NetBSD Problem Report #28459

From gson@araneus.fi  Mon Nov 29 13:57:42 2004
Return-Path: <gson@araneus.fi>
Received: from gusto.araneus.fi (gusto.araneus.fi [204.152.189.154])
	by narn.netbsd.org (Postfix) with ESMTP id A44FD251F89
	for <gnats-bugs@gnats.NetBSD.org>; Mon, 29 Nov 2004 13:57:41 +0000 (UTC)
Message-Id: <200411291357.iATDvY4Y027041@guava.araneus.fi>
Date: Mon, 29 Nov 2004 15:57:34 +0200 (EET)
From: gson@gson.org (Andreas Gustafsson)
Reply-To: gson@gson.org (Andreas Gustafsson)
To: gnats-bugs@netbsd.org
Subject: autri driver hangs for 30 seconds on close
X-Send-Pr-Version: 3.95

>Number:         28459
>Category:       kern
>Synopsis:       autri driver hangs for 30 seconds on close
>Confidential:   no
>Severity:       non-critical
>Priority:       medium
>Responsible:    nat
>State:          closed
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Mon Nov 29 14:00:00 +0000 2004
>Closed-Date:    Tue Jul 11 19:53:09 +0000 2017
>Last-Modified:  Tue Jul 11 19:53:09 +0000 2017
>Originator:     Andreas Gustafsson
>Release:        NetBSD-current as of Nov 28, 2004
>Organization:
>Environment:
System: NetBSD guam.araneus.fi 2.0B NetBSD 2.0B (GUAM) #0: Sun Nov 28 11:58:04 EET 2004 root@guava.araneus.fi:/usr/src/sys/arch/i386/compile/GUAM i386
Architecture: i386
Machine: i386
>Description:

When using a full-duplex audio application with the autri audio
driver, a close() of the audio device will hang for 30 seconds in
audio_drain() before finally returning.

This was seen on a Toshiba Libretto L2, which has an ALi M5451
sound chip.

>How-To-Repeat:

On a system with a sound chip that uses the autri driver, extract the
enclosed shell archive, type "make fullduplex" and then "sh test.sh".
It will play about a tenth of a second of silence while simultaneously
recording to /dev/null, and then attempt to exit.  Notice how the
process hangs for 30 seconds after printing the message "copied 5120
bytes".

# This is a shell archive.  Save it in a file, remove anything before
# this line, and then unpack it by entering "sh file".  Note, it may
# create directories; files and directories will be owned by you and
# have default permissions.
#
# This archive contains:
#
#	test.sh
#	fullduplex.cc
#
echo x - test.sh
sed 's/^X//' >test.sh << 'END-of-test.sh'
Xaudioctl -n -f /dev/sound -w \
X  record.rate=48000 record.channels=1 record.precision=16 \
X  record.encoding=slinear_le \
X  play.rate=48000 play.channels=1 play.precision=16 \
X  play.encoding=slinear_le 
X
Xdd if=/dev/zero bs=512 count=10 | ./fullduplex /dev/sound >/dev/null
END-of-test.sh
echo x - fullduplex.cc
sed 's/^X//' >fullduplex.cc << 'END-of-fullduplex.cc'
X// $Id: fdtunnel.cc,v 1.2 2002/08/18 02:41:28 gson Exp $
X//
X// Full duplex audio I/O for use in pipelines.
X//
X// (C) Copyright 1999-2004 Araneus Information Systems Oy.
X//
X// Simultaneously plays audio data from stdin and records onto stdout.
X//
X
X#include <stdio.h>
X
X#include <stdio.h>
X
X#include <sys/types.h>
X#include <sys/audioio.h>
X#include <sys/ioctl.h>
X#include <sys/mman.h>
X#include <sys/uio.h>
X#include <sys/param.h>
X#include <sys/file.h>
X
X#include <stdlib.h>
X#include <syslog.h>
X#include <string.h>
X#include <errno.h>
X#include <unistd.h>
X
X#define BSIZ 256
X
Xchar *progname;
X
X// Log an error message and exit
X
Xvoid error(char *s) {
X    fprintf(stderr, "%s: %s: %s\n", progname, s, strerror(errno));
X    exit(1);
X}
X
X// This structure represents a single (unidirectional) copy 
X// operation in progress, like a single instance of the "cat"
X// command.
X  
Xtypedef struct {
X    char buf[BSIZ];
X    int r;	/* read file descriptor */
X    int w;	/* write file descriptor */
X    char *p;	/* current position in buffer */
X    int n;	/* amount of unwritten data in buffer */
X    int eof;	/* 0 if open; 1 if eof received; 2 if eof sent */
X    int copied;
X} cat;
X
X// Do N simultaneous "cat" operations; return when they have all
X// seen EOF.
X
Xvoid copy(int catc, cat *catv)
X{
X    fd_set readfds, writefds, exceptfds;
X    int i, n;
X    
X    for (i = 0; i < catc; i++) {
X	cat *c = &catv[i];
X	c->n = 0;
X	c->eof = 0;
X	c->copied = 0;
X    }
X    
X    for (;;) {
X	int max;
X
X	FD_ZERO(&readfds); FD_ZERO(&writefds); FD_ZERO(&exceptfds);
X	max = -1;
X	
X	for (i = 0; i < catc; i++) {
X	    cat *c = &catv[i];
X	    if (c->n || c->eof == 1) {
X		FD_SET(c->w, &writefds);
X		if (c->w > max)
X		    max = c->w;
X	    } else if (!c->eof) {
X		FD_SET(c->r, &readfds);
X		if (c->r > max)
X		    max = c->r;
X	    }
X	}
X
X	if (catv[0].eof && catv[1].copied >= catv[0].copied)
X	    break;
X	    
X	n = select(max+1, &readfds, &writefds, &exceptfds,
X		   (struct timeval *) 0);
X	if (n == -1)
X	    error("select");
X
X	for (i = 0; i < catc; i++) {
X	    cat *c = &catv[i];
X	    if (FD_ISSET(c->r, &readfds)) {
X		n = read(c->r, c->buf, BSIZ);
X		if (n == 0) {
X		    fprintf(stderr, "eof reading fd %d\n", c->r);
X		    c->eof = 1;
X		} else if (n == -1) {
X		    fprintf(stderr, "error reading cat %d\n", i);
X		    error("read");
X		} else {
X		    c->n = n;
X		    c->p = c->buf;
X		}
X	    }
X	    if (FD_ISSET(c->w, &writefds)) {
X		if (c->n) {
X		    n = write(c->w, c->p, c->n);
X		    if (n == -1)
X			error("write");
X		    c->n -= n;
X		    c->p += n;
X		    c->copied += n;
X		}
X	    }	    
X	}
X    }
X}
X    
Xint main(int argc, char **argv) {
X    progname = argv[0];
X
X    int fd = open(argv[1], O_RDWR);
X    if (fd < 0) {
X	fprintf(stderr, "%s: open: %s\n",
X		argv[1], strerror(errno));
X	exit(1);
X    }
X
X    // Check audio device properties
X    int props;
X    int r = ioctl(fd, AUDIO_GETPROPS, &props);
X    if ((props & AUDIO_PROP_FULLDUPLEX) != AUDIO_PROP_FULLDUPLEX) {
X	fprintf(stderr, "Audio device does not support "
X		"full duplex.  Sorry.\n");
X	exit(1);
X    }
X    
X    // Set full duplex mode
X    int yes = 1;
X    r = ioctl(fd, AUDIO_SETFD, &yes);
X    if (r < 0) {
X	fprintf(stderr, "ioctl AUDIO_SETFD: %s\n",
X		strerror(errno));
X	exit(1);
X    }
X    
X    // Set an appropriate block size
X    audio_info_t info;	
X    AUDIO_INITINFO(&info);
X    info.blocksize = 4800;
X    // Due to a bug in the audio driver, the block size may not actually get
X    // set unless we set info.mode = AUMODE_RECORD | AUMODE_PLAY redundantly.
X    info.mode = AUMODE_RECORD | AUMODE_PLAY;
X    r = ioctl(fd, AUDIO_SETINFO, &info);
X    if (r < 0) {
X	fprintf(stderr, "ioctl AUDIO_SETINFO blocksize/mode: %s\n",
X		strerror(errno));
X	exit(1);
X    }
X
X    AUDIO_INITINFO(&info);
X    info.record.pause = 0;
X    r = ioctl(fd, AUDIO_SETINFO, &info);
X    if (r < 0) {
X	fprintf(stderr, "ioctl AUDIO_SETINFO unpause: %s\n",
X		strerror(errno));
X	exit(1);
X    }
X    
X    cat cats[2];
X    cats[0].r = 0;
X    cats[0].w = fd;
X    cats[1].r = fd;
X    cats[1].w = 1;
X    copy(2, cats);
X    fprintf(stderr, "copied %d bytes\n", cats[1].copied);
X    close(fd);
X    return 0;
X}
END-of-fullduplex.cc
exit

>Fix:

Not known.

>Release-Note:

>Audit-Trail:

Responsible-Changed-From-To: kern-bug-people->nat
Responsible-Changed-By: nat@NetBSD.org
Responsible-Changed-When: Wed, 21 Jun 2017 23:26:53 +0000
Responsible-Changed-Why:
Take.


State-Changed-From-To: open->feedback
State-Changed-By: nat@NetBSD.org
State-Changed-When: Wed, 21 Jun 2017 23:26:53 +0000
State-Changed-Why:
This should be fixed.  Does a kernel from -current exhibit the same problem?


From: Andreas Gustafsson <gson@gson.org>
To: nat@NetBSD.org, nonaka@NetBSD.org
Cc: gnats-bugs@NetBSD.org
Subject: Re: kern/28459 (autri driver hangs for 30 seconds on close)
Date: Thu, 22 Jun 2017 10:41:36 +0300

 nat@NetBSD.org wrote:
 > Synopsis: autri driver hangs for 30 seconds on close
 [...]
 > This should be fixed.  Does a kernel from -current exhibit the same problem?

 I'm currently unable to test becase the machine fails
 to boot due to the problem reported in PR 52266.
 Nonaka, any progress on that one?
 -- 
 Andreas Gustafsson, gson@gson.org

From: Andreas Gustafsson <gson@gson.org>
To: nat@NetBSD.org
Cc: gnats-bugs@NetBSD.org
Subject: Re: kern/28459 (autri driver hangs for 30 seconds on close)
Date: Tue, 11 Jul 2017 22:45:17 +0300

 Now that I have a work-around for PR 52266, I tried booting a -current
 kernel on the machine in case, which had not been turned on for 8
 years and still had NetBSD 4.99.30 installed.

 Sadly, this resulted in "panic: init died (signal 0, exit 1)".  As
 much as I would have liked to confirm that PR 28459 is now fixed,
 12 years after filing it, I'm afraid it will have to be closed as
 "submitter no longer able to test".
 -- 
 Andreas Gustafsson, gson@gson.org

State-Changed-From-To: feedback->closed
State-Changed-By: gson@NetBSD.org
State-Changed-When: Tue, 11 Jul 2017 19:53:09 +0000
State-Changed-Why:
Submitter no longer able to test.


>Unformatted:

Home	PR Database Search