On Mon, Jun 16, 2003 at 04:47:20AM -0700, David S. Miller wrote: > looking at /proc/interrupts, the irq for the tg3 (29) seems to be getting > about 10 interrupts per second...and no i don't have that much network > > This is absolutely normal, due to a chip bug that can deadlock > it we "kick it" 10 times per second and this forces an interrupt > even if no real events are pending. thanks. that kept me from spending a lot of time trying a direction that would have been ultimately fruitless. but unfortunately i haven't had any other bright ideas, and have made no real headway in solving this, despite hours of staring at it and littering the driver with debug prints. i only have one new piece of information to share, that hopefully will give someone a better bright idea. i yanked a 5701 card from another machine, it's 32-bit and 33mhz (which is different from the on-board 5704's they are 64-bit and 100mhz), and put it in a regular pci slot in the mobo. this 5701 card gets link just fine and passes packets very well. it is the same kernel binary, and therefore the same tg3 driver that didn't work for the on-board 5704's (in fact, the 5701 and 5704 can be seen to be failing concurrently). so it's something specific to the 5704's when on a 64-bit kernel (or perhaps just an x86_64 kernel). perhaps also specific to the 64-bit data path? i'll continue to flail at this, but if anyone has a bright idea please share. oh, and one more thing, i also tried the broadcom supplied driver (6.2.11). it failed identically: the switch shows link, the led on the back of the mobo shows link, the driver does not. and for reference, this is the tg3 driver as compiled in a 2.4.22-pre8 kernel. -- David W. Hankins "If you don't do it right the first time, Network Engineer you'll just have to do it again." -- Jack T. Hankins
Attachment:
pgp00072.pgp
Description: PGP signature
