+ if (uap->arg) {
+ vnode_setnocache(vp);
+ } else {
+ vnode_clearnocache(vp);
+ }
+
+ (void)vnode_put(vp);
+ }
+ goto outdrop;
+
+ case F_CHECK_OPENEVT:
+ if (fp->f_type != DTYPE_VNODE) {
+ error = EBADF;
+ goto out;
+ }
+ vp = (struct vnode *)fp->f_data;
+ proc_fdunlock(p);
+
+ if ((error = vnode_getwithref(vp)) == 0) {
+ *retval = vnode_is_openevt(vp);
+
+ if (uap->arg) {
+ vnode_set_openevt(vp);
+ } else {
+ vnode_clear_openevt(vp);
+ }
+
+ (void)vnode_put(vp);
+ }
+ goto outdrop;
+
+ case F_RDADVISE: {
+ struct radvisory ra_struct;
+
+ if (fp->f_type != DTYPE_VNODE) {
+ error = EBADF;
+ goto out;
+ }
+ vp = (struct vnode *)fp->f_data;
+ proc_fdunlock(p);
+
+ if ((error = copyin(argp, (caddr_t)&ra_struct, sizeof(ra_struct)))) {
+ goto outdrop;
+ }
+ if ((error = vnode_getwithref(vp)) == 0) {
+ error = VNOP_IOCTL(vp, F_RDADVISE, (caddr_t)&ra_struct, 0, &context);
+
+ (void)vnode_put(vp);
+ }
+ goto outdrop;
+ }
+
+ case F_FLUSH_DATA:
+
+ if (fp->f_type != DTYPE_VNODE) {
+ error = EBADF;
+ goto out;
+ }
+ vp = (struct vnode *)fp->f_data;
+ proc_fdunlock(p);
+
+ if ((error = vnode_getwithref(vp)) == 0) {
+ error = VNOP_FSYNC(vp, MNT_NOWAIT, &context);
+
+ (void)vnode_put(vp);
+ }
+ goto outdrop;
+
+ case F_LOG2PHYS:
+ case F_LOG2PHYS_EXT: {
+ struct log2phys l2p_struct = {}; /* structure for allocate command */
+ int devBlockSize;
+
+ off_t file_offset = 0;
+ size_t a_size = 0;
+ size_t run = 0;
+
+ if (uap->cmd == F_LOG2PHYS_EXT) {
+ error = copyin(argp, (caddr_t)&l2p_struct, sizeof(l2p_struct));
+ if (error) {
+ goto out;
+ }
+ file_offset = l2p_struct.l2p_devoffset;
+ } else {
+ file_offset = fp->f_offset;
+ }
+ if (fp->f_type != DTYPE_VNODE) {
+ error = EBADF;
+ goto out;
+ }
+ vp = (struct vnode *)fp->f_data;
+ proc_fdunlock(p);
+ if ((error = vnode_getwithref(vp))) {
+ goto outdrop;
+ }
+ error = VNOP_OFFTOBLK(vp, file_offset, &lbn);
+ if (error) {
+ (void)vnode_put(vp);
+ goto outdrop;
+ }
+ error = VNOP_BLKTOOFF(vp, lbn, &offset);
+ if (error) {
+ (void)vnode_put(vp);
+ goto outdrop;
+ }
+ devBlockSize = vfs_devblocksize(vnode_mount(vp));
+ if (uap->cmd == F_LOG2PHYS_EXT) {
+ if (l2p_struct.l2p_contigbytes < 0) {
+ vnode_put(vp);
+ error = EINVAL;
+ goto outdrop;
+ }
+
+ a_size = MIN((uint64_t)l2p_struct.l2p_contigbytes, SIZE_MAX);
+ } else {
+ a_size = devBlockSize;
+ }
+
+ error = VNOP_BLOCKMAP(vp, offset, a_size, &bn, &run, NULL, 0, &context);
+
+ (void)vnode_put(vp);
+
+ if (!error) {
+ l2p_struct.l2p_flags = 0; /* for now */
+ if (uap->cmd == F_LOG2PHYS_EXT) {
+ l2p_struct.l2p_contigbytes = run - (file_offset - offset);
+ } else {
+ l2p_struct.l2p_contigbytes = 0; /* for now */
+ }
+
+ /*
+ * The block number being -1 suggests that the file offset is not backed
+ * by any real blocks on-disk. As a result, just let it be passed back up wholesale.
+ */
+ if (bn == -1) {
+ /* Don't multiply it by the block size */
+ l2p_struct.l2p_devoffset = bn;
+ } else {
+ l2p_struct.l2p_devoffset = bn * devBlockSize;
+ l2p_struct.l2p_devoffset += file_offset - offset;
+ }
+ error = copyout((caddr_t)&l2p_struct, argp, sizeof(l2p_struct));
+ }
+ goto outdrop;
+ }
+ case F_GETPATH:
+ case F_GETPATH_NOFIRMLINK: {
+ char *pathbufp;
+ int pathlen;
+
+ if (fp->f_type != DTYPE_VNODE) {
+ error = EBADF;
+ goto out;
+ }
+ vp = (struct vnode *)fp->f_data;
+ proc_fdunlock(p);
+
+ pathlen = MAXPATHLEN;
+ MALLOC(pathbufp, char *, pathlen, M_TEMP, M_WAITOK);
+ if (pathbufp == NULL) {
+ error = ENOMEM;
+ goto outdrop;
+ }
+ if ((error = vnode_getwithref(vp)) == 0) {
+ if (uap->cmd == F_GETPATH_NOFIRMLINK) {
+ error = vn_getpath_ext(vp, NULL, pathbufp, &pathlen, VN_GETPATH_NO_FIRMLINK);
+ } else {
+ error = vn_getpath(vp, pathbufp, &pathlen);
+ }
+ (void)vnode_put(vp);
+
+ if (error == 0) {
+ error = copyout((caddr_t)pathbufp, argp, pathlen);
+ }
+ }
+ FREE(pathbufp, M_TEMP);
+ goto outdrop;
+ }
+
+ case F_PATHPKG_CHECK: {
+ char *pathbufp;
+ size_t pathlen;
+
+ if (fp->f_type != DTYPE_VNODE) {
+ error = EBADF;
+ goto out;
+ }
+ vp = (struct vnode *)fp->f_data;
+ proc_fdunlock(p);
+
+ pathlen = MAXPATHLEN;
+ pathbufp = kalloc(MAXPATHLEN);
+
+ if ((error = copyinstr(argp, pathbufp, MAXPATHLEN, &pathlen)) == 0) {
+ if ((error = vnode_getwithref(vp)) == 0) {
+ AUDIT_ARG(text, pathbufp);
+ error = vn_path_package_check(vp, pathbufp, pathlen, retval);
+
+ (void)vnode_put(vp);
+ }
+ }
+ kfree(pathbufp, MAXPATHLEN);
+ goto outdrop;
+ }
+
+ case F_CHKCLEAN: // used by regression tests to see if all dirty pages got cleaned by fsync()
+ case F_FULLFSYNC: // fsync + flush the journal + DKIOCSYNCHRONIZE
+ case F_BARRIERFSYNC: // fsync + barrier
+ case F_FREEZE_FS: // freeze all other fs operations for the fs of this fd
+ case F_THAW_FS: { // thaw all frozen fs operations for the fs of this fd
+ if (fp->f_type != DTYPE_VNODE) {
+ error = EBADF;
+ goto out;
+ }
+ vp = (struct vnode *)fp->f_data;
+ proc_fdunlock(p);
+
+ if ((error = vnode_getwithref(vp)) == 0) {
+ error = VNOP_IOCTL(vp, uap->cmd, (caddr_t)NULL, 0, &context);
+
+ (void)vnode_put(vp);
+ }
+ break;
+ }
+
+ /*
+ * SPI (private) for opening a file starting from a dir fd
+ */
+ case F_OPENFROM: {
+ struct user_fopenfrom fopen;
+ struct vnode_attr va;
+ struct nameidata nd;
+ int cmode;
+
+ /* Check if this isn't a valid file descriptor */
+ if ((fp->f_type != DTYPE_VNODE) ||
+ (fp->f_flag & FREAD) == 0) {
+ error = EBADF;
+ goto out;
+ }
+ vp = (struct vnode *)fp->f_data;
+ proc_fdunlock(p);
+
+ if (vnode_getwithref(vp)) {
+ error = ENOENT;
+ goto outdrop;
+ }
+
+ /* Only valid for directories */
+ if (vp->v_type != VDIR) {
+ vnode_put(vp);
+ error = ENOTDIR;
+ goto outdrop;
+ }
+
+ /* Get flags, mode and pathname arguments. */
+ if (IS_64BIT_PROCESS(p)) {
+ error = copyin(argp, &fopen, sizeof(fopen));
+ } else {
+ struct user32_fopenfrom fopen32;
+
+ error = copyin(argp, &fopen32, sizeof(fopen32));
+ fopen.o_flags = fopen32.o_flags;
+ fopen.o_mode = fopen32.o_mode;
+ fopen.o_pathname = CAST_USER_ADDR_T(fopen32.o_pathname);
+ }
+ if (error) {
+ vnode_put(vp);
+ goto outdrop;
+ }
+ AUDIT_ARG(fflags, fopen.o_flags);
+ AUDIT_ARG(mode, fopen.o_mode);
+ VATTR_INIT(&va);
+ /* Mask off all but regular access permissions */
+ cmode = ((fopen.o_mode & ~fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT;
+ VATTR_SET(&va, va_mode, cmode & ACCESSPERMS);
+
+ /* Start the lookup relative to the file descriptor's vnode. */
+ NDINIT(&nd, LOOKUP, OP_OPEN, USEDVP | FOLLOW | AUDITVNPATH1, UIO_USERSPACE,
+ fopen.o_pathname, &context);
+ nd.ni_dvp = vp;
+
+ error = open1(&context, &nd, fopen.o_flags, &va,
+ fileproc_alloc_init, NULL, retval);
+
+ vnode_put(vp);
+ break;
+ }
+ /*
+ * SPI (private) for unlinking a file starting from a dir fd
+ */
+ case F_UNLINKFROM: {
+ user_addr_t pathname;
+
+ /* Check if this isn't a valid file descriptor */
+ if ((fp->f_type != DTYPE_VNODE) ||
+ (fp->f_flag & FREAD) == 0) {
+ error = EBADF;
+ goto out;
+ }
+ vp = (struct vnode *)fp->f_data;
+ proc_fdunlock(p);
+
+ if (vnode_getwithref(vp)) {
+ error = ENOENT;
+ goto outdrop;
+ }
+
+ /* Only valid for directories */
+ if (vp->v_type != VDIR) {
+ vnode_put(vp);
+ error = ENOTDIR;
+ goto outdrop;
+ }
+
+ /* Get flags, mode and pathname arguments. */
+ if (IS_64BIT_PROCESS(p)) {
+ pathname = (user_addr_t)argp;
+ } else {
+ pathname = CAST_USER_ADDR_T(argp);
+ }
+
+ /* Start the lookup relative to the file descriptor's vnode. */
+ error = unlink1(&context, vp, pathname, UIO_USERSPACE, 0);
+
+ vnode_put(vp);
+ break;
+ }
+
+ case F_ADDSIGS:
+ case F_ADDFILESIGS:
+ case F_ADDFILESIGS_FOR_DYLD_SIM:
+ case F_ADDFILESIGS_RETURN:
+ {
+ struct cs_blob *blob = NULL;
+ struct user_fsignatures fs;
+ kern_return_t kr;
+ vm_offset_t kernel_blob_addr;
+ vm_size_t kernel_blob_size;
+ int blob_add_flags = 0;
+
+ if (fp->f_type != DTYPE_VNODE) {
+ error = EBADF;
+ goto out;
+ }
+ vp = (struct vnode *)fp->f_data;
+ proc_fdunlock(p);
+
+ if (uap->cmd == F_ADDFILESIGS_FOR_DYLD_SIM) {
+ blob_add_flags |= MAC_VNODE_CHECK_DYLD_SIM;
+ if ((p->p_csflags & CS_KILL) == 0) {
+ proc_lock(p);
+ p->p_csflags |= CS_KILL;
+ proc_unlock(p);
+ }
+ }
+
+ error = vnode_getwithref(vp);
+ if (error) {
+ goto outdrop;
+ }
+
+ if (IS_64BIT_PROCESS(p)) {
+ error = copyin(argp, &fs, sizeof(fs));
+ } else {
+ struct user32_fsignatures fs32;
+
+ error = copyin(argp, &fs32, sizeof(fs32));
+ fs.fs_file_start = fs32.fs_file_start;
+ fs.fs_blob_start = CAST_USER_ADDR_T(fs32.fs_blob_start);
+ fs.fs_blob_size = fs32.fs_blob_size;
+ }
+
+ if (error) {
+ vnode_put(vp);
+ goto outdrop;
+ }
+
+ /*
+ * First check if we have something loaded a this offset
+ */
+ blob = ubc_cs_blob_get(vp, CPU_TYPE_ANY, fs.fs_file_start);
+ if (blob != NULL) {
+ /* If this is for dyld_sim revalidate the blob */
+ if (uap->cmd == F_ADDFILESIGS_FOR_DYLD_SIM) {
+ error = ubc_cs_blob_revalidate(vp, blob, NULL, blob_add_flags);
+ if (error) {
+ blob = NULL;
+ if (error != EAGAIN) {
+ vnode_put(vp);
+ goto outdrop;
+ }
+ }
+ }
+ }
+
+ if (blob == NULL) {
+ /*
+ * An arbitrary limit, to prevent someone from mapping in a 20GB blob. This should cover
+ * our use cases for the immediate future, but note that at the time of this commit, some
+ * platforms are nearing 2MB blob sizes (with a prior soft limit of 2.5MB).
+ *
+ * We should consider how we can manage this more effectively; the above means that some
+ * platforms are using megabytes of memory for signing data; it merely hasn't crossed the
+ * threshold considered ridiculous at the time of this change.
+ */
+#define CS_MAX_BLOB_SIZE (40ULL * 1024ULL * 1024ULL)
+ if (fs.fs_blob_size > CS_MAX_BLOB_SIZE) {
+ error = E2BIG;
+ vnode_put(vp);
+ goto outdrop;
+ }
+
+ kernel_blob_size = CAST_DOWN(vm_size_t, fs.fs_blob_size);
+ kr = ubc_cs_blob_allocate(&kernel_blob_addr, &kernel_blob_size);
+ if (kr != KERN_SUCCESS || kernel_blob_size < fs.fs_blob_size) {
+ error = ENOMEM;
+ vnode_put(vp);
+ goto outdrop;
+ }
+
+ if (uap->cmd == F_ADDSIGS) {
+ error = copyin(fs.fs_blob_start,
+ (void *) kernel_blob_addr,
+ fs.fs_blob_size);
+ } else { /* F_ADDFILESIGS || F_ADDFILESIGS_RETURN || F_ADDFILESIGS_FOR_DYLD_SIM */
+ int resid;
+
+ error = vn_rdwr(UIO_READ,
+ vp,
+ (caddr_t) kernel_blob_addr,
+ kernel_blob_size,
+ fs.fs_file_start + fs.fs_blob_start,
+ UIO_SYSSPACE,
+ 0,
+ kauth_cred_get(),
+ &resid,
+ p);
+ if ((error == 0) && resid) {
+ /* kernel_blob_size rounded to a page size, but signature may be at end of file */
+ memset((void *)(kernel_blob_addr + (kernel_blob_size - resid)), 0x0, resid);
+ }
+ }
+
+ if (error) {
+ ubc_cs_blob_deallocate(kernel_blob_addr,
+ kernel_blob_size);
+ vnode_put(vp);
+ goto outdrop;
+ }
+
+ blob = NULL;
+ error = ubc_cs_blob_add(vp,
+ CPU_TYPE_ANY, /* not for a specific architecture */
+ fs.fs_file_start,
+ &kernel_blob_addr,
+ kernel_blob_size,
+ NULL,
+ blob_add_flags,
+ &blob);
+
+ /* ubc_blob_add() has consumed "kernel_blob_addr" if it is zeroed */
+ if (error) {
+ if (kernel_blob_addr) {
+ ubc_cs_blob_deallocate(kernel_blob_addr,
+ kernel_blob_size);
+ }
+ vnode_put(vp);
+ goto outdrop;
+ } else {
+#if CHECK_CS_VALIDATION_BITMAP
+ ubc_cs_validation_bitmap_allocate( vp );
+#endif
+ }
+ }
+
+ if (uap->cmd == F_ADDFILESIGS_RETURN || uap->cmd == F_ADDFILESIGS_FOR_DYLD_SIM) {
+ /*
+ * The first element of the structure is a
+ * off_t that happen to have the same size for
+ * all archs. Lets overwrite that.
+ */
+ off_t end_offset = 0;
+ if (blob) {
+ end_offset = blob->csb_end_offset;
+ }
+ error = copyout(&end_offset, argp, sizeof(end_offset));
+ }
+
+ (void) vnode_put(vp);
+ break;
+ }
+ case F_GETCODEDIR:
+ case F_FINDSIGS: {
+ error = ENOTSUP;
+ goto out;
+ }
+ case F_CHECK_LV: {
+ struct fileglob *fg;
+ fchecklv_t lv = {};
+
+ if (fp->f_type != DTYPE_VNODE) {
+ error = EBADF;
+ goto out;
+ }
+ fg = fp->f_fglob;
+ proc_fdunlock(p);
+
+ if (IS_64BIT_PROCESS(p)) {
+ error = copyin(argp, &lv, sizeof(lv));
+ } else {
+ struct user32_fchecklv lv32 = {};
+
+ error = copyin(argp, &lv32, sizeof(lv32));
+ lv.lv_file_start = lv32.lv_file_start;
+ lv.lv_error_message = (void *)(uintptr_t)lv32.lv_error_message;
+ lv.lv_error_message_size = lv32.lv_error_message_size;
+ }
+ if (error) {
+ goto outdrop;
+ }
+
+#if CONFIG_MACF
+ error = mac_file_check_library_validation(p, fg, lv.lv_file_start,
+ (user_long_t)lv.lv_error_message, lv.lv_error_message_size);
+#endif
+
+ break;
+ }
+#if CONFIG_PROTECT
+ case F_GETPROTECTIONCLASS: {
+ if (fp->f_type != DTYPE_VNODE) {
+ error = EBADF;
+ goto out;
+ }
+ vp = (struct vnode *)fp->f_data;
+
+ proc_fdunlock(p);
+
+ if (vnode_getwithref(vp)) {
+ error = ENOENT;
+ goto outdrop;
+ }
+
+ struct vnode_attr va;
+
+ VATTR_INIT(&va);
+ VATTR_WANTED(&va, va_dataprotect_class);
+ error = VNOP_GETATTR(vp, &va, &context);
+ if (!error) {
+ if (VATTR_IS_SUPPORTED(&va, va_dataprotect_class)) {
+ *retval = va.va_dataprotect_class;
+ } else {
+ error = ENOTSUP;
+ }
+ }
+
+ vnode_put(vp);
+ break;
+ }
+
+ case F_SETPROTECTIONCLASS: {
+ /* tmp must be a valid PROTECTION_CLASS_* */
+ tmp = CAST_DOWN_EXPLICIT(uint32_t, uap->arg);
+
+ if (fp->f_type != DTYPE_VNODE) {
+ error = EBADF;
+ goto out;
+ }
+ vp = (struct vnode *)fp->f_data;
+
+ proc_fdunlock(p);
+
+ if (vnode_getwithref(vp)) {
+ error = ENOENT;
+ goto outdrop;
+ }
+
+ /* Only go forward if you have write access */
+ vfs_context_t ctx = vfs_context_current();
+ if (vnode_authorize(vp, NULLVP, (KAUTH_VNODE_ACCESS | KAUTH_VNODE_WRITE_DATA), ctx) != 0) {
+ vnode_put(vp);
+ error = EBADF;
+ goto outdrop;
+ }
+
+ struct vnode_attr va;
+
+ VATTR_INIT(&va);
+ VATTR_SET(&va, va_dataprotect_class, tmp);
+
+ error = VNOP_SETATTR(vp, &va, ctx);
+
+ vnode_put(vp);
+ break;
+ }
+
+ case F_TRANSCODEKEY: {
+ if (fp->f_type != DTYPE_VNODE) {
+ error = EBADF;
+ goto out;
+ }
+
+ vp = (struct vnode *)fp->f_data;
+ proc_fdunlock(p);
+
+ if (vnode_getwithref(vp)) {
+ error = ENOENT;
+ goto outdrop;
+ }
+
+ cp_key_t k = {
+ .len = CP_MAX_WRAPPEDKEYSIZE,
+ };
+
+ MALLOC(k.key, char *, k.len, M_TEMP, M_WAITOK | M_ZERO);
+
+ error = VNOP_IOCTL(vp, F_TRANSCODEKEY, (caddr_t)&k, 1, &context);
+
+ vnode_put(vp);
+
+ if (error == 0) {
+ error = copyout(k.key, argp, k.len);
+ *retval = k.len;
+ }
+
+ FREE(k.key, M_TEMP);
+
+ break;
+ }
+
+ case F_GETPROTECTIONLEVEL: {
+ if (fp->f_type != DTYPE_VNODE) {
+ error = EBADF;
+ goto out;
+ }
+
+ vp = (struct vnode*) fp->f_data;
+ proc_fdunlock(p);
+
+ if (vnode_getwithref(vp)) {
+ error = ENOENT;
+ goto outdrop;
+ }
+
+ error = VNOP_IOCTL(vp, F_GETPROTECTIONLEVEL, (caddr_t)retval, 0, &context);
+
+ vnode_put(vp);
+ break;
+ }
+
+ case F_GETDEFAULTPROTLEVEL: {
+ if (fp->f_type != DTYPE_VNODE) {
+ error = EBADF;
+ goto out;
+ }
+
+ vp = (struct vnode*) fp->f_data;
+ proc_fdunlock(p);
+
+ if (vnode_getwithref(vp)) {
+ error = ENOENT;
+ goto outdrop;
+ }
+
+ /*
+ * if cp_get_major_vers fails, error will be set to proper errno
+ * and cp_version will still be 0.
+ */
+
+ error = VNOP_IOCTL(vp, F_GETDEFAULTPROTLEVEL, (caddr_t)retval, 0, &context);
+
+ vnode_put(vp);
+ break;
+ }
+
+#endif /* CONFIG_PROTECT */
+
+ case F_MOVEDATAEXTENTS: {
+ struct fileproc *fp2 = NULL;
+ struct vnode *src_vp = NULLVP;
+ struct vnode *dst_vp = NULLVP;
+ /* We need to grab the 2nd FD out of the argments before moving on. */
+ int fd2 = CAST_DOWN_EXPLICIT(int32_t, uap->arg);
+
+ error = priv_check_cred(kauth_cred_get(), PRIV_VFS_MOVE_DATA_EXTENTS, 0);
+ if (error) {
+ goto out;
+ }
+
+ if (fp->f_type != DTYPE_VNODE) {
+ error = EBADF;
+ goto out;
+ }
+
+ /*
+ * For now, special case HFS+ and APFS only, since this
+ * is SPI.
+ */
+ src_vp = (struct vnode *)fp->f_data;
+ if (src_vp->v_tag != VT_HFS && src_vp->v_tag != VT_APFS) {
+ error = ENOTSUP;
+ goto out;
+ }
+
+ /*
+ * Get the references before we start acquiring iocounts on the vnodes,
+ * while we still hold the proc fd lock
+ */
+ if ((error = fp_lookup(p, fd2, &fp2, 1))) {
+ error = EBADF;
+ goto out;
+ }
+ if (fp2->f_type != DTYPE_VNODE) {
+ fp_drop(p, fd2, fp2, 1);
+ error = EBADF;
+ goto out;
+ }
+ dst_vp = (struct vnode *)fp2->f_data;
+ if (dst_vp->v_tag != VT_HFS && dst_vp->v_tag != VT_APFS) {
+ fp_drop(p, fd2, fp2, 1);
+ error = ENOTSUP;
+ goto out;
+ }
+
+#if CONFIG_MACF
+ /* Re-do MAC checks against the new FD, pass in a fake argument */
+ error = mac_file_check_fcntl(proc_ucred(p), fp2->f_fglob, uap->cmd, 0);
+ if (error) {
+ fp_drop(p, fd2, fp2, 1);
+ goto out;
+ }
+#endif
+ /* Audit the 2nd FD */
+ AUDIT_ARG(fd, fd2);
+
+ proc_fdunlock(p);
+
+ if (vnode_getwithref(src_vp)) {
+ fp_drop(p, fd2, fp2, 0);
+ error = ENOENT;
+ goto outdrop;
+ }
+ if (vnode_getwithref(dst_vp)) {
+ vnode_put(src_vp);
+ fp_drop(p, fd2, fp2, 0);
+ error = ENOENT;
+ goto outdrop;
+ }
+
+ /*
+ * Basic asserts; validate they are not the same and that
+ * both live on the same filesystem.
+ */
+ if (dst_vp == src_vp) {
+ vnode_put(src_vp);
+ vnode_put(dst_vp);
+ fp_drop(p, fd2, fp2, 0);
+ error = EINVAL;
+ goto outdrop;
+ }
+
+ if (dst_vp->v_mount != src_vp->v_mount) {
+ vnode_put(src_vp);
+ vnode_put(dst_vp);
+ fp_drop(p, fd2, fp2, 0);
+ error = EXDEV;
+ goto outdrop;
+ }
+
+ /* Now we have a legit pair of FDs. Go to work */
+
+ /* Now check for write access to the target files */
+ if (vnode_authorize(src_vp, NULLVP,
+ (KAUTH_VNODE_ACCESS | KAUTH_VNODE_WRITE_DATA), &context) != 0) {
+ vnode_put(src_vp);
+ vnode_put(dst_vp);
+ fp_drop(p, fd2, fp2, 0);
+ error = EBADF;
+ goto outdrop;
+ }
+
+ if (vnode_authorize(dst_vp, NULLVP,
+ (KAUTH_VNODE_ACCESS | KAUTH_VNODE_WRITE_DATA), &context) != 0) {
+ vnode_put(src_vp);
+ vnode_put(dst_vp);
+ fp_drop(p, fd2, fp2, 0);
+ error = EBADF;
+ goto outdrop;
+ }
+
+ /* Verify that both vps point to files and not directories */
+ if (!vnode_isreg(src_vp) || !vnode_isreg(dst_vp)) {
+ error = EINVAL;
+ vnode_put(src_vp);
+ vnode_put(dst_vp);
+ fp_drop(p, fd2, fp2, 0);
+ goto outdrop;
+ }
+
+ /*
+ * The exchangedata syscall handler passes in 0 for the flags to VNOP_EXCHANGE.
+ * We'll pass in our special bit indicating that the new behavior is expected
+ */
+
+ error = VNOP_EXCHANGE(src_vp, dst_vp, FSOPT_EXCHANGE_DATA_ONLY, &context);
+
+ vnode_put(src_vp);
+ vnode_put(dst_vp);
+ fp_drop(p, fd2, fp2, 0);
+ break;
+ }
+
+ /*
+ * SPI for making a file compressed.
+ */
+ case F_MAKECOMPRESSED: {
+ uint32_t gcounter = CAST_DOWN_EXPLICIT(uint32_t, uap->arg);
+
+ if (fp->f_type != DTYPE_VNODE) {
+ error = EBADF;
+ goto out;
+ }
+
+ vp = (struct vnode*) fp->f_data;
+ proc_fdunlock(p);
+
+ /* get the vnode */
+ if (vnode_getwithref(vp)) {
+ error = ENOENT;
+ goto outdrop;
+ }
+
+ /* Is it a file? */
+ if ((vnode_isreg(vp) == 0) && (vnode_islnk(vp) == 0)) {
+ vnode_put(vp);
+ error = EBADF;
+ goto outdrop;
+ }
+
+ /* invoke ioctl to pass off to FS */
+ /* Only go forward if you have write access */
+ vfs_context_t ctx = vfs_context_current();
+ if (vnode_authorize(vp, NULLVP, (KAUTH_VNODE_ACCESS | KAUTH_VNODE_WRITE_DATA), ctx) != 0) {
+ vnode_put(vp);
+ error = EBADF;
+ goto outdrop;
+ }
+
+ error = VNOP_IOCTL(vp, uap->cmd, (caddr_t)&gcounter, 0, &context);
+
+ vnode_put(vp);
+ break;
+ }
+
+ /*
+ * SPI (private) for indicating to a filesystem that subsequent writes to
+ * the open FD will written to the Fastflow.
+ */
+ case F_SET_GREEDY_MODE:
+ /* intentionally drop through to the same handler as F_SETSTATIC.
+ * both fcntls should pass the argument and their selector into VNOP_IOCTL.
+ */
+
+ /*
+ * SPI (private) for indicating to a filesystem that subsequent writes to
+ * the open FD will represent static content.
+ */
+ case F_SETSTATICCONTENT: {
+ caddr_t ioctl_arg = NULL;
+
+ if (uap->arg) {
+ ioctl_arg = (caddr_t) 1;
+ }
+
+ if (fp->f_type != DTYPE_VNODE) {
+ error = EBADF;
+ goto out;
+ }
+ vp = (struct vnode *)fp->f_data;
+ proc_fdunlock(p);
+
+ error = vnode_getwithref(vp);
+ if (error) {
+ error = ENOENT;
+ goto outdrop;
+ }
+
+ /* Only go forward if you have write access */
+ vfs_context_t ctx = vfs_context_current();
+ if (vnode_authorize(vp, NULLVP, (KAUTH_VNODE_ACCESS | KAUTH_VNODE_WRITE_DATA), ctx) != 0) {
+ vnode_put(vp);
+ error = EBADF;
+ goto outdrop;
+ }
+
+ error = VNOP_IOCTL(vp, uap->cmd, ioctl_arg, 0, &context);
+ (void)vnode_put(vp);
+
+ break;
+ }
+
+ /*
+ * SPI (private) for indicating to the lower level storage driver that the
+ * subsequent writes should be of a particular IO type (burst, greedy, static),
+ * or other flavors that may be necessary.
+ */
+ case F_SETIOTYPE: {
+ caddr_t param_ptr;
+ uint32_t param;
+
+ if (uap->arg) {
+ /* extract 32 bits of flags from userland */
+ param_ptr = (caddr_t) uap->arg;
+ param = (uint32_t) param_ptr;
+ } else {
+ /* If no argument is specified, error out */
+ error = EINVAL;
+ goto out;
+ }
+
+ /*
+ * Validate the different types of flags that can be specified:
+ * all of them are mutually exclusive for now.
+ */
+ switch (param) {
+ case F_IOTYPE_ISOCHRONOUS:
+ break;
+
+ default:
+ error = EINVAL;
+ goto out;
+ }
+
+
+ if (fp->f_type != DTYPE_VNODE) {
+ error = EBADF;
+ goto out;
+ }
+ vp = (struct vnode *)fp->f_data;
+ proc_fdunlock(p);
+
+ error = vnode_getwithref(vp);
+ if (error) {
+ error = ENOENT;
+ goto outdrop;
+ }
+
+ /* Only go forward if you have write access */
+ vfs_context_t ctx = vfs_context_current();
+ if (vnode_authorize(vp, NULLVP, (KAUTH_VNODE_ACCESS | KAUTH_VNODE_WRITE_DATA), ctx) != 0) {
+ vnode_put(vp);
+ error = EBADF;
+ goto outdrop;
+ }
+
+ error = VNOP_IOCTL(vp, uap->cmd, param_ptr, 0, &context);
+ (void)vnode_put(vp);
+
+ break;
+ }
+
+ /*
+ * Set the vnode pointed to by 'fd'
+ * and tag it as the (potentially future) backing store
+ * for another filesystem
+ */
+ case F_SETBACKINGSTORE: {
+ if (fp->f_type != DTYPE_VNODE) {
+ error = EBADF;
+ goto out;
+ }
+
+ vp = (struct vnode *)fp->f_data;
+
+ if (vp->v_tag != VT_HFS) {
+ error = EINVAL;
+ goto out;
+ }
+ proc_fdunlock(p);
+
+ if (vnode_getwithref(vp)) {
+ error = ENOENT;
+ goto outdrop;
+ }
+
+ /* only proceed if you have write access */
+ vfs_context_t ctx = vfs_context_current();
+ if (vnode_authorize(vp, NULLVP, (KAUTH_VNODE_ACCESS | KAUTH_VNODE_WRITE_DATA), ctx) != 0) {
+ vnode_put(vp);
+ error = EBADF;
+ goto outdrop;
+ }
+
+
+ /* If arg != 0, set, otherwise unset */
+ if (uap->arg) {
+ error = VNOP_IOCTL(vp, uap->cmd, (caddr_t)1, 0, &context);
+ } else {
+ error = VNOP_IOCTL(vp, uap->cmd, (caddr_t)NULL, 0, &context);
+ }
+
+ vnode_put(vp);
+ break;
+ }
+
+ /*
+ * like F_GETPATH, but special semantics for
+ * the mobile time machine handler.
+ */
+ case F_GETPATH_MTMINFO: {
+ char *pathbufp;
+ int pathlen;
+
+ if (fp->f_type != DTYPE_VNODE) {
+ error = EBADF;
+ goto out;
+ }
+ vp = (struct vnode *)fp->f_data;
+ proc_fdunlock(p);
+
+ pathlen = MAXPATHLEN;
+ MALLOC(pathbufp, char *, pathlen, M_TEMP, M_WAITOK);
+ if (pathbufp == NULL) {
+ error = ENOMEM;
+ goto outdrop;
+ }
+ if ((error = vnode_getwithref(vp)) == 0) {
+ int backingstore = 0;
+
+ /* Check for error from vn_getpath before moving on */
+ if ((error = vn_getpath(vp, pathbufp, &pathlen)) == 0) {
+ if (vp->v_tag == VT_HFS) {
+ error = VNOP_IOCTL(vp, uap->cmd, (caddr_t) &backingstore, 0, &context);
+ }
+ (void)vnode_put(vp);
+
+ if (error == 0) {
+ error = copyout((caddr_t)pathbufp, argp, pathlen);
+ }
+ if (error == 0) {
+ /*
+ * If the copyout was successful, now check to ensure
+ * that this vnode is not a BACKINGSTORE vnode. mtmd
+ * wants the path regardless.
+ */
+ if (backingstore) {
+ error = EBUSY;
+ }
+ }
+ } else {
+ (void)vnode_put(vp);
+ }
+ }
+ FREE(pathbufp, M_TEMP);
+ goto outdrop;
+ }
+
+#if DEBUG || DEVELOPMENT
+ case F_RECYCLE:
+ if (fp->f_type != DTYPE_VNODE) {
+ error = EBADF;
+ goto out;
+ }
+ vp = (struct vnode *)fp->f_data;
+ proc_fdunlock(p);
+
+ vnode_recycle(vp);
+ break;
+#endif
+
+ default:
+ /*
+ * This is an fcntl() that we d not recognize at this level;
+ * if this is a vnode, we send it down into the VNOP_IOCTL
+ * for this vnode; this can include special devices, and will
+ * effectively overload fcntl() to send ioctl()'s.
+ */
+ if ((uap->cmd & IOC_VOID) && (uap->cmd & IOC_INOUT)) {
+ error = EINVAL;
+ goto out;
+ }
+
+ /* Catch any now-invalid fcntl() selectors */
+ switch (uap->cmd) {
+ case (int)APFSIOC_REVERT_TO_SNAPSHOT:
+ case (int)FSIOC_FIOSEEKHOLE:
+ case (int)FSIOC_FIOSEEKDATA:
+ case (int)FSIOC_CAS_BSDFLAGS:
+ case HFS_GET_BOOT_INFO:
+ case HFS_SET_BOOT_INFO:
+ case FIOPINSWAP:
+ case F_MARKDEPENDENCY:
+ case TIOCREVOKE:
+ error = EINVAL;
+ goto out;
+ default:
+ break;
+ }
+
+ if (fp->f_type != DTYPE_VNODE) {
+ error = EBADF;
+ goto out;
+ }
+ vp = (struct vnode *)fp->f_data;
+ proc_fdunlock(p);
+
+ if ((error = vnode_getwithref(vp)) == 0) {
+#define STK_PARAMS 128
+ char stkbuf[STK_PARAMS] = {0};
+ unsigned int size;
+ caddr_t data, memp;
+ /*
+ * For this to work properly, we have to copy in the
+ * ioctl() cmd argument if there is one; we must also
+ * check that a command parameter, if present, does
+ * not exceed the maximum command length dictated by
+ * the number of bits we have available in the command
+ * to represent a structure length. Finally, we have
+ * to copy the results back out, if it is that type of
+ * ioctl().
+ */
+ size = IOCPARM_LEN(uap->cmd);
+ if (size > IOCPARM_MAX) {
+ (void)vnode_put(vp);
+ error = EINVAL;
+ break;
+ }
+
+ memp = NULL;
+ if (size > sizeof(stkbuf)) {
+ if ((memp = (caddr_t)kalloc(size)) == 0) {
+ (void)vnode_put(vp);
+ error = ENOMEM;
+ goto outdrop;
+ }
+ data = memp;
+ } else {
+ data = &stkbuf[0];
+ }
+
+ if (uap->cmd & IOC_IN) {
+ if (size) {
+ /* structure */
+ error = copyin(argp, data, size);
+ if (error) {
+ (void)vnode_put(vp);
+ if (memp) {
+ kfree(memp, size);
+ }
+ goto outdrop;
+ }
+
+ /* Bzero the section beyond that which was needed */
+ if (size <= sizeof(stkbuf)) {
+ bzero((((uint8_t*)data) + size), (sizeof(stkbuf) - size));
+ }
+ } else {
+ /* int */
+ if (is64bit) {
+ *(user_addr_t *)data = argp;
+ } else {
+ *(uint32_t *)data = (uint32_t)argp;
+ }
+ };
+ } else if ((uap->cmd & IOC_OUT) && size) {
+ /*
+ * Zero the buffer so the user always
+ * gets back something deterministic.
+ */
+ bzero(data, size);
+ } else if (uap->cmd & IOC_VOID) {
+ if (is64bit) {
+ *(user_addr_t *)data = argp;
+ } else {
+ *(uint32_t *)data = (uint32_t)argp;
+ }
+ }
+
+ error = VNOP_IOCTL(vp, uap->cmd, CAST_DOWN(caddr_t, data), 0, &context);
+
+ (void)vnode_put(vp);
+
+ /* Copy any output data to user */
+ if (error == 0 && (uap->cmd & IOC_OUT) && size) {
+ error = copyout(data, argp, size);
+ }
+ if (memp) {
+ kfree(memp, size);
+ }
+ }
+ break;
+ }
+
+outdrop:
+ AUDIT_ARG(vnpath_withref, vp, ARG_VNODE1);
+ fp_drop(p, fd, fp, 0);
+ return error;
+out:
+ fp_drop(p, fd, fp, 1);
+ proc_fdunlock(p);
+ return error;
+}
+
+
+/*
+ * finishdup
+ *
+ * Description: Common code for dup, dup2, and fcntl(F_DUPFD).
+ *
+ * Parameters: p Process performing the dup
+ * old The fd to dup
+ * new The fd to dup it to
+ * fd_flags Flags to augment the new fd
+ * retval Pointer to the call return area
+ *
+ * Returns: 0 Success
+ * EBADF
+ * ENOMEM
+ *
+ * Implicit returns:
+ * *retval (modified) The new descriptor
+ *
+ * Locks: Assumes proc_fdlock for process pointing to fdp is held by
+ * the caller
+ *
+ * Notes: This function may drop and reacquire this lock; it is unsafe
+ * for a caller to assume that other state protected by the lock
+ * has not been subsequently changed out from under it.
+ */
+int
+finishdup(proc_t p,
+ struct filedesc *fdp, int old, int new, int fd_flags, int32_t *retval)
+{
+ struct fileproc *nfp;
+ struct fileproc *ofp;
+#if CONFIG_MACF
+ int error;
+#endif
+
+#if DIAGNOSTIC
+ proc_fdlock_assert(p, LCK_MTX_ASSERT_OWNED);
+#endif
+ if ((ofp = fdp->fd_ofiles[old]) == NULL ||
+ (fdp->fd_ofileflags[old] & UF_RESERVED)) {
+ fdrelse(p, new);
+ return EBADF;
+ }
+ fg_ref(ofp);
+
+#if CONFIG_MACF
+ error = mac_file_check_dup(proc_ucred(p), ofp->f_fglob, new);
+ if (error) {
+ fg_drop(ofp);
+ fdrelse(p, new);
+ return error;
+ }
+#endif
+
+ proc_fdunlock(p);
+
+ nfp = fileproc_alloc_init(NULL);
+
+ proc_fdlock(p);
+
+ if (nfp == NULL) {
+ fg_drop(ofp);
+ fdrelse(p, new);
+ return ENOMEM;
+ }
+
+ nfp->f_fglob = ofp->f_fglob;
+
+#if DIAGNOSTIC
+ if (fdp->fd_ofiles[new] != 0) {
+ panic("finishdup: overwriting fd_ofiles with new %d", new);
+ }
+ if ((fdp->fd_ofileflags[new] & UF_RESERVED) == 0) {
+ panic("finishdup: unreserved fileflags with new %d", new);
+ }
+#endif
+
+ if (new > fdp->fd_lastfile) {
+ fdp->fd_lastfile = new;
+ }
+ *fdflags(p, new) |= fd_flags;
+ procfdtbl_releasefd(p, new, nfp);
+ *retval = new;
+ return 0;
+}
+
+
+/*
+ * close
+ *
+ * Description: The implementation of the close(2) system call
+ *
+ * Parameters: p Process in whose per process file table
+ * the close is to occur
+ * uap->fd fd to be closed
+ * retval <unused>
+ *
+ * Returns: 0 Success
+ * fp_lookup:EBADF Bad file descriptor
+ * fp_guard_exception:??? Guarded file descriptor
+ * close_internal:EBADF
+ * close_internal:??? Anything returnable by a per-fileops
+ * close function
+ */
+int
+close(proc_t p, struct close_args *uap, int32_t *retval)
+{
+ __pthread_testcancel(1);
+ return close_nocancel(p, (struct close_nocancel_args *)uap, retval);
+}
+
+
+int
+close_nocancel(proc_t p, struct close_nocancel_args *uap, __unused int32_t *retval)
+{
+ struct fileproc *fp;
+ int fd = uap->fd;
+ int error;
+
+ AUDIT_SYSCLOSE(p, fd);
+
+ proc_fdlock(p);
+
+ if ((error = fp_lookup(p, fd, &fp, 1))) {
+ proc_fdunlock(p);
+ return error;
+ }
+
+ if (FP_ISGUARDED(fp, GUARD_CLOSE)) {
+ error = fp_guard_exception(p, fd, fp, kGUARD_EXC_CLOSE);
+ (void) fp_drop(p, fd, fp, 1);
+ proc_fdunlock(p);
+ return error;
+ }
+
+ error = close_internal_locked(p, fd, fp, 0);
+
+ proc_fdunlock(p);
+
+ return error;
+}
+
+
+/*
+ * close_internal_locked
+ *
+ * Close a file descriptor.
+ *
+ * Parameters: p Process in whose per process file table
+ * the close is to occur
+ * fd fd to be closed
+ * fp fileproc associated with the fd
+ *
+ * Returns: 0 Success
+ * EBADF fd already in close wait state
+ * closef_locked:??? Anything returnable by a per-fileops
+ * close function
+ *
+ * Locks: Assumes proc_fdlock for process is held by the caller and returns
+ * with lock held
+ *
+ * Notes: This function may drop and reacquire this lock; it is unsafe
+ * for a caller to assume that other state protected by the lock
+ * has not been subsequently changed out from under it.
+ */
+int
+close_internal_locked(proc_t p, int fd, struct fileproc *fp, int flags)
+{
+ struct filedesc *fdp = p->p_fd;
+ int error = 0;
+ int resvfd = flags & FD_DUP2RESV;
+
+
+#if DIAGNOSTIC
+ proc_fdlock_assert(p, LCK_MTX_ASSERT_OWNED);
+#endif
+
+ /* Keep people from using the filedesc while we are closing it */
+ procfdtbl_markclosefd(p, fd);
+
+
+ if ((fp->f_flags & FP_CLOSING) == FP_CLOSING) {
+ panic("close_internal_locked: being called on already closing fd");
+ }
+
+
+#if DIAGNOSTIC
+ if ((fdp->fd_ofileflags[fd] & UF_RESERVED) == 0) {
+ panic("close_internal: unreserved fileflags with fd %d", fd);
+ }
+#endif
+
+ fp->f_flags |= FP_CLOSING;
+
+ if ((fp->f_flags & FP_AIOISSUED) || kauth_authorize_fileop_has_listeners()) {
+ proc_fdunlock(p);
+
+ if ((fp->f_type == DTYPE_VNODE) && kauth_authorize_fileop_has_listeners()) {
+ /*
+ * call out to allow 3rd party notification of close.
+ * Ignore result of kauth_authorize_fileop call.
+ */
+ if (vnode_getwithref((vnode_t)fp->f_data) == 0) {
+ u_int fileop_flags = 0;
+ if ((fp->f_flags & FP_WRITTEN) != 0) {
+ fileop_flags |= KAUTH_FILEOP_CLOSE_MODIFIED;
+ }
+ kauth_authorize_fileop(fp->f_fglob->fg_cred, KAUTH_FILEOP_CLOSE,
+ (uintptr_t)fp->f_data, (uintptr_t)fileop_flags);
+ vnode_put((vnode_t)fp->f_data);
+ }
+ }
+ if (fp->f_flags & FP_AIOISSUED) {
+ /*
+ * cancel all async IO requests that can be cancelled.
+ */
+ _aio_close( p, fd );
+ }
+
+ proc_fdlock(p);
+ }
+
+ if (fd < fdp->fd_knlistsize) {
+ knote_fdclose(p, fd);
+ }
+
+ /* release the ref returned from fp_lookup before calling drain */
+ (void) os_ref_release_locked(&fp->f_iocount);
+ fileproc_drain(p, fp);
+
+ if (fp->f_flags & FP_WAITEVENT) {
+ (void)waitevent_close(p, fp);
+ }
+
+ if (resvfd == 0) {
+ _fdrelse(p, fd);
+ } else {
+ procfdtbl_reservefd(p, fd);
+ }
+
+ if (ENTR_SHOULDTRACE && fp->f_type == DTYPE_SOCKET) {
+ KERNEL_ENERGYTRACE(kEnTrActKernSocket, DBG_FUNC_END,
+ fd, 0, (int64_t)VM_KERNEL_ADDRPERM(fp->f_data));
+ }
+
+ error = closef_locked(fp, fp->f_fglob, p);
+ if ((fp->f_flags & FP_WAITCLOSE) == FP_WAITCLOSE) {
+ wakeup(&fp->f_flags);
+ }
+ fp->f_flags &= ~(FP_WAITCLOSE | FP_CLOSING);
+
+ proc_fdunlock(p);
+
+ fileproc_free(fp);
+
+ proc_fdlock(p);
+
+#if DIAGNOSTIC
+ if (resvfd != 0) {
+ if ((fdp->fd_ofileflags[fd] & UF_RESERVED) == 0) {
+ panic("close with reserved fd returns with freed fd:%d: proc: %p", fd, p);
+ }
+ }
+#endif
+
+ return error;
+}
+
+
+/*
+ * fstat1
+ *
+ * Description: Return status information about a file descriptor.
+ *
+ * Parameters: p The process doing the fstat
+ * fd The fd to stat
+ * ub The user stat buffer
+ * xsecurity The user extended security
+ * buffer, or 0 if none
+ * xsecurity_size The size of xsecurity, or 0
+ * if no xsecurity
+ * isstat64 Flag to indicate 64 bit version
+ * for inode size, etc.
+ *
+ * Returns: 0 Success
+ * EBADF
+ * EFAULT
+ * fp_lookup:EBADF Bad file descriptor
+ * vnode_getwithref:???
+ * copyout:EFAULT
+ * vnode_getwithref:???
+ * vn_stat:???
+ * soo_stat:???
+ * pipe_stat:???
+ * pshm_stat:???
+ * kqueue_stat:???
+ *
+ * Notes: Internal implementation for all other fstat() related
+ * functions
+ *
+ * XXX switch on node type is bogus; need a stat in struct
+ * XXX fileops instead.
+ */
+static int
+fstat1(proc_t p, int fd, user_addr_t ub, user_addr_t xsecurity, user_addr_t xsecurity_size, int isstat64)
+{
+ struct fileproc *fp;
+ union {
+ struct stat sb;
+ struct stat64 sb64;
+ } source;
+ union {
+ struct user64_stat user64_sb;
+ struct user32_stat user32_sb;
+ struct user64_stat64 user64_sb64;
+ struct user32_stat64 user32_sb64;
+ } dest;
+ int error, my_size;
+ file_type_t type;
+ caddr_t data;
+ kauth_filesec_t fsec;
+ user_size_t xsecurity_bufsize;
+ vfs_context_t ctx = vfs_context_current();
+ void * sbptr;
+
+
+ AUDIT_ARG(fd, fd);
+
+ if ((error = fp_lookup(p, fd, &fp, 0)) != 0) {
+ return error;
+ }
+ type = fp->f_type;
+ data = fp->f_data;
+ fsec = KAUTH_FILESEC_NONE;
+
+ sbptr = (void *)&source;
+
+ switch (type) {
+ case DTYPE_VNODE:
+ if ((error = vnode_getwithref((vnode_t)data)) == 0) {
+ /*
+ * If the caller has the file open, and is not
+ * requesting extended security information, we are
+ * going to let them get the basic stat information.
+ */
+ if (xsecurity == USER_ADDR_NULL) {
+ error = vn_stat_noauth((vnode_t)data, sbptr, NULL, isstat64, 0, ctx,
+ fp->f_fglob->fg_cred);
+ } else {
+ error = vn_stat((vnode_t)data, sbptr, &fsec, isstat64, 0, ctx);
+ }
+
+ AUDIT_ARG(vnpath, (struct vnode *)data, ARG_VNODE1);
+ (void)vnode_put((vnode_t)data);
+ }
+ break;
+
+#if SOCKETS
+ case DTYPE_SOCKET:
+ error = soo_stat((struct socket *)data, sbptr, isstat64);
+ break;
+#endif /* SOCKETS */
+
+ case DTYPE_PIPE:
+ error = pipe_stat((void *)data, sbptr, isstat64);
+ break;
+
+ case DTYPE_PSXSHM:
+ error = pshm_stat((void *)data, sbptr, isstat64);
+ break;
+
+ case DTYPE_KQUEUE:
+ error = kqueue_stat((void *)data, sbptr, isstat64, p);
+ break;
+
+ default:
+ error = EBADF;
+ goto out;
+ }
+ if (error == 0) {
+ caddr_t sbp;
+
+ if (isstat64 != 0) {
+ source.sb64.st_lspare = 0;
+ source.sb64.st_qspare[0] = 0LL;
+ source.sb64.st_qspare[1] = 0LL;
+
+ if (IS_64BIT_PROCESS(current_proc())) {
+ munge_user64_stat64(&source.sb64, &dest.user64_sb64);
+ my_size = sizeof(dest.user64_sb64);
+ sbp = (caddr_t)&dest.user64_sb64;
+ } else {
+ munge_user32_stat64(&source.sb64, &dest.user32_sb64);
+ my_size = sizeof(dest.user32_sb64);
+ sbp = (caddr_t)&dest.user32_sb64;
+ }
+ } else {
+ source.sb.st_lspare = 0;
+ source.sb.st_qspare[0] = 0LL;
+ source.sb.st_qspare[1] = 0LL;
+ if (IS_64BIT_PROCESS(current_proc())) {
+ munge_user64_stat(&source.sb, &dest.user64_sb);
+ my_size = sizeof(dest.user64_sb);
+ sbp = (caddr_t)&dest.user64_sb;
+ } else {
+ munge_user32_stat(&source.sb, &dest.user32_sb);
+ my_size = sizeof(dest.user32_sb);
+ sbp = (caddr_t)&dest.user32_sb;
+ }
+ }
+
+ error = copyout(sbp, ub, my_size);
+ }
+
+ /* caller wants extended security information? */
+ if (xsecurity != USER_ADDR_NULL) {
+ /* did we get any? */
+ if (fsec == KAUTH_FILESEC_NONE) {
+ if (susize(xsecurity_size, 0) != 0) {
+ error = EFAULT;
+ goto out;
+ }
+ } else {
+ /* find the user buffer size */
+ xsecurity_bufsize = fusize(xsecurity_size);
+
+ /* copy out the actual data size */
+ if (susize(xsecurity_size, KAUTH_FILESEC_COPYSIZE(fsec)) != 0) {
+ error = EFAULT;
+ goto out;
+ }
+
+ /* if the caller supplied enough room, copy out to it */
+ if (xsecurity_bufsize >= KAUTH_FILESEC_COPYSIZE(fsec)) {
+ error = copyout(fsec, xsecurity, KAUTH_FILESEC_COPYSIZE(fsec));
+ }
+ }
+ }
+out:
+ fp_drop(p, fd, fp, 0);
+ if (fsec != NULL) {
+ kauth_filesec_free(fsec);
+ }
+ return error;
+}
+
+
+/*
+ * fstat_extended
+ *
+ * Description: Extended version of fstat supporting returning extended
+ * security information
+ *
+ * Parameters: p The process doing the fstat
+ * uap->fd The fd to stat
+ * uap->ub The user stat buffer
+ * uap->xsecurity The user extended security
+ * buffer, or 0 if none
+ * uap->xsecurity_size The size of xsecurity, or 0
+ *
+ * Returns: 0 Success
+ * !0 Errno (see fstat1)
+ */
+int
+fstat_extended(proc_t p, struct fstat_extended_args *uap, __unused int32_t *retval)
+{
+ return fstat1(p, uap->fd, uap->ub, uap->xsecurity, uap->xsecurity_size, 0);
+}
+
+
+/*
+ * fstat
+ *
+ * Description: Get file status for the file associated with fd
+ *
+ * Parameters: p The process doing the fstat
+ * uap->fd The fd to stat
+ * uap->ub The user stat buffer
+ *
+ * Returns: 0 Success
+ * !0 Errno (see fstat1)
+ */
+int
+fstat(proc_t p, struct fstat_args *uap, __unused int32_t *retval)
+{
+ return fstat1(p, uap->fd, uap->ub, 0, 0, 0);
+}
+
+
+/*
+ * fstat64_extended
+ *
+ * Description: Extended version of fstat64 supporting returning extended
+ * security information
+ *
+ * Parameters: p The process doing the fstat
+ * uap->fd The fd to stat
+ * uap->ub The user stat buffer
+ * uap->xsecurity The user extended security
+ * buffer, or 0 if none
+ * uap->xsecurity_size The size of xsecurity, or 0
+ *
+ * Returns: 0 Success
+ * !0 Errno (see fstat1)
+ */
+int
+fstat64_extended(proc_t p, struct fstat64_extended_args *uap, __unused int32_t *retval)
+{
+ return fstat1(p, uap->fd, uap->ub, uap->xsecurity, uap->xsecurity_size, 1);
+}
+
+
+/*
+ * fstat64
+ *
+ * Description: Get 64 bit version of the file status for the file associated
+ * with fd
+ *
+ * Parameters: p The process doing the fstat
+ * uap->fd The fd to stat
+ * uap->ub The user stat buffer
+ *
+ * Returns: 0 Success
+ * !0 Errno (see fstat1)
+ */
+int
+fstat64(proc_t p, struct fstat64_args *uap, __unused int32_t *retval)
+{
+ return fstat1(p, uap->fd, uap->ub, 0, 0, 1);
+}
+
+
+/*
+ * fpathconf
+ *
+ * Description: Return pathconf information about a file descriptor.
+ *
+ * Parameters: p Process making the request
+ * uap->fd fd to get information about
+ * uap->name Name of information desired
+ * retval Pointer to the call return area
+ *
+ * Returns: 0 Success
+ * EINVAL
+ * fp_lookup:EBADF Bad file descriptor
+ * vnode_getwithref:???
+ * vn_pathconf:???
+ *
+ * Implicit returns:
+ * *retval (modified) Returned information (numeric)
+ */
+int
+fpathconf(proc_t p, struct fpathconf_args *uap, int32_t *retval)
+{
+ int fd = uap->fd;
+ struct fileproc *fp;
+ struct vnode *vp;
+ int error = 0;
+ file_type_t type;
+ caddr_t data;
+
+
+ AUDIT_ARG(fd, uap->fd);
+ if ((error = fp_lookup(p, fd, &fp, 0))) {
+ return error;
+ }
+ type = fp->f_type;
+ data = fp->f_data;
+
+ switch (type) {
+ case DTYPE_SOCKET:
+ if (uap->name != _PC_PIPE_BUF) {
+ error = EINVAL;
+ goto out;
+ }
+ *retval = PIPE_BUF;
+ error = 0;
+ goto out;
+
+ case DTYPE_PIPE:
+ if (uap->name != _PC_PIPE_BUF) {
+ error = EINVAL;
+ goto out;
+ }
+ *retval = PIPE_BUF;
+ error = 0;
+ goto out;
+
+ case DTYPE_VNODE:
+ vp = (struct vnode *)data;
+
+ if ((error = vnode_getwithref(vp)) == 0) {
+ AUDIT_ARG(vnpath, vp, ARG_VNODE1);
+
+ error = vn_pathconf(vp, uap->name, retval, vfs_context_current());
+
+ (void)vnode_put(vp);
+ }
+ goto out;
+
+ default:
+ error = EINVAL;
+ goto out;
+ }
+ /*NOTREACHED*/
+out:
+ fp_drop(p, fd, fp, 0);
+ return error;
+}
+
+/*
+ * Statistics counter for the number of times a process calling fdalloc()
+ * has resulted in an expansion of the per process open file table.
+ *
+ * XXX This would likely be of more use if it were per process
+ */
+int fdexpand;
+
+
+/*
+ * fdalloc
+ *
+ * Description: Allocate a file descriptor for the process.
+ *
+ * Parameters: p Process to allocate the fd in
+ * want The fd we would prefer to get
+ * result Pointer to fd we got
+ *
+ * Returns: 0 Success
+ * EMFILE
+ * ENOMEM
+ *
+ * Implicit returns:
+ * *result (modified) The fd which was allocated
+ */
+int
+fdalloc(proc_t p, int want, int *result)
+{
+ struct filedesc *fdp = p->p_fd;
+ int i;
+ int lim, last, numfiles, oldnfiles;
+ struct fileproc **newofiles, **ofiles;
+ char *newofileflags;
+
+ /*
+ * Search for a free descriptor starting at the higher
+ * of want or fd_freefile. If that fails, consider
+ * expanding the ofile array.
+ */
+#if DIAGNOSTIC
+ proc_fdlock_assert(p, LCK_MTX_ASSERT_OWNED);
+#endif
+
+ lim = min((int)p->p_rlimit[RLIMIT_NOFILE].rlim_cur, maxfiles);
+ for (;;) {
+ last = min(fdp->fd_nfiles, lim);
+ if ((i = want) < fdp->fd_freefile) {
+ i = fdp->fd_freefile;
+ }
+ for (; i < last; i++) {
+ if (fdp->fd_ofiles[i] == NULL && !(fdp->fd_ofileflags[i] & UF_RESERVED)) {
+ procfdtbl_reservefd(p, i);
+ if (i > fdp->fd_lastfile) {
+ fdp->fd_lastfile = i;
+ }
+ if (want <= fdp->fd_freefile) {
+ fdp->fd_freefile = i;
+ }
+ *result = i;
+ return 0;
+ }
+ }
+
+ /*
+ * No space in current array. Expand?
+ */
+ if (fdp->fd_nfiles >= lim) {
+ return EMFILE;
+ }
+ if (fdp->fd_nfiles < NDEXTENT) {
+ numfiles = NDEXTENT;
+ } else {
+ numfiles = 2 * fdp->fd_nfiles;
+ }
+ /* Enforce lim */
+ if (numfiles > lim) {
+ numfiles = lim;
+ }
+ proc_fdunlock(p);
+ MALLOC_ZONE(newofiles, struct fileproc **,
+ numfiles * OFILESIZE, M_OFILETABL, M_WAITOK);
+ proc_fdlock(p);
+ if (newofiles == NULL) {
+ return ENOMEM;
+ }
+ if (fdp->fd_nfiles >= numfiles) {
+ FREE_ZONE(newofiles, numfiles * OFILESIZE, M_OFILETABL);
+ continue;
+ }
+ newofileflags = (char *) &newofiles[numfiles];
+ /*
+ * Copy the existing ofile and ofileflags arrays
+ * and zero the new portion of each array.
+ */
+ oldnfiles = fdp->fd_nfiles;
+ (void) memcpy(newofiles, fdp->fd_ofiles,
+ oldnfiles * sizeof(*fdp->fd_ofiles));
+ (void) memset(&newofiles[oldnfiles], 0,
+ (numfiles - oldnfiles) * sizeof(*fdp->fd_ofiles));
+
+ (void) memcpy(newofileflags, fdp->fd_ofileflags,
+ oldnfiles * sizeof(*fdp->fd_ofileflags));
+ (void) memset(&newofileflags[oldnfiles], 0,
+ (numfiles - oldnfiles) *
+ sizeof(*fdp->fd_ofileflags));
+ ofiles = fdp->fd_ofiles;
+ fdp->fd_ofiles = newofiles;
+ fdp->fd_ofileflags = newofileflags;
+ fdp->fd_nfiles = numfiles;
+ FREE_ZONE(ofiles, oldnfiles * OFILESIZE, M_OFILETABL);
+ fdexpand++;
+ }
+}
+
+
+/*
+ * fdavail
+ *
+ * Description: Check to see whether n user file descriptors are available
+ * to the process p.
+ *
+ * Parameters: p Process to check in
+ * n The number of fd's desired
+ *
+ * Returns: 0 No
+ * 1 Yes
+ *
+ * Locks: Assumes proc_fdlock for process is held by the caller
+ *
+ * Notes: The answer only remains valid so long as the proc_fdlock is
+ * held by the caller.
+ */
+int
+fdavail(proc_t p, int n)
+{
+ struct filedesc *fdp = p->p_fd;
+ struct fileproc **fpp;
+ char *flags;
+ int i, lim;
+
+ lim = min((int)p->p_rlimit[RLIMIT_NOFILE].rlim_cur, maxfiles);
+ if ((i = lim - fdp->fd_nfiles) > 0 && (n -= i) <= 0) {
+ return 1;
+ }
+ fpp = &fdp->fd_ofiles[fdp->fd_freefile];
+ flags = &fdp->fd_ofileflags[fdp->fd_freefile];
+ for (i = fdp->fd_nfiles - fdp->fd_freefile; --i >= 0; fpp++, flags++) {
+ if (*fpp == NULL && !(*flags & UF_RESERVED) && --n <= 0) {
+ return 1;
+ }
+ }
+ return 0;
+}
+
+
+/*
+ * fdrelse
+ *
+ * Description: Legacy KPI wrapper function for _fdrelse
+ *
+ * Parameters: p Process in which fd lives
+ * fd fd to free
+ *
+ * Returns: void
+ *
+ * Locks: Assumes proc_fdlock for process is held by the caller
+ */
+void
+fdrelse(proc_t p, int fd)
+{
+ _fdrelse(p, fd);
+}
+
+
+/*
+ * fdgetf_noref
+ *
+ * Description: Get the fileproc pointer for the given fd from the per process
+ * open file table without taking an explicit reference on it.
+ *
+ * Parameters: p Process containing fd
+ * fd fd to obtain fileproc for
+ * resultfp Pointer to pointer return area
+ *
+ * Returns: 0 Success
+ * EBADF
+ *
+ * Implicit returns:
+ * *resultfp (modified) Pointer to fileproc pointer
+ *
+ * Locks: Assumes proc_fdlock for process is held by the caller
+ *
+ * Notes: Because there is no reference explicitly taken, the returned
+ * fileproc pointer is only valid so long as the proc_fdlock
+ * remains held by the caller.
+ */
+int
+fdgetf_noref(proc_t p, int fd, struct fileproc **resultfp)
+{
+ struct filedesc *fdp = p->p_fd;
+ struct fileproc *fp;
+
+ if (fd < 0 || fd >= fdp->fd_nfiles ||
+ (fp = fdp->fd_ofiles[fd]) == NULL ||
+ (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
+ return EBADF;
+ }
+ if (resultfp) {
+ *resultfp = fp;
+ }
+ return 0;
+}
+
+
+/*
+ * fp_getfvp
+ *
+ * Description: Get fileproc and vnode pointer for a given fd from the per
+ * process open file table of the specified process, and if
+ * successful, increment the f_iocount
+ *
+ * Parameters: p Process in which fd lives
+ * fd fd to get information for
+ * resultfp Pointer to result fileproc
+ * pointer area, or 0 if none
+ * resultvp Pointer to result vnode pointer
+ * area, or 0 if none
+ *
+ * Returns: 0 Success
+ * EBADF Bad file descriptor
+ * ENOTSUP fd does not refer to a vnode
+ *
+ * Implicit returns:
+ * *resultfp (modified) Fileproc pointer
+ * *resultvp (modified) vnode pointer
+ *
+ * Notes: The resultfp and resultvp fields are optional, and may be
+ * independently specified as NULL to skip returning information
+ *
+ * Locks: Internally takes and releases proc_fdlock
+ */
+int
+fp_getfvp(proc_t p, int fd, struct fileproc **resultfp, struct vnode **resultvp)
+{
+ struct filedesc *fdp = p->p_fd;
+ struct fileproc *fp;
+
+ proc_fdlock_spin(p);
+ if (fd < 0 || fd >= fdp->fd_nfiles ||
+ (fp = fdp->fd_ofiles[fd]) == NULL ||
+ (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
+ proc_fdunlock(p);
+ return EBADF;
+ }
+ if (fp->f_type != DTYPE_VNODE) {
+ proc_fdunlock(p);
+ return ENOTSUP;
+ }
+ os_ref_retain_locked(&fp->f_iocount);
+
+ if (resultfp) {
+ *resultfp = fp;
+ }
+ if (resultvp) {
+ *resultvp = (struct vnode *)fp->f_data;
+ }
+ proc_fdunlock(p);
+
+ return 0;
+}
+
+
+/*
+ * fp_getfvpandvid
+ *
+ * Description: Get fileproc, vnode pointer, and vid for a given fd from the
+ * per process open file table of the specified process, and if
+ * successful, increment the f_iocount
+ *
+ * Parameters: p Process in which fd lives
+ * fd fd to get information for
+ * resultfp Pointer to result fileproc
+ * pointer area, or 0 if none
+ * resultvp Pointer to result vnode pointer
+ * area, or 0 if none
+ * vidp Pointer to resuld vid area
+ *
+ * Returns: 0 Success
+ * EBADF Bad file descriptor
+ * ENOTSUP fd does not refer to a vnode
+ *
+ * Implicit returns:
+ * *resultfp (modified) Fileproc pointer
+ * *resultvp (modified) vnode pointer
+ * *vidp vid value
+ *
+ * Notes: The resultfp and resultvp fields are optional, and may be
+ * independently specified as NULL to skip returning information
+ *
+ * Locks: Internally takes and releases proc_fdlock
+ */
+int
+fp_getfvpandvid(proc_t p, int fd, struct fileproc **resultfp,
+ struct vnode **resultvp, uint32_t *vidp)
+{
+ struct filedesc *fdp = p->p_fd;
+ struct fileproc *fp;
+
+ proc_fdlock_spin(p);
+ if (fd < 0 || fd >= fdp->fd_nfiles ||
+ (fp = fdp->fd_ofiles[fd]) == NULL ||
+ (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
+ proc_fdunlock(p);
+ return EBADF;
+ }
+ if (fp->f_type != DTYPE_VNODE) {
+ proc_fdunlock(p);
+ return ENOTSUP;
+ }
+ os_ref_retain_locked(&fp->f_iocount);
+
+ if (resultfp) {
+ *resultfp = fp;
+ }
+ if (resultvp) {
+ *resultvp = (struct vnode *)fp->f_data;
+ }
+ if (vidp) {
+ *vidp = (uint32_t)vnode_vid((struct vnode *)fp->f_data);
+ }
+ proc_fdunlock(p);
+
+ return 0;
+}
+
+
+/*
+ * fp_getfsock
+ *
+ * Description: Get fileproc and socket pointer for a given fd from the
+ * per process open file table of the specified process, and if
+ * successful, increment the f_iocount
+ *
+ * Parameters: p Process in which fd lives
+ * fd fd to get information for
+ * resultfp Pointer to result fileproc
+ * pointer area, or 0 if none
+ * results Pointer to result socket
+ * pointer area, or 0 if none
+ *
+ * Returns: EBADF The file descriptor is invalid
+ * EOPNOTSUPP The file descriptor is not a socket
+ * 0 Success
+ *
+ * Implicit returns:
+ * *resultfp (modified) Fileproc pointer
+ * *results (modified) socket pointer
+ *
+ * Notes: EOPNOTSUPP should probably be ENOTSOCK; this function is only
+ * ever called from accept1().
+ */
+int
+fp_getfsock(proc_t p, int fd, struct fileproc **resultfp,
+ struct socket **results)
+{
+ struct filedesc *fdp = p->p_fd;
+ struct fileproc *fp;
+
+ proc_fdlock_spin(p);
+ if (fd < 0 || fd >= fdp->fd_nfiles ||
+ (fp = fdp->fd_ofiles[fd]) == NULL ||
+ (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
+ proc_fdunlock(p);
+ return EBADF;
+ }
+ if (fp->f_type != DTYPE_SOCKET) {
+ proc_fdunlock(p);
+ return EOPNOTSUPP;
+ }
+ os_ref_retain_locked(&fp->f_iocount);
+
+ if (resultfp) {
+ *resultfp = fp;
+ }
+ if (results) {
+ *results = (struct socket *)fp->f_data;
+ }
+ proc_fdunlock(p);
+
+ return 0;
+}
+
+
+/*
+ * fp_getfkq
+ *
+ * Description: Get fileproc and kqueue pointer for a given fd from the
+ * per process open file table of the specified process, and if
+ * successful, increment the f_iocount
+ *
+ * Parameters: p Process in which fd lives
+ * fd fd to get information for
+ * resultfp Pointer to result fileproc
+ * pointer area, or 0 if none
+ * resultkq Pointer to result kqueue
+ * pointer area, or 0 if none
+ *
+ * Returns: EBADF The file descriptor is invalid
+ * EBADF The file descriptor is not a socket
+ * 0 Success
+ *
+ * Implicit returns:
+ * *resultfp (modified) Fileproc pointer
+ * *resultkq (modified) kqueue pointer
+ *
+ * Notes: The second EBADF should probably be something else to make
+ * the error condition distinct.
+ */
+int
+fp_getfkq(proc_t p, int fd, struct fileproc **resultfp,
+ struct kqueue **resultkq)
+{
+ struct filedesc *fdp = p->p_fd;
+ struct fileproc *fp;
+
+ proc_fdlock_spin(p);
+ if (fd < 0 || fd >= fdp->fd_nfiles ||
+ (fp = fdp->fd_ofiles[fd]) == NULL ||
+ (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
+ proc_fdunlock(p);
+ return EBADF;
+ }
+ if (fp->f_type != DTYPE_KQUEUE) {
+ proc_fdunlock(p);
+ return EBADF;
+ }
+ os_ref_retain_locked(&fp->f_iocount);
+
+ if (resultfp) {
+ *resultfp = fp;
+ }
+ if (resultkq) {
+ *resultkq = (struct kqueue *)fp->f_data;
+ }
+ proc_fdunlock(p);
+
+ return 0;
+}
+
+
+/*
+ * fp_getfpshm
+ *
+ * Description: Get fileproc and POSIX shared memory pointer for a given fd
+ * from the per process open file table of the specified process
+ * and if successful, increment the f_iocount
+ *
+ * Parameters: p Process in which fd lives
+ * fd fd to get information for
+ * resultfp Pointer to result fileproc
+ * pointer area, or 0 if none
+ * resultpshm Pointer to result POSIX
+ * shared memory pointer
+ * pointer area, or 0 if none
+ *
+ * Returns: EBADF The file descriptor is invalid
+ * EBADF The file descriptor is not a POSIX
+ * shared memory area
+ * 0 Success
+ *
+ * Implicit returns:
+ * *resultfp (modified) Fileproc pointer
+ * *resultpshm (modified) POSIX shared memory pointer
+ *
+ * Notes: The second EBADF should probably be something else to make
+ * the error condition distinct.
+ */
+int
+fp_getfpshm(proc_t p, int fd, struct fileproc **resultfp,
+ struct pshmnode **resultpshm)
+{
+ struct filedesc *fdp = p->p_fd;
+ struct fileproc *fp;
+
+ proc_fdlock_spin(p);
+ if (fd < 0 || fd >= fdp->fd_nfiles ||
+ (fp = fdp->fd_ofiles[fd]) == NULL ||
+ (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
+ proc_fdunlock(p);
+ return EBADF;
+ }
+ if (fp->f_type != DTYPE_PSXSHM) {
+ proc_fdunlock(p);
+ return EBADF;
+ }
+ os_ref_retain_locked(&fp->f_iocount);
+
+ if (resultfp) {
+ *resultfp = fp;
+ }
+ if (resultpshm) {
+ *resultpshm = (struct pshmnode *)fp->f_data;
+ }
+ proc_fdunlock(p);
+
+ return 0;
+}
+
+
+/*
+ * fp_getfsem
+ *
+ * Description: Get fileproc and POSIX semaphore pointer for a given fd from
+ * the per process open file table of the specified process
+ * and if successful, increment the f_iocount
+ *
+ * Parameters: p Process in which fd lives
+ * fd fd to get information for
+ * resultfp Pointer to result fileproc
+ * pointer area, or 0 if none
+ * resultpsem Pointer to result POSIX
+ * semaphore pointer area, or
+ * 0 if none
+ *
+ * Returns: EBADF The file descriptor is invalid
+ * EBADF The file descriptor is not a POSIX
+ * semaphore
+ * 0 Success
+ *
+ * Implicit returns:
+ * *resultfp (modified) Fileproc pointer
+ * *resultpsem (modified) POSIX semaphore pointer
+ *
+ * Notes: The second EBADF should probably be something else to make
+ * the error condition distinct.
+ *
+ * In order to support unnamed POSIX semaphores, the named
+ * POSIX semaphores will have to move out of the per-process
+ * open filetable, and into a global table that is shared with
+ * unnamed POSIX semaphores, since unnamed POSIX semaphores
+ * are typically used by declaring instances in shared memory,
+ * and there's no other way to do this without changing the
+ * underlying type, which would introduce binary compatibility
+ * issues.
+ */
+int
+fp_getfpsem(proc_t p, int fd, struct fileproc **resultfp,
+ struct psemnode **resultpsem)
+{
+ struct filedesc *fdp = p->p_fd;
+ struct fileproc *fp;
+
+ proc_fdlock_spin(p);
+ if (fd < 0 || fd >= fdp->fd_nfiles ||
+ (fp = fdp->fd_ofiles[fd]) == NULL ||
+ (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
+ proc_fdunlock(p);
+ return EBADF;
+ }
+ if (fp->f_type != DTYPE_PSXSEM) {
+ proc_fdunlock(p);
+ return EBADF;
+ }
+ os_ref_retain_locked(&fp->f_iocount);
+
+ if (resultfp) {
+ *resultfp = fp;
+ }
+ if (resultpsem) {
+ *resultpsem = (struct psemnode *)fp->f_data;
+ }
+ proc_fdunlock(p);
+
+ return 0;
+}
+
+
+/*
+ * fp_getfpipe
+ *
+ * Description: Get fileproc and pipe pointer for a given fd from the
+ * per process open file table of the specified process
+ * and if successful, increment the f_iocount
+ *
+ * Parameters: p Process in which fd lives
+ * fd fd to get information for
+ * resultfp Pointer to result fileproc
+ * pointer area, or 0 if none
+ * resultpipe Pointer to result pipe
+ * pointer area, or 0 if none
+ *
+ * Returns: EBADF The file descriptor is invalid
+ * EBADF The file descriptor is not a socket
+ * 0 Success
+ *
+ * Implicit returns:
+ * *resultfp (modified) Fileproc pointer
+ * *resultpipe (modified) pipe pointer
+ *
+ * Notes: The second EBADF should probably be something else to make
+ * the error condition distinct.
+ */
+int
+fp_getfpipe(proc_t p, int fd, struct fileproc **resultfp,
+ struct pipe **resultpipe)
+{
+ struct filedesc *fdp = p->p_fd;
+ struct fileproc *fp;
+
+ proc_fdlock_spin(p);
+ if (fd < 0 || fd >= fdp->fd_nfiles ||
+ (fp = fdp->fd_ofiles[fd]) == NULL ||
+ (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
+ proc_fdunlock(p);
+ return EBADF;
+ }
+ if (fp->f_type != DTYPE_PIPE) {
+ proc_fdunlock(p);
+ return EBADF;
+ }
+ os_ref_retain_locked(&fp->f_iocount);
+
+ if (resultfp) {
+ *resultfp = fp;
+ }
+ if (resultpipe) {
+ *resultpipe = (struct pipe *)fp->f_data;
+ }
+ proc_fdunlock(p);
+
+ return 0;
+}
+
+
+/*
+ * fp_lookup
+ *
+ * Description: Get fileproc pointer for a given fd from the per process
+ * open file table of the specified process and if successful,
+ * increment the f_iocount
+ *
+ * Parameters: p Process in which fd lives
+ * fd fd to get information for
+ * resultfp Pointer to result fileproc
+ * pointer area, or 0 if none
+ * locked !0 if the caller holds the
+ * proc_fdlock, 0 otherwise
+ *
+ * Returns: 0 Success
+ * EBADF Bad file descriptor
+ *
+ * Implicit returns:
+ * *resultfp (modified) Fileproc pointer
+ *
+ * Locks: If the argument 'locked' is non-zero, then the caller is
+ * expected to have taken and held the proc_fdlock; if it is
+ * zero, than this routine internally takes and drops this lock.
+ */
+int
+fp_lookup(proc_t p, int fd, struct fileproc **resultfp, int locked)
+{
+ struct filedesc *fdp = p->p_fd;
+ struct fileproc *fp;
+
+ if (!locked) {
+ proc_fdlock_spin(p);
+ }
+ if (fd < 0 || fdp == NULL || fd >= fdp->fd_nfiles ||
+ (fp = fdp->fd_ofiles[fd]) == NULL ||
+ (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
+ if (!locked) {
+ proc_fdunlock(p);
+ }
+ return EBADF;
+ }
+ os_ref_retain_locked(&fp->f_iocount);
+
+ if (resultfp) {
+ *resultfp = fp;
+ }
+ if (!locked) {
+ proc_fdunlock(p);
+ }
+
+ return 0;
+}
+
+
+/*
+ * fp_tryswap
+ *
+ * Description: Swap the fileproc pointer for a given fd with a new
+ * fileproc pointer in the per-process open file table of
+ * the specified process. The fdlock must be held at entry.
+ * Iff the swap is successful, the old fileproc pointer is freed.
+ *
+ * Parameters: p Process containing the fd
+ * fd The fd of interest
+ * nfp Pointer to the newfp
+ *
+ * Returns: 0 Success
+ * EBADF Bad file descriptor
+ * EINTR Interrupted
+ * EKEEPLOOKING Other references were active, try again.
+ */
+int
+fp_tryswap(proc_t p, int fd, struct fileproc *nfp)
+{
+ struct fileproc *fp;
+ int error;
+
+ proc_fdlock_assert(p, LCK_MTX_ASSERT_OWNED);
+
+ if (0 != (error = fp_lookup(p, fd, &fp, 1))) {
+ return error;
+ }
+ /*
+ * At this point, our caller (change_guardedfd_np) has
+ * one f_iocount reference, and we just took another
+ * one to begin the replacement.
+ * fp and nfp have a +1 reference from allocation.
+ * Thus if no-one else is looking, f_iocount should be 3.
+ */
+ if (os_ref_get_count(&fp->f_iocount) < 3 ||
+ 1 != os_ref_get_count(&nfp->f_iocount)) {
+ panic("%s: f_iocount", __func__);
+ } else if (3 == os_ref_get_count(&fp->f_iocount)) {
+ /* Copy the contents of *fp, preserving the "type" of *nfp */
+
+ nfp->f_flags = (nfp->f_flags & FP_TYPEMASK) |
+ (fp->f_flags & ~FP_TYPEMASK);
+ os_ref_retain_locked(&nfp->f_iocount);
+ os_ref_retain_locked(&nfp->f_iocount);
+ nfp->f_fglob = fp->f_fglob;
+ nfp->f_wset = fp->f_wset;
+
+ p->p_fd->fd_ofiles[fd] = nfp;
+ fp_drop(p, fd, nfp, 1);
+
+ os_ref_release_live(&fp->f_iocount);
+ os_ref_release_live(&fp->f_iocount);
+ fileproc_free(fp);
+ } else {
+ /*
+ * Wait for all other active references to evaporate.
+ */
+ p->p_fpdrainwait = 1;
+ error = msleep(&p->p_fpdrainwait, &p->p_fdmlock,
+ PRIBIO | PCATCH, "tryswap fpdrain", NULL);
+ if (0 == error) {
+ /*
+ * Return an "internal" errno to trigger a full
+ * reevaluation of the change-guard attempt.
+ */
+ error = EKEEPLOOKING;
+ }
+ (void) fp_drop(p, fd, fp, 1);
+ }
+ return error;
+}
+
+
+/*
+ * fp_drop_written
+ *
+ * Description: Set the FP_WRITTEN flag on the fileproc and drop the I/O
+ * reference previously taken by calling fp_lookup et. al.
+ *
+ * Parameters: p Process in which the fd lives
+ * fd fd associated with the fileproc
+ * fp fileproc on which to set the
+ * flag and drop the reference
+ *
+ * Returns: 0 Success
+ * fp_drop:EBADF Bad file descriptor
+ *
+ * Locks: This function internally takes and drops the proc_fdlock for
+ * the supplied process
+ *
+ * Notes: The fileproc must correspond to the fd in the supplied proc
+ */
+int
+fp_drop_written(proc_t p, int fd, struct fileproc *fp)
+{
+ int error;
+
+ proc_fdlock_spin(p);
+
+ fp->f_flags |= FP_WRITTEN;
+
+ error = fp_drop(p, fd, fp, 1);
+
+ proc_fdunlock(p);
+
+ return error;
+}
+
+
+/*
+ * fp_drop_event
+ *
+ * Description: Set the FP_WAITEVENT flag on the fileproc and drop the I/O
+ * reference previously taken by calling fp_lookup et. al.
+ *
+ * Parameters: p Process in which the fd lives
+ * fd fd associated with the fileproc
+ * fp fileproc on which to set the
+ * flag and drop the reference
+ *
+ * Returns: 0 Success
+ * fp_drop:EBADF Bad file descriptor
+ *
+ * Locks: This function internally takes and drops the proc_fdlock for
+ * the supplied process
+ *
+ * Notes: The fileproc must correspond to the fd in the supplied proc
+ */
+int
+fp_drop_event(proc_t p, int fd, struct fileproc *fp)
+{
+ int error;
+
+ proc_fdlock_spin(p);
+
+ fp->f_flags |= FP_WAITEVENT;
+
+ error = fp_drop(p, fd, fp, 1);
+
+ proc_fdunlock(p);
+
+ return error;
+}
+
+
+/*
+ * fp_drop
+ *
+ * Description: Drop the I/O reference previously taken by calling fp_lookup
+ * et. al.
+ *
+ * Parameters: p Process in which the fd lives
+ * fd fd associated with the fileproc
+ * fp fileproc on which to set the
+ * flag and drop the reference
+ * locked flag to internally take and
+ * drop proc_fdlock if it is not
+ * already held by the caller
+ *
+ * Returns: 0 Success
+ * EBADF Bad file descriptor
+ *
+ * Locks: This function internally takes and drops the proc_fdlock for
+ * the supplied process if 'locked' is non-zero, and assumes that
+ * the caller already holds this lock if 'locked' is non-zero.
+ *
+ * Notes: The fileproc must correspond to the fd in the supplied proc
+ */
+int
+fp_drop(proc_t p, int fd, struct fileproc *fp, int locked)
+{
+ struct filedesc *fdp = p->p_fd;
+ int needwakeup = 0;
+
+ if (!locked) {
+ proc_fdlock_spin(p);
+ }
+ if ((fp == FILEPROC_NULL) && (fd < 0 || fd >= fdp->fd_nfiles ||
+ (fp = fdp->fd_ofiles[fd]) == NULL ||
+ ((fdp->fd_ofileflags[fd] & UF_RESERVED) &&
+ !(fdp->fd_ofileflags[fd] & UF_CLOSING)))) {
+ if (!locked) {
+ proc_fdunlock(p);
+ }
+ return EBADF;
+ }
+
+ if (1 == os_ref_release_locked(&fp->f_iocount)) {
+ if (fp->f_flags & FP_SELCONFLICT) {
+ fp->f_flags &= ~FP_SELCONFLICT;
+ }
+
+ if (p->p_fpdrainwait) {
+ p->p_fpdrainwait = 0;
+ needwakeup = 1;
+ }
+ }
+ if (!locked) {
+ proc_fdunlock(p);
+ }
+ if (needwakeup) {
+ wakeup(&p->p_fpdrainwait);
+ }
+
+ return 0;
+}
+
+
+/*
+ * file_vnode
+ *
+ * Description: Given an fd, look it up in the current process's per process
+ * open file table, and return its internal vnode pointer.
+ *
+ * Parameters: fd fd to obtain vnode from
+ * vpp pointer to vnode return area
+ *
+ * Returns: 0 Success
+ * EINVAL The fd does not refer to a
+ * vnode fileproc entry
+ * fp_lookup:EBADF Bad file descriptor
+ *
+ * Implicit returns:
+ * *vpp (modified) Returned vnode pointer
+ *
+ * Locks: This function internally takes and drops the proc_fdlock for
+ * the current process
+ *
+ * Notes: If successful, this function increments the f_iocount on the
+ * fd's corresponding fileproc.
+ *
+ * The fileproc referenced is not returned; because of this, care
+ * must be taken to not drop the last reference (e.g. by closing
+ * the file). This is inherently unsafe, since the reference may
+ * not be recoverable from the vnode, if there is a subsequent
+ * close that destroys the associate fileproc. The caller should
+ * therefore retain their own reference on the fileproc so that
+ * the f_iocount can be dropped subsequently. Failure to do this
+ * can result in the returned pointer immediately becoming invalid
+ * following the call.
+ *
+ * Use of this function is discouraged.
+ */
+int
+file_vnode(int fd, struct vnode **vpp)
+{
+ proc_t p = current_proc();
+ struct fileproc *fp;
+ int error;
+
+ proc_fdlock_spin(p);
+ if ((error = fp_lookup(p, fd, &fp, 1))) {
+ proc_fdunlock(p);
+ return error;
+ }
+ if (fp->f_type != DTYPE_VNODE) {
+ fp_drop(p, fd, fp, 1);
+ proc_fdunlock(p);
+ return EINVAL;
+ }
+ if (vpp != NULL) {
+ *vpp = (struct vnode *)fp->f_data;
+ }
+ proc_fdunlock(p);
+
+ return 0;
+}
+
+
+/*
+ * file_vnode_withvid
+ *
+ * Description: Given an fd, look it up in the current process's per process
+ * open file table, and return its internal vnode pointer.
+ *
+ * Parameters: fd fd to obtain vnode from
+ * vpp pointer to vnode return area
+ * vidp pointer to vid of the returned vnode
+ *
+ * Returns: 0 Success
+ * EINVAL The fd does not refer to a
+ * vnode fileproc entry
+ * fp_lookup:EBADF Bad file descriptor
+ *
+ * Implicit returns:
+ * *vpp (modified) Returned vnode pointer
+ *
+ * Locks: This function internally takes and drops the proc_fdlock for
+ * the current process
+ *
+ * Notes: If successful, this function increments the f_iocount on the
+ * fd's corresponding fileproc.
+ *
+ * The fileproc referenced is not returned; because of this, care
+ * must be taken to not drop the last reference (e.g. by closing
+ * the file). This is inherently unsafe, since the reference may
+ * not be recoverable from the vnode, if there is a subsequent
+ * close that destroys the associate fileproc. The caller should
+ * therefore retain their own reference on the fileproc so that
+ * the f_iocount can be dropped subsequently. Failure to do this
+ * can result in the returned pointer immediately becoming invalid
+ * following the call.
+ *
+ * Use of this function is discouraged.
+ */
+int
+file_vnode_withvid(int fd, struct vnode **vpp, uint32_t * vidp)
+{
+ proc_t p = current_proc();
+ struct fileproc *fp;
+ vnode_t vp;
+ int error;
+
+ proc_fdlock_spin(p);
+ if ((error = fp_lookup(p, fd, &fp, 1))) {
+ proc_fdunlock(p);
+ return error;
+ }
+ if (fp->f_type != DTYPE_VNODE) {
+ fp_drop(p, fd, fp, 1);
+ proc_fdunlock(p);
+ return EINVAL;
+ }
+ vp = (struct vnode *)fp->f_data;
+ if (vpp != NULL) {
+ *vpp = vp;
+ }
+
+ if ((vidp != NULL) && (vp != NULLVP)) {
+ *vidp = (uint32_t)vp->v_id;
+ }
+
+ proc_fdunlock(p);
+
+ return 0;
+}
+
+
+/*
+ * file_socket
+ *
+ * Description: Given an fd, look it up in the current process's per process
+ * open file table, and return its internal socket pointer.
+ *
+ * Parameters: fd fd to obtain vnode from
+ * sp pointer to socket return area
+ *
+ * Returns: 0 Success
+ * ENOTSOCK Not a socket
+ * fp_lookup:EBADF Bad file descriptor
+ *
+ * Implicit returns:
+ * *sp (modified) Returned socket pointer
+ *
+ * Locks: This function internally takes and drops the proc_fdlock for
+ * the current process
+ *
+ * Notes: If successful, this function increments the f_iocount on the
+ * fd's corresponding fileproc.
+ *
+ * The fileproc referenced is not returned; because of this, care
+ * must be taken to not drop the last reference (e.g. by closing
+ * the file). This is inherently unsafe, since the reference may
+ * not be recoverable from the socket, if there is a subsequent
+ * close that destroys the associate fileproc. The caller should
+ * therefore retain their own reference on the fileproc so that
+ * the f_iocount can be dropped subsequently. Failure to do this
+ * can result in the returned pointer immediately becoming invalid
+ * following the call.
+ *
+ * Use of this function is discouraged.
+ */
+int
+file_socket(int fd, struct socket **sp)
+{
+ proc_t p = current_proc();
+ struct fileproc *fp;
+ int error;
+
+ proc_fdlock_spin(p);
+ if ((error = fp_lookup(p, fd, &fp, 1))) {
+ proc_fdunlock(p);
+ return error;
+ }
+ if (fp->f_type != DTYPE_SOCKET) {
+ fp_drop(p, fd, fp, 1);
+ proc_fdunlock(p);
+ return ENOTSOCK;
+ }
+ *sp = (struct socket *)fp->f_data;
+ proc_fdunlock(p);
+
+ return 0;
+}
+
+
+/*
+ * file_flags
+ *
+ * Description: Given an fd, look it up in the current process's per process
+ * open file table, and return its fileproc's flags field.
+ *
+ * Parameters: fd fd whose flags are to be
+ * retrieved
+ * flags pointer to flags data area
+ *
+ * Returns: 0 Success
+ * ENOTSOCK Not a socket
+ * fp_lookup:EBADF Bad file descriptor
+ *
+ * Implicit returns:
+ * *flags (modified) Returned flags field
+ *
+ * Locks: This function internally takes and drops the proc_fdlock for
+ * the current process
+ *
+ * Notes: This function will internally increment and decrement the
+ * f_iocount of the fileproc as part of its operation.
+ */
+int
+file_flags(int fd, int *flags)
+{
+ proc_t p = current_proc();
+ struct fileproc *fp;
+ int error;
+
+ proc_fdlock_spin(p);
+ if ((error = fp_lookup(p, fd, &fp, 1))) {
+ proc_fdunlock(p);
+ return error;
+ }
+ *flags = (int)fp->f_flag;
+ fp_drop(p, fd, fp, 1);
+ proc_fdunlock(p);
+
+ return 0;
+}
+
+
+/*
+ * file_drop
+ *
+ * Description: Drop an iocount reference on an fd, and wake up any waiters
+ * for draining (i.e. blocked in fileproc_drain() called during
+ * the last attempt to close a file).
+ *
+ * Parameters: fd fd on which an ioreference is
+ * to be dropped
+ *
+ * Returns: 0 Success
+ * EBADF Bad file descriptor
+ *
+ * Description: Given an fd, look it up in the current process's per process
+ * open file table, and drop it's fileproc's f_iocount by one
+ *
+ * Notes: This is intended as a corresponding operation to the functions
+ * file_vnode() and file_socket() operations.
+ *
+ * Technically, the close reference is supposed to be protected
+ * by a fileproc_drain(), however, a drain will only block if
+ * the fd refers to a character device, and that device has had
+ * preparefileread() called on it. If it refers to something
+ * other than a character device, then the drain will occur and
+ * block each close attempt, rather than merely the last close.
+ *
+ * Since it's possible for an fd that refers to a character
+ * device to have an intermediate close followed by an open to
+ * cause a different file to correspond to that descriptor,
+ * unless there was a cautionary reference taken on the fileproc,
+ * this is an inherently unsafe function. This happens in the
+ * case where multiple fd's in a process refer to the same
+ * character device (e.g. stdin/out/err pointing to a tty, etc.).
+ *
+ * Use of this function is discouraged.
+ */
+int
+file_drop(int fd)
+{
+ struct fileproc *fp;
+ proc_t p = current_proc();
+ int needwakeup = 0;
+
+ proc_fdlock_spin(p);
+ if (fd < 0 || fd >= p->p_fd->fd_nfiles ||
+ (fp = p->p_fd->fd_ofiles[fd]) == NULL ||
+ ((p->p_fd->fd_ofileflags[fd] & UF_RESERVED) &&
+ !(p->p_fd->fd_ofileflags[fd] & UF_CLOSING))) {
+ proc_fdunlock(p);
+ return EBADF;
+ }
+
+ if (1 == os_ref_release_locked(&fp->f_iocount)) {
+ if (fp->f_flags & FP_SELCONFLICT) {
+ fp->f_flags &= ~FP_SELCONFLICT;
+ }
+
+ if (p->p_fpdrainwait) {
+ p->p_fpdrainwait = 0;
+ needwakeup = 1;
+ }
+ }
+ proc_fdunlock(p);
+
+ if (needwakeup) {
+ wakeup(&p->p_fpdrainwait);
+ }
+ return 0;
+}
+
+
+static int falloc_withalloc_locked(proc_t, struct fileproc **, int *,
+ vfs_context_t, struct fileproc * (*)(void *), void *, int);
+
+/*
+ * falloc
+ *
+ * Description: Allocate an entry in the per process open file table and
+ * return the corresponding fileproc and fd.
+ *
+ * Parameters: p The process in whose open file
+ * table the fd is to be allocated
+ * resultfp Pointer to fileproc pointer
+ * return area
+ * resultfd Pointer to fd return area
+ * ctx VFS context
+ *
+ * Returns: 0 Success
+ * falloc:ENFILE Too many open files in system
+ * falloc:EMFILE Too many open files in process
+ * falloc:ENOMEM M_FILEPROC or M_FILEGLOB zone
+ * exhausted
+ *
+ * Implicit returns:
+ * *resultfd (modified) Returned fileproc pointer
+ * *resultfd (modified) Returned fd
+ *
+ * Locks: This function takes and drops the proc_fdlock; if this lock
+ * is already held, use falloc_locked() instead.
+ *
+ * Notes: This function takes separate process and context arguments
+ * solely to support kern_exec.c; otherwise, it would take
+ * neither, and expect falloc_locked() to use the
+ * vfs_context_current() routine internally.
+ */
+int
+falloc(proc_t p, struct fileproc **resultfp, int *resultfd, vfs_context_t ctx)
+{
+ return falloc_withalloc(p, resultfp, resultfd, ctx,
+ fileproc_alloc_init, NULL);
+}
+
+/*
+ * Like falloc, but including the fileproc allocator and create-args
+ */
+int
+falloc_withalloc(proc_t p, struct fileproc **resultfp, int *resultfd,
+ vfs_context_t ctx, fp_allocfn_t fp_zalloc, void *arg)
+{
+ int error;
+
+ proc_fdlock(p);
+ error = falloc_withalloc_locked(p,
+ resultfp, resultfd, ctx, fp_zalloc, arg, 1);
+ proc_fdunlock(p);
+
+ return error;
+}
+
+/*
+ * "uninitialized" ops -- ensure fg->fg_ops->fo_type always exists
+ */
+static const struct fileops uninitops;
+
+/*
+ * falloc_locked
+ *
+ * Create a new open file structure and allocate
+ * a file descriptor for the process that refers to it.
+ *
+ * Returns: 0 Success
+ *
+ * Description: Allocate an entry in the per process open file table and
+ * return the corresponding fileproc and fd.
+ *
+ * Parameters: p The process in whose open file
+ * table the fd is to be allocated
+ * resultfp Pointer to fileproc pointer
+ * return area
+ * resultfd Pointer to fd return area
+ * ctx VFS context
+ * locked Flag to indicate whether the
+ * caller holds proc_fdlock
+ *
+ * Returns: 0 Success
+ * ENFILE Too many open files in system
+ * fdalloc:EMFILE Too many open files in process
+ * ENOMEM M_FILEPROC or M_FILEGLOB zone
+ * exhausted
+ * fdalloc:ENOMEM
+ *
+ * Implicit returns:
+ * *resultfd (modified) Returned fileproc pointer
+ * *resultfd (modified) Returned fd
+ *
+ * Locks: If the parameter 'locked' is zero, this function takes and
+ * drops the proc_fdlock; if non-zero, the caller must hold the
+ * lock.
+ *
+ * Notes: If you intend to use a non-zero 'locked' parameter, use the
+ * utility function falloc() instead.
+ *
+ * This function takes separate process and context arguments
+ * solely to support kern_exec.c; otherwise, it would take
+ * neither, and use the vfs_context_current() routine internally.
+ */
+int
+falloc_locked(proc_t p, struct fileproc **resultfp, int *resultfd,
+ vfs_context_t ctx, int locked)
+{
+ return falloc_withalloc_locked(p, resultfp, resultfd, ctx,
+ fileproc_alloc_init, NULL, locked);
+}
+
+static int
+falloc_withalloc_locked(proc_t p, struct fileproc **resultfp, int *resultfd,
+ vfs_context_t ctx, fp_allocfn_t fp_zalloc, void *crarg,
+ int locked)
+{
+ struct fileproc *fp;
+ struct fileglob *fg;
+ int error, nfd;
+
+ if (nfiles >= maxfiles) {
+ tablefull("file");
+ return ENFILE;
+ }
+
+ if (!locked) {
+ proc_fdlock(p);
+ }
+
+ if ((error = fdalloc(p, 0, &nfd))) {
+ if (!locked) {
+ proc_fdunlock(p);
+ }
+ return error;
+ }
+
+#if CONFIG_MACF
+ error = mac_file_check_create(proc_ucred(p));
+ if (error) {
+ if (!locked) {
+ proc_fdunlock(p);
+ }
+ return error;
+ }
+#endif
+
+ /*
+ * Allocate a new file descriptor.
+ * If the process has file descriptor zero open, add to the list
+ * of open files at that point, otherwise put it at the front of
+ * the list of open files.
+ */
+ proc_fdunlock(p);
+
+ fp = (*fp_zalloc)(crarg);
+ if (fp == NULL) {
+ if (locked) {
+ proc_fdlock(p);
+ }
+ return ENOMEM;
+ }
+ MALLOC_ZONE(fg, struct fileglob *, sizeof(struct fileglob), M_FILEGLOB, M_WAITOK);
+ if (fg == NULL) {
+ fileproc_free(fp);
+ if (locked) {
+ proc_fdlock(p);
+ }
+ return ENOMEM;
+ }
+ bzero(fg, sizeof(struct fileglob));
+ lck_mtx_init(&fg->fg_lock, file_lck_grp, file_lck_attr);
+
+ os_ref_retain_locked(&fp->f_iocount);
+ fg->fg_count = 1;
+ fg->fg_ops = &uninitops;
+ fp->f_fglob = fg;
+#if CONFIG_MACF
+ mac_file_label_init(fg);
+#endif
+
+ kauth_cred_ref(ctx->vc_ucred);
+
+ proc_fdlock(p);
+
+ fp->f_cred = ctx->vc_ucred;
+
+#if CONFIG_MACF
+ mac_file_label_associate(fp->f_cred, fg);
+#endif
+
+ OSAddAtomic(1, &nfiles);
+
+ p->p_fd->fd_ofiles[nfd] = fp;
+
+ if (!locked) {
+ proc_fdunlock(p);
+ }
+
+ if (resultfp) {
+ *resultfp = fp;
+ }
+ if (resultfd) {
+ *resultfd = nfd;
+ }
+
+ return 0;
+}
+
+
+/*
+ * fg_free
+ *
+ * Description: Free a file structure; drop the global open file count, and
+ * drop the credential reference, if the fileglob has one, and
+ * destroy the instance mutex before freeing
+ *
+ * Parameters: fg Pointer to fileglob to be
+ * freed
+ *
+ * Returns: void
+ */
+void
+fg_free(struct fileglob *fg)
+{
+ OSAddAtomic(-1, &nfiles);
+
+ if (fg->fg_vn_data) {
+ fg_vn_data_free(fg->fg_vn_data);
+ fg->fg_vn_data = NULL;
+ }
+
+ if (IS_VALID_CRED(fg->fg_cred)) {
+ kauth_cred_unref(&fg->fg_cred);
+ }
+ lck_mtx_destroy(&fg->fg_lock, file_lck_grp);
+
+#if CONFIG_MACF
+ mac_file_label_destroy(fg);
+#endif
+ FREE_ZONE(fg, sizeof *fg, M_FILEGLOB);
+}
+
+
+/*
+ * fg_get_vnode
+ *
+ * Description: Return vnode associated with the file structure, if
+ * any. The lifetime of the returned vnode is bound to
+ * the lifetime of the file structure.
+ *
+ * Parameters: fg Pointer to fileglob to
+ * inspect
+ *
+ * Returns: vnode_t
+ */
+vnode_t
+fg_get_vnode(struct fileglob *fg)
+{
+ if (FILEGLOB_DTYPE(fg) == DTYPE_VNODE) {
+ return (vnode_t)fg->fg_data;
+ } else {
+ return NULL;
+ }
+}
+
+/*
+ * fdexec
+ *
+ * Description: Perform close-on-exec processing for all files in a process
+ * that are either marked as close-on-exec, or which were in the
+ * process of being opened at the time of the execve
+ *
+ * Also handles the case (via posix_spawn()) where -all-
+ * files except those marked with "inherit" as treated as
+ * close-on-exec.
+ *
+ * Parameters: p Pointer to process calling
+ * execve
+ *
+ * Returns: void
+ *
+ * Locks: This function internally takes and drops proc_fdlock()
+ * But assumes tables don't grow/change while unlocked.
+ *
+ */
+void
+fdexec(proc_t p, short flags, int self_exec)
+{
+ struct filedesc *fdp = p->p_fd;
+ int i;
+ boolean_t cloexec_default = (flags & POSIX_SPAWN_CLOEXEC_DEFAULT) != 0;
+ thread_t self = current_thread();
+ struct uthread *ut = get_bsdthread_info(self);
+ struct kqworkq *dealloc_kqwq = NULL;
projects / apple / xnu.git / blobdiff