Hi Joe,
On Sa, 2015-03-21 at 22:59 -0500, Joe Hershberger wrote:
Hi Jörg,
On Sat, Mar 21, 2015 at 3:33 AM, Jörg Krause joerg.krause@embedded.rocks wrote:
Hi all,
there is an issue with the current network stack using netconsole. It's impossible to use network commands as TFTP inside netconsole, because they try to run as atomic network commands.
The issue was already reported by Stefano Babic in 2010: [U-Boot] NetConsole and network API http://lists.denx.de/pipermail/u-boot/2010-August/075535.html
I worked around this problem here:
http://lists.denx.de/pipermail/u-boot/2012-August/129913.html
I know about this patch, and I understand the problem it tries to workaround. But it does not work for using netconsole with another protocol like ping. When for example ping enters the NetLoop() it will halt the network device.
The problem for this is here: if (eth_is_on_demand_init() || protocol != NETCONS) { eth_halt(); ... }
eth_is_on_demand_init() returns correctly with false, because PING != NETCONS. But the second expression results in true, because PING != NETCONS.
I run into the same problem: [U-Boot] netconsole: USB Ethernet connection dropping with ping or tftpboot http://lists.denx.de/pipermail/u-boot/2015-February/203838.html
I didn't understand what about your case was not able to work given the workaround I implemented previously. What was different about it?
I have looked at the current network stack. The stack is based on the concept of atomic network commands. The implementation for netconsole looks very confusing.
There is no doubt that netconsole is quite confusing as it exists today.
I tried to fix the issue, but it did not work properly.
Sascha Hauer has reimplemented the network stack for Barebox: http://www.spinics.net/lists/u-boot-v2/msg00914.html
Looking at the current implementation of net.c looks very clean and well-designed.
Thanks for pointing this out. I hadn't gone to look at the network stack in barebox.
What do you think about porting this to U-Boot?
I can look into this. Naturally there are many other changes to u-boot network stack since the time barebox forked, so I expect such a port to be very intensive... most likely a near complete rewrite of Sascha's series, but I will investigate further.
I had a look at the patches and the code base is not entirely the same. But maybe we should just stick to the crucial points of the new network stack:
1) A network device gets initialized when it becomes the current one and gets brought down when it's not used anymore or prior booting to Linux so the connection remains active all the time.
My thoughts about this one: Bringing the device down can be done in eth_unregister(). Initializing for the current device can be done in eth_current_changed(). What I like even better is to make eth_current_changed() the sole place for calling init() and halt(). It would have to be extend to take at least two parameters, eth_current and new_current. If eth_current is null just bring up the new device, if new_current is null just bring down the current device.
2) Replace the NetLoop by a connection list where each protocol uses a connection to send and receive packets.
This induce to port all protocols to use the new network, of cause.
In my opinion it would be also good to do some clean ups and restructuring of code. Put more functions into net-utils, is there a need for eth_init and so on.
Best regards Jörg Krause