From: Darren Reed, Wed Dec 22 > Right. This seems to come up often - the Linux server is fragmenting > and sending out packets in an order that is the inverse of what everyone > else does and in an order that prevents "keep frags" from working.
I checked this with snoop, and indeed this is so for the external NFS server. I notified the administrator of that server. Now I tested this on another Solaris 7_x86 PC with one interface only and an internal Solaris (SPARC) NFS server, mounted with proto=udp. The (only) rules for testing were: pass out quick on elxl0 proto udp from THIS to NFSSERVER keep state keep frags block in log quick on elxl0 proto udp from NFSSERVER to THIS Here the fragments come in natural order, and with vers=2, no problems arise. But when using vers=3 (default), I still see some blocked fragments: 09:40:51.501154 elxl0 @0:1 b NFSSERVER -> THIS PR udp len 20 (1500) (frag 13483:[EMAIL PROTECTED]) IN 09:40:51.501472 elxl0 @0:1 b NFSSERVER -> THIS PR udp len 20 (1500) (frag 13483:[EMAIL PROTECTED]) IN 09:40:51.501694 elxl0 @0:1 b NFSSERVER -> THIS PR udp len 20 (1500) (frag 13483:[EMAIL PROTECTED]) IN etc.
