This patch fixes SMEM addressing issues when remote processors need to use
secure SMEM partitions.
Signed-off-by: Andy Gross <[email protected]>
---
.../devicetree/bindings/soc/qcom/qcom,smd.txt | 7 +++++++
drivers/soc/qcom/smd.c | 18 ++++++++++++++----
2 files changed, 21 insertions(+), 4 deletions(-)
diff --git a/Documentation/devicetree/bindings/soc/qcom/qcom,smd.txt b/Documentation/devicetree/bindings/soc/qcom/qcom,smd.txt
index f65c76d..3b60702 100644
--- a/Documentation/devicetree/bindings/soc/qcom/qcom,smd.txt
+++ b/Documentation/devicetree/bindings/soc/qcom/qcom,smd.txt
@@ -37,6 +37,12 @@ The edge is described by the following properties:
Definition: the identifier of the remote processor in the smd channel
allocation table
+- qcom,remote-pid:
+ Usage: required
+ Value type: <u32>
+ Definition: the identifier for the remote processor as known by the rest
+ of the system.
+
= SMD DEVICES
In turn, subnodes of the "edges" represent devices tied to SMD channels on that
@@ -68,6 +74,7 @@ The following example represents a smd node, with one edge representing the
interrupts = <0 168 1>;
qcom,ipc = <&apcs 8 0>;
qcom,smd-edge = <15>;
+ qcom,remote-pid = <0xffffffff>;
rpm_requests {
compatible = "qcom,rpm-msm8974";
diff --git a/drivers/soc/qcom/smd.c b/drivers/soc/qcom/smd.c
index 327adcf..bca8aba 100644
--- a/drivers/soc/qcom/smd.c
+++ b/drivers/soc/qcom/smd.c
@@ -111,6 +111,7 @@ struct qcom_smd_edge {
struct qcom_smd *smd;
struct device_node *of_node;
unsigned edge_id;
+ unsigned remote_pid;
int irq;
@@ -572,7 +573,7 @@ static irqreturn_t qcom_smd_edge_intr(int irq, void *data)
* have to scan if the amount of available space in smem have changed
* since last scan.
*/
- available = qcom_smem_get_free_space(edge->edge_id);
+ available = qcom_smem_get_free_space(edge->remote_pid);
if (available != edge->smem_available) {
edge->smem_available = available;
edge->need_rescan = true;
@@ -976,7 +977,8 @@ static struct qcom_smd_channel *qcom_smd_create_channel(struct qcom_smd_edge *ed
spin_lock_init(&channel->recv_lock);
init_waitqueue_head(&channel->fblockread_event);
- ret = qcom_smem_get(edge->edge_id, smem_info_item, (void **)&info, &info_size);
+ ret = qcom_smem_get(edge->remote_pid, smem_info_item, (void **)&info,
+ &info_size);
if (ret)
goto free_name_and_channel;
@@ -997,7 +999,8 @@ static struct qcom_smd_channel *qcom_smd_create_channel(struct qcom_smd_edge *ed
goto free_name_and_channel;
}
- ret = qcom_smem_get(edge->edge_id, smem_fifo_item, &fifo_base, &fifo_size);
+ ret = qcom_smem_get(edge->remote_pid, smem_fifo_item, &fifo_base,
+ &fifo_size);
if (ret)
goto free_name_and_channel;
@@ -1041,7 +1044,7 @@ static void qcom_discover_channels(struct qcom_smd_edge *edge)
int i;
for (tbl = 0; tbl < SMD_ALLOC_TBL_COUNT; tbl++) {
- ret = qcom_smem_get(edge->edge_id,
+ ret = qcom_smem_get(edge->remote_pid,
smem_items[tbl].alloc_tbl_id,
(void **)&alloc_tbl,
NULL);
@@ -1184,6 +1187,13 @@ static int qcom_smd_parse_edge(struct device *dev,
return -EINVAL;
}
+ key = "qcom,remote-pid";
+ ret = of_property_read_u32(node, key, &edge->remote_pid);
+ if (ret) {
+ dev_err(dev, "edge missing %s property\n", key);
+ return -EINVAL;
+ }
+
syscon_np = of_parse_phandle(node, "qcom,ipc", 0);
if (!syscon_np) {
dev_err(dev, "no qcom,ipc node\n");
--
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
hosted by The Linux Foundation
On Tue 25 Aug 13:36 PDT 2015, Andy Gross wrote:
> This patch fixes SMEM addressing issues when remote processors need to use
> secure SMEM partitions.
>
Right, sorry for missing that remote_pid and edge_pid isn't in sync...
> Signed-off-by: Andy Gross <[email protected]>
> ---
> .../devicetree/bindings/soc/qcom/qcom,smd.txt | 7 +++++++
> drivers/soc/qcom/smd.c | 18 ++++++++++++++----
> 2 files changed, 21 insertions(+), 4 deletions(-)
>
> diff --git a/Documentation/devicetree/bindings/soc/qcom/qcom,smd.txt b/Documentation/devicetree/bindings/soc/qcom/qcom,smd.txt
> index f65c76d..3b60702 100644
> --- a/Documentation/devicetree/bindings/soc/qcom/qcom,smd.txt
> +++ b/Documentation/devicetree/bindings/soc/qcom/qcom,smd.txt
> @@ -37,6 +37,12 @@ The edge is described by the following properties:
> Definition: the identifier of the remote processor in the smd channel
> allocation table
>
> +- qcom,remote-pid:
> + Usage: required
I would like to see this being optional, as this is not a property that
matters on any of the 32-bit systems (perhaps 8084?).
> + Value type: <u32>
> + Definition: the identifier for the remote processor as known by the rest
> + of the system.
> +
> = SMD DEVICES
>
> In turn, subnodes of the "edges" represent devices tied to SMD channels on that
> @@ -68,6 +74,7 @@ The following example represents a smd node, with one edge representing the
> interrupts = <0 168 1>;
> qcom,ipc = <&apcs 8 0>;
> qcom,smd-edge = <15>;
> + qcom,remote-pid = <0xffffffff>;
This looks messy, so let's make the property optional and make its
absence indicate the "global partition".
>
> rpm_requests {
> compatible = "qcom,rpm-msm8974";
> diff --git a/drivers/soc/qcom/smd.c b/drivers/soc/qcom/smd.c
[..]
> @@ -1184,6 +1187,13 @@ static int qcom_smd_parse_edge(struct device *dev,
> return -EINVAL;
> }
>
> + key = "qcom,remote-pid";
> + ret = of_property_read_u32(node, key, &edge->remote_pid);
> + if (ret) {
> + dev_err(dev, "edge missing %s property\n", key);
> + return -EINVAL;
> + }
> +
So I suggest you changing this to:
edge->remote_pid = QCOM_SMEM_HOST_ANY;
of_property_read_u32(node, "qcom,remote-pid", &edge->remote_pid);
> syscon_np = of_parse_phandle(node, "qcom,ipc", 0);
> if (!syscon_np) {
> dev_err(dev, "no qcom,ipc node\n");
Regards,
Bjorn
On Tue, Aug 25, 2015 at 08:46:22PM -0700, Bjorn Andersson wrote:
> On Tue 25 Aug 13:36 PDT 2015, Andy Gross wrote:
>
> > This patch fixes SMEM addressing issues when remote processors need to use
> > secure SMEM partitions.
> >
>
> Right, sorry for missing that remote_pid and edge_pid isn't in sync...
>
> > Signed-off-by: Andy Gross <[email protected]>
> > ---
> > .../devicetree/bindings/soc/qcom/qcom,smd.txt | 7 +++++++
> > drivers/soc/qcom/smd.c | 18 ++++++++++++++----
> > 2 files changed, 21 insertions(+), 4 deletions(-)
> >
> > diff --git a/Documentation/devicetree/bindings/soc/qcom/qcom,smd.txt b/Documentation/devicetree/bindings/soc/qcom/qcom,smd.txt
> > index f65c76d..3b60702 100644
> > --- a/Documentation/devicetree/bindings/soc/qcom/qcom,smd.txt
> > +++ b/Documentation/devicetree/bindings/soc/qcom/qcom,smd.txt
> > @@ -37,6 +37,12 @@ The edge is described by the following properties:
> > Definition: the identifier of the remote processor in the smd channel
> > allocation table
> >
> > +- qcom,remote-pid:
> > + Usage: required
>
> I would like to see this being optional, as this is not a property that
> matters on any of the 32-bit systems (perhaps 8084?).
I thought about that myself, and yeah I have to agree.
>
> > + Value type: <u32>
> > + Definition: the identifier for the remote processor as known by the rest
> > + of the system.
> > +
> > = SMD DEVICES
> >
> > In turn, subnodes of the "edges" represent devices tied to SMD channels on that
> > @@ -68,6 +74,7 @@ The following example represents a smd node, with one edge representing the
> > interrupts = <0 168 1>;
> > qcom,ipc = <&apcs 8 0>;
> > qcom,smd-edge = <15>;
> > + qcom,remote-pid = <0xffffffff>;
>
> This looks messy, so let's make the property optional and make its
> absence indicate the "global partition".
>
> >
> > rpm_requests {
> > compatible = "qcom,rpm-msm8974";
> > diff --git a/drivers/soc/qcom/smd.c b/drivers/soc/qcom/smd.c
> [..]
> > @@ -1184,6 +1187,13 @@ static int qcom_smd_parse_edge(struct device *dev,
> > return -EINVAL;
> > }
> >
> > + key = "qcom,remote-pid";
> > + ret = of_property_read_u32(node, key, &edge->remote_pid);
> > + if (ret) {
> > + dev_err(dev, "edge missing %s property\n", key);
> > + return -EINVAL;
> > + }
> > +
>
> So I suggest you changing this to:
>
> edge->remote_pid = QCOM_SMEM_HOST_ANY;
> of_property_read_u32(node, "qcom,remote-pid", &edge->remote_pid);
Agreed.
>
> > syscon_np = of_parse_phandle(node, "qcom,ipc", 0);
> > if (!syscon_np) {
> > dev_err(dev, "no qcom,ipc node\n");
>
> Regards,
> Bjorn
--
Qualcomm Innovation Center, Inc.
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project