From owner-svn-src-all@freebsd.org Thu Aug 29 07:50:29 2019 Return-Path: Delivered-To: svn-src-all@mailman.nyi.freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2610:1c1:1:606c::19:1]) by mailman.nyi.freebsd.org (Postfix) with ESMTP id 33FE5CD643; Thu, 29 Aug 2019 07:50:29 +0000 (UTC) (envelope-from kib@FreeBSD.org) Received: from mxrelay.nyi.freebsd.org (mxrelay.nyi.freebsd.org [IPv6:2610:1c1:1:606c::19:3]) (using TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits) server-signature RSA-PSS (4096 bits) client-signature RSA-PSS (4096 bits) client-digest SHA256) (Client CN "mxrelay.nyi.freebsd.org", Issuer "Let's Encrypt Authority X3" (verified OK)) by mx1.freebsd.org (Postfix) with ESMTPS id 46Jvsx1Jnrz4LvV; Thu, 29 Aug 2019 07:50:29 +0000 (UTC) (envelope-from kib@FreeBSD.org) Received: from repo.freebsd.org (repo.freebsd.org [IPv6:2610:1c1:1:6068::e6a:0]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (Client did not present a certificate) by mxrelay.nyi.freebsd.org (Postfix) with ESMTPS id 1009A1CB5E; Thu, 29 Aug 2019 07:50:29 +0000 (UTC) (envelope-from kib@FreeBSD.org) Received: from repo.freebsd.org ([127.0.1.37]) by repo.freebsd.org (8.15.2/8.15.2) with ESMTP id x7T7oSxI075725; Thu, 29 Aug 2019 07:50:28 GMT (envelope-from kib@FreeBSD.org) Received: (from kib@localhost) by repo.freebsd.org (8.15.2/8.15.2/Submit) id x7T7oPce075709; Thu, 29 Aug 2019 07:50:25 GMT (envelope-from kib@FreeBSD.org) Message-Id: <201908290750.x7T7oPce075709@repo.freebsd.org> X-Authentication-Warning: repo.freebsd.org: kib set sender to kib@FreeBSD.org using -f From: Konstantin Belousov Date: Thu, 29 Aug 2019 07:50:25 +0000 (UTC) To: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org Subject: svn commit: r351598 - in head/sys: cddl/contrib/opensolaris/uts/common/fs/zfs fs/cd9660 fs/devfs fs/ext2fs fs/fuse fs/msdosfs fs/nfsclient fs/smbfs fs/tmpfs fs/udf kern ufs/ufs vm X-SVN-Group: head X-SVN-Commit-Author: kib X-SVN-Commit-Paths: in head/sys: cddl/contrib/opensolaris/uts/common/fs/zfs fs/cd9660 fs/devfs fs/ext2fs fs/fuse fs/msdosfs fs/nfsclient fs/smbfs fs/tmpfs fs/udf kern ufs/ufs vm X-SVN-Commit-Revision: 351598 X-SVN-Commit-Repository: base MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit X-BeenThere: svn-src-all@freebsd.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: "SVN commit messages for the entire src tree \(except for " user" and " projects" \)" List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Thu, 29 Aug 2019 07:50:29 -0000 Author: kib Date: Thu Aug 29 07:50:25 2019 New Revision: 351598 URL: https://svnweb.freebsd.org/changeset/base/351598 Log: Rework v_object lifecycle for vnodes. Current implementation of vnode_create_vobject() and vnode_destroy_vobject() is written so that it prepared to handle the vm object destruction for live vnode. Practically, no filesystems use this, except for some remnants that were present in UFS till today. One of the consequences of that model is that each filesystem must call vnode_destroy_vobject() in VOP_RECLAIM() or earlier, as result all of them get rid of the v_object in reclaim. Move the call to vnode_destroy_vobject() to vgonel() before VOP_RECLAIM(). This makes v_object stable: either the object is NULL, or it is valid vm object till the vnode reclamation. Remove code from vnode_create_vobject() to handle races with the parallel destruction. Reviewed by: markj Tested by: pho Sponsored by: The FreeBSD Foundation Differential revision: https://reviews.freebsd.org/D21412 Modified: head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_vnops.c head/sys/fs/cd9660/cd9660_node.c head/sys/fs/devfs/devfs_vnops.c head/sys/fs/ext2fs/ext2_inode.c head/sys/fs/fuse/fuse_vnops.c head/sys/fs/msdosfs/msdosfs_denode.c head/sys/fs/nfsclient/nfs_clnode.c head/sys/fs/smbfs/smbfs_node.c head/sys/fs/tmpfs/tmpfs_vnops.c head/sys/fs/udf/udf_vnops.c head/sys/kern/vfs_subr.c head/sys/ufs/ufs/ufs_extern.h head/sys/ufs/ufs/ufs_inode.c head/sys/vm/vm_object.h head/sys/vm/vnode_pager.c Modified: head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_vnops.c ============================================================================== --- head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_vnops.c Thu Aug 29 07:45:23 2019 (r351597) +++ head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_vnops.c Thu Aug 29 07:50:25 2019 (r351598) @@ -5375,9 +5375,6 @@ zfs_freebsd_reclaim(ap) ASSERT(zp != NULL); - /* Destroy the vm object and flush associated pages. */ - vnode_destroy_vobject(vp); - /* * z_teardown_inactive_lock protects from a race with * zfs_znode_dmu_fini in zfsvfs_teardown during Modified: head/sys/fs/cd9660/cd9660_node.c ============================================================================== --- head/sys/fs/cd9660/cd9660_node.c Thu Aug 29 07:45:23 2019 (r351597) +++ head/sys/fs/cd9660/cd9660_node.c Thu Aug 29 07:50:25 2019 (r351598) @@ -92,10 +92,6 @@ cd9660_reclaim(ap) struct vnode *vp = ap->a_vp; /* - * Destroy the vm object and flush associated pages. - */ - vnode_destroy_vobject(vp); - /* * Remove the inode from its hash chain. */ vfs_hash_remove(vp); Modified: head/sys/fs/devfs/devfs_vnops.c ============================================================================== --- head/sys/fs/devfs/devfs_vnops.c Thu Aug 29 07:45:23 2019 (r351597) +++ head/sys/fs/devfs/devfs_vnops.c Thu Aug 29 07:50:25 2019 (r351598) @@ -1428,7 +1428,6 @@ devfs_reclaim(struct vop_reclaim_args *ap) vp->v_data = NULL; } mtx_unlock(&devfs_de_interlock); - vnode_destroy_vobject(vp); return (0); } Modified: head/sys/fs/ext2fs/ext2_inode.c ============================================================================== --- head/sys/fs/ext2fs/ext2_inode.c Thu Aug 29 07:45:23 2019 (r351597) +++ head/sys/fs/ext2fs/ext2_inode.c Thu Aug 29 07:50:25 2019 (r351598) @@ -639,6 +639,5 @@ ext2_reclaim(struct vop_reclaim_args *ap) vfs_hash_remove(vp); free(vp->v_data, M_EXT2NODE); vp->v_data = 0; - vnode_destroy_vobject(vp); return (0); } Modified: head/sys/fs/fuse/fuse_vnops.c ============================================================================== --- head/sys/fs/fuse/fuse_vnops.c Thu Aug 29 07:45:23 2019 (r351597) +++ head/sys/fs/fuse/fuse_vnops.c Thu Aug 29 07:50:25 2019 (r351598) @@ -1537,7 +1537,6 @@ fuse_vnop_reclaim(struct vop_reclaim_args *ap) fuse_vnode_setparent(vp, NULL); cache_purge(vp); vfs_hash_remove(vp); - vnode_destroy_vobject(vp); fuse_vnode_destroy(vp); return 0; Modified: head/sys/fs/msdosfs/msdosfs_denode.c ============================================================================== --- head/sys/fs/msdosfs/msdosfs_denode.c Thu Aug 29 07:45:23 2019 (r351597) +++ head/sys/fs/msdosfs/msdosfs_denode.c Thu Aug 29 07:50:25 2019 (r351598) @@ -552,10 +552,6 @@ msdosfs_reclaim(struct vop_reclaim_args *ap) #endif /* - * Destroy the vm object and flush associated pages. - */ - vnode_destroy_vobject(vp); - /* * Remove the denode from its hash chain. */ vfs_hash_remove(vp); Modified: head/sys/fs/nfsclient/nfs_clnode.c ============================================================================== --- head/sys/fs/nfsclient/nfs_clnode.c Thu Aug 29 07:45:23 2019 (r351597) +++ head/sys/fs/nfsclient/nfs_clnode.c Thu Aug 29 07:50:25 2019 (r351598) @@ -296,11 +296,6 @@ ncl_reclaim(struct vop_reclaim_args *ap) ncl_releasesillyrename(vp, ap->a_td); mtx_unlock(&np->n_mtx); - /* - * Destroy the vm object and flush associated pages. - */ - vnode_destroy_vobject(vp); - if (NFS_ISV4(vp) && vp->v_type == VREG) /* * We can now safely close any remaining NFSv4 Opens for Modified: head/sys/fs/smbfs/smbfs_node.c ============================================================================== --- head/sys/fs/smbfs/smbfs_node.c Thu Aug 29 07:45:23 2019 (r351597) +++ head/sys/fs/smbfs/smbfs_node.c Thu Aug 29 07:50:25 2019 (r351598) @@ -272,10 +272,6 @@ smbfs_reclaim(ap) KASSERT((np->n_flag & NOPEN) == 0, ("file not closed before reclaim")); - /* - * Destroy the vm object and flush associated pages. - */ - vnode_destroy_vobject(vp); dvp = (np->n_parent && (np->n_flag & NREFPARENT)) ? np->n_parent : NULL; Modified: head/sys/fs/tmpfs/tmpfs_vnops.c ============================================================================== --- head/sys/fs/tmpfs/tmpfs_vnops.c Thu Aug 29 07:45:23 2019 (r351597) +++ head/sys/fs/tmpfs/tmpfs_vnops.c Thu Aug 29 07:50:25 2019 (r351598) @@ -1317,8 +1317,6 @@ tmpfs_reclaim(struct vop_reclaim_args *v) if (vp->v_type == VREG) tmpfs_destroy_vobject(vp, node->tn_reg.tn_aobj); - else - vnode_destroy_vobject(vp); vp->v_object = NULL; if (tmpfs_use_nc(vp)) cache_purge(vp); Modified: head/sys/fs/udf/udf_vnops.c ============================================================================== --- head/sys/fs/udf/udf_vnops.c Thu Aug 29 07:45:23 2019 (r351597) +++ head/sys/fs/udf/udf_vnops.c Thu Aug 29 07:50:25 2019 (r351598) @@ -1266,11 +1266,6 @@ udf_reclaim(struct vop_reclaim_args *a) vp = a->a_vp; unode = VTON(vp); - /* - * Destroy the vm object and flush associated pages. - */ - vnode_destroy_vobject(vp); - if (unode != NULL) { vfs_hash_remove(vp); Modified: head/sys/kern/vfs_subr.c ============================================================================== --- head/sys/kern/vfs_subr.c Thu Aug 29 07:45:23 2019 (r351597) +++ head/sys/kern/vfs_subr.c Thu Aug 29 07:50:25 2019 (r351598) @@ -3481,9 +3481,9 @@ static void vgonel(struct vnode *vp) { struct thread *td; - int oweinact; - int active; struct mount *mp; + vm_object_t object; + bool active, oweinact; ASSERT_VOP_ELOCKED(vp, "vgonel"); ASSERT_VI_LOCKED(vp, "vgonel"); @@ -3503,8 +3503,8 @@ vgonel(struct vnode *vp) * Check to see if the vnode is in use. If so, we have to call * VOP_CLOSE() and VOP_INACTIVE(). */ - active = vp->v_usecount; - oweinact = (vp->v_iflag & VI_OWEINACT); + active = vp->v_usecount > 0; + oweinact = (vp->v_iflag & VI_OWEINACT) != 0; VI_UNLOCK(vp); vfs_notify_upper(vp, VFS_NOTIFY_UPPER_RECLAIM); @@ -3543,12 +3543,24 @@ vgonel(struct vnode *vp) ("vp %p bufobj not invalidated", vp)); /* - * For VMIO bufobj, BO_DEAD is set in vm_object_terminate() - * after the object's page queue is flushed. + * For VMIO bufobj, BO_DEAD is set later, or in + * vm_object_terminate() after the object's page queue is + * flushed. */ - if (vp->v_bufobj.bo_object == NULL) + object = vp->v_bufobj.bo_object; + if (object == NULL) vp->v_bufobj.bo_flag |= BO_DEAD; BO_UNLOCK(&vp->v_bufobj); + + /* + * Handle the VM part. Tmpfs handles v_object on its own (the + * OBJT_VNODE check). Nullfs or other bypassing filesystems + * should not touch the object borrowed from the lower vnode + * (the handle check). + */ + if (object != NULL && object->type == OBJT_VNODE && + object->handle == vp) + vnode_destroy_vobject(vp); /* * Reclaim the vnode. Modified: head/sys/ufs/ufs/ufs_extern.h ============================================================================== --- head/sys/ufs/ufs/ufs_extern.h Thu Aug 29 07:45:23 2019 (r351597) +++ head/sys/ufs/ufs/ufs_extern.h Thu Aug 29 07:50:25 2019 (r351598) @@ -79,7 +79,6 @@ int ufs_inactive(struct vop_inactive_args *); int ufs_init(struct vfsconf *); void ufs_itimes(struct vnode *vp); int ufs_lookup(struct vop_cachedlookup_args *); -void ufs_prepare_reclaim(struct vnode *vp); int ufs_readdir(struct vop_readdir_args *); int ufs_reclaim(struct vop_reclaim_args *); void ffs_snapgone(struct inode *); Modified: head/sys/ufs/ufs/ufs_inode.c ============================================================================== --- head/sys/ufs/ufs/ufs_inode.c Thu Aug 29 07:45:23 2019 (r351597) +++ head/sys/ufs/ufs/ufs_inode.c Thu Aug 29 07:50:25 2019 (r351598) @@ -179,18 +179,21 @@ out: return (error); } -void -ufs_prepare_reclaim(struct vnode *vp) +/* + * Reclaim an inode so that it can be used for other purposes. + */ +int +ufs_reclaim(ap) + struct vop_reclaim_args /* { + struct vnode *a_vp; + struct thread *a_td; + } */ *ap; { - struct inode *ip; + struct vnode *vp = ap->a_vp; + struct inode *ip = VTOI(vp); #ifdef QUOTA int i; -#endif - ip = VTOI(vp); - - vnode_destroy_vobject(vp); -#ifdef QUOTA for (i = 0; i < MAXQUOTAS; i++) { if (ip->i_dquot[i] != NODQUOT) { dqrele(vp, ip->i_dquot[i]); @@ -202,22 +205,6 @@ ufs_prepare_reclaim(struct vnode *vp) if (ip->i_dirhash != NULL) ufsdirhash_free(ip); #endif -} - -/* - * Reclaim an inode so that it can be used for other purposes. - */ -int -ufs_reclaim(ap) - struct vop_reclaim_args /* { - struct vnode *a_vp; - struct thread *a_td; - } */ *ap; -{ - struct vnode *vp = ap->a_vp; - struct inode *ip = VTOI(vp); - - ufs_prepare_reclaim(vp); if (ip->i_flag & IN_LAZYMOD) ip->i_flag |= IN_MODIFIED; Modified: head/sys/vm/vm_object.h ============================================================================== --- head/sys/vm/vm_object.h Thu Aug 29 07:45:23 2019 (r351597) +++ head/sys/vm/vm_object.h Thu Aug 29 07:50:25 2019 (r351598) @@ -189,7 +189,6 @@ struct vm_object { #define OBJ_TMPFS_DIRTY 0x0400 /* dirty tmpfs obj */ #define OBJ_COLORED 0x1000 /* pg_color is defined */ #define OBJ_ONEMAPPING 0x2000 /* One USE (a single, non-forked) mapping flag */ -#define OBJ_DISCONNECTWNT 0x4000 /* disconnect from vnode wanted */ #define OBJ_TMPFS 0x8000 /* has tmpfs vnode allocated */ /* Modified: head/sys/vm/vnode_pager.c ============================================================================== --- head/sys/vm/vnode_pager.c Thu Aug 29 07:45:23 2019 (r351597) +++ head/sys/vm/vnode_pager.c Thu Aug 29 07:50:25 2019 (r351598) @@ -147,17 +147,9 @@ vnode_create_vobject(struct vnode *vp, off_t isize, st if (!vn_isdisk(vp, NULL) && vn_canvmio(vp) == FALSE) return (0); - while ((object = vp->v_object) != NULL) { - VM_OBJECT_WLOCK(object); - if (!(object->flags & OBJ_DEAD)) { - VM_OBJECT_WUNLOCK(object); - return (0); - } - VOP_UNLOCK(vp, 0); - vm_object_set_flag(object, OBJ_DISCONNECTWNT); - VM_OBJECT_SLEEP(object, object, PDROP | PVM, "vodead", 0); - vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); - } + object = vp->v_object; + if (object != NULL) + return (0); if (size == 0) { if (vn_isdisk(vp, NULL)) { @@ -190,10 +182,11 @@ vnode_destroy_vobject(struct vnode *vp) struct vm_object *obj; obj = vp->v_object; - if (obj == NULL) + if (obj == NULL || obj->handle != vp) return; ASSERT_VOP_ELOCKED(vp, "vnode_destroy_vobject"); VM_OBJECT_WLOCK(obj); + MPASS(obj->type == OBJT_VNODE); umtx_shm_object_terminated(obj); if (obj->ref_count == 0) { /* @@ -223,9 +216,6 @@ vnode_destroy_vobject(struct vnode *vp) * prevented new waiters from referencing the dying * object. */ - KASSERT((obj->flags & OBJ_DISCONNECTWNT) == 0, - ("OBJ_DISCONNECTWNT set obj %p flags %x", - obj, obj->flags)); vp->v_object = NULL; VM_OBJECT_WUNLOCK(obj); } @@ -243,8 +233,6 @@ vnode_destroy_vobject(struct vnode *vp) /* * Allocate (or lookup) pager for a vnode. * Handle is a vnode pointer. - * - * MPSAFE */ vm_object_t vnode_pager_alloc(void *handle, vm_ooffset_t size, vm_prot_t prot, @@ -259,28 +247,18 @@ vnode_pager_alloc(void *handle, vm_ooffset_t size, vm_ if (handle == NULL) return (NULL); - vp = (struct vnode *) handle; - - /* - * If the object is being terminated, wait for it to - * go away. - */ + vp = (struct vnode *)handle; + ASSERT_VOP_LOCKED(vp, "vnode_pager_alloc"); + KASSERT(vp->v_usecount != 0, ("vnode_pager_alloc: no vnode reference")); retry: - while ((object = vp->v_object) != NULL) { - VM_OBJECT_WLOCK(object); - if ((object->flags & OBJ_DEAD) == 0) - break; - vm_object_set_flag(object, OBJ_DISCONNECTWNT); - VM_OBJECT_SLEEP(object, object, PDROP | PVM, "vadead", 0); - } + object = vp->v_object; - KASSERT(vp->v_usecount != 0, ("vnode_pager_alloc: no vnode reference")); - if (object == NULL) { /* * Add an object of the appropriate size */ - object = vm_object_allocate(OBJT_VNODE, OFF_TO_IDX(round_page(size))); + object = vm_object_allocate(OBJT_VNODE, + OFF_TO_IDX(round_page(size))); object->un_pager.vnp.vnp_size = size; object->un_pager.vnp.writemappings = 0; @@ -290,7 +268,7 @@ retry: VI_LOCK(vp); if (vp->v_object != NULL) { /* - * Object has been created while we were sleeping + * Object has been created while we were allocating. */ VI_UNLOCK(vp); VM_OBJECT_WLOCK(object); @@ -305,6 +283,7 @@ retry: vp->v_object = object; VI_UNLOCK(vp); } else { + VM_OBJECT_WLOCK(object); object->ref_count++; #if VM_NRESERVLEVEL > 0 vm_object_color(object, 0); @@ -334,10 +313,6 @@ vnode_pager_dealloc(vm_object_t object) object->handle = NULL; object->type = OBJT_DEAD; - if (object->flags & OBJ_DISCONNECTWNT) { - vm_object_clear_flag(object, OBJ_DISCONNECTWNT); - wakeup(object); - } ASSERT_VOP_ELOCKED(vp, "vnode_pager_dealloc"); if (object->un_pager.vnp.writemappings > 0) { object->un_pager.vnp.writemappings = 0;