Re: [PATCH] x86/split_lock: Restore warn mode (and add a new one) to avoid userspace regression

[Joshua Ashton]

On 9/28/22 17:54, Luck, Tony wrote:
> > So let's revamp the idea of having another option/mode for the split
> > lock detector, which is hereby called "seq" (based on the original
> > "sequential" naming in [0]). Also introduces a Kconfig option to give
> > the option of Linux vendors have a choice what mode should be their
> > default. While at it, fix/improve the documentation about bus locking.
> >
> > [1] https://github.com/doitsujin/dxvk/issues/2938
>
> Why not just use the workaround suggested in that bug report:
>
>     "so manual switching from default setting to split_lock_detect=off helps as workaround here"
>
> If you add this extra mode, I'm going to argue that the kernel default
> should be "seq" rather than "warn". So these game players will need
> to add a split_lock_detect=off (or warn) option.
>
> Has a bug report been filed against the God Of War game? Probably worth fixing,
> the performance penalty for split lock is only going to get worse as numbers of
> cores keeps increasing.

It's not just about God of War specifically.
There are many old titles that will never, ever, get updated to fix this 
problem.
These titles worked perfectly fine and were performant before.

I completely understand that split-locks are bad, and developers should 
fix their code, but applications paying penalties like this really 
doesn't make sense for the Linux desktop.

If server-oriented distributions want to set their default to `seq`, 
then that's completely fine, and I would completely encourage them doing 
that -- but I really don't think the Linux kernel should be essentially 
breaking compatibility with these applications by default and throwing 
desktop Linux usecases under the bus.

- Joshie 🐸✨

> -Tony