On 6/27/2023 6:09 PM, Manikanta Mylavarapu wrote:
On 6/24/2023 12:58 PM, Krzysztof Kozlowski wrote:
On 22/05/2023 00:28, Manikanta Mylavarapu wrote:
It adds support to bring up remoteproc's on multipd model.
Pd means protection domain. It's similar to process in Linux.
Here QDSP6 processor runs each wifi radio functionality on a
separate process. One process can't access other process
resources, so this is termed as PD i.e protection domain.
Here we have two pd's called root and user pd. We can correlate
Root pd as root and user pd as user in linux. Root pd has more
privileges than user pd. Root will provide services to user pd.
From remoteproc driver perspective, root pd corresponds to QDSP6
processor bring up and user pd corresponds to Wifi radio (WCSS)
bring up.
Here WCSS(user) PD is dependent on Q6(root) PD, so first
q6 pd should be up before wcss pd. After wcss pd goes down,
q6 pd should be turned off.
IPQ5018, IPQ9574 supports multipd remoteproc driver.
Signed-off-by: Manikanta Mylavarapu <quic_mmanikan@xxxxxxxxxxx>
---
Changes in V2:
- Common functionalities moved to seperate patches
- qcom_get_pd_asid() moved to mpd driver
- Last DMA block alone memset to zero
- Added diagram to show how userpd data is organized and sent to
trustzone
- Rewritten commit message since most of the content available
in cover page
- Removed 'remote_id' becuase it's not required for bring up.
drivers/remoteproc/Kconfig | 20 +
drivers/remoteproc/Makefile | 1 +
drivers/remoteproc/qcom_common.h | 9 +
drivers/remoteproc/qcom_q6v5_mpd.c | 677 ++++++++++++++++++++++++++++
drivers/soc/qcom/mdt_loader.c | 332 ++++++++++++++
include/linux/soc/qcom/mdt_loader.h | 19 +
6 files changed, 1058 insertions(+)
create mode 100644 drivers/remoteproc/qcom_q6v5_mpd.c
diff --git a/drivers/remoteproc/Kconfig b/drivers/remoteproc/Kconfig
index a850e9f486dd..44af5c36f67e 100644
--- a/drivers/remoteproc/Kconfig
+++ b/drivers/remoteproc/Kconfig
@@ -234,6 +234,26 @@ config QCOM_Q6V5_PAS
CDSP (Compute DSP), MPSS (Modem Peripheral SubSystem), and
SLPI (Sensor Low Power Island).
+config QCOM_Q6V5_MPD
+ tristate "Qualcomm Hexagon based MPD model Peripheral Image Loader"
+ depends on OF && ARCH_QCOM
+ depends on QCOM_SMEM
+ depends on RPMSG_QCOM_SMD || RPMSG_QCOM_SMD=n
+ depends on RPMSG_QCOM_GLINK_SMEM || RPMSG_QCOM_GLINK_SMEM=n
+ depends on QCOM_SYSMON || QCOM_SYSMON=n
+ depends on RPMSG_QCOM_GLINK || RPMSG_QCOM_GLINK=n
+ depends on QCOM_AOSS_QMP || QCOM_AOSS_QMP=n
+ select MFD_SYSCON
Do you really need this?
It's not required. I will remove. Thanks for catching.
+ select QCOM_MDT_LOADER
+ select QCOM_PIL_INFO
+ select QCOM_Q6V5_COMMON
+ select QCOM_RPROC_COMMON
+ select QCOM_SCM
+ help
+ Say y here to support the Qualcomm Secure Peripheral Image Loader
+ for the Hexagon based MultiPD model remote processors on e.g.
IPQ5018.
+ This is trustZone wireless subsystem.
+
...
I didn't understand. Can you please elaborate your comment?
+ int (*powerup_scm)(u32 peripheral);
+ int (*powerdown_scm)(u32 peripheral);
+};
+
+/**
+ * qcom_get_pd_asid() - get the pd asid number from DT node
From node name? NAK. It does not work like that. Node names can change
and you did not define this number as part of ABI.
Probably you wanted unit address.
Yeah i got your point. Each of the WCSS PD's are internally
represented in Q6 with their corresponding "spawn" bit numbers.
I will use same and remove the "PD-" hardcodings.
Is this fine ?
+ * @node: device tree node
+ *
+ * Returns asid if node name has 'pd' string
+ */
+s8 qcom_get_pd_asid(struct device_node *node)
+{
+ char *str;
+ s8 pd_asid;
+
+ if (!node)
+ return -EINVAL;
+
+ str = strstr(node->name, "pd");
+ if (!str)
+ return 0;
+
+ str += strlen("pd") + 1;
+ return kstrtos8(str, 10, &pd_asid) ? -EINVAL : pd_asid;
+}
+EXPORT_SYMBOL(qcom_get_pd_asid);
+
...
I didn't understand. Can you please elaborate your comment?
+
+static int q6_wcss_spawn_pd(struct rproc *rproc)
+{
+ int ret;
+ struct q6_wcss *wcss = rproc->priv;
+
+ ret = qcom_q6v5_request_spawn(&wcss->q6);
+ if (ret == -ETIMEDOUT) {
+ pr_err("%s spawn timedout\n", rproc->name);
dev_err
Sure, I will change to dev_err.
+ return ret;
+ }
+
+ ret = qcom_q6v5_wait_for_start(&wcss->q6, msecs_to_jiffies(10000));
+ if (ret == -ETIMEDOUT) {
+ pr_err("%s start timedout\n", rproc->name);
dev_err
Sure, I will change to dev_err.
+ wcss->q6.running = false;
+ return ret;
+ }
+ wcss->q6.running = true;
+ return ret;
+}
+
+static int wcss_ahb_pcie_pd_start(struct rproc *rproc)
+{
+ struct q6_wcss *wcss = rproc->priv;
+ const struct wcss_data *desc = wcss->desc;
+ int ret;
+
+ if (!desc->reset_seq)
+ return 0;
+
+ if (desc->powerup_scm) {
+ ret = desc->powerup_scm(desc->pasid);
+ if (ret) {
+ dev_err(wcss->dev, "failed to power up pd\n");
+ return ret;
+ }
+ }
+
+ ret = q6_wcss_spawn_pd(rproc);
+ if (ret)
+ return ret;
+
+ wcss->state = WCSS_NORMAL;
+ return ret;
+}
+
+static int q6_wcss_stop(struct rproc *rproc)
+{
+ struct q6_wcss *wcss = rproc->priv;
+ const struct wcss_data *desc = wcss->desc;
+ int ret;
+
+ ret = qcom_scm_pas_shutdown(desc->pasid);
+ if (ret) {
+ dev_err(wcss->dev, "not able to shutdown\n");
+ return ret;
+ }
+ qcom_q6v5_unprepare(&wcss->q6);
+
+ return 0;
+}
+
+static int wcss_ahb_pcie_pd_stop(struct rproc *rproc)
+{
+ struct q6_wcss *wcss = rproc->priv;
+ struct rproc *rpd_rproc = dev_get_drvdata(wcss->dev->parent);
+ const struct wcss_data *desc = wcss->desc;
+ int ret;
+
+ if (!desc->reset_seq)
+ goto shut_down_rpd;
+
+ if (rproc->state != RPROC_CRASHED && wcss->q6.stop_bit) {
+ ret = qcom_q6v5_request_stop(&wcss->q6, NULL);
+ if (ret) {
+ dev_err(&rproc->dev, "pd not stopped\n");
+ return ret;
+ }
+ }
+
+ if (desc->powerdown_scm) {
+ ret = desc->powerdown_scm(desc->pasid);
+ if (ret) {
+ dev_err(wcss->dev, "failed to power down pd\n");
+ return ret;
+ }
+ }
+
+shut_down_rpd:
+ rproc_shutdown(rpd_rproc);
+
+ wcss->state = WCSS_SHUTDOWN;
+ return 0;
+}
+
+static void *q6_wcss_da_to_va(struct rproc *rproc, u64 da, size_t len,
+ bool *is_iomem)
+{
+ struct q6_wcss *wcss = rproc->priv;
+ int offset;
+
+ offset = da - wcss->mem_reloc;
+ if (offset < 0 || offset + len > wcss->mem_size)
+ return NULL;
+
+ return wcss->mem_region + offset;
+}
+
+static int q6_wcss_load(struct rproc *rproc, const struct firmware *fw)
+{
+ struct q6_wcss *wcss = rproc->priv;
+ const struct firmware *fw_hdl;
+ int ret;
+ const struct wcss_data *desc = wcss->desc;
+ int loop;
+
+ ret = qcom_mdt_load(wcss->dev, fw, rproc->firmware,
+ desc->pasid, wcss->mem_region,
+ wcss->mem_phys, wcss->mem_size,
+ &wcss->mem_reloc);
+ if (ret)
+ return ret;
+
+ for (loop = 1; loop < MAX_FIRMWARE; loop++) {
+ if (!wcss->firmware[loop])
+ continue;
+
+ ret = request_firmware(&fw_hdl, wcss->firmware[loop],
+ wcss->dev);
+ if (ret)
+ continue;
+
+ ret = qcom_mdt_load_no_init(wcss->dev, fw_hdl,
+ wcss->firmware[loop], 0,
+ wcss->mem_region,
+ wcss->mem_phys,
+ wcss->mem_size,
+ &wcss->mem_reloc);
+
+ release_firmware(fw_hdl);
+
+ if (ret) {
+ dev_err(wcss->dev,
+ "can't load %s ret:%d\n", wcss->firmware[loop], ret);
+ return ret;
+ }
+ }
+ return 0;
+}
+
+/* This function load's userpd firmware. Since Userpd depends on rootpd
Use Linux coding style comments.
Sure, I will update.
+ * first bring up root pd and then load. User pd firmware load is
required
+ * only during user pd restart because root pd loads user pd FW pil
segments
+ * during it's bringup.
+ */
+static int wcss_ahb_pcie_pd_load(struct rproc *rproc, const struct
firmware *fw)
+{
+ struct q6_wcss *wcss = rproc->priv, *wcss_rpd;
+ struct rproc *rpd_rproc = dev_get_drvdata(wcss->dev->parent);
+ const struct wcss_data *desc = wcss->desc;
+ int ret;
+
+ wcss_rpd = rpd_rproc->priv;
+
+ /* Boot rootpd rproc */
That's not helpful - i tcopies the function name.
sure, i will remove comment.
+ ret = rproc_boot(rpd_rproc);
+ if (ret || wcss->state == WCSS_NORMAL)
+ return ret;
+
+ return desc->mdt_load_sec(wcss->dev, fw, rproc->firmware,
+ desc->pasid, wcss->mem_region,
+ wcss->mem_phys, wcss->mem_size,
+ &wcss->mem_reloc);
+}
+
...
I didn't understand. Can you please elaborate your comment?
+static int q6_get_inbound_irq(struct qcom_q6v5 *q6,
+ struct platform_device *pdev,
+ const char *int_name,
+ irqreturn_t (*handler)(int irq, void *data))
+{
+ int ret, irq;
+ char *interrupt, *tmp = (char *)int_name;
+ struct q6_wcss *wcss = q6->rproc->priv;
+
+ irq = platform_get_irq_byname(pdev, int_name);
+ if (irq < 0) {
+ if (irq != -EPROBE_DEFER)
+ dev_err(&pdev->dev,
+ "failed to retrieve %s IRQ: %d\n",
+ int_name, irq);
return dev_err_probe
I will use dev_err_probe.
+ return irq;
+ }
+
+ if (!strcmp(int_name, "fatal")) {
+ q6->fatal_irq = irq;
+ } else if (!strcmp(int_name, "stop-ack")) {
+ q6->stop_irq = irq;
+ tmp = "stop_ack";
+ } else if (!strcmp(int_name, "ready")) {
+ q6->ready_irq = irq;
+ } else if (!strcmp(int_name, "handover")) {
+ q6->handover_irq = irq;
+ } else if (!strcmp(int_name, "spawn-ack")) {
+ q6->spawn_irq = irq;
+ tmp = "spawn_ack";
+ } else {
+ dev_err(&pdev->dev, "unknown interrupt\n");
+ return -EINVAL;
+ }
This is over complicated method of getting interrupts. First, you can
pass here address of variable with interrupt to assign (*irq_field =
irq). Second, drop the names and get by index. Entries are fixed, not
flexible.
Sure, I will do it.
+
+ interrupt = devm_kzalloc(&pdev->dev, BUF_SIZE, GFP_KERNEL);
+ if (!interrupt)
+ return -ENOMEM;
+
+ snprintf(interrupt, BUF_SIZE, "q6v5_wcss_userpd%d_%s",
wcss->pd_asid, tmp);
+
+ ret = devm_request_threaded_irq(&pdev->dev, irq,
+ NULL, handler,
+ IRQF_TRIGGER_RISING | IRQF_ONESHOT,
+ interrupt, q6);
+ if (ret) {
+ dev_err(&pdev->dev, "failed to acquire %s irq\n", interrupt);
+ return ret;
+ }
+ return 0;
+}
+
+static int q6_get_outbound_irq(struct qcom_q6v5 *q6,
+ struct platform_device *pdev,
+ const char *int_name)
+{
+ struct qcom_smem_state *tmp_state;
+ unsigned bit;
+
+ tmp_state = qcom_smem_state_get(&pdev->dev, int_name, &bit);
+ if (IS_ERR(tmp_state)) {
+ dev_err(&pdev->dev, "failed to acquire %s state\n", int_name);
+ return PTR_ERR(tmp_state);
+ }
+
+ if (!strcmp(int_name, "stop")) {
+ q6->state = tmp_state;
+ q6->stop_bit = bit;
+ } else if (!strcmp(int_name, "spawn")) {
+ q6->spawn_state = tmp_state;
+ q6->spawn_bit = bit;
+ }
+
+ return 0;
+}
+
+static int init_irq(struct qcom_q6v5 *q6,
+ struct platform_device *pdev, struct rproc *rproc,
+ int crash_reason, const char *load_state,
+ void (*handover)(struct qcom_q6v5 *q6))
+{
+ int ret;
+
+ q6->rproc = rproc;
+ q6->dev = &pdev->dev;
+ q6->crash_reason = crash_reason;
+ q6->handover = handover;
+
+ init_completion(&q6->start_done);
+ init_completion(&q6->stop_done);
+ init_completion(&q6->spawn_done);
+
+ ret = q6_get_inbound_irq(q6, pdev, "fatal",
+ q6v5_fatal_interrupt);
+ if (ret)
+ return ret;
+
+ ret = q6_get_inbound_irq(q6, pdev, "ready",
+ q6v5_ready_interrupt);
+ if (ret)
+ return ret;
+
+ ret = q6_get_inbound_irq(q6, pdev, "stop-ack",
+ q6v5_stop_interrupt);
+ if (ret)
+ return ret;
+
+ ret = q6_get_inbound_irq(q6, pdev, "spawn-ack",
+ q6v5_spawn_interrupt);
+ if (ret)
+ return ret;
+
+ ret = q6_get_outbound_irq(q6, pdev, "stop");
+ if (ret)
+ return ret;
+
+ ret = q6_get_outbound_irq(q6, pdev, "spawn");
+ if (ret)
+ return ret;
+
+ return 0;
+}
+
+static int q6_wcss_probe(struct platform_device *pdev)
+{
+ const struct wcss_data *desc;
+ struct q6_wcss *wcss;
+ struct rproc *rproc;
+ int ret;
+ char *subdev_name;
+ const char **firmware;
+
+ desc = of_device_get_match_data(&pdev->dev);
+ if (!desc)
+ return -EINVAL;
+
+ firmware = devm_kcalloc(&pdev->dev, MAX_FIRMWARE,
+ sizeof(*firmware), GFP_KERNEL);
+ if (!firmware)
+ return -ENOMEM;
+
+ ret = of_property_read_string_array(pdev->dev.of_node,
"firmware-name",
+ firmware, MAX_FIRMWARE);
+ if (ret < 0)
+ return ret;
+
+ rproc = rproc_alloc(&pdev->dev, pdev->name, desc->ops,
+ firmware[0], sizeof(*wcss));
+ if (!rproc) {
+ dev_err(&pdev->dev, "failed to allocate rproc\n");
ENOMEM do not print errors. Why do you have to print something here?
Yeah, I will remove print.
+ return -ENOMEM;
+ }
+ wcss = rproc->priv;
+ wcss->dev = &pdev->dev;
+ wcss->desc = desc;
+ wcss->firmware = firmware;
+
+ ret = q6_alloc_memory_region(wcss);
+ if (ret)
+ goto free_rproc;
+
+ wcss->pd_asid = qcom_get_pd_asid(wcss->dev->of_node);
+ if (wcss->pd_asid < 0)
+ goto free_rproc;
+
+ if (desc->init_irq) {
+ ret = desc->init_irq(&wcss->q6, pdev, rproc,
+ desc->crash_reason_smem, NULL, NULL);
+ if (ret)
+ goto free_rproc;
+ }
+
+ if (desc->glink_subdev_required)
+ qcom_add_glink_subdev(rproc, &wcss->glink_subdev,
desc->ssr_name);
+
+ subdev_name = (char *)(desc->ssr_name ? desc->ssr_name :
pdev->name);
+ qcom_add_ssr_subdev(rproc, &wcss->ssr_subdev, subdev_name);
+
+ rproc->auto_boot = false;
+ ret = rproc_add(rproc);
+ if (ret)
+ goto free_rproc;
+
+ platform_set_drvdata(pdev, rproc);
+
+ ret = of_platform_populate(wcss->dev->of_node, NULL,
+ NULL, wcss->dev);
It is the same probe used for the children, right? So whom do they
populate?
I will add match table in 2nd param, so that childs, whose
compatible matches with match table will be populated.
Is this fine ?
+ if (ret) {
+ dev_err(&pdev->dev, "failed to populate wcss pd nodes\n");
dev_err_probe
Sure, i will use dev_err_probe
+ goto free_rproc;
+ }
+ return 0;
+
+free_rproc:
+ rproc_free(rproc);
+
+ return ret;
+}
+
+static int q6_wcss_remove(struct platform_device *pdev)
+{
+ struct rproc *rproc = platform_get_drvdata(pdev);
+ struct q6_wcss *wcss = rproc->priv;
+
+ qcom_q6v5_deinit(&wcss->q6);
+
+ rproc_del(rproc);
+ rproc_free(rproc);
+
+ return 0;
+}
Thanks & Regards,
Manikanta.
