This is a bug revealed in r41830.
First, the static variable `char nif[IFNAMSIZ]` of nl80211_phy2ifname()
would be zeroed out if the argument is "wlan0" or the like. This will
happen in the following call stack.
nl80211_get_scanlist("radio0", buf, len);
nl80211_phy2ifname("radio0") // return static var nif with content "wlan0"
nl80211_get_scanlist(nif, buf, len); // tail call
nl80211_get_mode(nif);
nl80211_phy2ifname(nif); // zero out nif
Later we try nl80211_ifadd("") which was supposed to create interface
"tmp.", but that won't happen because nl80211_msg() will put an invalid
ifidx 0 to the nlmsg.
Then iwinfo_ifup() and iwinfo_ifdown() would fail and happily
nl80211_get_scanlist() returned 0 and left *len undefined.
Signed-off-by: Yousong Zhou <yszhou4tech@gmail.com>
SVN-Revision: 42151
The GW5520 is a small form-factor single-board computer with the following
features:
* 70x100mm form-factor
* IMX6DL 800MHz SoC (IMX6Q optional)
* 512MB 32bit DDR3 SDRAM (up to 2GB optional)
* 256MB NAND FLASH (up to 2GB optional)
* Gateworks System Controller
* 2x front-panel Intel i210 GbE adapters with passive PoE support
* 2x MiniPCIe sockets with USB support
* 2x front-panel USB
* 1x rear-panel full-size HDMI connector
* 1x front-panel bi-color user LED
* 1x front-panel user pushbutton
* 1x rear-panel barrel jack for power
* 1x Application connector with:
* 2x TTL level UARTs
* 10x TTL level Digital IO
Signed-off-by: Tim Harvey <tharvey@gateworks.com>
SVN-Revision: 42148
The GW16083 Ethernet Expansion Mezzanine adds the following to supported
Gateworks baseboards:
* 7-port Ethernet Switch
* 4x RJ45 ports (ENET1-4) supporing 802.11af/at PoE (with optional PoE module)
* 2x RJ45 ports or SFP module (ENET5-6) (auto-selected)
This series adds support for a phy driver that adds support for ENET5/ENET6
PHY adding initialization for those PHY's and a polling mechanism that detects
SFP insertion and configuration.
Signed-off-by: Tim Harvey <tharvey@gateworks.com>
SVN-Revision: 42147
The GW16082 miniPCI Expansion Mezzanine has the INTA/B/C/D IRQ's reversed
from the PCI standard. This will soon be resolved in the bootloader via
devicetree, but in the meantime this will work around the issue.
Signed-off-by: Tim Harvey <tharvey@gateworks.com>
SVN-Revision: 42146
When looking for the first ipaddr also consider the current prefix just
like network_get_ipaddrs6 does. If ipv6-address was empty the function
did not return the first ipaddr even if the list was non-empty.
fixes commit 83e9122f88a002871d5cdf421cf6aa6052b7e006
Signed-off-by: Henning Schild <henning@hennsch.de>
SVN-Revision: 42139
Now that we migrated all users to dtb based detection, we can drop the
board fixup code.
Signed-off-by: Jonas Gorski <jogo@openwrt.org>
SVN-Revision: 42129
This requires individual images for each board version for now.
Linux partition was shrunk to ensure writing thewrong image won't
erase wifi calibration data.
Signed-off-by: Jonas Gorski <jogo@openwrt.org>
SVN-Revision: 42126
Add the required nodes to the dtsi files and code to prevent double
registration from the board support code.
Signed-off-by: Jonas Gorski <jogo@openwrt.org>
SVN-Revision: 42123
In preparation for switching to dtb based board identification, add
support for building lzma-loader and lzma cfe kernels with dtb
appended.
Signed-off-by: Jonas Gorski <jogo@openwrt.org>
SVN-Revision: 42121
Allow appending a dtb blob to the binary and use it for identifying the
board. Fall back to nvram based identification in case of no dtb passed.
Signed-off-by: Jonas Gorski <jogo@openwrt.org>
SVN-Revision: 42119
Utilize the new selective conntrack flushing facility to clear
out active conntrack entries referring to old IP addresses after
a firewall reload.
Signed-off-by: Jo-Philipp Wich <jow@openwrt.org>
SVN-Revision: 42114
Enable the separate feed repositories by default to align custom builds with the repositories
available on the snapshot download server.
Signed-off-by: Jo-Philipp Wich <jow@openwrt.org>
SVN-Revision: 42112
