Date: Fri, 9 Oct 2020 19:10:00 +0000 (UTC) From: Mateusz Guzik <mjg@FreeBSD.org> To: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org Subject: svn commit: r366582 - in head/sys: fs/deadfs kern Message-ID: <202010091910.099JA08f087371@repo.freebsd.org>
next in thread | raw e-mail | index | archive | help
Author: mjg Date: Fri Oct 9 19:10:00 2020 New Revision: 366582 URL: https://svnweb.freebsd.org/changeset/base/366582 Log: cache: fix vexec panic when racing against vgone Use of dead_vnodeops would result in a panic instead of returning the intended EOPNOTSUPP error. While here make sure to abort, not just try to return a partial result. The former allows the regular lookup to restart from scratch, while the latter makes it stuck with an unusable vnode. Reported by: kevans Modified: head/sys/fs/deadfs/dead_vnops.c head/sys/kern/vfs_cache.c Modified: head/sys/fs/deadfs/dead_vnops.c ============================================================================== --- head/sys/fs/deadfs/dead_vnops.c Fri Oct 9 18:30:49 2020 (r366581) +++ head/sys/fs/deadfs/dead_vnops.c Fri Oct 9 19:10:00 2020 (r366582) @@ -79,6 +79,7 @@ struct vop_vector dead_vnodeops = { .vop_vptocnp = VOP_EBADF, .vop_unset_text = dead_unset_text, .vop_write = dead_write, + .vop_fplookup_vexec = VOP_EOPNOTSUPP, }; VFS_VOP_VECTOR_REGISTER(dead_vnodeops); Modified: head/sys/kern/vfs_cache.c ============================================================================== --- head/sys/kern/vfs_cache.c Fri Oct 9 18:30:49 2020 (r366581) +++ head/sys/kern/vfs_cache.c Fri Oct 9 19:10:00 2020 (r366582) @@ -4040,32 +4040,46 @@ cache_fplookup_parse_advance(struct cache_fpl *fpl) } } +/* + * See the API contract for VOP_FPLOOKUP_VEXEC. + */ static int __noinline cache_fplookup_failed_vexec(struct cache_fpl *fpl, int error) { + struct vnode *dvp; + seqc_t dvp_seqc; + dvp = fpl->dvp; + dvp_seqc = fpl->dvp_seqc; + /* * Hack: they may be looking up foo/bar, where foo is a * regular file. In such a case we need to turn ENOTDIR, * but we may happen to get here with a different error. */ - if (fpl->dvp->v_type != VDIR) { + if (dvp->v_type != VDIR) { + /* + * The check here is predominantly to catch + * EOPNOTSUPP from dead_vnodeops. If the vnode + * gets doomed past this point it is going to + * fail seqc verification. + */ + if (VN_IS_DOOMED(dvp)) { + return (cache_fpl_aborted(fpl)); + } error = ENOTDIR; } switch (error) { case EAGAIN: - /* - * Can happen when racing against vgone. - * */ - case EOPNOTSUPP: - cache_fpl_partial(fpl); + if (!vn_seqc_consistent(dvp, dvp_seqc)) { + error = cache_fpl_aborted(fpl); + } else { + cache_fpl_partial(fpl); + } break; default: - /* - * See the API contract for VOP_FPLOOKUP_VEXEC. - */ - if (!vn_seqc_consistent(fpl->dvp, fpl->dvp_seqc)) { + if (!vn_seqc_consistent(dvp, dvp_seqc)) { error = cache_fpl_aborted(fpl); } else { cache_fpl_smr_exit(fpl);
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?202010091910.099JA08f087371>