Messages in this thread | | | Subject | Re: [RFC][PATCH] linux-2.5.34_vsyscall_A0 | From | john stultz <> | Date | 21 Oct 2002 10:15:57 -0700 |
| |
On Mon, 2002-10-21 at 06:18, Alan Cox wrote: > On Fri, 2002-10-18 at 17:13, Stephen Hemminger wrote: > > It would be great to rework the whole TSC time of day stuff to work with > > per cpu data and allow unsychronized TSC's like NUMA. The problem is > > that for fast user level access, there would need to be some way to find > > The timer isnt even necessarily constant rate. The tsc is a nice tool > for debugging. Using it as a clock was not in the long run brilliant. > Don't try and continue it further, we have ACPI and HPET and other > better solutions in upcoming PC hardware.
Yes, I also feel all this per-cpu TSC stuff is not the way to go (on top of all this per-cpu mapping, etc. you'd also have to round-robin the timer interrupt so each cpu has a last_tsc_low value, then if cpus are varying in freq you have to recalc that occasionally. it just gets messy.). The current vsyscall implementation uses the TSC because on 99% of the boxes out there, the TSC is in sync and works fine as time source (ie: normal gettimeofday uses it). On boxes that don't use the TSC, the vsyscall shouldn't be mapped in until we have an alternate HPET(equv) vsyscall solution.
I'll see if I can clean that up later today. I also didn't mind Andrea's suggestion for using a /proc entry to disable/check for vsyscalls, so I might give that a whirl as well.
thanks for all the feedback, everyone! -john
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |