Linux Audio

Check our new training course

Open-source upstreaming

Need help get the support for your hardware in upstream Linux?
Loading...
v3.15
 
   1#include <linux/ceph/ceph_debug.h>
 
   2
   3#include "super.h"
   4#include "mds_client.h"
   5
   6#include <linux/ceph/decode.h>
   7
   8#include <linux/xattr.h>
   9#include <linux/posix_acl_xattr.h>
  10#include <linux/slab.h>
  11
  12#define XATTR_CEPH_PREFIX "ceph."
  13#define XATTR_CEPH_PREFIX_LEN (sizeof (XATTR_CEPH_PREFIX) - 1)
  14
  15static int __remove_xattr(struct ceph_inode_info *ci,
  16			  struct ceph_inode_xattr *xattr);
  17
 
 
  18/*
  19 * List of handlers for synthetic system.* attributes. Other
  20 * attributes are handled directly.
  21 */
  22const struct xattr_handler *ceph_xattr_handlers[] = {
  23#ifdef CONFIG_CEPH_FS_POSIX_ACL
  24	&posix_acl_access_xattr_handler,
  25	&posix_acl_default_xattr_handler,
  26#endif
 
  27	NULL,
  28};
  29
  30static bool ceph_is_valid_xattr(const char *name)
  31{
  32	return !strncmp(name, XATTR_CEPH_PREFIX, XATTR_CEPH_PREFIX_LEN) ||
  33	       !strncmp(name, XATTR_SECURITY_PREFIX,
  34			XATTR_SECURITY_PREFIX_LEN) ||
  35	       !strncmp(name, XATTR_SYSTEM_PREFIX, XATTR_SYSTEM_PREFIX_LEN) ||
  36	       !strncmp(name, XATTR_TRUSTED_PREFIX, XATTR_TRUSTED_PREFIX_LEN) ||
  37	       !strncmp(name, XATTR_USER_PREFIX, XATTR_USER_PREFIX_LEN);
  38}
  39
  40/*
  41 * These define virtual xattrs exposing the recursive directory
  42 * statistics and layout metadata.
  43 */
  44struct ceph_vxattr {
  45	char *name;
  46	size_t name_size;	/* strlen(name) + 1 (for '\0') */
  47	size_t (*getxattr_cb)(struct ceph_inode_info *ci, char *val,
  48			      size_t size);
  49	bool readonly, hidden;
  50	bool (*exists_cb)(struct ceph_inode_info *ci);
  51};
  52
  53/* layouts */
  54
  55static bool ceph_vxattrcb_layout_exists(struct ceph_inode_info *ci)
  56{
  57	size_t s;
  58	char *p = (char *)&ci->i_layout;
  59
  60	for (s = 0; s < sizeof(ci->i_layout); s++, p++)
  61		if (*p)
  62			return true;
  63	return false;
  64}
  65
  66static size_t ceph_vxattrcb_layout(struct ceph_inode_info *ci, char *val,
  67				   size_t size)
  68{
  69	int ret;
  70	struct ceph_fs_client *fsc = ceph_sb_to_client(ci->vfs_inode.i_sb);
  71	struct ceph_osd_client *osdc = &fsc->client->osdc;
  72	s64 pool = ceph_file_layout_pg_pool(ci->i_layout);
 
  73	const char *pool_name;
 
  74	char buf[128];
 
 
 
 
  75
  76	dout("ceph_vxattrcb_layout %p\n", &ci->vfs_inode);
  77	down_read(&osdc->map_sem);
  78	pool_name = ceph_pg_pool_name_by_id(osdc->osdmap, pool);
  79	if (pool_name) {
  80		size_t len = strlen(pool_name);
  81		ret = snprintf(buf, sizeof(buf),
  82		"stripe_unit=%lld stripe_count=%lld object_size=%lld pool=",
  83		(unsigned long long)ceph_file_layout_su(ci->i_layout),
  84		(unsigned long long)ceph_file_layout_stripe_count(ci->i_layout),
  85	        (unsigned long long)ceph_file_layout_object_size(ci->i_layout));
  86		if (!size) {
  87			ret += len;
  88		} else if (ret + len > size) {
  89			ret = -ERANGE;
  90		} else {
  91			memcpy(val, buf, ret);
 
 
 
 
 
 
 
 
 
 
 
 
 
  92			memcpy(val + ret, pool_name, len);
  93			ret += len;
  94		}
  95	} else {
  96		ret = snprintf(buf, sizeof(buf),
  97		"stripe_unit=%lld stripe_count=%lld object_size=%lld pool=%lld",
  98		(unsigned long long)ceph_file_layout_su(ci->i_layout),
  99		(unsigned long long)ceph_file_layout_stripe_count(ci->i_layout),
 100	        (unsigned long long)ceph_file_layout_object_size(ci->i_layout),
 101		(unsigned long long)pool);
 102		if (size) {
 103			if (ret <= size)
 104				memcpy(val, buf, ret);
 105			else
 106				ret = -ERANGE;
 107		}
 108	}
 109	up_read(&osdc->map_sem);
 
 110	return ret;
 111}
 112
 113static size_t ceph_vxattrcb_layout_stripe_unit(struct ceph_inode_info *ci,
 114					       char *val, size_t size)
 115{
 116	return snprintf(val, size, "%lld",
 117			(unsigned long long)ceph_file_layout_su(ci->i_layout));
 118}
 119
 120static size_t ceph_vxattrcb_layout_stripe_count(struct ceph_inode_info *ci,
 121						char *val, size_t size)
 122{
 123	return snprintf(val, size, "%lld",
 124	       (unsigned long long)ceph_file_layout_stripe_count(ci->i_layout));
 125}
 126
 127static size_t ceph_vxattrcb_layout_object_size(struct ceph_inode_info *ci,
 128					       char *val, size_t size)
 129{
 130	return snprintf(val, size, "%lld",
 131	       (unsigned long long)ceph_file_layout_object_size(ci->i_layout));
 132}
 133
 134static size_t ceph_vxattrcb_layout_pool(struct ceph_inode_info *ci,
 135					char *val, size_t size)
 136{
 137	int ret;
 138	struct ceph_fs_client *fsc = ceph_sb_to_client(ci->vfs_inode.i_sb);
 139	struct ceph_osd_client *osdc = &fsc->client->osdc;
 140	s64 pool = ceph_file_layout_pg_pool(ci->i_layout);
 141	const char *pool_name;
 142
 143	down_read(&osdc->map_sem);
 144	pool_name = ceph_pg_pool_name_by_id(osdc->osdmap, pool);
 145	if (pool_name)
 146		ret = snprintf(val, size, "%s", pool_name);
 147	else
 148		ret = snprintf(val, size, "%lld", (unsigned long long)pool);
 149	up_read(&osdc->map_sem);
 
 
 
 
 
 
 
 
 
 
 
 
 150	return ret;
 151}
 152
 153/* directories */
 154
 155static size_t ceph_vxattrcb_dir_entries(struct ceph_inode_info *ci, char *val,
 156					size_t size)
 157{
 158	return snprintf(val, size, "%lld", ci->i_files + ci->i_subdirs);
 159}
 160
 161static size_t ceph_vxattrcb_dir_files(struct ceph_inode_info *ci, char *val,
 162				      size_t size)
 163{
 164	return snprintf(val, size, "%lld", ci->i_files);
 165}
 166
 167static size_t ceph_vxattrcb_dir_subdirs(struct ceph_inode_info *ci, char *val,
 168					size_t size)
 169{
 170	return snprintf(val, size, "%lld", ci->i_subdirs);
 171}
 172
 173static size_t ceph_vxattrcb_dir_rentries(struct ceph_inode_info *ci, char *val,
 174					 size_t size)
 175{
 176	return snprintf(val, size, "%lld", ci->i_rfiles + ci->i_rsubdirs);
 177}
 178
 179static size_t ceph_vxattrcb_dir_rfiles(struct ceph_inode_info *ci, char *val,
 180				       size_t size)
 181{
 182	return snprintf(val, size, "%lld", ci->i_rfiles);
 183}
 184
 185static size_t ceph_vxattrcb_dir_rsubdirs(struct ceph_inode_info *ci, char *val,
 186					 size_t size)
 187{
 188	return snprintf(val, size, "%lld", ci->i_rsubdirs);
 189}
 190
 191static size_t ceph_vxattrcb_dir_rbytes(struct ceph_inode_info *ci, char *val,
 192				       size_t size)
 193{
 194	return snprintf(val, size, "%lld", ci->i_rbytes);
 195}
 196
 197static size_t ceph_vxattrcb_dir_rctime(struct ceph_inode_info *ci, char *val,
 198				       size_t size)
 199{
 200	return snprintf(val, size, "%ld.09%ld", (long)ci->i_rctime.tv_sec,
 201			(long)ci->i_rctime.tv_nsec);
 202}
 203
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 204
 205#define CEPH_XATTR_NAME(_type, _name)	XATTR_CEPH_PREFIX #_type "." #_name
 206#define CEPH_XATTR_NAME2(_type, _name, _name2)	\
 207	XATTR_CEPH_PREFIX #_type "." #_name "." #_name2
 208
 209#define XATTR_NAME_CEPH(_type, _name)					\
 210	{								\
 211		.name = CEPH_XATTR_NAME(_type, _name),			\
 212		.name_size = sizeof (CEPH_XATTR_NAME(_type, _name)), \
 213		.getxattr_cb = ceph_vxattrcb_ ## _type ## _ ## _name, \
 214		.readonly = true,				\
 215		.hidden = false,				\
 216		.exists_cb = NULL,			\
 217	}
 218#define XATTR_LAYOUT_FIELD(_type, _name, _field)			\
 219	{								\
 220		.name = CEPH_XATTR_NAME2(_type, _name, _field),	\
 221		.name_size = sizeof (CEPH_XATTR_NAME2(_type, _name, _field)), \
 222		.getxattr_cb = ceph_vxattrcb_ ## _name ## _ ## _field, \
 223		.readonly = false,				\
 224		.hidden = true,			\
 225		.exists_cb = ceph_vxattrcb_layout_exists,	\
 226	}
 
 
 
 
 
 
 
 
 
 227
 228static struct ceph_vxattr ceph_dir_vxattrs[] = {
 229	{
 230		.name = "ceph.dir.layout",
 231		.name_size = sizeof("ceph.dir.layout"),
 232		.getxattr_cb = ceph_vxattrcb_layout,
 233		.readonly = false,
 234		.hidden = true,
 235		.exists_cb = ceph_vxattrcb_layout_exists,
 236	},
 237	XATTR_LAYOUT_FIELD(dir, layout, stripe_unit),
 238	XATTR_LAYOUT_FIELD(dir, layout, stripe_count),
 239	XATTR_LAYOUT_FIELD(dir, layout, object_size),
 240	XATTR_LAYOUT_FIELD(dir, layout, pool),
 
 241	XATTR_NAME_CEPH(dir, entries),
 242	XATTR_NAME_CEPH(dir, files),
 243	XATTR_NAME_CEPH(dir, subdirs),
 244	XATTR_NAME_CEPH(dir, rentries),
 245	XATTR_NAME_CEPH(dir, rfiles),
 246	XATTR_NAME_CEPH(dir, rsubdirs),
 247	XATTR_NAME_CEPH(dir, rbytes),
 248	XATTR_NAME_CEPH(dir, rctime),
 
 
 
 
 
 
 
 
 
 
 249	{ .name = NULL, 0 }	/* Required table terminator */
 250};
 251static size_t ceph_dir_vxattrs_name_size;	/* total size of all names */
 252
 253/* files */
 254
 255static struct ceph_vxattr ceph_file_vxattrs[] = {
 256	{
 257		.name = "ceph.file.layout",
 258		.name_size = sizeof("ceph.file.layout"),
 259		.getxattr_cb = ceph_vxattrcb_layout,
 260		.readonly = false,
 261		.hidden = true,
 262		.exists_cb = ceph_vxattrcb_layout_exists,
 263	},
 264	XATTR_LAYOUT_FIELD(file, layout, stripe_unit),
 265	XATTR_LAYOUT_FIELD(file, layout, stripe_count),
 266	XATTR_LAYOUT_FIELD(file, layout, object_size),
 267	XATTR_LAYOUT_FIELD(file, layout, pool),
 
 268	{ .name = NULL, 0 }	/* Required table terminator */
 269};
 270static size_t ceph_file_vxattrs_name_size;	/* total size of all names */
 271
 272static struct ceph_vxattr *ceph_inode_vxattrs(struct inode *inode)
 273{
 274	if (S_ISDIR(inode->i_mode))
 275		return ceph_dir_vxattrs;
 276	else if (S_ISREG(inode->i_mode))
 277		return ceph_file_vxattrs;
 278	return NULL;
 279}
 280
 281static size_t ceph_vxattrs_name_size(struct ceph_vxattr *vxattrs)
 282{
 283	if (vxattrs == ceph_dir_vxattrs)
 284		return ceph_dir_vxattrs_name_size;
 285	if (vxattrs == ceph_file_vxattrs)
 286		return ceph_file_vxattrs_name_size;
 287	BUG();
 288
 289	return 0;
 290}
 291
 292/*
 293 * Compute the aggregate size (including terminating '\0') of all
 294 * virtual extended attribute names in the given vxattr table.
 295 */
 296static size_t __init vxattrs_name_size(struct ceph_vxattr *vxattrs)
 297{
 298	struct ceph_vxattr *vxattr;
 299	size_t size = 0;
 300
 301	for (vxattr = vxattrs; vxattr->name; vxattr++)
 302		if (!vxattr->hidden)
 303			size += vxattr->name_size;
 304
 305	return size;
 306}
 307
 308/* Routines called at initialization and exit time */
 309
 310void __init ceph_xattr_init(void)
 311{
 312	ceph_dir_vxattrs_name_size = vxattrs_name_size(ceph_dir_vxattrs);
 313	ceph_file_vxattrs_name_size = vxattrs_name_size(ceph_file_vxattrs);
 314}
 315
 316void ceph_xattr_exit(void)
 317{
 318	ceph_dir_vxattrs_name_size = 0;
 319	ceph_file_vxattrs_name_size = 0;
 320}
 321
 322static struct ceph_vxattr *ceph_match_vxattr(struct inode *inode,
 323						const char *name)
 324{
 325	struct ceph_vxattr *vxattr = ceph_inode_vxattrs(inode);
 326
 327	if (vxattr) {
 328		while (vxattr->name) {
 329			if (!strcmp(vxattr->name, name))
 330				return vxattr;
 331			vxattr++;
 332		}
 333	}
 334
 335	return NULL;
 336}
 337
 338static int __set_xattr(struct ceph_inode_info *ci,
 339			   const char *name, int name_len,
 340			   const char *val, int val_len,
 341			   int flags, int update_xattr,
 342			   struct ceph_inode_xattr **newxattr)
 343{
 344	struct rb_node **p;
 345	struct rb_node *parent = NULL;
 346	struct ceph_inode_xattr *xattr = NULL;
 347	int c;
 348	int new = 0;
 349
 350	p = &ci->i_xattrs.index.rb_node;
 351	while (*p) {
 352		parent = *p;
 353		xattr = rb_entry(parent, struct ceph_inode_xattr, node);
 354		c = strncmp(name, xattr->name, min(name_len, xattr->name_len));
 355		if (c < 0)
 356			p = &(*p)->rb_left;
 357		else if (c > 0)
 358			p = &(*p)->rb_right;
 359		else {
 360			if (name_len == xattr->name_len)
 361				break;
 362			else if (name_len < xattr->name_len)
 363				p = &(*p)->rb_left;
 364			else
 365				p = &(*p)->rb_right;
 366		}
 367		xattr = NULL;
 368	}
 369
 370	if (update_xattr) {
 371		int err = 0;
 
 372		if (xattr && (flags & XATTR_CREATE))
 373			err = -EEXIST;
 374		else if (!xattr && (flags & XATTR_REPLACE))
 375			err = -ENODATA;
 376		if (err) {
 377			kfree(name);
 378			kfree(val);
 
 379			return err;
 380		}
 381		if (update_xattr < 0) {
 382			if (xattr)
 383				__remove_xattr(ci, xattr);
 384			kfree(name);
 
 385			return 0;
 386		}
 387	}
 388
 389	if (!xattr) {
 390		new = 1;
 391		xattr = *newxattr;
 392		xattr->name = name;
 393		xattr->name_len = name_len;
 394		xattr->should_free_name = update_xattr;
 395
 396		ci->i_xattrs.count++;
 397		dout("__set_xattr count=%d\n", ci->i_xattrs.count);
 398	} else {
 399		kfree(*newxattr);
 400		*newxattr = NULL;
 401		if (xattr->should_free_val)
 402			kfree((void *)xattr->val);
 403
 404		if (update_xattr) {
 405			kfree((void *)name);
 406			name = xattr->name;
 407		}
 408		ci->i_xattrs.names_size -= xattr->name_len;
 409		ci->i_xattrs.vals_size -= xattr->val_len;
 410	}
 411	ci->i_xattrs.names_size += name_len;
 412	ci->i_xattrs.vals_size += val_len;
 413	if (val)
 414		xattr->val = val;
 415	else
 416		xattr->val = "";
 417
 418	xattr->val_len = val_len;
 419	xattr->dirty = update_xattr;
 420	xattr->should_free_val = (val && update_xattr);
 421
 422	if (new) {
 423		rb_link_node(&xattr->node, parent, p);
 424		rb_insert_color(&xattr->node, &ci->i_xattrs.index);
 425		dout("__set_xattr_val p=%p\n", p);
 426	}
 427
 428	dout("__set_xattr_val added %llx.%llx xattr %p %s=%.*s\n",
 429	     ceph_vinop(&ci->vfs_inode), xattr, name, val_len, val);
 430
 431	return 0;
 432}
 433
 434static struct ceph_inode_xattr *__get_xattr(struct ceph_inode_info *ci,
 435			   const char *name)
 436{
 437	struct rb_node **p;
 438	struct rb_node *parent = NULL;
 439	struct ceph_inode_xattr *xattr = NULL;
 440	int name_len = strlen(name);
 441	int c;
 442
 443	p = &ci->i_xattrs.index.rb_node;
 444	while (*p) {
 445		parent = *p;
 446		xattr = rb_entry(parent, struct ceph_inode_xattr, node);
 447		c = strncmp(name, xattr->name, xattr->name_len);
 448		if (c == 0 && name_len > xattr->name_len)
 449			c = 1;
 450		if (c < 0)
 451			p = &(*p)->rb_left;
 452		else if (c > 0)
 453			p = &(*p)->rb_right;
 454		else {
 455			dout("__get_xattr %s: found %.*s\n", name,
 456			     xattr->val_len, xattr->val);
 457			return xattr;
 458		}
 459	}
 460
 461	dout("__get_xattr %s: not found\n", name);
 462
 463	return NULL;
 464}
 465
 466static void __free_xattr(struct ceph_inode_xattr *xattr)
 467{
 468	BUG_ON(!xattr);
 469
 470	if (xattr->should_free_name)
 471		kfree((void *)xattr->name);
 472	if (xattr->should_free_val)
 473		kfree((void *)xattr->val);
 474
 475	kfree(xattr);
 476}
 477
 478static int __remove_xattr(struct ceph_inode_info *ci,
 479			  struct ceph_inode_xattr *xattr)
 480{
 481	if (!xattr)
 482		return -ENODATA;
 483
 484	rb_erase(&xattr->node, &ci->i_xattrs.index);
 485
 486	if (xattr->should_free_name)
 487		kfree((void *)xattr->name);
 488	if (xattr->should_free_val)
 489		kfree((void *)xattr->val);
 490
 491	ci->i_xattrs.names_size -= xattr->name_len;
 492	ci->i_xattrs.vals_size -= xattr->val_len;
 493	ci->i_xattrs.count--;
 494	kfree(xattr);
 495
 496	return 0;
 497}
 498
 499static int __remove_xattr_by_name(struct ceph_inode_info *ci,
 500			   const char *name)
 501{
 502	struct rb_node **p;
 503	struct ceph_inode_xattr *xattr;
 504	int err;
 505
 506	p = &ci->i_xattrs.index.rb_node;
 507	xattr = __get_xattr(ci, name);
 508	err = __remove_xattr(ci, xattr);
 509	return err;
 510}
 511
 512static char *__copy_xattr_names(struct ceph_inode_info *ci,
 513				char *dest)
 514{
 515	struct rb_node *p;
 516	struct ceph_inode_xattr *xattr = NULL;
 517
 518	p = rb_first(&ci->i_xattrs.index);
 519	dout("__copy_xattr_names count=%d\n", ci->i_xattrs.count);
 520
 521	while (p) {
 522		xattr = rb_entry(p, struct ceph_inode_xattr, node);
 523		memcpy(dest, xattr->name, xattr->name_len);
 524		dest[xattr->name_len] = '\0';
 525
 526		dout("dest=%s %p (%s) (%d/%d)\n", dest, xattr, xattr->name,
 527		     xattr->name_len, ci->i_xattrs.names_size);
 528
 529		dest += xattr->name_len + 1;
 530		p = rb_next(p);
 531	}
 532
 533	return dest;
 534}
 535
 536void __ceph_destroy_xattrs(struct ceph_inode_info *ci)
 537{
 538	struct rb_node *p, *tmp;
 539	struct ceph_inode_xattr *xattr = NULL;
 540
 541	p = rb_first(&ci->i_xattrs.index);
 542
 543	dout("__ceph_destroy_xattrs p=%p\n", p);
 544
 545	while (p) {
 546		xattr = rb_entry(p, struct ceph_inode_xattr, node);
 547		tmp = p;
 548		p = rb_next(tmp);
 549		dout("__ceph_destroy_xattrs next p=%p (%.*s)\n", p,
 550		     xattr->name_len, xattr->name);
 551		rb_erase(tmp, &ci->i_xattrs.index);
 552
 553		__free_xattr(xattr);
 554	}
 555
 556	ci->i_xattrs.names_size = 0;
 557	ci->i_xattrs.vals_size = 0;
 558	ci->i_xattrs.index_version = 0;
 559	ci->i_xattrs.count = 0;
 560	ci->i_xattrs.index = RB_ROOT;
 561}
 562
 563static int __build_xattrs(struct inode *inode)
 564	__releases(ci->i_ceph_lock)
 565	__acquires(ci->i_ceph_lock)
 566{
 567	u32 namelen;
 568	u32 numattr = 0;
 569	void *p, *end;
 570	u32 len;
 571	const char *name, *val;
 572	struct ceph_inode_info *ci = ceph_inode(inode);
 573	int xattr_version;
 574	struct ceph_inode_xattr **xattrs = NULL;
 575	int err = 0;
 576	int i;
 577
 578	dout("__build_xattrs() len=%d\n",
 579	     ci->i_xattrs.blob ? (int)ci->i_xattrs.blob->vec.iov_len : 0);
 580
 581	if (ci->i_xattrs.index_version >= ci->i_xattrs.version)
 582		return 0; /* already built */
 583
 584	__ceph_destroy_xattrs(ci);
 585
 586start:
 587	/* updated internal xattr rb tree */
 588	if (ci->i_xattrs.blob && ci->i_xattrs.blob->vec.iov_len > 4) {
 589		p = ci->i_xattrs.blob->vec.iov_base;
 590		end = p + ci->i_xattrs.blob->vec.iov_len;
 591		ceph_decode_32_safe(&p, end, numattr, bad);
 592		xattr_version = ci->i_xattrs.version;
 593		spin_unlock(&ci->i_ceph_lock);
 594
 595		xattrs = kcalloc(numattr, sizeof(struct ceph_xattr *),
 596				 GFP_NOFS);
 597		err = -ENOMEM;
 598		if (!xattrs)
 599			goto bad_lock;
 600		memset(xattrs, 0, numattr*sizeof(struct ceph_xattr *));
 601		for (i = 0; i < numattr; i++) {
 602			xattrs[i] = kmalloc(sizeof(struct ceph_inode_xattr),
 603					    GFP_NOFS);
 604			if (!xattrs[i])
 605				goto bad_lock;
 606		}
 607
 608		spin_lock(&ci->i_ceph_lock);
 609		if (ci->i_xattrs.version != xattr_version) {
 610			/* lost a race, retry */
 611			for (i = 0; i < numattr; i++)
 612				kfree(xattrs[i]);
 613			kfree(xattrs);
 614			xattrs = NULL;
 615			goto start;
 616		}
 617		err = -EIO;
 618		while (numattr--) {
 619			ceph_decode_32_safe(&p, end, len, bad);
 620			namelen = len;
 621			name = p;
 622			p += len;
 623			ceph_decode_32_safe(&p, end, len, bad);
 624			val = p;
 625			p += len;
 626
 627			err = __set_xattr(ci, name, namelen, val, len,
 628					  0, 0, &xattrs[numattr]);
 629
 630			if (err < 0)
 631				goto bad;
 632		}
 633		kfree(xattrs);
 634	}
 635	ci->i_xattrs.index_version = ci->i_xattrs.version;
 636	ci->i_xattrs.dirty = false;
 637
 638	return err;
 639bad_lock:
 640	spin_lock(&ci->i_ceph_lock);
 641bad:
 642	if (xattrs) {
 643		for (i = 0; i < numattr; i++)
 644			kfree(xattrs[i]);
 645		kfree(xattrs);
 646	}
 647	ci->i_xattrs.names_size = 0;
 648	return err;
 649}
 650
 651static int __get_required_blob_size(struct ceph_inode_info *ci, int name_size,
 652				    int val_size)
 653{
 654	/*
 655	 * 4 bytes for the length, and additional 4 bytes per each xattr name,
 656	 * 4 bytes per each value
 657	 */
 658	int size = 4 + ci->i_xattrs.count*(4 + 4) +
 659			     ci->i_xattrs.names_size +
 660			     ci->i_xattrs.vals_size;
 661	dout("__get_required_blob_size c=%d names.size=%d vals.size=%d\n",
 662	     ci->i_xattrs.count, ci->i_xattrs.names_size,
 663	     ci->i_xattrs.vals_size);
 664
 665	if (name_size)
 666		size += 4 + 4 + name_size + val_size;
 667
 668	return size;
 669}
 670
 671/*
 672 * If there are dirty xattrs, reencode xattrs into the prealloc_blob
 673 * and swap into place.
 674 */
 675void __ceph_build_xattrs_blob(struct ceph_inode_info *ci)
 676{
 677	struct rb_node *p;
 678	struct ceph_inode_xattr *xattr = NULL;
 679	void *dest;
 680
 681	dout("__build_xattrs_blob %p\n", &ci->vfs_inode);
 682	if (ci->i_xattrs.dirty) {
 683		int need = __get_required_blob_size(ci, 0, 0);
 684
 685		BUG_ON(need > ci->i_xattrs.prealloc_blob->alloc_len);
 686
 687		p = rb_first(&ci->i_xattrs.index);
 688		dest = ci->i_xattrs.prealloc_blob->vec.iov_base;
 689
 690		ceph_encode_32(&dest, ci->i_xattrs.count);
 691		while (p) {
 692			xattr = rb_entry(p, struct ceph_inode_xattr, node);
 693
 694			ceph_encode_32(&dest, xattr->name_len);
 695			memcpy(dest, xattr->name, xattr->name_len);
 696			dest += xattr->name_len;
 697			ceph_encode_32(&dest, xattr->val_len);
 698			memcpy(dest, xattr->val, xattr->val_len);
 699			dest += xattr->val_len;
 700
 701			p = rb_next(p);
 702		}
 703
 704		/* adjust buffer len; it may be larger than we need */
 705		ci->i_xattrs.prealloc_blob->vec.iov_len =
 706			dest - ci->i_xattrs.prealloc_blob->vec.iov_base;
 707
 708		if (ci->i_xattrs.blob)
 709			ceph_buffer_put(ci->i_xattrs.blob);
 710		ci->i_xattrs.blob = ci->i_xattrs.prealloc_blob;
 711		ci->i_xattrs.prealloc_blob = NULL;
 712		ci->i_xattrs.dirty = false;
 713		ci->i_xattrs.version++;
 714	}
 715}
 716
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 717ssize_t __ceph_getxattr(struct inode *inode, const char *name, void *value,
 718		      size_t size)
 719{
 720	struct ceph_inode_info *ci = ceph_inode(inode);
 721	int err;
 722	struct ceph_inode_xattr *xattr;
 723	struct ceph_vxattr *vxattr = NULL;
 724
 725	if (!ceph_is_valid_xattr(name))
 726		return -ENODATA;
 727
 728	/* let's see if a virtual xattr was requested */
 729	vxattr = ceph_match_vxattr(inode, name);
 730	if (vxattr && !(vxattr->exists_cb && !vxattr->exists_cb(ci))) {
 731		err = vxattr->getxattr_cb(ci, value, size);
 
 
 
 
 
 732		return err;
 733	}
 734
 
 
 735	spin_lock(&ci->i_ceph_lock);
 736	dout("getxattr %p ver=%lld index_ver=%lld\n", inode,
 737	     ci->i_xattrs.version, ci->i_xattrs.index_version);
 738
 739	if (__ceph_caps_issued_mask(ci, CEPH_CAP_XATTR_SHARED, 1) &&
 740	    (ci->i_xattrs.index_version >= ci->i_xattrs.version)) {
 741		goto get_xattr;
 742	} else {
 743		spin_unlock(&ci->i_ceph_lock);
 
 
 
 
 
 
 
 
 744		/* get xattrs from mds (if we don't already have them) */
 745		err = ceph_do_getattr(inode, CEPH_STAT_CAP_XATTR);
 746		if (err)
 747			return err;
 
 748	}
 749
 750	spin_lock(&ci->i_ceph_lock);
 751
 752	err = __build_xattrs(inode);
 753	if (err < 0)
 754		goto out;
 755
 756get_xattr:
 757	err = -ENODATA;  /* == ENOATTR */
 758	xattr = __get_xattr(ci, name);
 759	if (!xattr)
 760		goto out;
 761
 762	err = -ERANGE;
 763	if (size && size < xattr->val_len)
 764		goto out;
 765
 766	err = xattr->val_len;
 767	if (size == 0)
 768		goto out;
 769
 770	memcpy(value, xattr->val, xattr->val_len);
 771
 
 
 
 772out:
 773	spin_unlock(&ci->i_ceph_lock);
 774	return err;
 775}
 776
 777ssize_t ceph_getxattr(struct dentry *dentry, const char *name, void *value,
 778		      size_t size)
 779{
 780	if (!strncmp(name, XATTR_SYSTEM_PREFIX, XATTR_SYSTEM_PREFIX_LEN))
 781		return generic_getxattr(dentry, name, value, size);
 782
 783	return __ceph_getxattr(dentry->d_inode, name, value, size);
 784}
 785
 786ssize_t ceph_listxattr(struct dentry *dentry, char *names, size_t size)
 787{
 788	struct inode *inode = dentry->d_inode;
 789	struct ceph_inode_info *ci = ceph_inode(inode);
 790	struct ceph_vxattr *vxattrs = ceph_inode_vxattrs(inode);
 791	u32 vir_namelen = 0;
 792	u32 namelen;
 793	int err;
 794	u32 len;
 795	int i;
 796
 797	spin_lock(&ci->i_ceph_lock);
 798	dout("listxattr %p ver=%lld index_ver=%lld\n", inode,
 799	     ci->i_xattrs.version, ci->i_xattrs.index_version);
 800
 801	if (__ceph_caps_issued_mask(ci, CEPH_CAP_XATTR_SHARED, 1) &&
 802	    (ci->i_xattrs.index_version >= ci->i_xattrs.version)) {
 803		goto list_xattr;
 804	} else {
 805		spin_unlock(&ci->i_ceph_lock);
 806		err = ceph_do_getattr(inode, CEPH_STAT_CAP_XATTR);
 807		if (err)
 808			return err;
 
 809	}
 810
 811	spin_lock(&ci->i_ceph_lock);
 812
 813	err = __build_xattrs(inode);
 814	if (err < 0)
 815		goto out;
 816
 817list_xattr:
 818	/*
 819	 * Start with virtual dir xattr names (if any) (including
 820	 * terminating '\0' characters for each).
 821	 */
 822	vir_namelen = ceph_vxattrs_name_size(vxattrs);
 823
 824	/* adding 1 byte per each variable due to the null termination */
 825	namelen = ci->i_xattrs.names_size + ci->i_xattrs.count;
 826	err = -ERANGE;
 827	if (size && vir_namelen + namelen > size)
 828		goto out;
 829
 830	err = namelen + vir_namelen;
 831	if (size == 0)
 832		goto out;
 833
 834	names = __copy_xattr_names(ci, names);
 835
 836	/* virtual xattr names, too */
 837	err = namelen;
 838	if (vxattrs) {
 839		for (i = 0; vxattrs[i].name; i++) {
 840			if (!vxattrs[i].hidden &&
 841			    !(vxattrs[i].exists_cb &&
 842			      !vxattrs[i].exists_cb(ci))) {
 843				len = sprintf(names, "%s", vxattrs[i].name);
 844				names += len + 1;
 845				err += len + 1;
 846			}
 847		}
 848	}
 849
 850out:
 851	spin_unlock(&ci->i_ceph_lock);
 852	return err;
 853}
 854
 855static int ceph_sync_setxattr(struct dentry *dentry, const char *name,
 856			      const char *value, size_t size, int flags)
 857{
 858	struct ceph_fs_client *fsc = ceph_sb_to_client(dentry->d_sb);
 859	struct inode *inode = dentry->d_inode;
 860	struct ceph_inode_info *ci = ceph_inode(inode);
 861	struct ceph_mds_request *req;
 862	struct ceph_mds_client *mdsc = fsc->mdsc;
 
 
 863	int err;
 864	int i, nr_pages;
 865	struct page **pages = NULL;
 866	void *kaddr;
 867
 868	/* copy value into some pages */
 869	nr_pages = calc_pages_for(0, size);
 870	if (nr_pages) {
 871		pages = kmalloc(sizeof(pages[0])*nr_pages, GFP_NOFS);
 872		if (!pages)
 873			return -ENOMEM;
 874		err = -ENOMEM;
 875		for (i = 0; i < nr_pages; i++) {
 876			pages[i] = __page_cache_alloc(GFP_NOFS);
 877			if (!pages[i]) {
 878				nr_pages = i;
 879				goto out;
 880			}
 881			kaddr = kmap(pages[i]);
 882			memcpy(kaddr, value + i*PAGE_CACHE_SIZE,
 883			       min(PAGE_CACHE_SIZE, size-i*PAGE_CACHE_SIZE));
 884		}
 885	}
 886
 887	dout("setxattr value=%.*s\n", (int)size, value);
 888
 889	if (!value)
 890		flags |= CEPH_XATTR_REMOVE;
 891
 892	/* do request */
 893	req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_SETXATTR,
 894				       USE_AUTH_MDS);
 895	if (IS_ERR(req)) {
 896		err = PTR_ERR(req);
 897		goto out;
 898	}
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 899	req->r_inode = inode;
 900	ihold(inode);
 901	req->r_inode_drop = CEPH_CAP_XATTR_SHARED;
 902	req->r_num_caps = 1;
 903	req->r_args.setxattr.flags = cpu_to_le32(flags);
 904	req->r_path2 = kstrdup(name, GFP_NOFS);
 905
 906	req->r_pages = pages;
 907	req->r_num_pages = nr_pages;
 908	req->r_data_len = size;
 909
 910	dout("xattr.ver (before): %lld\n", ci->i_xattrs.version);
 911	err = ceph_mdsc_do_request(mdsc, NULL, req);
 912	ceph_mdsc_put_request(req);
 913	dout("xattr.ver (after): %lld\n", ci->i_xattrs.version);
 914
 915out:
 916	if (pages) {
 917		for (i = 0; i < nr_pages; i++)
 918			__free_page(pages[i]);
 919		kfree(pages);
 920	}
 921	return err;
 922}
 923
 924int __ceph_setxattr(struct dentry *dentry, const char *name,
 925			const void *value, size_t size, int flags)
 926{
 927	struct inode *inode = dentry->d_inode;
 928	struct ceph_vxattr *vxattr;
 929	struct ceph_inode_info *ci = ceph_inode(inode);
 
 
 930	int issued;
 931	int err;
 932	int dirty = 0;
 933	int name_len = strlen(name);
 934	int val_len = size;
 935	char *newname = NULL;
 936	char *newval = NULL;
 937	struct ceph_inode_xattr *xattr = NULL;
 938	int required_blob_size;
 
 
 939
 940	if (!ceph_is_valid_xattr(name))
 941		return -EOPNOTSUPP;
 942
 943	vxattr = ceph_match_vxattr(inode, name);
 944	if (vxattr && vxattr->readonly)
 945		return -EOPNOTSUPP;
 
 
 
 
 946
 947	/* pass any unhandled ceph.* xattrs through to the MDS */
 948	if (!strncmp(name, XATTR_CEPH_PREFIX, XATTR_CEPH_PREFIX_LEN))
 949		goto do_sync_unlocked;
 950
 951	/* preallocate memory for xattr name, value, index node */
 952	err = -ENOMEM;
 953	newname = kmemdup(name, name_len + 1, GFP_NOFS);
 954	if (!newname)
 955		goto out;
 956
 957	if (val_len) {
 958		newval = kmemdup(value, val_len, GFP_NOFS);
 959		if (!newval)
 960			goto out;
 961	}
 962
 963	xattr = kmalloc(sizeof(struct ceph_inode_xattr), GFP_NOFS);
 964	if (!xattr)
 965		goto out;
 966
 
 
 
 
 967	spin_lock(&ci->i_ceph_lock);
 968retry:
 969	issued = __ceph_caps_issued(ci, NULL);
 970	dout("setxattr %p issued %s\n", inode, ceph_cap_string(issued));
 971	if (!(issued & CEPH_CAP_XATTR_EXCL))
 972		goto do_sync;
 
 
 
 
 
 
 
 
 
 
 
 
 973	__build_xattrs(inode);
 974
 975	required_blob_size = __get_required_blob_size(ci, name_len, val_len);
 976
 977	if (!ci->i_xattrs.prealloc_blob ||
 978	    required_blob_size > ci->i_xattrs.prealloc_blob->alloc_len) {
 979		struct ceph_buffer *blob;
 980
 981		spin_unlock(&ci->i_ceph_lock);
 982		dout(" preaallocating new blob size=%d\n", required_blob_size);
 983		blob = ceph_buffer_new(required_blob_size, GFP_NOFS);
 984		if (!blob)
 985			goto out;
 986		spin_lock(&ci->i_ceph_lock);
 987		if (ci->i_xattrs.prealloc_blob)
 988			ceph_buffer_put(ci->i_xattrs.prealloc_blob);
 989		ci->i_xattrs.prealloc_blob = blob;
 990		goto retry;
 991	}
 992
 993	err = __set_xattr(ci, newname, name_len, newval, val_len,
 994			  flags, value ? 1 : -1, &xattr);
 995
 996	if (!err) {
 997		dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_XATTR_EXCL);
 
 998		ci->i_xattrs.dirty = true;
 999		inode->i_ctime = CURRENT_TIME;
1000	}
1001
1002	spin_unlock(&ci->i_ceph_lock);
 
 
1003	if (dirty)
1004		__mark_inode_dirty(inode, dirty);
 
1005	return err;
1006
1007do_sync:
1008	spin_unlock(&ci->i_ceph_lock);
1009do_sync_unlocked:
1010	err = ceph_sync_setxattr(dentry, name, value, size, flags);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1011out:
 
1012	kfree(newname);
1013	kfree(newval);
1014	kfree(xattr);
1015	return err;
1016}
1017
1018int ceph_setxattr(struct dentry *dentry, const char *name,
1019		  const void *value, size_t size, int flags)
 
1020{
1021	if (ceph_snap(dentry->d_inode) != CEPH_NOSNAP)
1022		return -EROFS;
1023
1024	if (!strncmp(name, XATTR_SYSTEM_PREFIX, XATTR_SYSTEM_PREFIX_LEN))
1025		return generic_setxattr(dentry, name, value, size, flags);
1026
1027	return __ceph_setxattr(dentry, name, value, size, flags);
1028}
1029
1030static int ceph_send_removexattr(struct dentry *dentry, const char *name)
1031{
1032	struct ceph_fs_client *fsc = ceph_sb_to_client(dentry->d_sb);
1033	struct ceph_mds_client *mdsc = fsc->mdsc;
1034	struct inode *inode = dentry->d_inode;
1035	struct ceph_mds_request *req;
1036	int err;
1037
1038	req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_RMXATTR,
1039				       USE_AUTH_MDS);
1040	if (IS_ERR(req))
1041		return PTR_ERR(req);
1042	req->r_inode = inode;
1043	ihold(inode);
1044	req->r_inode_drop = CEPH_CAP_XATTR_SHARED;
1045	req->r_num_caps = 1;
1046	req->r_path2 = kstrdup(name, GFP_NOFS);
1047
1048	err = ceph_mdsc_do_request(mdsc, NULL, req);
1049	ceph_mdsc_put_request(req);
1050	return err;
1051}
1052
1053int __ceph_removexattr(struct dentry *dentry, const char *name)
 
 
 
1054{
1055	struct inode *inode = dentry->d_inode;
1056	struct ceph_vxattr *vxattr;
1057	struct ceph_inode_info *ci = ceph_inode(inode);
1058	int issued;
1059	int err;
1060	int required_blob_size;
1061	int dirty;
1062
1063	if (!ceph_is_valid_xattr(name))
1064		return -EOPNOTSUPP;
 
 
1065
1066	vxattr = ceph_match_vxattr(inode, name);
1067	if (vxattr && vxattr->readonly)
1068		return -EOPNOTSUPP;
1069
1070	/* pass any unhandled ceph.* xattrs through to the MDS */
1071	if (!strncmp(name, XATTR_CEPH_PREFIX, XATTR_CEPH_PREFIX_LEN))
1072		goto do_sync_unlocked;
1073
1074	err = -ENOMEM;
1075	spin_lock(&ci->i_ceph_lock);
1076retry:
1077	issued = __ceph_caps_issued(ci, NULL);
1078	dout("removexattr %p issued %s\n", inode, ceph_cap_string(issued));
1079
1080	if (!(issued & CEPH_CAP_XATTR_EXCL))
1081		goto do_sync;
1082	__build_xattrs(inode);
1083
1084	required_blob_size = __get_required_blob_size(ci, 0, 0);
1085
1086	if (!ci->i_xattrs.prealloc_blob ||
1087	    required_blob_size > ci->i_xattrs.prealloc_blob->alloc_len) {
1088		struct ceph_buffer *blob;
1089
1090		spin_unlock(&ci->i_ceph_lock);
1091		dout(" preaallocating new blob size=%d\n", required_blob_size);
1092		blob = ceph_buffer_new(required_blob_size, GFP_NOFS);
1093		if (!blob)
1094			goto out;
1095		spin_lock(&ci->i_ceph_lock);
1096		if (ci->i_xattrs.prealloc_blob)
1097			ceph_buffer_put(ci->i_xattrs.prealloc_blob);
1098		ci->i_xattrs.prealloc_blob = blob;
1099		goto retry;
1100	}
1101
1102	err = __remove_xattr_by_name(ceph_inode(inode), name);
1103
1104	dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_XATTR_EXCL);
1105	ci->i_xattrs.dirty = true;
1106	inode->i_ctime = CURRENT_TIME;
1107	spin_unlock(&ci->i_ceph_lock);
1108	if (dirty)
1109		__mark_inode_dirty(inode, dirty);
1110	return err;
1111do_sync:
1112	spin_unlock(&ci->i_ceph_lock);
1113do_sync_unlocked:
1114	err = ceph_send_removexattr(dentry, name);
1115out:
1116	return err;
1117}
1118
1119int ceph_removexattr(struct dentry *dentry, const char *name)
1120{
1121	if (ceph_snap(dentry->d_inode) != CEPH_NOSNAP)
1122		return -EROFS;
1123
1124	if (!strncmp(name, XATTR_SYSTEM_PREFIX, XATTR_SYSTEM_PREFIX_LEN))
1125		return generic_removexattr(dentry, name);
1126
1127	return __ceph_removexattr(dentry, name);
 
 
 
 
1128}
v4.17
   1// SPDX-License-Identifier: GPL-2.0
   2#include <linux/ceph/ceph_debug.h>
   3#include <linux/ceph/pagelist.h>
   4
   5#include "super.h"
   6#include "mds_client.h"
   7
   8#include <linux/ceph/decode.h>
   9
  10#include <linux/xattr.h>
  11#include <linux/posix_acl_xattr.h>
  12#include <linux/slab.h>
  13
  14#define XATTR_CEPH_PREFIX "ceph."
  15#define XATTR_CEPH_PREFIX_LEN (sizeof (XATTR_CEPH_PREFIX) - 1)
  16
  17static int __remove_xattr(struct ceph_inode_info *ci,
  18			  struct ceph_inode_xattr *xattr);
  19
  20static const struct xattr_handler ceph_other_xattr_handler;
  21
  22/*
  23 * List of handlers for synthetic system.* attributes. Other
  24 * attributes are handled directly.
  25 */
  26const struct xattr_handler *ceph_xattr_handlers[] = {
  27#ifdef CONFIG_CEPH_FS_POSIX_ACL
  28	&posix_acl_access_xattr_handler,
  29	&posix_acl_default_xattr_handler,
  30#endif
  31	&ceph_other_xattr_handler,
  32	NULL,
  33};
  34
  35static bool ceph_is_valid_xattr(const char *name)
  36{
  37	return !strncmp(name, XATTR_CEPH_PREFIX, XATTR_CEPH_PREFIX_LEN) ||
  38	       !strncmp(name, XATTR_SECURITY_PREFIX,
  39			XATTR_SECURITY_PREFIX_LEN) ||
 
  40	       !strncmp(name, XATTR_TRUSTED_PREFIX, XATTR_TRUSTED_PREFIX_LEN) ||
  41	       !strncmp(name, XATTR_USER_PREFIX, XATTR_USER_PREFIX_LEN);
  42}
  43
  44/*
  45 * These define virtual xattrs exposing the recursive directory
  46 * statistics and layout metadata.
  47 */
  48struct ceph_vxattr {
  49	char *name;
  50	size_t name_size;	/* strlen(name) + 1 (for '\0') */
  51	size_t (*getxattr_cb)(struct ceph_inode_info *ci, char *val,
  52			      size_t size);
  53	bool readonly, hidden;
  54	bool (*exists_cb)(struct ceph_inode_info *ci);
  55};
  56
  57/* layouts */
  58
  59static bool ceph_vxattrcb_layout_exists(struct ceph_inode_info *ci)
  60{
  61	struct ceph_file_layout *fl = &ci->i_layout;
  62	return (fl->stripe_unit > 0 || fl->stripe_count > 0 ||
  63		fl->object_size > 0 || fl->pool_id >= 0 ||
  64		rcu_dereference_raw(fl->pool_ns) != NULL);
 
 
 
  65}
  66
  67static size_t ceph_vxattrcb_layout(struct ceph_inode_info *ci, char *val,
  68				   size_t size)
  69{
 
  70	struct ceph_fs_client *fsc = ceph_sb_to_client(ci->vfs_inode.i_sb);
  71	struct ceph_osd_client *osdc = &fsc->client->osdc;
  72	struct ceph_string *pool_ns;
  73	s64 pool = ci->i_layout.pool_id;
  74	const char *pool_name;
  75	const char *ns_field = " pool_namespace=";
  76	char buf[128];
  77	size_t len, total_len = 0;
  78	int ret;
  79
  80	pool_ns = ceph_try_get_string(ci->i_layout.pool_ns);
  81
  82	dout("ceph_vxattrcb_layout %p\n", &ci->vfs_inode);
  83	down_read(&osdc->lock);
  84	pool_name = ceph_pg_pool_name_by_id(osdc->osdmap, pool);
  85	if (pool_name) {
  86		len = snprintf(buf, sizeof(buf),
  87		"stripe_unit=%u stripe_count=%u object_size=%u pool=",
  88		ci->i_layout.stripe_unit, ci->i_layout.stripe_count,
  89	        ci->i_layout.object_size);
  90		total_len = len + strlen(pool_name);
  91	} else {
  92		len = snprintf(buf, sizeof(buf),
  93		"stripe_unit=%u stripe_count=%u object_size=%u pool=%lld",
  94		ci->i_layout.stripe_unit, ci->i_layout.stripe_count,
  95	        ci->i_layout.object_size, (unsigned long long)pool);
  96		total_len = len;
  97	}
  98
  99	if (pool_ns)
 100		total_len += strlen(ns_field) + pool_ns->len;
 101
 102	if (!size) {
 103		ret = total_len;
 104	} else if (total_len > size) {
 105		ret = -ERANGE;
 106	} else {
 107		memcpy(val, buf, len);
 108		ret = len;
 109		if (pool_name) {
 110			len = strlen(pool_name);
 111			memcpy(val + ret, pool_name, len);
 112			ret += len;
 113		}
 114		if (pool_ns) {
 115			len = strlen(ns_field);
 116			memcpy(val + ret, ns_field, len);
 117			ret += len;
 118			memcpy(val + ret, pool_ns->str, pool_ns->len);
 119			ret += pool_ns->len;
 
 
 
 
 
 
 120		}
 121	}
 122	up_read(&osdc->lock);
 123	ceph_put_string(pool_ns);
 124	return ret;
 125}
 126
 127static size_t ceph_vxattrcb_layout_stripe_unit(struct ceph_inode_info *ci,
 128					       char *val, size_t size)
 129{
 130	return snprintf(val, size, "%u", ci->i_layout.stripe_unit);
 
 131}
 132
 133static size_t ceph_vxattrcb_layout_stripe_count(struct ceph_inode_info *ci,
 134						char *val, size_t size)
 135{
 136	return snprintf(val, size, "%u", ci->i_layout.stripe_count);
 
 137}
 138
 139static size_t ceph_vxattrcb_layout_object_size(struct ceph_inode_info *ci,
 140					       char *val, size_t size)
 141{
 142	return snprintf(val, size, "%u", ci->i_layout.object_size);
 
 143}
 144
 145static size_t ceph_vxattrcb_layout_pool(struct ceph_inode_info *ci,
 146					char *val, size_t size)
 147{
 148	int ret;
 149	struct ceph_fs_client *fsc = ceph_sb_to_client(ci->vfs_inode.i_sb);
 150	struct ceph_osd_client *osdc = &fsc->client->osdc;
 151	s64 pool = ci->i_layout.pool_id;
 152	const char *pool_name;
 153
 154	down_read(&osdc->lock);
 155	pool_name = ceph_pg_pool_name_by_id(osdc->osdmap, pool);
 156	if (pool_name)
 157		ret = snprintf(val, size, "%s", pool_name);
 158	else
 159		ret = snprintf(val, size, "%lld", (unsigned long long)pool);
 160	up_read(&osdc->lock);
 161	return ret;
 162}
 163
 164static size_t ceph_vxattrcb_layout_pool_namespace(struct ceph_inode_info *ci,
 165						  char *val, size_t size)
 166{
 167	int ret = 0;
 168	struct ceph_string *ns = ceph_try_get_string(ci->i_layout.pool_ns);
 169	if (ns) {
 170		ret = snprintf(val, size, "%.*s", (int)ns->len, ns->str);
 171		ceph_put_string(ns);
 172	}
 173	return ret;
 174}
 175
 176/* directories */
 177
 178static size_t ceph_vxattrcb_dir_entries(struct ceph_inode_info *ci, char *val,
 179					size_t size)
 180{
 181	return snprintf(val, size, "%lld", ci->i_files + ci->i_subdirs);
 182}
 183
 184static size_t ceph_vxattrcb_dir_files(struct ceph_inode_info *ci, char *val,
 185				      size_t size)
 186{
 187	return snprintf(val, size, "%lld", ci->i_files);
 188}
 189
 190static size_t ceph_vxattrcb_dir_subdirs(struct ceph_inode_info *ci, char *val,
 191					size_t size)
 192{
 193	return snprintf(val, size, "%lld", ci->i_subdirs);
 194}
 195
 196static size_t ceph_vxattrcb_dir_rentries(struct ceph_inode_info *ci, char *val,
 197					 size_t size)
 198{
 199	return snprintf(val, size, "%lld", ci->i_rfiles + ci->i_rsubdirs);
 200}
 201
 202static size_t ceph_vxattrcb_dir_rfiles(struct ceph_inode_info *ci, char *val,
 203				       size_t size)
 204{
 205	return snprintf(val, size, "%lld", ci->i_rfiles);
 206}
 207
 208static size_t ceph_vxattrcb_dir_rsubdirs(struct ceph_inode_info *ci, char *val,
 209					 size_t size)
 210{
 211	return snprintf(val, size, "%lld", ci->i_rsubdirs);
 212}
 213
 214static size_t ceph_vxattrcb_dir_rbytes(struct ceph_inode_info *ci, char *val,
 215				       size_t size)
 216{
 217	return snprintf(val, size, "%lld", ci->i_rbytes);
 218}
 219
 220static size_t ceph_vxattrcb_dir_rctime(struct ceph_inode_info *ci, char *val,
 221				       size_t size)
 222{
 223	return snprintf(val, size, "%ld.09%ld", (long)ci->i_rctime.tv_sec,
 224			(long)ci->i_rctime.tv_nsec);
 225}
 226
 227/* quotas */
 228
 229static bool ceph_vxattrcb_quota_exists(struct ceph_inode_info *ci)
 230{
 231	bool ret = false;
 232	spin_lock(&ci->i_ceph_lock);
 233	if ((ci->i_max_files || ci->i_max_bytes) &&
 234	    ci->i_vino.snap == CEPH_NOSNAP &&
 235	    ci->i_snap_realm &&
 236	    ci->i_snap_realm->ino == ci->i_vino.ino)
 237		ret = true;
 238	spin_unlock(&ci->i_ceph_lock);
 239	return ret;
 240}
 241
 242static size_t ceph_vxattrcb_quota(struct ceph_inode_info *ci, char *val,
 243				  size_t size)
 244{
 245	return snprintf(val, size, "max_bytes=%llu max_files=%llu",
 246			ci->i_max_bytes, ci->i_max_files);
 247}
 248
 249static size_t ceph_vxattrcb_quota_max_bytes(struct ceph_inode_info *ci,
 250					    char *val, size_t size)
 251{
 252	return snprintf(val, size, "%llu", ci->i_max_bytes);
 253}
 254
 255static size_t ceph_vxattrcb_quota_max_files(struct ceph_inode_info *ci,
 256					    char *val, size_t size)
 257{
 258	return snprintf(val, size, "%llu", ci->i_max_files);
 259}
 260
 261#define CEPH_XATTR_NAME(_type, _name)	XATTR_CEPH_PREFIX #_type "." #_name
 262#define CEPH_XATTR_NAME2(_type, _name, _name2)	\
 263	XATTR_CEPH_PREFIX #_type "." #_name "." #_name2
 264
 265#define XATTR_NAME_CEPH(_type, _name)					\
 266	{								\
 267		.name = CEPH_XATTR_NAME(_type, _name),			\
 268		.name_size = sizeof (CEPH_XATTR_NAME(_type, _name)), \
 269		.getxattr_cb = ceph_vxattrcb_ ## _type ## _ ## _name, \
 270		.readonly = true,				\
 271		.hidden = false,				\
 272		.exists_cb = NULL,			\
 273	}
 274#define XATTR_LAYOUT_FIELD(_type, _name, _field)			\
 275	{								\
 276		.name = CEPH_XATTR_NAME2(_type, _name, _field),	\
 277		.name_size = sizeof (CEPH_XATTR_NAME2(_type, _name, _field)), \
 278		.getxattr_cb = ceph_vxattrcb_ ## _name ## _ ## _field, \
 279		.readonly = false,				\
 280		.hidden = true,			\
 281		.exists_cb = ceph_vxattrcb_layout_exists,	\
 282	}
 283#define XATTR_QUOTA_FIELD(_type, _name)					\
 284	{								\
 285		.name = CEPH_XATTR_NAME(_type, _name),			\
 286		.name_size = sizeof(CEPH_XATTR_NAME(_type, _name)),	\
 287		.getxattr_cb = ceph_vxattrcb_ ## _type ## _ ## _name,	\
 288		.readonly = false,					\
 289		.hidden = true,						\
 290		.exists_cb = ceph_vxattrcb_quota_exists,		\
 291	}
 292
 293static struct ceph_vxattr ceph_dir_vxattrs[] = {
 294	{
 295		.name = "ceph.dir.layout",
 296		.name_size = sizeof("ceph.dir.layout"),
 297		.getxattr_cb = ceph_vxattrcb_layout,
 298		.readonly = false,
 299		.hidden = true,
 300		.exists_cb = ceph_vxattrcb_layout_exists,
 301	},
 302	XATTR_LAYOUT_FIELD(dir, layout, stripe_unit),
 303	XATTR_LAYOUT_FIELD(dir, layout, stripe_count),
 304	XATTR_LAYOUT_FIELD(dir, layout, object_size),
 305	XATTR_LAYOUT_FIELD(dir, layout, pool),
 306	XATTR_LAYOUT_FIELD(dir, layout, pool_namespace),
 307	XATTR_NAME_CEPH(dir, entries),
 308	XATTR_NAME_CEPH(dir, files),
 309	XATTR_NAME_CEPH(dir, subdirs),
 310	XATTR_NAME_CEPH(dir, rentries),
 311	XATTR_NAME_CEPH(dir, rfiles),
 312	XATTR_NAME_CEPH(dir, rsubdirs),
 313	XATTR_NAME_CEPH(dir, rbytes),
 314	XATTR_NAME_CEPH(dir, rctime),
 315	{
 316		.name = "ceph.quota",
 317		.name_size = sizeof("ceph.quota"),
 318		.getxattr_cb = ceph_vxattrcb_quota,
 319		.readonly = false,
 320		.hidden = true,
 321		.exists_cb = ceph_vxattrcb_quota_exists,
 322	},
 323	XATTR_QUOTA_FIELD(quota, max_bytes),
 324	XATTR_QUOTA_FIELD(quota, max_files),
 325	{ .name = NULL, 0 }	/* Required table terminator */
 326};
 327static size_t ceph_dir_vxattrs_name_size;	/* total size of all names */
 328
 329/* files */
 330
 331static struct ceph_vxattr ceph_file_vxattrs[] = {
 332	{
 333		.name = "ceph.file.layout",
 334		.name_size = sizeof("ceph.file.layout"),
 335		.getxattr_cb = ceph_vxattrcb_layout,
 336		.readonly = false,
 337		.hidden = true,
 338		.exists_cb = ceph_vxattrcb_layout_exists,
 339	},
 340	XATTR_LAYOUT_FIELD(file, layout, stripe_unit),
 341	XATTR_LAYOUT_FIELD(file, layout, stripe_count),
 342	XATTR_LAYOUT_FIELD(file, layout, object_size),
 343	XATTR_LAYOUT_FIELD(file, layout, pool),
 344	XATTR_LAYOUT_FIELD(file, layout, pool_namespace),
 345	{ .name = NULL, 0 }	/* Required table terminator */
 346};
 347static size_t ceph_file_vxattrs_name_size;	/* total size of all names */
 348
 349static struct ceph_vxattr *ceph_inode_vxattrs(struct inode *inode)
 350{
 351	if (S_ISDIR(inode->i_mode))
 352		return ceph_dir_vxattrs;
 353	else if (S_ISREG(inode->i_mode))
 354		return ceph_file_vxattrs;
 355	return NULL;
 356}
 357
 358static size_t ceph_vxattrs_name_size(struct ceph_vxattr *vxattrs)
 359{
 360	if (vxattrs == ceph_dir_vxattrs)
 361		return ceph_dir_vxattrs_name_size;
 362	if (vxattrs == ceph_file_vxattrs)
 363		return ceph_file_vxattrs_name_size;
 364	BUG_ON(vxattrs);
 
 365	return 0;
 366}
 367
 368/*
 369 * Compute the aggregate size (including terminating '\0') of all
 370 * virtual extended attribute names in the given vxattr table.
 371 */
 372static size_t __init vxattrs_name_size(struct ceph_vxattr *vxattrs)
 373{
 374	struct ceph_vxattr *vxattr;
 375	size_t size = 0;
 376
 377	for (vxattr = vxattrs; vxattr->name; vxattr++)
 378		if (!vxattr->hidden)
 379			size += vxattr->name_size;
 380
 381	return size;
 382}
 383
 384/* Routines called at initialization and exit time */
 385
 386void __init ceph_xattr_init(void)
 387{
 388	ceph_dir_vxattrs_name_size = vxattrs_name_size(ceph_dir_vxattrs);
 389	ceph_file_vxattrs_name_size = vxattrs_name_size(ceph_file_vxattrs);
 390}
 391
 392void ceph_xattr_exit(void)
 393{
 394	ceph_dir_vxattrs_name_size = 0;
 395	ceph_file_vxattrs_name_size = 0;
 396}
 397
 398static struct ceph_vxattr *ceph_match_vxattr(struct inode *inode,
 399						const char *name)
 400{
 401	struct ceph_vxattr *vxattr = ceph_inode_vxattrs(inode);
 402
 403	if (vxattr) {
 404		while (vxattr->name) {
 405			if (!strcmp(vxattr->name, name))
 406				return vxattr;
 407			vxattr++;
 408		}
 409	}
 410
 411	return NULL;
 412}
 413
 414static int __set_xattr(struct ceph_inode_info *ci,
 415			   const char *name, int name_len,
 416			   const char *val, int val_len,
 417			   int flags, int update_xattr,
 418			   struct ceph_inode_xattr **newxattr)
 419{
 420	struct rb_node **p;
 421	struct rb_node *parent = NULL;
 422	struct ceph_inode_xattr *xattr = NULL;
 423	int c;
 424	int new = 0;
 425
 426	p = &ci->i_xattrs.index.rb_node;
 427	while (*p) {
 428		parent = *p;
 429		xattr = rb_entry(parent, struct ceph_inode_xattr, node);
 430		c = strncmp(name, xattr->name, min(name_len, xattr->name_len));
 431		if (c < 0)
 432			p = &(*p)->rb_left;
 433		else if (c > 0)
 434			p = &(*p)->rb_right;
 435		else {
 436			if (name_len == xattr->name_len)
 437				break;
 438			else if (name_len < xattr->name_len)
 439				p = &(*p)->rb_left;
 440			else
 441				p = &(*p)->rb_right;
 442		}
 443		xattr = NULL;
 444	}
 445
 446	if (update_xattr) {
 447		int err = 0;
 448
 449		if (xattr && (flags & XATTR_CREATE))
 450			err = -EEXIST;
 451		else if (!xattr && (flags & XATTR_REPLACE))
 452			err = -ENODATA;
 453		if (err) {
 454			kfree(name);
 455			kfree(val);
 456			kfree(*newxattr);
 457			return err;
 458		}
 459		if (update_xattr < 0) {
 460			if (xattr)
 461				__remove_xattr(ci, xattr);
 462			kfree(name);
 463			kfree(*newxattr);
 464			return 0;
 465		}
 466	}
 467
 468	if (!xattr) {
 469		new = 1;
 470		xattr = *newxattr;
 471		xattr->name = name;
 472		xattr->name_len = name_len;
 473		xattr->should_free_name = update_xattr;
 474
 475		ci->i_xattrs.count++;
 476		dout("__set_xattr count=%d\n", ci->i_xattrs.count);
 477	} else {
 478		kfree(*newxattr);
 479		*newxattr = NULL;
 480		if (xattr->should_free_val)
 481			kfree((void *)xattr->val);
 482
 483		if (update_xattr) {
 484			kfree((void *)name);
 485			name = xattr->name;
 486		}
 487		ci->i_xattrs.names_size -= xattr->name_len;
 488		ci->i_xattrs.vals_size -= xattr->val_len;
 489	}
 490	ci->i_xattrs.names_size += name_len;
 491	ci->i_xattrs.vals_size += val_len;
 492	if (val)
 493		xattr->val = val;
 494	else
 495		xattr->val = "";
 496
 497	xattr->val_len = val_len;
 498	xattr->dirty = update_xattr;
 499	xattr->should_free_val = (val && update_xattr);
 500
 501	if (new) {
 502		rb_link_node(&xattr->node, parent, p);
 503		rb_insert_color(&xattr->node, &ci->i_xattrs.index);
 504		dout("__set_xattr_val p=%p\n", p);
 505	}
 506
 507	dout("__set_xattr_val added %llx.%llx xattr %p %s=%.*s\n",
 508	     ceph_vinop(&ci->vfs_inode), xattr, name, val_len, val);
 509
 510	return 0;
 511}
 512
 513static struct ceph_inode_xattr *__get_xattr(struct ceph_inode_info *ci,
 514			   const char *name)
 515{
 516	struct rb_node **p;
 517	struct rb_node *parent = NULL;
 518	struct ceph_inode_xattr *xattr = NULL;
 519	int name_len = strlen(name);
 520	int c;
 521
 522	p = &ci->i_xattrs.index.rb_node;
 523	while (*p) {
 524		parent = *p;
 525		xattr = rb_entry(parent, struct ceph_inode_xattr, node);
 526		c = strncmp(name, xattr->name, xattr->name_len);
 527		if (c == 0 && name_len > xattr->name_len)
 528			c = 1;
 529		if (c < 0)
 530			p = &(*p)->rb_left;
 531		else if (c > 0)
 532			p = &(*p)->rb_right;
 533		else {
 534			dout("__get_xattr %s: found %.*s\n", name,
 535			     xattr->val_len, xattr->val);
 536			return xattr;
 537		}
 538	}
 539
 540	dout("__get_xattr %s: not found\n", name);
 541
 542	return NULL;
 543}
 544
 545static void __free_xattr(struct ceph_inode_xattr *xattr)
 546{
 547	BUG_ON(!xattr);
 548
 549	if (xattr->should_free_name)
 550		kfree((void *)xattr->name);
 551	if (xattr->should_free_val)
 552		kfree((void *)xattr->val);
 553
 554	kfree(xattr);
 555}
 556
 557static int __remove_xattr(struct ceph_inode_info *ci,
 558			  struct ceph_inode_xattr *xattr)
 559{
 560	if (!xattr)
 561		return -ENODATA;
 562
 563	rb_erase(&xattr->node, &ci->i_xattrs.index);
 564
 565	if (xattr->should_free_name)
 566		kfree((void *)xattr->name);
 567	if (xattr->should_free_val)
 568		kfree((void *)xattr->val);
 569
 570	ci->i_xattrs.names_size -= xattr->name_len;
 571	ci->i_xattrs.vals_size -= xattr->val_len;
 572	ci->i_xattrs.count--;
 573	kfree(xattr);
 574
 575	return 0;
 576}
 577
 
 
 
 
 
 
 
 
 
 
 
 
 
 578static char *__copy_xattr_names(struct ceph_inode_info *ci,
 579				char *dest)
 580{
 581	struct rb_node *p;
 582	struct ceph_inode_xattr *xattr = NULL;
 583
 584	p = rb_first(&ci->i_xattrs.index);
 585	dout("__copy_xattr_names count=%d\n", ci->i_xattrs.count);
 586
 587	while (p) {
 588		xattr = rb_entry(p, struct ceph_inode_xattr, node);
 589		memcpy(dest, xattr->name, xattr->name_len);
 590		dest[xattr->name_len] = '\0';
 591
 592		dout("dest=%s %p (%s) (%d/%d)\n", dest, xattr, xattr->name,
 593		     xattr->name_len, ci->i_xattrs.names_size);
 594
 595		dest += xattr->name_len + 1;
 596		p = rb_next(p);
 597	}
 598
 599	return dest;
 600}
 601
 602void __ceph_destroy_xattrs(struct ceph_inode_info *ci)
 603{
 604	struct rb_node *p, *tmp;
 605	struct ceph_inode_xattr *xattr = NULL;
 606
 607	p = rb_first(&ci->i_xattrs.index);
 608
 609	dout("__ceph_destroy_xattrs p=%p\n", p);
 610
 611	while (p) {
 612		xattr = rb_entry(p, struct ceph_inode_xattr, node);
 613		tmp = p;
 614		p = rb_next(tmp);
 615		dout("__ceph_destroy_xattrs next p=%p (%.*s)\n", p,
 616		     xattr->name_len, xattr->name);
 617		rb_erase(tmp, &ci->i_xattrs.index);
 618
 619		__free_xattr(xattr);
 620	}
 621
 622	ci->i_xattrs.names_size = 0;
 623	ci->i_xattrs.vals_size = 0;
 624	ci->i_xattrs.index_version = 0;
 625	ci->i_xattrs.count = 0;
 626	ci->i_xattrs.index = RB_ROOT;
 627}
 628
 629static int __build_xattrs(struct inode *inode)
 630	__releases(ci->i_ceph_lock)
 631	__acquires(ci->i_ceph_lock)
 632{
 633	u32 namelen;
 634	u32 numattr = 0;
 635	void *p, *end;
 636	u32 len;
 637	const char *name, *val;
 638	struct ceph_inode_info *ci = ceph_inode(inode);
 639	int xattr_version;
 640	struct ceph_inode_xattr **xattrs = NULL;
 641	int err = 0;
 642	int i;
 643
 644	dout("__build_xattrs() len=%d\n",
 645	     ci->i_xattrs.blob ? (int)ci->i_xattrs.blob->vec.iov_len : 0);
 646
 647	if (ci->i_xattrs.index_version >= ci->i_xattrs.version)
 648		return 0; /* already built */
 649
 650	__ceph_destroy_xattrs(ci);
 651
 652start:
 653	/* updated internal xattr rb tree */
 654	if (ci->i_xattrs.blob && ci->i_xattrs.blob->vec.iov_len > 4) {
 655		p = ci->i_xattrs.blob->vec.iov_base;
 656		end = p + ci->i_xattrs.blob->vec.iov_len;
 657		ceph_decode_32_safe(&p, end, numattr, bad);
 658		xattr_version = ci->i_xattrs.version;
 659		spin_unlock(&ci->i_ceph_lock);
 660
 661		xattrs = kcalloc(numattr, sizeof(struct ceph_inode_xattr *),
 662				 GFP_NOFS);
 663		err = -ENOMEM;
 664		if (!xattrs)
 665			goto bad_lock;
 666
 667		for (i = 0; i < numattr; i++) {
 668			xattrs[i] = kmalloc(sizeof(struct ceph_inode_xattr),
 669					    GFP_NOFS);
 670			if (!xattrs[i])
 671				goto bad_lock;
 672		}
 673
 674		spin_lock(&ci->i_ceph_lock);
 675		if (ci->i_xattrs.version != xattr_version) {
 676			/* lost a race, retry */
 677			for (i = 0; i < numattr; i++)
 678				kfree(xattrs[i]);
 679			kfree(xattrs);
 680			xattrs = NULL;
 681			goto start;
 682		}
 683		err = -EIO;
 684		while (numattr--) {
 685			ceph_decode_32_safe(&p, end, len, bad);
 686			namelen = len;
 687			name = p;
 688			p += len;
 689			ceph_decode_32_safe(&p, end, len, bad);
 690			val = p;
 691			p += len;
 692
 693			err = __set_xattr(ci, name, namelen, val, len,
 694					  0, 0, &xattrs[numattr]);
 695
 696			if (err < 0)
 697				goto bad;
 698		}
 699		kfree(xattrs);
 700	}
 701	ci->i_xattrs.index_version = ci->i_xattrs.version;
 702	ci->i_xattrs.dirty = false;
 703
 704	return err;
 705bad_lock:
 706	spin_lock(&ci->i_ceph_lock);
 707bad:
 708	if (xattrs) {
 709		for (i = 0; i < numattr; i++)
 710			kfree(xattrs[i]);
 711		kfree(xattrs);
 712	}
 713	ci->i_xattrs.names_size = 0;
 714	return err;
 715}
 716
 717static int __get_required_blob_size(struct ceph_inode_info *ci, int name_size,
 718				    int val_size)
 719{
 720	/*
 721	 * 4 bytes for the length, and additional 4 bytes per each xattr name,
 722	 * 4 bytes per each value
 723	 */
 724	int size = 4 + ci->i_xattrs.count*(4 + 4) +
 725			     ci->i_xattrs.names_size +
 726			     ci->i_xattrs.vals_size;
 727	dout("__get_required_blob_size c=%d names.size=%d vals.size=%d\n",
 728	     ci->i_xattrs.count, ci->i_xattrs.names_size,
 729	     ci->i_xattrs.vals_size);
 730
 731	if (name_size)
 732		size += 4 + 4 + name_size + val_size;
 733
 734	return size;
 735}
 736
 737/*
 738 * If there are dirty xattrs, reencode xattrs into the prealloc_blob
 739 * and swap into place.
 740 */
 741void __ceph_build_xattrs_blob(struct ceph_inode_info *ci)
 742{
 743	struct rb_node *p;
 744	struct ceph_inode_xattr *xattr = NULL;
 745	void *dest;
 746
 747	dout("__build_xattrs_blob %p\n", &ci->vfs_inode);
 748	if (ci->i_xattrs.dirty) {
 749		int need = __get_required_blob_size(ci, 0, 0);
 750
 751		BUG_ON(need > ci->i_xattrs.prealloc_blob->alloc_len);
 752
 753		p = rb_first(&ci->i_xattrs.index);
 754		dest = ci->i_xattrs.prealloc_blob->vec.iov_base;
 755
 756		ceph_encode_32(&dest, ci->i_xattrs.count);
 757		while (p) {
 758			xattr = rb_entry(p, struct ceph_inode_xattr, node);
 759
 760			ceph_encode_32(&dest, xattr->name_len);
 761			memcpy(dest, xattr->name, xattr->name_len);
 762			dest += xattr->name_len;
 763			ceph_encode_32(&dest, xattr->val_len);
 764			memcpy(dest, xattr->val, xattr->val_len);
 765			dest += xattr->val_len;
 766
 767			p = rb_next(p);
 768		}
 769
 770		/* adjust buffer len; it may be larger than we need */
 771		ci->i_xattrs.prealloc_blob->vec.iov_len =
 772			dest - ci->i_xattrs.prealloc_blob->vec.iov_base;
 773
 774		if (ci->i_xattrs.blob)
 775			ceph_buffer_put(ci->i_xattrs.blob);
 776		ci->i_xattrs.blob = ci->i_xattrs.prealloc_blob;
 777		ci->i_xattrs.prealloc_blob = NULL;
 778		ci->i_xattrs.dirty = false;
 779		ci->i_xattrs.version++;
 780	}
 781}
 782
 783static inline int __get_request_mask(struct inode *in) {
 784	struct ceph_mds_request *req = current->journal_info;
 785	int mask = 0;
 786	if (req && req->r_target_inode == in) {
 787		if (req->r_op == CEPH_MDS_OP_LOOKUP ||
 788		    req->r_op == CEPH_MDS_OP_LOOKUPINO ||
 789		    req->r_op == CEPH_MDS_OP_LOOKUPPARENT ||
 790		    req->r_op == CEPH_MDS_OP_GETATTR) {
 791			mask = le32_to_cpu(req->r_args.getattr.mask);
 792		} else if (req->r_op == CEPH_MDS_OP_OPEN ||
 793			   req->r_op == CEPH_MDS_OP_CREATE) {
 794			mask = le32_to_cpu(req->r_args.open.mask);
 795		}
 796	}
 797	return mask;
 798}
 799
 800ssize_t __ceph_getxattr(struct inode *inode, const char *name, void *value,
 801		      size_t size)
 802{
 803	struct ceph_inode_info *ci = ceph_inode(inode);
 
 804	struct ceph_inode_xattr *xattr;
 805	struct ceph_vxattr *vxattr = NULL;
 806	int req_mask;
 807	int err;
 
 808
 809	/* let's see if a virtual xattr was requested */
 810	vxattr = ceph_match_vxattr(inode, name);
 811	if (vxattr) {
 812		err = ceph_do_getattr(inode, 0, true);
 813		if (err)
 814			return err;
 815		err = -ENODATA;
 816		if (!(vxattr->exists_cb && !vxattr->exists_cb(ci)))
 817			err = vxattr->getxattr_cb(ci, value, size);
 818		return err;
 819	}
 820
 821	req_mask = __get_request_mask(inode);
 822
 823	spin_lock(&ci->i_ceph_lock);
 824	dout("getxattr %p ver=%lld index_ver=%lld\n", inode,
 825	     ci->i_xattrs.version, ci->i_xattrs.index_version);
 826
 827	if (ci->i_xattrs.version == 0 ||
 828	    !((req_mask & CEPH_CAP_XATTR_SHARED) ||
 829	      __ceph_caps_issued_mask(ci, CEPH_CAP_XATTR_SHARED, 1))) {
 
 830		spin_unlock(&ci->i_ceph_lock);
 831
 832		/* security module gets xattr while filling trace */
 833		if (current->journal_info) {
 834			pr_warn_ratelimited("sync getxattr %p "
 835					    "during filling trace\n", inode);
 836			return -EBUSY;
 837		}
 838
 839		/* get xattrs from mds (if we don't already have them) */
 840		err = ceph_do_getattr(inode, CEPH_STAT_CAP_XATTR, true);
 841		if (err)
 842			return err;
 843		spin_lock(&ci->i_ceph_lock);
 844	}
 845
 
 
 846	err = __build_xattrs(inode);
 847	if (err < 0)
 848		goto out;
 849
 
 850	err = -ENODATA;  /* == ENOATTR */
 851	xattr = __get_xattr(ci, name);
 852	if (!xattr)
 853		goto out;
 854
 855	err = -ERANGE;
 856	if (size && size < xattr->val_len)
 857		goto out;
 858
 859	err = xattr->val_len;
 860	if (size == 0)
 861		goto out;
 862
 863	memcpy(value, xattr->val, xattr->val_len);
 864
 865	if (current->journal_info &&
 866	    !strncmp(name, XATTR_SECURITY_PREFIX, XATTR_SECURITY_PREFIX_LEN))
 867		ci->i_ceph_flags |= CEPH_I_SEC_INITED;
 868out:
 869	spin_unlock(&ci->i_ceph_lock);
 870	return err;
 871}
 872
 
 
 
 
 
 
 
 
 
 873ssize_t ceph_listxattr(struct dentry *dentry, char *names, size_t size)
 874{
 875	struct inode *inode = d_inode(dentry);
 876	struct ceph_inode_info *ci = ceph_inode(inode);
 877	struct ceph_vxattr *vxattrs = ceph_inode_vxattrs(inode);
 878	u32 vir_namelen = 0;
 879	u32 namelen;
 880	int err;
 881	u32 len;
 882	int i;
 883
 884	spin_lock(&ci->i_ceph_lock);
 885	dout("listxattr %p ver=%lld index_ver=%lld\n", inode,
 886	     ci->i_xattrs.version, ci->i_xattrs.index_version);
 887
 888	if (ci->i_xattrs.version == 0 ||
 889	    !__ceph_caps_issued_mask(ci, CEPH_CAP_XATTR_SHARED, 1)) {
 
 
 890		spin_unlock(&ci->i_ceph_lock);
 891		err = ceph_do_getattr(inode, CEPH_STAT_CAP_XATTR, true);
 892		if (err)
 893			return err;
 894		spin_lock(&ci->i_ceph_lock);
 895	}
 896
 
 
 897	err = __build_xattrs(inode);
 898	if (err < 0)
 899		goto out;
 
 
 900	/*
 901	 * Start with virtual dir xattr names (if any) (including
 902	 * terminating '\0' characters for each).
 903	 */
 904	vir_namelen = ceph_vxattrs_name_size(vxattrs);
 905
 906	/* adding 1 byte per each variable due to the null termination */
 907	namelen = ci->i_xattrs.names_size + ci->i_xattrs.count;
 908	err = -ERANGE;
 909	if (size && vir_namelen + namelen > size)
 910		goto out;
 911
 912	err = namelen + vir_namelen;
 913	if (size == 0)
 914		goto out;
 915
 916	names = __copy_xattr_names(ci, names);
 917
 918	/* virtual xattr names, too */
 919	err = namelen;
 920	if (vxattrs) {
 921		for (i = 0; vxattrs[i].name; i++) {
 922			if (!vxattrs[i].hidden &&
 923			    !(vxattrs[i].exists_cb &&
 924			      !vxattrs[i].exists_cb(ci))) {
 925				len = sprintf(names, "%s", vxattrs[i].name);
 926				names += len + 1;
 927				err += len + 1;
 928			}
 929		}
 930	}
 931
 932out:
 933	spin_unlock(&ci->i_ceph_lock);
 934	return err;
 935}
 936
 937static int ceph_sync_setxattr(struct inode *inode, const char *name,
 938			      const char *value, size_t size, int flags)
 939{
 940	struct ceph_fs_client *fsc = ceph_sb_to_client(inode->i_sb);
 
 941	struct ceph_inode_info *ci = ceph_inode(inode);
 942	struct ceph_mds_request *req;
 943	struct ceph_mds_client *mdsc = fsc->mdsc;
 944	struct ceph_pagelist *pagelist = NULL;
 945	int op = CEPH_MDS_OP_SETXATTR;
 946	int err;
 947
 948	if (size > 0) {
 949		/* copy value into pagelist */
 950		pagelist = kmalloc(sizeof(*pagelist), GFP_NOFS);
 951		if (!pagelist)
 
 
 
 
 952			return -ENOMEM;
 953
 954		ceph_pagelist_init(pagelist);
 955		err = ceph_pagelist_append(pagelist, value, size);
 956		if (err)
 957			goto out;
 958	} else if (!value) {
 959		if (flags & CEPH_XATTR_REPLACE)
 960			op = CEPH_MDS_OP_RMXATTR;
 961		else
 962			flags |= CEPH_XATTR_REMOVE;
 
 963	}
 964
 965	dout("setxattr value=%.*s\n", (int)size, value);
 966
 
 
 
 967	/* do request */
 968	req = ceph_mdsc_create_request(mdsc, op, USE_AUTH_MDS);
 
 969	if (IS_ERR(req)) {
 970		err = PTR_ERR(req);
 971		goto out;
 972	}
 973
 974	req->r_path2 = kstrdup(name, GFP_NOFS);
 975	if (!req->r_path2) {
 976		ceph_mdsc_put_request(req);
 977		err = -ENOMEM;
 978		goto out;
 979	}
 980
 981	if (op == CEPH_MDS_OP_SETXATTR) {
 982		req->r_args.setxattr.flags = cpu_to_le32(flags);
 983		req->r_pagelist = pagelist;
 984		pagelist = NULL;
 985	}
 986
 987	req->r_inode = inode;
 988	ihold(inode);
 
 989	req->r_num_caps = 1;
 990	req->r_inode_drop = CEPH_CAP_XATTR_SHARED;
 
 
 
 
 
 991
 992	dout("xattr.ver (before): %lld\n", ci->i_xattrs.version);
 993	err = ceph_mdsc_do_request(mdsc, NULL, req);
 994	ceph_mdsc_put_request(req);
 995	dout("xattr.ver (after): %lld\n", ci->i_xattrs.version);
 996
 997out:
 998	if (pagelist)
 999		ceph_pagelist_release(pagelist);
 
 
 
1000	return err;
1001}
1002
1003int __ceph_setxattr(struct inode *inode, const char *name,
1004			const void *value, size_t size, int flags)
1005{
 
1006	struct ceph_vxattr *vxattr;
1007	struct ceph_inode_info *ci = ceph_inode(inode);
1008	struct ceph_mds_client *mdsc = ceph_sb_to_client(inode->i_sb)->mdsc;
1009	struct ceph_cap_flush *prealloc_cf = NULL;
1010	int issued;
1011	int err;
1012	int dirty = 0;
1013	int name_len = strlen(name);
1014	int val_len = size;
1015	char *newname = NULL;
1016	char *newval = NULL;
1017	struct ceph_inode_xattr *xattr = NULL;
1018	int required_blob_size;
1019	bool check_realm = false;
1020	bool lock_snap_rwsem = false;
1021
1022	if (ceph_snap(inode) != CEPH_NOSNAP)
1023		return -EROFS;
1024
1025	vxattr = ceph_match_vxattr(inode, name);
1026	if (vxattr) {
1027		if (vxattr->readonly)
1028			return -EOPNOTSUPP;
1029		if (value && !strncmp(vxattr->name, "ceph.quota", 10))
1030			check_realm = true;
1031	}
1032
1033	/* pass any unhandled ceph.* xattrs through to the MDS */
1034	if (!strncmp(name, XATTR_CEPH_PREFIX, XATTR_CEPH_PREFIX_LEN))
1035		goto do_sync_unlocked;
1036
1037	/* preallocate memory for xattr name, value, index node */
1038	err = -ENOMEM;
1039	newname = kmemdup(name, name_len + 1, GFP_NOFS);
1040	if (!newname)
1041		goto out;
1042
1043	if (val_len) {
1044		newval = kmemdup(value, val_len, GFP_NOFS);
1045		if (!newval)
1046			goto out;
1047	}
1048
1049	xattr = kmalloc(sizeof(struct ceph_inode_xattr), GFP_NOFS);
1050	if (!xattr)
1051		goto out;
1052
1053	prealloc_cf = ceph_alloc_cap_flush();
1054	if (!prealloc_cf)
1055		goto out;
1056
1057	spin_lock(&ci->i_ceph_lock);
1058retry:
1059	issued = __ceph_caps_issued(ci, NULL);
1060	if (ci->i_xattrs.version == 0 || !(issued & CEPH_CAP_XATTR_EXCL))
 
1061		goto do_sync;
1062
1063	if (!lock_snap_rwsem && !ci->i_head_snapc) {
1064		lock_snap_rwsem = true;
1065		if (!down_read_trylock(&mdsc->snap_rwsem)) {
1066			spin_unlock(&ci->i_ceph_lock);
1067			down_read(&mdsc->snap_rwsem);
1068			spin_lock(&ci->i_ceph_lock);
1069			goto retry;
1070		}
1071	}
1072
1073	dout("setxattr %p issued %s\n", inode, ceph_cap_string(issued));
1074	__build_xattrs(inode);
1075
1076	required_blob_size = __get_required_blob_size(ci, name_len, val_len);
1077
1078	if (!ci->i_xattrs.prealloc_blob ||
1079	    required_blob_size > ci->i_xattrs.prealloc_blob->alloc_len) {
1080		struct ceph_buffer *blob;
1081
1082		spin_unlock(&ci->i_ceph_lock);
1083		dout(" preaallocating new blob size=%d\n", required_blob_size);
1084		blob = ceph_buffer_new(required_blob_size, GFP_NOFS);
1085		if (!blob)
1086			goto do_sync_unlocked;
1087		spin_lock(&ci->i_ceph_lock);
1088		if (ci->i_xattrs.prealloc_blob)
1089			ceph_buffer_put(ci->i_xattrs.prealloc_blob);
1090		ci->i_xattrs.prealloc_blob = blob;
1091		goto retry;
1092	}
1093
1094	err = __set_xattr(ci, newname, name_len, newval, val_len,
1095			  flags, value ? 1 : -1, &xattr);
1096
1097	if (!err) {
1098		dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_XATTR_EXCL,
1099					       &prealloc_cf);
1100		ci->i_xattrs.dirty = true;
1101		inode->i_ctime = current_time(inode);
1102	}
1103
1104	spin_unlock(&ci->i_ceph_lock);
1105	if (lock_snap_rwsem)
1106		up_read(&mdsc->snap_rwsem);
1107	if (dirty)
1108		__mark_inode_dirty(inode, dirty);
1109	ceph_free_cap_flush(prealloc_cf);
1110	return err;
1111
1112do_sync:
1113	spin_unlock(&ci->i_ceph_lock);
1114do_sync_unlocked:
1115	if (lock_snap_rwsem)
1116		up_read(&mdsc->snap_rwsem);
1117
1118	/* security module set xattr while filling trace */
1119	if (current->journal_info) {
1120		pr_warn_ratelimited("sync setxattr %p "
1121				    "during filling trace\n", inode);
1122		err = -EBUSY;
1123	} else {
1124		err = ceph_sync_setxattr(inode, name, value, size, flags);
1125		if (err >= 0 && check_realm) {
1126			/* check if snaprealm was created for quota inode */
1127			spin_lock(&ci->i_ceph_lock);
1128			if ((ci->i_max_files || ci->i_max_bytes) &&
1129			    !(ci->i_snap_realm &&
1130			      ci->i_snap_realm->ino == ci->i_vino.ino))
1131				err = -EOPNOTSUPP;
1132			spin_unlock(&ci->i_ceph_lock);
1133		}
1134	}
1135out:
1136	ceph_free_cap_flush(prealloc_cf);
1137	kfree(newname);
1138	kfree(newval);
1139	kfree(xattr);
1140	return err;
1141}
1142
1143static int ceph_get_xattr_handler(const struct xattr_handler *handler,
1144				  struct dentry *dentry, struct inode *inode,
1145				  const char *name, void *value, size_t size)
1146{
1147	if (!ceph_is_valid_xattr(name))
1148		return -EOPNOTSUPP;
1149	return __ceph_getxattr(inode, name, value, size);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1150}
1151
1152static int ceph_set_xattr_handler(const struct xattr_handler *handler,
1153				  struct dentry *unused, struct inode *inode,
1154				  const char *name, const void *value,
1155				  size_t size, int flags)
1156{
 
 
 
 
 
 
 
 
1157	if (!ceph_is_valid_xattr(name))
1158		return -EOPNOTSUPP;
1159	return __ceph_setxattr(inode, name, value, size, flags);
1160}
1161
1162static const struct xattr_handler ceph_other_xattr_handler = {
1163	.prefix = "",  /* match any name => handlers called with full name */
1164	.get = ceph_get_xattr_handler,
1165	.set = ceph_set_xattr_handler,
1166};
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1167
1168#ifdef CONFIG_SECURITY
1169bool ceph_security_xattr_wanted(struct inode *in)
1170{
1171	return in->i_security != NULL;
 
 
 
 
 
 
 
 
 
1172}
1173
1174bool ceph_security_xattr_deadlock(struct inode *in)
1175{
1176	struct ceph_inode_info *ci;
1177	bool ret;
1178	if (!in->i_security)
1179		return false;
1180	ci = ceph_inode(in);
1181	spin_lock(&ci->i_ceph_lock);
1182	ret = !(ci->i_ceph_flags & CEPH_I_SEC_INITED) &&
1183	      !(ci->i_xattrs.version > 0 &&
1184		__ceph_caps_issued_mask(ci, CEPH_CAP_XATTR_SHARED, 0));
1185	spin_unlock(&ci->i_ceph_lock);
1186	return ret;
1187}
1188#endif