Kernel 4.4 was ready for brcm47xx for almost a year now but I kept
postponing the bump due to problems with Linksys WRT300N v1.0. OpenWrt
and LEDE with 4.4 were hanging at the booting with the:
> Starting program at 0x80001000
(the last CFE message).
This was a permanent state, "make distclean" wasn't helping, I spent
hours debugging this and I was reliably reproducing the issue every
time. I also reported it on linux-mips ML in the thread:
> BCM4704 stopped booting with 4.4 (due to vmlinux size?)
After ~month I started working on WRT300N again. I got hangs as expected
every time I switched from 4.1 to 4.4. I started experimenting with:
1) TRX content (I tried dropping rootfs partition)
2) BZ_TEXT_START of lzma-loader
3) Flashing other variants of image: lzma compressed kernel (without a
loader), gzip compressed one, uncompressed one.
At some point I got rootfs-less image booting and after that I couldn't
reproduce problem anymore, even with a complete firmware. It seems like
hardware was in some locked/unstable state that got magically fixed.
I have LEDE working now, tested it even with "make distclean", it seems
we can bump kernel now. I'll keep testing it on WRT300N for some time.
Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
Linksys WRT300N V1 has pretty bugged CFE bootloader (it crashes in a lot
of situations) that doesn't accept .bin image.
Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
Local variable declarations outside of functions are illegal since the Busybox
update to v1.25.0, therfore remove them from the appropriate places.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
Refresh patches for all targets that support kernel 4.4.
Compile-tested on all targets that use kernel 4.4 and aren't marked broken.
Runtime-tested on ar71xx, octeon and x86/64.
Signed-off-by: Stijn Tintel <stijn@linux-ipv6.be>
There does not seem to be any meaningful difference in generated code.
This will save some time and space on snapshot builds
Signed-off-by: Felix Fietkau <nbd@nbd.name>
In kernel 4.7 there is upstreamed b53 driver using (mostly?) the same
symbols as our b53 does. Change our symbols so both drivers can coexist
in kernel tree.
Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Acked-by: Jonas Gorski <jonas.gorski@gmail.com>
The current code only partially invalidates both caches
because the cache size and cache-line size values are
incorrectly passed to the C code.
Fix the assembly code to pass the arguments in the correct
order.
Compile tested only.
Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
SVN-Revision: 49060
Apart from using our new building system there are 2 more changes:
1) Limit amount of images
So far we were generating all standard images (optimized one and two
with no loader) for every SUBTARGET. This is not needed, as e.g. the
only device requiring gzipped kernel is legacy Huawei E970.
2) Change output names
The new image building system requires specifying device name. This
forced picking some and resulted in:
openwrt-brcm47xx-$(SUBTARGET)-squashfs.trx
openwrt-brcm47xx-$(SUBTARGET)-squashfs-gz.trx
openwrt-brcm47xx-$(SUBTARGET)-squashfs-noloader-nodictionary.trx
becoming:
openwrt-brcm47xx-$(SUBTARGET)-standard-squashfs.trx
openwrt-brcm47xx-$(SUBTARGET)-standard-noloader-gz-squashfs.trx
openwrt-brcm47xx-$(SUBTARGET)-standard-noloader-nodictionarylzma-squashfs.trx
Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
SVN-Revision: 49006
This only drops WGR614 V9 which has 2 MiB flash and it's unlikely to get
any interest.
Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
SVN-Revision: 48975
Commit r47866 dropped default values which were handling WAN interface
at port 0. Fix it by handling 2 more cases on NVRAM values.
Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
SVN-Revision: 47932
Replace former uci-defaults.sh implementation with the uci-defaults-new.sh one
and update all users accordingly.
Signed-off-by: Jo-Philipp Wich <jow@openwrt.org>
SVN-Revision: 47867
compat-wireless/backports now contains a bcm47xx_nvram.h file to
backport some of the functions in it which are used by the bcmfmac
driver. This file just checks for the kernel versions and provide an
empty implementations on older kernel versions. This is OK on most
systems, but on bcm47xx / bcm53xx systems we want to call the real
functions here. This commit removes the file from backports in our
build process like we do it with the bcma and ssb header files. Instead
we add a recent version into our kernel so all code uses only one
header file. On bcm47xx / bcm53xx the real implementations of this code
will be used.
Reported-by: Hante Meuleman <meuleman@broadcom.com>
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
SVN-Revision: 47467
the trx utile uses a maximum image size of 7.2MB. There are brcm47xx
devices even with serial flash with bigger flash chips, but OpenWrt was
not able to create images for these devices. This patch provides an
additional parameter which increases this limit to 32 MB. There is a
warning in the trx utile code which suggests that bigger images could
overwrite the nvram partition on some devices, but normally the program
writing the image should make sure that it is safe to write it to the
flash.
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
SVN-Revision: 46872
This patch changes nothing on the behaviour, it just breaks long lines
with bin/trx to make it easier to add additional parameters.
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
SVN-Revision: 46871
Instead of letting each target define it themselves, create on in
include/image.mk and let the targets use it.
Signed-off-by: Jonas Gorski <jogo@openwrt.org>
SVN-Revision: 46596
We got check for nvram_len placed too early and the code never tried
reading CFE variables used on WGT634U. This is a regression introduced
in r45942 and reported in #20291.
Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
SVN-Revision: 46584
Extracting TRX to separated file in /tmp/ requires extra RAM which may
not be available on some chepaer devices. Instead of that lets pass a
proper extracting command to the default_do_upgrade.
Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
SVN-Revision: 45901
The mips74k subtarget of brcm47xx configures gcc to compile for mips32r2;
however, the generated kernel config for 3.14 and later kernels ends up
with CPU_MIPS32_R1 and CPU_MIPSR1 selected. The generated kernel config
for the 3.10 kernel (Barrier Breaker) properly selected CPU_MIPS32_R2 and
CPU_MIPSR2. Modify the default kernel config for mips74k to explicitly
select CPU_MIPS32_R2 and CPU_MIPSR2.
Signed-off-by: Nathan Hintz <nlhintz@hotmail.com>
Tested-by: Rafał Miłecki <zajec5@gmail.com>
SVN-Revision: 45469
First of all this patch was affecting other devices, not just Netgear
WNR834B. E.g. Linksys WRT300N v1 also has boardtype 0x0472.
This was breaking PCI on WRT300N, ssb couldn't detect core working in
hostmode (due to mips_busprobe32 failing).
Secondly this patch should not be needed for years now. From the analyze
of #8861 it seems it was needed just to make devices at slots other than
0 discoverable. Setting cardbus to 0 was simply making ssb allow access
to PCI config for devices other than bridge.
With the upstream commit a6c8462 (ssb: fix cardbus slot in hostmode) we
allow accessing device at slot 1 which should be enough for CardBus.
Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
SVN-Revision: 45293
KERNEL_IMAGE is used as target rule so reusing the same name causes:
Makefile:326: warning: overriding recipe for target `bin/brcm47xx/vmlinux.lzma'
Makefile:326: warning: ignoring old recipe for target `bin/brcm47xx/vmlinux.lzma'
Makefile:326: warning: overriding recipe for target `build_dir/target-mipsel_74kc+dsp2_uClibc-0.9.33.2/linux-brcm47xx_mips74k/vmlinux.lzma'
Makefile:326: warning: ignoring old recipe for target `build_dir/target-mipsel_74kc+dsp2_uClibc-0.9.33.2/linux-brcm47xx_mips74k/vmlinux.lzma'
Unfortunately this will cause copying vmlinux.lzma over and over like:
cp vmlinux.lzma FOO-kernel.bin
which is redundant on brcm47xx where we never modify kernel image.
Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
SVN-Revision: 45178
This makes sysupgrade reject firmware images for different devices.
Right now only support for 2 sample devices is implemented.
Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
SVN-Revision: 45170
Disable ARCH_NEEDS_CPU_IDLE_COUPLED by-default in generic config, since
only one platfrom (omap) needs them.
Signed-off-by: Sergey Ryazanov <ryazanov.s.a@gmail.com>
SVN-Revision: 44614
CONFIG_GENERIC_NET_UTILS is selected by CONFIG_NET and already enabled
in generic config, so we don't need this symbol in platform specific
configurations.
Signed-off-by: Sergey Ryazanov <ryazanov.s.a@gmail.com>
SVN-Revision: 44611
The flag is experimental and requires toolchain support, which we don't
provide. Also neither target has an FPU so it becomes meaningless
without the FPU emulator enabled.
Signed-off-by: Jonas Gorski <jogo@openwrt.org>
SVN-Revision: 44584
This will allow simple sysupgrade-s including simpler reverting to the
original firmware.
Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
SVN-Revision: 43799
This new argument is used right after starting regular preinit (which
happens if failsafe wasn't triggered). The main purpose of "preinit"
argument is to indicate that failsafe can be triggered, however we were
missing a way to inform user that we don't wait for a trigger anymore.
With this change it's clear when failsafe mode can be triggered.
Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
SVN-Revision: 43715
replace all occurences of LINUX_VERSION with the cleaner
approach. future kernel upgrades must mostly touch only
one file. the only platform left is netlogic, because it
uses a intermediate kernel 3.14.16
Signed-off-by: Bastian Bittorf <bittorf@bluebottle.com>
SVN-Revision: 43047
[base-files] shell-scripting: fix wrong usage of '==' operator
normally the '==' is used for invoking a regex parser and is a bashism.
all of the fixes just want to compare a string. the used busybox-ash
will silently "ignore" this mistake, but make it portable/clean at least.
this patch does not change the behavior/logic of the scripts.
Signed-off-by: Bastian Bittorf <bittorf@bluebottle.com>
SVN-Revision: 42911
There is a group of devices that lzma-loader doesn't work with. They
simply hang at "Starting program at 0x80001000" which is really hard to
debug and we didn't find any solution for this for years.
Broadcom doesn't use lzma-loader on these devices anyway. They decided
to drop lzma-loader and use less optimal LZMA compression that can be
handled by CFE itself (it doesn't use dictionary).
So support these devices we will need kernel compressed with different
parameters and trx without a loader.
Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
SVN-Revision: 42205
