Review request for 5049299

[Martin Buchholz]

[+ roland, jakub]

Hi Andrew,

I never tried using the clone() system call directly.
I was scared off by all the assembly code in glibc surrounding the system
call.
I went in the alternative direction of using vfork, which I have been
avoiding for the
past decade, but seems to work.

I continue to believe my glibc bug report is valid and contains an
excellent test case for glibc.  Perhaps other past maintainers of glibc's
process code could be persuaded to try to persuade Uli?

This is mostly for the good of glibc, not for java.  Even if glibc fixes
their bug,
we won't be able to make use of it for years.

OK, I'll try adding Roland and Jakub to this conversation.
Hope they're still at redhat.
http://sources.redhat.com/bugzilla/show_bug.cgi?id=10311

http://cr.openjdk.java.net/~martin/vfork-exec/

Martin

On Tue, Jun 23, 2009 at 07:33, Andrew Haley <aph at redhat.com> wrote:

> Martin Buchholz wrote:
> > clone-exec update:
> >
> > I submitted the changes for this, but jtreg tests failed on 32-bit Linux
> > (I had only tested on 64-bit Linux)
> >
> > We disabled (but did not roll back) the use of clone to allow the
> > TL integration to proceed.
> >
> > (As I promised elsewhere...)
> > I just filed a bug against upstream glibc demonstrating the problem
> > with clone(CLONE_VM).  You can see the small C program
> > in my bug report below.
> > Probably any discussion related just to the glibc bug
> > can occur on the public glibc bugzilla at
> > http://sources.redhat.com/bugzilla/show_bug.cgi?id=10311
> >
> > glibc maintainer Uli Drepper has already responded
> > saying
> >
> > "If you use clone() you're on your own."
> >
> > so if we are going to fix it, we'll have to do it ourselves.
> > Help from threading/kernel hackers appreciated.
>
> I can debug this.
>
> Please try first syscall(SYS_clone ...) to bypass the libc gubbins.
> That might be all you need.  If that doesn't help I'll have a look.
>
> Isn't there some point at which you have to say to a Linux user "Your
> system is simply misconfigured.  Fix the overcommit parameter and this
> problem will go away"  ?
>
> Andrew.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.openjdk.java.net/pipermail/core-libs-dev/attachments/20090627/45d0a848/attachment.html>