On Thu, Feb 02, 2017 at 07:39:06PM +0900, AKASHI Takahiro wrote:
> Mark,
>
> On Wed, Feb 01, 2017 at 06:25:06PM +0000, Mark Rutland wrote:
> > On Wed, Feb 01, 2017 at 06:00:08PM +0000, Mark Rutland wrote:
> > > On Wed, Feb 01, 2017 at 09:46:24PM +0900, AKASHI Takahiro wrote:
> > > > arch_kexec_protect_crashkres() and arch_kexec_unprotect_crashkres()
> > > > are meant to be called around kexec_load() in order to protect
> > > > the memory allocated for crash dump kernel once after it's loaded.
> > > >
> > > > The protection is implemented here by unmapping the region rather than
> > > > making it read-only.
> > > > To make the things work correctly, we also have to
> > > > - put the region in an isolated, page-level mapping initially, and
> > > > - move copying kexec's control_code_page to machine_kexec_prepare()
> > > >
> > > > Note that page-level mapping is also required to allow for shrinking
> > > > the size of memory, through /sys/kernel/kexec_crash_size, by any number
> > > > of multiple pages.
> > >
> > > Looking at kexec_crash_size_store(), I don't see where memory returned
> > > to the OS is mapped. AFAICT, if the region is protected when the user
> > > shrinks the region, the memory will not be mapped, yet handed over to
> > > the kernel for general allocation.
> > >
> > > Surely we need an arch-specific callback to handle that? e.g.
> > >
> > > arch_crash_release_region(unsigned long base, unsigned long size)
> > > {
> > > /*
> > > * Ensure the region is part of the linear map before we return
> > > * it to the OS. We won't unmap this again, so we can use block
> > > * mappings.
> > > */
> > > create_pgd_mapping(&init_mm, start, __phys_to_virt(start),
> > > size, PAGE_KERNEL, false);
> > > }
> > >
> > > ... which we'd call from crash_shrink_memory() before we freed the
> > > reserved pages.
> >
> > Another question is (how) does hyp map this region?
>
> I don't get your point here.
> Hyp mode does care only physical memory in intermediate address, doesn't it?
My concern was that hyp may map the region; and thus buggy code at hyp
can corrupt the region (and/or hyp may conflict w.r.t. attributes).
We mght have to ensure hyp doesn't map the crashkernel region, and to
case us pain, disallow freeing of any part of the region.
I'll dig into this.
Thanks,
Mark.
> If this is not a matter now, I will post v32 tomorrow :)
>
> -Takahiro AKASHI
>
>
> > Thanks,
> > Mark.
