[PATCH] efi_loader: Install the ACPI table from the bloblist

[Patrick Rudolph]

Hi Simon,
On Thu, Mar 6, 2025 at 5:11 PM Simon Glass <sjg at chromium.org> wrote:
>
> Hi Heinrich,
>
> On Thu, 6 Mar 2025 at 08:16, Heinrich Schuchardt <xypron.glpk at gmx.de> wrote:
> >
> > On 06.03.25 15:31, Simon Glass wrote:
> > > When BLOBLIST_TABLES is used, the ACPI tables are not currently added to
> > > the list of EFI tables. While we don't want to create a new memory
> >
> > EFI tables are added by InstallConfigurationTable().
> >
> > It seems this patch is not about the list of EFI configuration tables
> > but about the EFI memory map.
> >
> > Could you, please, update the commit message accordingly.
>
> It currently does not install the table. See the 'if
> (IS_ENABLED(CONFIG_BLOBLIST_TABLES))' below.
>
> >
> > > region, we do want to tell EFI about the tables.
> > >
> > > Fix this by covering this case. At some point the non-bloblist code can
> > > likely be removed.
> > >
> > > Signed-off-by: Simon Glass <sjg at chromium.org>
> > > Fixes: 3da59ee9579 ("efi_loader: Avoid mapping the ACPI tables twice")
> > > ---
> > >
> > >   lib/efi_loader/efi_acpi.c | 44 ++++++++++++++++++++++++++-------------
> > >   1 file changed, 30 insertions(+), 14 deletions(-)
> > >
> > > diff --git a/lib/efi_loader/efi_acpi.c b/lib/efi_loader/efi_acpi.c
> > > index ff305a6b13e..b8b3a3670da 100644
> > > --- a/lib/efi_loader/efi_acpi.c
> > > +++ b/lib/efi_loader/efi_acpi.c
> > > @@ -5,6 +5,9 @@
> > >    *  Copyright (C) 2018, Bin Meng <bmeng.cn at gmail.com>
> > >    */
> > >
> > > +#define LOG_CATEGORY LOGC_ACPI
> > > +
> > > +#include <bloblist.h>
> > >   #include <efi_loader.h>
> > >   #include <log.h>
> > >   #include <mapmem.h>
> > > @@ -32,25 +35,38 @@ efi_status_t efi_acpi_register(void)
> > >        * add_u_boot_and_runtime(). At some point that function could create a
> > >        * more detailed map.
> > >        */
> > > -     if (IS_ENABLED(CONFIG_BLOBLIST_TABLES))
> > > -             return EFI_SUCCESS;
> > > -
> > > -     /* Mark space used for tables */
> > > -     start = ALIGN_DOWN(gd->arch.table_start, EFI_PAGE_MASK);
> > > -     end = ALIGN(gd->arch.table_end, EFI_PAGE_MASK);
> > > -     ret = efi_add_memory_map(start, end - start, EFI_ACPI_RECLAIM_MEMORY);
> > > -     if (ret != EFI_SUCCESS)
> > > -             return ret;
> > > -     if (gd->arch.table_start_high) {
> > > -             start = ALIGN_DOWN(gd->arch.table_start_high, EFI_PAGE_MASK);
> > > -             end = ALIGN(gd->arch.table_end_high, EFI_PAGE_MASK);
> > > +     if (IS_ENABLED(CONFIG_BLOBLIST_TABLES)) {
> > > +             int size;
> > > +             void *tab = bloblist_get_blob(BLOBLISTT_ACPI_TABLES, &size);
> >
> > Why don't you set gd->arch.table_* in the bloblist case?
> > Wouldn't this be an easier fix?
>
> Yes, I did consider that as a quick hack. But I'm going to remove
> those fields. They are a pain to maintain and bloblist is easier.
>
> >
> > Best regards
> >
> > Heinrich
> >
> > > +
> > > +             if (!tab)
> > > +                     return EFI_NOT_FOUND;
> > > +             addr = map_to_sysmem(tab);
> > > +
> > > +             ret = efi_add_memory_map(addr, size,
> > > +                                      EFI_ACPI_RECLAIM_MEMORY);
> > > +             if (ret != EFI_SUCCESS)
> > > +                     return ret;
> > > +     } else {
> > > +             /* Mark space used for tables */
> > > +             start = ALIGN_DOWN(gd->arch.table_start, EFI_PAGE_MASK);
> > > +             end = ALIGN(gd->arch.table_end, EFI_PAGE_MASK);
> > >               ret = efi_add_memory_map(start, end - start,
> > >                                        EFI_ACPI_RECLAIM_MEMORY);
> > >               if (ret != EFI_SUCCESS)
> > >                       return ret;
> > > -     }
> > > +             if (gd->arch.table_start_high) {
> > > +                     start = ALIGN_DOWN(gd->arch.table_start_high,
> > > +                                        EFI_PAGE_MASK);
> > > +                     end = ALIGN(gd->arch.table_end_high, EFI_PAGE_MASK);
> > > +                     ret = efi_add_memory_map(start, end - start,
> > > +                                              EFI_ACPI_RECLAIM_MEMORY);
> > > +                     if (ret != EFI_SUCCESS)
> > > +                             return ret;
> > > +             }
> > >
> > > -     addr = gd_acpi_start();
> > > +             addr = gd_acpi_start();
> > > +     }
> > >       log_debug("EFI using ACPI tables at %lx\n", addr);
> > >
> > >       /* And expose them to our EFI payload */
> >
>
> Regards,
> Simon

I can confirm that your patch fixes booting on qemu-system-aarch64
sbsa-ref as the ACPI tables are visible to the OS again.

Signed-off-by: Patrick Rudolph <patrick.rudolph at 9elements.com>

Regards,
Patrick