qi.git
2 years agoFix initramfs loading bug introduced in f407b0626f0160fa8cca67c548c8de38c8353e16 master
Timo Juhani Lindfors [Fri, 11 Nov 2011 21:05:36 +0000 (23:05 +0200)]
Fix initramfs loading bug introduced in f407b0626f0160fa8cca67c548c8de38c8353e16

initramfs_len is unsigned so we can not assign return value of
read_file to it without checking for negative values that indicate
errors.

Signed-off-by: Timo Juhani Lindfors <timo.lindfors@iki.fi>
Signed-off-by: Paul Fertser <fercerpav@gmail.com>

2 years agogta02|gta01: change kernel parameters to boot from ubifs ubi0:rootfs
Martin Jansa [Tue, 5 Jan 2010 11:44:25 +0000 (12:44 +0100)]
gta02|gta01: change kernel parameters to boot from ubifs ubi0:rootfs

This assumes ubifs is always used on NAND instead of jffs2.

Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Signed-off-by: Paul Fertser <fercerpav@gmail.com>

2 years agogta02: add initramfs path to the platform (initrd-GTA02.bin)
Timo Juhani Lindfors [Fri, 4 Nov 2011 14:22:45 +0000 (16:22 +0200)]
gta02: add initramfs path to the platform (initrd-GTA02.bin)

Tested with 2.3 MB initrd created using debian mkinitramfs.

Signed-off-by: Timo Juhani Lindfors <timo.lindfors@iki.fi>
Signed-off-by: Paul Fertser <fercerpav@gmail.com>

2 years agoDo not abort the boot when initramfs path is specified but file is not there
Paul Fertser [Fri, 11 Nov 2011 19:22:14 +0000 (23:22 +0400)]
Do not abort the boot when initramfs path is specified but file is not there

Make read_file() return values more meaningful to distinguish between
file is not present and other error conditions. This allows us to
specify a path for the initramfs in board config file and to try
loading it only when it is actually present, continuing normally if it
is not.

Signed-off-by: Paul Fertser <fercerpav@gmail.com>

2 years agogta02: Try both uImage-GTA02.bin and uImage-GTA02.bin.2 on first partition
Timo Juhani Lindfors [Fri, 4 Nov 2011 14:22:43 +0000 (16:22 +0200)]
gta02: Try both uImage-GTA02.bin and uImage-GTA02.bin.2 on first partition

This patch makes it possible to choose between two different kernels
without having to have two SD card partitions.

Signed-off-by: Timo Juhani Lindfors <timo.lindfors@iki.fi>
Signed-off-by: Paul Fertser <fercerpav@gmail.com>

2 years agoPrint all compiler and linker invocations
Timo Juhani Lindfors [Fri, 4 Nov 2011 14:22:42 +0000 (16:22 +0200)]
Print all compiler and linker invocations

This is mostly needed since Debian policy is probably in the future
going to require build logs that show what flags have been passed to
compiler and linker. If needed, we can add support for V=1 flag in the
future.

Signed-off-by: Timo Juhani Lindfors <timo.lindfors@iki.fi>
Signed-off-by: Paul Fertser <fercerpav@gmail.com>

2 years agoRemove build dependency on git
Timo Juhani Lindfors [Fri, 4 Nov 2011 14:22:41 +0000 (16:22 +0200)]
Remove build dependency on git

If BUILD_HEAD and BUILD_BRANCH are passed as an argument the build no
longer calls git. This makes building qi possible from a source
tarball that no longer has ".git" without having to fear that git
finds some unrelated ".git" in some parent directory.

Signed-off-by: Timo Juhani Lindfors <timo.lindfors@iki.fi>
Signed-off-by: Paul Fertser <fercerpav@gmail.com>

2 years agogta02: replace a magic number with a define
Timo Juhani Lindfors [Fri, 4 Nov 2011 14:22:40 +0000 (16:22 +0200)]
gta02: replace a magic number with a define

This patch makes it easier to modify kernel source list in the future,
it does not cause any functional changes.

Signed-off-by: Timo Juhani Lindfors <timo.lindfors@iki.fi>
Signed-off-by: Paul Fertser <fercerpav@gmail.com>

2 years agoOutput \r before \n on serial console
Paul Fertser [Thu, 1 Sep 2011 18:06:04 +0000 (22:06 +0400)]
Output \r before \n on serial console

This makes it consistent with how Linux (and other bootloaders) work.

Reported-by: Timo Juhani Lindfors <timo.lindfors@iki.fi>
Signed-off-by: Paul Fertser <fercerpav@gmail.com>

2 years agoFix build with GCC 4.6 -Werror=unused-but-set-variable
Timo Juhani Lindfors [Thu, 1 Sep 2011 17:15:39 +0000 (20:15 +0300)]
Fix build with GCC 4.6 -Werror=unused-but-set-variable

This patch fixes the following build failures:

src/drivers/glamo-mmc.c: In function ‘mmc_write’:
src/drivers/glamo-mmc.c:492:6: error: variable ‘resp’ set but not used [-Werror=unused-but-set-variable]

src/cpu/s3c2442/s3c24xx-mci.c: In function ‘mmc_block_read’:
src/cpu/s3c2442/s3c24xx-mci.c:143:7: error: variable ‘resp’ set but not used [-Werror=unused-but-set-variable]

Signed-off-by: Timo Juhani Lindfors <timo.lindfors@iki.fi>

4 years agoFixed compilation for native toolchain
Paul Fertser [Tue, 5 Jan 2010 11:56:43 +0000 (14:56 +0300)]
Fixed compilation for native toolchain

These changes make it possible to compile Qi both natively and with the
help of cross-compiler (assuming you provide proper CROSS_COMPILE env
variable) with modern toolchains.

Reported-by: Paul Wise <pabs3@bonedaddy.net>
Signed-off-by: Paul Fertser <fercerpav@gmail.com>

4 years agoMakefile: use standard timestamp representation
Paul Fertser [Tue, 5 Jan 2010 11:54:45 +0000 (14:54 +0300)]
Makefile: use standard timestamp representation

Do not use locale-specific date representation since it might be confusing to
see on a serial console.

Signed-off-by: Paul Fertser <fercerpav@gmail.com>

4 years agoglamo-mmc: fix bogus MMC_IDLE test
Paul Fertser [Tue, 5 Jan 2010 11:53:10 +0000 (14:53 +0300)]
glamo-mmc: fix bogus MMC_IDLE test

Due to missing parentheses the test always evaluated to false, hence !MMC_IDLE
condition got ignored.

Signed-off-by: Paul Fertser <fercerpav@gmail.com>

4 years agoRevert "Remove loglevel from default kernel boot arguments"
Nelson Castillo [Wed, 10 Jun 2009 13:40:53 +0000 (10:40 -0300)]
Revert "Remove loglevel from default kernel boot arguments"

This reverts commit 01e049d704cfba97192efeed6afabd71cf7ca130.

It seems default loglevel cannot be set in kernel .config.

4 years agoRemove loglevel from default kernel boot arguments
Sebastian Krzyszkowiak [Sat, 6 Jun 2009 20:18:28 +0000 (22:18 +0200)]
Remove loglevel from default kernel boot arguments

Remove loglevel from default kernel boot arguments to
let kernel use its default configuration.

Signed-off-by: Sebastian Krzyszkowiak <seba.dos1@gmail.com>

4 years agoGTA02: set battery full charging current threshold to 2/32*Ichg
Paul Fertser [Tue, 2 Jun 2009 00:21:26 +0000 (04:21 +0400)]
GTA02: set battery full charging current threshold to 2/32*Ichg

In most conditions the charger is unable to detect battery full status
because gsm modem continiously draws current directly from the battery.
This makes it less sensitive and therefore will cutoff charging properly.

Signed-off-by: Paul Fertser <fercerpav@gmail.com>

4 years agoUse different MACs for the host and for the device ends of usb link
Paul Fertser [Thu, 28 May 2009 23:01:20 +0000 (20:01 -0300)]
Use different MACs for the host and for the device ends of usb link

This makes Qi consistent with current SHR boot script. It assigns
the MAC specified on the identity partition to the device and the
next address to the host, ie: from "g_ether.dev_addr=00:1F:11:01:58:67"
generate "g_ether.host_addr=00:1F:11:01:58:68".

Signed-off-by: Paul Fertser <fercerpav@gmail.com>

4 years agoAdded SDHC support
Paul Fertser [Sat, 23 May 2009 21:16:35 +0000 (01:16 +0400)]
Added SDHC support

Implemented and tested on Josch's GTA01Bv4

Signed-off-by: Paul Fertser <fercerpav@gmail.com>

4 years agoBad magic when booting from NAND
Dmitry Kurochkin [Thu, 9 Apr 2009 20:14:49 +0000 (00:14 +0400)]
Bad magic when booting from NAND

Qi fails to boot from NAND. Qi booting from SD goes fine. Booting NAND
from NOR works as well.

I have found the problem. The error was because we check if the
current block is bad or the next one is bad. And skip the current
block even if only the next one is bad. This way we skip good block
before a bad one.

This patch fixes the bug.

Signed-off-by: Dmitry Kurochkin <dmitry.kurochkin@gmail.com>

5 years agoqi-3d7k-more-suspend-gpio-meddle.patch
Andy Green [Thu, 12 Mar 2009 09:05:12 +0000 (09:05 +0000)]
qi-3d7k-more-suspend-gpio-meddle.patch

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoMy revised version of the partition tool
Christopher Hall [Wed, 11 Mar 2009 07:32:15 +0000 (07:32 +0000)]
My revised version of the partition tool

The fdisk and mkfs was factored out into a shell function
so that it is easier to create dirfferent partition layouts.

The default case was left the same as the original script

The partitions can be specified as percentages of the
available space (after Qi reserved space has been taken out)
by providing a comma separated list of values:
e.g. 25,0,50
Which would create partitions 1 and 3 with 25% and 50%
of the usable capacity. Partitions 2 and 4 would not be
assigned.

Signed-off-by: Christopher Hall <hsw@openmoko.com>

5 years agoqi-3d7k-additional-power-reg-config.patch
Andy Green [Wed, 11 Mar 2009 07:32:15 +0000 (07:32 +0000)]
qi-3d7k-additional-power-reg-config.patch

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-3d7k-suspend-gpio-tune.patch
Andy Green [Wed, 11 Mar 2009 07:32:14 +0000 (07:32 +0000)]
qi-3d7k-suspend-gpio-tune.patch

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-3d7k-gate-all-unused-clocks.patch
Andy Green [Tue, 10 Mar 2009 09:14:05 +0000 (09:14 +0000)]
qi-3d7k-gate-all-unused-clocks.patch

s3c6410 defaults to all clocks up, this leads to 120mA idle current at 5V.

Gating all the unused clocks off gets us down to 76mA idle current at 533Mhz
CPU clock down to 133MHz only helps by a further ~1.5mA.

During these tests camera unit was active with no camera attached, no EDGE.

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoPossible mistake in qi/6410-partition.sh
Christopher Hall [Mon, 9 Mar 2009 12:00:57 +0000 (12:00 +0000)]
Possible mistake in qi/6410-partition.sh

Hello Andy,

I am in the process of making a modified 6410-partition.sh
and I wonder if there is a mistake in the VFAT partition size:

Shouldn't the "+ $REARSECTORS" in the FATSECTORS calculation
be a subtraction,  otherwise the backupfs will overlap the
space reserved for Qi at the end of the SD Card.

Please look at this dif to see what I mean:

5 years agoqi-3d7k-GPK6-should-be-set-as-output.patch
Matt Hsu [Mon, 9 Mar 2009 08:51:06 +0000 (08:51 +0000)]
qi-3d7k-GPK6-should-be-set-as-output.patch

Hi Andy,

Sorry, previous patch to enable reset function of LCM is not working
stably somehow. :(
Since it's pin configuration is input. Attached patch is to set LCM
reset pin as output.

Cheers,

Matt

5 years agoclean-backslash-strings.patch
Werner Almesberger [Mon, 9 Mar 2009 08:51:05 +0000 (08:51 +0000)]
clean-backslash-strings.patch

We can get rid of quite a few backslashes since in modern C,
consecutive strings are concatenated at compile time.

Signed-off-by: Werner Almesberger <werner@openmoko.org>

5 years agoReturn-Path: <openmoko-kernel-bounces@lists.openmoko.org>
Andy Green [Sat, 7 Mar 2009 07:35:26 +0000 (07:35 +0000)]
Return-Path: <openmoko-kernel-bounces@lists.openmoko.org>
Received: from mail.openmoko.org ([unix socket])
by mail.openmoko.org (Cyrus v2.1.18-IPv6-Debian-2.1.18-5.1) with LMTP; Sat, 07 Mar 2009 06:46:20 +0000
X-Sieve: CMU Sieve 2.2
Return-path: <openmoko-kernel-bounces@lists.openmoko.org>
Received: from sita.openmoko.org ([88.198.124.203])
by mail.openmoko.org with esmtps (TLS-1.0:RSA_AES_256_CBC_SHA1:32)
(Exim 4.63)
(envelope-from <openmoko-kernel-bounces@lists.openmoko.org>)
id 1LfqIe-000690-LR
for andy@imap.openmoko.org; Sat, 07 Mar 2009 06:46:20 +0000
Received: from localhost ([127.0.0.1] helo=sita.openmoko.org)
by sita.openmoko.org with esmtp (Exim 4.63)
(envelope-from <openmoko-kernel-bounces@lists.openmoko.org>)
id 1LfqIA-0005iO-2x; Sat, 07 Mar 2009 07:45:50 +0100
Received: from imap.tw.openmoko.org ([59.124.115.149] helo=aakash.openmoko.org)
by sita.openmoko.org with esmtp (Exim 4.63)
(envelope-from <matt_hsu@openmoko.org>) id 1LfqI0-0005cg-Rz
for openmoko-kernel@lists.openmoko.org; Sat, 07 Mar 2009 07:45:44 +0100
Received: from [172.16.22.173] (helo=abacus-om.tw.openmoko.com
ident=Debian-exim)
by aakash.openmoko.org with esmtps (TLS-1.0:RSA_AES_256_CBC_SHA1:32)
(Exim 4.63) (envelope-from <matt_hsu@openmoko.org>)
id 1LfqHA-0007Xa-Eu
for openmoko-kernel@lists.openmoko.org; Sat, 07 Mar 2009 14:45:09 +0800
Received: from matt by abacus-om.tw.openmoko.com with local (Exim 4.69)
(envelope-from <matt_hsu@openmoko.org>) id 1Lfqto-0006XS-9X
for openmoko-kernel@lists.openmoko.org; Sat, 07 Mar 2009 15:24:44 +0800
From: Matt Hsu <matt_hsu@openmoko.org>
To: openmoko-kernel@lists.openmoko.org
Date: Sat,  7 Mar 2009 15:24:42 +0800
Message-Id: <1236410684-20364-2-git-send-email-matt_hsu@openmoko.org>
X-Mailer: git-send-email 1.5.6.5
In-Reply-To: <1236410684-20364-1-git-send-email-matt_hsu@openmoko.org>
References: <1236410684-20364-1-git-send-email-matt_hsu@openmoko.org>
X-Spam-Checker-Version: SpamAssassin 3.2.3 (2007-08-08) on sita.openmoko.org
X-Spam-Level:
X-Spam-Status: No, score=-2.6 required=5.0 tests=AWL,BAYES_00 autolearn=ham
version=3.2.3
Subject: [PATCH 1/3] qi/om_3d7k: deassert LCM_RESET pin before getting kernel.
X-BeenThere: openmoko-kernel@lists.openmoko.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Discussion regarding the OpenMoko Linux Kernel and boot loader
<openmoko-kernel.lists.openmoko.org>
List-Unsubscribe: <http://lists.openmoko.org/mailman/listinfo/openmoko-kernel>,
<mailto:openmoko-kernel-request@lists.openmoko.org?subject=unsubscribe>
List-Archive: <http://lists.openmoko.org/pipermail/openmoko-kernel>
List-Post: <mailto:openmoko-kernel@lists.openmoko.org>
List-Help: <mailto:openmoko-kernel-request@lists.openmoko.org?subject=help>
List-Subscribe: <http://lists.openmoko.org/mailman/listinfo/openmoko-kernel>,
<mailto:openmoko-kernel-request@lists.openmoko.org?subject=subscribe>
Sender: openmoko-kernel-bounces@lists.openmoko.org
Errors-To: openmoko-kernel-bounces@lists.openmoko.org

Since the reset callback function of probing jbt6k74 is
eliminated for some reasons in kernel. In order to support
both of LCM driver device, l1k002 and jbt6k74, just issue this reset in qi stage.

Signed-off-by: Matt Hsu <matt_hsu@openmoko.org>

5 years agoqi-3d7k-regulator-suspend-tune.patch
Andy Green [Sat, 7 Mar 2009 07:35:25 +0000 (07:35 +0000)]
qi-3d7k-regulator-suspend-tune.patch

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-3d7k-audit-gpio.patch
Andy Green [Sat, 7 Mar 2009 07:35:25 +0000 (07:35 +0000)]
qi-3d7k-audit-gpio.patch

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-rename-gta03-3d7k.patch
Andy Green [Mon, 2 Mar 2009 18:51:08 +0000 (18:51 +0000)]
qi-rename-gta03-3d7k.patch

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-gta02-revert-unpowered-gsm-UART-safe-at-boot.patch
Andy Green [Mon, 2 Mar 2009 12:21:38 +0000 (12:21 +0000)]
qi-gta02-revert-unpowered-gsm-UART-safe-at-boot.patch

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agosimplify QI GTA03 boot-SD size discovery
Werner Almesberger [Sat, 28 Feb 2009 11:04:47 +0000 (11:04 +0000)]
simplify QI GTA03 boot-SD size discovery

This patch uses fdisk instead of dmesg to determine the size of
the uSD card. This should be less fragile and is much simpler
than the previous solution.

Signed-off-by: Werner Almesberger <werner@openmoko.org>

5 years ago[QI] make GTA03 boot-SD on Debian
Werner Almesberger [Wed, 25 Feb 2009 15:09:43 +0000 (15:09 +0000)]
[QI] make GTA03 boot-SD on Debian

6410-partition-sd.sh didn't handle Debian, which is just like Ubuntu 7.
(Tested on "lenny". YMMV.)

Signed-off-by: Werner Almesberger <werner@openmoko.org>

5 years agoqi-fix-prepend-debug-space.patch
Andy Green [Mon, 23 Feb 2009 02:15:27 +0000 (02:15 +0000)]
qi-fix-prepend-debug-space.patch

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-gta02-a7-as-a6.patch
Andy Green [Mon, 23 Feb 2009 02:15:26 +0000 (02:15 +0000)]
qi-gta02-a7-as-a6.patch

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-gta02-fix-gph-uart0-pin.patch
Andy Green [Mon, 23 Feb 2009 02:15:26 +0000 (02:15 +0000)]
qi-gta02-fix-gph-uart0-pin.patch

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-fix-adding-space-to-char-not-to-string.patch
Andy Green [Fri, 20 Feb 2009 05:10:44 +0000 (05:10 +0000)]
qi-fix-adding-space-to-char-not-to-string.patch

Reported-by: Sean McNeil <sean@mcneil.com>
Signed-off-by: Andy Green <andy@openmoko.com>

5 years agofix-trailing-space-ethernet-mac.patch
Andy Green [Fri, 20 Feb 2009 04:57:47 +0000 (04:57 +0000)]
fix-trailing-space-ethernet-mac.patch

There's no trailing space after the last mac address addition, it
was broken before the recent patch adding the second mac address.

Also, we don't need to add the first space by hand if it's unconditionally
added in the string afterwards.

Reported-by: Werner Almesberger <werner@openmoko.org>
Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoRe: [PATCH 7/8] qi-add-gta02-indentity-part-parsing.patch
Robert Piasek [Fri, 20 Feb 2009 02:34:36 +0000 (02:34 +0000)]
Re: [PATCH 7/8] qi-add-gta02-indentity-part-parsing.patch

Andy,

On Tuesday 03 February 2009 18:12:50 Andy Green wrote:
> Now everything else is in place, we are able to mount
> the GTA02 "identity" partition and extract the USB Ethernet
> MAC Address from it, and add it to the kernel commandline.
>
> This causes the Ethernet gadget to use the same MAC address
> each boot, simplifying DHCP server situation.  The MAC
> address in the identity partition is globally unique from
> the factory.

Because of this patch I can no longer load g_ether module with host_addr= and
dev_addr= parameters.

The module is always loaded using factory mac for HOST and random mac for DEV.

That messes up my setup a bit, as I use NetworkManager 0.7 on FreeRunner
itself (and it expects DEV mac to be the same each time).

Would it be a problem to also include g_ether.dev_addr?

Attached patch solves my problem.

Rob

5 years agofix-6410-partition-initial-offset.patch
Andy Green [Thu, 19 Feb 2009 03:30:56 +0000 (03:30 +0000)]
fix-6410-partition-initial-offset.patch

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-gta02-compute-nand-kernel-partition-offset.patch
Andy Green [Tue, 3 Feb 2009 18:06:40 +0000 (18:06 +0000)]
qi-gta02-compute-nand-kernel-partition-offset.patch

There's one more thing that wants dynpart computed block offset -
the kernel offset used for NAND boot.

This patch moves the business end of the dynparts computation into
the port_init_gta02() so it's done earlier, and writes the block
index into the board_api struct NAND option's partition block start
field.

It also adds a field to the nand_dynparts struct so we can hold both
the true length and true block offset for each partition for later
use by the original code in append_device_specific_cmdline_gta02.

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-add-gta02-indentity-part-parsing.patch
Andy Green [Tue, 3 Feb 2009 18:06:39 +0000 (18:06 +0000)]
qi-add-gta02-indentity-part-parsing.patch

Now everything else is in place, we are able to mount
the GTA02 "identity" partition and extract the USB Ethernet
MAC Address from it, and add it to the kernel commandline.

This causes the Ethernet gadget to use the same MAC address
each boot, simplifying DHCP server situation.  The MAC
address in the identity partition is globally unique from
the factory.

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-improve-nand-read-512-granularity.patch
Andy Green [Tue, 3 Feb 2009 18:06:39 +0000 (18:06 +0000)]
qi-improve-nand-read-512-granularity.patch

Previously we insisted on NAND block granularity read addressing
and buffer length, even though we normalized the block indexing to
512 bytes to be compatible with ext2 and SD.

This patch improves the read functions so they are completely 512-
block friendly, any number of 512 byte blocks can be fetched from
any 512 byte boundary now.

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-gta02-compute-and-append-dynparts.patch
Andy Green [Tue, 3 Feb 2009 18:06:38 +0000 (18:06 +0000)]
qi-gta02-compute-and-append-dynparts.patch

This adds dynamic partition computation support for Qi on GTA02.
It's for compatability with existing GTA02 using U-Boot partition
scheme where a bad block in the previous partition moves on the
start of the next partition by one block.

It's important that Qi has no private state, so we need to compute
the NAND arrangement each boot.  It turns out this is extremely fast
using Qi's bad block code originally from Xinagfu.

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-add-append_device_specific_cmdline-API.patch
Andy Green [Tue, 3 Feb 2009 18:06:38 +0000 (18:06 +0000)]
qi-add-append_device_specific_cmdline-API.patch

Add a board API callback that allows a device-specific
commandline area to be created at the time the kernel
commandline is being composed.

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-add-hex-sprintf-type-functions.patch
Andy Green [Tue, 3 Feb 2009 18:06:37 +0000 (18:06 +0000)]
qi-add-hex-sprintf-type-functions.patch

We need to print hex into string buffers now, only in
phase 2

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-add-post-serial-init-api.patch
Andy Green [Tue, 3 Feb 2009 18:06:36 +0000 (18:06 +0000)]
qi-add-post-serial-init-api.patch

It can be useful to have a device API that can print device-specific things
after serial is initialized.

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-export-s3c2442-nand-bad-block-check.patch
Andy Green [Tue, 3 Feb 2009 18:06:36 +0000 (18:06 +0000)]
qi-export-s3c2442-nand-bad-block-check.patch

At least GTA02 specific code is now interested in bad blocks.
Rename the function and export it.

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-fix-block-init-cache-logic.patch
Andy Green [Sun, 1 Feb 2009 19:28:18 +0000 (19:28 +0000)]
qi-fix-block-init-cache-logic.patch

Changes in the last couple of weeks aimed at cleaning this
code broke the block device init cache logic.  This patch
restores the logic and reduces the card init failures when
there is no SD Card present to 1 regardless of the number
of partitions probed on the card.

Together with the reduction in Glamo card wait on init this
reduces the delay before trying NAND to 1/9th of before the
patches.

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-reduce-glamo-mmc-wait-delay.patch
Andy Green [Sun, 1 Feb 2009 19:28:18 +0000 (19:28 +0000)]
qi-reduce-glamo-mmc-wait-delay.patch

3000 retries is a very long time for an SD card to wake up,
far longer than should be necessary.  This patch reduces it
to 1000.

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoFix handling of 60 bytes long symlinks
Paul Fertser [Sat, 31 Jan 2009 17:23:30 +0000 (17:23 +0000)]
Fix handling of 60 bytes long symlinks

On ext2 the symlink can be stored in inode itself if it's not larger than
60 bytes. If the symlink path is exactly 60 bytes, then one more byte is
needed to store terminating NULL, therefore the path is placed in a
separate block.

5 years agoNone
Werner Almesberger [Sat, 31 Jan 2009 17:23:30 +0000 (17:23 +0000)]
None

The PMU initializes all GPIOs to inputs in NoPower, including GPIO2,
which drives GSM_ON and has an external pull-up. Furthermore, we may
have entered PMU.Standby with the modem up.

Unlike u-boot, qi didn't initialize the GPIOs. With this patch is
does.

Reported-by: Paul Ferster <fercerpav@gmail.com>
Signed-off-by: Werner Almesberger <werner@openmoko.org>

5 years agoqi-strip-trailing-newlines-in-append-file.patch
Andy Green [Sat, 31 Jan 2009 17:23:29 +0000 (17:23 +0000)]
qi-strip-trailing-newlines-in-append-file.patch

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoDescribe interactive UI in README
Paul Fertser [Fri, 23 Jan 2009 11:26:44 +0000 (11:26 +0000)]
Describe interactive UI in README

5 years agoGTA01: use POWER button to append debug parameters to the kernel command line
Paul Fertser [Mon, 19 Jan 2009 01:37:09 +0000 (01:37 +0000)]
GTA01: use POWER button to append debug parameters to the kernel command line

This patch is based on the one tested on GTA02, it compiles but i could not
verify if it works.

5 years agoGTA02: use POWER button to append debug parameters to the kernel command line
Paul Fertser [Mon, 19 Jan 2009 01:37:09 +0000 (01:37 +0000)]
GTA02: use POWER button to append debug parameters to the kernel command line

5 years agoAdd a dedicated function to query debug action to board_api
Paul Fertser [Mon, 19 Jan 2009 01:37:09 +0000 (01:37 +0000)]
Add a dedicated function to query debug action to board_api

On the devices where we have only 2 physical buttons, one of those buttons
(i.e. POWER) is connected directly to PMU. The other button (AUX) is
connected directly to S3C pin and we can get its state immediately, it is
currently used to skip a boot possibility. To allow user to debug boot
problems we can use the POWER button, but we cannot poll for it too many
times as it slows down the boot considerably, therefore a dedicated
function is needed.

5 years agoqi-change-partitioning.patch
Andy Green [Mon, 19 Jan 2009 01:37:08 +0000 (01:37 +0000)]
qi-change-partitioning.patch

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agozimage: add support for the zImage format
Werner Almesberger [Fri, 9 Jan 2009 04:32:53 +0000 (04:32 +0000)]
zimage: add support for the zImage format

If the kernel we found isn't an uImage, try to use it as a zImage.

Signed-off-by: Werner Almesberger <werner@openmoko.org>

5 years agozimage: separate uImage loader from try_this_kernel
Werner Almesberger [Fri, 9 Jan 2009 04:32:52 +0000 (04:32 +0000)]
zimage: separate uImage loader from try_this_kernel

Finally all the restructuring pays off: we can cleanly separate the
uImage loader from all the rest, which will make it easy to add
loaders for other kernel image formats.

Signed-off-by: Werner Almesberger <werner@openmoko.org>

5 years agophase2: use a typedefed type for the kernel function
Werner Almesberger [Fri, 9 Jan 2009 04:32:51 +0000 (04:32 +0000)]
phase2: use a typedefed type for the kernel function

This makes things a little easier to read, particularly when we return
this pointer from a function, which we'll do in the next patch.

Signed-off-by: Werner Almesberger <werner@openmoko.org>

5 years agophase2: use "static" wherever possible
Werner Almesberger [Fri, 9 Jan 2009 04:32:51 +0000 (04:32 +0000)]
phase2: use "static" wherever possible

Makes it easier to determine what can be safely changed.

Signed-off-by: Werner Almesberger <werner@openmoko.org>

5 years agophase2: block init buggy x
Werner Almesberger [Fri, 9 Jan 2009 04:32:50 +0000 (04:32 +0000)]
phase2: block init buggy x

Is it really correct to set last_block_init to the next entry ?
It would seem that the current (i.e., bogus) values are then just
used without further ado. Untested.

Not-Yet-Signed-off-by: Werner Almesberger <werner@openmoko.org>

5 years agophase2: separate block init from try_this_kernel
Werner Almesberger [Fri, 9 Jan 2009 04:32:49 +0000 (04:32 +0000)]
phase2: separate block init from try_this_kernel

As an added benefit, we can drop one level of indentation.

Signed-off-by: Werner Almesberger <werner@openmoko.org>

5 years agophase2: separate CRC from try_this_kernel
Werner Almesberger [Fri, 9 Jan 2009 04:32:49 +0000 (04:32 +0000)]
phase2: separate CRC from try_this_kernel

As an added benefit, we can drop one level of indentation.

Signed-off-by: Werner Almesberger <werner@openmoko.org>

5 years agophase2: separate partition scan from try_this_kernel
Werner Almesberger [Fri, 9 Jan 2009 04:32:48 +0000 (04:32 +0000)]
phase2: separate partition scan from try_this_kernel

As an added benefit, we can drop one level of indentation.

Signed-off-by: Werner Almesberger <werner@openmoko.org>

5 years agophase2: separate parameter setup from try_this_kernel
Werner Almesberger [Fri, 9 Jan 2009 04:32:47 +0000 (04:32 +0000)]
phase2: separate parameter setup from try_this_kernel

It's still huge, but less painful to read.

Signed-off-by: Werner Almesberger <werner@openmoko.org>

5 years agophase2: no space after function
Werner Almesberger [Fri, 9 Jan 2009 04:32:46 +0000 (04:32 +0000)]
phase2: no space after function

Putting a space between a function name and the parenthesis following
it ain't the One True K&R Way.

Signed-off-by: Werner Almesberger <werner@openmoko.org>

5 years agophase2: simplify the bootloader_second_phase mega-loop
Werner Almesberger [Fri, 9 Jan 2009 04:32:46 +0000 (04:32 +0000)]
phase2: simplify the bootloader_second_phase mega-loop

This puts the loop body in a new function called try_this_kernel.
As an added benefit, we can drop one level of indentation.

This change is hard to read as a patch. It gets better if one just
applies it and then looks at it with "git diff -w" or similar.

Signed-off-by: Werner Almesberger <werner@openmoko.org>

5 years agoFix dmesg line filtering if uptime is too low
Christopher Hall [Wed, 7 Jan 2009 10:07:46 +0000 (10:07 +0000)]
Fix dmesg line filtering if uptime is too low

On Ubuntu 8.04.1 there can be spaces inside the initial bracketed
uptime number after a reboot.  These spaces disappear once the uptime
is large enough.  Therefore, use sed to filter out these spaces so
that the cut command that extracts SD card size will always work.

Signed-off-by: Christopher Hall <hsw@openmoko.com>

5 years agoqi-gta02-correct-filter-polarity.patch
Andy Green [Thu, 18 Dec 2008 13:49:44 +0000 (13:49 +0000)]
qi-gta02-correct-filter-polarity.patch

Filter is enabled by b4 being 0 in each case, not 1

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-gta02-rootdelay.patch
Andy Green [Thu, 18 Dec 2008 13:49:33 +0000 (13:49 +0000)]
qi-gta02-rootdelay.patch

There's not enough time between Glamo init (now after pcf50633 init)
and the completion of machine init before we try to use the boot device
in the case of GTA02 and SD boot.  So we add rootdelay=1 to the SD
card boot cases

Signed-off-by: Andy Green <andy@openmoko.com>

5 years ago[PATCH] Some SD cards require more retries in mmc_init
Paul Fertser [Sat, 13 Dec 2008 20:34:46 +0000 (20:34 +0000)]
[PATCH] Some SD cards require more retries in mmc_init

This patch effectively doubles maximum time allowed for retries performed by
mmc_init. Also, the delay was shortened to lower the average time needed to
init a card. Without it, i was unable to boot from Transcend TS8GUSDHC4
(appending rootdelay=1 was necessary as well).

5 years agoqi-gta03-suspend-gpio.patch
Andy Green [Tue, 9 Dec 2008 11:04:57 +0000 (11:04 +0000)]
qi-gta03-suspend-gpio.patch

Change to input / pulldown on most GPIO.  With no battery and just
USB power, this gets us into suspend with 9.5mA at 5V consumption... but
Ben Dooks told that we don't put the mDDR into deep sleep yet in the
platform pm code, so this should come right down when we have that.

The suspend is otherwise real though, PWREN to the PMU goes down, and
when we wake the device Qi is able to see it is a resume wake and jumps
back into Linux, where we currently die due to issues on s3c6410 platform
code getting worked on.

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-gta03-revert-to-bin-sh.patch
Andy Green [Fri, 5 Dec 2008 23:15:48 +0000 (23:15 +0000)]
qi-gta03-revert-to-bin-sh.patch

Rootfs is not quite ready for /sbin/init

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-fix-gta03-default-wlan-power.patch
Andy Green [Fri, 5 Dec 2008 23:15:48 +0000 (23:15 +0000)]
qi-fix-gta03-default-wlan-power.patch

Make sure WLAN module power is off by default - this controls
a P-Channel MOSFET that gates all the power there

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-gta03-fix-charging.patch
Andy Green [Fri, 5 Dec 2008 23:15:47 +0000 (23:15 +0000)]
qi-gta03-fix-charging.patch

These two changes get charging working on GTA03 with new battery.

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-gta02-improve-default-reg-states.patch
Andy Green [Fri, 5 Dec 2008 23:15:47 +0000 (23:15 +0000)]
qi-gta02-improve-default-reg-states.patch

Need to let PWREN take care of more regulators really

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-s3c6410-add-resume-path-processing.patch
Andy Green [Fri, 5 Dec 2008 23:15:47 +0000 (23:15 +0000)]
qi-s3c6410-add-resume-path-processing.patch

First go at resume processing for 6410, can't test it until Ben Dooks
confirms operation of Linux side on SMDK

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-gta03-rootdelay.patch
Andy Green [Mon, 1 Dec 2008 11:16:17 +0000 (11:16 +0000)]
qi-gta03-rootdelay.patch

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-gta02-no-inidcators-if-battery-low.patch
Andy Green [Mon, 1 Dec 2008 01:26:07 +0000 (01:26 +0000)]
qi-gta02-no-inidcators-if-battery-low.patch

This patch turns on the SYS and BAT monitoring filters, and checks if the battery
meets the BAT OK threshold.  If it doesn't, which is the case if the battery is
not present, it disables the "indicator" (eg, LED, vibrator) stuff and holds the
CPU at 200MHz during the boot into Linux.

This allows the GTA02 A6 here to boot with no battery up to the point it is
going to bring up backlight, and this with 100mA limit on USB at PMU.  Enabling
the threshold filters for battery and SYS seems to have been critical in
getting any stability with this.

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-gta01-02-03-always-tty0-console.patch
Andy Green [Mon, 1 Dec 2008 01:26:07 +0000 (01:26 +0000)]
qi-gta01-02-03-always-tty0-console.patch

tty0 is the LCM... this patch changes the defaults so that the kernel
always treats tty0 as a console, but it also sets the loglevel so that
only KERN_ERR or worse will be printed there.  With matching changes
to the kernel, most of the noise at KERN_ERR is reduced to KERN_INFO,
so it gets the behviour the LCM by default is not cluttered with
messages unless they are important during boot.

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-fix-only-init-same-block-device-once.patch
Andy Green [Mon, 1 Dec 2008 01:26:07 +0000 (01:26 +0000)]
qi-fix-only-init-same-block-device-once.patch

If we're accessing the same device, we don't need to keep init-ing it

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-clean-fail-partition-on-mount-fail.patch
Andy Green [Mon, 1 Dec 2008 01:26:06 +0000 (01:26 +0000)]
qi-clean-fail-partition-on-mount-fail.patch

Failure to mount the filesystem makes us give up on the whole partition
the moment it happens

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-introduce-ui-indicators.patch
Andy Green [Mon, 1 Dec 2008 01:26:06 +0000 (01:26 +0000)]
qi-introduce-ui-indicators.patch

Allow a board to "indicate" events if it likes on whatever it has

On GTA02 we light AUX during boot and run the vibrator briefly when we
skip a partition or device.

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-gta02-remap-steppingstone.patch
Andy Green [Mon, 1 Dec 2008 01:26:06 +0000 (01:26 +0000)]
qi-gta02-remap-steppingstone.patch

Despite what the docs say, steppingstone is always resident at
0x40000000 on s3c2442, this patch changes our linker script to
stick all steppingstone code there.  The mapping of steppingstone
at 0x0 is broken by OM[] bus change dynamically caused for example
by GTA02 AUX button actuation.

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-s3c2442-enable-d-cache.patch
Andy Green [Mon, 1 Dec 2008 01:26:06 +0000 (01:26 +0000)]
qi-s3c2442-enable-d-cache.patch

Seems we have no D-Cache enabled before?  Also defeat AUX as EINT

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-commandline-debug-added-if-ui-key-down.patch
Andy Green [Mon, 1 Dec 2008 01:26:05 +0000 (01:26 +0000)]
qi-commandline-debug-added-if-ui-key-down.patch

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-kernel-source-skip-if-ui-key-down.patch
Andy Green [Mon, 1 Dec 2008 01:26:05 +0000 (01:26 +0000)]
qi-kernel-source-skip-if-ui-key-down.patch

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-gta01-add-ui-debounce.patch
Andy Green [Mon, 1 Dec 2008 01:26:04 +0000 (01:26 +0000)]
qi-gta01-add-ui-debounce.patch

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-gta03-add-ui-debounce.patch
Andy Green [Mon, 1 Dec 2008 01:26:04 +0000 (01:26 +0000)]
qi-gta03-add-ui-debounce.patch

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-gta02-add-ui-debounce.patch
Andy Green [Mon, 1 Dec 2008 01:26:04 +0000 (01:26 +0000)]
qi-gta02-add-ui-debounce.patch

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-add-ui-api.patch
Andy Green [Mon, 1 Dec 2008 01:26:03 +0000 (01:26 +0000)]
qi-add-ui-api.patch

Introduce the ability for boards to have UI inputs for purposes decided by
Qi

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-fix-nand-noboot-append-disable.patch
Andy Green [Fri, 28 Nov 2008 10:16:45 +0000 (10:16 +0000)]
qi-fix-nand-noboot-append-disable.patch

Recent addition of append and noboot file checks broke NAND

Reported-by: Micael Henriksson <micael.h@gmail.com>
Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-move-to-sbin-init.patch
Christopher Hall [Fri, 28 Nov 2008 10:16:45 +0000 (10:16 +0000)]
qi-move-to-sbin-init.patch

Signed-off-by: Christopher Hall <hsw@openmoko.org>

5 years agoqi-fixes-append.patch
Andy Green [Fri, 28 Nov 2008 10:16:45 +0000 (10:16 +0000)]
qi-fixes-append.patch

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-add-noboot-and-append-check.patch
Andy Green [Fri, 28 Nov 2008 10:16:45 +0000 (10:16 +0000)]
qi-add-noboot-and-append-check.patch

This adds two features that can be set in the rootfs that Qi is checking.

 - if it finds /boot/noboot-<device>, eg /boot/noboot-GTA02 then it will
   skip the rootfs for booting and try the next one if any

 - if it finds /boot/append-<device>, eg, /boot/append-GTA02 then it will
   append the contents of this file to the kernel commandline.

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-gta03-framebuffer-init-to-c.patch
Andy Green [Fri, 28 Nov 2008 10:16:45 +0000 (10:16 +0000)]
qi-gta03-framebuffer-init-to-c.patch

Signed-off-by: Andy Green <andy@openmoko.com>

5 years agoqi-update-README.patch
Andy Green [Fri, 28 Nov 2008 10:16:45 +0000 (10:16 +0000)]
qi-update-README.patch

Signed-off-by: Andy Green <andy@openmoko.com>