(
Jump to the original submission)
Tue 06 Nov 2007 12:05:27 PM UTC, comment #7:
You can close it.
|
Tue 06 Nov 2007 10:18:39 AM UTC, comment #6:
So, no objection to close this one?
|
Tue 06 Nov 2007 10:06:29 AM UTC, comment #5:
Actualy it is the same.
|
Tue 06 Nov 2007 09:57:48 AM UTC, comment #4:
Can you check if the record in coment#2 if the same problem than you (to avoid to duplicate records)?
|
Tue 06 Nov 2007 09:55:58 AM UTC, comment #3:
This document has good description of different approaches to zero window and keep alive implementations.
http://www.usenix.org/publications/library/proceedings/bos94/full_papers/lin.a
|
Tue 06 Nov 2007 09:41:34 AM UTC, comment #2:
>That's correct and it was already discussed (I belive we had a bug report about that around here somewhere, but I didn't find it, so I'll leave this open).
I suppose you think to :
bug #20511 : No persist timer (https://savannah.nongnu.org/bugs/?20511)
|
Tue 06 Nov 2007 09:36:38 AM UTC, comment #1:
You mean host2 is lwIP, in other words: lwIP doesn't have a timer probing for zero windows.
That's correct and it was already discussed (I belive we had a bug report about that around here somewhere, but I didn't find it, so I'll leave this open).
|
Tue 06 Nov 2007 09:09:36 AM UTC, original submission:
I think following case is not processed in lwip correct way.
Communication between host1 and host2.
Host1 advertises zero window.
Unacknoledged queue of the host2 is empty,
unsent queue contains some data.
In accordance with RFC-793 Section 3.7, host2 should
send zero-window probe, but it does not.
If host1 goes down, host2 is still sure, that connection is OK.
|
No files currently attached
Depends on the following items: None found
Items that depend on this one: None found
Follow 2 latest changes.