--On July 11, 2007 1:51:04 PM -0500 Josh Paetzel <[EMAIL PROTECTED]> wrote:
I was running the driver from STABLE until about March and it would bounce the link up and down a couple times a day even under relatively moderate load. (50-60 mbps) So perhaps there have been improvements since then, but I gave up dealing with it. The driver was utterly broken in 6.1-R (0.9.5), by the time 6.2-R was getting rolled the driver in STABLE would at least not tip over under TCP load, but it was trivial to wedge it with even moderate amounts of UDP. I eventually reached the conclusion (correct or not) that you can't fix crap hardware with a driver.
No offense, but I think that's the incorrect conclusion. It pains me to say it, but we have never had a single problem with the Broadcomms on Windows servers, and we have a boatload of them (at least 100). That seems to point to the driver being the source of the problem, not the hardware. Furthermore, when I first got the 1950, the NIC was completely unusable. It would lock up hard and require a reboot to function again. That problem was fixed in the next iteration of the driver, and, except for the link state problem, the NIC has functioned normally ever since. (I don't use jumbo frames.)
ISTM the driver is the source of all the problems associated with the card(s). I wish I knew enough to work on driver code, but I do not. I noticed that the guy writing the driver for FreeBSD works for Broadcomm. Perhaps he could comment?
Paul Schmehl ([EMAIL PROTECTED]) Senior Information Security Analyst The University of Texas at Dallas http://www.utdallas.edu/ir/security/
