From owner-freebsd-current  Fri Apr 14 20:31:48 1995
Return-Path: current-owner
Received: (from majordom@localhost)
          by freefall.cdrom.com (8.6.10/8.6.6) id UAA24816
          for current-outgoing; Fri, 14 Apr 1995 20:31:48 -0700
Received: from sequent.kiae.su (sequent.kiae.su [144.206.136.6])
          by freefall.cdrom.com (8.6.10/8.6.6) with SMTP id UAA24795
          for <current@FreeBSD.org>; Fri, 14 Apr 1995 20:31:33 -0700
Received: by sequent.kiae.su id AA29712
  (5.65.kiae-2 ); Sat, 15 Apr 1995 07:24:18 +0400
Received: by sequent.KIAE.su (UUMAIL/2.0); Sat, 15 Apr 95 07:24:18 +0400
Received: (from ache@localhost) by astral.msk.su (8.6.8/8.6.6) id DAA02679; Sat, 15 Apr 1995 03:30:17 +0400
To: current@FreeBSD.org, "John S. Dyson" <toor@jsdinc.root.com>
References: <199504141715.RAA05096@jsdinc.root.com>
In-Reply-To: <199504141715.RAA05096@jsdinc.root.com>;
    from "John S. Dyson" at Fri, 14 Apr 1995 17:15:53 GMT
Message-Id: <ZI6MmZlWWA@astral.msk.su>
Organization: Olahm Ha-Yetzirah
Date: Sat, 15 Apr 1995 03:30:14 +0400
X-Mailer: Mail/@ [v2.32 FreeBSD]
From: "Andrey A. Chernov, Black Mage" <ache@astral.msk.su>
X-Class: Fast
Subject: Re: Info on VM/VFS changes since 4.4Lite
Lines: 150
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Content-Length: 3977
Sender: current-owner@FreeBSD.org
Precedence: bulk

In message <199504141715.RAA05096@jsdinc.root.com> John S. Dyson
    writes:

>5)	The VM system and buffer cache has been merged.
>	Now mmap is fully coherent with the read/write system calls.  This
>	is an initial implementation, and the VOP_GETPAGE and VOP_PUTPAGE	
>	will be compatibly added soon (Probably V2.2).  For example, a
>	write to a file immediately causes the data to change immediately
>	in the address space of a process that might have the file mapped.

John, please, can you fix munmap() & msync() to modify modification times
as supposed and manpages promise? Now they don't touch modification
times at all. It cause impossibility to use mmap() inside INND, it
suppose modification time changing.

BTW: my local chief shows me that BSDI 2.0 can handle much more articles
quickly with INND then FreeBSD and call it "benchmark". The only one
reason exists: BSDI have INND compiled with mmap() and FreeBSD without
mmap() (it is configurable) due to modification times bug.

Here small test program to demonstrate this bug:

#include <sys/types.h>
#include <sys/stat.h>
#include <sys/file.h>
#include <sys/mman.h>
#include <stdio.h>

static char             ICDactpath[] = "testfile";
static char             *ICDactpointer;
static int              ICDactfd;
static int              ICDactsize;

#if	defined(MAP_FILE)
#define MAP__ARG	(MAP_FILE | MAP_SHARED)
#else
#define MAP__ARG	(MAP_SHARED)
#endif	/* defined(MAP_FILE) */

char *
ICDread()
{
    struct stat		Sb;
    int i;

    if ((ICDactfd = open(ICDactpath, O_WRONLY|O_TRUNC|O_CREAT, 0666)) < 0) {
	perror("open");
	fprintf(stderr, "read: cant open %s\n", ICDactpath);
	exit(1);
    }
    for (i = 0; i < 7; i++)
	write(ICDactfd, "1234567890", 10);
    if (close(ICDactfd) < 0) {
	perror("close");
	fprintf(stderr, "Close: cant close %s\n", ICDactpath);
	exit(1);
    }

    if ((ICDactfd = open(ICDactpath, O_RDWR)) < 0) {
	perror("open");
	fprintf(stderr, "read: cant open %s\n", ICDactpath);
	exit(1);
    }

    if (fstat(ICDactfd, &Sb) < 0) {
	perror("fstat");
	fprintf(stderr, "read: cant fstat %d %s\n", ICDactfd, ICDactpath);
	exit(1);
    }
    ICDactsize = Sb.st_size;
    ICDactpointer = mmap((caddr_t)0, ICDactsize, PROT_READ|PROT_WRITE,
			MAP__ARG, ICDactfd, (off_t)0);
    if (ICDactpointer == (char *)-1) {
	perror("mmap");
	fprintf(stderr, "read: cant mmap %d %s\n", ICDactfd, ICDactpath);
	exit(1);
    }

    return ICDactpointer;
}

ICDclose()
{
    if (ICDactpointer) {
	if (msync(ICDactpointer, ICDactsize, 0) < 0) {
	    perror("msync");
	    fprintf(stderr, "Close: cant msync %s in closeactive()\n",  ICDactpath);
	}

	if (munmap(ICDactpointer, ICDactsize) < 0) {
	    perror("munmap");
	    fprintf(stderr, "Close: cant munmap\n", ICDactpath);
	}
	ICDactpointer = NULL;
	if (close(ICDactfd) < 0) {
	   perror("close");
	   fprintf(stderr, "Close: cant close %s\n", ICDactpath);
	   exit(1);
	}
    }
}

ICDwrite()
{
    /* No-op. */
		    /* ICDactsize */
    if (msync(ICDactpointer, 0, 0) < 0) {
	perror("msync");
	fprintf(stderr, "Write: cant msync %s\n", ICDactpath);
    }
}

spy() {
	system("cat testfile; echo; ls -l testfile");
}

main()
{
	char *file_p;
	int to_sleep = 60;

	file_p = ICDread();

	fprintf(stderr, "Original file and time\n");
	spy();
	*(file_p + 20) = '\n';
#if 0
	fprintf(stderr, "After 1st write to mapped region\n");
	spy();
#endif
	sleep(to_sleep);
	*(file_p + 30) = '\n';
#if 0
	fprintf(stderr, "After 2nd write to mapped region and sleep()\n");
	spy();
#endif
	ICDwrite();
	fprintf(stderr, "After msync(..., 0)\n");
	spy();
	sleep(to_sleep);
	ICDclose();
	fprintf(stderr, "After sleep(), msync(..., size) and close()\n");
	spy();
	unlink(ICDactpath);
}
-- 
Andrey A. Chernov        : And I rest so composedly,  /Now, in my bed,
ache@astral.msk.su       : That any beholder  /Might fancy me dead -
FidoNet: 2:5020/230.3    : Might start at beholding me,  /Thinking me dead.
RELCOM Team,FreeBSD Team :         E.A.Poe         From "For Annie" 1849