Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Subject: Re: [PATCH] efi/libstub: Allow EFI_NOT_FOUND on LOAD_FILE2_PROTOCOL
 calls for initrd
To:     Ilias Apalodimas <ilias.apalodimas@linaro.org>,
        ard.biesheuvel@arm.com
Cc:     Ard Biesheuvel <ardb@kernel.org>,
        Arvind Sankar <nivedita@alum.mit.edu>,
        Ingo Molnar <mingo@kernel.org>, linux-efi@vger.kernel.org,
        linux-kernel@vger.kernel.org
References: <20201214170122.4569-1-ilias.apalodimas@linaro.org>
From:   Heinrich Schuchardt <xypron.glpk@gmx.de>
Message-ID: <aa66cc54-f987-cdcd-05a7-4b63aa8b422d@gmx.de>
Date:   Mon, 14 Dec 2020 18:39:21 +0100
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:78.0) Gecko/20100101
 Thunderbird/78.5.1
MIME-Version: 1.0
In-Reply-To: <20201214170122.4569-1-ilias.apalodimas@linaro.org>
Content-Type: text/plain; charset=utf-8
Content-Language: en-US
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On 14.12.20 18:01, Ilias Apalodimas wrote:
> At the moment the EFI stub tries to load an initrd from the
> cmdline provided option only if the LoadFile2 protocol does not exist
> on the initrd device path.
>
> This might prove problematic for EFI installers that need their own
> version of initrd to start the installation process and the firmware

Did you hit a real world case?

> installs the protocol but doesn't have a file to back it up (yet).
> Although some firmware implementations return EFI_NOT_FOUND, we
> currently return EFI_LOAD_ERROR in efi_load_initrd_dev_path() which
> stops the cmdline provided initrd to load.
>
> So let's change the behavior slightly here and explicitly respect the
> firmware in case it returns EFI_NOT_FOUND. This way we can load the
> cmdline provided initrd.
>
> Signed-off-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
> ---
>  drivers/firmware/efi/libstub/efi-stub-helper.c | 10 +++++++++-
>  1 file changed, 9 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/firmware/efi/libstub/efi-stub-helper.c b/drivers/fi=
rmware/efi/libstub/efi-stub-helper.c
> index aa8da0a49829..391aae2f0cde 100644
> --- a/drivers/firmware/efi/libstub/efi-stub-helper.c
> +++ b/drivers/firmware/efi/libstub/efi-stub-helper.c
> @@ -560,6 +560,7 @@ static const struct {
>   * * %EFI_SUCCESS if the initrd was loaded successfully, in which
>   *   case @load_addr and @load_size are assigned accordingly
>   * * %EFI_NOT_FOUND if no LoadFile2 protocol exists on the initrd devic=
e path
> + *   or if the firmware provides LoadFile2 but can't find a file to loa=
d
>   * * %EFI_INVALID_PARAMETER if load_addr =3D=3D NULL or load_size =3D=
=3D NULL
>   * * %EFI_OUT_OF_RESOURCES if memory allocation failed
>   * * %EFI_LOAD_ERROR in all other cases
> @@ -599,7 +600,14 @@ efi_status_t efi_load_initrd_dev_path(unsigned long=
 *load_addr,
>  				(void *)initrd_addr);
>  	if (status !=3D EFI_SUCCESS) {
>  		efi_free(initrd_size, initrd_addr);
> -		return EFI_LOAD_ERROR;
> +		/*
> +		 * Some firmware implementations might install the EFI

In U-Boot the filename is set a compile time. As the path may relate to
a removable medium, it would not make sense to check the existence of
the file when installing the protocol.

> +		 * protocol without checking the file is present and return
> +		 * EFI_NOT_FOUND when trying to load the file.
> +		 * If that's the case, allow the cmdline defined initrd to
> +		 * load.

U-Boot's implementation could also return EFI_NO_MEDIA if
CONFIG_EFI_INITRD_FILESPEC relates to a non-existent partition.

Why should we fall back to the command line in this case?

The whole idea of this protocol is to disallow the specification of an
initrd via the command line.

Best regards

Heinrich

> +		 */
> +		return status =3D=3D EFI_NOT_FOUND ? status : EFI_LOAD_ERROR;
>  	}
>
>  	*load_addr =3D initrd_addr;
>