I've got a weird problem with my network at home. I'm getting IP
defragmentation timeouts on NFS reads. The NFS server is running
2.0.37, the client 2.3.13. They both have 3Com 3C905B 100bTX cards in
them, and are connected via an ethernet crossover. The client is a
486dx4-100, and the server is a P166. I can't think of an easy way to
check if this only affects NFS or is all UDP.
I didn't see any problems when I had a 10baseT hub between them
(unfortunately the hub is dead now so I can't check :(), and I don't
see the same problems when I boot the client under 2.0.37. I did
notice netscrape was unusally slow (running over X11). The cable was
home-assembled, so is potentially dodgy, but a flood ping shows a zero
packet loss. What's more confusing, ICMP fragments seem to work fine
(ie, ping -s 2000 or so). Forcing the card to run at 10Mbps doesn't
seem to help.
This is what tcpdump shows on the 2.3.13 client:
01:55:50.414913 molybdenum.innsmouth > hydro.innsmouth: icmp: ip reassembly time exceeded [tos 0xc0]01:55:50.414913 molybdenum.innsmouth.1166040940 > hydro.innsmouth.nfs: 120 read fh Unknown/1 4096 bytes @ 0
01:55:50.414913 hydro.innsmouth > molybdenum.innsmouth: (frag 48162:1244@2960)
01:55:50.414913 hydro.innsmouth.nfs > molybdenum.innsmouth.1166040940: reply ok 1472 read (frag 48162:1480@0+)
01:55:56.414913 molybdenum.innsmouth.1166040940 > hydro.innsmouth.nfs: 120 read fh Unknown/1 4096 bytes @ 0
01:55:56.414913 hydro.innsmouth > molybdenum.innsmouth: (frag 48163:1244@2960)
01:55:56.414913 hydro.innsmouth.nfs > molybdenum.innsmouth.1166040940: reply ok 1472 read (frag 48163:1480@0+)
01:56:02.414913 molybdenum.innsmouth > hydro.innsmouth: icmp: ip reassembly time exceeded [tos 0xc0]01:56:02.414913 molybdenum.innsmouth.1166040940 > hydro.innsmouth.nfs: 120 read fh Unknown/1 4096 bytes @ 0
01:56:02.414913 hydro.innsmouth > molybdenum.innsmouth: (frag 48167:1244@2960)
01:56:02.414913 hydro.innsmouth.nfs > molybdenum.innsmouth.1166040940: reply ok 1472 read (frag 48167:1480@0+)
And this is what I see on the 2.0.37 server:
01:30:59.322445 molybdenum.innsmouth.1166040940 > hydro.innsmouth.nfs: 120 read [|nfs]
01:30:59.322445 hydro.innsmouth > molybdenum.innsmouth: (frag 49558:1244@2960)
01:30:59.322445 hydro.innsmouth > molybdenum.innsmouth: (frag 49558:1480@1480+)
01:30:59.332445 hydro.innsmouth.nfs > molybdenum.innsmouth.1166040940: reply ok 1472 read [|nfs] (frag 49558:1480@0+)
Does anyone have any ideas, or have any tips for diagnosing/fixing the
problem?
-- Sam Vilain, sam@whoever.com work: sam.vilain@unisys.com http://www.hydro.gen.nz home: sam@hydro.gen.nz Hi, I'm a .signature virus! Copy me into your ~/.signature, please!- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/