Skip to contentSkip to conversation

tee: optee: Fix dynamic shm pool allocations

PATCH Superseded Sumit Garg 2 participants 5 messages Oct 31, 2019 – Nov 8, 2019

Subsystems (2)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore

Sumit Garg

author
#0 lore raw

PATCH 1/1 tee: optee: Fix dynamic shm pool allocations

In case of dynamic shared memory pool, kernel memory allocated using dmabuf_mgr pool needs to be registered with OP-TEE prior to its usage during optee_open_session() or optee_invoke_func().

So fix dmabuf_mgr pool allocations via an additional call to optee_shm_register().

Fixes 9733b072a12a ("optee: allow to work without static shared memory")

---

Depends on patch: https://lkml.org/lkml/2019/7/30/506 that adds support to allow registration of kernel buffers with OP-TEE.

 drivers/tee/optee/shm_pool.c | 12 +++++++++++-
 1 file changed, 11 insertions(+), 1 deletion(-)
drivers/tee/optee/shm_pool.c +11 -1
index de1d9b8..0332a53 100644
@@ -17,6 +17,7 @@ static int pool_op_alloc(struct tee_shm_pool_mgr *poolm,
 {
 	unsigned int order = get_order(size);
 	struct page *page;
+	int rc = 0;
 
 	page = alloc_pages(GFP_KERNEL | __GFP_ZERO, order);
 	if (!page)
@@ -26,12 +27,21 @@ static int pool_op_alloc(struct tee_shm_pool_mgr *poolm,
 	shm->paddr = page_to_phys(page);
 	shm->size = PAGE_SIZE << order;
 
-	return 0;
+	if (shm->flags & TEE_SHM_DMA_BUF) {
+		shm->flags |= TEE_SHM_REGISTER;
+		rc = optee_shm_register(shm->ctx, shm, &page, 1 << order,
+					(unsigned long)shm->kaddr);
+	}
+
+	return rc;
 }
 
 static void pool_op_free(struct tee_shm_pool_mgr *poolm,
 			 struct tee_shm *shm)
 {
+	if (shm->flags & TEE_SHM_DMA_BUF)
+		optee_shm_unregister(shm->ctx, shm);
+
 	free_pages((unsigned long)shm->kaddr, get_order(shm->size));
 	shm->kaddr = NULL;
 }
Signature · 1 line
2.7.4
Tested-by Etienne Carriere on v1 1/1
#1 lore raw

Hi Sumit,

Quoted 9 lines · Sumit Garg

In case of dynamic shared memory pool, kernel memory allocated using dmabuf_mgr pool needs to be registered with OP-TEE prior to its usage during optee_open_session() or optee_invoke_func().

So fix dmabuf_mgr pool allocations via an additional call to optee_shm_register().

Fixes 9733b072a12a ("optee: allow to work without static shared memory")

Looks good, I'm picking this up. Etienne told me has tested this on some of his hardware so I'll add:

Thanks,
Jens

Quoted 46 lines

---

Depends on patch: https://lkml.org/lkml/2019/7/30/506 that adds support to allow registration of kernel buffers with OP-TEE.

 drivers/tee/optee/shm_pool.c | 12 +++++++++++-
 1 file changed, 11 insertions(+), 1 deletion(-)
drivers/tee/optee/shm_pool.c +11 -1
index de1d9b8..0332a53 100644
@@ -17,6 +17,7 @@ static int pool_op_alloc(struct tee_shm_pool_mgr *poolm,
 {
 	unsigned int order = get_order(size);
 	struct page *page;
+	int rc = 0;
 
 	page = alloc_pages(GFP_KERNEL | __GFP_ZERO, order);
 	if (!page)
@@ -26,12 +27,21 @@ static int pool_op_alloc(struct tee_shm_pool_mgr *poolm,
 	shm->paddr = page_to_phys(page);
 	shm->size = PAGE_SIZE << order;
 
-	return 0;
+	if (shm->flags & TEE_SHM_DMA_BUF) {
+		shm->flags |= TEE_SHM_REGISTER;
+		rc = optee_shm_register(shm->ctx, shm, &page, 1 << order,
+					(unsigned long)shm->kaddr);
+	}
+
+	return rc;
 }
 
 static void pool_op_free(struct tee_shm_pool_mgr *poolm,
 			 struct tee_shm *shm)
 {
+	if (shm->flags & TEE_SHM_DMA_BUF)
+		optee_shm_unregister(shm->ctx, shm);
+
 	free_pages((unsigned long)shm->kaddr, get_order(shm->size));
 	shm->kaddr = NULL;
 }
Signature · 1 line
2.7.4
author
#2 lore raw

Hi Jens,

Quoted 11 lines · Jens Wiklander

Hi Sumit,

Quoted 7 lines · Sumit Garg

In case of dynamic shared memory pool, kernel memory allocated using dmabuf_mgr pool needs to be registered with OP-TEE prior to its usage during optee_open_session() or optee_invoke_func().

So fix dmabuf_mgr pool allocations via an additional call to optee_shm_register().

Quoted 7 lines
Quoted 2 lines
Fixes 9733b072a12a ("optee: allow to work without static shared memory")

Looks good, I'm picking this up. Etienne told me has tested this on some of his hardware so I'll add:

Thanks for picking this up.

Quoted 5 lines
Quoted 5 lines

---

Depends on patch: https://lkml.org/lkml/2019/7/30/506 that adds support to allow registration of kernel buffers with OP-TEE.

You also need to pick up the dependency patch too. The latest v3 of this patch is here [1] although its unchanged from v2. Also, If you pick that up I will drop it from future revisions of Trusted Keys patchset [2].

[1] https://lkml.org/lkml/2019/10/31/431
[2] https://lkml.org/lkml/2019/10/31/430

-Sumit

Quoted 42 lines
Quoted 42 lines
 drivers/tee/optee/shm_pool.c | 12 +++++++++++-
 1 file changed, 11 insertions(+), 1 deletion(-)
drivers/tee/optee/shm_pool.c +11 -2
index de1d9b8..0332a53 100644
@@ -17,6 +17,7 @@ static int pool_op_alloc(struct tee_shm_pool_mgr *poolm,
 {
      unsigned int order = get_order(size);
      struct page *page;
+     int rc = 0;
 
      page = alloc_pages(GFP_KERNEL | __GFP_ZERO, order);
      if (!page)
@@ -26,12 +27,21 @@ static int pool_op_alloc(struct tee_shm_pool_mgr *poolm,
      shm->paddr = page_to_phys(page);
      shm->size = PAGE_SIZE << order;
 
-     return 0;
+     if (shm->flags & TEE_SHM_DMA_BUF) {
+             shm->flags |= TEE_SHM_REGISTER;
+             rc = optee_shm_register(shm->ctx, shm, &page, 1 << order,
+                                     (unsigned long)shm->kaddr);
+     }
+
+     return rc;
 }
 
 static void pool_op_free(struct tee_shm_pool_mgr *poolm,
                       struct tee_shm *shm)
 {
+     if (shm->flags & TEE_SHM_DMA_BUF)
+             optee_shm_unregister(shm->ctx, shm);
+
      free_pages((unsigned long)shm->kaddr, get_order(shm->size));
      shm->kaddr = NULL;
 }
--

2.7.4

#3 lore raw

Hi Sumit,

Quoted 31 lines · Sumit Garg

Hi Jens,

Quoted 11 lines · Jens Wiklander

Hi Sumit,

Quoted 7 lines · Sumit Garg

In case of dynamic shared memory pool, kernel memory allocated using dmabuf_mgr pool needs to be registered with OP-TEE prior to its usage during optee_open_session() or optee_invoke_func().

So fix dmabuf_mgr pool allocations via an additional call to optee_shm_register().

Quoted 7 lines
Quoted 2 lines
Fixes 9733b072a12a ("optee: allow to work without static shared memory")

Looks good, I'm picking this up. Etienne told me has tested this on some of his hardware so I'll add:

Thanks for picking this up.

Quoted 5 lines
Quoted 5 lines

---

Depends on patch: https://lkml.org/lkml/2019/7/30/506 that adds support to allow registration of kernel buffers with OP-TEE.

Quoted 7 lines

You also need to pick up the dependency patch too. The latest v3 of this patch is here [1] although its unchanged from v2. Also, If you pick that up I will drop it from future revisions of Trusted Keys patchset [2].

[1] https://lkml.org/lkml/2019/10/31/431
[2] https://lkml.org/lkml/2019/10/31/430

So I missed taking [1] into account. I think that it would make more sense to merge this patch ("tee: optee: Fix dynamic shm pool allocations") with [1] ("tee: optee: allow kernel pages to register as shm") into a new patch.

Cheers,
Jens

Quoted 45 lines

-Sumit

Quoted 42 lines
Quoted 42 lines
 drivers/tee/optee/shm_pool.c | 12 +++++++++++-
 1 file changed, 11 insertions(+), 1 deletion(-)
drivers/tee/optee/shm_pool.c +11 -2
index de1d9b8..0332a53 100644
@@ -17,6 +17,7 @@ static int pool_op_alloc(struct tee_shm_pool_mgr *poolm,
 {
      unsigned int order = get_order(size);
      struct page *page;
+     int rc = 0;
 
      page = alloc_pages(GFP_KERNEL | __GFP_ZERO, order);
      if (!page)
@@ -26,12 +27,21 @@ static int pool_op_alloc(struct tee_shm_pool_mgr *poolm,
      shm->paddr = page_to_phys(page);
      shm->size = PAGE_SIZE << order;
 
-     return 0;
+     if (shm->flags & TEE_SHM_DMA_BUF) {
+             shm->flags |= TEE_SHM_REGISTER;
+             rc = optee_shm_register(shm->ctx, shm, &page, 1 << order,
+                                     (unsigned long)shm->kaddr);
+     }
+
+     return rc;
 }
 
 static void pool_op_free(struct tee_shm_pool_mgr *poolm,
                       struct tee_shm *shm)
 {
+     if (shm->flags & TEE_SHM_DMA_BUF)
+             optee_shm_unregister(shm->ctx, shm);
+
      free_pages((unsigned long)shm->kaddr, get_order(shm->size));
      shm->kaddr = NULL;
 }
--

2.7.4

author
#4 lore raw

Hi Jens,

Quoted 40 lines · Jens Wiklander

Hi Sumit,

Quoted 36 lines · Sumit Garg

Hi Jens,

Quoted 11 lines · Jens Wiklander

Hi Sumit,

Quoted 7 lines · Sumit Garg

In case of dynamic shared memory pool, kernel memory allocated using dmabuf_mgr pool needs to be registered with OP-TEE prior to its usage during optee_open_session() or optee_invoke_func().

So fix dmabuf_mgr pool allocations via an additional call to optee_shm_register().

Quoted 7 lines
Quoted 2 lines
Fixes 9733b072a12a ("optee: allow to work without static shared memory")

Looks good, I'm picking this up. Etienne told me has tested this on some of his hardware so I'll add:

Thanks for picking this up.

Quoted 5 lines
Quoted 5 lines

---

Depends on patch: https://lkml.org/lkml/2019/7/30/506 that adds support to allow registration of kernel buffers with OP-TEE.

You also need to pick up the dependency patch too. The latest v3 of this patch is here [1] although its unchanged from v2. Also, If you pick that up I will drop it from future revisions of Trusted Keys patchset [2].

Quoted 7 lines
Quoted 2 lines

[1] https://lkml.org/lkml/2019/10/31/431
[2] https://lkml.org/lkml/2019/10/31/430

So I missed taking [1] into account. I think that it would make more sense to merge this patch ("tee: optee: Fix dynamic shm pool allocations") with [1] ("tee: optee: allow kernel pages to register as shm") into a new patch.

Sure, will send v2 as merged patch instead.

-Sumit

Quoted 48 lines

Cheers,
Jens

Quoted 45 lines

-Sumit

Quoted 42 lines
Quoted 42 lines
 drivers/tee/optee/shm_pool.c | 12 +++++++++++-
 1 file changed, 11 insertions(+), 1 deletion(-)
drivers/tee/optee/shm_pool.c +11 -2
index de1d9b8..0332a53 100644
@@ -17,6 +17,7 @@ static int pool_op_alloc(struct tee_shm_pool_mgr *poolm,
 {
      unsigned int order = get_order(size);
      struct page *page;
+     int rc = 0;
 
      page = alloc_pages(GFP_KERNEL | __GFP_ZERO, order);
      if (!page)
@@ -26,12 +27,21 @@ static int pool_op_alloc(struct tee_shm_pool_mgr *poolm,
      shm->paddr = page_to_phys(page);
      shm->size = PAGE_SIZE << order;
 
-     return 0;
+     if (shm->flags & TEE_SHM_DMA_BUF) {
+             shm->flags |= TEE_SHM_REGISTER;
+             rc = optee_shm_register(shm->ctx, shm, &page, 1 << order,
+                                     (unsigned long)shm->kaddr);
+     }
+
+     return rc;
 }
 
 static void pool_op_free(struct tee_shm_pool_mgr *poolm,
                       struct tee_shm *shm)
 {
+     if (shm->flags & TEE_SHM_DMA_BUF)
+             optee_shm_unregister(shm->ctx, shm);
+
      free_pages((unsigned long)shm->kaddr, get_order(shm->size));
      shm->kaddr = NULL;
 }
--

2.7.4