ARM: ux500: define serial port aliases

This enumerates the PL011 serial ports on the Ux500. This is necessary to do if we want to remove one of the serial ports, since userspace depends on console to be present on ttyAMA2 and we must not break userspace.

Cc Ulf Hansson <[email protected]>

Signed-off-by Linus Walleij <[email protected]>

---
ChangeLog v1->v2:
- Moved the alias definitions from the core SoC file to the toplevel DTS file, as requested by Arnd.

Quoted 5 lines · On Friday 10 July 2015 11:32:15 Linus Walleij wrote:

This enumerates the PL011 serial ports on the Ux500. This is necessary to do if we want to remove one of the serial ports, since userspace depends on console to be present on ttyAMA2 and we must not break userspace.

Quoted 8 lines

Cc Ulf Hansson <[email protected]>

Signed-off-by Linus Walleij <[email protected]>

---
ChangeLog v1->v2:
- Moved the alias definitions from the core SoC file to the toplevel DTS file, as requested by Arnd.

Do all of the boards expose all three uarts in the same order?

Quoted 15 lines · Arnd Bergmann

Quoted 5 lines · On Friday 10 July 2015 11:32:15 Linus Walleij wrote:

This enumerates the PL011 serial ports on the Ux500. This is necessary to do if we want to remove one of the serial ports, since userspace depends on console to be present on ttyAMA2 and we must not break userspace.

Quoted 8 lines

Cc Ulf Hansson <[email protected]>

Signed-off-by Linus Walleij <[email protected]>

---
ChangeLog v1->v2:
- Moved the alias definitions from the core SoC file to the toplevel DTS file, as requested by Arnd.

Quoted 1 line

Do all of the boards expose all three uarts in the same order?

Yes and that is the reason the patch series needs this one to come first.

What happened historically was that userspace was hard-coded to use ttyAMA2 as console. So for this reason, all boards had to define ttyAMA0, 1, 2 in their board file (later device tree) so they would be enumerated in the right order, relying on the DT to be parsed top to bottom of course.

So when I realized (in patch 2/2) that ttyAMA1 was not really used on some boards, it was impossible to remove it as userspace relied on the enumeration.

However I have to remove ttyAMA1 because it is fundamentally wrong (the RC pin is used for MMC!). Incidentally it worked to steal

But that breaks userspace since

Since the original code just defined all three serial ports for all boards by them including ste-dbx500.dtsi I first made the patch to that file, and then as you said it had to be done on a per-toplevel file, I did it in the individual DTS files instead.

Quoted 5 lines · On Monday 13 July 2015 10:21:51 Linus Walleij wrote:

Since the original code just defined all three serial ports for all boards by them including ste-dbx500.dtsi I first made the patch to that file, and then as you said it had to be done on a per-toplevel file, I did it in the individual DTS files instead.

I'm not following here. If you have boards that want port 1 to be disabled, shouldn't you remove that alias?

Quoted 9 lines · Arnd Bergmann

Quoted 5 lines · On Monday 13 July 2015 10:21:51 Linus Walleij wrote:

Since the original code just defined all three serial ports for all boards by them including ste-dbx500.dtsi I first made the patch to that file, and then as you said it had to be done on a per-toplevel file, I did it in the individual DTS files instead.

I'm not following here. If you have boards that want port 1 to be disabled, shouldn't you remove that alias?

No, that is done by not changing status = "disabled"; to status = "okay"; on a per-board basis.

This is what the second patch is doing.

The aliases are set also to disabled ports to ensure enumeration for the cases where a board file enabled a previously disabled UART.

Quoted 4 lines · On Thursday 16 July 2015 16:03:18 Linus Walleij wrote:

The aliases are set also to disabled ports to ensure enumeration for the cases where a board file enabled a previously disabled UART.

Well, both really. The status property tells the kernel to not ever touch a device, while the alias is what traditionally gets used to assign an OS visible device name to a device in the tree. This is more true for MacOS and AIX than Linux, which sometimes does its own naming, but we really should not have an alias pointing at a disabled device node.

Quoted 7 lines · Arnd Bergmann

Quoted 4 lines · On Thursday 16 July 2015 16:03:18 Linus Walleij wrote:

The aliases are set also to disabled ports to ensure enumeration for the cases where a board file enabled a previously disabled UART.

Quoted 6 lines

Well, both really. The status property tells the kernel to not ever touch a device, while the alias is what traditionally gets used to assign an OS visible device name to a device in the tree. This is more true for MacOS and AIX than Linux, which sometimes does its own naming, but we really should not have an alias pointing at a disabled device node.

OK ... I will make a patch that remove the aliases for the disabled nodes.

But this makes me think of whether the DT compiler should allow it or at least warn about it?

This enumerates the PL011 serial ports on the Ux500. This is necessary to do if we want to remove one of the serial ports, since userspace depends on console to be present on ttyAMA2 and we must not break userspace.

Cc Ulf Hansson <[email protected]>

Signed-off-by Linus Walleij <[email protected]>

---
ChangeLog v1->v2:
- Moved the alias definitions from the core SoC file to the toplevel DTS file, as requested by Arnd.

ARM SoC guys: if you find this patch OK, please apply this directly for fixes, as a prerequisite for patch 2/2, then apply 2/2.