Please note that diffs are not public domain; they are subject to the copyright notices on the relevant files. =================================================================== RCS file: /ftp/cvs/cvsroot/src/sys/miscfs/kernfs/kernfs_vnops.c,v rcsdiff: /ftp/cvs/cvsroot/src/sys/miscfs/kernfs/kernfs_vnops.c,v: warning: Unknown phrases like `commitid ...;' are present. retrieving revision 1.10 retrieving revision 1.128 diff -u -p -r1.10 -r1.128 --- src/sys/miscfs/kernfs/kernfs_vnops.c 1993/06/07 05:25:25 1.10 +++ src/sys/miscfs/kernfs/kernfs_vnops.c 2006/11/16 01:33:38 1.128 @@ -1,8 +1,10 @@ +/* $NetBSD: kernfs_vnops.c,v 1.128 2006/11/16 01:33:38 christos Exp $ */ + /* - * Copyright (c) 1990, 1992 Jan-Simon Pendry - * All rights reserved. + * Copyright (c) 1992, 1993 + * The Regents of the University of California. All rights reserved. * - * This code is derived from software contributed to Berkeley by + * This code is derived from software donated to Berkeley by * Jan-Simon Pendry. * * Redistribution and use in source and binary forms, with or without @@ -13,11 +15,7 @@ * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. - * 3. All advertising materials mentioning features or use of this software - * must display the following acknowledgement: - * This product includes software developed by the University of - * California, Berkeley and its contributors. - * 4. Neither the name of the University nor the names of its contributors + * 3. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * @@ -33,768 +31,1529 @@ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * - * $Id: kernfs_vnops.c,v 1.10 1993/06/07 05:25:25 cgd Exp $ + * @(#)kernfs_vnops.c 8.15 (Berkeley) 5/21/95 */ /* - * Kernel parameter filesystem + * Kernel parameter filesystem (/kern) */ -#include "param.h" -#include "systm.h" -#include "kernel.h" -#include "types.h" -#include "time.h" -#include "proc.h" -#include "file.h" -#include "vnode.h" -#include "stat.h" -#include "mount.h" -#include "namei.h" -#include "buf.h" -#include "miscfs/kernfs/kernfs.h" +#include +__KERNEL_RCSID(0, "$NetBSD: kernfs_vnops.c,v 1.128 2006/11/16 01:33:38 christos Exp $"); + +#ifdef _KERNEL_OPT +#include "opt_ipsec.h" +#endif + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include +#include + +#ifdef IPSEC +#include +#include +#include +#include +#include +#endif + +#include -#include "../ufs/dir.h" /* For readdir() XXX */ +#define KSTRING 256 /* Largest I/O available via this filesystem */ +#define UIO_MX 32 -struct kernfs_target kernfs_targets[] = { +#define READ_MODE (S_IRUSR|S_IRGRP|S_IROTH) +#define WRITE_MODE (S_IWUSR|S_IRUSR|S_IRGRP|S_IROTH) +#define UREAD_MODE (S_IRUSR) +#define DIR_MODE (S_IRUSR|S_IXUSR|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH) +#define UDIR_MODE (S_IRUSR|S_IXUSR) + +#define N(s) sizeof(s)-1, s +const struct kern_target kern_targets[] = { +/* NOTE: The name must be less than UIO_MX-16 chars in length */ + /* name data tag type ro/rw */ + { DT_DIR, N("."), 0, KFSkern, VDIR, DIR_MODE }, + { DT_DIR, N(".."), 0, KFSroot, VDIR, DIR_MODE }, + { DT_REG, N("boottime"), &boottime.tv_sec, KFSint, VREG, READ_MODE }, + /* XXXUNCONST */ + { DT_REG, N("copyright"), __UNCONST(copyright), + KFSstring, VREG, READ_MODE }, + { DT_REG, N("hostname"), 0, KFShostname, VREG, WRITE_MODE }, + { DT_REG, N("hz"), &hz, KFSint, VREG, READ_MODE }, +#ifdef IPSEC + { DT_DIR, N("ipsecsa"), 0, KFSipsecsadir, VDIR, UDIR_MODE }, + { DT_DIR, N("ipsecsp"), 0, KFSipsecspdir, VDIR, UDIR_MODE }, +#endif + { DT_REG, N("loadavg"), 0, KFSavenrun, VREG, READ_MODE }, + { DT_REG, N("msgbuf"), 0, KFSmsgbuf, VREG, READ_MODE }, + { DT_REG, N("pagesize"), &uvmexp.pagesize, KFSint, VREG, READ_MODE }, + { DT_REG, N("physmem"), &physmem, KFSint, VREG, READ_MODE }, +#if 0 + { DT_DIR, N("root"), 0, KFSnull, VDIR, DIR_MODE }, +#endif + { DT_BLK, N("rootdev"), &rootdev, KFSdevice, VBLK, READ_MODE }, + { DT_CHR, N("rrootdev"), &rrootdev, KFSdevice, VCHR, READ_MODE }, + { DT_REG, N("time"), 0, KFStime, VREG, READ_MODE }, + /* XXXUNCONST */ + { DT_REG, N("version"), __UNCONST(version), + KFSstring, VREG, READ_MODE }, +}; +const struct kern_target subdir_targets[] = { /* NOTE: The name must be less than UIO_MX-16 chars in length */ -DIR_TARGET(".", 0, KTT_NULL, KTM_DIR_PERMS ) -DIR_TARGET("..", 0, KTT_NULL, KTM_DIR_PERMS ) -REG_TARGET("copyright", copyright, KTT_STRING, KTM_RO_PERMS ) -REG_TARGET("hostname", 0, KTT_HOSTNAME, KTM_RW_PERMS ) -REG_TARGET("hz", &hz, KTT_INT, KTM_RO_PERMS ) -REG_TARGET("loadavg", 0, KTT_AVENRUN, KTM_RO_PERMS ) -REG_TARGET("physmem", &physmem, KTT_INT, KTM_RO_PERMS ) -DIR_TARGET("root", 0, KTT_NULL, KTM_DIR_PERMS ) -BLK_TARGET("rootdev", 0, KTT_NULL, KTM_RO_PERMS ) -CHR_TARGET("rrootdev", 0, KTT_NULL, KTM_RO_PERMS ) -REG_TARGET("time", 0, KTT_TIME, KTM_RO_PERMS ) -REG_TARGET("version", version, KTT_STRING, KTM_RO_PERMS ) + /* name data tag type ro/rw */ + { DT_DIR, N("."), 0, KFSsubdir, VDIR, DIR_MODE }, + { DT_DIR, N(".."), 0, KFSkern, VDIR, DIR_MODE }, }; +#ifdef IPSEC +const struct kern_target ipsecsa_targets[] = { +/* NOTE: The name must be less than UIO_MX-16 chars in length */ + /* name data tag type ro/rw */ + { DT_DIR, N("."), 0, KFSipsecsadir, VDIR, DIR_MODE }, + { DT_DIR, N(".."), 0, KFSkern, VDIR, DIR_MODE }, +}; +const struct kern_target ipsecsp_targets[] = { +/* NOTE: The name must be less than UIO_MX-16 chars in length */ + /* name data tag type ro/rw */ + { DT_DIR, N("."), 0, KFSipsecspdir, VDIR, DIR_MODE }, + { DT_DIR, N(".."), 0, KFSkern, VDIR, DIR_MODE }, +}; +const struct kern_target ipsecsa_kt = + { DT_DIR, N(""), 0, KFSipsecsa, VREG, UREAD_MODE }; +const struct kern_target ipsecsp_kt = + { DT_DIR, N(""), 0, KFSipsecsp, VREG, UREAD_MODE }; +#endif +#undef N +SIMPLEQ_HEAD(,dyn_kern_target) dyn_kern_targets = + SIMPLEQ_HEAD_INITIALIZER(dyn_kern_targets); +int nkern_targets = sizeof(kern_targets) / sizeof(kern_targets[0]); +const int static_nkern_targets = sizeof(kern_targets) / sizeof(kern_targets[0]); +#ifdef IPSEC +int nipsecsa_targets = sizeof(ipsecsa_targets) / sizeof(ipsecsa_targets[0]); +int nipsecsp_targets = sizeof(ipsecsp_targets) / sizeof(ipsecsp_targets[0]); +int nkern_dirs = 4; /* 2 extra subdirs */ +#else +int nkern_dirs = 2; +#endif + +int kernfs_try_fileop(kfstype, kfsfileop, void *, int); +int kernfs_try_xread(kfstype, const struct kernfs_node *, char **, + size_t, int); +int kernfs_try_xwrite(kfstype, const struct kernfs_node *, char *, + size_t, int); + +static int kernfs_default_xread(void *v); +static int kernfs_default_xwrite(void *v); +static int kernfs_default_fileop_getattr(void *); + +/* must include all fileop's */ +const struct kernfs_fileop kernfs_default_fileops[] = { + { .kf_fileop = KERNFS_XREAD }, + { .kf_fileop = KERNFS_XWRITE }, + { .kf_fileop = KERNFS_FILEOP_OPEN }, + { .kf_fileop = KERNFS_FILEOP_GETATTR, + .kf_vop = kernfs_default_fileop_getattr }, + { .kf_fileop = KERNFS_FILEOP_IOCTL }, + { .kf_fileop = KERNFS_FILEOP_CLOSE }, + { .kf_fileop = KERNFS_FILEOP_READ, + .kf_vop = kernfs_default_xread }, + { .kf_fileop = KERNFS_FILEOP_WRITE, + .kf_vop = kernfs_default_xwrite }, +}; + +int kernfs_lookup(void *); +#define kernfs_create genfs_eopnotsupp +#define kernfs_mknod genfs_eopnotsupp +int kernfs_open(void *); +int kernfs_close(void *); +int kernfs_access(void *); +int kernfs_getattr(void *); +int kernfs_setattr(void *); +int kernfs_read(void *); +int kernfs_write(void *); +#define kernfs_fcntl genfs_fcntl +int kernfs_ioctl(void *); +#define kernfs_poll genfs_poll +#define kernfs_revoke genfs_revoke +#define kernfs_fsync genfs_nullop +#define kernfs_seek genfs_nullop +#define kernfs_remove genfs_eopnotsupp +int kernfs_link(void *); +#define kernfs_rename genfs_eopnotsupp +#define kernfs_mkdir genfs_eopnotsupp +#define kernfs_rmdir genfs_eopnotsupp +int kernfs_symlink(void *); +int kernfs_readdir(void *); +#define kernfs_readlink genfs_eopnotsupp +#define kernfs_abortop genfs_abortop +int kernfs_inactive(void *); +int kernfs_reclaim(void *); +#define kernfs_lock genfs_lock +#define kernfs_unlock genfs_unlock +#define kernfs_bmap genfs_badop +#define kernfs_strategy genfs_badop +int kernfs_print(void *); +#define kernfs_islocked genfs_islocked +int kernfs_pathconf(void *); +#define kernfs_advlock genfs_einval +#define kernfs_bwrite genfs_eopnotsupp +#define kernfs_putpages genfs_putpages + +static int kernfs_xread(struct kernfs_node *, int, char **, + size_t, size_t *); +static int kernfs_xwrite(const struct kernfs_node *, char *, size_t); + +int (**kernfs_vnodeop_p)(void *); +const struct vnodeopv_entry_desc kernfs_vnodeop_entries[] = { + { &vop_default_desc, vn_default_error }, + { &vop_lookup_desc, kernfs_lookup }, /* lookup */ + { &vop_create_desc, kernfs_create }, /* create */ + { &vop_mknod_desc, kernfs_mknod }, /* mknod */ + { &vop_open_desc, kernfs_open }, /* open */ + { &vop_close_desc, kernfs_close }, /* close */ + { &vop_access_desc, kernfs_access }, /* access */ + { &vop_getattr_desc, kernfs_getattr }, /* getattr */ + { &vop_setattr_desc, kernfs_setattr }, /* setattr */ + { &vop_read_desc, kernfs_read }, /* read */ + { &vop_write_desc, kernfs_write }, /* write */ + { &vop_fcntl_desc, kernfs_fcntl }, /* fcntl */ + { &vop_ioctl_desc, kernfs_ioctl }, /* ioctl */ + { &vop_poll_desc, kernfs_poll }, /* poll */ + { &vop_revoke_desc, kernfs_revoke }, /* revoke */ + { &vop_fsync_desc, kernfs_fsync }, /* fsync */ + { &vop_seek_desc, kernfs_seek }, /* seek */ + { &vop_remove_desc, kernfs_remove }, /* remove */ + { &vop_link_desc, kernfs_link }, /* link */ + { &vop_rename_desc, kernfs_rename }, /* rename */ + { &vop_mkdir_desc, kernfs_mkdir }, /* mkdir */ + { &vop_rmdir_desc, kernfs_rmdir }, /* rmdir */ + { &vop_symlink_desc, kernfs_symlink }, /* symlink */ + { &vop_readdir_desc, kernfs_readdir }, /* readdir */ + { &vop_readlink_desc, kernfs_readlink }, /* readlink */ + { &vop_abortop_desc, kernfs_abortop }, /* abortop */ + { &vop_inactive_desc, kernfs_inactive }, /* inactive */ + { &vop_reclaim_desc, kernfs_reclaim }, /* reclaim */ + { &vop_lock_desc, kernfs_lock }, /* lock */ + { &vop_unlock_desc, kernfs_unlock }, /* unlock */ + { &vop_bmap_desc, kernfs_bmap }, /* bmap */ + { &vop_strategy_desc, kernfs_strategy }, /* strategy */ + { &vop_print_desc, kernfs_print }, /* print */ + { &vop_islocked_desc, kernfs_islocked }, /* islocked */ + { &vop_pathconf_desc, kernfs_pathconf }, /* pathconf */ + { &vop_advlock_desc, kernfs_advlock }, /* advlock */ + { &vop_bwrite_desc, kernfs_bwrite }, /* bwrite */ + { &vop_putpages_desc, kernfs_putpages }, /* putpages */ + { NULL, NULL } +}; +const struct vnodeopv_desc kernfs_vnodeop_opv_desc = + { &kernfs_vnodeop_p, kernfs_vnodeop_entries }; + +static inline int +kernfs_fileop_compare(struct kernfs_fileop *a, struct kernfs_fileop *b) +{ + if (a->kf_type < b->kf_type) + return -1; + if (a->kf_type > b->kf_type) + return 1; + if (a->kf_fileop < b->kf_fileop) + return -1; + if (a->kf_fileop > b->kf_fileop) + return 1; + return (0); +} + +SPLAY_HEAD(kfsfileoptree, kernfs_fileop) kfsfileoptree = + SPLAY_INITIALIZER(kfsfileoptree); +SPLAY_PROTOTYPE(kfsfileoptree, kernfs_fileop, kf_node, kernfs_fileop_compare); +SPLAY_GENERATE(kfsfileoptree, kernfs_fileop, kf_node, kernfs_fileop_compare); + +kfstype +kernfs_alloctype(int nkf, const struct kernfs_fileop *kf) +{ + static u_char nextfreetype = KFSlasttype; + struct kernfs_fileop *dkf, *fkf, skf; + int i; + + /* XXX need to keep track of dkf's memory if we support + deallocating types */ + dkf = malloc(sizeof(kernfs_default_fileops), M_TEMP, M_WAITOK); + memcpy(dkf, kernfs_default_fileops, sizeof(kernfs_default_fileops)); + + for (i = 0; i < sizeof(kernfs_default_fileops) / + sizeof(kernfs_default_fileops[0]); i++) { + dkf[i].kf_type = nextfreetype; + SPLAY_INSERT(kfsfileoptree, &kfsfileoptree, &dkf[i]); + } + + for (i = 0; i < nkf; i++) { + skf.kf_type = nextfreetype; + skf.kf_fileop = kf[i].kf_fileop; + if ((fkf = SPLAY_FIND(kfsfileoptree, &kfsfileoptree, &skf))) + fkf->kf_vop = kf[i].kf_vop; + } -static int nkernfs_targets = sizeof(kernfs_targets) / sizeof(kernfs_targets[0]); + return nextfreetype++; +} + +int +kernfs_try_fileop(kfstype type, kfsfileop fileop, void *v, int error) +{ + struct kernfs_fileop *kf, skf; + + skf.kf_type = type; + skf.kf_fileop = fileop; + if ((kf = SPLAY_FIND(kfsfileoptree, &kfsfileoptree, &skf))) + if (kf->kf_vop) + return kf->kf_vop(v); + return error; +} + +int +kernfs_try_xread(kfstype type, const struct kernfs_node *kfs, char **bfp, + size_t len, int error) +{ + struct kernfs_fileop *kf, skf; + + skf.kf_type = type; + skf.kf_fileop = KERNFS_XREAD; + if ((kf = SPLAY_FIND(kfsfileoptree, &kfsfileoptree, &skf))) + if (kf->kf_xread) + return kf->kf_xread(kfs, bfp, len); + return error; +} + +int +kernfs_try_xwrite(kfstype type, const struct kernfs_node *kfs, char *bf, + size_t len, int error) +{ + struct kernfs_fileop *kf, skf; + + skf.kf_type = type; + skf.kf_fileop = KERNFS_XWRITE; + if ((kf = SPLAY_FIND(kfsfileoptree, &kfsfileoptree, &skf))) + if (kf->kf_xwrite) + return kf->kf_xwrite(kfs, bf, len); + return error; +} + +int +kernfs_addentry(kernfs_parentdir_t *pkt, kernfs_entry_t *dkt) +{ + struct kernfs_subdir *ks, *parent; + + if (pkt == NULL) { + SIMPLEQ_INSERT_TAIL(&dyn_kern_targets, dkt, dkt_queue); + nkern_targets++; + if (dkt->dkt_kt.kt_vtype == VDIR) + nkern_dirs++; + } else { + parent = (struct kernfs_subdir *)pkt->kt_data; + SIMPLEQ_INSERT_TAIL(&parent->ks_entries, dkt, dkt_queue); + parent->ks_nentries++; + if (dkt->dkt_kt.kt_vtype == VDIR) + parent->ks_dirs++; + } + if (dkt->dkt_kt.kt_vtype == VDIR && dkt->dkt_kt.kt_data == NULL) { + ks = malloc(sizeof(struct kernfs_subdir), + M_TEMP, M_WAITOK); + SIMPLEQ_INIT(&ks->ks_entries); + ks->ks_nentries = 2; /* . and .. */ + ks->ks_dirs = 2; + ks->ks_parent = pkt ? pkt : &kern_targets[0]; + dkt->dkt_kt.kt_data = ks; + } + return 0; +} static int -kernfs_xread(kt, buf, len, lenp) - struct kernfs_target *kt; - char *buf; - int len; - int *lenp; +kernfs_xread(kfs, off, bufp, len, wrlen) + struct kernfs_node *kfs; + int off; + char **bufp; + size_t len; + size_t *wrlen; { - int xlen; + const struct kern_target *kt; +#ifdef IPSEC + struct mbuf *m; +#endif + int err; - switch (kt->kt_tag) { - case KTT_TIME: { + kt = kfs->kfs_kt; + + switch (kfs->kfs_type) { + case KFStime: { struct timeval tv; + microtime(&tv); - sprintf(buf, "%d %d\n", tv.tv_sec, tv.tv_usec); + snprintf(*bufp, len, "%ld %ld\n", tv.tv_sec, tv.tv_usec); break; } - case KTT_INT: { + case KFSint: { int *ip = kt->kt_data; - sprintf(buf, "%d\n", *ip); + + snprintf(*bufp, len, "%d\n", *ip); break; } - case KTT_STRING: { + case KFSstring: { char *cp = kt->kt_data; - int xlen = strlen(cp) + 1; - if (xlen >= len) - return (EINVAL); - - bcopy(cp, buf, xlen); + *bufp = cp; break; } - case KTT_HOSTNAME: { + case KFSmsgbuf: { + long n; + + /* + * deal with cases where the message buffer has + * become corrupted. + */ + if (!msgbufenabled || msgbufp->msg_magic != MSG_MAGIC) { + msgbufenabled = 0; + return (ENXIO); + } + + /* + * Note that reads of /kern/msgbuf won't necessarily yield + * consistent results, if the message buffer is modified + * while the read is in progress. The worst that can happen + * is that incorrect data will be read. There's no way + * that this can crash the system unless the values in the + * message buffer header are corrupted, but that'll cause + * the system to die anyway. + */ + if (off >= msgbufp->msg_bufs) { + *wrlen = 0; + return (0); + } + n = msgbufp->msg_bufx + off; + if (n >= msgbufp->msg_bufs) + n -= msgbufp->msg_bufs; + len = min(msgbufp->msg_bufs - n, msgbufp->msg_bufs - off); + *bufp = msgbufp->msg_bufc + n; + *wrlen = len; + return (0); + } + + case KFShostname: { char *cp = hostname; - int xlen = hostnamelen; + size_t xlen = hostnamelen; - if (xlen + 2 > len) /* extra space for null and newline */ + if (xlen >= (len - 2)) return (EINVAL); - bcopy(cp, buf, xlen); /* safer than sprintf */ - buf[xlen] = '\n'; - buf[xlen+1] = '\0'; + memcpy(*bufp, cp, xlen); + (*bufp)[xlen] = '\n'; + (*bufp)[xlen+1] = '\0'; + len = strlen(*bufp); break; } - case KTT_AVENRUN: - sprintf(buf, "%d %d %d %d\n", - averunnable[0], - averunnable[1], - averunnable[2], - FSCALE); + case KFSavenrun: + averunnable.fscale = FSCALE; + snprintf(*bufp, len, "%d %d %d %ld\n", + averunnable.ldavg[0], averunnable.ldavg[1], + averunnable.ldavg[2], averunnable.fscale); break; +#ifdef IPSEC + case KFSipsecsa: + /* + * Note that SA configuration could be changed during the + * read operation, resulting in garbled output. + */ + m = key_setdumpsa_spi(htonl(kfs->kfs_value)); + if (!m) + return (ENOBUFS); + if (off >= m->m_pkthdr.len) { + *wrlen = 0; + m_freem(m); + return (0); + } + if (len > m->m_pkthdr.len - off) + len = m->m_pkthdr.len - off; + m_copydata(m, off, len, *bufp); + *wrlen = len; + m_freem(m); + return (0); + + case KFSipsecsp: + /* + * Note that SP configuration could be changed during the + * read operation, resulting in garbled output. + */ + if (!kfs->kfs_v) { + struct secpolicy *sp; + + sp = key_getspbyid(kfs->kfs_value); + if (sp) + kfs->kfs_v = sp; + else + return (ENOENT); + } + m = key_setdumpsp((struct secpolicy *)kfs->kfs_v, + SADB_X_SPDGET, 0, 0); + if (!m) + return (ENOBUFS); + if (off >= m->m_pkthdr.len) { + *wrlen = 0; + m_freem(m); + return (0); + } + if (len > m->m_pkthdr.len - off) + len = m->m_pkthdr.len - off; + m_copydata(m, off, len, *bufp); + *wrlen = len; + m_freem(m); + return (0); +#endif + default: - return (EINVAL); + err = kernfs_try_xread(kfs->kfs_type, kfs, bufp, len, + EOPNOTSUPP); + if (err) + return err; } - *lenp = strlen(buf); + len = strlen(*bufp); + if (len <= off) + *wrlen = 0; + else { + *bufp += off; + *wrlen = len - off; + } return (0); } static int -kernfs_xwrite(kt, buf, len) - struct kernfs_target *kt; - char *buf; - int len; -{ - switch (kt->kt_tag) { - case KTT_HOSTNAME: { - if (buf[len-1] == '\n') +kernfs_xwrite(kfs, bf, len) + const struct kernfs_node *kfs; + char *bf; + size_t len; +{ + + switch (kfs->kfs_type) { + case KFShostname: + if (bf[len-1] == '\n') --len; - bcopy(buf, hostname, len); - /* kernfs_write set buf[value_passed_as_len] = \0. - * therefore, buf len (hostnamelen) = len. - */ - hostnamelen = len; - hostname[hostnamelen] = '\0'; /* null end of string. */ + memcpy(hostname, bf, len); + hostname[len] = '\0'; + hostnamelen = (size_t) len; return (0); - } default: - return (EIO); + return kernfs_try_xwrite(kfs->kfs_type, kfs, bf, len, EIO); } } + /* * vp is the current namei directory * ndp is the name to locate in that directory... */ -kernfs_lookup(dvp, ndp, p) - struct vnode *dvp; - struct nameidata *ndp; - struct proc *p; +int +kernfs_lookup(v) + void *v; { - char *pname = ndp->ni_ptr; - int error = ENOENT; - int i; - struct vnode *fvp; - -#ifdef KERNFS_DIAGNOSTIC - printf("kernfs_lookup(%s)\n", pname); + struct vop_lookup_args /* { + struct vnode * a_dvp; + struct vnode ** a_vpp; + struct componentname * a_cnp; + } */ *ap = v; + struct componentname *cnp = ap->a_cnp; + struct vnode **vpp = ap->a_vpp; + struct vnode *dvp = ap->a_dvp; + const char *pname = cnp->cn_nameptr; + const struct kernfs_node *kfs; + const struct kern_target *kt; + const struct dyn_kern_target *dkt; + const struct kernfs_subdir *ks; + int error, i, wantpunlock; +#ifdef IPSEC + char *ep; + u_int32_t id; #endif - if (ndp->ni_namelen == 1 && *pname == '.') { - ndp->ni_dvp = dvp; - ndp->ni_vp = dvp; + + *vpp = NULLVP; + cnp->cn_flags &= ~PDIRUNLOCK; + + if (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME) + return (EROFS); + + if (cnp->cn_namelen == 1 && *pname == '.') { + *vpp = dvp; VREF(dvp); - /*VOP_LOCK(dvp);*/ - return (0); - } - - if (ndp->ni_namelen == 4 && bcmp(pname, "root", 4) == 0) { - ndp->ni_dvp = dvp; - ndp->ni_vp = rootdir; - VREF(rootdir); - VOP_LOCK(rootdir); - return (0); - } - - /* - * /kern/rootdev is the root device - */ - if (ndp->ni_namelen == 7 && bcmp(pname, "rootdev", 7) == 0) { - if (vfinddev(rootdev, VBLK, &fvp)) - return (ENXIO); - ndp->ni_dvp = dvp; - ndp->ni_vp = fvp; - VREF(fvp); - VOP_LOCK(fvp); return (0); } - /* - * /kern/rrootdev is the root device - */ - if (ndp->ni_namelen == 8 && bcmp(pname, "rrootdev", 7) == 0) { - ndp->ni_dvp = dvp; - ndp->ni_vp = rrootdevvp; - VREF(rrootdevvp); - VOP_LOCK(rrootdevvp); - return (0); - } + wantpunlock = (~cnp->cn_flags & (LOCKPARENT | ISLASTCN)); + kfs = VTOKERN(dvp); + switch (kfs->kfs_type) { + case KFSkern: + /* + * Shouldn't get here with .. in the root node. + */ + if (cnp->cn_flags & ISDOTDOT) + return (EIO); + + for (i = 0; i < static_nkern_targets; i++) { + kt = &kern_targets[i]; + if (cnp->cn_namelen == kt->kt_namlen && + memcmp(kt->kt_name, pname, cnp->cn_namelen) == 0) + goto found; + } + SIMPLEQ_FOREACH(dkt, &dyn_kern_targets, dkt_queue) { + if (cnp->cn_namelen == dkt->dkt_kt.kt_namlen && + memcmp(dkt->dkt_kt.kt_name, pname, cnp->cn_namelen) == 0) { + kt = &dkt->dkt_kt; + goto found; + } + } + break; + + found: + error = kernfs_allocvp(dvp->v_mount, vpp, kt->kt_tag, kt, 0); + if ((error == 0) && wantpunlock) { + VOP_UNLOCK(dvp, 0); + cnp->cn_flags |= PDIRUNLOCK; + } + return (error); + + case KFSsubdir: + ks = (struct kernfs_subdir *)kfs->kfs_kt->kt_data; + if (cnp->cn_flags & ISDOTDOT) { + kt = ks->ks_parent; + goto found; + } + + SIMPLEQ_FOREACH(dkt, &ks->ks_entries, dkt_queue) { + if (cnp->cn_namelen == dkt->dkt_kt.kt_namlen && + memcmp(dkt->dkt_kt.kt_name, pname, cnp->cn_namelen) == 0) { + kt = &dkt->dkt_kt; + goto found; + } + } + break; + +#ifdef IPSEC + case KFSipsecsadir: + if (cnp->cn_flags & ISDOTDOT) { + kt = &kern_targets[0]; + goto found; + } + + for (i = 2; i < nipsecsa_targets; i++) { + kt = &ipsecsa_targets[i]; + if (cnp->cn_namelen == kt->kt_namlen && + memcmp(kt->kt_name, pname, cnp->cn_namelen) == 0) + goto found; + } - for (i = 0; i < nkernfs_targets; i++) { - struct kernfs_target *kt = &kernfs_targets[i]; - if (ndp->ni_namelen == strlen(kt->kt_name) && - bcmp(kt->kt_name, pname, ndp->ni_namelen) == 0) { - error = 0; + ep = NULL; + id = strtoul(pname, &ep, 10); + if (!ep || *ep || ep == pname) break; + + error = kernfs_allocvp(dvp->v_mount, vpp, KFSipsecsa, &ipsecsa_kt, id); + if ((error == 0) && wantpunlock) { + VOP_UNLOCK(dvp, 0); + cnp->cn_flags |= PDIRUNLOCK; } - } + return (error); -#ifdef KERNFS_DIAGNOSTIC - printf("kernfs_lookup: i = %d, error = %d\n", i, error); + case KFSipsecspdir: + if (cnp->cn_flags & ISDOTDOT) { + kt = &kern_targets[0]; + goto found; + } + + for (i = 2; i < nipsecsp_targets; i++) { + kt = &ipsecsp_targets[i]; + if (cnp->cn_namelen == kt->kt_namlen && + memcmp(kt->kt_name, pname, cnp->cn_namelen) == 0) + goto found; + } + + ep = NULL; + id = strtoul(pname, &ep, 10); + if (!ep || *ep || ep == pname) + break; + + error = kernfs_allocvp(dvp->v_mount, vpp, KFSipsecsp, &ipsecsp_kt, id); + if ((error == 0) && wantpunlock) { + VOP_UNLOCK(dvp, 0); + cnp->cn_flags |= PDIRUNLOCK; + } + return (error); #endif - if (error) - goto bad; + default: + return (ENOTDIR); + } + + return (cnp->cn_nameiop == LOOKUP ? ENOENT : EROFS); +} -#ifdef KERNFS_DIAGNOSTIC - printf("kernfs_lookup: allocate new vnode\n"); -#endif - error = getnewvnode(VT_KERNFS, dvp->v_mount, &kernfs_vnodeops, &fvp); - if (error) - goto bad; - VTOKERN(fvp)->kf_kt = &kernfs_targets[i]; - fvp->v_type = VTOKERN(fvp)->kf_kt->kt_vtype; - ndp->ni_dvp = dvp; - ndp->ni_vp = fvp; -#ifdef KERNFS_DIAGNOSTIC - printf("kernfs_lookup: newvp = %x\n", fvp); +int +kernfs_open(v) + void *v; +{ + struct vop_open_args /* { + struct vnode *a_vp; + int a_mode; + kauth_cred_t a_cred; + struct lwp *a_l; + } */ *ap = v; + struct kernfs_node *kfs = VTOKERN(ap->a_vp); +#ifdef IPSEC + struct mbuf *m; + struct secpolicy *sp; #endif - return (0); -bad:; - ndp->ni_dvp = dvp; - ndp->ni_vp = NULL; -#ifdef KERNFS_DIAGNOSTIC - printf("kernfs_lookup: error = %d\n", error); + switch (kfs->kfs_type) { +#ifdef IPSEC + case KFSipsecsa: + m = key_setdumpsa_spi(htonl(kfs->kfs_value)); + if (m) { + m_freem(m); + return (0); + } else + return (ENOENT); + + case KFSipsecsp: + sp = key_getspbyid(kfs->kfs_value); + if (sp) { + kfs->kfs_v = sp; + return (0); + } else + return (ENOENT); #endif - return (error); + + default: + return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_OPEN, + v, 0); + } } -kernfs_open(vp, mode, cred, p) - struct vnode *vp; - int mode; - struct ucred *cred; - struct proc *p; +int +kernfs_close(v) + void *v; { - /* if access succeeded, this always does, too */ + struct vop_close_args /* { + struct vnode *a_vp; + int a_fflag; + kauth_cred_t a_cred; + struct lwp *a_l; + } */ *ap = v; + struct kernfs_node *kfs = VTOKERN(ap->a_vp); + + switch (kfs->kfs_type) { +#ifdef IPSEC + case KFSipsecsp: + key_freesp((struct secpolicy *)kfs->kfs_v); + break; +#endif + + default: + return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_CLOSE, + v, 0); + } return (0); } -/* - * Check mode permission on target pointer. Mode is READ, WRITE or EXEC. - * The mode is shifted to select the owner/group/other fields. The - * super user is granted all permissions. - */ -kernfs_access(vp, mode, cred, p) - struct vnode *vp; - register int mode; - struct ucred *cred; - struct proc *p; -{ - struct kernfs_target *kt = VTOKERN(vp)->kf_kt; - register gid_t *gp; - int i, error; - -#ifdef KERN_DIAGNOSTIC - if (!VOP_ISLOCKED(vp)) { - vprint("kernfs_access: not locked", vp); - panic("kernfs_access: not locked"); - } -#endif - /* - * If you're the super-user, you always get access. - */ - if (cred->cr_uid == 0) - return (0); - /* - * Access check is based on only one of owner, group, public. - * If not owner, then check group. If not a member of the - * group, then check public access. - */ - if (cred->cr_uid != /* kt->kt_uid XXX */ 0) { - mode >>= 3; - gp = cred->cr_groups; - for (i = 0; i < cred->cr_ngroups; i++, gp++) - if (/* kt->kt_gid XXX */ 0 == *gp) - goto found; - mode >>= 3; -found: - ; - } - if ((kt->kt_perms & mode) == mode) - return (0); - return (EACCES); +int +kernfs_access(v) + void *v; +{ + struct vop_access_args /* { + struct vnode *a_vp; + int a_mode; + kauth_cred_t a_cred; + struct lwp *a_l; + } */ *ap = v; + struct vattr va; + int error; + + if ((error = VOP_GETATTR(ap->a_vp, &va, ap->a_cred, ap->a_l)) != 0) + return (error); + + return (vaccess(va.va_type, va.va_mode, va.va_uid, va.va_gid, + ap->a_mode, ap->a_cred)); } -kernfs_getattr(vp, vap, cred, p) - struct vnode *vp; - struct vattr *vap; - struct ucred *cred; - struct proc *p; +static int +kernfs_default_fileop_getattr(v) + void *v; { + struct vop_getattr_args /* { + struct vnode *a_vp; + struct vattr *a_vap; + kauth_cred_t a_cred; + struct lwp *a_l; + } */ *ap = v; + struct vattr *vap = ap->a_vap; + + vap->va_nlink = 1; + vap->va_bytes = vap->va_size = 0; + + return 0; +} + +int +kernfs_getattr(v) + void *v; +{ + struct vop_getattr_args /* { + struct vnode *a_vp; + struct vattr *a_vap; + kauth_cred_t a_cred; + struct lwp *a_l; + } */ *ap = v; + struct kernfs_node *kfs = VTOKERN(ap->a_vp); + struct kernfs_subdir *ks; + struct vattr *vap = ap->a_vap; int error = 0; - char strbuf[KSTRING]; - struct kernfs_target *kt = VTOKERN(vp)->kf_kt; + char strbuf[KSTRING], *bf; + size_t nread, total; - bzero((caddr_t) vap, sizeof(*vap)); - vattr_null(vap); - vap->va_uid = kt->kt_uid; - vap->va_gid = kt->kt_gid; - vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0]; - /* vap->va_qsize = 0; */ + VATTR_NULL(vap); + vap->va_type = ap->a_vp->v_type; + vap->va_uid = 0; + vap->va_gid = 0; + vap->va_mode = kfs->kfs_mode; + vap->va_fileid = kfs->kfs_fileno; + vap->va_flags = 0; + vap->va_size = 0; vap->va_blocksize = DEV_BSIZE; - microtime(&vap->va_atime); - vap->va_mtime = vap->va_atime; - vap->va_ctime = vap->va_ctime; + /* Make all times be current TOD, except for the "boottime" node. */ + if (kfs->kfs_kt && kfs->kfs_kt->kt_namlen == 8 && + !memcmp(kfs->kfs_kt->kt_name, "boottime", 8)) { + TIMEVAL_TO_TIMESPEC(&boottime, &vap->va_ctime); + } else { + getnanotime(&vap->va_ctime); + } + vap->va_atime = vap->va_mtime = vap->va_ctime; vap->va_gen = 0; vap->va_flags = 0; vap->va_rdev = 0; - /* vap->va_qbytes = 0; */ vap->va_bytes = 0; - vap->va_type = kt->kt_vtype; - vap->va_mode = kt->kt_perms; - if (vp->v_flag & VROOT) { -#ifdef KERNFS_DIAGNOSTIC - printf("kernfs_getattr: stat rootdir\n"); + switch (kfs->kfs_type) { + case KFSkern: + vap->va_nlink = nkern_dirs; + vap->va_bytes = vap->va_size = DEV_BSIZE; + break; + + case KFSroot: + vap->va_nlink = 1; + vap->va_bytes = vap->va_size = DEV_BSIZE; + break; + + case KFSsubdir: + ks = (struct kernfs_subdir *)kfs->kfs_kt->kt_data; + vap->va_nlink = ks->ks_dirs; + vap->va_bytes = vap->va_size = DEV_BSIZE; + break; + + case KFSnull: + case KFStime: + case KFSint: + case KFSstring: + case KFShostname: + case KFSavenrun: + case KFSdevice: + case KFSmsgbuf: +#ifdef IPSEC + case KFSipsecsa: + case KFSipsecsp: #endif + vap->va_nlink = 1; + total = 0; + do { + bf = strbuf; + error = kernfs_xread(kfs, total, &bf, + sizeof(strbuf), &nread); + total += nread; + } while (error == 0 && nread != 0); + vap->va_bytes = vap->va_size = total; + break; + +#ifdef IPSEC + case KFSipsecsadir: + case KFSipsecspdir: vap->va_nlink = 2; - vap->va_fileid = 2; - vap->va_size = DEV_BSIZE; - } else { -#ifdef KERNFS_DIAGNOSTIC - printf("kernfs_getattr: stat target %s\n", kt->kt_name); + vap->va_bytes = vap->va_size = DEV_BSIZE; + break; #endif - vap->va_nlink = 1; - vap->va_fileid = 3 + (kt - kernfs_targets) / sizeof(*kt); - error = kernfs_xread(kt, strbuf, sizeof(strbuf), &vap->va_size); + + default: + error = kernfs_try_fileop(kfs->kfs_type, + KERNFS_FILEOP_GETATTR, v, EINVAL); + break; } - vp->v_type = vap->va_type; -#ifdef KERNFS_DIAGNOSTIC - printf("kernfs_getattr: return error %d\n", error); -#endif return (error); } - -/* - * Change the mode on a file. - */ -kernfs_chmod1(vp, mode, p) - register struct vnode *vp; - register int mode; - struct proc *p; +/*ARGSUSED*/ +int +kernfs_setattr(void *v) { - register struct ucred *cred = p->p_ucred; - register struct kernfs_target *kt = VTOKERN(vp)->kf_kt; - int error; - - if ((mode & kt->kt_maxperms) != mode) /* can't set ro var to rw */ - return (EPERM); - if (cred->cr_uid != kt->kt_uid && - (error = suser(cred, &p->p_acflag))) - return (error); - if (cred->cr_uid) { - if (vp->v_type != VDIR && (mode & S_ISVTX)) - return (EFTYPE); - if (!groupmember(kt->kt_gid, cred) && (mode & S_ISGID)) - return (EPERM); - } - kt->kt_perms &= ~07777; - kt->kt_perms |= mode & 07777; -/* ip->i_flag |= ICHG;*/ - return (0); -} - -/* - * Perform chown operation on kernfs_target kt - */ -kernfs_chown1(vp, uid, gid, p) - register struct vnode *vp; - uid_t uid; - gid_t gid; - struct proc *p; -{ - register struct kernfs_target *kt = VTOKERN(vp)->kf_kt; - register struct ucred *cred = p->p_ucred; - uid_t ouid; - gid_t ogid; - int error = 0; - - if (uid == (u_short)VNOVAL) - uid = kt->kt_uid; - if (gid == (u_short)VNOVAL) - gid = kt->kt_gid; /* - * If we don't own the file, are trying to change the owner - * of the file, or are not a member of the target group, - * the caller must be superuser or the call fails. + * Silently ignore attribute changes. + * This allows for open with truncate to have no + * effect until some data is written. I want to + * do it this way because all writes are atomic. */ - if ((cred->cr_uid != kt->kt_uid || uid != kt->kt_uid || - !groupmember((gid_t)gid, cred)) && - (error = suser(cred, &p->p_acflag))) - return (error); - ouid = kt->kt_uid; - ogid = kt->kt_gid; - - kt->kt_uid = uid; - kt->kt_gid = gid; - -/* if (ouid != uid || ogid != gid) - ip->i_flag |= ICHG;*/ - if (ouid != uid && cred->cr_uid != 0) - kt->kt_perms &= ~S_ISUID; - if (ogid != gid && cred->cr_uid != 0) - kt->kt_perms &= ~S_ISGID; return (0); } -/* - * Set attribute vnode op. called from several syscalls - */ -kernfs_setattr(vp, vap, cred, p) - struct vnode *vp; - struct vattr *vap; - struct ucred *cred; - struct proc *p; +int +kernfs_default_xread(v) + void *v; { - int error = 0; + struct vop_read_args /* { + struct vnode *a_vp; + struct uio *a_uio; + int a_ioflag; + kauth_cred_t a_cred; + } */ *ap = v; + struct uio *uio = ap->a_uio; + struct kernfs_node *kfs = VTOKERN(ap->a_vp); + char strbuf[KSTRING], *bf; + int off; + size_t len; + int error; - /* - * Check for unsetable attributes. - */ - if ((vap->va_type != VNON) || (vap->va_nlink != VNOVAL) || - (vap->va_fsid != VNOVAL) || (vap->va_fileid != VNOVAL) || - (vap->va_blocksize != VNOVAL) || (vap->va_rdev != VNOVAL) || - ((int)vap->va_bytes != VNOVAL) || (vap->va_gen != VNOVAL)) { - return (EINVAL); - } - /* - * Go through the fields and update iff not VNOVAL. - */ - if (vap->va_uid != (u_short)VNOVAL || vap->va_gid != (u_short)VNOVAL) - if (error = kernfs_chown1(vp, vap->va_uid, vap->va_gid, p)) - return (error); - if (vap->va_size != VNOVAL) { - if (vp->v_type == VDIR) - return (EISDIR); - /* else just nod and smile... */ - } - if (vap->va_atime.tv_sec != VNOVAL || vap->va_mtime.tv_sec != VNOVAL) { -/* if (cred->cr_uid != ip->i_uid && - (error = suser(cred, &p->p_acflag))) - return (error); - if (vap->va_atime.tv_sec != VNOVAL) - ip->i_flag |= IACC; - if (vap->va_mtime.tv_sec != VNOVAL) - ip->i_flag |= IUPD; - ip->i_flag |= ICHG; - if (error = iupdat(ip, &vap->va_atime, &vap->va_mtime, 1)) - return (error); -*/ - } - if (vap->va_mode != (u_short)VNOVAL) - error = kernfs_chmod1(vp, (int)vap->va_mode, p); - if (vap->va_flags != VNOVAL) { -/* if (cred->cr_uid != ip->i_uid && - (error = suser(cred, &p->p_acflag))) - return (error); - if (cred->cr_uid == 0) { - ip->i_flags = vap->va_flags; - } else { - ip->i_flags &= 0xffff0000; - ip->i_flags |= (vap->va_flags & 0xffff); - } - ip->i_flag |= ICHG; -*/ - } + if (ap->a_vp->v_type == VDIR) + return (EOPNOTSUPP); + + off = (int)uio->uio_offset; + /* Don't allow negative offsets */ + if (off < 0) + return EINVAL; + + bf = strbuf; + if ((error = kernfs_xread(kfs, off, &bf, sizeof(strbuf), &len)) == 0) + error = uiomove(bf, len, uio); return (error); } -static int -kernfs_read(vp, uio, ioflag, cred) - struct vnode *vp; - struct uio *uio; - int ioflag; - struct ucred *cred; +int +kernfs_read(v) + void *v; { - struct kernfs_target *kt = VTOKERN(vp)->kf_kt; - char strbuf[KSTRING]; - int off = uio->uio_offset; - int len = 0; - char *cp = strbuf; - int error; -#ifdef KERNFS_DIAGNOSTIC - printf("kern_read %s\n", kt->kt_name); -#endif - - error = kernfs_xread(kt, strbuf, sizeof(strbuf), &len); - if (error) - return (error); - cp = strbuf + off; - len -= off; - return (uiomove(cp, len, uio)); + struct vop_read_args /* { + struct vnode *a_vp; + struct uio *a_uio; + int a_ioflag; + struct ucred *a_cred; + } */ *ap = v; + struct kernfs_node *kfs = VTOKERN(ap->a_vp); + + if (kfs->kfs_type < KFSlasttype) { + /* use default function */ + return kernfs_default_xread(v); + } + return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_READ, v, + EOPNOTSUPP); } static int -kernfs_write(vp, uio, ioflag, cred) - struct vnode *vp; - struct uio *uio; - int ioflag; - struct ucred *cred; +kernfs_default_xwrite(v) + void *v; { - struct kernfs_target *kt = VTOKERN(vp)->kf_kt; - char strbuf[KSTRING]; - int len = uio->uio_resid; - char *cp = strbuf; - int xlen; + struct vop_write_args /* { + struct vnode *a_vp; + struct uio *a_uio; + int a_ioflag; + kauth_cred_t a_cred; + } */ *ap = v; + struct kernfs_node *kfs = VTOKERN(ap->a_vp); + struct uio *uio = ap->a_uio; int error; + size_t xlen; + char strbuf[KSTRING]; if (uio->uio_offset != 0) return (EINVAL); xlen = min(uio->uio_resid, KSTRING-1); - error = uiomove(strbuf, xlen, uio); - if (error) + if ((error = uiomove(strbuf, xlen, uio)) != 0) return (error); if (uio->uio_resid != 0) return (EIO); strbuf[xlen] = '\0'; - return (kernfs_xwrite(kt, strbuf, xlen)); + xlen = strlen(strbuf); + return (kernfs_xwrite(kfs, strbuf, xlen)); +} + +int +kernfs_write(v) + void *v; +{ + struct vop_write_args /* { + struct vnode *a_vp; + struct uio *a_uio; + int a_ioflag; + kauth_cred_t a_cred; + } */ *ap = v; + struct kernfs_node *kfs = VTOKERN(ap->a_vp); + + if (kfs->kfs_type < KFSlasttype) { + /* use default function */ + return kernfs_default_xwrite(v); + } + return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_WRITE, v, + EOPNOTSUPP); } -kernfs_readdir(vp, uio, cred, eofflagp) +int +kernfs_ioctl(v) + void *v; +{ + struct vop_ioctl_args /* { + const struct vnodeop_desc *a_desc; + struct vnode *a_vp; + u_long a_command; + void *a_data; + int a_fflag; + kauth_cred_t a_cred; + struct lwp *a_l; + } */ *ap = v; + struct kernfs_node *kfs = VTOKERN(ap->a_vp); + + return kernfs_try_fileop(kfs->kfs_type, KERNFS_FILEOP_IOCTL, v, + EPASSTHROUGH); +} + +static int +kernfs_setdirentfileno_kt(struct dirent *d, const struct kern_target *kt, + u_int32_t value, struct vop_readdir_args *ap) +{ + struct kernfs_node *kfs; struct vnode *vp; - struct uio *uio; - struct ucred *cred; - int *eofflagp; + int error; + + if ((error = kernfs_allocvp(ap->a_vp->v_mount, &vp, kt->kt_tag, kt, + value)) != 0) + return error; + if (kt->kt_tag == KFSdevice) { + struct vattr va; + + error = VOP_GETATTR(vp, &va, ap->a_cred, curlwp); + if (error != 0) { + return error; + } + d->d_fileno = va.va_fileid; + } else { + kfs = VTOKERN(vp); + d->d_fileno = kfs->kfs_fileno; + } + vput(vp); + return 0; +} + +static int +kernfs_setdirentfileno(struct dirent *d, off_t entry, + struct kernfs_node *thisdir_kfs, const struct kern_target *parent_kt, + const struct kern_target *kt, struct vop_readdir_args *ap) { - struct filedesc *fdp; - int i; + const struct kern_target *ikt; int error; - i = uio->uio_offset / UIO_MX; - error = 0; - while (uio->uio_resid > 0) { -#ifdef KERNFS_DIAGNOSTIC - printf("kernfs_readdir: i = %d\n", i); + switch (entry) { + case 0: + d->d_fileno = thisdir_kfs->kfs_fileno; + return 0; + case 1: + ikt = parent_kt; + break; + default: + ikt = kt; + break; + } + if (ikt != thisdir_kfs->kfs_kt) { + if ((error = kernfs_setdirentfileno_kt(d, ikt, 0, ap)) != 0) + return error; + } else + d->d_fileno = thisdir_kfs->kfs_fileno; + return 0; +} + +int +kernfs_readdir(v) + void *v; +{ + struct vop_readdir_args /* { + struct vnode *a_vp; + struct uio *a_uio; + kauth_cred_t a_cred; + int *a_eofflag; + off_t **a_cookies; + int a_*ncookies; + } */ *ap = v; + struct uio *uio = ap->a_uio; + struct dirent d; + struct kernfs_node *kfs = VTOKERN(ap->a_vp); + const struct kern_target *kt; + const struct dyn_kern_target *dkt = NULL; + const struct kernfs_subdir *ks; + off_t i, j; + int error; + off_t *cookies = NULL; + int ncookies = 0, n; +#ifdef IPSEC + struct secasvar *sav, *sav2; + struct secpolicy *sp; #endif - if (i >= nkernfs_targets) { - *eofflagp = 1; + + if (uio->uio_resid < UIO_MX) + return (EINVAL); + if (uio->uio_offset < 0) + return (EINVAL); + + error = 0; + i = uio->uio_offset; + memset(&d, 0, sizeof(d)); + d.d_reclen = UIO_MX; + ncookies = uio->uio_resid / UIO_MX; + + switch (kfs->kfs_type) { + case KFSkern: + if (i >= nkern_targets) + return (0); + + if (ap->a_ncookies) { + ncookies = min(ncookies, (nkern_targets - i)); + cookies = malloc(ncookies * sizeof(off_t), M_TEMP, + M_WAITOK); + *ap->a_cookies = cookies; + } + + n = 0; + for (; i < nkern_targets && uio->uio_resid >= UIO_MX; i++) { + if (i < static_nkern_targets) + kt = &kern_targets[i]; + else { + if (dkt == NULL) { + dkt = SIMPLEQ_FIRST(&dyn_kern_targets); + for (j = static_nkern_targets; j < i && + dkt != NULL; j++) + dkt = SIMPLEQ_NEXT(dkt, dkt_queue); + if (j != i) + break; + } else { + dkt = SIMPLEQ_NEXT(dkt, dkt_queue); + } + if (dkt == NULL) + break; + kt = &dkt->dkt_kt; + } + if (kt->kt_tag == KFSdevice) { + dev_t *dp = kt->kt_data; + struct vnode *fvp; + + if (*dp == NODEV || + !vfinddev(*dp, kt->kt_vtype, &fvp)) + continue; + } + d.d_namlen = kt->kt_namlen; + if ((error = kernfs_setdirentfileno(&d, i, kfs, + &kern_targets[0], kt, ap)) != 0) + break; + memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1); + d.d_type = kt->kt_type; + if ((error = uiomove(&d, UIO_MX, uio)) != 0) + break; + if (cookies) + *cookies++ = i + 1; + n++; + } + ncookies = n; + break; + + case KFSroot: + if (i >= 2) + return 0; + + if (ap->a_ncookies) { + ncookies = min(ncookies, (2 - i)); + cookies = malloc(ncookies * sizeof(off_t), M_TEMP, + M_WAITOK); + *ap->a_cookies = cookies; + } + + n = 0; + for (; i < 2 && uio->uio_resid >= UIO_MX; i++) { + kt = &kern_targets[i]; + d.d_namlen = kt->kt_namlen; + d.d_fileno = KERNFS_FILENO(kt, kt->kt_tag, 0); + memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1); + d.d_type = kt->kt_type; + if ((error = uiomove(&d, UIO_MX, uio)) != 0) + break; + if (cookies) + *cookies++ = i + 1; + n++; + } + ncookies = n; + break; + + case KFSsubdir: + ks = (struct kernfs_subdir *)kfs->kfs_kt->kt_data; + if (i >= ks->ks_nentries) + return (0); + + if (ap->a_ncookies) { + ncookies = min(ncookies, (ks->ks_nentries - i)); + cookies = malloc(ncookies * sizeof(off_t), M_TEMP, + M_WAITOK); + *ap->a_cookies = cookies; + } + + dkt = SIMPLEQ_FIRST(&ks->ks_entries); + for (j = 0; j < i && dkt != NULL; j++) + dkt = SIMPLEQ_NEXT(dkt, dkt_queue); + n = 0; + for (; i < ks->ks_nentries && uio->uio_resid >= UIO_MX; i++) { + if (i < 2) + kt = &subdir_targets[i]; + else { + /* check if ks_nentries lied to us */ + if (dkt == NULL) + break; + kt = &dkt->dkt_kt; + dkt = SIMPLEQ_NEXT(dkt, dkt_queue); + } + if (kt->kt_tag == KFSdevice) { + dev_t *dp = kt->kt_data; + struct vnode *fvp; + + if (*dp == NODEV || + !vfinddev(*dp, kt->kt_vtype, &fvp)) + continue; + } + d.d_namlen = kt->kt_namlen; + if ((error = kernfs_setdirentfileno(&d, i, kfs, + ks->ks_parent, kt, ap)) != 0) + break; + memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1); + d.d_type = kt->kt_type; + if ((error = uiomove(&d, UIO_MX, uio)) != 0) + break; + if (cookies) + *cookies++ = i + 1; + n++; + } + ncookies = n; + break; + +#ifdef IPSEC + case KFSipsecsadir: + /* count SA in the system */ + n = 0; + TAILQ_FOREACH(sav, &satailq, tailq) { + for (sav2 = TAILQ_FIRST(&satailq); + sav2 != sav; + sav2 = TAILQ_NEXT(sav2, tailq)) { + if (sav->spi == sav2->spi) { + /* multiple SA with same SPI */ + break; + } + } + if (sav == sav2 || sav->spi != sav2->spi) + n++; + } + + if (i >= nipsecsa_targets + n) + return (0); + + if (ap->a_ncookies) { + ncookies = min(ncookies, (n - i)); + cookies = malloc(ncookies * sizeof(off_t), M_TEMP, + M_WAITOK); + *ap->a_cookies = cookies; + } + + n = 0; + for (; i < nipsecsa_targets && uio->uio_resid >= UIO_MX; i++) { + kt = &ipsecsa_targets[i]; + d.d_namlen = kt->kt_namlen; + if ((error = kernfs_setdirentfileno(&d, i, kfs, + &kern_targets[0], kt, ap)) != 0) + break; + memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1); + d.d_type = kt->kt_type; + if ((error = uiomove(&d, UIO_MX, uio)) != 0) + break; + if (cookies) + *cookies++ = i + 1; + n++; + } + if (error) { + ncookies = n; break; } - { - struct direct d; - struct direct *dp = &d; - struct kernfs_target *kt = &kernfs_targets[i]; - bzero((caddr_t) dp, UIO_MX); + TAILQ_FOREACH(sav, &satailq, tailq) { + for (sav2 = TAILQ_FIRST(&satailq); + sav2 != sav; + sav2 = TAILQ_NEXT(sav2, tailq)) { + if (sav->spi == sav2->spi) { + /* multiple SA with same SPI */ + break; + } + } + if (sav != sav2 && sav->spi == sav2->spi) + continue; + if (uio->uio_resid < UIO_MX) + break; + if ((error = kernfs_setdirentfileno_kt(&d, &ipsecsa_kt, + sav->spi, ap)) != 0) + break; + d.d_namlen = snprintf(d.d_name, sizeof(d.d_name), + "%u", ntohl(sav->spi)); + d.d_type = DT_REG; + if ((error = uiomove(&d, UIO_MX, uio)) != 0) + break; + if (cookies) + *cookies++ = i + 1; + n++; + i++; + } + ncookies = n; + break; + + case KFSipsecspdir: + /* count SP in the system */ + n = 0; + TAILQ_FOREACH(sp, &sptailq, tailq) + n++; + + if (i >= nipsecsp_targets + n) + return (0); + + if (ap->a_ncookies) { + ncookies = min(ncookies, (n - i)); + cookies = malloc(ncookies * sizeof(off_t), M_TEMP, + M_WAITOK); + *ap->a_cookies = cookies; + } - dp->d_namlen = strlen(kt->kt_name); - bcopy(kt->kt_name, dp->d_name, dp->d_namlen+1); + n = 0; + for (; i < nipsecsp_targets && uio->uio_resid >= UIO_MX; i++) { + kt = &ipsecsp_targets[i]; + d.d_namlen = kt->kt_namlen; + if ((error = kernfs_setdirentfileno(&d, i, kfs, + &kern_targets[0], kt, ap)) != 0) + break; + memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1); + d.d_type = kt->kt_type; + if ((error = uiomove(&d, UIO_MX, uio)) != 0) + break; + if (cookies) + *cookies++ = i + 1; + n++; + } + if (error) { + ncookies = n; + break; + } -#ifdef KERNFS_DIAGNOSTIC - printf("kernfs_readdir: name = %s, len = %d\n", - dp->d_name, dp->d_namlen); -#endif - /* - * Fill in the remaining fields - */ - dp->d_reclen = UIO_MX; - dp->d_ino = i + 3; - /* - * And ship to userland - */ - error = uiomove((caddr_t) dp, UIO_MX, uio); - if (error) + TAILQ_FOREACH(sp, &sptailq, tailq) { + if (uio->uio_resid < UIO_MX) + break; + if ((error = kernfs_setdirentfileno_kt(&d, &ipsecsp_kt, + sp->id, ap)) != 0) break; + d.d_namlen = snprintf(d.d_name, sizeof(d.d_name), + "%u", sp->id); + d.d_type = DT_REG; + if ((error = uiomove(&d, UIO_MX, uio)) != 0) + break; + if (cookies) + *cookies++ = i + 1; + n++; + i++; } - i++; + ncookies = n; + break; +#endif + + default: + error = ENOTDIR; + break; } - uio->uio_offset = i * UIO_MX; + if (ap->a_ncookies) { + if (error) { + if (cookies) + free(*ap->a_cookies, M_TEMP); + *ap->a_ncookies = 0; + *ap->a_cookies = NULL; + } else + *ap->a_ncookies = ncookies; + } + uio->uio_offset = i; return (error); } -kernfs_inactive(vp, p) - struct vnode *vp; - struct proc *p; +int +kernfs_inactive(v) + void *v; { - /* - * Clear out the v_type field to avoid - * nasty things happening in vgone(). - */ - vp->v_type = VNON; -#ifdef KERNFS_DIAGNOSTIC - printf("kernfs_inactive(%x)\n", vp); + struct vop_inactive_args /* { + struct vnode *a_vp; + struct lwp *a_l; + } */ *ap = v; + struct vnode *vp = ap->a_vp; + const struct kernfs_node *kfs = VTOKERN(ap->a_vp); +#ifdef IPSEC + struct mbuf *m; + struct secpolicy *sp; #endif + + VOP_UNLOCK(vp, 0); + switch (kfs->kfs_type) { +#ifdef IPSEC + case KFSipsecsa: + m = key_setdumpsa_spi(htonl(kfs->kfs_value)); + if (m) + m_freem(m); + else + vgone(vp); + break; + case KFSipsecsp: + sp = key_getspbyid(kfs->kfs_value); + if (sp) + key_freesp(sp); + else { + /* should never happen as we hold a refcnt */ + vgone(vp); + } + break; +#endif + default: + break; + } return (0); } -/* - * Print out the contents of a kernfs vnode. - */ -/* ARGSUSED */ -kernfs_print(vp) - struct vnode *vp; +int +kernfs_reclaim(v) + void *v; { - printf("tag VT_KERNFS, kernfs vnode\n"); -} + struct vop_reclaim_args /* { + struct vnode *a_vp; + } */ *ap = v; -/* - * kernfs vnode unsupported operation - */ -kernfs_enotsupp() -{ - return (EOPNOTSUPP); + return (kernfs_freevp(ap->a_vp)); } /* - * kernfs "should never get here" operation + * Return POSIX pathconf information applicable to special devices. */ -kernfs_badop() +int +kernfs_pathconf(v) + void *v; { - panic("kernfs: bad op"); + struct vop_pathconf_args /* { + struct vnode *a_vp; + int a_name; + register_t *a_retval; + } */ *ap = v; + + switch (ap->a_name) { + case _PC_LINK_MAX: + *ap->a_retval = LINK_MAX; + return (0); + case _PC_MAX_CANON: + *ap->a_retval = MAX_CANON; + return (0); + case _PC_MAX_INPUT: + *ap->a_retval = MAX_INPUT; + return (0); + case _PC_PIPE_BUF: + *ap->a_retval = PIPE_BUF; + return (0); + case _PC_CHOWN_RESTRICTED: + *ap->a_retval = 1; + return (0); + case _PC_VDISABLE: + *ap->a_retval = _POSIX_VDISABLE; + return (0); + case _PC_SYNC_IO: + *ap->a_retval = 1; + return (0); + default: + return (EINVAL); + } /* NOTREACHED */ } /* - * kernfs vnode null operation + * Print out the contents of a /dev/fd vnode. */ -kernfs_nullop() +/* ARGSUSED */ +int +kernfs_print(void *v) { + + printf("tag VT_KERNFS, kernfs vnode\n"); return (0); } -#define kernfs_create ((int (*) __P(( \ - struct nameidata *ndp, \ - struct vattr *vap, \ - struct proc *p))) kernfs_enotsupp) -#define kernfs_mknod ((int (*) __P(( \ - struct nameidata *ndp, \ - struct vattr *vap, \ - struct ucred *cred, \ - struct proc *p))) kernfs_enotsupp) -#define kernfs_close ((int (*) __P(( \ - struct vnode *vp, \ - int fflag, \ - struct ucred *cred, \ - struct proc *p))) nullop) -#define kernfs_ioctl ((int (*) __P(( \ - struct vnode *vp, \ - int command, \ - caddr_t data, \ - int fflag, \ - struct ucred *cred, \ - struct proc *p))) kernfs_enotsupp) -#define kernfs_select ((int (*) __P(( \ - struct vnode *vp, \ - int which, \ - int fflags, \ - struct ucred *cred, \ - struct proc *p))) kernfs_enotsupp) -#define kernfs_mmap ((int (*) __P(( \ - struct vnode *vp, \ - int fflags, \ - struct ucred *cred, \ - struct proc *p))) kernfs_enotsupp) -#define kernfs_fsync ((int (*) __P(( \ - struct vnode *vp, \ - int fflags, \ - struct ucred *cred, \ - int waitfor, \ - struct proc *p))) nullop) -#define kernfs_seek ((int (*) __P(( \ - struct vnode *vp, \ - off_t oldoff, \ - off_t newoff, \ - struct ucred *cred))) nullop) -#define kernfs_remove ((int (*) __P(( \ - struct nameidata *ndp, \ - struct proc *p))) kernfs_enotsupp) -#define kernfs_link ((int (*) __P(( \ - struct vnode *vp, \ - struct nameidata *ndp, \ - struct proc *p))) kernfs_enotsupp) -#define kernfs_rename ((int (*) __P(( \ - struct nameidata *fndp, \ - struct nameidata *tdnp, \ - struct proc *p))) kernfs_enotsupp) -#define kernfs_mkdir ((int (*) __P(( \ - struct nameidata *ndp, \ - struct vattr *vap, \ - struct proc *p))) kernfs_enotsupp) -#define kernfs_rmdir ((int (*) __P(( \ - struct nameidata *ndp, \ - struct proc *p))) kernfs_enotsupp) -#define kernfs_symlink ((int (*) __P(( \ - struct nameidata *ndp, \ - struct vattr *vap, \ - char *target, \ - struct proc *p))) kernfs_enotsupp) -#define kernfs_readlink ((int (*) __P(( \ - struct vnode *vp, \ - struct uio *uio, \ - struct ucred *cred))) kernfs_enotsupp) -#define kernfs_abortop ((int (*) __P(( \ - struct nameidata *ndp))) nullop) -#ifdef KERNFS_DIAGNOSTIC -int kernfs_reclaim(vp) -struct vnode *vp; +int +kernfs_link(v) + void *v; { - printf("kernfs_reclaim(%x)\n", vp); - return (0); + struct vop_link_args /* { + struct vnode *a_dvp; + struct vnode *a_vp; + struct componentname *a_cnp; + } */ *ap = v; + + VOP_ABORTOP(ap->a_dvp, ap->a_cnp); + vput(ap->a_dvp); + return (EROFS); +} + +int +kernfs_symlink(v) + void *v; +{ + struct vop_symlink_args /* { + struct vnode *a_dvp; + struct vnode **a_vpp; + struct componentname *a_cnp; + struct vattr *a_vap; + char *a_target; + } */ *ap = v; + + VOP_ABORTOP(ap->a_dvp, ap->a_cnp); + vput(ap->a_dvp); + return (EROFS); } -#else -#define kernfs_reclaim ((int (*) __P(( \ - struct vnode *vp))) nullop) -#endif -#define kernfs_lock ((int (*) __P(( \ - struct vnode *vp))) nullop) -#define kernfs_unlock ((int (*) __P(( \ - struct vnode *vp))) nullop) -#define kernfs_bmap ((int (*) __P(( \ - struct vnode *vp, \ - daddr_t bn, \ - struct vnode **vpp, \ - daddr_t *bnp))) kernfs_badop) -#define kernfs_strategy ((int (*) __P(( \ - struct buf *bp))) kernfs_badop) -#define kernfs_islocked ((int (*) __P(( \ - struct vnode *vp))) nullop) -#define kernfs_advlock ((int (*) __P(( \ - struct vnode *vp, \ - caddr_t id, \ - int op, \ - struct flock *fl, \ - int flags))) kernfs_enotsupp) - -struct vnodeops kernfs_vnodeops = { - kernfs_lookup, /* lookup */ - kernfs_create, /* create */ - kernfs_mknod, /* mknod */ - kernfs_open, /* open */ - kernfs_close, /* close */ - kernfs_access, /* access */ - kernfs_getattr, /* getattr */ - kernfs_setattr, /* setattr */ - kernfs_read, /* read */ - kernfs_write, /* write */ - kernfs_ioctl, /* ioctl */ - kernfs_select, /* select */ - kernfs_mmap, /* mmap */ - kernfs_fsync, /* fsync */ - kernfs_seek, /* seek */ - kernfs_remove, /* remove */ - kernfs_link, /* link */ - kernfs_rename, /* rename */ - kernfs_mkdir, /* mkdir */ - kernfs_rmdir, /* rmdir */ - kernfs_symlink, /* symlink */ - kernfs_readdir, /* readdir */ - kernfs_readlink, /* readlink */ - kernfs_abortop, /* abortop */ - kernfs_inactive, /* inactive */ - kernfs_reclaim, /* reclaim */ - kernfs_lock, /* lock */ - kernfs_unlock, /* unlock */ - kernfs_bmap, /* bmap */ - kernfs_strategy, /* strategy */ - kernfs_print, /* print */ - kernfs_islocked, /* islocked */ - kernfs_advlock, /* advlock */ -};