Re: [PATCH v8 11/14] tee: add tee_shm_alloc_cma_phys_mem()

[Robin Murphy <robin.murphy@arm.com>]

On 02/05/2025 10:59 am, Jens Wiklander wrote:
> Add tee_shm_alloc_cma_phys_mem() to allocate a physical memory using
> from the default CMA pool. The memory is represented by a tee_shm object
> using the new flag TEE_SHM_CMA_BUF to identify it as physical memory
> from CMA.

If and when it's possible to dynamically delegate any old kernel memory 
to the TEE, it's far from clear why that should involve poking around in 
CMA internals...

> Signed-off-by: Jens Wiklander <jens.wiklander@linaro.org>
> ---
>   drivers/tee/tee_shm.c    | 55 ++++++++++++++++++++++++++++++++++++++--
>   include/linux/tee_core.h |  4 +++
>   2 files changed, 57 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/tee/tee_shm.c b/drivers/tee/tee_shm.c
> index e1ed52ee0a16..faaa0a87bb18 100644
> --- a/drivers/tee/tee_shm.c
> +++ b/drivers/tee/tee_shm.c
> @@ -3,8 +3,11 @@
>    * Copyright (c) 2015-2017, 2019-2021 Linaro Limited
>    */
>   #include <linux/anon_inodes.h>
> +#include <linux/cma.h>
>   #include <linux/device.h>
>   #include <linux/dma-buf.h>
> +#include <linux/dma-map-ops.h>
> +#include <linux/highmem.h>
>   #include <linux/idr.h>
>   #include <linux/io.h>
>   #include <linux/mm.h>
> @@ -13,7 +16,6 @@
>   #include <linux/tee_core.h>
>   #include <linux/uaccess.h>
>   #include <linux/uio.h>
> -#include <linux/highmem.h>
>   #include "tee_private.h"
>   
>   static void shm_put_kernel_pages(struct page **pages, size_t page_count)
> @@ -49,7 +51,14 @@ static void tee_shm_release(struct tee_device *teedev, struct tee_shm *shm)
>   	struct tee_shm *parent_shm = NULL;
>   	void *p = shm;
>   
> -	if (shm->flags & TEE_SHM_DMA_BUF) {
> +	if (shm->flags & TEE_SHM_CMA_BUF) {
> +#if IS_ENABLED(CONFIG_CMA)
> +		struct page *page = phys_to_page(shm->paddr);
> +		struct cma *cma = dev_get_cma_area(&shm->ctx->teedev->dev);

If you want dma_contiguous_default_area as the commit message implies, 
use dma_contiguous_default_area. Appearing to support per-device CMA 
pools but relying on the device not having one is pretty yucky.

But again, why? If you want page-backed DMA-able memory, with all the 
other assumptions being made here, you may as well just rely on 
dma_alloc_pages(DMA_ATTR_SKIP_CPU_SYNC) doing what you want, while also 
being potentially more flexible for !CMA and non-invasive. Or at the 
very least, could the TEE delegation not be composed on top of the 
existing CMA heap allocator?

Thanks,
Robin.

> +
> +		cma_release(cma, page, shm->size / PAGE_SIZE);
> +#endif
> +	} else if (shm->flags & TEE_SHM_DMA_BUF) {
>   		struct tee_shm_dmabuf_ref *ref;
>   
>   		ref = container_of(shm, struct tee_shm_dmabuf_ref, shm);
> @@ -306,6 +315,48 @@ struct tee_shm *tee_shm_alloc_priv_buf(struct tee_context *ctx, size_t size)
>   }
>   EXPORT_SYMBOL_GPL(tee_shm_alloc_priv_buf);
>   
> +struct tee_shm *tee_shm_alloc_cma_phys_mem(struct tee_context *ctx,
> +					   size_t page_count, size_t align)
> +{
> +#if IS_ENABLED(CONFIG_CMA)
> +	struct tee_device *teedev = ctx->teedev;
> +	struct cma *cma = dev_get_cma_area(&teedev->dev);
> +	struct tee_shm *shm;
> +	struct page *page;
> +
> +	if (!tee_device_get(teedev))
> +		return ERR_PTR(-EINVAL);
> +
> +	page = cma_alloc(cma, page_count, align, true/*no_warn*/);
> +	if (!page)
> +		goto err_put_teedev;
> +
> +	shm = kzalloc(sizeof(*shm), GFP_KERNEL);
> +	if (!shm)
> +		goto err_cma_crelease;
> +
> +	refcount_set(&shm->refcount, 1);
> +	shm->ctx = ctx;
> +	shm->paddr = page_to_phys(page);
> +	shm->size = page_count * PAGE_SIZE;
> +	shm->flags = TEE_SHM_CMA_BUF;
> +
> +	teedev_ctx_get(ctx);
> +
> +	return shm;
> +
> +err_cma_crelease:
> +	cma_release(cma, page, page_count);
> +err_put_teedev:
> +	tee_device_put(teedev);
> +
> +	return ERR_PTR(-ENOMEM);
> +#else
> +	return ERR_PTR(-EINVAL);
> +#endif
> +}
> +EXPORT_SYMBOL_GPL(tee_shm_alloc_cma_phys_mem);
> +
>   int tee_dyn_shm_alloc_helper(struct tee_shm *shm, size_t size, size_t align,
>   			     int (*shm_register)(struct tee_context *ctx,
>   						 struct tee_shm *shm,
> diff --git a/include/linux/tee_core.h b/include/linux/tee_core.h
> index 02c07f661349..3a4e1b00fcc7 100644
> --- a/include/linux/tee_core.h
> +++ b/include/linux/tee_core.h
> @@ -29,6 +29,7 @@
>   #define TEE_SHM_POOL		BIT(2)  /* Memory allocated from pool */
>   #define TEE_SHM_PRIV		BIT(3)  /* Memory private to TEE driver */
>   #define TEE_SHM_DMA_BUF		BIT(4)	/* Memory with dma-buf handle */
> +#define TEE_SHM_CMA_BUF		BIT(5)	/* CMA allocated memory */
>   
>   #define TEE_DEVICE_FLAG_REGISTERED	0x1
>   #define TEE_MAX_DEV_NAME_LEN		32
> @@ -310,6 +311,9 @@ void *tee_get_drvdata(struct tee_device *teedev);
>    */
>   struct tee_shm *tee_shm_alloc_priv_buf(struct tee_context *ctx, size_t size);
>   
> +struct tee_shm *tee_shm_alloc_cma_phys_mem(struct tee_context *ctx,
> +					   size_t page_count, size_t align);
> +
>   int tee_dyn_shm_alloc_helper(struct tee_shm *shm, size_t size, size_t align,
>   			     int (*shm_register)(struct tee_context *ctx,
>   						 struct tee_shm *shm,