Hello Ketil, 

Thanks for taking a look!

On 12/09/2025 15:24, Ketil Johnsen wrote:
> On 25/07/2025 16:57, Lukas Zapolskas wrote:
>> From: Adrián Larumbe <[email protected]>
>>
>> The sampler aggregates counter and set requests coming from userspace
>> and mediates interactions with the FW interface, to ensure that user
>> sessions cannot override the global configuration.
>>
>>  From the top-level interface, the sampler supports two different types
>> of samples: clearing samples and regular samples. Clearing samples are
>> a special sample type that allow for the creation of a sampling
>> baseline, to ensure that a session does not obtain counter data from
>> before its creation.
>>
>> Upon receipt of a relevant interrupt, corresponding to one of the
>> three relevant bits of the GLB_ACK register, the sampler takes any
>> samples that occurred, and, based on the insert and extract indices,
>> accumulates them to an internal storage buffer after zero-extending
>> the counters from the 32-bit counters emitted by the hardware to
>> 64-bit counters for internal accumulation.
>>
>> When the performance counters are enabled, the FW ensures no counter
>> data is lost when entering and leaving non-counting regions by
>> producing automatic samples that do not correspond to a
>> GLB_REQ.PRFCNT_SAMPLE request. Such regions may be per hardware unit,
>> such as when a shader core powers down, or global. Most of these
>> events do not directly correspond to session sample requests, so any
>> intermediary counter data must be stored into a temporary
>> accumulation buffer.
>>
>> If there are sessions waiting for a sample, this accumulated buffer
>> will be taken, and emitted for each waiting client. During this phase,
>> information like the timestamps of sample request and sample emission,
>> type of the counter block and block index annotations are added to the
>> sample header and block headers. If no sessions are waiting for
>> a sample, this accumulation buffer is kept until the next time a
>> sample is requested.
>>
>> Special handling is needed for the PRFCNT_OVERFLOW interrupt, which is
>> an indication that the internal sample handling rate was insufficient.
>>
>> The sampler also maintains a buffer descriptor indicating the
>> structure of a firmware sample, since neither the firmware nor the
>> hardware give any indication of the sample structure, only that it
>> is composed out of three parts:
>>   - the metadata is an optional initial counter block on supporting
>>     firmware versions that contains a single counter, indicating the
>>     reason a sample was taken when entering global non-counting regions.
>>     This is used to provide coarse-grained information about why a
>>     sample was taken to userspace, to help userspace interpret
>>     variations in counter magnitude (the handling for this will be
>>     implemented in another patch series).
>>   - the firmware component of the sample is composed out of a global
>>     firmware counter block on supporting firmware versions.
>>   - the hardware component is the most sizeable of the three and
>>     contains a block of counters for each of the underlying hardware
>>     resources. It has a fixed structure that is described in the
>>     architecture specification, and contains the command stream
>>     hardware block(s), the tiler block(s), the MMU and L2 blocks
>>     (collectively named the memsys blocks) and the shader core blocks,
>>     in that order.
>> The structure of this buffer changes based on the firmware and
>> hardware combination, but is constant on a single system.
>>
>> This buffer descriptor also handles the sparseness of the shader cores,
>> wherein the physical core mask contains holes, but the memory allocated
>> for it is done based on the position of the most significant bit. In
>> cases with highly sparse core masks, this means that a lot of shader
>> counter blocks are empty, and must be skipped.
>>
>> The number of ring buffer slots is configurable through module param
>> to allow for a lower memory footprint on memory constrained systems.
>>
>> Signed-off-by: Adrián Larumbe <[email protected]>
>> Co-developed-by: Lukas Zapolskas <[email protected]>
>> Signed-off-by: Lukas Zapolskas <[email protected]>
>> ---
>>   drivers/gpu/drm/panthor/panthor_fw.c   |    6 +
>>   drivers/gpu/drm/panthor/panthor_fw.h   |    6 +-
>>   drivers/gpu/drm/panthor/panthor_perf.c | 1061 +++++++++++++++++++++++-
>>   drivers/gpu/drm/panthor/panthor_perf.h |    2 +
>>   4 files changed, 1068 insertions(+), 7 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/panthor/panthor_fw.c 
>> b/drivers/gpu/drm/panthor/panthor_fw.c
>> index 36f1034839c2..c3e1e5ed4f9c 100644
>> --- a/drivers/gpu/drm/panthor/panthor_fw.c
>> +++ b/drivers/gpu/drm/panthor/panthor_fw.c
>> @@ -22,6 +22,7 @@
>>   #include "panthor_gem.h"
>>   #include "panthor_gpu.h"
>>   #include "panthor_mmu.h"
>> +#include "panthor_perf.h"
>>   #include "panthor_regs.h"
>>   #include "panthor_sched.h"
>>   @@ -989,9 +990,12 @@ static void panthor_fw_init_global_iface(struct 
>> panthor_device *ptdev)
>>         /* Enable interrupts we care about. */
>>       glb_iface->input->ack_irq_mask = GLB_CFG_ALLOC_EN |
>> +                     GLB_PERFCNT_SAMPLE |
>>                        GLB_PING |
>>                        GLB_CFG_PROGRESS_TIMER |
>>                        GLB_CFG_POWEROFF_TIMER |
>> +                     GLB_PERFCNT_THRESHOLD |
>> +                     GLB_PERFCNT_OVERFLOW |
>>                        GLB_IDLE_EN |
>>                        GLB_IDLE;
>>   @@ -1022,6 +1026,8 @@ static void panthor_job_irq_handler(struct 
>> panthor_device *ptdev, u32 status)
>>           return;
>>         panthor_sched_report_fw_events(ptdev, status);
>> +
>> +    panthor_perf_report_irq(ptdev, status);
>>   }
>>   PANTHOR_IRQ_HANDLER(job, JOB, panthor_job_irq_handler);
>>   diff --git a/drivers/gpu/drm/panthor/panthor_fw.h 
>> b/drivers/gpu/drm/panthor/panthor_fw.h
>> index 8bcb933fa790..5a561e72e88b 100644
>> --- a/drivers/gpu/drm/panthor/panthor_fw.h
>> +++ b/drivers/gpu/drm/panthor/panthor_fw.h
>> @@ -198,6 +198,7 @@ struct panthor_fw_global_control_iface {
>>       u32 group_num;
>>       u32 group_stride;
>>   #define GLB_PERFCNT_FW_SIZE(x) ((((x) >> 16) << 8))
>> +#define GLB_PERFCNT_HW_SIZE(x) (((x) & GENMASK(15, 0)) << 8)
>>       u32 perfcnt_size;
>>       u32 instr_features;
>>   #define PERFCNT_FEATURES_MD_SIZE(x) (((x) & GENMASK(3, 0)) << 8)
>> @@ -210,7 +211,7 @@ struct panthor_fw_global_input_iface {
>>   #define GLB_CFG_ALLOC_EN            BIT(2)
>>   #define GLB_CFG_POWEROFF_TIMER            BIT(3)
>>   #define GLB_PROTM_ENTER                BIT(4)
>> -#define GLB_PERFCNT_EN                BIT(5)
>> +#define GLB_PERFCNT_ENABLE            BIT(5)
>>   #define GLB_PERFCNT_SAMPLE            BIT(6)
>>   #define GLB_COUNTER_EN                BIT(7)
>>   #define GLB_PING                BIT(8)
>> @@ -243,6 +244,9 @@ struct panthor_fw_global_input_iface {
>>       u64 perfcnt_base;
>>       u32 perfcnt_extract;
>>       u32 reserved3[3];
>> +#define GLB_PERFCNT_CONFIG_SIZE(x) ((x) & GENMASK(7, 0))
>> +#define GLB_PERFCNT_CONFIG_SET(x) (((x) & GENMASK(1, 0)) << 8)
>> +#define GLB_PERFCNT_METADATA_ENABLE BIT(10)
>>       u32 perfcnt_config;
>>       u32 perfcnt_csg_select;
>>       u32 perfcnt_fw_enable;
>> diff --git a/drivers/gpu/drm/panthor/panthor_perf.c 
>> b/drivers/gpu/drm/panthor/panthor_perf.c
>> index d1f984b3302a..fd16039d9244 100644
>> --- a/drivers/gpu/drm/panthor/panthor_perf.c
>> +++ b/drivers/gpu/drm/panthor/panthor_perf.c
>> @@ -10,6 +10,9 @@
>>     #include "panthor_device.h"
>>   #include "panthor_fw.h"
>> +#include "panthor_gem.h"
>> +#include "panthor_gpu.h"
>> +#include "panthor_mmu.h"
>>   #include "panthor_perf.h"
>>   #include "panthor_regs.h"
>>   @@ -20,6 +23,81 @@
>>    */
>>   #define PANTHOR_PERF_EM_BITS (BITS_PER_TYPE(u64) * 2)
>>   +/**
>> + * PANTHOR_CTR_TIMESTAMP_LO - The first architecturally mandated counter of 
>> every block type
>> + *                            contains the low 32-bits of the TIMESTAMP 
>> value.
>> + */
>> +#define PANTHOR_CTR_TIMESTAMP_LO (0)
>> +
>> +/**
>> + * PANTHOR_CTR_TIMESTAMP_HI - The register offset containinig the high 
>> 32-bits of the TIMESTAMP
>> + *                            value.
>> + */
>> +#define PANTHOR_CTR_TIMESTAMP_HI (1)
>> +
>> +/**
>> + * PANTHOR_CTR_PRFCNT_EN - The register offset containing the enable mask 
>> for the enabled counters
>> + *                         that were written to memory.
>> + */
>> +#define PANTHOR_CTR_PRFCNT_EN (2)
>> +
>> +/**
>> + * PANTHOR_HEADER_COUNTERS - The first four counters of every block type 
>> are architecturally
>> + *                           defined to be equivalent. The fourth counter 
>> is always reserved,
>> + *                           and should be zero and as such, does not have 
>> a separate define.
>> + *
>> + *                           These are the only four counters that are the 
>> same between different
>> + *                           blocks and are consistent between different 
>> architectures.
>> + */
>> +#define PANTHOR_HEADER_COUNTERS (4)
>> +
>> +/**
>> + * PANTHOR_CTR_SAMPLE_REASON - The metadata block has a single value in 
>> position three which
>> + *                             indicates the reason a sample was taken.
>> + */
>> +#define PANTHOR_CTR_SAMPLE_REASON (3)
>> +
>> +/**
>> + * PANTHOR_HW_COUNTER_SIZE - The size of a hardware counter in the FW ring 
>> buffer.
>> + */
>> +#define PANTHOR_HW_COUNTER_SIZE (sizeof(u32))
>> +
>> +/**
>> + * PANTHOR_PERF_RINGBUF_SLOTS_MIN - The minimum permitted number of slots 
>> in the Panthor perf
>> + *                                  ring buffer.
>> + */
>> +#define PANTHOR_PERF_RINGBUF_SLOTS_MIN (16)
>> +
>> +/**
>> + * PANTHOR_PERF_RINGBUF_SLOTS_MAX - The maximum permitted number of slots 
>> in the Panthor perf
>> + *                                  ring buffer.
>> + */
>> +#define PANTHOR_PERF_RINGBUF_SLOTS_MAX (256)
>> +
>> +static unsigned int perf_ringbuf_slots = 32;
>> +
>> +static int perf_ringbuf_slots_set(const char *val, const struct 
>> kernel_param *kp)
>> +{
>> +    unsigned int slots;
>> +    int ret = kstrtouint(val, 0, &slots);
>> +
>> +    if (ret)
>> +        return ret;
>> +
>> +    if (!is_power_of_2(slots))
>> +        return -EINVAL;
>> +
>> +    return param_set_uint_minmax(val, kp, 16, 256);
>> +}
>> +
>> +static const struct kernel_param_ops perf_ringbuf_ops = {
>> +    .set = perf_ringbuf_slots_set,
>> +    .get = param_get_uint,
>> +};
>> +module_param_cb(perf_ringbuf_slots, &perf_ringbuf_ops, &perf_ringbuf_slots, 
>> 0400);
>> +MODULE_PARM_DESC(perf_ringbuf_slots,
>> +         "Power of two slots allocated for the Panthor perf kernel-FW 
>> ringbuffer");
>> +
>>   enum panthor_perf_session_state {
>>       /** @PANTHOR_PERF_SESSION_ACTIVE: The session is active and can be 
>> used for sampling. */
>>       PANTHOR_PERF_SESSION_ACTIVE = 0,
>> @@ -65,6 +143,116 @@ enum session_sample_type {
>>       SAMPLE_TYPE_REGULAR,
>>   };
>>   +struct panthor_perf_buffer_descriptor {
>> +    /**
>> +     * @block_size: The size of a single block in the FW ring buffer, equal 
>> to
>> +     *              sizeof(u32) * counters_per_block.
>> +     */
>> +    size_t block_size;
>> +
>> +    /**
>> +     * @buffer_size: The total size of the buffer, equal to (#hardware 
>> blocks +
>> +     *               #firmware blocks) * block_size.
>> +     */
>> +    size_t buffer_size;
>> +
>> +    /**
>> +     * @available_blocks: Bitmask indicating the blocks supported by the 
>> hardware and firmware
>> +     *                    combination. Note that this can also include 
>> blocks that will not
>> +     *                    be exposed to the user.
>> +     */
>> +    DECLARE_BITMAP(available_blocks, DRM_PANTHOR_PERF_BLOCK_MAX);
>> +    struct {
>> +        /** @offset: Starting offset of a block of type @type in the FW 
>> ringbuffer. */
>> +        size_t offset;
>> +
>> +        /** @block_count: Number of blocks of the given @type, starting at 
>> @offset. */
>> +        size_t block_count;
>> +
>> +        /** @phys_mask: Bitmask of the physically available blocks. */
>> +        u64 phys_mask;
>> +    } blocks[DRM_PANTHOR_PERF_BLOCK_MAX];
>> +};
>> +
>> +/**
>> + * struct panthor_perf_sampler - Interface to de-multiplex firmware 
>> interaction and handle
>> + *                               global interactions.
>> + */
>> +struct panthor_perf_sampler {
>> +    /**
>> +     * @enabled_clients: The number of clients concurrently requesting 
>> samples. To ensure that
>> +     *                   one client cannot deny samples to another, we must 
>> ensure that clients
>> +     *                   are effectively reference counted.
>> +     */
>> +    atomic_t enabled_clients;
>> +
>> +    /**
>> +     * @sample_handled: Synchronization point between the interrupt bottom 
>> half and the
>> +     *                  main sampler interface. Must be re-armed solely on 
>> a new request
>> +     *                  coming to the sampler.
>> +     */
>> +    struct completion sample_handled;
>> +
>> +    /** @rb: Kernel BO in the FW AS containing the sample ringbuffer. */
>> +    struct panthor_kernel_bo *rb;
>> +
>> +    /**
>> +     * @sample_slots: Number of slots for samples in the FW ringbuffer. 
>> Could be static,
>> +     *          but may be useful to customize for low-memory devices.
>> +     */
>> +    size_t sample_slots;
>> +
>> +    /** @em: Combined enable mask for all of the active sessions. */
>> +    struct panthor_perf_enable_masks *em;
>> +
>> +    /**
>> +     * @desc: Buffer descriptor for a sample in the FW ringbuffer. Note 
>> that this buffer
>> +     *        at current time does some interesting things with the zeroth 
>> block type. On
>> +     *        newer FW revisions, the first counter block of the sample is 
>> the METADATA block,
>> +     *        which contains a single value indicating the reason the 
>> sample was taken (if
>> +     *        any). This block must not be exposed to userspace, as 
>> userspace does not
>> +     *        have sufficient context to interpret it. As such, this block 
>> type is not
>> +     *        added to the uAPI, but we still use it in the kernel.
>> +     */
>> +    struct panthor_perf_buffer_descriptor desc;
>> +
>> +    /**
>> +     * @sample: Pointer to an upscaled and annotated sample that may be 
>> emitted to userspace.
>> +     *          This is used both as an intermediate buffer to do the 
>> zero-extension of the
>> +     *          32-bit counters to 64-bits and as a storage buffer in case 
>> the sampler
>> +     *          requests an additional sample that was not requested by any 
>> of the top-level
>> +     *          sessions (for instance, when changing the enable masks).
>> +     */
>> +    u8 *sample;
>> +
>> +    /**
>> +     * @sampler_lock: Lock used to guard the list of sessions and the 
>> sampler configuration.
>> +     *                In particular, it guards the @session_list and the 
>> @em.
>> +     */
>> +    struct mutex sampler_lock;
>> +
>> +    /** @session_list: List of all sessions. */
>> +    struct list_head session_list;
>> +
>> +    /** @pend_lock: Lock used to guard the list of sessions with pending 
>> samples. */
>> +    spinlock_t pend_lock;
>> +
>> +    /** @pending_samples: List of sessions requesting samples. */
>> +    struct list_head pending_samples;
>> +
>> +    /** @sample_requested: A sample has been requested. */
>> +    bool sample_requested;
>> +
>> +    /** @set_config: The set that will be configured onto the hardware. */
>> +    u8 set_config;
>> +
>> +    /**
>> +     * @ptdev: Backpointer to the Panthor device, needed to ring the global 
>> doorbell and
>> +     *         interface with FW.
>> +     */
>> +    struct panthor_device *ptdev;
>> +};
>> +
>>   struct panthor_perf_session {
>>       DECLARE_BITMAP(state, PANTHOR_PERF_SESSION_MAX);
>>   @@ -186,6 +374,9 @@ struct panthor_perf {
>>        * @sessions: Global map of sessions, accessed by their ID.
>>        */
>>       struct xarray sessions;
>> +
>> +    /** @sampler: FW control interface. */
>> +    struct panthor_perf_sampler sampler;
>>   };
>>     static size_t get_annotated_block_size(size_t counters_per_block)
>> @@ -255,6 +446,23 @@ static struct panthor_perf_enable_masks 
>> *panthor_perf_create_em(struct drm_panth
>>       return em;
>>   }
>>   +static void panthor_perf_em_add(struct panthor_perf_enable_masks *const 
>> dst_em,
>> +                const struct panthor_perf_enable_masks *const src_em)
>> +{
>> +    size_t i = 0;
>> +
>> +    for (i = DRM_PANTHOR_PERF_BLOCK_FIRST; i <= 
>> DRM_PANTHOR_PERF_BLOCK_LAST; i++)
>> +        bitmap_or(dst_em->mask[i], dst_em->mask[i], src_em->mask[i], 
>> PANTHOR_PERF_EM_BITS);
>> +}
>> +
>> +static void panthor_perf_em_zero(struct panthor_perf_enable_masks *em)
>> +{
>> +    size_t i = 0;
>> +
>> +    for (i = DRM_PANTHOR_PERF_BLOCK_FIRST; i <= 
>> DRM_PANTHOR_PERF_BLOCK_LAST; i++)
>> +        bitmap_zero(em->mask[i], PANTHOR_PERF_EM_BITS);
>> +}
>> +
>>   static u64 session_read_extract_idx(struct panthor_perf_session *session)
>>   {
>>       const u64 slots = session->ringbuf_slots;
>> @@ -265,6 +473,12 @@ static u64 session_read_extract_idx(struct 
>> panthor_perf_session *session)
>>       return smp_load_acquire(&session->control->extract_idx) % slots;
>>   }
>>   +static void session_write_insert_idx(struct panthor_perf_session 
>> *session, u64 idx)
>> +{
>> +    /* Userspace needs the insert index to know where to look for the 
>> sample. */
>> +    smp_store_release(&session->control->insert_idx, idx);
>> +}
>> +
>>   static u64 session_read_insert_idx(struct panthor_perf_session *session)
>>   {
>>       const u64 slots = session->ringbuf_slots;
>> @@ -352,6 +566,758 @@ static struct panthor_perf_session 
>> *session_find(struct panthor_file *pfile,
>>       return session;
>>   }
>>   +static u32 compress_enable_mask(u32 counters, const unsigned long *const 
>> src)
>> +{
>> +    size_t i;
>> +    u32 result = 0;
>> +    const unsigned long ctr_per_bit = counters / BITS_PER_TYPE(u32);
>> +
>> +    for_each_set_bit(i, src, counters) {
>> +        const unsigned long pos = div_u64(i, ctr_per_bit);
>> +
>> +        result |= BIT(pos);
>> +    }
>> +
>> +    return result;
>> +}
>> +
>> +static void expand_enable_mask(u32 counters, u32 em, unsigned long *const 
>> dst)
>> +{
>> +    size_t i;
>> +    const unsigned long ctrs_per_bit = counters / BITS_PER_TYPE(u32);
>> +    DECLARE_BITMAP(emb, BITS_PER_TYPE(u32));
>> +
>> +    bitmap_from_arr32(emb, &em, BITS_PER_TYPE(u32));
>> +
>> +    for_each_set_bit(i, emb, BITS_PER_TYPE(u32))
>> +        bitmap_set(dst, i * ctrs_per_bit, ctrs_per_bit);
>> +}
>> +
>> +/**
>> + * panthor_perf_block_data - Identify the block index and type based on the 
>> offset.
>> + *
>> + * @desc:   FW buffer descriptor.
>> + * @offset: The current offset being examined.
>> + * @idx:    Pointer to an output index.
>> + * @type:   Pointer to an output block type.
>> + *
>> + * To disambiguate different types of blocks as well as different blocks of 
>> the same type,
>> + * the offset into the FW ringbuffer is used to uniquely identify the block 
>> being considered.
>> + *
>> + * In the future, this is a good time to identify whether a block will be 
>> empty,
>> + * allowing us to short-circuit its processing after emitting header 
>> information.
>> + *
>> + * Return: True if the current block is available, false otherwise.
>> + */
>> +static bool panthor_perf_block_data(struct panthor_perf_buffer_descriptor 
>> *const desc,
>> +                    size_t offset, u32 *const idx,
>> +                    enum drm_panthor_perf_block_type *type)
>> +{
>> +    unsigned long id;
>> +
>> +    for_each_set_bit(id, desc->available_blocks, 
>> DRM_PANTHOR_PERF_BLOCK_LAST) {
> 
> I think you mean DRM_PANTHOR_PERF_BLOCK_MAX here, otherwise bit for shader 
> block will not be checked.
> 

Thank you for spotting that! I think that may have been the issue that was 
preventing the shader core 
counters from being visible in Mesa. 

>> +        const size_t block_start = desc->blocks[id].offset;
>> +        const size_t block_count = desc->blocks[id].block_count;
>> +        const size_t block_end = desc->blocks[id].offset +
>> +            desc->block_size * block_count;
>> +
>> +        if (!block_count)
>> +            continue;
>> +
>> +        if (offset >= block_start && offset < block_end) {
>> +            const unsigned long phys_mask[] = {
>> +                BITMAP_FROM_U64(desc->blocks[id].phys_mask),
>> +            };
>> +            const size_t pos =
>> +                div_u64(offset - desc->blocks[id].offset, desc->block_size);
>> +
>> +            *type = id;
>> +
>> +            if (test_bit(pos, phys_mask)) {
>> +                const u64 mask = GENMASK_ULL(pos, 0);
>> +                const u64 zeroes = ~desc->blocks[id].phys_mask & mask;
>> +
>> +                *idx = pos - hweight64(zeroes);
>> +                return true;
>> +            }
>> +            return false;
>> +        }
>> +    }
>> +
>> +    return false;
>> +}
>> +
>> +static u32 panthor_perf_handle_sample(struct panthor_device *ptdev, u32 
>> extract_idx, u32 insert_idx)
>> +{
>> +    struct panthor_perf *perf = ptdev->perf;
>> +    struct panthor_perf_sampler *sampler = &ptdev->perf->sampler;
>> +    const size_t ann_block_size =
>> +        get_annotated_block_size(ptdev->perf_info.counters_per_block);
>> +    u32 i;
>> +
>> +    for (i = extract_idx; i != insert_idx; i++) {
>> +        u32 slot = i % sampler->sample_slots;
>> +        u8 *fw_sample = (u8 *)sampler->rb->kmap + slot * 
>> sampler->desc.buffer_size;
>> +
>> +        for (size_t fw_off = 0, ann_off = 
>> ptdev->perf_info.sample_header_size;
>> +                fw_off < sampler->desc.buffer_size;
>> +                fw_off += sampler->desc.block_size)
>> +
>> +        {
>> +            u32 idx = 0;
>> +            enum drm_panthor_perf_block_type type = 0;
>> +            DECLARE_BITMAP(expanded_em, PANTHOR_PERF_EM_BITS);
>> +            struct panthor_perf_counter_block *blk =
>> +                (typeof(blk))(perf->sampler.sample + ann_off);
>> +            u32 *const block = (u32 *)(fw_sample + fw_off);
>> +            const u32 prfcnt_en = block[PANTHOR_CTR_PRFCNT_EN];
>> +
>> +            if (!panthor_perf_block_data(&sampler->desc, fw_off, &idx, 
>> &type))
>> +                continue;
>> +
>> +            /**
>> +             * TODO Data from the metadata block must be used to populate 
>> the
>> +             * block state information.
>> +             */
>> +            if (type == DRM_PANTHOR_PERF_BLOCK_METADATA) {
>> +                /*
>> +                 * The host must clear the SAMPLE_REASON to acknowledge it 
>> has
>> +                 * consumed the sample.
>> +                 */
>> +                block[PANTHOR_CTR_SAMPLE_REASON] = 0;
>> +                continue;
>> +            }
>> +
>> +            expand_enable_mask(ptdev->perf_info.counters_per_block,
>> +                       prfcnt_en, expanded_em);
>> +
>> +            blk->header = (struct drm_panthor_perf_block_header) {
>> +                .clock = 0,
>> +                .block_idx = idx,
>> +                .block_type = type,
>> +                .block_states = DRM_PANTHOR_PERF_BLOCK_STATE_UNKNOWN
>> +            };
>> +            bitmap_to_arr64(blk->header.enable_mask, expanded_em, 
>> PANTHOR_PERF_EM_BITS);
>> +
>> +            /*
>> +             * The four header counters must be treated differently, 
>> because they are
>> +             * not additive. For the fourth, the assignment does not 
>> matter, as it
>> +             * is reserved and should be zero.
>> +             */
>> +            blk->counters[PANTHOR_CTR_TIMESTAMP_LO] = 
>> block[PANTHOR_CTR_TIMESTAMP_LO];
>> +            blk->counters[PANTHOR_CTR_TIMESTAMP_HI] = 
>> block[PANTHOR_CTR_TIMESTAMP_HI];
>> +            blk->counters[PANTHOR_CTR_PRFCNT_EN] = 
>> block[PANTHOR_CTR_PRFCNT_EN];
>> +
>> +            /*
>> +             * The host must clear PRFCNT_EN to acknowledge it has consumed 
>> the sample.
>> +             */
>> +            block[PANTHOR_CTR_PRFCNT_EN] = 0;
>> +
>> +            for (size_t k = PANTHOR_HEADER_COUNTERS;
>> +                    k < ptdev->perf_info.counters_per_block;
>> +                    k++)
>> +                blk->counters[k] += block[k];
>> +
>> +            ann_off += ann_block_size;
>> +        }
>> +    }
>> +
>> +    return i;
>> +}
>> +
>> +static size_t panthor_perf_get_fw_reported_size(struct panthor_device 
>> *ptdev)
>> +{
>> +    struct panthor_fw_global_iface *glb_iface = 
>> panthor_fw_get_glb_iface(ptdev);
>> +
>> +    size_t fw_size = GLB_PERFCNT_FW_SIZE(glb_iface->control->perfcnt_size);
>> +    size_t hw_size = GLB_PERFCNT_HW_SIZE(glb_iface->control->perfcnt_size);
>> +    size_t md_size = 
>> PERFCNT_FEATURES_MD_SIZE(glb_iface->control->perfcnt_features);
>> +
>> +    return md_size + fw_size + hw_size;
>> +}
>> +
>> +#define PANTHOR_PERF_SET_BLOCK_DESC_DATA(__desc, __type, __blk_count, 
>> __phys_mask, __offset) \
>> +    ({ \
>> +        (__desc)->blocks[(__type)].offset = (__offset); \
>> +        (__desc)->blocks[(__type)].block_count = (__blk_count);  \
>> +        (__desc)->blocks[(__type)].phys_mask = (__phys_mask);   \
>> +        if ((__blk_count))                                    \
>> +            set_bit((__type), (__desc)->available_blocks); \
>> +        (__offset) + ((__desc)->block_size) * (__blk_count); \
>> +     })
>> +
>> +static size_t get_reserved_sc_blocks(struct panthor_device *ptdev)
>> +{
>> +    const u64 sc_mask = ptdev->gpu_info.shader_present;
>> +
>> +    return fls64(sc_mask);
>> +}
>> +
>> +#define BLK_MASK(x) GENMASK_ULL((x) - 1, 0)
>> +
>> +static u64 get_sc_mask(struct panthor_device *ptdev)
>> +{
>> +    const u64 sc_mask = ptdev->gpu_info.shader_present;
>> +
>> +    return BLK_MASK(hweight64(sc_mask));
>> +}
>> +
>> +static int panthor_perf_setup_fw_buffer_desc(struct panthor_device *ptdev,
>> +                         struct panthor_perf_sampler *sampler)
>> +{
>> +    const struct drm_panthor_perf_info *const info = &ptdev->perf_info;
>> +    const size_t block_size = info->counters_per_block * 
>> PANTHOR_HW_COUNTER_SIZE;
>> +    struct panthor_perf_buffer_descriptor *desc = &sampler->desc;
>> +    const size_t fw_sample_size = panthor_perf_get_fw_reported_size(ptdev);
>> +    size_t offset = 0;
>> +
>> +    desc->block_size = block_size;
>> +
>> +    for (enum drm_panthor_perf_block_type type = 0; type < 
>> DRM_PANTHOR_PERF_BLOCK_MAX; type++) {
>> +        switch (type) {
>> +        case DRM_PANTHOR_PERF_BLOCK_METADATA:
>> +            if (info->flags & DRM_PANTHOR_PERF_BLOCK_STATES_SUPPORT)
>> +                offset = PANTHOR_PERF_SET_BLOCK_DESC_DATA(desc, type, 1,
>> +                                      BLK_MASK(1), offset);
>> +            break;
>> +        case DRM_PANTHOR_PERF_BLOCK_FW:
>> +            offset = PANTHOR_PERF_SET_BLOCK_DESC_DATA(desc, type, 
>> info->fw_blocks,
>> +                                  BLK_MASK(info->fw_blocks),
>> +                                  offset);
>> +            break;
>> +        case DRM_PANTHOR_PERF_BLOCK_CSHW:
>> +            offset = PANTHOR_PERF_SET_BLOCK_DESC_DATA(desc, type, 
>> info->cshw_blocks,
>> +                                  BLK_MASK(info->cshw_blocks),
>> +                                  offset);
>> +            break;
>> +        case DRM_PANTHOR_PERF_BLOCK_TILER:
>> +            offset = PANTHOR_PERF_SET_BLOCK_DESC_DATA(desc, type, 
>> info->tiler_blocks,
>> +                                  BLK_MASK(info->tiler_blocks),
>> +                                  offset);
>> +            break;
>> +        case DRM_PANTHOR_PERF_BLOCK_MEMSYS:
>> +            offset = PANTHOR_PERF_SET_BLOCK_DESC_DATA(desc, type, 
>> info->memsys_blocks,
>> +                                  BLK_MASK(info->memsys_blocks),
>> +                                  offset);
>> +            break;
>> +        case DRM_PANTHOR_PERF_BLOCK_SHADER:
>> +            offset = PANTHOR_PERF_SET_BLOCK_DESC_DATA(desc, type,
>> +                                  get_reserved_sc_blocks(ptdev),
>> +                                  get_sc_mask(ptdev), offset);
>> +            break;
>> +        case DRM_PANTHOR_PERF_BLOCK_MAX:
>> +            drm_WARN_ON_ONCE(&ptdev->base,
>> +                     "DRM_PANTHOR_PERF_BLOCK_MAX should be unreachable!");
>> +            break;
>> +        }
>> +    }
>> +
>> +    /* Computed size is not the same as the reported size, so we should not 
>> proceed in
>> +     * initializing the sampling session.
>> +     */
>> +    if (offset != fw_sample_size)
>> +        return -EINVAL;
>> +
>> +    desc->buffer_size = offset;
>> +
>> +    return 0;
>> +}
>> +
>> +static int panthor_perf_fw_stop_sampling(struct panthor_device *ptdev)
>> +{
>> +    struct panthor_fw_global_iface *glb_iface = 
>> panthor_fw_get_glb_iface(ptdev);
>> +    u32 acked;
>> +    int ret;
>> +
>> +    if (~READ_ONCE(glb_iface->input->req) & GLB_PERFCNT_ENABLE)
>> +        return 0;
>> +
>> +    panthor_fw_update_reqs(glb_iface, req, 0, GLB_PERFCNT_ENABLE);
>> +    gpu_write(ptdev, CSF_DOORBELL(CSF_GLB_DOORBELL_ID), 1);
>> +    ret = panthor_fw_glb_wait_acks(ptdev, GLB_PERFCNT_ENABLE, &acked, 100);
>> +    if (ret)
>> +        drm_warn(&ptdev->base, "Could not disable performance counters");
>> +
>> +    return ret;
>> +}
>> +
>> +static int panthor_perf_fw_start_sampling(struct panthor_device *ptdev)
>> +{
>> +    struct panthor_fw_global_iface *glb_iface = 
>> panthor_fw_get_glb_iface(ptdev);
>> +    u32 acked;
>> +    int ret;
>> +
>> +    if (READ_ONCE(glb_iface->input->req) & GLB_PERFCNT_ENABLE)
>> +        return 0;
>> +
>> +    /**
>> +     * The spec mandates that the host zero the PRFCNT_EXTRACT register 
>> before an enable
>> +     * operation, and each (re-)enable will require an enable-disable pair 
>> to program
>> +     * the new changes onto the FW interface.
>> +     */
>> +    WRITE_ONCE(glb_iface->input->perfcnt_extract, 0);
>> +
>> +    panthor_fw_update_reqs(glb_iface, req, GLB_PERFCNT_ENABLE, 
>> GLB_PERFCNT_ENABLE);
>> +    gpu_write(ptdev, CSF_DOORBELL(CSF_GLB_DOORBELL_ID), 1);
>> +    ret = panthor_fw_glb_wait_acks(ptdev, GLB_PERFCNT_ENABLE, &acked, 100);
>> +    if (ret)
>> +        drm_warn(&ptdev->base, "Could not enable performance counters");
>> +
>> +    return ret;
>> +}
>> +
>> +static void panthor_perf_fw_write_sampler_config(struct 
>> panthor_perf_sampler *sampler)
>> +{
>> +    struct panthor_fw_global_iface *glb_iface = 
>> panthor_fw_get_glb_iface(sampler->ptdev);
>> +    const u32 counters = sampler->ptdev->perf_info.counters_per_block;
>> +    const struct panthor_perf_enable_masks *const em = sampler->em;
>> +    u32 perfcnt_config;
>> +
>> +    glb_iface->input->perfcnt_csf_enable =
>> +        compress_enable_mask(counters, 
>> em->mask[DRM_PANTHOR_PERF_BLOCK_CSHW]);
>> +    glb_iface->input->perfcnt_shader_enable =
>> +        compress_enable_mask(counters, 
>> em->mask[DRM_PANTHOR_PERF_BLOCK_SHADER]);
>> +    glb_iface->input->perfcnt_mmu_l2_enable =
>> +        compress_enable_mask(counters, 
>> em->mask[DRM_PANTHOR_PERF_BLOCK_MEMSYS]);
>> +    glb_iface->input->perfcnt_tiler_enable =
>> +        compress_enable_mask(counters, 
>> em->mask[DRM_PANTHOR_PERF_BLOCK_TILER]);
>> +    glb_iface->input->perfcnt_fw_enable =
>> +        compress_enable_mask(counters, em->mask[DRM_PANTHOR_PERF_BLOCK_FW]);
>> +    glb_iface->input->perfcnt_csg_enable = 0;
>> +
>> +    WRITE_ONCE(glb_iface->input->perfcnt_as, 
>> panthor_vm_as(panthor_fw_vm(sampler->ptdev)));
>> +    WRITE_ONCE(glb_iface->input->perfcnt_base, 
>> panthor_kernel_bo_gpuva(sampler->rb));
>> +
>> +    perfcnt_config = GLB_PERFCNT_CONFIG_SIZE(perf_ringbuf_slots);
>> +    perfcnt_config |= GLB_PERFCNT_CONFIG_SET(sampler->set_config);
>> +    if (sampler->ptdev->perf_info.flags & 
>> DRM_PANTHOR_PERF_BLOCK_STATES_SUPPORT)
>> +        perfcnt_config |= GLB_PERFCNT_METADATA_ENABLE;
>> +
>> +    WRITE_ONCE(glb_iface->input->perfcnt_config, perfcnt_config);
>> +}
>> +
>> +static void session_populate_sample_header(struct panthor_perf_session 
>> *session,
>> +                       struct drm_panthor_perf_sample_header *hdr, u8 set)
>> +{
>> +    *hdr = (struct drm_panthor_perf_sample_header) {
>> +        .block_set = set,
>> +        .user_data = session->user_data,
>> +        .timestamp_start_ns = session->sample_start_ns,
>> +        /**
>> +         * TODO This should be changed to use the GPU clocks and the 
>> TIMESTAMP register,
>> +         * when support is added.
>> +         */
>> +        .timestamp_end_ns = ktime_get_raw_ns(),
>> +    };
>> +}
>> +
>> +/**
>> + * session_accumulate_sample - Accumulate the counters that are requested 
>> by the session
>> + *                             into the target buffer.
>> + *
>> + * @ptdev: Panthor device
>> + * @session: Perf session
>> + * @session_sample: Starting offset of the sample in the userspace mapping.
>> + * @sampler_sample: Starting offset of the sample in the sampler 
>> intermediate buffer.
>> + *
>> + * The hardware supports counter selection at the granularity of 1 bit per 
>> 4 counters, and there
>> + * is a single global FW frontend to program the counter requests from 
>> multiple sessions. This may
>> + * lead to a large disparity between the requested and provided counters 
>> for an individual client.
>> + * To remove this cross-talk, we patch out the counters that have not been 
>> requested by this
>> + * session and update the PRFCNT_EN, the header counter containing a 
>> bitmask of enabled counters,
>> + * accordingly.
>> + */
>> +static void session_accumulate_sample(struct panthor_device *ptdev,
>> +                      struct panthor_perf_session *session,
>> +                      u8 *session_sample, u8 *sampler_sample)
>> +{
>> +    const struct drm_panthor_perf_info *const perf_info = &ptdev->perf_info;
>> +    const u32 counters = perf_info->counters_per_block;
>> +    const size_t block_size = 
>> get_annotated_block_size(perf_info->counters_per_block);
>> +    const size_t sample_size = perf_info->sample_size;
>> +    const size_t sample_header_size = perf_info->sample_header_size;
>> +    const size_t data_size = sample_size - sample_header_size;
>> +    struct drm_panthor_perf_sample_header *hdr = 
>> (typeof(hdr))session_sample;
>> +
>> +    hdr->timestamp_end_ns = ktime_get_raw_ns();
>> +
>> +    session_sample += sample_header_size;
>> +    sampler_sample += sample_header_size;
>> +
>> +    for (size_t i = 0; i < data_size; i += block_size) {
>> +        DECLARE_BITMAP(enabled_ctrs, PANTHOR_PERF_EM_BITS);
>> +        struct panthor_perf_counter_block *dst_blk = 
>> (typeof(dst_blk))(session_sample + i);
>> +        struct panthor_perf_counter_block *src_blk = 
>> (typeof(src_blk))(sampler_sample + i);
>> +
>> +        bitmap_from_arr64(enabled_ctrs, dst_blk->header.enable_mask, 
>> PANTHOR_PERF_EM_BITS);
>> +        bitmap_clear(enabled_ctrs, 0, PANTHOR_HEADER_COUNTERS);
>> +
>> +        dst_blk->counters[PANTHOR_CTR_TIMESTAMP_HI] =
>> +            src_blk->counters[PANTHOR_CTR_TIMESTAMP_HI];
>> +        dst_blk->counters[PANTHOR_CTR_TIMESTAMP_LO] =
>> +            src_blk->counters[PANTHOR_CTR_TIMESTAMP_LO];
>> +
>> +        for (size_t ctr_idx = PANTHOR_HEADER_COUNTERS; ctr_idx < counters; 
>> ctr_idx++)
>> +            if (test_bit(ctr_idx, enabled_ctrs))
>> +                dst_blk->counters[ctr_idx] += src_blk->counters[ctr_idx];
>> +    }
>> +
>> +    hdr->timestamp_end_ns = ktime_get_raw_ns();
>> +    hdr->user_data = session->user_data;
>> +}
>> +
>> +static void panthor_perf_fw_request_sample(struct panthor_perf_sampler 
>> *sampler)
>> +{
>> +    struct panthor_fw_global_iface *glb_iface = 
>> panthor_fw_get_glb_iface(sampler->ptdev);
>> +
>> +    panthor_fw_toggle_reqs(glb_iface, req, ack, GLB_PERFCNT_SAMPLE);
>> +    gpu_write(sampler->ptdev, CSF_DOORBELL(CSF_GLB_DOORBELL_ID), 1);
>> +}
>> +
>> +/**
>> + * session_populate_sample - Write out a new sample into a previously 
>> populated slot in the user
>> + *                           ringbuffer and update both the header of the 
>> block and the PRFCNT_EN
>> + *                           counter to contain only the selected subset of 
>> counters for that block.
>> + *
>> + * @ptdev: Panthor device
>> + * @session: Perf session
>> + * @session_sample: Pointer aligned to the start of the data section of the 
>> sample in the targeted
>> + *                  slot.
>> + * @sampler_sample: Pointer aligned to the start of the data section of the 
>> intermediate sampler
>> + *                  buffer.
>> + *
>> + * When a new sample slot is targeted, it must be cleared of the data 
>> already existing there,
>> + * enabling a direct copy from the intermediate buffer and then zeroing out 
>> any counters
>> + * that are not required for the current session.
>> + */
>> +static void session_populate_sample(struct panthor_device *ptdev,
>> +                    struct panthor_perf_session *session, u8 
>> *session_sample,
>> +                    u8 *sampler_sample)
>> +{
>> +    const struct drm_panthor_perf_info *const perf_info = &ptdev->perf_info;
>> +
>> +    const size_t block_size = 
>> get_annotated_block_size(perf_info->counters_per_block);
>> +    const size_t sample_size = session_get_user_sample_size(perf_info);
>> +    const size_t sample_header_size = perf_info->sample_header_size;
>> +    const size_t data_size = sample_size - sample_header_size;
>> +    const u32 counters = perf_info->counters_per_block;
>> +
>> +    session_populate_sample_header(session,
>> +                       (struct drm_panthor_perf_sample_header 
>> *)session_sample,
>> +                       ptdev->perf->sampler.set_config);
>> +
>> +    session_sample += sample_header_size;
>> +
>> +    memcpy(session_sample, sampler_sample + sample_header_size, data_size);
>> +
>> +    for (size_t i = 0; i < data_size; i += block_size) {
>> +        DECLARE_BITMAP(em_diff, PANTHOR_PERF_EM_BITS);
>> +        struct panthor_perf_counter_block *blk = 
>> (typeof(blk))(session_sample + i);
>> +        enum drm_panthor_perf_block_type type = blk->header.block_type;
>> +        unsigned long *blk_em = session->enabled_counters->mask[type];
>> +
>> +        bitmap_from_arr64(em_diff, blk->header.enable_mask, 
>> PANTHOR_PERF_EM_BITS);
>> +
>> +        bitmap_andnot(em_diff, em_diff, blk_em, PANTHOR_PERF_EM_BITS);
>> +        bitmap_clear(em_diff, 0, PANTHOR_HEADER_COUNTERS);
>> +
>> +        blk->counters[PANTHOR_CTR_PRFCNT_EN] = 
>> compress_enable_mask(counters, blk_em);
>> +
>> +        for (size_t ctr_idx = PANTHOR_HEADER_COUNTERS; ctr_idx < counters; 
>> ctr_idx++) {
>> +            if (test_bit(ctr_idx, em_diff))
>> +                blk->counters[ctr_idx] = 0;
>> +        }
>> +
>> +        bitmap_to_arr64(&blk->header.enable_mask, blk_em, 
>> PANTHOR_PERF_EM_BITS);
>> +    }
>> +}
>> +
>> +static int session_copy_sample(struct panthor_device *ptdev, struct 
>> panthor_perf_session *session)
>> +{
>> +    struct panthor_perf *perf = ptdev->perf;
>> +    const size_t sample_size = 
>> session_get_user_sample_size(&ptdev->perf_info);
>> +    const u64 insert_idx = session_read_insert_idx(session);
>> +    const u64 extract_idx = session_read_extract_idx(session);
>> +    u8 *new_sample;
>> +
>> +    if (!CIRC_SPACE(insert_idx, extract_idx, session->ringbuf_slots))
>> +        return -ENOSPC;
>> +
>> +    if (READ_ONCE(session->pending_sample_request) == SAMPLE_TYPE_INITIAL)
>> +        return 0;
>> +
>> +    new_sample = session->samples + insert_idx * sample_size;
>> +
>> +    if (session->accum_idx != insert_idx) {
>> +        session_populate_sample(ptdev, session, new_sample, 
>> perf->sampler.sample);
>> +        session->accum_idx = insert_idx;
>> +    } else {
>> +        session_accumulate_sample(ptdev, session, new_sample, 
>> perf->sampler.sample);
>> +    }
>> +
>> +    return 0;
>> +}
>> +
>> +static void session_emit_sample(struct panthor_perf_session *session)
>> +{
>> +    const u64 insert_idx = session_read_insert_idx(session);
>> +    const enum session_sample_type type = 
>> READ_ONCE(session->pending_sample_request);
>> +
>> +    if (type == SAMPLE_TYPE_INITIAL || type == SAMPLE_TYPE_NONE)
>> +        goto reset_sample_request;
>> +
>> +    session_write_insert_idx(session, (insert_idx + 1) % 
>> session->ringbuf_slots);
>> +
>> +    /* Since we are about to notify userspace, we must ensure that all 
>> changes to memory
>> +     * are visible.
>> +     */
>> +    wmb();
>> +
>> +    eventfd_signal(session->eventfd);
>> +
>> +reset_sample_request:
>> +    WRITE_ONCE(session->pending_sample_request, SAMPLE_TYPE_NONE);
>> +}
>> +
>> +#define PRFCNT_IRQS (GLB_PERFCNT_OVERFLOW | GLB_PERFCNT_SAMPLE | 
>> GLB_PERFCNT_THRESHOLD)
>> +
>> +void panthor_perf_report_irq(struct panthor_device *ptdev, u32 status)
>> +{
>> +    struct panthor_perf *const perf = ptdev->perf;
>> +    struct panthor_perf_sampler *sampler;
>> +    struct panthor_fw_global_iface *glb_iface = 
>> panthor_fw_get_glb_iface(ptdev);
>> +    bool sample_requested;
>> +
>> +    if (!(status & JOB_INT_GLOBAL_IF))
>> +        return;
>> +
>> +    if (!perf)
>> +        return;
>> +
>> +    sampler = &perf->sampler;
>> +
>> +    const u32 ack = READ_ONCE(glb_iface->output->ack);
>> +    const u32 req = READ_ONCE(glb_iface->input->req);
>> +
>> +    scoped_guard(spinlock_irqsave, &sampler->pend_lock)
>> +        sample_requested = sampler->sample_requested;
>> +
>> +    /*
>> +     * TODO Fix up the error handling for overflow. Currently, the user is 
>> unblocked
>> +     * with a completely empty sample, whic is not the intended behaviour.
>> +     */
>> +    if (drm_WARN_ON_ONCE(&ptdev->base, (req ^ ack) & GLB_PERFCNT_OVERFLOW))
>> +        goto emit;
>> +
>> +    if ((sample_requested && (req & GLB_PERFCNT_SAMPLE) == (ack & 
>> GLB_PERFCNT_SAMPLE)) ||
>> +        ((req ^ ack) & GLB_PERFCNT_THRESHOLD)) {
>> +        const u32 extract_idx = 
>> READ_ONCE(glb_iface->input->perfcnt_extract);
>> +        const u32 insert_idx = READ_ONCE(glb_iface->output->perfcnt_insert);
>> +
>> +        /* If the sample was requested around a reset, some time may be 
>> needed
>> +         * for the FW interface to be updated, so we reschedule a sample
>> +         * and return immediately.
>> +         */
>> +        if (insert_idx == extract_idx) {
>> +            guard(spinlock_irqsave)(&sampler->pend_lock);
>> +            if (sampler->sample_requested)
>> +                panthor_perf_fw_request_sample(sampler);
>> +
>> +            return;
>> +        }
>> +
>> +        WRITE_ONCE(glb_iface->input->perfcnt_extract,
>> +               panthor_perf_handle_sample(ptdev, extract_idx, insert_idx));
>> +    }
>> +
>> +    scoped_guard(mutex, &sampler->sampler_lock)
>> +    {
>> +        struct list_head *pos;
>> +
>> +        list_for_each(pos, &sampler->session_list) {
>> +            struct panthor_perf_session *session = list_entry(pos,
>> +                    struct panthor_perf_session, sessions);
>> +
>> +            session_copy_sample(ptdev, session);
>> +        }
>> +    }
>> +
>> +emit:
>> +    scoped_guard(spinlock_irqsave, &sampler->pend_lock) {
>> +        struct list_head *pos, *tmp;
>> +
>> +        list_for_each_safe(pos, tmp, &sampler->pending_samples) {
>> +            struct panthor_perf_session *session = list_entry(pos,
>> +                    struct panthor_perf_session, pending);
>> +
>> +            session_emit_sample(session);
>> +            list_del(pos);
>> +            session_put(session);
>> +        }
>> +
>> +        sampler->sample_requested = false;
>> +    }
>> +
>> +    memset(sampler->sample, 0, 
>> session_get_user_sample_size(&ptdev->perf_info));
>> +    complete(&sampler->sample_handled);
>> +}
>> +
>> +static int panthor_perf_sampler_init(struct panthor_perf_sampler *sampler,
>> +                     struct panthor_device *ptdev)
>> +{
>> +    struct panthor_kernel_bo *bo;
>> +    u8 *sample;
>> +    int ret;
>> +
>> +    ret = panthor_perf_setup_fw_buffer_desc(ptdev, sampler);
>> +    if (ret) {
>> +        drm_err(&ptdev->base,
>> +            "Failed to setup descriptor for FW ring buffer, err = %d", ret);
>> +        return ret;
>> +    }
>> +
>> +    bo = panthor_kernel_bo_create(ptdev, panthor_fw_vm(ptdev),
>> +                      sampler->desc.buffer_size * perf_ringbuf_slots,
>> +                      DRM_PANTHOR_BO_NO_MMAP,
>> +                      DRM_PANTHOR_VM_BIND_OP_MAP_NOEXEC |
>> +                      DRM_PANTHOR_VM_BIND_OP_MAP_UNCACHED,
>> +                      PANTHOR_VM_KERNEL_AUTO_VA,
>> +                      "perf_ringbuf");
>> +
>> +    if (IS_ERR_OR_NULL(bo))
>> +        return IS_ERR(bo) ? PTR_ERR(bo) : -ENOMEM;
>> +
>> +    ret = panthor_kernel_bo_vmap(bo);
>> +    if (ret)
>> +        goto cleanup_bo;
>> +
>> +    sample = kzalloc(session_get_user_sample_size(&ptdev->perf_info), 
>> GFP_KERNEL);
>> +    if (ZERO_OR_NULL_PTR(sample)) {
>> +        ret = -ENOMEM;
>> +        goto cleanup_vmap;
>> +    }
>> +
>> +    sampler->rb = bo;
>> +    sampler->sample = sample;
>> +    sampler->sample_slots = perf_ringbuf_slots;
>> +    sampler->em = kzalloc(sizeof(*sampler->em), GFP_KERNEL);
>> +
>> +    mutex_init(&sampler->sampler_lock);
>> +    spin_lock_init(&sampler->pend_lock);
>> +    INIT_LIST_HEAD(&sampler->session_list);
>> +    INIT_LIST_HEAD(&sampler->pending_samples);
>> +    init_completion(&sampler->sample_handled);
>> +
>> +    sampler->ptdev = ptdev;
>> +
>> +    return 0;
>> +
>> +cleanup_vmap:
>> +    panthor_kernel_bo_vunmap(bo);
>> +
>> +cleanup_bo:
>> +    panthor_kernel_bo_destroy(bo);
>> +
>> +    return ret;
>> +}
>> +
>> +static void panthor_perf_sampler_term(struct panthor_perf_sampler *sampler)
>> +{
>> +    int ret;
>> +    bool requested;
>> +
>> +    scoped_guard(spinlock_irqsave, &sampler->pend_lock)
>> +        requested = sampler->sample_requested;
>> +
>> +    if (requested)
>> +        wait_for_completion_killable(&sampler->sample_handled);
>> +
>> +    ret = panthor_perf_fw_stop_sampling(sampler->ptdev);
>> +    if (ret)
>> +        drm_warn_once(&sampler->ptdev->base, "Sampler termination failed, 
>> ret = %d", ret);
>> +
>> +    /* At this point, the sampler enable mask is already zero, so it can be 
>> used directly. */
>> +    panthor_perf_fw_write_sampler_config(sampler);
>> +
>> +    kfree(sampler->em);
>> +    kfree(sampler->sample);
>> +
>> +    panthor_kernel_bo_destroy(sampler->rb);
>> +}
>> +
>> +static int panthor_perf_sampler_add(struct panthor_perf_sampler *sampler,
>> +                    struct panthor_perf_session *session, u8 set)
>> +{
>> +    int ret = 0;
>> +    struct panthor_perf_enable_masks *const session_em = 
>> session->enabled_counters;
>> +
>> +    guard(mutex)(&sampler->sampler_lock);
>> +
>> +    /* Early check for whether a new set can be configured. */
>> +    if (!atomic_read(&sampler->enabled_clients))
>> +        sampler->set_config = set;
>> +    else
>> +        if (sampler->set_config != set)
>> +            return -EBUSY;
>> +
>> +    panthor_perf_em_add(sampler->em, session_em);
>> +    ret = pm_runtime_resume_and_get(sampler->ptdev->base.dev);
>> +    if (ret)
>> +        return ret;
>> +
>> +    if (atomic_read(&sampler->enabled_clients)) {
>> +        ret = panthor_perf_fw_stop_sampling(sampler->ptdev);
>> +        if (ret)
>> +            return ret;
>> +    }
>> +
>> +    panthor_perf_fw_write_sampler_config(sampler);
>> +
>> +    ret = panthor_perf_fw_start_sampling(sampler->ptdev);
>> +    if (ret)
>> +        return ret;
>> +
>> +    session_get(session);
>> +    list_add_tail(&session->sessions, &sampler->session_list);
>> +    atomic_inc(&sampler->enabled_clients);
>> +
>> +    return 0;
>> +}
>> +
>> +static int panthor_perf_sampler_remove_session(struct panthor_perf_sampler 
>> *sampler,
>> +                           struct panthor_perf_session *session)
>> +{
>> +    int ret;
>> +    struct list_head *snode;
>> +
>> +    guard(mutex)(&sampler->sampler_lock);
>> +
>> +    list_del_init(&session->sessions);
>> +    session_put(session);
>> +
>> +    panthor_perf_em_zero(sampler->em);
>> +    list_for_each(snode, &sampler->session_list)
>> +    {
>> +        struct panthor_perf_session *session =
>> +            container_of(snode, typeof(*session), sessions);
>> +
>> +        panthor_perf_em_add(sampler->em, session->enabled_counters);
>> +    }
>> +
>> +    ret = panthor_perf_fw_stop_sampling(sampler->ptdev);
>> +    if (ret)
>> +        return ret;
>> +
>> +    atomic_dec(&sampler->enabled_clients);
>> +    pm_runtime_put_sync(sampler->ptdev->base.dev);
>> +
>> +    panthor_perf_fw_write_sampler_config(sampler);
>> +
>> +    if (atomic_read(&sampler->enabled_clients))
>> +        return panthor_perf_fw_start_sampling(sampler->ptdev);
>> +    return 0;
>> +}
>> +
>>   /**
>>    * panthor_perf_init - Initialize the performance counter subsystem.
>>    * @ptdev: Panthor device
>> @@ -382,6 +1348,10 @@ int panthor_perf_init(struct panthor_device *ptdev)
>>           .max = 1,
>>       };
>>   +    ret = panthor_perf_sampler_init(&perf->sampler, ptdev);
>> +    if (ret)
>> +        return ret;
>> +
>>       drm_info(&ptdev->base, "Performance counter subsystem initialized");
>>         ptdev->perf = no_free_ptr(perf);
>> @@ -389,6 +1359,70 @@ int panthor_perf_init(struct panthor_device *ptdev)
>>       return ret;
>>   }
>>   +static int sampler_request(struct panthor_perf_sampler *sampler,
>> +               struct panthor_perf_session *session, enum 
>> session_sample_type type)
>> +{
>> +    /*
>> +     * If a previous sample has not been handled yet, the session cannot 
>> request another
>> +     * sample. If this happens too often, the requested sample rate is too 
>> high.
>> +     */
>> +    if (READ_ONCE(session->pending_sample_request) != SAMPLE_TYPE_NONE)
>> +        return -EBUSY;
>> +
>> +    WRITE_ONCE(session->pending_sample_request, type);
>> +    session_get(session);
>> +
>> +    scoped_guard(spinlock_irqsave, &sampler->pend_lock) {
>> +        list_add_tail(&session->pending, &sampler->pending_samples);
>> +
>> +        if (!sampler->sample_requested) {
>> +            reinit_completion(&sampler->sample_handled);
>> +            sampler->sample_requested = true;
>> +            panthor_perf_fw_request_sample(sampler);
>> +        }
>> +    }
>> +
>> +    return 0;
>> +}
>> +
>> +/**
>> + * panthor_perf_sampler_request_initial - Request an initial sample.
>> + * @sampler: Panthor sampler
>> + * @session: Target session
>> + *
>> + * Perform a synchronous sample that gets immediately discarded. This sets 
>> a baseline at the point
>> + * of time a new session is started, to avoid having counters from before 
>> the session.
>> + */
>> +static int panthor_perf_sampler_request_initial(struct panthor_perf_sampler 
>> *sampler,
>> +                        struct panthor_perf_session *session)
>> +{
>> +    int ret = sampler_request(sampler, session, SAMPLE_TYPE_INITIAL);
>> +
>> +    if (ret)
>> +        return ret;
>> +
>> +    return wait_for_completion_timeout(&sampler->sample_handled,
>> +            msecs_to_jiffies(1000));
>> +}
>> +
>> +/**
>> + * panthor_perf_sampler_request_sample - Request a counter sample for the 
>> userspace client.
>> + * @sampler: Panthor sampler
>> + * @session: Target session
>> + *
>> + * A session that has already requested a sample cannot request another one 
>> until the previous
>> + * sample has been delivered.
>> + *
>> + * Return:
>> + * * %0       - The sample has been requested successfully.
>> + * * %-EBUSY  - The target session has already requested a sample and has 
>> not received it yet.
>> + */
>> +static int panthor_perf_sampler_request_sample(struct panthor_perf_sampler 
>> *sampler,
>> +                           struct panthor_perf_session *session)
>> +{
>> +    return sampler_request(sampler, session, SAMPLE_TYPE_REGULAR);
>> +}
>> +
>>   static int session_validate_set(u8 set)
>>   {
>>       if (set > DRM_PANTHOR_PERF_SET_TERTIARY)
>> @@ -513,6 +1547,10 @@ int panthor_perf_session_setup(struct drm_file *file, 
>> struct panthor_perf *perf,
>>       kref_init(&session->ref);
>>       session->enabled_counters = em;
>>   +    ret = panthor_perf_sampler_add(&perf->sampler, session, 
>> setup_args->block_set);
>> +    if (ret)
>> +        goto cleanup_xa_alloc;
>> +
>>       session->sample_freq_ns = setup_args->sample_freq_ns;
>>       session->user_sample_size = user_sample_size;
>>       session->ring_buf = ringbuffer;
>> @@ -523,6 +1561,9 @@ int panthor_perf_session_setup(struct drm_file *file, 
>> struct panthor_perf *perf,
>>         return session_id;
>>   +cleanup_xa_alloc:
>> +    xa_store(&perf->sessions, session_id, NULL, GFP_KERNEL);
>> +
>>   cleanup_em:
>>       kfree(em);
>>   @@ -550,6 +1591,8 @@ int panthor_perf_session_setup(struct drm_file *file, 
>> struct panthor_perf *perf,
>>   static int session_stop(struct panthor_perf *perf, struct 
>> panthor_perf_session *session,
>>               u64 user_data)
>>   {
>> +    int ret;
>> +
>>       if (!test_bit(PANTHOR_PERF_SESSION_ACTIVE, session->state))
>>           return 0;
>>   @@ -562,9 +1605,12 @@ static int session_stop(struct panthor_perf *perf, 
>> struct panthor_perf_session *
>>         session->user_data = user_data;
>>   +    ret = panthor_perf_sampler_request_sample(&perf->sampler, session);
>> +    if (ret)
>> +        return ret;
>> +
>>       clear_bit(PANTHOR_PERF_SESSION_ACTIVE, session->state);
>>   -    /* TODO Calls to the FW interface will go here in later patches. */
>>       return 0;
>>   }
>>   @@ -583,8 +1629,7 @@ static int session_start(struct panthor_perf *perf, 
>> struct panthor_perf_session
>>       if (session->sample_freq_ns)
>>           session->user_data = user_data;
>>   -    /* TODO Calls to the FW interface will go here in later patches. */
>> -    return 0;
>> +    return panthor_perf_sampler_request_initial(&perf->sampler, session);
>>   }
>>     static int session_sample(struct panthor_perf *perf, struct 
>> panthor_perf_session *session,
>> @@ -605,20 +1650,22 @@ static int session_sample(struct panthor_perf *perf, 
>> struct panthor_perf_session
>>        * the current sample, and one for a stop sample, since a stop command 
>> should
>>        * always be acknowledged by taking a final sample and stopping the 
>> session.
>>        */
>> -    if (CIRC_SPACE_TO_END(insert_idx, extract_idx, session->ringbuf_slots) 
>> < 2)
>> +    if (CIRC_SPACE(insert_idx, extract_idx, session->ringbuf_slots) < 2)
>>           return -EBUSY;
>>         session->sample_start_ns = ktime_get_raw_ns();
>>       session->user_data = user_data;
>>   -    return 0;
>> +    return panthor_perf_sampler_request_sample(&perf->sampler, session);
>>   }
>>     static int session_destroy(struct panthor_perf *perf, struct 
>> panthor_perf_session *session)
>>   {
>> +    int ret = panthor_perf_sampler_remove_session(&perf->sampler, session);
>> +
>>       session_put(session);
>>   -    return 0;
>> +    return ret;
>>   }
>>     static int session_teardown(struct panthor_perf *perf, struct 
>> panthor_perf_session *session)
>> @@ -825,6 +1872,8 @@ void panthor_perf_unplug(struct panthor_device *ptdev)
>>         xa_destroy(&perf->sessions);
>>   +    panthor_perf_sampler_term(&perf->sampler);
>> +
>>       kfree(ptdev->perf);
>>         ptdev->perf = NULL;
>> diff --git a/drivers/gpu/drm/panthor/panthor_perf.h 
>> b/drivers/gpu/drm/panthor/panthor_perf.h
>> index c46f94ed8a96..5a14854368eb 100644
>> --- a/drivers/gpu/drm/panthor/panthor_perf.h
>> +++ b/drivers/gpu/drm/panthor/panthor_perf.h
>> @@ -28,5 +28,7 @@ int panthor_perf_session_sample(struct panthor_file 
>> *pfile, struct panthor_perf
>>                   u32 sid, u64 user_data);
>>   void panthor_perf_session_destroy(struct panthor_file *pfile, struct 
>> panthor_perf *perf);
>>   +void panthor_perf_report_irq(struct panthor_device *ptdev, u32 status);
>> +
>>   #endif /* __PANTHOR_PERF_H__ */
>>   
> 
> -- 
> Cheers,
> Ketil

Reply via email to