xattr.c 27.6 KB
Newer Older
1
#include <linux/ceph/ceph_debug.h>
2
#include <linux/ceph/pagelist.h>
3

S
Sage Weil 已提交
4
#include "super.h"
5 6 7
#include "mds_client.h"

#include <linux/ceph/decode.h>
S
Sage Weil 已提交
8 9

#include <linux/xattr.h>
10
#include <linux/posix_acl_xattr.h>
11
#include <linux/slab.h>
S
Sage Weil 已提交
12

13 14 15
#define XATTR_CEPH_PREFIX "ceph."
#define XATTR_CEPH_PREFIX_LEN (sizeof (XATTR_CEPH_PREFIX) - 1)

16 17 18
static int __remove_xattr(struct ceph_inode_info *ci,
			  struct ceph_inode_xattr *xattr);

W
Wei Yongjun 已提交
19
static const struct xattr_handler ceph_other_xattr_handler;
20

G
Guangliang Zhao 已提交
21 22 23 24 25 26
/*
 * List of handlers for synthetic system.* attributes. Other
 * attributes are handled directly.
 */
const struct xattr_handler *ceph_xattr_handlers[] = {
#ifdef CONFIG_CEPH_FS_POSIX_ACL
27 28
	&posix_acl_access_xattr_handler,
	&posix_acl_default_xattr_handler,
G
Guangliang Zhao 已提交
29
#endif
30
	&ceph_other_xattr_handler,
G
Guangliang Zhao 已提交
31 32 33
	NULL,
};

S
Sage Weil 已提交
34 35
static bool ceph_is_valid_xattr(const char *name)
{
36
	return !strncmp(name, XATTR_CEPH_PREFIX, XATTR_CEPH_PREFIX_LEN) ||
37
	       !strncmp(name, XATTR_SECURITY_PREFIX,
S
Sage Weil 已提交
38 39 40 41 42 43 44 45 46
			XATTR_SECURITY_PREFIX_LEN) ||
	       !strncmp(name, XATTR_TRUSTED_PREFIX, XATTR_TRUSTED_PREFIX_LEN) ||
	       !strncmp(name, XATTR_USER_PREFIX, XATTR_USER_PREFIX_LEN);
}

/*
 * These define virtual xattrs exposing the recursive directory
 * statistics and layout metadata.
 */
47
struct ceph_vxattr {
S
Sage Weil 已提交
48
	char *name;
49
	size_t name_size;	/* strlen(name) + 1 (for '\0') */
S
Sage Weil 已提交
50 51
	size_t (*getxattr_cb)(struct ceph_inode_info *ci, char *val,
			      size_t size);
S
Sage Weil 已提交
52
	bool readonly, hidden;
S
Sage Weil 已提交
53
	bool (*exists_cb)(struct ceph_inode_info *ci);
S
Sage Weil 已提交
54 55
};

56 57 58 59
/* layouts */

static bool ceph_vxattrcb_layout_exists(struct ceph_inode_info *ci)
{
Y
Yan, Zheng 已提交
60 61 62 63
	struct ceph_file_layout *fl = &ci->i_layout;
	return (fl->stripe_unit > 0 || fl->stripe_count > 0 ||
		fl->object_size > 0 || fl->pool_id >= 0 ||
		rcu_dereference_raw(fl->pool_ns) != NULL);
64 65 66
}

static size_t ceph_vxattrcb_layout(struct ceph_inode_info *ci, char *val,
67
				   size_t size)
68 69 70
{
	struct ceph_fs_client *fsc = ceph_sb_to_client(ci->vfs_inode.i_sb);
	struct ceph_osd_client *osdc = &fsc->client->osdc;
Y
Yan, Zheng 已提交
71
	struct ceph_string *pool_ns;
72
	s64 pool = ci->i_layout.pool_id;
73
	const char *pool_name;
Y
Yan, Zheng 已提交
74
	const char *ns_field = " pool_namespace=";
75
	char buf[128];
Y
Yan, Zheng 已提交
76 77 78 79
	size_t len, total_len = 0;
	int ret;

	pool_ns = ceph_try_get_string(ci->i_layout.pool_ns);
80 81

	dout("ceph_vxattrcb_layout %p\n", &ci->vfs_inode);
82
	down_read(&osdc->lock);
83
	pool_name = ceph_pg_pool_name_by_id(osdc->osdmap, pool);
84
	if (pool_name) {
Y
Yan, Zheng 已提交
85
		len = snprintf(buf, sizeof(buf),
86 87 88
		"stripe_unit=%u stripe_count=%u object_size=%u pool=",
		ci->i_layout.stripe_unit, ci->i_layout.stripe_count,
	        ci->i_layout.object_size);
Y
Yan, Zheng 已提交
89
		total_len = len + strlen(pool_name);
90
	} else {
Y
Yan, Zheng 已提交
91
		len = snprintf(buf, sizeof(buf),
92 93 94
		"stripe_unit=%u stripe_count=%u object_size=%u pool=%lld",
		ci->i_layout.stripe_unit, ci->i_layout.stripe_count,
	        ci->i_layout.object_size, (unsigned long long)pool);
Y
Yan, Zheng 已提交
95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118
		total_len = len;
	}

	if (pool_ns)
		total_len += strlen(ns_field) + pool_ns->len;

	if (!size) {
		ret = total_len;
	} else if (total_len > size) {
		ret = -ERANGE;
	} else {
		memcpy(val, buf, len);
		ret = len;
		if (pool_name) {
			len = strlen(pool_name);
			memcpy(val + ret, pool_name, len);
			ret += len;
		}
		if (pool_ns) {
			len = strlen(ns_field);
			memcpy(val + ret, ns_field, len);
			ret += len;
			memcpy(val + ret, pool_ns->str, pool_ns->len);
			ret += pool_ns->len;
119 120
		}
	}
121
	up_read(&osdc->lock);
Y
Yan, Zheng 已提交
122
	ceph_put_string(pool_ns);
123 124 125
	return ret;
}

126 127 128
static size_t ceph_vxattrcb_layout_stripe_unit(struct ceph_inode_info *ci,
					       char *val, size_t size)
{
129
	return snprintf(val, size, "%u", ci->i_layout.stripe_unit);
130 131 132 133 134
}

static size_t ceph_vxattrcb_layout_stripe_count(struct ceph_inode_info *ci,
						char *val, size_t size)
{
135
	return snprintf(val, size, "%u", ci->i_layout.stripe_count);
136 137 138 139 140
}

static size_t ceph_vxattrcb_layout_object_size(struct ceph_inode_info *ci,
					       char *val, size_t size)
{
141
	return snprintf(val, size, "%u", ci->i_layout.object_size);
142 143 144 145 146 147 148 149
}

static size_t ceph_vxattrcb_layout_pool(struct ceph_inode_info *ci,
					char *val, size_t size)
{
	int ret;
	struct ceph_fs_client *fsc = ceph_sb_to_client(ci->vfs_inode.i_sb);
	struct ceph_osd_client *osdc = &fsc->client->osdc;
150
	s64 pool = ci->i_layout.pool_id;
151 152
	const char *pool_name;

153
	down_read(&osdc->lock);
154 155 156 157 158
	pool_name = ceph_pg_pool_name_by_id(osdc->osdmap, pool);
	if (pool_name)
		ret = snprintf(val, size, "%s", pool_name);
	else
		ret = snprintf(val, size, "%lld", (unsigned long long)pool);
159
	up_read(&osdc->lock);
160 161 162
	return ret;
}

Y
Yan, Zheng 已提交
163 164 165 166 167 168 169 170 171 172 173 174
static size_t ceph_vxattrcb_layout_pool_namespace(struct ceph_inode_info *ci,
						  char *val, size_t size)
{
	int ret = 0;
	struct ceph_string *ns = ceph_try_get_string(ci->i_layout.pool_ns);
	if (ns) {
		ret = snprintf(val, size, "%.*s", (int)ns->len, ns->str);
		ceph_put_string(ns);
	}
	return ret;
}

S
Sage Weil 已提交
175 176
/* directories */

177
static size_t ceph_vxattrcb_dir_entries(struct ceph_inode_info *ci, char *val,
S
Sage Weil 已提交
178 179 180 181 182
					size_t size)
{
	return snprintf(val, size, "%lld", ci->i_files + ci->i_subdirs);
}

183
static size_t ceph_vxattrcb_dir_files(struct ceph_inode_info *ci, char *val,
S
Sage Weil 已提交
184 185 186 187 188
				      size_t size)
{
	return snprintf(val, size, "%lld", ci->i_files);
}

189
static size_t ceph_vxattrcb_dir_subdirs(struct ceph_inode_info *ci, char *val,
S
Sage Weil 已提交
190 191 192 193 194
					size_t size)
{
	return snprintf(val, size, "%lld", ci->i_subdirs);
}

195
static size_t ceph_vxattrcb_dir_rentries(struct ceph_inode_info *ci, char *val,
S
Sage Weil 已提交
196 197 198 199 200
					 size_t size)
{
	return snprintf(val, size, "%lld", ci->i_rfiles + ci->i_rsubdirs);
}

201
static size_t ceph_vxattrcb_dir_rfiles(struct ceph_inode_info *ci, char *val,
S
Sage Weil 已提交
202 203 204 205 206
				       size_t size)
{
	return snprintf(val, size, "%lld", ci->i_rfiles);
}

207
static size_t ceph_vxattrcb_dir_rsubdirs(struct ceph_inode_info *ci, char *val,
S
Sage Weil 已提交
208 209 210 211 212
					 size_t size)
{
	return snprintf(val, size, "%lld", ci->i_rsubdirs);
}

213
static size_t ceph_vxattrcb_dir_rbytes(struct ceph_inode_info *ci, char *val,
S
Sage Weil 已提交
214 215 216 217 218
				       size_t size)
{
	return snprintf(val, size, "%lld", ci->i_rbytes);
}

219
static size_t ceph_vxattrcb_dir_rctime(struct ceph_inode_info *ci, char *val,
S
Sage Weil 已提交
220 221
				       size_t size)
{
222
	return snprintf(val, size, "%ld.09%ld", (long)ci->i_rctime.tv_sec,
S
Sage Weil 已提交
223 224 225
			(long)ci->i_rctime.tv_nsec);
}

226

227
#define CEPH_XATTR_NAME(_type, _name)	XATTR_CEPH_PREFIX #_type "." #_name
228 229
#define CEPH_XATTR_NAME2(_type, _name, _name2)	\
	XATTR_CEPH_PREFIX #_type "." #_name "." #_name2
230

S
Sage Weil 已提交
231 232 233 234 235 236 237
#define XATTR_NAME_CEPH(_type, _name)					\
	{								\
		.name = CEPH_XATTR_NAME(_type, _name),			\
		.name_size = sizeof (CEPH_XATTR_NAME(_type, _name)), \
		.getxattr_cb = ceph_vxattrcb_ ## _type ## _ ## _name, \
		.readonly = true,				\
		.hidden = false,				\
S
Sage Weil 已提交
238
		.exists_cb = NULL,			\
S
Sage Weil 已提交
239
	}
240 241 242 243 244 245 246 247 248
#define XATTR_LAYOUT_FIELD(_type, _name, _field)			\
	{								\
		.name = CEPH_XATTR_NAME2(_type, _name, _field),	\
		.name_size = sizeof (CEPH_XATTR_NAME2(_type, _name, _field)), \
		.getxattr_cb = ceph_vxattrcb_ ## _name ## _ ## _field, \
		.readonly = false,				\
		.hidden = true,			\
		.exists_cb = ceph_vxattrcb_layout_exists,	\
	}
249

250
static struct ceph_vxattr ceph_dir_vxattrs[] = {
S
Sage Weil 已提交
251 252 253 254 255
	{
		.name = "ceph.dir.layout",
		.name_size = sizeof("ceph.dir.layout"),
		.getxattr_cb = ceph_vxattrcb_layout,
		.readonly = false,
256
		.hidden = true,
S
Sage Weil 已提交
257 258
		.exists_cb = ceph_vxattrcb_layout_exists,
	},
259 260 261 262
	XATTR_LAYOUT_FIELD(dir, layout, stripe_unit),
	XATTR_LAYOUT_FIELD(dir, layout, stripe_count),
	XATTR_LAYOUT_FIELD(dir, layout, object_size),
	XATTR_LAYOUT_FIELD(dir, layout, pool),
Y
Yan, Zheng 已提交
263
	XATTR_LAYOUT_FIELD(dir, layout, pool_namespace),
264 265 266 267 268 269 270 271
	XATTR_NAME_CEPH(dir, entries),
	XATTR_NAME_CEPH(dir, files),
	XATTR_NAME_CEPH(dir, subdirs),
	XATTR_NAME_CEPH(dir, rentries),
	XATTR_NAME_CEPH(dir, rfiles),
	XATTR_NAME_CEPH(dir, rsubdirs),
	XATTR_NAME_CEPH(dir, rbytes),
	XATTR_NAME_CEPH(dir, rctime),
272
	{ .name = NULL, 0 }	/* Required table terminator */
S
Sage Weil 已提交
273
};
274
static size_t ceph_dir_vxattrs_name_size;	/* total size of all names */
S
Sage Weil 已提交
275 276 277

/* files */

278
static struct ceph_vxattr ceph_file_vxattrs[] = {
279 280 281 282 283
	{
		.name = "ceph.file.layout",
		.name_size = sizeof("ceph.file.layout"),
		.getxattr_cb = ceph_vxattrcb_layout,
		.readonly = false,
284
		.hidden = true,
285 286
		.exists_cb = ceph_vxattrcb_layout_exists,
	},
287 288 289 290
	XATTR_LAYOUT_FIELD(file, layout, stripe_unit),
	XATTR_LAYOUT_FIELD(file, layout, stripe_count),
	XATTR_LAYOUT_FIELD(file, layout, object_size),
	XATTR_LAYOUT_FIELD(file, layout, pool),
Y
Yan, Zheng 已提交
291
	XATTR_LAYOUT_FIELD(file, layout, pool_namespace),
292
	{ .name = NULL, 0 }	/* Required table terminator */
S
Sage Weil 已提交
293
};
294
static size_t ceph_file_vxattrs_name_size;	/* total size of all names */
S
Sage Weil 已提交
295

296
static struct ceph_vxattr *ceph_inode_vxattrs(struct inode *inode)
S
Sage Weil 已提交
297 298 299 300 301 302 303 304
{
	if (S_ISDIR(inode->i_mode))
		return ceph_dir_vxattrs;
	else if (S_ISREG(inode->i_mode))
		return ceph_file_vxattrs;
	return NULL;
}

305 306 307 308 309 310
static size_t ceph_vxattrs_name_size(struct ceph_vxattr *vxattrs)
{
	if (vxattrs == ceph_dir_vxattrs)
		return ceph_dir_vxattrs_name_size;
	if (vxattrs == ceph_file_vxattrs)
		return ceph_file_vxattrs_name_size;
Y
Yan, Zheng 已提交
311
	BUG_ON(vxattrs);
312 313 314 315 316 317 318 319 320 321 322 323 324
	return 0;
}

/*
 * Compute the aggregate size (including terminating '\0') of all
 * virtual extended attribute names in the given vxattr table.
 */
static size_t __init vxattrs_name_size(struct ceph_vxattr *vxattrs)
{
	struct ceph_vxattr *vxattr;
	size_t size = 0;

	for (vxattr = vxattrs; vxattr->name; vxattr++)
S
Sage Weil 已提交
325 326
		if (!vxattr->hidden)
			size += vxattr->name_size;
327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344

	return size;
}

/* Routines called at initialization and exit time */

void __init ceph_xattr_init(void)
{
	ceph_dir_vxattrs_name_size = vxattrs_name_size(ceph_dir_vxattrs);
	ceph_file_vxattrs_name_size = vxattrs_name_size(ceph_file_vxattrs);
}

void ceph_xattr_exit(void)
{
	ceph_dir_vxattrs_name_size = 0;
	ceph_file_vxattrs_name_size = 0;
}

345
static struct ceph_vxattr *ceph_match_vxattr(struct inode *inode,
S
Sage Weil 已提交
346 347
						const char *name)
{
348
	struct ceph_vxattr *vxattr = ceph_inode_vxattrs(inode);
349 350 351 352 353 354 355 356 357

	if (vxattr) {
		while (vxattr->name) {
			if (!strcmp(vxattr->name, name))
				return vxattr;
			vxattr++;
		}
	}

S
Sage Weil 已提交
358 359 360 361 362 363
	return NULL;
}

static int __set_xattr(struct ceph_inode_info *ci,
			   const char *name, int name_len,
			   const char *val, int val_len,
364
			   int flags, int update_xattr,
S
Sage Weil 已提交
365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392
			   struct ceph_inode_xattr **newxattr)
{
	struct rb_node **p;
	struct rb_node *parent = NULL;
	struct ceph_inode_xattr *xattr = NULL;
	int c;
	int new = 0;

	p = &ci->i_xattrs.index.rb_node;
	while (*p) {
		parent = *p;
		xattr = rb_entry(parent, struct ceph_inode_xattr, node);
		c = strncmp(name, xattr->name, min(name_len, xattr->name_len));
		if (c < 0)
			p = &(*p)->rb_left;
		else if (c > 0)
			p = &(*p)->rb_right;
		else {
			if (name_len == xattr->name_len)
				break;
			else if (name_len < xattr->name_len)
				p = &(*p)->rb_left;
			else
				p = &(*p)->rb_right;
		}
		xattr = NULL;
	}

393 394
	if (update_xattr) {
		int err = 0;
395

396 397 398 399 400 401 402
		if (xattr && (flags & XATTR_CREATE))
			err = -EEXIST;
		else if (!xattr && (flags & XATTR_REPLACE))
			err = -ENODATA;
		if (err) {
			kfree(name);
			kfree(val);
403
			kfree(*newxattr);
404 405
			return err;
		}
406 407 408 409
		if (update_xattr < 0) {
			if (xattr)
				__remove_xattr(ci, xattr);
			kfree(name);
410
			kfree(*newxattr);
411 412
			return 0;
		}
413 414
	}

S
Sage Weil 已提交
415 416 417 418 419
	if (!xattr) {
		new = 1;
		xattr = *newxattr;
		xattr->name = name;
		xattr->name_len = name_len;
420
		xattr->should_free_name = update_xattr;
S
Sage Weil 已提交
421 422 423 424 425 426 427 428 429

		ci->i_xattrs.count++;
		dout("__set_xattr count=%d\n", ci->i_xattrs.count);
	} else {
		kfree(*newxattr);
		*newxattr = NULL;
		if (xattr->should_free_val)
			kfree((void *)xattr->val);

430
		if (update_xattr) {
S
Sage Weil 已提交
431 432 433 434 435 436 437 438 439 440 441 442 443 444
			kfree((void *)name);
			name = xattr->name;
		}
		ci->i_xattrs.names_size -= xattr->name_len;
		ci->i_xattrs.vals_size -= xattr->val_len;
	}
	ci->i_xattrs.names_size += name_len;
	ci->i_xattrs.vals_size += val_len;
	if (val)
		xattr->val = val;
	else
		xattr->val = "";

	xattr->val_len = val_len;
445 446
	xattr->dirty = update_xattr;
	xattr->should_free_val = (val && update_xattr);
S
Sage Weil 已提交
447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465

	if (new) {
		rb_link_node(&xattr->node, parent, p);
		rb_insert_color(&xattr->node, &ci->i_xattrs.index);
		dout("__set_xattr_val p=%p\n", p);
	}

	dout("__set_xattr_val added %llx.%llx xattr %p %s=%.*s\n",
	     ceph_vinop(&ci->vfs_inode), xattr, name, val_len, val);

	return 0;
}

static struct ceph_inode_xattr *__get_xattr(struct ceph_inode_info *ci,
			   const char *name)
{
	struct rb_node **p;
	struct rb_node *parent = NULL;
	struct ceph_inode_xattr *xattr = NULL;
S
Sage Weil 已提交
466
	int name_len = strlen(name);
S
Sage Weil 已提交
467 468 469 470 471 472 473
	int c;

	p = &ci->i_xattrs.index.rb_node;
	while (*p) {
		parent = *p;
		xattr = rb_entry(parent, struct ceph_inode_xattr, node);
		c = strncmp(name, xattr->name, xattr->name_len);
S
Sage Weil 已提交
474 475
		if (c == 0 && name_len > xattr->name_len)
			c = 1;
S
Sage Weil 已提交
476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507
		if (c < 0)
			p = &(*p)->rb_left;
		else if (c > 0)
			p = &(*p)->rb_right;
		else {
			dout("__get_xattr %s: found %.*s\n", name,
			     xattr->val_len, xattr->val);
			return xattr;
		}
	}

	dout("__get_xattr %s: not found\n", name);

	return NULL;
}

static void __free_xattr(struct ceph_inode_xattr *xattr)
{
	BUG_ON(!xattr);

	if (xattr->should_free_name)
		kfree((void *)xattr->name);
	if (xattr->should_free_val)
		kfree((void *)xattr->val);

	kfree(xattr);
}

static int __remove_xattr(struct ceph_inode_info *ci,
			  struct ceph_inode_xattr *xattr)
{
	if (!xattr)
Y
Yan, Zheng 已提交
508
		return -ENODATA;
S
Sage Weil 已提交
509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576

	rb_erase(&xattr->node, &ci->i_xattrs.index);

	if (xattr->should_free_name)
		kfree((void *)xattr->name);
	if (xattr->should_free_val)
		kfree((void *)xattr->val);

	ci->i_xattrs.names_size -= xattr->name_len;
	ci->i_xattrs.vals_size -= xattr->val_len;
	ci->i_xattrs.count--;
	kfree(xattr);

	return 0;
}

static char *__copy_xattr_names(struct ceph_inode_info *ci,
				char *dest)
{
	struct rb_node *p;
	struct ceph_inode_xattr *xattr = NULL;

	p = rb_first(&ci->i_xattrs.index);
	dout("__copy_xattr_names count=%d\n", ci->i_xattrs.count);

	while (p) {
		xattr = rb_entry(p, struct ceph_inode_xattr, node);
		memcpy(dest, xattr->name, xattr->name_len);
		dest[xattr->name_len] = '\0';

		dout("dest=%s %p (%s) (%d/%d)\n", dest, xattr, xattr->name,
		     xattr->name_len, ci->i_xattrs.names_size);

		dest += xattr->name_len + 1;
		p = rb_next(p);
	}

	return dest;
}

void __ceph_destroy_xattrs(struct ceph_inode_info *ci)
{
	struct rb_node *p, *tmp;
	struct ceph_inode_xattr *xattr = NULL;

	p = rb_first(&ci->i_xattrs.index);

	dout("__ceph_destroy_xattrs p=%p\n", p);

	while (p) {
		xattr = rb_entry(p, struct ceph_inode_xattr, node);
		tmp = p;
		p = rb_next(tmp);
		dout("__ceph_destroy_xattrs next p=%p (%.*s)\n", p,
		     xattr->name_len, xattr->name);
		rb_erase(tmp, &ci->i_xattrs.index);

		__free_xattr(xattr);
	}

	ci->i_xattrs.names_size = 0;
	ci->i_xattrs.vals_size = 0;
	ci->i_xattrs.index_version = 0;
	ci->i_xattrs.count = 0;
	ci->i_xattrs.index = RB_ROOT;
}

static int __build_xattrs(struct inode *inode)
577 578
	__releases(ci->i_ceph_lock)
	__acquires(ci->i_ceph_lock)
S
Sage Weil 已提交
579 580 581 582 583 584 585 586 587
{
	u32 namelen;
	u32 numattr = 0;
	void *p, *end;
	u32 len;
	const char *name, *val;
	struct ceph_inode_info *ci = ceph_inode(inode);
	int xattr_version;
	struct ceph_inode_xattr **xattrs = NULL;
S
Sage Weil 已提交
588
	int err = 0;
S
Sage Weil 已提交
589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605
	int i;

	dout("__build_xattrs() len=%d\n",
	     ci->i_xattrs.blob ? (int)ci->i_xattrs.blob->vec.iov_len : 0);

	if (ci->i_xattrs.index_version >= ci->i_xattrs.version)
		return 0; /* already built */

	__ceph_destroy_xattrs(ci);

start:
	/* updated internal xattr rb tree */
	if (ci->i_xattrs.blob && ci->i_xattrs.blob->vec.iov_len > 4) {
		p = ci->i_xattrs.blob->vec.iov_base;
		end = p + ci->i_xattrs.blob->vec.iov_len;
		ceph_decode_32_safe(&p, end, numattr, bad);
		xattr_version = ci->i_xattrs.version;
606
		spin_unlock(&ci->i_ceph_lock);
S
Sage Weil 已提交
607

608
		xattrs = kcalloc(numattr, sizeof(struct ceph_inode_xattr *),
S
Sage Weil 已提交
609 610 611 612
				 GFP_NOFS);
		err = -ENOMEM;
		if (!xattrs)
			goto bad_lock;
I
Ilya Dryomov 已提交
613

S
Sage Weil 已提交
614 615 616 617 618 619 620
		for (i = 0; i < numattr; i++) {
			xattrs[i] = kmalloc(sizeof(struct ceph_inode_xattr),
					    GFP_NOFS);
			if (!xattrs[i])
				goto bad_lock;
		}

621
		spin_lock(&ci->i_ceph_lock);
S
Sage Weil 已提交
622 623 624 625 626
		if (ci->i_xattrs.version != xattr_version) {
			/* lost a race, retry */
			for (i = 0; i < numattr; i++)
				kfree(xattrs[i]);
			kfree(xattrs);
A
Alan Cox 已提交
627
			xattrs = NULL;
S
Sage Weil 已提交
628 629 630 631 632 633 634 635 636 637 638 639 640
			goto start;
		}
		err = -EIO;
		while (numattr--) {
			ceph_decode_32_safe(&p, end, len, bad);
			namelen = len;
			name = p;
			p += len;
			ceph_decode_32_safe(&p, end, len, bad);
			val = p;
			p += len;

			err = __set_xattr(ci, name, namelen, val, len,
641
					  0, 0, &xattrs[numattr]);
S
Sage Weil 已提交
642 643 644 645 646 647 648 649 650 651 652

			if (err < 0)
				goto bad;
		}
		kfree(xattrs);
	}
	ci->i_xattrs.index_version = ci->i_xattrs.version;
	ci->i_xattrs.dirty = false;

	return err;
bad_lock:
653
	spin_lock(&ci->i_ceph_lock);
S
Sage Weil 已提交
654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720
bad:
	if (xattrs) {
		for (i = 0; i < numattr; i++)
			kfree(xattrs[i]);
		kfree(xattrs);
	}
	ci->i_xattrs.names_size = 0;
	return err;
}

static int __get_required_blob_size(struct ceph_inode_info *ci, int name_size,
				    int val_size)
{
	/*
	 * 4 bytes for the length, and additional 4 bytes per each xattr name,
	 * 4 bytes per each value
	 */
	int size = 4 + ci->i_xattrs.count*(4 + 4) +
			     ci->i_xattrs.names_size +
			     ci->i_xattrs.vals_size;
	dout("__get_required_blob_size c=%d names.size=%d vals.size=%d\n",
	     ci->i_xattrs.count, ci->i_xattrs.names_size,
	     ci->i_xattrs.vals_size);

	if (name_size)
		size += 4 + 4 + name_size + val_size;

	return size;
}

/*
 * If there are dirty xattrs, reencode xattrs into the prealloc_blob
 * and swap into place.
 */
void __ceph_build_xattrs_blob(struct ceph_inode_info *ci)
{
	struct rb_node *p;
	struct ceph_inode_xattr *xattr = NULL;
	void *dest;

	dout("__build_xattrs_blob %p\n", &ci->vfs_inode);
	if (ci->i_xattrs.dirty) {
		int need = __get_required_blob_size(ci, 0, 0);

		BUG_ON(need > ci->i_xattrs.prealloc_blob->alloc_len);

		p = rb_first(&ci->i_xattrs.index);
		dest = ci->i_xattrs.prealloc_blob->vec.iov_base;

		ceph_encode_32(&dest, ci->i_xattrs.count);
		while (p) {
			xattr = rb_entry(p, struct ceph_inode_xattr, node);

			ceph_encode_32(&dest, xattr->name_len);
			memcpy(dest, xattr->name, xattr->name_len);
			dest += xattr->name_len;
			ceph_encode_32(&dest, xattr->val_len);
			memcpy(dest, xattr->val, xattr->val_len);
			dest += xattr->val_len;

			p = rb_next(p);
		}

		/* adjust buffer len; it may be larger than we need */
		ci->i_xattrs.prealloc_blob->vec.iov_len =
			dest - ci->i_xattrs.prealloc_blob->vec.iov_base;

S
Sage Weil 已提交
721 722
		if (ci->i_xattrs.blob)
			ceph_buffer_put(ci->i_xattrs.blob);
S
Sage Weil 已提交
723 724 725
		ci->i_xattrs.blob = ci->i_xattrs.prealloc_blob;
		ci->i_xattrs.prealloc_blob = NULL;
		ci->i_xattrs.dirty = false;
726
		ci->i_xattrs.version++;
S
Sage Weil 已提交
727 728 729
	}
}

Y
Yan, Zheng 已提交
730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746
static inline int __get_request_mask(struct inode *in) {
	struct ceph_mds_request *req = current->journal_info;
	int mask = 0;
	if (req && req->r_target_inode == in) {
		if (req->r_op == CEPH_MDS_OP_LOOKUP ||
		    req->r_op == CEPH_MDS_OP_LOOKUPINO ||
		    req->r_op == CEPH_MDS_OP_LOOKUPPARENT ||
		    req->r_op == CEPH_MDS_OP_GETATTR) {
			mask = le32_to_cpu(req->r_args.getattr.mask);
		} else if (req->r_op == CEPH_MDS_OP_OPEN ||
			   req->r_op == CEPH_MDS_OP_CREATE) {
			mask = le32_to_cpu(req->r_args.open.mask);
		}
	}
	return mask;
}

G
Guangliang Zhao 已提交
747
ssize_t __ceph_getxattr(struct inode *inode, const char *name, void *value,
S
Sage Weil 已提交
748 749 750 751
		      size_t size)
{
	struct ceph_inode_info *ci = ceph_inode(inode);
	struct ceph_inode_xattr *xattr;
752
	struct ceph_vxattr *vxattr = NULL;
Y
Yan, Zheng 已提交
753 754
	int req_mask;
	int err;
S
Sage Weil 已提交
755

S
Sage Weil 已提交
756 757
	/* let's see if a virtual xattr was requested */
	vxattr = ceph_match_vxattr(inode, name);
758
	if (vxattr) {
759 760 761
		err = ceph_do_getattr(inode, 0, true);
		if (err)
			return err;
762 763 764
		err = -ENODATA;
		if (!(vxattr->exists_cb && !vxattr->exists_cb(ci)))
			err = vxattr->getxattr_cb(ci, value, size);
765
		return err;
S
Sage Weil 已提交
766 767
	}

Y
Yan, Zheng 已提交
768 769
	req_mask = __get_request_mask(inode);

770 771 772 773
	spin_lock(&ci->i_ceph_lock);
	dout("getxattr %p ver=%lld index_ver=%lld\n", inode,
	     ci->i_xattrs.version, ci->i_xattrs.index_version);

774
	if (ci->i_xattrs.version == 0 ||
Y
Yan, Zheng 已提交
775 776
	    !((req_mask & CEPH_CAP_XATTR_SHARED) ||
	      __ceph_caps_issued_mask(ci, CEPH_CAP_XATTR_SHARED, 1))) {
777
		spin_unlock(&ci->i_ceph_lock);
Y
Yan, Zheng 已提交
778 779 780 781 782 783 784 785

		/* security module gets xattr while filling trace */
		if (current->journal_info != NULL) {
			pr_warn_ratelimited("sync getxattr %p "
					    "during filling trace\n", inode);
			return -EBUSY;
		}

S
Sage Weil 已提交
786
		/* get xattrs from mds (if we don't already have them) */
787
		err = ceph_do_getattr(inode, CEPH_STAT_CAP_XATTR, true);
S
Sage Weil 已提交
788 789
		if (err)
			return err;
790
		spin_lock(&ci->i_ceph_lock);
S
Sage Weil 已提交
791 792 793 794 795 796 797 798
	}

	err = __build_xattrs(inode);
	if (err < 0)
		goto out;

	err = -ENODATA;  /* == ENOATTR */
	xattr = __get_xattr(ci, name);
S
Sage Weil 已提交
799
	if (!xattr)
S
Sage Weil 已提交
800 801 802 803 804 805 806 807 808 809 810 811
		goto out;

	err = -ERANGE;
	if (size && size < xattr->val_len)
		goto out;

	err = xattr->val_len;
	if (size == 0)
		goto out;

	memcpy(value, xattr->val, xattr->val_len);

Y
Yan, Zheng 已提交
812 813 814
	if (current->journal_info != NULL &&
	    !strncmp(name, XATTR_SECURITY_PREFIX, XATTR_SECURITY_PREFIX_LEN))
		ci->i_ceph_flags |= CEPH_I_SEC_INITED;
S
Sage Weil 已提交
815
out:
816
	spin_unlock(&ci->i_ceph_lock);
S
Sage Weil 已提交
817 818 819 820 821
	return err;
}

ssize_t ceph_listxattr(struct dentry *dentry, char *names, size_t size)
{
822
	struct inode *inode = d_inode(dentry);
S
Sage Weil 已提交
823
	struct ceph_inode_info *ci = ceph_inode(inode);
824
	struct ceph_vxattr *vxattrs = ceph_inode_vxattrs(inode);
S
Sage Weil 已提交
825 826 827 828 829 830
	u32 vir_namelen = 0;
	u32 namelen;
	int err;
	u32 len;
	int i;

831
	spin_lock(&ci->i_ceph_lock);
S
Sage Weil 已提交
832 833 834
	dout("listxattr %p ver=%lld index_ver=%lld\n", inode,
	     ci->i_xattrs.version, ci->i_xattrs.index_version);

835 836
	if (ci->i_xattrs.version == 0 ||
	    !__ceph_caps_issued_mask(ci, CEPH_CAP_XATTR_SHARED, 1)) {
837
		spin_unlock(&ci->i_ceph_lock);
838
		err = ceph_do_getattr(inode, CEPH_STAT_CAP_XATTR, true);
S
Sage Weil 已提交
839 840
		if (err)
			return err;
841
		spin_lock(&ci->i_ceph_lock);
S
Sage Weil 已提交
842 843 844 845 846
	}

	err = __build_xattrs(inode);
	if (err < 0)
		goto out;
847 848 849 850 851 852
	/*
	 * Start with virtual dir xattr names (if any) (including
	 * terminating '\0' characters for each).
	 */
	vir_namelen = ceph_vxattrs_name_size(vxattrs);

S
Sage Weil 已提交
853
	/* adding 1 byte per each variable due to the null termination */
854
	namelen = ci->i_xattrs.names_size + ci->i_xattrs.count;
S
Sage Weil 已提交
855
	err = -ERANGE;
856
	if (size && vir_namelen + namelen > size)
S
Sage Weil 已提交
857 858
		goto out;

859
	err = namelen + vir_namelen;
S
Sage Weil 已提交
860 861 862 863 864 865
	if (size == 0)
		goto out;

	names = __copy_xattr_names(ci, names);

	/* virtual xattr names, too */
866 867
	err = namelen;
	if (vxattrs) {
S
Sage Weil 已提交
868
		for (i = 0; vxattrs[i].name; i++) {
869 870 871 872 873 874 875
			if (!vxattrs[i].hidden &&
			    !(vxattrs[i].exists_cb &&
			      !vxattrs[i].exists_cb(ci))) {
				len = sprintf(names, "%s", vxattrs[i].name);
				names += len + 1;
				err += len + 1;
			}
S
Sage Weil 已提交
876
		}
877
	}
S
Sage Weil 已提交
878 879

out:
880
	spin_unlock(&ci->i_ceph_lock);
S
Sage Weil 已提交
881 882 883
	return err;
}

884
static int ceph_sync_setxattr(struct inode *inode, const char *name,
S
Sage Weil 已提交
885 886
			      const char *value, size_t size, int flags)
{
887
	struct ceph_fs_client *fsc = ceph_sb_to_client(inode->i_sb);
S
Sage Weil 已提交
888 889
	struct ceph_inode_info *ci = ceph_inode(inode);
	struct ceph_mds_request *req;
890
	struct ceph_mds_client *mdsc = fsc->mdsc;
891
	struct ceph_pagelist *pagelist = NULL;
892
	int op = CEPH_MDS_OP_SETXATTR;
S
Sage Weil 已提交
893
	int err;
894

895
	if (size > 0) {
896 897 898
		/* copy value into pagelist */
		pagelist = kmalloc(sizeof(*pagelist), GFP_NOFS);
		if (!pagelist)
S
Sage Weil 已提交
899
			return -ENOMEM;
900 901 902 903 904

		ceph_pagelist_init(pagelist);
		err = ceph_pagelist_append(pagelist, value, size);
		if (err)
			goto out;
905
	} else if (!value) {
906 907 908 909
		if (flags & CEPH_XATTR_REPLACE)
			op = CEPH_MDS_OP_RMXATTR;
		else
			flags |= CEPH_XATTR_REMOVE;
S
Sage Weil 已提交
910 911 912 913 914
	}

	dout("setxattr value=%.*s\n", (int)size, value);

	/* do request */
915
	req = ceph_mdsc_create_request(mdsc, op, USE_AUTH_MDS);
J
Julia Lawall 已提交
916 917 918 919
	if (IS_ERR(req)) {
		err = PTR_ERR(req);
		goto out;
	}
920

S
Sage Weil 已提交
921
	req->r_path2 = kstrdup(name, GFP_NOFS);
922 923 924 925 926
	if (!req->r_path2) {
		ceph_mdsc_put_request(req);
		err = -ENOMEM;
		goto out;
	}
S
Sage Weil 已提交
927

928 929 930 931 932
	if (op == CEPH_MDS_OP_SETXATTR) {
		req->r_args.setxattr.flags = cpu_to_le32(flags);
		req->r_pagelist = pagelist;
		pagelist = NULL;
	}
S
Sage Weil 已提交
933

934 935 936 937 938
	req->r_inode = inode;
	ihold(inode);
	req->r_num_caps = 1;
	req->r_inode_drop = CEPH_CAP_XATTR_SHARED;

S
Sage Weil 已提交
939
	dout("xattr.ver (before): %lld\n", ci->i_xattrs.version);
940
	err = ceph_mdsc_do_request(mdsc, NULL, req);
S
Sage Weil 已提交
941 942 943 944
	ceph_mdsc_put_request(req);
	dout("xattr.ver (after): %lld\n", ci->i_xattrs.version);

out:
945 946
	if (pagelist)
		ceph_pagelist_release(pagelist);
S
Sage Weil 已提交
947 948 949
	return err;
}

950
int __ceph_setxattr(struct inode *inode, const char *name,
G
Guangliang Zhao 已提交
951
			const void *value, size_t size, int flags)
S
Sage Weil 已提交
952
{
953
	struct ceph_vxattr *vxattr;
S
Sage Weil 已提交
954
	struct ceph_inode_info *ci = ceph_inode(inode);
955
	struct ceph_mds_client *mdsc = ceph_sb_to_client(inode->i_sb)->mdsc;
956
	struct ceph_cap_flush *prealloc_cf = NULL;
957
	int issued;
S
Sage Weil 已提交
958
	int err;
959
	int dirty = 0;
S
Sage Weil 已提交
960 961 962 963 964 965
	int name_len = strlen(name);
	int val_len = size;
	char *newname = NULL;
	char *newval = NULL;
	struct ceph_inode_xattr *xattr = NULL;
	int required_blob_size;
966
	bool lock_snap_rwsem = false;
S
Sage Weil 已提交
967

968 969
	if (ceph_snap(inode) != CEPH_NOSNAP)
		return -EROFS;
S
Sage Weil 已提交
970

971 972 973
	vxattr = ceph_match_vxattr(inode, name);
	if (vxattr && vxattr->readonly)
		return -EOPNOTSUPP;
S
Sage Weil 已提交
974

975 976 977 978
	/* pass any unhandled ceph.* xattrs through to the MDS */
	if (!strncmp(name, XATTR_CEPH_PREFIX, XATTR_CEPH_PREFIX_LEN))
		goto do_sync_unlocked;

S
Sage Weil 已提交
979 980
	/* preallocate memory for xattr name, value, index node */
	err = -ENOMEM;
J
Julia Lawall 已提交
981
	newname = kmemdup(name, name_len + 1, GFP_NOFS);
S
Sage Weil 已提交
982 983 984 985
	if (!newname)
		goto out;

	if (val_len) {
986
		newval = kmemdup(value, val_len, GFP_NOFS);
S
Sage Weil 已提交
987 988 989 990 991 992 993 994
		if (!newval)
			goto out;
	}

	xattr = kmalloc(sizeof(struct ceph_inode_xattr), GFP_NOFS);
	if (!xattr)
		goto out;

995 996 997 998
	prealloc_cf = ceph_alloc_cap_flush();
	if (!prealloc_cf)
		goto out;

999
	spin_lock(&ci->i_ceph_lock);
S
Sage Weil 已提交
1000 1001
retry:
	issued = __ceph_caps_issued(ci, NULL);
1002
	if (ci->i_xattrs.version == 0 || !(issued & CEPH_CAP_XATTR_EXCL))
S
Sage Weil 已提交
1003
		goto do_sync;
1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015

	if (!lock_snap_rwsem && !ci->i_head_snapc) {
		lock_snap_rwsem = true;
		if (!down_read_trylock(&mdsc->snap_rwsem)) {
			spin_unlock(&ci->i_ceph_lock);
			down_read(&mdsc->snap_rwsem);
			spin_lock(&ci->i_ceph_lock);
			goto retry;
		}
	}

	dout("setxattr %p issued %s\n", inode, ceph_cap_string(issued));
S
Sage Weil 已提交
1016 1017 1018 1019 1020 1021
	__build_xattrs(inode);

	required_blob_size = __get_required_blob_size(ci, name_len, val_len);

	if (!ci->i_xattrs.prealloc_blob ||
	    required_blob_size > ci->i_xattrs.prealloc_blob->alloc_len) {
1022
		struct ceph_buffer *blob;
S
Sage Weil 已提交
1023

1024
		spin_unlock(&ci->i_ceph_lock);
S
Sage Weil 已提交
1025
		dout(" preaallocating new blob size=%d\n", required_blob_size);
S
Sage Weil 已提交
1026
		blob = ceph_buffer_new(required_blob_size, GFP_NOFS);
S
Sage Weil 已提交
1027
		if (!blob)
1028
			goto do_sync_unlocked;
1029
		spin_lock(&ci->i_ceph_lock);
S
Sage Weil 已提交
1030 1031
		if (ci->i_xattrs.prealloc_blob)
			ceph_buffer_put(ci->i_xattrs.prealloc_blob);
S
Sage Weil 已提交
1032 1033 1034 1035
		ci->i_xattrs.prealloc_blob = blob;
		goto retry;
	}

1036 1037
	err = __set_xattr(ci, newname, name_len, newval, val_len,
			  flags, value ? 1 : -1, &xattr);
1038

1039
	if (!err) {
1040 1041
		dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_XATTR_EXCL,
					       &prealloc_cf);
1042
		ci->i_xattrs.dirty = true;
1043
		inode->i_ctime = current_time(inode);
1044
	}
1045

1046
	spin_unlock(&ci->i_ceph_lock);
1047 1048
	if (lock_snap_rwsem)
		up_read(&mdsc->snap_rwsem);
1049 1050
	if (dirty)
		__mark_inode_dirty(inode, dirty);
1051
	ceph_free_cap_flush(prealloc_cf);
S
Sage Weil 已提交
1052 1053 1054
	return err;

do_sync:
1055
	spin_unlock(&ci->i_ceph_lock);
1056
do_sync_unlocked:
1057 1058
	if (lock_snap_rwsem)
		up_read(&mdsc->snap_rwsem);
Y
Yan, Zheng 已提交
1059 1060 1061 1062 1063 1064 1065

	/* security module set xattr while filling trace */
	if (current->journal_info != NULL) {
		pr_warn_ratelimited("sync setxattr %p "
				    "during filling trace\n", inode);
		err = -EBUSY;
	} else {
1066
		err = ceph_sync_setxattr(inode, name, value, size, flags);
Y
Yan, Zheng 已提交
1067
	}
S
Sage Weil 已提交
1068
out:
1069
	ceph_free_cap_flush(prealloc_cf);
S
Sage Weil 已提交
1070 1071 1072 1073 1074 1075
	kfree(newname);
	kfree(newval);
	kfree(xattr);
	return err;
}

1076 1077 1078
static int ceph_get_xattr_handler(const struct xattr_handler *handler,
				  struct dentry *dentry, struct inode *inode,
				  const char *name, void *value, size_t size)
G
Guangliang Zhao 已提交
1079
{
1080 1081 1082 1083
	if (!ceph_is_valid_xattr(name))
		return -EOPNOTSUPP;
	return __ceph_getxattr(inode, name, value, size);
}
G
Guangliang Zhao 已提交
1084

1085
static int ceph_set_xattr_handler(const struct xattr_handler *handler,
1086 1087 1088
				  struct dentry *unused, struct inode *inode,
				  const char *name, const void *value,
				  size_t size, int flags)
1089 1090 1091
{
	if (!ceph_is_valid_xattr(name))
		return -EOPNOTSUPP;
1092
	return __ceph_setxattr(inode, name, value, size, flags);
G
Guangliang Zhao 已提交
1093
}
Y
Yan, Zheng 已提交
1094

W
Wei Yongjun 已提交
1095
static const struct xattr_handler ceph_other_xattr_handler = {
1096 1097 1098 1099 1100
	.prefix = "",  /* match any name => handlers called with full name */
	.get = ceph_get_xattr_handler,
	.set = ceph_set_xattr_handler,
};

Y
Yan, Zheng 已提交
1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121
#ifdef CONFIG_SECURITY
bool ceph_security_xattr_wanted(struct inode *in)
{
	return in->i_security != NULL;
}

bool ceph_security_xattr_deadlock(struct inode *in)
{
	struct ceph_inode_info *ci;
	bool ret;
	if (in->i_security == NULL)
		return false;
	ci = ceph_inode(in);
	spin_lock(&ci->i_ceph_lock);
	ret = !(ci->i_ceph_flags & CEPH_I_SEC_INITED) &&
	      !(ci->i_xattrs.version > 0 &&
		__ceph_caps_issued_mask(ci, CEPH_CAP_XATTR_SHARED, 0));
	spin_unlock(&ci->i_ceph_lock);
	return ret;
}
#endif