Re: [RFC PATCH 1/1] soc: qcom: Add driver to read secondary bootloader (XBL) log

From: Ninad Naik
Date: Wed Aug 23 2023 - 07:16:43 EST


Hi Trilok,

On 8/22/2023 9:38 PM, Trilok Soni wrote:
On 8/22/2023 5:15 AM, Ninad Naik wrote:
Qualcomm secondary bootloader (XBL) boot log holds information to
identify various firmware configuration currently set on the SoC.
The XBL log is stored in a predefined reserved memory region.

What does "X" stands for here? From what you have described above it looks like SBL and not XBL.

Ack. I will change the commit text from "Secondary" to "eXtensible". Thank you.

This drivers provides a way to print XBL logs on the console. To
do so, it provides a debugfs entry which captures the logs stored
in this reserved memory region. This entry can now be used to read
and print the XBL logs to console.

User can use the below command to print XBL log to console:
         cat /sys/kernel/debug/xbl_log


It is not clear to me why these patches are posted as RFC. Please clarify. Are they not tested properly or just seeking some feedback and driver is not ready w/ all the features?

Hi Trilok, the driver is tested on sa8775 platform.The reason for posting as RFC is to seek feedback on possible improvements. One was primarily on the probing mechanism which I have pointers now from Pavan and Srini's suggestions.

Signed-off-by: Ninad Naik <quic_ninanaik@xxxxxxxxxxx>
---
  drivers/soc/qcom/Kconfig        |  13 +++
  drivers/soc/qcom/Makefile       |   1 +
  drivers/soc/qcom/dump_xbl_log.c | 139 ++++++++++++++++++++++++++++++++
  3 files changed, 153 insertions(+)
  create mode 100644 drivers/soc/qcom/dump_xbl_log.c

diff --git a/drivers/soc/qcom/Kconfig b/drivers/soc/qcom/Kconfig
index 715348869d04..4489d37e924d 100644
--- a/drivers/soc/qcom/Kconfig
+++ b/drivers/soc/qcom/Kconfig
@@ -291,4 +291,17 @@ config QCOM_INLINE_CRYPTO_ENGINE
      tristate
      select QCOM_SCM
+config QCOM_DUMP_XBL_LOG
+    tristate "Qualcomm driver to print XBL logs on console from debugfs"

Why you want to print these logs from the debugfs? What is the format of the logs? Can you post an example log?

This log is printed in plain ascii format and are accessed post boot up, so chose to access them through debugfs. I have uploaded the full logs at [1]

[1] https://gist.github.com/ninadnaik-quic/914e86f6a0acadade632dc45b727d87b
+    help
+      This driver is used to capture secondary bootloader (xbl) log
+      from a reserved memory region and provide a debugfs entry to read
+      logs captured from this memory region and print them on console.
+      User can use below command to print the xbl log on console:
+
+                cat /sys/kernel/debug/xbl_log
+
+      These logs help to identify firmware configuration information on
+      the SoC. The name of the built module will be dump_xbl_log
+
  endmenu
diff --git a/drivers/soc/qcom/Makefile b/drivers/soc/qcom/Makefile
index bbca2e1e55bb..aac088a1a0b6 100644
--- a/drivers/soc/qcom/Makefile
+++ b/drivers/soc/qcom/Makefile
@@ -32,3 +32,4 @@ obj-$(CONFIG_QCOM_KRYO_L2_ACCESSORS) += kryo-l2-accessors.o
  obj-$(CONFIG_QCOM_ICC_BWMON)    += icc-bwmon.o
  qcom_ice-objs            += ice.o
  obj-$(CONFIG_QCOM_INLINE_CRYPTO_ENGINE)    += qcom_ice.o
+obj-$(CONFIG_QCOM_DUMP_XBL_LOG)    += dump_xbl_log.o
diff --git a/drivers/soc/qcom/dump_xbl_log.c b/drivers/soc/qcom/dump_xbl_log.c
new file mode 100644
index 000000000000..ea335a5e660b
--- /dev/null
+++ b/drivers/soc/qcom/dump_xbl_log.c
@@ -0,0 +1,139 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright (c) 2021, The Linux Foundation. All rights reserved.
+ */
+
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/of.h>
+#include <linux/device.h>
+#include <linux/platform_device.h>
+#include <linux/memblock.h>
+#include <linux/of_address.h>
+#include <linux/err.h>
+#include <linux/debugfs.h>
+
+struct xbl_log_data {
+    struct device *dev;
+    size_t buf_size;
+    void __iomem *xbl_buf;
+    struct dentry *dbg_file;
+    struct debugfs_blob_wrapper dbg_data;
+};
+
+static int map_addr_range(struct device_node **parent, const char *name,
+              struct xbl_log_data *xbl_data)
+{
+    struct device_node *node;
+    struct resource res;
+    int ret;
+
+    node = of_find_node_by_name(*parent, name);
+    if (!node)
+        return -ENODEV;
+
+    ret = of_address_to_resource(node, 0, &res);
+    if (ret) {
+        dev_err(xbl_data->dev, "Failed to parse memory region\n");
+        return ret;
+    }
+    of_node_put(node);
+
+    if (!resource_size(&res)) {
+        dev_err(xbl_data->dev, "Failed to parse memory region size\n");
+        return -ENODEV;
+    }
+
+    xbl_data->buf_size = resource_size(&res) - 1;
+    xbl_data->xbl_buf = devm_memremap(xbl_data->dev, res.start,
+                      xbl_data->buf_size, MEMREMAP_WB);
+    if (!xbl_data->xbl_buf) {
+        dev_err(xbl_data->dev, "%s: memory remap failed\n", name);
+        return -ENOMEM;
+    }
+
+    return 0;
+}
+
+static int xbl_log_probe(struct platform_device *pdev)
+{
+    struct device *dev = &pdev->dev;
+    struct xbl_log_data *xbl_data;
+    struct device_node *parent;
+    int ret;
+
+    xbl_data = devm_kzalloc(dev, sizeof(*xbl_data), GFP_KERNEL);
+    if (!xbl_data)
+        return -ENOMEM;
+
+    xbl_data->dev = &pdev->dev;
+    platform_set_drvdata(pdev, xbl_data);
+
+    parent = of_find_node_by_path("/reserved-memory");
+    if (!parent) {
+        dev_err(xbl_data->dev, "reserved-memory node missing\n");
+        return -ENODEV;
+    }
+
+    ret = map_addr_range(&parent, "uefi-log", xbl_data);

Here you are calling it as uefi-log. Is it xbl-log or uefi-log? Please decide first.


The reason for using "uefi-log" here is because this node name is in accordance to the sa8775p.dtsi as seen in [2]

[2] https://github.com/torvalds/linux/blob/master/arch/arm64/boot/dts/qcom/sa8775p.dtsi#L354

So, in the next revision while adding the device tree bindings and corresponding compatible string, should this node name be changed as well?
+    if (ret)
+        goto put_node;
+
+    xbl_data->dbg_data.data = xbl_data->xbl_buf;
+    xbl_data->dbg_data.size = xbl_data->buf_size;
+    xbl_data->dbg_file = debugfs_create_blob("xbl_log", 0400, NULL,
+                         &xbl_data->dbg_data);
+    if (IS_ERR(xbl_data->dbg_file)) {
+        dev_err(xbl_data->dev, "failed to create debugfs entry\n");
+        ret = PTR_ERR(xbl_data->dbg_file);
+    }
+
+put_node:
+    of_node_put(parent);
+    return ret;
+}
+
+static int xbl_log_remove(struct platform_device *pdev)
+{
+    struct xbl_log_data *xbl_data = platform_get_drvdata(pdev);
+
+    debugfs_remove_recursive(xbl_data->dbg_file);
+    return 0;
+}
+
+static struct platform_driver xbl_log_driver = {
+    .probe = xbl_log_probe,
+    .remove = xbl_log_remove,
+    .driver = {
+           .name = "xbl-log",
+           },
+};
+
+static struct platform_device xbl_log_device = {
+    .name = "xbl-log",
+};
+
+static int __init xbl_log_init(void)
+{
+    int ret = 0;
+
+    ret = platform_driver_register(&xbl_log_driver);
+    if (!ret) {
+        ret = platform_device_register(&xbl_log_device);

I am puzzled here. Why?

Ack. I was registering a platform device to get the driver probed when built as module. I'll correct this according to comments from Srinivas and Pavan.

+        if (ret)
+            platform_driver_unregister(&xbl_log_driver);
+    }
+    return ret;
+}
+
+static void __exit xbl_log_exit(void)
+{
+    platform_device_unregister(&xbl_log_device);
+    platform_driver_unregister(&xbl_log_driver);
+}
+
+module_init(xbl_log_init);
+module_exit(xbl_log_exit);
+
+MODULE_DESCRIPTION("Qualcomm Technologies, Inc. (QTI) XBL log driver");
+MODULE_LICENSE("GPL");

Thanks,
Ninad