RE: [PATCH v2 3/4] perf: arm_pmu: Add workaround for FUJITSU-MONAKA Erratum E#030002

[Tomohiro Misono (Fujitsu)]

Hi Mark,
Thanks for the reviews.

> Subject: Re: [PATCH v2 3/4] perf: arm_pmu: Add workaround for FUJITSU-MONAKA Erratum E#030002
> 
> On Fri, Oct 16, 2026 at 07:00:47PM +0900, Tomohiro Misono wrote:
> > FUJITSU-MONAKA Erratum E#030002 affects FUJITSU-MONAKA CPU.
> >
> > On affected FUJITSU-MONAKA CPUs, after a PMU event counter is
> > stopped at a value just before overflow, an overflow might still
> > be incorrectly detected, resulting interrupt storm in some case.
> >
> > For example, if the following conditions are met:
> >   1) Running perf to count an event in user space only
> >   2) The event can also occur in kernel space
> >   3) The counter value is set just before the overflow
> >      (when using the -F option, initial counter value is -1)
> >
> > when the kernel is about to return to user space after enabling
> > the count, an overflow interrupt is incorrectly triggered by an
> > event that occurred in kernel space. Since the counter value is
> > not updated (correctly), the interrupt handler does nothing and
> > returns with the same condition. This leads to another wrong interrupt
> > and results in an interrupt storm and eventually a softlockup.
> 
> I don't fully understand this.
> 
> Surely the relevant PMOVS{CLR,SET}[n] bit set as if the counter n
> overflowed?
> 
> If that bit is set, the handler will certainly do *something*.
> 
> If that bit is not set, the interrupt shouldn't be asserted at all. When
> does it get de-asserted?

Sorry, my explanation wasn't clear. The interrupt will be de-asserted as usual.
The problem is that armpmu_event_set_period() will not update the counter
value if, for example, the counter is -1. Therefore, false overflow detection will continue.
With this patch, armpmu_event_set_period() will force the counter to -6 so that
false overflow detection won't occur in the kernel, since the maximum event
count at a time is 5.

> 
> > Work around the erratum by avoiding programming PMU event counters
> > with values within five counts of overflow.
> 
> Without more detail, I don't think that can be sufficient, because you
> can end up in a scenario where:
> 
> 1. The kernel confugures the counter to -5.
> 
> 2. The kernel returns to userspace.
> 
> 3. In userspace, 4 events occur, and are counted, leaving the counter at
>    -1.
> 
> 4. Some unrelated interrupt is taken, entering the kernel.
> 
> 5. Upon return to userspace, a kernel event occurs, triggering this
>    issue.
> 
> 6. The PMU interupt handler runs, does nothing, leading right back to
>    the original issue.
> 
> What am I missing?

When the interrupt is falsely triggered, armpmu_event_set_period() will
force the counter to -6 in this case too. Then, we can return to the user space
without triggering another false overflow.

I hope this answers your questions.
Best Regards,
Tomohiro Misono

> 
> Mark.
> 
> > Signed-off-by: Tomohiro Misono <misono.tomohiro@xxxxxxxxxxx>
> >
> > ---
> > Changes in v2:
> > - Add user impact explanation in the commit message
> > ---
> > ---
> >  Documentation/arch/arm64/silicon-errata.rst |  2 ++
> >  arch/arm64/Kconfig                          | 17 +++++++++++++++++
> >  arch/arm64/include/asm/cpucaps.h            |  2 ++
> >  arch/arm64/kernel/cpu_errata.c              | 10 ++++++++++
> >  arch/arm64/tools/cpucaps                    |  1 +
> >  drivers/perf/arm_pmu.c                      | 13 +++++++++++++
> >  6 files changed, 45 insertions(+)
> >
> > diff --git a/Documentation/arch/arm64/silicon-errata.rst
> b/Documentation/arch/arm64/silicon-errata.rst
> > index 527af20b26468..0efb24ec01b4c 100644
> > --- a/Documentation/arch/arm64/silicon-errata.rst
> > +++ b/Documentation/arch/arm64/silicon-errata.rst
> > @@ -377,6 +377,8 @@ stable kernels.
> >  +----------------+-----------------+-----------------+-----------------------------+
> >  | Fujitsu        | MONAKA          | E#030001        | FUJITSU_ERRATUM_030001      |
> >  +----------------+-----------------+-----------------+-----------------------------+
> > +| Fujitsu        | MONAKA          | E#030002        | FUJITSU_ERRATUM_030002      |
> > ++----------------+-----------------+-----------------+-----------------------------+
> >  +----------------+-----------------+-----------------+-----------------------------+
> >  | ASR            | ASR8601         | #8601001        | N/A                         |
> >  +----------------+-----------------+-----------------+-----------------------------+
> > diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig
> > index 5d1a17ab0f593..43fd265156e46 100644
> > --- a/arch/arm64/Kconfig
> > +++ b/arch/arm64/Kconfig
> > @@ -1361,6 +1361,23 @@ config FUJITSU_ERRATUM_030001
> >
> >  	  If unsure, say Y.
> >
> > +config FUJITSU_ERRATUM_030002
> > +	bool "FUJITSU-MONAKA erratum E#030002: PMU overflow may be incorrectly detected"
> > +	depends on ARM_PMUV3
> > +	default y
> > +	help
> > +	  This option adds a workaround for FUJITSU-MONAKA erratum E#030002.
> > +	  On affected FUJITSU-MONAKA CPUs, after a PMU event counter is
> > +	  stopped at a value just before overflow, an overflow might still
> > +	  be incorrectly detected.
> > +
> > +	  Work around the erratum by avoiding programming PMU event counters
> > +	  with values within five counts of overflow.
> > +
> > +	  The workaround only affects the FUJITSU-MONAKA.
> > +
> > +	  If unsure, say Y.
> > +
> >  config HISILICON_ERRATUM_161600802
> >  	bool "Hip07 161600802: Erroneous redistributor VLPI base"
> >  	default y
> > diff --git a/arch/arm64/include/asm/cpucaps.h b/arch/arm64/include/asm/cpucaps.h
> > index 9220b917ec207..c0a0452daf949 100644
> > --- a/arch/arm64/include/asm/cpucaps.h
> > +++ b/arch/arm64/include/asm/cpucaps.h
> > @@ -68,6 +68,8 @@ cpucap_is_possible(const unsigned int cap)
> >  		return IS_ENABLED(CONFIG_ARM64_ERRATUM_4193714);
> >  	case ARM64_WORKAROUND_FUJITSU_030001:
> >  		return IS_ENABLED(CONFIG_FUJITSU_ERRATUM_030001);
> > +	case ARM64_WORKAROUND_FUJITSU_030002:
> > +		return IS_ENABLED(CONFIG_FUJITSU_ERRATUM_030002);
> >  	case ARM64_MPAM:
> >  		/*
> >  		 * KVM MPAM support doesn't rely on the host kernel supporting MPAM.
> > diff --git a/arch/arm64/kernel/cpu_errata.c b/arch/arm64/kernel/cpu_errata.c
> > index a11135f83e858..6363b694b4478 100644
> > --- a/arch/arm64/kernel/cpu_errata.c
> > +++ b/arch/arm64/kernel/cpu_errata.c
> > @@ -1048,6 +1048,16 @@ const struct arm64_cpu_capabilities arm64_errata[] = {
> >  					{}
> >  				})),
> >  	},
> > +#endif
> > +#ifdef CONFIG_FUJITSU_ERRATUM_030002
> > +	{
> > +		.desc = "FUJITSU-MONAKA Erratum E#030002",
> > +		.capability = ARM64_WORKAROUND_FUJITSU_030002,
> > +		ERRATA_MIDR_RANGE_LIST(((const struct midr_range[]) {
> > +					MIDR_ALL_VERSIONS(MIDR_FUJITSU_MONAKA),
> > +					{}
> > +				})),
> > +	},
> >  #endif
> >  	{
> >  		.desc = "Known broken GICv3 SEIS implementation",
> > diff --git a/arch/arm64/tools/cpucaps b/arch/arm64/tools/cpucaps
> > index 1e57a0105395b..ef8ce70ab7121 100644
> > --- a/arch/arm64/tools/cpucaps
> > +++ b/arch/arm64/tools/cpucaps
> > @@ -133,3 +133,4 @@ WORKAROUND_SPECULATIVE_SSBS
> >  WORKAROUND_SPECULATIVE_UNPRIV_LOAD
> >  WORKAROUND_BROKEN_AMU_CONSTCNT
> >  WORKAROUND_FUJITSU_030001
> > +WORKAROUND_FUJITSU_030002
> > diff --git a/drivers/perf/arm_pmu.c b/drivers/perf/arm_pmu.c
> > index 1150695653892..c0b8580d80bae 100644
> > --- a/drivers/perf/arm_pmu.c
> > +++ b/drivers/perf/arm_pmu.c
> > @@ -234,6 +234,19 @@ int armpmu_event_set_period(struct perf_event *event)
> >  	if (left > (max_period >> 1))
> >  		left = (max_period >> 1);
> >
> > +#if IS_ENABLED(CONFIG_ARM64)
> > +	if (alternative_has_cap_unlikely(ARM64_WORKAROUND_FUJITSU_030002)) {
> > +		/*
> > +		 * Due to the errata, overflow might be incorrectly detected in the handler
> > +		 * when the counter is about to overflow in a single increment, resulting
> > +		 * interrupt storm in some case. As the counter can increase by up to
> > +		 * 5 at a time, set 6 here to prevent the condition.
> > +		 */
> > +		if (left > 0 && left < 6)
> > +			left = 6;
> > +	}
> > +#endif
> > +
> >  	local64_set(&hwc->prev_count, (u64)-left);
> >
> >  	armpmu->write_counter(event, (u64)(-left) & max_period);
> >
> > --
> > 2.43.0
> >