[PATCH v3 07/15] platform: andes: Add Andes custom PMU support

[Yu-Chien Peter Lin <peterlin@andestech.com>]

Hi Samuel,

Thanks for the review.

On Wed, Nov 22, 2023 at 08:24:49PM -0600, Samuel Holland wrote:
> On 2023-11-22 1:36 AM, Yu Chien Peter Lin wrote:
> > Before the ratification of Sscofpmf, the Andes PMU extension
> > was designed to support the sampling and filtering with hardware
> > performance counters (zihpm), it works with the current SBI PMU
> > extension and Linux SBI PMU driver.
> > 
> > We implement 1) the PMU device callbacks that update the
> > corresponding bits on custom CSRs, 2) extentions_init() to detect
> > the hardware support of Andes PMU and initialize the per-hart
> > PMU related CSR, and 3) pmu_init() to register PMU device and
> > populate event mappings (only called by coldboot hart).
> > 
> > Signed-off-by: Yu Chien Peter Lin <peterlin@andestech.com>
> > Reviewed-by: Leo Yu-Chi Liang <ycliang@andestech.com>
> > ---
> > Changes v1 -> v2:
> >   - Fix mode filtering in andes_hw_counter_filter_mode()
> >   - Return early if pmu is not supported in andes_pmu_init() (suggested by Prabhakar)
> >   - Don't grant write permissions via CSR_MCOUNTERWEN as not needed
> > Changes v2 -> v3:
> >   - Drop Anup's RB tag as we add andes_pmu_extensions_init() to initialize per-hart
> >     extension on scratch and move andes_pmu_init() to new added platform override pmu_init()
> > --- 
> >  platform/generic/andes/Kconfig             |   8 ++
> >  platform/generic/andes/andes_pmu.c         | 102 +++++++++++++++++++++
> >  platform/generic/andes/objects.mk          |   1 +
> >  platform/generic/include/andes/andes_pmu.h |  33 +++++++
> >  4 files changed, 144 insertions(+)
> >  create mode 100644 platform/generic/andes/andes_pmu.c
> >  create mode 100644 platform/generic/include/andes/andes_pmu.h
> > 
> > diff --git a/platform/generic/andes/Kconfig b/platform/generic/andes/Kconfig
> > index a91fb9c..5b2ed91 100644
> > --- a/platform/generic/andes/Kconfig
> > +++ b/platform/generic/andes/Kconfig
> > @@ -7,3 +7,11 @@ config ANDES45_PMA
> >  config ANDES_SBI
> >  	bool "Andes SBI support"
> >  	default n
> > +
> > +config ANDES_PMU
> > +	bool "Andes PMU extension (xandespmu) support"
> > +	default n
> > +	help
> > +	  Andes PMU extension supports the event counter overflow
> > +	  interrupt and mode filtering, similar to the standard
> > +	  Sscofpmf and Smcntrpmf.
> > diff --git a/platform/generic/andes/andes_pmu.c b/platform/generic/andes/andes_pmu.c
> > new file mode 100644
> > index 0000000..72003fd
> > --- /dev/null
> > +++ b/platform/generic/andes/andes_pmu.c
> > @@ -0,0 +1,102 @@
> > +// SPDX-License-Identifier: BSD-2-Clause
> > +/*
> > + * andes_pmu.c - Andes PMU device callbacks and platform overrides
> > + *
> > + * Copyright (C) 2023 Andes Technology Corporation
> > + */
> > +
> > +#include <andes/andes45.h>
> > +#include <andes/andes_hpm.h>
> > +#include <andes/andes_pmu.h>
> > +#include <sbi/sbi_bitops.h>
> > +#include <sbi/sbi_error.h>
> > +#include <sbi/sbi_pmu.h>
> > +
> > +static void andes_hw_counter_enable_irq(uint32_t ctr_idx)
> > +{
> > +	unsigned long mip_val;
> > +
> > +	if (ctr_idx >= SBI_PMU_HW_CTR_MAX)
> > +		return;
> > +
> > +	mip_val = csr_read(CSR_MIP);
> > +	if (!(mip_val & MIP_PMOVI))
> > +		csr_clear(CSR_MCOUNTEROVF, BIT(ctr_idx));
> > +
> > +	csr_set(CSR_MCOUNTERINTEN, BIT(ctr_idx));
> > +}
> > +
> > +static void andes_hw_counter_disable_irq(uint32_t ctr_idx)
> > +{
> > +	csr_clear(CSR_MCOUNTERINTEN, BIT(ctr_idx));
> > +}
> > +
> > +static void andes_hw_counter_filter_mode(unsigned long flags, int ctr_idx)
> > +{
> > +	if (flags & SBI_PMU_CFG_FLAG_SET_UINH)
> > +		csr_set(CSR_MCOUNTERMASK_U, BIT(ctr_idx));
> > +	else
> > +		csr_clear(CSR_MCOUNTERMASK_U, BIT(ctr_idx));
> > +
> > +	if (flags & SBI_PMU_CFG_FLAG_SET_SINH)
> > +		csr_set(CSR_MCOUNTERMASK_S, BIT(ctr_idx));
> > +	else
> > +		csr_clear(CSR_MCOUNTERMASK_S, BIT(ctr_idx));
> > +}
> > +
> > +static struct sbi_pmu_device andes_pmu = {
> > +	.name = "andes_pmu",
> > +	.hw_counter_enable_irq  = andes_hw_counter_enable_irq,
> > +	.hw_counter_disable_irq = andes_hw_counter_disable_irq,
> > +	/*
> > +	 * We set delegation of supervisor local interrupts via
> > +	 * 18th bit on mslideleg instead of mideleg, so leave
> > +	 * hw_counter_irq_bit() callback unimplemented.
> > +	 */
> > +	.hw_counter_irq_bit     = NULL,
> > +	.hw_counter_filter_mode = andes_hw_counter_filter_mode
> > +};
> > +
> > +int andes_pmu_extensions_init(const struct fdt_match *match,
> > +			      struct sbi_hart_features *hfeatures)
> > +{
> > +	struct sbi_scratch *scratch = sbi_scratch_thishart_ptr();
> > +
> > +	if (!has_andes_pmu())
> > +		return 0;
> > +
> > +	/*
> > +	 * Don't expect both Andes PMU and standard Sscofpmf/Smcntrpmf,
> > +	 * are supported as they serve the same purpose.
> > +	 */
> > +	if (sbi_hart_has_extension(scratch, SBI_HART_EXT_SSCOFPMF) ||
> > +		sbi_hart_has_extension(scratch, SBI_HART_EXT_SMCNTRPMF))
> > +		return SBI_EINVAL;
> 
> If Sscofpmf is supported, don't you want to successfully do nothing?

If Sscofpmf is supported, it will return early as has_andes_pmu() above
is false.

> > +	sbi_hart_update_extension(scratch, SBI_HART_EXT_XANDESPMU, true);
> > +
> > +	/* Inhibit all HPM counters in M-mode */
> > +	csr_write(CSR_MCOUNTERMASK_M, 0xfffffffd);
> > +	/* Delegate counter overflow interrupt to S-mode */
> > +	csr_write(CSR_MSLIDELEG, MIP_PMOVI);
> > +
> > +	return 0;
> > +}
> > +
> > +int andes_pmu_init(const struct fdt_match *match)
> > +{
> > +	int rc;
> > +	struct sbi_scratch *scratch = sbi_scratch_thishart_ptr();
> > +
> > +	if (sbi_hart_has_extension(scratch, SBI_HART_EXT_XANDESPMU))
> > +		sbi_pmu_set_device(&andes_pmu);
> > +
> > +	rc = fdt_pmu_setup(fdt_get_address());
> 
> generic_pmu_init() already calls fdt_pmu_setup().

OK will remove fdt_pmu_setup() here.

> > +	/*
> > +	 * Populate default mappings if device-tree doesn't
> > +	 * provide a valid pmu node.
> > +	 */
> > +	if (rc == SBI_ENOENT)
> > +		return andes_pmu_setup();
> 
> This function isn't added until patch 14, so this won't compile. Please make
> sure each patch compiles individually.

Sure, thank you for catching this.

Regards,
Peter Lin

> Regards,
> Samuel
>