Re: [PATCH v6 3/3] drm/doc/rfc: VM_BIND uapi definition

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On 30/06/2022 16:34, Jason Ekstrand wrote:
On Thu, Jun 30, 2022 at 10:14 AM Matthew Auld <matthew.auld@xxxxxxxxx <mailto:matthew.auld@xxxxxxxxx>> wrote:

    On 30/06/2022 06:11, Jason Ekstrand wrote:
     > On Sat, Jun 25, 2022 at 8:49 PM Niranjana Vishwanathapura
     > <niranjana.vishwanathapura@xxxxxxxxx
    <mailto:niranjana.vishwanathapura@xxxxxxxxx>
     > <mailto:niranjana.vishwanathapura@xxxxxxxxx
    <mailto:niranjana.vishwanathapura@xxxxxxxxx>>> wrote:
     >
     >     VM_BIND and related uapi definitions
     >
     >     v2: Reduce the scope to simple Mesa use case.
     >     v3: Expand VM_UNBIND documentation and add
     >          I915_GEM_VM_BIND/UNBIND_FENCE_VALID
     >          and I915_GEM_VM_BIND_TLB_FLUSH flags.
     >     v4: Remove I915_GEM_VM_BIND_TLB_FLUSH flag and add additional
     >          documentation for vm_bind/unbind.
     >     v5: Remove TLB flush requirement on VM_UNBIND.
     >          Add version support to stage implementation.
     >     v6: Define and use drm_i915_gem_timeline_fence structure for
     >          all timeline fences.
     >     v7: Rename I915_PARAM_HAS_VM_BIND to I915_PARAM_VM_BIND_VERSION.
     >          Update documentation on async vm_bind/unbind and versioning.
     >          Remove redundant vm_bind/unbind FENCE_VALID flag, execbuf3
     >          batch_count field and I915_EXEC3_SECURE flag.
     >
     >     Signed-off-by: Niranjana Vishwanathapura
     >     <niranjana.vishwanathapura@xxxxxxxxx
    <mailto:niranjana.vishwanathapura@xxxxxxxxx>
     >     <mailto:niranjana.vishwanathapura@xxxxxxxxx
    <mailto:niranjana.vishwanathapura@xxxxxxxxx>>>
     >     Reviewed-by: Daniel Vetter <daniel.vetter@xxxxxxxx
    <mailto:daniel.vetter@xxxxxxxx>
     >     <mailto:daniel.vetter@xxxxxxxx <mailto:daniel.vetter@xxxxxxxx>>>
     >     ---
     >       Documentation/gpu/rfc/i915_vm_bind.h | 280
    +++++++++++++++++++++++++++
     >       1 file changed, 280 insertions(+)
     >       create mode 100644 Documentation/gpu/rfc/i915_vm_bind.h
     >
     >     diff --git a/Documentation/gpu/rfc/i915_vm_bind.h
     >     b/Documentation/gpu/rfc/i915_vm_bind.h
     >     new file mode 100644
     >     index 000000000000..a93e08bceee6
     >     --- /dev/null
     >     +++ b/Documentation/gpu/rfc/i915_vm_bind.h
     >     @@ -0,0 +1,280 @@
     >     +/* SPDX-License-Identifier: MIT */
     >     +/*
     >     + * Copyright © 2022 Intel Corporation
     >     + */
     >     +
     >     +/**
     >     + * DOC: I915_PARAM_VM_BIND_VERSION
     >     + *
     >     + * VM_BIND feature version supported.
     >     + * See typedef drm_i915_getparam_t param.
     >     + *
     >     + * Specifies the VM_BIND feature version supported.
     >     + * The following versions of VM_BIND have been defined:
     >     + *
     >     + * 0: No VM_BIND support.
     >     + *
     >     + * 1: In VM_UNBIND calls, the UMD must specify the exact
    mappings
     >     created
     >     + *    previously with VM_BIND, the ioctl will not support
    unbinding
     >     multiple
     >     + *    mappings or splitting them. Similarly, VM_BIND calls
    will not
     >     replace
     >     + *    any existing mappings.
     >     + *
     >     + * 2: The restrictions on unbinding partial or multiple
    mappings is
     >     + *    lifted, Similarly, binding will replace any mappings
    in the
     >     given range.
     >     + *
     >     + * See struct drm_i915_gem_vm_bind and struct
    drm_i915_gem_vm_unbind.
     >     + */
     >     +#define I915_PARAM_VM_BIND_VERSION     57
     >     +
     >     +/**
     >     + * DOC: I915_VM_CREATE_FLAGS_USE_VM_BIND
     >     + *
     >     + * Flag to opt-in for VM_BIND mode of binding during VM
    creation.
     >     + * See struct drm_i915_gem_vm_control flags.
     >     + *
     >     + * The older execbuf2 ioctl will not support VM_BIND mode of
    operation.
     >     + * For VM_BIND mode, we have new execbuf3 ioctl which will not
     >     accept any
     >     + * execlist (See struct drm_i915_gem_execbuffer3 for more
    details).
     >     + */
     >     +#define I915_VM_CREATE_FLAGS_USE_VM_BIND       (1 << 0)
     >     +
     >     +/* VM_BIND related ioctls */
     >     +#define DRM_I915_GEM_VM_BIND           0x3d
     >     +#define DRM_I915_GEM_VM_UNBIND         0x3e
     >     +#define DRM_I915_GEM_EXECBUFFER3       0x3f
     >     +
     >     +#define DRM_IOCTL_I915_GEM_VM_BIND
     >       DRM_IOWR(DRM_COMMAND_BASE + DRM_I915_GEM_VM_BIND, struct
     >     drm_i915_gem_vm_bind)
     >     +#define DRM_IOCTL_I915_GEM_VM_UNBIND
     >       DRM_IOWR(DRM_COMMAND_BASE + DRM_I915_GEM_VM_UNBIND, struct
     >     drm_i915_gem_vm_bind)
     >     +#define DRM_IOCTL_I915_GEM_EXECBUFFER3
     >       DRM_IOWR(DRM_COMMAND_BASE + DRM_I915_GEM_EXECBUFFER3, struct
     >     drm_i915_gem_execbuffer3)
     >     +
     >     +/**
     >     + * struct drm_i915_gem_timeline_fence - An input or output
    timeline
     >     fence.
     >     + *
     >     + * The operation will wait for input fence to signal.
     >     + *
     >     + * The returned output fence will be signaled after the
    completion
     >     of the
     >     + * operation.
     >     + */
     >     +struct drm_i915_gem_timeline_fence {
     >     +       /** @handle: User's handle for a drm_syncobj to wait
    on or
     >     signal. */
     >     +       __u32 handle;
     >     +
     >     +       /**
     >     +        * @flags: Supported flags are:
     >     +        *
     >     +        * I915_TIMELINE_FENCE_WAIT:
     >     +        * Wait for the input fence before the operation.
     >     +        *
     >     +        * I915_TIMELINE_FENCE_SIGNAL:
     >     +        * Return operation completion fence as output.
     >     +        */
     >     +       __u32 flags;
     >     +#define I915_TIMELINE_FENCE_WAIT            (1 << 0)
     >     +#define I915_TIMELINE_FENCE_SIGNAL          (1 << 1)
     >     +#define __I915_TIMELINE_FENCE_UNKNOWN_FLAGS
     >     (-(I915_TIMELINE_FENCE_SIGNAL << 1))
     >     +
     >     +       /**
     >     +        * @value: A point in the timeline.
     >     +        * Value must be 0 for a binary drm_syncobj. A Value
    of 0 for a
     >     +        * timeline drm_syncobj is invalid as it turns a
    drm_syncobj
     >     into a
     >     +        * binary one.
     >     +        */
     >     +       __u64 value;
     >     +};
     >     +
     >     +/**
     >     + * struct drm_i915_gem_vm_bind - VA to object mapping to bind.
     >     + *
     >     + * This structure is passed to VM_BIND ioctl and specifies the
     >     mapping of GPU
     >     + * virtual address (VA) range to the section of an object that
     >     should be bound
     >     + * in the device page table of the specified address space (VM).
     >     + * The VA range specified must be unique (ie., not currently
    bound)
     >     and can
     >     + * be mapped to whole object or a section of the object (partial
     >     binding).
     >     + * Multiple VA mappings can be created to the same section
    of the
     >     object
     >     + * (aliasing).
     >     + *
     >     + * The @start, @offset and @length must be 4K page aligned.
    However
     >     the DG2
     >     + * and XEHPSDV has 64K page size for device local-memory and has
     >     compact page
     >     + * table. On those platforms, for binding device local-memory
     >     objects, the
     >     + * @start must be 2M aligned, @offset and @length must be
    64K aligned.
     >
     >
     > This is not acceptable.  We need 64K granularity.  This includes the
     > starting address, the BO offset, and the length.  Why?  The tl;dr is
     > that it's a requirement for about 50% of D3D12 apps if we want
    them to
     > run on Linux via D3D12.  A longer explanation follows.  I don't
     > necessarily expect kernel folks to get all the details but hopefully
     > I'll have left enough of a map that some of the Intel Mesa folks can
     > help fill in details.
     >
     > Many modern D3D12 apps have a hard requirement on Tier2 tiled
     > resources.  This is a feature that Intel has supported in the D3D12
     > driver since Skylake.  In order to implement this feature, VKD3D
     > requires the various sparseResidencyImage* and
    sparseResidency*Sampled
     > Vulkan features.  If we want those apps to work (there's getting
    to be
     > quite a few of them), we need to implement the Vulkan sparse
    residency
     > features.
     > |
     > |
     > What is sparse residency?  I'm glad you asked!  The sparse residency
     > features allow a client to separately bind each miplevel or array
    slice
     > of an image to a chunk of device memory independently, without
    affecting
     > any other areas of the image.  Once you get to a high enough
    miplevel
     > that everything fits inside a single sparse image block (that's a
     > technical Vulkan term you can search for in the spec), you can
    enter a
     > "miptail" which contains all the remaining miplevels in a single
    sparse
     > image block.
     >
     > The term "sparse image block" is what the Vulkan spec uses.  On
    Intel
> hardware and in the docs, it's what we call a "tile". Specifically, the
     > image needs to use Yf or Ys tiling on SKL-TGL or a Tile64 on
    DG2+.  This
     > is because Tile4 and legacy X and Y-tiling don't provide any
    guarantees
     > about page alignment for slices.  Yf, Ys, and Tile64, on the
    other hand,
     > align all slices of the image to a tile boundary, allowing us to map
     > memory to different slices independently, assuming we have 64K
    (or 4K
     > for Yf) VM_BIND granularity.  (4K isn't actually a requirement for
     > SKL-TGL; we can use Ys all the time which has 64K tiles but
    there's no
     > reason to not support 4K alignments on integrated.)
     >
     > Someone may be tempted to ask, "Can't we wiggle the strides
    around or
     > something to make it work?"  I thought about that and no, you can't.
     > The problem here is LOD2+.  Sure, you can have a stride such that
    the
     > image is a multiple of 2M worth of tiles across.  That'll work
    fine for
     > LOD0 and LOD1; both will be 2M aligned.  However, LOD2 won't be and
     > there's no way to control that.  The hardware will place it to
    the right
     > of LOD1 by ROUND_UP(width, tile_width) pixels and there's nothing
    you
     > can do about that.  If that position doesn't happen to hit a 2M
     > boundary, you're out of luck.
     >
     > I hope that explanation provides enough detail.  Sadly, this is
    one of
     > those things which has a lot of moving pieces all over different
    bits of
     > the hardware and various APIs and they all have to work together
    just
     > right for it to all come out in the end.  But, yeah, we really
    need 64K
     > aligned binding if we want VKD3D to work.

    Just to confirm, the new model would be to enforce 64K GTT alignment
    for
    lmem pages, and then for smem pages we would only require 4K alignment,
    but with the added restriction that userspace will never try to mix the
    two (lmem vs smem) within the same 2M va range (page-table). The kernel
    will verify this and throw an error if needed. This model should work
    with the above?


Mesa doesn't have full control over BO placement so I don't think we can guarantee quite as much as you want there.  We can guarantee, I think, that we never place LMEM-only and SMEM-only in the same 2M block. However, most BOs will be LMEM+SMEM (with a preference for LMEM) and then it'll be up to the kernel to sort out any issues.  Is that reasonable?

That seems tricky for the lmem + smem case. On DG2 the hw design is such that you can't have 64K and 4K GTT pages within the same page-table, since the entire page-table is either operating in 64K or 4K GTT page mode (there is some special bit on the PDE that we need to toggle to turn on/off the 64K mode).


--Jason

     >
     > --Jason
     >
     >     + * Also, for such mappings, i915 will reserve the whole 2M range
     >     for it so as
     >     + * to not allow multiple mappings in that 2M range (Compact page
     >     tables do not
     >     + * allow 64K page and 4K page bindings in the same 2M range).
     >     + *
     >     + * Error code -EINVAL will be returned if @start, @offset and
     >     @length are not
     >     + * properly aligned. In version 1 (See
    I915_PARAM_VM_BIND_VERSION),
     >     error code
     >     + * -ENOSPC will be returned if the VA range specified can't be
     >     reserved.
     >     + *
     >     + * VM_BIND/UNBIND ioctl calls executed on different CPU threads
     >     concurrently
     >     + * are not ordered. Furthermore, parts of the VM_BIND
    operation can
     >     be done
     >     + * asynchronously, if valid @fence is specified.
     >     + */
     >     +struct drm_i915_gem_vm_bind {
     >     +       /** @vm_id: VM (address space) id to bind */
     >     +       __u32 vm_id;
     >     +
     >     +       /** @handle: Object handle */
     >     +       __u32 handle;
     >     +
     >     +       /** @start: Virtual Address start to bind */
     >     +       __u64 start;
     >     +
     >     +       /** @offset: Offset in object to bind */
     >     +       __u64 offset;
     >     +
     >     +       /** @length: Length of mapping to bind */
     >     +       __u64 length;
     >     +
     >     +       /**
     >     +        * @flags: Supported flags are:
     >     +        *
     >     +        * I915_GEM_VM_BIND_READONLY:
     >     +        * Mapping is read-only.
     >     +        *
     >     +        * I915_GEM_VM_BIND_CAPTURE:
     >     +        * Capture this mapping in the dump upon GPU error.
     >     +        */
     >     +       __u64 flags;
     >     +#define I915_GEM_VM_BIND_READONLY      (1 << 1)
     >     +#define I915_GEM_VM_BIND_CAPTURE       (1 << 2)
     >     +
     >     +       /**
     >     +        * @fence: Timeline fence for bind completion signaling.
     >     +        *
     >     +        * It is an out fence, hence using
    I915_TIMELINE_FENCE_WAIT flag
     >     +        * is invalid, and an error will be returned.
     >     +        */
     >     +       struct drm_i915_gem_timeline_fence fence;
     >     +
     >     +       /**
     >     +        * @extensions: Zero-terminated chain of extensions.
     >     +        *
     >     +        * For future extensions. See struct i915_user_extension.
     >     +        */
     >     +       __u64 extensions;
     >     +};
     >     +
     >     +/**
     >     + * struct drm_i915_gem_vm_unbind - VA to object mapping to
    unbind.
     >     + *
     >     + * This structure is passed to VM_UNBIND ioctl and specifies the
     >     GPU virtual
     >     + * address (VA) range that should be unbound from the device
    page
     >     table of the
     >     + * specified address space (VM). VM_UNBIND will force unbind the
     >     specified
     >     + * range from device page table without waiting for any GPU
    job to
     >     complete.
     >     + * It is UMDs responsibility to ensure the mapping is no
    longer in
     >     use before
     >     + * calling VM_UNBIND.
     >     + *
     >     + * If the specified mapping is not found, the ioctl will simply
     >     return without
     >     + * any error.
     >     + *
     >     + * VM_BIND/UNBIND ioctl calls executed on different CPU threads
     >     concurrently
     >     + * are not ordered. Furthermore, parts of the VM_UNBIND
    operation
     >     can be done
     >     + * asynchronously, if valid @fence is specified.
     >     + */
     >     +struct drm_i915_gem_vm_unbind {
     >     +       /** @vm_id: VM (address space) id to bind */
     >     +       __u32 vm_id;
     >     +
     >     +       /** @rsvd: Reserved, MBZ */
     >     +       __u32 rsvd;
     >     +
     >     +       /** @start: Virtual Address start to unbind */
     >     +       __u64 start;
     >     +
     >     +       /** @length: Length of mapping to unbind */
     >     +       __u64 length;
     >     +
     >     +       /** @flags: Currently reserved, MBZ */
     >     +       __u64 flags;
     >     +
     >     +       /**
     >     +        * @fence: Timeline fence for unbind completion
    signaling.
     >     +        *
     >     +        * It is an out fence, hence using
    I915_TIMELINE_FENCE_WAIT flag
     >     +        * is invalid, and an error will be returned.
     >     +        */
     >     +       struct drm_i915_gem_timeline_fence fence;
     >     +
     >     +       /**
     >     +        * @extensions: Zero-terminated chain of extensions.
     >     +        *
     >     +        * For future extensions. See struct i915_user_extension.
     >     +        */
     >     +       __u64 extensions;
     >     +};
     >     +
     >     +/**
     >     + * struct drm_i915_gem_execbuffer3 - Structure for
     >     DRM_I915_GEM_EXECBUFFER3
     >     + * ioctl.
     >     + *
     >     + * DRM_I915_GEM_EXECBUFFER3 ioctl only works in VM_BIND mode and
     >     VM_BIND mode
     >     + * only works with this ioctl for submission.
     >     + * See I915_VM_CREATE_FLAGS_USE_VM_BIND.
     >     + */
     >     +struct drm_i915_gem_execbuffer3 {
     >     +       /**
     >     +        * @ctx_id: Context id
     >     +        *
     >     +        * Only contexts with user engine map are allowed.
     >     +        */
     >     +       __u32 ctx_id;
     >     +
     >     +       /**
     >     +        * @engine_idx: Engine index
     >     +        *
     >     +        * An index in the user engine map of the context
    specified
     >     by @ctx_id.
     >     +        */
     >     +       __u32 engine_idx;
     >     +
     >     +       /**
     >     +        * @batch_address: Batch gpu virtual address/es.
     >     +        *
     >     +        * For normal submission, it is the gpu virtual
    address of
     >     the batch
     >     +        * buffer. For parallel submission, it is a pointer to an
     >     array of
     >     +        * batch buffer gpu virtual addresses with array size
    equal
     >     to the
     >     +        * number of (parallel) engines involved in that
    submission (See
     >     +        * struct i915_context_engines_parallel_submit).
     >     +        */
     >     +       __u64 batch_address;
     >     +
     >     +       /** @flags: Currently reserved, MBZ */
     >     +       __u64 flags;
     >     +
     >     +       /** @rsvd1: Reserved, MBZ */
     >     +       __u32 rsvd1;
     >     +
     >     +       /** @fence_count: Number of fences in
    @timeline_fences array. */
     >     +       __u32 fence_count;
     >     +
     >     +       /**
     >     +        * @timeline_fences: Pointer to an array of timeline
    fences.
     >     +        *
     >     +        * Timeline fences are of format struct
     >     drm_i915_gem_timeline_fence.
     >     +        */
     >     +       __u64 timeline_fences;
     >     +
     >     +       /** @rsvd2: Reserved, MBZ */
     >     +       __u64 rsvd2;
     >     +
     >     +       /**
     >     +        * @extensions: Zero-terminated chain of extensions.
     >     +        *
     >     +        * For future extensions. See struct i915_user_extension.
     >     +        */
     >     +       __u64 extensions;
     >     +};
     >     +
     >     +/**
     >     + * struct drm_i915_gem_create_ext_vm_private - Extension to make
     >     the object
     >     + * private to the specified VM.
     >     + *
     >     + * See struct drm_i915_gem_create_ext.
     >     + */
     >     +struct drm_i915_gem_create_ext_vm_private {
     >     +#define I915_GEM_CREATE_EXT_VM_PRIVATE         2
     >     +       /** @base: Extension link. See struct
    i915_user_extension. */
     >     +       struct i915_user_extension base;
     >     +
     >     +       /** @vm_id: Id of the VM to which the object is
    private */
     >     +       __u32 vm_id;
     >     +};
     >     --
     >     2.21.0.rc0.32.g243a4c7e27
     >




[Index of Archives]     [Linux DRI Users]     [Linux Intel Graphics]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]     [XFree86]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Linux Kernel]     [Linux SCSI]     [XFree86]
  Powered by Linux