Skip to contentSkip to conversation

msm: add minimal board file for HTC Dream device

PATCH In review Pavel Machek 4 participants 6 messages Nov 2, 2009 – Nov 16, 2009

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 2/3 msm: add minimal board file for HTC Dream device

This is just enough to get the device booting and serial console working. Sufficient for debugging further MSM7k/Dream Support. This will support HTC Dream / T-Mobile G1 / Android ADP1 (which are all the same hardware, known as "trout" to the ARM machine database).

---
arch/arm/mach-msm/Kconfig | 6 +++

 arch/arm/mach-msm/Makefile      |    1 +
 arch/arm/mach-msm/board-dream.c |   84 +++++++++++++++++++++++++++++++++++++++
 arch/arm/mach-msm/board-dream.h |    5 ++
 4 files changed, 96 insertions(+), 0 deletions(-)
 create mode 100644 arch/arm/mach-msm/board-dream.c
 create mode 100644 arch/arm/mach-msm/board-dream.h
arch/arm/mach-msm/Kconfig +6 -0
index 35f2a90..f780086 100644
@@ -34,4 +34,10 @@ config MACH_HALIBUT
 	help
 	  Support for the Qualcomm SURF7201A eval board.
 
+config MACH_TROUT
+	default y
+	bool "HTC Dream (aka trout)"
+	help
+	  Support for the HTC Dream, T-Mobile G1, Android ADP1 devices.
+
 endif
arch/arm/mach-msm/Makefile +1 -0
index 1aa4700..91e6f5c 100644
@@ -6,3 +6,4 @@ obj-y += clock.o clock-7x01a.o
 
 obj-$(CONFIG_MACH_HALIBUT) += board-halibut.o
 
+obj-$(CONFIG_MACH_TROUT) += board-dream.o
arch/arm/mach-msm/board-dream.c new file +84 -0

84 lines not on this page · raw message

arch/arm/mach-msm/board-dream.h new file +5 -0
new file mode 100644
index 0000000..4f345a5
@@ -0,0 +1,5 @@
+
+#define TROUT_CPLD_BASE   0xE8100000
+#define TROUT_CPLD_START  0x98000000
+#define TROUT_CPLD_SIZE   SZ_4K
+
Signature · 2 lines
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
#1 lore raw

arch/arm/mach-msm/board-dream.c line 78

Quoted 5 lines · Pavel Machek
arch/arm/mach-msm/board-dream.c new file +5 -0
+MACHINE_START(TROUT, "HTC Dream")
+#ifdef CONFIG_MSM_DEBUG_UART
+	.phys_io	= MSM_DEBUG_UART_PHYS,
+	.io_pg_offst	= ((MSM_DEBUG_UART_BASE) >> 18) & 0xfffc,
+#endif

Why is this ifdef'd? Two reasons why not to do this:

1. MSM_DEBUG_UART is always defined, so the conditional is pointless

2. the low level page table building code does not expect these to be unset (or even set to an invalid value). Leaving them unset and having DEBUG_LL enabled is a recipe to prevent booting.

To put it another way, these two fields aren't optional.

#2 lore raw
Quoted 10 lines · Russell King - ARM Linux

arch/arm/mach-msm/board-dream.c line 78

Quoted 5 lines · Pavel Machek
arch/arm/mach-msm/board-dream.c new file +5 -0
+MACHINE_START(TROUT, "HTC Dream")
+#ifdef CONFIG_MSM_DEBUG_UART
+ ? ? .phys_io ? ? ? ?= MSM_DEBUG_UART_PHYS,
+ ? ? .io_pg_offst ? ?= ((MSM_DEBUG_UART_BASE) >> 18) & 0xfffc,
+#endif

Why is this ifdef'd? ?Two reasons why not to do this:

1. MSM_DEBUG_UART is always defined, so the conditional is pointless

For a default dream build, definitely, but not always on every msm7k target (some devices may not have a serial port that can safely be chattered at by the debug code). For this board file, no harm to require it.

Quoted 3 lines

2. the low level page table building code does not expect these to be ? unset (or even set to an invalid value). ?Leaving them unset and ? having DEBUG_LL enabled is a recipe to prevent booting.

What's the best way to handle a situation where there is no valid debug uart? Could the arch/platform require DEBUG_LL be unset via Kconfig directives if it is configured in a way where there is no valid debug uart?

Quoted 2 lines

To put it another way, these two fields aren't optional.

Thanks,

Brian

#3 lore raw
Quoted 4 lines · Brian Swetland

What's the best way to handle a situation where there is no valid debug uart? Could the arch/platform require DEBUG_LL be unset via Kconfig directives if it is configured in a way where there is no valid debug uart?

You're the first to have that issue. Everyone else has a UART they can always direct stuff at.

However, I believe you're making the issue far larger than it really is.

1. If you define the DEBUG_LL macros to be empty, printascii() etc will not touch any mapping.

2. If no mapping is going to be touched by printascii(), does it matter whether a UART is mapped via this early mapping stuff?

The answer to (2) is no.

So, you can still arrange for these fields to be populated with a valid value even if you don't intend to use the resulting mapping. And so there's no need to force DEBUG_LL to be unset.

If you really have no values you can use, make sure you set io_pg_offst to 0x3ffc - the last offset in the L1 page table. This will cause the code to write a single dummy entry at the very top of the page table, which will then be overwritten by the generic ARM arch code for its own use.

#4 lore raw
Quoted 10 lines · Russell King - ARM Linux
Quoted 4 lines · Brian Swetland

What's the best way to handle a situation where there is no valid debug uart?  Could the arch/platform require DEBUG_LL be unset via Kconfig directives if it is configured in a way where there is no valid debug uart?

You're the first to have that issue.  Everyone else has a UART they can always direct stuff at.

However, I believe you're making the issue far larger than it really is.

I actually haven't run into it on the hardware I've worked on (often the debug uart isn't exposed anywhere useful, but it's usually there), but have seen designs where it would happen (all uarts assigned to some non-debug function). Just curious about how to deal with such a situation correctly.

Quoted 12 lines

1. If you define the DEBUG_LL macros to be empty, printascii() etc will not touch any mapping.

2. If no mapping is going to be touched by printascii(), does it matter whether a UART is mapped via this early mapping stuff?

The answer to (2) is no.

So, you can still arrange for these fields to be populated with a valid value even if you don't intend to use the resulting mapping.  And so there's no need to force DEBUG_LL to be unset.

Quoted 5 lines

If you really have no values you can use, make sure you set io_pg_offst to 0x3ffc - the last offset in the L1 page table.  This will cause the code to write a single dummy entry at the very top of the page table, which will then be overwritten by the generic ARM arch code for its own use.

Pointing it at UART1 with the DEBUG_LL macros empty in the event of no debug uart available works great.

Thanks,

Brian

#5 lore raw
In reply to Russell King - ARM Linux
Quoted 8 lines · Russell King - ARM Linux
Quoted 4 lines · Brian Swetland

What's the best way to handle a situation where there is no valid debug uart? Could the arch/platform require DEBUG_LL be unset via Kconfig directives if it is configured in a way where there is no valid debug uart?

You're the first to have that issue. Everyone else has a UART they can always direct stuff at.

Just for the record, there are machine files without a debug UART, especially in the smartphone area.

In the PXA subworld I know, the UARTs are often used for :
- a GSM chip (pxa ffuart)
- a Bluetooth chip (pxa btuart)
- a GPS chip (pxa stuart / hwuart)
- an IR chip (pxa stuart / hwuart)

I didn't check through the mach-pxa/*.c, but mioa701.c accounts for one, and I think magician and hx4700 would be part of it too. That also means kernel debugging is _really_ painfull for these platforms.

Cheers.

--
Robert