Buffer Sharing and Synchronization (dma-buf)¶

The dma-buf subsystem provides the framework for sharing buffers for hardware (DMA) access across multiple device drivers and subsystems, and for synchronizing asynchronous hardware access.

This is used, for example, by drm "prime" multi-GPU support, but is of course not limited to GPU use cases.

The three main components of this are: (1) dma-buf, representing a sg_table and exposed to userspace as a file descriptor to allow passing between devices, (2) fence, which provides a mechanism to signal when one device has finished access, and (3) reservation, which manages the shared or exclusive fence(s) associated with the buffer.

Shared DMA Buffers¶

This document serves as a guide to device-driver writers on what is the dma-buf buffer sharing API, how to use it for exporting and using shared buffers.

Any device driver which wishes to be a part of DMA buffer sharing, can do so as either the 'exporter' of buffers, or the 'user' or 'importer' of buffers.

Say a driver A wants to use buffers created by driver B, then we call B as the exporter, and A as buffer-user/importer.

The exporter

implements and manages operations in struct dma_buf_ops for the buffer,

allows other users to share the buffer by using dma_buf sharing APIs,

manages the details of buffer allocation, wrapped in a struct dma_buf,

decides about the actual backing storage where this allocation happens,

and takes care of any migration of scatterlist - for all (shared) users of this buffer.

The buffer-user

is one of (many) sharing users of the buffer.

doesn't need to worry about how the buffer is allocated, or where.

and needs a mechanism to get access to the scatterlist that makes up this buffer in memory, mapped into its own address space, so it can access the same area of memory. This interface is provided by struct dma_buf_attachment.

Any exporters or users of the dma-buf buffer sharing framework must have a 'select DMA_SHARED_BUFFER' in their respective Kconfigs.

Userspace Interface Notes¶

Mostly a DMA buffer file descriptor is simply an opaque object for userspace, and hence the generic interface exposed is very minimal. There's a few things to consider though:

Since kernel 3.12 the dma-buf FD supports the llseek system call, but only with offset=0 and whence=SEEK_END|SEEK_SET. SEEK_SET is supported to allow the usual size discover pattern size = SEEK_END(0); SEEK_SET(0). Every other llseek operation will report -EINVAL.

If llseek on dma-buf FDs isn't support the kernel will report -ESPIPE for all cases. Userspace can use this to detect support for discovering the dma-buf size using llseek.
In order to avoid fd leaks on exec, the FD_CLOEXEC flag must be set on the file descriptor. This is not just a resource leak, but a potential security hole. It could give the newly exec'd application access to buffers, via the leaked fd, to which it should otherwise not be permitted access.

The problem with doing this via a separate fcntl() call, versus doing it atomically when the fd is created, is that this is inherently racy in a multi-threaded app[3]. The issue is made worse when it is library code opening/creating the file descriptor, as the application may not even be aware of the fd's.

To avoid this problem, userspace must have a way to request O_CLOEXEC flag be set when the dma-buf fd is created. So any API provided by the exporting driver to create a dmabuf fd must provide a way to let userspace control setting of O_CLOEXEC flag passed in to dma_buf_fd().
Memory mapping the contents of the DMA buffer is also supported. See the discussion below on CPU Access to DMA Buffer Objects for the full details.
The DMA buffer FD is also pollable, see Implicit Fence Poll Support below for details.
The DMA buffer FD also supports a few dma-buf-specific ioctls, see DMA Buffer ioctls below for details.

Basic Operation and Device DMA Access¶

For device DMA access to a shared DMA buffer the usual sequence of operations is fairly simple:

The exporter defines his exporter instance using DEFINE_DMA_BUF_EXPORT_INFO() and calls dma_buf_export() to wrap a private buffer object into a dma_buf. It then exports that dma_buf to userspace as a file descriptor by calling dma_buf_fd().
Userspace passes this file-descriptors to all drivers it wants this buffer to share with: First the file descriptor is converted to a dma_buf using dma_buf_get(). Then the buffer is attached to the device using dma_buf_attach().

Up to this stage the exporter is still free to migrate or reallocate the backing storage.
Once the buffer is attached to all devices userspace can initiate DMA access to the shared buffer. In the kernel this is done by calling dma_buf_map_attachment() and dma_buf_unmap_attachment().
Once a driver is done with a shared buffer it needs to call dma_buf_detach() (after cleaning up any mappings) and then release the reference acquired with dma_buf_get() by calling dma_buf_put().

For the detailed semantics exporters are expected to implement see dma_buf_ops.

CPU Access to DMA Buffer Objects¶

There are multiple reasons for supporting CPU access to a dma buffer object:

Fallback operations in the kernel, for example when a device is connected over USB and the kernel needs to shuffle the data around first before sending it away. Cache coherency is handled by bracketing any transactions with calls to dma_buf_begin_cpu_access() and dma_buf_end_cpu_access() access.

Since for most kernel internal dma-buf accesses need the entire buffer, a vmap interface is introduced. Note that on very old 32-bit architectures vmalloc space might be limited and result in vmap calls failing.

Interfaces:
```
void \*dma_buf_vmap(struct dma_buf \*dmabuf, struct iosys_map \*map)
void dma_buf_vunmap(struct dma_buf \*dmabuf, struct iosys_map \*map)
```
The vmap call can fail if there is no vmap support in the exporter, or if it runs out of vmalloc space. Note that the dma-buf layer keeps a reference count for all vmap access and calls down into the exporter's vmap function only when no vmapping exists, and only unmaps it once. Protection against concurrent vmap/vunmap calls is provided by taking the dma_buf.lock mutex.
For full compatibility on the importer side with existing userspace interfaces, which might already support mmap'ing buffers. This is needed in many processing pipelines (e.g. feeding a software rendered image into a hardware pipeline, thumbnail creation, snapshots, ...). Also, Android's ION framework already supported this and for DMA buffer file descriptors to replace ION buffers mmap support was needed.

There is no special interfaces, userspace simply calls mmap on the dma-buf fd. But like for CPU access there's a need to bracket the actual access, which is handled by the ioctl (DMA_BUF_IOCTL_SYNC). Note that DMA_BUF_IOCTL_SYNC can fail with -EAGAIN or -EINTR, in which case it must be restarted.

Some systems might need some sort of cache coherency management e.g. when CPU and GPU domains are being accessed through dma-buf at the same time. To circumvent this problem there are begin/end coherency markers, that forward directly to existing dma-buf device drivers vfunc hooks. Userspace can make use of those markers through the DMA_BUF_IOCTL_SYNC ioctl. The sequence would be used like following:
mmap dma-buf fd

for each drawing/upload cycle in CPU 1. SYNC_START ioctl, 2. read/write to mmap area 3. SYNC_END ioctl. This can be repeated as often as you want (with the new data being consumed by say the GPU or the scanout device)

munmap once you don't need the buffer any more
For correctness and optimal performance, it is always required to use SYNC_START and SYNC_END before and after, respectively, when accessing the mapped address. Userspace cannot rely on coherent access, even when there are systems where it just works without calling these ioctls.
And as a CPU fallback in userspace processing pipelines.

Similar to the motivation for kernel cpu access it is again important that the userspace code of a given importing subsystem can use the same interfaces with a imported dma-buf buffer object as with a native buffer object. This is especially important for drm where the userspace part of contemporary OpenGL, X, and other drivers is huge, and reworking them to use a different way to mmap a buffer rather invasive.

The assumption in the current dma-buf interfaces is that redirecting the initial mmap is all that's needed. A survey of some of the existing subsystems shows that no driver seems to do any nefarious thing like syncing up with outstanding asynchronous processing on the device or allocating special resources at fault time. So hopefully this is good enough, since adding interfaces to intercept pagefaults and allow pte shootdowns would increase the complexity quite a bit.

Interface:
```
int dma_buf_mmap(struct dma_buf \*, struct vm_area_struct \*,
               unsigned long);
```
If the importing subsystem simply provides a special-purpose mmap call to set up a mapping in userspace, calling do_mmap with dma_buf.file will equally achieve that for a dma-buf object.

Implicit Fence Poll Support¶

To support cross-device and cross-driver synchronization of buffer access implicit fences (represented internally in the kernel with struct dma_fence) can be attached to a dma_buf. The glue for that and a few related things are provided in the dma_resv structure.

Userspace can query the state of these implicitly tracked fences using poll() and related system calls:

Checking for EPOLLIN, i.e. read access, can be use to query the state of the most recent write or exclusive fence.
Checking for EPOLLOUT, i.e. write access, can be used to query the state of all attached fences, shared and exclusive ones.

Note that this only signals the completion of the respective fences, i.e. the DMA transfers are complete. Cache flushing and any other necessary preparations before CPU access can begin still need to happen.

As an alternative to poll(), the set of fences on DMA buffer can be exported as a sync_file using dma_buf_sync_file_export.

DMA-BUF statistics¶

/sys/kernel/debug/dma_buf/bufinfo provides an overview of every DMA-BUF in the system. However, since debugfs is not safe to be mounted in production, procfs and sysfs can be used to gather DMA-BUF statistics on production systems.

The /proc/<pid>/fdinfo/<fd> files in procfs can be used to gather information about DMA-BUF fds. Detailed documentation about the interface is present in The /proc Filesystem.

Unfortunately, the existing procfs interfaces can only provide information about the DMA-BUFs for which processes hold fds or have the buffers mmapped into their address space. This necessitated the creation of the DMA-BUF sysfs statistics interface to provide per-buffer information on production systems.

The interface at /sys/kernel/dmabuf/buffers exposes information about every DMA-BUF when CONFIG_DMABUF_SYSFS_STATS is enabled.

The following stats are exposed by the interface:

/sys/kernel/dmabuf/buffers/<inode_number>/exporter_name
/sys/kernel/dmabuf/buffers/<inode_number>/size

The information in the interface can also be used to derive per-exporter statistics. The data from the interface can be gathered on error conditions or other important events to provide a snapshot of DMA-BUF usage. It can also be collected periodically by telemetry to monitor various metrics.

Detailed documentation about the interface is present in Documentation/ABI/testing/sysfs-kernel-dmabuf-buffers.

DMA Buffer ioctls¶

struct dma_buf_sync¶: Synchronize with CPU access.

Definition:

struct dma_buf_sync {
    __u64 flags;
};

Members

flags

Set of access flags

DMA_BUF_SYNC_START:: Indicates the start of a map access session.
DMA_BUF_SYNC_END:: Indicates the end of a map access session.
DMA_BUF_SYNC_READ:: Indicates that the mapped DMA buffer will be read by the client via the CPU map.
DMA_BUF_SYNC_WRITE:: Indicates that the mapped DMA buffer will be written by the client via the CPU map.
DMA_BUF_SYNC_RW:: An alias for DMA_BUF_SYNC_READ | DMA_BUF_SYNC_WRITE.

Description

When a DMA buffer is accessed from the CPU via mmap, it is not always possible to guarantee coherency between the CPU-visible map and underlying memory. To manage coherency, DMA_BUF_IOCTL_SYNC must be used to bracket any CPU access to give the kernel the chance to shuffle memory around if needed.

Prior to accessing the map, the client must call DMA_BUF_IOCTL_SYNC with DMA_BUF_SYNC_START and the appropriate read/write flags. Once the access is complete, the client should call DMA_BUF_IOCTL_SYNC with DMA_BUF_SYNC_END and the same read/write flags.

The synchronization provided via DMA_BUF_IOCTL_SYNC only provides cache coherency. It does not prevent other processes or devices from accessing the memory at the same time. If synchronization with a GPU or other device driver is required, it is the client's responsibility to wait for buffer to be ready for reading or writing before calling this ioctl with DMA_BUF_SYNC_START. Likewise, the client must ensure that follow-up work is not submitted to GPU or other device driver until after this ioctl has been called with DMA_BUF_SYNC_END?

If the driver or API with which the client is interacting uses implicit synchronization, waiting for prior work to complete can be done via poll() on the DMA buffer file descriptor. If the driver or API requires explicit synchronization, the client may have to wait on a sync_file or other synchronization primitive outside the scope of the DMA buffer API.

struct dma_buf_export_sync_file¶: Get a sync_file from a dma-buf

Definition:

struct dma_buf_export_sync_file {
    __u32 flags;
    __s32 fd;
};

Members

flags

Read/write flags

Must be DMA_BUF_SYNC_READ, DMA_BUF_SYNC_WRITE, or both.

If DMA_BUF_SYNC_READ is set and DMA_BUF_SYNC_WRITE is not set, the returned sync file waits on any writers of the dma-buf to complete. Waiting on the returned sync file is equivalent to poll() with POLLIN.

If DMA_BUF_SYNC_WRITE is set, the returned sync file waits on any users of the dma-buf (read or write) to complete. Waiting on the returned sync file is equivalent to poll() with POLLOUT. If both DMA_BUF_SYNC_WRITE and DMA_BUF_SYNC_READ are set, this is equivalent to just DMA_BUF_SYNC_WRITE.

fd

Returned sync file descriptor

Description

Userspace can perform a DMA_BUF_IOCTL_EXPORT_SYNC_FILE to retrieve the current set of fences on a dma-buf file descriptor as a sync_file. CPU waits via poll() or other driver-specific mechanisms typically wait on whatever fences are on the dma-buf at the time the wait begins. This is similar except that it takes a snapshot of the current fences on the dma-buf for waiting later instead of waiting immediately. This is useful for modern graphics APIs such as Vulkan which assume an explicit synchronization model but still need to inter-operate with dma-buf.

The intended usage pattern is the following:

Export a sync_file with flags corresponding to the expected GPU usage via DMA_BUF_IOCTL_EXPORT_SYNC_FILE.

Submit rendering work which uses the dma-buf. The work should wait on the exported sync file before rendering and produce another sync_file when complete.

Import the rendering-complete sync_file into the dma-buf with flags corresponding to the GPU usage via DMA_BUF_IOCTL_IMPORT_SYNC_FILE.

Unlike doing implicit synchronization via a GPU kernel driver's exec ioctl, the above is not a single atomic operation. If userspace wants to ensure ordering via these fences, it is the respnosibility of userspace to use locks or other mechanisms to ensure that no other context adds fences or submits work between steps 1 and 3 above.

struct dma_buf_import_sync_file¶: Insert a sync_file into a dma-buf

Definition:

struct dma_buf_import_sync_file {
    __u32 flags;
    __s32 fd;
};

Members

flags

Read/write flags

Must be DMA_BUF_SYNC_READ, DMA_BUF_SYNC_WRITE, or both.

If DMA_BUF_SYNC_READ is set and DMA_BUF_SYNC_WRITE is not set, this inserts the sync_file as a read-only fence. Any subsequent implicitly synchronized writes to this dma-buf will wait on this fence but reads will not.

If DMA_BUF_SYNC_WRITE is set, this inserts the sync_file as a write fence. All subsequent implicitly synchronized access to this dma-buf will wait on this fence.

fd

Sync file descriptor

Description

Userspace can perform a DMA_BUF_IOCTL_IMPORT_SYNC_FILE to insert a sync_file into a dma-buf for the purposes of implicit synchronization with other dma-buf consumers. This allows clients using explicitly synchronized APIs such as Vulkan to inter-op with dma-buf consumers which expect implicit synchronization such as OpenGL or most media drivers/video.

DMA-BUF locking convention¶

In order to avoid deadlock situations between dma-buf exports and importers, all dma-buf API users must follow the common dma-buf locking convention.

Convention for importers

Importers must hold the dma-buf reservation lock when calling these functions:
Importers must not hold the dma-buf reservation lock when calling these functions:

Convention for exporters

These dma_buf_ops callbacks are invoked with unlocked dma-buf reservation and exporter can take the lock:
These dma_buf_ops callbacks are invoked with locked dma-buf reservation and exporter can't take the lock:
Exporters must hold the dma-buf reservation lock when calling these functions:
- dma_buf_move_notify()

Kernel Functions and Structures Reference¶

struct dma_buf *dma_buf_export(const struct dma_buf_export_info *exp_info)¶: Creates a new dma_buf, and associates an anon file with this buffer, so it can be exported. Also connect the allocator specific data and ops to the buffer. Additionally, provide a name string for exporter; useful in debugging.

Parameters

const struct dma_buf_export_info *exp_info: [in] holds all the export related information provided by the exporter. see struct dma_buf_export_info for further details.

Description

Returns, on success, a newly created struct dma_buf object, which wraps the supplied private data and operations for struct dma_buf_ops. On either missing ops, or error in allocating struct dma_buf, will return negative error.

For most cases the easiest way to create exp_info is through the DEFINE_DMA_BUF_EXPORT_INFO macro.

int dma_buf_fd(struct dma_buf *dmabuf, int flags)¶: returns a file descriptor for the given struct dma_buf

Parameters

struct dma_buf *dmabuf: [in] pointer to dma_buf for which fd is required.
int flags: [in] flags to give to fd

Description

On success, returns an associated 'fd'. Else, returns error.

struct dma_buf *dma_buf_get(int fd)¶: returns the struct dma_buf related to an fd

Parameters

int fd: [in] fd associated with the struct dma_buf to be returned

Description

On success, returns the struct dma_buf associated with an fd; uses file's refcounting done by fget to increase refcount. returns ERR_PTR otherwise.

void dma_buf_put(struct dma_buf *dmabuf)¶: decreases refcount of the buffer

Parameters

struct dma_buf *dmabuf: [in] buffer to reduce refcount of

Description

Uses file's refcounting done implicitly by fput().

If, as a result of this call, the refcount becomes 0, the 'release' file operation related to this fd is called. It calls dma_buf_ops.release vfunc in turn, and frees the memory allocated for dmabuf when exported.

struct dma_buf_attachment *dma_buf_dynamic_attach(struct dma_buf *dmabuf, struct device *dev, const struct dma_buf_attach_ops *importer_ops, void *importer_priv)¶: Add the device to dma_buf's attachments list

Parameters

struct dma_buf *dmabuf: [in] buffer to attach device to.
struct device *dev: [in] device to be attached.
const struct dma_buf_attach_ops *importer_ops: [in] importer operations for the attachment
void *importer_priv: [in] importer private pointer for the attachment

Description

Returns struct dma_buf_attachment pointer for this attachment. Attachments must be cleaned up by calling dma_buf_detach().

Optionally this calls dma_buf_ops.attach to allow device-specific attach functionality.

A pointer to newly created dma_buf_attachment on success, or a negative error code wrapped into a pointer on failure.

Note that this can fail if the backing storage of dmabuf is in a place not accessible to dev, and cannot be moved to a more suitable place. This is indicated with the error code -EBUSY.

Return

struct dma_buf_attachment *dma_buf_attach(struct dma_buf *dmabuf, struct device *dev)¶: Wrapper for dma_buf_dynamic_attach

Parameters

struct dma_buf *dmabuf: [in] buffer to attach device to.
struct device *dev: [in] device to be attached.

Description

Wrapper to call dma_buf_dynamic_attach() for drivers which still use a static mapping.

void dma_buf_detach(struct dma_buf *dmabuf, struct dma_buf_attachment *attach)¶: Remove the given attachment from dmabuf's attachments list

Parameters

struct dma_buf *dmabuf: [in] buffer to detach from.
struct dma_buf_attachment *attach: [in] attachment to be detached; is free'd after this call.

Description

Clean up a device attachment obtained by calling dma_buf_attach().

Optionally this calls dma_buf_ops.detach for device-specific detach.

int dma_buf_pin(struct dma_buf_attachment *attach)¶: Lock down the DMA-buf

Parameters

struct dma_buf_attachment *attach: [in] attachment which should be pinned

Description

Only dynamic importers (who set up attach with dma_buf_dynamic_attach()) may call this, and only for limited use cases like scanout and not for temporary pin operations. It is not permitted to allow userspace to pin arbitrary amounts of buffers through this interface.

Buffers must be unpinned by calling dma_buf_unpin().

Return

0 on success, negative error code on failure.

void dma_buf_unpin(struct dma_buf_attachment *attach)¶: Unpin a DMA-buf

Parameters

struct dma_buf_attachment *attach: [in] attachment which should be unpinned

Description

This unpins a buffer pinned by dma_buf_pin() and allows the exporter to move any mapping of attach again and inform the importer through dma_buf_attach_ops.move_notify.

struct sg_table *dma_buf_map_attachment(struct dma_buf_attachment *attach, enum dma_data_direction direction)¶: Returns the scatterlist table of the attachment; mapped into _device_ address space. Is a wrapper for map_dma_buf() of the dma_buf_ops.

Parameters

struct dma_buf_attachment *attach: [in] attachment whose scatterlist is to be returned
enum dma_data_direction direction: [in] direction of DMA transfer

Description

Returns sg_table containing the scatterlist to be returned; returns ERR_PTR on error. May return -EINTR if it is interrupted by a signal.

On success, the DMA addresses and lengths in the returned scatterlist are PAGE_SIZE aligned.

A mapping must be unmapped by using dma_buf_unmap_attachment(). Note that the underlying backing storage is pinned for as long as a mapping exists, therefore users/importers should not hold onto a mapping for undue amounts of time.

Important: Dynamic importers must wait for the exclusive fence of the struct dma_resv attached to the DMA-BUF first.

struct sg_table *dma_buf_map_attachment_unlocked(struct dma_buf_attachment *attach, enum dma_data_direction direction)¶: Returns the scatterlist table of the attachment; mapped into _device_ address space. Is a wrapper for map_dma_buf() of the dma_buf_ops.

Parameters

struct dma_buf_attachment *attach: [in] attachment whose scatterlist is to be returned
enum dma_data_direction direction: [in] direction of DMA transfer

Description

Unlocked variant of dma_buf_map_attachment().

void dma_buf_unmap_attachment(struct dma_buf_attachment *attach, struct sg_table *sg_table, enum dma_data_direction direction)¶: unmaps and decreases usecount of the buffer;might deallocate the scatterlist associated. Is a wrapper for unmap_dma_buf() of dma_buf_ops.

Parameters

struct dma_buf_attachment *attach: [in] attachment to unmap buffer from
struct sg_table *sg_table: [in] scatterlist info of the buffer to unmap
enum dma_data_direction direction: [in] direction of DMA transfer

Description

This unmaps a DMA mapping for attached obtained by dma_buf_map_attachment().

void dma_buf_unmap_attachment_unlocked(struct dma_buf_attachment *attach, struct sg_table *sg_table, enum dma_data_direction direction)¶: unmaps and decreases usecount of the buffer;might deallocate the scatterlist associated. Is a wrapper for unmap_dma_buf() of dma_buf_ops.

Parameters

struct dma_buf_attachment *attach: [in] attachment to unmap buffer from
struct sg_table *sg_table: [in] scatterlist info of the buffer to unmap
enum dma_data_direction direction: [in] direction of DMA transfer

Description

Unlocked variant of dma_buf_unmap_attachment().

void dma_buf_move_notify(struct dma_buf *dmabuf)¶: notify attachments that DMA-buf is moving

Parameters

struct dma_buf *dmabuf: [in] buffer which is moving

Description

Informs all attachments that they need to destroy and recreate all their mappings.

int dma_buf_begin_cpu_access(struct dma_buf *dmabuf, enum dma_data_direction direction)¶: Must be called before accessing a dma_buf from the cpu in the kernel context. Calls begin_cpu_access to allow exporter-specific preparations. Coherency is only guaranteed in the specified range for the specified access direction.

Parameters

struct dma_buf *dmabuf: [in] buffer to prepare cpu access for.
enum dma_data_direction direction: [in] direction of access.

Description

After the cpu access is complete the caller should call dma_buf_end_cpu_access(). Only when cpu access is bracketed by both calls is it guaranteed to be coherent with other DMA access.

This function will also wait for any DMA transactions tracked through implicit synchronization in dma_buf.resv. For DMA transactions with explicit synchronization this function will only ensure cache coherency, callers must ensure synchronization with such DMA transactions on their own.

Can return negative error values, returns 0 on success.

int dma_buf_end_cpu_access(struct dma_buf *dmabuf, enum dma_data_direction direction)¶: Must be called after accessing a dma_buf from the cpu in the kernel context. Calls end_cpu_access to allow exporter-specific actions. Coherency is only guaranteed in the specified range for the specified access direction.

Parameters

struct dma_buf *dmabuf: [in] buffer to complete cpu access for.
enum dma_data_direction direction: [in] direction of access.

Description

This terminates CPU access started with dma_buf_begin_cpu_access().

Can return negative error values, returns 0 on success.

int dma_buf_mmap(struct dma_buf *dmabuf, struct vm_area_struct *vma, unsigned long pgoff)¶: Setup up a userspace mmap with the given vma

Parameters

struct dma_buf *dmabuf: [in] buffer that should back the vma
struct vm_area_struct *vma: [in] vma for the mmap
unsigned long pgoff: [in] offset in pages where this mmap should start within the dma-buf buffer.

Description

This function adjusts the passed in vma so that it points at the file of the dma_buf operation. It also adjusts the starting pgoff and does bounds checking on the size of the vma. Then it calls the exporters mmap function to set up the mapping.

Can return negative error values, returns 0 on success.

int dma_buf_vmap(struct dma_buf *dmabuf, struct iosys_map *map)¶: Create virtual mapping for the buffer object into kernel address space. Same restrictions as for vmap and friends apply.

Parameters

struct dma_buf *dmabuf: [in] buffer to vmap
struct iosys_map *map: [out] returns the vmap pointer

Description

This call may fail due to lack of virtual mapping address space. These calls are optional in drivers. The intended use for them is for mapping objects linear in kernel space for high use objects.

To ensure coherency users must call dma_buf_begin_cpu_access() and dma_buf_end_cpu_access() around any cpu access performed through this mapping.

Returns 0 on success, or a negative errno code otherwise.

int dma_buf_vmap_unlocked(struct dma_buf *dmabuf, struct iosys_map *map)¶: Create virtual mapping for the buffer object into kernel address space. Same restrictions as for vmap and friends apply.

Parameters

struct dma_buf *dmabuf: [in] buffer to vmap
struct iosys_map *map: [out] returns the vmap pointer

Description

Unlocked version of dma_buf_vmap()

Returns 0 on success, or a negative errno code otherwise.

void dma_buf_vunmap(struct dma_buf *dmabuf, struct iosys_map *map)¶: Unmap a vmap obtained by dma_buf_vmap.

Parameters

struct dma_buf *dmabuf: [in] buffer to vunmap
struct iosys_map *map: [in] vmap pointer to vunmap

void dma_buf_vunmap_unlocked(struct dma_buf *dmabuf, struct iosys_map *map)¶: Unmap a vmap obtained by dma_buf_vmap.

Parameters

struct dma_buf *dmabuf: [in] buffer to vunmap
struct iosys_map *map: [in] vmap pointer to vunmap

struct dma_buf_ops¶: operations possible on struct dma_buf

Definition:

struct dma_buf_ops {
    bool cache_sgt_mapping;
    int (*attach)(struct dma_buf *, struct dma_buf_attachment *);
    void (*detach)(struct dma_buf *, struct dma_buf_attachment *);
    int (*pin)(struct dma_buf_attachment *attach);
    void (*unpin)(struct dma_buf_attachment *attach);
    struct sg_table * (*map_dma_buf)(struct dma_buf_attachment *, enum dma_data_direction);
    void (*unmap_dma_buf)(struct dma_buf_attachment *,struct sg_table *, enum dma_data_direction);
    void (*release)(struct dma_buf *);
    int (*begin_cpu_access)(struct dma_buf *, enum dma_data_direction);
    int (*end_cpu_access)(struct dma_buf *, enum dma_data_direction);
    int (*mmap)(struct dma_buf *, struct vm_area_struct *vma);
    int (*vmap)(struct dma_buf *dmabuf, struct iosys_map *map);
    void (*vunmap)(struct dma_buf *dmabuf, struct iosys_map *map);
};

Members

cache_sgt_mapping

If true the framework will cache the first mapping made for each attachment. This avoids creating mappings for attachments multiple times.

attach

This is called from dma_buf_attach() to make sure that a given dma_buf_attachment.dev can access the provided dma_buf. Exporters which support buffer objects in special locations like VRAM or device-specific carveout areas should check whether the buffer could be move to system memory (or directly accessed by the provided device), and otherwise need to fail the attach operation.

The exporter should also in general check whether the current allocation fulfills the DMA constraints of the new device. If this is not the case, and the allocation cannot be moved, it should also fail the attach operation.

Any exporter-private housekeeping data can be stored in the dma_buf_attachment.priv pointer.

This callback is optional.

Returns:

0 on success, negative error code on failure. It might return -EBUSY to signal that backing storage is already allocated and incompatible with the requirements of requesting device.

detach

This is called by dma_buf_detach() to release a dma_buf_attachment. Provided so that exporters can clean up any housekeeping for an dma_buf_attachment.

This callback is optional.

pin

This is called by dma_buf_pin() and lets the exporter know that the DMA-buf can't be moved any more. Ideally, the exporter should pin the buffer so that it is generally accessible by all devices.

This is called with the dmabuf.resv object locked and is mutual exclusive with cache_sgt_mapping.

This is called automatically for non-dynamic importers from dma_buf_attach().

Note that similar to non-dynamic exporters in their map_dma_buf callback the driver must guarantee that the memory is available for use and cleared of any old data by the time this function returns. Drivers which pipeline their buffer moves internally must wait for all moves and clears to complete.

Returns:

0 on success, negative error code on failure.

unpin

This is called by dma_buf_unpin() and lets the exporter know that the DMA-buf can be moved again.

This is called with the dmabuf->resv object locked and is mutual exclusive with cache_sgt_mapping.

This callback is optional.

map_dma_buf

This is called by dma_buf_map_attachment() and is used to map a shared dma_buf into device address space, and it is mandatory. It can only be called if attach has been called successfully.

This call may sleep, e.g. when the backing storage first needs to be allocated, or moved to a location suitable for all currently attached devices.

Note that any specific buffer attributes required for this function should get added to device_dma_parameters accessible via device.dma_params from the dma_buf_attachment. The attach callback should also check these constraints.

If this is being called for the first time, the exporter can now choose to scan through the list of attachments for this buffer, collate the requirements of the attached devices, and choose an appropriate backing storage for the buffer.

Based on enum dma_data_direction, it might be possible to have multiple users accessing at the same time (for reading, maybe), or any other kind of sharing that the exporter might wish to make available to buffer-users.

This is always called with the dmabuf->resv object locked when the dynamic_mapping flag is true.

Note that for non-dynamic exporters the driver must guarantee that that the memory is available for use and cleared of any old data by the time this function returns. Drivers which pipeline their buffer moves internally must wait for all moves and clears to complete. Dynamic exporters do not need to follow this rule: For non-dynamic importers the buffer is already pinned through pin, which has the same requirements. Dynamic importers otoh are required to obey the dma_resv fences.

Returns:

A sg_table scatter list of the backing storage of the DMA buffer, already mapped into the device address space of the device attached with the provided dma_buf_attachment. The addresses and lengths in the scatter list are PAGE_SIZE aligned.

On failure, returns a negative error value wrapped into a pointer. May also return -EINTR when a signal was received while being blocked.

Note that exporters should not try to cache the scatter list, or return the same one for multiple calls. Caching is done either by the DMA-BUF code (for non-dynamic importers) or the importer. Ownership of the scatter list is transferred to the caller, and returned by unmap_dma_buf.

unmap_dma_buf

This is called by dma_buf_unmap_attachment() and should unmap and release the sg_table allocated in map_dma_buf, and it is mandatory. For static dma_buf handling this might also unpin the backing storage if this is the last mapping of the DMA buffer.

release

Called after the last dma_buf_put to release the dma_buf, and mandatory.

begin_cpu_access

This is called from dma_buf_begin_cpu_access() and allows the exporter to ensure that the memory is actually coherent for cpu access. The exporter also needs to ensure that cpu access is coherent for the access direction. The direction can be used by the exporter to optimize the cache flushing, i.e. access with a different direction (read instead of write) might return stale or even bogus data (e.g. when the exporter needs to copy the data to temporary storage).

Note that this is both called through the DMA_BUF_IOCTL_SYNC IOCTL command for userspace mappings established through mmap, and also for kernel mappings established with vmap.

This callback is optional.

Returns:

0 on success or a negative error code on failure. This can for example fail when the backing storage can't be allocated. Can also return -ERESTARTSYS or -EINTR when the call has been interrupted and needs to be restarted.

end_cpu_access

This is called from dma_buf_end_cpu_access() when the importer is done accessing the CPU. The exporter can use this to flush caches and undo anything else done in begin_cpu_access.

This callback is optional.

Returns:

0 on success or a negative error code on failure. Can return -ERESTARTSYS or -EINTR when the call has been interrupted and needs to be restarted.

mmap

This callback is used by the dma_buf_mmap() function

Note that the mapping needs to be incoherent, userspace is expected to bracket CPU access using the DMA_BUF_IOCTL_SYNC interface.

Because dma-buf buffers have invariant size over their lifetime, the dma-buf core checks whether a vma is too large and rejects such mappings. The exporter hence does not need to duplicate this check. Drivers do not need to check this themselves.

If an exporter needs to manually flush caches and hence needs to fake coherency for mmap support, it needs to be able to zap all the ptes pointing at the backing storage. Now linux mm needs a struct address_space associated with the struct file stored in vma->vm_file to do that with the function unmap_mapping_range. But the dma_buf framework only backs every dma_buf fd with the anon_file struct file, i.e. all dma_bufs share the same file.

Hence exporters need to setup their own file (and address_space) association by setting vma->vm_file and adjusting vma->vm_pgoff in the dma_buf mmap callback. In the specific case of a gem driver the exporter could use the shmem file already provided by gem (and set vm_pgoff = 0). Exporters can then zap ptes by unmapping the corresponding range of the struct address_space associated with their own file.

This callback is optional.

Returns:

0 on success or a negative error code on failure.

vmap

[optional] creates a virtual mapping for the buffer into kernel address space. Same restrictions as for vmap and friends apply.

vunmap

[optional] unmaps a vmap from the buffer

struct dma_buf¶: shared buffer object

Definition:

struct dma_buf {
    size_t size;
    struct file *file;
    struct list_head attachments;
    const struct dma_buf_ops *ops;
    unsigned vmapping_counter;
    struct iosys_map vmap_ptr;
    const char *exp_name;
    const char *name;
    spinlock_t name_lock;
    struct module *owner;
    struct list_head list_node;
    void *priv;
    struct dma_resv *resv;
    wait_queue_head_t poll;
    struct dma_buf_poll_cb_t {
        struct dma_fence_cb cb;
        wait_queue_head_t *poll;
        __poll_t active;
    } cb_in, cb_out;
#ifdef CONFIG_DMABUF_SYSFS_STATS;
    struct dma_buf_sysfs_entry {
        struct kobject kobj;
        struct dma_buf *dmabuf;
    } *sysfs_entry;
#endif;
};

Members

size

Size of the buffer; invariant over the lifetime of the buffer.

file

File pointer used for sharing buffers across, and for refcounting. See dma_buf_get() and dma_buf_put().

attachments

List of dma_buf_attachment that denotes all devices attached, protected by dma_resv lock resv.

ops

dma_buf_ops associated with this buffer object.

vmapping_counter

Used internally to refcnt the vmaps returned by dma_buf_vmap(). Protected by lock.

vmap_ptr

The current vmap ptr if vmapping_counter > 0. Protected by lock.

exp_name

Name of the exporter; useful for debugging. See the DMA_BUF_SET_NAME IOCTL.

name

Userspace-provided name; useful for accounting and debugging, protected by dma_resv_lock() on resv and name_lock for read access.

name_lock

Spinlock to protect name access for read access.

owner

Pointer to exporter module; used for refcounting when exporter is a kernel module.

list_node

node for dma_buf accounting and debugging.

priv

exporter specific private data for this buffer object.

resv

Reservation object linked to this dma-buf.

IMPLICIT SYNCHRONIZATION RULES:

Drivers which support implicit synchronization of buffer access as e.g. exposed in Implicit Fence Poll Support must follow the below rules.

Drivers must add a read fence through dma_resv_add_fence() with the DMA_RESV_USAGE_READ flag for anything the userspace API considers a read access. This highly depends upon the API and window system.
Similarly drivers must add a write fence through dma_resv_add_fence() with the DMA_RESV_USAGE_WRITE flag for anything the userspace API considers write access.
Drivers may just always add a write fence, since that only causes unnecessary synchronization, but no correctness issues.
Some drivers only expose a synchronous userspace API with no pipelining across drivers. These do not set any fences for their access. An example here is v4l.
Driver should use dma_resv_usage_rw() when retrieving fences as dependency for implicit synchronization.

DYNAMIC IMPORTER RULES:

Dynamic importers, see dma_buf_attachment_is_dynamic(), have additional constraints on how they set up fences:

Dynamic importers must obey the write fences and wait for them to signal before allowing access to the buffer's underlying storage through the device.
Dynamic importers should set fences for any access that they can't disable immediately from their dma_buf_attach_ops.move_notify callback.

IMPORTANT:

All drivers and memory management related functions must obey the struct dma_resv rules, specifically the rules for updating and obeying fences. See enum dma_resv_usage for further descriptions.

poll

for userspace poll support

cb_in

for userspace poll support

cb_out

for userspace poll support

sysfs_entry

For exposing information about this buffer in sysfs. See also DMA-BUF statistics for the uapi this enables.

Description

This represents a shared buffer, created by calling dma_buf_export(). The userspace representation is a normal file descriptor, which can be created by calling dma_buf_fd().

Shared dma buffers are reference counted using dma_buf_put() and get_dma_buf().

Device DMA access is handled by the separate struct dma_buf_attachment.

struct dma_buf_attach_ops¶: importer operations for an attachment

Definition:

struct dma_buf_attach_ops {
    bool allow_peer2peer;
    void (*move_notify)(struct dma_buf_attachment *attach);
};

Members

allow_peer2peer

If this is set to true the importer must be able to handle peer resources without struct pages.

move_notify

[optional] notification that the DMA-buf is moving

If this callback is provided the framework can avoid pinning the backing store while mappings exists.

This callback is called with the lock of the reservation object associated with the dma_buf held and the mapping function must be called with this lock held as well. This makes sure that no mapping is created concurrently with an ongoing move operation.

Mappings stay valid and are not directly affected by this callback. But the DMA-buf can now be in a different physical location, so all mappings should be destroyed and re-created as soon as possible.

New mappings can be created after this callback returns, and will point to the new location of the DMA-buf.

Description

Attachment operations implemented by the importer.

struct dma_buf_attachment¶: holds device-buffer attachment data

Definition:

struct dma_buf_attachment {
    struct dma_buf *dmabuf;
    struct device *dev;
    struct list_head node;
    struct sg_table *sgt;
    enum dma_data_direction dir;
    bool peer2peer;
    const struct dma_buf_attach_ops *importer_ops;
    void *importer_priv;
    void *priv;
};

Members

dmabuf: buffer for this attachment.
dev: device attached to the buffer.
node: list of dma_buf_attachment, protected by dma_resv lock of the dmabuf.
sgt: cached mapping.
dir: direction of cached mapping.
peer2peer: true if the importer can handle peer resources without pages.
importer_ops: importer operations for this attachment, if provided dma_buf_map/unmap_attachment() must be called with the dma_resv lock held.
importer_priv: importer specific attachment data.
priv: exporter specific attachment data.

Description

This structure holds the attachment information between the dma_buf buffer and its user device(s). The list contains one attachment struct per device attached to the buffer.

An attachment is created by calling dma_buf_attach(), and released again by calling dma_buf_detach(). The DMA mapping itself needed to initiate a transfer is created by dma_buf_map_attachment() and freed again by calling dma_buf_unmap_attachment().

struct dma_buf_export_info¶: holds information needed to export a dma_buf

Definition:

struct dma_buf_export_info {
    const char *exp_name;
    struct module *owner;
    const struct dma_buf_ops *ops;
    size_t size;
    int flags;
    struct dma_resv *resv;
    void *priv;
};

Members

exp_name: name of the exporter - useful for debugging.
owner: pointer to exporter module - used for refcounting kernel module
ops: Attach allocator-defined dma buf ops to the new buffer
size: Size of the buffer - invariant over the lifetime of the buffer
flags: mode flags for the file
resv: reservation-object, NULL to allocate default one
priv: Attach private data of allocator to this buffer

Description

This structure holds the information required to export the buffer. Used with dma_buf_export() only.

DEFINE_DMA_BUF_EXPORT_INFO¶

DEFINE_DMA_BUF_EXPORT_INFO (name)

helper macro for exporters

Parameters

name: export-info name

Description

DEFINE_DMA_BUF_EXPORT_INFO macro defines the struct dma_buf_export_info, zeroes it out and pre-populates exp_name in it.

void get_dma_buf(struct dma_buf *dmabuf)¶: convenience wrapper for get_file.

Parameters

struct dma_buf *dmabuf: [in] pointer to dma_buf

Description

Increments the reference count on the dma-buf, needed in case of drivers that either need to create additional references to the dmabuf on the kernel side. For example, an exporter that needs to keep a dmabuf ptr so that subsequent exports don't create a new dmabuf.

bool dma_buf_is_dynamic(struct dma_buf *dmabuf)¶: check if a DMA-buf uses dynamic mappings.

Parameters

struct dma_buf *dmabuf: the DMA-buf to check

Description

Returns true if a DMA-buf exporter wants to be called with the dma_resv locked for the map/unmap callbacks, false if it doesn't wants to be called with the lock held.

bool dma_buf_attachment_is_dynamic(struct dma_buf_attachment *attach)¶: check if a DMA-buf attachment uses dynamic mappings

Parameters

struct dma_buf_attachment *attach: the DMA-buf attachment to check

Description

Returns true if a DMA-buf importer wants to call the map/unmap functions with the dma_resv lock held.

Reservation Objects¶

The reservation object provides a mechanism to manage a container of dma_fence object associated with a resource. A reservation object can have any number of fences attaches to it. Each fence carries an usage parameter determining how the operation represented by the fence is using the resource. The RCU mechanism is used to protect read access to fences from locked write-side updates.

See struct dma_resv for more details.

void dma_resv_init(struct dma_resv *obj)¶: initialize a reservation object

Parameters

struct dma_resv *obj: the reservation object

void dma_resv_fini(struct dma_resv *obj)¶: destroys a reservation object

Parameters

struct dma_resv *obj: the reservation object

int dma_resv_reserve_fences(struct dma_resv *obj, unsigned int num_fences)¶: Reserve space to add fences to a dma_resv object.

Parameters

struct dma_resv *obj: reservation object
unsigned int num_fences: number of fences we want to add

Description

Should be called before dma_resv_add_fence(). Must be called with obj locked through dma_resv_lock().

Note that the preallocated slots need to be re-reserved if obj is unlocked at any time before calling dma_resv_add_fence(). This is validated when CONFIG_DEBUG_MUTEXES is enabled.

RETURNS Zero for success, or -errno

void dma_resv_reset_max_fences(struct dma_resv *obj)¶: reset fences for debugging

Parameters

struct dma_resv *obj: the dma_resv object to reset

Description

Reset the number of pre-reserved fence slots to test that drivers do correct slot allocation using dma_resv_reserve_fences(). See also dma_resv_list.max_fences.

void dma_resv_add_fence(struct dma_resv *obj, struct dma_fence *fence, enum dma_resv_usage usage)¶: Add a fence to the dma_resv obj

Parameters

struct dma_resv *obj: the reservation object
struct dma_fence *fence: the fence to add
enum dma_resv_usage usage: how the fence is used, see enum dma_resv_usage

Description

Add a fence to a slot, obj must be locked with dma_resv_lock(), and dma_resv_reserve_fences() has been called.

See also dma_resv.fence for a discussion of the semantics.

void dma_resv_replace_fences(struct dma_resv *obj, uint64_t context, struct dma_fence *replacement, enum dma_resv_usage usage)¶: replace fences in the dma_resv obj

Parameters

struct dma_resv *obj: the reservation object
uint64_t context: the context of the fences to replace
struct dma_fence *replacement: the new fence to use instead
enum dma_resv_usage usage: how the new fence is used, see enum dma_resv_usage

Description

Replace fences with a specified context with a new fence. Only valid if the operation represented by the original fence has no longer access to the resources represented by the dma_resv object when the new fence completes.

And example for using this is replacing a preemption fence with a page table update fence which makes the resource inaccessible.

struct dma_fence *dma_resv_iter_first_unlocked(struct dma_resv_iter *cursor)¶: first fence in an unlocked dma_resv obj.

Parameters

struct dma_resv_iter *cursor: the cursor with the current position

Description

Subsequent fences are iterated with dma_resv_iter_next_unlocked().

Beware that the iterator can be restarted. Code which accumulates statistics or similar needs to check for this with dma_resv_iter_is_restarted(). For this reason prefer the locked dma_resv_iter_first() whenver possible.

Returns the first fence from an unlocked dma_resv obj.

struct dma_fence *dma_resv_iter_next_unlocked(struct dma_resv_iter *cursor)¶: next fence in an unlocked dma_resv obj.

Parameters

struct dma_resv_iter *cursor: the cursor with the current position

Description

Beware that the iterator can be restarted. Code which accumulates statistics or similar needs to check for this with dma_resv_iter_is_restarted(). For this reason prefer the locked dma_resv_iter_next() whenver possible.

Returns the next fence from an unlocked dma_resv obj.

struct dma_fence *dma_resv_iter_first(struct dma_resv_iter *cursor)¶: first fence from a locked dma_resv object

Parameters

struct dma_resv_iter *cursor: cursor to record the current position

Description

Subsequent fences are iterated with dma_resv_iter_next_unlocked().

Return the first fence in the dma_resv object while holding the dma_resv.lock.

struct dma_fence *dma_resv_iter_next(struct dma_resv_iter *cursor)¶: next fence from a locked dma_resv object

Parameters

struct dma_resv_iter *cursor: cursor to record the current position

Description

Return the next fences from the dma_resv object while holding the dma_resv.lock.

int dma_resv_copy_fences(struct dma_resv *dst, struct dma_resv *src)¶: Copy all fences from src to dst.

Parameters

struct dma_resv *dst: the destination reservation object
struct dma_resv *src: the source reservation object

Description

Copy all fences from src to dst. dst-lock must be held.

int dma_resv_get_fences(struct dma_resv *obj, enum dma_resv_usage usage, unsigned int *num_fences, struct dma_fence ***fences)¶: Get an object's fences fences without update side lock held

Parameters

struct dma_resv *obj: the reservation object
enum dma_resv_usage usage: controls which fences to include, see enum dma_resv_usage.
unsigned int *num_fences: the number of fences returned
struct dma_fence ***fences: the array of fence ptrs returned (array is krealloc'd to the required size, and must be freed by caller)

Description

Retrieve all fences from the reservation object. Returns either zero or -ENOMEM.

int dma_resv_get_singleton(struct dma_resv *obj, enum dma_resv_usage usage, struct dma_fence **fence)¶: Get a single fence for all the fences

Parameters

struct dma_resv *obj: the reservation object
enum dma_resv_usage usage: controls which fences to include, see enum dma_resv_usage.
struct dma_fence **fence: the resulting fence

Description

Get a single fence representing all the fences inside the resv object. Returns either 0 for success or -ENOMEM.

Warning: This can't be used like this when adding the fence back to the resv object since that can lead to stack corruption when finalizing the dma_fence_array.

Returns 0 on success and negative error values on failure.

long dma_resv_wait_timeout(struct dma_resv *obj, enum dma_resv_usage usage, bool intr, unsigned long timeout)¶: Wait on reservation's objects fences

Parameters

struct dma_resv *obj: the reservation object
enum dma_resv_usage usage: controls which fences to include, see enum dma_resv_usage.
bool intr: if true, do interruptible wait
unsigned long timeout: timeout value in jiffies or zero to return immediately

Description

Callers are not required to hold specific locks, but maybe hold dma_resv_lock() already RETURNS Returns -ERESTARTSYS if interrupted, 0 if the wait timed out, or greater than zero on success.

void dma_resv_set_deadline(struct dma_resv *obj, enum dma_resv_usage usage, ktime_t deadline)¶: Set a deadline on reservation's objects fences

Parameters

struct dma_resv *obj: the reservation object
enum dma_resv_usage usage: controls which fences to include, see enum dma_resv_usage.
ktime_t deadline: the requested deadline (MONOTONIC)

Description

May be called without holding the dma_resv lock. Sets deadline on all fences filtered by usage.

bool dma_resv_test_signaled(struct dma_resv *obj, enum dma_resv_usage usage)¶: Test if a reservation object's fences have been signaled.

Parameters

struct dma_resv *obj: the reservation object
enum dma_resv_usage usage: controls which fences to include, see enum dma_resv_usage.

Description

Callers are not required to hold specific locks, but maybe hold dma_resv_lock() already.

RETURNS

True if all fences signaled, else false.

void dma_resv_describe(struct dma_resv *obj, struct seq_file *seq)¶: Dump description of the resv object into seq_file

Parameters

struct dma_resv *obj: the reservation object
struct seq_file *seq: the seq_file to dump the description into

Description

Dump a textual description of the fences inside an dma_resv object into the seq_file.

enum dma_resv_usage¶: how the fences from a dma_resv obj are used

Constants

DMA_RESV_USAGE_KERNEL

For in kernel memory management only.

This should only be used for things like copying or clearing memory with a DMA hardware engine for the purpose of kernel memory management.

Drivers always must wait for those fences before accessing the resource protected by the dma_resv object. The only exception for that is when the resource is known to be locked down in place by pinning it previously.

DMA_RESV_USAGE_WRITE

Implicit write synchronization.

This should only be used for userspace command submissions which add an implicit write dependency.

DMA_RESV_USAGE_READ

Implicit read synchronization.

This should only be used for userspace command submissions which add an implicit read dependency.

DMA_RESV_USAGE_BOOKKEEP

No implicit sync.

This should be used by submissions which don't want to participate in any implicit synchronization.

The most common case are preemption fences, page table updates, TLB flushes as well as explicit synced user submissions.

Explicit synced user user submissions can be promoted to DMA_RESV_USAGE_READ or DMA_RESV_USAGE_WRITE as needed using dma_buf_import_sync_file() when implicit synchronization should become necessary after initial adding of the fence.

Description

This enum describes the different use cases for a dma_resv object and controls which fences are returned when queried.

An important fact is that there is the order KERNEL<WRITE<READ<BOOKKEEP and when the dma_resv object is asked for fences for one use case the fences for the lower use case are returned as well.

For example when asking for WRITE fences then the KERNEL fences are returned as well. Similar when asked for READ fences then both WRITE and KERNEL fences are returned as well.

Already used fences can be promoted in the sense that a fence with DMA_RESV_USAGE_BOOKKEEP could become DMA_RESV_USAGE_READ by adding it again with this usage. But fences can never be degraded in the sense that a fence with DMA_RESV_USAGE_WRITE could become DMA_RESV_USAGE_READ.

enum dma_resv_usage dma_resv_usage_rw(bool write)¶: helper for implicit sync

Parameters

bool write: true if we create a new implicit sync write

Description

This returns the implicit synchronization usage for write or read accesses, see enum dma_resv_usage and dma_buf.resv.

struct dma_resv¶: a reservation object manages fences for a buffer

Definition:

struct dma_resv {
    struct ww_mutex lock;
    struct dma_resv_list __rcu *fences;
};

Members

lock

Update side lock. Don't use directly, instead use the wrapper functions like dma_resv_lock() and dma_resv_unlock().

Drivers which use the reservation object to manage memory dynamically also use this lock to protect buffer object state like placement, allocation policies or throughout command submission.

fences

Array of fences which where added to the dma_resv object

A new fence is added by calling dma_resv_add_fence(). Since this often needs to be done past the point of no return in command submission it cannot fail, and therefore sufficient slots need to be reserved by calling dma_resv_reserve_fences().

Description

This is a container for dma_fence objects which needs to handle multiple use cases.

One use is to synchronize cross-driver access to a struct dma_buf, either for dynamic buffer management or just to handle implicit synchronization between different users of the buffer in userspace. See dma_buf.resv for a more in-depth discussion.

The other major use is to manage access and locking within a driver in a buffer based memory manager. struct ttm_buffer_object is the canonical example here, since this is where reservation objects originated from. But use in drivers is spreading and some drivers also manage struct drm_gem_object with the same scheme.

struct dma_resv_iter¶: current position into the dma_resv fences

Definition:

struct dma_resv_iter {
    struct dma_resv *obj;
    enum dma_resv_usage usage;
    struct dma_fence *fence;
    enum dma_resv_usage fence_usage;
    unsigned int index;
    struct dma_resv_list *fences;
    unsigned int num_fences;
    bool is_restarted;
};

Members

obj: The dma_resv object we iterate over
usage: Return fences with this usage or lower.
fence: the currently handled fence
fence_usage: the usage of the current fence
index: index into the shared fences
fences: the shared fences; private, MUST not dereference
num_fences: number of fences
is_restarted: true if this is the first returned fence

Description

Don't touch this directly in the driver, use the accessor function instead.

IMPORTANT

When using the lockless iterators like dma_resv_iter_next_unlocked() or dma_resv_for_each_fence_unlocked() beware that the iterator can be restarted. Code which accumulates statistics or similar needs to check for this with dma_resv_iter_is_restarted().

void dma_resv_iter_begin(struct dma_resv_iter *cursor, struct dma_resv *obj, enum dma_resv_usage usage)¶: initialize a dma_resv_iter object

Parameters

struct dma_resv_iter *cursor: The dma_resv_iter object to initialize
struct dma_resv *obj: The dma_resv object which we want to iterate over
enum dma_resv_usage usage: controls which fences to include, see enum dma_resv_usage.

void dma_resv_iter_end(struct dma_resv_iter *cursor)¶: cleanup a dma_resv_iter object

Parameters

struct dma_resv_iter *cursor: the dma_resv_iter object which should be cleaned up

Description

Make sure that the reference to the fence in the cursor is properly dropped.

enum dma_resv_usage dma_resv_iter_usage(struct dma_resv_iter *cursor)¶: Return the usage of the current fence

Parameters

struct dma_resv_iter *cursor: the cursor of the current position

Description

Returns the usage of the currently processed fence.

bool dma_resv_iter_is_restarted(struct dma_resv_iter *cursor)¶: test if this is the first fence after a restart

Parameters

struct dma_resv_iter *cursor: the cursor with the current position

Description

Return true if this is the first fence in an iteration after a restart.

dma_resv_for_each_fence_unlocked¶

dma_resv_for_each_fence_unlocked (cursor, fence)

unlocked fence iterator

Parameters

cursor: a struct dma_resv_iter pointer
fence: the current fence

Description

Iterate over the fences in a struct dma_resv object without holding the dma_resv.lock and using RCU instead. The cursor needs to be initialized with dma_resv_iter_begin() and cleaned up with dma_resv_iter_end(). Inside the iterator a reference to the dma_fence is held and the RCU lock dropped.

Beware that the iterator can be restarted when the struct dma_resv for cursor is modified. Code which accumulates statistics or similar needs to check for this with dma_resv_iter_is_restarted(). For this reason prefer the lock iterator dma_resv_for_each_fence() whenever possible.

dma_resv_for_each_fence¶

dma_resv_for_each_fence (cursor, obj, usage, fence)

fence iterator

Parameters

cursor: a struct dma_resv_iter pointer
obj: a dma_resv object pointer
usage: controls which fences to return
fence: the current fence

Description

Iterate over the fences in a struct dma_resv object while holding the dma_resv.lock. all_fences controls if the shared fences are returned as well. The cursor initialisation is part of the iterator and the fence stays valid as long as the lock is held and so no extra reference to the fence is taken.

int dma_resv_lock(struct dma_resv *obj, struct ww_acquire_ctx *ctx)¶: lock the reservation object

Parameters

struct dma_resv *obj: the reservation object
struct ww_acquire_ctx *ctx: the locking context

Description

Locks the reservation object for exclusive access and modification. Note, that the lock is only against other writers, readers will run concurrently with a writer under RCU. The seqlock is used to notify readers if they overlap with a writer.

As the reservation object may be locked by multiple parties in an undefined order, a #ww_acquire_ctx is passed to unwind if a cycle is detected. See ww_mutex_lock() and ww_acquire_init(). A reservation object may be locked by itself by passing NULL as ctx.

When a die situation is indicated by returning -EDEADLK all locks held by ctx must be unlocked and then dma_resv_lock_slow() called on obj.

Unlocked by calling dma_resv_unlock().

See also dma_resv_lock_interruptible() for the interruptible variant.

int dma_resv_lock_interruptible(struct dma_resv *obj, struct ww_acquire_ctx *ctx)¶: lock the reservation object

Parameters

struct dma_resv *obj: the reservation object
struct ww_acquire_ctx *ctx: the locking context

Description

Locks the reservation object interruptible for exclusive access and modification. Note, that the lock is only against other writers, readers will run concurrently with a writer under RCU. The seqlock is used to notify readers if they overlap with a writer.

As the reservation object may be locked by multiple parties in an undefined order, a #ww_acquire_ctx is passed to unwind if a cycle is detected. See ww_mutex_lock() and ww_acquire_init(). A reservation object may be locked by itself by passing NULL as ctx.

When a die situation is indicated by returning -EDEADLK all locks held by ctx must be unlocked and then dma_resv_lock_slow_interruptible() called on obj.

Unlocked by calling dma_resv_unlock().

void dma_resv_lock_slow(struct dma_resv *obj, struct ww_acquire_ctx *ctx)¶: slowpath lock the reservation object

Parameters

struct dma_resv *obj: the reservation object
struct ww_acquire_ctx *ctx: the locking context

Description

Acquires the reservation object after a die case. This function will sleep until the lock becomes available. See dma_resv_lock() as well.

See also dma_resv_lock_slow_interruptible() for the interruptible variant.

int dma_resv_lock_slow_interruptible(struct dma_resv *obj, struct ww_acquire_ctx *ctx)¶: slowpath lock the reservation object, interruptible

Parameters

struct dma_resv *obj: the reservation object
struct ww_acquire_ctx *ctx: the locking context

Description

Acquires the reservation object interruptible after a die case. This function will sleep until the lock becomes available. See dma_resv_lock_interruptible() as well.

bool dma_resv_trylock(struct dma_resv *obj)¶: trylock the reservation object

Parameters

struct dma_resv *obj: the reservation object

Description

Tries to lock the reservation object for exclusive access and modification. Note, that the lock is only against other writers, readers will run concurrently with a writer under RCU. The seqlock is used to notify readers if they overlap with a writer.

Also note that since no context is provided, no deadlock protection is possible, which is also not needed for a trylock.

Returns true if the lock was acquired, false otherwise.

bool dma_resv_is_locked(struct dma_resv *obj)¶: is the reservation object locked

Parameters

struct dma_resv *obj: the reservation object

Description

Returns true if the mutex is locked, false if unlocked.

struct ww_acquire_ctx *dma_resv_locking_ctx(struct dma_resv *obj)¶: returns the context used to lock the object

Parameters

struct dma_resv *obj: the reservation object

Description

Returns the context used to lock a reservation object or NULL if no context was used or the object is not locked at all.

WARNING: This interface is pretty horrible, but TTM needs it because it doesn't pass the struct ww_acquire_ctx around in some very long callchains. Everyone else just uses it to check whether they're holding a reservation or not.

void dma_resv_unlock(struct dma_resv *obj)¶: unlock the reservation object

Parameters

struct dma_resv *obj: the reservation object

Description

Unlocks the reservation object following exclusive access.

DMA Fences¶

DMA fences, represented by struct dma_fence, are the kernel internal synchronization primitive for DMA operations like GPU rendering, video encoding/decoding, or displaying buffers on a screen.

A fence is initialized using dma_fence_init() and completed using dma_fence_signal(). Fences are associated with a context, allocated through dma_fence_context_alloc(), and all fences on the same context are fully ordered.

Since the purposes of fences is to facilitate cross-device and cross-application synchronization, there's multiple ways to use one:

Individual fences can be exposed as a sync_file, accessed as a file descriptor from userspace, created by calling sync_file_create(). This is called explicit fencing, since userspace passes around explicit synchronization points.
Some subsystems also have their own explicit fencing primitives, like drm_syncobj. Compared to sync_file, a drm_syncobj allows the underlying fence to be updated.
Then there's also implicit fencing, where the synchronization points are implicitly passed around as part of shared dma_buf instances. Such implicit fences are stored in struct dma_resv through the dma_buf.resv pointer.

DMA Fence Cross-Driver Contract¶

Since dma_fence provide a cross driver contract, all drivers must follow the same rules:

Fences must complete in a reasonable time. Fences which represent kernels and shaders submitted by userspace, which could run forever, must be backed up by timeout and gpu hang recovery code. Minimally that code must prevent further command submission and force complete all in-flight fences, e.g. when the driver or hardware do not support gpu reset, or if the gpu reset failed for some reason. Ideally the driver supports gpu recovery which only affects the offending userspace context, and no other userspace submissions.
Drivers may have different ideas of what completion within a reasonable time means. Some hang recovery code uses a fixed timeout, others a mix between observing forward progress and increasingly strict timeouts. Drivers should not try to second guess timeout handling of fences from other drivers.
To ensure there's no deadlocks of dma_fence_wait() against other locks drivers should annotate all code required to reach dma_fence_signal(), which completes the fences, with dma_fence_begin_signalling() and dma_fence_end_signalling().
Drivers are allowed to call dma_fence_wait() while holding dma_resv_lock(). This means any code required for fence completion cannot acquire a dma_resv lock. Note that this also pulls in the entire established locking hierarchy around dma_resv_lock() and dma_resv_unlock().
Drivers are allowed to call dma_fence_wait() from their shrinker callbacks. This means any code required for fence completion cannot allocate memory with GFP_KERNEL.
Drivers are allowed to call dma_fence_wait() from their mmu_notifier respectively mmu_interval_notifier callbacks. This means any code required for fence completeion cannot allocate memory with GFP_NOFS or GFP_NOIO. Only GFP_ATOMIC is permissible, which might fail.

Note that only GPU drivers have a reasonable excuse for both requiring mmu_interval_notifier and shrinker callbacks at the same time as having to track asynchronous compute work using dma_fence. No driver outside of drivers/gpu should ever call dma_fence_wait() in such contexts.

DMA Fence Signalling Annotations¶

Proving correctness of all the kernel code around dma_fence through code review and testing is tricky for a few reasons:

It is a cross-driver contract, and therefore all drivers must follow the same rules for lock nesting order, calling contexts for various functions and anything else significant for in-kernel interfaces. But it is also impossible to test all drivers in a single machine, hence brute-force N vs. N testing of all combinations is impossible. Even just limiting to the possible combinations is infeasible.
There is an enormous amount of driver code involved. For render drivers there's the tail of command submission, after fences are published, scheduler code, interrupt and workers to process job completion, and timeout, gpu reset and gpu hang recovery code. Plus for integration with core mm with have mmu_notifier, respectively mmu_interval_notifier, and shrinker. For modesetting drivers there's the commit tail functions between when fences for an atomic modeset are published, and when the corresponding vblank completes, including any interrupt processing and related workers. Auditing all that code, across all drivers, is not feasible.
Due to how many other subsystems are involved and the locking hierarchies this pulls in there is extremely thin wiggle-room for driver-specific differences. dma_fence interacts with almost all of the core memory handling through page fault handlers via dma_resv, dma_resv_lock() and dma_resv_unlock(). On the other side it also interacts through all allocation sites through mmu_notifier and shrinker.

Furthermore lockdep does not handle cross-release dependencies, which means any deadlocks between dma_fence_wait() and dma_fence_signal() can't be caught at runtime with some quick testing. The simplest example is one thread waiting on a dma_fence while holding a lock:

lock(A);
dma_fence_wait(B);
unlock(A);

while the other thread is stuck trying to acquire the same lock, which prevents it from signalling the fence the previous thread is stuck waiting on:

lock(A);
unlock(A);
dma_fence_signal(B);

By manually annotating all code relevant to signalling a dma_fence we can teach lockdep about these dependencies, which also helps with the validation headache since now lockdep can check all the rules for us:

cookie = dma_fence_begin_signalling();
lock(A);
unlock(A);
dma_fence_signal(B);
dma_fence_end_signalling(cookie);

For using dma_fence_begin_signalling() and dma_fence_end_signalling() to annotate critical sections the following rules need to be observed:

All code necessary to complete a dma_fence must be annotated, from the point where a fence is accessible to other threads, to the point where dma_fence_signal() is called. Un-annotated code can contain deadlock issues, and due to the very strict rules and many corner cases it is infeasible to catch these just with review or normal stress testing.
struct dma_resv deserves a special note, since the readers are only protected by rcu. This means the signalling critical section starts as soon as the new fences are installed, even before dma_resv_unlock() is called.
The only exception are fast paths and opportunistic signalling code, which calls dma_fence_signal() purely as an optimization, but is not required to guarantee completion of a dma_fence. The usual example is a wait IOCTL which calls dma_fence_signal(), while the mandatory completion path goes through a hardware interrupt and possible job completion worker.
To aid composability of code, the annotations can be freely nested, as long as the overall locking hierarchy is consistent. The annotations also work both in interrupt and process context. Due to implementation details this requires that callers pass an opaque cookie from dma_fence_begin_signalling() to dma_fence_end_signalling().
Validation against the cross driver contract is implemented by priming lockdep with the relevant hierarchy at boot-up. This means even just testing with a single device is enough to validate a driver, at least as far as deadlocks with dma_fence_wait() against dma_fence_signal() are concerned.

DMA Fence Deadline Hints¶

In an ideal world, it would be possible to pipeline a workload sufficiently that a utilization based device frequency governor could arrive at a minimum frequency that meets the requirements of the use-case, in order to minimize power consumption. But in the real world there are many workloads which defy this ideal. For example, but not limited to:

Workloads that ping-pong between device and CPU, with alternating periods of CPU waiting for device, and device waiting on CPU. This can result in devfreq and cpufreq seeing idle time in their respective domains and in result reduce frequency.
Workloads that interact with a periodic time based deadline, such as double buffered GPU rendering vs vblank sync'd page flipping. In this scenario, missing a vblank deadline results in an increase in idle time on the GPU (since it has to wait an additional vblank period), sending a signal to the GPU's devfreq to reduce frequency, when in fact the opposite is what is needed.

To this end, deadline hint(s) can be set on a dma_fence via dma_fence_set_deadline. The deadline hint provides a way for the waiting driver, or userspace, to convey an appropriate sense of urgency to the signaling driver.

A deadline hint is given in absolute ktime (CLOCK_MONOTONIC for userspace facing APIs). The time could either be some point in the future (such as the vblank based deadline for page-flipping, or the start of a compositor's composition cycle), or the current time to indicate an immediate deadline hint (Ie. forward progress cannot be made until this fence is signaled).

Multiple deadlines may be set on a given fence, even in parallel. See the documentation for dma_fence_ops.set_deadline.

The deadline hint is just that, a hint. The driver that created the fence may react by increasing frequency, making different scheduling choices, etc. Or doing nothing at all.

DMA Fences Functions Reference¶

struct dma_fence *dma_fence_get_stub(void)¶: return a signaled fence

Parameters

void: no arguments

Description

Return a stub fence which is already signaled. The fence's timestamp corresponds to the first time after boot this function is called.

struct dma_fence *dma_fence_allocate_private_stub(ktime_t timestamp)¶: return a private, signaled fence

Parameters

ktime_t timestamp: timestamp when the fence was signaled

Description

Return a newly allocated and signaled stub fence.

u64 dma_fence_context_alloc(unsigned num)¶: allocate an array of fence contexts

Parameters

unsigned num: amount of contexts to allocate

Description

This function will return the first index of the number of fence contexts allocated. The fence context is used for setting dma_fence.context to a unique number by passing the context to dma_fence_init().

bool dma_fence_begin_signalling(void)¶: begin a critical DMA fence signalling section

Parameters

void: no arguments

Description

Drivers should use this to annotate the beginning of any code section required to eventually complete dma_fence by calling dma_fence_signal().

The end of these critical sections are annotated with dma_fence_end_signalling().

Opaque cookie needed by the implementation, which needs to be passed to dma_fence_end_signalling().

Return

void dma_fence_end_signalling(bool cookie)¶: end a critical DMA fence signalling section

Parameters

bool cookie: opaque cookie from dma_fence_begin_signalling()

Description

Closes a critical section annotation opened by dma_fence_begin_signalling().

int dma_fence_signal_timestamp_locked(struct dma_fence *fence, ktime_t timestamp)¶: signal completion of a fence

Parameters

struct dma_fence *fence: the fence to signal
ktime_t timestamp: fence signal timestamp in kernel's CLOCK_MONOTONIC time domain

Description

Signal completion for software callbacks on a fence, this will unblock dma_fence_wait() calls and run all the callbacks added with dma_fence_add_callback(). Can be called multiple times, but since a fence can only go from the unsignaled to the signaled state and not back, it will only be effective the first time. Set the timestamp provided as the fence signal timestamp.

Unlike dma_fence_signal_timestamp(), this function must be called with dma_fence.lock held.

Returns 0 on success and a negative error value when fence has been signalled already.

int dma_fence_signal_timestamp(struct dma_fence *fence, ktime_t timestamp)¶: signal completion of a fence

Parameters

struct dma_fence *fence: the fence to signal
ktime_t timestamp: fence signal timestamp in kernel's CLOCK_MONOTONIC time domain

Description

Signal completion for software callbacks on a fence, this will unblock dma_fence_wait() calls and run all the callbacks added with dma_fence_add_callback(). Can be called multiple times, but since a fence can only go from the unsignaled to the signaled state and not back, it will only be effective the first time. Set the timestamp provided as the fence signal timestamp.

Returns 0 on success and a negative error value when fence has been signalled already.

int dma_fence_signal_locked(struct dma_fence *fence)¶: signal completion of a fence

Parameters

struct dma_fence *fence: the fence to signal

Description

Signal completion for software callbacks on a fence, this will unblock dma_fence_wait() calls and run all the callbacks added with dma_fence_add_callback(). Can be called multiple times, but since a fence can only go from the unsignaled to the signaled state and not back, it will only be effective the first time.

Unlike dma_fence_signal(), this function must be called with dma_fence.lock held.

Returns 0 on success and a negative error value when fence has been signalled already.

int dma_fence_signal(struct dma_fence *fence)¶: signal completion of a fence

Parameters

struct dma_fence *fence: the fence to signal

Description

Signal completion for software callbacks on a fence, this will unblock dma_fence_wait() calls and run all the callbacks added with dma_fence_add_callback(). Can be called multiple times, but since a fence can only go from the unsignaled to the signaled state and not back, it will only be effective the first time.

Returns 0 on success and a negative error value when fence has been signalled already.

signed long dma_fence_wait_timeout(struct dma_fence *fence, bool intr, signed long timeout)¶: sleep until the fence gets signaled or until timeout elapses

Parameters

struct dma_fence *fence: the fence to wait on
bool intr: if true, do an interruptible wait
signed long timeout: timeout value in jiffies, or MAX_SCHEDULE_TIMEOUT

Description

Returns -ERESTARTSYS if interrupted, 0 if the wait timed out, or the remaining timeout in jiffies on success. Other error values may be returned on custom implementations.

Performs a synchronous wait on this fence. It is assumed the caller directly or indirectly (buf-mgr between reservation and committing) holds a reference to the fence, otherwise the fence might be freed before return, resulting in undefined behavior.

void dma_fence_release(struct kref *kref)¶: default relese function for fences

Parameters

struct kref *kref: dma_fence.recfount

Description

This is the default release functions for dma_fence. Drivers shouldn't call this directly, but instead call dma_fence_put().

void dma_fence_free(struct dma_fence *fence)¶: default release function for dma_fence.

Parameters

struct dma_fence *fence: fence to release

Description

This is the default implementation for dma_fence_ops.release. It calls kfree_rcu() on fence.

void dma_fence_enable_sw_signaling(struct dma_fence *fence)¶: enable signaling on fence

Parameters

struct dma_fence *fence: the fence to enable

Description

This will request for sw signaling to be enabled, to make the fence complete as soon as possible. This calls dma_fence_ops.enable_signaling internally.

int dma_fence_add_callback(struct dma_fence *fence, struct dma_fence_cb *cb, dma_fence_func_t func)¶: add a callback to be called when the fence is signaled

Parameters

struct dma_fence *fence: the fence to wait on
struct dma_fence_cb *cb: the callback to register
dma_fence_func_t func: the function to call

Description

Add a software callback to the fence. The caller should keep a reference to the fence.

cb will be initialized by dma_fence_add_callback(), no initialization by the caller is required. Any number of callbacks can be registered to a fence, but a callback can only be registered to one fence at a time.

If fence is already signaled, this function will return -ENOENT (and not call the callback).

Note that the callback can be called from an atomic context or irq context.

Returns 0 in case of success, -ENOENT if the fence is already signaled and -EINVAL in case of error.

int dma_fence_get_status(struct dma_fence *fence)¶: returns the status upon completion

Parameters

struct dma_fence *fence: the dma_fence to query

Description

This wraps dma_fence_get_status_locked() to return the error status condition on a signaled fence. See dma_fence_get_status_locked() for more details.

Returns 0 if the fence has not yet been signaled, 1 if the fence has been signaled without an error condition, or a negative error code if the fence has been completed in err.

bool dma_fence_remove_callback(struct dma_fence *fence, struct dma_fence_cb *cb)¶: remove a callback from the signaling list

Parameters

struct dma_fence *fence: the fence to wait on
struct dma_fence_cb *cb: the callback to remove

Description

Remove a previously queued callback from the fence. This function returns true if the callback is successfully removed, or false if the fence has already been signaled.

WARNING: Cancelling a callback should only be done if you really know what you're doing, since deadlocks and race conditions could occur all too easily. For this reason, it should only ever be done on hardware lockup recovery, with a reference held to the fence.

Behaviour is undefined if cb has not been added to fence using dma_fence_add_callback() beforehand.

signed long dma_fence_default_wait(struct dma_fence *fence, bool intr, signed long timeout)¶: default sleep until the fence gets signaled or until timeout elapses

Parameters

struct dma_fence *fence: the fence to wait on
bool intr: if true, do an interruptible wait
signed long timeout: timeout value in jiffies, or MAX_SCHEDULE_TIMEOUT

Description

Returns -ERESTARTSYS if interrupted, 0 if the wait timed out, or the remaining timeout in jiffies on success. If timeout is zero the value one is returned if the fence is already signaled for consistency with other functions taking a jiffies timeout.

signed long dma_fence_wait_any_timeout(struct dma_fence **fences, uint32_t count, bool intr, signed long timeout, uint32_t *idx)¶: sleep until any fence gets signaled or until timeout elapses

Parameters

struct dma_fence **fences: array of fences to wait on
uint32_t count: number of fences to wait on
bool intr: if true, do an interruptible wait
signed long timeout: timeout value in jiffies, or MAX_SCHEDULE_TIMEOUT
uint32_t *idx: used to store the first signaled fence index, meaningful only on positive return

Description

Returns -EINVAL on custom fence wait implementation, -ERESTARTSYS if interrupted, 0 if the wait timed out, or the remaining timeout in jiffies on success.

Synchronous waits for the first fence in the array to be signaled. The caller needs to hold a reference to all fences in the array, otherwise a fence might be freed before return, resulting in undefined behavior.

void dma_fence_set_deadline(struct dma_fence *fence, ktime_t deadline)¶: set desired fence-wait deadline hint

Parameters

struct dma_fence *fence: the fence that is to be waited on
ktime_t deadline: the time by which the waiter hopes for the fence to be signaled

Description

Give the fence signaler a hint about an upcoming deadline, such as vblank, by which point the waiter would prefer the fence to be signaled by. This is intended to give feedback to the fence signaler to aid in power management decisions, such as boosting GPU frequency if a periodic vblank deadline is approaching but the fence is not yet signaled..

void dma_fence_describe(struct dma_fence *fence, struct seq_file *seq)¶: Dump fence describtion into seq_file

Parameters

struct dma_fence *fence: the 6fence to describe
struct seq_file *seq: the seq_file to put the textual description into

Description

Dump a textual description of the fence and it's state into the seq_file.

void dma_fence_init(struct dma_fence *fence, const struct dma_fence_ops *ops, spinlock_t *lock, u64 context, u64 seqno)¶: Initialize a custom fence.

Parameters

struct dma_fence *fence: the fence to initialize
const struct dma_fence_ops *ops: the dma_fence_ops for operations on this fence
spinlock_t *lock: the irqsafe spinlock to use for locking this fence
u64 context: the execution context this fence is run on
u64 seqno: a linear increasing sequence number for this context

Description

Initializes an allocated fence, the caller doesn't have to keep its refcount after committing with this fence, but it will need to hold a refcount again if dma_fence_ops.enable_signaling gets called.

context and seqno are used for easy comparison between fences, allowing to check which fence is later by simply using dma_fence_later().

struct dma_fence¶: software synchronization primitive

Definition:

struct dma_fence {
    spinlock_t *lock;
    const struct dma_fence_ops *ops;
    union {
        struct list_head cb_list;
        ktime_t timestamp;
        struct rcu_head rcu;
    };
    u64 context;
    u64 seqno;
    unsigned long flags;
    struct kref refcount;
    int error;
};

Members

lock: spin_lock_irqsave used for locking
ops: dma_fence_ops associated with this fence
{unnamed_union}: anonymous
cb_list: list of all callbacks to call
timestamp: Timestamp when the fence was signaled.
rcu: used for releasing fence with kfree_rcu
context: execution context this fence belongs to, returned by dma_fence_context_alloc()
seqno: the sequence number of this fence inside the execution context, can be compared to decide which fence would be signaled later.
flags: A mask of DMA_FENCE_FLAG_* defined below
refcount: refcount for this fence
error: Optional, only valid if < 0, must be set before calling dma_fence_signal, indicates that the fence has completed with an error.

Description

the flags member must be manipulated and read using the appropriate atomic ops (bit_*), so taking the spinlock will not be needed most of the time.

DMA_FENCE_FLAG_SIGNALED_BIT - fence is already signaled DMA_FENCE_FLAG_TIMESTAMP_BIT - timestamp recorded for fence signaling DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT - enable_signaling might have been called DMA_FENCE_FLAG_USER_BITS - start of the unused bits, can be used by the implementer of the fence for its own purposes. Can be used in different ways by different fence implementers, so do not rely on this.

Since atomic bitops are used, this is not guaranteed to be the case. Particularly, if the bit was set, but dma_fence_signal was called right before this bit was set, it would have been able to set the DMA_FENCE_FLAG_SIGNALED_BIT, before enable_signaling was called. Adding a check for DMA_FENCE_FLAG_SIGNALED_BIT after setting DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT closes this race, and makes sure that after dma_fence_signal was called, any enable_signaling call will have either been completed, or never called at all.

struct dma_fence_cb¶: callback for dma_fence_add_callback()

Definition:

struct dma_fence_cb {
    struct list_head node;
    dma_fence_func_t func;
};

Members

node: used by dma_fence_add_callback() to append this struct to fence::cb_list
func: dma_fence_func_t to call

Description

This struct will be initialized by dma_fence_add_callback(), additional data can be passed along by embedding dma_fence_cb in another struct.

struct dma_fence_ops¶: operations implemented for fence

Definition:

struct dma_fence_ops {
    bool use_64bit_seqno;
    const char * (*get_driver_name)(struct dma_fence *fence);
    const char * (*get_timeline_name)(struct dma_fence *fence);
    bool (*enable_signaling)(struct dma_fence *fence);
    bool (*signaled)(struct dma_fence *fence);
    signed long (*wait)(struct dma_fence *fence, bool intr, signed long timeout);
    void (*release)(struct dma_fence *fence);
    void (*fence_value_str)(struct dma_fence *fence, char *str, int size);
    void (*timeline_value_str)(struct dma_fence *fence, char *str, int size);
    void (*set_deadline)(struct dma_fence *fence, ktime_t deadline);
};

Members

use_64bit_seqno

True if this dma_fence implementation uses 64bit seqno, false otherwise.

get_driver_name

Returns the driver name. This is a callback to allow drivers to compute the name at runtime, without having it to store permanently for each fence, or build a cache of some sort.

This callback is mandatory.

get_timeline_name

Return the name of the context this fence belongs to. This is a callback to allow drivers to compute the name at runtime, without having it to store permanently for each fence, or build a cache of some sort.

This callback is mandatory.

enable_signaling

Enable software signaling of fence.

For fence implementations that have the capability for hw->hw signaling, they can implement this op to enable the necessary interrupts, or insert commands into cmdstream, etc, to avoid these costly operations for the common case where only hw->hw synchronization is required. This is called in the first dma_fence_wait() or dma_fence_add_callback() path to let the fence implementation know that there is another driver waiting on the signal (ie. hw->sw case).

This function can be called from atomic context, but not from irq context, so normal spinlocks can be used.

A return value of false indicates the fence already passed, or some failure occurred that made it impossible to enable signaling. True indicates successful enabling.

dma_fence.error may be set in enable_signaling, but only when false is returned.

Since many implementations can call dma_fence_signal() even when before enable_signaling has been called there's a race window, where the dma_fence_signal() might result in the final fence reference being released and its memory freed. To avoid this, implementations of this callback should grab their own reference using dma_fence_get(), to be released when the fence is signalled (through e.g. the interrupt handler).

This callback is optional. If this callback is not present, then the driver must always have signaling enabled.

signaled

Peek whether the fence is signaled, as a fastpath optimization for e.g. dma_fence_wait() or dma_fence_add_callback(). Note that this callback does not need to make any guarantees beyond that a fence once indicates as signalled must always return true from this callback. This callback may return false even if the fence has completed already, in this case information hasn't propogated throug the system yet. See also dma_fence_is_signaled().

May set dma_fence.error if returning true.

This callback is optional.

wait

Custom wait implementation, defaults to dma_fence_default_wait() if not set.

Deprecated and should not be used by new implementations. Only used by existing implementations which need special handling for their hardware reset procedure.

Must return -ERESTARTSYS if the wait is intr = true and the wait was interrupted, and remaining jiffies if fence has signaled, or 0 if wait timed out. Can also return other error values on custom implementations, which should be treated as if the fence is signaled. For example a hardware lockup could be reported like that.

release

Called on destruction of fence to release additional resources. Can be called from irq context. This callback is optional. If it is NULL, then dma_fence_free() is instead called as the default implementation.

fence_value_str

Callback to fill in free-form debug info specific to this fence, like the sequence number.

This callback is optional.

timeline_value_str

Fills in the current value of the timeline as a string, like the sequence number. Note that the specific fence passed to this function should not matter, drivers should only use it to look up the corresponding timeline structures.

set_deadline

Callback to allow a fence waiter to inform the fence signaler of an upcoming deadline, such as vblank, by which point the waiter would prefer the fence to be signaled by. This is intended to give feedback to the fence signaler to aid in power management decisions, such as boosting GPU frequency.

This is called without dma_fence.lock held, it can be called multiple times and from any context. Locking is up to the callee if it has some state to manage. If multiple deadlines are set, the expectation is to track the soonest one. If the deadline is before the current time, it should be interpreted as an immediate deadline.

This callback is optional.

void dma_fence_put(struct dma_fence *fence)¶: decreases refcount of the fence

Parameters

struct dma_fence *fence: fence to reduce refcount of

struct dma_fence *dma_fence_get(struct dma_fence *fence)¶: increases refcount of the fence

Parameters

struct dma_fence *fence: fence to increase refcount of

Description

Returns the same fence, with refcount increased by 1.

struct dma_fence *dma_fence_get_rcu(struct dma_fence *fence)¶: get a fence from a dma_resv_list with rcu read lock

Parameters

struct dma_fence *fence: fence to increase refcount of

Description

Function returns NULL if no refcount could be obtained, or the fence.

struct dma_fence *dma_fence_get_rcu_safe(struct dma_fence __rcu **fencep)¶: acquire a reference to an RCU tracked fence

Parameters

struct dma_fence __rcu **fencep: pointer to fence to increase refcount of

Description

Function returns NULL if no refcount could be obtained, or the fence. This function handles acquiring a reference to a fence that may be reallocated within the RCU grace period (such as with SLAB_TYPESAFE_BY_RCU), so long as the caller is using RCU on the pointer to the fence.

An alternative mechanism is to employ a seqlock to protect a bunch of fences, such as used by struct dma_resv. When using a seqlock, the seqlock must be taken before and checked after a reference to the fence is acquired (as shown here).

The caller is required to hold the RCU read lock.

bool dma_fence_is_signaled_locked(struct dma_fence *fence)¶: Return an indication if the fence is signaled yet.

Parameters

struct dma_fence *fence: the fence to check

Description

Returns true if the fence was already signaled, false if not. Since this function doesn't enable signaling, it is not guaranteed to ever return true if dma_fence_add_callback(), dma_fence_wait() or dma_fence_enable_sw_signaling() haven't been called before.

This function requires dma_fence.lock to be held.

DMA Fence Array¶

struct dma_fence_array *dma_fence_array_create(int num_fences, struct dma_fence **fences, u64 context, unsigned seqno, bool signal_on_any)¶: Create a custom fence array

Parameters

int num_fences: [in] number of fences to add in the array
struct dma_fence **fences: [in] array containing the fences
u64 context: [in] fence context to use
unsigned seqno: [in] sequence number to use
bool signal_on_any: [in] signal on any fence in the array

Description

Allocate a dma_fence_array object and initialize the base fence with dma_fence_init(). In case of error it returns NULL.

The caller should allocate the fences array with num_fences size and fill it with the fences it wants to add to the object. Ownership of this array is taken and dma_fence_put() is used on each fence on release.

If signal_on_any is true the fence array signals if any fence in the array signals, otherwise it signals when all fences in the array signal.

bool dma_fence_match_context(struct dma_fence *fence, u64 context)¶: Check if all fences are from the given context

Parameters

struct dma_fence *fence: [in] fence or fence array
u64 context: [in] fence context to check all fences against

Description

Checks the provided fence or, for a fence array, all fences in the array against the given context. Returns false if any fence is from a different context.

struct dma_fence_array_cb¶: callback helper for fence array

Definition:

struct dma_fence_array_cb {
    struct dma_fence_cb cb;
    struct dma_fence_array *array;
};

Members

cb: fence callback structure for signaling
array: reference to the parent fence array object

struct dma_fence_array¶: fence to represent an array of fences

Definition:

struct dma_fence_array {
    struct dma_fence base;
    spinlock_t lock;
    unsigned num_fences;
    atomic_t num_pending;
    struct dma_fence **fences;
    struct irq_work work;
};

Members

base: fence base class
lock: spinlock for fence handling
num_fences: number of fences in the array
num_pending: fences in the array still pending
fences: array of the fences
work: internal irq_work function

struct dma_fence_array *to_dma_fence_array(struct dma_fence *fence)¶: cast a fence to a dma_fence_array

Parameters

struct dma_fence *fence: fence to cast to a dma_fence_array

Description

Returns NULL if the fence is not a dma_fence_array, or the dma_fence_array otherwise.

dma_fence_array_for_each¶

dma_fence_array_for_each (fence, index, head)

iterate over all fences in array

Parameters

fence: current fence
index: index into the array
head: potential dma_fence_array object

Description

Test if array is a dma_fence_array object and if yes iterate over all fences in the array. If not just iterate over the fence in array itself.

For a deep dive iterator see dma_fence_unwrap_for_each().

DMA Fence Chain¶

struct dma_fence *dma_fence_chain_walk(struct dma_fence *fence)¶: chain walking function

Parameters

struct dma_fence *fence: current chain node

Description

Walk the chain to the next node. Returns the next fence or NULL if we are at the end of the chain. Garbage collects chain nodes which are already signaled.

int dma_fence_chain_find_seqno(struct dma_fence **pfence, uint64_t seqno)¶: find fence chain node by seqno

Parameters

struct dma_fence **pfence: pointer to the chain node where to start
uint64_t seqno: the sequence number to search for

Description

Advance the fence pointer to the chain node which will signal this sequence number. If no sequence number is provided then this is a no-op.

Returns EINVAL if the fence is not a chain node or the sequence number has not yet advanced far enough.

void dma_fence_chain_init(struct dma_fence_chain *chain, struct dma_fence *prev, struct dma_fence *fence, uint64_t seqno)¶: initialize a fence chain

Parameters

struct dma_fence_chain *chain: the chain node to initialize
struct dma_fence *prev: the previous fence
struct dma_fence *fence: the current fence
uint64_t seqno: the sequence number to use for the fence chain

Description

Initialize a new chain node and either start a new chain or add the node to the existing chain of the previous fence.

struct dma_fence_chain¶: fence to represent an node of a fence chain

Definition:

struct dma_fence_chain {
    struct dma_fence base;
    struct dma_fence __rcu *prev;
    u64 prev_seqno;
    struct dma_fence *fence;
    union {
        struct dma_fence_cb cb;
        struct irq_work work;
    };
    spinlock_t lock;
};

Members

base

fence base class

prev

previous fence of the chain

prev_seqno

original previous seqno before garbage collection

fence

encapsulated fence

{unnamed_union}

anonymous

cb

callback for signaling

This is used to add the callback for signaling the complection of the fence chain. Never used at the same time as the irq work.

work

irq work item for signaling

Irq work structure to allow us to add the callback without running into lock inversion. Never used at the same time as the callback.

lock

spinlock for fence handling

struct dma_fence_chain *to_dma_fence_chain(struct dma_fence *fence)¶: cast a fence to a dma_fence_chain

Parameters

struct dma_fence *fence: fence to cast to a dma_fence_array

Description

Returns NULL if the fence is not a dma_fence_chain, or the dma_fence_chain otherwise.

struct dma_fence *dma_fence_chain_contained(struct dma_fence *fence)¶: return the contained fence

Parameters

struct dma_fence *fence: the fence to test

Description

If the fence is a dma_fence_chain the function returns the fence contained inside the chain object, otherwise it returns the fence itself.

struct dma_fence_chain *dma_fence_chain_alloc(void)¶

Parameters

void: no arguments

Description

Returns a new struct dma_fence_chain object or NULL on failure.

void dma_fence_chain_free(struct dma_fence_chain *chain)¶

Parameters

struct dma_fence_chain *chain: chain node to free

Description

Frees up an allocated but not used struct dma_fence_chain object. This doesn't need an RCU grace period since the fence was never initialized nor published. After dma_fence_chain_init() has been called the fence must be released by calling dma_fence_put(), and not through this function.

dma_fence_chain_for_each¶

dma_fence_chain_for_each (iter, head)

iterate over all fences in chain

Parameters

iter: current fence
head: starting point

Description

Iterate over all fences in the chain. We keep a reference to the current fence while inside the loop which must be dropped when breaking out.

For a deep dive iterator see dma_fence_unwrap_for_each().

DMA Fence unwrap¶

struct dma_fence_unwrap¶: cursor into the container structure

Definition:

struct dma_fence_unwrap {
    struct dma_fence *chain;
    struct dma_fence *array;
    unsigned int index;
};

Members

chain: potential dma_fence_chain, but can be other fence as well
array: potential dma_fence_array, but can be other fence as well
index: last returned index if array is really a dma_fence_array

Description

Should be used with dma_fence_unwrap_for_each() iterator macro.

dma_fence_unwrap_for_each¶

dma_fence_unwrap_for_each (fence, cursor, head)

iterate over all fences in containers

Parameters

fence: current fence
cursor: current position inside the containers
head: starting point for the iterator

Description

Unwrap dma_fence_chain and dma_fence_array containers and deep dive into all potential fences in them. If head is just a normal fence only that one is returned.

dma_fence_unwrap_merge¶

dma_fence_unwrap_merge (...)

unwrap and merge fences

Parameters

...: variable arguments

Description

All fences given as parameters are unwrapped and merged back together as flat dma_fence_array. Useful if multiple containers need to be merged together.

Implemented as a macro to allocate the necessary arrays on the stack and account the stack frame size to the caller.

Returns NULL on memory allocation failure, a dma_fence object representing all the given fences otherwise.

DMA Fence Sync File¶

struct sync_file *sync_file_create(struct dma_fence *fence)¶: creates a sync file

Parameters

struct dma_fence *fence: fence to add to the sync_fence

Description

Creates a sync_file containg fence. This function acquires and additional reference of fence for the newly-created sync_file, if it succeeds. The sync_file can be released with fput(sync_file->file). Returns the sync_file or NULL in case of error.

struct dma_fence *sync_file_get_fence(int fd)¶: get the fence related to the sync_file fd

Parameters

int fd: sync_file fd to get the fence from

Description

Ensures fd references a valid sync_file and returns a fence that represents all fence in the sync_file. On error NULL is returned.

struct sync_file¶: sync file to export to the userspace

Definition:

struct sync_file {
    struct file             *file;
    char user_name[32];
#ifdef CONFIG_DEBUG_FS;
    struct list_head        sync_file_list;
#endif;
    wait_queue_head_t wq;
    unsigned long           flags;
    struct dma_fence        *fence;
    struct dma_fence_cb cb;
};

Members

file: file representing this fence
user_name: Name of the sync file provided by userspace, for merged fences. Otherwise generated through driver callbacks (in which case the entire array is 0).
sync_file_list: membership in global file list
wq: wait queue for fence signaling
flags: flags for the sync_file
fence: fence with the fences in the sync_file
cb: fence callback information

Description

flags: POLL_ENABLED: whether userspace is currently poll()'ing or not

DMA Fence Sync File uABI¶

struct sync_merge_data¶: SYNC_IOC_MERGE: merge two fences

Definition:

struct sync_merge_data {
    char name[32];
    __s32 fd2;
    __s32 fence;
    __u32 flags;
    __u32 pad;
};

Members

name: name of new fence
fd2: file descriptor of second fence
fence: returns the fd of the new fence to userspace
flags: merge_data flags
pad: padding for 64-bit alignment, should always be zero

Description

Creates a new fence containing copies of the sync_pts in both the calling fd and sync_merge_data.fd2. Returns the new fence's fd in sync_merge_data.fence

struct sync_fence_info¶: detailed fence information

Definition:

struct sync_fence_info {
    char obj_name[32];
    char driver_name[32];
    __s32 status;
    __u32 flags;
    __u64 timestamp_ns;
};

Members

obj_name: name of parent sync_timeline
driver_name: name of driver implementing the parent
status: status of the fence 0:active 1:signaled <0:error
flags: fence_info flags
timestamp_ns: timestamp of status change in nanoseconds

struct sync_file_info¶: SYNC_IOC_FILE_INFO: get detailed information on a sync_file

Definition:

struct sync_file_info {
    char name[32];
    __s32 status;
    __u32 flags;
    __u32 num_fences;
    __u32 pad;
    __u64 sync_fence_info;
};

Members

name: name of fence
status: status of fence. 1: signaled 0:active <0:error
flags: sync_file_info flags
num_fences: number of fences in the sync_file
pad: padding for 64-bit alignment, should always be zero
sync_fence_info: pointer to array of struct sync_fence_info with all fences in the sync_file

Description

Takes a struct sync_file_info. If num_fences is 0, the field is updated with the actual number of fences. If num_fences is > 0, the system will use the pointer provided on sync_fence_info to return up to num_fences of struct sync_fence_info, with detailed fence information.

Indefinite DMA Fences¶

At various times struct dma_fence with an indefinite time until dma_fence_wait() finishes have been proposed. Examples include:

Future fences, used in HWC1 to signal when a buffer isn't used by the display any longer, and created with the screen update that makes the buffer visible. The time this fence completes is entirely under userspace's control.
Proxy fences, proposed to handle &drm_syncobj for which the fence has not yet been set. Used to asynchronously delay command submission.
Userspace fences or gpu futexes, fine-grained locking within a command buffer that userspace uses for synchronization across engines or with the CPU, which are then imported as a DMA fence for integration into existing winsys protocols.
Long-running compute command buffers, while still using traditional end of batch DMA fences for memory management instead of context preemption DMA fences which get reattached when the compute job is rescheduled.

Common to all these schemes is that userspace controls the dependencies of these fences and controls when they fire. Mixing indefinite fences with normal in-kernel DMA fences does not work, even when a fallback timeout is included to protect against malicious userspace:

Only the kernel knows about all DMA fence dependencies, userspace is not aware of dependencies injected due to memory management or scheduler decisions.
Only userspace knows about all dependencies in indefinite fences and when exactly they will complete, the kernel has no visibility.

Furthermore the kernel has to be able to hold up userspace command submission for memory management needs, which means we must support indefinite fences being dependent upon DMA fences. If the kernel also support indefinite fences in the kernel like a DMA fence, like any of the above proposal would, there is the potential for deadlocks.

This means that the kernel might accidentally create deadlocks through memory management dependencies which userspace is unaware of, which randomly hangs workloads until the timeout kicks in. Workloads, which from userspace's perspective, do not contain a deadlock. In such a mixed fencing architecture there is no single entity with knowledge of all dependencies. Therefore preventing such deadlocks from within the kernel is not possible.

The only solution to avoid dependencies loops is by not allowing indefinite fences in the kernel. This means:

No future fences, proxy fences or userspace fences imported as DMA fences, with or without a timeout.
No DMA fences that signal end of batchbuffer for command submission where userspace is allowed to use userspace fencing or long running compute workloads. This also means no implicit fencing for shared buffers in these cases.

Recoverable Hardware Page Faults Implications¶

Modern hardware supports recoverable page faults, which has a lot of implications for DMA fences.

First, a pending page fault obviously holds up the work that's running on the accelerator and a memory allocation is usually required to resolve the fault. But memory allocations are not allowed to gate completion of DMA fences, which means any workload using recoverable page faults cannot use DMA fences for synchronization. Synchronization fences controlled by userspace must be used instead.

On GPUs this poses a problem, because current desktop compositor protocols on Linux rely on DMA fences, which means without an entirely new userspace stack built on top of userspace fences, they cannot benefit from recoverable page faults. Specifically this means implicit synchronization will not be possible. The exception is when page faults are only used as migration hints and never to on-demand fill a memory request. For now this means recoverable page faults on GPUs are limited to pure compute workloads.

Furthermore GPUs usually have shared resources between the 3D rendering and compute side, like compute units or command submission engines. If both a 3D job with a DMA fence and a compute workload using recoverable page faults are pending they could deadlock:

The 3D workload might need to wait for the compute job to finish and release hardware resources first.
The compute workload might be stuck in a page fault, because the memory allocation is waiting for the DMA fence of the 3D workload to complete.

There are a few options to prevent this problem, one of which drivers need to ensure:

Compute workloads can always be preempted, even when a page fault is pending and not yet repaired. Not all hardware supports this.
DMA fence workloads and workloads which need page fault handling have independent hardware resources to guarantee forward progress. This could be achieved through e.g. through dedicated engines and minimal compute unit reservations for DMA fence workloads.
The reservation approach could be further refined by only reserving the hardware resources for DMA fence workloads when they are in-flight. This must cover the time from when the DMA fence is visible to other threads up to moment when fence is completed through dma_fence_signal().
As a last resort, if the hardware provides no useful reservation mechanics, all workloads must be flushed from the GPU when switching between jobs requiring DMA fences or jobs requiring page fault handling: This means all DMA fences must complete before a compute job with page fault handling can be inserted into the scheduler queue. And vice versa, before a DMA fence can be made visible anywhere in the system, all compute workloads must be preempted to guarantee all pending GPU page faults are flushed.
Only a fairly theoretical option would be to untangle these dependencies when allocating memory to repair hardware page faults, either through separate memory blocks or runtime tracking of the full dependency graph of all DMA fences. This results very wide impact on the kernel, since resolving the page on the CPU side can itself involve a page fault. It is much more feasible and robust to limit the impact of handling hardware page faults to the specific driver.

Note that workloads that run on independent hardware like copy engines or other GPUs do not have any impact. This allows us to keep using DMA fences internally in the kernel even for resolving hardware page faults, e.g. by using copy engines to clear or copy memory needed to resolve the page fault.

In some ways this page fault problem is a special case of the Infinite DMA Fences discussions: Infinite fences from compute workloads are allowed to depend on DMA fences, but not the other way around. And not even the page fault problem is new, because some other CPU thread in userspace might hit a page fault which holds up a userspace fence - supporting page faults on GPUs doesn't anything fundamentally new.

The Linux Kernel

Contents

This Page