SCIENTIFIC-LINUX-USERS Archives

February 2009

SCIENTIFIC-LINUX-USERS@LISTSERV.FNAL.GOV

Options: Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
From:
Jon Peatfield <[log in to unmask]>
Reply To:
Jon Peatfield <[log in to unmask]>
Date:
Thu, 12 Feb 2009 21:41:03 +0000
Content-Type:
TEXT/PLAIN
Parts/Attachments:
TEXT/PLAIN (70 lines)
On Wed, 11 Feb 2009, Miles O'Neal wrote:

> Connie Sieh said...
>
> |We have been experiencing intermittent network failures on systems running
> |SLF47 2.6.9-78.0.1 kernel/tg3 3.86 driver .  They waiting for the tcp to
> |finish, which never happens.  The failures are load-related.
> |
> |This error happened from time to time over the life of these nodes
> |but they were operating more or less stably under SLF45/x86_64, tg3 3.77
> |driver.
>
> All I know is that I have found the Broadcom Linux drivers to be
> so unstable between releases that I try hard to avoid them.  They
> will work just fine with one release, but the next release we try
> may be a nightmare.
>
> We have a variety of servers from a few years ago that came with
> Broadcom on-board NICs.  Each worked with the initial install it
> had been certified with.  OS upgrades was painful, and in each
> case we gave up after several hours of fighting drivers and
> installed NIC cards using another vendor's chip.
>
> The only systems we've ever had that were rock solid with Broadcom
> are some Dell Optiplex 330n desktops.  But we've only run EL4.4 on
> them, and we're nervous about the switch to 5.x .  The NIC type was
> the one thing we forgot to check on the eval before doing the bulk
> order. 8^/

We have plenty of Optiplex boxes running sl5 and the 620, 647 models have 
broadcom chips though maybe not the same as the 330.  For what it's worth 
these are the lspci {,-n} lines for a 620 and a 745 for you to compare:

620:
  02:00.0 Ethernet controller: Broadcom Corporation NetXtreme BCM5751 Gigabit Ethernet PCI Express (rev 01)
  02:00.0 0200: 14e4:1677 (rev 01)

745:
  03:00.0 Ethernet controller: Broadcom Corporation NetXtreme BCM5754 Gigabit Ethernet PCI Express (rev 02)
  03:00.0 0200: 14e4:167a (rev 02)

those seem to work flawlessly with the sl5 stock kernels though we might 
have needed sl51 to support the chip in the 745.  Of course for the 755, 
760 Dell have gone back to using Intel chips (and the 760 needs the sl53 
kernels but otherwise seems fine).

We have another higher end desktop using the tg3 on sl5 which is also 
fine:

Dell Precision 490:
  0b:00.0 Ethernet controller: Broadcom Corporation NetXtreme BCM5752 Gigabit Ethernet PCI Express (rev 02)
  0b:00.0 0200: 14e4:1600 (rev 02)

We also have a Dell PE860 (cheap server) which uses the tg3 driver:

04:00.0 Ethernet controller: Broadcom Corporation NetXtreme BCM5721 Gigabit Ethernet PCI Express (rev 11)
05:00.0 Ethernet controller: Broadcom Corporation NetXtreme BCM5721 Gigabit Ethernet PCI Express (rev 11)

04:00.0 Class 0200: 14e4:1659 (rev 11)
05:00.0 Class 0200: 14e4:1659 (rev 11)

and that seems solid enough, though it is running a custom kernel on an 
sl3 base (it is a firewall keeping user's laptops off our main networks so 
I'd get complaints if it was failing).

None of the 'non-Dell' machines we have (working with slX) seem to use the 
tg3 driver for some reason.  I'm not sure if that is co-incidence or not.

  -- Jon

ATOM RSS1 RSS2