[PATCH 0/4] ARM: tegra: Automatic UART selection for earlyprintk

[Doug Anderson]

(repost, since gmail seems to have mangled my message)

For the series,

Tested-by: Doug Anderson <dianders at chromium.org>

Thank you for picking this up.

A few overall comments on the patch set:
1. This patch has the side effect of requiring the use of zImage to
   get the proper behavior.  If we ever try to move decompression
   of the kernel to the bootloader (as has been talked about, since
   u-boot supports it) we'll lose earlyprintk behavior.  ...of course,
   this does match how some other ARM boards do things.  I'm OK with
   this.
2. I'm slightly concerned about cramming stuff into IRAM, since
   I don't have a good concept of what IRAM adresses are used when.
   When I grep for IRAM users in our kernel, I see
   TEGRA_AVP_RESUME_ADDR is using the base of IRAM.  We're not going
   to be hosing suspend/resume or anything else with this CL, are we?
   In the very least, I'd perfer to see a #define for the '0x100' and
   the '0x55415254' so that we can correlate the constants in
   uncompress.h and debug-macro.S

Tested:
Tested the series on a Tegra2-based board (Kaen) with
CONFIG_TEGRA_DEBUG_UART_NONE defined on the current ToT chromium-os
tree with the following pre-changes:
1. Reverted hack from our tree: 63b76a4638f3934e69c8724881060f121845184c
2. Added "arm/tegra: Support Tegra30 in decompressor UART setup" as
   described in the patch header.
3. Adjusted patch 4 to apply cleanly to our older "common.c"
4. Added missing #include <linux/kernel.h> to uncompress.h, as
   indicated by swarren on tegra IRC.

Also tried stopping in the bootloader and clearing the 'D' with:
  mw.b 0x7000605c 0
...and validated that I missed the earlyprintks (as expected) but
still booted.

---

On Wed, Jan 4, 2012 at 3:45 PM, Doug Anderson <dianders at chromium.org> wrote:
> For the series,
> Tested-by: Doug Anderson <dianders at chromium.org>
> Thank you for picking this up.
> A few overall comments on the patch set:1. This patch has the side
> effect of requiring the use of zImage to    get the proper behavior.
> If we ever try to move decompression   of the kernel to the bootloader
> (as has been talked about, since   u-boot supports it) we'll lose
> earlyprintk behavior.  ...of course,    this does match how some other
> ARM boards do things.  I'm OK with   this.2. I'm slightly concerned
> about cramming stuff into IRAM, since   I don't have a good concept of
> what IRAM adresses are used when.   When I grep for IRAM users in our
> kernel, I see    TEGRA_AVP_RESUME_ADDR is using the base of IRAM.
> We're not going   to be hosing suspend/resume or anything else with
> this CL, are we?   In the very least, I'd perfer to see a #define for
> the '0x100' and    the '0x55415254' so that we can correlate the
> constants in    uncompress.h and debug-macro.S
> Tested:Tested the series on a Tegra2-based board (Kaen)
> withCONFIG_TEGRA_DEBUG_UART_NONE defined on the current ToT
> chromium-os tree with the following pre-changes:1. Reverted hack from
> our tree: 63b76a4638f3934e69c8724881060f121845184c2. Added "arm/tegra:
> Support Tegra30 in decompressor UART setup" as   described in the
> patch header.3. Adjusted patch 4 to apply cleanly to our older
> "common.c"4. Added missing #include <linux/kernel.h> to uncompress.h,
> as    indicated by swarren on tegra IRC.
> Also tried stopping in the bootloader and clearing the 'D' with:  mw.b
> 0x7000605c 0...and validated that I missed the earlyprintks (as
> expected) butstill booted.
> ---
> On Tue, Jan 3, 2012 at 2:25 PM, Stephen Warren <swarren at nvidia.com> wrote:
>> This series allows a Tegra zImage to automatically select the UART to use
>> for the zImage decompressor and early printk, based on information from
>> the bootloader. The code scans each UART, looking for one that is clocked,
>> is out of reset, and that the bootloader has written 'D' (for 'D'ebug) to
>> the scratch register.
>>
>> At present, the ChromiumOS branch of U-Boot sets up UARTs in this way. If
>> this patch is accepted into the kernel, I'll upstream the U-Boot change to
>> mainline U-Boot.
>>
>> This is important, since different boards use different UARTs, and it's
>> nice not to have to configure the kernel differently when switching
>> boards.
>>
>> This patch series depends on patch "[PATCH V2] arm/tegra: Support Tegra30
>> in decompressor UART setup" for context.
>>
>> Doug Anderson (2):
>>  ARM: tegra: Introduce define DEBUG_UART_SHIFT
>>  ARM: tegra: uncompress.h: Store UART address in a variable
>>
>> Stephen Warren (2):
>>  ARM: tegra: uncompress.h: Choose a UART at runtime
>>  ARM: tegra: Pass uncompress.h UART selection to DEBUG_LL
>>
>>  arch/arm/mach-tegra/common.c                   |   17 ++++
>>  arch/arm/mach-tegra/include/mach/debug-macro.S |   87 +++++++++++++++++---
>>  arch/arm/mach-tegra/include/mach/uncompress.h  |  108 +++++++++++++++++++++---
>>  3 files changed, 190 insertions(+), 22 deletions(-)
>>