Re: [PATCH] Improvement of behaviour when -s 0 is used

[Guy Harris <guy () alum mit edu>]

On Oct 14, 2011, at 2:51 PM, Magnus Gille wrote:

> I came across an issue with tcpdump where the linux kernel couldn't allocate
> memory properly when we ran tcpdump -s 0 on one of our boxes. Tcpdump sets
> snaplen to 65535 if -s 0 is provided and this became a problem for us, to
> get around this I changed the behaviour to query what the MTU actually is.

See the responses to the "tcpdump -s 0 improvement" message you sent on 17 October on the same topic (presumably 
because the 14 October message didn't show up - there was a problem with the tcpdump-workers mailing list, so messages 
had gotten stuck).
 
> I made this patch against git current,

This belongs in libpcap, not tcpdump; the Git trunk of libpcap should already be querying the MTU in cases where we can 
rely on it (currently, only Ethernet, and only if no segmentation/desegmentation offloading is being done).  See my 
response to the 17 October message for reasons why.-
This is the tcpdump-workers list.
Visit https://cod.sandelman.ca/ to unsubscribe.