[PATCH 2/6] efi_loader: Add headers for EDK2 StandAloneMM communication

[Heinrich Schuchardt]

On 5/6/20 9:12 PM, Ilias Apalodimas wrote:
> From: Sughosh Ganu <sughosh.ganu at linaro.org>
>
> In Arm devices OP-TEE has the ability to run StandAloneMM (from EDK2)
> in a separate partition and handle UEFI variables.
> A following patch introduces this functionality.
>
> Add the headers needed for OP-TEE <--> StandAloneMM communication
>
> Signed-off-by: Sughosh Ganu <sughosh.ganu at linaro.org>
> Signed-off-by: Ilias Apalodimas <ilias.apalodimas at linaro.org>
> ---
>  include/mm_communication.h | 28 ++++++++++++++
>  include/mm_variable.h      | 78 ++++++++++++++++++++++++++++++++++++++
>  2 files changed, 106 insertions(+)
>  create mode 100644 include/mm_communication.h
>  create mode 100644 include/mm_variable.h
>
> diff --git a/include/mm_communication.h b/include/mm_communication.h
> new file mode 100644
> index 000000000000..fb4c91103400
> --- /dev/null
> +++ b/include/mm_communication.h
> @@ -0,0 +1,28 @@
> +/* SPDX-License-Identifier: GPL-2.0+ */

BSD-2-Clause-Patent is compatible with GPL-2. So relicensing as GPL
should be ok.

> +/*
> + *  Headers for EFI variable service via StandAloneMM, EDK2 application running
> + *  in OP-TEE
> + *
> + *  Copyright (C) 2020 Linaro Ltd. <sughosh.ganu at linaro.org>
> + *  Copyright (C) 2020 Linaro Ltd. <ilias.apalodimas at linaro.org>

EDK2, MdePkg/Include/Protocol/MmCommunication.h has:
Copyright (c) 2017, Intel Corporation. All rights reserved.

Why did you replace their copyright by yours?
Who is the original copyright holder of the MM module?

> + */
> +
> +#if !defined _MM_COMMUNICATION_H_

I would use #ifndef here.

> +#define _MM_COMMUNICATION_H_
> +
> +/* defined in EDK2 MmCommunication.h */

Please, add a description of the structure, cf.
https://www.kernel.org/doc/html/latest/doc-guide/kernel-doc.html#structure-union-and-enumeration-documentation

> +struct mm_communicate_header {
> +	efi_guid_t header_guid;
> +	size_t     message_len;
> +	u8         data[1];

In C11 you can use data[].

> +};
> +
> +#define MM_COMMUNICATE_HEADER_SIZE \
> +	(offsetof(struct mm_communicate_header, data))

SMM_COMMUNICATE_HEADER_SIZE?
Why not simply use data[] and sizeof(struct mm_communicate_header)
instead of defining this constant.

> +
> +#define MM_RET_SUCCESS         0

Why don't you use the same names as in EDK2, e.g.
ARM_SMC_MM_RET_SUCCESS?

Please, add ARM_SMC_MM_RET_NOT_SUPPORTED.

> +#define MM_RET_INVALID_PARAMS -2
> +#define MM_RET_DENIED         -3
> +#define MM_RET_NO_MEMORY      -4
> +
> +#endif /* _MM_COMMUNICATION_H_*/
> diff --git a/include/mm_variable.h b/include/mm_variable.h
> new file mode 100644
> index 000000000000..f56c52597629
> --- /dev/null
> +++ b/include/mm_variable.h
> @@ -0,0 +1,78 @@
> +/* SPDX-License-Identifier: GPL-2.0+ */
> +/*
> + *  Headers for EFI variable service via StandAloneMM, EDK2 application running
> + *  in OP-TEE
> + *
> + *  Copyright (C) 2020 Linaro Ltd. <sughosh.ganu at linaro.org>
> + *  Copyright (C) 2020 Linaro Ltd. <ilias.apalodimas at linaro.org>

If you copied this from SmmVariableCommon.h shouldn't you mention:
Copyright (c) 2011 - 2019, Intel Corporation. All rights reserved.

> + */
> +
> +#if !defined _MM_VARIABLE_H_

#ifndef would be shorter.

> +#define _MM_VARIABLE_H_
> +
> +#include <part_efi.h>
> +
> +/* defined in EDK2 SmmVariableCommon.h */

Please, add a description of the structure, cf.
https://www.kernel.org/doc/html/latest/doc-guide/kernel-doc.html#structure-union-and-enumeration-documentation

> +struct mm_variable_communicate_header {
> +	efi_uintn_t  function;
> +	efi_status_t ret_status;
> +	u8           data[1];
> +};
> +
> +#define MM_VARIABLE_COMMUNICATE_SIZE \
> +	(offsetof(struct mm_variable_communicate_header, data))

Why not the original name SMM_VARIABLE_COMMUNICATE_HEADER_SIZE?

> +
> +#define MM_VARIABLE_FUNCTION_GET_VARIABLE			1

MdeModulePkg/Include/Guid/SmmVariableCommon.h uses
SMM_VARIABLE_FUNCTION_GET_VARIABLE. Why invent a new name?

> +
> +#define MM_VARIABLE_FUNCTION_GET_NEXT_VARIABLE_NAME		2
> +
> +#define MM_VARIABLE_FUNCTION_SET_VARIABLE			3
> +
> +#define MM_VARIABLE_FUNCTION_QUERY_VARIABLE_INFO		4
> +
> +#define MM_VARIABLE_FUNCTION_READY_TO_BOOT			5
> +
> +#define MM_VARIABLE_FUNCTION_EXIT_BOOT_SERVICE			6
> +
> +#define MM_VARIABLE_FUNCTION_GET_STATISTICS			7
> +
> +#define MM_VARIABLE_FUNCTION_LOCK_VARIABLE			8
> +
> +#define MM_VARIABLE_FUNCTION_VAR_CHECK_VARIABLE_PROPERTY_SET	9
> +
> +#define MM_VARIABLE_FUNCTION_VAR_CHECK_VARIABLE_PROPERTY_GET	10
> +
> +#define MM_VARIABLE_FUNCTION_GET_PAYLOAD_SIZE			11

Values 12 - 14 are missing. I think we should provide all values.

> +

Missing structure description.

> +struct mm_variable_access {

smm_variable_access?

> +	efi_guid_t  guid;
> +	efi_uintn_t data_size;
> +	efi_uintn_t name_size;
> +	u32         attr;
> +	u16         name[1];

This name[1] was needed in old compilers. It is not needed anymore in
C11.  For a variable length structure component, please, use name[].

> +};
> +
> +#define MM_VARIABLE_ACCESS_HEADER_SIZE \
> +	(offsetof(struct mm_variable_access, name))

If you have name[] as component, you can use sizeof(struct
smm_variable_access) instead of this define.

> +

Structure description missing.

> +struct mm_variable_payload_size {
> +	efi_uintn_t size;
> +};

In EDK2 PayloadSize is simply UINTN.

Isn't this structure superfluous? Can't we directly pass a type UINTN
variable instead and get rid of a level of indirection?

> +

Structure description missing.

> +struct mm_variable_getnext {
> +	efi_guid_t  guid;
> +	efi_uintn_t name_size;
> +	u16         name[1];

name[]?

> +};
> +
> +#define MM_VARIABLE_GET_NEXT_HEADER_SIZE \
> +	(offsetof(struct mm_variable_getnext, name))
> +

Structure description missing.

You could think about merging the two include files into one.

Best regards

Heinrich

> +struct mm_variable_query_info {
> +	u64 max_variable_storage;
> +	u64 remaining_variable_storage;
> +	u64 max_variable_size;
> +	u32 attr;
> +};
> +
> +#endif /* _MM_VARIABLE_H_ */
>