[U-Boot-Users] SCC Ethernet and "Linux crashes by early packets" ???
Hi there,
the CHANGELOG contains the lines
* Make sure SCC Ethernet is always stopped by the time we boot Linux to avoid Linux crashes by early packets coming in.
in "Changes for U-Boot 1.0.2".
My two question:
1.) How can I find out which changes in the CVS do excactly this? 2.) Do we need the same with 8xx FEC Ethernet?
I am asking because I am expierencing sporadic crashes when booting linux from U-Boot 1.0.0. But only when I init the ethernet (e.g. by doing "dhcp" or "tftp") before booting...
In message 40757661.50905@imc-berlin.de you wrote:
the CHANGELOG contains the lines
- Make sure SCC Ethernet is always stopped by the time we boot Linux to avoid Linux crashes by early packets coming in.
in "Changes for U-Boot 1.0.2".
My two question:
1.) How can I find out which changes in the CVS do excactly this?
Checking the CVS logs?
Try: cvs diff -r1.3 -r1.4 post/ether.c
2.) Do we need the same with 8xx FEC Ethernet?
I don't think so. Note that the ethernet interface was left in a non- stopped state by the POST code. Normal network use will (should) correctly shut down the interface.
Best regards,
Wolfgang Denk
Wolfgang Denk wrote:
In message 40757661.50905@imc-berlin.de you wrote:
the CHANGELOG contains the lines
- Make sure SCC Ethernet is always stopped by the time we boot Linux to avoid Linux crashes by early packets coming in.
in "Changes for U-Boot 1.0.2".
My two question:
1.) How can I find out which changes in the CVS do excactly this?
Checking the CVS logs?
Maybe I am to stupid. Of course I can find the line
"Make sure SCC Ethernet is always stopped by the time we boot Linux"
in the CVS logs. Actually it's there 128 times. But I could not find out which file excactly was changed for this purpose!
Try: cvs diff -r1.3 -r1.4 post/ether.c
And I never thought of POST code! Thanks!
2.) Do we need the same with 8xx FEC Ethernet?
I don't think so. Note that the ethernet interface was left in a non- stopped state by the POST code. Normal network use will (should) correctly shut down the interface.
Hmm. Then I have another problem and start a new thread.
Thanks.
Steven
In message 407674A7.7040805@imc-berlin.de you wrote:
Checking the CVS logs?
Maybe I am to stupid. Of course I can find the line
"Make sure SCC Ethernet is always stopped by the time we boot Linux"
in the CVS logs. Actually it's there 128 times. But I could not find out which file excactly was changed for this purpose!
Yes, I know; some other files were also changed during this checkin. Buit if you look at the patchset, it seems obvious to me that there was only one netweork related source file included. [You are using cvsps, aren't you?]
Try: cvs diff -r1.3 -r1.4 post/ether.c
And I never thought of POST code!
So did we [not]. Don't ask how loong it took to trace down this problem, which apperared just for one customer, and just in their network environment [with a lot of windoze broadcasts], and never in our lab [which is 100% Micro$oft-free].
Hmm. Then I have another problem and start a new thread.
Good luck. May your pesticide be efficient ;-)
Best regards,
Wolfgang Denk
Wolfgang Denk wrote:
In message 40757661.50905@imc-berlin.de you wrote:
the CHANGELOG contains the lines
- Make sure SCC Ethernet is always stopped by the time we boot Linux to avoid Linux crashes by early packets coming in.
in "Changes for U-Boot 1.0.2".
My two question:
1.) How can I find out which changes in the CVS do excactly this?
Checking the CVS logs?
Try: cvs diff -r1.3 -r1.4 post/ether.c
2.) Do we need the same with 8xx FEC Ethernet?
I don't think so. Note that the ethernet interface was left in a non- stopped state by the POST code. Normal network use will (should) correctly shut down the interface.
Actually we most certainly do. I was bitten by the same bug the past week and have a patch ready; I'm just waiting for my other patches to be included in order to sent it.
When we leave any ethernet on with DMA capability, there is a small window of time before the new code reprograms the controller and stops the DMA.
In the mean time if a packet is received (typically a broadcast) it will write over in the physical memory where u-boot had placed it's buffers. It's a crapshoot. Most of the times no-one will notice since the possibility of actually overwritting something important is very small.
Best regards,
Wolfgang Denk
Regards
Pantelis
Pantelis Antoniou wrote:
I don't think so. Note that the ethernet interface was left in a non- stopped state by the POST code. Normal network use will (should) correctly shut down the interface.
Actually we most certainly do. I was bitten by the same bug the past week and have a patch ready; I'm just waiting for my other patches to be included in order to sent it.
Maybe you want to send this patch to me. So I could test if this fixes things for me as well?
Thanks!
participants (3)
-
Pantelis Antoniou -
Steven Scholz -
Wolfgang Denk