Re: [PATCH v2 03/10] target/arm: Don't mishandle count when enabling or disabling PMU counters

[Peter Maydell]

On Mon, 3 Oct 2022 at 09:55, Alex Bennée <alex.bennee@linaro.org> wrote:
>
>
> Peter Maydell <peter.maydell@linaro.org> writes:
>
> > The PMU cycle and event counter infrastructure design requires that
> > operations on the PMU register fields are wrapped in pmu_op_start()
> > and pmu_op_finish() calls (or their more specific pmmcntr and
> > pmevcntr equivalents).  This includes any changes to registers which
> > affect whether the counter should be enabled or disabled, but we
> > forgot to do this.
> >
> > The effect of this bug is that in sequences like:
> >  * disable the cycle counter (PMCCNTR) using the PMCNTEN register
> >  * write a value such as 0xfffff000 to the PMCCNTR
> >  * restart the counter by writing to PMCNTEN
> > the value written to the cycle counter is corrupted, and it starts
> > counting from the wrong place. (Essentially, we fail to record that
> > the QEMU_CLOCK_VIRTUAL timestamp when the counter should be considered
> > to have started counting is the point when PMCNTEN is written to enable
> > the counter.)
> >
> > Add the necessary bracketing calls, so that updates to the various
> > registers which affect whether the PMU is counting are handled
> > correctly.
> >
> > Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
> > Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
>
> I'm not sure why but this commit seems to be breaking a bunch of avocado
> tests for me, including the TCG plugin ones:
>
>   ➜  ./tests/venv/bin/avocado run 
> tests/avocado/tcg_plugins.py:test_aarch64_virt_insn_icount
>   JOB ID     : 0f5647d95f678e73fc01730cf9f8d7f80118443e
>   JOB LOG    : 
> /home/alex/avocado/job-results/job-2022-10-02T20.19-0f5647d/job.log
>    (1/1) 
> tests/avocado/tcg_plugins.py:PluginKernelNormal.test_aarch64_virt_insn_icount:
>  INTERRUPTED: Test interrupted by SIGTERM\nRunner error occurred: Timeout 
> reached\nOrigi
>   nal status: ERROR\n{'name': 
> '1-tests/avocado/tcg_plugins.py:PluginKernelNormal.test_aarch64_virt_insn_icount',
>  'logdir': '/home/alex/avocado/job-results/job-2022-10-02T20.19
>   -0f5647d/te... (120.43 s)
>   RESULTS    : PASS 0 | ERROR 0 | FAIL 0 | SKIP 0 | WARN 0 | INTERRUPT 1 | 
> CANCEL 0
>   JOB TIME   : 120.72 s

Known issue, fixed by
https://patchew.org/QEMU/20220930133511.2112734-1-peter.maydell@linaro.org/20220930133511.2112734-2-peter.maydell@linaro.org/
in a pending pullreq.

(I have no idea why avocado reports this as a timeout, because
what actually happens is the QEMU binary prints an error
message and exits with non-zero exit status.)

-- PMM