brintos

brintos / linux-shallow public Read only

0
0
Text · 26.9 KiB · 317934c Raw
819 lines · plain
1.. SPDX-License-Identifier: GPL-2.02 3====================4Filesystem Mount API5====================6 7.. CONTENTS8 9 (1) Overview.10 11 (2) The filesystem context.12 13 (3) The filesystem context operations.14 15 (4) Filesystem context security.16 17 (5) VFS filesystem context API.18 19 (6) Superblock creation helpers.20 21 (7) Parameter description.22 23 (8) Parameter helper functions.24 25 26Overview27========28 29The creation of new mounts is now to be done in a multistep process:30 31 (1) Create a filesystem context.32 33 (2) Parse the parameters and attach them to the context.  Parameters are34     expected to be passed individually from userspace, though legacy binary35     parameters can also be handled.36 37 (3) Validate and pre-process the context.38 39 (4) Get or create a superblock and mountable root.40 41 (5) Perform the mount.42 43 (6) Return an error message attached to the context.44 45 (7) Destroy the context.46 47To support this, the file_system_type struct gains two new fields::48 49	int (*init_fs_context)(struct fs_context *fc);50	const struct fs_parameter_description *parameters;51 52The first is invoked to set up the filesystem-specific parts of a filesystem53context, including the additional space, and the second points to the54parameter description for validation at registration time and querying by a55future system call.56 57Note that security initialisation is done *after* the filesystem is called so58that the namespaces may be adjusted first.59 60 61The Filesystem context62======================63 64The creation and reconfiguration of a superblock is governed by a filesystem65context.  This is represented by the fs_context structure::66 67	struct fs_context {68		const struct fs_context_operations *ops;69		struct file_system_type *fs_type;70		void			*fs_private;71		struct dentry		*root;72		struct user_namespace	*user_ns;73		struct net		*net_ns;74		const struct cred	*cred;75		char			*source;76		char			*subtype;77		void			*security;78		void			*s_fs_info;79		unsigned int		sb_flags;80		unsigned int		sb_flags_mask;81		unsigned int		s_iflags;82		enum fs_context_purpose	purpose:8;83		...84	};85 86The fs_context fields are as follows:87 88   * ::89 90       const struct fs_context_operations *ops91 92     These are operations that can be done on a filesystem context (see93     below).  This must be set by the ->init_fs_context() file_system_type94     operation.95 96   * ::97 98       struct file_system_type *fs_type99 100     A pointer to the file_system_type of the filesystem that is being101     constructed or reconfigured.  This retains a reference on the type owner.102 103   * ::104 105       void *fs_private106 107     A pointer to the file system's private data.  This is where the filesystem108     will need to store any options it parses.109 110   * ::111 112       struct dentry *root113 114     A pointer to the root of the mountable tree (and indirectly, the115     superblock thereof).  This is filled in by the ->get_tree() op.  If this116     is set, an active reference on root->d_sb must also be held.117 118   * ::119 120       struct user_namespace *user_ns121       struct net *net_ns122 123     There are a subset of the namespaces in use by the invoking process.  They124     retain references on each namespace.  The subscribed namespaces may be125     replaced by the filesystem to reflect other sources, such as the parent126     mount superblock on an automount.127 128   * ::129 130       const struct cred *cred131 132     The mounter's credentials.  This retains a reference on the credentials.133 134   * ::135 136       char *source137 138     This specifies the source.  It may be a block device (e.g. /dev/sda1) or139     something more exotic, such as the "host:/path" that NFS desires.140 141   * ::142 143       char *subtype144 145     This is a string to be added to the type displayed in /proc/mounts to146     qualify it (used by FUSE).  This is available for the filesystem to set if147     desired.148 149   * ::150 151       void *security152 153     A place for the LSMs to hang their security data for the superblock.  The154     relevant security operations are described below.155 156   * ::157 158       void *s_fs_info159 160     The proposed s_fs_info for a new superblock, set in the superblock by161     sget_fc().  This can be used to distinguish superblocks.162 163   * ::164 165       unsigned int sb_flags166       unsigned int sb_flags_mask167 168     Which bits SB_* flags are to be set/cleared in super_block::s_flags.169 170   * ::171 172       unsigned int s_iflags173 174     These will be bitwise-OR'd with s->s_iflags when a superblock is created.175 176   * ::177 178       enum fs_context_purpose179 180     This indicates the purpose for which the context is intended.  The181     available values are:182 183	==========================	======================================184	FS_CONTEXT_FOR_MOUNT,		New superblock for explicit mount185	FS_CONTEXT_FOR_SUBMOUNT		New automatic submount of extant mount186	FS_CONTEXT_FOR_RECONFIGURE	Change an existing mount187	==========================	======================================188 189The mount context is created by calling vfs_new_fs_context() or190vfs_dup_fs_context() and is destroyed with put_fs_context().  Note that the191structure is not refcounted.192 193VFS, security and filesystem mount options are set individually with194vfs_parse_mount_option().  Options provided by the old mount(2) system call as195a page of data can be parsed with generic_parse_monolithic().196 197When mounting, the filesystem is allowed to take data from any of the pointers198and attach it to the superblock (or whatever), provided it clears the pointer199in the mount context.200 201The filesystem is also allowed to allocate resources and pin them with the202mount context.  For instance, NFS might pin the appropriate protocol version203module.204 205 206The Filesystem Context Operations207=================================208 209The filesystem context points to a table of operations::210 211	struct fs_context_operations {212		void (*free)(struct fs_context *fc);213		int (*dup)(struct fs_context *fc, struct fs_context *src_fc);214		int (*parse_param)(struct fs_context *fc,215				   struct fs_parameter *param);216		int (*parse_monolithic)(struct fs_context *fc, void *data);217		int (*get_tree)(struct fs_context *fc);218		int (*reconfigure)(struct fs_context *fc);219	};220 221These operations are invoked by the various stages of the mount procedure to222manage the filesystem context.  They are as follows:223 224   * ::225 226	void (*free)(struct fs_context *fc);227 228     Called to clean up the filesystem-specific part of the filesystem context229     when the context is destroyed.  It should be aware that parts of the230     context may have been removed and NULL'd out by ->get_tree().231 232   * ::233 234	int (*dup)(struct fs_context *fc, struct fs_context *src_fc);235 236     Called when a filesystem context has been duplicated to duplicate the237     filesystem-private data.  An error may be returned to indicate failure to238     do this.239 240     .. Warning::241 242         Note that even if this fails, put_fs_context() will be called243	 immediately thereafter, so ->dup() *must* make the244	 filesystem-private data safe for ->free().245 246   * ::247 248	int (*parse_param)(struct fs_context *fc,249			   struct fs_parameter *param);250 251     Called when a parameter is being added to the filesystem context.  param252     points to the key name and maybe a value object.  VFS-specific options253     will have been weeded out and fc->sb_flags updated in the context.254     Security options will also have been weeded out and fc->security updated.255 256     The parameter can be parsed with fs_parse() and fs_lookup_param().  Note257     that the source(s) are presented as parameters named "source".258 259     If successful, 0 should be returned or a negative error code otherwise.260 261   * ::262 263	int (*parse_monolithic)(struct fs_context *fc, void *data);264 265     Called when the mount(2) system call is invoked to pass the entire data266     page in one go.  If this is expected to be just a list of "key[=val]"267     items separated by commas, then this may be set to NULL.268 269     The return value is as for ->parse_param().270 271     If the filesystem (e.g. NFS) needs to examine the data first and then272     finds it's the standard key-val list then it may pass it off to273     generic_parse_monolithic().274 275   * ::276 277	int (*get_tree)(struct fs_context *fc);278 279     Called to get or create the mountable root and superblock, using the280     information stored in the filesystem context (reconfiguration goes via a281     different vector).  It may detach any resources it desires from the282     filesystem context and transfer them to the superblock it creates.283 284     On success it should set fc->root to the mountable root and return 0.  In285     the case of an error, it should return a negative error code.286 287     The phase on a userspace-driven context will be set to only allow this to288     be called once on any particular context.289 290   * ::291 292	int (*reconfigure)(struct fs_context *fc);293 294     Called to effect reconfiguration of a superblock using information stored295     in the filesystem context.  It may detach any resources it desires from296     the filesystem context and transfer them to the superblock.  The297     superblock can be found from fc->root->d_sb.298 299     On success it should return 0.  In the case of an error, it should return300     a negative error code.301 302     .. Note:: reconfigure is intended as a replacement for remount_fs.303 304 305Filesystem context Security306===========================307 308The filesystem context contains a security pointer that the LSMs can use for309building up a security context for the superblock to be mounted.  There are a310number of operations used by the new mount code for this purpose:311 312   * ::313 314	int security_fs_context_alloc(struct fs_context *fc,315				      struct dentry *reference);316 317     Called to initialise fc->security (which is preset to NULL) and allocate318     any resources needed.  It should return 0 on success or a negative error319     code on failure.320 321     reference will be non-NULL if the context is being created for superblock322     reconfiguration (FS_CONTEXT_FOR_RECONFIGURE) in which case it indicates323     the root dentry of the superblock to be reconfigured.  It will also be324     non-NULL in the case of a submount (FS_CONTEXT_FOR_SUBMOUNT) in which case325     it indicates the automount point.326 327   * ::328 329	int security_fs_context_dup(struct fs_context *fc,330				    struct fs_context *src_fc);331 332     Called to initialise fc->security (which is preset to NULL) and allocate333     any resources needed.  The original filesystem context is pointed to by334     src_fc and may be used for reference.  It should return 0 on success or a335     negative error code on failure.336 337   * ::338 339	void security_fs_context_free(struct fs_context *fc);340 341     Called to clean up anything attached to fc->security.  Note that the342     contents may have been transferred to a superblock and the pointer cleared343     during get_tree.344 345   * ::346 347	int security_fs_context_parse_param(struct fs_context *fc,348					    struct fs_parameter *param);349 350     Called for each mount parameter, including the source.  The arguments are351     as for the ->parse_param() method.  It should return 0 to indicate that352     the parameter should be passed on to the filesystem, 1 to indicate that353     the parameter should be discarded or an error to indicate that the354     parameter should be rejected.355 356     The value pointed to by param may be modified (if a string) or stolen357     (provided the value pointer is NULL'd out).  If it is stolen, 1 must be358     returned to prevent it being passed to the filesystem.359 360   * ::361 362	int security_fs_context_validate(struct fs_context *fc);363 364     Called after all the options have been parsed to validate the collection365     as a whole and to do any necessary allocation so that366     security_sb_get_tree() and security_sb_reconfigure() are less likely to367     fail.  It should return 0 or a negative error code.368 369     In the case of reconfiguration, the target superblock will be accessible370     via fc->root.371 372   * ::373 374	int security_sb_get_tree(struct fs_context *fc);375 376     Called during the mount procedure to verify that the specified superblock377     is allowed to be mounted and to transfer the security data there.  It378     should return 0 or a negative error code.379 380   * ::381 382	void security_sb_reconfigure(struct fs_context *fc);383 384     Called to apply any reconfiguration to an LSM's context.  It must not385     fail.  Error checking and resource allocation must be done in advance by386     the parameter parsing and validation hooks.387 388   * ::389 390	int security_sb_mountpoint(struct fs_context *fc,391			           struct path *mountpoint,392				   unsigned int mnt_flags);393 394     Called during the mount procedure to verify that the root dentry attached395     to the context is permitted to be attached to the specified mountpoint.396     It should return 0 on success or a negative error code on failure.397 398 399VFS Filesystem context API400==========================401 402There are four operations for creating a filesystem context and one for403destroying a context:404 405   * ::406 407       struct fs_context *fs_context_for_mount(struct file_system_type *fs_type,408					       unsigned int sb_flags);409 410     Allocate a filesystem context for the purpose of setting up a new mount,411     whether that be with a new superblock or sharing an existing one.  This412     sets the superblock flags, initialises the security and calls413     fs_type->init_fs_context() to initialise the filesystem private data.414 415     fs_type specifies the filesystem type that will manage the context and416     sb_flags presets the superblock flags stored therein.417 418   * ::419 420       struct fs_context *fs_context_for_reconfigure(421		struct dentry *dentry,422		unsigned int sb_flags,423		unsigned int sb_flags_mask);424 425     Allocate a filesystem context for the purpose of reconfiguring an426     existing superblock.  dentry provides a reference to the superblock to be427     configured.  sb_flags and sb_flags_mask indicate which superblock flags428     need changing and to what.429 430   * ::431 432       struct fs_context *fs_context_for_submount(433		struct file_system_type *fs_type,434		struct dentry *reference);435 436     Allocate a filesystem context for the purpose of creating a new mount for437     an automount point or other derived superblock.  fs_type specifies the438     filesystem type that will manage the context and the reference dentry439     supplies the parameters.  Namespaces are propagated from the reference440     dentry's superblock also.441 442     Note that it's not a requirement that the reference dentry be of the same443     filesystem type as fs_type.444 445   * ::446 447        struct fs_context *vfs_dup_fs_context(struct fs_context *src_fc);448 449     Duplicate a filesystem context, copying any options noted and duplicating450     or additionally referencing any resources held therein.  This is available451     for use where a filesystem has to get a mount within a mount, such as NFS4452     does by internally mounting the root of the target server and then doing a453     private pathwalk to the target directory.454 455     The purpose in the new context is inherited from the old one.456 457   * ::458 459       void put_fs_context(struct fs_context *fc);460 461     Destroy a filesystem context, releasing any resources it holds.  This462     calls the ->free() operation.  This is intended to be called by anyone who463     created a filesystem context.464 465     .. Warning::466 467        filesystem contexts are not refcounted, so this causes unconditional468	destruction.469 470In all the above operations, apart from the put op, the return is a mount471context pointer or a negative error code.472 473For the remaining operations, if an error occurs, a negative error code will be474returned.475 476   * ::477 478        int vfs_parse_fs_param(struct fs_context *fc,479			       struct fs_parameter *param);480 481     Supply a single mount parameter to the filesystem context.  This includes482     the specification of the source/device which is specified as the "source"483     parameter (which may be specified multiple times if the filesystem484     supports that).485 486     param specifies the parameter key name and the value.  The parameter is487     first checked to see if it corresponds to a standard mount flag (in which488     case it is used to set an SB_xxx flag and consumed) or a security option489     (in which case the LSM consumes it) before it is passed on to the490     filesystem.491 492     The parameter value is typed and can be one of:493 494	====================		=============================495	fs_value_is_flag		Parameter not given a value496	fs_value_is_string		Value is a string497	fs_value_is_blob		Value is a binary blob498	fs_value_is_filename		Value is a filename* + dirfd499	fs_value_is_file		Value is an open file (file*)500	====================		=============================501 502     If there is a value, that value is stored in a union in the struct in one503     of param->{string,blob,name,file}.  Note that the function may steal and504     clear the pointer, but then becomes responsible for disposing of the505     object.506 507   * ::508 509       int vfs_parse_fs_string(struct fs_context *fc, const char *key,510			       const char *value, size_t v_size);511 512     A wrapper around vfs_parse_fs_param() that copies the value string it is513     passed.514 515   * ::516 517       int generic_parse_monolithic(struct fs_context *fc, void *data);518 519     Parse a sys_mount() data page, assuming the form to be a text list520     consisting of key[=val] options separated by commas.  Each item in the521     list is passed to vfs_mount_option().  This is the default when the522     ->parse_monolithic() method is NULL.523 524   * ::525 526       int vfs_get_tree(struct fs_context *fc);527 528     Get or create the mountable root and superblock, using the parameters in529     the filesystem context to select/configure the superblock.  This invokes530     the ->get_tree() method.531 532   * ::533 534       struct vfsmount *vfs_create_mount(struct fs_context *fc);535 536     Create a mount given the parameters in the specified filesystem context.537     Note that this does not attach the mount to anything.538 539 540Superblock Creation Helpers541===========================542 543A number of VFS helpers are available for use by filesystems for the creation544or looking up of superblocks.545 546   * ::547 548       struct super_block *549       sget_fc(struct fs_context *fc,550	       int (*test)(struct super_block *sb, struct fs_context *fc),551	       int (*set)(struct super_block *sb, struct fs_context *fc));552 553     This is the core routine.  If test is non-NULL, it searches for an554     existing superblock matching the criteria held in the fs_context, using555     the test function to match them.  If no match is found, a new superblock556     is created and the set function is called to set it up.557 558     Prior to the set function being called, fc->s_fs_info will be transferred559     to sb->s_fs_info - and fc->s_fs_info will be cleared if set returns560     success (ie. 0).561 562The following helpers all wrap sget_fc():563 564	(1) vfs_get_single_super565 566	    Only one such superblock may exist in the system.  Any further567	    attempt to get a new superblock gets this one (and any parameter568	    differences are ignored).569 570	(2) vfs_get_keyed_super571 572	    Multiple superblocks of this type may exist and they're keyed on573	    their s_fs_info pointer (for example this may refer to a574	    namespace).575 576	(3) vfs_get_independent_super577 578	    Multiple independent superblocks of this type may exist.  This579	    function never matches an existing one and always creates a new580	    one.581 582 583Parameter Description584=====================585 586Parameters are described using structures defined in linux/fs_parser.h.587There's a core description struct that links everything together::588 589	struct fs_parameter_description {590		const struct fs_parameter_spec *specs;591		const struct fs_parameter_enum *enums;592	};593 594For example::595 596	enum {597		Opt_autocell,598		Opt_bar,599		Opt_dyn,600		Opt_foo,601		Opt_source,602	};603 604	static const struct fs_parameter_description afs_fs_parameters = {605		.specs		= afs_param_specs,606		.enums		= afs_param_enums,607	};608 609The members are as follows:610 611 (1) ::612 613       const struct fs_parameter_specification *specs;614 615     Table of parameter specifications, terminated with a null entry, where the616     entries are of type::617 618	struct fs_parameter_spec {619		const char		*name;620		u8			opt;621		enum fs_parameter_type	type:8;622		unsigned short		flags;623	};624 625     The 'name' field is a string to match exactly to the parameter key (no626     wildcards, patterns and no case-independence) and 'opt' is the value that627     will be returned by the fs_parser() function in the case of a successful628     match.629 630     The 'type' field indicates the desired value type and must be one of:631 632	=======================	=======================	=====================633	TYPE NAME		EXPECTED VALUE		RESULT IN634	=======================	=======================	=====================635	fs_param_is_flag	No value		n/a636	fs_param_is_bool	Boolean value		result->boolean637	fs_param_is_u32		32-bit unsigned int	result->uint_32638	fs_param_is_u32_octal	32-bit octal int	result->uint_32639	fs_param_is_u32_hex	32-bit hex int		result->uint_32640	fs_param_is_s32		32-bit signed int	result->int_32641	fs_param_is_u64		64-bit unsigned int	result->uint_64642	fs_param_is_enum	Enum value name 	result->uint_32643	fs_param_is_string	Arbitrary string	param->string644	fs_param_is_blob	Binary blob		param->blob645	fs_param_is_blockdev	Blockdev path		* Needs lookup646	fs_param_is_path	Path			* Needs lookup647	fs_param_is_fd		File descriptor		result->int_32648	fs_param_is_uid		User ID (u32)           result->uid649	fs_param_is_gid		Group ID (u32)          result->gid650	=======================	=======================	=====================651 652     Note that if the value is of fs_param_is_bool type, fs_parse() will try653     to match any string value against "0", "1", "no", "yes", "false", "true".654 655     Each parameter can also be qualified with 'flags':656 657	=======================	================================================658	fs_param_v_optional	The value is optional659	fs_param_neg_with_no	result->negated set if key is prefixed with "no"660	fs_param_neg_with_empty	result->negated set if value is ""661	fs_param_deprecated	The parameter is deprecated.662	=======================	================================================663 664     These are wrapped with a number of convenience wrappers:665 666	=======================	===============================================667	MACRO			SPECIFIES668	=======================	===============================================669	fsparam_flag()		fs_param_is_flag670	fsparam_flag_no()	fs_param_is_flag, fs_param_neg_with_no671	fsparam_bool()		fs_param_is_bool672	fsparam_u32()		fs_param_is_u32673	fsparam_u32oct()	fs_param_is_u32_octal674	fsparam_u32hex()	fs_param_is_u32_hex675	fsparam_s32()		fs_param_is_s32676	fsparam_u64()		fs_param_is_u64677	fsparam_enum()		fs_param_is_enum678	fsparam_string()	fs_param_is_string679	fsparam_blob()		fs_param_is_blob680	fsparam_bdev()		fs_param_is_blockdev681	fsparam_path()		fs_param_is_path682	fsparam_fd()		fs_param_is_fd683	fsparam_uid()		fs_param_is_uid684	fsparam_gid()		fs_param_is_gid685	=======================	===============================================686 687     all of which take two arguments, name string and option number - for688     example::689 690	static const struct fs_parameter_spec afs_param_specs[] = {691		fsparam_flag	("autocell",	Opt_autocell),692		fsparam_flag	("dyn",		Opt_dyn),693		fsparam_string	("source",	Opt_source),694		fsparam_flag_no	("foo",		Opt_foo),695		{}696	};697 698     An addition macro, __fsparam() is provided that takes an additional pair699     of arguments to specify the type and the flags for anything that doesn't700     match one of the above macros.701 702 (2) ::703 704       const struct fs_parameter_enum *enums;705 706     Table of enum value names to integer mappings, terminated with a null707     entry.  This is of type::708 709	struct fs_parameter_enum {710		u8		opt;711		char		name[14];712		u8		value;713	};714 715     Where the array is an unsorted list of { parameter ID, name }-keyed716     elements that indicate the value to map to, e.g.::717 718	static const struct fs_parameter_enum afs_param_enums[] = {719		{ Opt_bar,   "x",      1},720		{ Opt_bar,   "y",      23},721		{ Opt_bar,   "z",      42},722	};723 724     If a parameter of type fs_param_is_enum is encountered, fs_parse() will725     try to look the value up in the enum table and the result will be stored726     in the parse result.727 728The parser should be pointed to by the parser pointer in the file_system_type729struct as this will provide validation on registration (if730CONFIG_VALIDATE_FS_PARSER=y) and will allow the description to be queried from731userspace using the fsinfo() syscall.732 733 734Parameter Helper Functions735==========================736 737A number of helper functions are provided to help a filesystem or an LSM738process the parameters it is given.739 740   * ::741 742       int lookup_constant(const struct constant_table tbl[],743			   const char *name, int not_found);744 745     Look up a constant by name in a table of name -> integer mappings.  The746     table is an array of elements of the following type::747 748	struct constant_table {749		const char	*name;750		int		value;751	};752 753     If a match is found, the corresponding value is returned.  If a match754     isn't found, the not_found value is returned instead.755 756   * ::757 758       bool validate_constant_table(const struct constant_table *tbl,759				    size_t tbl_size,760				    int low, int high, int special);761 762     Validate a constant table.  Checks that all the elements are appropriately763     ordered, that there are no duplicates and that the values are between low764     and high inclusive, though provision is made for one allowable special765     value outside of that range.  If no special value is required, special766     should just be set to lie inside the low-to-high range.767 768     If all is good, true is returned.  If the table is invalid, errors are769     logged to the kernel log buffer and false is returned.770 771   * ::772 773       bool fs_validate_description(const struct fs_parameter_description *desc);774 775     This performs some validation checks on a parameter description.  It776     returns true if the description is good and false if it is not.  It will777     log errors to the kernel log buffer if validation fails.778 779   * ::780 781        int fs_parse(struct fs_context *fc,782		     const struct fs_parameter_description *desc,783		     struct fs_parameter *param,784		     struct fs_parse_result *result);785 786     This is the main interpreter of parameters.  It uses the parameter787     description to look up a parameter by key name and to convert that to an788     option number (which it returns).789 790     If successful, and if the parameter type indicates the result is a791     boolean, integer, enum, uid, or gid type, the value is converted by this792     function and the result stored in793     result->{boolean,int_32,uint_32,uint_64,uid,gid}.794 795     If a match isn't initially made, the key is prefixed with "no" and no796     value is present then an attempt will be made to look up the key with the797     prefix removed.  If this matches a parameter for which the type has flag798     fs_param_neg_with_no set, then a match will be made and result->negated799     will be set to true.800 801     If the parameter isn't matched, -ENOPARAM will be returned; if the802     parameter is matched, but the value is erroneous, -EINVAL will be803     returned; otherwise the parameter's option number will be returned.804 805   * ::806 807       int fs_lookup_param(struct fs_context *fc,808			   struct fs_parameter *value,809			   bool want_bdev,810			   unsigned int flags,811			   struct path *_path);812 813     This takes a parameter that carries a string or filename type and attempts814     to do a path lookup on it.  If the parameter expects a blockdev, a check815     is made that the inode actually represents one.816 817     Returns 0 if successful and ``*_path`` will be set; returns a negative818     error code if not.819