Date: Mon, 31 Aug 1998 12:02:49 +0300 From: Alexander Litvin <archer@lucky.net> To: Stefan Eggers <seggers@semyam.dinoco.de> Cc: current@FreeBSD.ORG Subject: Re: IPFW showing extra lines Message-ID: <19980831120249.58455@carrier.kiev.ua> In-Reply-To: <199808302055.WAA02656@semyam.dinoco.de>
next in thread | previous in thread | raw e-mail | index | archive | help
In article <199808302055.WAA02656@semyam.dinoco.de> you wrote:
>> Specifically, ipfw -at l now shows a total of 1024 lines, whereas I
>> should only have about 15 firewall rules set up.
SE> The socket options handling code was renovated lately and totally
SE> restructured removing the need for mbuf's there.
SE> Taking a hard look at where it might fail I think it's a line in
SE> sooptcopyout in sys/kern/uipc_socket.c causing this. The length of
SE> the result is only copied into its rightful place if the buffer for
SE> the result is too small. Sound's weird, doesn't it?
SE> That can explain the problem you have with ipfw. It not only outputs
SE> the rules the kernel gave back but also a lot more reading junk from
SE> uninitialised memory as the length it gets back is the length of the
SE> buffer ipfw allocated and not the length of the result. In this case
SE> it seems to be freshly allocated memory filled with zero by the OS
SE> resulting in these zero rules.
SE> Thus I think the fix is to always set the result length which I expect
SE> the attached patch (TOTALLY UNTESTED!!!) to do. It's a little bit too
SE> late for me to test it today (about an hour till midnight) but maybe
SE> someone else can say if I'm on the right track or not.
Well, I'm not anywhere near expert, but manpage for getsockopt says:
The parameters optval and optlen are used to access option values for
setsockopt(). For getsockopt() they identify a buffer in which the value
for the requested option(s) are to be returned. For getsockopt(), optlen
is a value-result parameter, initially containing the size of the buffer
pointed to by optval, and modified on return to indicate the actual size
of the value returned.
But, if sopt->sopt_valsize (read: optlen in getsockopt(2)) is less
than the actual size of data which sooptcopyout() is asked to copy,
the code copies optlen bytes of data. It is correct, as optlen is
the size of the buffer.
But also in such situation, sooptcopyout() modifies optlen to be the
(larger) size of the data which sooptcopyout() was asked to copy.
Either code or manpage is wrong.
SE> Stefan.
---
Q: How many Zen masters does it take to screw in a light bulb?
A: None. The Universe spins the bulb, and the Zen master stays out
of the way.
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?19980831120249.58455>