[Gcl-devel] Re: sgc issue

[Camm Maguire]

Greetings!

I know of one condition which can cause this -- running under an
SELinux box (i.e. certain redhat boxes come with this by default)
configured in strict mode.  By default, the setup instructs the kernel
to deny any application the right to make allocated memory executable,
which is indispensable to GCL.  Turning off strict mode (I think
'kickstart' can do this) is a workaround.  There is also likely a less
severe but more involved workaround (not yet tested) posted earlier on
gcl-devel -- rebuilding the SELinux policy with the memory protect
access explicitly granted (cannot remember the flag name at present,
but I can dig it up if you want.)  Most drastically, one can simply
disable SELinux, at least temporarily to test.

Please keep me posted if this does not work.  Alas, there is no
workaround at GCL configure time possible for a system which refuses
executable memory access, as we load and execute code on the fly all
the time.

Take care,

Matt Kaufmann <address@hidden> writes:

> Hi, Camm --
> 
> I've gotten a report from a user of ACL2 built on GCL that he is getting a
> "couldn't protect" errorl.  I suggested that he evaluate (si::sgc-on nil), and
> he tells me that this causes the error itself.  Do you have any suggestions on
> how to avoid this error entirely, especially by doing something at startup?
> 
> Thanks --
> -- Matt
>    X-Cam-SpamDetails: Not scanned
>    X-Cam-AntiVirus: No virus found
>    X-Cam-ScannerInfo: http://www.cam.ac.uk/cs/email/scanner/
>    Date: Thu, 05 Jan 2006 10:35:57 +0000
>    From: James Reynolds <address@hidden>
>    User-Agent: Thunderbird 1.5 (X11/20051025)
>    CC: address@hidden
>    Content-Type: text/plain; charset=ISO-8859-1; format=flowed
>    X-SpamAssassin-Status: No, hits=-2.6 required=5.0
>    X-UTCS-Spam-Status: No, hits=-182 required=180
> 
>    Thanks Matt, that definitely seems to be the cause of the error, as 
>    trying (si::sgc-on nil) causes the error itself! Is there any way of 
>    turning this off outside of acl2?
> 
>    James
> 
>    Matt Kaufmann wrote:
>    > Thanks, Mike.  I don't recall ever seeing a "couldn't protect" error, so 
> a log
>    > might help me understand.  I don't think that's from ACL2, as the 
> following
>    > produces no output:
>    >
>    >   fgrep -i "Couldn't protect" *.lisp
>    >
>    > Perhaps it's a problem with the underlying Lisp.  Actually, I just did a 
> google
>    > search and I have a suspicion that the underlying Lisp is GCL and that 
> it's a
>    > problem with the stratified garbage collector (which marks pages in a 
> way that
>    > maybe is not supported by the OS on the machine James is using).  As an
>    > experiment, he can try this:
>    >
>    >   :q ; exit the ACL2 loop
>    >   (si::sgc-on nil) ; turn off SGC
>    >   (lp) ; re-enter the ACL2 loop
>    >
>    > SGC is nice for performance, but I can imagine that in practice it won't 
> hurt
>    > much to turn it off.
>    >
>    > -- Matt
>    >    cc: address@hidden
>    >    Date: Wed, 04 Jan 2006 19:12:29 +0000
>    >    From: Mike Gordon <address@hidden>
>    >    X-SpamAssassin-Status: No, hits=-2.6 required=5.0
>    >    X-UTCS-Spam-Status: No, hits=-172 required=180
>    >
>    >
>    >    Oops .. here's James' original message ...
>    >
>    >    ------- Forwarded Message
>    >
>    >    Date: Wed, 04 Jan 2006 18:10:26 +0000
>    >    From: James Reynolds <address@hidden>
>    >    To: Mike Gordon <address@hidden>
>    >    Subject: ACL2 "Couldn't protect" error
>    >
>    >    Hi Mike,
>    >
>    >    I seem to be having some problems with the ACL2 evaluator, its 
> producing 
>    >    a "Couldn't protect" error whilst defining functions, but only 
>    >    intermittently. It quite often fails on the following functions:
>    >
>    >    (defun fracwidth (x) (cdr x))
>    >
>    >    and
>    >
>    >    (defun exponent (x) (cadr x)).
>    >
>    >    Unfortunately I can't seem to find the "Couldn't protect" error 
> anywhere 
>    >    in the online documentation, and trying various forms of guards on 
> the 
>    >    functions doesn't seem to work. I have a hunch its an install issue, 
>    >    have you seen anything like this on your machine?
>    >
>    >    Thanks,
>    >    James
>    >
>    >    ------- End of Forwarded Message
>    >
>    >   
> 
> 
> 

-- 
Camm Maguire                                            address@hidden
==========================================================================
"The earth is but one country, and mankind its citizens."  --  Baha'u'llah