drm: Add the TTM GPU memory manager subsystem.

TTM is a GPU memory manager subsystem designed for use with GPU
devices with various memory types (On-card VRAM, AGP,
PCI apertures etc.). It's essentially a helper library that assists
the DRM driver in creating and managing persistent buffer objects.

TTM manages placement of data and CPU map setup and teardown on
data movement. It can also optionally manage synchronization of
data on a per-buffer-object level.

TTM takes care to provide an always valid virtual user-space address
to a buffer object which makes user-space sub-allocation of
big buffer objects feasible.

TTM uses a fine-grained per buffer-object locking scheme, taking
care to release all relevant locks when waiting for the GPU.
Although this implies some locking overhead, it's probably a big
win for devices with multiple command submission mechanisms, since
the lock contention will be minimal.

TTM can be used with whatever user-space interface the driver
chooses, including GEM. It's used by the upcoming Radeon KMS DRM driver
and is also the GPU memory management core of various new experimental
DRM drivers.

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

1 files changed, 618 insertions, 0 deletions

diff --git a/include/drm/ttm/ttm_bo_api.h b/include/drm/ttm/ttm_bo_api.h
new file mode 100644
index 000000000000..cd22ab4b495c
--- /dev/null
+++ b/include/drm/ttm/ttm_bo_api.h
@@ -0,0 +1,618 @@
+/**************************************************************************
+ *
+ * Copyright (c) 2006-2009 VMware, Inc., Palo Alto, CA., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+/*
+ * Authors: Thomas Hellstrom <thellstrom-at-vmware-dot-com>
+ */
+
+#ifndef _TTM_BO_API_H_
+#define _TTM_BO_API_H_
+
+#include "drm_hashtab.h"
+#include <linux/kref.h>
+#include <linux/list.h>
+#include <linux/wait.h>
+#include <linux/mutex.h>
+#include <linux/mm.h>
+#include <linux/rbtree.h>
+#include <linux/bitmap.h>
+
+struct ttm_bo_device;
+
+struct drm_mm_node;
+
+/**
+ * struct ttm_mem_reg
+ *
+ * @mm_node: Memory manager node.
+ * @size: Requested size of memory region.
+ * @num_pages: Actual size of memory region in pages.
+ * @page_alignment: Page alignment.
+ * @placement: Placement flags.
+ *
+ * Structure indicating the placement and space resources used by a
+ * buffer object.
+ */
+
+struct ttm_mem_reg {
+	struct drm_mm_node *mm_node;
+	unsigned long size;
+	unsigned long num_pages;
+	uint32_t page_alignment;
+	uint32_t mem_type;
+	uint32_t placement;
+};
+
+/**
+ * enum ttm_bo_type
+ *
+ * @ttm_bo_type_device:	These are 'normal' buffers that can
+ * be mmapped by user space. Each of these bos occupy a slot in the
+ * device address space, that can be used for normal vm operations.
+ *
+ * @ttm_bo_type_user: These are user-space memory areas that are made
+ * available to the GPU by mapping the buffer pages into the GPU aperture
+ * space. These buffers cannot be mmaped from the device address space.
+ *
+ * @ttm_bo_type_kernel: These buffers are like ttm_bo_type_device buffers,
+ * but they cannot be accessed from user-space. For kernel-only use.
+ */
+
+enum ttm_bo_type {
+	ttm_bo_type_device,
+	ttm_bo_type_user,
+	ttm_bo_type_kernel
+};
+
+struct ttm_tt;
+
+/**
+ * struct ttm_buffer_object
+ *
+ * @bdev: Pointer to the buffer object device structure.
+ * @buffer_start: The virtual user-space start address of ttm_bo_type_user
+ * buffers.
+ * @type: The bo type.
+ * @destroy: Destruction function. If NULL, kfree is used.
+ * @num_pages: Actual number of pages.
+ * @addr_space_offset: Address space offset.
+ * @acc_size: Accounted size for this object.
+ * @kref: Reference count of this buffer object. When this refcount reaches
+ * zero, the object is put on the delayed delete list.
+ * @list_kref: List reference count of this buffer object. This member is
+ * used to avoid destruction while the buffer object is still on a list.
+ * Lru lists may keep one refcount, the delayed delete list, and kref != 0
+ * keeps one refcount. When this refcount reaches zero,
+ * the object is destroyed.
+ * @event_queue: Queue for processes waiting on buffer object status change.
+ * @lock: spinlock protecting mostly synchronization members.
+ * @proposed_placement: Proposed placement for the buffer. Changed only by the
+ * creator prior to validation as opposed to bo->mem.proposed_flags which is
+ * changed by the implementation prior to a buffer move if it wants to outsmart
+ * the buffer creator / user. This latter happens, for example, at eviction.
+ * @mem: structure describing current placement.
+ * @persistant_swap_storage: Usually the swap storage is deleted for buffers
+ * pinned in physical memory. If this behaviour is not desired, this member
+ * holds a pointer to a persistant shmem object.
+ * @ttm: TTM structure holding system pages.
+ * @evicted: Whether the object was evicted without user-space knowing.
+ * @cpu_writes: For synchronization. Number of cpu writers.
+ * @lru: List head for the lru list.
+ * @ddestroy: List head for the delayed destroy list.
+ * @swap: List head for swap LRU list.
+ * @val_seq: Sequence of the validation holding the @reserved lock.
+ * Used to avoid starvation when many processes compete to validate the
+ * buffer. This member is protected by the bo_device::lru_lock.
+ * @seq_valid: The value of @val_seq is valid. This value is protected by
+ * the bo_device::lru_lock.
+ * @reserved: Deadlock-free lock used for synchronization state transitions.
+ * @sync_obj_arg: Opaque argument to synchronization object function.
+ * @sync_obj: Pointer to a synchronization object.
+ * @priv_flags: Flags describing buffer object internal state.
+ * @vm_rb: Rb node for the vm rb tree.
+ * @vm_node: Address space manager node.
+ * @offset: The current GPU offset, which can have different meanings
+ * depending on the memory type. For SYSTEM type memory, it should be 0.
+ * @cur_placement: Hint of current placement.
+ *
+ * Base class for TTM buffer object, that deals with data placement and CPU
+ * mappings. GPU mappings are really up to the driver, but for simpler GPUs
+ * the driver can usually use the placement offset @offset directly as the
+ * GPU virtual address. For drivers implementing multiple
+ * GPU memory manager contexts, the driver should manage the address space
+ * in these contexts separately and use these objects to get the correct
+ * placement and caching for these GPU maps. This makes it possible to use
+ * these objects for even quite elaborate memory management schemes.
+ * The destroy member, the API visibility of this object makes it possible
+ * to derive driver specific types.
+ */
+
+struct ttm_buffer_object {
+	/**
+	 * Members constant at init.
+	 */
+
+	struct ttm_bo_device *bdev;
+	unsigned long buffer_start;
+	enum ttm_bo_type type;
+	void (*destroy) (struct ttm_buffer_object *);
+	unsigned long num_pages;
+	uint64_t addr_space_offset;
+	size_t acc_size;
+
+	/**
+	* Members not needing protection.
+	*/
+
+	struct kref kref;
+	struct kref list_kref;
+	wait_queue_head_t event_queue;
+	spinlock_t lock;
+
+	/**
+	 * Members protected by the bo::reserved lock.
+	 */
+
+	uint32_t proposed_placement;
+	struct ttm_mem_reg mem;
+	struct file *persistant_swap_storage;
+	struct ttm_tt *ttm;
+	bool evicted;
+
+	/**
+	 * Members protected by the bo::reserved lock only when written to.
+	 */
+
+	atomic_t cpu_writers;
+
+	/**
+	 * Members protected by the bdev::lru_lock.
+	 */
+
+	struct list_head lru;
+	struct list_head ddestroy;
+	struct list_head swap;
+	uint32_t val_seq;
+	bool seq_valid;
+
+	/**
+	 * Members protected by the bdev::lru_lock
+	 * only when written to.
+	 */
+
+	atomic_t reserved;
+
+
+	/**
+	 * Members protected by the bo::lock
+	 */
+
+	void *sync_obj_arg;
+	void *sync_obj;
+	unsigned long priv_flags;
+
+	/**
+	 * Members protected by the bdev::vm_lock
+	 */
+
+	struct rb_node vm_rb;
+	struct drm_mm_node *vm_node;
+
+
+	/**
+	 * Special members that are protected by the reserve lock
+	 * and the bo::lock when written to. Can be read with
+	 * either of these locks held.
+	 */
+
+	unsigned long offset;
+	uint32_t cur_placement;
+};
+
+/**
+ * struct ttm_bo_kmap_obj
+ *
+ * @virtual: The current kernel virtual address.
+ * @page: The page when kmap'ing a single page.
+ * @bo_kmap_type: Type of bo_kmap.
+ *
+ * Object describing a kernel mapping. Since a TTM bo may be located
+ * in various memory types with various caching policies, the
+ * mapping can either be an ioremap, a vmap, a kmap or part of a
+ * premapped region.
+ */
+
+struct ttm_bo_kmap_obj {
+	void *virtual;
+	struct page *page;
+	enum {
+		ttm_bo_map_iomap,
+		ttm_bo_map_vmap,
+		ttm_bo_map_kmap,
+		ttm_bo_map_premapped,
+	} bo_kmap_type;
+};
+
+/**
+ * ttm_bo_reference - reference a struct ttm_buffer_object
+ *
+ * @bo: The buffer object.
+ *
+ * Returns a refcounted pointer to a buffer object.
+ */
+
+static inline struct ttm_buffer_object *
+ttm_bo_reference(struct ttm_buffer_object *bo)
+{
+	kref_get(&bo->kref);
+	return bo;
+}
+
+/**
+ * ttm_bo_wait - wait for buffer idle.
+ *
+ * @bo:  The buffer object.
+ * @interruptible:  Use interruptible wait.
+ * @no_wait:  Return immediately if buffer is busy.
+ *
+ * This function must be called with the bo::mutex held, and makes
+ * sure any previous rendering to the buffer is completed.
+ * Note: It might be necessary to block validations before the
+ * wait by reserving the buffer.
+ * Returns -EBUSY if no_wait is true and the buffer is busy.
+ * Returns -ERESTART if interrupted by a signal.
+ */
+extern int ttm_bo_wait(struct ttm_buffer_object *bo, bool lazy,
+		       bool interruptible, bool no_wait);
+/**
+ * ttm_buffer_object_validate
+ *
+ * @bo: The buffer object.
+ * @proposed_placement: Proposed_placement for the buffer object.
+ * @interruptible: Sleep interruptible if sleeping.
+ * @no_wait: Return immediately if the buffer is busy.
+ *
+ * Changes placement and caching policy of the buffer object
+ * according to bo::proposed_flags.
+ * Returns
+ * -EINVAL on invalid proposed_flags.
+ * -ENOMEM on out-of-memory condition.
+ * -EBUSY if no_wait is true and buffer busy.
+ * -ERESTART if interrupted by a signal.
+ */
+extern int ttm_buffer_object_validate(struct ttm_buffer_object *bo,
+				      uint32_t proposed_placement,
+				      bool interruptible, bool no_wait);
+/**
+ * ttm_bo_unref
+ *
+ * @bo: The buffer object.
+ *
+ * Unreference and clear a pointer to a buffer object.
+ */
+extern void ttm_bo_unref(struct ttm_buffer_object **bo);
+
+/**
+ * ttm_bo_synccpu_write_grab
+ *
+ * @bo: The buffer object:
+ * @no_wait: Return immediately if buffer is busy.
+ *
+ * Synchronizes a buffer object for CPU RW access. This means
+ * blocking command submission that affects the buffer and
+ * waiting for buffer idle. This lock is recursive.
+ * Returns
+ * -EBUSY if the buffer is busy and no_wait is true.
+ * -ERESTART if interrupted by a signal.
+ */
+
+extern int
+ttm_bo_synccpu_write_grab(struct ttm_buffer_object *bo, bool no_wait);
+/**
+ * ttm_bo_synccpu_write_release:
+ *
+ * @bo : The buffer object.
+ *
+ * Releases a synccpu lock.
+ */
+extern void ttm_bo_synccpu_write_release(struct ttm_buffer_object *bo);
+
+/**
+ * ttm_buffer_object_init
+ *
+ * @bdev: Pointer to a ttm_bo_device struct.
+ * @bo: Pointer to a ttm_buffer_object to be initialized.
+ * @size: Requested size of buffer object.
+ * @type: Requested type of buffer object.
+ * @flags: Initial placement flags.
+ * @page_alignment: Data alignment in pages.
+ * @buffer_start: Virtual address of user space data backing a
+ * user buffer object.
+ * @interruptible: If needing to sleep to wait for GPU resources,
+ * sleep interruptible.
+ * @persistant_swap_storage: Usually the swap storage is deleted for buffers
+ * pinned in physical memory. If this behaviour is not desired, this member
+ * holds a pointer to a persistant shmem object. Typically, this would
+ * point to the shmem object backing a GEM object if TTM is used to back a
+ * GEM user interface.
+ * @acc_size: Accounted size for this object.
+ * @destroy: Destroy function. Use NULL for kfree().
+ *
+ * This function initializes a pre-allocated struct ttm_buffer_object.
+ * As this object may be part of a larger structure, this function,
+ * together with the @destroy function,
+ * enables driver-specific objects derived from a ttm_buffer_object.
+ * On successful return, the object kref and list_kref are set to 1.
+ * Returns
+ * -ENOMEM: Out of memory.
+ * -EINVAL: Invalid placement flags.
+ * -ERESTART: Interrupted by signal while sleeping waiting for resources.
+ */
+
+extern int ttm_buffer_object_init(struct ttm_bo_device *bdev,
+				  struct ttm_buffer_object *bo,
+				  unsigned long size,
+				  enum ttm_bo_type type,
+				  uint32_t flags,
+				  uint32_t page_alignment,
+				  unsigned long buffer_start,
+				  bool interrubtible,
+				  struct file *persistant_swap_storage,
+				  size_t acc_size,
+				  void (*destroy) (struct ttm_buffer_object *));
+/**
+ * ttm_bo_synccpu_object_init
+ *
+ * @bdev: Pointer to a ttm_bo_device struct.
+ * @bo: Pointer to a ttm_buffer_object to be initialized.
+ * @size: Requested size of buffer object.
+ * @type: Requested type of buffer object.
+ * @flags: Initial placement flags.
+ * @page_alignment: Data alignment in pages.
+ * @buffer_start: Virtual address of user space data backing a
+ * user buffer object.
+ * @interruptible: If needing to sleep while waiting for GPU resources,
+ * sleep interruptible.
+ * @persistant_swap_storage: Usually the swap storage is deleted for buffers
+ * pinned in physical memory. If this behaviour is not desired, this member
+ * holds a pointer to a persistant shmem object. Typically, this would
+ * point to the shmem object backing a GEM object if TTM is used to back a
+ * GEM user interface.
+ * @p_bo: On successful completion *p_bo points to the created object.
+ *
+ * This function allocates a ttm_buffer_object, and then calls
+ * ttm_buffer_object_init on that object.
+ * The destroy function is set to kfree().
+ * Returns
+ * -ENOMEM: Out of memory.
+ * -EINVAL: Invalid placement flags.
+ * -ERESTART: Interrupted by signal while waiting for resources.
+ */
+
+extern int ttm_buffer_object_create(struct ttm_bo_device *bdev,
+				    unsigned long size,
+				    enum ttm_bo_type type,
+				    uint32_t flags,
+				    uint32_t page_alignment,
+				    unsigned long buffer_start,
+				    bool interruptible,
+				    struct file *persistant_swap_storage,
+				    struct ttm_buffer_object **p_bo);
+
+/**
+ * ttm_bo_check_placement
+ *
+ * @bo: the buffer object.
+ * @set_flags: placement flags to set.
+ * @clr_flags: placement flags to clear.
+ *
+ * Performs minimal validity checking on an intended change of
+ * placement flags.
+ * Returns
+ * -EINVAL: Intended change is invalid or not allowed.
+ */
+
+extern int ttm_bo_check_placement(struct ttm_buffer_object *bo,
+				  uint32_t set_flags, uint32_t clr_flags);
+
+/**
+ * ttm_bo_init_mm
+ *
+ * @bdev: Pointer to a ttm_bo_device struct.
+ * @mem_type: The memory type.
+ * @p_offset: offset for managed area in pages.
+ * @p_size: size managed area in pages.
+ *
+ * Initialize a manager for a given memory type.
+ * Note: if part of driver firstopen, it must be protected from a
+ * potentially racing lastclose.
+ * Returns:
+ * -EINVAL: invalid size or memory type.
+ * -ENOMEM: Not enough memory.
+ * May also return driver-specified errors.
+ */
+
+extern int ttm_bo_init_mm(struct ttm_bo_device *bdev, unsigned type,
+			  unsigned long p_offset, unsigned long p_size);
+/**
+ * ttm_bo_clean_mm
+ *
+ * @bdev: Pointer to a ttm_bo_device struct.
+ * @mem_type: The memory type.
+ *
+ * Take down a manager for a given memory type after first walking
+ * the LRU list to evict any buffers left alive.
+ *
+ * Normally, this function is part of lastclose() or unload(), and at that
+ * point there shouldn't be any buffers left created by user-space, since
+ * there should've been removed by the file descriptor release() method.
+ * However, before this function is run, make sure to signal all sync objects,
+ * and verify that the delayed delete queue is empty. The driver must also
+ * make sure that there are no NO_EVICT buffers present in this memory type
+ * when the call is made.
+ *
+ * If this function is part of a VT switch, the caller must make sure that
+ * there are no appications currently validating buffers before this
+ * function is called. The caller can do that by first taking the
+ * struct ttm_bo_device::ttm_lock in write mode.
+ *
+ * Returns:
+ * -EINVAL: invalid or uninitialized memory type.
+ * -EBUSY: There are still buffers left in this memory type.
+ */
+
+extern int ttm_bo_clean_mm(struct ttm_bo_device *bdev, unsigned mem_type);
+
+/**
+ * ttm_bo_evict_mm
+ *
+ * @bdev: Pointer to a ttm_bo_device struct.
+ * @mem_type: The memory type.
+ *
+ * Evicts all buffers on the lru list of the memory type.
+ * This is normally part of a VT switch or an
+ * out-of-memory-space-due-to-fragmentation handler.
+ * The caller must make sure that there are no other processes
+ * currently validating buffers, and can do that by taking the
+ * struct ttm_bo_device::ttm_lock in write mode.
+ *
+ * Returns:
+ * -EINVAL: Invalid or uninitialized memory type.
+ * -ERESTART: The call was interrupted by a signal while waiting to
+ * evict a buffer.
+ */
+
+extern int ttm_bo_evict_mm(struct ttm_bo_device *bdev, unsigned mem_type);
+
+/**
+ * ttm_kmap_obj_virtual
+ *
+ * @map: A struct ttm_bo_kmap_obj returned from ttm_bo_kmap.
+ * @is_iomem: Pointer to an integer that on return indicates 1 if the
+ * virtual map is io memory, 0 if normal memory.
+ *
+ * Returns the virtual address of a buffer object area mapped by ttm_bo_kmap.
+ * If *is_iomem is 1 on return, the virtual address points to an io memory area,
+ * that should strictly be accessed by the iowriteXX() and similar functions.
+ */
+
+static inline void *ttm_kmap_obj_virtual(struct ttm_bo_kmap_obj *map,
+					 bool *is_iomem)
+{
+	*is_iomem = (map->bo_kmap_type == ttm_bo_map_iomap ||
+		     map->bo_kmap_type == ttm_bo_map_premapped);
+	return map->virtual;
+}
+
+/**
+ * ttm_bo_kmap
+ *
+ * @bo: The buffer object.
+ * @start_page: The first page to map.
+ * @num_pages: Number of pages to map.
+ * @map: pointer to a struct ttm_bo_kmap_obj representing the map.
+ *
+ * Sets up a kernel virtual mapping, using ioremap, vmap or kmap to the
+ * data in the buffer object. The ttm_kmap_obj_virtual function can then be
+ * used to obtain a virtual address to the data.
+ *
+ * Returns
+ * -ENOMEM: Out of memory.
+ * -EINVAL: Invalid range.
+ */
+
+extern int ttm_bo_kmap(struct ttm_buffer_object *bo, unsigned long start_page,
+		       unsigned long num_pages, struct ttm_bo_kmap_obj *map);
+
+/**
+ * ttm_bo_kunmap
+ *
+ * @map: Object describing the map to unmap.
+ *
+ * Unmaps a kernel map set up by ttm_bo_kmap.
+ */
+
+extern void ttm_bo_kunmap(struct ttm_bo_kmap_obj *map);
+
+#if 0
+#endif
+
+/**
+ * ttm_fbdev_mmap - mmap fbdev memory backed by a ttm buffer object.
+ *
+ * @vma:       vma as input from the fbdev mmap method.
+ * @bo:        The bo backing the address space. The address space will
+ * have the same size as the bo, and start at offset 0.
+ *
+ * This function is intended to be called by the fbdev mmap method
+ * if the fbdev address space is to be backed by a bo.
+ */
+
+extern int ttm_fbdev_mmap(struct vm_area_struct *vma,
+			  struct ttm_buffer_object *bo);
+
+/**
+ * ttm_bo_mmap - mmap out of the ttm device address space.
+ *
+ * @filp:      filp as input from the mmap method.
+ * @vma:       vma as input from the mmap method.
+ * @bdev:      Pointer to the ttm_bo_device with the address space manager.
+ *
+ * This function is intended to be called by the device mmap method.
+ * if the device address space is to be backed by the bo manager.
+ */
+
+extern int ttm_bo_mmap(struct file *filp, struct vm_area_struct *vma,
+		       struct ttm_bo_device *bdev);
+
+/**
+ * ttm_bo_io
+ *
+ * @bdev:      Pointer to the struct ttm_bo_device.
+ * @filp:      Pointer to the struct file attempting to read / write.
+ * @wbuf:      User-space pointer to address of buffer to write. NULL on read.
+ * @rbuf:      User-space pointer to address of buffer to read into.
+ * Null on write.
+ * @count:     Number of bytes to read / write.
+ * @f_pos:     Pointer to current file position.
+ * @write:     1 for read, 0 for write.
+ *
+ * This function implements read / write into ttm buffer objects, and is
+ * intended to
+ * be called from the fops::read and fops::write method.
+ * Returns:
+ * See man (2) write, man(2) read. In particular,
+ * the function may return -EINTR if
+ * interrupted by a signal.
+ */
+
+extern ssize_t ttm_bo_io(struct ttm_bo_device *bdev, struct file *filp,
+			 const char __user *wbuf, char __user *rbuf,
+			 size_t count, loff_t *f_pos, bool write);
+
+extern void ttm_bo_swapout_all(struct ttm_bo_device *bdev);
+
+#endif