Re: [PATCH 5/7] efi_loader: PARTITION_UUIDS should be optional

[Tom Rini <trini@konsulko.com>]

[-- Attachment #1: Type: text/plain, Size: 3432 bytes --]

On Thu, Apr 21, 2022 at 09:57:50AM +0900, AKASHI Takahiro wrote:
> On Wed, Apr 20, 2022 at 09:37:39AM +0200, Heinrich Schuchardt wrote:
> > On 4/19/22 03:01, AKASHI Takahiro wrote:
> > > In the current implementation, partition table support (either GPT or DOS)
> > > is not mandatory. So CONFIG_PARTITION_UUIDS should not be enabled
> > > (selected) unconditionally.
> > > 
> > > Fixes: commit 17f8cda505e3 ("efi_loader: set partition GUID in device path for SIG_TYPE_GUID")
> > > Signed-off-by: AKASHI Takahiro <takahiro.akashi@linaro.org>
> > > ---
> > >   lib/efi_loader/Kconfig           |  2 +-
> > >   lib/efi_loader/efi_device_path.c | 11 ++++++++++-
> > >   2 files changed, 11 insertions(+), 2 deletions(-)
> > > 
> > > diff --git a/lib/efi_loader/Kconfig b/lib/efi_loader/Kconfig
> > > index d50cd2563d3d..bc518d7a413b 100644
> > > --- a/lib/efi_loader/Kconfig
> > > +++ b/lib/efi_loader/Kconfig
> > > @@ -15,7 +15,7 @@ config EFI_LOADER
> > >   	depends on !EFI_APP
> > >   	default y if !ARM || SYS_CPU = armv7 || SYS_CPU = armv8
> > >   	select LIB_UUID
> > > -	select PARTITION_UUIDS
> > > +	imply PARTITION_UUIDS
> > >   	select HAVE_BLOCK_DEVICE
> > >   	select REGEX
> > >   	imply FAT
> > > diff --git a/lib/efi_loader/efi_device_path.c b/lib/efi_loader/efi_device_path.c
> > > index 0542aaae16c7..01fb53ae8c40 100644
> > > --- a/lib/efi_loader/efi_device_path.c
> > > +++ b/lib/efi_loader/efi_device_path.c
> > > @@ -864,11 +864,20 @@ static void *dp_part_node(void *buf, struct blk_desc *desc, int part)
> > >   			break;
> > >   		case SIG_TYPE_GUID:
> > 
> > This can only be reached for CONFIG_EFI_PARTITION=y.
> 
> I know, but the code won't be exercised anyway if !CONFIG_EFI_PARTITION.
> 
> > >   			hddp->signature_type = 2;
> > > +#if CONFIG_IS_ENABLED(PARTITION_UUIDS)
> > > +			/* info.uuid exists only with PARTITION_UUIDS */
> > >   			if (uuid_str_to_bin(info.uuid,
> > > -					    hddp->partition_signature, 1))
> > > +					    hddp->partition_signature, 1)) {
> > 
> > uuid_guid_get_bin() does not write any byte if info.uuid is not valid.
> 
> What matters here is that "uuid" in struct disk_partition is defined
> only if CONFIG_PARTITION_UUIDS.
> 
> > >   				log_warning(
> > >   					"Partition no. %d: invalid guid: %s\n",
> > >   					part, info.uuid);
> > 
> > > +				memset(hddp->partition_signature, 0,
> > > +				       sizeof(hddp->partition_signature));
> > 
> > dp_alloc() already has zeroed out the memory. Please, remove the
> > superfluous code.
> > 
> > > +			}
> > > +#else
> > > +			memset(hddp->partition_signature, 0,
> > > +			       sizeof(hddp->partition_signature));
> > 
> > ditto.
> > 
> > 
> > This does not conform to the UEFI specification. Why should we support it?
> 
> As I said in my previous comment at patch#6,
> CONFIG_EFI_PARTITION can be seen optional under the current implementation.
> 
> > 
> > Shouldn't we select CONFIG_PARTITION_UUIDS if CONFIG_EFI_LOADER=y and
> > CONFIG_EFI_PARTITION=y.
> 
> Unfortunately some platforms enable (in other words, "select") PARTITION_UUIDS
> unconditionally even without explicitly enabling EFI_LOADER or EFI_PARTITION.

I wouldn't assume that they're correct in doing so and it might have
been how some other problem was addressed at the time to fix incorrect /
not yet implemented in Kconfig at the time dependencies.

-- 
Tom

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 659 bytes --]