by Peter A. Ko » Thu, 06 Jan 1994 04:00:39
I have a socket that is set non-blocking with
fcntl(fd, F_SETFL, FNDELAY);
and would like to set it back to blocking. What do I do? Also, since this
is the "BSD" way what is now the portable manner of setting sockets
blocking/non-blocking?
--
Peter A. Korp
Assistant Scientist
Argonne National Laboratory
by John Agos » Thu, 06 Jan 1994 05:16:47
I thought this was a pretty good question, so I looked into it a little.
On HP-UX this is done with an ioctl():
FIOSNBIO If the int with the address arg is non-zero, the socket
is put into non-blocking mode. Otherwise, the socket
is put into blocking mode. Blocking mode is the
default. The FIONBIO request is equivalent to the
FIOSNBIO request, although using FIONBIO is not
recommended. See accept(2), connect(2), recv(2), and
send(2) for an explanation of how non-blocking mode is
used.
I looked at an HP-UX and a Solaris system to see if there were
commonalities in this area. The answer, I found a FIONBIO ioctl() on
both systems. As stated above, HP-UX recommends using FIOSNBIO instead
of FIONBIO. I could not find the man page information on the Solaris
for ioctl() non/blocking socket settings, but I would guess it is
probably FIONBIO.
At first I thought they would be the same since both are SYSV systems,
BUT . . . these a BSD sockets, so there is no standard. :-(
--
John Agosta
-------------------------------------------------------------------------
by Casper H.S. D » Thu, 06 Jan 1994 05:45:15
POSIX seems to suggest:Quote:>I have a socket that is set non-blocking with
> fcntl(fd, F_SETFL, FNDELAY);
>and would like to set it back to blocking. What do I do? Also, since this
>is the "BSD" way what is now the portable manner of setting sockets
>blocking/non-blocking?
fcntl(fd, F_SETFL, fcntl(fd, G_GETFL, 0) & ~O_NONBLOCK);
(Get the currect fd flags, unset the O_NONBLOCK flag and
use the resulting value as the new flags).
Casper
by Oliver Lauma » Thu, 06 Jan 1994 22:34:08
The portable manner probably is to write the code in such a way (usingQuote:> > I have a socket that is set non-blocking [..] and would like to set it
> > back to blocking. What do I do? Also, since [using FNDELAY] is the "BSD"
> > way what is now the portable manner [..]?
> fcntl(fd, F_SETFL, fcntl(fd, G_GETFL, 0) & ~O_NONBLOCK);
Note that not all UNIX versions that support sockets also have O_NONBLOCK;
4.3BSD, for instance, doesn't.
Also, there are subtle differences between O_NDELAY and O_NONBLOCK; on
some systems that support both flags (such as SunOS 5.x), using
O_NDELAY causes future read() calls to return zero if the call would
have blocked, while using O_NONBLOCK causes read() calls to fail with
EAGAIN or EWOULDBLOCK. You have to keep this in mind when writing a
fully portable program that uses non-blocking I/O.
by Casper H.S. D » Thu, 06 Jan 1994 23:14:04
>> fcntl(fd, F_SETFL, fcntl(fd, G_GETFL, 0) & ~O_NONBLOCK);
>The portable manner probably is to write the code in such a way (using
>conditional compilation) that either O_NDELAY *or* O_NONBLOCK can be
>used, depending on which of these flags is available on the platform.
4.3BSD is pretty old now, I felt stuff like this sort-of goes w/oQuote:>Note that not all UNIX versions that support sockets also have O_NONBLOCK;
>4.3BSD, for instance, doesn't.
#if <has POSIX>
<use POSIX way>
#elif <has BSD way>
<use BSD>
#elif <has SysV way>
<use SysV>
#endif
This is where the people of SysV goofed: letting read return 0Quote:>Also, there are subtle differences between O_NDELAY and O_NONBLOCK; on
>some systems that support both flags (such as SunOS 5.x), using
>O_NDELAY causes future read() calls to return zero if the call would
>have blocked, while using O_NONBLOCK causes read() calls to fail with
>EAGAIN or EWOULDBLOCK. You have to keep this in mind when writing a
>fully portable program that uses non-blocking I/O.
Casper
by Per Hedela » Fri, 07 Jan 1994 19:22:51
And of course, SunOS 5 has to give us those broken semantics for theQuote:>In that case, it O_NONBLOCK is the preferred solution, if it
>is defined. O_NONBLOCK has POSIX defined semantics. System V
>and BSD don't agree on what O_NDELAY means. You don't want SysV
>O_NDELAY: It is indistinguishable from EOF.
So, does SunOS 4.x[.y] fall in the POSIX clause in this particular case?Quote:>#if <has POSIX>
> <use POSIX way>
>#elif <has BSD way>
> <use BSD>
>#elif <has SysV way>
> <use SysV>
>#endif
--Per Hedeland
...uunet!erix.ericsson.se!per
by Casper H.S. D » Fri, 07 Jan 1994 21:02:31
And who had it first? BSD 4.? Sys III ? Sys Vr?
What is *really* broken in Solaris 2.x is this:
/usr/ucb/cc fcntl.c
./a.out
O_NDELAY: read returns 0
O_NONBLOCK: read returns -1
(fcntl.c is a small program that tries O_NDELAY and O_NONBLOCK
on stdin and reads from stdin. See also the ucb include files)
O_NONBLOCK is, I think, a POSIX invention, because O_NDELAY hadQuote:>>#if <has POSIX>
>> <use POSIX way>
>>#elif <has BSD way>
>> <use BSD>
>>#elif <has SysV way>
>> <use SysV>
>>#endif
>So, does SunOS 4.x[.y] fall in the POSIX clause in this particular case?
>fcntl(fd, F_SETFL, O_NONBLOCK) seems to work as expected on at least
>4.1.x, but O_NONBLOCK isn't documented as allowed for this purpose
>anywhere I can see (e.g. fcntl(2) or fcntl(5)).
What to choose when O_NONBLOCK isn't defined?
There is no way to choose correctly unless you know what enviroment
you're in. Even then, if you're out of luck it's SysV w/o O_NONBLOCK.
In that case you have to distinguish between reads returning 0 (EOF)
and reads returning 0 (Operation would block).
Casper
by W. Richard Steve » Fri, 07 Jan 1994 22:40:12
by Per Hedela » Fri, 07 Jan 1994 23:35:57
Unfortunately I don't have a 4.3BSD system running anymore, but I seem toQuote:>I must admit that I *really* like to know which idiot decided
>to implement O_NDELAY is this way that *totally* breaks the
>semantics of read(2).
>And who had it first? BSD 4.? Sys III ? Sys Vr?
Well, the problem with SunOS 4 is as I mentioned that O_NONBLOCK existsQuote:>It seems that the best choice therefor is to test for O_NONBLOCK
>and use that if it exists. It has well-defined behaviour.
F_SETFL Set descriptor status flags (see fcntl(5) for
definitions). The following flags are the
only ones whose values may change: O_APPEND,
O_SYNC, and O_NDELAY, and the FASYNC, FNDE-
LAY, and FNBIO flags defined in <fcntl.h>.
- of course none of those flags are defined in <fcntl.h> but in
<sys/fcntlcom.h>, #included by fcntl.h, and none of them are the same as
O_NONBLOCK, and fcntlcom.h also claims that FAPPEND and FSYNC (which are
the same as O_APPEND and O_SYNC), and FNONBIO (!) (which is the same as
O_NONBLOCK!) work for the F_SETFL fcntl. Hmm, I guess if I didn't lose
myself somewhere in that maze, and include files can be regarded as
documentation, that the answer is that fcntl(fd, F_SETFL, O_NONBLOCK)
*is* (sort of:-) documented to work...
--Per Hedeland
...uunet!erix.ericsson.se!per
by Brandon S. Allbe » Sat, 08 Jan 1994 02:07:43
Try again. BSD4.[12] called O_NDELAY FNDELAY.
++Brandon
--
"MSDOS didn't get as bad as it is overnight -- it took over ten years
Do not taunt Happy Fun Coder. (seen on the Net...)
by Per Hedela » Tue, 18 Jan 1994 23:12:32
Not sure how that particular piece of information contradicts anything I
said, however if you mean to imply that fcntl FNDELAY was the "standard"
way of setting non-blocking mode, you're probably right - at least that
was what was documented in 4.3BSD, whereas ioctl FIONBIO wasn't (I
checked my last 4.3 dump:-).
Anyway, ioctl FIONBIO certainly existed in 4.3BSD, and according to the
source did *exactly* the same thing as fcntl FNDELAY, and O_NDELAY
existed too, #defined to FNDELAY, but only documented for open(2). And
of course either of fcntl/ioctl caused read(2) to return -1 with errno
== EWOULDBLOCK, which was the question I tried to answer...
--Per Hedeland
...uunet!erix.ericsson.se!per
1. Non blocking socket blocks; says 'read would block' ?
Hi all,
I have three sockets in server which I am reading in a continuous loop,
so don't want to wait on any one. I have designated them as 'non blocking'
but when run the server exits on the very first read() and says 'would
block'. (Is it the same as 'EWOULDBLOCK' mentioned in R. Steven's book ?)
I don't want that I should get any error msg and exit, in fact the
whole purpose of nonblocking is that if something is available, read it
else proceed in the loop and come back to read next time.
The client (sender) is sending msgs on all three sockets. The sockets
are conn_oriented. I am using SunOS 4.1. The small code part is below:
int emer_s, env_s, sens_s; /* socket fds */
//----------- making sockets nonblocking ******
if ((res=fcntl(emer_s,F_SETFL,FNDELAY)) < 0)
{
perror("fcntl res = -1");
exit(1);
}
if ((res=fcntl(env_s,F_SETFL,FNDELAY)) < 0)
{
perror("fcntl res = -1");
exit(1);
}
if ((res=fcntl(sens_s,F_SETFL,FNDELAY)) < 0)
{
perror("fcntl res = -1");
exit(1);
}
for(; ;)
{
if ((cc=read(emer_s,(char*)&gen_struct,size)) < size)
{
perror("read error");
exit(1);
}
......... do something
if ((cc=read(env_s,(char*)&gen_struct,size)) < size)
{
perror("read error");
exit(1);
}
......... do something
if ((cc=read(sens_s,(char*)&gen_struct,size)) < size)
{
perror("read error");
exit(1);
}
......... do something
}
Any help is appreciated.
hashmi
--
-----
Atiqullah Hashmi
UTA (Univ. of Texas at Arlington)
2. 1.0.9 Slackware 2.0 lockup with 86C801 (S3) VGA Chipset
3. Blocking and Non-Blocking socket
5. Non-blocking socket reads block! (Bug?)
7. Difference between blocking and non-blocking socket ?
8. Python Imaging Library and Sketch SUCCESS!
9. What is difference of SYNC, ASYNC, BLOCKING, NON-BLOCKING sockets?
10. Non-blocking sockets on SunOS 4 are blocking
11. select() behavior for Blocking and non-blocking sockets.
12. How do I set a socket to non-blocking?
13. changing socket to blocking after non blocking connect