svn commit: r185548 - in head: lib/libutil sys sys/contrib/pf sys/kern sys/sys usr.bin/procstat

Peter Wemm peter at FreeBSD.org
Mon Dec 1 22:50:27 PST 2008


Author: peter
Date: Tue Dec  2 06:50:26 2008
New Revision: 185548
URL: http://svn.freebsd.org/changeset/base/185548

Log:
  Merge user/peter/kinfo branch as of r185547 into head.
  
  This changes struct kinfo_filedesc and kinfo_vmentry such that they are
  same on both 32 and 64 bit platforms like i386/amd64 and won't require
  sysctl wrapping.
  
  Two new OIDs are assigned.  The old ones are available under
  COMPAT_FREEBSD7 - but it isn't that simple.  The superceded interface
  was never actually released on 7.x.
  
  The other main change is to pack the data passed to userland via the
  sysctl.  kf_structsize and kve_structsize are reduced for the copyout.
  If you have a process with 100,000+ sockets open, the unpacked records
  require a 132MB+ copyout.  With packing, it is "only" ~35MB.  (Still
  seriously unpleasant, but not quite as devastating).  A similar problem
  exists for the vmentry structure - have lots and lots of shared libraries
  and small mmaps and its copyout gets expensive too.
  
  My immediate problem is valgrind.  It traditionally achieves this
  functionality by parsing procfs output, in a packed format.  Secondly, when
  tracing 32 bit binaries on amd64 under valgrind, it uses a cross compiled
  32 bit binary which ran directly into the differing data structures in 32
  vs 64 bit mode.  (valgrind uses this to track file descriptor operations
  and this therefore affected every single 32 bit binary)
  
  I've added two utility functions to libutil to unpack the structures into
  a fixed record length and to make it a little more convenient to use.

Added:
  head/lib/libutil/kinfo_getfile.c
     - copied unchanged from r185545, user/peter/kinfo/lib/libutil/kinfo_getfile.c
  head/lib/libutil/kinfo_getvmmap.c
     - copied unchanged from r185545, user/peter/kinfo/lib/libutil/kinfo_getvmmap.c
Modified:
  head/lib/libutil/   (props changed)
  head/lib/libutil/Makefile
  head/lib/libutil/libutil.h
  head/sys/   (props changed)
  head/sys/contrib/pf/   (props changed)
  head/sys/kern/kern_descrip.c
  head/sys/kern/kern_proc.c
  head/sys/sys/sysctl.h
  head/sys/sys/user.h
  head/usr.bin/procstat/   (props changed)
  head/usr.bin/procstat/Makefile
  head/usr.bin/procstat/procstat_files.c
  head/usr.bin/procstat/procstat_vm.c

Modified: head/lib/libutil/Makefile
==============================================================================
--- head/lib/libutil/Makefile	Tue Dec  2 06:34:08 2008	(r185547)
+++ head/lib/libutil/Makefile	Tue Dec  2 06:50:26 2008	(r185548)
@@ -9,7 +9,8 @@ LIB=	util
 SHLIB_MAJOR= 7
 
 SRCS=	_secure_path.c auth.c expand_number.c flopen.c fparseln.c gr_util.c \
-	hexdump.c humanize_number.c kld.c login.c login_auth.c login_cap.c \
+	hexdump.c humanize_number.c kinfo_getfile.c kinfo_getvmmap.c kld.c \
+	login.c login_auth.c login_cap.c \
 	login_class.c login_crypt.c login_ok.c login_times.c login_tty.c \
 	logout.c logwtmp.c pidfile.c property.c pty.c pw_util.c realhostname.c \
 	stub.c trimdomain.c uucplock.c

Copied: head/lib/libutil/kinfo_getfile.c (from r185545, user/peter/kinfo/lib/libutil/kinfo_getfile.c)
==============================================================================
--- /dev/null	00:00:00 1970	(empty, because file is newly added)
+++ head/lib/libutil/kinfo_getfile.c	Tue Dec  2 06:50:26 2008	(r185548, copy of r185545, user/peter/kinfo/lib/libutil/kinfo_getfile.c)
@@ -0,0 +1,72 @@
+#include <sys/cdefs.h>
+__FBSDID("$FreeBSD$");
+
+#include <sys/types.h>
+#include <sys/user.h>
+#include <sys/sysctl.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include "libutil.h"
+
+struct kinfo_file *
+kinfo_getfile(pid_t pid, int *cntp)
+{
+	int mib[4];
+	int error;
+	int cnt;
+	size_t len;
+	char *buf, *bp, *eb;
+	struct kinfo_file *kif, *kp, *kf;
+
+	len = 0;
+	mib[0] = CTL_KERN;
+	mib[1] = KERN_PROC;
+	mib[2] = KERN_PROC_FILEDESC;
+	mib[3] = pid;
+
+	error = sysctl(mib, 4, NULL, &len, NULL, 0);
+	if (error)
+		return (0);
+	len = len * 4 / 3;
+	buf = malloc(len);
+	if (buf == NULL)
+		return (0);
+	error = sysctl(mib, 4, buf, &len, NULL, 0);
+	if (error) {
+		free(buf);
+		return (0);
+	}
+	/* Pass 1: count items */
+	cnt = 0;
+	bp = buf;
+	eb = buf + len;
+	while (bp < eb) {
+		kf = (struct kinfo_file *)bp;
+		bp += kf->kf_structsize;
+		cnt++;
+	}
+
+	kif = calloc(cnt, sizeof(*kif));
+	if (kif == NULL) {
+		free(buf);
+		return (0);
+	}
+	bp = buf;
+	eb = buf + len;
+	kp = kif;
+	/* Pass 2: unpack */
+	while (bp < eb) {
+		kf = (struct kinfo_file *)bp;
+		/* Copy/expand into pre-zeroed buffer */
+		memcpy(kp, kf, kf->kf_structsize);
+		/* Advance to next packed record */
+		bp += kf->kf_structsize;
+		/* Set field size to fixed length, advance */
+		kp->kf_structsize = sizeof(*kp);
+		kp++;
+	}
+	free(buf);
+	*cntp = cnt;
+	return (kif);	/* Caller must free() return value */
+}

Copied: head/lib/libutil/kinfo_getvmmap.c (from r185545, user/peter/kinfo/lib/libutil/kinfo_getvmmap.c)
==============================================================================
--- /dev/null	00:00:00 1970	(empty, because file is newly added)
+++ head/lib/libutil/kinfo_getvmmap.c	Tue Dec  2 06:50:26 2008	(r185548, copy of r185545, user/peter/kinfo/lib/libutil/kinfo_getvmmap.c)
@@ -0,0 +1,72 @@
+#include <sys/cdefs.h>
+__FBSDID("$FreeBSD$");
+
+#include <sys/types.h>
+#include <sys/user.h>
+#include <sys/sysctl.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include "libutil.h"
+
+struct kinfo_vmentry *
+kinfo_getvmmap(pid_t pid, int *cntp)
+{
+	int mib[4];
+	int error;
+	int cnt;
+	size_t len;
+	char *buf, *bp, *eb;
+	struct kinfo_vmentry *kiv, *kp, *kv;
+
+	len = 0;
+	mib[0] = CTL_KERN;
+	mib[1] = KERN_PROC;
+	mib[2] = KERN_PROC_VMMAP;
+	mib[3] = pid;
+
+	error = sysctl(mib, 4, NULL, &len, NULL, 0);
+	if (error)
+		return (0);
+	len = len * 4 / 3;
+	buf = malloc(len);
+	if (buf == NULL)
+		return (0);
+	error = sysctl(mib, 4, buf, &len, NULL, 0);
+	if (error) {
+		free(buf);
+		return (0);
+	}
+	/* Pass 1: count items */
+	cnt = 0;
+	bp = buf;
+	eb = buf + len;
+	while (bp < eb) {
+		kv = (struct kinfo_vmentry *)bp;
+		bp += kv->kve_structsize;
+		cnt++;
+	}
+
+	kiv = calloc(cnt, sizeof(*kiv));
+	if (kiv == NULL) {
+		free(buf);
+		return (0);
+	}
+	bp = buf;
+	eb = buf + len;
+	kp = kiv;
+	/* Pass 2: unpack */
+	while (bp < eb) {
+		kv = (struct kinfo_vmentry *)bp;
+		/* Copy/expand into pre-zeroed buffer */
+		memcpy(kp, kv, kv->kve_structsize);
+		/* Advance to next packed record */
+		bp += kv->kve_structsize;
+		/* Set field size to fixed length, advance */
+		kp->kve_structsize = sizeof(*kp);
+		kp++;
+	}
+	free(buf);
+	*cntp = cnt;
+	return (kiv);	/* Caller must free() return value */
+}

Modified: head/lib/libutil/libutil.h
==============================================================================
--- head/lib/libutil/libutil.h	Tue Dec  2 06:34:08 2008	(r185547)
+++ head/lib/libutil/libutil.h	Tue Dec  2 06:50:26 2008	(r185548)
@@ -64,6 +64,8 @@ struct termios;
 struct winsize;
 struct utmp;
 struct in_addr;
+struct kinfo_file;
+struct kinfo_vmentry;
 
 __BEGIN_DECLS
 void	clean_environment(const char * const *_white,
@@ -100,6 +102,10 @@ int	realhostname_sa(char *host, size_t h
 
 int	kld_isloaded(const char *name);
 int	kld_load(const char *name);
+struct kinfo_file *
+	kinfo_getfile(pid_t _pid, int *_cntp);
+struct kinfo_vmentry *
+	kinfo_getvmmap(pid_t _pid, int *_cntp);
 
 #ifdef _STDIO_H_	/* avoid adding new includes */
 char   *fparseln(FILE *, size_t *, size_t *, const char[3], int);

Modified: head/sys/kern/kern_descrip.c
==============================================================================
--- head/sys/kern/kern_descrip.c	Tue Dec  2 06:34:08 2008	(r185547)
+++ head/sys/kern/kern_descrip.c	Tue Dec  2 06:50:26 2008	(r185548)
@@ -2509,6 +2509,259 @@ sysctl_kern_file(SYSCTL_HANDLER_ARGS)
 SYSCTL_PROC(_kern, KERN_FILE, file, CTLTYPE_OPAQUE|CTLFLAG_RD,
     0, 0, sysctl_kern_file, "S,xfile", "Entire file table");
 
+#ifdef KINFO_OFILE_SIZE
+CTASSERT(sizeof(struct kinfo_ofile) == KINFO_OFILE_SIZE);
+#endif
+
+#ifdef COMPAT_FREEBSD7
+static int
+export_vnode_for_osysctl(struct vnode *vp, int type,
+    struct kinfo_ofile *kif, struct filedesc *fdp, struct sysctl_req *req)
+{
+	int error;
+	char *fullpath, *freepath;
+	int vfslocked;
+
+	bzero(kif, sizeof(*kif));
+	kif->kf_structsize = sizeof(*kif);
+
+	vref(vp);
+	kif->kf_fd = type;
+	kif->kf_type = KF_TYPE_VNODE;
+	/* This function only handles directories. */
+	KASSERT(vp->v_type == VDIR, ("export_vnode_for_osysctl: vnode not directory"));
+	kif->kf_vnode_type = KF_VTYPE_VDIR;
+
+	/*
+	 * This is not a true file descriptor, so we set a bogus refcount
+	 * and offset to indicate these fields should be ignored.
+	 */
+	kif->kf_ref_count = -1;
+	kif->kf_offset = -1;
+
+	freepath = NULL;
+	fullpath = "-";
+	FILEDESC_SUNLOCK(fdp);
+	vn_fullpath(curthread, vp, &fullpath, &freepath);
+	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
+	vrele(vp);
+	VFS_UNLOCK_GIANT(vfslocked);
+	strlcpy(kif->kf_path, fullpath, sizeof(kif->kf_path));
+	if (freepath != NULL)
+		free(freepath, M_TEMP);
+	error = SYSCTL_OUT(req, kif, sizeof(*kif));
+	FILEDESC_SLOCK(fdp);
+	return (error);
+}
+
+/*
+ * Get per-process file descriptors for use by procstat(1), et al.
+ */
+static int
+sysctl_kern_proc_ofiledesc(SYSCTL_HANDLER_ARGS)
+{
+	char *fullpath, *freepath;
+	struct kinfo_ofile *kif;
+	struct filedesc *fdp;
+	int error, i, *name;
+	struct socket *so;
+	struct vnode *vp;
+	struct file *fp;
+	struct proc *p;
+	struct tty *tp;
+	int vfslocked;
+
+	name = (int *)arg1;
+	if ((p = pfind((pid_t)name[0])) == NULL)
+		return (ESRCH);
+	if ((error = p_candebug(curthread, p))) {
+		PROC_UNLOCK(p);
+		return (error);
+	}
+	fdp = fdhold(p);
+	PROC_UNLOCK(p);
+	if (fdp == NULL)
+		return (ENOENT);
+	kif = malloc(sizeof(*kif), M_TEMP, M_WAITOK);
+	FILEDESC_SLOCK(fdp);
+	if (fdp->fd_cdir != NULL)
+		export_vnode_for_osysctl(fdp->fd_cdir, KF_FD_TYPE_CWD, kif,
+				fdp, req);
+	if (fdp->fd_rdir != NULL)
+		export_vnode_for_osysctl(fdp->fd_rdir, KF_FD_TYPE_ROOT, kif,
+				fdp, req);
+	if (fdp->fd_jdir != NULL)
+		export_vnode_for_osysctl(fdp->fd_jdir, KF_FD_TYPE_JAIL, kif,
+				fdp, req);
+	for (i = 0; i < fdp->fd_nfiles; i++) {
+		if ((fp = fdp->fd_ofiles[i]) == NULL)
+			continue;
+		bzero(kif, sizeof(*kif));
+		kif->kf_structsize = sizeof(*kif);
+		vp = NULL;
+		so = NULL;
+		tp = NULL;
+		kif->kf_fd = i;
+		switch (fp->f_type) {
+		case DTYPE_VNODE:
+			kif->kf_type = KF_TYPE_VNODE;
+			vp = fp->f_vnode;
+			break;
+
+		case DTYPE_SOCKET:
+			kif->kf_type = KF_TYPE_SOCKET;
+			so = fp->f_data;
+			break;
+
+		case DTYPE_PIPE:
+			kif->kf_type = KF_TYPE_PIPE;
+			break;
+
+		case DTYPE_FIFO:
+			kif->kf_type = KF_TYPE_FIFO;
+			vp = fp->f_vnode;
+			vref(vp);
+			break;
+
+		case DTYPE_KQUEUE:
+			kif->kf_type = KF_TYPE_KQUEUE;
+			break;
+
+		case DTYPE_CRYPTO:
+			kif->kf_type = KF_TYPE_CRYPTO;
+			break;
+
+		case DTYPE_MQUEUE:
+			kif->kf_type = KF_TYPE_MQUEUE;
+			break;
+
+		case DTYPE_SHM:
+			kif->kf_type = KF_TYPE_SHM;
+			break;
+
+		case DTYPE_SEM:
+			kif->kf_type = KF_TYPE_SEM;
+			break;
+
+		case DTYPE_PTS:
+			kif->kf_type = KF_TYPE_PTS;
+			tp = fp->f_data;
+			break;
+
+		default:
+			kif->kf_type = KF_TYPE_UNKNOWN;
+			break;
+		}
+		kif->kf_ref_count = fp->f_count;
+		if (fp->f_flag & FREAD)
+			kif->kf_flags |= KF_FLAG_READ;
+		if (fp->f_flag & FWRITE)
+			kif->kf_flags |= KF_FLAG_WRITE;
+		if (fp->f_flag & FAPPEND)
+			kif->kf_flags |= KF_FLAG_APPEND;
+		if (fp->f_flag & FASYNC)
+			kif->kf_flags |= KF_FLAG_ASYNC;
+		if (fp->f_flag & FFSYNC)
+			kif->kf_flags |= KF_FLAG_FSYNC;
+		if (fp->f_flag & FNONBLOCK)
+			kif->kf_flags |= KF_FLAG_NONBLOCK;
+		if (fp->f_flag & O_DIRECT)
+			kif->kf_flags |= KF_FLAG_DIRECT;
+		if (fp->f_flag & FHASLOCK)
+			kif->kf_flags |= KF_FLAG_HASLOCK;
+		kif->kf_offset = fp->f_offset;
+		if (vp != NULL) {
+			vref(vp);
+			switch (vp->v_type) {
+			case VNON:
+				kif->kf_vnode_type = KF_VTYPE_VNON;
+				break;
+			case VREG:
+				kif->kf_vnode_type = KF_VTYPE_VREG;
+				break;
+			case VDIR:
+				kif->kf_vnode_type = KF_VTYPE_VDIR;
+				break;
+			case VBLK:
+				kif->kf_vnode_type = KF_VTYPE_VBLK;
+				break;
+			case VCHR:
+				kif->kf_vnode_type = KF_VTYPE_VCHR;
+				break;
+			case VLNK:
+				kif->kf_vnode_type = KF_VTYPE_VLNK;
+				break;
+			case VSOCK:
+				kif->kf_vnode_type = KF_VTYPE_VSOCK;
+				break;
+			case VFIFO:
+				kif->kf_vnode_type = KF_VTYPE_VFIFO;
+				break;
+			case VBAD:
+				kif->kf_vnode_type = KF_VTYPE_VBAD;
+				break;
+			default:
+				kif->kf_vnode_type = KF_VTYPE_UNKNOWN;
+				break;
+			}
+			/*
+			 * It is OK to drop the filedesc lock here as we will
+			 * re-validate and re-evaluate its properties when
+			 * the loop continues.
+			 */
+			freepath = NULL;
+			fullpath = "-";
+			FILEDESC_SUNLOCK(fdp);
+			vn_fullpath(curthread, vp, &fullpath, &freepath);
+			vfslocked = VFS_LOCK_GIANT(vp->v_mount);
+			vrele(vp);
+			VFS_UNLOCK_GIANT(vfslocked);
+			strlcpy(kif->kf_path, fullpath,
+			    sizeof(kif->kf_path));
+			if (freepath != NULL)
+				free(freepath, M_TEMP);
+			FILEDESC_SLOCK(fdp);
+		}
+		if (so != NULL) {
+			struct sockaddr *sa;
+
+			if (so->so_proto->pr_usrreqs->pru_sockaddr(so, &sa)
+			    == 0 && sa->sa_len <= sizeof(kif->kf_sa_local)) {
+				bcopy(sa, &kif->kf_sa_local, sa->sa_len);
+				free(sa, M_SONAME);
+			}
+			if (so->so_proto->pr_usrreqs->pru_peeraddr(so, &sa)
+			    == 00 && sa->sa_len <= sizeof(kif->kf_sa_peer)) {
+				bcopy(sa, &kif->kf_sa_peer, sa->sa_len);
+				free(sa, M_SONAME);
+			}
+			kif->kf_sock_domain =
+			    so->so_proto->pr_domain->dom_family;
+			kif->kf_sock_type = so->so_type;
+			kif->kf_sock_protocol = so->so_proto->pr_protocol;
+		}
+		if (tp != NULL) {
+			strlcpy(kif->kf_path, tty_devname(tp),
+			    sizeof(kif->kf_path));
+		}
+		error = SYSCTL_OUT(req, kif, sizeof(*kif));
+		if (error)
+			break;
+	}
+	FILEDESC_SUNLOCK(fdp);
+	fddrop(fdp);
+	free(kif, M_TEMP);
+	return (0);
+}
+
+static SYSCTL_NODE(_kern_proc, KERN_PROC_OFILEDESC, ofiledesc, CTLFLAG_RD,
+    sysctl_kern_proc_ofiledesc, "Process ofiledesc entries");
+#endif	/* COMPAT_FREEBSD7 */
+
+#ifdef KINFO_FILE_SIZE
+CTASSERT(sizeof(struct kinfo_file) == KINFO_FILE_SIZE);
+#endif
+
 static int
 export_vnode_for_sysctl(struct vnode *vp, int type,
     struct kinfo_file *kif, struct filedesc *fdp, struct sysctl_req *req)
@@ -2518,7 +2771,6 @@ export_vnode_for_sysctl(struct vnode *vp
 	int vfslocked;
 
 	bzero(kif, sizeof(*kif));
-	kif->kf_structsize = sizeof(*kif);
 
 	vref(vp);
 	kif->kf_fd = type;
@@ -2544,7 +2796,11 @@ export_vnode_for_sysctl(struct vnode *vp
 	strlcpy(kif->kf_path, fullpath, sizeof(kif->kf_path));
 	if (freepath != NULL)
 		free(freepath, M_TEMP);
-	error = SYSCTL_OUT(req, kif, sizeof(*kif));
+	/* Pack record size down */
+	kif->kf_structsize = offsetof(struct kinfo_file, kf_path) +
+	    strlen(kif->kf_path) + 1;
+	kif->kf_structsize = roundup(kif->kf_structsize, sizeof(uint64_t));
+	error = SYSCTL_OUT(req, kif, kif->kf_structsize);
 	FILEDESC_SLOCK(fdp);
 	return (error);
 }
@@ -2592,7 +2848,6 @@ sysctl_kern_proc_filedesc(SYSCTL_HANDLER
 		if ((fp = fdp->fd_ofiles[i]) == NULL)
 			continue;
 		bzero(kif, sizeof(*kif));
-		kif->kf_structsize = sizeof(*kif);
 		vp = NULL;
 		so = NULL;
 		tp = NULL;
@@ -2739,7 +2994,12 @@ sysctl_kern_proc_filedesc(SYSCTL_HANDLER
 			strlcpy(kif->kf_path, tty_devname(tp),
 			    sizeof(kif->kf_path));
 		}
-		error = SYSCTL_OUT(req, kif, sizeof(*kif));
+		/* Pack record size down */
+		kif->kf_structsize = offsetof(struct kinfo_file, kf_path) +
+		    strlen(kif->kf_path) + 1;
+		kif->kf_structsize = roundup(kif->kf_structsize,
+		    sizeof(uint64_t));
+		error = SYSCTL_OUT(req, kif, kif->kf_structsize);
 		if (error)
 			break;
 	}

Modified: head/sys/kern/kern_proc.c
==============================================================================
--- head/sys/kern/kern_proc.c	Tue Dec  2 06:34:08 2008	(r185547)
+++ head/sys/kern/kern_proc.c	Tue Dec  2 06:50:26 2008	(r185548)
@@ -32,6 +32,7 @@
 #include <sys/cdefs.h>
 __FBSDID("$FreeBSD$");
 
+#include "opt_compat.h"
 #include "opt_ddb.h"
 #include "opt_kdtrace.h"
 #include "opt_ktrace.h"
@@ -1337,13 +1338,18 @@ sysctl_kern_proc_sv_name(SYSCTL_HANDLER_
 	return (sysctl_handle_string(oidp, sv_name, 0, req));
 }
 
+#ifdef KINFO_OVMENTRY_SIZE
+CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
+#endif
+
+#ifdef COMPAT_FREEBSD7
 static int
-sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
+sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
 {
 	vm_map_entry_t entry, tmp_entry;
 	unsigned int last_timestamp;
 	char *fullpath, *freepath;
-	struct kinfo_vmentry *kve;
+	struct kinfo_ovmentry *kve;
 	struct vattr va;
 	struct ucred *cred;
 	int error, *name;
@@ -1497,6 +1503,176 @@ sysctl_kern_proc_vmmap(SYSCTL_HANDLER_AR
 	free(kve, M_TEMP);
 	return (error);
 }
+#endif	/* COMPAT_FREEBSD7 */
+
+#ifdef KINFO_VMENTRY_SIZE
+CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
+#endif
+
+static int
+sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
+{
+	vm_map_entry_t entry, tmp_entry;
+	unsigned int last_timestamp;
+	char *fullpath, *freepath;
+	struct kinfo_vmentry *kve;
+	struct vattr va;
+	struct ucred *cred;
+	int error, *name;
+	struct vnode *vp;
+	struct proc *p;
+	vm_map_t map;
+
+	name = (int *)arg1;
+	if ((p = pfind((pid_t)name[0])) == NULL)
+		return (ESRCH);
+	if (p->p_flag & P_WEXIT) {
+		PROC_UNLOCK(p);
+		return (ESRCH);
+	}
+	if ((error = p_candebug(curthread, p))) {
+		PROC_UNLOCK(p);
+		return (error);
+	}
+	_PHOLD(p);
+	PROC_UNLOCK(p);
+
+	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
+
+	map = &p->p_vmspace->vm_map;	/* XXXRW: More locking required? */
+	vm_map_lock_read(map);
+	for (entry = map->header.next; entry != &map->header;
+	    entry = entry->next) {
+		vm_object_t obj, tobj, lobj;
+		vm_offset_t addr;
+		int vfslocked;
+
+		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
+			continue;
+
+		bzero(kve, sizeof(*kve));
+
+		kve->kve_private_resident = 0;
+		obj = entry->object.vm_object;
+		if (obj != NULL) {
+			VM_OBJECT_LOCK(obj);
+			if (obj->shadow_count == 1)
+				kve->kve_private_resident =
+				    obj->resident_page_count;
+		}
+		kve->kve_resident = 0;
+		addr = entry->start;
+		while (addr < entry->end) {
+			if (pmap_extract(map->pmap, addr))
+				kve->kve_resident++;
+			addr += PAGE_SIZE;
+		}
+
+		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
+			if (tobj != obj)
+				VM_OBJECT_LOCK(tobj);
+			if (lobj != obj)
+				VM_OBJECT_UNLOCK(lobj);
+			lobj = tobj;
+		}
+
+		kve->kve_fileid = 0;
+		kve->kve_fsid = 0;
+		freepath = NULL;
+		fullpath = "";
+		if (lobj) {
+			vp = NULL;
+			switch(lobj->type) {
+			case OBJT_DEFAULT:
+				kve->kve_type = KVME_TYPE_DEFAULT;
+				break;
+			case OBJT_VNODE:
+				kve->kve_type = KVME_TYPE_VNODE;
+				vp = lobj->handle;
+				vref(vp);
+				break;
+			case OBJT_SWAP:
+				kve->kve_type = KVME_TYPE_SWAP;
+				break;
+			case OBJT_DEVICE:
+				kve->kve_type = KVME_TYPE_DEVICE;
+				break;
+			case OBJT_PHYS:
+				kve->kve_type = KVME_TYPE_PHYS;
+				break;
+			case OBJT_DEAD:
+				kve->kve_type = KVME_TYPE_DEAD;
+				break;
+			default:
+				kve->kve_type = KVME_TYPE_UNKNOWN;
+				break;
+			}
+			if (lobj != obj)
+				VM_OBJECT_UNLOCK(lobj);
+
+			kve->kve_ref_count = obj->ref_count;
+			kve->kve_shadow_count = obj->shadow_count;
+			VM_OBJECT_UNLOCK(obj);
+			if (vp != NULL) {
+				vn_fullpath(curthread, vp, &fullpath,
+				    &freepath);
+				cred = curthread->td_ucred;
+				vfslocked = VFS_LOCK_GIANT(vp->v_mount);
+				vn_lock(vp, LK_SHARED | LK_RETRY);
+				if (VOP_GETATTR(vp, &va, cred) == 0) {
+					kve->kve_fileid = va.va_fileid;
+					kve->kve_fsid = va.va_fsid;
+				}
+				vput(vp);
+				VFS_UNLOCK_GIANT(vfslocked);
+			}
+		} else {
+			kve->kve_type = KVME_TYPE_NONE;
+			kve->kve_ref_count = 0;
+			kve->kve_shadow_count = 0;
+		}
+
+		kve->kve_start = entry->start;
+		kve->kve_end = entry->end;
+		kve->kve_offset = entry->offset;
+
+		if (entry->protection & VM_PROT_READ)
+			kve->kve_protection |= KVME_PROT_READ;
+		if (entry->protection & VM_PROT_WRITE)
+			kve->kve_protection |= KVME_PROT_WRITE;
+		if (entry->protection & VM_PROT_EXECUTE)
+			kve->kve_protection |= KVME_PROT_EXEC;
+
+		if (entry->eflags & MAP_ENTRY_COW)
+			kve->kve_flags |= KVME_FLAG_COW;
+		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
+			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
+
+		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
+		if (freepath != NULL)
+			free(freepath, M_TEMP);
+
+		last_timestamp = map->timestamp;
+		vm_map_unlock_read(map);
+		/* Pack record size down */
+		kve->kve_structsize = offsetof(struct kinfo_vmentry, kve_path) +
+		    strlen(kve->kve_path) + 1;
+		kve->kve_structsize = roundup(kve->kve_structsize,
+		    sizeof(uint64_t));
+		error = SYSCTL_OUT(req, kve, kve->kve_structsize);
+		vm_map_lock_read(map);
+		if (error)
+			break;
+		if (last_timestamp + 1 != map->timestamp) {
+			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
+			entry = tmp_entry;
+		}
+	}
+	vm_map_unlock_read(map);
+	PRELE(p);
+	free(kve, M_TEMP);
+	return (error);
+}
 
 #if defined(STACK) || defined(DDB)
 static int
@@ -1669,6 +1845,11 @@ static SYSCTL_NODE(_kern_proc, (KERN_PRO
 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
 	CTLFLAG_RD, sysctl_kern_proc, "Return process table, no threads");
 
+#ifdef COMPAT_FREEBSD7
+static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD,
+	sysctl_kern_proc_ovmmap, "Old Process vm map entries");
+#endif
+
 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD,
 	sysctl_kern_proc_vmmap, "Process vm map entries");
 

Modified: head/sys/sys/sysctl.h
==============================================================================
--- head/sys/sys/sysctl.h	Tue Dec  2 06:34:08 2008	(r185547)
+++ head/sys/sys/sysctl.h	Tue Dec  2 06:50:26 2008	(r185548)
@@ -540,13 +540,16 @@ TAILQ_HEAD(sysctl_ctx_list, sysctl_ctx_e
 #define	KERN_PROC_RGID		10	/* by real group id */
 #define	KERN_PROC_GID		11	/* by effective group id */
 #define	KERN_PROC_PATHNAME	12	/* path to executable */
-#define	KERN_PROC_VMMAP		13	/* VM map entries for process */
-#define	KERN_PROC_FILEDESC	14	/* File descriptors for process */
+#define	KERN_PROC_OVMMAP	13	/* Old VM map entries for process */
+#define	KERN_PROC_OFILEDESC	14	/* Old file descriptors for process */
 #define	KERN_PROC_KSTACK	15	/* Kernel stacks for process */
 #define	KERN_PROC_INC_THREAD	0x10	/*
 					 * modifier for pid, pgrp, tty,
 					 * uid, ruid, gid, rgid and proc
+					 * This effectively uses 16-31
 					 */
+#define	KERN_PROC_VMMAP		32	/* VM map entries for process */
+#define	KERN_PROC_FILEDESC	33	/* File descriptors for process */
 
 /*
  * KERN_IPC identifiers

Modified: head/sys/sys/user.h
==============================================================================
--- head/sys/sys/user.h	Tue Dec  2 06:34:08 2008	(r185547)
+++ head/sys/sys/user.h	Tue Dec  2 06:50:26 2008	(r185548)
@@ -277,20 +277,55 @@ struct user {
 #define	KF_FLAG_DIRECT		0x00000040
 #define	KF_FLAG_HASLOCK		0x00000080
 
-struct kinfo_file {
+/*
+ * Old format.  Has variable hidden padding due to alignment.
+ * This is a compatability hack for pre-build 7.1 packages.
+ */
+#if defined(__amd64__)
+#define	KINFO_OFILE_SIZE	1328
+#endif
+#if defined(__i386__)
+#define	KINFO_OFILE_SIZE	1324
+#endif
+
+struct kinfo_ofile {
 	int	kf_structsize;			/* Size of kinfo_file. */
 	int	kf_type;			/* Descriptor type. */
 	int	kf_fd;				/* Array index. */
 	int	kf_ref_count;			/* Reference count. */
 	int	kf_flags;			/* Flags. */
+	/* XXX Hidden alignment padding here on amd64 */
 	off_t	kf_offset;			/* Seek location. */
 	int	kf_vnode_type;			/* Vnode type. */
 	int	kf_sock_domain;			/* Socket domain. */
 	int	kf_sock_type;			/* Socket type. */
 	int	kf_sock_protocol;		/* Socket protocol. */
-	char	kf_path[PATH_MAX];		/* Path to file, if any. */
+	char	kf_path[PATH_MAX];	/* Path to file, if any. */
+	struct sockaddr_storage kf_sa_local;	/* Socket address. */
+	struct sockaddr_storage	kf_sa_peer;	/* Peer address. */
+};
+
+#if defined(__amd64__) || defined(__i386__)
+#define	KINFO_FILE_SIZE	1392
+#endif
+
+struct kinfo_file {
+	int	kf_structsize;			/* Variable size of record. */
+	int	kf_type;			/* Descriptor type. */
+	int	kf_fd;				/* Array index. */
+	int	kf_ref_count;			/* Reference count. */
+	int	kf_flags;			/* Flags. */
+	int	_kf_pad0;			/* Round to 64 bit alignment */
+	uint64_t kf_offset;			/* Seek location. */
+	int	kf_vnode_type;			/* Vnode type. */
+	int	kf_sock_domain;			/* Socket domain. */
+	int	kf_sock_type;			/* Socket type. */
+	int	kf_sock_protocol;		/* Socket protocol. */
 	struct sockaddr_storage kf_sa_local;	/* Socket address. */
 	struct sockaddr_storage	kf_sa_peer;	/* Peer address. */
+	int	_kf_ispare[16];			/* Space for more stuff. */
+	/* Truncated before copyout in sysctl */
+	char	kf_path[PATH_MAX];		/* Path to file, if any. */
 };
 
 /*
@@ -313,11 +348,18 @@ struct kinfo_file {
 #define	KVME_FLAG_COW		0x00000001
 #define	KVME_FLAG_NEEDS_COPY	0x00000002
 
-struct kinfo_vmentry {
+#if defined(__amd64__)
+#define	KINFO_OVMENTRY_SIZE	1168
+#endif
+#if defined(__i386__)
+#define	KINFO_OVMENTRY_SIZE	1128
+#endif
+
+struct kinfo_ovmentry {
 	int	 kve_structsize;		/* Size of kinfo_vmmapentry. */
 	int	 kve_type;			/* Type of map entry. */
-	void	*kve_start;			/* Starting pointer. */
-	void	*kve_end;			/* Finishing pointer. */
+	void	*kve_start;			/* Starting address. */
+	void	*kve_end;			/* Finishing address. */
 	int	 kve_flags;			/* Flags on map entry. */
 	int	 kve_resident;			/* Number of resident pages. */
 	int	 kve_private_resident;		/* Number of private pages. */
@@ -327,11 +369,35 @@ struct kinfo_vmentry {
 	char	 kve_path[PATH_MAX];		/* Path to VM obj, if any. */
 	void	*_kve_pspare[8];		/* Space for more stuff. */
 	off_t	 kve_offset;			/* Mapping offset in object */
-	uint64_t kve_fileid;			/* inode number of vnode */
+	uint64_t kve_fileid;			/* inode number if vnode */
 	dev_t	 kve_fsid;			/* dev_t of vnode location */
 	int	 _kve_ispare[3];		/* Space for more stuff. */
 };
 
+#if defined(__amd64__) || defined(__i386__)
+#define	KINFO_VMENTRY_SIZE	1160
+#endif
+
+struct kinfo_vmentry {
+	int	 kve_structsize;		/* Variable size of record. */
+	int	 kve_type;			/* Type of map entry. */
+	uint64_t kve_start;			/* Starting address. */
+	uint64_t kve_end;			/* Finishing address. */
+	uint64_t kve_offset;			/* Mapping offset in object */
+	uint64_t kve_fileid;			/* inode number if vnode */
+	uint32_t kve_fsid;			/* dev_t of vnode location */
+	int	 kve_flags;			/* Flags on map entry. */
+	int	 kve_resident;			/* Number of resident pages. */
+	int	 kve_private_resident;		/* Number of private pages. */
+	int	 kve_protection;		/* Protection bitmask. */
+	int	 kve_ref_count;			/* VM obj ref count. */
+	int	 kve_shadow_count;		/* VM obj shadow count. */
+	int	 _kve_pad0;			/* 64bit align next field */
+	int	 _kve_ispare[16];		/* Space for more stuff. */
+	/* Truncated before copyout in sysctl */
+	char	 kve_path[PATH_MAX];		/* Path to VM obj, if any. */
+};
+
 /*
  * The KERN_PROC_KSTACK sysctl allows a process to dump the kernel stacks of
  * another process as a series of entries.  Each stack is represented by a
@@ -343,12 +409,15 @@ struct kinfo_vmentry {
 #define	KKST_STATE_SWAPPED	1		/* Stack swapped out. */
 #define	KKST_STATE_RUNNING	2		/* Stack ephemeral. */
 
+#if defined(__amd64__) || defined(__i386__)
+#define	KINFO_KSTACK_SIZE	1096
+#endif
+
 struct kinfo_kstack {
 	lwpid_t	 kkst_tid;			/* ID of thread. */
 	int	 kkst_state;			/* Validity of stack. */
 	char	 kkst_trace[KKST_MAXLEN];	/* String representing stack. */
-	void	*_kkst_pspare[8];		/* Space for more stuff. */
-	int	 _kkst_ispare[8];		/* Space for more stuff. */
+	int	 _kkst_ispare[16];		/* Space for more stuff. */
 };
 
 #endif

Modified: head/usr.bin/procstat/Makefile
==============================================================================
--- head/usr.bin/procstat/Makefile	Tue Dec  2 06:34:08 2008	(r185547)
+++ head/usr.bin/procstat/Makefile	Tue Dec  2 06:50:26 2008	(r185548)
@@ -12,4 +12,8 @@ SRCS=	procstat.c		\
 	procstat_threads.c	\
 	procstat_vm.c
 
+LDADD+=	-lutil
+DPADD+=	${LIBUTIL}
+WARNS?=	4
+
 .include <bsd.prog.mk>

Modified: head/usr.bin/procstat/procstat_files.c
==============================================================================
--- head/usr.bin/procstat/procstat_files.c	Tue Dec  2 06:34:08 2008	(r185547)
+++ head/usr.bin/procstat/procstat_files.c	Tue Dec  2 06:50:26 2008	(r185548)
@@ -41,6 +41,7 @@
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
+#include <libutil.h>
 
 #include "procstat.h"
 
@@ -134,42 +135,18 @@ void
 procstat_files(pid_t pid, struct kinfo_proc *kipp)
 {
 	struct kinfo_file *freep, *kif;
-	int error, name[4];
-	unsigned int i;
+	int i, cnt;
 	const char *str;
-	size_t len;
 
 	if (!hflag)
 		printf("%5s %-16s %4s %1s %1s %-8s %3s %7s %-3s %-12s\n",
 		    "PID", "COMM", "FD", "T", "V", "FLAGS", "REF", "OFFSET",
 		    "PRO", "NAME");
 
-	name[0] = CTL_KERN;
-	name[1] = KERN_PROC;
-	name[2] = KERN_PROC_FILEDESC;
-	name[3] = pid;
-
-	error = sysctl(name, 4, NULL, &len, NULL, 0);
-	if (error < 0 && errno != ESRCH && errno != EPERM) {
-		warn("sysctl: kern.proc.filedesc: %d", pid);
-		return;
-	}
-	if (error < 0)
-		return;
-
-	freep = kif = malloc(len);
-	if (kif == NULL)
-		err(-1, "malloc");
-
-	if (sysctl(name, 4, kif, &len, NULL, 0) < 0) {
-		warn("sysctl: kern.proc.filedesc %d", pid);
-		free(freep);
-		return;
-	}
-
-	for (i = 0; i < len / sizeof(*kif); i++, kif++) {
-		if (kif->kf_structsize != sizeof(*kif))
-			errx(-1, "kinfo_file mismatch");
+	freep = kinfo_getfile(pid, &cnt);
+	for (i = 0; i < cnt; i++) {
+		kif = &freep[i];
+		
 		printf("%5d ", pid);
 		printf("%-16s ", kipp->ki_comm);
 		switch (kif->kf_fd) {

Modified: head/usr.bin/procstat/procstat_vm.c
==============================================================================
--- head/usr.bin/procstat/procstat_vm.c	Tue Dec  2 06:34:08 2008	(r185547)
+++ head/usr.bin/procstat/procstat_vm.c	Tue Dec  2 06:50:26 2008	(r185548)
@@ -34,6 +34,7 @@
 #include <errno.h>
 #include <stdio.h>
 #include <stdlib.h>
+#include <libutil.h>
 
 #include "procstat.h"
 
@@ -41,10 +42,9 @@ void
 procstat_vm(pid_t pid, struct kinfo_proc *kipp __unused)
 {
 	struct kinfo_vmentry *freep, *kve;
-	int error, name[4], ptrwidth;
-	unsigned int i;
+	int ptrwidth;
+	int i, cnt;
 	const char *str;
-	size_t len;
 
 	ptrwidth = 2*sizeof(void *) + 2;
 	if (!hflag)
@@ -52,38 +52,9 @@ procstat_vm(pid_t pid, struct kinfo_proc
 		    "PID", ptrwidth, "START", ptrwidth, "END", "PRT", "RES",
 		    "PRES", "REF", "SHD", "FL", "TP", "PATH");
 
-	name[0] = CTL_KERN;
-	name[1] = KERN_PROC;
-	name[2] = KERN_PROC_VMMAP;
-	name[3] = pid;
-
-	len = 0;
-	error = sysctl(name, 4, NULL, &len, NULL, 0);
-	if (error < 0 && errno != ESRCH && errno != EPERM) {
-		warn("sysctl: kern.proc.vmmap: %d", pid);
-		return;
-	}
-	if (error < 0)
-		return;
-
-	/*
-	 * Especially if running procstat -sv, we may need room for more
-	 * mappings when printing than were present when we queried, so pad
-	 * out the allocation a bit.
-	 */
-	len += sizeof(*kve) * 3;
-	freep = kve = malloc(len);
-	if (kve == NULL)
-		err(-1, "malloc");
-	if (sysctl(name, 4, kve, &len, NULL, 0) < 0) {
-		warn("sysctl: kern.proc.vmmap: %d", pid);
-		free(freep);
-		return;
-	}
-
-	for (i = 0; i < (len / sizeof(*kve)); i++, kve++) {
-		if (kve->kve_structsize != sizeof(*kve))
-			errx(-1, "kinfo_vmentry structure mismatch");
+	freep = kinfo_getvmmap(pid, &cnt);
+	for (i = 0; i < cnt; i++) {
+		kve = &freep[i];
 		printf("%5d ", pid);
 		printf("%*p ", ptrwidth, kve->kve_start);

*** DIFF OUTPUT TRUNCATED AT 1000 LINES ***


More information about the svn-src-head mailing list