Date: Wed, 18 Sep 1996 10:08:53 -0600 From: "Mike Durian" <durian@plutotech.com> To: Michael Smith <msmith@atrad.adelaide.edu.au> Cc: freebsd-hackers@freebsd.org Subject: Re: Special Cycles on the PCI bus Message-ID: <199609181608.KAA12961@pluto.plutotech.com> In-Reply-To: Your message of "Wed, 18 Sep 1996 12:15:12 %2B0930."
next in thread | raw e-mail | index | archive | help
On Wed, 18 Sep 1996 12:15:12 +0930, Michael Smith <msmith@atrad.adelaide.edu.au> wrote: >Mike Durian stands accused of saying: > >There is regular timer-initiated activity with the 'ep' driver which >would appear to correlate with the symptoms you describe below. I ran the tests again in a system with the ethernet card physically removed. The special cycles still appeared. >Is there any other activity (eg. starting of timers) which happens at >some interval before this? Yes, initclocks() is called before the special cycles appear. In init_main.c:main(), the special cycles appear during the call to (*mountroot)(). >Is it possible that you're arriving here >by coincidence or are you breakpointing beforehand? We localized the problem to VOP_OPEN using DDB and breakpoints. We set a breakpoint before the VOP_OPEN call and after. The special cycles only appeared at the second breakpoint. >Did you have the 'ep' driver active at this point in time? If you >disable the 'ep' driver using userconfig, do these cycles still show up? The special cycles show up even if the 3c509 is removed from the system. >There might also be other activity hung off the timer. True, but we were seeing the special cycles immediately (or rather ~11ms) after both the 10ms clock and the 8ms clock. Not just the 10ms clock which is used for the timeout callbacks. >Are they causing you serious grief? They will. We're building custom hardware that uses 4 PCI busses and *lots* of bandwidth. We want to minimize the PCI bus traffic as much as possible. Since none of the chips we are using claim to produce these special cycles (and certainly not illegal ones), they are most likely being produced by some side effect. Perhaps some code is accidentally writing to an undocumented register in the Triton chipset that causes the special cycles. In any case, the special cycles are most likely symptoms of some software bug that might have other unwanted effects. Even if the special cycles are themselves innocuous, we should at least determine why they are generated. mike
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199609181608.KAA12961>