Message-ID: <33f9822a-fbb5-47e1-ab5c-97b30511a97f@redhat.com>
Date: Tue, 11 Nov 2025 09:26:22 +1000
From: Gavin Shan <gshan@...hat.com>
To: Ben Horgan <ben.horgan@....com>, james.morse@....com
Cc: amitsinght@...vell.com, baisheng.gao@...soc.com,
 baolin.wang@...ux.alibaba.com, bobo.shaobowang@...wei.com,
 carl@...amperecomputing.com, catalin.marinas@....com, dakr@...nel.org,
 dave.martin@....com, david@...hat.com, dfustini@...libre.com,
 fenghuay@...dia.com, gregkh@...uxfoundation.org, guohanjun@...wei.com,
 jeremy.linton@....com, jonathan.cameron@...wei.com, kobak@...dia.com,
 lcherian@...vell.com, lenb@...nel.org, linux-acpi@...r.kernel.org,
 linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org,
 lpieralisi@...nel.org, peternewman@...gle.com, quic_jiles@...cinc.com,
 rafael@...nel.org, robh@...nel.org, rohit.mathew@....com,
 scott@...amperecomputing.com, sdonthineni@...dia.com, sudeep.holla@....com,
 tan.shaopeng@...itsu.com, will@...nel.org, xhao@...ux.alibaba.com,
 Shaopeng Tan <tan.shaopeng@...fujitsu.com>
Subject: Re: [PATCH 14/33] arm_mpam: Probe hardware to find the supported
 partid/pmg values

Hi Ben,

On 11/9/25 10:43 AM, Gavin Shan wrote:
> On 11/7/25 10:34 PM, Ben Horgan wrote:
>> From: James Morse <james.morse@....com>
>>
>> CPUs can generate traffic with a range of PARTID and PMG values,
>> but each MSC may also have its own maximum size for these fields.
>> Before MPAM can be used, the driver needs to probe each RIS on
>> each MSC, to find the system-wide smallest value that can be used.
>> The limits from requestors (e.g. CPUs) also need taking into account.
>>
>> While doing this, RIS entries that firmware didn't describe are created
>> under MPAM_CLASS_UNKNOWN.
>>
>> This adds the low level MSC write accessors.
>>
>> While we're here, implement the mpam_register_requestor() call
>> for the arch code to register the CPU limits. Future callers of this
>> will tell us about the SMMU and ITS.
>>
>> Signed-off-by: James Morse <james.morse@....com>
>> Reviewed-by: Jonathan Cameron <jonathan.cameron@...wei.com>
>> Reviewed-by: Ben Horgan <ben.horgan@....com>
>> Tested-by: Fenghua Yu <fenghuay@...dia.com>
>> Tested-by: Shaopeng Tan <tan.shaopeng@...fujitsu.com>
>> Tested-by: Peter Newman <peternewman@...gle.com>
>> Signed-off-by: Ben Horgan <ben.horgan@....com>
>> ---
>> Changes since v3:
>>  From Jonathan:
>> Stray comma in printk
>> Unnecessary braces
>> ---
>>   drivers/resctrl/mpam_devices.c  | 148 +++++++++++++++++++++++++++++++-
>>   drivers/resctrl/mpam_internal.h |   6 ++
>>   include/linux/arm_mpam.h        |  14 +++
>>   3 files changed, 167 insertions(+), 1 deletion(-)
>>
> 
> One comment related to 'partid_max_init', but this looks good to me in
> either way:
> 
> Reviewed-by: Gavin Shan <gshan@...hat.com>
> 
>> diff --git a/drivers/resctrl/mpam_devices.c b/drivers/resctrl/mpam_devices.c
>> index 4162a7a57626..e1e26cb350f7 100644
>> --- a/drivers/resctrl/mpam_devices.c
>> +++ b/drivers/resctrl/mpam_devices.c
>> @@ -6,6 +6,7 @@
>>   #include <linux/acpi.h>
>>   #include <linux/atomic.h>
>>   #include <linux/arm_mpam.h>
>> +#include <linux/bitfield.h>
>>   #include <linux/cacheinfo.h>
>>   #include <linux/cpu.h>
>>   #include <linux/cpumask.h>
>> @@ -42,6 +43,15 @@ static atomic_t mpam_num_msc;
>>   static int mpam_cpuhp_state;
>>   static DEFINE_MUTEX(mpam_cpuhp_state_lock);
>> +/*
>> + * The smallest common values for any CPU or MSC in the system.
>> + * Generating traffic outside this range will result in screaming interrupts.
>> + */
>> +u16 mpam_partid_max;
>> +u8 mpam_pmg_max;
>> +static bool partid_max_init, partid_max_published;
>> +static DEFINE_SPINLOCK(partid_max_lock);
>> +
> 
> If mpam_partid_max and mpam_pmg_max are initialized to USHRT_MAX and 255 here,
> the state related to partid_max_init can be removed...
> 
>>   /*
>>    * mpam is enabled once all devices have been probed from CPU online callbacks,
>>    * scheduled via this work_struct. If access to an MSC depends on a CPU that
>> @@ -142,6 +152,70 @@ static inline u32 _mpam_read_partsel_reg(struct mpam_msc *msc, u16 reg)
>>   #define mpam_read_partsel_reg(msc, reg) _mpam_read_partsel_reg(msc, MPAMF_##reg)
>> +static void __mpam_write_reg(struct mpam_msc *msc, u16 reg, u32 val)
>> +{
>> +    WARN_ON_ONCE(reg + sizeof(u32) >= msc->mapped_hwpage_sz);
>> +    WARN_ON_ONCE(!cpumask_test_cpu(smp_processor_id(), &msc->accessibility));
>> +
>> +    writel_relaxed(val, msc->mapped_hwpage + reg);
>> +}
>> +
>> +static inline void _mpam_write_partsel_reg(struct mpam_msc *msc, u16 reg, u32 val)
>> +{
>> +    lockdep_assert_held_once(&msc->part_sel_lock);
>> +    __mpam_write_reg(msc, reg, val);
>> +}
>> +
>> +#define mpam_write_partsel_reg(msc, reg, val)  _mpam_write_partsel_reg(msc, MPAMCFG_##reg, val)
>> +
>> +static u64 mpam_msc_read_idr(struct mpam_msc *msc)
>> +{
>> +    u64 idr_high = 0, idr_low;
>> +
>> +    lockdep_assert_held(&msc->part_sel_lock);
>> +
>> +    idr_low = mpam_read_partsel_reg(msc, IDR);
>> +    if (FIELD_GET(MPAMF_IDR_EXT, idr_low))
>> +        idr_high = mpam_read_partsel_reg(msc, IDR + 4);
>> +
>> +    return (idr_high << 32) | idr_low;
>> +}
>> +
>> +static void __mpam_part_sel_raw(u32 partsel, struct mpam_msc *msc)
>> +{
>> +    lockdep_assert_held(&msc->part_sel_lock);
>> +
>> +    mpam_write_partsel_reg(msc, PART_SEL, partsel);
>> +}
>> +
>> +static void __mpam_part_sel(u8 ris_idx, u16 partid, struct mpam_msc *msc)
>> +{
>> +    u32 partsel = FIELD_PREP(MPAMCFG_PART_SEL_RIS, ris_idx) |
>> +              FIELD_PREP(MPAMCFG_PART_SEL_PARTID_SEL, partid);
>> +
>> +    __mpam_part_sel_raw(partsel, msc);
>> +}
>> +
>> +int mpam_register_requestor(u16 partid_max, u8 pmg_max)
>> +{
>> +    guard(spinlock)(&partid_max_lock);
>> +    if (!partid_max_init) {
>> +        mpam_partid_max = partid_max;
>> +        mpam_pmg_max = pmg_max;
>> +        partid_max_init = true;
>> +    } else if (!partid_max_published) {
>> +        mpam_partid_max = min(mpam_partid_max, partid_max);
>> +        mpam_pmg_max = min(mpam_pmg_max, pmg_max);
>> +    } else {
>> +        /* New requestors can't lower the values */
>> +        if (partid_max < mpam_partid_max || pmg_max < mpam_pmg_max)
>> +            return -EBUSY;
>> +    }
>> +
>> +    return 0;
>> +}
>> +EXPORT_SYMBOL(mpam_register_requestor);
>> +
> 
> With mpam_partid_max and mpam_pmg_max initialized to USHRT_MAX and 255, this
> can be:
> 
>      if (!partid_max_published) {
>              ...
>          } else {
>              ...
>          }
> 
>>   static struct mpam_vmsc *
>>   mpam_vmsc_alloc(struct mpam_component *comp, struct mpam_msc *msc)
>>   {
>> @@ -451,9 +525,35 @@ int mpam_ris_create(struct mpam_msc *msc, u8 ris_idx,
>>       return err;
>>   }
>> +static struct mpam_msc_ris *mpam_get_or_create_ris(struct mpam_msc *msc,
>> +                           u8 ris_idx)
>> +{
>> +    int err;
>> +    struct mpam_msc_ris *ris;
>> +
>> +    lockdep_assert_held(&mpam_list_lock);
>> +
>> +    if (!test_bit(ris_idx, &msc->ris_idxs)) {
>> +        err = mpam_ris_create_locked(msc, ris_idx, MPAM_CLASS_UNKNOWN,
>> +                         0, 0);
>> +        if (err)
>> +            return ERR_PTR(err);
>> +    }
>> +
>> +    list_for_each_entry(ris, &msc->ris, msc_list) {
>> +        if (ris->ris_idx == ris_idx)
>> +            return ris;
>> +    }
>> +
>> +    return ERR_PTR(-ENOENT);
>> +}
>> +
>>   static int mpam_msc_hw_probe(struct mpam_msc *msc)
>>   {
>>       u64 idr;
>> +    u16 partid_max;
>> +    u8 ris_idx, pmg_max;
>> +    struct mpam_msc_ris *ris;
>>       struct device *dev = &msc->pdev->dev;
>>       lockdep_assert_held(&msc->probe_lock);
>> @@ -464,6 +564,40 @@ static int mpam_msc_hw_probe(struct mpam_msc *msc)
>>           return -EIO;
>>       }
>> +    /* Grab an IDR value to find out how many RIS there are */
>> +    mutex_lock(&msc->part_sel_lock);
>> +    idr = mpam_msc_read_idr(msc);
>> +    mutex_unlock(&msc->part_sel_lock);
>> +
>> +    msc->ris_max = FIELD_GET(MPAMF_IDR_RIS_MAX, idr);
>> +
>> +    /* Use these values so partid/pmg always starts with a valid value */
>> +    msc->partid_max = FIELD_GET(MPAMF_IDR_PARTID_MAX, idr);
>> +    msc->pmg_max = FIELD_GET(MPAMF_IDR_PMG_MAX, idr);
>> +
>> +    for (ris_idx = 0; ris_idx <= msc->ris_max; ris_idx++) {
>> +        mutex_lock(&msc->part_sel_lock);
>> +        __mpam_part_sel(ris_idx, 0, msc);
>> +        idr = mpam_msc_read_idr(msc);
>> +        mutex_unlock(&msc->part_sel_lock);
>> +
>> +        partid_max = FIELD_GET(MPAMF_IDR_PARTID_MAX, idr);
>> +        pmg_max = FIELD_GET(MPAMF_IDR_PMG_MAX, idr);
>> +        msc->partid_max = min(msc->partid_max, partid_max);
>> +        msc->pmg_max = min(msc->pmg_max, pmg_max);
>> +
>> +        mutex_lock(&mpam_list_lock);
>> +        ris = mpam_get_or_create_ris(msc, ris_idx);
>> +        mutex_unlock(&mpam_list_lock);
>> +        if (IS_ERR(ris))
>> +            return PTR_ERR(ris);
>> +    }
>> +
>> +    spin_lock(&partid_max_lock);
>> +    mpam_partid_max = min(mpam_partid_max, msc->partid_max);
>> +    mpam_pmg_max = min(mpam_pmg_max, msc->pmg_max);
>> +    spin_unlock(&partid_max_lock);
>> +

mpam_register_requestor() could be used here to avoid the capacities
(maximal PARTIDs and PMGs) are unexpectedly lowered.


>>       msc->probed = true;
>>       return 0;
>> @@ -686,10 +820,20 @@ static struct platform_driver mpam_msc_driver = {
>>   static void mpam_enable_once(void)
>>   {
>> +    /*
>> +     * Once the cpuhp callbacks have been changed, mpam_partid_max can no
>> +     * longer change.
>> +     */
>> +    spin_lock(&partid_max_lock);
>> +    partid_max_published = true;
>> +    spin_unlock(&partid_max_lock);
>> +
>>       mpam_register_cpuhp_callbacks(mpam_cpu_online, mpam_cpu_offline,
>>                         "mpam:online");
>> -    pr_info("MPAM enabled\n");
>> +    /* Use printk() to avoid the pr_fmt adding the function name. */
>> +    printk(KERN_INFO "MPAM enabled with %u PARTIDs and %u PMGs\n",
>> +           mpam_partid_max + 1, mpam_pmg_max + 1);
>>   }
>>   void mpam_disable(struct work_struct *ignored)
>> @@ -756,4 +900,6 @@ static int __init mpam_msc_driver_init(void)
>>       return platform_driver_register(&mpam_msc_driver);
>>   }
>> +
>> +/* Must occur after arm64_mpam_register_cpus() from arch_initcall() */
>>   subsys_initcall(mpam_msc_driver_init);
>> diff --git a/drivers/resctrl/mpam_internal.h b/drivers/resctrl/mpam_internal.h
>> index 4e1538d29783..768a58a3ab27 100644
>> --- a/drivers/resctrl/mpam_internal.h
>> +++ b/drivers/resctrl/mpam_internal.h
>> @@ -49,6 +49,8 @@ struct mpam_msc {
>>        */
>>       struct mutex        probe_lock;
>>       bool            probed;
>> +    u16            partid_max;
>> +    u8            pmg_max;
>>       unsigned long        ris_idxs;
>>       u32            ris_max;
>> @@ -138,6 +140,10 @@ struct mpam_msc_ris {
>>   extern struct srcu_struct mpam_srcu;
>>   extern struct list_head mpam_classes;
>> +/* System wide partid/pmg values */
>> +extern u16 mpam_partid_max;
>> +extern u8 mpam_pmg_max;
>> +
>>   /* Scheduled work callback to enable mpam once all MSC have been probed */
>>   void mpam_enable(struct work_struct *work);
>>   void mpam_disable(struct work_struct *work);
>> diff --git a/include/linux/arm_mpam.h b/include/linux/arm_mpam.h
>> index 5a3aab6bb1d4..c7246cfeb091 100644
>> --- a/include/linux/arm_mpam.h
>> +++ b/include/linux/arm_mpam.h
>> @@ -49,4 +49,18 @@ static inline int mpam_ris_create(struct mpam_msc *msc, u8 ris_idx,
>>   }
>>   #endif
>> +/**
>> + * mpam_register_requestor() - Register a requestor with the MPAM driver
>> + * @partid_max:        The maximum PARTID value the requestor can generate.
>> + * @pmg_max:        The maximum PMG value the requestor can generate.
>> + *
>> + * Registers a requestor with the MPAM driver to ensure the chosen system-wide
>> + * minimum PARTID and PMG values will allow the requestors features to be used.
>> + *
>> + * Returns an error if the registration is too late, and a larger PARTID/PMG
>> + * value has been advertised to user-space. In this case the requestor should
>> + * not use its MPAM features. Returns 0 on success.
>> + */
>> +int mpam_register_requestor(u16 partid_max, u8 pmg_max);
>> +
>>   #endif /* __LINUX_ARM_MPAM_H */

Thanks,
Gavin

Powered by blists - more mailing lists