Date: Sun, 23 Dec 2001 09:00:46 +0100 From: Bernd Walter <ticso@cicely9.cicely.de> To: Matthew Jacob <mjacob@feral.com> Cc: Wilko Bulte <wkb@freebie.xs4all.nl>, freebsd-alpha@FreeBSD.ORG Subject: Re: FreeBSD 4.5 alpha buyglet Message-ID: <20011223080046.GB61241@cicely9.cicely.de> In-Reply-To: <Pine.BSF.4.21.0112222233290.2338-100000@beppo> References: <20011223032201.GA61241@cicely9.cicely.de> <Pine.BSF.4.21.0112222233290.2338-100000@beppo>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sat, Dec 22, 2001 at 10:37:55PM -0800, Matthew Jacob wrote:
>
>
> On Sun, 23 Dec 2001, Bernd Walter wrote:
>
> > On Sat, Dec 22, 2001 at 05:12:32PM -0800, Matthew Jacob wrote:
> > >
> > > >Yes, see Message-ID: <20011221175252.A25617@freebie.xs4all.nl>
> > > >on -alpha.
> > >
> > > Ooops- so right you are. Anyway, I looked at this just a bit- seems a compiler
> > > error to me- it shouldn't be generating ldq's for those tags because they
> > > couldn't possibly be at 8 byte offsets given the structures involved. Oh well.
> >
> > The two variables it fails on are defined as beeing long but only 32 bit
> > aligned. I'm more about beleaving the sysctl tranfer misaligns them. But I
> > have currently no system where I see this problem so I can't be shure. The
> > sysctl receive buffer is malloc'ed so it should be 64 bit aligned.
>
> Hmm? I used gdb to find the misaligned reference to be, e.g.:
>
> ***> if (ifm->ifm_data.ifi_metric)
> printf(" metric %ld", ifm->ifm_data.ifi_metric);
> ***> if (ifm->ifm_data.ifi_mtu)
> printf(" mtu %ld", ifm->ifm_data.ifi_mtu);
> putchar('\n');
>
> struct if_msghdr {
> u_short ifm_msglen; /* to skip over non-understood messages */
> u_char ifm_version; /* future binary compatability */
> u_char ifm_type; /* message type */
> int ifm_addrs; /* like rtm_addrs */
> int ifm_flags; /* value of if_flags */
> u_short ifm_index; /* index for associated ifp */
> offset: 14
> struct if_data ifm_data;/* statistics and other data about if */
> };
>
>
> ...
>
>
> struct if_data {
> /* generic interface information */
> u_char ifi_type; /* ethernet, tokenring, etc */
> u_char ifi_physical; /* e.g., AUI, Thinnet, 10base-T, etc
> */
> u_char ifi_addrlen; /* media address length */
> u_char ifi_hdrlen; /* media header length */
> u_char ifi_recvquota; /* polling quota for receive intrs */
> u_char ifi_xmitquota; /* polling quota for xmit intrs */
>
> offset: 6
>
> u_long ifi_mtu; /* maximum transmission unit */
^^^^^^^^^^^^^^^^
> u_long ifi_metric; /* routing metric (external only) */
^^^^^^^^^^^^^^^^^^^
> u_long ifi_baudrate; /* linespeed */
>
>
> So, if gdb was right and dereferencing ifi_mtu or ifi_metric is generating
> a ldq, then the compiler seems to be broken.
What else would you use for loading a 64 bit value?
If it starts with the second or later interface I would asume it's
because several structs are concatenated in the sysctl output and read
as one.
If one of them is not n*8 bytes long you get an alignment offset.
But as I don't have more debug output than posted on the list I can
only guess.
--
B.Walter COSMO-Project http://www.cosmo-project.de
ticso@cicely.de Usergroup info@cosmo-project.de
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-alpha" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20011223080046.GB61241>