[PATCH v7 0/4] Add memory ownership switch support and enable mss rproc on msm8996

[PATCH v7 0/4] Add memory ownership switch support and enable mss rproc on msm8996

[Avaneesh Kumar Dwivedi]

This patch set does following
	1- Adds new scm call which helps in stage two translation of a memory region
	   so that memory ownership sharing and switching can be achieved on armv8 and later.
	2- Enable mss remoteproc on msm8996

Changes w.r.t. last patchset
	1- Uses dma_alloc_coherent() that dma_alloc_attrs()
	2- Fixes warning from checkpatch with --strict option
	3- Fixes other minor comments

Avaneesh Kumar Dwivedi (4):
  firmware: scm: Add new SCM call API for switching memory ownership
  remoteproc: qcom: refactor mss fw image loading sequence
  remoteproc: qcom: Make secure world call for mem ownership switch
  remoteproc: qcom: Add support for mss remoteproc on msm8996

 .../devicetree/bindings/remoteproc/qcom,q6v5.txt   |   1 +
 drivers/firmware/qcom_scm-32.c                     |   7 +
 drivers/firmware/qcom_scm-64.c                     |  27 ++
 drivers/firmware/qcom_scm.c                        |  94 +++++++
 drivers/firmware/qcom_scm.h                        |   5 +
 drivers/remoteproc/qcom_q6v5_pil.c                 | 273 ++++++++++++++++++---
 include/linux/qcom_scm.h                           |  16 ++
 7 files changed, 385 insertions(+), 38 deletions(-)

-- 
Qualcomm India Private Limited, on behalf of Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project.

[PATCH v7 1/4] firmware: scm: Add new SCM call API for switching memory ownership

[Avaneesh Kumar Dwivedi]

Two different processors on a SOC need to switch memory ownership
during load/unload. To enable this, second level memory map table
need to be updated, which is done by secure layer.
This patch adds the interface for making secure monitor call for
memory ownership switching request.

Signed-off-by: Avaneesh Kumar Dwivedi <akdwived@codeaurora.org>
---
 drivers/firmware/qcom_scm-32.c |  7 ++++
 drivers/firmware/qcom_scm-64.c | 27 ++++++++++++
 drivers/firmware/qcom_scm.c    | 94 ++++++++++++++++++++++++++++++++++++++++++
 drivers/firmware/qcom_scm.h    |  5 +++
 include/linux/qcom_scm.h       | 16 +++++++
 5 files changed, 149 insertions(+)

diff --git a/drivers/firmware/qcom_scm-32.c b/drivers/firmware/qcom_scm-32.c
index 93e3b96..758b910 100644
--- a/drivers/firmware/qcom_scm-32.c
+++ b/drivers/firmware/qcom_scm-32.c
@@ -596,3 +596,10 @@ int __qcom_scm_iommu_secure_ptbl_init(struct device *dev, u64 addr, u32 size,
 {
 	return -ENODEV;
 }
+
+int __qcom_scm_assign_mem(struct device *dev, phys_addr_t mem_region,
+			  size_t mem_sz, phys_addr_t src, size_t src_sz,
+			  phys_addr_t dest, size_t dest_sz)
+{
+	return -ENODEV;
+}
diff --git a/drivers/firmware/qcom_scm-64.c b/drivers/firmware/qcom_scm-64.c
index 6e6d561..6bdbeae 100644
--- a/drivers/firmware/qcom_scm-64.c
+++ b/drivers/firmware/qcom_scm-64.c
@@ -439,3 +439,30 @@ int __qcom_scm_iommu_secure_ptbl_init(struct device *dev, u64 addr, u32 size,
 
 	return ret;
 }
+
+int __qcom_scm_assign_mem(struct device *dev, phys_addr_t mem_region,
+			  size_t mem_sz, phys_addr_t src, size_t src_sz,
+			  phys_addr_t dest, size_t dest_sz)
+{
+	int ret;
+	struct qcom_scm_desc desc = {0};
+	struct arm_smccc_res res;
+
+	desc.args[0] = mem_region;
+	desc.args[1] = mem_sz;
+	desc.args[2] = src;
+	desc.args[3] = src_sz;
+	desc.args[4] = dest;
+	desc.args[5] = dest_sz;
+	desc.args[6] = 0;
+
+	desc.arginfo = QCOM_SCM_ARGS(7, QCOM_SCM_RO, QCOM_SCM_VAL,
+				     QCOM_SCM_RO, QCOM_SCM_VAL, QCOM_SCM_RO,
+				     QCOM_SCM_VAL, QCOM_SCM_VAL);
+
+	ret = qcom_scm_call(dev, QCOM_SCM_SVC_MP,
+			    QCOM_MEM_PROT_ASSIGN_ID,
+			    &desc, &res);
+
+	return ret ? : res.a1;
+}
diff --git a/drivers/firmware/qcom_scm.c b/drivers/firmware/qcom_scm.c
index bb16510..009a42d 100644
--- a/drivers/firmware/qcom_scm.c
+++ b/drivers/firmware/qcom_scm.c
@@ -40,6 +40,18 @@ struct qcom_scm {
 	struct reset_controller_dev reset;
 };
 
+struct qcom_scm_current_perm_info {
+	__le32 vmid;
+	__le32 perm;
+	__le64 ctx;
+	__le32 ctx_size;
+};
+
+struct qcom_scm_mem_map_info {
+	__le64 mem_addr;
+	__le64 mem_size;
+};
+
 static struct qcom_scm *__scm;
 
 static int qcom_scm_clk_enable(void)
@@ -292,6 +304,88 @@ int qcom_scm_pas_shutdown(u32 peripheral)
 }
 EXPORT_SYMBOL(qcom_scm_pas_shutdown);
 
+/**
+ * qcom_scm_assign_mem() - Make a secure call to reassign memory ownership
+ *
+ * @mem_addr: mem region whose ownership need to be reassigned
+ * @mem_sz:   size of the region.
+ * @srcvm:    vmid for current set of owners, each set bit in
+ *            flag indicate a unique owner
+ * @newvm:    array having new owners and corrsponding permission
+ *            flags
+ * @dest_cnt: number of owners in next set.
+ * Return next set of owners on success.
+ */
+int qcom_scm_assign_mem(phys_addr_t mem_addr, size_t mem_sz, int srcvm,
+			struct qcom_scm_vmperm *newvm, int dest_cnt)
+{
+	struct qcom_scm_current_perm_info *destvm;
+	struct qcom_scm_mem_map_info *mem;
+	phys_addr_t memory_phys;
+	phys_addr_t dest_phys;
+	phys_addr_t src_phys;
+	size_t mem_all_sz;
+	size_t memory_sz;
+	size_t dest_sz;
+	size_t src_sz;
+	int next_vm;
+	__le32 *src;
+	void *ptr;
+	int ret;
+	int len;
+	int i;
+
+	src_sz = hweight_long(srcvm) * sizeof(*src);
+	memory_sz = sizeof(*mem);
+	dest_sz = dest_cnt * sizeof(*destvm);
+	mem_all_sz = src_sz + memory_sz + dest_sz;
+
+	ptr = dma_alloc_coherent(__scm->dev, ALIGN(mem_all_sz, SZ_64),
+				 &src_phys, GFP_KERNEL);
+
+	if (!ptr)
+		return -ENOMEM;
+
+	/* Fill source vmid detail */
+	src = ptr;
+	len = hweight_long(srcvm);
+	for (i = 0; i < len; i++) {
+		src[i] = cpu_to_le32(ffs(srcvm) - 1);
+		srcvm ^= 1 << (ffs(srcvm) - 1);
+	}
+
+	/* Fill details of mem buff to map */
+	mem = ptr + ALIGN(src_sz, SZ_64);
+	memory_phys = src_phys + ALIGN(src_sz, SZ_64);
+	mem[0].mem_addr = cpu_to_le64(mem_addr);
+	mem[0].mem_size = cpu_to_le64(mem_sz);
+
+	next_vm = 0;
+	/* Fill details of next vmid detail */
+	destvm = ptr + ALIGN(memory_sz, SZ_64) + ALIGN(src_sz, SZ_64);
+	dest_phys = memory_phys + ALIGN(memory_sz, SZ_64);
+	for (i = 0; i < dest_cnt; i++) {
+		destvm[i].vmid = cpu_to_le32(newvm[i].vmid);
+		destvm[i].perm = cpu_to_le32(newvm[i].perm);
+		destvm[i].ctx = 0;
+		destvm[i].ctx_size = 0;
+		next_vm |= BIT(newvm[i].vmid);
+	}
+
+	ret = __qcom_scm_assign_mem(__scm->dev, memory_phys, memory_sz,
+				    src_phys, src_sz, dest_phys, dest_sz);
+	dma_free_coherent(__scm->dev, ALIGN(mem_all_sz, SZ_64),
+			  ptr, src_phys);
+	if (ret != 0) {
+		dev_err(__scm->dev,
+			"Assign memory protection call failed %d.\n", ret);
+		return -EINVAL;
+	} else {
+		return next_vm;
+	}
+}
+EXPORT_SYMBOL(qcom_scm_assign_mem);
+
 static int qcom_scm_pas_reset_assert(struct reset_controller_dev *rcdev,
 				     unsigned long idx)
 {
diff --git a/drivers/firmware/qcom_scm.h b/drivers/firmware/qcom_scm.h
index 9bea691..fe54b7b 100644
--- a/drivers/firmware/qcom_scm.h
+++ b/drivers/firmware/qcom_scm.h
@@ -95,5 +95,10 @@ extern int __qcom_scm_iommu_secure_ptbl_size(struct device *dev, u32 spare,
 					     size_t *size);
 extern int __qcom_scm_iommu_secure_ptbl_init(struct device *dev, u64 addr,
 					     u32 size, u32 spare);
+#define QCOM_MEM_PROT_ASSIGN_ID	0x16
+extern int  __qcom_scm_assign_mem(struct device *dev,
+				  phys_addr_t mem_region, size_t mem_sz,
+				  phys_addr_t src, size_t src_sz,
+				  phys_addr_t dest, size_t dest_sz);
 
 #endif
diff --git a/include/linux/qcom_scm.h b/include/linux/qcom_scm.h
index e538047..a9c284d 100644
--- a/include/linux/qcom_scm.h
+++ b/include/linux/qcom_scm.h
@@ -23,6 +23,19 @@ struct qcom_scm_hdcp_req {
 	u32 val;
 };
 
+struct qcom_scm_vmperm {
+	int vmid;
+	int perm;
+};
+
+#define QCOM_SCM_VMID_HLOS       0x3
+#define QCOM_SCM_VMID_MSS_MSA    0xF
+#define QCOM_SCM_PERM_READ       0x4
+#define QCOM_SCM_PERM_WRITE      0x2
+#define QCOM_SCM_PERM_EXEC       0x1
+#define QCOM_SCM_PERM_RW (QCOM_SCM_PERM_READ | QCOM_SCM_PERM_WRITE)
+#define QCOM_SCM_PERM_RWX (QCOM_SCM_PERM_RW | QCOM_SCM_PERM_EXEC)
+
 #if IS_ENABLED(CONFIG_QCOM_SCM)
 extern int qcom_scm_set_cold_boot_addr(void *entry, const cpumask_t *cpus);
 extern int qcom_scm_set_warm_boot_addr(void *entry, const cpumask_t *cpus);
@@ -37,6 +50,9 @@ extern int qcom_scm_pas_mem_setup(u32 peripheral, phys_addr_t addr,
 				  phys_addr_t size);
 extern int qcom_scm_pas_auth_and_reset(u32 peripheral);
 extern int qcom_scm_pas_shutdown(u32 peripheral);
+extern int qcom_scm_assign_mem(phys_addr_t mem_addr, size_t mem_sz,
+			       int src, struct qcom_scm_vmperm *newvm,
+			       int dest_cnt);
 extern void qcom_scm_cpu_power_down(u32 flags);
 extern u32 qcom_scm_get_version(void);
 extern int qcom_scm_set_remote_state(u32 state, u32 id);
-- 
Qualcomm India Private Limited, on behalf of Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project.

Re: [PATCH v7 1/4] firmware: scm: Add new SCM call API for switching memory ownership

[Bjorn Andersson]

On Fri 21 Jul 03:49 PDT 2017, Avaneesh Kumar Dwivedi wrote:

> Two different processors on a SOC need to switch memory ownership
> during load/unload. To enable this, second level memory map table
> need to be updated, which is done by secure layer.
> This patch adds the interface for making secure monitor call for
> memory ownership switching request.
> 

As I reported to you a while back I finally managed to use this to get
the modem on db820c up and running (with "all" IPCROUTER services
showing up). So let's try to resurrect and get this merged!


In addition I've successfully used this patch in extending the rmtfs
shared memory driver to allow setting up the ownership of that memory.

[..]
> diff --git a/drivers/firmware/qcom_scm.c b/drivers/firmware/qcom_scm.c
> index bb16510..009a42d 100644
> --- a/drivers/firmware/qcom_scm.c
> +++ b/drivers/firmware/qcom_scm.c
> @@ -40,6 +40,18 @@ struct qcom_scm {
>  	struct reset_controller_dev reset;
>  };
>  
> +struct qcom_scm_current_perm_info {
> +	__le32 vmid;
> +	__le32 perm;
> +	__le64 ctx;
> +	__le32 ctx_size;
> +};

I learned the hard way that this struct is supposed to be 24 bytes, so
please add a __le32 padding; at the end of this.

[..]
> +int qcom_scm_assign_mem(phys_addr_t mem_addr, size_t mem_sz, int srcvm,
> +			struct qcom_scm_vmperm *newvm, int dest_cnt)
> +{

It turns out that the standard way of calling this (shown by the
remoteproc and rmtfs-memory driver implementations) is:

	ret = qcom_scm_assign_mem(mem, len, curr_vm, perms, sizeof(perms));
	if (ret < 0)
		fail();
	curr_vm = ret;

I therefor suggest that we make one more adjustment to the prototype in
the form of taking srcvm as a pointer. And as this is now a bitmask it
should be made an unsigned int. I.e. 

int qcom_scm_assign_mem(phys_addr_t mem_addr, size_t mem_sz, unsigned int srcvm,
			struct qcom_scm_vmperm *newvm, int dest_cnt)

> +	struct qcom_scm_current_perm_info *destvm;
> +	struct qcom_scm_mem_map_info *mem;
> +	phys_addr_t memory_phys;
> +	phys_addr_t dest_phys;
> +	phys_addr_t src_phys;
> +	size_t mem_all_sz;
> +	size_t memory_sz;
> +	size_t dest_sz;
> +	size_t src_sz;
> +	int next_vm;
> +	__le32 *src;
> +	void *ptr;
> +	int ret;
> +	int len;
> +	int i;
> +
> +	src_sz = hweight_long(srcvm) * sizeof(*src);
> +	memory_sz = sizeof(*mem);
> +	dest_sz = dest_cnt * sizeof(*destvm);
> +	mem_all_sz = src_sz + memory_sz + dest_sz;

ALIGN(x + y + z, 64) <= ALIGN(x, 64) + ALIGN(y, 64) + ALIGN(z, 64) 

So please replace this with:

	ptr_sz = ALIGN(src_sz, SZ_64) + ALIGN(memory_sz, SZ_64) +
		 ALIGN(dest_sz, SZ_64);

(renaming the variable to ptr_sz saves you some line wraps as well)

> +
> +	ptr = dma_alloc_coherent(__scm->dev, ALIGN(mem_all_sz, SZ_64),
> +				 &src_phys, GFP_KERNEL);
> +
> +	if (!ptr)
> +		return -ENOMEM;
> +
> +	/* Fill source vmid detail */
> +	src = ptr;
> +	len = hweight_long(srcvm);
> +	for (i = 0; i < len; i++) {
> +		src[i] = cpu_to_le32(ffs(srcvm) - 1);
> +		srcvm ^= 1 << (ffs(srcvm) - 1);
> +	}
> +
> +	/* Fill details of mem buff to map */
> +	mem = ptr + ALIGN(src_sz, SZ_64);
> +	memory_phys = src_phys + ALIGN(src_sz, SZ_64);
> +	mem[0].mem_addr = cpu_to_le64(mem_addr);
> +	mem[0].mem_size = cpu_to_le64(mem_sz);
> +
> +	next_vm = 0;
> +	/* Fill details of next vmid detail */
> +	destvm = ptr + ALIGN(memory_sz, SZ_64) + ALIGN(src_sz, SZ_64);
> +	dest_phys = memory_phys + ALIGN(memory_sz, SZ_64);

For clarity it would be nice if you keep the math for virtual and
physical addresses the same; so add another variable "ptr_phys" and
replace this with:
	dest_phys = ptr_phys + ALIGN(src_sz, 64) + ALIGN(memory_sz, 64);

> +	for (i = 0; i < dest_cnt; i++) {
> +		destvm[i].vmid = cpu_to_le32(newvm[i].vmid);
> +		destvm[i].perm = cpu_to_le32(newvm[i].perm);
> +		destvm[i].ctx = 0;
> +		destvm[i].ctx_size = 0;
> +		next_vm |= BIT(newvm[i].vmid);
> +	}
> +
> +	ret = __qcom_scm_assign_mem(__scm->dev, memory_phys, memory_sz,
> +				    src_phys, src_sz, dest_phys, dest_sz);
> +	dma_free_coherent(__scm->dev, ALIGN(mem_all_sz, SZ_64),
> +			  ptr, src_phys);
> +	if (ret != 0) {
> +		dev_err(__scm->dev,
> +			"Assign memory protection call failed %d.\n", ret);
> +		return -EINVAL;
> +	} else {
> +		return next_vm;
> +	}

Replace this with:

	if (ret) {
		dev_err(__scm->dev,
			"Assign memory protection call failed %d.\n", ret);
		return -EINVAL;
	}

	*srcvm = next_vm;
	return 0;


> +}
> +EXPORT_SYMBOL(qcom_scm_assign_mem);
> +

Regards,
Bjorn

Re: [PATCH v7 1/4] firmware: scm: Add new SCM call API for switching memory ownership

[Dwivedi, Avaneesh Kumar (avani)]

On 10/12/2017 11:41 PM, Bjorn Andersson wrote:
> On Fri 21 Jul 03:49 PDT 2017, Avaneesh Kumar Dwivedi wrote:
>
>> Two different processors on a SOC need to switch memory ownership
>> during load/unload. To enable this, second level memory map table
>> need to be updated, which is done by secure layer.
>> This patch adds the interface for making secure monitor call for
>> memory ownership switching request.
>>
> As I reported to you a while back I finally managed to use this to get
> the modem on db820c up and running (with "all" IPCROUTER services
> showing up). So let's try to resurrect and get this merged!
>
>
> In addition I've successfully used this patch in extending the rmtfs
> shared memory driver to allow setting up the ownership of that memory.
>
> [..]
Thanks for reviving this patch series Bjorn!

Will try to take it closure this time.

>> diff --git a/drivers/firmware/qcom_scm.c b/drivers/firmware/qcom_scm.c
>> index bb16510..009a42d 100644
>> --- a/drivers/firmware/qcom_scm.c
>> +++ b/drivers/firmware/qcom_scm.c
>> @@ -40,6 +40,18 @@ struct qcom_scm {
>>   	struct reset_controller_dev reset;
>>   };
>>   
>> +struct qcom_scm_current_perm_info {
>> +	__le32 vmid;
>> +	__le32 perm;
>> +	__le64 ctx;
>> +	__le32 ctx_size;
>> +};
> I learned the hard way that this struct is supposed to be 24 bytes, so
> please add a __le32 padding; at the end of this.
>
> [..]
OK.
>> +int qcom_scm_assign_mem(phys_addr_t mem_addr, size_t mem_sz, int srcvm,
>> +			struct qcom_scm_vmperm *newvm, int dest_cnt)
>> +{
> It turns out that the standard way of calling this (shown by the
> remoteproc and rmtfs-memory driver implementations) is:
>
> 	ret = qcom_scm_assign_mem(mem, len, curr_vm, perms, sizeof(perms));
> 	if (ret < 0)
> 		fail();
> 	curr_vm = ret;
>
> I therefor suggest that we make one more adjustment to the prototype in
> the form of taking srcvm as a pointer. And as this is now a bitmask it
> should be made an unsigned int. I.e.
>
> int qcom_scm_assign_mem(phys_addr_t mem_addr, size_t mem_sz, unsigned int srcvm,
> 			struct qcom_scm_vmperm *newvm, int dest_cnt)
>
>> +	struct qcom_scm_current_perm_info *destvm;
>> +	struct qcom_scm_mem_map_info *mem;
>> +	phys_addr_t memory_phys;
>> +	phys_addr_t dest_phys;
>> +	phys_addr_t src_phys;
>> +	size_t mem_all_sz;
>> +	size_t memory_sz;
>> +	size_t dest_sz;
>> +	size_t src_sz;
>> +	int next_vm;
>> +	__le32 *src;
>> +	void *ptr;
>> +	int ret;
>> +	int len;
>> +	int i;
>> +
>> +	src_sz = hweight_long(srcvm) * sizeof(*src);
>> +	memory_sz = sizeof(*mem);
>> +	dest_sz = dest_cnt * sizeof(*destvm);
>> +	mem_all_sz = src_sz + memory_sz + dest_sz;
> ALIGN(x + y + z, 64) <= ALIGN(x, 64) + ALIGN(y, 64) + ALIGN(z, 64)
>
> So please replace this with:
>
> 	ptr_sz = ALIGN(src_sz, SZ_64) + ALIGN(memory_sz, SZ_64) +
> 		 ALIGN(dest_sz, SZ_64);
>
> (renaming the variable to ptr_sz saves you some line wraps as well)
Ok, Will try to incorporate your idea, and if any concern will revert 
back soon.
>> +
>> +	ptr = dma_alloc_coherent(__scm->dev, ALIGN(mem_all_sz, SZ_64),
>> +				 &src_phys, GFP_KERNEL);
>> +
>> +	if (!ptr)
>> +		return -ENOMEM;
>> +
>> +	/* Fill source vmid detail */
>> +	src = ptr;
>> +	len = hweight_long(srcvm);
>> +	for (i = 0; i < len; i++) {
>> +		src[i] = cpu_to_le32(ffs(srcvm) - 1);
>> +		srcvm ^= 1 << (ffs(srcvm) - 1);
>> +	}
>> +
>> +	/* Fill details of mem buff to map */
>> +	mem = ptr + ALIGN(src_sz, SZ_64);
>> +	memory_phys = src_phys + ALIGN(src_sz, SZ_64);
>> +	mem[0].mem_addr = cpu_to_le64(mem_addr);
>> +	mem[0].mem_size = cpu_to_le64(mem_sz);
>> +
>> +	next_vm = 0;
>> +	/* Fill details of next vmid detail */
>> +	destvm = ptr + ALIGN(memory_sz, SZ_64) + ALIGN(src_sz, SZ_64);
>> +	dest_phys = memory_phys + ALIGN(memory_sz, SZ_64);
> For clarity it would be nice if you keep the math for virtual and
> physical addresses the same; so add another variable "ptr_phys" and
> replace this with:
> 	dest_phys = ptr_phys + ALIGN(src_sz, 64) + ALIGN(memory_sz, 64);
OK.
>
>> +	for (i = 0; i < dest_cnt; i++) {
>> +		destvm[i].vmid = cpu_to_le32(newvm[i].vmid);
>> +		destvm[i].perm = cpu_to_le32(newvm[i].perm);
>> +		destvm[i].ctx = 0;
>> +		destvm[i].ctx_size = 0;
>> +		next_vm |= BIT(newvm[i].vmid);
>> +	}
>> +
>> +	ret = __qcom_scm_assign_mem(__scm->dev, memory_phys, memory_sz,
>> +				    src_phys, src_sz, dest_phys, dest_sz);
>> +	dma_free_coherent(__scm->dev, ALIGN(mem_all_sz, SZ_64),
>> +			  ptr, src_phys);
>> +	if (ret != 0) {
>> +		dev_err(__scm->dev,
>> +			"Assign memory protection call failed %d.\n", ret);
>> +		return -EINVAL;
>> +	} else {
>> +		return next_vm;
>> +	}
> Replace this with:
>
> 	if (ret) {
> 		dev_err(__scm->dev,
> 			"Assign memory protection call failed %d.\n", ret);
> 		return -EINVAL;
> 	}
>
> 	*srcvm = next_vm;
> 	return 0;
OK.
>
>> +}
>> +EXPORT_SYMBOL(qcom_scm_assign_mem);
>> +
> Regards,
> Bjorn
> --
> To unsubscribe from this list: send the line "unsubscribe linux-remoteproc" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

-- 
Qualcomm India Private Limited, on behalf of Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project.

[PATCH v7 2/4] remoteproc: qcom: refactor mss fw image loading sequence

[Avaneesh Kumar Dwivedi]

This patch refactor code to first load all firmware blobs
and then update modem proc to authenticate and boot fw.

Signed-off-by: Avaneesh Kumar Dwivedi <akdwived@codeaurora.org>
---
 drivers/remoteproc/qcom_q6v5_pil.c | 19 ++++++++-----------
 1 file changed, 8 insertions(+), 11 deletions(-)

diff --git a/drivers/remoteproc/qcom_q6v5_pil.c b/drivers/remoteproc/qcom_q6v5_pil.c
index 8fd697a..f5f8850 100644
--- a/drivers/remoteproc/qcom_q6v5_pil.c
+++ b/drivers/remoteproc/qcom_q6v5_pil.c
@@ -503,7 +503,7 @@ static int q6v5_mpss_load(struct q6v5 *qproc)
 	bool relocate = false;
 	char seg_name[10];
 	ssize_t offset;
-	size_t size;
+	size_t size = 0;
 	void *ptr;
 	int ret;
 	int i;
@@ -541,7 +541,7 @@ static int q6v5_mpss_load(struct q6v5 *qproc)
 	}
 
 	mpss_reloc = relocate ? min_addr : qproc->mpss_phys;
-
+	/* Load firmware segments */
 	for (i = 0; i < ehdr->e_phnum; i++) {
 		phdr = &phdrs[i];
 
@@ -574,18 +574,15 @@ static int q6v5_mpss_load(struct q6v5 *qproc)
 			memset(ptr + phdr->p_filesz, 0,
 			       phdr->p_memsz - phdr->p_filesz);
 		}
-
-		size = readl(qproc->rmb_base + RMB_PMI_CODE_LENGTH_REG);
-		if (!size) {
-			boot_addr = relocate ? qproc->mpss_phys : min_addr;
-			writel(boot_addr, qproc->rmb_base + RMB_PMI_CODE_START_REG);
-			writel(RMB_CMD_LOAD_READY, qproc->rmb_base + RMB_MBA_COMMAND_REG);
-		}
-
 		size += phdr->p_memsz;
-		writel(size, qproc->rmb_base + RMB_PMI_CODE_LENGTH_REG);
 	}
 
+	/* Transfer ownership of modem ddr region with q6*/
+	boot_addr = relocate ? qproc->mpss_phys : min_addr;
+	writel(boot_addr, qproc->rmb_base + RMB_PMI_CODE_START_REG);
+	writel(RMB_CMD_LOAD_READY, qproc->rmb_base + RMB_MBA_COMMAND_REG);
+	writel(size, qproc->rmb_base + RMB_PMI_CODE_LENGTH_REG);
+
 	ret = q6v5_rmb_mba_wait(qproc, RMB_MBA_AUTH_COMPLETE, 10000);
 	if (ret == -ETIMEDOUT)
 		dev_err(qproc->dev, "MPSS authentication timed out\n");
-- 
Qualcomm India Private Limited, on behalf of Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project.

Re: [PATCH v7 2/4] remoteproc: qcom: refactor mss fw image loading sequence

[Bjorn Andersson]

On Fri 21 Jul 03:49 PDT 2017, Avaneesh Kumar Dwivedi wrote:

> This patch refactor code to first load all firmware blobs
> and then update modem proc to authenticate and boot fw.
> 

Acked-by: Bjorn Andersson <bjorn.andersson@linaro.org>

Regards,
Bjorn

> Signed-off-by: Avaneesh Kumar Dwivedi <akdwived@codeaurora.org>
> ---
>  drivers/remoteproc/qcom_q6v5_pil.c | 19 ++++++++-----------
>  1 file changed, 8 insertions(+), 11 deletions(-)
> 
> diff --git a/drivers/remoteproc/qcom_q6v5_pil.c b/drivers/remoteproc/qcom_q6v5_pil.c
> index 8fd697a..f5f8850 100644
> --- a/drivers/remoteproc/qcom_q6v5_pil.c
> +++ b/drivers/remoteproc/qcom_q6v5_pil.c
> @@ -503,7 +503,7 @@ static int q6v5_mpss_load(struct q6v5 *qproc)
>  	bool relocate = false;
>  	char seg_name[10];
>  	ssize_t offset;
> -	size_t size;
> +	size_t size = 0;
>  	void *ptr;
>  	int ret;
>  	int i;
> @@ -541,7 +541,7 @@ static int q6v5_mpss_load(struct q6v5 *qproc)
>  	}
>  
>  	mpss_reloc = relocate ? min_addr : qproc->mpss_phys;
> -
> +	/* Load firmware segments */
>  	for (i = 0; i < ehdr->e_phnum; i++) {
>  		phdr = &phdrs[i];
>  
> @@ -574,18 +574,15 @@ static int q6v5_mpss_load(struct q6v5 *qproc)
>  			memset(ptr + phdr->p_filesz, 0,
>  			       phdr->p_memsz - phdr->p_filesz);
>  		}
> -
> -		size = readl(qproc->rmb_base + RMB_PMI_CODE_LENGTH_REG);
> -		if (!size) {
> -			boot_addr = relocate ? qproc->mpss_phys : min_addr;
> -			writel(boot_addr, qproc->rmb_base + RMB_PMI_CODE_START_REG);
> -			writel(RMB_CMD_LOAD_READY, qproc->rmb_base + RMB_MBA_COMMAND_REG);
> -		}
> -
>  		size += phdr->p_memsz;
> -		writel(size, qproc->rmb_base + RMB_PMI_CODE_LENGTH_REG);
>  	}
>  
> +	/* Transfer ownership of modem ddr region with q6*/
> +	boot_addr = relocate ? qproc->mpss_phys : min_addr;
> +	writel(boot_addr, qproc->rmb_base + RMB_PMI_CODE_START_REG);
> +	writel(RMB_CMD_LOAD_READY, qproc->rmb_base + RMB_MBA_COMMAND_REG);
> +	writel(size, qproc->rmb_base + RMB_PMI_CODE_LENGTH_REG);
> +
>  	ret = q6v5_rmb_mba_wait(qproc, RMB_MBA_AUTH_COMPLETE, 10000);
>  	if (ret == -ETIMEDOUT)
>  		dev_err(qproc->dev, "MPSS authentication timed out\n");
> -- 
> Qualcomm India Private Limited, on behalf of Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
> a Linux Foundation Collaborative Project.
> 

[PATCH v7 3/4] remoteproc: qcom: Make secure world call for mem ownership switch

[Avaneesh Kumar Dwivedi]

MSS proc on msm8996 can not access fw loaded region without stage
second translation of memory pages where mpss image are loaded.
This patch in order to enable mss boot on msm8996 invoke scm call
to switch or share ownership between apps and modem.

Signed-off-by: Avaneesh Kumar Dwivedi <akdwived@codeaurora.org>
---
 drivers/remoteproc/qcom_q6v5_pil.c | 90 ++++++++++++++++++++++++++++++++++++--
 1 file changed, 87 insertions(+), 3 deletions(-)

diff --git a/drivers/remoteproc/qcom_q6v5_pil.c b/drivers/remoteproc/qcom_q6v5_pil.c
index f5f8850..a82b62a 100644
--- a/drivers/remoteproc/qcom_q6v5_pil.c
+++ b/drivers/remoteproc/qcom_q6v5_pil.c
@@ -110,6 +110,7 @@ struct rproc_hexagon_res {
 	struct qcom_mss_reg_res *active_supply;
 	char **proxy_clk_names;
 	char **active_clk_names;
+	bool need_mem_protection;
 };
 
 struct q6v5 {
@@ -151,8 +152,11 @@ struct q6v5 {
 	phys_addr_t mpss_reloc;
 	void *mpss_region;
 	size_t mpss_size;
+	bool need_mem_protection;
 
 	struct qcom_rproc_subdev smd_subdev;
+	int mpss_perm;
+	int mba_perm;
 };
 
 static int q6v5_regulator_init(struct device *dev, struct reg_info *regs,
@@ -288,6 +292,36 @@ static struct resource_table *q6v5_find_rsc_table(struct rproc *rproc,
 	return &table;
 }
 
+static int q6v5_xfer_mem_ownership(struct q6v5 *qproc, int *current_perm,
+				   bool remote_owner, phys_addr_t addr,
+				   size_t size)
+{
+	struct qcom_scm_vmperm next;
+	int ret;
+
+	if (!qproc->need_mem_protection)
+		return 0;
+	if (remote_owner && *current_perm == BIT(QCOM_SCM_VMID_MSS_MSA))
+		return 0;
+	if (!remote_owner && *current_perm == BIT(QCOM_SCM_VMID_HLOS))
+		return 0;
+
+	next.vmid = remote_owner ? QCOM_SCM_VMID_MSS_MSA : QCOM_SCM_VMID_HLOS;
+	next.perm = remote_owner ? QCOM_SCM_PERM_RW : QCOM_SCM_PERM_RWX;
+
+	ret = qcom_scm_assign_mem(addr, ALIGN(size, SZ_4K),
+				  *current_perm, &next, 1);
+	if (ret < 0) {
+		pr_err("Failed to assign memory access in range %p to %p to %s ret = %d\n",
+		       (void *)addr, (void *)(addr + size),
+		       remote_owner ? "mss" : "hlos", ret);
+		return ret;
+	}
+
+	*current_perm = ret;
+	return 0;
+}
+
 static int q6v5_load(struct rproc *rproc, const struct firmware *fw)
 {
 	struct q6v5 *qproc = rproc->priv;
@@ -450,6 +484,8 @@ static int q6v5_mpss_init_image(struct q6v5 *qproc, const struct firmware *fw)
 {
 	unsigned long dma_attrs = DMA_ATTR_FORCE_CONTIGUOUS;
 	dma_addr_t phys;
+	int mdata_perm;
+	int xferop_ret;
 	void *ptr;
 	int ret;
 
@@ -461,6 +497,12 @@ static int q6v5_mpss_init_image(struct q6v5 *qproc, const struct firmware *fw)
 
 	memcpy(ptr, fw->data, fw->size);
 
+	/* Hypervisor mapping to access metadata by modem */
+	mdata_perm = BIT(QCOM_SCM_VMID_HLOS);
+	ret = q6v5_xfer_mem_ownership(qproc, &mdata_perm,
+				      1, phys, fw->size);
+	if (ret)
+		return -EAGAIN;
 	writel(phys, qproc->rmb_base + RMB_PMI_META_DATA_REG);
 	writel(RMB_CMD_META_DATA_READY, qproc->rmb_base + RMB_MBA_COMMAND_REG);
 
@@ -470,6 +512,12 @@ static int q6v5_mpss_init_image(struct q6v5 *qproc, const struct firmware *fw)
 	else if (ret < 0)
 		dev_err(qproc->dev, "MPSS header authentication failed: %d\n", ret);
 
+	/* Metadata authentication done, remove modem access */
+	xferop_ret = q6v5_xfer_mem_ownership(qproc, &mdata_perm,
+					     0, phys, fw->size);
+	if (xferop_ret)
+		dev_warn(qproc->dev,
+			 "mdt buffer not reclaimed system may become unstable\n");
 	dma_free_attrs(qproc->dev, fw->size, ptr, phys, dma_attrs);
 
 	return ret < 0 ? ret : 0;
@@ -579,6 +627,10 @@ static int q6v5_mpss_load(struct q6v5 *qproc)
 
 	/* Transfer ownership of modem ddr region with q6*/
 	boot_addr = relocate ? qproc->mpss_phys : min_addr;
+	ret = q6v5_xfer_mem_ownership(qproc, &qproc->mpss_perm, 1,
+				      qproc->mpss_phys, qproc->mpss_size);
+	if (ret)
+		return -EAGAIN;
 	writel(boot_addr, qproc->rmb_base + RMB_PMI_CODE_START_REG);
 	writel(RMB_CMD_LOAD_READY, qproc->rmb_base + RMB_MBA_COMMAND_REG);
 	writel(size, qproc->rmb_base + RMB_PMI_CODE_LENGTH_REG);
@@ -598,6 +650,7 @@ static int q6v5_mpss_load(struct q6v5 *qproc)
 static int q6v5_start(struct rproc *rproc)
 {
 	struct q6v5 *qproc = (struct q6v5 *)rproc->priv;
+	int xfermemop_ret;
 	int ret;
 
 	ret = q6v5_regulator_enable(qproc, qproc->proxy_regs,
@@ -633,6 +686,12 @@ static int q6v5_start(struct rproc *rproc)
 		goto assert_reset;
 	}
 
+	/* Assign MBA image access in DDR to q6 */
+	xfermemop_ret = q6v5_xfer_mem_ownership(qproc, &qproc->mba_perm, 1,
+						qproc->mba_phys,
+						qproc->mba_size);
+	if (xfermemop_ret)
+		goto assert_reset;
 	writel(qproc->mba_phys, qproc->rmb_base + RMB_MBA_IMAGE_REG);
 
 	ret = q6v5proc_reset(qproc);
@@ -654,16 +713,22 @@ static int q6v5_start(struct rproc *rproc)
 
 	ret = q6v5_mpss_load(qproc);
 	if (ret)
-		goto halt_axi_ports;
+		goto reclaim_mem;
 
 	ret = wait_for_completion_timeout(&qproc->start_done,
 					  msecs_to_jiffies(5000));
 	if (ret == 0) {
 		dev_err(qproc->dev, "start timed out\n");
 		ret = -ETIMEDOUT;
-		goto halt_axi_ports;
+		goto reclaim_mem;
 	}
 
+	xfermemop_ret = q6v5_xfer_mem_ownership(qproc, &qproc->mba_perm, 0,
+						qproc->mba_phys,
+						qproc->mba_size);
+	if (xfermemop_ret)
+		dev_err(qproc->dev,
+			"Failed to reclaim mba buffer system may become unstable\n");
 	qproc->running = true;
 
 	q6v5_clk_disable(qproc->dev, qproc->proxy_clks,
@@ -673,12 +738,23 @@ static int q6v5_start(struct rproc *rproc)
 
 	return 0;
 
+reclaim_mem:
+	xfermemop_ret = q6v5_xfer_mem_ownership(qproc, &qproc->mpss_perm,
+						0, qproc->mpss_phys,
+						qproc->mpss_size);
+	WARN_ON(xfermemop_ret);
 halt_axi_ports:
 	q6v5proc_halt_axi_port(qproc, qproc->halt_map, qproc->halt_q6);
 	q6v5proc_halt_axi_port(qproc, qproc->halt_map, qproc->halt_modem);
 	q6v5proc_halt_axi_port(qproc, qproc->halt_map, qproc->halt_nc);
 	q6v5_clk_disable(qproc->dev, qproc->active_clks,
 			 qproc->active_clk_count);
+	xfermemop_ret = q6v5_xfer_mem_ownership(qproc, &qproc->mba_perm, 0,
+						qproc->mba_phys,
+						qproc->mba_size);
+	if (xfermemop_ret)
+		dev_err(qproc->dev, "Failed to reclaim mba buffer, system may become unstable\n");
+
 assert_reset:
 	reset_control_assert(qproc->mss_restart);
 disable_vdd:
@@ -698,6 +774,7 @@ static int q6v5_stop(struct rproc *rproc)
 {
 	struct q6v5 *qproc = (struct q6v5 *)rproc->priv;
 	int ret;
+	u32 val;
 
 	qproc->running = false;
 
@@ -715,6 +792,9 @@ static int q6v5_stop(struct rproc *rproc)
 	q6v5proc_halt_axi_port(qproc, qproc->halt_map, qproc->halt_modem);
 	q6v5proc_halt_axi_port(qproc, qproc->halt_map, qproc->halt_nc);
 
+	ret = q6v5_xfer_mem_ownership(qproc, &qproc->mpss_perm, 0,
+				      qproc->mpss_phys, qproc->mpss_size);
+	WARN_ON(ret);
 	reset_control_assert(qproc->mss_restart);
 	q6v5_clk_disable(qproc->dev, qproc->active_clks,
 			 qproc->active_clk_count);
@@ -1012,6 +1092,7 @@ static int q6v5_probe(struct platform_device *pdev)
 	if (ret)
 		goto free_rproc;
 
+	qproc->need_mem_protection = desc->need_mem_protection;
 	ret = q6v5_request_irq(qproc, pdev, "wdog", q6v5_wdog_interrupt);
 	if (ret < 0)
 		goto free_rproc;
@@ -1033,7 +1114,8 @@ static int q6v5_probe(struct platform_device *pdev)
 		ret = PTR_ERR(qproc->state);
 		goto free_rproc;
 	}
-
+	qproc->mpss_perm = BIT(QCOM_SCM_VMID_HLOS);
+	qproc->mba_perm = BIT(QCOM_SCM_VMID_HLOS);
 	qcom_add_smd_subdev(rproc, &qproc->smd_subdev);
 
 	ret = rproc_add(rproc);
@@ -1087,6 +1169,7 @@ static int q6v5_remove(struct platform_device *pdev)
 		"mem",
 		NULL
 	},
+	.need_mem_protection = false,
 };
 
 static const struct rproc_hexagon_res msm8974_mss = {
@@ -1124,6 +1207,7 @@ static int q6v5_remove(struct platform_device *pdev)
 		"mem",
 		NULL
 	},
+	.need_mem_protection = false,
 };
 
 static const struct of_device_id q6v5_of_match[] = {
-- 
Qualcomm India Private Limited, on behalf of Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project.

Re: [PATCH v7 3/4] remoteproc: qcom: Make secure world call for mem ownership switch

[Bjorn Andersson]

On Fri 21 Jul 03:49 PDT 2017, Avaneesh Kumar Dwivedi wrote:

> MSS proc on msm8996 can not access fw loaded region without stage
> second translation of memory pages where mpss image are loaded.
> This patch in order to enable mss boot on msm8996 invoke scm call
> to switch or share ownership between apps and modem.
> 
> Signed-off-by: Avaneesh Kumar Dwivedi <akdwived@codeaurora.org>
> ---
>  drivers/remoteproc/qcom_q6v5_pil.c | 90 ++++++++++++++++++++++++++++++++++++--
>  1 file changed, 87 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/remoteproc/qcom_q6v5_pil.c b/drivers/remoteproc/qcom_q6v5_pil.c
[..]
> +static int q6v5_xfer_mem_ownership(struct q6v5 *qproc, int *current_perm,
> +				   bool remote_owner, phys_addr_t addr,
> +				   size_t size)
> +{
> +	struct qcom_scm_vmperm next;
> +	int ret;
> +
> +	if (!qproc->need_mem_protection)
> +		return 0;
> +	if (remote_owner && *current_perm == BIT(QCOM_SCM_VMID_MSS_MSA))
> +		return 0;
> +	if (!remote_owner && *current_perm == BIT(QCOM_SCM_VMID_HLOS))
> +		return 0;
> +
> +	next.vmid = remote_owner ? QCOM_SCM_VMID_MSS_MSA : QCOM_SCM_VMID_HLOS;
> +	next.perm = remote_owner ? QCOM_SCM_PERM_RW : QCOM_SCM_PERM_RWX;
> +
> +	ret = qcom_scm_assign_mem(addr, ALIGN(size, SZ_4K),
> +				  *current_perm, &next, 1);
> +	if (ret < 0) {
> +		pr_err("Failed to assign memory access in range %p to %p to %s ret = %d\n",
> +		       (void *)addr, (void *)(addr + size),
> +		       remote_owner ? "mss" : "hlos", ret);

qcom_scm_assign_mem() also prints an error when this happens, there's no
need for that. I'm happy with you dropping the print from
qcom_scm_assign_mem() and keeping this one.

> +		return ret;
> +	}
> +
> +	*current_perm = ret;
> +	return 0;
> +}
> +
>  static int q6v5_load(struct rproc *rproc, const struct firmware *fw)
>  {
>  	struct q6v5 *qproc = rproc->priv;
> @@ -450,6 +484,8 @@ static int q6v5_mpss_init_image(struct q6v5 *qproc, const struct firmware *fw)
>  {
>  	unsigned long dma_attrs = DMA_ATTR_FORCE_CONTIGUOUS;
>  	dma_addr_t phys;
> +	int mdata_perm;
> +	int xferop_ret;
>  	void *ptr;
>  	int ret;
>  
> @@ -461,6 +497,12 @@ static int q6v5_mpss_init_image(struct q6v5 *qproc, const struct firmware *fw)
>  
>  	memcpy(ptr, fw->data, fw->size);
>  
> +	/* Hypervisor mapping to access metadata by modem */
> +	mdata_perm = BIT(QCOM_SCM_VMID_HLOS);
> +	ret = q6v5_xfer_mem_ownership(qproc, &mdata_perm,
> +				      1, phys, fw->size);

Please use "true" instead of "1".

> +	if (ret)
> +		return -EAGAIN;

Please keep a empty line here, to keep the "paragraphs" or "chunks" of
code split up.

>  	writel(phys, qproc->rmb_base + RMB_PMI_META_DATA_REG);
>  	writel(RMB_CMD_META_DATA_READY, qproc->rmb_base + RMB_MBA_COMMAND_REG);
>  

Regards,
Bjorn

Re: [PATCH v7 3/4] remoteproc: qcom: Make secure world call for mem ownership switch

[Dwivedi, Avaneesh Kumar (avani)]

On 10/12/2017 11:48 PM, Bjorn Andersson wrote:
> On Fri 21 Jul 03:49 PDT 2017, Avaneesh Kumar Dwivedi wrote:
>
>> MSS proc on msm8996 can not access fw loaded region without stage
>> second translation of memory pages where mpss image are loaded.
>> This patch in order to enable mss boot on msm8996 invoke scm call
>> to switch or share ownership between apps and modem.
>>
>> Signed-off-by: Avaneesh Kumar Dwivedi <akdwived@codeaurora.org>
>> ---
>>   drivers/remoteproc/qcom_q6v5_pil.c | 90 ++++++++++++++++++++++++++++++++++++--
>>   1 file changed, 87 insertions(+), 3 deletions(-)
>>
>> diff --git a/drivers/remoteproc/qcom_q6v5_pil.c b/drivers/remoteproc/qcom_q6v5_pil.c
> [..]
>> +static int q6v5_xfer_mem_ownership(struct q6v5 *qproc, int *current_perm,
>> +				   bool remote_owner, phys_addr_t addr,
>> +				   size_t size)
>> +{
>> +	struct qcom_scm_vmperm next;
>> +	int ret;
>> +
>> +	if (!qproc->need_mem_protection)
>> +		return 0;
>> +	if (remote_owner && *current_perm == BIT(QCOM_SCM_VMID_MSS_MSA))
>> +		return 0;
>> +	if (!remote_owner && *current_perm == BIT(QCOM_SCM_VMID_HLOS))
>> +		return 0;
>> +
>> +	next.vmid = remote_owner ? QCOM_SCM_VMID_MSS_MSA : QCOM_SCM_VMID_HLOS;
>> +	next.perm = remote_owner ? QCOM_SCM_PERM_RW : QCOM_SCM_PERM_RWX;
>> +
>> +	ret = qcom_scm_assign_mem(addr, ALIGN(size, SZ_4K),
>> +				  *current_perm, &next, 1);
>> +	if (ret < 0) {
>> +		pr_err("Failed to assign memory access in range %p to %p to %s ret = %d\n",
>> +		       (void *)addr, (void *)(addr + size),
>> +		       remote_owner ? "mss" : "hlos", ret);
> qcom_scm_assign_mem() also prints an error when this happens, there's no
> need for that. I'm happy with you dropping the print from
> qcom_scm_assign_mem() and keeping this one.
>
>> +		return ret;
>> +	}
>> +
>> +	*current_perm = ret;
>> +	return 0;
>> +}
>> +
>>   static int q6v5_load(struct rproc *rproc, const struct firmware *fw)
>>   {
>>   	struct q6v5 *qproc = rproc->priv;
>> @@ -450,6 +484,8 @@ static int q6v5_mpss_init_image(struct q6v5 *qproc, const struct firmware *fw)
>>   {
>>   	unsigned long dma_attrs = DMA_ATTR_FORCE_CONTIGUOUS;
>>   	dma_addr_t phys;
>> +	int mdata_perm;
>> +	int xferop_ret;
>>   	void *ptr;
>>   	int ret;
>>   
>> @@ -461,6 +497,12 @@ static int q6v5_mpss_init_image(struct q6v5 *qproc, const struct firmware *fw)
>>   
>>   	memcpy(ptr, fw->data, fw->size);
>>   
>> +	/* Hypervisor mapping to access metadata by modem */
>> +	mdata_perm = BIT(QCOM_SCM_VMID_HLOS);
>> +	ret = q6v5_xfer_mem_ownership(qproc, &mdata_perm,
>> +				      1, phys, fw->size);
> Please use "true" instead of "1".
OK Sure, will do it.
>> +	if (ret)
>> +		return -EAGAIN;
> Please keep a empty line here, to keep the "paragraphs" or "chunks" of
> code split up.
OK.
>
>>   	writel(phys, qproc->rmb_base + RMB_PMI_META_DATA_REG);
>>   	writel(RMB_CMD_META_DATA_READY, qproc->rmb_base + RMB_MBA_COMMAND_REG);
>>   
> Regards,
> Bjorn
> --
> To unsubscribe from this list: send the line "unsubscribe linux-remoteproc" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

-- 
Qualcomm India Private Limited, on behalf of Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project.

[PATCH v7 4/4] remoteproc: qcom: Add support for mss remoteproc on msm8996

[Avaneesh Kumar Dwivedi]

This patch add support for mss boot on msm8996. Major changes
include initializing mss rproc for msm8996, making appropriate
change for executing mss reset sequence etc.

Signed-off-by: Avaneesh Kumar Dwivedi <akdwived@codeaurora.org>
---
 .../devicetree/bindings/remoteproc/qcom,q6v5.txt   |   1 +
 drivers/remoteproc/qcom_q6v5_pil.c                 | 164 ++++++++++++++++++---
 2 files changed, 141 insertions(+), 24 deletions(-)

diff --git a/Documentation/devicetree/bindings/remoteproc/qcom,q6v5.txt b/Documentation/devicetree/bindings/remoteproc/qcom,q6v5.txt
index 92347fe..87a8e51 100644
--- a/Documentation/devicetree/bindings/remoteproc/qcom,q6v5.txt
+++ b/Documentation/devicetree/bindings/remoteproc/qcom,q6v5.txt
@@ -11,6 +11,7 @@ on the Qualcomm Hexagon core.
 		    "qcom,msm8916-mss-pil",
 		    "qcom,msm8974-mss-pil"
 
+		    "qcom,msm8996-mss-pil"
 - reg:
 	Usage: required
 	Value type: <prop-encoded-array>
diff --git a/drivers/remoteproc/qcom_q6v5_pil.c b/drivers/remoteproc/qcom_q6v5_pil.c
index a82b62a..808ee45 100644
--- a/drivers/remoteproc/qcom_q6v5_pil.c
+++ b/drivers/remoteproc/qcom_q6v5_pil.c
@@ -32,6 +32,7 @@
 #include <linux/soc/qcom/mdt_loader.h>
 #include <linux/soc/qcom/smem.h>
 #include <linux/soc/qcom/smem_state.h>
+#include <linux/iopoll.h>
 
 #include "remoteproc_internal.h"
 #include "qcom_common.h"
@@ -64,6 +65,8 @@
 #define QDSP6SS_RESET_REG		0x014
 #define QDSP6SS_GFMUX_CTL_REG		0x020
 #define QDSP6SS_PWR_CTL_REG		0x030
+#define QDSP6SS_MEM_PWR_CTL		0x0B0
+#define QDSP6SS_STRAP_ACC		0x110
 
 /* AXI Halt Register Offsets */
 #define AXI_HALTREQ_REG			0x0
@@ -92,6 +95,15 @@
 #define QDSS_BHS_ON			BIT(21)
 #define QDSS_LDO_BYP			BIT(22)
 
+/* QDSP6v56 parameters */
+#define QDSP6v56_LDO_BYP		BIT(25)
+#define QDSP6v56_BHS_ON		BIT(24)
+#define QDSP6v56_CLAMP_WL		BIT(21)
+#define QDSP6v56_CLAMP_QMC_MEM		BIT(22)
+#define HALT_CHECK_MAX_LOOPS		200
+#define QDSP6SS_XO_CBCR		0x0038
+#define QDSP6SS_ACC_OVERRIDE_VAL		0x20
+
 struct reg_info {
 	struct regulator *reg;
 	int uV;
@@ -110,6 +122,7 @@ struct rproc_hexagon_res {
 	struct qcom_mss_reg_res *active_supply;
 	char **proxy_clk_names;
 	char **active_clk_names;
+	int version;
 	bool need_mem_protection;
 };
 
@@ -157,6 +170,13 @@ struct q6v5 {
 	struct qcom_rproc_subdev smd_subdev;
 	int mpss_perm;
 	int mba_perm;
+	int version;
+};
+
+enum {
+	MSS_MSM8916,
+	MSS_MSM8974,
+	MSS_MSM8996,
 };
 
 static int q6v5_regulator_init(struct device *dev, struct reg_info *regs,
@@ -386,33 +406,98 @@ static int q6v5proc_reset(struct q6v5 *qproc)
 {
 	u32 val;
 	int ret;
+	int i;
 
-	/* Assert resets, stop core */
-	val = readl(qproc->reg_base + QDSP6SS_RESET_REG);
-	val |= (Q6SS_CORE_ARES | Q6SS_BUS_ARES_ENABLE | Q6SS_STOP_CORE);
-	writel(val, qproc->reg_base + QDSP6SS_RESET_REG);
 
-	/* Enable power block headswitch, and wait for it to stabilize */
-	val = readl(qproc->reg_base + QDSP6SS_PWR_CTL_REG);
-	val |= QDSS_BHS_ON | QDSS_LDO_BYP;
-	writel(val, qproc->reg_base + QDSP6SS_PWR_CTL_REG);
-	udelay(1);
-
-	/*
-	 * Turn on memories. L2 banks should be done individually
-	 * to minimize inrush current.
-	 */
-	val = readl(qproc->reg_base + QDSP6SS_PWR_CTL_REG);
-	val |= Q6SS_SLP_RET_N | Q6SS_L2TAG_SLP_NRET_N |
-		Q6SS_ETB_SLP_NRET_N | Q6SS_L2DATA_STBY_N;
-	writel(val, qproc->reg_base + QDSP6SS_PWR_CTL_REG);
-	val |= Q6SS_L2DATA_SLP_NRET_N_2;
-	writel(val, qproc->reg_base + QDSP6SS_PWR_CTL_REG);
-	val |= Q6SS_L2DATA_SLP_NRET_N_1;
-	writel(val, qproc->reg_base + QDSP6SS_PWR_CTL_REG);
-	val |= Q6SS_L2DATA_SLP_NRET_N_0;
-	writel(val, qproc->reg_base + QDSP6SS_PWR_CTL_REG);
+	if (qproc->version == MSS_MSM8996) {
+		/* Override the ACC value if required */
+		writel(QDSP6SS_ACC_OVERRIDE_VAL,
+		       qproc->reg_base + QDSP6SS_STRAP_ACC);
+
+		/* Assert resets, stop core */
+		val = readl(qproc->reg_base + QDSP6SS_RESET_REG);
+		val |= Q6SS_CORE_ARES | Q6SS_BUS_ARES_ENABLE | Q6SS_STOP_CORE;
+		writel(val, qproc->reg_base + QDSP6SS_RESET_REG);
+
+		/* BHS require xo cbcr to be enabled */
+		val = readl(qproc->reg_base + QDSP6SS_XO_CBCR);
+		val |= 0x1;
+		writel(val, qproc->reg_base + QDSP6SS_XO_CBCR);
 
+		/* Read CLKOFF bit to go low indicating CLK is enabled */
+		ret = readl_poll_timeout(qproc->reg_base + QDSP6SS_XO_CBCR,
+					 val, !(val & BIT(31)), 1,
+					 HALT_CHECK_MAX_LOOPS);
+		if (ret) {
+			dev_err(qproc->dev,
+				"xo cbcr enabling timed out (rc:%d)\n", ret);
+			return ret;
+		}
+		/* Enable power block headswitch and wait for it to stabilize */
+		val = readl(qproc->reg_base + QDSP6SS_PWR_CTL_REG);
+		val |= QDSP6v56_BHS_ON;
+		writel(val, qproc->reg_base + QDSP6SS_PWR_CTL_REG);
+		val |= readl(qproc->reg_base + QDSP6SS_PWR_CTL_REG);
+		udelay(1);
+
+		/* Put LDO in bypass mode */
+		val |= QDSP6v56_LDO_BYP;
+		writel(val, qproc->reg_base + QDSP6SS_PWR_CTL_REG);
+
+		/* Deassert QDSP6 compiler memory clamp */
+		val = readl(qproc->reg_base + QDSP6SS_PWR_CTL_REG);
+		val &= ~QDSP6v56_CLAMP_QMC_MEM;
+		writel(val, qproc->reg_base + QDSP6SS_PWR_CTL_REG);
+
+		/* Deassert memory peripheral sleep and L2 memory standby */
+		val |= Q6SS_L2DATA_STBY_N | Q6SS_SLP_RET_N;
+		writel(val, qproc->reg_base + QDSP6SS_PWR_CTL_REG);
+
+		/* Turn on L1, L2, ETB and JU memories 1 at a time */
+		val = readl(qproc->reg_base + QDSP6SS_MEM_PWR_CTL);
+		for (i = 19; i >= 0; i--) {
+			val |= BIT(i);
+			writel(val, qproc->reg_base +
+						QDSP6SS_MEM_PWR_CTL);
+			/*
+			 * Read back value to ensure the write is done then
+			 * wait for 1us for both memory peripheral and data
+			 * array to turn on.
+			 */
+			val |= readl(qproc->reg_base + QDSP6SS_MEM_PWR_CTL);
+			udelay(1);
+		}
+		/* Remove word line clamp */
+		val = readl(qproc->reg_base + QDSP6SS_PWR_CTL_REG);
+		val &= ~QDSP6v56_CLAMP_WL;
+		writel(val, qproc->reg_base + QDSP6SS_PWR_CTL_REG);
+	} else {
+		/* Assert resets, stop core */
+		val = readl(qproc->reg_base + QDSP6SS_RESET_REG);
+		val |= Q6SS_CORE_ARES | Q6SS_BUS_ARES_ENABLE | Q6SS_STOP_CORE;
+		writel(val, qproc->reg_base + QDSP6SS_RESET_REG);
+
+		/* Enable power block headswitch and wait for it to stabilize */
+		val = readl(qproc->reg_base + QDSP6SS_PWR_CTL_REG);
+		val |= QDSS_BHS_ON | QDSS_LDO_BYP;
+		writel(val, qproc->reg_base + QDSP6SS_PWR_CTL_REG);
+		val |= readl(qproc->reg_base + QDSP6SS_PWR_CTL_REG);
+		udelay(1);
+		/*
+		 * Turn on memories. L2 banks should be done individually
+		 * to minimize inrush current.
+		 */
+		val = readl(qproc->reg_base + QDSP6SS_PWR_CTL_REG);
+		val |= Q6SS_SLP_RET_N | Q6SS_L2TAG_SLP_NRET_N |
+			Q6SS_ETB_SLP_NRET_N | Q6SS_L2DATA_STBY_N;
+		writel(val, qproc->reg_base + QDSP6SS_PWR_CTL_REG);
+		val |= Q6SS_L2DATA_SLP_NRET_N_2;
+		writel(val, qproc->reg_base + QDSP6SS_PWR_CTL_REG);
+		val |= Q6SS_L2DATA_SLP_NRET_N_1;
+		writel(val, qproc->reg_base + QDSP6SS_PWR_CTL_REG);
+		val |= Q6SS_L2DATA_SLP_NRET_N_0;
+		writel(val, qproc->reg_base + QDSP6SS_PWR_CTL_REG);
+	}
 	/* Remove IO clamp */
 	val &= ~Q6SS_CLAMP_IO;
 	writel(val, qproc->reg_base + QDSP6SS_PWR_CTL_REG);
@@ -792,6 +877,15 @@ static int q6v5_stop(struct rproc *rproc)
 	q6v5proc_halt_axi_port(qproc, qproc->halt_map, qproc->halt_modem);
 	q6v5proc_halt_axi_port(qproc, qproc->halt_map, qproc->halt_nc);
 
+	if (qproc->version == MSS_MSM8996) {
+		/*
+		 * To avoid high MX current during LPASS/MSS restart.
+		 */
+		val = readl(qproc->reg_base + QDSP6SS_PWR_CTL_REG);
+		val |= Q6SS_CLAMP_IO | QDSP6v56_CLAMP_WL |
+			QDSP6v56_CLAMP_QMC_MEM;
+		writel(val, qproc->reg_base + QDSP6SS_PWR_CTL_REG);
+	}
 	ret = q6v5_xfer_mem_ownership(qproc, &qproc->mpss_perm, 0,
 				      qproc->mpss_phys, qproc->mpss_size);
 	WARN_ON(ret);
@@ -1092,6 +1186,7 @@ static int q6v5_probe(struct platform_device *pdev)
 	if (ret)
 		goto free_rproc;
 
+	qproc->version = desc->version;
 	qproc->need_mem_protection = desc->need_mem_protection;
 	ret = q6v5_request_irq(qproc, pdev, "wdog", q6v5_wdog_interrupt);
 	if (ret < 0)
@@ -1142,6 +1237,24 @@ static int q6v5_remove(struct platform_device *pdev)
 	return 0;
 }
 
+static const struct rproc_hexagon_res msm8996_mss = {
+	.hexagon_mba_image = "mba.mbn",
+	.proxy_clk_names = (char*[]){
+			"xo",
+			"pnoc",
+			NULL
+	},
+	.active_clk_names = (char*[]){
+			"iface",
+			"bus",
+			"mem",
+			"gpll0_mss_clk",
+			NULL
+	},
+	.need_mem_protection = true,
+	.version = MSS_MSM8996,
+};
+
 static const struct rproc_hexagon_res msm8916_mss = {
 	.hexagon_mba_image = "mba.mbn",
 	.proxy_supply = (struct qcom_mss_reg_res[]) {
@@ -1170,6 +1283,7 @@ static int q6v5_remove(struct platform_device *pdev)
 		NULL
 	},
 	.need_mem_protection = false,
+	.version = MSS_MSM8916,
 };
 
 static const struct rproc_hexagon_res msm8974_mss = {
@@ -1208,12 +1322,14 @@ static int q6v5_remove(struct platform_device *pdev)
 		NULL
 	},
 	.need_mem_protection = false,
+	.version = MSS_MSM8974,
 };
 
 static const struct of_device_id q6v5_of_match[] = {
 	{ .compatible = "qcom,q6v5-pil", .data = &msm8916_mss},
 	{ .compatible = "qcom,msm8916-mss-pil", .data = &msm8916_mss},
 	{ .compatible = "qcom,msm8974-mss-pil", .data = &msm8974_mss},
+	{ .compatible = "qcom,msm8996-mss-pil", .data = &msm8996_mss},
 	{ },
 };
 MODULE_DEVICE_TABLE(of, q6v5_of_match);
-- 
Qualcomm India Private Limited, on behalf of Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project.

Re: [PATCH v7 4/4] remoteproc: qcom: Add support for mss remoteproc on msm8996

[Bjorn Andersson]

On Fri 21 Jul 03:49 PDT 2017, Avaneesh Kumar Dwivedi wrote:

> This patch add support for mss boot on msm8996. Major changes
> include initializing mss rproc for msm8996, making appropriate
> change for executing mss reset sequence etc.
> 
> Signed-off-by: Avaneesh Kumar Dwivedi <akdwived@codeaurora.org>
> ---
>  .../devicetree/bindings/remoteproc/qcom,q6v5.txt   |   1 +
>  drivers/remoteproc/qcom_q6v5_pil.c                 | 164 ++++++++++++++++++---
>  2 files changed, 141 insertions(+), 24 deletions(-)
> 
> diff --git a/Documentation/devicetree/bindings/remoteproc/qcom,q6v5.txt b/Documentation/devicetree/bindings/remoteproc/qcom,q6v5.txt
> index 92347fe..87a8e51 100644
> --- a/Documentation/devicetree/bindings/remoteproc/qcom,q6v5.txt
> +++ b/Documentation/devicetree/bindings/remoteproc/qcom,q6v5.txt
> @@ -11,6 +11,7 @@ on the Qualcomm Hexagon core.
>  		    "qcom,msm8916-mss-pil",
>  		    "qcom,msm8974-mss-pil"
>  
> +		    "qcom,msm8996-mss-pil"

I like the empty line, but it should be after the list, not in the
middle.

>  - reg:

Please fix this and add my Acked-by

Regards,
Bjorn

Re: [PATCH v7 4/4] remoteproc: qcom: Add support for mss remoteproc on msm8996

[Dwivedi, Avaneesh Kumar (avani)]

On 10/12/2017 11:50 PM, Bjorn Andersson wrote:
> On Fri 21 Jul 03:49 PDT 2017, Avaneesh Kumar Dwivedi wrote:
>
>> This patch add support for mss boot on msm8996. Major changes
>> include initializing mss rproc for msm8996, making appropriate
>> change for executing mss reset sequence etc.
>>
>> Signed-off-by: Avaneesh Kumar Dwivedi <akdwived@codeaurora.org>
>> ---
>>   .../devicetree/bindings/remoteproc/qcom,q6v5.txt   |   1 +
>>   drivers/remoteproc/qcom_q6v5_pil.c                 | 164 ++++++++++++++++++---
>>   2 files changed, 141 insertions(+), 24 deletions(-)
>>
>> diff --git a/Documentation/devicetree/bindings/remoteproc/qcom,q6v5.txt b/Documentation/devicetree/bindings/remoteproc/qcom,q6v5.txt
>> index 92347fe..87a8e51 100644
>> --- a/Documentation/devicetree/bindings/remoteproc/qcom,q6v5.txt
>> +++ b/Documentation/devicetree/bindings/remoteproc/qcom,q6v5.txt
>> @@ -11,6 +11,7 @@ on the Qualcomm Hexagon core.
>>   		    "qcom,msm8916-mss-pil",
>>   		    "qcom,msm8974-mss-pil"
>>   
>> +		    "qcom,msm8996-mss-pil"
> I like the empty line, but it should be after the list, not in the
> middle.
>
>>   - reg:
> Please fix this and add my Acked-by
Sure will do, just want to ask, when i am sending updated patches, 
should i sent all the 4 patches in series or should skip those which are 
acked by you.
>
> Regards,
> Bjorn
> --
> To unsubscribe from this list: send the line "unsubscribe linux-arm-msm" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

-- 
Qualcomm India Private Limited, on behalf of Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project.

Re: [PATCH v7 4/4] remoteproc: qcom: Add support for mss remoteproc on msm8996

[Bjorn Andersson]

On Mon 16 Oct 06:19 PDT 2017, Dwivedi, Avaneesh Kumar (avani) wrote:
> On 10/12/2017 11:50 PM, Bjorn Andersson wrote:
[..]
> > Please fix this and add my Acked-by
> Sure will do, just want to ask, when i am sending updated patches, should i
> sent all the 4 patches in series or should skip those which are acked by
> you.

Please include all the patches in each submission (unless the maintainer
tells you that he/she applied some of them already).

Make sure to add any "Acked-by", "Reviewed-by" and "Tested-by" on your
patches, so that people know that they are already reviewed/tested.


And finally, when resending patches, add a section between "---" and the
diff-stat describing changes since the last version - that makes it
easier to see what changed since last time and will help the reviewer
look at the "new" things.

Thanks,
Bjorn

Re: [PATCH v7 4/4] remoteproc: qcom: Add support for mss remoteproc on msm8996

[Sricharan R]

Hi Bjorn,

On 10/17/2017 12:28 AM, Bjorn Andersson wrote:
> On Mon 16 Oct 06:19 PDT 2017, Dwivedi, Avaneesh Kumar (avani) wrote:
>> On 10/12/2017 11:50 PM, Bjorn Andersson wrote:
> [..]
>>> Please fix this and add my Acked-by
>> Sure will do, just want to ask, when i am sending updated patches, should i
>> sent all the 4 patches in series or should skip those which are acked by
>> you.
> 
> Please include all the patches in each submission (unless the maintainer
> tells you that he/she applied some of them already).
> 
> Make sure to add any "Acked-by", "Reviewed-by" and "Tested-by" on your
> patches, so that people know that they are already reviewed/tested.
> 
> 
> And finally, when resending patches, add a section between "---" and the
> diff-stat describing changes since the last version - that makes it
> easier to see what changed since last time and will help the reviewer
> look at the "new" things.

 So i was having the ipq8074 q6 remoteproc support based on this series last time.
 With the msm8996 now getting cleared, i will rebase my patches and repost.
 Otherwise, just hope that you were ok with the approach [1].

[1] https://www.mail-archive.com/linux-kernel@vger.kernel.org/msg1481961.html


Regards,
 Sricharan

-- 
"QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by The Linux Foundation

---
This email has been checked for viruses by Avast antivirus software.
https://www.avast.com/antivirus