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 1.67 retrieving revision 1.99 diff -u -p -r1.67 -r1.99 --- src/sys/miscfs/kernfs/kernfs_vnops.c 1999/08/25 14:42:35 1.67 +++ src/sys/miscfs/kernfs/kernfs_vnops.c 2004/04/29 16:10:55 1.99 @@ -1,4 +1,4 @@ -/* $NetBSD: kernfs_vnops.c,v 1.67 1999/08/25 14:42:35 sommerfeld Exp $ */ +/* $NetBSD: kernfs_vnops.c,v 1.99 2004/04/29 16:10:55 jrf Exp $ */ /* * Copyright (c) 1992, 1993 @@ -15,11 +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. * @@ -42,11 +38,17 @@ * Kernel parameter filesystem (/kern) */ +#include +__KERNEL_RCSID(0, "$NetBSD: kernfs_vnops.c,v 1.99 2004/04/29 16:10:55 jrf Exp $"); + +#ifdef _KERNEL_OPT +#include "opt_ipsec.h" +#endif + #include #include #include #include -#include #include #include #include @@ -62,7 +64,13 @@ #include #include -#include +#ifdef IPSEC +#include +#include +#include +#include +#include +#endif #include @@ -72,37 +80,65 @@ #define READ_MODE (S_IRUSR|S_IRGRP|S_IROTH) #define WRITE_MODE (S_IWUSR|S_IRUSR|S_IRGRP|S_IROTH) #define DIR_MODE (S_IRUSR|S_IXUSR|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH) +#define UDIR_MODE (S_IRUSR|S_IXUSR) -struct kern_target kern_targets[] = { -/* NOTE: The name must be less than UIO_MX-16 chars in length */ #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, KTT_NULL, VDIR, DIR_MODE }, - { DT_DIR, N(".."), 0, KTT_NULL, VDIR, DIR_MODE }, - { DT_REG, N("boottime"), &boottime.tv_sec, KTT_INT, VREG, READ_MODE }, - { DT_REG, N("copyright"), copyright, KTT_STRING, VREG, READ_MODE }, - { DT_REG, N("hostname"), 0, KTT_HOSTNAME, VREG, WRITE_MODE }, - { DT_REG, N("hz"), &hz, KTT_INT, VREG, READ_MODE }, - { DT_REG, N("loadavg"), 0, KTT_AVENRUN, VREG, READ_MODE }, - { DT_REG, N("msgbuf"), 0, KTT_MSGBUF, VREG, READ_MODE }, - { DT_REG, N("pagesize"), &uvmexp.pagesize, KTT_INT, VREG, READ_MODE }, - { DT_REG, N("physmem"), &physmem, KTT_INT, VREG, READ_MODE }, + { 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 }, + /* XXX cast away const */ + { DT_REG, N("copyright"), (void *)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, KTT_NULL, VDIR, DIR_MODE }, + { DT_DIR, N("root"), 0, KFSnull, VDIR, DIR_MODE }, #endif - { DT_BLK, N("rootdev"), &rootdev, KTT_DEVICE, VBLK, READ_MODE }, - { DT_CHR, N("rrootdev"), &rrootdev, KTT_DEVICE, VCHR, READ_MODE }, - { DT_REG, N("time"), 0, KTT_TIME, VREG, READ_MODE }, - { DT_REG, N("version"), version, KTT_STRING, VREG, READ_MODE }, -#undef N + { 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 }, + /* XXX cast away const */ + { DT_REG, N("version"), (void *)version, + KFSstring, VREG, READ_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 }, }; -static int nkern_targets = sizeof(kern_targets) / sizeof(kern_targets[0]); +#endif +#undef N +int 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]); +#endif + int kernfs_lookup __P((void *)); -#define kernfs_create genfs_eopnotsupp_rele -#define kernfs_mknod genfs_eopnotsupp_rele -#define kernfs_open genfs_nullop -#define kernfs_close genfs_nullop +#define kernfs_create genfs_eopnotsupp +#define kernfs_mknod genfs_eopnotsupp +int kernfs_open __P((void *)); +int kernfs_close __P((void *)); int kernfs_access __P((void *)); int kernfs_getattr __P((void *)); int kernfs_setattr __P((void *)); @@ -112,14 +148,13 @@ int kernfs_write __P((void *)); #define kernfs_ioctl genfs_enoioctl #define kernfs_poll genfs_poll #define kernfs_revoke genfs_revoke -#define kernfs_mmap genfs_eopnotsupp #define kernfs_fsync genfs_nullop #define kernfs_seek genfs_nullop -#define kernfs_remove genfs_eopnotsupp_rele +#define kernfs_remove genfs_eopnotsupp int kernfs_link __P((void *)); -#define kernfs_rename genfs_eopnotsupp_rele -#define kernfs_mkdir genfs_eopnotsupp_rele -#define kernfs_rmdir genfs_eopnotsupp_rele +#define kernfs_rename genfs_eopnotsupp +#define kernfs_mkdir genfs_eopnotsupp +#define kernfs_rmdir genfs_eopnotsupp int kernfs_symlink __P((void *)); int kernfs_readdir __P((void *)); #define kernfs_readlink genfs_eopnotsupp @@ -140,12 +175,13 @@ int kernfs_pathconf __P((void *)); #define kernfs_truncate genfs_eopnotsupp #define kernfs_update genfs_nullop #define kernfs_bwrite genfs_eopnotsupp +#define kernfs_putpages genfs_putpages -int kernfs_xread __P((struct kern_target *, int, char **, int)); -int kernfs_xwrite __P((struct kern_target *, char *, int)); +static int kernfs_xread __P((struct kernfs_node *, int, char **, size_t, size_t *)); +static int kernfs_xwrite __P((const struct kernfs_node *, char *, size_t)); int (**kernfs_vnodeop_p) __P((void *)); -struct vnodeopv_entry_desc kernfs_vnodeop_entries[] = { +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 */ @@ -161,7 +197,6 @@ struct vnodeopv_entry_desc kernfs_vnodeo { &vop_ioctl_desc, kernfs_ioctl }, /* ioctl */ { &vop_poll_desc, kernfs_poll }, /* poll */ { &vop_revoke_desc, kernfs_revoke }, /* revoke */ - { &vop_mmap_desc, kernfs_mmap }, /* mmap */ { &vop_fsync_desc, kernfs_fsync }, /* fsync */ { &vop_seek_desc, kernfs_seek }, /* seek */ { &vop_remove_desc, kernfs_remove }, /* remove */ @@ -189,43 +224,51 @@ struct vnodeopv_entry_desc kernfs_vnodeo { &vop_truncate_desc, kernfs_truncate }, /* truncate */ { &vop_update_desc, kernfs_update }, /* update */ { &vop_bwrite_desc, kernfs_bwrite }, /* bwrite */ - { (struct vnodeop_desc*)NULL, (int(*) __P((void *)))NULL } + { &vop_putpages_desc, kernfs_putpages }, /* putpages */ + { NULL, NULL } }; -struct vnodeopv_desc kernfs_vnodeop_opv_desc = +const struct vnodeopv_desc kernfs_vnodeop_opv_desc = { &kernfs_vnodeop_p, kernfs_vnodeop_entries }; -int -kernfs_xread(kt, off, bufp, len) - struct kern_target *kt; +static int +kernfs_xread(kfs, off, bufp, len, wrlen) + struct kernfs_node *kfs; int off; char **bufp; - int len; + size_t len; + size_t *wrlen; { + const struct kern_target *kt; +#ifdef IPSEC + struct mbuf *m; +#endif + + kt = kfs->kfs_kt; - switch (kt->kt_tag) { - case KTT_TIME: { + switch (kfs->kfs_type) { + case KFStime: { struct timeval tv; microtime(&tv); - sprintf(*bufp, "%ld %ld\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(*bufp, "%d\n", *ip); + snprintf(*bufp, len, "%d\n", *ip); break; } - case KTT_STRING: { + case KFSstring: { char *cp = kt->kt_data; *bufp = cp; break; } - case KTT_MSGBUF: { + case KFSmsgbuf: { long n; /* @@ -246,61 +289,121 @@ kernfs_xread(kt, off, bufp, len) * message buffer header are corrupted, but that'll cause * the system to die anyway. */ - if (off >= msgbufp->msg_bufs) + 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; - return (len); + *wrlen = len; + return (0); } - case KTT_HOSTNAME: { + case KFShostname: { char *cp = hostname; int xlen = hostnamelen; - if (xlen >= (len-2)) + if (xlen >= (len - 2)) return (EINVAL); memcpy(*bufp, cp, xlen); (*bufp)[xlen] = '\n'; (*bufp)[xlen+1] = '\0'; + len = strlen(*bufp); break; } - case KTT_AVENRUN: + case KFSavenrun: averunnable.fscale = FSCALE; - sprintf(*bufp, "%d %d %d %ld\n", + 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: + *wrlen = 0; return (0); } len = strlen(*bufp); if (len <= off) - return (0); - *bufp += off; - return (len - off); + *wrlen = 0; + else { + *bufp += off; + *wrlen = len - off; + } + return (0); } -int -kernfs_xwrite(kt, buf, len) - struct kern_target *kt; +static int +kernfs_xwrite(kfs, buf, len) + const struct kernfs_node *kfs; char *buf; - int len; + size_t len; { - switch (kt->kt_tag) { - case KTT_HOSTNAME: + switch (kfs->kfs_type) { + case KFShostname: if (buf[len-1] == '\n') --len; memcpy(hostname, buf, len); hostname[len] = '\0'; - hostnamelen = len; + hostnamelen = (size_t) len; return (0); default: @@ -326,14 +429,12 @@ kernfs_lookup(v) struct vnode **vpp = ap->a_vpp; struct vnode *dvp = ap->a_dvp; const char *pname = cnp->cn_nameptr; - struct kern_target *kt; - struct vnode *fvp; + const struct kernfs_node *kfs; + const struct kern_target *kt; int error, i, wantpunlock; - -#ifdef KERNFS_DIAGNOSTIC - printf("kernfs_lookup(%p)\n", ap); - printf("kernfs_lookup(dp = %p, vpp = %p, cnp = %p)\n", dvp, vpp, ap->a_cnp); - printf("kernfs_lookup(%s)\n", pname); +#ifdef IPSEC + char *ep; + u_int32_t id; #endif *vpp = NULLVP; @@ -348,73 +449,158 @@ kernfs_lookup(v) return (0); } - /* - * This code only supports a flat directory, so we don't - * need to worry about .. - */ + 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); -#if 0 - if (cnp->cn_namelen == 4 && memcmp(pname, "root", 4) == 0) { - *vpp = rootdir; - VREF(rootdir); - vn_lock(rootdir, LK_SHARED | LK_RETRY); - return (0); - } -#endif + for (i = 0; i < 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; + } + break; - wantpunlock = (~cnp->cn_flags & (LOCKPARENT | ISLASTCN)); + 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); - for (kt = kern_targets, i = 0; i < nkern_targets; kt++, i++) { - if (cnp->cn_namelen == kt->kt_namlen && - memcmp(kt->kt_name, pname, cnp->cn_namelen) == 0) - goto found; - } +#ifdef IPSEC + case KFSipsecsadir: + for (i = 0; i < nipsecsa_targets; i++) { + kt = &ipsecsa_targets[i]; + if (cnp->cn_namelen == kt->kt_namlen && + memcmp(kt->kt_name, pname, cnp->cn_namelen) == 0) { + 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); + } + } -#ifdef KERNFS_DIAGNOSTIC - printf("kernfs_lookup: i = %d, failed", i); -#endif + ep = NULL; + id = strtoul(pname, &ep, 10); + if (!ep || *ep || ep == pname) + break; - return (cnp->cn_nameiop == LOOKUP ? ENOENT : EROFS); + error = kernfs_allocvp(dvp->v_mount, vpp, KFSipsecsa, NULL, id); + if ((error == 0) && wantpunlock) { + VOP_UNLOCK(dvp, 0); + cnp->cn_flags |= PDIRUNLOCK; + } + return (error); -found: - if (kt->kt_tag == KTT_DEVICE) { - dev_t *dp = kt->kt_data; - loop: - if (*dp == NODEV || !vfinddev(*dp, kt->kt_vtype, &fvp)) { - return (ENOENT); + case KFSipsecspdir: + for (i = 0; i < nipsecsp_targets; i++) { + kt = &ipsecsp_targets[i]; + if (cnp->cn_namelen == kt->kt_namlen && + memcmp(kt->kt_name, pname, cnp->cn_namelen) == 0) { + 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); + } } - *vpp = fvp; - if (vget(fvp, LK_EXCLUSIVE)) - goto loop; - if (wantpunlock) { + + ep = NULL; + id = strtoul(pname, &ep, 10); + if (!ep || *ep || ep == pname) + break; + + error = kernfs_allocvp(dvp->v_mount, vpp, KFSipsecsp, NULL, id); + if ((error == 0) && wantpunlock) { VOP_UNLOCK(dvp, 0); cnp->cn_flags |= PDIRUNLOCK; } - return (0); + return (error); +#endif + + default: + return (ENOTDIR); } -#ifdef KERNFS_DIAGNOSTIC - printf("kernfs_lookup: allocate new vnode\n"); + return (cnp->cn_nameiop == LOOKUP ? ENOENT : EROFS); +} + +int +kernfs_open(v) + void *v; +{ + struct vop_open_args /* { + struct vnode *a_vp; + int a_mode; + struct ucred *a_cred; + struct proc *a_p; + } */ *ap = v; + struct kernfs_node *kfs = VTOKERN(ap->a_vp); +#ifdef IPSEC + struct mbuf *m; + struct secpolicy *sp; +#endif + + 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 - error = getnewvnode(VT_KERNFS, dvp->v_mount, kernfs_vnodeop_p, &fvp); - if (error) { - return (error); + + default: + return (0); } +} - MALLOC(fvp->v_data, void *, sizeof(struct kernfs_node), M_TEMP, - M_WAITOK); - VTOKERN(fvp)->kf_kt = kt; - fvp->v_type = kt->kt_vtype; - vn_lock(fvp, LK_EXCLUSIVE | LK_RETRY); - *vpp = fvp; - -#ifdef KERNFS_DIAGNOSTIC - printf("kernfs_lookup: newvp = %p\n", fvp); -#endif - if (wantpunlock) { - VOP_UNLOCK(dvp, 0); - cnp->cn_flags |= PDIRUNLOCK; +int +kernfs_close(v) + void *v; +{ + struct vop_close_args /* { + struct vnode *a_vp; + int a_fflag; + struct ucred *a_cred; + struct proc *a_p; + } */ *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: + break; } + return (0); } @@ -428,18 +614,14 @@ kernfs_access(v) struct ucred *a_cred; struct proc *a_p; } */ *ap = v; - struct vnode *vp = ap->a_vp; - mode_t mode; + struct vattr va; + int error; - if (vp->v_flag & VROOT) { - mode = DIR_MODE; - } else { - struct kern_target *kt = VTOKERN(vp)->kf_kt; - mode = kt->kt_mode; - } + if ((error = VOP_GETATTR(ap->a_vp, &va, ap->a_cred, ap->a_p)) != 0) + return (error); - return (vaccess(vp->v_type, mode, (uid_t)0, (gid_t)0, ap->a_mode, - ap->a_cred)); + return (vaccess(va.va_type, va.va_mode, va.va_uid, va.va_gid, + ap->a_mode, ap->a_cred)); } int @@ -452,57 +634,90 @@ kernfs_getattr(v) struct ucred *a_cred; struct proc *a_p; } */ *ap = v; - struct vnode *vp = ap->a_vp; + struct kernfs_node *kfs = VTOKERN(ap->a_vp); struct vattr *vap = ap->a_vap; - struct timeval tv; int error = 0; char strbuf[KSTRING], *buf; + size_t nread, total; - memset((caddr_t) vap, 0, sizeof(*vap)); - vattr_null(vap); + VATTR_NULL(vap); + vap->va_type = ap->a_vp->v_type; vap->va_uid = 0; vap->va_gid = 0; - vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[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(&tv); - TIMEVAL_TO_TIMESPEC(&tv, &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. + * Avoid microtime(9), it's slow. + * We don't guard the read from time(9) with splclock(9) since we + * don't actually need to be THAT sure the access is atomic. + */ + 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 { + TIMEVAL_TO_TIMESPEC(&time, &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_bytes = 0; - if (vp->v_flag & VROOT) { -#ifdef KERNFS_DIAGNOSTIC - printf("kernfs_getattr: stat rootdir\n"); -#endif - vap->va_type = VDIR; - vap->va_mode = DIR_MODE; + switch (kfs->kfs_type) { + case KFSkern: +#ifdef IPSEC + vap->va_nlink = 4; /* 2 extra subdirs */ +#else vap->va_nlink = 2; - vap->va_fileid = 2; - vap->va_size = DEV_BSIZE; - } else { - struct kern_target *kt = VTOKERN(vp)->kf_kt; - int nbytes, total; -#ifdef KERNFS_DIAGNOSTIC - printf("kernfs_getattr: stat target %s\n", kt->kt_name); #endif - vap->va_type = kt->kt_vtype; - vap->va_mode = kt->kt_mode; + 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 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; - vap->va_fileid = 1 + (kt - kern_targets); total = 0; - while (buf = strbuf, - nbytes = kernfs_xread(kt, total, &buf, sizeof(strbuf))) - total += nbytes; - vap->va_size = total; - } + do { + buf = strbuf; + error = kernfs_xread(kfs, total, &buf, + sizeof(strbuf), &nread); + total += nread; + } while (error == 0 && nread != 0); + vap->va_bytes = vap->va_size = total; + break; -#ifdef KERNFS_DIAGNOSTIC - printf("kernfs_getattr: return error %d\n", error); +#ifdef IPSEC + case KFSipsecsadir: + case KFSipsecspdir: + vap->va_nlink = 2; + vap->va_bytes = vap->va_size = DEV_BSIZE; + break; #endif + + default: + error = EINVAL; + break; + } + return (error); } @@ -511,6 +726,7 @@ int kernfs_setattr(v) void *v; { + /* * Silently ignore attribute changes. * This allows for open with truncate to have no @@ -530,34 +746,21 @@ kernfs_read(v) int a_ioflag; struct ucred *a_cred; } */ *ap = v; - struct vnode *vp = ap->a_vp; struct uio *uio = ap->a_uio; - struct kern_target *kt; + struct kernfs_node *kfs = VTOKERN(ap->a_vp); char strbuf[KSTRING], *buf; - int off, len; + off_t off; + size_t len; int error; - if (vp->v_type == VDIR) + if (ap->a_vp->v_type == VDIR) return (EOPNOTSUPP); - kt = VTOKERN(vp)->kf_kt; - -#ifdef KERNFS_DIAGNOSTIC - printf("kern_read %s\n", kt->kt_name); -#endif - off = uio->uio_offset; -#if 0 - while (buf = strbuf, -#else - if (buf = strbuf, -#endif - len = kernfs_xread(kt, off, &buf, sizeof(strbuf))) { - if ((error = uiomove(buf, len, uio)) != 0) - return (error); - off += len; - } - return (0); + buf = strbuf; + if ((error = kernfs_xread(kfs, off, &buf, sizeof(strbuf), &len)) == 0) + error = uiomove(buf, len, uio); + return (error); } int @@ -570,17 +773,11 @@ kernfs_write(v) int a_ioflag; struct ucred *a_cred; } */ *ap = v; - struct vnode *vp = ap->a_vp; + struct kernfs_node *kfs = VTOKERN(ap->a_vp); struct uio *uio = ap->a_uio; - struct kern_target *kt; int error, xlen; char strbuf[KSTRING]; - if (vp->v_type == VDIR) - return (EOPNOTSUPP); - - kt = VTOKERN(vp)->kf_kt; - if (uio->uio_offset != 0) return (EINVAL); @@ -593,7 +790,7 @@ kernfs_write(v) strbuf[xlen] = '\0'; xlen = strlen(strbuf); - return (kernfs_xwrite(kt, strbuf, xlen)); + return (kernfs_xwrite(kfs, strbuf, xlen)); } int @@ -610,14 +807,16 @@ kernfs_readdir(v) } */ *ap = v; struct uio *uio = ap->a_uio; struct dirent d; - struct kern_target *kt; + struct kernfs_node *kfs = VTOKERN(ap->a_vp); + const struct kern_target *kt; off_t i; int error; off_t *cookies = NULL; - int ncookies = 0, nc = 0; - - if (ap->a_vp->v_type != VDIR) - return (ENOTDIR); + int ncookies = 0, n; +#ifdef IPSEC + struct secasvar *sav, *sav2; + struct secpolicy *sp; +#endif if (uio->uio_resid < UIO_MX) return (EINVAL); @@ -626,51 +825,212 @@ kernfs_readdir(v) error = 0; i = uio->uio_offset; - - if (i >= nkern_targets) - return 0; - - memset((caddr_t)&d, 0, UIO_MX); + memset(&d, 0, sizeof(d)); d.d_reclen = UIO_MX; + ncookies = uio->uio_resid / UIO_MX; - if (ap->a_ncookies) { - nc = uio->uio_resid / UIO_MX; - nc = min(nc, (nkern_targets - i)); - MALLOC(cookies, off_t *, nc * sizeof(off_t), M_TEMP, - M_WAITOK); - *ap->a_cookies = cookies; - } + switch (kfs->kfs_type) { + case KFSkern: + if (i >= nkern_targets) + return (0); - for (kt = &kern_targets[i]; - uio->uio_resid >= UIO_MX && i < nkern_targets; kt++, i++) { -#ifdef KERNFS_DIAGNOSTIC - printf("kernfs_readdir: i = %d\n", i); -#endif + if (ap->a_ncookies) { + ncookies = min(ncookies, (nkern_targets - i)); + cookies = malloc(ncookies * sizeof(off_t), M_TEMP, + M_WAITOK); + *ap->a_cookies = cookies; + } - if (kt->kt_tag == KTT_DEVICE) { - dev_t *dp = kt->kt_data; - struct vnode *fvp; + n = 0; + for (; i < nkern_targets && uio->uio_resid >= UIO_MX; i++) { + kt = &kern_targets[i]; + 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 (i < 2) + d.d_fileno = KERNFS_FILENO(&kern_targets[0], + kern_targets[0].kt_tag, 0); + else + 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 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; - if (*dp == NODEV || !vfinddev(*dp, kt->kt_vtype, &fvp)) +#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; + 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++; + } + if (error) { + ncookies = n; + break; + } + + 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; + d.d_fileno = KERNFS_FILENO(kfs->kfs_kt, kfs->kfs_type, + kfs->kfs_cookie); + 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++; - d.d_fileno = i + 3; - d.d_namlen = kt->kt_namlen; - memcpy(d.d_name, kt->kt_name, kt->kt_namlen + 1); - d.d_type = kt->kt_type; + if (i >= 2 + 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; + } - if ((error = uiomove((caddr_t)&d, UIO_MX, uio)) != 0) + n = 0; + for (; i < nipsecsp_targets && uio->uio_resid >= UIO_MX; i++) { + kt = &ipsecsp_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++; + } + if (error) { + ncookies = n; break; - if (cookies) { - *cookies++ = i + 1; - ncookies++; } + + TAILQ_FOREACH(sp, &sptailq, tailq) { + if (uio->uio_resid < UIO_MX) + break; + d.d_fileno = KERNFS_FILENO(kfs->kfs_kt, kfs->kfs_type, + kfs->kfs_cookie); + 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++; + } + ncookies = n; + break; +#endif + + default: + error = ENOTDIR; + break; } if (ap->a_ncookies) { if (error) { - FREE(*ap->a_cookies, M_TEMP); + if (cookies) + free(*ap->a_cookies, M_TEMP); *ap->a_ncookies = 0; *ap->a_cookies = NULL; } else @@ -690,16 +1050,35 @@ kernfs_inactive(v) struct proc *a_p; } */ *ap = v; struct vnode *vp = ap->a_vp; - -#ifdef KERNFS_DIAGNOSTIC - printf("kernfs_inactive(%p)\n", vp); + const struct kernfs_node *kfs = VTOKERN(ap->a_vp); +#ifdef IPSEC + struct mbuf *m; + struct secpolicy *sp; #endif - /* - * Clear out the v_type field to avoid - * nasty things happening in vgone(). - */ + VOP_UNLOCK(vp, 0); - vp->v_type = VNON; + 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); } @@ -710,16 +1089,8 @@ kernfs_reclaim(v) struct vop_reclaim_args /* { struct vnode *a_vp; } */ *ap = v; - struct vnode *vp = ap->a_vp; -#ifdef KERNFS_DIAGNOSTIC - printf("kernfs_reclaim(%p)\n", vp); -#endif - if (vp->v_data) { - FREE(vp->v_data, M_TEMP); - vp->v_data = 0; - } - return (0); + return (kernfs_freevp(ap->a_vp)); } /*