r6040_rx() was rewritten.
Further changes:
- fix a bug in hex2str(), which overwrote the source with a zero
- removed rx_free_desc counter; allocate skb inside r6040_rx() on the fly
- r6040_rx: don't take the error from the MLSR register (may be too late) but from the descriptor (we may have to allow these errors (via MCR0) there
- r6040_alloc_rxbufs() may fail due to lack of skb, added a return value
- changed some debugs
- more defines
- version: 0.18 -> 0.19
Signed-Off by Joerg Albert <jal2@gmx.de>
git-svn-id: svn://svn.openwrt.org/openwrt/trunk@11508 3c298f89-4303-0410-b956-a3cf2f4a3e73
- re-added the parent parameter to specify a netdevice to steal the MAC address from (optional).
- fixed null pointer access in r6040_rx() (lp->dev init. was missing)
- fixed "scheduling while atomic" endless loop if tx_timeout() was called (caused by the re-allocs of the buffers)
- fixed tx timeouts (tx irq weren't re-enabled on rx irq)
It adds some defines and cleans the code a bit (IMHO).
Signed-Off by Joerg Albert <jal2@gmx.de>
git-svn-id: svn://svn.openwrt.org/openwrt/trunk@11507 3c298f89-4303-0410-b956-a3cf2f4a3e73
this is the first of three patches to fix the ethernet driver of RDC R32xx applied to svn r10754.
It adds debugs and make some procedures (r6040_*) global symbols in order to help with debugging kernel oops. Version is increased from 0.16 to 0.18 to avoid conflicts with the driver variant metioned in the X-WRT forum (call 0.17.3) Tested on a Sitecom WL-153.
Signed-Off-By: Joerg Albert <jal2@gmx.de>
git-svn-id: svn://svn.openwrt.org/openwrt/trunk@11506 3c298f89-4303-0410-b956-a3cf2f4a3e73
This is the bcm57xx package. I have tested default vlan functions,
but I dont have the equipment to test more advanced setups. The default
vlan setup seems to be working fine. I also added the activate_gpio
parameter which will make the driver activate the switch via gpio before
probing for it.
I'm not sure which method is best for autoload. For the wrt350n, I
need the activate_gpio parameter. But its probably not a good idea
to add that to the autoload file. On a system without a bcm57xx switch,
isn't it a bad idea to mess with the gpios looking for the switch? Ideally,
wouldn't it be best to load the bcm57xx module from broadcom-diag, after
it has determined which router its on? I tried using 'request_module' from
there, but had no success. For now, I am relying on preinit to load
the bcm57xx module with activate_gpio param, after it has failed to load
switch_robo and switch_adm.
Signed-off-by: Ben Pfountz <netprince (at) vt (dot) edu>
git-svn-id: svn://svn.openwrt.org/openwrt/trunk@11471 3c298f89-4303-0410-b956-a3cf2f4a3e73
I noticed my wrt350n would not reliably reboot after entering
the reboot command. I found this code in the source for the
wrt600n. It corrects the problem, and the wrt350n reboots
reliably now.
Signed-off-by: Ben Pfountz <netprince (at) vt (dot) edu>
git-svn-id: svn://svn.openwrt.org/openwrt/trunk@11470 3c298f89-4303-0410-b956-a3cf2f4a3e73
This patch prevents switch-robo.c from attempting robo_probe on a port
that is already registered. robo_probe will adjust kernel reference counts
if it detects a switch on the port. If this patch wasn't applied, the
wrt350n would hang on reboot, waiting for the network driver reference count
to reach zero indefinitely.
Signed-off-by: Ben Pfountz <netprince (at) vt (dot) edu>
git-svn-id: svn://svn.openwrt.org/openwrt/trunk@11469 3c298f89-4303-0410-b956-a3cf2f4a3e73
This patch allows the bcm57xx module to work correctly with failsafe mode.
insmod doesn't return an error when a module loads but doesn't detect a switch.
I added the check_module function to load the module, then make sure
it doesn't just exit immediately. This allows preinit to only attempt to
load the bcm57xx module when switch-robo and switch-adm dont detect a switch.
The activate_gpio parameter to bcm57xx simply instructs the module to attempt
to activate the switch via gpio before probing for the switch.
Tested on wrt350n.
Signed-off-by: Ben Pfountz <netprince (at) vt (dot) edu>
git-svn-id: svn://svn.openwrt.org/openwrt/trunk@11468 3c298f89-4303-0410-b956-a3cf2f4a3e73
Update the netconfig script to support boardtype 0x478. I've tested this
on the wrt350n, hopefully it will match the 600n as well.
Signed-off-by: Ben Pfountz <netprince (at) vt (dot) edu>
git-svn-id: svn://svn.openwrt.org/openwrt/trunk@11467 3c298f89-4303-0410-b956-a3cf2f4a3e73
I've been working to finish up the bcm57xx module package nbd
posted a few months ago. I am no expert, just had some spare
time and some motivation. Here is the background:
https://dev.openwrt.org/ticket/2744
This first patch disables the bcm57xx gpio setup in broadcom-diag.
The switch needs to be initialized by the driver so the driver can
then reset the switch ASAP. If the switch isn't reset quickly enough,
it will forward packets between the WAN and LAN, which will cause
problems with modems that only allow one mac to access the internet.
Tested on wrt350n.
Signed-off-by: Ben Pfountz <netprince (at) vt (dot) edu>
git-svn-id: svn://svn.openwrt.org/openwrt/trunk@11466 3c298f89-4303-0410-b956-a3cf2f4a3e73
