[Bug dynamic-link/11787] Program with large TLS segments fail.

["carlos at redhat dot com" <sourceware-bugzilla at sourceware dot org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=11787

Carlos O'Donell <carlos at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |NEW

--- Comment #44 from Carlos O'Donell <carlos at redhat dot com> ---
(In reply to Steven Schlansker from comment #43)
> The JVM is having trouble too, see the following core-libs-dev discussion:
> 
> http://mail.openjdk.java.net/pipermail/core-libs-dev/2015-December/037557.
> html
> 
> If you link the JVM as a library and use thread local storage, it is very
> hard for JVM internal threads to ensure that their stack is usable.
> 
> From the Rust discussion, they link that the Go runtime had the exact same
> problem.
> 
> How many languages are going to have to work around this broken behavior
> before there is a glibc fix? :(

I have not been working on a fix for this because of other priorities. Sorry
about that. We need to see submissions like this to raise the priority of a
given issue beyond just that of chrome having problems in certain scenarios.

Seeing that Rust and Java are both having problems means we should do something
about this. It has been 5 years since we discovered the issue, but like all
things it's not trivial and we need to rewrite the NPTL stack layout code.

I'm moving this back to NEW for now since I'm not going to work on it, but I'll
see what we can do internally at Red Hat to prioritize this.

As always "Patches welcome", but I understand that the code in question that
needs to be changed is very very hairy, and requires a senior community
developer to make or review the changes.

Thanks again for raising the visibility.

-- 
You are receiving this mail because:
You are on the CC list for the bug.