support for kernel events on vnodes, from jlemon@freebsd.org, okay art@

5 files changed, 137 insertions, 15 deletions

diff --git a/sys/kern/kern_event.c b/sys/kern/kern_event.c
index 4c5fed5e9ac..d407ba8ad27 100644
--- a/sys/kern/kern_event.c
+++ b/sys/kern/kern_event.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: kern_event.c,v 1.5 2000/11/18 22:16:49 provos Exp $	*/
+/*	$OpenBSD: kern_event.c,v 1.6 2000/11/21 21:49:57 provos Exp $	*/
 
 /*-
  * Copyright (c) 1999,2000 Jonathan Lemon <jlemon@FreeBSD.org>
@@ -100,8 +100,8 @@ extern struct filterops so_rwfiltops[];
 extern struct filterops pipe_rwfiltops[];
 #ifdef notyet
 extern struct filterops fifo_rwfiltops[];
-extern struct filterops vn_rwfiltops[];
 #endif
+extern struct filterops vn_rwfiltops[];
 
 struct filterops kq_rwfiltops[] = {
     { 1, filt_kqattach, filt_kqdetach, filt_kqueue },
@@ -111,8 +111,8 @@ struct filterops kq_rwfiltops[] = {
 extern struct filterops sig_filtops;
 #ifdef notyet
 extern struct filterops aio_filtops;
-extern struct filterops vn_filtops;
 #endif
+extern struct filterops vn_filtops;
 
 struct filterops rwtype_filtops =
 	{ 1, filt_rwtypattach, NULL, NULL };
@@ -125,7 +125,7 @@ struct filterops proc_filtops =
  */
 struct filterops *rwtypfilt_sw[] = {
 	NULL,				/* 0 */
-	NULL, /* vn_rwfiltops, */	/* DTYPE_VNODE */
+	vn_rwfiltops,			/* DTYPE_VNODE */
 	so_rwfiltops,			/* DTYPE_SOCKET */
 	pipe_rwfiltops,			/* DTYPE_PIPE */
 	/* fifo_rwfiltops, */		/* DTYPE_FIFO */
@@ -139,7 +139,7 @@ struct filterops *sysfilt_ops[] = {
 	&rwtype_filtops,		/* EVFILT_READ */
 	&rwtype_filtops,		/* EVFILT_WRITE */
 	NULL, /*&aio_filtops,*/		/* EVFILT_AIO */
-	NULL, /*&vn_filtops,*/		/* EVFILT_VNODE */
+	&vn_filtops,			/* EVFILT_VNODE */
 	&proc_filtops,			/* EVFILT_PROC */
 	&sig_filtops,			/* EVFILT_SIGNAL */
 };
diff --git a/sys/kern/vfs_vnops.c b/sys/kern/vfs_vnops.c
index b9435c46086..6bfcf9614ab 100644
--- a/sys/kern/vfs_vnops.c
+++ b/sys/kern/vfs_vnops.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: vfs_vnops.c,v 1.26 2000/04/25 02:10:04 deraadt Exp $	*/
+/*	$OpenBSD: vfs_vnops.c,v 1.27 2000/11/21 21:49:57 provos Exp $	*/
 /*	$NetBSD: vfs_vnops.c,v 1.20 1996/02/04 02:18:41 christos Exp $	*/
 
 /*
@@ -61,6 +61,9 @@
 #include <uvm/uvm_extern.h>
 #endif
 
+#include <ufs/ufs/quota.h>
+#include <ufs/ufs/inode.h>
+
 int	vn_read __P((struct file *fp, off_t *off, struct uio *uio, 
 	    struct ucred *cred));
 int	vn_write __P((struct file *fp, off_t *off, struct uio *uio, 
@@ -73,6 +76,25 @@ int	vn_ioctl __P((struct file *fp, u_long com, caddr_t data,
 struct 	fileops vnops =
 	{ vn_read, vn_write, vn_ioctl, vn_select, vn_closefile };
 
+static int	filt_nullattach(struct knote *kn);
+static int	filt_vnattach(struct knote *kn);
+static void	filt_vndetach(struct knote *kn);
+static int	filt_vnode(struct knote *kn, long hint);
+static int	filt_vnread(struct knote *kn, long hint);
+
+struct filterops vn_filtops =
+	{ 1, filt_vnattach, filt_vndetach, filt_vnode };
+
+/*
+ * XXX
+ * filt_vnread is ufs-specific, so the attach routine should really
+ * switch out to different filterops based on the vn filetype
+ */
+struct filterops vn_rwfiltops[] = {
+	{ 1, filt_vnattach, filt_vndetach, filt_vnread },
+	{ 1, filt_nullattach, NULL, NULL },
+};
+
 /*
  * Common code for vnode open operations.
  * Check permissions, and call the VOP_OPEN or VOP_CREATE routine.
@@ -508,3 +530,64 @@ vn_closefile(fp, p)
 	return (vn_close(((struct vnode *)fp->f_data), fp->f_flag,
 		fp->f_cred, p));
 }
+
+static int
+filt_vnattach(struct knote *kn)
+{
+	struct vnode *vp;
+
+	if (kn->kn_fp->f_type != DTYPE_VNODE)
+		return (EBADF);
+
+	vp = (struct vnode *)kn->kn_fp->f_data;
+
+	/*
+	 * XXX
+	 * this is a hack simply to cause the filter attach to fail
+	 * for non-ufs filesystems, until the support for them is done.
+	 */
+	if ((vp)->v_tag != VT_UFS || (vp)->v_type == VFIFO)
+		return (EOPNOTSUPP);
+
+	simple_lock(&vp->v_selectinfo.vsi_lock);
+	SLIST_INSERT_HEAD(&vp->v_selectinfo.vsi_selinfo.si_note, kn, kn_selnext);
+	simple_unlock(&vp->v_selectinfo.vsi_lock);
+
+	return (0);
+}
+
+static void
+filt_vndetach(struct knote *kn)
+{
+	struct vnode *vp = (struct vnode *)kn->kn_fp->f_data;
+
+	simple_lock(&vp->v_selectinfo.vsi_lock);
+	SLIST_REMOVE(&vp->v_selectinfo.vsi_selinfo.si_note,
+	    kn, knote, kn_selnext);
+	simple_unlock(&vp->v_selectinfo.vsi_lock);
+}
+
+static int
+filt_vnode(struct knote *kn, long hint)
+{
+	if (kn->kn_sfflags & hint)
+		kn->kn_fflags |= hint;
+	return (kn->kn_fflags != 0);
+}
+
+static int
+filt_nullattach(struct knote *kn)
+{
+	return (ENXIO);
+}
+
+/*ARGSUSED*/
+static int
+filt_vnread(struct knote *kn, long hint)
+{
+	struct vnode *vp = (struct vnode *)kn->kn_fp->f_data;
+	struct inode *ip = VTOI(vp);
+
+	kn->kn_data = ip->i_ffs_size - kn->kn_fp->f_offset;
+	return (kn->kn_data != 0);
+}
diff --git a/sys/sys/vnode.h b/sys/sys/vnode.h
index 47ffb2e8af3..9d5685c67e4 100644
--- a/sys/sys/vnode.h
+++ b/sys/sys/vnode.h
@@ -1,4 +1,4 @@
-/*	$OpenBSD: vnode.h,v 1.25 2000/06/17 17:16:05 provos Exp $	*/
+/*	$OpenBSD: vnode.h,v 1.26 2000/11/21 21:49:56 provos Exp $	*/
 /*	$NetBSD: vnode.h,v 1.38 1996/02/29 20:59:05 cgd Exp $	*/
 
 /*
@@ -38,6 +38,7 @@
 
 #include <sys/queue.h>
 #include <sys/lock.h>
+#include <sys/select.h>
 #ifdef UVM			/* XXX: clean up includes later */
 #include <vm/pglist.h>		/* XXX */
 #include <vm/vm_param.h>	/* XXX */
@@ -126,6 +127,10 @@ struct vnode {
 #endif
 	enum	vtagtype v_tag;			/* type of underlying data */
 	void 	*v_data;			/* private data for fs */
+	struct {
+		struct	simplelock vsi_lock;	/* lock to protect below */
+		struct	selinfo vsi_selinfo;	/* identity of poller(s) */
+	} v_selectinfo;
 };
 #define	v_mountedhere	v_un.vu_mountedhere
 #define	v_socket	v_un.vu_socket
diff --git a/sys/ufs/ufs/ufs_readwrite.c b/sys/ufs/ufs/ufs_readwrite.c
index 7a9ce5a44eb..4eb92eb225c 100644
--- a/sys/ufs/ufs/ufs_readwrite.c
+++ b/sys/ufs/ufs/ufs_readwrite.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: ufs_readwrite.c,v 1.15 1999/02/26 03:35:18 art Exp $	*/
+/*	$OpenBSD: ufs_readwrite.c,v 1.16 2000/11/21 21:49:57 provos Exp $	*/
 /*	$NetBSD: ufs_readwrite.c,v 1.9 1996/05/11 18:27:57 mycroft Exp $	*/
 
 /*-
@@ -57,6 +57,11 @@
 #define MAXFILESIZE		fs->fs_maxfilesize
 #endif
 
+#include <sys/event.h>
+
+#define VN_KNOTE(vp, b) \
+	KNOTE((struct klist *)&vp->v_selectinfo.vsi_selinfo.si_note, (b))
+
 /*
  * Vnode op for reading.
  */
@@ -186,9 +191,10 @@ WRITE(v)
 	struct proc *p;
 	daddr_t lbn;
 	off_t osize;
-	int blkoffset, error, flags, ioflag, resid, size, xfersize;
+	int blkoffset, error, extended, flags, ioflag, resid, size, xfersize;
 	struct timespec ts;
 
+	extended = 0;
 	ioflag = ap->a_ioflag;
 	uio = ap->a_uio;
 	vp = ap->a_vp;
@@ -257,6 +263,7 @@ WRITE(v)
 #else
 			vnode_pager_setsize(vp, (u_long)ip->i_ffs_size);
 #endif
+			extended = 1;
 		}
 #if defined(UVM)
 		(void)uvm_vnp_uncache(vp);
@@ -298,6 +305,8 @@ WRITE(v)
 	 */
 	if (resid > uio->uio_resid && ap->a_cred && ap->a_cred->cr_uid != 0)
 		ip->i_ffs_mode &= ~(ISUID | ISGID);
+	if (resid > uio->uio_resid)
+		VN_KNOTE(vp, NOTE_WRITE | (extended ? NOTE_EXTEND : 0));
 	if (error) {
 		if (ioflag & IO_UNIT) {
 			(void)VOP_TRUNCATE(vp, osize,
diff --git a/sys/ufs/ufs/ufs_vnops.c b/sys/ufs/ufs/ufs_vnops.c
index 5575728b316..2a648f44883 100644
--- a/sys/ufs/ufs/ufs_vnops.c
+++ b/sys/ufs/ufs/ufs_vnops.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: ufs_vnops.c,v 1.27 1999/03/09 21:16:28 art Exp $	*/
+/*	$OpenBSD: ufs_vnops.c,v 1.28 2000/11/21 21:49:57 provos Exp $	*/
 /*	$NetBSD: ufs_vnops.c,v 1.18 1996/05/11 18:28:04 mycroft Exp $	*/
 
 /*
@@ -56,6 +56,7 @@
 #include <sys/malloc.h>
 #include <sys/dirent.h>
 #include <sys/lockf.h>
+#include <sys/event.h>
 
 #include <vm/vm.h>
 
@@ -93,6 +94,8 @@ union _qcvt {
 	tmp.val[_QUAD_LOWWORD] = (l); \
 	(q) = tmp.qcvt; \
 }
+#define VN_KNOTE(vp, b) \
+	KNOTE(&vp->v_selectinfo.vsi_selinfo.si_note, (b))
 
 
 /*
@@ -120,9 +123,15 @@ ufs_create(v)
 		struct componentname *a_cnp;
 		struct vattr *a_vap;
 	} */ *ap = v;
-	return
+	int error;
+
+	error =
 	    ufs_makeinode(MAKEIMODE(ap->a_vap->va_type, ap->a_vap->va_mode),
 			  ap->a_dvp, ap->a_vpp, ap->a_cnp);
+	if (error)
+		return (error);
+	VN_KNOTE(ap->a_dvp, NOTE_WRITE);
+	return (0);
 }
 
 /*
@@ -148,6 +157,7 @@ ufs_mknod(v)
 	     ufs_makeinode(MAKEIMODE(vap->va_type, vap->va_mode),
 			   ap->a_dvp, vpp, ap->a_cnp)) != 0)
 		return (error);
+	VN_KNOTE(ap->a_dvp, NOTE_WRITE);
 	ip = VTOI(*vpp);
 	ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
 	if (vap->va_rdev != VNOVAL) {
@@ -433,6 +443,7 @@ ufs_setattr(v)
 			return (EROFS);
 		error = ufs_chmod(vp, (int)vap->va_mode, cred, p);
 	}
+	VN_KNOTE(vp, NOTE_ATTRIB);
 	return (error);
 }
 
@@ -677,11 +688,14 @@ ufs_remove(v)
 
 	ip = VTOI(vp);
 	if (vp->v_type == VDIR || (ip->i_ffs_flags & (IMMUTABLE | APPEND)) ||
-	    (VTOI(dvp)->i_ffs_flags & APPEND))
+	    (VTOI(dvp)->i_ffs_flags & APPEND)) {
 		error = EPERM;
-	else
-		error = ufs_dirremove(dvp, ip, ap->a_cnp->cn_flags, 0);
-
+		goto out;
+	}
+	error = ufs_dirremove(dvp, ip, ap->a_cnp->cn_flags, 0);
+	VN_KNOTE(vp, NOTE_DELETE);
+	VN_KNOTE(dvp, NOTE_WRITE);
+ out:
 	if (dvp == vp)
 		vrele(vp);
 	else
@@ -756,6 +770,8 @@ ufs_link(v)
 		ip->i_flag |= IN_CHANGE;
 	}
 	FREE(cnp->cn_pnbuf, M_NAMEI);
+	VN_KNOTE(vp, NOTE_LINK);
+	VN_KNOTE(dvp, NOTE_WRITE);
 out1:
 	if (dvp != vp)
 		VOP_UNLOCK(vp, 0, p);
@@ -973,6 +989,7 @@ abortit:
 		oldparent = dp->i_number;
 		doingdirectory = 1;
 	}
+	VN_KNOTE(fdvp, NOTE_WRITE);		/* XXX right place? */
 	/* Why? */
 	vrele(fdvp);
 
@@ -1079,6 +1096,7 @@ abortit:
 			}
 			goto bad;
 		}
+		VN_KNOTE(tdvp, NOTE_WRITE);
 		vput(tdvp);
 	} else {
 		if (xp->i_dev != dp->i_dev || xp->i_dev != ip->i_dev)
@@ -1153,7 +1171,9 @@ abortit:
 			        tcnp->cn_cred, tcnp->cn_proc)) != 0)
 				goto bad;
                 }
+		VN_KNOTE(tdvp, NOTE_WRITE);
 	        vput(tdvp);
+		VN_KNOTE(tvp, NOTE_DELETE);
 		vput(tvp);
 		xp = NULL;
 	}
@@ -1205,6 +1225,7 @@ abortit:
 		error = ufs_dirremove(fdvp, xp, fcnp->cn_flags, 0);
 		xp->i_flag &= ~IN_RENAME;
 	}
+	VN_KNOTE(fvp, NOTE_RENAME);
 	if (dp)
 		vput(fdvp);
 	if (xp)
@@ -1368,6 +1389,7 @@ ufs_mkdir(v)
   
 bad:
         if (error == 0) {
+		VN_KNOTE(dvp, NOTE_WRITE);
                 *ap->a_vpp = tvp;
         } else {
                 dp->i_effnlink--;
@@ -1449,6 +1471,7 @@ ufs_rmdir(v)
 	 */
 	if ((error = ufs_dirremove(dvp, ip, cnp->cn_flags, 1)) != 0)
 		goto out;
+	VN_KNOTE(dvp, NOTE_WRITE | NOTE_LINK);
 	cache_purge(dvp);
         /*
 	 * Truncate inode. The only stuff left in the directory is "." and
@@ -1471,6 +1494,7 @@ ufs_rmdir(v)
 	}
 	cache_purge(vp);
 out:
+	VN_KNOTE(vp, NOTE_DELETE);
         vput(dvp);
 	vput(vp);
 	return (error);
@@ -1498,6 +1522,7 @@ ufs_symlink(v)
 			      vpp, ap->a_cnp);
 	if (error)
 		return (error);
+	VN_KNOTE(ap->a_dvp, NOTE_WRITE);
 	vp = *vpp;
 	len = strlen(ap->a_target);
 	if (len < vp->v_mount->mnt_maxsymlinklen) {