mdsmap.c 4.51 KB
Newer Older
1
#include <linux/ceph/ceph_debug.h>
Sage Weil's avatar
Sage Weil committed
2 3 4 5 6 7 8

#include <linux/bug.h>
#include <linux/err.h>
#include <linux/random.h>
#include <linux/slab.h>
#include <linux/types.h>

9 10 11
#include <linux/ceph/mdsmap.h>
#include <linux/ceph/messenger.h>
#include <linux/ceph/decode.h>
Sage Weil's avatar
Sage Weil committed
12 13 14 15 16 17 18 19 20 21 22

#include "super.h"


/*
 * choose a random mds that is "up" (i.e. has a state > 0), or -1.
 */
int ceph_mdsmap_get_random_mds(struct ceph_mdsmap *m)
{
	int n = 0;
	int i;
23 24 25 26

	/* special case for one mds */
	if (1 == m->m_max_mds && m->m_info[0].state > 0)
		return 0;
Sage Weil's avatar
Sage Weil committed
27 28 29 30 31 32 33 34 35

	/* count */
	for (i = 0; i < m->m_max_mds; i++)
		if (m->m_info[i].state > 0)
			n++;
	if (n == 0)
		return -1;

	/* pick */
36
	n = prandom_u32() % n;
Sage Weil's avatar
Sage Weil committed
37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53
	i = 0;
	for (i = 0; n > 0; i++, n--)
		while (m->m_info[i].state <= 0)
			i++;

	return i;
}

/*
 * Decode an MDS map
 *
 * Ignore any fields we don't care about (there are quite a few of
 * them).
 */
struct ceph_mdsmap *ceph_mdsmap_decode(void **p, void *end)
{
	struct ceph_mdsmap *m;
54
	const void *start = *p;
Sage Weil's avatar
Sage Weil committed
55 56 57 58 59 60 61 62 63
	int i, j, n;
	int err = -EINVAL;
	u16 version;

	m = kzalloc(sizeof(*m), GFP_NOFS);
	if (m == NULL)
		return ERR_PTR(-ENOMEM);

	ceph_decode_16_safe(p, end, version, bad);
64 65 66 67
	if (version > 3) {
		pr_warning("got mdsmap version %d > 3, failing", version);
		goto bad;
	}
Sage Weil's avatar
Sage Weil committed
68 69

	ceph_decode_need(p, end, 8*sizeof(u32) + sizeof(u64), bad);
70 71 72 73 74 75 76 77
	m->m_epoch = ceph_decode_32(p);
	m->m_client_epoch = ceph_decode_32(p);
	m->m_last_failure = ceph_decode_32(p);
	m->m_root = ceph_decode_32(p);
	m->m_session_timeout = ceph_decode_32(p);
	m->m_session_autoclose = ceph_decode_32(p);
	m->m_max_file_size = ceph_decode_64(p);
	m->m_max_mds = ceph_decode_32(p);
Sage Weil's avatar
Sage Weil committed
78 79 80 81 82 83

	m->m_info = kcalloc(m->m_max_mds, sizeof(*m->m_info), GFP_NOFS);
	if (m->m_info == NULL)
		goto badmem;

	/* pick out active nodes from mds_info (state > 0) */
84
	n = ceph_decode_32(p);
Sage Weil's avatar
Sage Weil committed
85
	for (i = 0; i < n; i++) {
86
		u64 global_id;
Sage Weil's avatar
Sage Weil committed
87 88 89 90 91 92 93
		u32 namelen;
		s32 mds, inc, state;
		u64 state_seq;
		u8 infoversion;
		struct ceph_entity_addr addr;
		u32 num_export_targets;
		void *pexport_targets = NULL;
94
		struct ceph_timespec laggy_since;
95
		struct ceph_mds_info *info;
Sage Weil's avatar
Sage Weil committed
96

97 98
		ceph_decode_need(p, end, sizeof(u64)*2 + 1 + sizeof(u32), bad);
		global_id = ceph_decode_64(p);
99
		infoversion = ceph_decode_8(p);
100
		*p += sizeof(u64);
101
		namelen = ceph_decode_32(p);  /* skip mds name */
Sage Weil's avatar
Sage Weil committed
102 103 104
		*p += namelen;

		ceph_decode_need(p, end,
105
				 4*sizeof(u32) + sizeof(u64) +
Sage Weil's avatar
Sage Weil committed
106 107
				 sizeof(addr) + sizeof(struct ceph_timespec),
				 bad);
108 109 110 111
		mds = ceph_decode_32(p);
		inc = ceph_decode_32(p);
		state = ceph_decode_32(p);
		state_seq = ceph_decode_64(p);
112 113
		ceph_decode_copy(p, &addr, sizeof(addr));
		ceph_decode_addr(&addr);
114
		ceph_decode_copy(p, &laggy_since, sizeof(laggy_since));
Sage Weil's avatar
Sage Weil committed
115 116
		*p += sizeof(u32);
		ceph_decode_32_safe(p, end, namelen, bad);
117
		*p += namelen;
Sage Weil's avatar
Sage Weil committed
118 119 120
		if (infoversion >= 2) {
			ceph_decode_32_safe(p, end, num_export_targets, bad);
			pexport_targets = *p;
121
			*p += num_export_targets * sizeof(u32);
Sage Weil's avatar
Sage Weil committed
122 123 124 125
		} else {
			num_export_targets = 0;
		}

126
		dout("mdsmap_decode %d/%d %lld mds%d.%d %s %s\n",
127 128
		     i+1, n, global_id, mds, inc,
		     ceph_pr_addr(&addr.in_addr),
Sage Weil's avatar
Sage Weil committed
129
		     ceph_mds_state_name(state));
130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150

		if (mds < 0 || mds >= m->m_max_mds || state <= 0)
			continue;

		info = &m->m_info[mds];
		info->global_id = global_id;
		info->state = state;
		info->addr = addr;
		info->laggy = (laggy_since.tv_sec != 0 ||
			       laggy_since.tv_nsec != 0);
		info->num_export_targets = num_export_targets;
		if (num_export_targets) {
			info->export_targets = kcalloc(num_export_targets,
						       sizeof(u32), GFP_NOFS);
			if (info->export_targets == NULL)
				goto badmem;
			for (j = 0; j < num_export_targets; j++)
				info->export_targets[j] =
				       ceph_decode_32(&pexport_targets);
		} else {
			info->export_targets = NULL;
Sage Weil's avatar
Sage Weil committed
151 152 153 154 155 156
		}
	}

	/* pg_pools */
	ceph_decode_32_safe(p, end, n, bad);
	m->m_num_data_pg_pools = n;
157
	m->m_data_pg_pools = kcalloc(n, sizeof(u64), GFP_NOFS);
Sage Weil's avatar
Sage Weil committed
158 159
	if (!m->m_data_pg_pools)
		goto badmem;
160
	ceph_decode_need(p, end, sizeof(u64)*(n+1), bad);
Sage Weil's avatar
Sage Weil committed
161
	for (i = 0; i < n; i++)
162 163
		m->m_data_pg_pools[i] = ceph_decode_64(p);
	m->m_cas_pg_pool = ceph_decode_64(p);
Sage Weil's avatar
Sage Weil committed
164 165 166 167 168 169 170 171 172

	/* ok, we don't care about the rest. */
	dout("mdsmap_decode success epoch %u\n", m->m_epoch);
	return m;

badmem:
	err = -ENOMEM;
bad:
	pr_err("corrupt mdsmap\n");
173 174 175
	print_hex_dump(KERN_DEBUG, "mdsmap: ",
		       DUMP_PREFIX_OFFSET, 16, 1,
		       start, end - start, true);
Sage Weil's avatar
Sage Weil committed
176
	ceph_mdsmap_destroy(m);
177
	return ERR_PTR(err);
Sage Weil's avatar
Sage Weil committed
178 179 180 181 182 183 184 185 186 187 188 189
}

void ceph_mdsmap_destroy(struct ceph_mdsmap *m)
{
	int i;

	for (i = 0; i < m->m_max_mds; i++)
		kfree(m->m_info[i].export_targets);
	kfree(m->m_info);
	kfree(m->m_data_pg_pools);
	kfree(m);
}