[PATCH 1/3] drm/amdgpu: use HMM mirror callback to replace mmu notifier v6

Christian König ckoenig.leichtzumerken at gmail.com
Mon Feb 4 18:09:17 UTC 2019


Am 04.02.19 um 18:17 schrieb Yang, Philip:
>
> On 2019-02-04 10:18 a.m., Christian König wrote:
>> Am 04.02.19 um 16:06 schrieb Yang, Philip:
>>> Replace our MMU notifier with hmm_mirror_ops.sync_cpu_device_pagetables
>>> callback. Enable CONFIG_HMM and CONFIG_HMM_MIRROR as a dependency in
>>> DRM_AMDGPU_USERPTR Kconfig.
>>>
>>> It supports both KFD userptr and gfx userptr paths.
>>>
>>> The depdent HMM patchset from Jérôme Glisse are all merged into 4.20.0
>>> kernel now.
>>>
>>> Change-Id: Ie62c3c5e3c5b8521ab3b438d1eff2aa2a003835e
>>> Signed-off-by: Philip Yang <Philip.Yang at amd.com>
>>> ---
>>>    drivers/gpu/drm/amd/amdgpu/Kconfig     |   6 +-
>>>    drivers/gpu/drm/amd/amdgpu/Makefile    |   2 +-
>>>    drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c | 139 +++++++++++--------------
>>>    drivers/gpu/drm/amd/amdgpu/amdgpu_mn.h |   2 +-
>>>    4 files changed, 67 insertions(+), 82 deletions(-)
>>>
>>> diff --git a/drivers/gpu/drm/amd/amdgpu/Kconfig
>>> b/drivers/gpu/drm/amd/amdgpu/Kconfig
>>> index 9221e5489069..960a63355705 100644
>>> --- a/drivers/gpu/drm/amd/amdgpu/Kconfig
>>> +++ b/drivers/gpu/drm/amd/amdgpu/Kconfig
>>> @@ -26,10 +26,10 @@ config DRM_AMDGPU_CIK
>>>    config DRM_AMDGPU_USERPTR
>>>        bool "Always enable userptr write support"
>>>        depends on DRM_AMDGPU
>>> -    select MMU_NOTIFIER
>>> +    select HMM_MIRROR
>>>        help
>>> -      This option selects CONFIG_MMU_NOTIFIER if it isn't already
>>> -      selected to enabled full userptr support.
>>> +      This option selects CONFIG_HMM and CONFIG_HMM_MIRROR if it
>>> +      isn't already selected to enabled full userptr support.
>>>    config DRM_AMDGPU_GART_DEBUGFS
>>>        bool "Allow GART access through debugfs"
>>> diff --git a/drivers/gpu/drm/amd/amdgpu/Makefile
>>> b/drivers/gpu/drm/amd/amdgpu/Makefile
>>> index 466da5954a68..851001ced5e8 100644
>>> --- a/drivers/gpu/drm/amd/amdgpu/Makefile
>>> +++ b/drivers/gpu/drm/amd/amdgpu/Makefile
>>> @@ -172,7 +172,7 @@ endif
>>>    amdgpu-$(CONFIG_COMPAT) += amdgpu_ioc32.o
>>>    amdgpu-$(CONFIG_VGA_SWITCHEROO) += amdgpu_atpx_handler.o
>>>    amdgpu-$(CONFIG_ACPI) += amdgpu_acpi.o
>>> -amdgpu-$(CONFIG_MMU_NOTIFIER) += amdgpu_mn.o
>>> +amdgpu-$(CONFIG_HMM_MIRROR) += amdgpu_mn.o
>>>    include $(FULL_AMD_PATH)/powerplay/Makefile
>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c
>>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c
>>> index 3e6823fdd939..5d518d2bb9be 100644
>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c
>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c
>>> @@ -45,7 +45,7 @@
>>>    #include <linux/firmware.h>
>>>    #include <linux/module.h>
>>> -#include <linux/mmu_notifier.h>
>>> +#include <linux/hmm.h>
>>>    #include <linux/interval_tree.h>
>>>    #include <drm/drmP.h>
>>>    #include <drm/drm.h>
>>> @@ -58,7 +58,6 @@
>>>     *
>>>     * @adev: amdgpu device pointer
>>>     * @mm: process address space
>>> - * @mn: MMU notifier structure
>>>     * @type: type of MMU notifier
>>>     * @work: destruction work item
>>>     * @node: hash table node to find structure by adev and mn
>>> @@ -66,6 +65,7 @@
>>>     * @objects: interval tree containing amdgpu_mn_nodes
>>>     * @read_lock: mutex for recursive locking of @lock
>>>     * @recursion: depth of recursion
>>> + * @mirror: HMM mirror function support
>>>     *
>>>     * Data for each amdgpu device and process address space.
>>>     */
>>> @@ -73,7 +73,6 @@ struct amdgpu_mn {
>>>        /* constant after initialisation */
>>>        struct amdgpu_device    *adev;
>>>        struct mm_struct    *mm;
>>> -    struct mmu_notifier    mn;
>>>        enum amdgpu_mn_type    type;
>>>        /* only used on destruction */
>>> @@ -87,6 +86,9 @@ struct amdgpu_mn {
>>>        struct rb_root_cached    objects;
>>>        struct mutex        read_lock;
>>>        atomic_t        recursion;
>> With HMM we don't need this any more. Please remove it and simplify
>> amdgpu_mn_read_lock() and amdgpu_mn_read_unlock().
>>
> Thanks, this makes sense because HMM uses hmm->mirror_sem to serialize
> invalidate range.
>
> amn->read_lock is also not needed anymore, this was used to protect
> atomic operations for atomic_inc_return(amn->recursion) and
> down_read(amn->lock).
>
> Just one amn->lock is needed to sync amdgpu_cs_submit and userptr
> update. I will submit another patch.

Sounds good. I will take a closer look at the other patches tomorrow.

Thanks,
Christian.

>
> Philip
>
>> Apart from that looks good to me,
>> Christian.
>>
>>> +
>>> +    /* HMM mirror */
>>> +    struct hmm_mirror    mirror;
>>>    };
>>>    /**
>>> @@ -103,7 +105,7 @@ struct amdgpu_mn_node {
>>>    };
>>>    /**
>>> - * amdgpu_mn_destroy - destroy the MMU notifier
>>> + * amdgpu_mn_destroy - destroy the HMM mirror
>>>     *
>>>     * @work: previously sheduled work item
>>>     *
>>> @@ -129,28 +131,26 @@ static void amdgpu_mn_destroy(struct work_struct
>>> *work)
>>>        }
>>>        up_write(&amn->lock);
>>>        mutex_unlock(&adev->mn_lock);
>>> -    mmu_notifier_unregister_no_release(&amn->mn, amn->mm);
>>> +
>>> +    hmm_mirror_unregister(&amn->mirror);
>>>        kfree(amn);
>>>    }
>>>    /**
>>> - * amdgpu_mn_release - callback to notify about mm destruction
>>> + * amdgpu_hmm_mirror_release - callback to notify about mm destruction
>>>     *
>>> - * @mn: our notifier
>>> - * @mm: the mm this callback is about
>>> + * @mirror: the HMM mirror (mm) this callback is about
>>>     *
>>> - * Shedule a work item to lazy destroy our notifier.
>>> + * Shedule a work item to lazy destroy HMM mirror.
>>>     */
>>> -static void amdgpu_mn_release(struct mmu_notifier *mn,
>>> -                  struct mm_struct *mm)
>>> +static void amdgpu_hmm_mirror_release(struct hmm_mirror *mirror)
>>>    {
>>> -    struct amdgpu_mn *amn = container_of(mn, struct amdgpu_mn, mn);
>>> +    struct amdgpu_mn *amn = container_of(mirror, struct amdgpu_mn,
>>> mirror);
>>>        INIT_WORK(&amn->work, amdgpu_mn_destroy);
>>>        schedule_work(&amn->work);
>>>    }
>>> -
>>>    /**
>>>     * amdgpu_mn_lock - take the write side lock for this notifier
>>>     *
>>> @@ -237,141 +237,126 @@ static void amdgpu_mn_invalidate_node(struct
>>> amdgpu_mn_node *node,
>>>    /**
>>>     * amdgpu_mn_invalidate_range_start_gfx - callback to notify about
>>> mm change
>>>     *
>>> - * @mn: our notifier
>>> - * @range: mmu notifier context
>>> + * @mirror: the hmm_mirror (mm) is about to update
>>> + * @update: the update start, end address
>>>     *
>>>     * Block for operations on BOs to finish and mark pages as accessed and
>>>     * potentially dirty.
>>>     */
>>> -static int amdgpu_mn_invalidate_range_start_gfx(struct mmu_notifier *mn,
>>> -            const struct mmu_notifier_range *range)
>>> +static int amdgpu_mn_sync_pagetables_gfx(struct hmm_mirror *mirror,
>>> +            const struct hmm_update *update)
>>>    {
>>> -    struct amdgpu_mn *amn = container_of(mn, struct amdgpu_mn, mn);
>>> +    struct amdgpu_mn *amn = container_of(mirror, struct amdgpu_mn,
>>> mirror);
>>> +    unsigned long start = update->start;
>>> +    unsigned long end = update->end;
>>> +    bool blockable = update->blockable;
>>>        struct interval_tree_node *it;
>>> -    unsigned long end;
>>>        /* notification is exclusive, but interval is inclusive */
>>> -    end = range->end - 1;
>>> +    end -= 1;
>>>        /* TODO we should be able to split locking for interval tree and
>>>         * amdgpu_mn_invalidate_node
>>>         */
>>> -    if (amdgpu_mn_read_lock(amn, range->blockable))
>>> +    if (amdgpu_mn_read_lock(amn, blockable))
>>>            return -EAGAIN;
>>> -    it = interval_tree_iter_first(&amn->objects, range->start, end);
>>> +    it = interval_tree_iter_first(&amn->objects, start, end);
>>>        while (it) {
>>>            struct amdgpu_mn_node *node;
>>> -        if (!range->blockable) {
>>> +        if (!blockable) {
>>>                amdgpu_mn_read_unlock(amn);
>>>                return -EAGAIN;
>>>            }
>>>            node = container_of(it, struct amdgpu_mn_node, it);
>>> -        it = interval_tree_iter_next(it, range->start, end);
>>> +        it = interval_tree_iter_next(it, start, end);
>>> -        amdgpu_mn_invalidate_node(node, range->start, end);
>>> +        amdgpu_mn_invalidate_node(node, start, end);
>>>        }
>>> +    amdgpu_mn_read_unlock(amn);
>>> +
>>>        return 0;
>>>    }
>>>    /**
>>>     * amdgpu_mn_invalidate_range_start_hsa - callback to notify about
>>> mm change
>>>     *
>>> - * @mn: our notifier
>>> - * @mm: the mm this callback is about
>>> - * @start: start of updated range
>>> - * @end: end of updated range
>>> + * @mirror: the hmm_mirror (mm) is about to update
>>> + * @update: the update start, end address
>>>     *
>>>     * We temporarily evict all BOs between start and end. This
>>>     * necessitates evicting all user-mode queues of the process. The BOs
>>>     * are restorted in amdgpu_mn_invalidate_range_end_hsa.
>>>     */
>>> -static int amdgpu_mn_invalidate_range_start_hsa(struct mmu_notifier *mn,
>>> -            const struct mmu_notifier_range *range)
>>> +static int amdgpu_mn_sync_pagetables_hsa(struct hmm_mirror *mirror,
>>> +            const struct hmm_update *update)
>>>    {
>>> -    struct amdgpu_mn *amn = container_of(mn, struct amdgpu_mn, mn);
>>> +    struct amdgpu_mn *amn = container_of(mirror, struct amdgpu_mn,
>>> mirror);
>>> +    unsigned long start = update->start;
>>> +    unsigned long end = update->end;
>>> +    bool blockable = update->blockable;
>>>        struct interval_tree_node *it;
>>> -    unsigned long end;
>>>        /* notification is exclusive, but interval is inclusive */
>>> -    end = range->end - 1;
>>> +    end -= 1;
>>> -    if (amdgpu_mn_read_lock(amn, range->blockable))
>>> +    if (amdgpu_mn_read_lock(amn, blockable))
>>>            return -EAGAIN;
>>> -    it = interval_tree_iter_first(&amn->objects, range->start, end);
>>> +    it = interval_tree_iter_first(&amn->objects, start, end);
>>>        while (it) {
>>>            struct amdgpu_mn_node *node;
>>>            struct amdgpu_bo *bo;
>>> -        if (!range->blockable) {
>>> +        if (!blockable) {
>>>                amdgpu_mn_read_unlock(amn);
>>>                return -EAGAIN;
>>>            }
>>>            node = container_of(it, struct amdgpu_mn_node, it);
>>> -        it = interval_tree_iter_next(it, range->start, end);
>>> +        it = interval_tree_iter_next(it, start, end);
>>>            list_for_each_entry(bo, &node->bos, mn_list) {
>>>                struct kgd_mem *mem = bo->kfd_bo;
>>>                if (amdgpu_ttm_tt_affect_userptr(bo->tbo.ttm,
>>> -                             range->start,
>>> -                             end))
>>> -                amdgpu_amdkfd_evict_userptr(mem, range->mm);
>>> +                             start, end))
>>> +                amdgpu_amdkfd_evict_userptr(mem, amn->mm);
>>>            }
>>>        }
>>> +    amdgpu_mn_read_unlock(amn);
>>> +
>>>        return 0;
>>>    }
>>> -/**
>>> - * amdgpu_mn_invalidate_range_end - callback to notify about mm change
>>> - *
>>> - * @mn: our notifier
>>> - * @mm: the mm this callback is about
>>> - * @start: start of updated range
>>> - * @end: end of updated range
>>> - *
>>> - * Release the lock again to allow new command submissions.
>>> +/* Low bits of any reasonable mm pointer will be unused due to struct
>>> + * alignment. Use these bits to make a unique key from the mm pointer
>>> + * and notifier type.
>>>     */
>>> -static void amdgpu_mn_invalidate_range_end(struct mmu_notifier *mn,
>>> -            const struct mmu_notifier_range *range)
>>> -{
>>> -    struct amdgpu_mn *amn = container_of(mn, struct amdgpu_mn, mn);
>>> -
>>> -    amdgpu_mn_read_unlock(amn);
>>> -}
>>> +#define AMDGPU_MN_KEY(mm, type) ((unsigned long)(mm) + (type))
>>> -static const struct mmu_notifier_ops amdgpu_mn_ops[] = {
>>> +static struct hmm_mirror_ops amdgpu_hmm_mirror_ops[] = {
>>>        [AMDGPU_MN_TYPE_GFX] = {
>>> -        .release = amdgpu_mn_release,
>>> -        .invalidate_range_start = amdgpu_mn_invalidate_range_start_gfx,
>>> -        .invalidate_range_end = amdgpu_mn_invalidate_range_end,
>>> +        .sync_cpu_device_pagetables = amdgpu_mn_sync_pagetables_gfx,
>>> +        .release = amdgpu_hmm_mirror_release
>>>        },
>>>        [AMDGPU_MN_TYPE_HSA] = {
>>> -        .release = amdgpu_mn_release,
>>> -        .invalidate_range_start = amdgpu_mn_invalidate_range_start_hsa,
>>> -        .invalidate_range_end = amdgpu_mn_invalidate_range_end,
>>> +        .sync_cpu_device_pagetables = amdgpu_mn_sync_pagetables_hsa,
>>> +        .release = amdgpu_hmm_mirror_release
>>>        },
>>>    };
>>> -/* Low bits of any reasonable mm pointer will be unused due to struct
>>> - * alignment. Use these bits to make a unique key from the mm pointer
>>> - * and notifier type.
>>> - */
>>> -#define AMDGPU_MN_KEY(mm, type) ((unsigned long)(mm) + (type))
>>> -
>>>    /**
>>> - * amdgpu_mn_get - create notifier context
>>> + * amdgpu_mn_get - create HMM mirror context
>>>     *
>>>     * @adev: amdgpu device pointer
>>>     * @type: type of MMU notifier context
>>>     *
>>> - * Creates a notifier context for current->mm.
>>> + * Creates a HMM mirror context for current->mm.
>>>     */
>>>    struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev,
>>>                    enum amdgpu_mn_type type)
>>> @@ -401,12 +386,12 @@ struct amdgpu_mn *amdgpu_mn_get(struct
>>> amdgpu_device *adev,
>>>        amn->mm = mm;
>>>        init_rwsem(&amn->lock);
>>>        amn->type = type;
>>> -    amn->mn.ops = &amdgpu_mn_ops[type];
>>>        amn->objects = RB_ROOT_CACHED;
>>>        mutex_init(&amn->read_lock);
>>>        atomic_set(&amn->recursion, 0);
>>> -    r = __mmu_notifier_register(&amn->mn, mm);
>>> +    amn->mirror.ops = &amdgpu_hmm_mirror_ops[type];
>>> +    r = hmm_mirror_register(&amn->mirror, mm);
>>>        if (r)
>>>            goto free_amn;
>>> @@ -432,7 +417,7 @@ struct amdgpu_mn *amdgpu_mn_get(struct
>>> amdgpu_device *adev,
>>>     * @bo: amdgpu buffer object
>>>     * @addr: userptr addr we should monitor
>>>     *
>>> - * Registers an MMU notifier for the given BO at the specified address.
>>> + * Registers an HMM mirror for the given BO at the specified address.
>>>     * Returns 0 on success, -ERRNO if anything goes wrong.
>>>     */
>>>    int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr)
>>> @@ -488,11 +473,11 @@ int amdgpu_mn_register(struct amdgpu_bo *bo,
>>> unsigned long addr)
>>>    }
>>>    /**
>>> - * amdgpu_mn_unregister - unregister a BO for notifier updates
>>> + * amdgpu_mn_unregister - unregister a BO for HMM mirror updates
>>>     *
>>>     * @bo: amdgpu buffer object
>>>     *
>>> - * Remove any registration of MMU notifier updates from the buffer
>>> object.
>>> + * Remove any registration of HMM mirror updates from the buffer object.
>>>     */
>>>    void amdgpu_mn_unregister(struct amdgpu_bo *bo)
>>>    {
>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.h
>>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.h
>>> index eb0f432f78fe..0a51fd00021c 100644
>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.h
>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.h
>>> @@ -34,7 +34,7 @@ enum amdgpu_mn_type {
>>>        AMDGPU_MN_TYPE_HSA,
>>>    };
>>> -#if defined(CONFIG_MMU_NOTIFIER)
>>> +#if defined(CONFIG_HMM_MIRROR)
>>>    void amdgpu_mn_lock(struct amdgpu_mn *mn);
>>>    void amdgpu_mn_unlock(struct amdgpu_mn *mn);
>>>    struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev,
> _______________________________________________
> amd-gfx mailing list
> amd-gfx at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/amd-gfx



More information about the amd-gfx mailing list