Thread Local Storage

[Doug Rabson]

On Monday 29 March 2004 22:26, Daniel Eischen wrote:
> On Mon, 29 Mar 2004, Doug Rabson wrote:
> > On Monday 29 March 2004 20:36, Daniel Eischen wrote:
> > > On Mon, 29 Mar 2004, Doug Rabson wrote:
> > > > I've been spending a bit of time recently familiarising myself
> > > > with this TLS stuff and trying out a few things. I've been
> > > > playing with rtld and I have a prototype patch which implements
> > > > enough TLS support to let a non-threaded program which uses
> > > > static TLS work. With a tiny bit more work I can have limited
> > > > support for dynamic TLS as well (not for dlopen'ed modules yet
> > > > though). Is there a p4 tree for this stuff yet? I'd like to
> > > > check in what I have sometime.
> > > >
> > > > I've also been looking at libpthread and I can see some
> > > > potential problems with it. Currently libpthread on i386 uses
> > > > %gs to point at a struct kcb which seems to be a per-kse
> > > > structure. This structure contains a pointer to a per-thread
> > > > struct tcb and this pointer is managed by the userland context
> > > > switch code. Other arches are similar, e.g. ia64 uses $tp to
> > > > point at struct kcb.
> > > >
> > > > The problem with TLS is that the i386 ABI needs %gs to point at
> > > > the TLS
> > >
> > > There are 2 different methods allowed for the i386 ABI.
> > > We want to use the other method in which there is an
> > > additional indirection.  The current i386 libpthread
> > > stuff adheres to this method.
> >
> > Surely the GNU TLS ABI is preferable? It generates much smaller
> > code and needs many fewer relocations.
>
> No, we don't want an LDT for every thread and don't want
> to force a syscall for a thread switch.

But the code it generates is at least twice the size for dynamic TLS. It 
seems that the GNU people have done a better job defining the TLS abi 
for i386.

You don't need a syscall at thread switch if you do something like:

	_thread_switch(...)
	{
		if (tcb doesn't have LDT entry) {
			if (!free LDT entries)
				steal LDT entry from non-running thread;
			allocate LDT entry and point it at TLS goop for tcb.
		}
		load_gs(tcb's LDT sel);
		...
	}

I just have this feeling that the GNU ABI is going to get far better 
testing and support in the future since thats what linux uses.