1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /* Provide a way to create a superblock configuration context within the kernel
3 * that allows a superblock to be set up prior to mounting.
5 * Copyright (C) 2017 Red Hat, Inc. All Rights Reserved.
6 * Written by David Howells (dhowells@redhat.com)
9 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
10 #include <linux/module.h>
11 #include <linux/fs_context.h>
12 #include <linux/fs_parser.h>
14 #include <linux/mount.h>
15 #include <linux/nsproxy.h>
16 #include <linux/slab.h>
17 #include <linux/magic.h>
18 #include <linux/security.h>
19 #include <linux/mnt_namespace.h>
20 #include <linux/pid_namespace.h>
21 #include <linux/user_namespace.h>
22 #include <net/net_namespace.h>
23 #include <asm/sections.h>
27 enum legacy_fs_param {
28 LEGACY_FS_UNSET_PARAMS,
29 LEGACY_FS_MONOLITHIC_PARAMS,
30 LEGACY_FS_INDIVIDUAL_PARAMS,
33 struct legacy_fs_context {
34 char *legacy_data; /* Data page for legacy filesystems */
36 enum legacy_fs_param param_type;
39 static int legacy_init_fs_context(struct fs_context *fc);
41 static const struct constant_table common_set_sb_flag[] = {
42 { "dirsync", SB_DIRSYNC },
43 { "lazytime", SB_LAZYTIME },
44 { "mand", SB_MANDLOCK },
46 { "sync", SB_SYNCHRONOUS },
50 static const struct constant_table common_clear_sb_flag[] = {
51 { "async", SB_SYNCHRONOUS },
52 { "nolazytime", SB_LAZYTIME },
53 { "nomand", SB_MANDLOCK },
59 * Check for a common mount option that manipulates s_flags.
61 static int vfs_parse_sb_flag(struct fs_context *fc, const char *key)
65 token = lookup_constant(common_set_sb_flag, key, 0);
67 fc->sb_flags |= token;
68 fc->sb_flags_mask |= token;
72 token = lookup_constant(common_clear_sb_flag, key, 0);
74 fc->sb_flags &= ~token;
75 fc->sb_flags_mask |= token;
83 * vfs_parse_fs_param_source - Handle setting "source" via parameter
84 * @fc: The filesystem context to modify
85 * @param: The parameter
87 * This is a simple helper for filesystems to verify that the "source" they
90 * Returns 0 on success, -ENOPARAM if this is not "source" parameter, and
91 * -EINVAL otherwise. In the event of failure, supplementary error information
94 int vfs_parse_fs_param_source(struct fs_context *fc, struct fs_parameter *param)
96 if (strcmp(param->key, "source") != 0)
99 if (param->type != fs_value_is_string)
100 return invalf(fc, "Non-string source");
103 return invalf(fc, "Multiple sources");
105 fc->source = param->string;
106 param->string = NULL;
109 EXPORT_SYMBOL(vfs_parse_fs_param_source);
112 * vfs_parse_fs_param - Add a single parameter to a superblock config
113 * @fc: The filesystem context to modify
114 * @param: The parameter
116 * A single mount option in string form is applied to the filesystem context
117 * being set up. Certain standard options (for example "ro") are translated
118 * into flag bits without going to the filesystem. The active security module
119 * is allowed to observe and poach options. Any other options are passed over
120 * to the filesystem to parse.
122 * This may be called multiple times for a context.
124 * Returns 0 on success and a negative error code on failure. In the event of
125 * failure, supplementary error information may have been set.
127 int vfs_parse_fs_param(struct fs_context *fc, struct fs_parameter *param)
132 return invalf(fc, "Unnamed parameter\n");
134 ret = vfs_parse_sb_flag(fc, param->key);
135 if (ret != -ENOPARAM)
138 ret = security_fs_context_parse_param(fc, param);
139 if (ret != -ENOPARAM)
140 /* Param belongs to the LSM or is disallowed by the LSM; so
141 * don't pass to the FS.
145 if (fc->ops->parse_param) {
146 ret = fc->ops->parse_param(fc, param);
147 if (ret != -ENOPARAM)
151 /* If the filesystem doesn't take any arguments, give it the
152 * default handling of source.
154 ret = vfs_parse_fs_param_source(fc, param);
155 if (ret != -ENOPARAM)
158 return invalf(fc, "%s: Unknown parameter '%s'",
159 fc->fs_type->name, param->key);
161 EXPORT_SYMBOL(vfs_parse_fs_param);
164 * vfs_parse_fs_string - Convenience function to just parse a string.
166 int vfs_parse_fs_string(struct fs_context *fc, const char *key,
167 const char *value, size_t v_size)
171 struct fs_parameter param = {
173 .type = fs_value_is_flag,
178 param.string = kmemdup_nul(value, v_size, GFP_KERNEL);
181 param.type = fs_value_is_string;
184 ret = vfs_parse_fs_param(fc, ¶m);
188 EXPORT_SYMBOL(vfs_parse_fs_string);
191 * generic_parse_monolithic - Parse key[=val][,key[=val]]* mount data
192 * @ctx: The superblock configuration to fill in.
193 * @data: The data to parse
195 * Parse a blob of data that's in key[=val][,key[=val]]* form. This can be
196 * called from the ->monolithic_mount_data() fs_context operation.
198 * Returns 0 on success or the error returned by the ->parse_option() fs_context
199 * operation on failure.
201 int generic_parse_monolithic(struct fs_context *fc, void *data)
203 char *options = data, *key;
209 ret = security_sb_eat_lsm_opts(options, &fc->security);
213 while ((key = strsep(&options, ",")) != NULL) {
216 char *value = strchr(key, '=');
222 v_len = strlen(value);
224 ret = vfs_parse_fs_string(fc, key, value, v_len);
232 EXPORT_SYMBOL(generic_parse_monolithic);
235 * alloc_fs_context - Create a filesystem context.
236 * @fs_type: The filesystem type.
237 * @reference: The dentry from which this one derives (or NULL)
238 * @sb_flags: Filesystem/superblock flags (SB_*)
239 * @sb_flags_mask: Applicable members of @sb_flags
240 * @purpose: The purpose that this configuration shall be used for.
242 * Open a filesystem and create a mount context. The mount context is
243 * initialised with the supplied flags and, if a submount/automount from
244 * another superblock (referred to by @reference) is supplied, may have
245 * parameters such as namespaces copied across from that superblock.
247 static struct fs_context *alloc_fs_context(struct file_system_type *fs_type,
248 struct dentry *reference,
249 unsigned int sb_flags,
250 unsigned int sb_flags_mask,
251 enum fs_context_purpose purpose)
253 int (*init_fs_context)(struct fs_context *);
254 struct fs_context *fc;
257 fc = kzalloc(sizeof(struct fs_context), GFP_KERNEL);
259 return ERR_PTR(-ENOMEM);
261 fc->purpose = purpose;
262 fc->sb_flags = sb_flags;
263 fc->sb_flags_mask = sb_flags_mask;
264 fc->fs_type = get_filesystem(fs_type);
265 fc->cred = get_current_cred();
266 fc->net_ns = get_net(current->nsproxy->net_ns);
267 fc->log.prefix = fs_type->name;
269 mutex_init(&fc->uapi_mutex);
272 case FS_CONTEXT_FOR_MOUNT:
273 fc->user_ns = get_user_ns(fc->cred->user_ns);
275 case FS_CONTEXT_FOR_SUBMOUNT:
276 fc->user_ns = get_user_ns(reference->d_sb->s_user_ns);
278 case FS_CONTEXT_FOR_RECONFIGURE:
279 atomic_inc(&reference->d_sb->s_active);
280 fc->user_ns = get_user_ns(reference->d_sb->s_user_ns);
281 fc->root = dget(reference);
285 /* TODO: Make all filesystems support this unconditionally */
286 init_fs_context = fc->fs_type->init_fs_context;
287 if (!init_fs_context)
288 init_fs_context = legacy_init_fs_context;
290 ret = init_fs_context(fc);
293 fc->need_free = true;
301 struct fs_context *fs_context_for_mount(struct file_system_type *fs_type,
302 unsigned int sb_flags)
304 return alloc_fs_context(fs_type, NULL, sb_flags, 0,
305 FS_CONTEXT_FOR_MOUNT);
307 EXPORT_SYMBOL(fs_context_for_mount);
309 struct fs_context *fs_context_for_reconfigure(struct dentry *dentry,
310 unsigned int sb_flags,
311 unsigned int sb_flags_mask)
313 return alloc_fs_context(dentry->d_sb->s_type, dentry, sb_flags,
314 sb_flags_mask, FS_CONTEXT_FOR_RECONFIGURE);
316 EXPORT_SYMBOL(fs_context_for_reconfigure);
318 struct fs_context *fs_context_for_submount(struct file_system_type *type,
319 struct dentry *reference)
321 return alloc_fs_context(type, reference, 0, 0, FS_CONTEXT_FOR_SUBMOUNT);
323 EXPORT_SYMBOL(fs_context_for_submount);
325 void fc_drop_locked(struct fs_context *fc)
327 struct super_block *sb = fc->root->d_sb;
330 deactivate_locked_super(sb);
333 static void legacy_fs_context_free(struct fs_context *fc);
336 * vfs_dup_fc_config: Duplicate a filesystem context.
337 * @src_fc: The context to copy.
339 struct fs_context *vfs_dup_fs_context(struct fs_context *src_fc)
341 struct fs_context *fc;
344 if (!src_fc->ops->dup)
345 return ERR_PTR(-EOPNOTSUPP);
347 fc = kmemdup(src_fc, sizeof(struct fs_context), GFP_KERNEL);
349 return ERR_PTR(-ENOMEM);
351 mutex_init(&fc->uapi_mutex);
353 fc->fs_private = NULL;
354 fc->s_fs_info = NULL;
357 get_filesystem(fc->fs_type);
359 get_user_ns(fc->user_ns);
362 refcount_inc(&fc->log.log->usage);
364 /* Can't call put until we've called ->dup */
365 ret = fc->ops->dup(fc, src_fc);
369 ret = security_fs_context_dup(fc, src_fc);
378 EXPORT_SYMBOL(vfs_dup_fs_context);
381 * logfc - Log a message to a filesystem context
382 * @fc: The filesystem context to log to.
383 * @fmt: The format of the buffer.
385 void logfc(struct fc_log *log, const char *prefix, char level, const char *fmt, ...)
388 struct va_format vaf = {.fmt = fmt, .va = &va};
394 printk(KERN_WARNING "%s%s%pV\n", prefix ? prefix : "",
395 prefix ? ": " : "", &vaf);
398 printk(KERN_ERR "%s%s%pV\n", prefix ? prefix : "",
399 prefix ? ": " : "", &vaf);
402 printk(KERN_NOTICE "%s%s%pV\n", prefix ? prefix : "",
403 prefix ? ": " : "", &vaf);
407 unsigned int logsize = ARRAY_SIZE(log->buffer);
409 char *q = kasprintf(GFP_KERNEL, "%c %s%s%pV\n", level,
410 prefix ? prefix : "",
411 prefix ? ": " : "", &vaf);
413 index = log->head & (logsize - 1);
414 BUILD_BUG_ON(sizeof(log->head) != sizeof(u8) ||
415 sizeof(log->tail) != sizeof(u8));
416 if ((u8)(log->head - log->tail) == logsize) {
417 /* The buffer is full, discard the oldest message */
418 if (log->need_free & (1 << index))
419 kfree(log->buffer[index]);
423 log->buffer[index] = q ? q : "OOM: Can't store error string";
425 log->need_free |= 1 << index;
427 log->need_free &= ~(1 << index);
432 EXPORT_SYMBOL(logfc);
435 * Free a logging structure.
437 static void put_fc_log(struct fs_context *fc)
439 struct fc_log *log = fc->log.log;
443 if (refcount_dec_and_test(&log->usage)) {
445 for (i = 0; i <= 7; i++)
446 if (log->need_free & (1 << i))
447 kfree(log->buffer[i]);
454 * put_fs_context - Dispose of a superblock configuration context.
455 * @fc: The context to dispose of.
457 void put_fs_context(struct fs_context *fc)
459 struct super_block *sb;
465 deactivate_super(sb);
468 if (fc->need_free && fc->ops && fc->ops->free)
471 security_free_mnt_opts(&fc->security);
473 put_user_ns(fc->user_ns);
476 put_filesystem(fc->fs_type);
480 EXPORT_SYMBOL(put_fs_context);
483 * Free the config for a filesystem that doesn't support fs_context.
485 static void legacy_fs_context_free(struct fs_context *fc)
487 struct legacy_fs_context *ctx = fc->fs_private;
490 if (ctx->param_type == LEGACY_FS_INDIVIDUAL_PARAMS)
491 kfree(ctx->legacy_data);
497 * Duplicate a legacy config.
499 static int legacy_fs_context_dup(struct fs_context *fc, struct fs_context *src_fc)
501 struct legacy_fs_context *ctx;
502 struct legacy_fs_context *src_ctx = src_fc->fs_private;
504 ctx = kmemdup(src_ctx, sizeof(*src_ctx), GFP_KERNEL);
508 if (ctx->param_type == LEGACY_FS_INDIVIDUAL_PARAMS) {
509 ctx->legacy_data = kmemdup(src_ctx->legacy_data,
510 src_ctx->data_size, GFP_KERNEL);
511 if (!ctx->legacy_data) {
517 fc->fs_private = ctx;
522 * Add a parameter to a legacy config. We build up a comma-separated list of
525 static int legacy_parse_param(struct fs_context *fc, struct fs_parameter *param)
527 struct legacy_fs_context *ctx = fc->fs_private;
528 unsigned int size = ctx->data_size;
532 ret = vfs_parse_fs_param_source(fc, param);
533 if (ret != -ENOPARAM)
536 if (ctx->param_type == LEGACY_FS_MONOLITHIC_PARAMS)
537 return invalf(fc, "VFS: Legacy: Can't mix monolithic and individual options");
539 switch (param->type) {
540 case fs_value_is_string:
541 len = 1 + param->size;
543 case fs_value_is_flag:
544 len += strlen(param->key);
547 return invalf(fc, "VFS: Legacy: Parameter type for '%s' not supported",
551 if (len > PAGE_SIZE - 2 - size)
552 return invalf(fc, "VFS: Legacy: Cumulative options too large");
553 if (strchr(param->key, ',') ||
554 (param->type == fs_value_is_string &&
555 memchr(param->string, ',', param->size)))
556 return invalf(fc, "VFS: Legacy: Option '%s' contained comma",
558 if (!ctx->legacy_data) {
559 ctx->legacy_data = kmalloc(PAGE_SIZE, GFP_KERNEL);
560 if (!ctx->legacy_data)
564 ctx->legacy_data[size++] = ',';
565 len = strlen(param->key);
566 memcpy(ctx->legacy_data + size, param->key, len);
568 if (param->type == fs_value_is_string) {
569 ctx->legacy_data[size++] = '=';
570 memcpy(ctx->legacy_data + size, param->string, param->size);
573 ctx->legacy_data[size] = '\0';
574 ctx->data_size = size;
575 ctx->param_type = LEGACY_FS_INDIVIDUAL_PARAMS;
580 * Add monolithic mount data.
582 static int legacy_parse_monolithic(struct fs_context *fc, void *data)
584 struct legacy_fs_context *ctx = fc->fs_private;
586 if (ctx->param_type != LEGACY_FS_UNSET_PARAMS) {
587 pr_warn("VFS: Can't mix monolithic and individual options\n");
591 ctx->legacy_data = data;
592 ctx->param_type = LEGACY_FS_MONOLITHIC_PARAMS;
593 if (!ctx->legacy_data)
596 if (fc->fs_type->fs_flags & FS_BINARY_MOUNTDATA)
598 return security_sb_eat_lsm_opts(ctx->legacy_data, &fc->security);
602 * Get a mountable root with the legacy mount command.
604 static int legacy_get_tree(struct fs_context *fc)
606 struct legacy_fs_context *ctx = fc->fs_private;
607 struct super_block *sb;
610 root = fc->fs_type->mount(fc->fs_type, fc->sb_flags,
611 fc->source, ctx->legacy_data);
613 return PTR_ERR(root);
625 static int legacy_reconfigure(struct fs_context *fc)
627 struct legacy_fs_context *ctx = fc->fs_private;
628 struct super_block *sb = fc->root->d_sb;
630 if (!sb->s_op->remount_fs)
633 return sb->s_op->remount_fs(sb, &fc->sb_flags,
634 ctx ? ctx->legacy_data : NULL);
637 const struct fs_context_operations legacy_fs_context_ops = {
638 .free = legacy_fs_context_free,
639 .dup = legacy_fs_context_dup,
640 .parse_param = legacy_parse_param,
641 .parse_monolithic = legacy_parse_monolithic,
642 .get_tree = legacy_get_tree,
643 .reconfigure = legacy_reconfigure,
647 * Initialise a legacy context for a filesystem that doesn't support
650 static int legacy_init_fs_context(struct fs_context *fc)
652 fc->fs_private = kzalloc(sizeof(struct legacy_fs_context), GFP_KERNEL);
655 fc->ops = &legacy_fs_context_ops;
659 int parse_monolithic_mount_data(struct fs_context *fc, void *data)
661 int (*monolithic_mount_data)(struct fs_context *, void *);
663 monolithic_mount_data = fc->ops->parse_monolithic;
664 if (!monolithic_mount_data)
665 monolithic_mount_data = generic_parse_monolithic;
667 return monolithic_mount_data(fc, data);
671 * Clean up a context after performing an action on it and put it into a state
672 * from where it can be used to reconfigure a superblock.
674 * Note that here we do only the parts that can't fail; the rest is in
675 * finish_clean_context() below and in between those fs_context is marked
676 * FS_CONTEXT_AWAITING_RECONF. The reason for splitup is that after
677 * successful mount or remount we need to report success to userland.
678 * Trying to do full reinit (for the sake of possible subsequent remount)
679 * and failing to allocate memory would've put us into a nasty situation.
680 * So here we only discard the old state and reinitialization is left
681 * until we actually try to reconfigure.
683 void vfs_clean_context(struct fs_context *fc)
685 if (fc->need_free && fc->ops && fc->ops->free)
687 fc->need_free = false;
688 fc->fs_private = NULL;
689 fc->s_fs_info = NULL;
691 security_free_mnt_opts(&fc->security);
695 fc->purpose = FS_CONTEXT_FOR_RECONFIGURE;
696 fc->phase = FS_CONTEXT_AWAITING_RECONF;
699 int finish_clean_context(struct fs_context *fc)
703 if (fc->phase != FS_CONTEXT_AWAITING_RECONF)
706 if (fc->fs_type->init_fs_context)
707 error = fc->fs_type->init_fs_context(fc);
709 error = legacy_init_fs_context(fc);
710 if (unlikely(error)) {
711 fc->phase = FS_CONTEXT_FAILED;
714 fc->need_free = true;
715 fc->phase = FS_CONTEXT_RECONF_PARAMS;