--=-=-=


Pavel Cahyna <pavel.cahyna@st.cuni.cz> writes:

> Please use "bt/l" in ddb next time you repeat this bug. I saw that
> sometimes (maybe always?) one frame is missed by gdb, but shows in ddb
> when the kernel panics with uvm_fault (which is usually a symptom of a
> NULL-pointer dereference). See for example PR kern/28669:
[....]
> Clearly the fr_coalesce function (where the panic really happens) is
> missing from the gdb backtrace.
>
> Bye	Pavel
>
> P.S. I should really write a document like "best debugging practices" as
> they are highly non-obvious.

I think you can find the offending function by looking at the trapframe and
examine where in the function it fell over.

(gdb) x/i ipaddr

I've never understod why gdb can't print source-code lines for disas but I
have to use obdump -Ssl --prefix-addr for that.

Love


--=-=-=
Content-Type: application/pgp-signature

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.5 (NetBSD)
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=4+R3
-----END PGP SIGNATURE-----
--=-=-=--