Re: Possible regression with file madvise(MADV_COLLAPSE)

[Avi Kivity <avi@scylladb.com>]

On Fri, 2024-10-11 at 15:29 -0700, Yang Shi wrote:
> On Wed, Oct 9, 2024 at 9:04 AM Avi Kivity <avi@scylladb.com> wrote:
> > 
> > On Linux 6.10.10 with CONFIG_READ_ONLY_THP_FOR_FS=y,
> > madvise(MADV_COLLAPSE) on  program text fails with EINVAL.
> > 
> > To reproduce, compile the reproducer with
> > 
> > clang -g -o text-hugepage  text-hugepage.c \
> >         -fuse-ld=lld \
> >         -Wl,-zcommon-page-size=2097152 -Wl,-zmax-page-size=2097152
> > \
> >         -Wl,-z,separate-loadable-segments
> > 
> > and run:
> 
> Didn't clang make the page cache dirty?
> 
> Having sync between clang and the execution made the problem go away
> for me.
> 

I see it even with sync (and msync just before the madvise calls).


Tracing shows this (last lines before syscall exit):

|          hpage_collapse_scan_file() {
|            __rcu_read_lock();
|            __rcu_read_unlock();
|          }


so, it's not clear what the root cause is.