in modern CPUs...

TBH I do not remember exactly where i got this, but what would you
suggest?

Maybe:
  uart_iotype_io()


Hugo.

Re: [PATCH v2 02/15] serial: core: add uart_iotype_mmio/legacy_io helper functions

[Andy Shevchenko]

On Thu, Apr 30, 2026 at 11:30:11AM -0400, Hugo Villeneuve wrote:
> On Thu, 30 Apr 2026 17:08:25 +0200
> Andy Shevchenko <andriy.shevchenko@intel.com> wrote:
> > On Tue, Apr 28, 2026 at 01:53:48PM -0400, Hugo Villeneuve wrote:

...

> > > +bool uart_iotype_legacy_io(enum uart_iotype iotype)
> > 
> > Why do we use 'legacy'? Still in use in modern CPUs...
> 
> TBH I do not remember exactly where i got this, but what would you
> suggest?
> 
> Maybe:
>   uart_iotype_io()

Yep!

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v2 02/15] serial: core: add uart_iotype_mmio/legacy_io helper functions

[Maciej W. Rozycki]

On Thu, 30 Apr 2026, Andy Shevchenko wrote:

> > +bool uart_iotype_legacy_io(enum uart_iotype iotype)
> 
> Why do we use 'legacy'? Still in use in modern CPUs...

 Deprecated in PCIe and not available in numerous systems.  Also actually 
called "legacy" in some serial port datasheets aged ~20 years now.  While 
some contemporary CPUs indeed retain the port I/O address space, it's for 
legacy use anyway, you don't want to rely on it in new designs.

  Maciej

Re: [PATCH v2 02/15] serial: core: add uart_iotype_mmio/legacy_io helper functions

[Andy Shevchenko]

On Sun, May 03, 2026 at 12:25:24AM +0100, Maciej W. Rozycki wrote:
> On Thu, 30 Apr 2026, Andy Shevchenko wrote:
> 
> > > +bool uart_iotype_legacy_io(enum uart_iotype iotype)
> > 
> > Why do we use 'legacy'? Still in use in modern CPUs...
> 
>  Deprecated in PCIe and not available in numerous systems.  Also actually 
> called "legacy" in some serial port datasheets aged ~20 years now.  While 
> some contemporary CPUs indeed retain the port I/O address space, it's for 
> legacy use anyway, you don't want to rely on it in new designs.

For the holder of the new (modern) CPU which supports the IO ports, this is
definitely not a legacy interface despite on whatever PCIe or other datasheets
call it.

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v2 02/15] serial: core: add uart_iotype_mmio/legacy_io helper functions

[Maciej W. Rozycki]

On Mon, 4 May 2026, Andy Shevchenko wrote:

> > > Why do we use 'legacy'? Still in use in modern CPUs...
> > 
> >  Deprecated in PCIe and not available in numerous systems.  Also actually 
> > called "legacy" in some serial port datasheets aged ~20 years now.  While 
> > some contemporary CPUs indeed retain the port I/O address space, it's for 
> > legacy use anyway, you don't want to rely on it in new designs.
> 
> For the holder of the new (modern) CPU which supports the IO ports, this is
> definitely not a legacy interface despite on whatever PCIe or other datasheets
> call it.

 I appreciate your point of view, however I disagree that the presence of 
an interface in a contemporary chip makes the interface modern.

 I see the port I/O space so much legacy as say the PC/AT DMA controller 
(8237 pair), which is similarly present in current x86 chipsets.  If this 
stuff was not present, such as say the PC/AT interrupt controller (8259A 
pair), which I believe has been removed from some x86 system designs, then 
it would be obsolete/removed rather than legacy.

 It is analogous to PCI/e systems that lack a southbridge and are called 
"legacy-free", as the whole southbridge stuff, the main consumer of the 
port I/O space still remaining in use, is legacy nowadays (the other one I 
know of being the 8255-based PC parallel port, which has been considered a 
legacy interface as well, even though you can still buy and plug one into 
a modern PCIe system).

 NB I have a couple of modern x86 CPUs around too that support the port 
I/O space, but it doesn't change my view as to the nomenclature.

 FWIW,

  Maciej

Re: [PATCH v2 02/15] serial: core: add uart_iotype_mmio/legacy_io helper functions

[Andy Shevchenko]

On Mon, May 04, 2026 at 12:44:41PM +0100, Maciej W. Rozycki wrote:
> On Mon, 4 May 2026, Andy Shevchenko wrote:

...

> > > > Why do we use 'legacy'? Still in use in modern CPUs...
> > > 
> > >  Deprecated in PCIe and not available in numerous systems.  Also actually 
> > > called "legacy" in some serial port datasheets aged ~20 years now.  While 
> > > some contemporary CPUs indeed retain the port I/O address space, it's for 
> > > legacy use anyway, you don't want to rely on it in new designs.
> > 
> > For the holder of the new (modern) CPU which supports the IO ports, this is
> > definitely not a legacy interface despite on whatever PCIe or other datasheets
> > call it.
> 
>  I appreciate your point of view, however I disagree that the presence of an
>  interface in a contemporary chip makes the interface modern.

I appreciate a bike shedding.

>  I see the port I/O space so much legacy as say the PC/AT DMA controller 
> (8237 pair), which is similarly present in current x86 chipsets.  If this 
> stuff was not present, such as say the PC/AT interrupt controller (8259A 
> pair), which I believe has been removed from some x86 system designs, then 
> it would be obsolete/removed rather than legacy.
> 
>  It is analogous to PCI/e systems that lack a southbridge and are called 
> "legacy-free", as the whole southbridge stuff, the main consumer of the 
> port I/O space still remaining in use, is legacy nowadays (the other one I 
> know of being the 8255-based PC parallel port, which has been considered a 
> legacy interface as well, even though you can still buy and plug one into 
> a modern PCIe system).
> 
>  NB I have a couple of modern x86 CPUs around too that support the port 
> I/O space, but it doesn't change my view as to the nomenclature.

Is the regular user assumed to go to dig to the mailing list (at best) or drawn
into the search of the documentation for all this? In their perspective they
possess modern CPU with support of that technology. If the same user possesses
the HW uses that ("legacy") technology, it doesn't make so in their eyes.

On top of that the additional word occupies more space in the code, making it
harder to read. It's just a redundant word. I see no value at all of using it.

P.S.
I see no value in this discussion. And I see no point to add a word where it's
not needed at all.

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v2 02/15] serial: core: add uart_iotype_mmio/legacy_io helper functions

[Maciej W. Rozycki]

On Mon, 4 May 2026, Andy Shevchenko wrote:

> I see no value in this discussion. And I see no point to add a word where it's
> not needed at all.

 Agreed.  I just pointed out that the use of "legacy" in this context is 
not incorrect.  And "regular users" won't see any of this stuff anyway.

  Maciej

Re: [PATCH v2 02/15] serial: core: add uart_iotype_mmio/legacy_io helper functions

[Andy Shevchenko]

On Mon, May 04, 2026 at 04:20:13PM +0100, Maciej W. Rozycki wrote:
> On Mon, 4 May 2026, Andy Shevchenko wrote:
> 
> > I see no value in this discussion. And I see no point to add a word where it's
> > not needed at all.
> 
>  Agreed.  I just pointed out that the use of "legacy" in this context is 
> not incorrect.  And "regular users" won't see any of this stuff anyway.

Good, thanks for the point!

-- 
With Best Regards,
Andy Shevchenko

[PATCH v2 03/15] serial: core: use uart_iotype_*() to simplify uart_match_port()

[Hugo Villeneuve]

From: Hugo Villeneuve <hvilleneuve@dimonoff.com>

Make use of new functions uart_iotype_mmio() and uart_iotype_legacy_io()
to simplify and improve code readability.

Signed-off-by: Hugo Villeneuve <hvilleneuve@dimonoff.com>
---
 drivers/tty/serial/serial_core.c | 16 ++++------------
 1 file changed, 4 insertions(+), 12 deletions(-)

diff --git a/drivers/tty/serial/serial_core.c b/drivers/tty/serial/serial_core.c
index c177054f9515e1cb5a1e557c0dac096cb1cbd8cf..06a774f93652d65bc8e0f1ed912aa3e5c95e2bfa 100644
--- a/drivers/tty/serial/serial_core.c
+++ b/drivers/tty/serial/serial_core.c
@@ -3222,22 +3222,14 @@ bool uart_match_port(const struct uart_port *port1,
 {
 	if (port1->iotype != port2->iotype)
 		return false;
-
-	switch (port1->iotype) {
-	case UPIO_PORT:
+	else if (port1->iotype == UPIO_PORT)
 		return port1->iobase == port2->iobase;
-	case UPIO_HUB6:
+	else if (port1->iotype == UPIO_HUB6)
 		return hub6_match_port(port1, port2);
-	case UPIO_MEM:
-	case UPIO_MEM16:
-	case UPIO_MEM32:
-	case UPIO_MEM32BE:
-	case UPIO_AU:
-	case UPIO_TSI:
+	else if (uart_iotype_mmio(port1->iotype))
 		return port1->mapbase == port2->mapbase;
-	default:
+	else
 		return false;
-	}
 }
 EXPORT_SYMBOL(uart_match_port);
 

-- 
2.47.3

[PATCH v2 04/15] serial: core: use uart_iotype_*() to simplify uart_line_info()

[Hugo Villeneuve]

From: Hugo Villeneuve <hvilleneuve@dimonoff.com>

Make use of new functions uart_iotype_mmio() and uart_iotype_legacy_io()
to simplify and improve code readability. This will prevent displaying
irrelevant information for future IO types (ex: UPIO_BUS), while also
addressing the (eventually) invalid check for "iotype >= UPIO_MEM".

This also allows us to remove the confusing cast to (unsigned long long)
for iobase which is defined as an unsigned long, and use %pa to display the
mapbase pointer, as it is done in earlycon_print_info().

Signed-off-by: Hugo Villeneuve <hvilleneuve@dimonoff.com>
---
On 32-bit cpu:
  no changes.
On 64-bit cpu:
  # cat /proc/tty/driver/sci
  before
    serinfo:1.0 driver revision:
    0: uart:scif mmio:0x1004B800 irq:35 tx:11192 rx:37 RTS|CTS|DTR|DSR|CD
  after:
    serinfo:1.0 driver revision:
    0: uart:scif mmio:0x000000001004b800 irq:35 tx:11339 rx:35 RTS|CTS|DTR|DSR|CD
---
 drivers/tty/serial/serial_core.c | 16 ++++++++--------
 1 file changed, 8 insertions(+), 8 deletions(-)

diff --git a/drivers/tty/serial/serial_core.c b/drivers/tty/serial/serial_core.c
index 06a774f93652d65bc8e0f1ed912aa3e5c95e2bfa..c5ef2dd00a9c9ad57c4010c05daea34fd2053676 100644
--- a/drivers/tty/serial/serial_core.c
+++ b/drivers/tty/serial/serial_core.c
@@ -2004,7 +2004,6 @@ static void uart_line_info(struct seq_file *m, struct uart_state *state)
 	struct uart_port *uport;
 	char stat_buf[32];
 	unsigned int status;
-	int mmio;
 
 	guard(mutex)(&port->mutex);
 
@@ -2012,13 +2011,14 @@ static void uart_line_info(struct seq_file *m, struct uart_state *state)
 	if (!uport)
 		return;
 
-	mmio = uport->iotype >= UPIO_MEM;
-	seq_printf(m, "%u: uart:%s %s%08llX irq:%u",
-			uport->line, uart_type(uport),
-			mmio ? "mmio:0x" : "port:",
-			mmio ? (unsigned long long)uport->mapbase
-			     : (unsigned long long)uport->iobase,
-			uport->irq);
+	seq_printf(m, "%u: uart:%s", uport->line, uart_type(uport));
+
+	if (uart_iotype_mmio(uport->iotype))
+		seq_printf(m, " mmio:%pa", &uport->mapbase);
+	else if (uart_iotype_legacy_io(uport->iotype))
+		seq_printf(m, " port:%08lX", uport->iobase);
+
+	seq_printf(m, " irq:%u", uport->irq);
 
 	if (uport->type == PORT_UNKNOWN) {
 		seq_putc(m, '\n');

-- 
2.47.3

[PATCH v2 05/15] serial: core: replace snprintf with more robust scnprintf

[Hugo Villeneuve]

From: Hugo Villeneuve <hvilleneuve@dimonoff.com>

Use scnprintf() so we could perhaps one day get rid of snprintf() entirely.

Suggested-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Signed-off-by: Hugo Villeneuve <hvilleneuve@dimonoff.com>
---
 drivers/tty/serial/serial_core.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/drivers/tty/serial/serial_core.c b/drivers/tty/serial/serial_core.c
index c5ef2dd00a9c9ad57c4010c05daea34fd2053676..3c909b8a82135a6245e98a9ae6bb050fc32590cc 100644
--- a/drivers/tty/serial/serial_core.c
+++ b/drivers/tty/serial/serial_core.c
@@ -2487,11 +2487,11 @@ uart_report_port(struct uart_driver *drv, struct uart_port *port)
 
 	switch (port->iotype) {
 	case UPIO_PORT:
-		snprintf(address, sizeof(address), "I/O 0x%lx", port->iobase);
+		scnprintf(address, sizeof(address), "I/O 0x%lx", port->iobase);
 		break;
 	case UPIO_HUB6:
-		snprintf(address, sizeof(address),
-			 "I/O 0x%lx offset 0x%x", port->iobase, port->hub6);
+		scnprintf(address, sizeof(address),
+			  "I/O 0x%lx offset 0x%x", port->iobase, port->hub6);
 		break;
 	case UPIO_MEM:
 	case UPIO_MEM16:
@@ -2499,8 +2499,8 @@ uart_report_port(struct uart_driver *drv, struct uart_port *port)
 	case UPIO_MEM32BE:
 	case UPIO_AU:
 	case UPIO_TSI:
-		snprintf(address, sizeof(address),
-			 "MMIO 0x%llx", (unsigned long long)port->mapbase);
+		scnprintf(address, sizeof(address),
+			  "MMIO 0x%llx", (unsigned long long)port->mapbase);
 		break;
 	default:
 		strscpy(address, "*unknown*", sizeof(address));

-- 
2.47.3

Re: [PATCH v2 05/15] serial: core: replace snprintf with more robust scnprintf

[Andy Shevchenko]

On Tue, Apr 28, 2026 at 01:53:51PM -0400, Hugo Villeneuve wrote:
> 
> Use scnprintf() so we could perhaps one day get rid of snprintf() entirely.

Ah, now I understand the approach in earlycon. Hmm... Still not sure which one
I prefer (it's not related to the contents of this patch anyway).

...

>  	default:
>  		strscpy(address, "*unknown*", sizeof(address));

Side note: This may use 2-argument strscpy().

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v2 05/15] serial: core: replace snprintf with more robust scnprintf

[Hugo Villeneuve]

On Thu, 30 Apr 2026 17:07:11 +0200
Andy Shevchenko <andriy.shevchenko@intel.com> wrote:

> On Tue, Apr 28, 2026 at 01:53:51PM -0400, Hugo Villeneuve wrote:
> > 
> > Use scnprintf() so we could perhaps one day get rid of snprintf() entirely.
> 
> Ah, now I understand the approach in earlycon. Hmm... Still not sure which one
> I prefer (it's not related to the contents of this patch anyway).
> 
> ...
> 
> >  	default:
> >  		strscpy(address, "*unknown*", sizeof(address));
> 
> Side note: This may use 2-argument strscpy().

This section is now removed in patch:
  serial: core: use uart_iotype_*() to simplify uart_report_port()

But interesting to know this 2-argument variant :)


-- 
Hugo Villeneuve

Re: [PATCH v2 05/15] serial: core: replace snprintf with more robust scnprintf

[Andy Shevchenko]

On Mon, May 04, 2026 at 11:27:12AM -0400, Hugo Villeneuve wrote:
> On Thu, 30 Apr 2026 17:07:11 +0200
> Andy Shevchenko <andriy.shevchenko@intel.com> wrote:
> > On Tue, Apr 28, 2026 at 01:53:51PM -0400, Hugo Villeneuve wrote:

...

> > >  	default:
> > >  		strscpy(address, "*unknown*", sizeof(address));
> > 
> > Side note: This may use 2-argument strscpy().
> 
> This section is now removed in patch:
>   serial: core: use uart_iotype_*() to simplify uart_report_port()
> 
> But interesting to know this 2-argument variant :)

Yep, we have a handful of functions (actually macros) that use some magic
to allow supply different number of arguments to the same macro and behave
accordingly.

-- 
With Best Regards,
Andy Shevchenko

[PATCH v2 06/15] serial: core: fix indentation/alignment

[Hugo Villeneuve]

From: Hugo Villeneuve <hvilleneuve@dimonoff.com>

Fixes the following checkpatch warnings:
    CHECK: Alignment should match open parenthesis

Signed-off-by: Hugo Villeneuve <hvilleneuve@dimonoff.com>
---
Without this, the next patch will have a checkpatch warning.
---
 drivers/tty/serial/serial_core.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/drivers/tty/serial/serial_core.c b/drivers/tty/serial/serial_core.c
index 3c909b8a82135a6245e98a9ae6bb050fc32590cc..e9986f9221f946c9075a8eea7ac28d2ad818f095 100644
--- a/drivers/tty/serial/serial_core.c
+++ b/drivers/tty/serial/serial_core.c
@@ -2508,10 +2508,10 @@ uart_report_port(struct uart_driver *drv, struct uart_port *port)
 	}
 
 	pr_info("%s%s%s at %s (irq = %u, base_baud = %u) is a %s\n",
-	       port->dev ? dev_name(port->dev) : "",
-	       port->dev ? ": " : "",
-	       port->name,
-	       address, port->irq, port->uartclk / 16, uart_type(port));
+		port->dev ? dev_name(port->dev) : "",
+		port->dev ? ": " : "",
+		port->name,
+		address, port->irq, port->uartclk / 16, uart_type(port));
 
 	/* The magic multiplier feature is a bit obscure, so report it too.  */
 	if (port->flags & UPF_MAGIC_MULTIPLIER)

-- 
2.47.3

[PATCH v2 07/15] serial: core: use uart_iotype_*() to simplify uart_report_port()

[Hugo Villeneuve]

From: Hugo Villeneuve <hvilleneuve@dimonoff.com>

Make use of new functions uart_iotype_mmio() and uart_iotype_legacy_io()
to simplify and improve code readability. It will also prevent displaying
irrelevant I/O information for future IO types (ex: UPIO_BUS). Use %pa to
display the mapbase pointer, as it is done in earlycon_print_info().

Signed-off-by: Hugo Villeneuve <hvilleneuve@dimonoff.com>
---
On 32-bit cpu:
  no changes.
On 64-bit cpu:
  before:
    1004b800.serial: ttySC0 at MMIO 0x1004b800 (irq = 35, base_baud = 0) is a scif
    1004bc00.serial: ttySC1 at MMIO 0x1004bc00 (irq = 40, base_baud = 0) is a scif
  after:
    1004b800.serial: ttySC0 at MMIO 0x000000001004b800 (irq = 35, base_baud = 0) is a scif
    1004bc00.serial: ttySC1 at MMIO 0x000000001004bc00 (irq = 40, base_baud = 0) is a scif
---
 drivers/tty/serial/serial_core.c | 32 ++++++++++----------------------
 1 file changed, 10 insertions(+), 22 deletions(-)

diff --git a/drivers/tty/serial/serial_core.c b/drivers/tty/serial/serial_core.c
index e9986f9221f946c9075a8eea7ac28d2ad818f095..d0d4dba41fb9ee43403391d9d599abc1d64b48ec 100644
--- a/drivers/tty/serial/serial_core.c
+++ b/drivers/tty/serial/serial_core.c
@@ -2483,31 +2483,19 @@ EXPORT_SYMBOL(uart_resume_port);
 static inline void
 uart_report_port(struct uart_driver *drv, struct uart_port *port)
 {
-	char address[64];
+	char address[64] = "";
 
-	switch (port->iotype) {
-	case UPIO_PORT:
-		scnprintf(address, sizeof(address), "I/O 0x%lx", port->iobase);
-		break;
-	case UPIO_HUB6:
-		scnprintf(address, sizeof(address),
-			  "I/O 0x%lx offset 0x%x", port->iobase, port->hub6);
-		break;
-	case UPIO_MEM:
-	case UPIO_MEM16:
-	case UPIO_MEM32:
-	case UPIO_MEM32BE:
-	case UPIO_AU:
-	case UPIO_TSI:
-		scnprintf(address, sizeof(address),
-			  "MMIO 0x%llx", (unsigned long long)port->mapbase);
-		break;
-	default:
-		strscpy(address, "*unknown*", sizeof(address));
-		break;
+	if (uart_iotype_mmio(port->iotype))
+		scnprintf(address, sizeof(address), " at MMIO %pa", &port->mapbase);
+	else if (uart_iotype_legacy_io(port->iotype)) {
+		if (port->iotype == UPIO_PORT)
+			scnprintf(address, sizeof(address), " at I/O 0x%lx", port->iobase);
+		else if (port->iotype == UPIO_HUB6)
+			scnprintf(address, sizeof(address), " at I/O 0x%lx offset 0x%x",
+				  port->iobase, port->hub6);
 	}
 
-	pr_info("%s%s%s at %s (irq = %u, base_baud = %u) is a %s\n",
+	pr_info("%s%s%s%s (irq = %u, base_baud = %u) is a %s\n",
 		port->dev ? dev_name(port->dev) : "",
 		port->dev ? ": " : "",
 		port->name,

-- 
2.47.3

[PATCH v2 08/15] serial: earlycon: use uart_iotype_*() to simplify code

[Hugo Villeneuve]

From: Hugo Villeneuve <hvilleneuve@dimonoff.com>

Make use of new functions uart_iotype_mmio() and uart_iotype_legacy_io()
to simplify and improve code readability.

Signed-off-by: Hugo Villeneuve <hvilleneuve@dimonoff.com>
---
Using uart_iotype_mmio() now includes the case for UPIO_TSI and UPIO_AU,
which fell before this patch in the else clause.
UPIO_TSI is no longer used by any driver, so not an issue.
UPIO_AU, if this is even possible for earlycon, falls into the "MMIO" category
in uart_line_info() and uart_report_port()...
---
 drivers/tty/serial/earlycon.c | 24 ++++++++++++------------
 1 file changed, 12 insertions(+), 12 deletions(-)

diff --git a/drivers/tty/serial/earlycon.c b/drivers/tty/serial/earlycon.c
index ab9af37f6cda35ea2e3ea966fdac8ba5c4475cb2..e799feaa14bf1ac7c1a5677cd84490e7c486449b 100644
--- a/drivers/tty/serial/earlycon.c
+++ b/drivers/tty/serial/earlycon.c
@@ -75,19 +75,19 @@ static void __init earlycon_print_info(struct earlycon_device *device)
 {
 	struct console *earlycon = device->con;
 	struct uart_port *port = &device->port;
+	char address[64] = "";
 
-	if (port->iotype == UPIO_MEM || port->iotype == UPIO_MEM16 ||
-	    port->iotype == UPIO_MEM32 || port->iotype == UPIO_MEM32BE)
-		pr_info("%s%d at MMIO%s %pa (options '%s')\n",
-			earlycon->name, earlycon->index,
-			(port->iotype == UPIO_MEM) ? "" :
-			(port->iotype == UPIO_MEM16) ? "16" :
-			(port->iotype == UPIO_MEM32) ? "32" : "32be",
-			&port->mapbase, device->options);
-	else
-		pr_info("%s%d at I/O port 0x%lx (options '%s')\n",
-			earlycon->name, earlycon->index,
-			port->iobase, device->options);
+	if (uart_iotype_mmio(port->iotype))
+		scnprintf(address, sizeof(address), " at MMIO%s %pa",
+			  (port->iotype == UPIO_MEM) ? "" :
+			  (port->iotype == UPIO_MEM16) ? "16" :
+			  (port->iotype == UPIO_MEM32) ? "32" : "32be",
+			  &port->mapbase);
+	else if (uart_iotype_legacy_io(port->iotype))
+		scnprintf(address, sizeof(address), " at I/O port 0x%lx", port->iobase);
+
+	pr_info("%s%d%s (options '%s')\n", earlycon->name, earlycon->index,
+		address, device->options);
 }
 
 static int __init parse_options(struct earlycon_device *device, char *options)

-- 
2.47.3

Re: [PATCH v2 08/15] serial: earlycon: use uart_iotype_*() to simplify code

[Andy Shevchenko]

On Tue, Apr 28, 2026 at 01:53:54PM -0400, Hugo Villeneuve wrote:

> Make use of new functions uart_iotype_mmio() and uart_iotype_legacy_io()
> to simplify and improve code readability.

...

> +	char address[64] = "";

TBH, I prefer two pr_info() calls as that approach
- doesn't require temporary buffer
- doesn't use heavy s*printf() on top of the existing printing
- doesn't limit the flexibility of each of the strings (64 might
  become not enough in some cases, however unlikely to happen)

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v2 08/15] serial: earlycon: use uart_iotype_*() to simplify code

[Hugo Villeneuve]

On Thu, 30 Apr 2026 17:03:35 +0200
Andy Shevchenko <andriy.shevchenko@intel.com> wrote:

> On Tue, Apr 28, 2026 at 01:53:54PM -0400, Hugo Villeneuve wrote:
> 
> > Make use of new functions uart_iotype_mmio() and uart_iotype_legacy_io()
> > to simplify and improve code readability.
> 
> ...
> 
> > +	char address[64] = "";
> 
> TBH, I prefer two pr_info() calls as that approach
> - doesn't require temporary buffer
> - doesn't use heavy s*printf() on top of the existing printing
> - doesn't limit the flexibility of each of the strings (64 might
>   become not enough in some cases, however unlikely to happen)

Note that this approach is already used in uart_report_port().

Also this patch may be reworked if patch "serial: uniformize serial
port I/O infos display" is deemed ok...

-- 
Hugo Villeneuve

Re: [PATCH v2 08/15] serial: earlycon: use uart_iotype_*() to simplify code

[Andy Shevchenko]

On Mon, May 04, 2026 at 11:40:01AM -0400, Hugo Villeneuve wrote:
> On Thu, 30 Apr 2026 17:03:35 +0200
> Andy Shevchenko <andriy.shevchenko@intel.com> wrote:
> > On Tue, Apr 28, 2026 at 01:53:54PM -0400, Hugo Villeneuve wrote:

...

> > > +	char address[64] = "";
> > 
> > TBH, I prefer two pr_info() calls as that approach
> > - doesn't require temporary buffer
> > - doesn't use heavy s*printf() on top of the existing printing
> > - doesn't limit the flexibility of each of the strings (64 might
> >   become not enough in some cases, however unlikely to happen)
> 
> Note that this approach is already used in uart_report_port().

Yeah, yeah... I noticed that after review.

> Also this patch may be reworked if patch "serial: uniformize serial
> port I/O infos display" is deemed ok...

-- 
With Best Regards,
Andy Shevchenko

[PATCH v2 09/15] serial: 8250: use uart_iotype_*() to simplify code

[Hugo Villeneuve]

From: Hugo Villeneuve <hvilleneuve@dimonoff.com>

Make use of new functions uart_iotype_mmio() and uart_iotype_legacy_io()
to simplify and improve code readability.

Signed-off-by: Hugo Villeneuve <hvilleneuve@dimonoff.com>
---
 drivers/tty/serial/8250/8250_port.c | 43 ++++++-------------------------------
 1 file changed, 7 insertions(+), 36 deletions(-)

diff --git a/drivers/tty/serial/8250/8250_port.c b/drivers/tty/serial/8250/8250_port.c
index af78cc02f38e719573becd0aea226f7790555a3e..20b7de120b1bab596de22d09fdeff8361afd8b17 100644
--- a/drivers/tty/serial/8250/8250_port.c
+++ b/drivers/tty/serial/8250/8250_port.c
@@ -472,16 +472,10 @@ static void set_io_from_upio(struct uart_port *p)
 static void
 serial_port_out_sync(struct uart_port *p, int offset, int value)
 {
-	switch (p->iotype) {
-	case UPIO_MEM:
-	case UPIO_MEM16:
-	case UPIO_MEM32:
-	case UPIO_MEM32BE:
-	case UPIO_AU:
+	if (uart_iotype_mmio(p->iotype)) {
 		p->serial_out(p, offset, value);
 		p->serial_in(p, UART_LCR);	/* safe, no side-effects */
-		break;
-	default:
+	} else {
 		p->serial_out(p, offset, value);
 	}
 }
@@ -2863,13 +2857,7 @@ static int serial8250_request_std_resource(struct uart_8250_port *up)
 	unsigned int size = serial8250_port_size(up);
 	struct uart_port *port = &up->port;
 
-	switch (port->iotype) {
-	case UPIO_AU:
-	case UPIO_TSI:
-	case UPIO_MEM32:
-	case UPIO_MEM32BE:
-	case UPIO_MEM16:
-	case UPIO_MEM:
+	if (uart_iotype_mmio(port->iotype)) {
 		if (!port->mapbase)
 			return -EINVAL;
 
@@ -2883,14 +2871,9 @@ static int serial8250_request_std_resource(struct uart_8250_port *up)
 				return -ENOMEM;
 			}
 		}
-		return 0;
-	case UPIO_HUB6:
-	case UPIO_PORT:
+	} else if (uart_iotype_legacy_io(port->iotype)) {
 		if (!request_region(port->iobase, size, "serial"))
 			return -EBUSY;
-		return 0;
-	case UPIO_UNKNOWN:
-		break;
 	}
 
 	return 0;
@@ -2901,15 +2884,9 @@ static void serial8250_release_std_resource(struct uart_8250_port *up)
 	unsigned int size = serial8250_port_size(up);
 	struct uart_port *port = &up->port;
 
-	switch (port->iotype) {
-	case UPIO_AU:
-	case UPIO_TSI:
-	case UPIO_MEM32:
-	case UPIO_MEM32BE:
-	case UPIO_MEM16:
-	case UPIO_MEM:
+	if (uart_iotype_mmio(port->iotype)) {
 		if (!port->mapbase)
-			break;
+			return;
 
 		if (port->flags & UPF_IOREMAP) {
 			iounmap(port->membase);
@@ -2917,14 +2894,8 @@ static void serial8250_release_std_resource(struct uart_8250_port *up)
 		}
 
 		release_mem_region(port->mapbase, size);
-		break;
-
-	case UPIO_HUB6:
-	case UPIO_PORT:
+	} else if (uart_iotype_legacy_io(port->iotype)) {
 		release_region(port->iobase, size);
-		break;
-	case UPIO_UNKNOWN:
-		break;
 	}
 }
 

-- 
2.47.3

[PATCH v2 10/15] serial: 8250_rsa: use uart_iotype_*() to simplify code

[Hugo Villeneuve]

From: Hugo Villeneuve <hvilleneuve@dimonoff.com>

Make use of new functions uart_iotype_mmio() and uart_iotype_legacy_io()
to simplify and improve code readability, as well as avoid some variables
init if the iotype is not valid.

Signed-off-by: Hugo Villeneuve <hvilleneuve@dimonoff.com>
---
 drivers/tty/serial/8250/8250_rsa.c | 40 ++++++++++++++++++--------------------
 1 file changed, 19 insertions(+), 21 deletions(-)

diff --git a/drivers/tty/serial/8250/8250_rsa.c b/drivers/tty/serial/8250/8250_rsa.c
index fff9395948e33be6b08355bace70275b7515b608..010ba5af81a08e737e0198d06d150bbe470e6688 100644
--- a/drivers/tty/serial/8250/8250_rsa.c
+++ b/drivers/tty/serial/8250/8250_rsa.c
@@ -19,35 +19,33 @@ static const struct uart_ops *core_port_base_ops;
 static int rsa8250_request_resource(struct uart_8250_port *up)
 {
 	struct uart_port *port = &up->port;
-	unsigned long start = UART_RSA_BASE << port->regshift;
-	unsigned int size = 8 << port->regshift;
+	unsigned long start;
+	unsigned int size;
 
-	switch (port->iotype) {
-	case UPIO_HUB6:
-	case UPIO_PORT:
-		start += port->iobase;
-		if (!request_region(start, size, "serial-rsa"))
-			return -EBUSY;
-		return 0;
-	default:
+	if (!uart_iotype_legacy_io(port->iotype))
 		return -EINVAL;
-	}
+
+	start = UART_RSA_BASE << port->regshift;
+	start += port->iobase;
+	size = 8 << port->regshift;
+
+	if (!request_region(start, size, "serial-rsa"))
+		return -EBUSY;
+	return 0;
 }
 
 static void rsa8250_release_resource(struct uart_8250_port *up)
 {
 	struct uart_port *port = &up->port;
-	unsigned long offset = UART_RSA_BASE << port->regshift;
-	unsigned int size = 8 << port->regshift;
+	unsigned long offset;
+	unsigned int size;
 
-	switch (port->iotype) {
-	case UPIO_HUB6:
-	case UPIO_PORT:
-		release_region(port->iobase + offset, size);
-		break;
-	default:
-		break;
-	}
+	if (!uart_iotype_legacy_io(port->iotype))
+		return;
+
+	offset = UART_RSA_BASE << port->regshift;
+	size = 8 << port->regshift;
+	release_region(port->iobase + offset, size);
 }
 
 static void univ8250_config_port(struct uart_port *port, int flags)

-- 
2.47.3

[PATCH v2 11/15] serial: core: add new I/O type for SPI and I2C bus devices

[Hugo Villeneuve]

From: Hugo Villeneuve <hvilleneuve@dimonoff.com>

I2C/SPI serial drivers don't use the following struct uart_port variables:
    port->membase
    port->mapbase
    port->iobase

However, they are forced to set membase to a non-zero value so that
uart_configure_port() will succeed because of the following check:

    /* If there isn't a port here, don't do anything further. */
    if (!port->iobase && !port->mapbase && !port->membase)
        return;

Add a new I/O type for SPI and I2C bus devices to remove the need to
implement the kind of above-mentioned ambiguous workarounds to make them
work. Now that UART report functions are using uart_iotype_*() functions,
no more irrelevant I/O information are being printed for UPIO_BUS iotypes.

Signed-off-by: Hugo Villeneuve <hvilleneuve@dimonoff.com>
---
With this change, uart_match_port() will always return true for UPIO_BUS
types, and this function is not used by any of the SPI/I2C drivers.
---
 drivers/tty/serial/serial_core.c | 11 ++++++-----
 include/linux/serial_core.h      |  1 +
 include/uapi/linux/serial.h      |  1 +
 3 files changed, 8 insertions(+), 5 deletions(-)

diff --git a/drivers/tty/serial/serial_core.c b/drivers/tty/serial/serial_core.c
index d0d4dba41fb9ee43403391d9d599abc1d64b48ec..0e015477848504e37afb34c8088957083f1662c7 100644
--- a/drivers/tty/serial/serial_core.c
+++ b/drivers/tty/serial/serial_core.c
@@ -2516,11 +2516,10 @@ uart_configure_port(struct uart_driver *drv, struct uart_state *state,
 {
 	unsigned int flags;
 
-	/*
-	 * If there isn't a port here, don't do anything further.
-	 */
-	if (!port->iobase && !port->mapbase && !port->membase)
-		return;
+	/* If there isn't a port here, don't do anything further. */
+	if (uart_iotype_mmio(port->iotype) || uart_iotype_legacy_io(port->iotype))
+		if (!port->iobase && !port->mapbase && !port->membase)
+			return;
 
 	/*
 	 * Now do the auto configuration stuff.  Note that config_port
@@ -3216,6 +3215,8 @@ bool uart_match_port(const struct uart_port *port1,
 		return hub6_match_port(port1, port2);
 	else if (uart_iotype_mmio(port1->iotype))
 		return port1->mapbase == port2->mapbase;
+	else if (port1->iotype == UPIO_BUS)
+		return true;
 	else
 		return false;
 }
diff --git a/include/linux/serial_core.h b/include/linux/serial_core.h
index fc3f5ebe389658c197ffc105ce4ac11cacef59bb..626dd939c53c9f920d71aadd360ec0ea0bacce0d 100644
--- a/include/linux/serial_core.h
+++ b/include/linux/serial_core.h
@@ -437,6 +437,7 @@ enum uart_iotype {
 	UPIO_TSI	= SERIAL_IO_TSI,	/* Tsi108/109 type IO */
 	UPIO_MEM32BE	= SERIAL_IO_MEM32BE,	/* 32b big endian */
 	UPIO_MEM16	= SERIAL_IO_MEM16,	/* 16b little endian */
+	UPIO_BUS	= SERIAL_IO_BUS,	/* Serial bus I/O access (ex: SPI, I2C) */
 };
 
 struct uart_port {
diff --git a/include/uapi/linux/serial.h b/include/uapi/linux/serial.h
index de9b4733607e6b61b08ff7089ff90070168ff4a2..e6f61538fc2837a264d27942afaaf3f12e743445 100644
--- a/include/uapi/linux/serial.h
+++ b/include/uapi/linux/serial.h
@@ -72,6 +72,7 @@ struct serial_struct {
 #define SERIAL_IO_TSI	  5
 #define SERIAL_IO_MEM32BE 6
 #define SERIAL_IO_MEM16	7
+#define SERIAL_IO_BUS	8
 
 #define UART_CLEAR_FIFO		0x01
 #define UART_USE_FIFO		0x02

-- 
2.47.3

[PATCH v2 12/15] serial: sc16is7xx: use new UPIO_BUS as iotype

[Hugo Villeneuve]

From: Hugo Villeneuve <hvilleneuve@dimonoff.com>

Now that we have a new UPIO_BUS I/O type, use it to register our serial
port and remove ambiguous membase/iobase workaround.

Note that commit 5da6b1c079e6 ("sc16is7xx: Set iobase to device index")
used the iobase field as an index within the device to allow infering
the order through sysfs, but this is no longer needed since
commit 1ef2c2df1199 ("serial: core: Fix serial core controller port name
to show controller id").

Signed-off-by: Hugo Villeneuve <hvilleneuve@dimonoff.com>
---
This means that displaying iomem_base will now always be zero:

cat /sys/class/tty/ttySC0/iomem_base
0x0
cat /sys/class/tty/ttySC1/iomem_base
0x0
...

But the index can be properly displayed with this instead (example with two
sc16is7xx devices):
$> ls -al /sys/class/tty/ttySC*/device
... /sys/class/tty/ttySC0/device -> ../../../spi1.0:0.0
... /sys/class/tty/ttySC1/device -> ../../../spi1.0:0.1
... /sys/class/tty/ttySC2/device -> ../../../spi3.0:0.0
... /sys/class/tty/ttySC3/device -> ../../../spi3.0:0.1
---
 drivers/tty/serial/sc16is7xx.c | 9 +--------
 1 file changed, 1 insertion(+), 8 deletions(-)

diff --git a/drivers/tty/serial/sc16is7xx.c b/drivers/tty/serial/sc16is7xx.c
index 1fd64a47341d8263c82fcadd90a4c2e549193e19..4b638e69f36f2d20253d8de31f52ecfe349aa39f 100644
--- a/drivers/tty/serial/sc16is7xx.c
+++ b/drivers/tty/serial/sc16is7xx.c
@@ -1472,14 +1472,7 @@ static int sc16is7xx_setup_channel(struct sc16is7xx_one *one, int i,
 	port->type	= PORT_SC16IS7XX;
 	port->fifosize	= SC16IS7XX_FIFO_SIZE;
 	port->flags	= UPF_FIXED_TYPE | UPF_LOW_LATENCY;
-	port->iobase	= i;
-	/*
-	 * Use all ones as membase to make sure uart_configure_port() in
-	 * serial_core.c does not abort for SPI/I2C devices where the
-	 * membase address is not applicable.
-	 */
-	port->membase	= (void __iomem *)~0;
-	port->iotype	= UPIO_PORT;
+	port->iotype	= UPIO_BUS;
 	port->rs485_config = sc16is7xx_config_rs485;
 	port->rs485_supported = sc16is7xx_rs485_supported;
 	port->ops	= &sc16is7xx_ops;

-- 
2.47.3

[PATCH v2 13/15] serial: max310x: use new UPIO_BUS as iotype

[Hugo Villeneuve]

From: Hugo Villeneuve <hvilleneuve@dimonoff.com>

Now that we have a new UPIO_BUS I/O type, use it to register our serial
port and remove obscure membase/iobase workaround.

Signed-off-by: Hugo Villeneuve <hvilleneuve@dimonoff.com>
---
 drivers/tty/serial/max310x.c | 9 +--------
 1 file changed, 1 insertion(+), 8 deletions(-)

diff --git a/drivers/tty/serial/max310x.c b/drivers/tty/serial/max310x.c
index ac7d3f197c3a5ce3531d5607f48e21a807314021..cd0496e307091dae1f53911abf86726f52d5770d 100644
--- a/drivers/tty/serial/max310x.c
+++ b/drivers/tty/serial/max310x.c
@@ -1380,14 +1380,7 @@ static int max310x_probe(struct device *dev, const struct max310x_devtype *devty
 		s->p[i].port.type	= PORT_MAX310X;
 		s->p[i].port.fifosize	= MAX310X_FIFO_SIZE;
 		s->p[i].port.flags	= UPF_FIXED_TYPE | UPF_LOW_LATENCY;
-		s->p[i].port.iotype	= UPIO_PORT;
-		s->p[i].port.iobase	= i;
-		/*
-		 * Use all ones as membase to make sure uart_configure_port() in
-		 * serial_core.c does not abort for SPI/I2C devices where the
-		 * membase address is not applicable.
-		 */
-		s->p[i].port.membase	= (void __iomem *)~0;
+		s->p[i].port.iotype	= UPIO_BUS;
 		s->p[i].port.uartclk	= uartclk;
 		s->p[i].port.rs485_config = max310x_rs485_config;
 		s->p[i].port.rs485_supported = max310x_rs485_supported;

-- 
2.47.3

[PATCH v2 14/15] serial: max3100: use new UPIO_BUS as iotype

[Hugo Villeneuve]

From: Hugo Villeneuve <hvilleneuve@dimonoff.com>

Now that we have a new UPIO_BUS I/O type, use it to register our serial
port. This allows the driver to work properly when using DT where
membase/iobase are not set.

Signed-off-by: Hugo Villeneuve <hvilleneuve@dimonoff.com>
---
 drivers/tty/serial/max3100.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/tty/serial/max3100.c b/drivers/tty/serial/max3100.c
index 475b0a6efce4bec36aa4b94ccff83934dbdf9246..17a2ff410305142798ddbbb22b46f18eadce91aa 100644
--- a/drivers/tty/serial/max3100.c
+++ b/drivers/tty/serial/max3100.c
@@ -725,6 +725,7 @@ static int max3100_probe(struct spi_device *spi)
 	max3100s[i]->port.ops = &max3100_ops;
 	max3100s[i]->port.flags = UPF_SKIP_TEST | UPF_BOOT_AUTOCONF;
 	max3100s[i]->port.line = i;
+	max3100s[i]->port.iotype = UPIO_BUS;
 	max3100s[i]->port.type = PORT_MAX3100;
 	max3100s[i]->port.dev = &spi->dev;
 

-- 
2.47.3

[PATCH v2 15/15] serial: uniformize serial port I/O infos display

[Hugo Villeneuve]

From: Hugo Villeneuve <hvilleneuve@dimonoff.com>

Uniformize serial port I/O infos display from three different functions
that display mostly the same information, but with some variations, by
adding a common function.

Signed-off-by: Hugo Villeneuve <hvilleneuve@dimonoff.com>
---
Maybe the width information could be dropped entirely?
---
 drivers/tty/serial/earlycon.c    | 13 +++-------
 drivers/tty/serial/serial_core.c | 52 +++++++++++++++++++++++++++-------------
 include/linux/serial_core.h      |  1 +
 3 files changed, 40 insertions(+), 26 deletions(-)

diff --git a/drivers/tty/serial/earlycon.c b/drivers/tty/serial/earlycon.c
index e799feaa14bf1ac7c1a5677cd84490e7c486449b..ce740cdc7cebffca0b4f6be98b8bca66540af72d 100644
--- a/drivers/tty/serial/earlycon.c
+++ b/drivers/tty/serial/earlycon.c
@@ -75,19 +75,12 @@ static void __init earlycon_print_info(struct earlycon_device *device)
 {
 	struct console *earlycon = device->con;
 	struct uart_port *port = &device->port;
-	char address[64] = "";
+	char ioinfos[64];
 
-	if (uart_iotype_mmio(port->iotype))
-		scnprintf(address, sizeof(address), " at MMIO%s %pa",
-			  (port->iotype == UPIO_MEM) ? "" :
-			  (port->iotype == UPIO_MEM16) ? "16" :
-			  (port->iotype == UPIO_MEM32) ? "32" : "32be",
-			  &port->mapbase);
-	else if (uart_iotype_legacy_io(port->iotype))
-		scnprintf(address, sizeof(address), " at I/O port 0x%lx", port->iobase);
+	uart_get_ioinfos(port, ioinfos, sizeof(ioinfos));
 
 	pr_info("%s%d%s (options '%s')\n", earlycon->name, earlycon->index,
-		address, device->options);
+		ioinfos, device->options);
 }
 
 static int __init parse_options(struct earlycon_device *device, char *options)
diff --git a/drivers/tty/serial/serial_core.c b/drivers/tty/serial/serial_core.c
index 0e015477848504e37afb34c8088957083f1662c7..b9a9225d99b616a7f691a8f990fd77ab02609012 100644
--- a/drivers/tty/serial/serial_core.c
+++ b/drivers/tty/serial/serial_core.c
@@ -2002,6 +2002,7 @@ static void uart_line_info(struct seq_file *m, struct uart_state *state)
 	struct tty_port *port = &state->port;
 	enum uart_pm_state pm_state;
 	struct uart_port *uport;
+	char ioinfos[64];
 	char stat_buf[32];
 	unsigned int status;
 
@@ -2013,10 +2014,8 @@ static void uart_line_info(struct seq_file *m, struct uart_state *state)
 
 	seq_printf(m, "%u: uart:%s", uport->line, uart_type(uport));
 
-	if (uart_iotype_mmio(uport->iotype))
-		seq_printf(m, " mmio:%pa", &uport->mapbase);
-	else if (uart_iotype_legacy_io(uport->iotype))
-		seq_printf(m, " port:%08lX", uport->iobase);
+	uart_get_ioinfos(uport, ioinfos, sizeof(ioinfos));
+	seq_printf(m, "%s", ioinfos);
 
 	seq_printf(m, " irq:%u", uport->irq);
 
@@ -2480,26 +2479,47 @@ int uart_resume_port(struct uart_driver *drv, struct uart_port *uport)
 }
 EXPORT_SYMBOL(uart_resume_port);
 
+static const char *uart_get_mmio_width(struct uart_port *port)
+{
+	switch (port->iotype) {
+	case UPIO_MEM16:
+		return "16";
+	case UPIO_MEM32:
+	case UPIO_MEM32BE:
+		return "32be";
+	case UPIO_AU:
+	case UPIO_MEM:
+	default:
+		return "";
+	}
+}
+
+void uart_get_ioinfos(struct uart_port *port, char *buf, size_t size)
+{
+	buf[0] = '\0';
+
+	if (uart_iotype_mmio(port->iotype)) {
+		scnprintf(buf, size, " MMIO%s:%pa", uart_get_mmio_width(port), &port->mapbase);
+	} else if (uart_iotype_legacy_io(port->iotype)) {
+		if (port->iotype == UPIO_PORT)
+			scnprintf(buf, size, " I/O:0x%lx", port->iobase);
+		else if (port->iotype == UPIO_HUB6)
+			scnprintf(buf, size, " I/O:0x%lx, offset 0x%x", port->iobase, port->hub6);
+	}
+}
+EXPORT_SYMBOL(uart_get_ioinfos);
+
 static inline void
 uart_report_port(struct uart_driver *drv, struct uart_port *port)
 {
-	char address[64] = "";
+	char ioinfos[64];
 
-	if (uart_iotype_mmio(port->iotype))
-		scnprintf(address, sizeof(address), " at MMIO %pa", &port->mapbase);
-	else if (uart_iotype_legacy_io(port->iotype)) {
-		if (port->iotype == UPIO_PORT)
-			scnprintf(address, sizeof(address), " at I/O 0x%lx", port->iobase);
-		else if (port->iotype == UPIO_HUB6)
-			scnprintf(address, sizeof(address), " at I/O 0x%lx offset 0x%x",
-				  port->iobase, port->hub6);
-	}
+	uart_get_ioinfos(port, ioinfos, sizeof(ioinfos));
 
 	pr_info("%s%s%s%s (irq = %u, base_baud = %u) is a %s\n",
 		port->dev ? dev_name(port->dev) : "",
 		port->dev ? ": " : "",
-		port->name,
-		address, port->irq, port->uartclk / 16, uart_type(port));
+		port->name, ioinfos, port->irq, port->uartclk / 16, uart_type(port));
 
 	/* The magic multiplier feature is a bit obscure, so report it too.  */
 	if (port->flags & UPF_MAGIC_MULTIPLIER)
diff --git a/include/linux/serial_core.h b/include/linux/serial_core.h
index 626dd939c53c9f920d71aadd360ec0ea0bacce0d..1289e812b40f3d7b7a02ff1a87eaf649e71e0d2f 100644
--- a/include/linux/serial_core.h
+++ b/include/linux/serial_core.h
@@ -1308,6 +1308,7 @@ static inline int uart_handle_break(struct uart_port *port)
 
 int uart_get_rs485_mode(struct uart_port *port);
 
+void uart_get_ioinfos(struct uart_port *port, char *buf, size_t size);
 bool uart_iotype_mmio(enum uart_iotype iotype);
 bool uart_iotype_legacy_io(enum uart_iotype iotype);
 

-- 
2.47.3