Le 12/08/2022 à 14:47, Srinivasa Rao Mandadapu a écrit :
Update pil driver with SMMU mapping for allowing authorised
memory access to ADSP firmware, by reading required memory
regions either from device tree file or from resource table
embedded in ADSP binary header.
Signed-off-by: Srinivasa Rao Mandadapu <quic_srivasam-jfJNa2p1gH1BDgjK7y7TUQ@xxxxxxxxxxxxxxxx>
---
Changes since V3:
-- Rename is_adsp_sb_needed to adsp_sandbox_needed.
-- Add smmu unmapping in error case and in adsp stop.
Changes since V2:
-- Replace platform_bus_type with adsp->dev->bus.
-- Use API of_parse_phandle_with_args() instead of of_parse_phandle_with_fixed_args().
-- Replace adsp->is_wpss with adsp->is_adsp.
-- Update error handling in adsp_start().
drivers/remoteproc/qcom_q6v5_adsp.c | 172 +++++++++++++++++++++++++++++++++++-
1 file changed, 170 insertions(+), 2 deletions(-)
diff --git a/drivers/remoteproc/qcom_q6v5_adsp.c b/drivers/remoteproc/qcom_q6v5_adsp.c
index b0a63a0..ca45d2c 100644
--- a/drivers/remoteproc/qcom_q6v5_adsp.c
+++ b/drivers/remoteproc/qcom_q6v5_adsp.c
@@ -9,6 +9,7 @@
#include <linux/firmware.h>
#include <linux/interrupt.h>
#include <linux/io.h>
+#include <linux/iommu.h>
#include <linux/iopoll.h>
#include <linux/kernel.h>
#include <linux/mfd/syscon.h>
@@ -48,6 +49,8 @@
#define LPASS_PWR_ON_REG 0x10
#define LPASS_HALTREQ_REG 0x0
+#define SID_MASK_DEFAULT 0xF
+
#define QDSP6SS_XO_CBCR 0x38
#define QDSP6SS_CORE_CBCR 0x20
#define QDSP6SS_SLEEP_CBCR 0x3c
@@ -78,7 +81,7 @@ struct adsp_pil_data {
struct qcom_adsp {
struct device *dev;
struct rproc *rproc;
-
+ struct iommu_domain *iommu_dom;
struct qcom_q6v5 q6v5;
struct clk *xo;
@@ -333,6 +336,155 @@ static int adsp_load(struct rproc *rproc, const struct firmware *fw)
return 0;
}
+static void adsp_of_unmap_smmu(struct iommu_domain *iommu_dom, const __be32 *prop, int len)
+{
+ unsigned long mem_phys;
+ unsigned long iova;
+ unsigned int mem_size;
+ int access_level;
+ int i;
+
+ for (i = 0; i < len; i++) {
+ iova = be32_to_cpu(prop[i++]);
+ mem_phys = be32_to_cpu(prop[i++]);
+ mem_size = be32_to_cpu(prop[i++]);
+ access_level = be32_to_cpu(prop[i]);
+ iommu_unmap(iommu_dom, iova, mem_size);
+ }
+}
+
+static void adsp_rproc_unmap_smmu(struct rproc *rproc, int len)
+{
+ struct fw_rsc_devmem *rsc_fw;
+ struct fw_rsc_hdr *hdr;
+ int offset;
+ int i;
+
+ for (i = 0; i < len; i++) {
+ offset = rproc->table_ptr->offset[i];
+ hdr = (void *)rproc->table_ptr + offset;
+ rsc_fw = (struct fw_rsc_devmem *)hdr + sizeof(*hdr);
+
+ iommu_unmap(rproc->domain, rsc_fw->da, rsc_fw->len);
+ }
+}
+
+static void adsp_unmap_smmu(struct rproc *rproc)
+{
When I proposed a adsp_unmap_smmu() function, the idea was to undo
everything that is donne by adsp_map_smmu().
iommu_domain_alloc() and iommu_map(adsp->iommu_dom, ..) are not undone here.
If this make sense, it would improve the semantic, simplify the
'adsp_smmu_unmap' label in adsp_start() and avoid what looks like a leak
to me in adsp_stop().
+ struct qcom_adsp *adsp = (struct qcom_adsp *)rproc->priv;
+ const __be32 *prop;
+ unsigned int len;
+
+ prop = of_get_property(adsp->dev->of_node, "qcom,adsp-memory-regions", &len);
+ if (prop) {
In the allocation path, you have a "len /= sizeof(__be32);" which is not
here. Is it needed?
You call adsp_unmap_smmu() from the error handling path of
adsp_map_smmu(). If needed, maybe it should be part of adsp_of_unmap_smmu()?
+ adsp_of_unmap_smmu(adsp->iommu_dom, prop, len);
+ } else {
+ if (rproc->table_ptr)
+ adsp_rproc_unmap_smmu(rproc, rproc->table_ptr->num);
+ }
+}
+
+static int adsp_map_smmu(struct qcom_adsp *adsp, struct rproc *rproc)
+{
+ struct of_phandle_args args;
+ struct fw_rsc_devmem *rsc_fw;
+ struct fw_rsc_hdr *hdr;
+ const __be32 *prop;
+ long long sid;
+ unsigned long mem_phys;
+ unsigned long iova;
+ unsigned int mem_size;
+ unsigned int flag;
+ unsigned int len;
+ int access_level;
+ int offset;
+ int ret;
+ int rc;
Are ret and rc both needed?
+ int i;
+
+ rc = of_parse_phandle_with_args(adsp->dev->of_node, "iommus", "#iommu-cells", 0, &args);
+ if (rc < 0)
+ sid = -1;
+ else
+ sid = args.args[0] & SID_MASK_DEFAULT;
+
+ adsp->iommu_dom = iommu_domain_alloc(adsp->dev->bus);
+ if (!adsp->iommu_dom) {
+ dev_err(adsp->dev, "failed to allocate iommu domain\n");
+ ret = -ENOMEM;
+ goto domain_free;
+ }
+
+ ret = iommu_attach_device(adsp->iommu_dom, adsp->dev);
+ if (ret) {
+ dev_err(adsp->dev, "could not attach device ret = %d\n", ret);
+ ret = -EBUSY;
+ goto detach_device;
+ }
+
+ /* Add SID configuration for ADSP Firmware to SMMU */
+ adsp->mem_phys = adsp->mem_phys | (sid << 32);
+
+ ret = iommu_map(adsp->iommu_dom, adsp->mem_phys, adsp->mem_phys,
+ adsp->mem_size, IOMMU_READ | IOMMU_WRITE);
+ if (ret) {
+ dev_err(adsp->dev, "Unable to map ADSP Physical Memory\n");
+ goto sid_unmap;
+ }
+
+ prop = of_get_property(adsp->dev->of_node, "qcom,adsp-memory-regions", &len);
+ if (prop) {
+ len /= sizeof(__be32);
+ for (i = 0; i < len; i++) {
+ iova = be32_to_cpu(prop[i++]);
+ mem_phys = be32_to_cpu(prop[i++]);
+ mem_size = be32_to_cpu(prop[i++]);
+ access_level = be32_to_cpu(prop[i]);
+
+ if (access_level)
+ flag = IOMMU_READ | IOMMU_WRITE;
+ else
+ flag = IOMMU_READ;
+
+ ret = iommu_map(adsp->iommu_dom, iova, mem_phys, mem_size, flag);
+ if (ret) {
+ dev_err(adsp->dev, "failed to map addr = %p mem_size = %x\n",
+ &(mem_phys), mem_size);
+ goto smmu_unmap;
+ }
+ }
+ } else {
+ if (!rproc->table_ptr)
+ goto sid_unmap;
+
+ for (i = 0; i < rproc->table_ptr->num; i++) {
+ offset = rproc->table_ptr->offset[i];
+ hdr = (void *)rproc->table_ptr + offset;
+ rsc_fw = (struct fw_rsc_devmem *)hdr + sizeof(*hdr);
+
+ ret = iommu_map(rproc->domain, rsc_fw->da, rsc_fw->pa,
+ rsc_fw->len, rsc_fw->flags);
+ if (ret) {
+ pr_err("%s; unable to map adsp memory address\n", __func__);
+ goto rproc_smmu_unmap;
+ }
+ }
+ }
If you introduce a adsp_of_unmap_smmu() and adsp_rproc_unmap_smmu(),
would it make things more readable to have the same kind of functions
when allocating the resources?
Symmetry often helps.
+ return 0;
Add an empty new line here?
+rproc_smmu_unmap:
+ adsp_rproc_unmap_smmu(rproc, i);
+ goto sid_unmap;
+smmu_unmap:
+ adsp_of_unmap_smmu(adsp->iommu_dom, prop, i);
+sid_unmap:
+ iommu_unmap(adsp->iommu_dom, adsp->mem_phys, adsp->mem_size);
+detach_device:
+ iommu_domain_free(adsp->iommu_dom);
+domain_free:
+ return ret;
+}
+
+
static int adsp_start(struct rproc *rproc)
{
struct qcom_adsp *adsp = (struct qcom_adsp *)rproc->priv;
@@ -343,9 +495,16 @@ static int adsp_start(struct rproc *rproc)
if (ret)
return ret;
+ if (adsp->adsp_sandbox_needed) {
+ ret = adsp_map_smmu(adsp, rproc);
+ if (ret) {
+ dev_err(adsp->dev, "ADSP smmu mapping failed\n");
+ goto disable_irqs;
+ }
+ }
ret = clk_prepare_enable(adsp->xo);
if (ret)
- goto disable_irqs;
+ goto adsp_smmu_unmap;
ret = qcom_rproc_pds_enable(adsp, adsp->proxy_pds,
adsp->proxy_pd_count);
@@ -401,6 +560,12 @@ static int adsp_start(struct rproc *rproc)
qcom_rproc_pds_disable(adsp, adsp->proxy_pds, adsp->proxy_pd_count);
disable_xo_clk:
clk_disable_unprepare(adsp->xo);
+adsp_smmu_unmap:
+ if (adsp->adsp_sandbox_needed) {
+ iommu_unmap(adsp->iommu_dom, adsp->mem_phys, adsp->mem_size);
+ adsp_unmap_smmu(rproc);
+ iommu_domain_free(adsp->iommu_dom);
+ }
disable_irqs:
qcom_q6v5_unprepare(&adsp->q6v5);
@@ -429,6 +594,9 @@ static int adsp_stop(struct rproc *rproc)
if (ret)
dev_err(adsp->dev, "failed to shutdown: %d\n", ret);
+ if (adsp->adsp_sandbox_needed)
+ adsp_unmap_smmu(rproc);
No need to call iommu_unmap() and iommu_domain_free() here?
(this is the same comment as the one in adsp_rproc_unmap_smmu(). This is
just a blind guess based on symmetry of the code.)
+
handover = qcom_q6v5_unprepare(&adsp->q6v5);
if (handover)
qcom_adsp_pil_handover(&adsp->q6v5);
