Bug 15164

Summary:	rl0 driver, "no free buffer space avalaible"
Product:	Base System	Reporter:	fraysse <fraysse>
Component:	kern	Assignee:	Alfred Perlstein <alfred>
Status:	Closed FIXED
Severity:	Affects Only Me
Priority:	Normal
Version:	Unspecified
Hardware:	Any
OS:	Any

Description fraysse 1999-11-29 14:40:01 UTC

We have a server here which rl0 driver (the ethernet card is a Realtek 8139 10/100BaseTX) 
freezes from time to time.
It seems to be related to both bugs #12394 and #9537.
We get "no free buffer space avalaible" when trying to 
ping another computer and the only way to fix the problem is to restart 
the rl0 driver by doing a "ifconfig rl0 down; ifconfig rl0 up".
We think it might come from an overload of the server due to the number 
of emails it sends to the administrators for system administration.

Fix: 

Restart the driver :
"ifconfig rl0 down; ifconfig rl0 up"
How-To-Repeat: No real 100% sure way. Just let the server run and send tons of emails.
The driver should stop at some point.

Comment 1 ji 1999-11-29 15:02:19 UTC

I've had the same problem with the xl driver.  It appears that the
driver is occasionally consuming mbufs clusters faster than it can
release them.  This is not a memory leak -- eventually all the buffers
are returned.  Increasing NMBCLUSTERS seems to solve that problem:

options         NMBCLUSTERS=16384

/ji

Comment 2 williamr 2000-05-18 09:06:18 UTC

We've had simular problems with intel 10/100 cards...

Another fix was to change the link from it's 10 to a 100 and it went away

Comment 3 Alfred Perlstein freebsd_committer

2000-06-06 10:31:39 UTC

State Changed
From-To: open->feedback

I'm quite sure I've seen a lot of work go into the rl and xl drivers since 
3.2, I'd like to know if more recent versions of FreeBSD have addressed 
this problem so that I may close this problem report.

Comment 4 Alfred Perlstein freebsd_committer

2000-06-06 10:31:39 UTC

Responsible Changed
From-To: freebsd-bugs->alfred

I'm taking care of this.

Comment 5 Alfred Perlstein freebsd_committer

2000-06-07 08:27:06 UTC

State Changed
From-To: feedback->closed

Submitter unable to reproduce problem as environment has changed, there 
have been no recent problem reports with this extremely common chipset and 
there have been numerous fixes applied to the rl driver since the report 
was initially opened.

Comment 6 Stavros Grigorakakis 2002-05-02 14:09:18 UTC

Unfortunately i am having the same problem on a 4.5fbsd with 2 realtek
ethernet cards.
It is getting very annoying.

Ways to repeat the problem:
-Just try uploading a large file (ftp) from a machine direct linked to the
net (100Mbps) and it stops.

I am trying to monitor the problem
Up to now I am monitor only one of the 2 cards.
I have the following indications (not yet proofs)

-While the card stopped functioning, i tried pinging outside and i sniifed
the net through another
machine and detected that the card was actually sending packets and
packets were coming back to it .!

-I have installed a simple script to monitor whenever i loose contact (and
restore it through ifconfigs).
The result is that i monitored net losses at times when the card was not
receiveing heavy load (no load at all infact) .
or at least the traffic was not directed to that machine.!

I am trying to investigate more details on this problem but i would very
much appreciate some help!

The card i investigate is sitting comfortably on irq 5
rl0: <RealTek 8139
10/100BaseTX> port 0xdc00-0xdcff mem 0xdc000000-0xdc0000ff irq 5 at device
11.0 on pci0
rl0: Ethernet address: 00:c0:df:25:20:fc


Any help of Any kind is welcomed

Stavros