Date: Sun, 28 Oct 2001 13:18:37 -0500 (EST) From: Robert Watson <rwatson@FreeBSD.org> To: sobomax@FreeBSD.org Cc: cvs-committers@FreeBSD.org, cvs-all@FreeBSD.org Subject: Re: cvs commit: ports/devel/ORBit Makefile ports/devel/ORBit/files patch-src::IIOP::giop-msg-buffer.c Message-ID: <Pine.NEB.3.96L.1011028122718.15623C-100000@fledge.watson.org> In-Reply-To: <E15xtOb-0002Me-00@smtp2.port.ru>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sun, 28 Oct 2001, Maxim Sobolev wrote:
> > My local Linux box has:
> >
> > ...
> > linux/uio.h: * UIO_MAXIOV shall be at least 16 1003.1g (5.4.1.1)
> > ...
> > #define UIO_MAXIOV 1024
> > #if 0
> > #define UIO_MAXIOV 16 /* Maximum iovec's in one operation
> > 16 matches BSD */
> > /* Beg pardon: BSD has 1024 --ANK */
> > #endif
>
> I doubt that Linux is really enforces this limit, because if it does,
> then the ORBit would be broken there as well. BTW, could you please do
> me a favor and run the following program on your Linux box and send me
> its output? It tests and reports real value of the limit.
<snip>
Here's the output:
phoenix:/tmp/tmp> gcc -Wall -g -o o o.c
o.c: In function `main':
o.c:10: warning: implicit declaration of function `exit'
o.c:13: warning: implicit declaration of function `close'
phoenix:/tmp/tmp> ./o
16385
Interestingly, from the Linux source on that box:
linux-2.4/fs/read_write.c:
/*
* First get the "struct iovec" from user memory and
* verify all the pointers
*/
ret = 0;
if (!count)
goto out_nofree;
ret = -EINVAL;
if (count > UIO_MAXIOV)
goto out_nofree;
if (!file->f_op)
goto out_nofree;
if (count > UIO_FASTIOV) {
ret = -ENOMEM;
iov = kmalloc(count*sizeof(struct iovec), GFP_KERNEL);
if (!iov)
goto out_nofree;
}
So it really looks like, for a file readv or writev, it should reject high
count values. Likewise the socket code:
linux-2.4/net/socket.c:
/* do not move before msg_sys is valid */
err = -EINVAL;
if (msg_sys.msg_iovlen > UIO_MAXIOV)
goto out_put;
/* Check whether to allocate the iovec area*/
err = -ENOMEM;
iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
if (msg_sys.msg_iovlen > UIO_FASTIOV) {
iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
if (!iov)
goto out_put;
}
Pointing it at a file instead of a device doesn't seem to make a
difference.
Not sure what to think :-).
Robert N M Watson FreeBSD Core Team, TrustedBSD Project
robert@fledge.watson.org NAI Labs, Safeport Network Services
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe cvs-all" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.NEB.3.96L.1011028122718.15623C-100000>