TCP/IP: Slow packets every so often, even with the TCP patch.

TCP/IP: Slow packets every so often, even with the TCP patch.

Post by Larry Rosenm » Tue, 23 Jul 1996 04:00:00



I have a 2.5 X86 Pentium 133 running as a WWW server.

when tracerouteing to it, and when refreshing or grabbing pages,
every 10th or so packet seems to take 200+ ms when normally it's
30ms over my ISDN link.

This system has the Recommended cluster (INCLUDING the TCP fix).

it does this with BOTH a Intel (eepro0) and SMC UltraPower (dnet0)
Ethernet card.

Anyone see this?

Any Ideas?

(Please POST and MAIL your responses, as I don't always get down this
far in my newsrc.)

Thanx!
--
Larry Rosenman                      http://www.lerctr.org/~ler

US Mail: 900 Lake Isle Circle, Irving, TX 75060-7726

 
 
 

TCP/IP: Slow packets every so often, even with the TCP patch.

Post by Peter Marela » Tue, 23 Jul 1996 04:00:00



> I have a 2.5 X86 Pentium 133 running as a WWW server.

> when tracerouteing to it, and when refreshing or grabbing pages,
> every 10th or so packet seems to take 200+ ms when normally it's
> 30ms over my ISDN link.

> This system has the Recommended cluster (INCLUDING the TCP fix).

> it does this with BOTH a Intel (eepro0) and SMC UltraPower (dnet0)
> Ethernet card.

> Anyone see this?

> Any Ideas?

> (Please POST and MAIL your responses, as I don't always get down this
> far in my newsrc.)

I suggest you use snoop to see what is going down the wire at the time.

Peter Marelas

 
 
 

1. Tracing TCP/IP packets from NIC to TCP

I've asked questions on this topic in the previous e-mails Subj:"Help me
get up to speed, please" and Subj:"From __cpu_raise_softirq() to
net_rx_action()".  I understand the packet flow from the NIC all the way
to the ip_rcv() function (and a little further).  So, maybe somebody can
fill in another little gap for me.

ip_rcv() call NF_HOOK( PF_INET, NF_IP_PRE_ROUTING, skb, dev, NULL,
ip_rcv_finish) which checks all netfilter hooks that are currently
registered in the PREROUTING hook.  All packets that are accepted by
these hook functions get passed to the ip_rcv_finish() function.  In
ip_rcv_finish, we call ip_route_input() which sets skb->dst.  Then
ip_rv_finish() returns skb->dst->input(skb).

I don't understand enough about the kernel's routing mechanisms to
figure out how these packets (sk_buffs) make their way to the tcp
stack.  Can somebody fill me in on exactly what the skb->dst structure
does and where the rt_hash_table array gets setup.  If I could just
follow this sk_buff until I start hitting code from tcp.c I'd be more
than satisfied.

Thanks,
   Josh

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in

More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

2. installing WHIP

3. How to convert TCP/IP packet to IPX packet and visa-versa ?

4. Tell what process is running on port

5. Diamond Viper VLB Grafx Accelerator?!?

6. !!DDE->TCP/IP, REAL TIME DATA PUMP, Trans.DDE App.DATA via TCP/IP

7. rperf 1.6 patch01 make: *** No way to make target `kernel.h.in'. Stop.

8. SLS X wants tcp/ip, tcp/ip won't compile, what now?