Skip to contentSkip to conversation

HELP ! Why my dev->hard_start_xmit get called three times for one ping packet?

DISCUSSION ZI ZHOU 2 participants 3 messages Apr 15, 2004

lore
Share & tools
lore

ZI ZHOU

author
#0 lore raw

Hi,

I've written a ethernet device driver based on 2.6.4 sis190.c, and I am testing it on my 2.4.25 machine. I am currently testing the transmit part and couldn't figure out the scenorio I've seen. Can someone shed some light PLEASE !

After I configure the ethernet interface(it's a PCI card plugged in my debian PC), I run command ' ping -c 1 192.168.0.100', (the interface is configured with IP address 192.168.0.35), I used smartbit to look at the wire and found 3 ping packets. From the debugging info, I can see dev->hard_start_xmit is called three times, the DMA engine seems to be acting correctly, and each time hard_start_xmit is invoked the skb address is different. How can this happen? Could it be sth related to tx timeout? I am not very clear about what happens between IP send and dev->hard_start_xmit besides knowing there is a queue for skb, and if device is busy, the skb will be re-queued? Can someone give me some idea about how the driver communicate with the upper layer software about the status of its transmit ? There used to be tbusy I remember, but I've lost track of what happen there after.

Btw, I run different number of ping packets, here is some result, Haven't been able to make sense out of it yet.

ping -c 1/2/3  -> recv 3 packets
ping -c 4/5/6  -> recv 6 packets
ping -c 7      -> recv 8 packets

Thanks & your help is greatly appreciated !

Zi

__________________________________
Do you Yahoo!?
Yahoo! Tax Center - File online by April 15th
http://taxes.yahoo.com/filing.html

#1 lore raw

ZI> After I configure the ethernet interface(it's a PCI card ZI> plugged in my debian PC), I run command ' ping -c 1 ZI> 192.168.0.100', (the interface is configured with IP address ZI> 192.168.0.35), I used smartbit to look at the wire and found 3 ZI> ping packets. From the debugging info, I can see ZI> dev->hard_start_xmit is called three times, the DMA engine ZI> seems to be acting correctly, and each time hard_start_xmit is ZI> invoked the skb address is different. How can this happen?

Are any of the 3 packets actually arp packets?

- Roland

author
#2 lore raw
Quoted 4 lines

Are any of the 3 packets actually arp packets?

- Roland

Yes, they are arp packets and are exactly the same.

Zi

__________________________________
Do you Yahoo!?
Yahoo! Tax Center - File online by April 15th
http://taxes.yahoo.com/filing.html