Linux commit d75b1ade567ffab ("net: less interrupt masking in NAPI")
chose the samantics of the napi poll return value, which caused napi to
not resume interrups/polling anymore.
Fix the return value to restore expected behaviour.
Signed-off-by: Jonas Gorski <jogo@openwrt.org>
SVN-Revision: 44576
Legacy probe passed the IRQSTAT registers instead of the IRQMASK
registers causing all register accesses to be offset by 16 bytes.
Signed-off-by: Jonas Gorski <jogo@openwrt.org>
SVN-Revision: 44575
The driver expected data then dir, but both dts and legacy code passed
dir then data. Fix this by making the driver expect the registers in
ascending order, i.e. dir then data.
Signed-off-by: Jonas Gorski <jogo@openwrt.org>
SVN-Revision: 44574
Now that all gpio users without named gpios use DT, we do not need
to fix the gpio controller bases anymore.
Signed-off-by: Jonas Gorski <jogo@openwrt.org>
SVN-Revision: 44567
Register buttons and leds through DT for all available dts,
and remove them from the board files.
Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
[jogo: remove leds/buttons from board files
fix wrong led polarities for dsl-274xb-c2, cpva642, p870hw
comment out spi-gpio and associated leds]
Signed-off-by: Jonas Gorski <jogo@openwrt.org>
SVN-Revision: 44566
Add a generic mmio gpio controller based driver and probe it
through device tree.
Use aliases for base calculation until we converted all users to
device tree or named gpios.
Convert bcm63xx_enet's ephy-reset gpio to use a named gpio.
While at it, remove the duplicate reset gpio defintion for livebox.
Signed-off-by: Jonas Gorski <jogo@openwrt.org>
SVN-Revision: 44565
Add a dtsi for bcm3368, add a dts and profile for cvg834g, and convert
it to use ImageDTB. Since HCS requires more arguments, enhance the max
arguments of the call.
The image name is intentionally left blank to prevent non-initramfs
images to be built, as they currently contain no rootfs and consequently
won't work.
Signed-off-by: Jonas Gorski <jogo@openwrt.org>
SVN-Revision: 44561
Most 16 MiB RAM devices don't even load elf kernels, so it's safe to
assume we have at least 32 MB RAM. This fixes tftp booting with the
default package set as this already produces an uncompressed
kernel > 8 MiB.
New limit is 8 MiB compressed / 24 MiB compressed.
Signed-off-by: Jonas Gorski <jogo@openwrt.org>
SVN-Revision: 44558
Remove the need for the header file to be exported - we don't need most
of it anyway; all we care about are the offset of the rootfs length and
header crc fields.
Signed-off-by: Jonas Gorski <jogo@openwrt.org>
SVN-Revision: 44557
We do not need to align the start of read only rootfs's to erase blocks.
This allows us to write the squashfs rootfs directly behind the kernel,
potentially freeing up one erase block.
We still need to align for jffs2, so add a flag for imagetag to
optionally align the rootfs start.
Signed-off-by: Jonas Gorski <jogo@openwrt.org>
SVN-Revision: 44556
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
According to the vendor tarball, the TD-w8900GB's flash has 64k erase
block size, but CFE spans two blocks. So fixup the image offset
accordingly but keep block size at its default (64k).
Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
[jogo: add commit message, add image offset, change nvram offset]
Signed-off-by: Jonas Gorski <jogo@openwrt.org>
SVN-Revision: 43572
Add the required nodes for the interrupt controllers and register
them through DT when a DTB is present.
Signed-off-by: Jonas Gorski <jogo@openwrt.org>
SVN-Revision: 43457
Add irq-domain aware irqchip drivers for the irq controllers of bcm63xx
and switch to use them.
Signed-off-by: Jonas Gorski <jogo@openwrt.org>
SVN-Revision: 43454
If there is no board support in the kernel, it does not make sense to
build images for devices. So drop any images for board ids for which
there are nc corresponding board_info structs in the kernel.
Signed-off-by: Jonas Gorski <jogo@openwrt.org>
SVN-Revision: 43364
