Hi,
On Tue, Dec 13, 2022 at 10:01:06AM -0800, Ricardo Koller wrote:
> On Tue, Dec 13, 2022 at 04:43:38PM +0000, Alexandru Elisei wrote:
> > Hi,
> >
> > On Tue, Dec 13, 2022 at 08:21:24AM -0800, Ricardo Koller wrote:
> > > On Tue, Dec 13, 2022 at 12:36:14PM +0000, Alexandru Elisei wrote:
> > > > Hi,
> > > >
> > > > Some more comments below.
> > > >
> > > > On Fri, Dec 02, 2022 at 04:55:25AM +0000, Ricardo Koller wrote:
> > > > > PMUv3p5 uses 64-bit counters irrespective of whether the PMU is
> > > > > configured
> > > > > for overflowing at 32 or 64-bits. The consequence is that tests that
> > > > > check
> > > > > the counter values after overflowing should not assume that values
> > > > > will be
> > > > > wrapped around 32-bits: they overflow into the other half of the
> > > > > 64-bit
> > > > > counters on PMUv3p5.
> > > > >
> > > > > Fix tests by correctly checking overflowing-counters against the
> > > > > expected
> > > > > 64-bit value.
> > > > >
> > > > > Signed-off-by: Ricardo Koller <ricar...@google.com>
> > > > > ---
> > > > > arm/pmu.c | 29 ++++++++++++++++++-----------
> > > > > 1 file changed, 18 insertions(+), 11 deletions(-)
> > > > >
> > > > > diff --git a/arm/pmu.c b/arm/pmu.c
> > > > > index cd47b14..eeac984 100644
> > > > > --- a/arm/pmu.c
> > > > > +++ b/arm/pmu.c
> > > > > @@ -54,10 +54,10 @@
> > > > > #define EXT_COMMON_EVENTS_LOW 0x4000
> > > > > #define EXT_COMMON_EVENTS_HIGH 0x403F
> > > > >
> > > > > -#define ALL_SET 0xFFFFFFFF
> > > > > -#define ALL_CLEAR 0x0
> > > > > -#define PRE_OVERFLOW 0xFFFFFFF0
> > > > > -#define PRE_OVERFLOW2 0xFFFFFFDC
> > > > > +#define ALL_SET 0x00000000FFFFFFFFULL
> > > > > +#define ALL_CLEAR 0x0000000000000000ULL
> > > > > +#define PRE_OVERFLOW 0x00000000FFFFFFF0ULL
> > > > > +#define PRE_OVERFLOW2 0x00000000FFFFFFDCULL
> > > > >
> > > > > #define PMU_PPI 23
> > > > >
> > > > > @@ -538,6 +538,7 @@ static void test_mem_access(void)
> > > > > static void test_sw_incr(void)
> > > > > {
> > > > > uint32_t events[] = {SW_INCR, SW_INCR};
> > > > > + uint64_t cntr0;
> > > > > int i;
> > > > >
> > > > > if (!satisfy_prerequisites(events, ARRAY_SIZE(events)))
> > > > > @@ -572,9 +573,9 @@ static void test_sw_incr(void)
> > > > > write_sysreg(0x3, pmswinc_el0);
> > > > >
> > > > > isb();
> > > > > - report(read_regn_el0(pmevcntr, 0) == 84, "counter #1 after +
> > > > > 100 SW_INCR");
> > > > > - report(read_regn_el0(pmevcntr, 1) == 100,
> > > > > - "counter #0 after + 100 SW_INCR");
> > > > > + cntr0 = (pmu.version < ID_DFR0_PMU_V3_8_5) ? 84 : PRE_OVERFLOW
> > > > > + 100;
> > > > > + report(read_regn_el0(pmevcntr, 0) == cntr0, "counter #0 after +
> > > > > 100 SW_INCR");
> > > > > + report(read_regn_el0(pmevcntr, 1) == 100, "counter #1 after +
> > > > > 100 SW_INCR");
> > > > > report_info("counter values after 100 SW_INCR #0=%ld #1=%ld",
> > > > > read_regn_el0(pmevcntr, 0), read_regn_el0(pmevcntr,
> > > > > 1));
> > > > > report(read_sysreg(pmovsclr_el0) == 0x1,
> > > > > @@ -584,6 +585,7 @@ static void test_sw_incr(void)
> > > > > static void test_chained_counters(void)
> > > > > {
> > > > > uint32_t events[] = {CPU_CYCLES, CHAIN};
> > > > > + uint64_t cntr1;
> > > > >
> > > > > if (!satisfy_prerequisites(events, ARRAY_SIZE(events)))
> > > > > return;
> > > > > @@ -618,13 +620,16 @@ static void test_chained_counters(void)
> > > > >
> > > > > precise_instrs_loop(22, pmu.pmcr_ro | PMU_PMCR_E);
> > > > > report_info("overflow reg = 0x%lx", read_sysreg(pmovsclr_el0));
> > > > > - report(!read_regn_el0(pmevcntr, 1), "CHAIN counter #1 wrapped");
> > > > > + cntr1 = (pmu.version < ID_DFR0_PMU_V3_8_5) ? 0 : ALL_SET + 1;
> > > > > + report(read_regn_el0(pmevcntr, 1) == cntr1, "CHAIN counter #1
> > > > > wrapped");
> > > >
> > > > It looks to me like the intention of the test was to check that the
> > > > counter
> > > > programmed with the CHAIN event wraps, judging from the report message.
> > > >
> > >
> > > Ah, right. Yeah, that message is confusing. It should be the short
> > > version of "Inrementing at 32-bits resulted in the right value".
> > >
> > > > I think it would be interesting to keep that by programming counter #1
> > > > with
> > > > ~0ULL when PMUv3p5 (maybe call it ALL_SET64?) and test the counter value
> > > > against 0.
> > >
> > > The last commit adds tests using ALL_SET64. Tests can be run in two
> > > modes: overflow_at_64bits and not. However, this test,
> > > test_chained_counters(), and all other chained tests only use the
> > > !overflow_at_64bits mode (even after the last commit). The reason is
> > > that there are no CHAIN events when overflowing at 64-bits (more details
> > > in the commit message). But, don't worry, there are lots of tests that
> > > check wrapping at 64-bits (overflow_at_64bits=true).
> >
> > What I was suggesting is this (change on top of this series, not on top of
> > this patch, to get access to ALL_SET_AT):
>
> Ooh, I see, I agree: it would be better to check that the odd counter
> increments from ~0ULL to 0 when using 64-bit counters.
>
> >
> > diff --git a/arm/pmu.c b/arm/pmu.c
> > index 3cb563b1abe4..fd7f20fc6c52 100644
> > --- a/arm/pmu.c
> > +++ b/arm/pmu.c
> > @@ -607,7 +607,6 @@ static void test_sw_incr(bool overflow_at_64bits)
> > static void test_chained_counters(bool overflow_at_64bits)
> > {
> > uint32_t events[] = {CPU_CYCLES, CHAIN};
> > - uint64_t cntr1;
> >
> > if (!satisfy_prerequisites(events, ARRAY_SIZE(events),
> > overflow_at_64bits))
> > @@ -639,12 +638,11 @@ static void test_chained_counters(bool
> > overflow_at_64bits)
> > report(read_sysreg(pmovsclr_el0) == 0x1, "overflow recorded for
> > chained incr #2");
> >
> > write_regn_el0(pmevcntr, 0, PRE_OVERFLOW);
> > - write_regn_el0(pmevcntr, 1, ALL_SET);
> > + write_regn_el0(pmevcntr, 1, ALL_SET_AT(overflow_at_64bits));
>
> The only change is that this should be:
>
> ALL_SET_AT(pmu.version >= ID_DFR0_PMU_V3_8_5)
>
> Because "overflow_at_64bits" implies PMCR_EL0.LP = 1.
Right, and test_chained_counters() is never called with overflow_at_64bits
== true.
How about renaming the parameter overflow_at_64bits -> unused (or something
like that) to make it clear that the function does the same thing
regardless of the value?
Thanks,
Alex
>
> >
> > precise_instrs_loop(22, pmu.pmcr_ro | PMU_PMCR_E);
> > report_info("overflow reg = 0x%lx", read_sysreg(pmovsclr_el0));
> > - cntr1 = (pmu.version < ID_DFR0_PMU_V3_8_5) ? 0 : ALL_SET + 1;
> > - report(read_regn_el0(pmevcntr, 1) == cntr1, "CHAIN counter #1
> > wrapped");
> > + report(read_regn_el0(pmevcntr, 1) == 0, "CHAIN counter #1 wrapped");
> >
> > report(read_sysreg(pmovsclr_el0) == 0x3, "overflow on even and odd
> > counters");
> > }
> >
> > The counters are 64bit, but the CHAIN event should still work because
> > PMCR_EL0.LP is 0, according to the event description in the Arm ARM (ARM
> > DDI 0487I.a, page D17-6461).
> >
> > Thanks,
> > Alex
> >
> > >
> > > > Alternatively, the report message can be modified, and "wrapped"
> > > > replaced with "incremented" (or something like that), to avoid
> > > > confusion.
> > > >
> > > > > +
> > > > > report(read_sysreg(pmovsclr_el0) == 0x3, "overflow on even and
> > > > > odd counters");
> > > > > }
> > > > >
> > > > > static void test_chained_sw_incr(void)
> > > > > {
> > > > > uint32_t events[] = {SW_INCR, CHAIN};
> > > > > + uint64_t cntr0, cntr1;
> > > > > int i;
> > > > >
> > > > > if (!satisfy_prerequisites(events, ARRAY_SIZE(events)))
> > > > > @@ -665,10 +670,12 @@ static void test_chained_sw_incr(void)
> > > > > write_sysreg(0x1, pmswinc_el0);
> > > > >
> > > > > isb();
> > > > > + cntr0 = (pmu.version < ID_DFR0_PMU_V3_8_5) ? 0 : ALL_SET + 1;
> > > > > + cntr1 = (pmu.version < ID_DFR0_PMU_V3_8_5) ? 84 : PRE_OVERFLOW
> > > > > + 100;
> > > > > report((read_sysreg(pmovsclr_el0) == 0x3) &&
> > > > > - (read_regn_el0(pmevcntr, 1) == 0) &&
> > > > > - (read_regn_el0(pmevcntr, 0) == 84),
> > > > > - "expected overflows and values after 100
> > > > > SW_INCR/CHAIN");
> > > > > + (read_regn_el0(pmevcntr, 1) == cntr0) &&
> > > > > + (read_regn_el0(pmevcntr, 0) == cntr1),
> > > >
> > > > This is hard to parse, it would be better if counter 0 is compared
> > > > against
> > > > cntr0 and counter 1 against cntr1.
> > >
> > > Ah, yes, code is correct but that's indeed confusing.
> > >
> > > >
> > > > Also, same suggestion here, looks like the test wants to check that the
> > > > odd-numbered counter wraps around when counting the CHAIN event.
> > >
> > > Ack. Will update for v2.
> > >
> > > Thanks!
> > > Ricardo
> > >
> > > >
> > > > Thanks,
> > > > Alex
> > > >
> > > > > + "expected overflows and values after 100 SW_INCR/CHAIN");
> > > > > report_info("overflow=0x%lx, #0=%ld #1=%ld",
> > > > > read_sysreg(pmovsclr_el0),
> > > > > read_regn_el0(pmevcntr, 0), read_regn_el0(pmevcntr,
> > > > > 1));
> > > > > }
> > > > > --
> > > > > 2.39.0.rc0.267.gcb52ba06e7-goog
> > > > >
_______________________________________________
kvmarm mailing list
kvmarm@lists.cs.columbia.edu
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
