Re: Squid AIX notes for the FAQ

On Fri, 3 Jan 2003 19:27, Robert Collins wrote:
> From Gerard Eviston:
>

A few qualifications follow for the sake of correctness.

> <cgonzalez> hi mate
> <cgonzalez> a few notes for when you get back ...
> <cgonzalez> my seemingly bizarre coredumps on aix are solved
> <cgonzalez> for a >256MB RSS on aix the xcoff header needs to have
> o_maxdata set
> <cgonzalez> oops not rss as such, just stack+heap+ublock+kernelstack

From the docs: per-process information including user data, user stack, kernel
stack and user block.

> <cgonzalez> with ibm's ld this is done by linking with
> -bmaxdata:{insert_nonzero_number_here}

Without the { } braces.

> <cgonzalez> with gld too bad so sad it seems but echo '\0200\0\0\0'|dd
> of=squid bs=4 count=1 seek=19 conv=notrunc doest the trick

This should be an explicit /usr/bin/echo '\0200\0\0\0' ... not the shell echo
for obvious reasons.

> <cgonzalez> thats straight out of ibm's docco and was how i got it to
> work
> <cgonzalez> the thing that threw me was that the core dumps were not all
> 256MB, they varied quite a bit
> <cgonzalez> stack and other fluctuations i guess
> <cgonzalez> side effects: possibly a significantly larger backing store
> (this wasnt clear to me from the docs)
> <cgonzalez> the nozero maxdata enables large program support
> <cgonzalez> bmaxdata's valuse is unused unless another flag is set that
> enables vlery large program support
>

Perhaps someone more skilled in AIX would care to consider/explain whether the
DSA flag and (hence) Very Large Program support are (un)necessary.

Or even better, I'll try to find a public URL for the relevant documentation.

> I'm not sure where this should go in the FAQ, probably in OS
> weirdness's.
>
> Rob

Regards
Gerard
Received on Fri Jan 03 2003 - 04:53:46 MST