On 2007.06.20 14:31:43 +0200, Andi Kleen wrote: > On Wednesday 20 June 2007 12:35:56 Björn Steinbrink wrote: > > The performance counter allocator is tied to the LAPIC NMI watchdog, > > That's not true. It's completely independent. It just happens to be in > the same file, but it has no direct functional ties to the watchdog.
The allocator relies on the wd_ops structure, which is only initialized if the LAPIC NMI watchdog is enabled. If it is not enabled, the functions that convert from absolute perfctr/evntsel addresses to offsets from the respective base address break, and always return 0. > > Fix the performance counter allocator by making it independent of the > > LAPIC NMI watchdog. This also fixes a bug in the LAPIC NMI watchdog > > which allocated the wrong performance counter on CPUs with PerfMon > > support. > > Combining code movement with functional fixes makes it impossible > to review properly. Don't do that please. Ok. Björn _______________________________________________ perfmon mailing list [email protected] http://www.hpl.hp.com/hosted/linux/mail-archives/perfmon/
