linux-nvgpu/drivers/gpu/nvgpu/common/mm/buddy_allocator.c

/*
 * Copyright (c) 2016-2017, NVIDIA CORPORATION.  All rights reserved.
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms and conditions of the GNU General Public License,
 * version 2, as published by the Free Software Foundation.
 *
 * This program is distributed in the hope it will be useful, but WITHOUT
 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
 * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
 * more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program.  If not, see <http://www.gnu.org/licenses/>.
 */

#include <nvgpu/allocator.h>
#include <nvgpu/kmem.h>
#include <nvgpu/bug.h>
#include <nvgpu/log2.h>

#include "gk20a/mm_gk20a.h"
#include "gk20a/platform_gk20a.h"

#include "buddy_allocator_priv.h"

/* Some other buddy allocator functions. */
static struct nvgpu_buddy *balloc_free_buddy(struct nvgpu_buddy_allocator *a,
					     u64 addr);
static void balloc_coalesce(struct nvgpu_buddy_allocator *a,
			    struct nvgpu_buddy *b);
static void __balloc_do_free_fixed(struct nvgpu_buddy_allocator *a,
				   struct nvgpu_fixed_alloc *falloc);

/*
 * This function is not present in older kernel's list.h code.
 */
#ifndef list_last_entry
#define list_last_entry(ptr, type, member)	\
	list_entry((ptr)->prev, type, member)
#endif

/*
 * GPU buddy allocator for various address spaces.
 *
 * Current limitations:
 *   o  A fixed allocation could potentially be made that borders PDEs with
 *      different PTE sizes. This would require that fixed buffer to have
 *      different sized PTEs for different parts of the allocation. Probably
 *      best to just require PDE alignment for fixed address allocs.
 *
 *   o  It is currently possible to make an allocator that has a buddy alignment
 *      out of sync with the PDE block size alignment. A simple example is a
 *      32GB address space starting at byte 1. Every buddy is shifted off by 1
 *      which means each buddy corresponf to more than one actual GPU page. The
 *      best way to fix this is probably just require PDE blocksize alignment
 *      for the start of the address space. At the moment all allocators are
 *      easily PDE aligned so this hasn't been a problem.
 */

/*
 * Pick a suitable maximum order for this allocator.
 *
 * Hueristic: Just guessing that the best max order is the largest single
 * block that will fit in the address space.
 */
static void balloc_compute_max_order(struct nvgpu_buddy_allocator *a)
{
	u64 true_max_order = ilog2(a->blks);

	if (a->max_order == 0) {
		a->max_order = true_max_order;
		return;
	}

	if (a->max_order > true_max_order)
		a->max_order = true_max_order;
	if (a->max_order > GPU_BALLOC_MAX_ORDER)
		a->max_order = GPU_BALLOC_MAX_ORDER;
}

/*
 * Since we can only allocate in chucks of a->blk_size we need to trim off
 * any excess data that is not aligned to a->blk_size.
 */
static void balloc_allocator_align(struct nvgpu_buddy_allocator *a)
{
	a->start = ALIGN(a->base, a->blk_size);
	WARN_ON(a->start != a->base);
	a->end   = (a->base + a->length) & ~(a->blk_size - 1);
	a->count = a->end - a->start;
	a->blks  = a->count >> a->blk_shift;
}

/*
 * Pass NULL for parent if you want a top level buddy.
 */
static struct nvgpu_buddy *balloc_new_buddy(struct nvgpu_buddy_allocator *a,
					    struct nvgpu_buddy *parent,
					    u64 start, u64 order)
{
	struct nvgpu_buddy *new_buddy;

	new_buddy = nvgpu_kmem_cache_alloc(a->buddy_cache);
	if (!new_buddy)
		return NULL;

	memset(new_buddy, 0, sizeof(struct nvgpu_buddy));

	new_buddy->parent = parent;
	new_buddy->start = start;
	new_buddy->order = order;
	new_buddy->end = start + (1 << order) * a->blk_size;
	new_buddy->pte_size = BALLOC_PTE_SIZE_ANY;

	return new_buddy;
}

static void __balloc_buddy_list_add(struct nvgpu_buddy_allocator *a,
				    struct nvgpu_buddy *b,
				    struct nvgpu_list_node *list)
{
	if (buddy_is_in_list(b)) {
		alloc_dbg(balloc_owner(a),
			  "Oops: adding added buddy (%llu:0x%llx)\n",
			  b->order, b->start);
		BUG();
	}

	/*
	 * Add big PTE blocks to the tail, small to the head for GVA spaces.
	 * This lets the code that checks if there are available blocks check
	 * without cycling through the entire list.
	 */
	if (a->flags & GPU_ALLOC_GVA_SPACE &&
	    b->pte_size == gmmu_page_size_big)
		nvgpu_list_add_tail(&b->buddy_entry, list);
	else
		nvgpu_list_add(&b->buddy_entry, list);

	buddy_set_in_list(b);
}

static void __balloc_buddy_list_rem(struct nvgpu_buddy_allocator *a,
				    struct nvgpu_buddy *b)
{
	if (!buddy_is_in_list(b)) {
		alloc_dbg(balloc_owner(a),
			  "Oops: removing removed buddy (%llu:0x%llx)\n",
			  b->order, b->start);
		BUG();
	}

	nvgpu_list_del(&b->buddy_entry);
	buddy_clr_in_list(b);
}

/*
 * Add a buddy to one of the buddy lists and deal with the necessary
 * book keeping. Adds the buddy to the list specified by the buddy's order.
 */
static void balloc_blist_add(struct nvgpu_buddy_allocator *a,
			     struct nvgpu_buddy *b)
{
	__balloc_buddy_list_add(a, b, balloc_get_order_list(a, b->order));
	a->buddy_list_len[b->order]++;
}

static void balloc_blist_rem(struct nvgpu_buddy_allocator *a,
			     struct nvgpu_buddy *b)
{
	__balloc_buddy_list_rem(a, b);
	a->buddy_list_len[b->order]--;
}

static u64 balloc_get_order(struct nvgpu_buddy_allocator *a, u64 len)
{
	if (len == 0)
		return 0;

	len--;
	len >>= a->blk_shift;

	return fls(len);
}

static u64 __balloc_max_order_in(struct nvgpu_buddy_allocator *a,
				 u64 start, u64 end)
{
	u64 size = (end - start) >> a->blk_shift;

	if (size > 0)
		return min_t(u64, ilog2(size), a->max_order);
	else
		return GPU_BALLOC_MAX_ORDER;
}

/*
 * Initialize the buddy lists.
 */
static int balloc_init_lists(struct nvgpu_buddy_allocator *a)
{
	int i;
	u64 bstart, bend, order;
	struct nvgpu_buddy *buddy;

	bstart = a->start;
	bend = a->end;

	/* First make sure the LLs are valid. */
	for (i = 0; i < GPU_BALLOC_ORDER_LIST_LEN; i++)
		nvgpu_init_list_node(balloc_get_order_list(a, i));

	while (bstart < bend) {
		order = __balloc_max_order_in(a, bstart, bend);

		buddy = balloc_new_buddy(a, NULL, bstart, order);
		if (!buddy)
			goto cleanup;

		balloc_blist_add(a, buddy);
		bstart += balloc_order_to_len(a, order);
	}

	return 0;

cleanup:
	for (i = 0; i < GPU_BALLOC_ORDER_LIST_LEN; i++) {
		if (!nvgpu_list_empty(balloc_get_order_list(a, i))) {
			buddy = nvgpu_list_first_entry(
					balloc_get_order_list(a, i),
					nvgpu_buddy, buddy_entry);
			balloc_blist_rem(a, buddy);
			nvgpu_kmem_cache_free(a->buddy_cache, buddy);
		}
	}

	return -ENOMEM;
}

/*
 * Clean up and destroy the passed allocator.
 */
static void nvgpu_buddy_allocator_destroy(struct nvgpu_allocator *__a)
{
	int i;
	struct nvgpu_rbtree_node *node = NULL;
	struct nvgpu_buddy *bud;
	struct nvgpu_fixed_alloc *falloc;
	struct nvgpu_buddy_allocator *a = __a->priv;

	alloc_lock(__a);

#ifdef CONFIG_DEBUG_FS
	nvgpu_fini_alloc_debug(__a);
#endif

	/*
	 * Free the fixed allocs first.
	 */
	nvgpu_rbtree_enum_start(0, &node, a->fixed_allocs);
	while (node) {
		falloc = nvgpu_fixed_alloc_from_rbtree_node(node);

		nvgpu_rbtree_unlink(node, &a->fixed_allocs);
		__balloc_do_free_fixed(a, falloc);

		nvgpu_rbtree_enum_start(0, &node, a->fixed_allocs);
	}

	/*
	 * And now free all outstanding allocations.
	 */
	nvgpu_rbtree_enum_start(0, &node, a->alloced_buddies);
	while (node) {
		bud = nvgpu_buddy_from_rbtree_node(node);

		balloc_free_buddy(a, bud->start);
		balloc_blist_add(a, bud);
		balloc_coalesce(a, bud);

		nvgpu_rbtree_enum_start(0, &node, a->alloced_buddies);
	}

	/*
	 * Now clean up the unallocated buddies.
	 */
	for (i = 0; i < GPU_BALLOC_ORDER_LIST_LEN; i++) {
		BUG_ON(a->buddy_list_alloced[i] != 0);

		while (!nvgpu_list_empty(balloc_get_order_list(a, i))) {
			bud = nvgpu_list_first_entry(
						balloc_get_order_list(a, i),
						nvgpu_buddy, buddy_entry);
			balloc_blist_rem(a, bud);
			nvgpu_kmem_cache_free(a->buddy_cache, bud);
		}

		if (a->buddy_list_len[i] != 0) {
			pr_info("Excess buddies!!! (%d: %llu)\n",
				i, a->buddy_list_len[i]);
			BUG();
		}
		if (a->buddy_list_split[i] != 0) {
			pr_info("Excess split nodes!!! (%d: %llu)\n",
				i, a->buddy_list_split[i]);
			BUG();
		}
		if (a->buddy_list_alloced[i] != 0) {
			pr_info("Excess alloced nodes!!! (%d: %llu)\n",
				i, a->buddy_list_alloced[i]);
			BUG();
		}
	}

	nvgpu_kmem_cache_destroy(a->buddy_cache);
	nvgpu_kfree(nvgpu_alloc_to_gpu(__a), a);

	alloc_unlock(__a);
}

/*
 * Combine the passed buddy if possible. The pointer in @b may not be valid
 * after this as the buddy may be freed.
 *
 * @a must be locked.
 */
static void balloc_coalesce(struct nvgpu_buddy_allocator *a,
			    struct nvgpu_buddy *b)
{
	struct nvgpu_buddy *parent;

	if (buddy_is_alloced(b) || buddy_is_split(b))
		return;

	/*
	 * If both our buddy and I are both not allocated and not split then
	 * we can coalesce ourselves.
	 */
	if (!b->buddy)
		return;
	if (buddy_is_alloced(b->buddy) || buddy_is_split(b->buddy))
		return;

	parent = b->parent;

	balloc_blist_rem(a, b);
	balloc_blist_rem(a, b->buddy);

	buddy_clr_split(parent);
	a->buddy_list_split[parent->order]--;
	balloc_blist_add(a, parent);

	/*
	 * Recursively coalesce as far as we can go.
	 */
	balloc_coalesce(a, parent);

	/* Clean up the remains. */
	nvgpu_kmem_cache_free(a->buddy_cache, b->buddy);
	nvgpu_kmem_cache_free(a->buddy_cache, b);
}

/*
 * Split a buddy into two new buddies who are 1/2 the size of the parent buddy.
 *
 * @a must be locked.
 */
static int balloc_split_buddy(struct nvgpu_buddy_allocator *a,
			      struct nvgpu_buddy *b, int pte_size)
{
	struct nvgpu_buddy *left, *right;
	u64 half;

	left = balloc_new_buddy(a, b, b->start, b->order - 1);
	if (!left)
		return -ENOMEM;

	half = (b->end - b->start) / 2;

	right = balloc_new_buddy(a, b, b->start + half, b->order - 1);
	if (!right) {
		nvgpu_kmem_cache_free(a->buddy_cache, left);
		return -ENOMEM;
	}

	buddy_set_split(b);
	a->buddy_list_split[b->order]++;

	b->left = left;
	b->right = right;
	left->buddy = right;
	right->buddy = left;
	left->parent = b;
	right->parent = b;

	/* PTE considerations. */
	if (a->flags & GPU_ALLOC_GVA_SPACE &&
	    left->order <= a->pte_blk_order) {
		left->pte_size = pte_size;
		right->pte_size = pte_size;
	}

	balloc_blist_rem(a, b);
	balloc_blist_add(a, left);
	balloc_blist_add(a, right);

	return 0;
}

/*
 * Place the passed buddy into the RB tree for allocated buddies. Never fails
 * unless the passed entry is a duplicate which is a bug.
 *
 * @a must be locked.
 */
static void balloc_alloc_buddy(struct nvgpu_buddy_allocator *a,
			       struct nvgpu_buddy *b)
{
	b->alloced_entry.key_start = b->start;
	b->alloced_entry.key_end = b->end;

	nvgpu_rbtree_insert(&b->alloced_entry, &a->alloced_buddies);

	buddy_set_alloced(b);
	a->buddy_list_alloced[b->order]++;
}

/*
 * Remove the passed buddy from the allocated buddy RB tree. Returns the
 * deallocated buddy for further processing.
 *
 * @a must be locked.
 */
static struct nvgpu_buddy *balloc_free_buddy(struct nvgpu_buddy_allocator *a,
					     u64 addr)
{
	struct nvgpu_rbtree_node *node = NULL;
	struct nvgpu_buddy *bud;

	nvgpu_rbtree_search(addr, &node, a->alloced_buddies);
	if (!node)
		return NULL;

	bud = nvgpu_buddy_from_rbtree_node(node);

	nvgpu_rbtree_unlink(node, &a->alloced_buddies);
	buddy_clr_alloced(bud);
	a->buddy_list_alloced[bud->order]--;

	return bud;
}

/*
 * Find a suitable buddy for the given order and PTE type (big or little).
 */
static struct nvgpu_buddy *__balloc_find_buddy(struct nvgpu_buddy_allocator *a,
					       u64 order, int pte_size)
{
	struct nvgpu_buddy *bud;

	if (order > a->max_order ||
	    nvgpu_list_empty(balloc_get_order_list(a, order)))
		return NULL;

	if (a->flags & GPU_ALLOC_GVA_SPACE &&
	    pte_size == gmmu_page_size_big)
		bud = nvgpu_list_last_entry(balloc_get_order_list(a, order),
				      nvgpu_buddy, buddy_entry);
	else
		bud = nvgpu_list_first_entry(balloc_get_order_list(a, order),
				       nvgpu_buddy, buddy_entry);

	if (pte_size != BALLOC_PTE_SIZE_ANY &&
	    pte_size != bud->pte_size &&
	    bud->pte_size != BALLOC_PTE_SIZE_ANY)
		return NULL;

	return bud;
}

/*
 * Allocate a suitably sized buddy. If no suitable buddy exists split higher
 * order buddies until we have a suitable buddy to allocate.
 *
 * For PDE grouping add an extra check to see if a buddy is suitable: that the
 * buddy exists in a PDE who's PTE size is reasonable
 *
 * @a must be locked.
 */
static u64 __balloc_do_alloc(struct nvgpu_buddy_allocator *a,
			     u64 order, int pte_size)
{
	u64 split_order;
	struct nvgpu_buddy *bud = NULL;

	split_order = order;
	while (split_order <= a->max_order &&
	       !(bud = __balloc_find_buddy(a, split_order, pte_size)))
		split_order++;

	/* Out of memory! */
	if (!bud)
		return 0;

	while (bud->order != order) {
		if (balloc_split_buddy(a, bud, pte_size))
			return 0; /* No mem... */
		bud = bud->left;
	}

	balloc_blist_rem(a, bud);
	balloc_alloc_buddy(a, bud);

	return bud->start;
}

/*
 * See if the passed range is actually available for allocation. If so, then
 * return 1, otherwise return 0.
 *
 * TODO: Right now this uses the unoptimal approach of going through all
 * outstanding allocations and checking their base/ends. This could be better.
 */
static int balloc_is_range_free(struct nvgpu_buddy_allocator *a,
				u64 base, u64 end)
{
	struct nvgpu_rbtree_node *node = NULL;
	struct nvgpu_buddy *bud;

	nvgpu_rbtree_enum_start(0, &node, a->alloced_buddies);
	if (!node)
		return 1; /* No allocs yet. */

	bud = nvgpu_buddy_from_rbtree_node(node);

	while (bud->start < end) {
		if ((bud->start > base && bud->start < end) ||
		    (bud->end   > base && bud->end   < end))
			return 0;

		nvgpu_rbtree_enum_next(&node, node);
		if (!node)
			break;
		bud = nvgpu_buddy_from_rbtree_node(node);
	}

	return 1;
}

static void balloc_alloc_fixed(struct nvgpu_buddy_allocator *a,
			       struct nvgpu_fixed_alloc *f)
{
	f->alloced_entry.key_start = f->start;
	f->alloced_entry.key_end = f->end;

	nvgpu_rbtree_insert(&f->alloced_entry, &a->fixed_allocs);
}

/*
 * Remove the passed buddy from the allocated buddy RB tree. Returns the
 * deallocated buddy for further processing.
 *
 * @a must be locked.
 */
static struct nvgpu_fixed_alloc *balloc_free_fixed(
	struct nvgpu_buddy_allocator *a, u64 addr)
{
	struct nvgpu_fixed_alloc *falloc;
	struct nvgpu_rbtree_node *node = NULL;

	nvgpu_rbtree_search(addr, &node, a->fixed_allocs);
	if (!node)
		return NULL;

	falloc = nvgpu_fixed_alloc_from_rbtree_node(node);

	nvgpu_rbtree_unlink(node, &a->fixed_allocs);

	return falloc;
}

/*
 * Find the parent range - doesn't necessarily need the parent to actually exist
 * as a buddy. Finding an existing parent comes later...
 */
static void __balloc_get_parent_range(struct nvgpu_buddy_allocator *a,
				      u64 base, u64 order,
				      u64 *pbase, u64 *porder)
{
	u64 base_mask;
	u64 shifted_base = balloc_base_shift(a, base);

	order++;
	base_mask = ~((a->blk_size << order) - 1);

	shifted_base &= base_mask;

	*pbase = balloc_base_unshift(a, shifted_base);
	*porder = order;
}

/*
 * Makes a buddy at the passed address. This will make all parent buddies
 * necessary for this buddy to exist as well.
 */
static struct nvgpu_buddy *__balloc_make_fixed_buddy(
	struct nvgpu_buddy_allocator *a, u64 base, u64 order, int pte_size)
{
	struct nvgpu_buddy *bud = NULL;
	struct nvgpu_list_node *order_list;
	u64 cur_order = order, cur_base = base;

	/*
	 * Algo:
	 *  1. Keep jumping up a buddy order until we find the real buddy that
	 *     this buddy exists in.
	 *  2. Then work our way down through the buddy tree until we hit a dead
	 *     end.
	 *  3. Start splitting buddies until we split to the one we need to
	 *     make.
	 */
	while (cur_order <= a->max_order) {
		int found = 0;

		order_list = balloc_get_order_list(a, cur_order);
		nvgpu_list_for_each_entry(bud, order_list,
					nvgpu_buddy, buddy_entry) {
			if (bud->start == cur_base) {
				/*
				 * Make sure page size matches if it's smaller
				 * than a PDE sized buddy.
				 */
				if (bud->order <= a->pte_blk_order &&
				    bud->pte_size != BALLOC_PTE_SIZE_ANY &&
				    bud->pte_size != pte_size) {
					/* Welp, that's the end of that. */
					alloc_dbg(balloc_owner(a),
						  "Fixed buddy PTE "
						  "size mismatch!\n");
					return NULL;
				}

				found = 1;
				break;
			}
		}

		if (found)
			break;

		__balloc_get_parent_range(a, cur_base, cur_order,
					  &cur_base, &cur_order);
	}

	if (cur_order > a->max_order) {
		alloc_dbg(balloc_owner(a), "No buddy for range ???\n");
		return NULL;
	}

	/* Split this buddy as necessary until we get the target buddy. */
	while (bud->start != base || bud->order != order) {
		if (balloc_split_buddy(a, bud, pte_size)) {
			alloc_dbg(balloc_owner(a),
				  "split buddy failed? {0x%llx, %llu}\n",
				  bud->start, bud->order);
			balloc_coalesce(a, bud);
			return NULL;
		}

		if (base < bud->right->start)
			bud = bud->left;
		else
			bud = bud->right;

	}

	return bud;
}

static u64 __balloc_do_alloc_fixed(struct nvgpu_buddy_allocator *a,
				   struct nvgpu_fixed_alloc *falloc,
				   u64 base, u64 len, int pte_size)
{
	u64 shifted_base, inc_base;
	u64 align_order;

	shifted_base = balloc_base_shift(a, base);
	if (shifted_base == 0)
		align_order = __fls(len >> a->blk_shift);
	else
		align_order = min_t(u64,
				    __ffs(shifted_base >> a->blk_shift),
				    __fls(len >> a->blk_shift));

	if (align_order > a->max_order) {
		alloc_dbg(balloc_owner(a),
			  "Align order too big: %llu > %llu\n",
			  align_order, a->max_order);
		return 0;
	}

	/*
	 * Generate a list of buddies that satisfy this allocation.
	 */
	inc_base = shifted_base;
	while (inc_base < (shifted_base + len)) {
		u64 order_len = balloc_order_to_len(a, align_order);
		u64 remaining;
		struct nvgpu_buddy *bud;

		bud = __balloc_make_fixed_buddy(a,
					balloc_base_unshift(a, inc_base),
					align_order, pte_size);
		if (!bud) {
			alloc_dbg(balloc_owner(a),
				  "Fixed buddy failed: {0x%llx, %llu}!\n",
				  balloc_base_unshift(a, inc_base),
				  align_order);
			goto err_and_cleanup;
		}

		balloc_blist_rem(a, bud);
		balloc_alloc_buddy(a, bud);
		__balloc_buddy_list_add(a, bud, &falloc->buddies);

		/* Book keeping. */
		inc_base += order_len;
		remaining = (shifted_base + len) - inc_base;
		align_order = __ffs(inc_base >> a->blk_shift);

		/* If we don't have much left - trim down align_order. */
		if (balloc_order_to_len(a, align_order) > remaining)
			align_order = __balloc_max_order_in(a, inc_base,
							inc_base + remaining);
	}

	return base;

err_and_cleanup:
	while (!nvgpu_list_empty(&falloc->buddies)) {
		struct nvgpu_buddy *bud = nvgpu_list_first_entry(
						&falloc->buddies,
						nvgpu_buddy, buddy_entry);

		__balloc_buddy_list_rem(a, bud);
		balloc_free_buddy(a, bud->start);
		nvgpu_kmem_cache_free(a->buddy_cache, bud);
	}

	return 0;
}

static void __balloc_do_free_fixed(struct nvgpu_buddy_allocator *a,
				   struct nvgpu_fixed_alloc *falloc)
{
	struct nvgpu_buddy *bud;

	while (!nvgpu_list_empty(&falloc->buddies)) {
		bud = nvgpu_list_first_entry(&falloc->buddies,
				       nvgpu_buddy,
				       buddy_entry);
		__balloc_buddy_list_rem(a, bud);

		balloc_free_buddy(a, bud->start);
		balloc_blist_add(a, bud);
		a->bytes_freed += balloc_order_to_len(a, bud->order);

		/*
		 * Attemp to defrag the allocation.
		 */
		balloc_coalesce(a, bud);
	}

	nvgpu_kfree(nvgpu_alloc_to_gpu(a->owner), falloc);
}

/*
 * Allocate memory from the passed allocator.
 */
static u64 nvgpu_buddy_balloc(struct nvgpu_allocator *__a, u64 len)
{
	u64 order, addr;
	int pte_size;
	struct nvgpu_buddy_allocator *a = __a->priv;

	nvgpu_alloc_trace_func();

	alloc_lock(__a);

	order = balloc_get_order(a, len);

	if (order > a->max_order) {
		alloc_unlock(__a);
		alloc_dbg(balloc_owner(a), "Alloc fail\n");
		nvgpu_alloc_trace_func_done();
		return 0;
	}

	if (a->flags & GPU_ALLOC_GVA_SPACE)
		pte_size = __get_pte_size(a->vm, 0, len);
	else
		pte_size = BALLOC_PTE_SIZE_ANY;

	addr = __balloc_do_alloc(a, order, pte_size);

	if (addr) {
		a->bytes_alloced += len;
		a->bytes_alloced_real += balloc_order_to_len(a, order);
		alloc_dbg(balloc_owner(a),
			  "Alloc 0x%-10llx %3lld:0x%-10llx pte_size=%s\n",
			  addr, order, len,
			  pte_size == gmmu_page_size_big   ? "big" :
			  pte_size == gmmu_page_size_small ? "small" :
			  "NA/any");
	} else {
		alloc_dbg(balloc_owner(a), "Alloc failed: no mem!\n");
	}

	a->alloc_made = 1;

	alloc_unlock(__a);

	nvgpu_alloc_trace_func_done();
	return addr;
}

/*
 * Requires @__a to be locked.
 */
static u64 __nvgpu_balloc_fixed_buddy(struct nvgpu_allocator *__a,
				      u64 base, u64 len, u32 page_size)
{
	int pte_size = BALLOC_PTE_SIZE_ANY;
	u64 ret, real_bytes = 0;
	struct nvgpu_buddy *bud;
	struct nvgpu_fixed_alloc *falloc = NULL;
	struct nvgpu_buddy_allocator *a = __a->priv;

	nvgpu_alloc_trace_func();

	/* If base isn't aligned to an order 0 block, fail. */
	if (base & (a->blk_size - 1))
		goto fail;

	if (len == 0)
		goto fail;

	/* Check that the page size is valid. */
	if (a->flags & GPU_ALLOC_GVA_SPACE && a->vm->big_pages) {
		if (page_size == a->vm->big_page_size)
			pte_size = gmmu_page_size_big;
		else if (page_size == SZ_4K)
			pte_size = gmmu_page_size_small;
		else
			goto fail;
	}

	falloc = nvgpu_kmalloc(nvgpu_alloc_to_gpu(__a), sizeof(*falloc));
	if (!falloc)
		goto fail;

	nvgpu_init_list_node(&falloc->buddies);
	falloc->start = base;
	falloc->end = base + len;

	if (!balloc_is_range_free(a, base, base + len)) {
		alloc_dbg(balloc_owner(a),
			  "Range not free: 0x%llx -> 0x%llx\n",
			  base, base + len);
		goto fail_unlock;
	}

	ret = __balloc_do_alloc_fixed(a, falloc, base, len, pte_size);
	if (!ret) {
		alloc_dbg(balloc_owner(a),
			  "Alloc-fixed failed ?? 0x%llx -> 0x%llx\n",
			  base, base + len);
		goto fail_unlock;
	}

	balloc_alloc_fixed(a, falloc);

	list_for_each_entry(bud, &falloc->buddies, buddy_entry)
		real_bytes += (bud->end - bud->start);

	a->bytes_alloced += len;
	a->bytes_alloced_real += real_bytes;

	alloc_dbg(balloc_owner(a), "Alloc (fixed) 0x%llx\n", base);

	nvgpu_alloc_trace_func_done();
	return base;

fail_unlock:
	alloc_unlock(__a);
fail:
	nvgpu_kfree(nvgpu_alloc_to_gpu(__a), falloc);
	nvgpu_alloc_trace_func_done();
	return 0;
}

/*
 * Allocate a fixed address allocation. The address of the allocation is @base
 * and the length is @len. This is not a typical buddy allocator operation and
 * as such has a high posibility of failure if the address space is heavily in
 * use.
 *
 * Please do not use this function unless _absolutely_ necessary.
 */
static u64 nvgpu_balloc_fixed_buddy(struct nvgpu_allocator *__a,
				    u64 base, u64 len, u32 page_size)
{
	u64 alloc;
	struct nvgpu_buddy_allocator *a = __a->priv;

	alloc_lock(__a);
	alloc = __nvgpu_balloc_fixed_buddy(__a, base, len, page_size);
	a->alloc_made = 1;
	alloc_unlock(__a);

	return alloc;
}

/*
 * Free the passed allocation.
 */
static void nvgpu_buddy_bfree(struct nvgpu_allocator *__a, u64 addr)
{
	struct nvgpu_buddy *bud;
	struct nvgpu_fixed_alloc *falloc;
	struct nvgpu_buddy_allocator *a = __a->priv;

	nvgpu_alloc_trace_func();

	if (!addr) {
		nvgpu_alloc_trace_func_done();
		return;
	}

	alloc_lock(__a);

	/*
	 * First see if this is a fixed alloc. If not fall back to a regular
	 * buddy.
	 */
	falloc = balloc_free_fixed(a, addr);
	if (falloc) {
		__balloc_do_free_fixed(a, falloc);
		goto done;
	}

	bud = balloc_free_buddy(a, addr);
	if (!bud)
		goto done;

	balloc_blist_add(a, bud);
	a->bytes_freed += balloc_order_to_len(a, bud->order);

	/*
	 * Attemp to defrag the allocation.
	 */
	balloc_coalesce(a, bud);

done:
	alloc_unlock(__a);
	alloc_dbg(balloc_owner(a), "Free 0x%llx\n", addr);
	nvgpu_alloc_trace_func_done();
	return;
}

static bool nvgpu_buddy_reserve_is_possible(struct nvgpu_buddy_allocator *a,
					    struct nvgpu_alloc_carveout *co)
{
	struct nvgpu_alloc_carveout *tmp;
	u64 co_base, co_end;

	co_base = co->base;
	co_end  = co->base + co->length;

	/*
	 * Not the fastest approach but we should not have that many carveouts
	 * for any reasonable allocator.
	 */
	nvgpu_list_for_each_entry(tmp, &a->co_list,
				nvgpu_alloc_carveout, co_entry) {
		if ((co_base >= tmp->base &&
		     co_base < (tmp->base + tmp->length)) ||
		    (co_end >= tmp->base &&
		     co_end < (tmp->base + tmp->length)))
			return false;
	}

	return true;
}

/*
 * Carveouts can only be reserved before any regular allocations have been
 * made.
 */
static int nvgpu_buddy_reserve_co(struct nvgpu_allocator *__a,
				  struct nvgpu_alloc_carveout *co)
{
	struct nvgpu_buddy_allocator *a = __a->priv;
	u64 addr;
	int err = 0;

	if (co->base < a->start || (co->base + co->length) > a->end ||
	    a->alloc_made)
		return -EINVAL;

	alloc_lock(__a);

	if (!nvgpu_buddy_reserve_is_possible(a, co)) {
		err = -EBUSY;
		goto done;
	}

	/* Should not be possible to fail... */
	addr = __nvgpu_balloc_fixed_buddy(__a, co->base, co->length, 0);
	if (!addr) {
		err = -ENOMEM;
		pr_warn("%s: Failed to reserve a valid carveout!\n", __func__);
		goto done;
	}

	nvgpu_list_add(&co->co_entry, &a->co_list);

done:
	alloc_unlock(__a);
	return err;
}

/*
 * Carveouts can be release at any time.
 */
static void nvgpu_buddy_release_co(struct nvgpu_allocator *__a,
				   struct nvgpu_alloc_carveout *co)
{
	alloc_lock(__a);

	nvgpu_list_del(&co->co_entry);
	nvgpu_free(__a, co->base);

	alloc_unlock(__a);
}

static u64 nvgpu_buddy_alloc_length(struct nvgpu_allocator *a)
{
	struct nvgpu_buddy_allocator *ba = a->priv;

	return ba->length;
}

static u64 nvgpu_buddy_alloc_base(struct nvgpu_allocator *a)
{
	struct nvgpu_buddy_allocator *ba = a->priv;

	return ba->start;
}

static int nvgpu_buddy_alloc_inited(struct nvgpu_allocator *a)
{
	struct nvgpu_buddy_allocator *ba = a->priv;
	int inited = ba->initialized;

	rmb();
	return inited;
}

static u64 nvgpu_buddy_alloc_end(struct nvgpu_allocator *a)
{
	struct nvgpu_buddy_allocator *ba = a->priv;

	return ba->end;
}

static u64 nvgpu_buddy_alloc_space(struct nvgpu_allocator *a)
{
	struct nvgpu_buddy_allocator *ba = a->priv;
	u64 space;

	alloc_lock(a);
	space = ba->end - ba->start -
		(ba->bytes_alloced_real - ba->bytes_freed);
	alloc_unlock(a);

	return space;
}

#ifdef CONFIG_DEBUG_FS
/*
 * Print the buddy allocator top level stats. If you pass @s as NULL then the
 * stats are printed to the kernel log. This lets this code be used for
 * debugging purposes internal to the allocator.
 */
static void nvgpu_buddy_print_stats(struct nvgpu_allocator *__a,
				    struct seq_file *s, int lock)
{
	int i = 0;
	struct nvgpu_rbtree_node *node = NULL;
	struct nvgpu_fixed_alloc *falloc;
	struct nvgpu_alloc_carveout *tmp;
	struct nvgpu_buddy_allocator *a = __a->priv;

	__alloc_pstat(s, __a, "base = %llu, limit = %llu, blk_size = %llu\n",
		      a->base, a->length, a->blk_size);
	__alloc_pstat(s, __a, "Internal params:\n");
	__alloc_pstat(s, __a, "  start = 0x%llx\n", a->start);
	__alloc_pstat(s, __a, "  end   = 0x%llx\n", a->end);
	__alloc_pstat(s, __a, "  count = 0x%llx\n", a->count);
	__alloc_pstat(s, __a, "  blks  = 0x%llx\n", a->blks);
	__alloc_pstat(s, __a, "  max_order = %llu\n", a->max_order);

	if (lock)
		alloc_lock(__a);

	if (!nvgpu_list_empty(&a->co_list)) {
		__alloc_pstat(s, __a, "\n");
		__alloc_pstat(s, __a, "Carveouts:\n");
		nvgpu_list_for_each_entry(tmp, &a->co_list,
					nvgpu_alloc_carveout, co_entry)
			__alloc_pstat(s, __a,
				      "  CO %2d: %-20s 0x%010llx + 0x%llx\n",
				      i++, tmp->name, tmp->base, tmp->length);
	}

	__alloc_pstat(s, __a, "\n");
	__alloc_pstat(s, __a, "Buddy blocks:\n");
	__alloc_pstat(s, __a, "  Order   Free    Alloced   Split\n");
	__alloc_pstat(s, __a, "  -----   ----    -------   -----\n");

	for (i = a->max_order; i >= 0; i--) {
		if (a->buddy_list_len[i] == 0 &&
		    a->buddy_list_alloced[i] == 0 &&
		    a->buddy_list_split[i] == 0)
			continue;

		__alloc_pstat(s, __a, "  %3d     %-7llu %-9llu %llu\n", i,
			      a->buddy_list_len[i],
			      a->buddy_list_alloced[i],
			      a->buddy_list_split[i]);
	}

	__alloc_pstat(s, __a, "\n");

	nvgpu_rbtree_enum_start(0, &node, a->fixed_allocs);
	i = 1;
	while (node) {
		falloc = nvgpu_fixed_alloc_from_rbtree_node(node);

		__alloc_pstat(s, __a, "Fixed alloc (%d): [0x%llx -> 0x%llx]\n",
			      i, falloc->start, falloc->end);

		nvgpu_rbtree_enum_next(&node, a->fixed_allocs);
	}

	__alloc_pstat(s, __a, "\n");
	__alloc_pstat(s, __a, "Bytes allocated:        %llu\n",
		      a->bytes_alloced);
	__alloc_pstat(s, __a, "Bytes allocated (real): %llu\n",
		      a->bytes_alloced_real);
	__alloc_pstat(s, __a, "Bytes freed:            %llu\n",
		      a->bytes_freed);

	if (lock)
		alloc_unlock(__a);
}
#endif

static const struct nvgpu_allocator_ops buddy_ops = {
	.alloc		= nvgpu_buddy_balloc,
	.free		= nvgpu_buddy_bfree,

	.alloc_fixed	= nvgpu_balloc_fixed_buddy,
	/* .free_fixed not needed. */

	.reserve_carveout	= nvgpu_buddy_reserve_co,
	.release_carveout	= nvgpu_buddy_release_co,

	.base		= nvgpu_buddy_alloc_base,
	.length		= nvgpu_buddy_alloc_length,
	.end		= nvgpu_buddy_alloc_end,
	.inited		= nvgpu_buddy_alloc_inited,
	.space		= nvgpu_buddy_alloc_space,

	.fini		= nvgpu_buddy_allocator_destroy,

#ifdef CONFIG_DEBUG_FS
	.print_stats	= nvgpu_buddy_print_stats,
#endif
};

/*
 * Initialize a buddy allocator. Returns 0 on success. This allocator does
 * not necessarily manage bytes. It manages distinct ranges of resources. This
 * allows the allocator to work for things like comp_tags, semaphores, etc.
 *
 * @allocator: Ptr to an allocator struct to init.
 * @vm: GPU VM to associate this allocator with. Can be NULL. Will be used to
 *      get PTE size for GVA spaces.
 * @name: Name of the allocator. Doesn't have to be static storage.
 * @base: The base address of the resource pool being managed.
 * @size: Number of resources in the pool.
 * @blk_size: Minimum number of resources to allocate at once. For things like
 *            semaphores this is 1. For GVA this might be as much as 64k. This
 *            corresponds to order 0. Must be power of 2.
 * @max_order: Pick a maximum order. If you leave this as 0, the buddy allocator
 *             will try and pick a reasonable max order.
 * @flags: Extra flags necessary. See GPU_BALLOC_*.
 */
int __nvgpu_buddy_allocator_init(struct gk20a *g, struct nvgpu_allocator *__a,
				 struct vm_gk20a *vm, const char *name,
				 u64 base, u64 size, u64 blk_size,
				 u64 max_order, u64 flags)
{
	int err;
	u64 pde_size;
	struct nvgpu_buddy_allocator *a;

	/* blk_size must be greater than 0 and a power of 2. */
	if (blk_size == 0)
		return -EINVAL;
	if (blk_size & (blk_size - 1))
		return -EINVAL;

	if (max_order > GPU_BALLOC_MAX_ORDER)
		return -EINVAL;

	/* If this is to manage a GVA space we need a VM. */
	if (flags & GPU_ALLOC_GVA_SPACE && !vm)
		return -EINVAL;

	a = nvgpu_kzalloc(g, sizeof(struct nvgpu_buddy_allocator));
	if (!a)
		return -ENOMEM;

	err = __nvgpu_alloc_common_init(__a, g, name, a, false, &buddy_ops);
	if (err)
		goto fail;

	a->base = base;
	a->length = size;
	a->blk_size = blk_size;
	a->blk_shift = __ffs(blk_size);
	a->owner = __a;

	/*
	 * If base is 0 then modfy base to be the size of one block so that we
	 * can return errors by returning addr == 0.
	 */
	if (a->base == 0) {
		a->base = a->blk_size;
		a->length -= a->blk_size;
	}

	a->vm = vm;
	if (flags & GPU_ALLOC_GVA_SPACE) {
		pde_size = ((u64)vm->big_page_size) << 10;
		a->pte_blk_order = balloc_get_order(a, pde_size);
	}

	/*
	 * When we have a GVA space with big_pages enabled the size and base
	 * must be PDE aligned. If big_pages are not enabled then this
	 * requirement is not necessary.
	 */
	if (flags & GPU_ALLOC_GVA_SPACE && vm->big_pages &&
	    (base & ((vm->big_page_size << 10) - 1) ||
	     size & ((vm->big_page_size << 10) - 1)))
		return -EINVAL;

	a->flags = flags;
	a->max_order = max_order;

	balloc_allocator_align(a);
	balloc_compute_max_order(a);

	a->buddy_cache = nvgpu_kmem_cache_create(g, sizeof(struct nvgpu_buddy));
	if (!a->buddy_cache) {
		err = -ENOMEM;
		goto fail;
	}

	a->alloced_buddies = NULL;
	a->fixed_allocs = NULL;
	nvgpu_init_list_node(&a->co_list);
	err = balloc_init_lists(a);
	if (err)
		goto fail;

	wmb();
	a->initialized = 1;

#ifdef CONFIG_DEBUG_FS
	nvgpu_init_alloc_debug(g, __a);
#endif
	alloc_dbg(__a, "New allocator: type      buddy\n");
	alloc_dbg(__a, "               base      0x%llx\n", a->base);
	alloc_dbg(__a, "               size      0x%llx\n", a->length);
	alloc_dbg(__a, "               blk_size  0x%llx\n", a->blk_size);
	if (flags & GPU_ALLOC_GVA_SPACE)
		alloc_dbg(balloc_owner(a),
		       "               pde_size  0x%llx\n",
			  balloc_order_to_len(a, a->pte_blk_order));
	alloc_dbg(__a, "               max_order %llu\n", a->max_order);
	alloc_dbg(__a, "               flags     0x%llx\n", a->flags);

	return 0;

fail:
	if (a->buddy_cache)
		nvgpu_kmem_cache_destroy(a->buddy_cache);
	nvgpu_kfree(g, a);
	return err;
}

int nvgpu_buddy_allocator_init(struct gk20a *g, struct nvgpu_allocator *a,
			       const char *name, u64 base, u64 size,
			       u64 blk_size, u64 flags)
{
	return __nvgpu_buddy_allocator_init(g, a, NULL, name,
					    base, size, blk_size, 0, 0);
}