ACK: FADT table checksum corrupted when FACS table is faked

[Colin Ian King]

On 19/11/15 14:28, Jiri Vohanka wrote:
> Hello,
> 
> I discovered a problem in src/lib/src/fwts_acpi_tables.c that causes
> a false FAIL in the 'checksum' test.
> 
> If a FACS ACPI table is not found, a fake one is created and a fake
> physical
> address is assigned to it. This fake physical address is written to
> FIRMWARE_CTRL and X_FIRMWARE_CTRL fields of FADT table:
> 
>                facs_fake_addr = fwts_fake_physical_addr(size);
>                fadt->firmware_control = (uint32_t)facs_fake_addr;
>                if (fadt->header.length >= 140)
>                        fadt->x_firmware_ctrl = (uint64_t)facs_fake_addr;
> 
> (By the way, the ACPI 6.0 specification states that only one of
> FIRMWARE_CTRL,
> X_FIRMWARE_CTRL could be nonzero.)
> The problem is that the checksum of the FADT table is not updated to
> reflect
> the change in 'firmware_control' and 'x_firmware_ctrl' fields.
> 
> I created a patch that changes the update of physical address of FACS in
> FADT.
> It tries to use the physical address from 'x_firmware_ctrl' and
> 'firmware_control' fields of FADT table. If both fields are zero then
> a fake physical address is assigned to FACS, written to FADT table and the
> checksum of FADT is updated.
> 
> Unpatched test:
> No FACS found, fwts has faked one instead.
> ...
> FAILED [MEDIUM] ACPITableChecksum: Test 1, Table FACP has incorrect
> checksum,
> expected 0xe8, got 0xcf.
> 
> Patched test:
> No FACS found, fwts has faked one instead.
> ...
> PASSED: Test 1, Table FACP has correct checksum 0xcf
> 
> Regards,
> Jiri Vohanka
> 
> 
I've verified this works against the fwts regression tests and also by
dumping my ACPI tables, and removing the FACP, and running

fwts --dumpfile=acpidump.log checksum -

this patch fixes the issue, so I'm happy to ACK it. I

BTW, Alex, I don't think it needs to be re-sent. I can apply it using
git-am into my fwts git repo.

Acked-by: Colin Ian King <colin.king at canonical.com>