futex(0x1ffffff81300000) on risc-v -> mm panic

[Nam Cao]

On Wed, Jun 18, 2025 at 12:58:44PM -0700, Palmer Dabbelt wrote:
> On Wed, 18 Jun 2025 06:25:59 PDT (-0700), namcao at linutronix.de wrote:
> > On Wed, Jun 18, 2025 at 07:10:28AM -0400, rtm at csail.mit.edu wrote:
> > > This program on risc-v:
> > > 
> > > main(){
> > >   futex((void*) 0x1ffffff81300000, FUTEX_WAIT, 0, 0, 0, 0);
> > > }
> > > 
> > > results in:
> > > 
> > > BUG: Bad page state in process a.out  pfn:81500
> > 
> > I cannot reproduce this issue. Can you please share your .config?
> 
> It's also not reproducing for me -- or even compiling for that matter, but I
> stuck one that compiles over at
> <https://github.com/palmer-dabbelt/palmer-tests/blob/main/rtm-futex.c> (I
> figured this was a good excuse to start piling up some tests).
> 
> > Also, your kernel seems to have some commits on top. Are you sure none of
> > those is the reason?
> 
> This is on my fixes branch right now, with defconfig + CONFIG_DEBUG_VM=y.
> So maybe it's one of these patches, or moybe it's some other configuration
> thing?

I can reproduce this with QEMU and Debian trixie. defconfig + DEBUG_VM=y is
enough. I have to "search for the address", as OP showed in the other
email, because 0x1ffffff81300000 doesn't trigger any issue for me.

I can also reproduce on 6.14 and 6.15, so not a recent regression.

I will look closer tomorrow.

Best regards,
Nam