[linux-2.6-block.git] / net / ceph / crush / mapper.c


#ifdef __KERNEL__
# include <linux/string.h>
# include <linux/slab.h>
# include <linux/bug.h>
# include <linux/kernel.h>
# ifndef dprintk
#  define dprintk(args...)
# endif
#else
# include <string.h>
# include <stdio.h>
# include <stdlib.h>
# include <assert.h>
# define BUG_ON(x) assert(!(x))
# define dprintk(args...) /* printf(args) */
# define kmalloc(x, f) malloc(x)
# define kfree(x) free(x)
#endif

#include <linux/crush/crush.h>
#include <linux/crush/hash.h>
#include <linux/crush/mapper.h>

/*
 * Implement the core CRUSH mapping algorithm.
 */

/**
 * crush_find_rule - find a crush_rule id for a given ruleset, type, and size.
 * @map: the crush_map
 * @ruleset: the storage ruleset id (user defined)
 * @type: storage ruleset type (user defined)
 * @size: output set size
 */
int crush_find_rule(const struct crush_map *map, int ruleset, int type, int size)
{
	__u32 i;

	for (i = 0; i < map->max_rules; i++) {
		if (map->rules[i] &&
		    map->rules[i]->mask.ruleset == ruleset &&
		    map->rules[i]->mask.type == type &&
		    map->rules[i]->mask.min_size <= size &&
		    map->rules[i]->mask.max_size >= size)
			return i;
	}
	return -1;
}


/*
 * bucket choose methods
 *
 * For each bucket algorithm, we have a "choose" method that, given a
 * crush input @x and replica position (usually, position in output set) @r,
 * will produce an item in the bucket.
 */

/*
 * Choose based on a random permutation of the bucket.
 *
 * We used to use some prime number arithmetic to do this, but it
 * wasn't very random, and had some other bad behaviors.  Instead, we
 * calculate an actual random permutation of the bucket members.
 * Since this is expensive, we optimize for the r=0 case, which
 * captures the vast majority of calls.
 */
static int bucket_perm_choose(struct crush_bucket *bucket,
			      int x, int r)
{
	unsigned int pr = r % bucket->size;
	unsigned int i, s;

	/* start a new permutation if @x has changed */
	if (bucket->perm_x != (__u32)x || bucket->perm_n == 0) {
		dprintk("bucket %d new x=%d\n", bucket->id, x);
		bucket->perm_x = x;

		/* optimize common r=0 case */
		if (pr == 0) {
			s = crush_hash32_3(bucket->hash, x, bucket->id, 0) %
				bucket->size;
			bucket->perm[0] = s;
			bucket->perm_n = 0xffff;   /* magic value, see below */
			goto out;
		}

		for (i = 0; i < bucket->size; i++)
			bucket->perm[i] = i;
		bucket->perm_n = 0;
	} else if (bucket->perm_n == 0xffff) {
		/* clean up after the r=0 case above */
		for (i = 1; i < bucket->size; i++)
			bucket->perm[i] = i;
		bucket->perm[bucket->perm[0]] = 0;
		bucket->perm_n = 1;
	}

	/* calculate permutation up to pr */
	for (i = 0; i < bucket->perm_n; i++)
		dprintk(" perm_choose have %d: %d\n", i, bucket->perm[i]);
	while (bucket->perm_n <= pr) {
		unsigned int p = bucket->perm_n;
		/* no point in swapping the final entry */
		if (p < bucket->size - 1) {
			i = crush_hash32_3(bucket->hash, x, bucket->id, p) %
				(bucket->size - p);
			if (i) {
				unsigned int t = bucket->perm[p + i];
				bucket->perm[p + i] = bucket->perm[p];
				bucket->perm[p] = t;
			}
			dprintk(" perm_choose swap %d with %d\n", p, p+i);
		}
		bucket->perm_n++;
	}
	for (i = 0; i < bucket->size; i++)
		dprintk(" perm_choose  %d: %d\n", i, bucket->perm[i]);

	s = bucket->perm[pr];
out:
	dprintk(" perm_choose %d sz=%d x=%d r=%d (%d) s=%d\n", bucket->id,
		bucket->size, x, r, pr, s);
	return bucket->items[s];
}

/* uniform */
static int bucket_uniform_choose(struct crush_bucket_uniform *bucket,
				 int x, int r)
{
	return bucket_perm_choose(&bucket->h, x, r);
}

/* list */
static int bucket_list_choose(struct crush_bucket_list *bucket,
			      int x, int r)
{
	int i;

	for (i = bucket->h.size-1; i >= 0; i--) {
		__u64 w = crush_hash32_4(bucket->h.hash,x, bucket->h.items[i],
					 r, bucket->h.id);
		w &= 0xffff;
		dprintk("list_choose i=%d x=%d r=%d item %d weight %x "
			"sw %x rand %llx",
			i, x, r, bucket->h.items[i], bucket->item_weights[i],
			bucket->sum_weights[i], w);
		w *= bucket->sum_weights[i];
		w = w >> 16;
		/*dprintk(" scaled %llx\n", w);*/
		if (w < bucket->item_weights[i])
			return bucket->h.items[i];
	}

	dprintk("bad list sums for bucket %d\n", bucket->h.id);
	return bucket->h.items[0];
}


/* (binary) tree */
static int height(int n)
{
	int h = 0;
	while ((n & 1) == 0) {
		h++;
		n = n >> 1;
	}
	return h;
}

static int left(int x)
{
	int h = height(x);
	return x - (1 << (h-1));
}

static int right(int x)
{
	int h = height(x);
	return x + (1 << (h-1));
}

static int terminal(int x)
{
	return x & 1;
}

static int bucket_tree_choose(struct crush_bucket_tree *bucket,
			      int x, int r)
{
	int n, l;
	__u32 w;
	__u64 t;

	/* start at root */
	n = bucket->num_nodes >> 1;

	while (!terminal(n)) {
		/* pick point in [0, w) */
		w = bucket->node_weights[n];
		t = (__u64)crush_hash32_4(bucket->h.hash, x, n, r,
					  bucket->h.id) * (__u64)w;
		t = t >> 32;

		/* descend to the left or right? */
		l = left(n);
		if (t < bucket->node_weights[l])
			n = l;
		else
			n = right(n);
	}

	return bucket->h.items[n >> 1];
}


/* straw */

static int bucket_straw_choose(struct crush_bucket_straw *bucket,
			       int x, int r)
{
	__u32 i;
	int high = 0;
	__u64 high_draw = 0;
	__u64 draw;

	for (i = 0; i < bucket->h.size; i++) {
		draw = crush_hash32_3(bucket->h.hash, x, bucket->h.items[i], r);
		draw &= 0xffff;
		draw *= bucket->straws[i];
		if (i == 0 || draw > high_draw) {
			high = i;
			high_draw = draw;
		}
	}
	return bucket->h.items[high];
}

static int crush_bucket_choose(struct crush_bucket *in, int x, int r)
{
	dprintk(" crush_bucket_choose %d x=%d r=%d\n", in->id, x, r);
	BUG_ON(in->size == 0);
	switch (in->alg) {
	case CRUSH_BUCKET_UNIFORM:
		return bucket_uniform_choose((struct crush_bucket_uniform *)in,
					  x, r);
	case CRUSH_BUCKET_LIST:
		return bucket_list_choose((struct crush_bucket_list *)in,
					  x, r);
	case CRUSH_BUCKET_TREE:
		return bucket_tree_choose((struct crush_bucket_tree *)in,
					  x, r);
	case CRUSH_BUCKET_STRAW:
		return bucket_straw_choose((struct crush_bucket_straw *)in,
					   x, r);
	default:
		dprintk("unknown bucket %d alg %d\n", in->id, in->alg);
		return in->items[0];
	}
}

/*
 * true if device is marked "out" (failed, fully offloaded)
 * of the cluster
 */
static int is_out(const struct crush_map *map, const __u32 *weight, int item, int x)
{
	if (weight[item] >= 0x10000)
		return 0;
	if (weight[item] == 0)
		return 1;
	if ((crush_hash32_2(CRUSH_HASH_RJENKINS1, x, item) & 0xffff)
	    < weight[item])
		return 0;
	return 1;
}

/**
 * crush_choose - choose numrep distinct items of given type
 * @map: the crush_map
 * @bucket: the bucket we are choose an item from
 * @x: crush input value
 * @numrep: the number of items to choose
 * @type: the type of item to choose
 * @out: pointer to output vector
 * @outpos: our position in that vector
 * @firstn: true if choosing "first n" items, false if choosing "indep"
 * @recurse_to_leaf: true if we want one device under each item of given type
 * @out2: second output vector for leaf items (if @recurse_to_leaf)
 */
static int crush_choose(const struct crush_map *map,
			struct crush_bucket *bucket,
			const __u32 *weight,
			int x, int numrep, int type,
			int *out, int outpos,
			int firstn, int recurse_to_leaf,
			int *out2)
{
	int rep;
	unsigned int ftotal, flocal;
	int retry_descent, retry_bucket, skip_rep;
	struct crush_bucket *in = bucket;
	int r;
	int i;
	int item = 0;
	int itemtype;
	int collide, reject;

	dprintk("CHOOSE%s bucket %d x %d outpos %d numrep %d\n", recurse_to_leaf ? "_LEAF" : "",
		bucket->id, x, outpos, numrep);

	for (rep = outpos; rep < numrep; rep++) {
		/* keep trying until we get a non-out, non-colliding item */
		ftotal = 0;
		skip_rep = 0;
		do {
			retry_descent = 0;
			in = bucket;               /* initial bucket */

			/* choose through intervening buckets */
			flocal = 0;
			do {
				collide = 0;
				retry_bucket = 0;
				r = rep;
				if (in->alg == CRUSH_BUCKET_UNIFORM) {
					/* be careful */
					if (firstn || (__u32)numrep >= in->size)
						/* r' = r + f_total */
						r += ftotal;
					else if (in->size % numrep == 0)
						/* r'=r+(n+1)*f_local */
						r += (numrep+1) *
							(flocal+ftotal);
					else
						/* r' = r + n*f_local */
						r += numrep * (flocal+ftotal);
				} else {
					if (firstn)
						/* r' = r + f_total */
						r += ftotal;
					else
						/* r' = r + n*f_local */
						r += numrep * (flocal+ftotal);
				}

				/* bucket choose */
				if (in->size == 0) {
					reject = 1;
					goto reject;
				}
				if (map->choose_local_fallback_tries > 0 &&
				    flocal >= (in->size>>1) &&
				    flocal > map->choose_local_fallback_tries)
					item = bucket_perm_choose(in, x, r);
				else
					item = crush_bucket_choose(in, x, r);
				if (item >= map->max_devices) {
					dprintk("   bad item %d\n", item);
					skip_rep = 1;
					break;
				}

				/* desired type? */
				if (item < 0)
					itemtype = map->buckets[-1-item]->type;
				else
					itemtype = 0;
				dprintk("  item %d type %d\n", item, itemtype);

				/* keep going? */
				if (itemtype != type) {
					if (item >= 0 ||
					    (-1-item) >= map->max_buckets) {
						dprintk("   bad item type %d\n", type);
						skip_rep = 1;
						break;
					}
					in = map->buckets[-1-item];
					retry_bucket = 1;
					continue;
				}

				/* collision? */
				for (i = 0; i < outpos; i++) {
					if (out[i] == item) {
						collide = 1;
						break;
					}
				}

				reject = 0;
				if (recurse_to_leaf) {
					if (item < 0) {
						if (crush_choose(map,
							 map->buckets[-1-item],
							 weight,
							 x, outpos+1, 0,
							 out2, outpos,
							 firstn, 0,
							 NULL) <= outpos)
							/* didn't get leaf */
							reject = 1;
					} else {
						/* we already have a leaf! */
						out2[outpos] = item;
					}
				}

				if (!reject) {
					/* out? */
					if (itemtype == 0)
						reject = is_out(map, weight,
								item, x);
					else
						reject = 0;
				}

reject:
				if (reject || collide) {
					ftotal++;
					flocal++;

					if (collide && flocal <= map->choose_local_tries)
						/* retry locally a few times */
						retry_bucket = 1;
					else if (map->choose_local_fallback_tries > 0 &&
						 flocal <= in->size + map->choose_local_fallback_tries)
						/* exhaustive bucket search */
						retry_bucket = 1;
					else if (ftotal <= map->choose_total_tries)
						/* then retry descent */
						retry_descent = 1;
					else
						/* else give up */
						skip_rep = 1;
					dprintk("  reject %d  collide %d  "
						"ftotal %u  flocal %u\n",
						reject, collide, ftotal,
						flocal);
				}
			} while (retry_bucket);
		} while (retry_descent);

		if (skip_rep) {
			dprintk("skip rep\n");
			continue;
		}

		dprintk("CHOOSE got %d\n", item);
		out[outpos] = item;
		outpos++;
	}

	dprintk("CHOOSE returns %d\n", outpos);
	return outpos;
}


/**
 * crush_do_rule - calculate a mapping with the given input and rule
 * @map: the crush_map
 * @ruleno: the rule id
 * @x: hash input
 * @result: pointer to result vector
 * @result_max: maximum result size
 */
int crush_do_rule(const struct crush_map *map,
		  int ruleno, int x, int *result, int result_max,
		  const __u32 *weight)
{
	int result_len;
	int a[CRUSH_MAX_SET];
	int b[CRUSH_MAX_SET];
	int c[CRUSH_MAX_SET];
	int recurse_to_leaf;
	int *w;
	int wsize = 0;
	int *o;
	int osize;
	int *tmp;
	struct crush_rule *rule;
	__u32 step;
	int i, j;
	int numrep;
	int firstn;

	if ((__u32)ruleno >= map->max_rules) {
		dprintk(" bad ruleno %d\n", ruleno);
		return 0;
	}

	rule = map->rules[ruleno];
	result_len = 0;
	w = a;
	o = b;

	for (step = 0; step < rule->len; step++) {
		struct crush_rule_step *curstep = &rule->steps[step];

		firstn = 0;
		switch (curstep->op) {
		case CRUSH_RULE_TAKE:
			w[0] = curstep->arg1;
			wsize = 1;
			break;

		case CRUSH_RULE_CHOOSE_LEAF_FIRSTN:
		case CRUSH_RULE_CHOOSE_FIRSTN:
			firstn = 1;
			/* fall through */
		case CRUSH_RULE_CHOOSE_LEAF_INDEP:
		case CRUSH_RULE_CHOOSE_INDEP:
			if (wsize == 0)
				break;

			recurse_to_leaf =
				curstep->op ==
				 CRUSH_RULE_CHOOSE_LEAF_FIRSTN ||
				curstep->op ==
				CRUSH_RULE_CHOOSE_LEAF_INDEP;

			/* reset output */
			osize = 0;

			for (i = 0; i < wsize; i++) {
				/*
				 * see CRUSH_N, CRUSH_N_MINUS macros.
				 * basically, numrep <= 0 means relative to
				 * the provided result_max
				 */
				numrep = curstep->arg1;
				if (numrep <= 0) {
					numrep += result_max;
					if (numrep <= 0)
						continue;
				}
				j = 0;
				osize += crush_choose(map,
						      map->buckets[-1-w[i]],
						      weight,
						      x, numrep,
						      curstep->arg2,
						      o+osize, j,
						      firstn,
						      recurse_to_leaf, c+osize);
			}

			if (recurse_to_leaf)
				/* copy final _leaf_ values to output set */
				memcpy(o, c, osize*sizeof(*o));

			/* swap t and w arrays */
			tmp = o;
			o = w;
			w = tmp;
			wsize = osize;
			break;


		case CRUSH_RULE_EMIT:
			for (i = 0; i < wsize && result_len < result_max; i++) {
				result[result_len] = w[i];
				result_len++;
			}
			wsize = 0;
			break;

		default:
			dprintk(" unknown op %d at step %d\n",
				curstep->op, step);
			break;
		}
	}
	return result_len;
}
Commit	Line	Data
5ecc0a0f SW	1
	2	#ifdef __KERNEL__
	3	# include <linux/string.h>
	4	# include <linux/slab.h>
	5	# include <linux/bug.h>
	6	# include <linux/kernel.h>
	7	# ifndef dprintk
	8	# define dprintk(args...)
	9	# endif
	10	#else
	11	# include <string.h>
	12	# include <stdio.h>
	13	# include <stdlib.h>
	14	# include <assert.h>
	15	# define BUG_ON(x) assert(!(x))
	16	# define dprintk(args...) /* printf(args) */
	17	# define kmalloc(x, f) malloc(x)
	18	# define kfree(x) free(x)
	19	#endif
	20
3d14c5d2 YS	21	#include <linux/crush/crush.h>
3d14c5d2 YS	22	#include <linux/crush/hash.h>
feb50ac1	23	#include <linux/crush/mapper.h>
5ecc0a0f SW	24
	25	/*
	26	* Implement the core CRUSH mapping algorithm.
	27	*/
	28
	29	/**
	30	* crush_find_rule - find a crush_rule id for a given ruleset, type, and size.
	31	* @map: the crush_map
	32	* @ruleset: the storage ruleset id (user defined)
	33	* @type: storage ruleset type (user defined)
	34	* @size: output set size
	35	*/
8b12d47b	36	int crush_find_rule(const struct crush_map *map, int ruleset, int type, int size)
5ecc0a0f	37	{
8b12d47b	38	__u32 i;
5ecc0a0f SW	39
	40	for (i = 0; i < map->max_rules; i++) {
	41	if (map->rules[i] &&
	42	map->rules[i]->mask.ruleset == ruleset &&
	43	map->rules[i]->mask.type == type &&
	44	map->rules[i]->mask.min_size <= size &&
	45	map->rules[i]->mask.max_size >= size)
	46	return i;
	47	}
	48	return -1;
	49	}
	50
	51
	52	/*
	53	* bucket choose methods
	54	*
	55	* For each bucket algorithm, we have a "choose" method that, given a
	56	* crush input @x and replica position (usually, position in output set) @r,
	57	* will produce an item in the bucket.
	58	*/
	59
	60	/*
	61	* Choose based on a random permutation of the bucket.
	62	*
	63	* We used to use some prime number arithmetic to do this, but it
	64	* wasn't very random, and had some other bad behaviors. Instead, we
	65	* calculate an actual random permutation of the bucket members.
	66	* Since this is expensive, we optimize for the r=0 case, which
	67	* captures the vast majority of calls.
	68	*/
	69	static int bucket_perm_choose(struct crush_bucket *bucket,
	70	int x, int r)
	71	{
95c96174 ED	72	unsigned int pr = r % bucket->size;
95c96174 ED	73	unsigned int i, s;
5ecc0a0f SW	74
5ecc0a0f SW	75	/* start a new permutation if @x has changed */
8b12d47b	76	if (bucket->perm_x != (__u32)x \|\| bucket->perm_n == 0) {
5ecc0a0f SW	77	dprintk("bucket %d new x=%d\n", bucket->id, x);
	78	bucket->perm_x = x;
	79
	80	/* optimize common r=0 case */
	81	if (pr == 0) {
fb690390	82	s = crush_hash32_3(bucket->hash, x, bucket->id, 0) %
5ecc0a0f SW	83	bucket->size;
	84	bucket->perm[0] = s;
	85	bucket->perm_n = 0xffff; /* magic value, see below */
	86	goto out;
	87	}
	88
	89	for (i = 0; i < bucket->size; i++)
	90	bucket->perm[i] = i;
	91	bucket->perm_n = 0;
	92	} else if (bucket->perm_n == 0xffff) {
	93	/* clean up after the r=0 case above */
	94	for (i = 1; i < bucket->size; i++)
	95	bucket->perm[i] = i;
	96	bucket->perm[bucket->perm[0]] = 0;
	97	bucket->perm_n = 1;
	98	}
	99
	100	/* calculate permutation up to pr */
	101	for (i = 0; i < bucket->perm_n; i++)
	102	dprintk(" perm_choose have %d: %d\n", i, bucket->perm[i]);
	103	while (bucket->perm_n <= pr) {
95c96174	104	unsigned int p = bucket->perm_n;
5ecc0a0f SW	105	/* no point in swapping the final entry */
5ecc0a0f SW	106	if (p < bucket->size - 1) {
fb690390	107	i = crush_hash32_3(bucket->hash, x, bucket->id, p) %
5ecc0a0f SW	108	(bucket->size - p);
5ecc0a0f SW	109	if (i) {
95c96174	110	unsigned int t = bucket->perm[p + i];
5ecc0a0f SW	111	bucket->perm[p + i] = bucket->perm[p];
	112	bucket->perm[p] = t;
	113	}
	114	dprintk(" perm_choose swap %d with %d\n", p, p+i);
	115	}
	116	bucket->perm_n++;
	117	}
	118	for (i = 0; i < bucket->size; i++)
	119	dprintk(" perm_choose %d: %d\n", i, bucket->perm[i]);
	120
	121	s = bucket->perm[pr];
	122	out:
	123	dprintk(" perm_choose %d sz=%d x=%d r=%d (%d) s=%d\n", bucket->id,
	124	bucket->size, x, r, pr, s);
	125	return bucket->items[s];
	126	}
	127
	128	/* uniform */
	129	static int bucket_uniform_choose(struct crush_bucket_uniform *bucket,
	130	int x, int r)
	131	{
	132	return bucket_perm_choose(&bucket->h, x, r);
	133	}
	134
	135	/* list */
	136	static int bucket_list_choose(struct crush_bucket_list *bucket,
	137	int x, int r)
	138	{
	139	int i;
	140
	141	for (i = bucket->h.size-1; i >= 0; i--) {
fb690390 SW	142	__u64 w = crush_hash32_4(bucket->h.hash,x, bucket->h.items[i],
fb690390 SW	143	r, bucket->h.id);
5ecc0a0f SW	144	w &= 0xffff;
	145	dprintk("list_choose i=%d x=%d r=%d item %d weight %x "
	146	"sw %x rand %llx",
	147	i, x, r, bucket->h.items[i], bucket->item_weights[i],
	148	bucket->sum_weights[i], w);
	149	w *= bucket->sum_weights[i];
	150	w = w >> 16;
	151	/dprintk(" scaled %llx\n", w);/
	152	if (w < bucket->item_weights[i])
	153	return bucket->h.items[i];
	154	}
	155
a1f4895b SW	156	dprintk("bad list sums for bucket %d\n", bucket->h.id);
a1f4895b SW	157	return bucket->h.items[0];
5ecc0a0f SW	158	}
	159
	160
	161	/* (binary) tree */
	162	static int height(int n)
	163	{
	164	int h = 0;
	165	while ((n & 1) == 0) {
	166	h++;
	167	n = n >> 1;
	168	}
	169	return h;
	170	}
	171
	172	static int left(int x)
	173	{
	174	int h = height(x);
	175	return x - (1 << (h-1));
	176	}
	177
	178	static int right(int x)
	179	{
	180	int h = height(x);
	181	return x + (1 << (h-1));
	182	}
	183
	184	static int terminal(int x)
	185	{
	186	return x & 1;
	187	}
	188
	189	static int bucket_tree_choose(struct crush_bucket_tree *bucket,
	190	int x, int r)
	191	{
	192	int n, l;
	193	__u32 w;
	194	__u64 t;
	195
	196	/* start at root */
	197	n = bucket->num_nodes >> 1;
	198
	199	while (!terminal(n)) {
	200	/* pick point in [0, w) */
	201	w = bucket->node_weights[n];
fb690390 SW	202	t = (__u64)crush_hash32_4(bucket->h.hash, x, n, r,
fb690390 SW	203	bucket->h.id) * (__u64)w;
5ecc0a0f SW	204	t = t >> 32;
	205
	206	/* descend to the left or right? */
	207	l = left(n);
	208	if (t < bucket->node_weights[l])
	209	n = l;
	210	else
	211	n = right(n);
	212	}
	213
	214	return bucket->h.items[n >> 1];
	215	}
	216
	217
	218	/* straw */
	219
	220	static int bucket_straw_choose(struct crush_bucket_straw *bucket,
	221	int x, int r)
	222	{
8b12d47b	223	__u32 i;
5ecc0a0f SW	224	int high = 0;
	225	__u64 high_draw = 0;
	226	__u64 draw;
	227
	228	for (i = 0; i < bucket->h.size; i++) {
fb690390	229	draw = crush_hash32_3(bucket->h.hash, x, bucket->h.items[i], r);
5ecc0a0f SW	230	draw &= 0xffff;
	231	draw *= bucket->straws[i];
	232	if (i == 0 \|\| draw > high_draw) {
	233	high = i;
	234	high_draw = draw;
	235	}
	236	}
	237	return bucket->h.items[high];
	238	}
	239
	240	static int crush_bucket_choose(struct crush_bucket *in, int x, int r)
	241	{
a1a31e73	242	dprintk(" crush_bucket_choose %d x=%d r=%d\n", in->id, x, r);
a1f4895b	243	BUG_ON(in->size == 0);
5ecc0a0f SW	244	switch (in->alg) {
	245	case CRUSH_BUCKET_UNIFORM:
	246	return bucket_uniform_choose((struct crush_bucket_uniform *)in,
	247	x, r);
	248	case CRUSH_BUCKET_LIST:
	249	return bucket_list_choose((struct crush_bucket_list *)in,
	250	x, r);
	251	case CRUSH_BUCKET_TREE:
	252	return bucket_tree_choose((struct crush_bucket_tree *)in,
	253	x, r);
	254	case CRUSH_BUCKET_STRAW:
	255	return bucket_straw_choose((struct crush_bucket_straw *)in,
	256	x, r);
	257	default:
a1f4895b	258	dprintk("unknown bucket %d alg %d\n", in->id, in->alg);
50b885b9	259	return in->items[0];
5ecc0a0f SW	260	}
	261	}
	262
	263	/*
	264	* true if device is marked "out" (failed, fully offloaded)
	265	* of the cluster
	266	*/
8b12d47b	267	static int is_out(const struct crush_map map, const __u32 weight, int item, int x)
5ecc0a0f	268	{
153a1093	269	if (weight[item] >= 0x10000)
5ecc0a0f SW	270	return 0;
	271	if (weight[item] == 0)
	272	return 1;
fb690390 SW	273	if ((crush_hash32_2(CRUSH_HASH_RJENKINS1, x, item) & 0xffff)
fb690390 SW	274	< weight[item])
5ecc0a0f SW	275	return 0;
	276	return 1;
	277	}
	278
	279	/**
	280	* crush_choose - choose numrep distinct items of given type
	281	* @map: the crush_map
	282	* @bucket: the bucket we are choose an item from
	283	* @x: crush input value
	284	* @numrep: the number of items to choose
	285	* @type: the type of item to choose
	286	* @out: pointer to output vector
	287	* @outpos: our position in that vector
	288	* @firstn: true if choosing "first n" items, false if choosing "indep"
	289	* @recurse_to_leaf: true if we want one device under each item of given type
	290	* @out2: second output vector for leaf items (if @recurse_to_leaf)
	291	*/
8b12d47b	292	static int crush_choose(const struct crush_map *map,
5ecc0a0f	293	struct crush_bucket *bucket,
8b12d47b	294	const __u32 *weight,
5ecc0a0f SW	295	int x, int numrep, int type,
	296	int *out, int outpos,
	297	int firstn, int recurse_to_leaf,
	298	int *out2)
	299	{
	300	int rep;
8b12d47b	301	unsigned int ftotal, flocal;
5ecc0a0f SW	302	int retry_descent, retry_bucket, skip_rep;
	303	struct crush_bucket *in = bucket;
	304	int r;
	305	int i;
b28813a6	306	int item = 0;
5ecc0a0f SW	307	int itemtype;
5ecc0a0f SW	308	int collide, reject;
a1a31e73 SW	309
	310	dprintk("CHOOSE%s bucket %d x %d outpos %d numrep %d\n", recurse_to_leaf ? "_LEAF" : "",
	311	bucket->id, x, outpos, numrep);
5ecc0a0f SW	312
	313	for (rep = outpos; rep < numrep; rep++) {
	314	/* keep trying until we get a non-out, non-colliding item */
	315	ftotal = 0;
	316	skip_rep = 0;
	317	do {
	318	retry_descent = 0;
	319	in = bucket; /* initial bucket */
	320
	321	/* choose through intervening buckets */
	322	flocal = 0;
	323	do {
b28813a6	324	collide = 0;
5ecc0a0f SW	325	retry_bucket = 0;
	326	r = rep;
	327	if (in->alg == CRUSH_BUCKET_UNIFORM) {
	328	/* be careful */
8b12d47b	329	if (firstn \|\| (__u32)numrep >= in->size)
5ecc0a0f SW	330	/* r' = r + f_total */
	331	r += ftotal;
	332	else if (in->size % numrep == 0)
	333	/* r'=r+(n+1)f_local /
	334	r += (numrep+1) *
	335	(flocal+ftotal);
	336	else
	337	/* r' = r + nf_local /
	338	r += numrep * (flocal+ftotal);
	339	} else {
	340	if (firstn)
	341	/* r' = r + f_total */
	342	r += ftotal;
	343	else
	344	/* r' = r + nf_local /
	345	r += numrep * (flocal+ftotal);
	346	}
	347
	348	/* bucket choose */
b28813a6 SW	349	if (in->size == 0) {
	350	reject = 1;
	351	goto reject;
	352	}
546f04ef SW	353	if (map->choose_local_fallback_tries > 0 &&
	354	flocal >= (in->size>>1) &&
	355	flocal > map->choose_local_fallback_tries)
5ecc0a0f SW	356	item = bucket_perm_choose(in, x, r);
	357	else
	358	item = crush_bucket_choose(in, x, r);
a1f4895b SW	359	if (item >= map->max_devices) {
	360	dprintk(" bad item %d\n", item);
	361	skip_rep = 1;
	362	break;
	363	}
5ecc0a0f SW	364
	365	/* desired type? */
	366	if (item < 0)
	367	itemtype = map->buckets[-1-item]->type;
	368	else
	369	itemtype = 0;
	370	dprintk(" item %d type %d\n", item, itemtype);
	371
	372	/* keep going? */
	373	if (itemtype != type) {
a1f4895b SW	374	if (item >= 0 \|\|
	375	(-1-item) >= map->max_buckets) {
	376	dprintk(" bad item type %d\n", type);
	377	skip_rep = 1;
	378	break;
	379	}
5ecc0a0f	380	in = map->buckets[-1-item];
55bda7aa	381	retry_bucket = 1;
5ecc0a0f SW	382	continue;
	383	}
	384
	385	/* collision? */
5ecc0a0f SW	386	for (i = 0; i < outpos; i++) {
	387	if (out[i] == item) {
	388	collide = 1;
	389	break;
	390	}
	391	}
	392
a1a31e73 SW	393	reject = 0;
	394	if (recurse_to_leaf) {
	395	if (item < 0) {
	396	if (crush_choose(map,
	397	map->buckets[-1-item],
	398	weight,
	399	x, outpos+1, 0,
	400	out2, outpos,
	401	firstn, 0,
	402	NULL) <= outpos)
	403	/* didn't get leaf */
	404	reject = 1;
	405	} else {
	406	/* we already have a leaf! */
	407	out2[outpos] = item;
	408	}
	409	}
	410
	411	if (!reject) {
5ecc0a0f SW	412	/* out? */
	413	if (itemtype == 0)
	414	reject = is_out(map, weight,
	415	item, x);
	416	else
	417	reject = 0;
	418	}
	419
b28813a6	420	reject:
5ecc0a0f SW	421	if (reject \|\| collide) {
	422	ftotal++;
	423	flocal++;
	424
546f04ef	425	if (collide && flocal <= map->choose_local_tries)
5ecc0a0f SW	426	/* retry locally a few times */
5ecc0a0f SW	427	retry_bucket = 1;
546f04ef SW	428	else if (map->choose_local_fallback_tries > 0 &&
546f04ef SW	429	flocal <= in->size + map->choose_local_fallback_tries)
5ecc0a0f SW	430	/* exhaustive bucket search */
5ecc0a0f SW	431	retry_bucket = 1;
546f04ef	432	else if (ftotal <= map->choose_total_tries)
5ecc0a0f SW	433	/* then retry descent */
	434	retry_descent = 1;
	435	else
	436	/* else give up */
	437	skip_rep = 1;
	438	dprintk(" reject %d collide %d "
8b12d47b	439	"ftotal %u flocal %u\n",
5ecc0a0f SW	440	reject, collide, ftotal,
	441	flocal);
	442	}
	443	} while (retry_bucket);
	444	} while (retry_descent);
	445
	446	if (skip_rep) {
	447	dprintk("skip rep\n");
	448	continue;
	449	}
	450
a1a31e73	451	dprintk("CHOOSE got %d\n", item);
5ecc0a0f SW	452	out[outpos] = item;
	453	outpos++;
	454	}
	455
a1a31e73	456	dprintk("CHOOSE returns %d\n", outpos);
5ecc0a0f SW	457	return outpos;
	458	}
	459
	460
	461	/**
	462	* crush_do_rule - calculate a mapping with the given input and rule
	463	* @map: the crush_map
	464	* @ruleno: the rule id
	465	* @x: hash input
	466	* @result: pointer to result vector
	467	* @result_max: maximum result size
5ecc0a0f	468	*/
8b12d47b	469	int crush_do_rule(const struct crush_map *map,
5ecc0a0f	470	int ruleno, int x, int *result, int result_max,
41ebcc09	471	const __u32 *weight)
5ecc0a0f SW	472	{
5ecc0a0f SW	473	int result_len;
5ecc0a0f SW	474	int a[CRUSH_MAX_SET];
	475	int b[CRUSH_MAX_SET];
	476	int c[CRUSH_MAX_SET];
	477	int recurse_to_leaf;
	478	int *w;
	479	int wsize = 0;
	480	int *o;
	481	int osize;
	482	int *tmp;
	483	struct crush_rule *rule;
8b12d47b	484	__u32 step;
5ecc0a0f SW	485	int i, j;
	486	int numrep;
	487	int firstn;
5ecc0a0f	488
a1f4895b SW	489	if ((__u32)ruleno >= map->max_rules) {
	490	dprintk(" bad ruleno %d\n", ruleno);
	491	return 0;
	492	}
5ecc0a0f SW	493
	494	rule = map->rules[ruleno];
	495	result_len = 0;
	496	w = a;
	497	o = b;
	498
5ecc0a0f	499	for (step = 0; step < rule->len; step++) {
0668216e SW	500	struct crush_rule_step *curstep = &rule->steps[step];
0668216e SW	501
5ecc0a0f	502	firstn = 0;
0668216e	503	switch (curstep->op) {
5ecc0a0f	504	case CRUSH_RULE_TAKE:
0668216e	505	w[0] = curstep->arg1;
5ecc0a0f SW	506	wsize = 1;
	507	break;
	508
	509	case CRUSH_RULE_CHOOSE_LEAF_FIRSTN:
	510	case CRUSH_RULE_CHOOSE_FIRSTN:
	511	firstn = 1;
0668216e	512	/* fall through */
5ecc0a0f SW	513	case CRUSH_RULE_CHOOSE_LEAF_INDEP:
5ecc0a0f SW	514	case CRUSH_RULE_CHOOSE_INDEP:
a1f4895b SW	515	if (wsize == 0)
a1f4895b SW	516	break;
5ecc0a0f SW	517
5ecc0a0f SW	518	recurse_to_leaf =
0668216e	519	curstep->op ==
5ecc0a0f	520	CRUSH_RULE_CHOOSE_LEAF_FIRSTN \|\|
0668216e	521	curstep->op ==
5ecc0a0f SW	522	CRUSH_RULE_CHOOSE_LEAF_INDEP;
	523
	524	/* reset output */
	525	osize = 0;
	526
	527	for (i = 0; i < wsize; i++) {
	528	/*
	529	* see CRUSH_N, CRUSH_N_MINUS macros.
	530	* basically, numrep <= 0 means relative to
	531	* the provided result_max
	532	*/
0668216e	533	numrep = curstep->arg1;
5ecc0a0f SW	534	if (numrep <= 0) {
	535	numrep += result_max;
	536	if (numrep <= 0)
	537	continue;
	538	}
	539	j = 0;
5ecc0a0f SW	540	osize += crush_choose(map,
	541	map->buckets[-1-w[i]],
	542	weight,
	543	x, numrep,
0668216e	544	curstep->arg2,
5ecc0a0f SW	545	o+osize, j,
	546	firstn,
	547	recurse_to_leaf, c+osize);
	548	}
	549
	550	if (recurse_to_leaf)
	551	/* copy final _leaf_ values to output set */
	552	memcpy(o, c, osizesizeof(o));
	553
	554	/* swap t and w arrays */
	555	tmp = o;
	556	o = w;
	557	w = tmp;
	558	wsize = osize;
	559	break;
	560
	561
	562	case CRUSH_RULE_EMIT:
	563	for (i = 0; i < wsize && result_len < result_max; i++) {
	564	result[result_len] = w[i];
	565	result_len++;
	566	}
	567	wsize = 0;
	568	break;
	569
	570	default:
a1f4895b SW	571	dprintk(" unknown op %d at step %d\n",
	572	curstep->op, step);
	573	break;
5ecc0a0f SW	574	}
5ecc0a0f SW	575	}
f1932fc1	576	return result_len;
5ecc0a0f SW	577	}
	578
	579