VNODEOPS(9) NetBSD Kernel Developer's Manual VNODEOPS(9)
NAME
vnodeops, VOP_LOOKUP, VOP_CREATE, VOP_MKNOD, VOP_OPEN, VOP_CLOSE, VOP_ACCESS, VOP_GETATTR, VOP_SETATTR, VOP_READ, VOP_WRITE, VOP_FALLOCATE, VOP_FDISCARD, VOP_IOCTL, VOP_FCNTL, VOP_POLL, VOP_KQFILTER, VOP_REVOKE, VOP_MMAP, VOP_FSYNC, VOP_SEEK, VOP_REMOVE, VOP_LINK, VOP_RENAME, VOP_MKDIR, VOP_RMDIR, VOP_SYMLINK, VOP_READDIR, VOP_READLINK, VOP_ABORTOP, VOP_INACTIVE, VOP_RECLAIM, VOP_LOCK, VOP_UNLOCK, VOP_ISLOCKED, VOP_BMAP, VOP_PRINT, VOP_PATHCONF, VOP_ADVLOCK, VOP_WHITEOUT, VOP_GETPAGES, VOP_PUTPAGES, VOP_STRATEGY, VOP_BWRITE, VOP_GETEXTATTR, VOP_SETEXTATTR, VOP_LISTEXTATTR, VOP_DELETEEXTATTR -- vnode operations
SYNOPSIS
#include <sys/param.h> #include <sys/buf.h> #include <sys/dirent.h> #include <sys/vnode.h> #include <sys/mount.h> #include <sys/namei.h> #include <sys/unistd.h> #include <sys/fcntl.h> #include <sys/lockf.h> #include <sys/extattr.h> int VOP_LOOKUP(struct vnode *dvp, struct vnode **vpp, struct componentname *cnp); int VOP_CREATE(struct vnode *dvp, struct vnode **vpp, struct componentname *cnp, struct vattr *vap); int VOP_MKNOD(struct vnode *dvp, struct vnode **vpp, struct componentname *cnp, struct vattr *vap); int VOP_OPEN(struct vnode *vp, int mode, kauth_cred_t cred); int VOP_CLOSE(struct vnode *vp, int fflag, kauth_cred_t cred); int VOP_ACCESS(struct vnode *vp, int mode, kauth_cred_t cred); int VOP_GETATTR(struct vnode *vp, struct vattr *vap, kauth_cred_t cred); int VOP_SETATTR(struct vnode *vp, struct vattr *vap, kauth_cred_t cred); int VOP_READ(struct vnode *vp, struct uio *uio, int ioflag, kauth_cred_t cred); int VOP_WRITE(struct vnode *vp, struct uio *uio, int ioflag, kauth_cred_t cred); int VOP_FALLOCATE(struct vnode *vp, off_t pos, off_t len); int VOP_FDISCARD(struct vnode *vp, off_t pos, off_t len); int VOP_IOCTL(struct vnode *vp, u_long command, void *data, int fflag, kauth_cred_t cred); int VOP_FCNTL(struct vnode *vp, u_int command, void *data, int fflag, kauth_cred_t cred); int VOP_POLL(struct vnode *vp, int events); int VOP_KQFILTER(struct vnode *vp, struct knote *kn); int VOP_REVOKE(struct vnode *vp, int flags); int VOP_MMAP(struct vnode *vp, vm_prot_t prot, kauth_cred_t cred); int VOP_FSYNC(struct vnode *vp, kauth_cred_t cred, int flags, off_t offlo, off_t offhi); int VOP_SEEK(struct vnode *vp, off_t oldoff, off_t newoff, kauth_cred_t cred); int VOP_REMOVE(struct vnode *dvp, struct vnode *vp, struct componentname *cnp); int VOP_LINK(struct vnode *dvp, struct vnode *vp, struct componentname *cnp); int VOP_RENAME(struct vnode *fdvp, struct vnode *fvp, struct componentname *fcnp, struct vnode *tdvp, struct vnode *tvp, struct componentname *tcnp); int VOP_MKDIR(struct vnode *dvp, struct vnode **vpp, struct componentname *cnp, struct vattr *vap); int VOP_RMDIR(struct vnode *dvp, struct vnode *vp, struct componentname *cnp); int VOP_SYMLINK(struct vnode *dvp, struct vnode **vpp, struct componentname *cnp, struct vattr *vap, char *target); int VOP_READDIR(struct vnode *vp, struct uio *uio, kauth_cred_t cred, int *eofflag, off_t **cookies, int *ncookies); int VOP_READLINK(struct vnode *vp, struct uio *uio, kauth_cred_t cred); int VOP_ABORTOP(struct vnode *dvp, struct componentname *cnp); int VOP_INACTIVE(struct vnode *vp); int VOP_RECLAIM(struct vnode *vp); int VOP_LOCK(struct vnode *vp, int flags); int VOP_UNLOCK(struct vnode *vp); int VOP_ISLOCKED(struct vnode *vp); int VOP_BMAP(struct vnode *vp, daddr_t bn, struct vnode **vpp, daddr_t *bnp, int *runp); int VOP_PRINT(struct vnode *vp); int VOP_PATHCONF(struct vnode *vp, int name, register_t *retval); int VOP_ADVLOCK(struct vnode *vp, void *id, int op, struct flock *fl, int flags); int VOP_WHITEOUT(struct vnode *dvp, struct componentname *cnp, int flags); int VOP_GETPAGES(struct vnode *vp, voff_t offset, struct vm_page **m, int *count, int centeridx, vm_prot_t access_type, int advice, int flags); int VOP_PUTPAGES(struct vnode *vp, voff_t offlo, voff_t offhi, int flags); int VOP_STRATEGY(struct vnode *vp, struct buf *bp); int VOP_BWRITE(struct vnode *vp, struct buf *bp); int VOP_GETEXTATTR(struct vnode *vp, int attrnamespace, const char *name, struct uio *uio, size_t *size, kauth_cred_t cred); int VOP_SETEXTATTR(struct vnode *vp, int attrnamespace, const char *name, struct uio *uio, kauth_cred_t cred); int VOP_LISTEXTATTR(struct vnode *vp, int attrnamespace, struct uio *uio, size_t *size, kauth_cred_t cred); int VOP_DELETEEXTATTR(struct vnode *vp, int attrnamespace, const char *name, kauth_cred_t cred); Not all header files are required for each function.
DESCRIPTION
The vnode operations vector describes what operations can be done to the file associated with the vnode. The system maintains one vnode opera- tions vector for each file system type configured into the kernel. The vnode operations vector contains a pointer to a function for each opera- tion supported by the file system. Many of the functions described in the vnode operations vector are closely related to their corresponding system calls. In most cases, they are called as a result of the system call associated with the operation being invoked. Functions in the vnode operations vector are invoked using specialized macros. The following table gives a summary of the operations. Macro Description VOP_LOOKUP Lookup file name in name cache VOP_CREATE Create a new file VOP_MKNOD Make a new device VOP_OPEN Open a file VOP_CLOSE Close a file VOP_ACCESS Determine file accessibility VOP_GETATTR Get file attributes VOP_SETATTR Set file attributes VOP_READ Read from a file VOP_WRITE Write to a file VOP_FALLOCATE Allocate backing for a file VOP_FDISCARD Discard backing for a file VOP_IOCTL Perform device-specific I/O VOP_FCNTL Perform file control VOP_POLL Test if poll event has occurred VOP_KQFILTER Register a knote VOP_REVOKE Eliminate vnode activity VOP_MMAP Map file into user address space VOP_FSYNC Flush pending data to disk VOP_SEEK Test if file is seekable VOP_REMOVE Remove a file VOP_LINK Link a file VOP_RENAME Rename a file VOP_MKDIR Make a new directory VOP_RMDIR Remove a directory VOP_SYMLINK Create a symbolic link VOP_READDIR Read directory entry VOP_READLINK Read contents of a symlink VOP_ABORTOP Abort pending operation VOP_INACTIVE Release the inactive vnode VOP_RECLAIM Reclaim vnode for another file VOP_LOCK Sleep until vnode lock is free VOP_UNLOCK Wake up process sleeping on lock VOP_ISLOCKED Test if vnode is locked VOP_BMAP Logical block number conversion VOP_PRINT Print debugging information VOP_PATHCONF Return POSIX pathconf data VOP_ADVLOCK Advisory record locking VOP_WHITEOUT Whiteout vnode VOP_GETPAGES Read VM pages from file VOP_PUTPAGES Write VM pages to file VOP_STRATEGY Read/write a file system buffer VOP_BWRITE Write a file system buffer VOP_GETEXTATTR Get extended attribute VOP_SETEXTATTR Set extended attribute VOP_LISTEXTATTR List extended attributes VOP_DELETEEXTATTR Remove extended attribute The implementation details of the vnode operations vector are not quite what is described here. If the file system type does not support a specific operation, it must nevertheless assign an appropriate stub in the vnode operations vector to do the minimum required of it. In most cases, such functions either do nothing or return an error value to the effect that it is not supported. Many of the functions in the vnode operations vector take a componentname structure. It is used to encapsulate many parameters into a single func- tion argument. It has the following structure: struct componentname { /* * Arguments to lookup. */ uint32_t cn_nameiop; /* namei operation */ uint32_t cn_flags; /* flags to namei */ kauth_cred_t cn_cred; /* credentials */ /* * Shared between lookup and commit routines. */ const char *cn_nameptr; /* pointer to looked up name */ size_t cn_namelen; /* length of looked up component */ size_t cn_consume; /* chars to consume in lookup() */ }; The top half of the structure is used exclusively for the pathname lookups using VOP_LOOKUP() and is initialized by the caller. The seman- tics of the lookup are affected by the lookup operation specified in cn_nameiop and the flags specified in cn_flags. Valid operations are: LOOKUP perform name lookup only CREATE set up for file creation DELETE set up for file deletion RENAME set up for file renaming OPMASK mask for operation Valid values for cn->cn_flags are: LOCKLEAF lock inode on return LOCKPARENT want parent vnode returned locked NOCACHE name must not be left in name cache (see namecache(9)) FOLLOW follow symbolic links NOFOLLOW do not follow symbolic links (pseudo) MODMASK mask of operational modifiers No vnode operations may be called from interrupt context. Most opera- tions also require the vnode to be locked on entry. To prevent dead- locks, when acquiring locks on multiple vnodes, the lock of parent direc- tory must be acquired before the lock on the child directory. Vnode operations for a file system type generally should not be called directly from the kernel, but accessed indirectly through the high-level convenience functions discussed in vnsubr(9).
FUNCTIONS
VOP_LOOKUP(dvp, vpp, cnp) Lookup a single pathname component in a given directory. The argument dvp is the locked vnode of the directory to search and cnp is the pathname component to be searched for. If the path- name component is found, the address of the resulting unlocked vnode is returned in vpp. The operation specified in cnp->cn_nameiop indicates VOP_LOOKUP() the reason for requesting the lookup and uses it to cache file system type specific infor- mation in the vnode for subsequent operations. There are three types of lookups: ".", ".." (ISDOTDOT), and reg- ular. If the pathname component being searched for is ".", then dvp has an extra reference added to it and it is returned in *vpp. For other pathname components, VOP_LOOKUP() checks the accessibility of the directory and searches the name cache for the pathname component. See namecache(9). If the pathname is not found in the name cache, the directory is searched for the pathname. The resulting unlocked vnode is returned in vpp. dvp is always returned locked. On failure *vpp is NULL, and *dvp is left locked. If the opera- tion is successful *vpp is unlocked and zero is returned. Typi- cally, if *vpp and dvp are the same vnode the caller will need to release twice (decrement the reference count) and unlock once. VOP_CREATE(dvp, vpp, cnp, vap) Create a new file in a given directory. The argument dvp is the locked vnode of the directory to create the new file in and cnp is the pathname component of the new file. The argument vap specifies the attributes that the new file should be created with. If the file is successfully created, the address of the resulting unlocked vnode is returned in vpp and zero is returned. This function is called after VOP_LOOKUP() when a file is being created. Normally, VOP_LOOKUP() will have set the SAVENAME flag in cnp->cn_flags to keep the memory pointed to by cnp->cn_pnbuf valid. If an error is detected when creating the file, this memory is released. If the file is created successfully it will be released unless the SAVESTART flags in specified in cnp->cn_flags. VOP_MKNOD(dvp, vpp, cnp, vap) Make a new device-special file in a given directory. The argu- ment dvp is the locked vnode of the directory to create the new device-special file in and cnp is the pathname component of the new device-special file. The argument vap specifies the attributes that the new device-special file should be created with. If the file is successfully created, the address of the resulting unlocked vnode is returned in vpp and zero is returned. This function is called after VOP_LOOKUP() when a device-special file is being created. Normally, VOP_LOOKUP() will have set the SAVENAME flag in cnp->cn_flags to keep the memory pointed to by cnp->cn_pnbuf valid. If an error is detected when creating the device-special file, this memory is released. If the device- special file is created successfully it will be released unless the SAVESTART flags in specified in cnp->cn_flags. VOP_OPEN(vp, mode, cred) Open a file. The argument vp is the vnode of the file to open and mode specifies the access mode required by the calling process. The calling credentials are specified by cred. The access mode is a set of flags, including FREAD, FWRITE, O_NON- BLOCK, O_APPEND, etc. VOP_OPEN() must be called before a file can be accessed by a thread. The vnode reference count is incremented. VOP_OPEN() expects the vnode vp to be locked on entry and will leave it locked on return. If the operation is successful zero is returned, otherwise an appropriate error code is returned. VOP_CLOSE(vp, fflag, cred) Close a file. The argument vp is the vnode of the file to close and fflag specifies the access mode by the calling process. The possible flags are FREAD, FWRITE and FNONBLOCK. The calling credentials are specified by cred. VOP_CLOSE() frees resources allocated by VOP_OPEN(). The vnode vp will be locked on entry and should remain locked on return. VOP_ACCESS(vp, mode, cred) Determine the accessibility (permissions) of the file against the specified credentials. The argument vp is the vnode of the file to check, mode is the type of access required and cred con- tains the user credentials to check. The argument mode is a mask which can contain VREAD, VWRITE or VEXEC. If the file is accessible in the specified way, zero is returned, otherwise an appropriate error code is returned. The vnode vp will be locked on entry and should remain locked on return. VOP_GETATTR(vp, vap, cred) Get specific vnode attributes on a file. The argument vp is the vnode of the file to get the attributes for. The argument cred specifies the calling credentials. VOP_GETATTR() uses the file system type specific data object vp->v_data to reference the underlying file attributes. Attributes associated with the file are collected by setting the required attribute bits in vap->va_mask. The attributes are returned in vap. Attributes which are not available are set to the value VNOVAL. For more information on vnode attributes see vattr(9). Histori- cally it was considered acceptable to call VOP_GETATTR() without first locking the vnode. This usage is deprecated. The vnode vp will be locked on entry and should remain locked on return. VOP_SETATTR(vp, vap, cred) Set specific vnode attributes on a file. The argument vp is the locked vnode of the file to set the attributes for. The argu- ment cred specifies the calling credentials. VOP_SETATTR() uses the file system type specific data object vp->v_data to refer- ence the underlying file attributes. The new attributes are defined in vap. Attributes associated with the file are set by setting the required attribute bits in vap->va_mask. Attributes which are not being modified by VOP_SETATTR() should be set to the value VNOVAL. If the operation is successful zero is returned, otherwise an appropriate error is returned. For more information on vnode attributes see vattr(9). VOP_READ(vp, uio, ioflag, cred) Read the contents of a file. The argument vp is the vnode of the file to read from, uio is the location to read the data into, ioflag is a set of flags and cred are the credentials of the calling process. The ioflag argument is used to give directives and hints to the file system. When attempting a read, the high 16 bits are used to provide a read-ahead hint (in unit of file system blocks) that the file system should attempt. The low 16 bits are a bit mask which can contain the following flags: IO_UNIT do I/O as atomic unit IO_APPEND append write to end IO_SYNC sync I/O file integrity completion IO_NODELOCKED underlying node already locked IO_NDELAY FNDELAY flag set in file table IO_DSYNC sync I/O data integrity completion IO_ALTSEMANTICS use alternate I/O semantics IO_NORMAL operate on regular data IO_EXT operate on extended attributes IO_DIRECT do not buffer data in the kernel Zero is returned on success, otherwise an error is returned. The vnode should be locked on entry and remains locked on exit. VOP_WRITE(vp, uio, ioflag, cred) Write to a file. The argument vp is the vnode of the file to write to, uio is the location of the data to write, ioflag is a set of flags and cred are the credentials of the calling process. The ioflag argument is used to give directives and hints to the file system. The low 16 bits are a bit mask which can contain the same flags as VOP_READ(). Zero is returned on success, otherwise an error is returned. The vnode should be locked on entry and remains locked on exit. VOP_FALLOCATE(vp, pos, len) Allocate backing store. The argument vp is the vnode for the file. The pos and len arguments (specified in bytes) name an extent within the file. The blocks underlying this range, rounding up at the top and down at the bottom if needed, are checked; if no physical storage is allocated, a physical block is allocated and zeroed. This operation removes ``holes'' from files. VOP_FDISCARD(vp, pos, len) Discard backing store. The argument vp is the vnode for the file. The pos and len arguments (specified in bytes) name an extent within the file. The blocks underlying this range, rounding down at the top and up at the bottom if needed, are checked. If any physical storage is used, it is deallocated. This operation creates ``holes'' in files. Discarded blocks of regular files read back afterwards as zeroes. On devices, the underlying discard-block operation if any (e.g. ATA TRIM) is issued. The device handles this as it sees fit. In particular it is not guaranteed that discarded blocks on devices will be zeroed; reading a discarded block might produce zeros, or ones, or the previously existing data, or some other data, or trash. VOP_IOCTL(vp, command, data, fflag, cred) Perform device-specific I/O. The argument vp is the locked vnode of the file, normally representing a device. The argument command specifies the device-specific operation to perform and cnp provides extra data for the specified operation. The argu- ment fflags is a set of flags. The argument cred is the caller's credentials. If the operation is successful, zero is returned, otherwise an appropriate error code is returned. Most file systems do not supply a function for VOP_IOCTL(). This function implements the ioctl(2) system call. VOP_FCNTL(vp, command, data, fflag, cred) Perform file control. The argument vp is the locked vnode of the file. The argument command specifies the operation to per- form and cnp provides extra data for the specified operation. The argument fflags is a set of flags. The argument cred is the caller's credentials. If the operation is successful, zero is returned, otherwise an appropriate error code is returned. VOP_POLL(vp, events) Test if a poll event has occurred. The argument vp is the vnode of the file to poll. It returns any events of interest as spec- ified by events that may have occurred for the file. The argu- ment events is a set of flags as specified by poll(2). The vnode vp remains unlocked throughout the whole operation. VOP_KQFILTER(vp, kn) Register a knote kn with the vnode vn. If the operation is suc- cessful zero is returned, otherwise an appropriate error code is returned. The vnode vp remains unlocked throughout the whole operation. VOP_REVOKE(vp, flags) Eliminate all activity associated with the vnode vp. The argu- ment flags is a set of flags. If REVOKEALL is set in flags all vnodes aliased to the vnode vp are also eliminated. If the operation is successful zero is returned, otherwise an appropri- ate error is returned. The vnode vp remains unlocked throughout the whole operation. VOP_MMAP(vp, prot, cred) Inform file system that vp is in the process of being memory mapped. The argument prot specifies the vm access protection the vnode is going to be mapped with. The argument cred is the caller's credentials. If the file system allows the memory map- ping, zero is returned, otherwise an appropriate error code is returned. Most file systems do not supply a function for VOP_MMAP() and use genfs_mmap() to default for success. Only file systems which do not integrate with the page cache at all typically want to disallow memory mapping. VOP_FSYNC(vp, cred, flags, offlo, offhi) Flush pending data buffers for a file to disk. The argument vp is the locked vnode of the file for flush. The argument cred is the caller's credentials. The argument flags is a set of flags. If FSYNC_WAIT is specified in flags, the function should wait for I/O to complete before returning. The argument offlo and offhi specify the range of file to flush. If the operation is successful zero is returned, otherwise an appropriate error code is returned. This function implements the sync(2) and fsync(2) system calls. VOP_SEEK(vp, oldoff, newoff, cred) Test if the file is seekable for the specified offset newoff. The argument vp is the locked vnode of the file to test. For most file systems this function simply tests if newoff is valid. If the specified newoff is less than zero, the function returns error code EINVAL. VOP_REMOVE(dvp, vp, cnp) Remove a file. The argument dvp is the locked vnode of the directory to remove the file from and vp is the locked vnode of the file to remove. The argument cnp is the pathname component about the file to remove. If the operation is successful zero is returned, otherwise an appropriate error code is returned. Both dvp and vp are locked on entry and are to be unlocked before returning. VOP_LINK(dvp, vp, cnp) Link to a file. The argument dvp is the locked node of the directory to create the new link and vp is the vnode of the file to be linked. The argument cnp is the pathname component of the new link. If the operation is successful zero is returned, oth- erwise an error code is returned. The directory vnode dvp should be locked on entry and will be released and unlocked on return. The vnode vp should not be locked on entry and will remain unlocked on return. VOP_RENAME(fdvp, fvp, fcnp, tdvp, tvp, tcnp) Rename a file. The argument fdvp is the vnode of the old parent directory containing in the file to be renamed and fvp is the vnode of the file to be renamed. The argument fcnp is the path- name component about the file to be renamed. The argument tdvp is the vnode of the new directory of the target file and tvp is the vnode of the target file (if it exists). The argument tcnp is the pathname component about the file's new name. If the operation is successful zero is returned, otherwise an error code is returned. The caller must hold the target file system's rename lock. The source directory and file vnodes should be unlocked and their reference counts should be incremented before entry. The target directory and file vnodes should both be locked on entry. VOP_RENAME() updates the reference counts prior to returning. Because of the complexity and nastiness of the interface, please do not write new code that calls VOP_RENAME() directly until such time as ongoing cleanup work reaches a point where the interface has been rendered halfway sane. VOP_MKDIR(dvp, vpp, cnp, vap) Make a new directory in a given directory. The argument dvp is the locked vnode of the directory to create the new directory in and cnp is the pathname component of the new directory. The argument vap specifies the attributes that the new directory should be created with. If the file is successfully created, the address of the resulting unlocked vnode is returned in vpp and zero is returned. This function is called after VOP_LOOKUP() when a directory is being created. Normally, VOP_LOOKUP() will have set the SAVE- NAME flag in cnp->cn_flags to keep the memory pointed to by cnp->cn_pnbuf valid. If an error is detected when creating the directory, this memory is released. If the directory is created successfully it will be released unless the SAVESTART flags in specified in cnp->cn_flags. VOP_RMDIR(dvp, vp, cnp) Remove a directory in a given directory. The argument dvp is the locked vnode of the directory to remove the directory from and vp is the locked vnode of the directory to remove. The argument cnp is the pathname component of the directory. Zero is returned on success, otherwise an error code is returned. Both dvp and vp should be locked on entry and will be released and unlocked on return. VOP_SYMLINK(dvp, vpp, cnp, vap, target) Create a symbolic link in a given directory. The argument dvp is the locked vnode of the directory to create the symbolic link in and cnp is the pathname component of the symbolic link. The argument vap specifies the attributes that the symbolic link should be created with and target specifies the pathname of the target of the symbolic link. If the symbolic link is success- fully created, the address of the resulting unlocked vnode is returned in vpp and zero is returned. This function is called after VOP_LOOKUP() when a symbolic link is being created. Normally, VOP_LOOKUP() will have set the SAVENAME flag in cnp->cn_flags to keep the memory pointed to by cnp->cn_pnbuf valid. If an error is detected when creating the symbolic link, this memory is released. If the symbolic link is created successfully it will be released unless the SAVESTART flags in specified in cnp->cn_flags. VOP_READDIR(vp, uio, cred, eofflag, cookies, ncookies) Read directory entry. The argument vp is the vnode of the directory to read the contents of and uio is the destination location to read the contents into. The argument cred is the caller's credentials. The argument eofflag is the pointer to a flag which is set by VOP_READDIR() to indicate an end-of-file condition. If eofflag is NULL, the end-of-file condition is not returned. The arguments cookies and ncookies specify the addresses for the list and number of directory seek cookies gen- erated for NFS. Both cookies and ncookies should be NULL if they aren't required to be returned by VOP_READDIR(). The directory contents are read into struct dirent structures and uio->uio_offset is set to the offset of the next unread direc- tory entry. This offset may be used in a following invocation to continue a sequential read of the directory contents. If the operation is successful zero is returned, otherwise an appropri- ate error code is returned. The directory should be locked on entry and will remain locked on return. In case ncookies and cookies are supplied, one cookie should be returned per directory entry. The value of the cookie for each directory entry should be the offset within the directory where the on-disk version of the following directory entry starts. That is, for each directory entry i, the corresponding cookie should refer to the offset of directory entry i + 1. Note that the cookies array must be allocated by the callee using the M_TEMP malloc type as callers of VOP_READDIR() must be able to free the allocation. VOP_READLINK(vp, uio, cred) Read the contents of a symbolic link. The argument vp is the locked vnode of the symlink and uio is the destination location to read the contents into. The argument cred is the credentials of the caller. If the operation is successful zero is returned, otherwise an error code is returned. The vnode should be locked on entry and will remain locked on return. VOP_ABORTOP(dvp, cnp) Abort pending operation on vnode dvp and free resources allo- cated in cnp. This operation is rarely implemented in file systems and genfs_abortop() is typically used instead. VOP_INACTIVE(vp) Release the inactive vnode. VOP_INACTIVE() is called when the kernel is no longer using the vnode. This may be because the reference count reaches zero or it may be that the file system is being forcibly unmounted while there are open files. It can be used to reclaim space for open but deleted files. The argu- ment vp is the locked vnode to be released. If the operation is successful zero is returned, otherwise an appropriate error code is returned. The vnode vp must be locked on entry, and will be unlocked on return. VOP_RECLAIM(vp) Reclaim the vnode for another file system. VOP_RECLAIM() is called when a vnode is being reused for a different file system. Any file system specific resources associated with the vnode should be freed. The argument vp is the vnode to be reclaimed. If the operation is successful zero is returned, otherwise an appropriate error code is returned. The vnode vp should not be locked on entry, and will remain unlocked on return. VOP_LOCK(vp, flags) Sleep until vnode lock is free. The argument vp is the vnode of the file to be locked. The argument flags is LK_EXCLUSIVE to take the lock exclusively or LK_SHARED to take a shared lock. If flags contains LK_NOWAIT and the lock is busy, the operation will return immediately with an error code. If flags contains LK_RETRY this is a hint the caller wants the lock on dead vnodes too. If the operation is successful zero is returned, otherwise an appropriate error code is returned. VOP_LOCK() is used to serialize access to the file system such as to prevent two writes to the same file from happening at the same time. Kernel code should use vn_lock(9) to lock a vnode rather than calling VOP_LOCK() directly. VOP_UNLOCK(vp) Wake up process sleeping on lock. The argument vp is the vnode of the file to be unlocked. If the operation is successful zero is returned, otherwise an appropriate error code is returned. VOP_UNLOCK() is used to serialize access to the file system such as to prevent two writes to the same file from happening at the same time. VOP_ISLOCKED(vp) Test if the vnode vp is locked. Possible return values are LK_EXCLUSIVE, LK_SHARED or 0 for lock held exclusively by the calling thread, shared lock held by anyone or unlocked, respec- tively. This function must never be used to make locking decisions at run time: it is provided only for diagnostic purposes. VOP_BMAP(vp, bn, vpp, bnp, runp) Convert the logical block number bn of a file specified by vnode vp to its physical block number on the disk. The physical block is returned in bnp. In case the logical block is not allocated, -1 is used. If vpp is not NULL, the vnode of the device vnode for the file system is returned in the address specified by vpp. If runp is not NULL, the number of contiguous blocks starting from the next block after the queried block will be returned in runp. VOP_PRINT(vp) Print debugging information. The argument vp is the vnode to print. If the operation is successful zero is returned, other- wise an appropriate error code is returned. VOP_PATHCONF(vp, name, retval) Implement POSIX pathconf(2) and fpathconf(2) support. The argu- ment vp is the locked vnode to get information about. The argu- ment name specified the type of information to return. The information is returned in the address specified by retval. Valid values for name are: _PC_LINK_MAX return the maximum number of links to a file _PC_NAME_MAX return the maximum number of bytes in a file name _PC_PATH_MAX return the maximum number of bytes in a pathname _PC_PIPE_BUF return the maximum number of bytes which will be written atomically to a pipe _PC_CHOWN_RESTRICTED return 1 if appropriate privileges are required for the chown(2) system call, otherwise zero _PC_NO_TRUNC return 0 if file names longer than {NAME_MAX} are silently truncated If name is recognized, *retval is set to the specified value and zero is returned, otherwise an appropriate error is returned. VOP_ADVLOCK(vp, id, op, fl, flags) Manipulate Advisory record locks on a vnode. The argument vp is the vnode on which locks are manipulated. The argument id is the id token which is changing the lock and op is the fcntl(2) operation to perform. Valid values are: F_SETLK set lock F_GETLK get the first conflicted lock F_UNLCK clear lock The argument fl is a description of the lock. In the case of SEEK_CUR, The caller should add the current file offset to fl->l_start beforehand. VOP_ADVLOCK() treats SEEK_CUR as SEEK_SET. The argument flags is the set of flags. Valid values are: F_WAIT wait until lock is granted F_FLOCK use flock(2) semantics for lock F_POSIX use POSIX semantics for lock If the operation is successful zero is returned, otherwise an appropriate error is returned. VOP_WHITEOUT(dvp, cnp, flags) Whiteout pathname component in directory with vnode dvp. The argument cnp specifies the pathname component to whiteout. The vnode dvp should be locked on entry and will remain locked on return. VOP_GETPAGES(vp, offset, m, count, centeridx, access_type, advice, flags) Read VM pages from file. The argument vp is the locked vnode to read the VM pages from. The argument offset is offset in the file to start accessing and m is an array of VM pages. The argument count points a variable that specifies the number of pages to read. If the operation is successful zero is returned, otherwise an appropriate error code is returned. If PGO_LOCKED is specified in flags, VOP_GETPAGES() might return less pages than requested. In that case, the variable pointed to by count will be updated. This function is primarily used by the page-fault handing mecha- nism. VOP_PUTPAGES(vp, offlo, offhi, flags) Write modified (dirty) VM pages to file. The argument vp is the vnode to write the VM pages to. The vnode's vm object lock (v_uobj.vmobjlock) must be held by the caller and will be released upon return. The arguments offlo and offhi specify the range of VM pages to write. In case offhi is given as 0, all pages at and after the start offset offlo belonging the vnode vp will be written. The argument flags controls the behavior of the routine and takes the vm pager's flags (PGO_ -prefixed). If the operation is successful zero is returned, otherwise an appropriate error code is returned. The function is primarily used by the pageout handling mechanism and is commonly implemented indirectly by genfs_putpages() with the help of VOP_STRATEGY() and VOP_BMAP(). VOP_STRATEGY(vp, bp) Read/write a file system buffer. The argument vp is the vnode to read/write to. The argument bp is the buffer to be read or written. VOP_STRATEGY() will either read or write data to the file depending on the value of bp->b_flags. If the operation is successful zero is returned, otherwise an appropriate error code is returned. VOP_BWRITE(vp, bp) Write a file system buffer. The argument vp is the vnode to write to. The argument bp specifies the buffer to be written. If the operation is successful zero is returned, otherwise an appropriate error code is returned. VOP_GETEXTATTR(vp, attrnamespace, name, uio, size, cred) Get an extended attribute. The argument vp is the locked vnode of the file or directory from which to retrieve the attribute. The argument attrnamespace specifies the extended attribute namespace. The argument name is a nul-terminated character string naming the attribute to retrieve. The argument uio, if not NULL, specifies where the extended attribute value is to be written. The argument size, if not NULL, will contain the num- ber of bytes required to read all of the attribute data upon return. In most cases, uio will be NULL when size is not, and vice versa. The argument cred specifies the user credentials to use when authorizing the request. VOP_SETEXTATTR(vp, attrnamespace, name, uio, cred) Set an extended attribute. The argument vp is the locked vnode of the file or directory to which to store the attribute. The argument namespace specifies the extended attribute namespace. The argument name is a nul-terminated character string naming the attribute to store. The argument uio specifies the source of the extended attribute data. The argument cred specifies the user credentials to use when authorizing the request. VOP_LISTEXTATTR(vp, attrnamespace, uio, size, cred) Retrieve the list of extended attributes. The argument vp is the locked vnode of the file or directory whose attributes are to be listed. The argument attrnamespace specifies the extended attribute namespace. The argument uio, if not NULL, specifies where the extended attribute list is to be written. The argu- ment size, if not NULL, will contain the number of bytes required to read all of the attribute names upon return. In most cases, uio will be NULL when size is not, and vice versa. The argument cred specifies the user credentials to use when authorizing the request. VOP_DELETEEXTATTR(vp, attrnamespace, name, cred) Remove attribute name from file associated with vp. The argu- ment attrnamespace specifies the extended attribute namespace. If full removal is not supported, the file system should return EOPNOTSUPP to allow the caller to zero out the value with VOP_SETEXTATTR(). The vnode vp should be locked on entry and will remain locked on return.
FILES
src/sys/kern/vnode_if.src contains the list of vnode functions, their definitions and an exact locking protocol.
ERRORS
[EACCES] Access for the specified operation is denied. [EDQUOT] Quota exceeded. [EINVAL] attempt to read from an illegal offset in the direc- tory; unrecognized input [EIO] a read error occurred while reading the directory or reading the contents of a symbolic link [EJUSTRETURN] A CREATE or RENAME operation would be successful. [ENOATTR] The requested attribute is not defined for this vnode. [ENOENT] The component was not found in the directory. [ENOSPC] The file system is full. [ENOTDIR] The vnode does not represent a directory. [ENOTEMPTY] attempt to remove a directory which is not empty [EPERM] an attempt was made to change an immutable file [EROFS] the file system is read-only
SEE ALSO
extattr(9), intro(9), namei(9), vattr(9), vfs(9), vfsops(9), vnode(9)
HISTORY
The vnode operations vector, its functions and the corresponding macros appeared in 4.3BSD. NetBSD 7.1_STABLE July 21, 2014 NetBSD 7.1_STABLE
Powered by man-cgi (2024-08-26). Maintained for NetBSD by Kimmo Suominen. Based on man-cgi by Panagiotis Christias.