Re: [PATCH v4 1/5] efi: Export boot-services code and data as debugfs-blobs

[Hans de Goede]

Hi,

On 26-04-18 23:35, Ard Biesheuvel wrote:
> On 26 April 2018 at 23:02, Hans de Goede <hdegoede@xxxxxxxxxx> wrote:
> > Hi,
> >
> >
> > On 26-04-18 18:51, Ard Biesheuvel wrote:
> > > On 26 April 2018 at 14:06, Hans de Goede <hdegoede@xxxxxxxxxx> wrote:
> > > > Sometimes it is useful to be able to dump the efi boot-services code and
> > > > data. This commit adds these as debugfs-blobs to /sys/kernel/debug/efi,
> > > > but only if efi=debug is passed on the kernel-commandline as this
> > > > requires
> > > > not freeing those memory-regions, which costs 20+ MB of RAM.
> > > >
> > > > Reviewed-by: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx>
> > > > Signed-off-by: Hans de Goede <hdegoede@xxxxxxxxxx>
> > > > ---
> > > > Changes in v4:
> > > > -Add new EFI_BOOT_SERVICES flag and use it to determine if the
> > > > boot-services
> > > >    memory segments are available (and thus if it makes sense to register
> > > > the
> > > >    debugfs bits for them)
> > > >
> > > > Changes in v2:
> > > > -Do not call pr_err on debugfs call failures
> > > > ---
> > > >    arch/x86/platform/efi/efi.c    |  1 +
> > > >    arch/x86/platform/efi/quirks.c |  4 +++
> > > >    drivers/firmware/efi/efi.c     | 53 ++++++++++++++++++++++++++++++++++
> > > >    include/linux/efi.h            |  1 +
> > > >    4 files changed, 59 insertions(+)
> > > >
> > > > diff --git a/arch/x86/platform/efi/efi.c b/arch/x86/platform/efi/efi.c
> > > > index 9061babfbc83..568b7ee3d323 100644
> > > > --- a/arch/x86/platform/efi/efi.c
> > > > +++ b/arch/x86/platform/efi/efi.c
> > > > @@ -208,6 +208,7 @@ int __init efi_memblock_x86_reserve_range(void)
> > > >                efi.memmap.desc_version);
> > > >
> > > >           memblock_reserve(pmap, efi.memmap.nr_map *
> > > > efi.memmap.desc_size);
> > > > +       set_bit(EFI_BOOT_SERVICES, &efi.flags);
> > >
> > > I think it would be better if the flag conveys whether boot services
> > > regions are being preserved, because they will always exist when
> > > EFI_BOOT is set.
> > > The name should then reflect that as well, e.g., EFI_PRESERVE_BS_REGIONS.
> >
> >
> > Ok, I will rename the flag to EFI_PRESERVE_BS_REGIONS for v5
> > (I'm going to wait a bit with sending out v5 to give others a change
> >   to comment on v4).
> >
> > > >           return 0;
> > > >    }
> > > > diff --git a/arch/x86/platform/efi/quirks.c
> > > > b/arch/x86/platform/efi/quirks.c
> > > > index 36c1f8b9f7e0..16bdb9e3b343 100644
> > > > --- a/arch/x86/platform/efi/quirks.c
> > > > +++ b/arch/x86/platform/efi/quirks.c
> > > > @@ -376,6 +376,10 @@ void __init efi_free_boot_services(void)
> > > >           int num_entries = 0;
> > > >           void *new, *new_md;
> > > >
> > > > +       /* Keep all regions for /sys/kernel/debug/efi */
> > > > +       if (efi_enabled(EFI_DBG))
> > > > +               return;
> > > > +
> > >
> > >
> > > Why is this only necessary when EFI_DBG is enabled? How are you
> > > ensuring that the firmware is still in memory when the subsequent
> > > patches start relying on that?
> >
> >
> > The 2nd patch in this series makes init/main.c call
> > efi_check_for_embedded_firmwares() before efi_free_boot_services(),
> > efi_check_for_embedded_firmwares() then walks the dmi_system_id-s
> > "registered" (its a static list) by drivers and if their is a dmi_match
> > searches for the firmware described by the dmi_system_id.driver_data
> > ptr. If a firmware gets found it gets memdup-ed, so that we do not
> > have to keep all of the boot-services code around.
>
> Right, thanks for clearing that up.
>
> So that means that preserving the boot regions is really only
> necessary if you want to inspect them via debugfs, and the firmware
> loader does not rely on that. I missed that part.
>
> That means the only reason we have the new flag is to ensure that the
> shared debugfs code only exposes the boot services regions if they
> were preserved to begin with by the arch code, right?

Mostly right, since we have the flag now anyways I'm also using
it as a condition to call efi_check_for_embedded_firmwares(),
efi_check_for_embedded_firmwares() needs to happen after mm_init() (*)
and on non x86 the boot-services code/data is long gone then so
there is nothing for efi_check_for_embedded_firmwares() to look at.

> If so, after the flag rename:
>
> Acked-by: Ard Biesheuvel <ard.biesheuvel@xxxxxxxxxx>

I assume that the "mostly right" is good enough and I'm going to
add your Acked-by for the next version. Let me know if you've
any objections against that.

Regards,

Hans


*) efi_check_for_embedded_firmwares() mremap()s the boot-services
memory segments to scan then and they are too big for early_mremap()