This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: Regarding https://sourceware.org/bugzilla/show_bug.cgi?id=22850
- From: Carlos O'Donell <carlos at redhat dot com>
- To: kamlesh kumar <kamleshbhalui at gmail dot com>, libc-alpha at sourceware dot org
- Date: Mon, 21 Oct 2019 08:59:20 -0400
- Subject: Re: Regarding https://sourceware.org/bugzilla/show_bug.cgi?id=22850
- References: <CABKRkghmq-1zqPaMMSTDpx+swzTerH5gvGLhWwRBw2SNnieMLQ@mail.gmail.com>
On 10/21/19 5:34 AM, kamlesh kumar wrote:
> I'm looking at the subjected issue and to start with we are thinking
> to move out the tcb from the thread stack
That sounds great.
> but we are not sure about like where we can room this tcb structure
> (since we are new to libc ) and we would like to know your suggestion
> on this and did anyone tried this implementation and would like to
> know cons and pros of this implementation and etc .
As new contributors please review the "Contribution Checklist":
https://sourceware.org/glibc/wiki/Contribution%20checklist
Bug 22850 is a difficult bug that covers multiple subsystems and will
touch all architectures. It requires considerable skill to get all the
details right.
So far there is no attempted implementation in glibc where the TCB
and associated data is split from the thread stack.
A related bug is:
"Program with large TLS segments fail."
https://sourceware.org/bugzilla/show_bug.cgi?id=11787
Removing the struct pthread, the TCB, and TLS data from the stack
would make the stack extent computation much easier.
Someone needs to design a solution and present it to the community
with their evalution of the pros/cons of such a design.
--
Cheers,
Carlos.