Re: [PATCH 4/9] dma-buf: heaps: Initialise MediaTek secure heap

[Vijayanand Jitta <quic_vjitta@quicinc.com>]

On 10/20/2023 3:29 PM, Yong Wu (吴勇) wrote:
> On Thu, 2023-10-19 at 10:15 +0530, Vijayanand Jitta wrote:
>>  	 
>> External email : Please do not click links or open attachments until
>> you have verified the sender or the content.
>>  
>>
>> On 9/11/2023 8:00 AM, Yong Wu wrote:
>>> Initialise a mtk_svp heap. Currently just add a null heap, Prepare
>> for
>>> the later patches.
>>>
>>> Signed-off-by: Yong Wu <yong.wu@mediatek.com>
>>> ---
>>>  drivers/dma-buf/heaps/Kconfig           |  8 ++
>>>  drivers/dma-buf/heaps/Makefile          |  1 +
>>>  drivers/dma-buf/heaps/mtk_secure_heap.c | 99
>> +++++++++++++++++++++++++
>>>  3 files changed, 108 insertions(+)
>>>  create mode 100644 drivers/dma-buf/heaps/mtk_secure_heap.c
>>>
>>> diff --git a/drivers/dma-buf/heaps/Kconfig b/drivers/dma-
>> buf/heaps/Kconfig
>>> index a5eef06c4226..729c0cf3eb7c 100644
>>> --- a/drivers/dma-buf/heaps/Kconfig
>>> +++ b/drivers/dma-buf/heaps/Kconfig
>>> @@ -12,3 +12,11 @@ config DMABUF_HEAPS_CMA
>>>    Choose this option to enable dma-buf CMA heap. This heap is
>> backed
>>>    by the Contiguous Memory Allocator (CMA). If your system has
>> these
>>>    regions, you should say Y here.
>>> +
>>> +config DMABUF_HEAPS_MTK_SECURE
>>> +bool "DMA-BUF MediaTek Secure Heap"
>>> +depends on DMABUF_HEAPS && TEE
>>> +help
>>> +  Choose this option to enable dma-buf MediaTek secure heap for
>> Secure
>>> +  Video Path. This heap is backed by TEE client interfaces. If in
>>> +  doubt, say N.
>>> diff --git a/drivers/dma-buf/heaps/Makefile b/drivers/dma-
>> buf/heaps/Makefile
>>> index 974467791032..df559dbe33fe 100644
>>> --- a/drivers/dma-buf/heaps/Makefile
>>> +++ b/drivers/dma-buf/heaps/Makefile
>>> @@ -1,3 +1,4 @@
>>>  # SPDX-License-Identifier: GPL-2.0
>>>  obj-$(CONFIG_DMABUF_HEAPS_SYSTEM)+= system_heap.o
>>>  obj-$(CONFIG_DMABUF_HEAPS_CMA)+= cma_heap.o
>>> +obj-$(CONFIG_DMABUF_HEAPS_MTK_SECURE)+= mtk_secure_heap.o
>>> diff --git a/drivers/dma-buf/heaps/mtk_secure_heap.c b/drivers/dma-
>> buf/heaps/mtk_secure_heap.c
>>> new file mode 100644
>>> index 000000000000..bbf1c8dce23e
>>> --- /dev/null
>>> +++ b/drivers/dma-buf/heaps/mtk_secure_heap.c
>>> @@ -0,0 +1,99 @@
>>> +// SPDX-License-Identifier: GPL-2.0
>>> +/*
>>> + * DMABUF mtk_secure_heap exporter
>>> + *
>>> + * Copyright (C) 2023 MediaTek Inc.
>>> + */
>>> +
>>> +#include <linux/dma-buf.h>
>>> +#include <linux/dma-heap.h>
>>> +#include <linux/err.h>
>>> +#include <linux/module.h>
>>> +#include <linux/slab.h>
>>> +
>>> +/*
>>> + * MediaTek secure (chunk) memory type
>>> + *
>>> + * @KREE_MEM_SEC_CM_TZ: static chunk memory carved out for
>> trustzone.
>>> + */
>>> +enum kree_mem_type {
>>> +KREE_MEM_SEC_CM_TZ = 1,
>>> +};
>>> +
>>> +struct mtk_secure_heap_buffer {
>>> +struct dma_heap*heap;
>>> +size_tsize;
>>> +};
>>> +
>>> +struct mtk_secure_heap {
>>> +const char*name;
>>> +const enum kree_mem_type mem_type;
>>> +};
>>> +
>>> +static struct dma_buf *
>>> +mtk_sec_heap_allocate(struct dma_heap *heap, size_t size,
>>> +      unsigned long fd_flags, unsigned long heap_flags)
>>> +{
>>> +struct mtk_secure_heap_buffer *sec_buf;
>>> +DEFINE_DMA_BUF_EXPORT_INFO(exp_info);
>>> +struct dma_buf *dmabuf;
>>> +int ret;
>>> +
>>> +sec_buf = kzalloc(sizeof(*sec_buf), GFP_KERNEL);
>>
>> As we know, kzalloc can only allocate 4MB at max. So, secure heap has
>> this limitation.
>> can we have a way to allocate more memory in secure heap ? maybe
>> similar to how system heap does?
> 
> This is just the size of a internal structure. I guess you mean the
> secure memory size here. Regarding secure memory allocating flow, our
> flow may be different with yours.
> 
> Let me explain our flow, we have two secure buffer types(heaps).
> a) mtk_svp
> b) mtk_svp_cma which requires the cma binding.
> 
> The memory management of both is inside the TEE. We only need to tell
> the TEE which type and size of buffer we want, and then the TEE will
> perform and return the memory handle to the kernel. The
> kzalloc/alloc_pages is for the normal buffers.
> 
> Regarding the CMA buffer, we only call cma_alloc once, and its
> management is also within the TEE.
> 

Thanks for the details.

I see for mvp_svp, allocation is also specific to TEE, as TEE takes
care of allocation as well. 

I was thinking if allocation path can also be made generic ? without having
dependency on TEE.
For eg : A case where we want to allocate from kernel and secure that memory,
the current secure heap design can't be used. 

Also i suppose TEE allocates contiguous memory for mtk_svp ? or does it support
scattered memory ?

>>
>> Thanks,
>> Vijay
>>