Date: Fri, 5 May 2000 22:24:52 +0200 From: Gerhard Sittig <Gerhard.Sittig@gmx.net> To: freebsd-stable@FreeBSD.ORG Subject: Re: Debugging Kernel/System Crashes, can anyone help?? Message-ID: <20000505222452.T11903@speedy.gsinet> In-Reply-To: <4.3.2.20000503221040.02051d40@207.227.119.2>; from jeff-ml@mountin.net on Thu, May 04, 2000 at 05:00:35PM -0500 References: <20000503185936.E72341@ethereal.net> <20000504095941.B18453@freebie.lemis.com> <200005040124.VAA55655@account.abs.net> <20000504111136.B22025@freebie.lemis.com> <20000503185936.E72341@ethereal.net> <20000504114013.E22025@freebie.lemis.com> <4.3.2.20000503221040.02051d40@207.227.119.2>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, May 04, 2000 at 17:00 -0500, Jeffrey J. Mountin wrote:
>
> [ ... installing nonstripped kernel when building -g ... ]
>
> Running the non-stripped kernel uses more memory and isn't
> there also a performance issue. Also what is gained by running
> it, as long as you have the complimentary debug kernel in
> /var/crash (or wherever) for the stripped boot kernel.
Are you sure about these facts? I'm tempted to speak against
both points: The symbols are in the executable file only and
won't make it into physical memory unless they're referenced
(which should happen only when linkers or debuggers need them).
And the performance penalty comes from extra checks (inserted by
the compiler or #ifdef'ed) or different stack frames for
compiling with "-g", _not_ from not stripping symbols off.
So "nothing's gained" except a little more consistency or - to
put it in other words - there's a loss of potential irritation :)
And in addition as was stated before, keeping a stripped and an
unstripped version on your disk is more expensive in terms of
space than just holding one non-stripped version.
And what I missed in the previous discussion: How "normal" is it
to have more than three kernel images around so that their
symbols fill up a (too small sized?) root filesystem? The only
environments that come to mind are kernel developer machines or
"toys" for curious people compiling kernels "for fun". Both of
them should know what they do. I feel everyday users would just
have one GENERIC (stripped) kernel around from install time and
one or two generations of selfmade kernels -- once the new one
has proven to work the former might disappear, but I never see a
need in keeping more of them around.
> Pretty sure past discussions say nothing is gained, but some is
> lost by running kernel.debug rather than the stripped kernel.
> Maybe I missed something.
I wouldn't mix "debug" and "stripped" in this way -- they're
completely orthogonal to me. Unless *I* missed a point :)
> I do agree with building, but not installing, a debug kernel.
I thought the discussion went like this:
- When everything's fine, run a non-debug kernel.
- When you have problems you feel like digging into, build a
debug kernel.
- To gain real knowledge from the debugging session, you need
symbols.
Please correct me if I'm wrong.
But this leads me to "What's the use in compiling with "-g" and
stripping symbols off the installed copy when you have to keep
the unstripped version somewhere anyway?". Do unstripped
executables really eat more resources when run? Does it matter
in today's memory management environments (demand paging) if the
symbol's segment is "loaded" at all)?
virtually yours 82D1 9B9C 01DC 4FB4 D7B4 61BE 3F49 4F77 72DE DA76
Gerhard Sittig true | mail -s "get gpg key" Gerhard.Sittig@gmx.net
--
If you don't understand or are scared by any of the above
ask your parents or an adult to help you.
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-stable" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20000505222452.T11903>