[Bug libc/29115] vfork()-based posix_spawn() has more failure modes than fork()-based one

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

https://sourceware.org/bugzilla/show_bug.cgi?id=29115

--- Comment #11 from Carlos O'Donell <carlos at redhat dot com> ---
(In reply to Florian Weimer from comment #9)
> (In reply to Carlos O'Donell from comment #8)
> > (In reply to Florian Weimer from comment #6)
> > > CLONE_NEWTIME is as specified today fundamentally incompatible with real
> > > vfork and the vDSO. It just does not work. Entering the new namespace
> > > requires a new vDSO data mapping, and that conflicts with vfork using the
> > > same address space.
> > 
> > The kernel already has per-cpu data in the vDSO.
> 
> Uh, since when? I thought that Linux didn't do per-CPU page tables.

So, this is a stretch, but on x86 you use GDT to get the per-CPU data.

Is this not what we could call per-cpu data in a distinct address space?

> > The vDSO doesn't follow any concept of a single address space for the
> > process.
> > 
> > The vDSO is not a part of POSIX and so doesn't have to follow any vfork
> > semantic requirements.
> > 
> > What prevents the kernel from making a new vDSO data mapping?
> 
> It requires creating a new VM for the vfork process, while preserving
> existing shared VM semantics in other regards. That seems difficult?

I don't know until a kernel developer tells me this is difficult :-)

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