"could not allocate 65535 byte recieve ring" messages

[Ron Arts]

Cedric Puddy wrote:
> 
> Hello folks,
> 
> My apologies if this is an old topic (I have not yet
> located the list archive, and am therefore posting
> a rather less informed that I would normally like to
> be).
> 
> In any case, I have recently started working with a
> resonably sized sample set of Linux 2.2.5 boxes (six)
> equiped with two RTL8139 NICs apiece.
> 
> The message "Could not allocate 65535 byte recieve ring"
> comes up occationally on boot (accompanied by the card

The only thing I have to say here is: me too.

I used a pretty non-standard configuration (7 PCI RealTeks in
one machine), I used kernel 2.0.37, afterwards switched
to 2.2.13 but still experience lock-ups.
I switched to Intel Ether Express Pro cards, this improved
things, but the machine still locks up when I do >4 hour
of full ethernet load on all 7 cards.
This happens on two identical machines, so it can't be 
a hardware malfunction, I thought it might be some hardware 
design problem (there aren't many backplanes with 7 PCI slots)

By the way I receive this message on more than one cards.
And indeed, it's not consistent. On some reboots it's there,
and sometimes it's not. Which is not to say the cards function
properly after that, because I get total kernel lockups after
10 minutes of network load.

Ron Arts

> not initializing), or very often when I attempt to
> re-initialize a cad while the system is running.  I
> cannot reliably reproduce this problem, though I have
> been able to produce it at least ounce on every system
> I have tried installing the cards in.  When a
> system has multiple cards, it usually only affects one
> card.  I do not have this problem with other cards,
> and other services do not fail.
> 
> I upgraded some systems to 2.2.13, and continued to
> have the problem.
> 
> I had someone with the appropriate experience and
> skill look through the kernel code and driver code
> in an attempt to figure out "who is at fault",
> and the verdict there was that the driver appeared
> to be making a perfectly ordinary request for
> a chunk of kernel memory, and that the kernel
> appearred to be refusing on account of believing
> itself to have instufficent resources to
> allocate the memory.
> 
> This would sound to me like a "general kernel issue",
> except that I have been unable to reproduce similar
> problems with my "mottly crew" of 3c905b, NE2000 (ISA),
> racals, and SMC cards.
> 
> I am therefore very interested in hearing if this
> is a "known issue" of some kind against either the
> driver or someother piece of code, if there is
> possibly some workaround that might be had (say,
> reducing the size of the recieve ring??), and
> so on.
> 
> Thanks to everyone who works on these drivers -
> though the only name I know is Donald Becker
> (those driver pages have been really handy!).
> 
>         Regards,
> 
>         -Cedric
> 
> -
> |  CCj/ClearLine - Unix/NT Administration and TCP/IP Network Services
> |  118 Louisa Street, Kitchener, Ontario, N2H 5M3, 519-741-2157
> \____________________________________________________________________
>    Cedric Puddy, IS Director            cedric@thinkers.org
>      PGP Key Available at:              http://www.thinkers.org/cedric
> 
>  | To unsubscribe, send mail to Majordomo@cesdis.gsfc.nasa.gov, and within the
>  |  body of the mail, include only the text:
>  |   unsubscribe this-list-name youraddress@wherever.org
>  | You will be unsubscribed as speedily as possible.
 | To unsubscribe, send mail to Majordomo@cesdis.gsfc.nasa.gov, and within the
 |  body of the mail, include only the text:
 |   unsubscribe this-list-name youraddress@wherever.org
 | You will be unsubscribed as speedily as possible.