Re: fxp(4) vs "too much" RAM?

["John D. Baker" <jdbaker%mylinuxisp.com@localhost>]

On Fri, 9 Jun 2017, Felix Deichmann wrote:

> Am 09.06.2017 um 08:37 schrieb John D. Baker:
> > If I install 3GB or less RAM, the on-board fxp(4) interface works
> > properly when NetBSD boots.
> > 
> > Observed with NetBSD/amd64-7.1_STABLE (local disk) and -current
> > (8.99.1 via netboot--almost).
> > 
> > Sound familiar to anyone?  Need more details?
> 
> Rings a bell and -- if not already fixed -- could be related to (rather
> longish thread):
>     http://mail-index.netbsd.org/port-xen/2009/10/30/msg005523.html
> 
> Synopsis is that bus_dma(9) usage in fxp could be a problem, or the
> device itself (64-bit DMA).
> 
> Manuel also posted a patch, no idea if it helps:
>     http://mail-index.netbsd.org/port-xen/2009/11/05/msg005541.html

I skimmed through that thread and it seems very likely to be the case.
I gathered from followups after the patch was posted that it did not help.

With 4GB in the machine, I ran through my entire stash of PCI ethernet
add-in cards and found the following:

  fxp fails (i8255[09] PCI add-in, i82562G on-board)
  pcn fails (lots of timeouts, may be failed HW?)
  rtk works (8139[CD])
  tlp works (LinkSys clone)
   ex works (3c905cx-txnm)
   re works (8169)
   wm works
   sk works in -8/-current (but see PR kern/49154), fails in 7.1

-- 
|/"\ John D. Baker, KN5UKS               NetBSD     Darwin/MacOS X
|\ / jdbaker[snail]mylinuxisp[flyspeck]com    OpenBSD            FreeBSD
| X  No HTML/proprietary data in email.   BSD just sits there and works!
|/ \ GPGkeyID:  D703 4A7E 479F 63F8 D3F4  BD99 9572 8F23 E4AD 1645