[PATCH 2/3] drm/amd: Convert amdgpu to use suballocation helper.

Thomas Hellström (Intel) thomas_os at shipmail.org
Thu Feb 23 16:37:05 UTC 2023


On 2/23/23 17:22, Christian König wrote:
> Am 23.02.23 um 15:29 schrieb Thomas Hellström:
>>
>> On 2/23/23 12:15, Christian König wrote:
>>> Am 23.02.23 um 11:57 schrieb Thomas Hellström:
>>>> From: Maarten Lankhorst <maarten.lankhorst at linux.intel.com>
>>>>
>>>> Now that we have a generic suballocation helper, Use it in amdgpu.
>>>> For lines that get moved or changed, also fix up pre-existing style 
>>>> issues.
>>>>
>>>> Signed-off-by: Maarten Lankhorst <maarten.lankhorst at linux.intel.com>
>>>> Co-developed-by: Thomas Hellström <thomas.hellstrom at linux.intel.com>
>>>> Signed-off-by: Thomas Hellström <thomas.hellstrom at linux.intel.com>
>>>> ---
>>>>   drivers/gpu/drm/Kconfig                    |   1 +
>>>>   drivers/gpu/drm/amd/amdgpu/Kconfig         |   1 +
>>>>   drivers/gpu/drm/amd/amdgpu/amdgpu.h        |  26 +-
>>>>   drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c     |   5 +-
>>>>   drivers/gpu/drm/amd/amdgpu/amdgpu_object.h |  23 +-
>>>>   drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h   |   3 +-
>>>>   drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c     | 324 
>>>> ++-------------------
>>>>   7 files changed, 46 insertions(+), 337 deletions(-)
>>>>
>>>> diff --git a/drivers/gpu/drm/Kconfig b/drivers/gpu/drm/Kconfig
>>>> index 8fbe57407c60..73ddfdf3a894 100644
>>>> --- a/drivers/gpu/drm/Kconfig
>>>> +++ b/drivers/gpu/drm/Kconfig
>>>> @@ -77,6 +77,7 @@ config DRM_KUNIT_TEST
>>>>       select DRM_DISPLAY_HELPER
>>>>       select DRM_LIB_RANDOM
>>>>       select DRM_KMS_HELPER
>>>> +    select DRM_SUBALLOC_HELPER
>>>>       select DRM_BUDDY
>>>>       select DRM_EXPORT_FOR_TESTS if m
>>>>       select DRM_KUNIT_TEST_HELPERS
>>>
>>> This looks like it's misplaced, apart from that the patch looks good 
>>> to me.
>>
>> Looks like a TAB vs spaces issue. The resulting file looks correct. 
>> Also added the same select for Radeon in the following patch which 
>> was forgotten.
>
>
> That wasn't what I meant. This here is the patch to change amdgpu, but 
> you are adding the dependency to the KUNIT test.
>
> It looks like that adding this line should be in patch #1, not patch #2.

Ah yes, hmm. that change shouldn't be there at all. It probably strayed 
from the Kunit test that I haven't posted yet.

Will do a v4. We need to get rid of those integer divisions as well.

/Thomas

>
> Regards,
> Christian.
>
>>
>> Added your R-B to all patches, even if it wasn't exlicit for this 
>> one. Please let me know if I misunderstood that one.
>>
>> Thanks,
>>
>> Thomas
>>
>>
>>>
>>> Regards,
>>> Christian.
>>>
>>>> diff --git a/drivers/gpu/drm/amd/amdgpu/Kconfig 
>>>> b/drivers/gpu/drm/amd/amdgpu/Kconfig
>>>> index 5341b6b242c3..0ed12171450b 100644
>>>> --- a/drivers/gpu/drm/amd/amdgpu/Kconfig
>>>> +++ b/drivers/gpu/drm/amd/amdgpu/Kconfig
>>>> @@ -18,6 +18,7 @@ config DRM_AMDGPU
>>>>       select BACKLIGHT_CLASS_DEVICE
>>>>       select INTERVAL_TREE
>>>>       select DRM_BUDDY
>>>> +    select DRM_SUBALLOC_HELPER
>>>>       # amdgpu depends on ACPI_VIDEO when ACPI is enabled, for 
>>>> select to work
>>>>       # ACPI_VIDEO's dependencies must also be selected.
>>>>       select INPUT if ACPI
>>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h 
>>>> b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
>>>> index 164141bc8b4a..dda88090f044 100644
>>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
>>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
>>>> @@ -424,29 +424,11 @@ struct amdgpu_clock {
>>>>    * alignment).
>>>>    */
>>>>   -#define AMDGPU_SA_NUM_FENCE_LISTS    32
>>>> -
>>>>   struct amdgpu_sa_manager {
>>>> -    wait_queue_head_t    wq;
>>>> -    struct amdgpu_bo    *bo;
>>>> -    struct list_head    *hole;
>>>> -    struct list_head    flist[AMDGPU_SA_NUM_FENCE_LISTS];
>>>> -    struct list_head    olist;
>>>> -    unsigned        size;
>>>> -    uint64_t        gpu_addr;
>>>> -    void            *cpu_ptr;
>>>> -    uint32_t        domain;
>>>> -    uint32_t        align;
>>>> -};
>>>> -
>>>> -/* sub-allocation buffer */
>>>> -struct amdgpu_sa_bo {
>>>> -    struct list_head        olist;
>>>> -    struct list_head        flist;
>>>> -    struct amdgpu_sa_manager    *manager;
>>>> -    unsigned            soffset;
>>>> -    unsigned            eoffset;
>>>> -    struct dma_fence            *fence;
>>>> +    struct drm_suballoc_manager    base;
>>>> +    struct amdgpu_bo        *bo;
>>>> +    uint64_t            gpu_addr;
>>>> +    void                *cpu_ptr;
>>>>   };
>>>>     int amdgpu_fence_slab_init(void);
>>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c 
>>>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c
>>>> index bcccc348dbe2..df7eb0b7c4b9 100644
>>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c
>>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c
>>>> @@ -69,7 +69,7 @@ int amdgpu_ib_get(struct amdgpu_device *adev, 
>>>> struct amdgpu_vm *vm,
>>>>         if (size) {
>>>>           r = amdgpu_sa_bo_new(&adev->ib_pools[pool_type],
>>>> -                      &ib->sa_bo, size, 256);
>>>> +                     &ib->sa_bo, size);
>>>>           if (r) {
>>>>               dev_err(adev->dev, "failed to get a new IB (%d)\n", r);
>>>>               return r;
>>>> @@ -309,8 +309,7 @@ int amdgpu_ib_pool_init(struct amdgpu_device 
>>>> *adev)
>>>>         for (i = 0; i < AMDGPU_IB_POOL_MAX; i++) {
>>>>           r = amdgpu_sa_bo_manager_init(adev, &adev->ib_pools[i],
>>>> -                          AMDGPU_IB_POOL_SIZE,
>>>> -                          AMDGPU_GPU_PAGE_SIZE,
>>>> +                          AMDGPU_IB_POOL_SIZE, 256,
>>>>                             AMDGPU_GEM_DOMAIN_GTT);
>>>>           if (r)
>>>>               goto error;
>>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h 
>>>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
>>>> index 93207badf83f..5a85726ce853 100644
>>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
>>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
>>>> @@ -336,15 +336,22 @@ uint32_t 
>>>> amdgpu_bo_get_preferred_domain(struct amdgpu_device *adev,
>>>>   /*
>>>>    * sub allocation
>>>>    */
>>>> +static inline struct amdgpu_sa_manager *
>>>> +to_amdgpu_sa_manager(struct drm_suballoc_manager *manager)
>>>> +{
>>>> +    return container_of(manager, struct amdgpu_sa_manager, base);
>>>> +}
>>>>   -static inline uint64_t amdgpu_sa_bo_gpu_addr(struct amdgpu_sa_bo 
>>>> *sa_bo)
>>>> +static inline uint64_t amdgpu_sa_bo_gpu_addr(struct drm_suballoc 
>>>> *sa_bo)
>>>>   {
>>>> -    return sa_bo->manager->gpu_addr + sa_bo->soffset;
>>>> +    return to_amdgpu_sa_manager(sa_bo->manager)->gpu_addr +
>>>> +        drm_suballoc_soffset(sa_bo);
>>>>   }
>>>>   -static inline void * amdgpu_sa_bo_cpu_addr(struct amdgpu_sa_bo 
>>>> *sa_bo)
>>>> +static inline void *amdgpu_sa_bo_cpu_addr(struct drm_suballoc *sa_bo)
>>>>   {
>>>> -    return sa_bo->manager->cpu_ptr + sa_bo->soffset;
>>>> +    return to_amdgpu_sa_manager(sa_bo->manager)->cpu_ptr +
>>>> +        drm_suballoc_soffset(sa_bo);
>>>>   }
>>>>     int amdgpu_sa_bo_manager_init(struct amdgpu_device *adev,
>>>> @@ -355,11 +362,11 @@ void amdgpu_sa_bo_manager_fini(struct 
>>>> amdgpu_device *adev,
>>>>   int amdgpu_sa_bo_manager_start(struct amdgpu_device *adev,
>>>>                         struct amdgpu_sa_manager *sa_manager);
>>>>   int amdgpu_sa_bo_new(struct amdgpu_sa_manager *sa_manager,
>>>> -             struct amdgpu_sa_bo **sa_bo,
>>>> -             unsigned size, unsigned align);
>>>> +             struct drm_suballoc **sa_bo,
>>>> +             unsigned int size);
>>>>   void amdgpu_sa_bo_free(struct amdgpu_device *adev,
>>>> -                  struct amdgpu_sa_bo **sa_bo,
>>>> -                  struct dma_fence *fence);
>>>> +               struct drm_suballoc **sa_bo,
>>>> +               struct dma_fence *fence);
>>>>   #if defined(CONFIG_DEBUG_FS)
>>>>   void amdgpu_sa_bo_dump_debug_info(struct amdgpu_sa_manager 
>>>> *sa_manager,
>>>>                        struct seq_file *m);
>>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h 
>>>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h
>>>> index 3989e755a5b4..018f36b10de8 100644
>>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h
>>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h
>>>> @@ -27,6 +27,7 @@
>>>>   #include <drm/amdgpu_drm.h>
>>>>   #include <drm/gpu_scheduler.h>
>>>>   #include <drm/drm_print.h>
>>>> +#include <drm/drm_suballoc.h>
>>>>     struct amdgpu_device;
>>>>   struct amdgpu_ring;
>>>> @@ -92,7 +93,7 @@ enum amdgpu_ib_pool_type {
>>>>   };
>>>>     struct amdgpu_ib {
>>>> -    struct amdgpu_sa_bo        *sa_bo;
>>>> +    struct drm_suballoc        *sa_bo;
>>>>       uint32_t            length_dw;
>>>>       uint64_t            gpu_addr;
>>>>       uint32_t            *ptr;
>>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c 
>>>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c
>>>> index 524d10b21041..c6b4337eb20c 100644
>>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c
>>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c
>>>> @@ -44,327 +44,63 @@
>>>>     #include "amdgpu.h"
>>>>   -static void amdgpu_sa_bo_remove_locked(struct amdgpu_sa_bo *sa_bo);
>>>> -static void amdgpu_sa_bo_try_free(struct amdgpu_sa_manager 
>>>> *sa_manager);
>>>> -
>>>>   int amdgpu_sa_bo_manager_init(struct amdgpu_device *adev,
>>>>                     struct amdgpu_sa_manager *sa_manager,
>>>> -                  unsigned size, u32 align, u32 domain)
>>>> +                  unsigned int size, u32 suballoc_align, u32 domain)
>>>>   {
>>>> -    int i, r;
>>>> -
>>>> -    init_waitqueue_head(&sa_manager->wq);
>>>> -    sa_manager->bo = NULL;
>>>> -    sa_manager->size = size;
>>>> -    sa_manager->domain = domain;
>>>> -    sa_manager->align = align;
>>>> -    sa_manager->hole = &sa_manager->olist;
>>>> -    INIT_LIST_HEAD(&sa_manager->olist);
>>>> -    for (i = 0; i < AMDGPU_SA_NUM_FENCE_LISTS; ++i)
>>>> -        INIT_LIST_HEAD(&sa_manager->flist[i]);
>>>> +    int r;
>>>>   -    r = amdgpu_bo_create_kernel(adev, size, align, domain, 
>>>> &sa_manager->bo,
>>>> -                &sa_manager->gpu_addr, &sa_manager->cpu_ptr);
>>>> +    r = amdgpu_bo_create_kernel(adev, size, AMDGPU_GPU_PAGE_SIZE, 
>>>> domain,
>>>> +                    &sa_manager->bo, &sa_manager->gpu_addr,
>>>> +                    &sa_manager->cpu_ptr);
>>>>       if (r) {
>>>>           dev_err(adev->dev, "(%d) failed to allocate bo for 
>>>> manager\n", r);
>>>>           return r;
>>>>       }
>>>>   -    memset(sa_manager->cpu_ptr, 0, sa_manager->size);
>>>> +    memset(sa_manager->cpu_ptr, 0, size);
>>>> +    drm_suballoc_manager_init(&sa_manager->base, size, 
>>>> suballoc_align);
>>>>       return r;
>>>>   }
>>>>     void amdgpu_sa_bo_manager_fini(struct amdgpu_device *adev,
>>>>                      struct amdgpu_sa_manager *sa_manager)
>>>>   {
>>>> -    struct amdgpu_sa_bo *sa_bo, *tmp;
>>>> -
>>>>       if (sa_manager->bo == NULL) {
>>>>           dev_err(adev->dev, "no bo for sa manager\n");
>>>>           return;
>>>>       }
>>>>   -    if (!list_empty(&sa_manager->olist)) {
>>>> -        sa_manager->hole = &sa_manager->olist,
>>>> -        amdgpu_sa_bo_try_free(sa_manager);
>>>> -        if (!list_empty(&sa_manager->olist)) {
>>>> -            dev_err(adev->dev, "sa_manager is not empty, clearing 
>>>> anyway\n");
>>>> -        }
>>>> -    }
>>>> -    list_for_each_entry_safe(sa_bo, tmp, &sa_manager->olist, olist) {
>>>> -        amdgpu_sa_bo_remove_locked(sa_bo);
>>>> -    }
>>>> +    drm_suballoc_manager_fini(&sa_manager->base);
>>>>         amdgpu_bo_free_kernel(&sa_manager->bo, 
>>>> &sa_manager->gpu_addr, &sa_manager->cpu_ptr);
>>>> -    sa_manager->size = 0;
>>>>   }
>>>>   -static void amdgpu_sa_bo_remove_locked(struct amdgpu_sa_bo *sa_bo)
>>>> -{
>>>> -    struct amdgpu_sa_manager *sa_manager = sa_bo->manager;
>>>> -    if (sa_manager->hole == &sa_bo->olist) {
>>>> -        sa_manager->hole = sa_bo->olist.prev;
>>>> -    }
>>>> -    list_del_init(&sa_bo->olist);
>>>> -    list_del_init(&sa_bo->flist);
>>>> -    dma_fence_put(sa_bo->fence);
>>>> -    kfree(sa_bo);
>>>> -}
>>>> -
>>>> -static void amdgpu_sa_bo_try_free(struct amdgpu_sa_manager 
>>>> *sa_manager)
>>>> +int amdgpu_sa_bo_new(struct amdgpu_sa_manager *sa_manager,
>>>> +             struct drm_suballoc **sa_bo,
>>>> +             unsigned int size)
>>>>   {
>>>> -    struct amdgpu_sa_bo *sa_bo, *tmp;
>>>> +    struct drm_suballoc *sa = drm_suballoc_new(&sa_manager->base, 
>>>> size,
>>>> +                           GFP_KERNEL, true, 0);
>>>>   -    if (sa_manager->hole->next == &sa_manager->olist)
>>>> -        return;
>>>> +    if (IS_ERR(sa)) {
>>>> +        *sa_bo = NULL;
>>>>   -    sa_bo = list_entry(sa_manager->hole->next, struct 
>>>> amdgpu_sa_bo, olist);
>>>> -    list_for_each_entry_safe_from(sa_bo, tmp, &sa_manager->olist, 
>>>> olist) {
>>>> -        if (sa_bo->fence == NULL ||
>>>> -            !dma_fence_is_signaled(sa_bo->fence)) {
>>>> -            return;
>>>> -        }
>>>> -        amdgpu_sa_bo_remove_locked(sa_bo);
>>>> +        return PTR_ERR(sa);
>>>>       }
>>>> -}
>>>>   -static inline unsigned amdgpu_sa_bo_hole_soffset(struct 
>>>> amdgpu_sa_manager *sa_manager)
>>>> -{
>>>> -    struct list_head *hole = sa_manager->hole;
>>>> -
>>>> -    if (hole != &sa_manager->olist) {
>>>> -        return list_entry(hole, struct amdgpu_sa_bo, olist)->eoffset;
>>>> -    }
>>>> +    *sa_bo = sa;
>>>>       return 0;
>>>>   }
>>>>   -static inline unsigned amdgpu_sa_bo_hole_eoffset(struct 
>>>> amdgpu_sa_manager *sa_manager)
>>>> -{
>>>> -    struct list_head *hole = sa_manager->hole;
>>>> -
>>>> -    if (hole->next != &sa_manager->olist) {
>>>> -        return list_entry(hole->next, struct amdgpu_sa_bo, 
>>>> olist)->soffset;
>>>> -    }
>>>> -    return sa_manager->size;
>>>> -}
>>>> -
>>>> -static bool amdgpu_sa_bo_try_alloc(struct amdgpu_sa_manager 
>>>> *sa_manager,
>>>> -                   struct amdgpu_sa_bo *sa_bo,
>>>> -                   unsigned size, unsigned align)
>>>> -{
>>>> -    unsigned soffset, eoffset, wasted;
>>>> -
>>>> -    soffset = amdgpu_sa_bo_hole_soffset(sa_manager);
>>>> -    eoffset = amdgpu_sa_bo_hole_eoffset(sa_manager);
>>>> -    wasted = (align - (soffset % align)) % align;
>>>> -
>>>> -    if ((eoffset - soffset) >= (size + wasted)) {
>>>> -        soffset += wasted;
>>>> -
>>>> -        sa_bo->manager = sa_manager;
>>>> -        sa_bo->soffset = soffset;
>>>> -        sa_bo->eoffset = soffset + size;
>>>> -        list_add(&sa_bo->olist, sa_manager->hole);
>>>> -        INIT_LIST_HEAD(&sa_bo->flist);
>>>> -        sa_manager->hole = &sa_bo->olist;
>>>> -        return true;
>>>> -    }
>>>> -    return false;
>>>> -}
>>>> -
>>>> -/**
>>>> - * amdgpu_sa_event - Check if we can stop waiting
>>>> - *
>>>> - * @sa_manager: pointer to the sa_manager
>>>> - * @size: number of bytes we want to allocate
>>>> - * @align: alignment we need to match
>>>> - *
>>>> - * Check if either there is a fence we can wait for or
>>>> - * enough free memory to satisfy the allocation directly
>>>> - */
>>>> -static bool amdgpu_sa_event(struct amdgpu_sa_manager *sa_manager,
>>>> -                unsigned size, unsigned align)
>>>> -{
>>>> -    unsigned soffset, eoffset, wasted;
>>>> -    int i;
>>>> -
>>>> -    for (i = 0; i < AMDGPU_SA_NUM_FENCE_LISTS; ++i)
>>>> -        if (!list_empty(&sa_manager->flist[i]))
>>>> -            return true;
>>>> -
>>>> -    soffset = amdgpu_sa_bo_hole_soffset(sa_manager);
>>>> -    eoffset = amdgpu_sa_bo_hole_eoffset(sa_manager);
>>>> -    wasted = (align - (soffset % align)) % align;
>>>> -
>>>> -    if ((eoffset - soffset) >= (size + wasted)) {
>>>> -        return true;
>>>> -    }
>>>> -
>>>> -    return false;
>>>> -}
>>>> -
>>>> -static bool amdgpu_sa_bo_next_hole(struct amdgpu_sa_manager 
>>>> *sa_manager,
>>>> -                   struct dma_fence **fences,
>>>> -                   unsigned *tries)
>>>> -{
>>>> -    struct amdgpu_sa_bo *best_bo = NULL;
>>>> -    unsigned i, soffset, best, tmp;
>>>> -
>>>> -    /* if hole points to the end of the buffer */
>>>> -    if (sa_manager->hole->next == &sa_manager->olist) {
>>>> -        /* try again with its beginning */
>>>> -        sa_manager->hole = &sa_manager->olist;
>>>> -        return true;
>>>> -    }
>>>> -
>>>> -    soffset = amdgpu_sa_bo_hole_soffset(sa_manager);
>>>> -    /* to handle wrap around we add sa_manager->size */
>>>> -    best = sa_manager->size * 2;
>>>> -    /* go over all fence list and try to find the closest sa_bo
>>>> -     * of the current last
>>>> -     */
>>>> -    for (i = 0; i < AMDGPU_SA_NUM_FENCE_LISTS; ++i) {
>>>> -        struct amdgpu_sa_bo *sa_bo;
>>>> -
>>>> -        fences[i] = NULL;
>>>> -
>>>> -        if (list_empty(&sa_manager->flist[i]))
>>>> -            continue;
>>>> -
>>>> -        sa_bo = list_first_entry(&sa_manager->flist[i],
>>>> -                     struct amdgpu_sa_bo, flist);
>>>> -
>>>> -        if (!dma_fence_is_signaled(sa_bo->fence)) {
>>>> -            fences[i] = sa_bo->fence;
>>>> -            continue;
>>>> -        }
>>>> -
>>>> -        /* limit the number of tries each ring gets */
>>>> -        if (tries[i] > 2) {
>>>> -            continue;
>>>> -        }
>>>> -
>>>> -        tmp = sa_bo->soffset;
>>>> -        if (tmp < soffset) {
>>>> -            /* wrap around, pretend it's after */
>>>> -            tmp += sa_manager->size;
>>>> -        }
>>>> -        tmp -= soffset;
>>>> -        if (tmp < best) {
>>>> -            /* this sa bo is the closest one */
>>>> -            best = tmp;
>>>> -            best_bo = sa_bo;
>>>> -        }
>>>> -    }
>>>> -
>>>> -    if (best_bo) {
>>>> -        uint32_t idx = best_bo->fence->context;
>>>> -
>>>> -        idx %= AMDGPU_SA_NUM_FENCE_LISTS;
>>>> -        ++tries[idx];
>>>> -        sa_manager->hole = best_bo->olist.prev;
>>>> -
>>>> -        /* we knew that this one is signaled,
>>>> -           so it's save to remote it */
>>>> -        amdgpu_sa_bo_remove_locked(best_bo);
>>>> -        return true;
>>>> -    }
>>>> -    return false;
>>>> -}
>>>> -
>>>> -int amdgpu_sa_bo_new(struct amdgpu_sa_manager *sa_manager,
>>>> -             struct amdgpu_sa_bo **sa_bo,
>>>> -             unsigned size, unsigned align)
>>>> -{
>>>> -    struct dma_fence *fences[AMDGPU_SA_NUM_FENCE_LISTS];
>>>> -    unsigned tries[AMDGPU_SA_NUM_FENCE_LISTS];
>>>> -    unsigned count;
>>>> -    int i, r;
>>>> -    signed long t;
>>>> -
>>>> -    if (WARN_ON_ONCE(align > sa_manager->align))
>>>> -        return -EINVAL;
>>>> -
>>>> -    if (WARN_ON_ONCE(size > sa_manager->size))
>>>> -        return -EINVAL;
>>>> -
>>>> -    *sa_bo = kmalloc(sizeof(struct amdgpu_sa_bo), GFP_KERNEL);
>>>> -    if (!(*sa_bo))
>>>> -        return -ENOMEM;
>>>> -    (*sa_bo)->manager = sa_manager;
>>>> -    (*sa_bo)->fence = NULL;
>>>> -    INIT_LIST_HEAD(&(*sa_bo)->olist);
>>>> -    INIT_LIST_HEAD(&(*sa_bo)->flist);
>>>> -
>>>> -    spin_lock(&sa_manager->wq.lock);
>>>> -    do {
>>>> -        for (i = 0; i < AMDGPU_SA_NUM_FENCE_LISTS; ++i)
>>>> -            tries[i] = 0;
>>>> -
>>>> -        do {
>>>> -            amdgpu_sa_bo_try_free(sa_manager);
>>>> -
>>>> -            if (amdgpu_sa_bo_try_alloc(sa_manager, *sa_bo,
>>>> -                           size, align)) {
>>>> -                spin_unlock(&sa_manager->wq.lock);
>>>> -                return 0;
>>>> -            }
>>>> -
>>>> -            /* see if we can skip over some allocations */
>>>> -        } while (amdgpu_sa_bo_next_hole(sa_manager, fences, tries));
>>>> -
>>>> -        for (i = 0, count = 0; i < AMDGPU_SA_NUM_FENCE_LISTS; ++i)
>>>> -            if (fences[i])
>>>> -                fences[count++] = dma_fence_get(fences[i]);
>>>> -
>>>> -        if (count) {
>>>> -            spin_unlock(&sa_manager->wq.lock);
>>>> -            t = dma_fence_wait_any_timeout(fences, count, false,
>>>> -                               MAX_SCHEDULE_TIMEOUT,
>>>> -                               NULL);
>>>> -            for (i = 0; i < count; ++i)
>>>> -                dma_fence_put(fences[i]);
>>>> -
>>>> -            r = (t > 0) ? 0 : t;
>>>> -            spin_lock(&sa_manager->wq.lock);
>>>> -        } else {
>>>> -            /* if we have nothing to wait for block */
>>>> -            r = wait_event_interruptible_locked(
>>>> -                sa_manager->wq,
>>>> -                amdgpu_sa_event(sa_manager, size, align)
>>>> -            );
>>>> -        }
>>>> -
>>>> -    } while (!r);
>>>> -
>>>> -    spin_unlock(&sa_manager->wq.lock);
>>>> -    kfree(*sa_bo);
>>>> -    *sa_bo = NULL;
>>>> -    return r;
>>>> -}
>>>> -
>>>> -void amdgpu_sa_bo_free(struct amdgpu_device *adev, struct 
>>>> amdgpu_sa_bo **sa_bo,
>>>> +void amdgpu_sa_bo_free(struct amdgpu_device *adev, struct 
>>>> drm_suballoc **sa_bo,
>>>>                  struct dma_fence *fence)
>>>>   {
>>>> -    struct amdgpu_sa_manager *sa_manager;
>>>> -
>>>>       if (sa_bo == NULL || *sa_bo == NULL) {
>>>>           return;
>>>>       }
>>>>   -    sa_manager = (*sa_bo)->manager;
>>>> -    spin_lock(&sa_manager->wq.lock);
>>>> -    if (fence && !dma_fence_is_signaled(fence)) {
>>>> -        uint32_t idx;
>>>> -
>>>> -        (*sa_bo)->fence = dma_fence_get(fence);
>>>> -        idx = fence->context % AMDGPU_SA_NUM_FENCE_LISTS;
>>>> -        list_add_tail(&(*sa_bo)->flist, &sa_manager->flist[idx]);
>>>> -    } else {
>>>> -        amdgpu_sa_bo_remove_locked(*sa_bo);
>>>> -    }
>>>> -    wake_up_all_locked(&sa_manager->wq);
>>>> -    spin_unlock(&sa_manager->wq.lock);
>>>> +    drm_suballoc_free(*sa_bo, fence);
>>>>       *sa_bo = NULL;
>>>>   }
>>>>   @@ -373,26 +109,8 @@ void amdgpu_sa_bo_free(struct amdgpu_device 
>>>> *adev, struct amdgpu_sa_bo **sa_bo,
>>>>   void amdgpu_sa_bo_dump_debug_info(struct amdgpu_sa_manager 
>>>> *sa_manager,
>>>>                     struct seq_file *m)
>>>>   {
>>>> -    struct amdgpu_sa_bo *i;
>>>> -
>>>> -    spin_lock(&sa_manager->wq.lock);
>>>> -    list_for_each_entry(i, &sa_manager->olist, olist) {
>>>> -        uint64_t soffset = i->soffset + sa_manager->gpu_addr;
>>>> -        uint64_t eoffset = i->eoffset + sa_manager->gpu_addr;
>>>> -        if (&i->olist == sa_manager->hole) {
>>>> -            seq_printf(m, ">");
>>>> -        } else {
>>>> -            seq_printf(m, " ");
>>>> -        }
>>>> -        seq_printf(m, "[0x%010llx 0x%010llx] size %8lld",
>>>> -               soffset, eoffset, eoffset - soffset);
>>>> +    struct drm_printer p = drm_seq_file_printer(m);
>>>>   -        if (i->fence)
>>>> -            seq_printf(m, " protected by 0x%016llx on context %llu",
>>>> -                   i->fence->seqno, i->fence->context);
>>>> -
>>>> -        seq_printf(m, "\n");
>>>> -    }
>>>> -    spin_unlock(&sa_manager->wq.lock);
>>>> +    drm_suballoc_dump_debug_info(&sa_manager->base, &p, 
>>>> sa_manager->gpu_addr);
>>>>   }
>>>>   #endif
>>>


More information about the dri-devel mailing list