lkml.org 
[lkml]   [2008]   [Sep]   [4]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
    Patch in this message
    /
    Date
    From
    SubjectRe: [RFC patch 0/4] TSC calibration improvements

    > +static unsigned long quick_pit_calibrate(void)
    > +{
    [...]
    > + if (pit_expect_msb(0xff)) {
    > + int i;
    > + u64 t1, t2, delta;
    > + unsigned char expect = 0xfe;
    > +
    > + t1 = get_cycles();
    > + for (i = 0; i < QUICK_PIT_ITERATIONS; i++, expect--) {
    > + if (!pit_expect_msb(expect))
    > + goto failed;
    > + }
    > + t2 = get_cycles();

    hm, unless i'm missing something i think here we still have a small
    window for an SMI or some virtualization delay to slip in and cause
    massive inaccuracy: if the delay happens _after_ the last
    pit_expect_msb() and _before_ the external get_cycles() call. Right?

    i fixed that by adding one more pit_expect_msb() call.

    plus i think QUICK_PIT_ITERATIONS is quite close to overflowing 255
    which is built into the u32 'expect' variable (the MSB will only
    overflow to 10 bits or so) - so i've added a BUILD_BUG_ON() to make sure
    anyone tuning QUICK_PIT_MS above 60msec or so would get a build error
    instead of some hard(er) to track down calibration error.

    but it's getting late here so please double-check me ... The commit is
    below.

    Ingo

    ------------>
    From 40d2650256289d3ba59c4fd146b86b972db6ec40 Mon Sep 17 00:00:00 2001
    From: Ingo Molnar <mingo@elte.hu>
    Date: Thu, 4 Sep 2008 22:47:47 +0200
    Subject: [PATCH] x86: quick TSC calibration, improve

    - make sure the final TSC timestamp is reliable too

    - make sure nobody increases QUICK_PIT_MS so that
    QUICK_PIT_ITERATIONS can get larger than 0xff, breaking the iteration.
    (It would take about 60 msecs to reach that limit.)

    Signed-off-by: Ingo Molnar <mingo@elte.hu>
    ---
    arch/x86/kernel/tsc.c | 11 +++++++++++
    1 files changed, 11 insertions(+), 0 deletions(-)

    diff --git a/arch/x86/kernel/tsc.c b/arch/x86/kernel/tsc.c
    index 839070b..4832a40 100644
    --- a/arch/x86/kernel/tsc.c
    +++ b/arch/x86/kernel/tsc.c
    @@ -304,6 +304,11 @@ static unsigned long quick_pit_calibrate(void)
    outb(0xff, 0x42);
    outb(0xff, 0x42);

    + /*
    + * The iteration assumes that expect never goes below zero:
    + */
    + BUILD_BUG_ON(QUICK_PIT_ITERATIONS >= 0xff);
    +
    if (pit_expect_msb(0xff)) {
    int i;
    u64 t1, t2, delta;
    @@ -317,6 +322,12 @@ static unsigned long quick_pit_calibrate(void)
    t2 = get_cycles();

    /*
    + * Make sure we can rely on the second TSC timestamp:
    + */
    + if (!pit_expect_msb(--expect))
    + goto failed;
    +
    + /*
    * Ok, if we get here, then we've seen the
    * MSB of the PIT decrement QUICK_PIT_ITERATIONS
    * times, and each MSB had many hits, so we never

    \
     
     \ /
      Last update: 2008-09-04 22:55    [W:0.025 / U:91.704 seconds]
    ©2003-2016 Jasper Spaans. hosted at Digital OceanAdvertise on this site