puffs_path(3) [netbsd man page]

PUFFS_PATH(3)						   BSD Library Functions Manual 					     PUFFS_PATH(3)

NAME

     puffs_path -- puffs pathbuilding routines

LIBRARY

     puffs Convenience Library (libpuffs, -lpuffs)

SYNOPSIS

     #include <puffs.h>

     int
     pu_pathbuild_fn(struct puffs_usermount *pu, const struct puffs_pathobj *po_dir, const struct puffs_pathobj *po_comp, size_t offset,
	 struct puffs_pathobj *po_new);

     int
     pu_pathtransform_fn(struct puffs_usermount *pu, const struct puffs_pathobj *po_base, const struct puffs_cn *pcn,
	 struct puffs_pathobj *po_new);

     int
     pu_pathcmp_fn(struct puffs_usermount *pu, struct puffs_pathobj *po1, struct puffs_pathobj *po2, size_t checklen, int checkprefix);

     void
     pu_pathfree_fn(struct puffs_usermount *pu, struct puffs_pathobj *po);

     int
     pu_namemod_fn(struct puffs_usermount *pu, struct puffs_pathobj *po_dir, struct puffs_cn *pcn);

     struct puffs_pathobj *
     puffs_getrootpathobj(struct puffs_usermount *pu);

DESCRIPTION

     The puffs library has the ability to provide full pathnames for backends which require them.  Normal file systems should be constructed with-
     out the file system node tied to a file name and should not used routines described herein.  An example of a file system where the backend
     requires filenames is mount_psshfs(8).

     The features described here are enabled by passing PUFFS_FLAG_BUILDPATH to puffs_init().  This facility requires to use puffs nodes to store
     the contents of the pathname.  Either the address of the operation cookie must directly be that of the puffs node, or puffs_set_cmap() must
     be used to set a mapping function from the cookie to the puffs node associated with the cookie.  Finally, the root node for the file system
     must be set using puffs_setroot() and the root path object retrieved using puffs_getrootpathobj() and initialized.

     There are two different places a filename can be retrieved from.  It is available for each puffs node after the node has been registered with
     the framework, i.e.  after the routine creating the node returns.	In other words, there is a window between the node is created and when the
     pathname is available and multithreaded file systems must take this into account.	The second place where a pathname is available is from the
     componentname struct puffs_pcn in operations which are passed one.  These can be retrieved using the convenience macros PNPATH() and
     PCNPATH() for node and componentname, respectively.  The type of object they return is void *.

     By default the framework manages "regular" filenames, which consist of directory names separated by "/" and a final component.  If the file
     system wishes to use pathnames of this format, all it has to do it enable the feature.  Everything else, including bookkeeping for node and
     directory renames, is done by the library.  The callback routines described next provide the ability to build non-standard pathnames.  A
     pu_foo_fn() callback is set using the puffs_set_foo() routine.

     This manual page is still unfinished.  Please take a number and wait in line.

SEE ALSO

     puffs(3), puffs_node(3), mount_psshfs(8), mount_sysctlfs(8)

BSD
								 December 27, 2007							       BSD

PUFFS_CRED(3)						   BSD Library Functions Manual 					     PUFFS_CRED(3)

NAME

     puffs_cred -- puffs credential and access control routines

LIBRARY

     puffs Convenience Library (libpuffs, -lpuffs)

SYNOPSIS

     #include <puffs.h>

     int
     puffs_cred_getuid(const struct puffs_cred *pcr, uid_t *uid);

     int
     puffs_cred_getgid(const struct puffs_cred *pcr, gid_t *gid);

     int
     puffs_cred_getgroups(const struct puffs_cred *pcr, gid_t *gids, short *ngids);

     bool
     puffs_cred_isuid(const struct puffs_cred *pcr, uid_t uid);

     bool
     puffs_cred_hasgroup(const struct puffs_cred *pcr, gid_t gid);

     bool
     puffs_cred_iskernel(const struct puffs_cred *pcr);

     bool
     puffs_cred_isfs(const struct puffs_cred *pcr);

     bool
     puffs_cred_isjuggernaut(const struct puffs_cred *pcr);

     int
     puffs_access(enum vtype type, mode_t file_mode, uid_t owner, gid_t group, mode_t access_mode, const struct puffs_cred *pcr);

     int
     puffs_access_chown(uid_t owner, gid_t group, uid_t newowner, gid_t newgroup, const struct puffs_cred *pcr);

     int
     puffs_access_chmod(uid_t owner, gid_t group, enum vtype type, mode_t newmode, const struct puffs_cred *pcr);

     int
     puffs_access_times(uid_t owner, gid_t group, mode_t file_mode, int va_utimes_null, const struct puffs_cred *pcr);

DESCRIPTION

     These functions can be used to check operation credentials and perform access control.  The structure struct puffs_cred can contain two types
     of credentials: ones belonging to users and ones belonging to the kernel.	The latter is further divided into generic kernel credentials and
     file system credentials.  The general rule is that these should be treated as more powerful than superuser credentials, but the file system
     is free to treat them as it sees fit.

   Credentials
     The puffs_cred_get() family of functions fetch the uid or gid(s) from the given credential cookie.  They return 0 for success or -1 for an
     error and set errno.  An error happens when the credentials represent kernel or file system credentials and do not contain an uid or gid(s).

     For puffs_cred_getgroups(), the argument gids should point to an array with room for *ngids elements.

     The puffs_cred_is() family of functions return 1 if the truth value of the function for pcr is true and 0 if it is false.	The function
     puffs_cred_isjuggernaut() is true if pcr describes superuser, kernel or file system credentials.

   Access Control
     To help the programmers task of emulating normal kernel file system access control semantics, several helper functions are provided to check
     if access is allowed.  They return 0 if access should be permitted or an errno value to indicate that access should be denied with the
     returned value.

     puffs_access() is equivalent to the kernel vaccess() function.  The arguments specify current information of the file to be tested with the
     exception of access_mode, which is a combination of PUFFS_VREAD, PUFFS_VWRITE and PUFFS_VEXEC indicating the desired file access mode.

     The rest of the functions provide UFS semantics for operations.  puffs_access_chown() checks if it is permissible to chown a file with the
     current uid and gid to the new uid and gid with the credentials of pcr.

     puffs_access_chmod() checks against permission to chmod a file of type type to the mode newmode.

     Finally, puffs_access_times() checks if it is allowable to update the timestamps of a file.  The argument va_utimes_null signals if the flags
     VA_UTIMES_NULL was set in va_vaflags of struct vattr.  If coming from a path where this information is unavailable, passing 0 as this argu-
     ment restricts the permission of modification to the owner and superuser.	Otherwise the function checks for write permissions to the node
     and returns success if the caller has write permissions.

SEE ALSO

     puffs(3), vnode(9)

BSD
								 October 18, 2007							       BSD