Author: mjg Date: Thu Feb 20 16:58:19 2020 New Revision: 358172 URL: https://svnweb.freebsd.org/changeset/base/358172
Log: vfs: add realpathat syscall realpath(3) is used a lot e.g., by clang and is a major source of getcwd and fstatat calls. This can be done more efficiently in the kernel. This works by performing a regular lookup while saving the name and found parent directory. If the terminal vnode is a directory we can resolve it using usual means. Otherwise we can use the name saved by lookup and resolve the parent. See the review for sample syscall counts. Reviewed by: kib Differential Revision: https://reviews.freebsd.org/D23574 Modified: head/lib/libc/stdlib/realpath.c head/libexec/rtld-elf/rtld-libc/Makefile.inc head/sys/bsm/audit_kevents.h head/sys/compat/freebsd32/syscalls.master head/sys/kern/syscalls.master head/sys/kern/vfs_cache.c head/sys/security/audit/audit_bsm.c head/sys/sys/param.h head/usr.bin/truss/syscalls.c Modified: head/lib/libc/stdlib/realpath.c ============================================================================== --- head/lib/libc/stdlib/realpath.c Thu Feb 20 16:38:30 2020 (r358171) +++ head/lib/libc/stdlib/realpath.c Thu Feb 20 16:58:19 2020 (r358172) @@ -42,14 +42,19 @@ __FBSDID("$FreeBSD$"); #include <stdlib.h> #include <string.h> #include <unistd.h> +#include <fcntl.h> #include "un-namespace.h" +#include "libc_private.h" +extern int __realpathat(int fd, const char *path, char *buf, size_t size, + int flags); + /* * Find the real name of path, by removing all ".", ".." and symlink * components. Returns (resolved) on success, or (NULL) on failure, * in which case the path which caused trouble is left in (resolved). */ -static char * +static char * __noinline realpath1(const char *path, char *resolved) { struct stat sb; @@ -222,6 +227,10 @@ realpath(const char * __restrict path, char * __restri m = resolved = malloc(PATH_MAX); if (resolved == NULL) return (NULL); + } + if (__getosreldate() >= 1300080) { + if (__realpathat(AT_FDCWD, path, resolved, PATH_MAX, 0) == 0) + return (resolved); } res = realpath1(path, resolved); if (res == NULL) Modified: head/libexec/rtld-elf/rtld-libc/Makefile.inc ============================================================================== --- head/libexec/rtld-elf/rtld-libc/Makefile.inc Thu Feb 20 16:38:30 2020 (r358171) +++ head/libexec/rtld-elf/rtld-libc/Makefile.inc Thu Feb 20 16:58:19 2020 (r358172) @@ -49,7 +49,7 @@ _libc_other_objects= sigsetjmp lstat stat fstat fstata sysarch __sysctl issetugid __getcwd utrace \ thr_self thr_kill pread mmap lseek _exit _fstat _fstatat _fstatfs \ getdirentries _getdirentries _close _fcntl _open _openat _read \ - _sigprocmask _write readlink _setjmp setjmp setjmperr + _sigprocmask _write readlink __realpathat _setjmp setjmp setjmperr # Finally add additional architecture-dependent libc dependencies Modified: head/sys/bsm/audit_kevents.h ============================================================================== --- head/sys/bsm/audit_kevents.h Thu Feb 20 16:38:30 2020 (r358171) +++ head/sys/bsm/audit_kevents.h Thu Feb 20 16:58:19 2020 (r358172) @@ -657,6 +657,7 @@ #define AUE_LGETUUID 43261 /* CADETS. */ #define AUE_EXECVEAT 43262 /* FreeBSD/Linux. */ #define AUE_SHMRENAME 43263 /* FreeBSD-specific. */ +#define AUE_REALPATHAT 43264 /* FreeBSD-specific. */ /* * Darwin BSM uses a number of AUE_O_* definitions, which are aliased to the Modified: head/sys/compat/freebsd32/syscalls.master ============================================================================== --- head/sys/compat/freebsd32/syscalls.master Thu Feb 20 16:38:30 2020 (r358171) +++ head/sys/compat/freebsd32/syscalls.master Thu Feb 20 16:58:19 2020 (r358172) @@ -1160,5 +1160,7 @@ 572 AUE_SHMRENAME NOPROTO { int shm_rename(const char *path_from, \ const char *path_to, int flags); } 573 AUE_NULL NOPROTO { int sigfastblock(int cmd, uint32_t *ptr); } +574 AUE_REALPATHAT NOPROTO { int __realpathat(int fd, const char *path, \ + char *buf, size_t size, int flags); } ; vim: syntax=off Modified: head/sys/kern/syscalls.master ============================================================================== --- head/sys/kern/syscalls.master Thu Feb 20 16:38:30 2020 (r358171) +++ head/sys/kern/syscalls.master Thu Feb 20 16:58:19 2020 (r358172) @@ -3218,6 +3218,15 @@ _Inout_opt_ uint32_t *ptr ); } +574 AUE_REALPATHAT STD { + int __realpathat( + int fd, + _In_z_ const char *path, + _Out_writes_z_(size) char *buf, + size_t size, + int flags + ); + } ; Please copy any additions and changes to the following compatability tables: ; sys/compat/freebsd32/syscalls.master Modified: head/sys/kern/vfs_cache.c ============================================================================== --- head/sys/kern/vfs_cache.c Thu Feb 20 16:38:30 2020 (r358171) +++ head/sys/kern/vfs_cache.c Thu Feb 20 16:58:19 2020 (r358172) @@ -42,6 +42,7 @@ __FBSDID("$FreeBSD$"); #include <sys/param.h> #include <sys/systm.h> +#include <sys/capsicum.h> #include <sys/counter.h> #include <sys/filedesc.h> #include <sys/fnv_hash.h> @@ -387,8 +388,12 @@ STATNODE_COUNTER(shrinking_skipped, "Number of times shrinking was already in progress"); static void cache_zap_locked(struct namecache *ncp, bool neg_locked); -static int vn_fullpath1(struct thread *td, struct vnode *vp, struct vnode *rdir, +static int vn_fullpath_hardlink(struct thread *td, struct nameidata *ndp, char **retbuf, + char **freebuf, size_t *buflen); +static int vn_fullpath_any(struct thread *td, struct vnode *vp, struct vnode *rdir, char *buf, char **retbuf, size_t *buflen); +static int vn_fullpath_dir(struct thread *td, struct vnode *vp, struct vnode *rdir, + char *buf, char **retbuf, size_t *len, bool slash_prefixed, size_t addend); static MALLOC_DEFINE(M_VFSCACHE, "vfscache", "VFS name cache entries"); @@ -2201,7 +2206,7 @@ vn_getcwd(struct thread *td, char *buf, char **retbuf, rdir = fdp->fd_rdir; vrefact(rdir); FILEDESC_SUNLOCK(fdp); - error = vn_fullpath1(td, cdir, rdir, buf, retbuf, buflen); + error = vn_fullpath_any(td, cdir, rdir, buf, retbuf, buflen); vrele(rdir); vrele(cdir); @@ -2212,6 +2217,37 @@ vn_getcwd(struct thread *td, char *buf, char **retbuf, return (error); } +static int +kern___realpathat(struct thread *td, int fd, const char *path, char *buf, + size_t size, int flags, enum uio_seg pathseg) +{ + struct nameidata nd; + char *retbuf, *freebuf; + int error; + + if (flags != 0) + return (EINVAL); + NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | SAVENAME | WANTPARENT | AUDITVNODE1, + pathseg, path, fd, &cap_fstat_rights, td); + if ((error = namei(&nd)) != 0) + return (error); + error = vn_fullpath_hardlink(td, &nd, &retbuf, &freebuf, &size); + if (error == 0) { + error = copyout(retbuf, buf, size); + free(freebuf, M_TEMP); + } + NDFREE(&nd, 0); + return (error); +} + +int +sys___realpathat(struct thread *td, struct __realpathat_args *uap) +{ + + return (kern___realpathat(td, uap->fd, uap->path, uap->buf, uap->size, + uap->flags, UIO_USERSPACE)); +} + /* * Retrieve the full filesystem path that correspond to a vnode from the name * cache (if available) @@ -2235,7 +2271,7 @@ vn_fullpath(struct thread *td, struct vnode *vn, char rdir = fdp->fd_rdir; vrefact(rdir); FILEDESC_SUNLOCK(fdp); - error = vn_fullpath1(td, vn, rdir, buf, retbuf, &buflen); + error = vn_fullpath_any(td, vn, rdir, buf, retbuf, &buflen); vrele(rdir); if (!error) @@ -2263,7 +2299,7 @@ vn_fullpath_global(struct thread *td, struct vnode *vn return (EINVAL); buflen = MAXPATHLEN; buf = malloc(buflen, M_TEMP, M_WAITOK); - error = vn_fullpath1(td, vn, rootvnode, buf, retbuf, &buflen); + error = vn_fullpath_any(td, vn, rootvnode, buf, retbuf, &buflen); if (!error) *freebuf = buf; else @@ -2334,40 +2370,40 @@ vn_vptocnp(struct vnode **vp, struct ucred *cred, char } /* - * The magic behind vn_getcwd() and vn_fullpath(). + * Resolve a directory to a pathname. + * + * The name of the directory can always be found in the namecache or fetched + * from the filesystem. There is also guaranteed to be only one parent, meaning + * we can just follow vnodes up until we find the root. + * + * The vnode must be referenced. */ static int -vn_fullpath1(struct thread *td, struct vnode *vp, struct vnode *rdir, - char *buf, char **retbuf, size_t *len) +vn_fullpath_dir(struct thread *td, struct vnode *vp, struct vnode *rdir, + char *buf, char **retbuf, size_t *len, bool slash_prefixed, size_t addend) { - int error, slash_prefixed; #ifdef KDTRACE_HOOKS struct vnode *startvp = vp; #endif struct vnode *vp1; size_t buflen; + int error; + VNPASS(vp->v_type == VDIR || VN_IS_DOOMED(vp), vp); + VNPASS(vp->v_usecount > 0, vp); + buflen = *len; - buflen--; - buf[buflen] = '\0'; + if (!slash_prefixed) { + MPASS(*len >= 2); + buflen--; + buf[buflen] = '\0'; + } + error = 0; - slash_prefixed = 0; SDT_PROBE1(vfs, namecache, fullpath, entry, vp); counter_u64_add(numfullpathcalls, 1); - vref(vp); - if (vp->v_type != VDIR) { - error = vn_vptocnp(&vp, td->td_ucred, buf, &buflen); - if (error) - return (error); - if (buflen == 0) { - vrele(vp); - return (ENOMEM); - } - buf[--buflen] = '/'; - slash_prefixed = 1; - } while (vp != rdir && vp != rootvnode) { /* * The vp vnode must be already fully constructed, @@ -2420,7 +2456,7 @@ vn_fullpath1(struct thread *td, struct vnode *vp, stru break; } buf[--buflen] = '/'; - slash_prefixed = 1; + slash_prefixed = true; } if (error) return (error); @@ -2437,10 +2473,126 @@ vn_fullpath1(struct thread *td, struct vnode *vp, stru counter_u64_add(numfullpathfound, 1); vrele(vp); - SDT_PROBE3(vfs, namecache, fullpath, return, 0, startvp, buf + buflen); *retbuf = buf + buflen; + SDT_PROBE3(vfs, namecache, fullpath, return, 0, startvp, *retbuf); *len -= buflen; + *len += addend; return (0); +} + +/* + * Resolve an arbitrary vnode to a pathname. + * + * Note 2 caveats: + * - hardlinks are not tracked, thus if the vnode is not a directory this can + * resolve to a different path than the one used to find it + * - namecache is not mandatory, meaning names are not guaranteed to be added + * (in which case resolving fails) + */ +static int +vn_fullpath_any(struct thread *td, struct vnode *vp, struct vnode *rdir, + char *buf, char **retbuf, size_t *buflen) +{ + size_t orig_buflen; + bool slash_prefixed; + int error; + + if (*buflen < 2) + return (EINVAL); + + orig_buflen = *buflen; + + vref(vp); + slash_prefixed = false; + if (vp->v_type != VDIR) { + *buflen -= 1; + buf[*buflen] = '\0'; + error = vn_vptocnp(&vp, td->td_ucred, buf, buflen); + if (error) + return (error); + if (*buflen == 0) { + vrele(vp); + return (ENOMEM); + } + *buflen -= 1; + buf[*buflen] = '/'; + slash_prefixed = true; + } + + return (vn_fullpath_dir(td, vp, rdir, buf, retbuf, buflen, slash_prefixed, + orig_buflen - *buflen)); +} + +/* + * Resolve an arbitrary vnode to a pathname (taking care of hardlinks). + * + * Since the namecache does not track handlings, the caller is expected to first + * look up the target vnode with SAVENAME | WANTPARENT flags passed to namei. + * + * Then we have 2 cases: + * - if the found vnode is a directory, the path can be constructed just by + * fullowing names up the chain + * - otherwise we populate the buffer with the saved name and start resolving + * from the parent + */ +static int +vn_fullpath_hardlink(struct thread *td, struct nameidata *ndp, char **retbuf, + char **freebuf, size_t *buflen) +{ + char *buf, *tmpbuf; + struct filedesc *fdp; + struct vnode *rdir; + struct componentname *cnp; + struct vnode *vp; + size_t addend; + int error; + bool slash_prefixed; + + if (*buflen < 2) + return (EINVAL); + if (*buflen > MAXPATHLEN) + *buflen = MAXPATHLEN; + + slash_prefixed = false; + + buf = malloc(*buflen, M_TEMP, M_WAITOK); + fdp = td->td_proc->p_fd; + FILEDESC_SLOCK(fdp); + rdir = fdp->fd_rdir; + vrefact(rdir); + FILEDESC_SUNLOCK(fdp); + + addend = 0; + vp = ndp->ni_vp; + if (vp->v_type != VDIR) { + cnp = &ndp->ni_cnd; + addend = cnp->cn_namelen + 2; + if (*buflen < addend) { + error = ENOMEM; + goto out_bad; + } + *buflen -= addend; + tmpbuf = buf + *buflen; + tmpbuf[0] = '/'; + memcpy(&tmpbuf[1], cnp->cn_nameptr, cnp->cn_namelen); + tmpbuf[addend - 1] = '\0'; + slash_prefixed = true; + vp = ndp->ni_dvp; + } + + vref(vp); + error = vn_fullpath_dir(td, vp, rdir, buf, retbuf, buflen, slash_prefixed, addend); + if (error != 0) + goto out_bad; + + vrele(rdir); + *freebuf = buf; + + return (0); +out_bad: + vrele(rdir); + free(buf, M_TEMP); + return (error); } struct vnode * Modified: head/sys/security/audit/audit_bsm.c ============================================================================== --- head/sys/security/audit/audit_bsm.c Thu Feb 20 16:38:30 2020 (r358171) +++ head/sys/security/audit/audit_bsm.c Thu Feb 20 16:58:19 2020 (r358172) @@ -830,6 +830,7 @@ kaudit_to_bsm(struct kaudit_record *kar, struct au_rec case AUE_UNLINK: case AUE_UNLINKAT: case AUE_UTIMES: + case AUE_REALPATHAT: ATFD1_TOKENS(1); UPATH1_VNODE1_TOKENS; break; Modified: head/sys/sys/param.h ============================================================================== --- head/sys/sys/param.h Thu Feb 20 16:38:30 2020 (r358171) +++ head/sys/sys/param.h Thu Feb 20 16:58:19 2020 (r358172) @@ -60,7 +60,7 @@ * in the range 5 to 9. */ #undef __FreeBSD_version -#define __FreeBSD_version 1300079 /* Master, propagated to newvers */ +#define __FreeBSD_version 1300080 /* Master, propagated to newvers */ /* * __FreeBSD_kernel__ indicates that this system uses the kernel of FreeBSD, Modified: head/usr.bin/truss/syscalls.c ============================================================================== --- head/usr.bin/truss/syscalls.c Thu Feb 20 16:38:30 2020 (r358171) +++ head/usr.bin/truss/syscalls.c Thu Feb 20 16:58:19 2020 (r358172) @@ -115,6 +115,9 @@ static struct syscall decoded_syscalls[] = { .args = { { Int, 0 }, { Int, 1 }, { CapRights | OUT, 2 } } }, { .name = "__getcwd", .ret_type = 1, .nargs = 2, .args = { { Name | OUT, 0 }, { Int, 1 } } }, + { .name = "__realpathat", .ret_type = 1, .nargs = 5, + .args = { { Atfd, 0 }, { Name | IN, 1 }, { Name | OUT, 2 }, + { Sizet, 3 }, { Int, 4} } }, { .name = "_umtx_op", .ret_type = 1, .nargs = 5, .args = { { Ptr, 0 }, { Umtxop, 1 }, { LongHex, 2 }, { Ptr, 3 }, { Ptr, 4 } } }, _______________________________________________ [email protected] mailing list https://lists.freebsd.org/mailman/listinfo/svn-src-head To unsubscribe, send any mail to "[email protected]"
