Re: [x86 pmap changes] CVS commit: src/sys/arch

[Chavdar Ivanov <ci4ic4%gmail.com@localhost>]

BTW, on a just-built 9.99.36, qemu-nvmm now works. However, it seems
at least three times slower than before.

Chavdar

On Sat, 11 Jan 2020 at 20:59, Andrew Doran <ad%netbsd.org@localhost> wrote:
>
> On Fri, Jan 10, 2020 at 06:56:29PM +0000, Andrew Doran wrote:
>
> > On Thu, Jan 09, 2020 at 05:46:13PM +0100, Maxime Villard wrote:
> >
> > > This FPU issue should be fixed in the latest nvmm_x86_vmx.c, we still have
> > > STTS/CLTS (not needed but for debugging) as part of context switches, and
> > > when overhauling the FPU code I overlooked that VMX needs special CR0_TS
> > > care that SVM doesn't need.
> > >
> > > Note that dropping STTS/CLTS would probably increase cswitch performance,
> > > because updating cr0 is costly.
> > >
> > > Having said that, I am still hitting a KASSERT related to pmap:
> > >
> > >     kernel diagnostic assertion "ptp->wire_count == 1" failed file
> > >             ".../x86/x86/pmap.c", line 1969
> > >     pmap_freepages
> > >     pmap_ept_free_ptp
> > >     pmap_ept_remove
> > >     pmap_remove
> > >     uvm_unmap_remove
> > >     uvmspace_free
> > >     nvmm_ioctl
> > >     sys_ioctl
> >
> > Taking a look.  I think it's probably related to the changes I made to
> > further defer freeing PTPs (ugly but necessary for yamt-pagecache).  I think
> > I see a more elegant way to handle that.
>
> Yes that's the problem. I have a fix and will commit soon.
>
> Andrew



-- 
----