[PATCH v7 0/5] Add support for debugfs based RAS DES feature in PCIe DW

[PATCH v7 0/5] Add support for debugfs based RAS DES feature in PCIe DW

[Shradha Todi]

DesignWare controller provides a vendor specific extended capability
called RASDES as an IP feature. This extended capability  provides
hardware information like:
 - Debug registers to know the state of the link or controller. 
 - Error injection mechanisms to inject various PCIe errors including
   sequence number, CRC
 - Statistical counters to know how many times a particular event
   occurred

However, in Linux we do not have any generic or custom support to be
able to use this feature in an efficient manner. This is the reason we
are proposing this framework. Debug and bring up time of high-speed IPs
are highly dependent on costlier hardware analyzers and this solution
will in some ways help to reduce the HW analyzer usage.

The debugfs entries can be used to get information about underlying
hardware and can be shared with user space. Separate debugfs entries has
been created to cater to all the DES hooks provided by the controller.
The debugfs entries interacts with the RASDES registers in the required
sequence and provides the meaningful data to the user. This eases the
effort to understand and use the register information for debugging.

This series creates a generic debugfs framework for DesignWare PCIe
controllers where other debug features apart from RASDES can also be
added as and when required.

v7:
    - Moved the patches to make finding VSEC IDs common from Mani's patchset [1]
      into this series to remove dependancy as discussed
    - Addressed style related change requests from v6

v6: https://lore.kernel.org/all/20250214105007.97582-1-shradha.t@samsung.com/
    - Addressed Niklas's comment to make vsec ID finding similar to perf
    - Minor changes in the driver to make the debugfs file common and
      not specefic to RASDES so that other developers can add debug
      related features to this file.

v5: https://lore.kernel.org/all/20250121111421.35437-1-shradha.t@samsung.com/
    - Addressed Fan's comment to split the patches for easier review
    - Addressed Bjorn's comment to fix vendor specific cap search
    - Addressed style related change requests from v4
    - Added rasdes debugfs init call to common designware files for host
      and EP.

v4: https://lore.kernel.org/lkml/20241206074456.17401-1-shradha.t@samsung.com/
    - Addressed comments from Manivannan, Bjorn and Jonathan
    - Addressed style related change requests from v3
    - Added Documentation under Documentation/ABI/testing and kdoc stype
      comments wherever required for better understanding
    - Enhanced error injection to include all possible error groups
    - Removed debugfs init call from common designware file and left it
      up to individual platform drivers to init/deinit as required.

v3: https://lore.kernel.org/all/20240625093813.112555-1-shradha.t@samsung.com/
    - v2 had suggestions about moving this framework to perf/EDAC instead of a
      controller specific debugfs but after discussions we decided to go ahead
      with the same. Rebased and posted v3 with minor style changes.

v2: https://lore.kernel.org/lkml/20231130115044.53512-1-shradha.t@samsung.com/
    - Addressed comments from Krzysztof Wilczyński, Bjorn Helgaas and
      posted v2 with a changed implementation for a better code design

v1: https://lore.kernel.org/all/20210518174618.42089-1-shradha.t@samsung.com/T/

[1] https://lore.kernel.org/all/20250218-pcie-qcom-ptm-v1-0-16d7e480d73e@linaro.org/

Manivannan Sadhasivam (1):
  perf/dwc_pcie: Move common DWC struct definitions to 'pcie-dwc.h'

Shradha Todi (4):
  PCI: dwc: Add helper to find the Vendor Specific Extended Capability
    (VSEC)
  Add debugfs based silicon debug support in DWC
  Add debugfs based error injection support in DWC
  Add debugfs based statistical counter support in DWC

 Documentation/ABI/testing/debugfs-dwc-pcie    | 144 +++++
 MAINTAINERS                                   |   1 +
 drivers/pci/controller/dwc/Kconfig            |  10 +
 drivers/pci/controller/dwc/Makefile           |   1 +
 .../controller/dwc/pcie-designware-debugfs.c  | 564 ++++++++++++++++++
 .../pci/controller/dwc/pcie-designware-ep.c   |   5 +
 .../pci/controller/dwc/pcie-designware-host.c |   6 +
 drivers/pci/controller/dwc/pcie-designware.c  |  46 ++
 drivers/pci/controller/dwc/pcie-designware.h  |  21 +
 drivers/perf/dwc_pcie_pmu.c                   |  25 +-
 include/linux/pcie-dwc.h                      |  36 ++
 11 files changed, 837 insertions(+), 22 deletions(-)
 create mode 100644 Documentation/ABI/testing/debugfs-dwc-pcie
 create mode 100644 drivers/pci/controller/dwc/pcie-designware-debugfs.c
 create mode 100644 include/linux/pcie-dwc.h

-- 
2.17.1

[PATCH v7 1/5] perf/dwc_pcie: Move common DWC struct definitions to 'pcie-dwc.h'

[Shradha Todi]

From: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>

Since these are common to all Desginware PCIe IPs, move them to a new
header 'pcie-dwc.h', so that other drivers like debugfs, perf and sysfs
could make use of them.

Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Shradha Todi <shradha.t@samsung.com>
---
 MAINTAINERS                 |  1 +
 drivers/perf/dwc_pcie_pmu.c | 25 +++----------------------
 include/linux/pcie-dwc.h    | 34 ++++++++++++++++++++++++++++++++++
 3 files changed, 38 insertions(+), 22 deletions(-)
 create mode 100644 include/linux/pcie-dwc.h

diff --git a/MAINTAINERS b/MAINTAINERS
index 3864d473f52f..6474a2d83de4 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -18167,6 +18167,7 @@ S:	Maintained
 F:	Documentation/devicetree/bindings/pci/snps,dw-pcie-ep.yaml
 F:	Documentation/devicetree/bindings/pci/snps,dw-pcie.yaml
 F:	drivers/pci/controller/dwc/*designware*
+F:	include/linux/pcie-dwc.h
 
 PCI DRIVER FOR TI DRA7XX/J721E
 M:	Vignesh Raghavendra <vigneshr@ti.com>
diff --git a/drivers/perf/dwc_pcie_pmu.c b/drivers/perf/dwc_pcie_pmu.c
index cccecae9823f..da30f2c2d674 100644
--- a/drivers/perf/dwc_pcie_pmu.c
+++ b/drivers/perf/dwc_pcie_pmu.c
@@ -13,6 +13,7 @@
 #include <linux/errno.h>
 #include <linux/kernel.h>
 #include <linux/list.h>
+#include <linux/pcie-dwc.h>
 #include <linux/perf_event.h>
 #include <linux/pci.h>
 #include <linux/platform_device.h>
@@ -99,26 +100,6 @@ struct dwc_pcie_dev_info {
 	struct list_head dev_node;
 };
 
-struct dwc_pcie_pmu_vsec_id {
-	u16 vendor_id;
-	u16 vsec_id;
-	u8 vsec_rev;
-};
-
-/*
- * VSEC IDs are allocated by the vendor, so a given ID may mean different
- * things to different vendors.  See PCIe r6.0, sec 7.9.5.2.
- */
-static const struct dwc_pcie_pmu_vsec_id dwc_pcie_pmu_vsec_ids[] = {
-	{ .vendor_id = PCI_VENDOR_ID_ALIBABA,
-	  .vsec_id = 0x02, .vsec_rev = 0x4 },
-	{ .vendor_id = PCI_VENDOR_ID_AMPERE,
-	  .vsec_id = 0x02, .vsec_rev = 0x4 },
-	{ .vendor_id = PCI_VENDOR_ID_QCOM,
-	  .vsec_id = 0x02, .vsec_rev = 0x4 },
-	{} /* terminator */
-};
-
 static ssize_t cpumask_show(struct device *dev,
 					 struct device_attribute *attr,
 					 char *buf)
@@ -529,14 +510,14 @@ static void dwc_pcie_unregister_pmu(void *data)
 
 static u16 dwc_pcie_des_cap(struct pci_dev *pdev)
 {
-	const struct dwc_pcie_pmu_vsec_id *vid;
+	const struct dwc_pcie_vsec_id *vid;
 	u16 vsec;
 	u32 val;
 
 	if (!pci_is_pcie(pdev) || !(pci_pcie_type(pdev) == PCI_EXP_TYPE_ROOT_PORT))
 		return 0;
 
-	for (vid = dwc_pcie_pmu_vsec_ids; vid->vendor_id; vid++) {
+	for (vid = dwc_pcie_rasdes_vsec_ids; vid->vendor_id; vid++) {
 		vsec = pci_find_vsec_capability(pdev, vid->vendor_id,
 						vid->vsec_id);
 		if (vsec) {
diff --git a/include/linux/pcie-dwc.h b/include/linux/pcie-dwc.h
new file mode 100644
index 000000000000..40f3545731c8
--- /dev/null
+++ b/include/linux/pcie-dwc.h
@@ -0,0 +1,34 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Copyright (C) 2021-2023 Alibaba Inc.
+ *
+ * Copyright 2025 Linaro Ltd.
+ * Author: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
+ */
+
+#ifndef LINUX_PCIE_DWC_H
+#define LINUX_PCIE_DWC_H
+
+#include <linux/pci_ids.h>
+
+struct dwc_pcie_vsec_id {
+	u16 vendor_id;
+	u16 vsec_id;
+	u8 vsec_rev;
+};
+
+/*
+ * VSEC IDs are allocated by the vendor, so a given ID may mean different
+ * things to different vendors.  See PCIe r6.0, sec 7.9.5.2.
+ */
+static const struct dwc_pcie_vsec_id dwc_pcie_rasdes_vsec_ids[] = {
+	{ .vendor_id = PCI_VENDOR_ID_ALIBABA,
+	  .vsec_id = 0x02, .vsec_rev = 0x4 },
+	{ .vendor_id = PCI_VENDOR_ID_AMPERE,
+	  .vsec_id = 0x02, .vsec_rev = 0x4 },
+	{ .vendor_id = PCI_VENDOR_ID_QCOM,
+	  .vsec_id = 0x02, .vsec_rev = 0x4 },
+	{} /* terminator */
+};
+
+#endif /* LINUX_PCIE_DWC_H */
-- 
2.17.1

[PATCH v7 2/5] PCI: dwc: Add helper to find the Vendor Specific Extended Capability (VSEC)

[Shradha Todi]

dw_pcie_find_vsec_capability() is used by upcoming DWC APIs to find the
VSEC capabilities like PTM, RAS etc.

Co-developed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Shradha Todi <shradha.t@samsung.com>
---
 drivers/pci/controller/dwc/pcie-designware.c | 40 ++++++++++++++++++++
 1 file changed, 40 insertions(+)

diff --git a/drivers/pci/controller/dwc/pcie-designware.c b/drivers/pci/controller/dwc/pcie-designware.c
index 145e7f579072..a7c0671c6715 100644
--- a/drivers/pci/controller/dwc/pcie-designware.c
+++ b/drivers/pci/controller/dwc/pcie-designware.c
@@ -16,6 +16,7 @@
 #include <linux/gpio/consumer.h>
 #include <linux/ioport.h>
 #include <linux/of.h>
+#include <linux/pcie-dwc.h>
 #include <linux/platform_device.h>
 #include <linux/sizes.h>
 #include <linux/types.h>
@@ -283,6 +284,45 @@ u16 dw_pcie_find_ext_capability(struct dw_pcie *pci, u8 cap)
 }
 EXPORT_SYMBOL_GPL(dw_pcie_find_ext_capability);
 
+static u16 __dw_pcie_find_vsec_capability(struct dw_pcie *pci, u16 vendor_id,
+					  u16 vsec_id)
+{
+	u16 vsec = 0;
+	u32 header;
+
+	if (vendor_id != dw_pcie_readw_dbi(pci, PCI_VENDOR_ID))
+		return 0;
+
+	while ((vsec = dw_pcie_find_next_ext_capability(pci, vsec,
+						       PCI_EXT_CAP_ID_VNDR))) {
+		header = dw_pcie_readl_dbi(pci, vsec + PCI_VNDR_HEADER);
+		if (PCI_VNDR_HEADER_ID(header) == vsec_id)
+			return vsec;
+	}
+
+	return 0;
+}
+
+static u16 dw_pcie_find_vsec_capability(struct dw_pcie *pci,
+					const struct dwc_pcie_vsec_id *vsec_ids)
+{
+	const struct dwc_pcie_vsec_id *vid;
+	u16 vsec;
+	u32 header;
+
+	for (vid = vsec_ids; vid->vendor_id; vid++) {
+		vsec = __dw_pcie_find_vsec_capability(pci, vid->vendor_id,
+						      vid->vsec_id);
+		if (vsec) {
+			header = dw_pcie_readl_dbi(pci, vsec + PCI_VNDR_HEADER);
+			if (PCI_VNDR_HEADER_REV(header) == vid->vsec_rev)
+				return vsec;
+		}
+	}
+
+	return 0;
+}
+
 int dw_pcie_read(void __iomem *addr, int size, u32 *val)
 {
 	if (!IS_ALIGNED((uintptr_t)addr, size)) {
-- 
2.17.1

[PATCH v7 3/5] Add debugfs based silicon debug support in DWC

[Shradha Todi]

Add support to provide silicon debug interface to userspace. This set
of debug registers are part of the RASDES feature present in DesignWare
PCIe controllers.

Signed-off-by: Shradha Todi <shradha.t@samsung.com>
---
 Documentation/ABI/testing/debugfs-dwc-pcie    |  13 ++
 drivers/pci/controller/dwc/Kconfig            |  10 +
 drivers/pci/controller/dwc/Makefile           |   1 +
 .../controller/dwc/pcie-designware-debugfs.c  | 176 ++++++++++++++++++
 .../pci/controller/dwc/pcie-designware-ep.c   |   5 +
 .../pci/controller/dwc/pcie-designware-host.c |   6 +
 drivers/pci/controller/dwc/pcie-designware.c  |   6 +
 drivers/pci/controller/dwc/pcie-designware.h  |  21 +++
 include/linux/pcie-dwc.h                      |   2 +
 9 files changed, 240 insertions(+)
 create mode 100644 Documentation/ABI/testing/debugfs-dwc-pcie
 create mode 100644 drivers/pci/controller/dwc/pcie-designware-debugfs.c

diff --git a/Documentation/ABI/testing/debugfs-dwc-pcie b/Documentation/ABI/testing/debugfs-dwc-pcie
new file mode 100644
index 000000000000..e8ed34e988ef
--- /dev/null
+++ b/Documentation/ABI/testing/debugfs-dwc-pcie
@@ -0,0 +1,13 @@
+What:		/sys/kernel/debug/dwc_pcie_<dev>/rasdes_debug/lane_detect
+Date:		Feburary 2025
+Contact:	Shradha Todi <shradha.t@samsung.com>
+Description:	(RW) Write the lane number to be checked for detection.	Read
+		will return whether PHY indicates receiver detection on the
+		selected lane. The default selected lane is Lane0.
+
+What:		/sys/kernel/debug/dwc_pcie_<dev>/rasdes_debug/rx_valid
+Date:		Feburary 2025
+Contact:	Shradha Todi <shradha.t@samsung.com>
+Description:	(RW) Write the lane number to be checked as valid or invalid. Read
+		will return the status of PIPE RXVALID signal of the selected lane.
+		The default selected lane is Lane0.
diff --git a/drivers/pci/controller/dwc/Kconfig b/drivers/pci/controller/dwc/Kconfig
index b6d6778b0698..48a10428a492 100644
--- a/drivers/pci/controller/dwc/Kconfig
+++ b/drivers/pci/controller/dwc/Kconfig
@@ -6,6 +6,16 @@ menu "DesignWare-based PCIe controllers"
 config PCIE_DW
 	bool
 
+config PCIE_DW_DEBUGFS
+	default y
+	depends on DEBUG_FS
+	depends on PCIE_DW_HOST || PCIE_DW_EP
+	bool "DWC PCIe debugfs entries"
+	help
+	  Enables debugfs entries for the DW PCIe Controller. These entries
+	  provide all debug features related to DW controller including the RAS
+	  DES features to help in debug, error injection and statistical counters.
+
 config PCIE_DW_HOST
 	bool
 	select PCIE_DW
diff --git a/drivers/pci/controller/dwc/Makefile b/drivers/pci/controller/dwc/Makefile
index a8308d9ea986..54565eedc52c 100644
--- a/drivers/pci/controller/dwc/Makefile
+++ b/drivers/pci/controller/dwc/Makefile
@@ -1,5 +1,6 @@
 # SPDX-License-Identifier: GPL-2.0
 obj-$(CONFIG_PCIE_DW) += pcie-designware.o
+obj-$(CONFIG_PCIE_DW_DEBUGFS) += pcie-designware-debugfs.o
 obj-$(CONFIG_PCIE_DW_HOST) += pcie-designware-host.o
 obj-$(CONFIG_PCIE_DW_EP) += pcie-designware-ep.o
 obj-$(CONFIG_PCIE_DW_PLAT) += pcie-designware-plat.o
diff --git a/drivers/pci/controller/dwc/pcie-designware-debugfs.c b/drivers/pci/controller/dwc/pcie-designware-debugfs.c
new file mode 100644
index 000000000000..3887a6996706
--- /dev/null
+++ b/drivers/pci/controller/dwc/pcie-designware-debugfs.c
@@ -0,0 +1,176 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Synopsys DesignWare PCIe controller debugfs driver
+ *
+ * Copyright (C) 2025 Samsung Electronics Co., Ltd.
+ *		http://www.samsung.com
+ *
+ * Author: Shradha Todi <shradha.t@samsung.com>
+ */
+
+#include <linux/debugfs.h>
+
+#include "pcie-designware.h"
+
+#define SD_STATUS_L1LANE_REG		0xb0
+#define PIPE_RXVALID			BIT(18)
+#define PIPE_DETECT_LANE		BIT(17)
+#define LANE_SELECT			GENMASK(3, 0)
+
+#define DWC_DEBUGFS_BUF_MAX		128
+
+/**
+ * struct dwc_pcie_rasdes_info - Stores controller common information
+ * @ras_cap_offset: RAS DES vendor specific extended capability offset
+ * @reg_event_lock: Mutex used for RASDES shadow event registers
+ *
+ * Any parameter constant to all files of the debugfs hierarchy for a single controller
+ * will be stored in this struct. It is allocated and assigned to controller specific
+ * struct dw_pcie during initialization.
+ */
+struct dwc_pcie_rasdes_info {
+	u32 ras_cap_offset;
+	struct mutex reg_event_lock;
+};
+
+static ssize_t lane_detect_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
+{
+	struct dw_pcie *pci = file->private_data;
+	struct dwc_pcie_rasdes_info *rinfo = pci->debugfs->rasdes_info;
+	char debugfs_buf[DWC_DEBUGFS_BUF_MAX];
+	ssize_t pos;
+	u32 val;
+
+	val = dw_pcie_readl_dbi(pci, rinfo->ras_cap_offset + SD_STATUS_L1LANE_REG);
+	val = FIELD_GET(PIPE_DETECT_LANE, val);
+	if (val)
+		pos = scnprintf(debugfs_buf, DWC_DEBUGFS_BUF_MAX, "Lane Detected\n");
+	else
+		pos = scnprintf(debugfs_buf, DWC_DEBUGFS_BUF_MAX, "Lane Undetected\n");
+
+	return simple_read_from_buffer(buf, count, ppos, debugfs_buf, pos);
+}
+
+static ssize_t lane_detect_write(struct file *file, const char __user *buf,
+				 size_t count, loff_t *ppos)
+{
+	struct dw_pcie *pci = file->private_data;
+	struct dwc_pcie_rasdes_info *rinfo = pci->debugfs->rasdes_info;
+	u32 lane, val;
+
+	val = kstrtou32_from_user(buf, count, 0, &lane);
+	if (val)
+		return val;
+
+	val = dw_pcie_readl_dbi(pci, rinfo->ras_cap_offset + SD_STATUS_L1LANE_REG);
+	val &= ~(LANE_SELECT);
+	val |= FIELD_PREP(LANE_SELECT, lane);
+	dw_pcie_writel_dbi(pci, rinfo->ras_cap_offset + SD_STATUS_L1LANE_REG, val);
+
+	return count;
+}
+
+static ssize_t rx_valid_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
+{
+	struct dw_pcie *pci = file->private_data;
+	struct dwc_pcie_rasdes_info *rinfo = pci->debugfs->rasdes_info;
+	char debugfs_buf[DWC_DEBUGFS_BUF_MAX];
+	ssize_t pos;
+	u32 val;
+
+	val = dw_pcie_readl_dbi(pci, rinfo->ras_cap_offset + SD_STATUS_L1LANE_REG);
+	val = FIELD_GET(PIPE_RXVALID, val);
+	if (val)
+		pos = scnprintf(debugfs_buf, DWC_DEBUGFS_BUF_MAX, "RX Valid\n");
+	else
+		pos = scnprintf(debugfs_buf, DWC_DEBUGFS_BUF_MAX, "RX Invalid\n");
+
+	return simple_read_from_buffer(buf, count, ppos, debugfs_buf, pos);
+}
+
+static ssize_t rx_valid_write(struct file *file, const char __user *buf, size_t count, loff_t *ppos)
+{
+	return lane_detect_write(file, buf, count, ppos);
+}
+
+#define dwc_debugfs_create(name)			\
+debugfs_create_file(#name, 0644, rasdes_debug, pci,	\
+			&dbg_ ## name ## _fops)
+
+#define DWC_DEBUGFS_FOPS(name)					\
+static const struct file_operations dbg_ ## name ## _fops = {	\
+	.open = simple_open,				\
+	.read = name ## _read,				\
+	.write = name ## _write				\
+}
+
+DWC_DEBUGFS_FOPS(lane_detect);
+DWC_DEBUGFS_FOPS(rx_valid);
+
+static void dwc_pcie_rasdes_debugfs_deinit(struct dw_pcie *pci)
+{
+	struct dwc_pcie_rasdes_info *rinfo = pci->debugfs->rasdes_info;
+
+	mutex_destroy(&rinfo->reg_event_lock);
+}
+
+static int dwc_pcie_rasdes_debugfs_init(struct dw_pcie *pci, struct dentry *dir)
+{
+	struct dentry *rasdes_debug;
+	struct dwc_pcie_rasdes_info *rasdes_info;
+	struct device *dev = pci->dev;
+	int ras_cap;
+
+	ras_cap = dw_pcie_find_rasdes_capability(pci);
+	if (!ras_cap) {
+		dev_dbg(dev, "no RASDES capability available\n");
+		return -ENODEV;
+	}
+
+	rasdes_info = devm_kzalloc(dev, sizeof(*rasdes_info), GFP_KERNEL);
+	if (!rasdes_info)
+		return -ENOMEM;
+
+	/* Create subdirectories for Debug, Error injection, Statistics */
+	rasdes_debug = debugfs_create_dir("rasdes_debug", dir);
+
+	mutex_init(&rasdes_info->reg_event_lock);
+	rasdes_info->ras_cap_offset = ras_cap;
+	pci->debugfs->rasdes_info = rasdes_info;
+
+	/* Create debugfs files for Debug subdirectory */
+	dwc_debugfs_create(lane_detect);
+	dwc_debugfs_create(rx_valid);
+
+	return 0;
+}
+
+void dwc_pcie_debugfs_deinit(struct dw_pcie *pci)
+{
+	dwc_pcie_rasdes_debugfs_deinit(pci);
+	debugfs_remove_recursive(pci->debugfs->debug_dir);
+}
+
+int dwc_pcie_debugfs_init(struct dw_pcie *pci)
+{
+	char dirname[DWC_DEBUGFS_BUF_MAX];
+	struct device *dev = pci->dev;
+	struct debugfs_info *debugfs;
+	struct dentry *dir;
+	int ret;
+
+	/* Create main directory for each platform driver */
+	snprintf(dirname, DWC_DEBUGFS_BUF_MAX, "dwc_pcie_%s", dev_name(dev));
+	dir = debugfs_create_dir(dirname, NULL);
+	debugfs = devm_kzalloc(dev, sizeof(*debugfs), GFP_KERNEL);
+	if (!debugfs)
+		return -ENOMEM;
+
+	debugfs->debug_dir = dir;
+	pci->debugfs = debugfs;
+	ret = dwc_pcie_rasdes_debugfs_init(pci, dir);
+	if (ret)
+		dev_dbg(dev, "RASDES debugfs init failed\n");
+
+	return 0;
+}
diff --git a/drivers/pci/controller/dwc/pcie-designware-ep.c b/drivers/pci/controller/dwc/pcie-designware-ep.c
index 72418160e658..f9d7f3f989ad 100644
--- a/drivers/pci/controller/dwc/pcie-designware-ep.c
+++ b/drivers/pci/controller/dwc/pcie-designware-ep.c
@@ -814,6 +814,7 @@ void dw_pcie_ep_cleanup(struct dw_pcie_ep *ep)
 {
 	struct dw_pcie *pci = to_dw_pcie_from_ep(ep);
 
+	dwc_pcie_debugfs_deinit(pci);
 	dw_pcie_edma_remove(pci);
 }
 EXPORT_SYMBOL_GPL(dw_pcie_ep_cleanup);
@@ -989,6 +990,10 @@ int dw_pcie_ep_init_registers(struct dw_pcie_ep *ep)
 
 	dw_pcie_ep_init_non_sticky_registers(pci);
 
+	ret = dwc_pcie_debugfs_init(pci);
+	if (ret)
+		goto err_remove_edma;
+
 	return 0;
 
 err_remove_edma:
diff --git a/drivers/pci/controller/dwc/pcie-designware-host.c b/drivers/pci/controller/dwc/pcie-designware-host.c
index ffaded8f2df7..2081e8c72d12 100644
--- a/drivers/pci/controller/dwc/pcie-designware-host.c
+++ b/drivers/pci/controller/dwc/pcie-designware-host.c
@@ -548,6 +548,10 @@ int dw_pcie_host_init(struct dw_pcie_rp *pp)
 	if (pp->ops->post_init)
 		pp->ops->post_init(pp);
 
+	ret = dwc_pcie_debugfs_init(pci);
+	if (ret)
+		goto err_stop_link;
+
 	return 0;
 
 err_stop_link:
@@ -572,6 +576,8 @@ void dw_pcie_host_deinit(struct dw_pcie_rp *pp)
 {
 	struct dw_pcie *pci = to_dw_pcie_from_pp(pp);
 
+	dwc_pcie_debugfs_deinit(pci);
+
 	pci_stop_root_bus(pp->bridge->bus);
 	pci_remove_root_bus(pp->bridge->bus);
 
diff --git a/drivers/pci/controller/dwc/pcie-designware.c b/drivers/pci/controller/dwc/pcie-designware.c
index a7c0671c6715..3d1d95d9e380 100644
--- a/drivers/pci/controller/dwc/pcie-designware.c
+++ b/drivers/pci/controller/dwc/pcie-designware.c
@@ -323,6 +323,12 @@ static u16 dw_pcie_find_vsec_capability(struct dw_pcie *pci,
 	return 0;
 }
 
+u16 dw_pcie_find_rasdes_capability(struct dw_pcie *pci)
+{
+	return dw_pcie_find_vsec_capability(pci, dwc_pcie_rasdes_vsec_ids);
+}
+EXPORT_SYMBOL_GPL(dw_pcie_find_rasdes_capability);
+
 int dw_pcie_read(void __iomem *addr, int size, u32 *val)
 {
 	if (!IS_ALIGNED((uintptr_t)addr, size)) {
diff --git a/drivers/pci/controller/dwc/pcie-designware.h b/drivers/pci/controller/dwc/pcie-designware.h
index 501d9ddfea16..7f9807d4e5de 100644
--- a/drivers/pci/controller/dwc/pcie-designware.h
+++ b/drivers/pci/controller/dwc/pcie-designware.h
@@ -437,6 +437,11 @@ struct dw_pcie_ops {
 	void	(*stop_link)(struct dw_pcie *pcie);
 };
 
+struct debugfs_info {
+	struct dentry		*debug_dir;
+	void			*rasdes_info;
+};
+
 struct dw_pcie {
 	struct device		*dev;
 	void __iomem		*dbi_base;
@@ -465,6 +470,7 @@ struct dw_pcie {
 	struct reset_control_bulk_data	core_rsts[DW_PCIE_NUM_CORE_RSTS];
 	struct gpio_desc		*pe_rst;
 	bool			suspended;
+	struct debugfs_info	*debugfs;
 };
 
 #define to_dw_pcie_from_pp(port) container_of((port), struct dw_pcie, pp)
@@ -478,6 +484,7 @@ void dw_pcie_version_detect(struct dw_pcie *pci);
 
 u8 dw_pcie_find_capability(struct dw_pcie *pci, u8 cap);
 u16 dw_pcie_find_ext_capability(struct dw_pcie *pci, u8 cap);
+u16 dw_pcie_find_rasdes_capability(struct dw_pcie *pci);
 
 int dw_pcie_read(void __iomem *addr, int size, u32 *val);
 int dw_pcie_write(void __iomem *addr, int size, u32 val);
@@ -806,4 +813,18 @@ dw_pcie_ep_get_func_from_ep(struct dw_pcie_ep *ep, u8 func_no)
 	return NULL;
 }
 #endif
+
+#ifdef CONFIG_PCIE_DW_DEBUGFS
+int dwc_pcie_debugfs_init(struct dw_pcie *pci);
+void dwc_pcie_debugfs_deinit(struct dw_pcie *pci);
+#else
+static inline int dwc_pcie_debugfs_init(struct dw_pcie *pci)
+{
+	return 0;
+}
+static inline void dwc_pcie_debugfs_deinit(struct dw_pcie *pci)
+{
+}
+#endif
+
 #endif /* _PCIE_DESIGNWARE_H */
diff --git a/include/linux/pcie-dwc.h b/include/linux/pcie-dwc.h
index 40f3545731c8..6436e7fadc75 100644
--- a/include/linux/pcie-dwc.h
+++ b/include/linux/pcie-dwc.h
@@ -28,6 +28,8 @@ static const struct dwc_pcie_vsec_id dwc_pcie_rasdes_vsec_ids[] = {
 	  .vsec_id = 0x02, .vsec_rev = 0x4 },
 	{ .vendor_id = PCI_VENDOR_ID_QCOM,
 	  .vsec_id = 0x02, .vsec_rev = 0x4 },
+	{ .vendor_id = PCI_VENDOR_ID_SAMSUNG,
+	  .vsec_id = 0x02, .vsec_rev = 0x4 },
 	{} /* terminator */
 };
 
-- 
2.17.1

[PATCH v7 4/5] Add debugfs based error injection support in DWC

[Shradha Todi]

Add support to provide error injection interface to userspace. This set
of debug registers are part of the RASDES feature present in DesignWare
PCIe controllers.

Signed-off-by: Shradha Todi <shradha.t@samsung.com>
---
 Documentation/ABI/testing/debugfs-dwc-pcie    |  70 ++++++++
 .../controller/dwc/pcie-designware-debugfs.c  | 165 +++++++++++++++++-
 2 files changed, 233 insertions(+), 2 deletions(-)

diff --git a/Documentation/ABI/testing/debugfs-dwc-pcie b/Documentation/ABI/testing/debugfs-dwc-pcie
index e8ed34e988ef..6ee0897fe753 100644
--- a/Documentation/ABI/testing/debugfs-dwc-pcie
+++ b/Documentation/ABI/testing/debugfs-dwc-pcie
@@ -11,3 +11,73 @@ Contact:	Shradha Todi <shradha.t@samsung.com>
 Description:	(RW) Write the lane number to be checked as valid or invalid. Read
 		will return the status of PIPE RXVALID signal of the selected lane.
 		The default selected lane is Lane0.
+
+What:		/sys/kernel/debug/dwc_pcie_<dev>/rasdes_err_inj/<error>
+Date:		Feburary 2025
+Contact:	Shradha Todi <shradha.t@samsung.com>
+Description:	rasdes_err_inj is the directory which can be used to inject errors in the
+		system. The possible errors that can be injected are:
+
+		1) TLP LCRC error injection TX Path - tx_lcrc
+		2) 16b CRC error injection of ACK/NAK DLLP - b16_crc_dllp
+		3) 16b CRC error injection of Update-FC DLLP - b16_crc_upd_fc
+		4) TLP ECRC error injection TX Path - tx_ecrc
+		5) TLP's FCRC error injection TX Path - fcrc_tlp
+		6) Parity error of TSOS - parity_tsos
+		7) Parity error on SKPOS - parity_skpos
+		8) LCRC error injection RX Path - rx_lcrc
+		9) ECRC error injection RX Path - rx_ecrc
+		10) TLPs SEQ# error - tlp_err_seq
+		11) DLLPS ACK/NAK SEQ# error - ack_nak_dllp_seq
+		12) ACK/NAK DLLPs transmission block - ack_nak_dllp
+		13) UpdateFC DLLPs transmission block - upd_fc_dllp
+		14) Always transmission for NAK DLLP - nak_dllp
+		15) Invert SYNC header - inv_sync_hdr_sym
+		16) COM/PAD TS1 order set - com_pad_ts1
+		17) COM/PAD TS2 order set - com_pad_ts2
+		18) COM/FTS FTS order set - com_fts
+		19) COM/IDL E-idle order set - com_idl
+		20) END/EDB symbol - end_edb
+		21) STP/SDP symbol - stp_sdp
+		22) COM/SKP SKP order set - com_skp
+		23) Posted TLP Header credit value control - posted_tlp_hdr
+		24) Non-Posted TLP Header credit value control - non_post_tlp_hdr
+		25) Completion TLP Header credit value control - cmpl_tlp_hdr
+		26) Posted TLP Data credit value control - posted_tlp_data
+		27) Non-Posted TLP Data credit value control - non_post_tlp_data
+		28) Completion TLP Data credit value control - cmpl_tlp_data
+		29) Generates duplicate TLPs - duplicate_dllp
+		30) Generates Nullified TLPs - nullified_tlp
+
+		(WO) Write to the attribute will prepare controller to inject the respective
+		error in the next transmission of data. Parameter required to write will
+		change in the following ways:
+
+		i) Errors 9) - 10) are sequence errors. The write command for these will be
+
+			echo <count> <diff> > /sys/kernel/debug/dwc_pcie_<dev>/rasdes_err_inj/<error>
+
+			<count>
+				Number of errors to be injected
+			<diff>
+				The difference to add or subtract from natural sequence number to
+				generate sequence error. Range (-4095 : 4095)
+
+		ii) Errors 23) - 28) are credit value error insertions. Write command:
+
+			echo <count> <diff> <vc> > /sys/kernel/debug/dwc_pcie_<dev>/rasdes_err_inj/<error>
+
+			<count>
+				Number of errors to be injected
+			<diff>
+				The difference to add or subtract from UpdateFC credit value.
+				Range (-4095 : 4095)
+			<vc>
+				Target VC number
+
+		iii) All other errors. Write command:
+
+			echo <count> > /sys/kernel/debug/dwc_pcie_<dev>/rasdes_err_inj/<error>
+
+			<count>
+				Number of errors to be injected
diff --git a/drivers/pci/controller/dwc/pcie-designware-debugfs.c b/drivers/pci/controller/dwc/pcie-designware-debugfs.c
index 3887a6996706..b7260edd2336 100644
--- a/drivers/pci/controller/dwc/pcie-designware-debugfs.c
+++ b/drivers/pci/controller/dwc/pcie-designware-debugfs.c
@@ -17,6 +17,20 @@
 #define PIPE_DETECT_LANE		BIT(17)
 #define LANE_SELECT			GENMASK(3, 0)
 
+#define ERR_INJ0_OFF			0x34
+#define EINJ_VAL_DIFF			GENMASK(28, 16)
+#define EINJ_VC_NUM			GENMASK(14, 12)
+#define EINJ_TYPE_SHIFT			8
+#define EINJ0_TYPE			GENMASK(11, 8)
+#define EINJ1_TYPE			BIT(8)
+#define EINJ2_TYPE			GENMASK(9, 8)
+#define EINJ3_TYPE			GENMASK(10, 8)
+#define EINJ4_TYPE			GENMASK(10, 8)
+#define EINJ5_TYPE			BIT(8)
+#define EINJ_COUNT			GENMASK(7, 0)
+
+#define ERR_INJ_ENABLE_REG		0x30
+
 #define DWC_DEBUGFS_BUF_MAX		128
 
 /**
@@ -33,6 +47,72 @@ struct dwc_pcie_rasdes_info {
 	struct mutex reg_event_lock;
 };
 
+/**
+ * struct dwc_pcie_rasdes_priv - Stores file specific private data information
+ * @pci: Reference to the dw_pcie structure
+ * @idx: Index to point to specific file related information in array of structs
+ *
+ * All debugfs files will have this struct as its private data.
+ */
+struct dwc_pcie_rasdes_priv {
+	struct dw_pcie *pci;
+	int idx;
+};
+
+/**
+ * struct dwc_pcie_err_inj - Store details about each error injection supported by DWC RASDES
+ * @name: Name of the error that can be injected
+ * @err_inj_group: Group number to which the error belongs to. Value can range from 0 - 5
+ * @err_inj_type: Each group can have multiple types of error
+ */
+struct dwc_pcie_err_inj {
+	const char *name;
+	u32 err_inj_group;
+	u32 err_inj_type;
+};
+
+static const struct dwc_pcie_err_inj err_inj_list[] = {
+	{"tx_lcrc", 0x0, 0x0},
+	{"b16_crc_dllp", 0x0, 0x1},
+	{"b16_crc_upd_fc", 0x0, 0x2},
+	{"tx_ecrc", 0x0, 0x3},
+	{"fcrc_tlp", 0x0, 0x4},
+	{"parity_tsos", 0x0, 0x5},
+	{"parity_skpos", 0x0, 0x6},
+	{"rx_lcrc", 0x0, 0x8},
+	{"rx_ecrc", 0x0, 0xb},
+	{"tlp_err_seq", 0x1, 0x0},
+	{"ack_nak_dllp_seq", 0x1, 0x1},
+	{"ack_nak_dllp", 0x2, 0x0},
+	{"upd_fc_dllp", 0x2, 0x1},
+	{"nak_dllp", 0x2, 0x2},
+	{"inv_sync_hdr_sym", 0x3, 0x0},
+	{"com_pad_ts1", 0x3, 0x1},
+	{"com_pad_ts2", 0x3, 0x2},
+	{"com_fts", 0x3, 0x3},
+	{"com_idl", 0x3, 0x4},
+	{"end_edb", 0x3, 0x5},
+	{"stp_sdp", 0x3, 0x6},
+	{"com_skp", 0x3, 0x7},
+	{"posted_tlp_hdr", 0x4, 0x0},
+	{"non_post_tlp_hdr", 0x4, 0x1},
+	{"cmpl_tlp_hdr", 0x4, 0x2},
+	{"posted_tlp_data", 0x4, 0x4},
+	{"non_post_tlp_data", 0x4, 0x5},
+	{"cmpl_tlp_data", 0x4, 0x6},
+	{"duplicate_dllp", 0x5, 0x0},
+	{"nullified_tlp", 0x5, 0x1},
+};
+
+static const u32 err_inj_type_mask[] = {
+	EINJ0_TYPE,
+	EINJ1_TYPE,
+	EINJ2_TYPE,
+	EINJ3_TYPE,
+	EINJ4_TYPE,
+	EINJ5_TYPE,
+};
+
 static ssize_t lane_detect_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
 {
 	struct dw_pcie *pci = file->private_data;
@@ -93,6 +173,63 @@ static ssize_t rx_valid_write(struct file *file, const char __user *buf, size_t
 	return lane_detect_write(file, buf, count, ppos);
 }
 
+static ssize_t err_inj_write(struct file *file, const char __user *buf, size_t count, loff_t *ppos)
+{
+	struct dwc_pcie_rasdes_priv *pdata = file->private_data;
+	struct dw_pcie *pci = pdata->pci;
+	struct dwc_pcie_rasdes_info *rinfo = pci->debugfs->rasdes_info;
+	u32 val, counter, vc_num, err_group, type_mask;
+	int val_diff = 0;
+	char *kern_buf;
+
+	err_group = err_inj_list[pdata->idx].err_inj_group;
+	type_mask = err_inj_type_mask[err_group];
+
+	kern_buf = memdup_user_nul(buf, count);
+	if (IS_ERR(kern_buf))
+		return PTR_ERR(kern_buf);
+
+	if (err_group == 4) {
+		val = sscanf(kern_buf, "%u %d %u", &counter, &val_diff, &vc_num);
+		if ((val != 3) || (val_diff < -4095 || val_diff > 4095)) {
+			kfree(kern_buf);
+			return -EINVAL;
+		}
+	} else if (err_group == 1) {
+		val = sscanf(kern_buf, "%u %d", &counter, &val_diff);
+		if ((val != 2) || (val_diff < -4095 || val_diff > 4095)) {
+			kfree(kern_buf);
+			return -EINVAL;
+		}
+	} else {
+		val = kstrtou32(kern_buf, 0, &counter);
+		if (val) {
+			kfree(kern_buf);
+			return val;
+		}
+	}
+
+	val = dw_pcie_readl_dbi(pci, rinfo->ras_cap_offset + ERR_INJ0_OFF + (0x4 * err_group));
+	val &= ~(type_mask | EINJ_COUNT);
+	val |= ((err_inj_list[pdata->idx].err_inj_type << EINJ_TYPE_SHIFT) & type_mask);
+	val |= FIELD_PREP(EINJ_COUNT, counter);
+
+	if (err_group == 1 || err_group == 4) {
+		val &= ~(EINJ_VAL_DIFF);
+		val |= FIELD_PREP(EINJ_VAL_DIFF, val_diff);
+	}
+	if (err_group == 4) {
+		val &= ~(EINJ_VC_NUM);
+		val |= FIELD_PREP(EINJ_VC_NUM, vc_num);
+	}
+
+	dw_pcie_writel_dbi(pci, rinfo->ras_cap_offset + ERR_INJ0_OFF + (0x4 * err_group), val);
+	dw_pcie_writel_dbi(pci, rinfo->ras_cap_offset + ERR_INJ_ENABLE_REG, (0x1 << err_group));
+
+	kfree(kern_buf);
+	return count;
+}
+
 #define dwc_debugfs_create(name)			\
 debugfs_create_file(#name, 0644, rasdes_debug, pci,	\
 			&dbg_ ## name ## _fops)
@@ -107,6 +244,11 @@ static const struct file_operations dbg_ ## name ## _fops = {	\
 DWC_DEBUGFS_FOPS(lane_detect);
 DWC_DEBUGFS_FOPS(rx_valid);
 
+static const struct file_operations dwc_pcie_err_inj_ops = {
+	.open = simple_open,
+	.write = err_inj_write,
+};
+
 static void dwc_pcie_rasdes_debugfs_deinit(struct dw_pcie *pci)
 {
 	struct dwc_pcie_rasdes_info *rinfo = pci->debugfs->rasdes_info;
@@ -116,10 +258,11 @@ static void dwc_pcie_rasdes_debugfs_deinit(struct dw_pcie *pci)
 
 static int dwc_pcie_rasdes_debugfs_init(struct dw_pcie *pci, struct dentry *dir)
 {
-	struct dentry *rasdes_debug;
+	struct dentry *rasdes_debug, *rasdes_err_inj;
 	struct dwc_pcie_rasdes_info *rasdes_info;
+	struct dwc_pcie_rasdes_priv *priv_tmp;
 	struct device *dev = pci->dev;
-	int ras_cap;
+	int ras_cap, i, ret;
 
 	ras_cap = dw_pcie_find_rasdes_capability(pci);
 	if (!ras_cap) {
@@ -133,6 +276,7 @@ static int dwc_pcie_rasdes_debugfs_init(struct dw_pcie *pci, struct dentry *dir)
 
 	/* Create subdirectories for Debug, Error injection, Statistics */
 	rasdes_debug = debugfs_create_dir("rasdes_debug", dir);
+	rasdes_err_inj = debugfs_create_dir("rasdes_err_inj", dir);
 
 	mutex_init(&rasdes_info->reg_event_lock);
 	rasdes_info->ras_cap_offset = ras_cap;
@@ -142,7 +286,24 @@ static int dwc_pcie_rasdes_debugfs_init(struct dw_pcie *pci, struct dentry *dir)
 	dwc_debugfs_create(lane_detect);
 	dwc_debugfs_create(rx_valid);
 
+	/* Create debugfs files for Error injection subdirectory */
+	for (i = 0; i < ARRAY_SIZE(err_inj_list); i++) {
+		priv_tmp = devm_kzalloc(dev, sizeof(*priv_tmp), GFP_KERNEL);
+		if (!priv_tmp) {
+			ret = -ENOMEM;
+			goto err_deinit;
+		}
+
+		priv_tmp->idx = i;
+		priv_tmp->pci = pci;
+		debugfs_create_file(err_inj_list[i].name, 0200, rasdes_err_inj, priv_tmp,
+				    &dwc_pcie_err_inj_ops);
+	}
 	return 0;
+
+err_deinit:
+	dwc_pcie_rasdes_debugfs_deinit(pci);
+	return ret;
 }
 
 void dwc_pcie_debugfs_deinit(struct dw_pcie *pci)
-- 
2.17.1

[PATCH v7 5/5] Add debugfs based statistical counter support in DWC

[Shradha Todi]

Add support to provide statistical counter interface to userspace. This set
of debug registers are part of the RASDES feature present in DesignWare
PCIe controllers.

Signed-off-by: Shradha Todi <shradha.t@samsung.com>
---
 Documentation/ABI/testing/debugfs-dwc-pcie    |  61 +++++
 .../controller/dwc/pcie-designware-debugfs.c  | 229 +++++++++++++++++-
 2 files changed, 289 insertions(+), 1 deletion(-)

diff --git a/Documentation/ABI/testing/debugfs-dwc-pcie b/Documentation/ABI/testing/debugfs-dwc-pcie
index 6ee0897fe753..650a89b0511e 100644
--- a/Documentation/ABI/testing/debugfs-dwc-pcie
+++ b/Documentation/ABI/testing/debugfs-dwc-pcie
@@ -81,3 +81,64 @@ Description:	rasdes_err_inj is the directory which can be used to inject errors
 
 			<count>
 				Number of errors to be injected
+
+What:		/sys/kernel/debug/dwc_pcie_<dev>/rasdes_event_counters/<event>/counter_enable
+Date:		Feburary 2025
+Contact:	Shradha Todi <shradha.t@samsung.com>
+Description:	rasdes_event_counters is the directory which can be used to collect
+		statistical data about the number of times a certain event has occurred
+		in the controller. The list of possible events are:
+
+		1) EBUF Overflow
+		2) EBUF Underrun
+		3) Decode Error
+		4) Running Disparity Error
+		5) SKP OS Parity Error
+		6) SYNC Header Error
+		7) Rx Valid De-assertion
+		8) CTL SKP OS Parity Error
+		9) 1st Retimer Parity Error
+		10) 2nd Retimer Parity Error
+		11) Margin CRC and Parity Error
+		12) Detect EI Infer
+		13) Receiver Error
+		14) RX Recovery Req
+		15) N_FTS Timeout
+		16) Framing Error
+		17) Deskew Error
+		18) Framing Error In L0
+		19) Deskew Uncompleted Error
+		20) Bad TLP
+		21) LCRC Error
+		22) Bad DLLP
+		23) Replay Number Rollover
+		24) Replay Timeout
+		25) Rx Nak DLLP
+		26) Tx Nak DLLP
+		27) Retry TLP
+		28) FC Timeout
+		29) Poisoned TLP
+		30) ECRC Error
+		31) Unsupported Request
+		32) Completer Abort
+		33) Completion Timeout
+		34) EBUF SKP Add
+		35) EBUF SKP Del
+
+		(RW) Write 1 to enable the event counter and write 0 to disable the event counter.
+		Read will return whether the counter is currently enabled or disabled. Counter is
+		disabled by default.
+
+What:		/sys/kernel/debug/dwc_pcie_<dev>/rasdes_event_counters/<event>/counter_value
+Date:		Feburary 2025
+Contact:	Shradha Todi <shradha.t@samsung.com>
+Description:	(RO) Read will return the current value of the event counter. To reset the counter,
+		counter should be disabled and enabled back using the 'counter_enable' attribute.
+
+What:		/sys/kernel/debug/dwc_pcie_<dev>/rasdes_event_counters/<event>/lane_select
+Date:		Feburary 2025
+Contact:	Shradha Todi <shradha.t@samsung.com>
+Description:	(RW) Some lanes in the event list are lane specific events. These include
+		events 1) - 11) and 34) - 35).
+		Write lane number for which counter needs to be enabled/disabled/dumped.
+		Read will return the current selected lane number. Lane0 is selected by default.
diff --git a/drivers/pci/controller/dwc/pcie-designware-debugfs.c b/drivers/pci/controller/dwc/pcie-designware-debugfs.c
index b7260edd2336..dca1e9999113 100644
--- a/drivers/pci/controller/dwc/pcie-designware-debugfs.c
+++ b/drivers/pci/controller/dwc/pcie-designware-debugfs.c
@@ -31,6 +31,17 @@
 
 #define ERR_INJ_ENABLE_REG		0x30
 
+#define RAS_DES_EVENT_COUNTER_DATA_REG	0xc
+
+#define RAS_DES_EVENT_COUNTER_CTRL_REG	0x8
+#define EVENT_COUNTER_GROUP_SELECT	GENMASK(27, 24)
+#define EVENT_COUNTER_EVENT_SELECT	GENMASK(23, 16)
+#define EVENT_COUNTER_LANE_SELECT	GENMASK(11, 8)
+#define EVENT_COUNTER_STATUS		BIT(7)
+#define EVENT_COUNTER_ENABLE		GENMASK(4, 2)
+#define PER_EVENT_ON			0x3
+#define PER_EVENT_OFF			0x1
+
 #define DWC_DEBUGFS_BUF_MAX		128
 
 /**
@@ -113,6 +124,61 @@ static const u32 err_inj_type_mask[] = {
 	EINJ5_TYPE,
 };
 
+/**
+ * struct dwc_pcie_event_counter - Store details about each event counter supported in DWC RASDES
+ * @name: Name of the error counter
+ * @group_no: Group number that the event belongs to. Value ranges from 0 - 4
+ * @event_no: Event number of the particular event. Value ranges from -
+ *		Group 0: 0 - 10
+ *		Group 1: 5 - 13
+ *		Group 2: 0 - 7
+ *		Group 3: 0 - 5
+ *		Group 4: 0 - 1
+ */
+struct dwc_pcie_event_counter {
+	const char *name;
+	u32 group_no;
+	u32 event_no;
+};
+
+static const struct dwc_pcie_event_counter event_list[] = {
+	{"ebuf_overflow", 0x0, 0x0},
+	{"ebuf_underrun", 0x0, 0x1},
+	{"decode_err", 0x0, 0x2},
+	{"running_disparity_err", 0x0, 0x3},
+	{"skp_os_parity_err", 0x0, 0x4},
+	{"sync_header_err", 0x0, 0x5},
+	{"rx_valid_deassertion", 0x0, 0x6},
+	{"ctl_skp_os_parity_err", 0x0, 0x7},
+	{"retimer_parity_err_1st", 0x0, 0x8},
+	{"retimer_parity_err_2nd", 0x0, 0x9},
+	{"margin_crc_parity_err", 0x0, 0xA},
+	{"detect_ei_infer", 0x1, 0x5},
+	{"receiver_err", 0x1, 0x6},
+	{"rx_recovery_req", 0x1, 0x7},
+	{"n_fts_timeout", 0x1, 0x8},
+	{"framing_err", 0x1, 0x9},
+	{"deskew_err", 0x1, 0xa},
+	{"framing_err_in_l0", 0x1, 0xc},
+	{"deskew_uncompleted_err", 0x1, 0xd},
+	{"bad_tlp", 0x2, 0x0},
+	{"lcrc_err", 0x2, 0x1},
+	{"bad_dllp", 0x2, 0x2},
+	{"replay_num_rollover", 0x2, 0x3},
+	{"replay_timeout", 0x2, 0x4},
+	{"rx_nak_dllp", 0x2, 0x5},
+	{"tx_nak_dllp", 0x2, 0x6},
+	{"retry_tlp", 0x2, 0x7},
+	{"fc_timeout", 0x3, 0x0},
+	{"poisoned_tlp", 0x3, 0x1},
+	{"ecrc_error", 0x3, 0x2},
+	{"unsupported_request", 0x3, 0x3},
+	{"completer_abort", 0x3, 0x4},
+	{"completion_timeout", 0x3, 0x5},
+	{"ebuf_skp_add", 0x4, 0x0},
+	{"ebuf_skp_del", 0x4, 0x1},
+};
+
 static ssize_t lane_detect_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
 {
 	struct dw_pcie *pci = file->private_data;
@@ -230,6 +296,127 @@ static ssize_t err_inj_write(struct file *file, const char __user *buf, size_t c
 	return count;
 }
 
+static void set_event_number(struct dwc_pcie_rasdes_priv *pdata, struct dw_pcie *pci,
+			     struct dwc_pcie_rasdes_info *rinfo)
+{
+	u32 val;
+
+	val = dw_pcie_readl_dbi(pci, rinfo->ras_cap_offset + RAS_DES_EVENT_COUNTER_CTRL_REG);
+	val &= ~EVENT_COUNTER_ENABLE;
+	val &= ~(EVENT_COUNTER_GROUP_SELECT | EVENT_COUNTER_EVENT_SELECT);
+	val |= FIELD_PREP(EVENT_COUNTER_GROUP_SELECT, event_list[pdata->idx].group_no);
+	val |= FIELD_PREP(EVENT_COUNTER_EVENT_SELECT, event_list[pdata->idx].event_no);
+	dw_pcie_writel_dbi(pci, rinfo->ras_cap_offset + RAS_DES_EVENT_COUNTER_CTRL_REG, val);
+}
+
+static ssize_t counter_enable_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
+{
+	struct dwc_pcie_rasdes_priv *pdata = file->private_data;
+	struct dw_pcie *pci = pdata->pci;
+	struct dwc_pcie_rasdes_info *rinfo = pci->debugfs->rasdes_info;
+	char debugfs_buf[DWC_DEBUGFS_BUF_MAX];
+	ssize_t pos;
+	u32 val;
+
+	mutex_lock(&rinfo->reg_event_lock);
+	set_event_number(pdata, pci, rinfo);
+	val = dw_pcie_readl_dbi(pci, rinfo->ras_cap_offset + RAS_DES_EVENT_COUNTER_CTRL_REG);
+	mutex_unlock(&rinfo->reg_event_lock);
+	val = FIELD_GET(EVENT_COUNTER_STATUS, val);
+	if (val)
+		pos = scnprintf(debugfs_buf, DWC_DEBUGFS_BUF_MAX, "Counter Enabled\n");
+	else
+		pos = scnprintf(debugfs_buf, DWC_DEBUGFS_BUF_MAX, "Counter Disabled\n");
+
+	return simple_read_from_buffer(buf, count, ppos, debugfs_buf, pos);
+}
+
+static ssize_t counter_enable_write(struct file *file, const char __user *buf,
+				    size_t count, loff_t *ppos)
+{
+	struct dwc_pcie_rasdes_priv *pdata = file->private_data;
+	struct dw_pcie *pci = pdata->pci;
+	struct dwc_pcie_rasdes_info *rinfo = pci->debugfs->rasdes_info;
+	u32 val, enable;
+
+	val = kstrtou32_from_user(buf, count, 0, &enable);
+	if (val)
+		return val;
+
+	mutex_lock(&rinfo->reg_event_lock);
+	set_event_number(pdata, pci, rinfo);
+	val = dw_pcie_readl_dbi(pci, rinfo->ras_cap_offset + RAS_DES_EVENT_COUNTER_CTRL_REG);
+	if (enable)
+		val |= FIELD_PREP(EVENT_COUNTER_ENABLE, PER_EVENT_ON);
+	else
+		val |= FIELD_PREP(EVENT_COUNTER_ENABLE, PER_EVENT_OFF);
+
+	dw_pcie_writel_dbi(pci, rinfo->ras_cap_offset + RAS_DES_EVENT_COUNTER_CTRL_REG, val);
+	mutex_unlock(&rinfo->reg_event_lock);
+
+	return count;
+}
+
+static ssize_t counter_lane_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
+{
+	struct dwc_pcie_rasdes_priv *pdata = file->private_data;
+	struct dw_pcie *pci = pdata->pci;
+	struct dwc_pcie_rasdes_info *rinfo = pci->debugfs->rasdes_info;
+	char debugfs_buf[DWC_DEBUGFS_BUF_MAX];
+	ssize_t pos;
+	u32 val;
+
+	mutex_lock(&rinfo->reg_event_lock);
+	set_event_number(pdata, pci, rinfo);
+	val = dw_pcie_readl_dbi(pci, rinfo->ras_cap_offset + RAS_DES_EVENT_COUNTER_CTRL_REG);
+	mutex_unlock(&rinfo->reg_event_lock);
+	val = FIELD_GET(EVENT_COUNTER_LANE_SELECT, val);
+	pos = scnprintf(debugfs_buf, DWC_DEBUGFS_BUF_MAX, "Lane: %d\n", val);
+
+	return simple_read_from_buffer(buf, count, ppos, debugfs_buf, pos);
+}
+
+static ssize_t counter_lane_write(struct file *file, const char __user *buf,
+				  size_t count, loff_t *ppos)
+{
+	struct dwc_pcie_rasdes_priv *pdata = file->private_data;
+	struct dw_pcie *pci = pdata->pci;
+	struct dwc_pcie_rasdes_info *rinfo = pci->debugfs->rasdes_info;
+	u32 val, lane;
+
+	val = kstrtou32_from_user(buf, count, 0, &lane);
+	if (val)
+		return val;
+
+	mutex_lock(&rinfo->reg_event_lock);
+	set_event_number(pdata, pci, rinfo);
+	val = dw_pcie_readl_dbi(pci, rinfo->ras_cap_offset + RAS_DES_EVENT_COUNTER_CTRL_REG);
+	val &= ~(EVENT_COUNTER_LANE_SELECT);
+	val |= FIELD_PREP(EVENT_COUNTER_LANE_SELECT, lane);
+	dw_pcie_writel_dbi(pci, rinfo->ras_cap_offset + RAS_DES_EVENT_COUNTER_CTRL_REG, val);
+	mutex_unlock(&rinfo->reg_event_lock);
+
+	return count;
+}
+
+static ssize_t counter_value_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
+{
+	struct dwc_pcie_rasdes_priv *pdata = file->private_data;
+	struct dw_pcie *pci = pdata->pci;
+	struct dwc_pcie_rasdes_info *rinfo = pci->debugfs->rasdes_info;
+	char debugfs_buf[DWC_DEBUGFS_BUF_MAX];
+	ssize_t pos;
+	u32 val;
+
+	mutex_lock(&rinfo->reg_event_lock);
+	set_event_number(pdata, pci, rinfo);
+	val = dw_pcie_readl_dbi(pci, rinfo->ras_cap_offset + RAS_DES_EVENT_COUNTER_DATA_REG);
+	mutex_unlock(&rinfo->reg_event_lock);
+	pos = scnprintf(debugfs_buf, DWC_DEBUGFS_BUF_MAX, "Counter value: %d\n", val);
+
+	return simple_read_from_buffer(buf, count, ppos, debugfs_buf, pos);
+}
+
 #define dwc_debugfs_create(name)			\
 debugfs_create_file(#name, 0644, rasdes_debug, pci,	\
 			&dbg_ ## name ## _fops)
@@ -249,6 +436,23 @@ static const struct file_operations dwc_pcie_err_inj_ops = {
 	.write = err_inj_write,
 };
 
+static const struct file_operations dwc_pcie_counter_enable_ops = {
+	.open = simple_open,
+	.read = counter_enable_read,
+	.write = counter_enable_write,
+};
+
+static const struct file_operations dwc_pcie_counter_lane_ops = {
+	.open = simple_open,
+	.read = counter_lane_read,
+	.write = counter_lane_write,
+};
+
+static const struct file_operations dwc_pcie_counter_value_ops = {
+	.open = simple_open,
+	.read = counter_value_read,
+};
+
 static void dwc_pcie_rasdes_debugfs_deinit(struct dw_pcie *pci)
 {
 	struct dwc_pcie_rasdes_info *rinfo = pci->debugfs->rasdes_info;
@@ -258,7 +462,7 @@ static void dwc_pcie_rasdes_debugfs_deinit(struct dw_pcie *pci)
 
 static int dwc_pcie_rasdes_debugfs_init(struct dw_pcie *pci, struct dentry *dir)
 {
-	struct dentry *rasdes_debug, *rasdes_err_inj;
+	struct dentry *rasdes_debug, *rasdes_err_inj, *rasdes_event_counter, *rasdes_events;
 	struct dwc_pcie_rasdes_info *rasdes_info;
 	struct dwc_pcie_rasdes_priv *priv_tmp;
 	struct device *dev = pci->dev;
@@ -277,6 +481,7 @@ static int dwc_pcie_rasdes_debugfs_init(struct dw_pcie *pci, struct dentry *dir)
 	/* Create subdirectories for Debug, Error injection, Statistics */
 	rasdes_debug = debugfs_create_dir("rasdes_debug", dir);
 	rasdes_err_inj = debugfs_create_dir("rasdes_err_inj", dir);
+	rasdes_event_counter = debugfs_create_dir("rasdes_event_counter", dir);
 
 	mutex_init(&rasdes_info->reg_event_lock);
 	rasdes_info->ras_cap_offset = ras_cap;
@@ -299,6 +504,28 @@ static int dwc_pcie_rasdes_debugfs_init(struct dw_pcie *pci, struct dentry *dir)
 		debugfs_create_file(err_inj_list[i].name, 0200, rasdes_err_inj, priv_tmp,
 				    &dwc_pcie_err_inj_ops);
 	}
+
+	/* Create debugfs files for Statistical counter subdirectory */
+	for (i = 0; i < ARRAY_SIZE(event_list); i++) {
+		priv_tmp = devm_kzalloc(dev, sizeof(*priv_tmp), GFP_KERNEL);
+		if (!priv_tmp) {
+			ret = -ENOMEM;
+			goto err_deinit;
+		}
+
+		priv_tmp->idx = i;
+		priv_tmp->pci = pci;
+		rasdes_events = debugfs_create_dir(event_list[i].name, rasdes_event_counter);
+		if (event_list[i].group_no == 0 || event_list[i].group_no == 4) {
+			debugfs_create_file("lane_select", 0644, rasdes_events,
+					    priv_tmp, &dwc_pcie_counter_lane_ops);
+		}
+		debugfs_create_file("counter_value", 0444, rasdes_events, priv_tmp,
+				    &dwc_pcie_counter_value_ops);
+		debugfs_create_file("counter_enable", 0644, rasdes_events, priv_tmp,
+				    &dwc_pcie_counter_enable_ops);
+	}
+
 	return 0;
 
 err_deinit:
-- 
2.17.1

Re: [PATCH v7 3/5] Add debugfs based silicon debug support in DWC

[Manivannan Sadhasivam]

On Fri, Feb 21, 2025 at 06:45:46PM +0530, Shradha Todi wrote:
> Add support to provide silicon debug interface to userspace. This set
> of debug registers are part of the RASDES feature present in DesignWare
> PCIe controllers.
> 
> Signed-off-by: Shradha Todi <shradha.t@samsung.com>

Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>

- Mani

> ---
>  Documentation/ABI/testing/debugfs-dwc-pcie    |  13 ++
>  drivers/pci/controller/dwc/Kconfig            |  10 +
>  drivers/pci/controller/dwc/Makefile           |   1 +
>  .../controller/dwc/pcie-designware-debugfs.c  | 176 ++++++++++++++++++
>  .../pci/controller/dwc/pcie-designware-ep.c   |   5 +
>  .../pci/controller/dwc/pcie-designware-host.c |   6 +
>  drivers/pci/controller/dwc/pcie-designware.c  |   6 +
>  drivers/pci/controller/dwc/pcie-designware.h  |  21 +++
>  include/linux/pcie-dwc.h                      |   2 +
>  9 files changed, 240 insertions(+)
>  create mode 100644 Documentation/ABI/testing/debugfs-dwc-pcie
>  create mode 100644 drivers/pci/controller/dwc/pcie-designware-debugfs.c
> 
> diff --git a/Documentation/ABI/testing/debugfs-dwc-pcie b/Documentation/ABI/testing/debugfs-dwc-pcie
> new file mode 100644
> index 000000000000..e8ed34e988ef
> --- /dev/null
> +++ b/Documentation/ABI/testing/debugfs-dwc-pcie
> @@ -0,0 +1,13 @@
> +What:		/sys/kernel/debug/dwc_pcie_<dev>/rasdes_debug/lane_detect
> +Date:		Feburary 2025
> +Contact:	Shradha Todi <shradha.t@samsung.com>
> +Description:	(RW) Write the lane number to be checked for detection.	Read
> +		will return whether PHY indicates receiver detection on the
> +		selected lane. The default selected lane is Lane0.
> +
> +What:		/sys/kernel/debug/dwc_pcie_<dev>/rasdes_debug/rx_valid
> +Date:		Feburary 2025
> +Contact:	Shradha Todi <shradha.t@samsung.com>
> +Description:	(RW) Write the lane number to be checked as valid or invalid. Read
> +		will return the status of PIPE RXVALID signal of the selected lane.
> +		The default selected lane is Lane0.
> diff --git a/drivers/pci/controller/dwc/Kconfig b/drivers/pci/controller/dwc/Kconfig
> index b6d6778b0698..48a10428a492 100644
> --- a/drivers/pci/controller/dwc/Kconfig
> +++ b/drivers/pci/controller/dwc/Kconfig
> @@ -6,6 +6,16 @@ menu "DesignWare-based PCIe controllers"
>  config PCIE_DW
>  	bool
>  
> +config PCIE_DW_DEBUGFS
> +	default y
> +	depends on DEBUG_FS
> +	depends on PCIE_DW_HOST || PCIE_DW_EP
> +	bool "DWC PCIe debugfs entries"
> +	help
> +	  Enables debugfs entries for the DW PCIe Controller. These entries
> +	  provide all debug features related to DW controller including the RAS
> +	  DES features to help in debug, error injection and statistical counters.
> +
>  config PCIE_DW_HOST
>  	bool
>  	select PCIE_DW
> diff --git a/drivers/pci/controller/dwc/Makefile b/drivers/pci/controller/dwc/Makefile
> index a8308d9ea986..54565eedc52c 100644
> --- a/drivers/pci/controller/dwc/Makefile
> +++ b/drivers/pci/controller/dwc/Makefile
> @@ -1,5 +1,6 @@
>  # SPDX-License-Identifier: GPL-2.0
>  obj-$(CONFIG_PCIE_DW) += pcie-designware.o
> +obj-$(CONFIG_PCIE_DW_DEBUGFS) += pcie-designware-debugfs.o
>  obj-$(CONFIG_PCIE_DW_HOST) += pcie-designware-host.o
>  obj-$(CONFIG_PCIE_DW_EP) += pcie-designware-ep.o
>  obj-$(CONFIG_PCIE_DW_PLAT) += pcie-designware-plat.o
> diff --git a/drivers/pci/controller/dwc/pcie-designware-debugfs.c b/drivers/pci/controller/dwc/pcie-designware-debugfs.c
> new file mode 100644
> index 000000000000..3887a6996706
> --- /dev/null
> +++ b/drivers/pci/controller/dwc/pcie-designware-debugfs.c
> @@ -0,0 +1,176 @@
> +// SPDX-License-Identifier: GPL-2.0
> +/*
> + * Synopsys DesignWare PCIe controller debugfs driver
> + *
> + * Copyright (C) 2025 Samsung Electronics Co., Ltd.
> + *		http://www.samsung.com
> + *
> + * Author: Shradha Todi <shradha.t@samsung.com>
> + */
> +
> +#include <linux/debugfs.h>
> +
> +#include "pcie-designware.h"
> +
> +#define SD_STATUS_L1LANE_REG		0xb0
> +#define PIPE_RXVALID			BIT(18)
> +#define PIPE_DETECT_LANE		BIT(17)
> +#define LANE_SELECT			GENMASK(3, 0)
> +
> +#define DWC_DEBUGFS_BUF_MAX		128
> +
> +/**
> + * struct dwc_pcie_rasdes_info - Stores controller common information
> + * @ras_cap_offset: RAS DES vendor specific extended capability offset
> + * @reg_event_lock: Mutex used for RASDES shadow event registers
> + *
> + * Any parameter constant to all files of the debugfs hierarchy for a single controller
> + * will be stored in this struct. It is allocated and assigned to controller specific
> + * struct dw_pcie during initialization.
> + */
> +struct dwc_pcie_rasdes_info {
> +	u32 ras_cap_offset;
> +	struct mutex reg_event_lock;
> +};
> +
> +static ssize_t lane_detect_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
> +{
> +	struct dw_pcie *pci = file->private_data;
> +	struct dwc_pcie_rasdes_info *rinfo = pci->debugfs->rasdes_info;
> +	char debugfs_buf[DWC_DEBUGFS_BUF_MAX];
> +	ssize_t pos;
> +	u32 val;
> +
> +	val = dw_pcie_readl_dbi(pci, rinfo->ras_cap_offset + SD_STATUS_L1LANE_REG);
> +	val = FIELD_GET(PIPE_DETECT_LANE, val);
> +	if (val)
> +		pos = scnprintf(debugfs_buf, DWC_DEBUGFS_BUF_MAX, "Lane Detected\n");
> +	else
> +		pos = scnprintf(debugfs_buf, DWC_DEBUGFS_BUF_MAX, "Lane Undetected\n");
> +
> +	return simple_read_from_buffer(buf, count, ppos, debugfs_buf, pos);
> +}
> +
> +static ssize_t lane_detect_write(struct file *file, const char __user *buf,
> +				 size_t count, loff_t *ppos)
> +{
> +	struct dw_pcie *pci = file->private_data;
> +	struct dwc_pcie_rasdes_info *rinfo = pci->debugfs->rasdes_info;
> +	u32 lane, val;
> +
> +	val = kstrtou32_from_user(buf, count, 0, &lane);
> +	if (val)
> +		return val;
> +
> +	val = dw_pcie_readl_dbi(pci, rinfo->ras_cap_offset + SD_STATUS_L1LANE_REG);
> +	val &= ~(LANE_SELECT);
> +	val |= FIELD_PREP(LANE_SELECT, lane);
> +	dw_pcie_writel_dbi(pci, rinfo->ras_cap_offset + SD_STATUS_L1LANE_REG, val);
> +
> +	return count;
> +}
> +
> +static ssize_t rx_valid_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
> +{
> +	struct dw_pcie *pci = file->private_data;
> +	struct dwc_pcie_rasdes_info *rinfo = pci->debugfs->rasdes_info;
> +	char debugfs_buf[DWC_DEBUGFS_BUF_MAX];
> +	ssize_t pos;
> +	u32 val;
> +
> +	val = dw_pcie_readl_dbi(pci, rinfo->ras_cap_offset + SD_STATUS_L1LANE_REG);
> +	val = FIELD_GET(PIPE_RXVALID, val);
> +	if (val)
> +		pos = scnprintf(debugfs_buf, DWC_DEBUGFS_BUF_MAX, "RX Valid\n");
> +	else
> +		pos = scnprintf(debugfs_buf, DWC_DEBUGFS_BUF_MAX, "RX Invalid\n");
> +
> +	return simple_read_from_buffer(buf, count, ppos, debugfs_buf, pos);
> +}
> +
> +static ssize_t rx_valid_write(struct file *file, const char __user *buf, size_t count, loff_t *ppos)
> +{
> +	return lane_detect_write(file, buf, count, ppos);
> +}
> +
> +#define dwc_debugfs_create(name)			\
> +debugfs_create_file(#name, 0644, rasdes_debug, pci,	\
> +			&dbg_ ## name ## _fops)
> +
> +#define DWC_DEBUGFS_FOPS(name)					\
> +static const struct file_operations dbg_ ## name ## _fops = {	\
> +	.open = simple_open,				\
> +	.read = name ## _read,				\
> +	.write = name ## _write				\
> +}
> +
> +DWC_DEBUGFS_FOPS(lane_detect);
> +DWC_DEBUGFS_FOPS(rx_valid);
> +
> +static void dwc_pcie_rasdes_debugfs_deinit(struct dw_pcie *pci)
> +{
> +	struct dwc_pcie_rasdes_info *rinfo = pci->debugfs->rasdes_info;
> +
> +	mutex_destroy(&rinfo->reg_event_lock);
> +}
> +
> +static int dwc_pcie_rasdes_debugfs_init(struct dw_pcie *pci, struct dentry *dir)
> +{
> +	struct dentry *rasdes_debug;
> +	struct dwc_pcie_rasdes_info *rasdes_info;
> +	struct device *dev = pci->dev;
> +	int ras_cap;
> +
> +	ras_cap = dw_pcie_find_rasdes_capability(pci);
> +	if (!ras_cap) {
> +		dev_dbg(dev, "no RASDES capability available\n");
> +		return -ENODEV;
> +	}
> +
> +	rasdes_info = devm_kzalloc(dev, sizeof(*rasdes_info), GFP_KERNEL);
> +	if (!rasdes_info)
> +		return -ENOMEM;
> +
> +	/* Create subdirectories for Debug, Error injection, Statistics */
> +	rasdes_debug = debugfs_create_dir("rasdes_debug", dir);
> +
> +	mutex_init(&rasdes_info->reg_event_lock);
> +	rasdes_info->ras_cap_offset = ras_cap;
> +	pci->debugfs->rasdes_info = rasdes_info;
> +
> +	/* Create debugfs files for Debug subdirectory */
> +	dwc_debugfs_create(lane_detect);
> +	dwc_debugfs_create(rx_valid);
> +
> +	return 0;
> +}
> +
> +void dwc_pcie_debugfs_deinit(struct dw_pcie *pci)
> +{
> +	dwc_pcie_rasdes_debugfs_deinit(pci);
> +	debugfs_remove_recursive(pci->debugfs->debug_dir);
> +}
> +
> +int dwc_pcie_debugfs_init(struct dw_pcie *pci)
> +{
> +	char dirname[DWC_DEBUGFS_BUF_MAX];
> +	struct device *dev = pci->dev;
> +	struct debugfs_info *debugfs;
> +	struct dentry *dir;
> +	int ret;
> +
> +	/* Create main directory for each platform driver */
> +	snprintf(dirname, DWC_DEBUGFS_BUF_MAX, "dwc_pcie_%s", dev_name(dev));
> +	dir = debugfs_create_dir(dirname, NULL);
> +	debugfs = devm_kzalloc(dev, sizeof(*debugfs), GFP_KERNEL);
> +	if (!debugfs)
> +		return -ENOMEM;
> +
> +	debugfs->debug_dir = dir;
> +	pci->debugfs = debugfs;
> +	ret = dwc_pcie_rasdes_debugfs_init(pci, dir);
> +	if (ret)
> +		dev_dbg(dev, "RASDES debugfs init failed\n");
> +
> +	return 0;
> +}
> diff --git a/drivers/pci/controller/dwc/pcie-designware-ep.c b/drivers/pci/controller/dwc/pcie-designware-ep.c
> index 72418160e658..f9d7f3f989ad 100644
> --- a/drivers/pci/controller/dwc/pcie-designware-ep.c
> +++ b/drivers/pci/controller/dwc/pcie-designware-ep.c
> @@ -814,6 +814,7 @@ void dw_pcie_ep_cleanup(struct dw_pcie_ep *ep)
>  {
>  	struct dw_pcie *pci = to_dw_pcie_from_ep(ep);
>  
> +	dwc_pcie_debugfs_deinit(pci);
>  	dw_pcie_edma_remove(pci);
>  }
>  EXPORT_SYMBOL_GPL(dw_pcie_ep_cleanup);
> @@ -989,6 +990,10 @@ int dw_pcie_ep_init_registers(struct dw_pcie_ep *ep)
>  
>  	dw_pcie_ep_init_non_sticky_registers(pci);
>  
> +	ret = dwc_pcie_debugfs_init(pci);
> +	if (ret)
> +		goto err_remove_edma;
> +
>  	return 0;
>  
>  err_remove_edma:
> diff --git a/drivers/pci/controller/dwc/pcie-designware-host.c b/drivers/pci/controller/dwc/pcie-designware-host.c
> index ffaded8f2df7..2081e8c72d12 100644
> --- a/drivers/pci/controller/dwc/pcie-designware-host.c
> +++ b/drivers/pci/controller/dwc/pcie-designware-host.c
> @@ -548,6 +548,10 @@ int dw_pcie_host_init(struct dw_pcie_rp *pp)
>  	if (pp->ops->post_init)
>  		pp->ops->post_init(pp);
>  
> +	ret = dwc_pcie_debugfs_init(pci);
> +	if (ret)
> +		goto err_stop_link;
> +
>  	return 0;
>  
>  err_stop_link:
> @@ -572,6 +576,8 @@ void dw_pcie_host_deinit(struct dw_pcie_rp *pp)
>  {
>  	struct dw_pcie *pci = to_dw_pcie_from_pp(pp);
>  
> +	dwc_pcie_debugfs_deinit(pci);
> +
>  	pci_stop_root_bus(pp->bridge->bus);
>  	pci_remove_root_bus(pp->bridge->bus);
>  
> diff --git a/drivers/pci/controller/dwc/pcie-designware.c b/drivers/pci/controller/dwc/pcie-designware.c
> index a7c0671c6715..3d1d95d9e380 100644
> --- a/drivers/pci/controller/dwc/pcie-designware.c
> +++ b/drivers/pci/controller/dwc/pcie-designware.c
> @@ -323,6 +323,12 @@ static u16 dw_pcie_find_vsec_capability(struct dw_pcie *pci,
>  	return 0;
>  }
>  
> +u16 dw_pcie_find_rasdes_capability(struct dw_pcie *pci)
> +{
> +	return dw_pcie_find_vsec_capability(pci, dwc_pcie_rasdes_vsec_ids);
> +}
> +EXPORT_SYMBOL_GPL(dw_pcie_find_rasdes_capability);
> +
>  int dw_pcie_read(void __iomem *addr, int size, u32 *val)
>  {
>  	if (!IS_ALIGNED((uintptr_t)addr, size)) {
> diff --git a/drivers/pci/controller/dwc/pcie-designware.h b/drivers/pci/controller/dwc/pcie-designware.h
> index 501d9ddfea16..7f9807d4e5de 100644
> --- a/drivers/pci/controller/dwc/pcie-designware.h
> +++ b/drivers/pci/controller/dwc/pcie-designware.h
> @@ -437,6 +437,11 @@ struct dw_pcie_ops {
>  	void	(*stop_link)(struct dw_pcie *pcie);
>  };
>  
> +struct debugfs_info {
> +	struct dentry		*debug_dir;
> +	void			*rasdes_info;
> +};
> +
>  struct dw_pcie {
>  	struct device		*dev;
>  	void __iomem		*dbi_base;
> @@ -465,6 +470,7 @@ struct dw_pcie {
>  	struct reset_control_bulk_data	core_rsts[DW_PCIE_NUM_CORE_RSTS];
>  	struct gpio_desc		*pe_rst;
>  	bool			suspended;
> +	struct debugfs_info	*debugfs;
>  };
>  
>  #define to_dw_pcie_from_pp(port) container_of((port), struct dw_pcie, pp)
> @@ -478,6 +484,7 @@ void dw_pcie_version_detect(struct dw_pcie *pci);
>  
>  u8 dw_pcie_find_capability(struct dw_pcie *pci, u8 cap);
>  u16 dw_pcie_find_ext_capability(struct dw_pcie *pci, u8 cap);
> +u16 dw_pcie_find_rasdes_capability(struct dw_pcie *pci);
>  
>  int dw_pcie_read(void __iomem *addr, int size, u32 *val);
>  int dw_pcie_write(void __iomem *addr, int size, u32 val);
> @@ -806,4 +813,18 @@ dw_pcie_ep_get_func_from_ep(struct dw_pcie_ep *ep, u8 func_no)
>  	return NULL;
>  }
>  #endif
> +
> +#ifdef CONFIG_PCIE_DW_DEBUGFS
> +int dwc_pcie_debugfs_init(struct dw_pcie *pci);
> +void dwc_pcie_debugfs_deinit(struct dw_pcie *pci);
> +#else
> +static inline int dwc_pcie_debugfs_init(struct dw_pcie *pci)
> +{
> +	return 0;
> +}
> +static inline void dwc_pcie_debugfs_deinit(struct dw_pcie *pci)
> +{
> +}
> +#endif
> +
>  #endif /* _PCIE_DESIGNWARE_H */
> diff --git a/include/linux/pcie-dwc.h b/include/linux/pcie-dwc.h
> index 40f3545731c8..6436e7fadc75 100644
> --- a/include/linux/pcie-dwc.h
> +++ b/include/linux/pcie-dwc.h
> @@ -28,6 +28,8 @@ static const struct dwc_pcie_vsec_id dwc_pcie_rasdes_vsec_ids[] = {
>  	  .vsec_id = 0x02, .vsec_rev = 0x4 },
>  	{ .vendor_id = PCI_VENDOR_ID_QCOM,
>  	  .vsec_id = 0x02, .vsec_rev = 0x4 },
> +	{ .vendor_id = PCI_VENDOR_ID_SAMSUNG,
> +	  .vsec_id = 0x02, .vsec_rev = 0x4 },
>  	{} /* terminator */
>  };
>  
> -- 
> 2.17.1
> 

-- 
மணிவண்ணன் சதாசிவம்

Re: [PATCH v7 4/5] Add debugfs based error injection support in DWC

[Manivannan Sadhasivam]

On Fri, Feb 21, 2025 at 06:45:47PM +0530, Shradha Todi wrote:
> Add support to provide error injection interface to userspace. This set
> of debug registers are part of the RASDES feature present in DesignWare
> PCIe controllers.
> 
> Signed-off-by: Shradha Todi <shradha.t@samsung.com>

Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>

- Mani

> ---
>  Documentation/ABI/testing/debugfs-dwc-pcie    |  70 ++++++++
>  .../controller/dwc/pcie-designware-debugfs.c  | 165 +++++++++++++++++-
>  2 files changed, 233 insertions(+), 2 deletions(-)
> 
> diff --git a/Documentation/ABI/testing/debugfs-dwc-pcie b/Documentation/ABI/testing/debugfs-dwc-pcie
> index e8ed34e988ef..6ee0897fe753 100644
> --- a/Documentation/ABI/testing/debugfs-dwc-pcie
> +++ b/Documentation/ABI/testing/debugfs-dwc-pcie
> @@ -11,3 +11,73 @@ Contact:	Shradha Todi <shradha.t@samsung.com>
>  Description:	(RW) Write the lane number to be checked as valid or invalid. Read
>  		will return the status of PIPE RXVALID signal of the selected lane.
>  		The default selected lane is Lane0.
> +
> +What:		/sys/kernel/debug/dwc_pcie_<dev>/rasdes_err_inj/<error>
> +Date:		Feburary 2025
> +Contact:	Shradha Todi <shradha.t@samsung.com>
> +Description:	rasdes_err_inj is the directory which can be used to inject errors in the
> +		system. The possible errors that can be injected are:
> +
> +		1) TLP LCRC error injection TX Path - tx_lcrc
> +		2) 16b CRC error injection of ACK/NAK DLLP - b16_crc_dllp
> +		3) 16b CRC error injection of Update-FC DLLP - b16_crc_upd_fc
> +		4) TLP ECRC error injection TX Path - tx_ecrc
> +		5) TLP's FCRC error injection TX Path - fcrc_tlp
> +		6) Parity error of TSOS - parity_tsos
> +		7) Parity error on SKPOS - parity_skpos
> +		8) LCRC error injection RX Path - rx_lcrc
> +		9) ECRC error injection RX Path - rx_ecrc
> +		10) TLPs SEQ# error - tlp_err_seq
> +		11) DLLPS ACK/NAK SEQ# error - ack_nak_dllp_seq
> +		12) ACK/NAK DLLPs transmission block - ack_nak_dllp
> +		13) UpdateFC DLLPs transmission block - upd_fc_dllp
> +		14) Always transmission for NAK DLLP - nak_dllp
> +		15) Invert SYNC header - inv_sync_hdr_sym
> +		16) COM/PAD TS1 order set - com_pad_ts1
> +		17) COM/PAD TS2 order set - com_pad_ts2
> +		18) COM/FTS FTS order set - com_fts
> +		19) COM/IDL E-idle order set - com_idl
> +		20) END/EDB symbol - end_edb
> +		21) STP/SDP symbol - stp_sdp
> +		22) COM/SKP SKP order set - com_skp
> +		23) Posted TLP Header credit value control - posted_tlp_hdr
> +		24) Non-Posted TLP Header credit value control - non_post_tlp_hdr
> +		25) Completion TLP Header credit value control - cmpl_tlp_hdr
> +		26) Posted TLP Data credit value control - posted_tlp_data
> +		27) Non-Posted TLP Data credit value control - non_post_tlp_data
> +		28) Completion TLP Data credit value control - cmpl_tlp_data
> +		29) Generates duplicate TLPs - duplicate_dllp
> +		30) Generates Nullified TLPs - nullified_tlp
> +
> +		(WO) Write to the attribute will prepare controller to inject the respective
> +		error in the next transmission of data. Parameter required to write will
> +		change in the following ways:
> +
> +		i) Errors 9) - 10) are sequence errors. The write command for these will be
> +
> +			echo <count> <diff> > /sys/kernel/debug/dwc_pcie_<dev>/rasdes_err_inj/<error>
> +
> +			<count>
> +				Number of errors to be injected
> +			<diff>
> +				The difference to add or subtract from natural sequence number to
> +				generate sequence error. Range (-4095 : 4095)
> +
> +		ii) Errors 23) - 28) are credit value error insertions. Write command:
> +
> +			echo <count> <diff> <vc> > /sys/kernel/debug/dwc_pcie_<dev>/rasdes_err_inj/<error>
> +
> +			<count>
> +				Number of errors to be injected
> +			<diff>
> +				The difference to add or subtract from UpdateFC credit value.
> +				Range (-4095 : 4095)
> +			<vc>
> +				Target VC number
> +
> +		iii) All other errors. Write command:
> +
> +			echo <count> > /sys/kernel/debug/dwc_pcie_<dev>/rasdes_err_inj/<error>
> +
> +			<count>
> +				Number of errors to be injected
> diff --git a/drivers/pci/controller/dwc/pcie-designware-debugfs.c b/drivers/pci/controller/dwc/pcie-designware-debugfs.c
> index 3887a6996706..b7260edd2336 100644
> --- a/drivers/pci/controller/dwc/pcie-designware-debugfs.c
> +++ b/drivers/pci/controller/dwc/pcie-designware-debugfs.c
> @@ -17,6 +17,20 @@
>  #define PIPE_DETECT_LANE		BIT(17)
>  #define LANE_SELECT			GENMASK(3, 0)
>  
> +#define ERR_INJ0_OFF			0x34
> +#define EINJ_VAL_DIFF			GENMASK(28, 16)
> +#define EINJ_VC_NUM			GENMASK(14, 12)
> +#define EINJ_TYPE_SHIFT			8
> +#define EINJ0_TYPE			GENMASK(11, 8)
> +#define EINJ1_TYPE			BIT(8)
> +#define EINJ2_TYPE			GENMASK(9, 8)
> +#define EINJ3_TYPE			GENMASK(10, 8)
> +#define EINJ4_TYPE			GENMASK(10, 8)
> +#define EINJ5_TYPE			BIT(8)
> +#define EINJ_COUNT			GENMASK(7, 0)
> +
> +#define ERR_INJ_ENABLE_REG		0x30
> +
>  #define DWC_DEBUGFS_BUF_MAX		128
>  
>  /**
> @@ -33,6 +47,72 @@ struct dwc_pcie_rasdes_info {
>  	struct mutex reg_event_lock;
>  };
>  
> +/**
> + * struct dwc_pcie_rasdes_priv - Stores file specific private data information
> + * @pci: Reference to the dw_pcie structure
> + * @idx: Index to point to specific file related information in array of structs
> + *
> + * All debugfs files will have this struct as its private data.
> + */
> +struct dwc_pcie_rasdes_priv {
> +	struct dw_pcie *pci;
> +	int idx;
> +};
> +
> +/**
> + * struct dwc_pcie_err_inj - Store details about each error injection supported by DWC RASDES
> + * @name: Name of the error that can be injected
> + * @err_inj_group: Group number to which the error belongs to. Value can range from 0 - 5
> + * @err_inj_type: Each group can have multiple types of error
> + */
> +struct dwc_pcie_err_inj {
> +	const char *name;
> +	u32 err_inj_group;
> +	u32 err_inj_type;
> +};
> +
> +static const struct dwc_pcie_err_inj err_inj_list[] = {
> +	{"tx_lcrc", 0x0, 0x0},
> +	{"b16_crc_dllp", 0x0, 0x1},
> +	{"b16_crc_upd_fc", 0x0, 0x2},
> +	{"tx_ecrc", 0x0, 0x3},
> +	{"fcrc_tlp", 0x0, 0x4},
> +	{"parity_tsos", 0x0, 0x5},
> +	{"parity_skpos", 0x0, 0x6},
> +	{"rx_lcrc", 0x0, 0x8},
> +	{"rx_ecrc", 0x0, 0xb},
> +	{"tlp_err_seq", 0x1, 0x0},
> +	{"ack_nak_dllp_seq", 0x1, 0x1},
> +	{"ack_nak_dllp", 0x2, 0x0},
> +	{"upd_fc_dllp", 0x2, 0x1},
> +	{"nak_dllp", 0x2, 0x2},
> +	{"inv_sync_hdr_sym", 0x3, 0x0},
> +	{"com_pad_ts1", 0x3, 0x1},
> +	{"com_pad_ts2", 0x3, 0x2},
> +	{"com_fts", 0x3, 0x3},
> +	{"com_idl", 0x3, 0x4},
> +	{"end_edb", 0x3, 0x5},
> +	{"stp_sdp", 0x3, 0x6},
> +	{"com_skp", 0x3, 0x7},
> +	{"posted_tlp_hdr", 0x4, 0x0},
> +	{"non_post_tlp_hdr", 0x4, 0x1},
> +	{"cmpl_tlp_hdr", 0x4, 0x2},
> +	{"posted_tlp_data", 0x4, 0x4},
> +	{"non_post_tlp_data", 0x4, 0x5},
> +	{"cmpl_tlp_data", 0x4, 0x6},
> +	{"duplicate_dllp", 0x5, 0x0},
> +	{"nullified_tlp", 0x5, 0x1},
> +};
> +
> +static const u32 err_inj_type_mask[] = {
> +	EINJ0_TYPE,
> +	EINJ1_TYPE,
> +	EINJ2_TYPE,
> +	EINJ3_TYPE,
> +	EINJ4_TYPE,
> +	EINJ5_TYPE,
> +};
> +
>  static ssize_t lane_detect_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
>  {
>  	struct dw_pcie *pci = file->private_data;
> @@ -93,6 +173,63 @@ static ssize_t rx_valid_write(struct file *file, const char __user *buf, size_t
>  	return lane_detect_write(file, buf, count, ppos);
>  }
>  
> +static ssize_t err_inj_write(struct file *file, const char __user *buf, size_t count, loff_t *ppos)
> +{
> +	struct dwc_pcie_rasdes_priv *pdata = file->private_data;
> +	struct dw_pcie *pci = pdata->pci;
> +	struct dwc_pcie_rasdes_info *rinfo = pci->debugfs->rasdes_info;
> +	u32 val, counter, vc_num, err_group, type_mask;
> +	int val_diff = 0;
> +	char *kern_buf;
> +
> +	err_group = err_inj_list[pdata->idx].err_inj_group;
> +	type_mask = err_inj_type_mask[err_group];
> +
> +	kern_buf = memdup_user_nul(buf, count);
> +	if (IS_ERR(kern_buf))
> +		return PTR_ERR(kern_buf);
> +
> +	if (err_group == 4) {
> +		val = sscanf(kern_buf, "%u %d %u", &counter, &val_diff, &vc_num);
> +		if ((val != 3) || (val_diff < -4095 || val_diff > 4095)) {
> +			kfree(kern_buf);
> +			return -EINVAL;
> +		}
> +	} else if (err_group == 1) {
> +		val = sscanf(kern_buf, "%u %d", &counter, &val_diff);
> +		if ((val != 2) || (val_diff < -4095 || val_diff > 4095)) {
> +			kfree(kern_buf);
> +			return -EINVAL;
> +		}
> +	} else {
> +		val = kstrtou32(kern_buf, 0, &counter);
> +		if (val) {
> +			kfree(kern_buf);
> +			return val;
> +		}
> +	}
> +
> +	val = dw_pcie_readl_dbi(pci, rinfo->ras_cap_offset + ERR_INJ0_OFF + (0x4 * err_group));
> +	val &= ~(type_mask | EINJ_COUNT);
> +	val |= ((err_inj_list[pdata->idx].err_inj_type << EINJ_TYPE_SHIFT) & type_mask);
> +	val |= FIELD_PREP(EINJ_COUNT, counter);
> +
> +	if (err_group == 1 || err_group == 4) {
> +		val &= ~(EINJ_VAL_DIFF);
> +		val |= FIELD_PREP(EINJ_VAL_DIFF, val_diff);
> +	}
> +	if (err_group == 4) {
> +		val &= ~(EINJ_VC_NUM);
> +		val |= FIELD_PREP(EINJ_VC_NUM, vc_num);
> +	}
> +
> +	dw_pcie_writel_dbi(pci, rinfo->ras_cap_offset + ERR_INJ0_OFF + (0x4 * err_group), val);
> +	dw_pcie_writel_dbi(pci, rinfo->ras_cap_offset + ERR_INJ_ENABLE_REG, (0x1 << err_group));
> +
> +	kfree(kern_buf);
> +	return count;
> +}
> +
>  #define dwc_debugfs_create(name)			\
>  debugfs_create_file(#name, 0644, rasdes_debug, pci,	\
>  			&dbg_ ## name ## _fops)
> @@ -107,6 +244,11 @@ static const struct file_operations dbg_ ## name ## _fops = {	\
>  DWC_DEBUGFS_FOPS(lane_detect);
>  DWC_DEBUGFS_FOPS(rx_valid);
>  
> +static const struct file_operations dwc_pcie_err_inj_ops = {
> +	.open = simple_open,
> +	.write = err_inj_write,
> +};
> +
>  static void dwc_pcie_rasdes_debugfs_deinit(struct dw_pcie *pci)
>  {
>  	struct dwc_pcie_rasdes_info *rinfo = pci->debugfs->rasdes_info;
> @@ -116,10 +258,11 @@ static void dwc_pcie_rasdes_debugfs_deinit(struct dw_pcie *pci)
>  
>  static int dwc_pcie_rasdes_debugfs_init(struct dw_pcie *pci, struct dentry *dir)
>  {
> -	struct dentry *rasdes_debug;
> +	struct dentry *rasdes_debug, *rasdes_err_inj;
>  	struct dwc_pcie_rasdes_info *rasdes_info;
> +	struct dwc_pcie_rasdes_priv *priv_tmp;
>  	struct device *dev = pci->dev;
> -	int ras_cap;
> +	int ras_cap, i, ret;
>  
>  	ras_cap = dw_pcie_find_rasdes_capability(pci);
>  	if (!ras_cap) {
> @@ -133,6 +276,7 @@ static int dwc_pcie_rasdes_debugfs_init(struct dw_pcie *pci, struct dentry *dir)
>  
>  	/* Create subdirectories for Debug, Error injection, Statistics */
>  	rasdes_debug = debugfs_create_dir("rasdes_debug", dir);
> +	rasdes_err_inj = debugfs_create_dir("rasdes_err_inj", dir);
>  
>  	mutex_init(&rasdes_info->reg_event_lock);
>  	rasdes_info->ras_cap_offset = ras_cap;
> @@ -142,7 +286,24 @@ static int dwc_pcie_rasdes_debugfs_init(struct dw_pcie *pci, struct dentry *dir)
>  	dwc_debugfs_create(lane_detect);
>  	dwc_debugfs_create(rx_valid);
>  
> +	/* Create debugfs files for Error injection subdirectory */
> +	for (i = 0; i < ARRAY_SIZE(err_inj_list); i++) {
> +		priv_tmp = devm_kzalloc(dev, sizeof(*priv_tmp), GFP_KERNEL);
> +		if (!priv_tmp) {
> +			ret = -ENOMEM;
> +			goto err_deinit;
> +		}
> +
> +		priv_tmp->idx = i;
> +		priv_tmp->pci = pci;
> +		debugfs_create_file(err_inj_list[i].name, 0200, rasdes_err_inj, priv_tmp,
> +				    &dwc_pcie_err_inj_ops);
> +	}
>  	return 0;
> +
> +err_deinit:
> +	dwc_pcie_rasdes_debugfs_deinit(pci);
> +	return ret;
>  }
>  
>  void dwc_pcie_debugfs_deinit(struct dw_pcie *pci)
> -- 
> 2.17.1
> 

-- 
மணிவண்ணன் சதாசிவம்

Re: [PATCH v7 5/5] Add debugfs based statistical counter support in DWC

[Manivannan Sadhasivam]

On Fri, Feb 21, 2025 at 06:45:48PM +0530, Shradha Todi wrote:
> Add support to provide statistical counter interface to userspace. This set
> of debug registers are part of the RASDES feature present in DesignWare
> PCIe controllers.
> 
> Signed-off-by: Shradha Todi <shradha.t@samsung.com>

Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>

- Mani

> ---
>  Documentation/ABI/testing/debugfs-dwc-pcie    |  61 +++++
>  .../controller/dwc/pcie-designware-debugfs.c  | 229 +++++++++++++++++-
>  2 files changed, 289 insertions(+), 1 deletion(-)
> 
> diff --git a/Documentation/ABI/testing/debugfs-dwc-pcie b/Documentation/ABI/testing/debugfs-dwc-pcie
> index 6ee0897fe753..650a89b0511e 100644
> --- a/Documentation/ABI/testing/debugfs-dwc-pcie
> +++ b/Documentation/ABI/testing/debugfs-dwc-pcie
> @@ -81,3 +81,64 @@ Description:	rasdes_err_inj is the directory which can be used to inject errors
>  
>  			<count>
>  				Number of errors to be injected
> +
> +What:		/sys/kernel/debug/dwc_pcie_<dev>/rasdes_event_counters/<event>/counter_enable
> +Date:		Feburary 2025
> +Contact:	Shradha Todi <shradha.t@samsung.com>
> +Description:	rasdes_event_counters is the directory which can be used to collect
> +		statistical data about the number of times a certain event has occurred
> +		in the controller. The list of possible events are:
> +
> +		1) EBUF Overflow
> +		2) EBUF Underrun
> +		3) Decode Error
> +		4) Running Disparity Error
> +		5) SKP OS Parity Error
> +		6) SYNC Header Error
> +		7) Rx Valid De-assertion
> +		8) CTL SKP OS Parity Error
> +		9) 1st Retimer Parity Error
> +		10) 2nd Retimer Parity Error
> +		11) Margin CRC and Parity Error
> +		12) Detect EI Infer
> +		13) Receiver Error
> +		14) RX Recovery Req
> +		15) N_FTS Timeout
> +		16) Framing Error
> +		17) Deskew Error
> +		18) Framing Error In L0
> +		19) Deskew Uncompleted Error
> +		20) Bad TLP
> +		21) LCRC Error
> +		22) Bad DLLP
> +		23) Replay Number Rollover
> +		24) Replay Timeout
> +		25) Rx Nak DLLP
> +		26) Tx Nak DLLP
> +		27) Retry TLP
> +		28) FC Timeout
> +		29) Poisoned TLP
> +		30) ECRC Error
> +		31) Unsupported Request
> +		32) Completer Abort
> +		33) Completion Timeout
> +		34) EBUF SKP Add
> +		35) EBUF SKP Del
> +
> +		(RW) Write 1 to enable the event counter and write 0 to disable the event counter.
> +		Read will return whether the counter is currently enabled or disabled. Counter is
> +		disabled by default.
> +
> +What:		/sys/kernel/debug/dwc_pcie_<dev>/rasdes_event_counters/<event>/counter_value
> +Date:		Feburary 2025
> +Contact:	Shradha Todi <shradha.t@samsung.com>
> +Description:	(RO) Read will return the current value of the event counter. To reset the counter,
> +		counter should be disabled and enabled back using the 'counter_enable' attribute.
> +
> +What:		/sys/kernel/debug/dwc_pcie_<dev>/rasdes_event_counters/<event>/lane_select
> +Date:		Feburary 2025
> +Contact:	Shradha Todi <shradha.t@samsung.com>
> +Description:	(RW) Some lanes in the event list are lane specific events. These include
> +		events 1) - 11) and 34) - 35).
> +		Write lane number for which counter needs to be enabled/disabled/dumped.
> +		Read will return the current selected lane number. Lane0 is selected by default.
> diff --git a/drivers/pci/controller/dwc/pcie-designware-debugfs.c b/drivers/pci/controller/dwc/pcie-designware-debugfs.c
> index b7260edd2336..dca1e9999113 100644
> --- a/drivers/pci/controller/dwc/pcie-designware-debugfs.c
> +++ b/drivers/pci/controller/dwc/pcie-designware-debugfs.c
> @@ -31,6 +31,17 @@
>  
>  #define ERR_INJ_ENABLE_REG		0x30
>  
> +#define RAS_DES_EVENT_COUNTER_DATA_REG	0xc
> +
> +#define RAS_DES_EVENT_COUNTER_CTRL_REG	0x8
> +#define EVENT_COUNTER_GROUP_SELECT	GENMASK(27, 24)
> +#define EVENT_COUNTER_EVENT_SELECT	GENMASK(23, 16)
> +#define EVENT_COUNTER_LANE_SELECT	GENMASK(11, 8)
> +#define EVENT_COUNTER_STATUS		BIT(7)
> +#define EVENT_COUNTER_ENABLE		GENMASK(4, 2)
> +#define PER_EVENT_ON			0x3
> +#define PER_EVENT_OFF			0x1
> +
>  #define DWC_DEBUGFS_BUF_MAX		128
>  
>  /**
> @@ -113,6 +124,61 @@ static const u32 err_inj_type_mask[] = {
>  	EINJ5_TYPE,
>  };
>  
> +/**
> + * struct dwc_pcie_event_counter - Store details about each event counter supported in DWC RASDES
> + * @name: Name of the error counter
> + * @group_no: Group number that the event belongs to. Value ranges from 0 - 4
> + * @event_no: Event number of the particular event. Value ranges from -
> + *		Group 0: 0 - 10
> + *		Group 1: 5 - 13
> + *		Group 2: 0 - 7
> + *		Group 3: 0 - 5
> + *		Group 4: 0 - 1
> + */
> +struct dwc_pcie_event_counter {
> +	const char *name;
> +	u32 group_no;
> +	u32 event_no;
> +};
> +
> +static const struct dwc_pcie_event_counter event_list[] = {
> +	{"ebuf_overflow", 0x0, 0x0},
> +	{"ebuf_underrun", 0x0, 0x1},
> +	{"decode_err", 0x0, 0x2},
> +	{"running_disparity_err", 0x0, 0x3},
> +	{"skp_os_parity_err", 0x0, 0x4},
> +	{"sync_header_err", 0x0, 0x5},
> +	{"rx_valid_deassertion", 0x0, 0x6},
> +	{"ctl_skp_os_parity_err", 0x0, 0x7},
> +	{"retimer_parity_err_1st", 0x0, 0x8},
> +	{"retimer_parity_err_2nd", 0x0, 0x9},
> +	{"margin_crc_parity_err", 0x0, 0xA},
> +	{"detect_ei_infer", 0x1, 0x5},
> +	{"receiver_err", 0x1, 0x6},
> +	{"rx_recovery_req", 0x1, 0x7},
> +	{"n_fts_timeout", 0x1, 0x8},
> +	{"framing_err", 0x1, 0x9},
> +	{"deskew_err", 0x1, 0xa},
> +	{"framing_err_in_l0", 0x1, 0xc},
> +	{"deskew_uncompleted_err", 0x1, 0xd},
> +	{"bad_tlp", 0x2, 0x0},
> +	{"lcrc_err", 0x2, 0x1},
> +	{"bad_dllp", 0x2, 0x2},
> +	{"replay_num_rollover", 0x2, 0x3},
> +	{"replay_timeout", 0x2, 0x4},
> +	{"rx_nak_dllp", 0x2, 0x5},
> +	{"tx_nak_dllp", 0x2, 0x6},
> +	{"retry_tlp", 0x2, 0x7},
> +	{"fc_timeout", 0x3, 0x0},
> +	{"poisoned_tlp", 0x3, 0x1},
> +	{"ecrc_error", 0x3, 0x2},
> +	{"unsupported_request", 0x3, 0x3},
> +	{"completer_abort", 0x3, 0x4},
> +	{"completion_timeout", 0x3, 0x5},
> +	{"ebuf_skp_add", 0x4, 0x0},
> +	{"ebuf_skp_del", 0x4, 0x1},
> +};
> +
>  static ssize_t lane_detect_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
>  {
>  	struct dw_pcie *pci = file->private_data;
> @@ -230,6 +296,127 @@ static ssize_t err_inj_write(struct file *file, const char __user *buf, size_t c
>  	return count;
>  }
>  
> +static void set_event_number(struct dwc_pcie_rasdes_priv *pdata, struct dw_pcie *pci,
> +			     struct dwc_pcie_rasdes_info *rinfo)
> +{
> +	u32 val;
> +
> +	val = dw_pcie_readl_dbi(pci, rinfo->ras_cap_offset + RAS_DES_EVENT_COUNTER_CTRL_REG);
> +	val &= ~EVENT_COUNTER_ENABLE;
> +	val &= ~(EVENT_COUNTER_GROUP_SELECT | EVENT_COUNTER_EVENT_SELECT);
> +	val |= FIELD_PREP(EVENT_COUNTER_GROUP_SELECT, event_list[pdata->idx].group_no);
> +	val |= FIELD_PREP(EVENT_COUNTER_EVENT_SELECT, event_list[pdata->idx].event_no);
> +	dw_pcie_writel_dbi(pci, rinfo->ras_cap_offset + RAS_DES_EVENT_COUNTER_CTRL_REG, val);
> +}
> +
> +static ssize_t counter_enable_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
> +{
> +	struct dwc_pcie_rasdes_priv *pdata = file->private_data;
> +	struct dw_pcie *pci = pdata->pci;
> +	struct dwc_pcie_rasdes_info *rinfo = pci->debugfs->rasdes_info;
> +	char debugfs_buf[DWC_DEBUGFS_BUF_MAX];
> +	ssize_t pos;
> +	u32 val;
> +
> +	mutex_lock(&rinfo->reg_event_lock);
> +	set_event_number(pdata, pci, rinfo);
> +	val = dw_pcie_readl_dbi(pci, rinfo->ras_cap_offset + RAS_DES_EVENT_COUNTER_CTRL_REG);
> +	mutex_unlock(&rinfo->reg_event_lock);
> +	val = FIELD_GET(EVENT_COUNTER_STATUS, val);
> +	if (val)
> +		pos = scnprintf(debugfs_buf, DWC_DEBUGFS_BUF_MAX, "Counter Enabled\n");
> +	else
> +		pos = scnprintf(debugfs_buf, DWC_DEBUGFS_BUF_MAX, "Counter Disabled\n");
> +
> +	return simple_read_from_buffer(buf, count, ppos, debugfs_buf, pos);
> +}
> +
> +static ssize_t counter_enable_write(struct file *file, const char __user *buf,
> +				    size_t count, loff_t *ppos)
> +{
> +	struct dwc_pcie_rasdes_priv *pdata = file->private_data;
> +	struct dw_pcie *pci = pdata->pci;
> +	struct dwc_pcie_rasdes_info *rinfo = pci->debugfs->rasdes_info;
> +	u32 val, enable;
> +
> +	val = kstrtou32_from_user(buf, count, 0, &enable);
> +	if (val)
> +		return val;
> +
> +	mutex_lock(&rinfo->reg_event_lock);
> +	set_event_number(pdata, pci, rinfo);
> +	val = dw_pcie_readl_dbi(pci, rinfo->ras_cap_offset + RAS_DES_EVENT_COUNTER_CTRL_REG);
> +	if (enable)
> +		val |= FIELD_PREP(EVENT_COUNTER_ENABLE, PER_EVENT_ON);
> +	else
> +		val |= FIELD_PREP(EVENT_COUNTER_ENABLE, PER_EVENT_OFF);
> +
> +	dw_pcie_writel_dbi(pci, rinfo->ras_cap_offset + RAS_DES_EVENT_COUNTER_CTRL_REG, val);
> +	mutex_unlock(&rinfo->reg_event_lock);
> +
> +	return count;
> +}
> +
> +static ssize_t counter_lane_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
> +{
> +	struct dwc_pcie_rasdes_priv *pdata = file->private_data;
> +	struct dw_pcie *pci = pdata->pci;
> +	struct dwc_pcie_rasdes_info *rinfo = pci->debugfs->rasdes_info;
> +	char debugfs_buf[DWC_DEBUGFS_BUF_MAX];
> +	ssize_t pos;
> +	u32 val;
> +
> +	mutex_lock(&rinfo->reg_event_lock);
> +	set_event_number(pdata, pci, rinfo);
> +	val = dw_pcie_readl_dbi(pci, rinfo->ras_cap_offset + RAS_DES_EVENT_COUNTER_CTRL_REG);
> +	mutex_unlock(&rinfo->reg_event_lock);
> +	val = FIELD_GET(EVENT_COUNTER_LANE_SELECT, val);
> +	pos = scnprintf(debugfs_buf, DWC_DEBUGFS_BUF_MAX, "Lane: %d\n", val);
> +
> +	return simple_read_from_buffer(buf, count, ppos, debugfs_buf, pos);
> +}
> +
> +static ssize_t counter_lane_write(struct file *file, const char __user *buf,
> +				  size_t count, loff_t *ppos)
> +{
> +	struct dwc_pcie_rasdes_priv *pdata = file->private_data;
> +	struct dw_pcie *pci = pdata->pci;
> +	struct dwc_pcie_rasdes_info *rinfo = pci->debugfs->rasdes_info;
> +	u32 val, lane;
> +
> +	val = kstrtou32_from_user(buf, count, 0, &lane);
> +	if (val)
> +		return val;
> +
> +	mutex_lock(&rinfo->reg_event_lock);
> +	set_event_number(pdata, pci, rinfo);
> +	val = dw_pcie_readl_dbi(pci, rinfo->ras_cap_offset + RAS_DES_EVENT_COUNTER_CTRL_REG);
> +	val &= ~(EVENT_COUNTER_LANE_SELECT);
> +	val |= FIELD_PREP(EVENT_COUNTER_LANE_SELECT, lane);
> +	dw_pcie_writel_dbi(pci, rinfo->ras_cap_offset + RAS_DES_EVENT_COUNTER_CTRL_REG, val);
> +	mutex_unlock(&rinfo->reg_event_lock);
> +
> +	return count;
> +}
> +
> +static ssize_t counter_value_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
> +{
> +	struct dwc_pcie_rasdes_priv *pdata = file->private_data;
> +	struct dw_pcie *pci = pdata->pci;
> +	struct dwc_pcie_rasdes_info *rinfo = pci->debugfs->rasdes_info;
> +	char debugfs_buf[DWC_DEBUGFS_BUF_MAX];
> +	ssize_t pos;
> +	u32 val;
> +
> +	mutex_lock(&rinfo->reg_event_lock);
> +	set_event_number(pdata, pci, rinfo);
> +	val = dw_pcie_readl_dbi(pci, rinfo->ras_cap_offset + RAS_DES_EVENT_COUNTER_DATA_REG);
> +	mutex_unlock(&rinfo->reg_event_lock);
> +	pos = scnprintf(debugfs_buf, DWC_DEBUGFS_BUF_MAX, "Counter value: %d\n", val);
> +
> +	return simple_read_from_buffer(buf, count, ppos, debugfs_buf, pos);
> +}
> +
>  #define dwc_debugfs_create(name)			\
>  debugfs_create_file(#name, 0644, rasdes_debug, pci,	\
>  			&dbg_ ## name ## _fops)
> @@ -249,6 +436,23 @@ static const struct file_operations dwc_pcie_err_inj_ops = {
>  	.write = err_inj_write,
>  };
>  
> +static const struct file_operations dwc_pcie_counter_enable_ops = {
> +	.open = simple_open,
> +	.read = counter_enable_read,
> +	.write = counter_enable_write,
> +};
> +
> +static const struct file_operations dwc_pcie_counter_lane_ops = {
> +	.open = simple_open,
> +	.read = counter_lane_read,
> +	.write = counter_lane_write,
> +};
> +
> +static const struct file_operations dwc_pcie_counter_value_ops = {
> +	.open = simple_open,
> +	.read = counter_value_read,
> +};
> +
>  static void dwc_pcie_rasdes_debugfs_deinit(struct dw_pcie *pci)
>  {
>  	struct dwc_pcie_rasdes_info *rinfo = pci->debugfs->rasdes_info;
> @@ -258,7 +462,7 @@ static void dwc_pcie_rasdes_debugfs_deinit(struct dw_pcie *pci)
>  
>  static int dwc_pcie_rasdes_debugfs_init(struct dw_pcie *pci, struct dentry *dir)
>  {
> -	struct dentry *rasdes_debug, *rasdes_err_inj;
> +	struct dentry *rasdes_debug, *rasdes_err_inj, *rasdes_event_counter, *rasdes_events;
>  	struct dwc_pcie_rasdes_info *rasdes_info;
>  	struct dwc_pcie_rasdes_priv *priv_tmp;
>  	struct device *dev = pci->dev;
> @@ -277,6 +481,7 @@ static int dwc_pcie_rasdes_debugfs_init(struct dw_pcie *pci, struct dentry *dir)
>  	/* Create subdirectories for Debug, Error injection, Statistics */
>  	rasdes_debug = debugfs_create_dir("rasdes_debug", dir);
>  	rasdes_err_inj = debugfs_create_dir("rasdes_err_inj", dir);
> +	rasdes_event_counter = debugfs_create_dir("rasdes_event_counter", dir);
>  
>  	mutex_init(&rasdes_info->reg_event_lock);
>  	rasdes_info->ras_cap_offset = ras_cap;
> @@ -299,6 +504,28 @@ static int dwc_pcie_rasdes_debugfs_init(struct dw_pcie *pci, struct dentry *dir)
>  		debugfs_create_file(err_inj_list[i].name, 0200, rasdes_err_inj, priv_tmp,
>  				    &dwc_pcie_err_inj_ops);
>  	}
> +
> +	/* Create debugfs files for Statistical counter subdirectory */
> +	for (i = 0; i < ARRAY_SIZE(event_list); i++) {
> +		priv_tmp = devm_kzalloc(dev, sizeof(*priv_tmp), GFP_KERNEL);
> +		if (!priv_tmp) {
> +			ret = -ENOMEM;
> +			goto err_deinit;
> +		}
> +
> +		priv_tmp->idx = i;
> +		priv_tmp->pci = pci;
> +		rasdes_events = debugfs_create_dir(event_list[i].name, rasdes_event_counter);
> +		if (event_list[i].group_no == 0 || event_list[i].group_no == 4) {
> +			debugfs_create_file("lane_select", 0644, rasdes_events,
> +					    priv_tmp, &dwc_pcie_counter_lane_ops);
> +		}
> +		debugfs_create_file("counter_value", 0444, rasdes_events, priv_tmp,
> +				    &dwc_pcie_counter_value_ops);
> +		debugfs_create_file("counter_enable", 0644, rasdes_events, priv_tmp,
> +				    &dwc_pcie_counter_enable_ops);
> +	}
> +
>  	return 0;
>  
>  err_deinit:
> -- 
> 2.17.1
> 

-- 
மணிவண்ணன் சதாசிவம்

Re: [PATCH v7 0/5] Add support for debugfs based RAS DES feature in PCIe DW

[Niklas Cassel]

Hello Shradha,

On Fri, Feb 21, 2025 at 06:45:43PM +0530, Shradha Todi wrote:
> DesignWare controller provides a vendor specific extended capability
> called RASDES as an IP feature. This extended capability  provides
> hardware information like:
>  - Debug registers to know the state of the link or controller. 
>  - Error injection mechanisms to inject various PCIe errors including
>    sequence number, CRC
>  - Statistical counters to know how many times a particular event
>    occurred
> 
> However, in Linux we do not have any generic or custom support to be
> able to use this feature in an efficient manner. This is the reason we
> are proposing this framework. Debug and bring up time of high-speed IPs
> are highly dependent on costlier hardware analyzers and this solution
> will in some ways help to reduce the HW analyzer usage.
> 
> The debugfs entries can be used to get information about underlying
> hardware and can be shared with user space. Separate debugfs entries has
> been created to cater to all the DES hooks provided by the controller.
> The debugfs entries interacts with the RASDES registers in the required
> sequence and provides the meaningful data to the user. This eases the
> effort to understand and use the register information for debugging.
> 
> This series creates a generic debugfs framework for DesignWare PCIe
> controllers where other debug features apart from RASDES can also be
> added as and when required.
> 
> v7:
>     - Moved the patches to make finding VSEC IDs common from Mani's patchset [1]
>       into this series to remove dependancy as discussed
>     - Addressed style related change requests from v6

I tested this series, and one thing that I noticed:

# for f in /sys/kernel/debug/dwc_pcie_a40000000.pcie/rasdes_event_counter/*/counter_enable; do echo 1 > $f; done

# grep "" /sys/kernel/debug/dwc_pcie_a40000000.pcie/rasdes_event_counter/*/* | grep Disabled
/sys/kernel/debug/dwc_pcie_a40000000.pcie/rasdes_event_counter/ctl_skp_os_parity_err/counter_enable:Counter Disabled
/sys/kernel/debug/dwc_pcie_a40000000.pcie/rasdes_event_counter/deskew_uncompleted_err/counter_enable:Counter Disabled
/sys/kernel/debug/dwc_pcie_a40000000.pcie/rasdes_event_counter/framing_err_in_l0/counter_enable:Counter Disabled
/sys/kernel/debug/dwc_pcie_a40000000.pcie/rasdes_event_counter/margin_crc_parity_err/counter_enable:Counter Disabled
/sys/kernel/debug/dwc_pcie_a40000000.pcie/rasdes_event_counter/retimer_parity_err_1st/counter_enable:Counter Disabled
/sys/kernel/debug/dwc_pcie_a40000000.pcie/rasdes_event_counter/retimer_parity_err_2nd/counter_enable:Counter Disabled

that there are some events that cannot be enabled when testing on my platform,
rk3588, perhaps this is because my version of the DWC IP does not have these
events.

(Because all the other events can be enabled successfully:
# grep "" /sys/kernel/debug/dwc_pcie_a40000000.pcie/rasdes_event_counter/*/* | grep Enabled | wc -l
29
)


So the question is, how do we want to handle that?

E.g. counter_enable_write() could theoretically read back the
dw_pcie_readl_dbi(pci, rinfo->ras_cap_offset + RAS_DES_EVENT_COUNTER_CTRL_REG);
register after doing the
ww_pcie_writel_dbi(pci, rinfo->ras_cap_offset + RAS_DES_EVENT_COUNTER_CTRL_REG, val);

to actually check if it could enable the event.

If counter_enable_write() could not enable the specific event, should it
perhaps return a failure to user space?

Or, do we want to keep the current behavior of just letting counter_enable_write()
return success, even for events that are not supported by the specific DWC PCIe
implementation?


Kind regards,
Niklas

Re: [PATCH v7 0/5] Add support for debugfs based RAS DES feature in PCIe DW

[Manivannan Sadhasivam]

On Mon, Feb 24, 2025 at 06:08:26PM +0100, Niklas Cassel wrote:
> Hello Shradha,
> 
> On Fri, Feb 21, 2025 at 06:45:43PM +0530, Shradha Todi wrote:
> > DesignWare controller provides a vendor specific extended capability
> > called RASDES as an IP feature. This extended capability  provides
> > hardware information like:
> >  - Debug registers to know the state of the link or controller. 
> >  - Error injection mechanisms to inject various PCIe errors including
> >    sequence number, CRC
> >  - Statistical counters to know how many times a particular event
> >    occurred
> > 
> > However, in Linux we do not have any generic or custom support to be
> > able to use this feature in an efficient manner. This is the reason we
> > are proposing this framework. Debug and bring up time of high-speed IPs
> > are highly dependent on costlier hardware analyzers and this solution
> > will in some ways help to reduce the HW analyzer usage.
> > 
> > The debugfs entries can be used to get information about underlying
> > hardware and can be shared with user space. Separate debugfs entries has
> > been created to cater to all the DES hooks provided by the controller.
> > The debugfs entries interacts with the RASDES registers in the required
> > sequence and provides the meaningful data to the user. This eases the
> > effort to understand and use the register information for debugging.
> > 
> > This series creates a generic debugfs framework for DesignWare PCIe
> > controllers where other debug features apart from RASDES can also be
> > added as and when required.
> > 
> > v7:
> >     - Moved the patches to make finding VSEC IDs common from Mani's patchset [1]
> >       into this series to remove dependancy as discussed
> >     - Addressed style related change requests from v6
> 
> I tested this series, and one thing that I noticed:
> 
> # for f in /sys/kernel/debug/dwc_pcie_a40000000.pcie/rasdes_event_counter/*/counter_enable; do echo 1 > $f; done
> 
> # grep "" /sys/kernel/debug/dwc_pcie_a40000000.pcie/rasdes_event_counter/*/* | grep Disabled
> /sys/kernel/debug/dwc_pcie_a40000000.pcie/rasdes_event_counter/ctl_skp_os_parity_err/counter_enable:Counter Disabled
> /sys/kernel/debug/dwc_pcie_a40000000.pcie/rasdes_event_counter/deskew_uncompleted_err/counter_enable:Counter Disabled
> /sys/kernel/debug/dwc_pcie_a40000000.pcie/rasdes_event_counter/framing_err_in_l0/counter_enable:Counter Disabled
> /sys/kernel/debug/dwc_pcie_a40000000.pcie/rasdes_event_counter/margin_crc_parity_err/counter_enable:Counter Disabled
> /sys/kernel/debug/dwc_pcie_a40000000.pcie/rasdes_event_counter/retimer_parity_err_1st/counter_enable:Counter Disabled
> /sys/kernel/debug/dwc_pcie_a40000000.pcie/rasdes_event_counter/retimer_parity_err_2nd/counter_enable:Counter Disabled
> 
> that there are some events that cannot be enabled when testing on my platform,
> rk3588, perhaps this is because my version of the DWC IP does not have these
> events.
> 
> (Because all the other events can be enabled successfully:
> # grep "" /sys/kernel/debug/dwc_pcie_a40000000.pcie/rasdes_event_counter/*/* | grep Enabled | wc -l
> 29
> )
> 
> 
> So the question is, how do we want to handle that?
>

This is a really good question.
 
> E.g. counter_enable_write() could theoretically read back the
> dw_pcie_readl_dbi(pci, rinfo->ras_cap_offset + RAS_DES_EVENT_COUNTER_CTRL_REG);
> register after doing the
> ww_pcie_writel_dbi(pci, rinfo->ras_cap_offset + RAS_DES_EVENT_COUNTER_CTRL_REG, val);
> 
> to actually check if it could enable the event.
> 
> If counter_enable_write() could not enable the specific event, should it
> perhaps return a failure to user space?
> 

Yes, it would be appropriate to return -EOPNOTSUPP in that case. But I'd like to
merge this series asap. So this patch can come on top of this series.

- Mani

-- 
மணிவண்ணன் சதாசிவம்

Re: [PATCH v7 0/5] Add support for debugfs based RAS DES feature in PCIe DW

[Krzysztof Wilczyński]

Hello,

> DesignWare controller provides a vendor specific extended capability
> called RASDES as an IP feature. This extended capability  provides
> hardware information like:
>  - Debug registers to know the state of the link or controller. 
>  - Error injection mechanisms to inject various PCIe errors including
>    sequence number, CRC
>  - Statistical counters to know how many times a particular event
>    occurred
> 
> However, in Linux we do not have any generic or custom support to be
> able to use this feature in an efficient manner. This is the reason we
> are proposing this framework. Debug and bring up time of high-speed IPs
> are highly dependent on costlier hardware analyzers and this solution
> will in some ways help to reduce the HW analyzer usage.
> 
> The debugfs entries can be used to get information about underlying
> hardware and can be shared with user space. Separate debugfs entries has
> been created to cater to all the DES hooks provided by the controller.
> The debugfs entries interacts with the RASDES registers in the required
> sequence and provides the meaningful data to the user. This eases the
> effort to understand and use the register information for debugging.
> 
> This series creates a generic debugfs framework for DesignWare PCIe
> controllers where other debug features apart from RASDES can also be
> added as and when required.

Applied to controller/dwc, thank you!

	Krzysztof

Re: [PATCH v7 0/5] Add support for debugfs based RAS DES feature in PCIe DW

[Krzysztof Wilczyński]

Hello,

[...]
> Yes, it would be appropriate to return -EOPNOTSUPP in that case. But I'd like to
> merge this series asap. So this patch can come on top of this series.

I pulled the series so that we can get some mileage out of the 0-day bot,
so to speak, and also get some testing via the linux-next tree.

That said, while looking at the code, there have been bits where I wanted
to get some clarification.  Nothing will be a blocker.

	Krzysztof

Re: [PATCH v7 0/5] Add support for debugfs based RAS DES feature in PCIe DW

[Niklas Cassel]

On Tue, Feb 25, 2025 at 01:58:35PM +0530, Manivannan Sadhasivam wrote:
> On Mon, Feb 24, 2025 at 06:08:26PM +0100, Niklas Cassel wrote:
> > Hello Shradha,
> > 
> > On Fri, Feb 21, 2025 at 06:45:43PM +0530, Shradha Todi wrote:
> > > DesignWare controller provides a vendor specific extended capability
> > > called RASDES as an IP feature. This extended capability  provides
> > > hardware information like:
> > >  - Debug registers to know the state of the link or controller. 
> > >  - Error injection mechanisms to inject various PCIe errors including
> > >    sequence number, CRC
> > >  - Statistical counters to know how many times a particular event
> > >    occurred
> > > 
> > > However, in Linux we do not have any generic or custom support to be
> > > able to use this feature in an efficient manner. This is the reason we
> > > are proposing this framework. Debug and bring up time of high-speed IPs
> > > are highly dependent on costlier hardware analyzers and this solution
> > > will in some ways help to reduce the HW analyzer usage.
> > > 
> > > The debugfs entries can be used to get information about underlying
> > > hardware and can be shared with user space. Separate debugfs entries has
> > > been created to cater to all the DES hooks provided by the controller.
> > > The debugfs entries interacts with the RASDES registers in the required
> > > sequence and provides the meaningful data to the user. This eases the
> > > effort to understand and use the register information for debugging.
> > > 
> > > This series creates a generic debugfs framework for DesignWare PCIe
> > > controllers where other debug features apart from RASDES can also be
> > > added as and when required.
> > > 
> > > v7:
> > >     - Moved the patches to make finding VSEC IDs common from Mani's patchset [1]
> > >       into this series to remove dependancy as discussed
> > >     - Addressed style related change requests from v6
> > 
> > I tested this series, and one thing that I noticed:
> > 
> > # for f in /sys/kernel/debug/dwc_pcie_a40000000.pcie/rasdes_event_counter/*/counter_enable; do echo 1 > $f; done
> > 
> > # grep "" /sys/kernel/debug/dwc_pcie_a40000000.pcie/rasdes_event_counter/*/* | grep Disabled
> > /sys/kernel/debug/dwc_pcie_a40000000.pcie/rasdes_event_counter/ctl_skp_os_parity_err/counter_enable:Counter Disabled
> > /sys/kernel/debug/dwc_pcie_a40000000.pcie/rasdes_event_counter/deskew_uncompleted_err/counter_enable:Counter Disabled
> > /sys/kernel/debug/dwc_pcie_a40000000.pcie/rasdes_event_counter/framing_err_in_l0/counter_enable:Counter Disabled
> > /sys/kernel/debug/dwc_pcie_a40000000.pcie/rasdes_event_counter/margin_crc_parity_err/counter_enable:Counter Disabled
> > /sys/kernel/debug/dwc_pcie_a40000000.pcie/rasdes_event_counter/retimer_parity_err_1st/counter_enable:Counter Disabled
> > /sys/kernel/debug/dwc_pcie_a40000000.pcie/rasdes_event_counter/retimer_parity_err_2nd/counter_enable:Counter Disabled
> > 
> > that there are some events that cannot be enabled when testing on my platform,
> > rk3588, perhaps this is because my version of the DWC IP does not have these
> > events.
> > 
> > (Because all the other events can be enabled successfully:
> > # grep "" /sys/kernel/debug/dwc_pcie_a40000000.pcie/rasdes_event_counter/*/* | grep Enabled | wc -l
> > 29
> > )
> > 
> > 
> > So the question is, how do we want to handle that?
> >
> 
> This is a really good question.
>  
> > E.g. counter_enable_write() could theoretically read back the
> > dw_pcie_readl_dbi(pci, rinfo->ras_cap_offset + RAS_DES_EVENT_COUNTER_CTRL_REG);
> > register after doing the
> > ww_pcie_writel_dbi(pci, rinfo->ras_cap_offset + RAS_DES_EVENT_COUNTER_CTRL_REG, val);
> > 
> > to actually check if it could enable the event.
> > 
> > If counter_enable_write() could not enable the specific event, should it
> > perhaps return a failure to user space?
> > 
> 
> Yes, it would be appropriate to return -EOPNOTSUPP in that case. But I'd like to
> merge this series asap. So this patch can come on top of this series.

I agree that returning an error is probably the nicest thing.

However, this series has been picked up already :)

Is there anyone who volunteers on implementing the proposed feature?

If you have time, it would be interesting to see if you see the same behavior
on QCOM SoCs. (Assuming that your DWC PCIe controller does not implement all
events that Samsung DWC PCIe controller does.)


Kind regards,
Niklas

Re: [PATCH v7 1/5] perf/dwc_pcie: Move common DWC struct definitions to 'pcie-dwc.h'

[Krzysztof Wilczyński]

Hello,

> Since these are common to all Desginware PCIe IPs, move them to a new
> header 'pcie-dwc.h', so that other drivers like debugfs, perf and sysfs
> could make use of them.
> 
> Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
> Signed-off-by: Shradha Todi <shradha.t@samsung.com>

We are still missing feedback from the perf maintainers.

Especially, as I would like to take this patch via the PCI tree, if there
is no objection.

Thank you, Niklas, for pointing this out.  Appreciated.

	Krzysztof

Re: [PATCH v7 0/5] Add support for debugfs based RAS DES feature in PCIe DW

[Manivannan Sadhasivam]

On Tue, Feb 25, 2025 at 03:35:25PM +0100, Niklas Cassel wrote:
> On Tue, Feb 25, 2025 at 01:58:35PM +0530, Manivannan Sadhasivam wrote:
> > On Mon, Feb 24, 2025 at 06:08:26PM +0100, Niklas Cassel wrote:
> > > Hello Shradha,
> > > 
> > > On Fri, Feb 21, 2025 at 06:45:43PM +0530, Shradha Todi wrote:
> > > > DesignWare controller provides a vendor specific extended capability
> > > > called RASDES as an IP feature. This extended capability  provides
> > > > hardware information like:
> > > >  - Debug registers to know the state of the link or controller. 
> > > >  - Error injection mechanisms to inject various PCIe errors including
> > > >    sequence number, CRC
> > > >  - Statistical counters to know how many times a particular event
> > > >    occurred
> > > > 
> > > > However, in Linux we do not have any generic or custom support to be
> > > > able to use this feature in an efficient manner. This is the reason we
> > > > are proposing this framework. Debug and bring up time of high-speed IPs
> > > > are highly dependent on costlier hardware analyzers and this solution
> > > > will in some ways help to reduce the HW analyzer usage.
> > > > 
> > > > The debugfs entries can be used to get information about underlying
> > > > hardware and can be shared with user space. Separate debugfs entries has
> > > > been created to cater to all the DES hooks provided by the controller.
> > > > The debugfs entries interacts with the RASDES registers in the required
> > > > sequence and provides the meaningful data to the user. This eases the
> > > > effort to understand and use the register information for debugging.
> > > > 
> > > > This series creates a generic debugfs framework for DesignWare PCIe
> > > > controllers where other debug features apart from RASDES can also be
> > > > added as and when required.
> > > > 
> > > > v7:
> > > >     - Moved the patches to make finding VSEC IDs common from Mani's patchset [1]
> > > >       into this series to remove dependancy as discussed
> > > >     - Addressed style related change requests from v6
> > > 
> > > I tested this series, and one thing that I noticed:
> > > 
> > > # for f in /sys/kernel/debug/dwc_pcie_a40000000.pcie/rasdes_event_counter/*/counter_enable; do echo 1 > $f; done
> > > 
> > > # grep "" /sys/kernel/debug/dwc_pcie_a40000000.pcie/rasdes_event_counter/*/* | grep Disabled
> > > /sys/kernel/debug/dwc_pcie_a40000000.pcie/rasdes_event_counter/ctl_skp_os_parity_err/counter_enable:Counter Disabled
> > > /sys/kernel/debug/dwc_pcie_a40000000.pcie/rasdes_event_counter/deskew_uncompleted_err/counter_enable:Counter Disabled
> > > /sys/kernel/debug/dwc_pcie_a40000000.pcie/rasdes_event_counter/framing_err_in_l0/counter_enable:Counter Disabled
> > > /sys/kernel/debug/dwc_pcie_a40000000.pcie/rasdes_event_counter/margin_crc_parity_err/counter_enable:Counter Disabled
> > > /sys/kernel/debug/dwc_pcie_a40000000.pcie/rasdes_event_counter/retimer_parity_err_1st/counter_enable:Counter Disabled
> > > /sys/kernel/debug/dwc_pcie_a40000000.pcie/rasdes_event_counter/retimer_parity_err_2nd/counter_enable:Counter Disabled
> > > 
> > > that there are some events that cannot be enabled when testing on my platform,
> > > rk3588, perhaps this is because my version of the DWC IP does not have these
> > > events.
> > > 
> > > (Because all the other events can be enabled successfully:
> > > # grep "" /sys/kernel/debug/dwc_pcie_a40000000.pcie/rasdes_event_counter/*/* | grep Enabled | wc -l
> > > 29
> > > )
> > > 
> > > 
> > > So the question is, how do we want to handle that?
> > >
> > 
> > This is a really good question.
> >  
> > > E.g. counter_enable_write() could theoretically read back the
> > > dw_pcie_readl_dbi(pci, rinfo->ras_cap_offset + RAS_DES_EVENT_COUNTER_CTRL_REG);
> > > register after doing the
> > > ww_pcie_writel_dbi(pci, rinfo->ras_cap_offset + RAS_DES_EVENT_COUNTER_CTRL_REG, val);
> > > 
> > > to actually check if it could enable the event.
> > > 
> > > If counter_enable_write() could not enable the specific event, should it
> > > perhaps return a failure to user space?
> > > 
> > 
> > Yes, it would be appropriate to return -EOPNOTSUPP in that case. But I'd like to
> > merge this series asap. So this patch can come on top of this series.
> 
> I agree that returning an error is probably the nicest thing.
> 
> However, this series has been picked up already :)
> 
> Is there anyone who volunteers on implementing the proposed feature?
> 

I did and submitted the fix [1].

> If you have time, it would be interesting to see if you see the same behavior
> on QCOM SoCs. (Assuming that your DWC PCIe controller does not implement all
> events that Samsung DWC PCIe controller does.)
> 

Yeah, I observed the same behavior on the Qcom platform, though only 2 counters
were not supported. But I also noticed a null ptr dereference due to refclk
dependency, so submitted a fix for that also.

- Mani

[1] https://lore.kernel.org/linux-pci/20250225171239.19574-1-manivannan.sadhasivam@linaro.org/

-- 
மணிவண்ணன் சதாசிவம்

Re: [PATCH v7 1/5] perf/dwc_pcie: Move common DWC struct definitions to 'pcie-dwc.h'

[Shuai Xue]

在 2025/2/21 21:15, Shradha Todi 写道:
> From: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
> 
> Since these are common to all Desginware PCIe IPs, move them to a new
> header 'pcie-dwc.h', so that other drivers like debugfs, perf and sysfs
> could make use of them.
> 
> Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
> Signed-off-by: Shradha Todi <shradha.t@samsung.com>
> ---
>   MAINTAINERS                 |  1 +
>   drivers/perf/dwc_pcie_pmu.c | 25 +++----------------------
>   include/linux/pcie-dwc.h    | 34 ++++++++++++++++++++++++++++++++++
>   3 files changed, 38 insertions(+), 22 deletions(-)
>   create mode 100644 include/linux/pcie-dwc.h
> 
> diff --git a/MAINTAINERS b/MAINTAINERS
> index 3864d473f52f..6474a2d83de4 100644
> --- a/MAINTAINERS
> +++ b/MAINTAINERS
> @@ -18167,6 +18167,7 @@ S:	Maintained
>   F:	Documentation/devicetree/bindings/pci/snps,dw-pcie-ep.yaml
>   F:	Documentation/devicetree/bindings/pci/snps,dw-pcie.yaml
>   F:	drivers/pci/controller/dwc/*designware*
> +F:	include/linux/pcie-dwc.h
>   
>   PCI DRIVER FOR TI DRA7XX/J721E
>   M:	Vignesh Raghavendra <vigneshr@ti.com>
> diff --git a/drivers/perf/dwc_pcie_pmu.c b/drivers/perf/dwc_pcie_pmu.c
> index cccecae9823f..da30f2c2d674 100644
> --- a/drivers/perf/dwc_pcie_pmu.c
> +++ b/drivers/perf/dwc_pcie_pmu.c
> @@ -13,6 +13,7 @@
>   #include <linux/errno.h>
>   #include <linux/kernel.h>
>   #include <linux/list.h>
> +#include <linux/pcie-dwc.h>
>   #include <linux/perf_event.h>
>   #include <linux/pci.h>
>   #include <linux/platform_device.h>
> @@ -99,26 +100,6 @@ struct dwc_pcie_dev_info {
>   	struct list_head dev_node;
>   };
>   
> -struct dwc_pcie_pmu_vsec_id {
> -	u16 vendor_id;
> -	u16 vsec_id;
> -	u8 vsec_rev;
> -};
> -
> -/*
> - * VSEC IDs are allocated by the vendor, so a given ID may mean different
> - * things to different vendors.  See PCIe r6.0, sec 7.9.5.2.
> - */
> -static const struct dwc_pcie_pmu_vsec_id dwc_pcie_pmu_vsec_ids[] = {
> -	{ .vendor_id = PCI_VENDOR_ID_ALIBABA,
> -	  .vsec_id = 0x02, .vsec_rev = 0x4 },
> -	{ .vendor_id = PCI_VENDOR_ID_AMPERE,
> -	  .vsec_id = 0x02, .vsec_rev = 0x4 },
> -	{ .vendor_id = PCI_VENDOR_ID_QCOM,
> -	  .vsec_id = 0x02, .vsec_rev = 0x4 },
> -	{} /* terminator */
> -};
> -
>   static ssize_t cpumask_show(struct device *dev,
>   					 struct device_attribute *attr,
>   					 char *buf)
> @@ -529,14 +510,14 @@ static void dwc_pcie_unregister_pmu(void *data)
>   
>   static u16 dwc_pcie_des_cap(struct pci_dev *pdev)
>   {
> -	const struct dwc_pcie_pmu_vsec_id *vid;
> +	const struct dwc_pcie_vsec_id *vid;
>   	u16 vsec;
>   	u32 val;
>   
>   	if (!pci_is_pcie(pdev) || !(pci_pcie_type(pdev) == PCI_EXP_TYPE_ROOT_PORT))
>   		return 0;
>   
> -	for (vid = dwc_pcie_pmu_vsec_ids; vid->vendor_id; vid++) {
> +	for (vid = dwc_pcie_rasdes_vsec_ids; vid->vendor_id; vid++) {
>   		vsec = pci_find_vsec_capability(pdev, vid->vendor_id,
>   						vid->vsec_id);
>   		if (vsec) {
> diff --git a/include/linux/pcie-dwc.h b/include/linux/pcie-dwc.h
> new file mode 100644
> index 000000000000..40f3545731c8
> --- /dev/null
> +++ b/include/linux/pcie-dwc.h
> @@ -0,0 +1,34 @@
> +/* SPDX-License-Identifier: GPL-2.0 */
> +/*
> + * Copyright (C) 2021-2023 Alibaba Inc.
> + *
> + * Copyright 2025 Linaro Ltd.
> + * Author: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
> + */
> +
> +#ifndef LINUX_PCIE_DWC_H
> +#define LINUX_PCIE_DWC_H
> +
> +#include <linux/pci_ids.h>
> +
> +struct dwc_pcie_vsec_id {
> +	u16 vendor_id;
> +	u16 vsec_id;
> +	u8 vsec_rev;
> +};
> +
> +/*
> + * VSEC IDs are allocated by the vendor, so a given ID may mean different
> + * things to different vendors.  See PCIe r6.0, sec 7.9.5.2.
> + */
> +static const struct dwc_pcie_vsec_id dwc_pcie_rasdes_vsec_ids[] = {
> +	{ .vendor_id = PCI_VENDOR_ID_ALIBABA,
> +	  .vsec_id = 0x02, .vsec_rev = 0x4 },
> +	{ .vendor_id = PCI_VENDOR_ID_AMPERE,
> +	  .vsec_id = 0x02, .vsec_rev = 0x4 },
> +	{ .vendor_id = PCI_VENDOR_ID_QCOM,
> +	  .vsec_id = 0x02, .vsec_rev = 0x4 },
> +	{} /* terminator */
> +};
> +
> +#endif /* LINUX_PCIE_DWC_H */

LGTM. Thanks.

Reviewed-by: Shuai Xue <xueshuai@linux.alibaba.com>

Shuai

Re: [PATCH v7 1/5] perf/dwc_pcie: Move common DWC struct definitions to 'pcie-dwc.h'

[Krzysztof Wilczyński]

Hello,

> > Since these are common to all Desginware PCIe IPs, move them to a new
> > header 'pcie-dwc.h', so that other drivers like debugfs, perf and sysfs
> > could make use of them.
[...]
> LGTM. Thanks.
> 
> Reviewed-by: Shuai Xue <xueshuai@linux.alibaba.com>

Thank you!

	Krzysztof

Re: [PATCH v7 0/5] Add support for debugfs based RAS DES feature in PCIe DW

[Hrishikesh Deleep]

>Subject: [PATCH v7 0/5] Add support for debugfs based RAS DES feature in
> PCIe DW
>Date: Fri, 21 Feb 2025 18:45:43 +0530
>Message-Id: <20250221131548.59616-1-shradha.t@samsung.com>
>X-Mailer: git-send-email 2.17.1
>Precedence: bulk
>X-Mailing-List: linux-kernel@vger.kernel.org
>List-Id: <linux-kernel.vger.kernel.org>
>List-Subscribe: <mailto:linux-kernel+subscribe@vger.kernel.org>
>List-Unsubscribe: <mailto:linux-kernel+unsubscribe@vger.kernel.org>
>MIME-Version: 1.0
>Content-Transfer-Encoding: 8bit
>X-Brightmail-Tracker: H4sIAAAAAAAAA+NgFnrBJsWRmVeSWpSXmKPExsWy7bCmum7f2p3pBpev8Fhcaf/NbjH9sKLF
>	kqYMi2MTVjBbNK2+y2qx4stMdotVC6+xWTT0/Ga12PT4GqvF5V1z2CzOzjvOZnFl6zoWi5Y/
>	LSwWd1s6WS2WXr/IZPF3215Gi0Vbv7BbLGx+yWjxf88Odovew7UWLXdMLd7/3MzmIOaxeMUU
>	Vo8189YweuycdZfdY8GmUo+WI29ZPTat6mTzuHNtD5vHzoeWHk+uTGfy2Lyk3qNvyypGj8+b
>	5AJ4orJtMlITU1KLFFLzkvNTMvPSbZW8g+Od403NDAx1DS0tzJUU8hJzU22VXHwCdN0yc4C+
>	VVIoS8wpBQoFJBYXK+nb2RTll5akKmTkF5fYKqUWpOQUmBToFSfmFpfmpevlpZZYGRoYGJkC
>	FSZkZ9w5vYalYLZqxYb7X5gaGB/IdjFyckgImEi83vKXqYuRi0NIYDejxLsX7UwgCSGBT4wS
>	u6dkQySA7IMTF7PDdKyZ08QKkdjJKHFn3nso5wujxL2+j2BVbAJaEo1fu5hBbBGBVkaJI8/E
>	QIqYBbYxSxxYsQIsISwQJPFjXRsLiM0ioCrx/e98oGYODl4BK4k9K+0htslLrN5wAKycV0BQ
>	4uTMJ2DlzEDx5q2zmSFqvnBIfJnNB2G7SCzadQPqUmGJV8e3QNlSEp/f7WWDsNMlVm6eAdWb
>	I/Ft8xImCNte4sCVOSwgJzALaEqs36UPEZaVmHpqHRPEWj6J3t9PoMp5JXbMg7GVJb783cMC
>	YUtKzDt2mRXC9pDYufkh2EghgViJ/v+2ExjlZyF5ZhaSZ2YhLF7AyLyKUTK1oDg3PTXZtMAw
>	L7UcHq3J+bmbGMFJXstlB+ON+f/0DjEycTAeYpTgYFYS4dUt2ZEuxJuSWFmVWpQfX1Sak1p8
>	iNEUGMATmaVEk/OBeSavJN7QxNLAxMzMzMTS2MxQSZy3eWdLupBAemJJanZqakFqEUwfEwen
>	VAOTyWph2UVGD0z1j0+NvfvEdnGQ413ON1OnCFx4vMfJeqvN3Mp7Kx86NXwKsN+sdkziZXTt
>	kesTTnswSfQ/W1W2UrvUr+Mf14yGGbL9V1cY802NfhC1Su7CI57OFPH5lz5En/srKjG9MXbJ
>	wbvLF2vMuPtkX2nEE4bFtdvOlN1YvEY1xMF6apLnj5CG0sSSXn3PgKUsajzVWQ/TfU7n/lD5
>	bfjrwVJLhqPcRqoTSkq8zi29HTM1wO3dCcNfq6we9H4Sbrx8ef67MN+3L8NLzveyuAo5iV1+
>	s8+BVbngfu6khvO+c9YriUadC2QxkJvBY3eiVHQx3yuDqBXzeubVRsnkfzzpt9tl19/cGTOv
>	1GuzKrEUZyQaajEXFScCAMosolJ7BAAA
>X-Brightmail-Tracker: H4sIAAAAAAAAA+NgFlrJIsWRmVeSWpSXmKPExsWy7bCSvG537Y50g2MLlC2utP9mt5h+WNFi
>	SVOGxbEJK5gtmlbfZbVY8WUmu8WqhdfYLBp6frNabHp8jdXi8q45bBZn5x1ns7iydR2LRcuf
>	FhaLuy2drBZLr19ksvi7bS+jxaKtX9gtFja/ZLT4v2cHu0Xv4VqLljumFu9/bmZzEPNYvGIK
>	q8eaeWsYPXbOusvusWBTqUfLkbesHptWdbJ53Lm2h81j50NLjydXpjN5bF5S79G3ZRWjx+dN
>	cgE8UVw2Kak5mWWpRfp2CVwZd06vYSmYrVqx4f4XpgbGB7JdjJwcEgImEmvmNLF2MXJxCAls
>	Z5Q4cm4FE0RCUuLzxXVQtrDEyn/P2SGKPjFKnLi3HyzBJqAl0fi1ixkkISLQySix98g7sCpm
>	gXPMEjM/tzCCVAkLBEicWbwezGYRUJX4/nc+UBEHB6+AlcSelfYQG+QlVm84wAxi8woISpyc
>	+YQFpIRZQF1i/TwhkDAzUEnz1tnMExj5ZyGpmoVQNQtJ1QJG5lWMoqkFxbnpuckFhnrFibnF
>	pXnpesn5uZsYwbGpFbSDcdn6v3qHGJk4GA8xSnAwK4nw6pbsSBfiTUmsrEotyo8vKs1JLT7E
>	KM3BoiTOq5zTmSIkkJ5YkpqdmlqQWgSTZeLglGpg4pjMtu2b3EL3R9/vnl2nEvFC55TM2YSA
>	my6J94qZeTt8782a/COa5zfzt5k3prcfseJ7luLkekH78Adlm03yN7mZkk1XMM56/M3uTPar
>	ex8KP295FMh6s2nhnKm/5osxdX5TKq3b61nX9iT1YXNdyKerQsdqYjgT+VaEJK99GX3A6He8
>	1XIWZ2vz9ffnsnQutJ2jtMajR/n7bM0NmbNmena/SZXhnfzD5alN2PFWrd0vK8Je8Blf3/b1
>	x9bwBX5/Yl/s/XhNzGKZrer2uvBbUQqvGNOWb17bYhX/xT3W2q3tpEjcHY7SxiIpxtc63+us
>	rinukH9y8IUFl3ej4vrW9/u/5z0wPMPHsW9quEntZSWW4oxEQy3mouJEAI4aFoM8AwAA
>X-CMS-MailID: 20250221132011epcas5p4dea1e9ae5c09afaabcd1822f3a7d15c5
>X-Msg-Generator: CA
>Content-Type: text/plain; charset="utf-8"
>X-Sendblock-Type: REQ_APPROVE
>CMS-TYPE: 105P
>DLP-Filter: Pass
>X-CFilter-Loop: Reflected
>X-CMS-RootMailID: 20250221132011epcas5p4dea1e9ae5c09afaabcd1822f3a7d15c5
>References: <CGME20250221132011epcas5p4dea1e9ae5c09afaabcd1822f3a7d15c5@epcas5p4.samsung.com>
>
>DesignWare controller provides a vendor specific extended capability
>called RASDES as an IP feature. This extended capability  provides
>hardware information like:
> - Debug registers to know the state of the link or controller. 
> - Error injection mechanisms to inject various PCIe errors including
>   sequence number, CRC
> - Statistical counters to know how many times a particular event
>   occurred
>
>However, in Linux we do not have any generic or custom support to be
>able to use this feature in an efficient manner. This is the reason we
>are proposing this framework. Debug and bring up time of high-speed IPs
>are highly dependent on costlier hardware analyzers and this solution
>will in some ways help to reduce the HW analyzer usage.
>
>The debugfs entries can be used to get information about underlying
>hardware and can be shared with user space. Separate debugfs entries has
>been created to cater to all the DES hooks provided by the controller.
>The debugfs entries interacts with the RASDES registers in the required
>sequence and provides the meaningful data to the user. This eases the
>effort to understand and use the register information for debugging.
>
>This series creates a generic debugfs framework for DesignWare PCIe
>controllers where other debug features apart from RASDES can also be
>added as and when required.
>
>v7:
>    - Moved the patches to make finding VSEC IDs common from Mani's patchset [1]
>      into this series to remove dependancy as discussed
>    - Addressed style related change requests from v6
>
>v6: https://lore.kernel.org/all/20250214105007.97582-1-shradha.t@samsung.com/
>    - Addressed Niklas's comment to make vsec ID finding similar to perf
>    - Minor changes in the driver to make the debugfs file common and
>      not specefic to RASDES so that other developers can add debug
>      related features to this file.
>
>v5: https://lore.kernel.org/all/20250121111421.35437-1-shradha.t@samsung.com/
>    - Addressed Fan's comment to split the patches for easier review
>    - Addressed Bjorn's comment to fix vendor specific cap search
>    - Addressed style related change requests from v4
>    - Added rasdes debugfs init call to common designware files for host
>      and EP.
>
>v4: https://lore.kernel.org/lkml/20241206074456.17401-1-shradha.t@samsung.com/
>    - Addressed comments from Manivannan, Bjorn and Jonathan
>    - Addressed style related change requests from v3
>    - Added Documentation under Documentation/ABI/testing and kdoc stype
>      comments wherever required for better understanding
>    - Enhanced error injection to include all possible error groups
>    - Removed debugfs init call from common designware file and left it
>      up to individual platform drivers to init/deinit as required.
>
>v3: https://lore.kernel.org/all/20240625093813.112555-1-shradha.t@samsung.com/
>    - v2 had suggestions about moving this framework to perf/EDAC instead of a
>      controller specific debugfs but after discussions we decided to go ahead
>      with the same. Rebased and posted v3 with minor style changes.
>
>v2: https://lore.kernel.org/lkml/20231130115044.53512-1-shradha.t@samsung.com/
>    - Addressed comments from Krzysztof Wilczyński, Bjorn Helgaas and
>      posted v2 with a changed implementation for a better code design
>
>v1: https://lore.kernel.org/all/20210518174618.42089-1-shradha.t@samsung.com/T/
>
>[1] https://lore.kernel.org/all/20250218-pcie-qcom-ptm-v1-0-16d7e480d73e@linaro.org/
>
>Manivannan Sadhasivam (1):
>  perf/dwc_pcie: Move common DWC struct definitions to 'pcie-dwc.h'
>
>Shradha Todi (4):
>  PCI: dwc: Add helper to find the Vendor Specific Extended Capability
>    (VSEC)
>  Add debugfs based silicon debug support in DWC
>  Add debugfs based error injection support in DWC
>  Add debugfs based statistical counter support in DWC
>
> Documentation/ABI/testing/debugfs-dwc-pcie    | 144 +++++
> MAINTAINERS                                   |   1 +
> drivers/pci/controller/dwc/Kconfig            |  10 +
> drivers/pci/controller/dwc/Makefile           |   1 +
> .../controller/dwc/pcie-designware-debugfs.c  | 564 ++++++++++++++++++
> .../pci/controller/dwc/pcie-designware-ep.c   |   5 +
> .../pci/controller/dwc/pcie-designware-host.c |   6 +
> drivers/pci/controller/dwc/pcie-designware.c  |  46 ++
> drivers/pci/controller/dwc/pcie-designware.h  |  21 +
> drivers/perf/dwc_pcie_pmu.c                   |  25 +-
> include/linux/pcie-dwc.h                      |  36 ++
> 11 files changed, 837 insertions(+), 22 deletions(-)
> create mode 100644 Documentation/ABI/testing/debugfs-dwc-pcie
> create mode 100644 drivers/pci/controller/dwc/pcie-designware-debugfs.c
> create mode 100644 include/linux/pcie-dwc.h
>
>-- 
>2.17.1
>
>

Thanks for adding the support!

Tested the patch in one of our internal SoC with DesginWare PCIe controller. The patch works fine.

Tested-by: Hrishikesh Deleep <hrishikesh.d@samsung.com>

Re: [PATCH v7 4/5] Add debugfs based error injection support in DWC

[Krzysztof Wilczyński]

Hello,

[...]
> +		29) Generates duplicate TLPs - duplicate_dllp
> +		30) Generates Nullified TLPs - nullified_tlp

Would the above field called "duplicate_dllp" for duplicate TLPs be
a potential typo?  Perhaps this should be called "duplicate_tlp"?

I wanted to make sure we have the correct field name.

Thank you!

	Krzysztof

Re: [PATCH v7 1/5] perf/dwc_pcie: Move common DWC struct definitions to 'pcie-dwc.h'

[Fan Ni]

On Fri, Feb 21, 2025 at 06:45:44PM +0530, Shradha Todi wrote:
> From: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
> 
> Since these are common to all Desginware PCIe IPs, move them to a new
> header 'pcie-dwc.h', so that other drivers like debugfs, perf and sysfs
> could make use of them.
> 
> Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
> Signed-off-by: Shradha Todi <shradha.t@samsung.com>

Reviewed-by: Fan Ni <fan.ni@samsung.com>

> ---
>  MAINTAINERS                 |  1 +
>  drivers/perf/dwc_pcie_pmu.c | 25 +++----------------------
>  include/linux/pcie-dwc.h    | 34 ++++++++++++++++++++++++++++++++++
>  3 files changed, 38 insertions(+), 22 deletions(-)
>  create mode 100644 include/linux/pcie-dwc.h
> 
> diff --git a/MAINTAINERS b/MAINTAINERS
> index 3864d473f52f..6474a2d83de4 100644
> --- a/MAINTAINERS
> +++ b/MAINTAINERS
> @@ -18167,6 +18167,7 @@ S:	Maintained
>  F:	Documentation/devicetree/bindings/pci/snps,dw-pcie-ep.yaml
>  F:	Documentation/devicetree/bindings/pci/snps,dw-pcie.yaml
>  F:	drivers/pci/controller/dwc/*designware*
> +F:	include/linux/pcie-dwc.h
>  
>  PCI DRIVER FOR TI DRA7XX/J721E
>  M:	Vignesh Raghavendra <vigneshr@ti.com>
> diff --git a/drivers/perf/dwc_pcie_pmu.c b/drivers/perf/dwc_pcie_pmu.c
> index cccecae9823f..da30f2c2d674 100644
> --- a/drivers/perf/dwc_pcie_pmu.c
> +++ b/drivers/perf/dwc_pcie_pmu.c
> @@ -13,6 +13,7 @@
>  #include <linux/errno.h>
>  #include <linux/kernel.h>
>  #include <linux/list.h>
> +#include <linux/pcie-dwc.h>
>  #include <linux/perf_event.h>
>  #include <linux/pci.h>
>  #include <linux/platform_device.h>
> @@ -99,26 +100,6 @@ struct dwc_pcie_dev_info {
>  	struct list_head dev_node;
>  };
>  
> -struct dwc_pcie_pmu_vsec_id {
> -	u16 vendor_id;
> -	u16 vsec_id;
> -	u8 vsec_rev;
> -};
> -
> -/*
> - * VSEC IDs are allocated by the vendor, so a given ID may mean different
> - * things to different vendors.  See PCIe r6.0, sec 7.9.5.2.
> - */
> -static const struct dwc_pcie_pmu_vsec_id dwc_pcie_pmu_vsec_ids[] = {
> -	{ .vendor_id = PCI_VENDOR_ID_ALIBABA,
> -	  .vsec_id = 0x02, .vsec_rev = 0x4 },
> -	{ .vendor_id = PCI_VENDOR_ID_AMPERE,
> -	  .vsec_id = 0x02, .vsec_rev = 0x4 },
> -	{ .vendor_id = PCI_VENDOR_ID_QCOM,
> -	  .vsec_id = 0x02, .vsec_rev = 0x4 },
> -	{} /* terminator */
> -};
> -
>  static ssize_t cpumask_show(struct device *dev,
>  					 struct device_attribute *attr,
>  					 char *buf)
> @@ -529,14 +510,14 @@ static void dwc_pcie_unregister_pmu(void *data)
>  
>  static u16 dwc_pcie_des_cap(struct pci_dev *pdev)
>  {
> -	const struct dwc_pcie_pmu_vsec_id *vid;
> +	const struct dwc_pcie_vsec_id *vid;
>  	u16 vsec;
>  	u32 val;
>  
>  	if (!pci_is_pcie(pdev) || !(pci_pcie_type(pdev) == PCI_EXP_TYPE_ROOT_PORT))
>  		return 0;
>  
> -	for (vid = dwc_pcie_pmu_vsec_ids; vid->vendor_id; vid++) {
> +	for (vid = dwc_pcie_rasdes_vsec_ids; vid->vendor_id; vid++) {
>  		vsec = pci_find_vsec_capability(pdev, vid->vendor_id,
>  						vid->vsec_id);
>  		if (vsec) {
> diff --git a/include/linux/pcie-dwc.h b/include/linux/pcie-dwc.h
> new file mode 100644
> index 000000000000..40f3545731c8
> --- /dev/null
> +++ b/include/linux/pcie-dwc.h
> @@ -0,0 +1,34 @@
> +/* SPDX-License-Identifier: GPL-2.0 */
> +/*
> + * Copyright (C) 2021-2023 Alibaba Inc.
> + *
> + * Copyright 2025 Linaro Ltd.
> + * Author: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
> + */
> +
> +#ifndef LINUX_PCIE_DWC_H
> +#define LINUX_PCIE_DWC_H
> +
> +#include <linux/pci_ids.h>
> +
> +struct dwc_pcie_vsec_id {
> +	u16 vendor_id;
> +	u16 vsec_id;
> +	u8 vsec_rev;
> +};
> +
> +/*
> + * VSEC IDs are allocated by the vendor, so a given ID may mean different
> + * things to different vendors.  See PCIe r6.0, sec 7.9.5.2.
> + */
> +static const struct dwc_pcie_vsec_id dwc_pcie_rasdes_vsec_ids[] = {
> +	{ .vendor_id = PCI_VENDOR_ID_ALIBABA,
> +	  .vsec_id = 0x02, .vsec_rev = 0x4 },
> +	{ .vendor_id = PCI_VENDOR_ID_AMPERE,
> +	  .vsec_id = 0x02, .vsec_rev = 0x4 },
> +	{ .vendor_id = PCI_VENDOR_ID_QCOM,
> +	  .vsec_id = 0x02, .vsec_rev = 0x4 },
> +	{} /* terminator */
> +};
> +
> +#endif /* LINUX_PCIE_DWC_H */
> -- 
> 2.17.1
> 

Re: [PATCH v7 2/5] PCI: dwc: Add helper to find the Vendor Specific Extended Capability (VSEC)

[Fan Ni]

On Fri, Feb 21, 2025 at 06:45:45PM +0530, Shradha Todi wrote:
> dw_pcie_find_vsec_capability() is used by upcoming DWC APIs to find the
How are about "Add dw_pcie_find_ext_capability(), which will be used by
..."

Other than that,


Reviewed-by: Fan Ni <fan.ni@samsung.com>

> VSEC capabilities like PTM, RAS etc.
> 
> Co-developed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
> Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
> Signed-off-by: Shradha Todi <shradha.t@samsung.com>
> ---
>  drivers/pci/controller/dwc/pcie-designware.c | 40 ++++++++++++++++++++
>  1 file changed, 40 insertions(+)
> 
> diff --git a/drivers/pci/controller/dwc/pcie-designware.c b/drivers/pci/controller/dwc/pcie-designware.c
> index 145e7f579072..a7c0671c6715 100644
> --- a/drivers/pci/controller/dwc/pcie-designware.c
> +++ b/drivers/pci/controller/dwc/pcie-designware.c
> @@ -16,6 +16,7 @@
>  #include <linux/gpio/consumer.h>
>  #include <linux/ioport.h>
>  #include <linux/of.h>
> +#include <linux/pcie-dwc.h>
>  #include <linux/platform_device.h>
>  #include <linux/sizes.h>
>  #include <linux/types.h>
> @@ -283,6 +284,45 @@ u16 dw_pcie_find_ext_capability(struct dw_pcie *pci, u8 cap)
>  }
>  EXPORT_SYMBOL_GPL(dw_pcie_find_ext_capability);
>  
> +static u16 __dw_pcie_find_vsec_capability(struct dw_pcie *pci, u16 vendor_id,
> +					  u16 vsec_id)
> +{
> +	u16 vsec = 0;
> +	u32 header;
> +
> +	if (vendor_id != dw_pcie_readw_dbi(pci, PCI_VENDOR_ID))
> +		return 0;
> +
> +	while ((vsec = dw_pcie_find_next_ext_capability(pci, vsec,
> +						       PCI_EXT_CAP_ID_VNDR))) {
> +		header = dw_pcie_readl_dbi(pci, vsec + PCI_VNDR_HEADER);
> +		if (PCI_VNDR_HEADER_ID(header) == vsec_id)
> +			return vsec;
> +	}
> +
> +	return 0;
> +}
> +
> +static u16 dw_pcie_find_vsec_capability(struct dw_pcie *pci,
> +					const struct dwc_pcie_vsec_id *vsec_ids)
> +{
> +	const struct dwc_pcie_vsec_id *vid;
> +	u16 vsec;
> +	u32 header;
> +
> +	for (vid = vsec_ids; vid->vendor_id; vid++) {
> +		vsec = __dw_pcie_find_vsec_capability(pci, vid->vendor_id,
> +						      vid->vsec_id);
> +		if (vsec) {
> +			header = dw_pcie_readl_dbi(pci, vsec + PCI_VNDR_HEADER);
> +			if (PCI_VNDR_HEADER_REV(header) == vid->vsec_rev)
> +				return vsec;
> +		}
> +	}
> +
> +	return 0;
> +}
> +
>  int dw_pcie_read(void __iomem *addr, int size, u32 *val)
>  {
>  	if (!IS_ALIGNED((uintptr_t)addr, size)) {
> -- 
> 2.17.1
> 

Re: [PATCH v7 3/5] Add debugfs based silicon debug support in DWC

[Fan Ni]

On Fri, Feb 21, 2025 at 06:45:46PM +0530, Shradha Todi wrote:
> Add support to provide silicon debug interface to userspace. This set
> of debug registers are part of the RASDES feature present in DesignWare
> PCIe controllers.
> 
> Signed-off-by: Shradha Todi <shradha.t@samsung.com>

One comment inline.
> ---
>  Documentation/ABI/testing/debugfs-dwc-pcie    |  13 ++
>  drivers/pci/controller/dwc/Kconfig            |  10 +
>  drivers/pci/controller/dwc/Makefile           |   1 +
>  .../controller/dwc/pcie-designware-debugfs.c  | 176 ++++++++++++++++++
>  .../pci/controller/dwc/pcie-designware-ep.c   |   5 +
>  .../pci/controller/dwc/pcie-designware-host.c |   6 +
>  drivers/pci/controller/dwc/pcie-designware.c  |   6 +
>  drivers/pci/controller/dwc/pcie-designware.h  |  21 +++
>  include/linux/pcie-dwc.h                      |   2 +
>  9 files changed, 240 insertions(+)
>  create mode 100644 Documentation/ABI/testing/debugfs-dwc-pcie
>  create mode 100644 drivers/pci/controller/dwc/pcie-designware-debugfs.c
> 

...

> +
> +static void dwc_pcie_rasdes_debugfs_deinit(struct dw_pcie *pci)
> +{
> +	struct dwc_pcie_rasdes_info *rinfo = pci->debugfs->rasdes_info;
> +
> +	mutex_destroy(&rinfo->reg_event_lock);
> +}
> +
> +static int dwc_pcie_rasdes_debugfs_init(struct dw_pcie *pci, struct dentry *dir)
> +{
> +	struct dentry *rasdes_debug;
> +	struct dwc_pcie_rasdes_info *rasdes_info;
> +	struct device *dev = pci->dev;
> +	int ras_cap;
> +
> +	ras_cap = dw_pcie_find_rasdes_capability(pci);
> +	if (!ras_cap) {
> +		dev_dbg(dev, "no RASDES capability available\n");
> +		return -ENODEV;
> +	}
> +
> +	rasdes_info = devm_kzalloc(dev, sizeof(*rasdes_info), GFP_KERNEL);
> +	if (!rasdes_info)
> +		return -ENOMEM;
> +
> +	/* Create subdirectories for Debug, Error injection, Statistics */
> +	rasdes_debug = debugfs_create_dir("rasdes_debug", dir);
> +
> +	mutex_init(&rasdes_info->reg_event_lock);
> +	rasdes_info->ras_cap_offset = ras_cap;
> +	pci->debugfs->rasdes_info = rasdes_info;
> +
> +	/* Create debugfs files for Debug subdirectory */
> +	dwc_debugfs_create(lane_detect);
> +	dwc_debugfs_create(rx_valid);
> +
> +	return 0;
> +}
> +
> +void dwc_pcie_debugfs_deinit(struct dw_pcie *pci)
> +{
> +	dwc_pcie_rasdes_debugfs_deinit(pci);
> +	debugfs_remove_recursive(pci->debugfs->debug_dir);
> +}
> +
> +int dwc_pcie_debugfs_init(struct dw_pcie *pci)
> +{
> +	char dirname[DWC_DEBUGFS_BUF_MAX];
> +	struct device *dev = pci->dev;
> +	struct debugfs_info *debugfs;
> +	struct dentry *dir;
> +	int ret;
> +
> +	/* Create main directory for each platform driver */
> +	snprintf(dirname, DWC_DEBUGFS_BUF_MAX, "dwc_pcie_%s", dev_name(dev));
> +	dir = debugfs_create_dir(dirname, NULL);
> +	debugfs = devm_kzalloc(dev, sizeof(*debugfs), GFP_KERNEL);
> +	if (!debugfs)
> +		return -ENOMEM;
> +
> +	debugfs->debug_dir = dir;
> +	pci->debugfs = debugfs;
> +	ret = dwc_pcie_rasdes_debugfs_init(pci, dir);
> +	if (ret)
> +		dev_dbg(dev, "RASDES debugfs init failed\n");

What will happen if ret != 0? still return 0? 

Fan
> +
> +	return 0;
> +}
> diff --git a/drivers/pci/controller/dwc/pcie-designware-ep.c b/drivers/pci/controller/dwc/pcie-designware-ep.c
> index 72418160e658..f9d7f3f989ad 100644
> --- a/drivers/pci/controller/dwc/pcie-designware-ep.c
> +++ b/drivers/pci/controller/dwc/pcie-designware-ep.c
> @@ -814,6 +814,7 @@ void dw_pcie_ep_cleanup(struct dw_pcie_ep *ep)
>  {
>  	struct dw_pcie *pci = to_dw_pcie_from_ep(ep);
>  
> +	dwc_pcie_debugfs_deinit(pci);
>  	dw_pcie_edma_remove(pci);
>  }
>  EXPORT_SYMBOL_GPL(dw_pcie_ep_cleanup);
> @@ -989,6 +990,10 @@ int dw_pcie_ep_init_registers(struct dw_pcie_ep *ep)
>  
>  	dw_pcie_ep_init_non_sticky_registers(pci);
>  
> +	ret = dwc_pcie_debugfs_init(pci);
> +	if (ret)
> +		goto err_remove_edma;
> +
>  	return 0;
>  
>  err_remove_edma:
> diff --git a/drivers/pci/controller/dwc/pcie-designware-host.c b/drivers/pci/controller/dwc/pcie-designware-host.c
> index ffaded8f2df7..2081e8c72d12 100644
> --- a/drivers/pci/controller/dwc/pcie-designware-host.c
> +++ b/drivers/pci/controller/dwc/pcie-designware-host.c
> @@ -548,6 +548,10 @@ int dw_pcie_host_init(struct dw_pcie_rp *pp)
>  	if (pp->ops->post_init)
>  		pp->ops->post_init(pp);
>  
> +	ret = dwc_pcie_debugfs_init(pci);
> +	if (ret)
> +		goto err_stop_link;
> +
>  	return 0;
>  
>  err_stop_link:
> @@ -572,6 +576,8 @@ void dw_pcie_host_deinit(struct dw_pcie_rp *pp)
>  {
>  	struct dw_pcie *pci = to_dw_pcie_from_pp(pp);
>  
> +	dwc_pcie_debugfs_deinit(pci);
> +
>  	pci_stop_root_bus(pp->bridge->bus);
>  	pci_remove_root_bus(pp->bridge->bus);
>  
> diff --git a/drivers/pci/controller/dwc/pcie-designware.c b/drivers/pci/controller/dwc/pcie-designware.c
> index a7c0671c6715..3d1d95d9e380 100644
> --- a/drivers/pci/controller/dwc/pcie-designware.c
> +++ b/drivers/pci/controller/dwc/pcie-designware.c
> @@ -323,6 +323,12 @@ static u16 dw_pcie_find_vsec_capability(struct dw_pcie *pci,
>  	return 0;
>  }
>  
> +u16 dw_pcie_find_rasdes_capability(struct dw_pcie *pci)
> +{
> +	return dw_pcie_find_vsec_capability(pci, dwc_pcie_rasdes_vsec_ids);
> +}
> +EXPORT_SYMBOL_GPL(dw_pcie_find_rasdes_capability);
> +
>  int dw_pcie_read(void __iomem *addr, int size, u32 *val)
>  {
>  	if (!IS_ALIGNED((uintptr_t)addr, size)) {
> diff --git a/drivers/pci/controller/dwc/pcie-designware.h b/drivers/pci/controller/dwc/pcie-designware.h
> index 501d9ddfea16..7f9807d4e5de 100644
> --- a/drivers/pci/controller/dwc/pcie-designware.h
> +++ b/drivers/pci/controller/dwc/pcie-designware.h
> @@ -437,6 +437,11 @@ struct dw_pcie_ops {
>  	void	(*stop_link)(struct dw_pcie *pcie);
>  };
>  
> +struct debugfs_info {
> +	struct dentry		*debug_dir;
> +	void			*rasdes_info;
> +};
> +
>  struct dw_pcie {
>  	struct device		*dev;
>  	void __iomem		*dbi_base;
> @@ -465,6 +470,7 @@ struct dw_pcie {
>  	struct reset_control_bulk_data	core_rsts[DW_PCIE_NUM_CORE_RSTS];
>  	struct gpio_desc		*pe_rst;
>  	bool			suspended;
> +	struct debugfs_info	*debugfs;
>  };
>  
>  #define to_dw_pcie_from_pp(port) container_of((port), struct dw_pcie, pp)
> @@ -478,6 +484,7 @@ void dw_pcie_version_detect(struct dw_pcie *pci);
>  
>  u8 dw_pcie_find_capability(struct dw_pcie *pci, u8 cap);
>  u16 dw_pcie_find_ext_capability(struct dw_pcie *pci, u8 cap);
> +u16 dw_pcie_find_rasdes_capability(struct dw_pcie *pci);
>  
>  int dw_pcie_read(void __iomem *addr, int size, u32 *val);
>  int dw_pcie_write(void __iomem *addr, int size, u32 val);
> @@ -806,4 +813,18 @@ dw_pcie_ep_get_func_from_ep(struct dw_pcie_ep *ep, u8 func_no)
>  	return NULL;
>  }
>  #endif
> +
> +#ifdef CONFIG_PCIE_DW_DEBUGFS
> +int dwc_pcie_debugfs_init(struct dw_pcie *pci);
> +void dwc_pcie_debugfs_deinit(struct dw_pcie *pci);
> +#else
> +static inline int dwc_pcie_debugfs_init(struct dw_pcie *pci)
> +{
> +	return 0;
> +}
> +static inline void dwc_pcie_debugfs_deinit(struct dw_pcie *pci)
> +{
> +}
> +#endif
> +
>  #endif /* _PCIE_DESIGNWARE_H */
> diff --git a/include/linux/pcie-dwc.h b/include/linux/pcie-dwc.h
> index 40f3545731c8..6436e7fadc75 100644
> --- a/include/linux/pcie-dwc.h
> +++ b/include/linux/pcie-dwc.h
> @@ -28,6 +28,8 @@ static const struct dwc_pcie_vsec_id dwc_pcie_rasdes_vsec_ids[] = {
>  	  .vsec_id = 0x02, .vsec_rev = 0x4 },
>  	{ .vendor_id = PCI_VENDOR_ID_QCOM,
>  	  .vsec_id = 0x02, .vsec_rev = 0x4 },
> +	{ .vendor_id = PCI_VENDOR_ID_SAMSUNG,
> +	  .vsec_id = 0x02, .vsec_rev = 0x4 },
>  	{} /* terminator */
>  };
>  
> -- 
> 2.17.1
> 

Re: [PATCH v7 4/5] Add debugfs based error injection support in DWC

[Fan Ni]

On Fri, Feb 21, 2025 at 06:45:47PM +0530, Shradha Todi wrote:
> Add support to provide error injection interface to userspace. This set
> of debug registers are part of the RASDES feature present in DesignWare
> PCIe controllers.
> 
> Signed-off-by: Shradha Todi <shradha.t@samsung.com>
> ---

LGTM.

Reviewed-by: Fan Ni <fan.ni@samsung.com>

>  Documentation/ABI/testing/debugfs-dwc-pcie    |  70 ++++++++
>  .../controller/dwc/pcie-designware-debugfs.c  | 165 +++++++++++++++++-
>  2 files changed, 233 insertions(+), 2 deletions(-)
> 
> diff --git a/Documentation/ABI/testing/debugfs-dwc-pcie b/Documentation/ABI/testing/debugfs-dwc-pcie
> index e8ed34e988ef..6ee0897fe753 100644
> --- a/Documentation/ABI/testing/debugfs-dwc-pcie
> +++ b/Documentation/ABI/testing/debugfs-dwc-pcie
> @@ -11,3 +11,73 @@ Contact:	Shradha Todi <shradha.t@samsung.com>
>  Description:	(RW) Write the lane number to be checked as valid or invalid. Read
>  		will return the status of PIPE RXVALID signal of the selected lane.
>  		The default selected lane is Lane0.
> +
> +What:		/sys/kernel/debug/dwc_pcie_<dev>/rasdes_err_inj/<error>
> +Date:		Feburary 2025
> +Contact:	Shradha Todi <shradha.t@samsung.com>
> +Description:	rasdes_err_inj is the directory which can be used to inject errors in the
> +		system. The possible errors that can be injected are:
> +
> +		1) TLP LCRC error injection TX Path - tx_lcrc
> +		2) 16b CRC error injection of ACK/NAK DLLP - b16_crc_dllp
> +		3) 16b CRC error injection of Update-FC DLLP - b16_crc_upd_fc
> +		4) TLP ECRC error injection TX Path - tx_ecrc
> +		5) TLP's FCRC error injection TX Path - fcrc_tlp
> +		6) Parity error of TSOS - parity_tsos
> +		7) Parity error on SKPOS - parity_skpos
> +		8) LCRC error injection RX Path - rx_lcrc
> +		9) ECRC error injection RX Path - rx_ecrc
> +		10) TLPs SEQ# error - tlp_err_seq
> +		11) DLLPS ACK/NAK SEQ# error - ack_nak_dllp_seq
> +		12) ACK/NAK DLLPs transmission block - ack_nak_dllp
> +		13) UpdateFC DLLPs transmission block - upd_fc_dllp
> +		14) Always transmission for NAK DLLP - nak_dllp
> +		15) Invert SYNC header - inv_sync_hdr_sym
> +		16) COM/PAD TS1 order set - com_pad_ts1
> +		17) COM/PAD TS2 order set - com_pad_ts2
> +		18) COM/FTS FTS order set - com_fts
> +		19) COM/IDL E-idle order set - com_idl
> +		20) END/EDB symbol - end_edb
> +		21) STP/SDP symbol - stp_sdp
> +		22) COM/SKP SKP order set - com_skp
> +		23) Posted TLP Header credit value control - posted_tlp_hdr
> +		24) Non-Posted TLP Header credit value control - non_post_tlp_hdr
> +		25) Completion TLP Header credit value control - cmpl_tlp_hdr
> +		26) Posted TLP Data credit value control - posted_tlp_data
> +		27) Non-Posted TLP Data credit value control - non_post_tlp_data
> +		28) Completion TLP Data credit value control - cmpl_tlp_data
> +		29) Generates duplicate TLPs - duplicate_dllp
> +		30) Generates Nullified TLPs - nullified_tlp
> +
> +		(WO) Write to the attribute will prepare controller to inject the respective
> +		error in the next transmission of data. Parameter required to write will
> +		change in the following ways:
> +
> +		i) Errors 9) - 10) are sequence errors. The write command for these will be
> +
> +			echo <count> <diff> > /sys/kernel/debug/dwc_pcie_<dev>/rasdes_err_inj/<error>
> +
> +			<count>
> +				Number of errors to be injected
> +			<diff>
> +				The difference to add or subtract from natural sequence number to
> +				generate sequence error. Range (-4095 : 4095)
> +
> +		ii) Errors 23) - 28) are credit value error insertions. Write command:
> +
> +			echo <count> <diff> <vc> > /sys/kernel/debug/dwc_pcie_<dev>/rasdes_err_inj/<error>
> +
> +			<count>
> +				Number of errors to be injected
> +			<diff>
> +				The difference to add or subtract from UpdateFC credit value.
> +				Range (-4095 : 4095)
> +			<vc>
> +				Target VC number
> +
> +		iii) All other errors. Write command:
> +
> +			echo <count> > /sys/kernel/debug/dwc_pcie_<dev>/rasdes_err_inj/<error>
> +
> +			<count>
> +				Number of errors to be injected
> diff --git a/drivers/pci/controller/dwc/pcie-designware-debugfs.c b/drivers/pci/controller/dwc/pcie-designware-debugfs.c
> index 3887a6996706..b7260edd2336 100644
> --- a/drivers/pci/controller/dwc/pcie-designware-debugfs.c
> +++ b/drivers/pci/controller/dwc/pcie-designware-debugfs.c
> @@ -17,6 +17,20 @@
>  #define PIPE_DETECT_LANE		BIT(17)
>  #define LANE_SELECT			GENMASK(3, 0)
>  
> +#define ERR_INJ0_OFF			0x34
> +#define EINJ_VAL_DIFF			GENMASK(28, 16)
> +#define EINJ_VC_NUM			GENMASK(14, 12)
> +#define EINJ_TYPE_SHIFT			8
> +#define EINJ0_TYPE			GENMASK(11, 8)
> +#define EINJ1_TYPE			BIT(8)
> +#define EINJ2_TYPE			GENMASK(9, 8)
> +#define EINJ3_TYPE			GENMASK(10, 8)
> +#define EINJ4_TYPE			GENMASK(10, 8)
> +#define EINJ5_TYPE			BIT(8)
> +#define EINJ_COUNT			GENMASK(7, 0)
> +
> +#define ERR_INJ_ENABLE_REG		0x30
> +
>  #define DWC_DEBUGFS_BUF_MAX		128
>  
>  /**
> @@ -33,6 +47,72 @@ struct dwc_pcie_rasdes_info {
>  	struct mutex reg_event_lock;
>  };
>  
> +/**
> + * struct dwc_pcie_rasdes_priv - Stores file specific private data information
> + * @pci: Reference to the dw_pcie structure
> + * @idx: Index to point to specific file related information in array of structs
> + *
> + * All debugfs files will have this struct as its private data.
> + */
> +struct dwc_pcie_rasdes_priv {
> +	struct dw_pcie *pci;
> +	int idx;
> +};
> +
> +/**
> + * struct dwc_pcie_err_inj - Store details about each error injection supported by DWC RASDES
> + * @name: Name of the error that can be injected
> + * @err_inj_group: Group number to which the error belongs to. Value can range from 0 - 5
> + * @err_inj_type: Each group can have multiple types of error
> + */
> +struct dwc_pcie_err_inj {
> +	const char *name;
> +	u32 err_inj_group;
> +	u32 err_inj_type;
> +};
> +
> +static const struct dwc_pcie_err_inj err_inj_list[] = {
> +	{"tx_lcrc", 0x0, 0x0},
> +	{"b16_crc_dllp", 0x0, 0x1},
> +	{"b16_crc_upd_fc", 0x0, 0x2},
> +	{"tx_ecrc", 0x0, 0x3},
> +	{"fcrc_tlp", 0x0, 0x4},
> +	{"parity_tsos", 0x0, 0x5},
> +	{"parity_skpos", 0x0, 0x6},
> +	{"rx_lcrc", 0x0, 0x8},
> +	{"rx_ecrc", 0x0, 0xb},
> +	{"tlp_err_seq", 0x1, 0x0},
> +	{"ack_nak_dllp_seq", 0x1, 0x1},
> +	{"ack_nak_dllp", 0x2, 0x0},
> +	{"upd_fc_dllp", 0x2, 0x1},
> +	{"nak_dllp", 0x2, 0x2},
> +	{"inv_sync_hdr_sym", 0x3, 0x0},
> +	{"com_pad_ts1", 0x3, 0x1},
> +	{"com_pad_ts2", 0x3, 0x2},
> +	{"com_fts", 0x3, 0x3},
> +	{"com_idl", 0x3, 0x4},
> +	{"end_edb", 0x3, 0x5},
> +	{"stp_sdp", 0x3, 0x6},
> +	{"com_skp", 0x3, 0x7},
> +	{"posted_tlp_hdr", 0x4, 0x0},
> +	{"non_post_tlp_hdr", 0x4, 0x1},
> +	{"cmpl_tlp_hdr", 0x4, 0x2},
> +	{"posted_tlp_data", 0x4, 0x4},
> +	{"non_post_tlp_data", 0x4, 0x5},
> +	{"cmpl_tlp_data", 0x4, 0x6},
> +	{"duplicate_dllp", 0x5, 0x0},
> +	{"nullified_tlp", 0x5, 0x1},
> +};
> +
> +static const u32 err_inj_type_mask[] = {
> +	EINJ0_TYPE,
> +	EINJ1_TYPE,
> +	EINJ2_TYPE,
> +	EINJ3_TYPE,
> +	EINJ4_TYPE,
> +	EINJ5_TYPE,
> +};
> +
>  static ssize_t lane_detect_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
>  {
>  	struct dw_pcie *pci = file->private_data;
> @@ -93,6 +173,63 @@ static ssize_t rx_valid_write(struct file *file, const char __user *buf, size_t
>  	return lane_detect_write(file, buf, count, ppos);
>  }
>  
> +static ssize_t err_inj_write(struct file *file, const char __user *buf, size_t count, loff_t *ppos)
> +{
> +	struct dwc_pcie_rasdes_priv *pdata = file->private_data;
> +	struct dw_pcie *pci = pdata->pci;
> +	struct dwc_pcie_rasdes_info *rinfo = pci->debugfs->rasdes_info;
> +	u32 val, counter, vc_num, err_group, type_mask;
> +	int val_diff = 0;
> +	char *kern_buf;
> +
> +	err_group = err_inj_list[pdata->idx].err_inj_group;
> +	type_mask = err_inj_type_mask[err_group];
> +
> +	kern_buf = memdup_user_nul(buf, count);
> +	if (IS_ERR(kern_buf))
> +		return PTR_ERR(kern_buf);
> +
> +	if (err_group == 4) {
> +		val = sscanf(kern_buf, "%u %d %u", &counter, &val_diff, &vc_num);
> +		if ((val != 3) || (val_diff < -4095 || val_diff > 4095)) {
> +			kfree(kern_buf);
> +			return -EINVAL;
> +		}
> +	} else if (err_group == 1) {
> +		val = sscanf(kern_buf, "%u %d", &counter, &val_diff);
> +		if ((val != 2) || (val_diff < -4095 || val_diff > 4095)) {
> +			kfree(kern_buf);
> +			return -EINVAL;
> +		}
> +	} else {
> +		val = kstrtou32(kern_buf, 0, &counter);
> +		if (val) {
> +			kfree(kern_buf);
> +			return val;
> +		}
> +	}
> +
> +	val = dw_pcie_readl_dbi(pci, rinfo->ras_cap_offset + ERR_INJ0_OFF + (0x4 * err_group));
> +	val &= ~(type_mask | EINJ_COUNT);
> +	val |= ((err_inj_list[pdata->idx].err_inj_type << EINJ_TYPE_SHIFT) & type_mask);
> +	val |= FIELD_PREP(EINJ_COUNT, counter);
> +
> +	if (err_group == 1 || err_group == 4) {
> +		val &= ~(EINJ_VAL_DIFF);
> +		val |= FIELD_PREP(EINJ_VAL_DIFF, val_diff);
> +	}
> +	if (err_group == 4) {
> +		val &= ~(EINJ_VC_NUM);
> +		val |= FIELD_PREP(EINJ_VC_NUM, vc_num);
> +	}
> +
> +	dw_pcie_writel_dbi(pci, rinfo->ras_cap_offset + ERR_INJ0_OFF + (0x4 * err_group), val);
> +	dw_pcie_writel_dbi(pci, rinfo->ras_cap_offset + ERR_INJ_ENABLE_REG, (0x1 << err_group));
> +
> +	kfree(kern_buf);
> +	return count;
> +}
> +
>  #define dwc_debugfs_create(name)			\
>  debugfs_create_file(#name, 0644, rasdes_debug, pci,	\
>  			&dbg_ ## name ## _fops)
> @@ -107,6 +244,11 @@ static const struct file_operations dbg_ ## name ## _fops = {	\
>  DWC_DEBUGFS_FOPS(lane_detect);
>  DWC_DEBUGFS_FOPS(rx_valid);
>  
> +static const struct file_operations dwc_pcie_err_inj_ops = {
> +	.open = simple_open,
> +	.write = err_inj_write,
> +};
> +
>  static void dwc_pcie_rasdes_debugfs_deinit(struct dw_pcie *pci)
>  {
>  	struct dwc_pcie_rasdes_info *rinfo = pci->debugfs->rasdes_info;
> @@ -116,10 +258,11 @@ static void dwc_pcie_rasdes_debugfs_deinit(struct dw_pcie *pci)
>  
>  static int dwc_pcie_rasdes_debugfs_init(struct dw_pcie *pci, struct dentry *dir)
>  {
> -	struct dentry *rasdes_debug;
> +	struct dentry *rasdes_debug, *rasdes_err_inj;
>  	struct dwc_pcie_rasdes_info *rasdes_info;
> +	struct dwc_pcie_rasdes_priv *priv_tmp;
>  	struct device *dev = pci->dev;
> -	int ras_cap;
> +	int ras_cap, i, ret;
>  
>  	ras_cap = dw_pcie_find_rasdes_capability(pci);
>  	if (!ras_cap) {
> @@ -133,6 +276,7 @@ static int dwc_pcie_rasdes_debugfs_init(struct dw_pcie *pci, struct dentry *dir)
>  
>  	/* Create subdirectories for Debug, Error injection, Statistics */
>  	rasdes_debug = debugfs_create_dir("rasdes_debug", dir);
> +	rasdes_err_inj = debugfs_create_dir("rasdes_err_inj", dir);
>  
>  	mutex_init(&rasdes_info->reg_event_lock);
>  	rasdes_info->ras_cap_offset = ras_cap;
> @@ -142,7 +286,24 @@ static int dwc_pcie_rasdes_debugfs_init(struct dw_pcie *pci, struct dentry *dir)
>  	dwc_debugfs_create(lane_detect);
>  	dwc_debugfs_create(rx_valid);
>  
> +	/* Create debugfs files for Error injection subdirectory */
> +	for (i = 0; i < ARRAY_SIZE(err_inj_list); i++) {
> +		priv_tmp = devm_kzalloc(dev, sizeof(*priv_tmp), GFP_KERNEL);
> +		if (!priv_tmp) {
> +			ret = -ENOMEM;
> +			goto err_deinit;
> +		}
> +
> +		priv_tmp->idx = i;
> +		priv_tmp->pci = pci;
> +		debugfs_create_file(err_inj_list[i].name, 0200, rasdes_err_inj, priv_tmp,
> +				    &dwc_pcie_err_inj_ops);
> +	}
>  	return 0;
> +
> +err_deinit:
> +	dwc_pcie_rasdes_debugfs_deinit(pci);
> +	return ret;
>  }
>  
>  void dwc_pcie_debugfs_deinit(struct dw_pcie *pci)
> -- 
> 2.17.1
> 

Re: [PATCH v7 5/5] Add debugfs based statistical counter support in DWC

[Fan Ni]

On Fri, Feb 21, 2025 at 06:45:48PM +0530, Shradha Todi wrote:
> Add support to provide statistical counter interface to userspace. This set
> of debug registers are part of the RASDES feature present in DesignWare
> PCIe controllers.
> 
One comment inline.
> Signed-off-by: Shradha Todi <shradha.t@samsung.com>
> ---
>  Documentation/ABI/testing/debugfs-dwc-pcie    |  61 +++++
>  .../controller/dwc/pcie-designware-debugfs.c  | 229 +++++++++++++++++-
>  2 files changed, 289 insertions(+), 1 deletion(-)
> 
> diff --git a/Documentation/ABI/testing/debugfs-dwc-pcie b/Documentation/ABI/testing/debugfs-dwc-pcie
> index 6ee0897fe753..650a89b0511e 100644
> --- a/Documentation/ABI/testing/debugfs-dwc-pcie
> +++ b/Documentation/ABI/testing/debugfs-dwc-pcie
> @@ -81,3 +81,64 @@ Description:	rasdes_err_inj is the directory which can be used to inject errors
>  
>  			<count>
>  				Number of errors to be injected

...

> +
> +static ssize_t counter_value_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
> +{
> +	struct dwc_pcie_rasdes_priv *pdata = file->private_data;
> +	struct dw_pcie *pci = pdata->pci;
> +	struct dwc_pcie_rasdes_info *rinfo = pci->debugfs->rasdes_info;
> +	char debugfs_buf[DWC_DEBUGFS_BUF_MAX];
> +	ssize_t pos;
> +	u32 val;
> +
> +	mutex_lock(&rinfo->reg_event_lock);
> +	set_event_number(pdata, pci, rinfo);
> +	val = dw_pcie_readl_dbi(pci, rinfo->ras_cap_offset + RAS_DES_EVENT_COUNTER_DATA_REG);
> +	mutex_unlock(&rinfo->reg_event_lock);
> +	pos = scnprintf(debugfs_buf, DWC_DEBUGFS_BUF_MAX, "Counter value: %d\n", val);
> +
> +	return simple_read_from_buffer(buf, count, ppos, debugfs_buf, pos);
> +}

Do we need to check whether the counter is enabled or not for the event
before retrieving the counter value?

Fan
> +
>  #define dwc_debugfs_create(name)			\
>  debugfs_create_file(#name, 0644, rasdes_debug, pci,	\
>  			&dbg_ ## name ## _fops)
> @@ -249,6 +436,23 @@ static const struct file_operations dwc_pcie_err_inj_ops = {
>  	.write = err_inj_write,
>  };
>  
> +static const struct file_operations dwc_pcie_counter_enable_ops = {
> +	.open = simple_open,
> +	.read = counter_enable_read,
> +	.write = counter_enable_write,
> +};
> +
> +static const struct file_operations dwc_pcie_counter_lane_ops = {
> +	.open = simple_open,
> +	.read = counter_lane_read,
> +	.write = counter_lane_write,
> +};
> +
> +static const struct file_operations dwc_pcie_counter_value_ops = {
> +	.open = simple_open,
> +	.read = counter_value_read,
> +};
> +
>  static void dwc_pcie_rasdes_debugfs_deinit(struct dw_pcie *pci)
>  {
>  	struct dwc_pcie_rasdes_info *rinfo = pci->debugfs->rasdes_info;
> @@ -258,7 +462,7 @@ static void dwc_pcie_rasdes_debugfs_deinit(struct dw_pcie *pci)
>  
>  static int dwc_pcie_rasdes_debugfs_init(struct dw_pcie *pci, struct dentry *dir)
>  {
> -	struct dentry *rasdes_debug, *rasdes_err_inj;
> +	struct dentry *rasdes_debug, *rasdes_err_inj, *rasdes_event_counter, *rasdes_events;
>  	struct dwc_pcie_rasdes_info *rasdes_info;
>  	struct dwc_pcie_rasdes_priv *priv_tmp;
>  	struct device *dev = pci->dev;
> @@ -277,6 +481,7 @@ static int dwc_pcie_rasdes_debugfs_init(struct dw_pcie *pci, struct dentry *dir)
>  	/* Create subdirectories for Debug, Error injection, Statistics */
>  	rasdes_debug = debugfs_create_dir("rasdes_debug", dir);
>  	rasdes_err_inj = debugfs_create_dir("rasdes_err_inj", dir);
> +	rasdes_event_counter = debugfs_create_dir("rasdes_event_counter", dir);
>  
>  	mutex_init(&rasdes_info->reg_event_lock);
>  	rasdes_info->ras_cap_offset = ras_cap;
> @@ -299,6 +504,28 @@ static int dwc_pcie_rasdes_debugfs_init(struct dw_pcie *pci, struct dentry *dir)
>  		debugfs_create_file(err_inj_list[i].name, 0200, rasdes_err_inj, priv_tmp,
>  				    &dwc_pcie_err_inj_ops);
>  	}
> +
> +	/* Create debugfs files for Statistical counter subdirectory */
> +	for (i = 0; i < ARRAY_SIZE(event_list); i++) {
> +		priv_tmp = devm_kzalloc(dev, sizeof(*priv_tmp), GFP_KERNEL);
> +		if (!priv_tmp) {
> +			ret = -ENOMEM;
> +			goto err_deinit;
> +		}
> +
> +		priv_tmp->idx = i;
> +		priv_tmp->pci = pci;
> +		rasdes_events = debugfs_create_dir(event_list[i].name, rasdes_event_counter);
> +		if (event_list[i].group_no == 0 || event_list[i].group_no == 4) {
> +			debugfs_create_file("lane_select", 0644, rasdes_events,
> +					    priv_tmp, &dwc_pcie_counter_lane_ops);
> +		}
> +		debugfs_create_file("counter_value", 0444, rasdes_events, priv_tmp,
> +				    &dwc_pcie_counter_value_ops);
> +		debugfs_create_file("counter_enable", 0644, rasdes_events, priv_tmp,
> +				    &dwc_pcie_counter_enable_ops);
> +	}
> +
>  	return 0;
>  
>  err_deinit:
> -- 
> 2.17.1
> 

Re: [PATCH v7 5/5] Add debugfs based statistical counter support in DWC

[Krzysztof Wilczyński]

Hello,

[...]
> > +static ssize_t counter_value_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
> > +{
> > +	struct dwc_pcie_rasdes_priv *pdata = file->private_data;
> > +	struct dw_pcie *pci = pdata->pci;
> > +	struct dwc_pcie_rasdes_info *rinfo = pci->debugfs->rasdes_info;
> > +	char debugfs_buf[DWC_DEBUGFS_BUF_MAX];
> > +	ssize_t pos;
> > +	u32 val;
> > +
> > +	mutex_lock(&rinfo->reg_event_lock);
> > +	set_event_number(pdata, pci, rinfo);
> > +	val = dw_pcie_readl_dbi(pci, rinfo->ras_cap_offset + RAS_DES_EVENT_COUNTER_DATA_REG);
> > +	mutex_unlock(&rinfo->reg_event_lock);
> > +	pos = scnprintf(debugfs_buf, DWC_DEBUGFS_BUF_MAX, "Counter value: %d\n", val);
> > +
> > +	return simple_read_from_buffer(buf, count, ppos, debugfs_buf, pos);
> > +}
> 
> Do we need to check whether the counter is enabled or not for the event
> before retrieving the counter value?

I believe, we have a patch that aims to address, have a look at:

  https://lore.kernel.org/linux-pci/20250225171239.19574-1-manivannan.sadhasivam@linaro.org

Thank you!

	Krzysztof

Re: [PATCH v7 3/5] Add debugfs based silicon debug support in DWC

[Krzysztof Wilczyński]

Hello,

[...]
> > +int dwc_pcie_debugfs_init(struct dw_pcie *pci)
> > +{
> > +	char dirname[DWC_DEBUGFS_BUF_MAX];
> > +	struct device *dev = pci->dev;
> > +	struct debugfs_info *debugfs;
> > +	struct dentry *dir;
> > +	int ret;
> > +
> > +	/* Create main directory for each platform driver */
> > +	snprintf(dirname, DWC_DEBUGFS_BUF_MAX, "dwc_pcie_%s", dev_name(dev));
> > +	dir = debugfs_create_dir(dirname, NULL);
> > +	debugfs = devm_kzalloc(dev, sizeof(*debugfs), GFP_KERNEL);
> > +	if (!debugfs)
> > +		return -ENOMEM;
> > +
> > +	debugfs->debug_dir = dir;
> > +	pci->debugfs = debugfs;
> > +	ret = dwc_pcie_rasdes_debugfs_init(pci, dir);
> > +	if (ret)
> > +		dev_dbg(dev, "RASDES debugfs init failed\n");
> 
> What will happen if ret != 0? still return 0? 

Given that callers of dwc_pcie_debugfs_init() check for errors,
this probably should correctly bubble up any failure coming from
dwc_pcie_rasdes_debugfs_init().

I made updates to the code directly on the current branch, have a look:

  https://web.git.kernel.org/pub/scm/linux/kernel/git/pci/pci.git/commit/?h=controller/dwc&id=1ff54f4cbaed9ec6994844967c36cf7ada4cbe5e

Let me know if this is OK with you.

Good catch, thank you!

	Krzysztof

Re: [PATCH v7 0/5] Add support for debugfs based RAS DES feature in PCIe DW

[Krzysztof Wilczyński]

Hello,

[...]
> Applied to controller/dwc, thank you!

Updated the current branch with a few missing "Reviewed-by" tags from Fan Ni.

Thank you!

	Krzysztof

Re: [PATCH v7 0/5] Add support for debugfs based RAS DES feature in PCIe DW

[Krzysztof Wilczyński]

Hello,

> Thanks for adding the support!
> 
> Tested the patch in one of our internal SoC with DesginWare PCIe controller. The patch works fine.

I assume this was for an entire series, not any specific patch.

> Tested-by: Hrishikesh Deleep <hrishikesh.d@samsung.com>

Thank you for testing!  I will add your tag to the current branch.

	Krzysztof

Re: [PATCH v7 3/5] Add debugfs based silicon debug support in DWC

[Fan Ni]

On Tue, Mar 04, 2025 at 04:46:47AM +0900, Krzysztof Wilczyński wrote:
> Hello,
> 
> [...]
> > > +int dwc_pcie_debugfs_init(struct dw_pcie *pci)
> > > +{
> > > +	char dirname[DWC_DEBUGFS_BUF_MAX];
> > > +	struct device *dev = pci->dev;
> > > +	struct debugfs_info *debugfs;
> > > +	struct dentry *dir;
> > > +	int ret;
> > > +
> > > +	/* Create main directory for each platform driver */
> > > +	snprintf(dirname, DWC_DEBUGFS_BUF_MAX, "dwc_pcie_%s", dev_name(dev));
> > > +	dir = debugfs_create_dir(dirname, NULL);
> > > +	debugfs = devm_kzalloc(dev, sizeof(*debugfs), GFP_KERNEL);
> > > +	if (!debugfs)
> > > +		return -ENOMEM;
> > > +
> > > +	debugfs->debug_dir = dir;
> > > +	pci->debugfs = debugfs;
> > > +	ret = dwc_pcie_rasdes_debugfs_init(pci, dir);
> > > +	if (ret)
> > > +		dev_dbg(dev, "RASDES debugfs init failed\n");
> > 
> > What will happen if ret != 0? still return 0? 
> 
> Given that callers of dwc_pcie_debugfs_init() check for errors,
> this probably should correctly bubble up any failure coming from
> dwc_pcie_rasdes_debugfs_init().
> 
> I made updates to the code directly on the current branch, have a look:
> 
>   https://web.git.kernel.org/pub/scm/linux/kernel/git/pci/pci.git/commit/?h=controller/dwc&id=1ff54f4cbaed9ec6994844967c36cf7ada4cbe5e
> 
> Let me know if this is OK with you.

It looks good to me.

Feel free to add,
Reviewed-by: Fan Ni <fan.ni@samsung.com>

Fan
> 
> Good catch, thank you!
> 
> 	Krzysztof

Re: [PATCH v7 5/5] Add debugfs based statistical counter support in DWC

[Fan Ni]

On Tue, Mar 04, 2025 at 04:42:28AM +0900, Krzysztof Wilczyński wrote:
> Hello,
> 
> [...]
> > > +static ssize_t counter_value_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
> > > +{
> > > +	struct dwc_pcie_rasdes_priv *pdata = file->private_data;
> > > +	struct dw_pcie *pci = pdata->pci;
> > > +	struct dwc_pcie_rasdes_info *rinfo = pci->debugfs->rasdes_info;
> > > +	char debugfs_buf[DWC_DEBUGFS_BUF_MAX];
> > > +	ssize_t pos;
> > > +	u32 val;
> > > +
> > > +	mutex_lock(&rinfo->reg_event_lock);
> > > +	set_event_number(pdata, pci, rinfo);
> > > +	val = dw_pcie_readl_dbi(pci, rinfo->ras_cap_offset + RAS_DES_EVENT_COUNTER_DATA_REG);
> > > +	mutex_unlock(&rinfo->reg_event_lock);
> > > +	pos = scnprintf(debugfs_buf, DWC_DEBUGFS_BUF_MAX, "Counter value: %d\n", val);
> > > +
> > > +	return simple_read_from_buffer(buf, count, ppos, debugfs_buf, pos);
> > > +}
> > 
> > Do we need to check whether the counter is enabled or not for the event
> > before retrieving the counter value?
> 
> I believe, we have a patch that aims to address, have a look at:
> 
>   https://lore.kernel.org/linux-pci/20250225171239.19574-1-manivannan.sadhasivam@linaro.org

Maybe I missed something, that seems to fix counter_enable_read(), but
here is to retrieve counter value. 
How dw_pcie_readl_dbi() can return something like "Counter Disabled"? 

Fan
> 
> Thank you!
> 
> 	Krzysztof

Re: [PATCH v7 3/5] Add debugfs based silicon debug support in DWC

[Krzysztof Wilczyński]

Hello,

[...]
> > > What will happen if ret != 0? still return 0? 
> > 
> > Given that callers of dwc_pcie_debugfs_init() check for errors,
> > this probably should correctly bubble up any failure coming from
> > dwc_pcie_rasdes_debugfs_init().
> > 
> > I made updates to the code directly on the current branch, have a look:
> > 
> >   https://web.git.kernel.org/pub/scm/linux/kernel/git/pci/pci.git/commit/?h=controller/dwc&id=1ff54f4cbaed9ec6994844967c36cf7ada4cbe5e
> > 
> > Let me know if this is OK with you.
> 
> It looks good to me.
> 
> Feel free to add,
> Reviewed-by: Fan Ni <fan.ni@samsung.com>

Will do.  Thank you.

	Krzysztof

Re: [PATCH v7 4/5] Add debugfs based error injection support in DWC

[Krzysztof Wilczyński]

Hello,

[...]
> > +		29) Generates duplicate TLPs - duplicate_dllp
> > +		30) Generates Nullified TLPs - nullified_tlp
> 
> Would the above field called "duplicate_dllp" for duplicate TLPs be
> a potential typo?  Perhaps this should be called "duplicate_tlp"?
> 
> I wanted to make sure we have the correct field name.

Fan or Shradha, any thoughts on this?

Would the problem be with the name of the property of the description?

Thank you!

	Krzysztof

Re: [PATCH v7 3/5] Add debugfs based silicon debug support in DWC

[Geert Uytterhoeven]

Hi Krzysztof,

This patch is now commit 1ff54f4cbaed9ec6 ("PCI: dwc: Add debugfs
based Silicon Debug support for DWC") in pci/next (next-20250304).

On Mon, 3 Mar 2025 at 20:47, Krzysztof Wilczyński <kw@linux.com> wrote:
> [...]
> > > +int dwc_pcie_debugfs_init(struct dw_pcie *pci)
> > > +{
> > > +   char dirname[DWC_DEBUGFS_BUF_MAX];
> > > +   struct device *dev = pci->dev;
> > > +   struct debugfs_info *debugfs;
> > > +   struct dentry *dir;
> > > +   int ret;
> > > +
> > > +   /* Create main directory for each platform driver */
> > > +   snprintf(dirname, DWC_DEBUGFS_BUF_MAX, "dwc_pcie_%s", dev_name(dev));
> > > +   dir = debugfs_create_dir(dirname, NULL);
> > > +   debugfs = devm_kzalloc(dev, sizeof(*debugfs), GFP_KERNEL);
> > > +   if (!debugfs)
> > > +           return -ENOMEM;
> > > +
> > > +   debugfs->debug_dir = dir;
> > > +   pci->debugfs = debugfs;
> > > +   ret = dwc_pcie_rasdes_debugfs_init(pci, dir);
> > > +   if (ret)
> > > +           dev_dbg(dev, "RASDES debugfs init failed\n");
> >
> > What will happen if ret != 0? still return 0?

And that is exactly what happens on Gray Hawk Single with R-Car
V4M: dw_pcie_find_rasdes_capability() returns NULL, causing
dwc_pcie_rasdes_debugfs_init() to return -ENODEV.

> Given that callers of dwc_pcie_debugfs_init() check for errors,

Debugfs issues should never be propagated upstream!

> this probably should correctly bubble up any failure coming from
> dwc_pcie_rasdes_debugfs_init().
>
> I made updates to the code directly on the current branch, have a look:

So while applying, you changed this like:

            ret = dwc_pcie_rasdes_debugfs_init(pci, dir);
    -       if (ret)
    -               dev_dbg(dev, "RASDES debugfs init failed\n");
    +       if (ret) {
    +               dev_err(dev, "failed to initialize RAS DES debugfs\n");
    +               return ret;
    +       }

            return 0;

Hence this is now a fatal error, causing the probe to fail.
Unfortunately something fails during cleanup:

    pcie-rcar-gen4 e65d0000.pcie: failed to initialize RAS DES debugfs
    ------------[ cut here ]------------
    WARNING: CPU: 3 PID: 36 at kernel/irq/irqdomain.c:393
irq_domain_remove+0xa8/0xb0
    CPU: 3 UID: 0 PID: 36 Comm: kworker/u16:1 Not tainted
6.14.0-rc1-arm64-renesas-00134-g12c8c1363538 #2884
    Hardware name: Renesas Gray Hawk Single board based on r8a779h0 (DT)
    Workqueue: async async_run_entry_fn
    pstate: 60400005 (nZCv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
    pc : irq_domain_remove+0xa8/0xb0
    lr : irq_domain_remove+0x2c/0xb0
    sp : ffff8000819b3b10
    x29: ffff8000819b3b10 x28: 0000000000000000 x27: 0000000000000000
    x26: ffff00044011d800 x25: ffff80008053294c x24: ffff000441740400
    x23: ffff0004413a30f0 x22: ffff0004413a3130 x21: ffff0004413a3130
    x20: ffff8000815c0ec8 x19: ffff0004415f8240 x18: 00000000ffffffff
    x17: 6775626564205345 x16: 0000000000000020 x15: ffff8000819b37b0
    x14: 0000000000000004 x13: ffff8000813e9dd8 x12: 0000000000000000
    x11: ffff0004404b6448 x10: ffff800080e85400 x9 : 1fffe00088334301
    x8 : 0000000000000001 x7 : ffff0004419a1800 x6 : ffff0004419a1808
    x5 : ffff000441349030 x4 : fffffffffffffdc1 x3 : 0000000000000000
    x2 : ffff0004403e0000 x1 : 0000000000000000 x0 : ffff00044134f630
    Call trace:
     irq_domain_remove+0xa8/0xb0 (P)
     dw_pcie_host_init+0x394/0x710
     rcar_gen4_pcie_probe+0x104/0x2f8
     platform_probe+0x64/0xbc
     really_probe+0xb8/0x294
     __driver_probe_device+0x74/0x124
     driver_probe_device+0x3c/0x158
     __device_attach_driver+0xd4/0x154
     bus_for_each_drv+0x84/0xe0
     __device_attach_async_helper+0xac/0xd0
     async_run_entry_fn+0x30/0xd8
     process_one_work+0x144/0x280
     worker_thread+0x2c4/0x3cc
     kthread+0x128/0x1e0
     ret_from_fork+0x10/0x20
    ---[ end trace 0000000000000000 ]---

Worse, the PCI bus is still registered, so running "lspci" causes an Oops:

    Unable to handle kernel NULL pointer dereference at virtual
address 0000000000000004
    Mem abort info:
      ESR = 0x0000000096000004
      EC = 0x25: DABT (current EL), IL = 32 bits
      SET = 0, FnV = 0
      EA = 0, S1PTW = 0
      FSC = 0x04: level 0 translation fault
    Data abort info:
      ISV = 0, ISS = 0x00000004, ISS2 = 0x00000000
      CM = 0, WnR = 0, TnD = 0, TagAccess = 0
      GCS = 0, Overlay = 0, DirtyBit = 0, Xs = 0
    user pgtable: 4k pages, 48-bit VAs, pgdp=0000000483b53000
    [0000000000000004] pgd=0000000000000000, p4d=0000000000000000
    Internal error: Oops: 0000000096000004 [#1] PREEMPT SMP
    CPU: 3 UID: 0 PID: 707 Comm: lspci Tainted: G        W
6.14.0-rc1-arm64-renesas-00134-g12c8c1363538 #2884
    Tainted: [W]=WARN
    Hardware name: Renesas Gray Hawk Single board based on r8a779h0 (DT)
    pstate: 204000c5 (nzCv daIF +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
    pc : pci_generic_config_read+0x34/0xac
    lr : pci_generic_config_read+0x20/0xac
    sp : ffff8000825cbbf0
    x29: ffff8000825cbbf0 x28: ffff0004491c4b84 x27: 0000000000000004
    x26: 000000000000000f x25: ffff0004491c4b80 x24: 0000000000000040
    x23: 0000000000000040 x22: ffff8000825cbc64 x21: ffff8000816fb4f8
    x20: ffff8000825cbc14 x19: 0000000000000004 x18: 0000000000000000
    x17: 0000000000000000 x16: 0000000000000000 x15: 0000000000000000
    x14: 0000000000000000 x13: 0000000000000000 x12: 0000000000000000
    x11: 0000000000000000 x10: 0000000000000000 x9 : 0000000000000000
    x8 : 0000000000000000 x7 : ffff000442c653c0 x6 : ffff8000805163d0
    x5 : ffff8000804f3334 x4 : ffff8000825cbc14 x3 : ffff800080531990
    x2 : 0000000000000004 x1 : 0000000000000000 x0 : 0000000000000004
    Call trace:
     pci_generic_config_read+0x34/0xac (P)
     pci_user_read_config_dword+0x78/0x118
     pci_read_config+0xe4/0x29c
     sysfs_kf_bin_read+0x8c/0x9c
     kernfs_fop_read_iter+0x9c/0x19c
     vfs_read+0x24c/0x330
     __arm64_sys_pread64+0xac/0xc8
     invoke_syscall+0x44/0x100
     el0_svc_common.constprop.0+0x3c/0xd4
     do_el0_svc+0x18/0x20
     el0_svc+0x24/0xa8
     el0t_64_sync_handler+0x104/0x130
     el0t_64_sync+0x154/0x158
    Code: 7100067f 540002a0 71000a7f 54000160 (b9400000)
    ---[ end trace 0000000000000000 ]---
    note: lspci[707] exited with irqs disabled
    note: lspci[707] exited with preempt_count 1

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

Re: [PATCH v7 3/5] Add debugfs based silicon debug support in DWC

[Geert Uytterhoeven]

Hi Krzysztof,

(CC corrected)

This patch is now commit 1ff54f4cbaed9ec6 ("PCI: dwc: Add debugfs
based Silicon Debug support for DWC") in pci/next (next-20250304).

On Mon, 3 Mar 2025 at 20:47, Krzysztof Wilczyński <kw@linux.com> wrote:
> [...]
> > > +int dwc_pcie_debugfs_init(struct dw_pcie *pci)
> > > +{
> > > +   char dirname[DWC_DEBUGFS_BUF_MAX];
> > > +   struct device *dev = pci->dev;
> > > +   struct debugfs_info *debugfs;
> > > +   struct dentry *dir;
> > > +   int ret;
> > > +
> > > +   /* Create main directory for each platform driver */
> > > +   snprintf(dirname, DWC_DEBUGFS_BUF_MAX, "dwc_pcie_%s", dev_name(dev));
> > > +   dir = debugfs_create_dir(dirname, NULL);
> > > +   debugfs = devm_kzalloc(dev, sizeof(*debugfs), GFP_KERNEL);
> > > +   if (!debugfs)
> > > +           return -ENOMEM;
> > > +
> > > +   debugfs->debug_dir = dir;
> > > +   pci->debugfs = debugfs;
> > > +   ret = dwc_pcie_rasdes_debugfs_init(pci, dir);
> > > +   if (ret)
> > > +           dev_dbg(dev, "RASDES debugfs init failed\n");
> >
> > What will happen if ret != 0? still return 0?

And that is exactly what happens on Gray Hawk Single with R-Car
V4M: dw_pcie_find_rasdes_capability() returns NULL, causing
dwc_pcie_rasdes_debugfs_init() to return -ENODEV.

> Given that callers of dwc_pcie_debugfs_init() check for errors,

Debugfs issues should never be propagated upstream!

> this probably should correctly bubble up any failure coming from
> dwc_pcie_rasdes_debugfs_init().
>
> I made updates to the code directly on the current branch, have a look:

So while applying, you changed this like:

            ret = dwc_pcie_rasdes_debugfs_init(pci, dir);
    -       if (ret)
    -               dev_dbg(dev, "RASDES debugfs init failed\n");
    +       if (ret) {
    +               dev_err(dev, "failed to initialize RAS DES debugfs\n");
    +               return ret;
    +       }

            return 0;

Hence this is now a fatal error, causing the probe to fail.
Unfortunately something fails during cleanup:

    pcie-rcar-gen4 e65d0000.pcie: failed to initialize RAS DES debugfs
    ------------[ cut here ]------------
    WARNING: CPU: 3 PID: 36 at kernel/irq/irqdomain.c:393
irq_domain_remove+0xa8/0xb0
    CPU: 3 UID: 0 PID: 36 Comm: kworker/u16:1 Not tainted
6.14.0-rc1-arm64-renesas-00134-g12c8c1363538 #2884
    Hardware name: Renesas Gray Hawk Single board based on r8a779h0 (DT)
    Workqueue: async async_run_entry_fn
    pstate: 60400005 (nZCv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
    pc : irq_domain_remove+0xa8/0xb0
    lr : irq_domain_remove+0x2c/0xb0
    sp : ffff8000819b3b10
    x29: ffff8000819b3b10 x28: 0000000000000000 x27: 0000000000000000
    x26: ffff00044011d800 x25: ffff80008053294c x24: ffff000441740400
    x23: ffff0004413a30f0 x22: ffff0004413a3130 x21: ffff0004413a3130
    x20: ffff8000815c0ec8 x19: ffff0004415f8240 x18: 00000000ffffffff
    x17: 6775626564205345 x16: 0000000000000020 x15: ffff8000819b37b0
    x14: 0000000000000004 x13: ffff8000813e9dd8 x12: 0000000000000000
    x11: ffff0004404b6448 x10: ffff800080e85400 x9 : 1fffe00088334301
    x8 : 0000000000000001 x7 : ffff0004419a1800 x6 : ffff0004419a1808
    x5 : ffff000441349030 x4 : fffffffffffffdc1 x3 : 0000000000000000
    x2 : ffff0004403e0000 x1 : 0000000000000000 x0 : ffff00044134f630
    Call trace:
     irq_domain_remove+0xa8/0xb0 (P)
     dw_pcie_host_init+0x394/0x710
     rcar_gen4_pcie_probe+0x104/0x2f8
     platform_probe+0x64/0xbc
     really_probe+0xb8/0x294
     __driver_probe_device+0x74/0x124
     driver_probe_device+0x3c/0x158
     __device_attach_driver+0xd4/0x154
     bus_for_each_drv+0x84/0xe0
     __device_attach_async_helper+0xac/0xd0
     async_run_entry_fn+0x30/0xd8
     process_one_work+0x144/0x280
     worker_thread+0x2c4/0x3cc
     kthread+0x128/0x1e0
     ret_from_fork+0x10/0x20
    ---[ end trace 0000000000000000 ]---

Worse, the PCI bus is still registered, so running "lspci" causes an Oops:

    Unable to handle kernel NULL pointer dereference at virtual
address 0000000000000004
    Mem abort info:
      ESR = 0x0000000096000004
      EC = 0x25: DABT (current EL), IL = 32 bits
      SET = 0, FnV = 0
      EA = 0, S1PTW = 0
      FSC = 0x04: level 0 translation fault
    Data abort info:
      ISV = 0, ISS = 0x00000004, ISS2 = 0x00000000
      CM = 0, WnR = 0, TnD = 0, TagAccess = 0
      GCS = 0, Overlay = 0, DirtyBit = 0, Xs = 0
    user pgtable: 4k pages, 48-bit VAs, pgdp=0000000483b53000
    [0000000000000004] pgd=0000000000000000, p4d=0000000000000000
    Internal error: Oops: 0000000096000004 [#1] PREEMPT SMP
    CPU: 3 UID: 0 PID: 707 Comm: lspci Tainted: G
W6.14.0-rc1-arm64-renesas-00134-g12c8c1363538 #2884
    Tainted: [W]=WARN
    Hardware name: Renesas Gray Hawk Single board based on r8a779h0 (DT)
    pstate: 204000c5 (nzCv daIF +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
    pc : pci_generic_config_read+0x34/0xac
    lr : pci_generic_config_read+0x20/0xac
    sp : ffff8000825cbbf0
    x29: ffff8000825cbbf0 x28: ffff0004491c4b84 x27: 0000000000000004
    x26: 000000000000000f x25: ffff0004491c4b80 x24: 0000000000000040
    x23: 0000000000000040 x22: ffff8000825cbc64 x21: ffff8000816fb4f8
    x20: ffff8000825cbc14 x19: 0000000000000004 x18: 0000000000000000
    x17: 0000000000000000 x16: 0000000000000000 x15: 0000000000000000
    x14: 0000000000000000 x13: 0000000000000000 x12: 0000000000000000
    x11: 0000000000000000 x10: 0000000000000000 x9 : 0000000000000000
    x8 : 0000000000000000 x7 : ffff000442c653c0 x6 : ffff8000805163d0
    x5 : ffff8000804f3334 x4 : ffff8000825cbc14 x3 : ffff800080531990
    x2 : 0000000000000004 x1 : 0000000000000000 x0 : 0000000000000004
    Call trace:
     pci_generic_config_read+0x34/0xac (P)
     pci_user_read_config_dword+0x78/0x118
     pci_read_config+0xe4/0x29c
     sysfs_kf_bin_read+0x8c/0x9c
     kernfs_fop_read_iter+0x9c/0x19c
     vfs_read+0x24c/0x330
     __arm64_sys_pread64+0xac/0xc8
     invoke_syscall+0x44/0x100
     el0_svc_common.constprop.0+0x3c/0xd4
     do_el0_svc+0x18/0x20
     el0_svc+0x24/0xa8
     el0t_64_sync_handler+0x104/0x130
     el0t_64_sync+0x154/0x158
    Code: 7100067f 540002a0 71000a7f 54000160 (b9400000)
    ---[ end trace 0000000000000000 ]---
    note: lspci[707] exited with irqs disabled
    note: lspci[707] exited with preempt_count 1

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

Re: [PATCH v7 3/5] Add debugfs based silicon debug support in DWC

[Manivannan Sadhasivam]

+ Geert (who reported the regression in -next)

On Tue, Mar 04, 2025 at 04:46:47AM +0900, Krzysztof Wilczyński wrote:
> Hello,
> 
> [...]
> > > +int dwc_pcie_debugfs_init(struct dw_pcie *pci)
> > > +{
> > > +	char dirname[DWC_DEBUGFS_BUF_MAX];
> > > +	struct device *dev = pci->dev;
> > > +	struct debugfs_info *debugfs;
> > > +	struct dentry *dir;
> > > +	int ret;
> > > +
> > > +	/* Create main directory for each platform driver */
> > > +	snprintf(dirname, DWC_DEBUGFS_BUF_MAX, "dwc_pcie_%s", dev_name(dev));
> > > +	dir = debugfs_create_dir(dirname, NULL);
> > > +	debugfs = devm_kzalloc(dev, sizeof(*debugfs), GFP_KERNEL);
> > > +	if (!debugfs)
> > > +		return -ENOMEM;
> > > +
> > > +	debugfs->debug_dir = dir;
> > > +	pci->debugfs = debugfs;
> > > +	ret = dwc_pcie_rasdes_debugfs_init(pci, dir);
> > > +	if (ret)
> > > +		dev_dbg(dev, "RASDES debugfs init failed\n");
> > 
> > What will happen if ret != 0? still return 0? 
> 
> Given that callers of dwc_pcie_debugfs_init() check for errors,
> this probably should correctly bubble up any failure coming from
> dwc_pcie_rasdes_debugfs_init().
> 
> I made updates to the code directly on the current branch, have a look:
> 
>   https://web.git.kernel.org/pub/scm/linux/kernel/git/pci/pci.git/commit/?h=controller/dwc&id=1ff54f4cbaed9ec6994844967c36cf7ada4cbe5e
> 
> Let me know if this is OK with you.
> 

If the SoC has no RASDES capability, then this call is bound to fail (which will
break existing platforms). I'd propose to return 0 if
dw_pcie_find_rasdes_capability() fails in addition to this change:

diff --git a/drivers/pci/controller/dwc/pcie-designware-debugfs.c b/drivers/pci/controller/dwc/pcie-designware-debugfs.c
index dca1e9999113..7277a21e30d5 100644
--- a/drivers/pci/controller/dwc/pcie-designware-debugfs.c
+++ b/drivers/pci/controller/dwc/pcie-designware-debugfs.c
@@ -471,7 +471,7 @@ static int dwc_pcie_rasdes_debugfs_init(struct dw_pcie *pci, struct dentry *dir)
        ras_cap = dw_pcie_find_rasdes_capability(pci);
        if (!ras_cap) {
                dev_dbg(dev, "no RASDES capability available\n");
-               return -ENODEV;
+               return 0;
        }
 
        rasdes_info = devm_kzalloc(dev, sizeof(*rasdes_info), GFP_KERNEL);

This will fix the regressions like the one reported by Geert:

https://lore.kernel.org/linux-pci/CAMuHMdWuCJAd-mCpCoseThureCKnnep4T-Z0h1_WJ1BOf2ZeDg@mail.gmail.com/

- Mani

-- 
மணிவண்ணன் சதாசிவம்

Re: [PATCH v7 4/5] Add debugfs based error injection support in DWC

[Manivannan Sadhasivam]

On Mon, Mar 03, 2025 at 06:52:00PM +0900, Krzysztof Wilczyński wrote:
> Hello,
> 
> [...]
> > +		29) Generates duplicate TLPs - duplicate_dllp
> > +		30) Generates Nullified TLPs - nullified_tlp
> 
> Would the above field called "duplicate_dllp" for duplicate TLPs be
> a potential typo?  Perhaps this should be called "duplicate_tlp"?
> 

Looks like a typo. As per Synopsys documentation, there is only 'duplicate TLP'
field.

Good catch!

- Mani

-- 
மணிவண்ணன் சதாசிவம்

Re: [PATCH v7 5/5] Add debugfs based statistical counter support in DWC

[Manivannan Sadhasivam]

On Mon, Mar 03, 2025 at 01:03:12PM -0800, Fan Ni wrote:
> On Tue, Mar 04, 2025 at 04:42:28AM +0900, Krzysztof Wilczyński wrote:
> > Hello,
> > 
> > [...]
> > > > +static ssize_t counter_value_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
> > > > +{
> > > > +	struct dwc_pcie_rasdes_priv *pdata = file->private_data;
> > > > +	struct dw_pcie *pci = pdata->pci;
> > > > +	struct dwc_pcie_rasdes_info *rinfo = pci->debugfs->rasdes_info;
> > > > +	char debugfs_buf[DWC_DEBUGFS_BUF_MAX];
> > > > +	ssize_t pos;
> > > > +	u32 val;
> > > > +
> > > > +	mutex_lock(&rinfo->reg_event_lock);
> > > > +	set_event_number(pdata, pci, rinfo);
> > > > +	val = dw_pcie_readl_dbi(pci, rinfo->ras_cap_offset + RAS_DES_EVENT_COUNTER_DATA_REG);
> > > > +	mutex_unlock(&rinfo->reg_event_lock);
> > > > +	pos = scnprintf(debugfs_buf, DWC_DEBUGFS_BUF_MAX, "Counter value: %d\n", val);
> > > > +
> > > > +	return simple_read_from_buffer(buf, count, ppos, debugfs_buf, pos);
> > > > +}
> > > 
> > > Do we need to check whether the counter is enabled or not for the event
> > > before retrieving the counter value?
> > 
> > I believe, we have a patch that aims to address, have a look at:
> > 
> >   https://lore.kernel.org/linux-pci/20250225171239.19574-1-manivannan.sadhasivam@linaro.org
> 
> Maybe I missed something, that seems to fix counter_enable_read(), but
> here is to retrieve counter value. 
> How dw_pcie_readl_dbi() can return something like "Counter Disabled"? 
> 

Only way to know if a counter is enabled by reading back the status. And that is
what the patch is doing.

- Mani

-- 
மணிவண்ணன் சதாசிவம்

Re: [PATCH v7 4/5] Add debugfs based error injection support in DWC

[Krzysztof Wilczyński]

Hello,

[....]
> > > +		29) Generates duplicate TLPs - duplicate_dllp
> > > +		30) Generates Nullified TLPs - nullified_tlp
> > 
> > Would the above field called "duplicate_dllp" for duplicate TLPs be
> > a potential typo?  Perhaps this should be called "duplicate_tlp"?
> > 
> 
> Looks like a typo. As per Synopsys documentation, there is only 'duplicate TLP'
> field.
> 
> Good catch!

Updated.  Thank you!

	Krzysztof

Re: [PATCH v7 3/5] Add debugfs based silicon debug support in DWC

[Krzysztof Wilczyński]

Hello,

> This patch is now commit 1ff54f4cbaed9ec6 ("PCI: dwc: Add debugfs
> based Silicon Debug support for DWC") in pci/next (next-20250304).
> 
> On Mon, 3 Mar 2025 at 20:47, Krzysztof Wilczyński <kw@linux.com> wrote:
> > [...]
> > > > +int dwc_pcie_debugfs_init(struct dw_pcie *pci)
> > > > +{
> > > > +   char dirname[DWC_DEBUGFS_BUF_MAX];
> > > > +   struct device *dev = pci->dev;
> > > > +   struct debugfs_info *debugfs;
> > > > +   struct dentry *dir;
> > > > +   int ret;
> > > > +
> > > > +   /* Create main directory for each platform driver */
> > > > +   snprintf(dirname, DWC_DEBUGFS_BUF_MAX, "dwc_pcie_%s", dev_name(dev));
> > > > +   dir = debugfs_create_dir(dirname, NULL);
> > > > +   debugfs = devm_kzalloc(dev, sizeof(*debugfs), GFP_KERNEL);
> > > > +   if (!debugfs)
> > > > +           return -ENOMEM;
> > > > +
> > > > +   debugfs->debug_dir = dir;
> > > > +   pci->debugfs = debugfs;
> > > > +   ret = dwc_pcie_rasdes_debugfs_init(pci, dir);
> > > > +   if (ret)
> > > > +           dev_dbg(dev, "RASDES debugfs init failed\n");
> > >
> > > What will happen if ret != 0? still return 0?
> 
> And that is exactly what happens on Gray Hawk Single with R-Car
> V4M: dw_pcie_find_rasdes_capability() returns NULL, causing
> dwc_pcie_rasdes_debugfs_init() to return -ENODEV.

Thank you for testing and for catching this issue.  Much appreciated.

> > Given that callers of dwc_pcie_debugfs_init() check for errors,
> 
> Debugfs issues should never be propagated upstream!

Makes complete sense.  Sorry for breaking things here!

> > this probably should correctly bubble up any failure coming from
> > dwc_pcie_rasdes_debugfs_init().
> >
> > I made updates to the code directly on the current branch, have a look:
> 
> So while applying, you changed this like:
> 
>             ret = dwc_pcie_rasdes_debugfs_init(pci, dir);
>     -       if (ret)
>     -               dev_dbg(dev, "RASDES debugfs init failed\n");
>     +       if (ret) {
>     +               dev_err(dev, "failed to initialize RAS DES debugfs\n");
>     +               return ret;
>     +       }
> 
>             return 0;
> 
> Hence this is now a fatal error, causing the probe to fail.

I removed the changed, and also move the log level to be a warning, per:

  https://web.git.kernel.org/pub/scm/linux/kernel/git/pci/pci.git/commit/?h=controller/dwc&id=c6759a967e69aba16aef0d92f43e527b112e98a5

Would this be acceptable here?

Mani, would this be acceptable to you, too?  Given that you posted the
following recently:

  https://lore.kernel.org/linux-pci/20250303200055.GA1881771@rocinante/T/#mab9cbd5834390d259afea056eee9a73d8c3b435f

That said, perhaps moving the log level to a debug would be better served here.

	Krzysztof

RE: [PATCH v7 3/5] Add debugfs based silicon debug support in DWC

[Shradha Todi]

> -----Original Message-----
> From: Krzysztof Wilczyński <kw@linux.com>
> Sent: 04 March 2025 21:17
> To: Geert Uytterhoeven <geert@linux-m68k.org>
> Cc: Fan Ni <nifan.cxl@gmail.com>; Shradha Todi <shradha.t@samsung.com>; linux-kernel@vger.kernel.org; linux-
> pci@vger.kernel.org; linux-arm-kernel@lists.infradead.org; linux-perf-users@vger.kernel.org; manivannan.sadhasivam@linaro.org;
> lpieralisi@kernel.org; robh@kernel.org; bhelgaas@google.com; jingoohan1@gmail.com; Jonathan.Cameron@huawei.com;
> a.manzanares@samsung.com; pankaj.dubey@samsung.com; cassel@kernel.org; 18255117159@163.com;
> xueshuai@linux.alibaba.com; renyu.zj@linux.alibaba.com; will@kernel.org; mark.rutland@arm.com; Yoshihiro Shimoda
> <yoshihiro.shimoda.uh@renesas.com>; Linux-Renesas <linux-renesas-soc@vger.kernel.org>
> Subject: Re: [PATCH v7 3/5] Add debugfs based silicon debug support in DWC
> 
> Hello,
> 
> > This patch is now commit 1ff54f4cbaed9ec6 ("PCI: dwc: Add debugfs
> > based Silicon Debug support for DWC") in pci/next (next-20250304).
> >
> > On Mon, 3 Mar 2025 at 20:47, Krzysztof Wilczyński <kw@linux.com> wrote:
> > > [...]
> > > > > +int dwc_pcie_debugfs_init(struct dw_pcie *pci) {
> > > > > +   char dirname[DWC_DEBUGFS_BUF_MAX];
> > > > > +   struct device *dev = pci->dev;
> > > > > +   struct debugfs_info *debugfs;
> > > > > +   struct dentry *dir;
> > > > > +   int ret;
> > > > > +
> > > > > +   /* Create main directory for each platform driver */
> > > > > +   snprintf(dirname, DWC_DEBUGFS_BUF_MAX, "dwc_pcie_%s", dev_name(dev));
> > > > > +   dir = debugfs_create_dir(dirname, NULL);
> > > > > +   debugfs = devm_kzalloc(dev, sizeof(*debugfs), GFP_KERNEL);
> > > > > +   if (!debugfs)
> > > > > +           return -ENOMEM;
> > > > > +
> > > > > +   debugfs->debug_dir = dir;
> > > > > +   pci->debugfs = debugfs;
> > > > > +   ret = dwc_pcie_rasdes_debugfs_init(pci, dir);
> > > > > +   if (ret)
> > > > > +           dev_dbg(dev, "RASDES debugfs init failed\n");
> > > >
> > > > What will happen if ret != 0? still return 0?
> >
> > And that is exactly what happens on Gray Hawk Single with R-Car
> > V4M: dw_pcie_find_rasdes_capability() returns NULL, causing
> > dwc_pcie_rasdes_debugfs_init() to return -ENODEV.
> 
> Thank you for testing and for catching this issue.  Much appreciated.
> 
> > > Given that callers of dwc_pcie_debugfs_init() check for errors,
> >
> > Debugfs issues should never be propagated upstream!
> 
> Makes complete sense.  Sorry for breaking things here!
> 
> > > this probably should correctly bubble up any failure coming from
> > > dwc_pcie_rasdes_debugfs_init().
> > >
> > > I made updates to the code directly on the current branch, have a look:
> >
> > So while applying, you changed this like:
> >
> >             ret = dwc_pcie_rasdes_debugfs_init(pci, dir);
> >     -       if (ret)
> >     -               dev_dbg(dev, "RASDES debugfs init failed\n");
> >     +       if (ret) {
> >     +               dev_err(dev, "failed to initialize RAS DES debugfs\n");
> >     +               return ret;
> >     +       }
> >
> >             return 0;
> >
> > Hence this is now a fatal error, causing the probe to fail.
> 
> I removed the changed, and also move the log level to be a warning, per:
> 

Hey Krzysztof,
I think we shouldn't move the log level to be a WARN. I believe many controllers might not support
RAS DES feature in their design and giving a warn dump would draw unnecessary attention.
My opinion is to silently let it fail unless the user is actually interested in getting the RAS DES feature up.
We can wait for Mani's response though. But good catch to also add the error type, that's definitely a more
informative error log.

> 
> https://web.git.kernel.org/pub/scm/linux/kernel/git/pci/pci.git/commit/?h=controller/dwc&id=c6759a967e69aba16aef0d92f43e527b
> 112e98a5
> 
> Would this be acceptable here?
> 
> Mani, would this be acceptable to you, too?  Given that you posted the following recently:
> 
>   https://lore.kernel.org/linux-pci/20250303200055.GA1881771@rocinante/T/#mab9cbd5834390d259afea056eee9a73d8c3b435f
> 
> That said, perhaps moving the log level to a debug would be better served here.
> 
> 	Krzysztof

RE: [PATCH v7 4/5] Add debugfs based error injection support in DWC

[Shradha Todi]

> -----Original Message-----
> From: Krzysztof Wilczyński <kw@linux.com>
> Sent: 04 March 2025 21:05
> To: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
> Cc: Shradha Todi <shradha.t@samsung.com>; linux-kernel@vger.kernel.org; linux-pci@vger.kernel.org; linux-arm-
> kernel@lists.infradead.org; linux-perf-users@vger.kernel.org; lpieralisi@kernel.org; robh@kernel.org; bhelgaas@google.com;
> jingoohan1@gmail.com; Jonathan.Cameron@huawei.com; fan.ni@samsung.com; nifan.cxl@gmail.com;
> a.manzanares@samsung.com; pankaj.dubey@samsung.com; cassel@kernel.org; 18255117159@163.com;
> xueshuai@linux.alibaba.com; renyu.zj@linux.alibaba.com; will@kernel.org; mark.rutland@arm.com
> Subject: Re: [PATCH v7 4/5] Add debugfs based error injection support in DWC
> 
> Hello,
> 
> [....]
> > > > +		29) Generates duplicate TLPs - duplicate_dllp
> > > > +		30) Generates Nullified TLPs - nullified_tlp
> > >
> > > Would the above field called "duplicate_dllp" for duplicate TLPs be
> > > a potential typo?  Perhaps this should be called "duplicate_tlp"?
> > >
> >
> > Looks like a typo. As per Synopsys documentation, there is only 'duplicate TLP'
> > field.
> >
> > Good catch!
> 
> Updated.  Thank you!
> 

Sorry, this was a typo. Krzysztof, we need another change for this typo.

diff --git a/drivers/pci/controller/dwc/pcie-designware-debugfs.c b/drivers/pci/controller/dwc/pcie-designware-debugfs.c
index 729ac14ff700..4d77db3ca6ae 100644
--- a/drivers/pci/controller/dwc/pcie-designware-debugfs.c
+++ b/drivers/pci/controller/dwc/pcie-designware-debugfs.c
@@ -113,7 +113,7 @@ static const struct dwc_pcie_err_inj err_inj_list[] = {
        {"posted_tlp_data", 0x4, 0x4},
        {"non_post_tlp_data", 0x4, 0x5},
        {"cmpl_tlp_data", 0x4, 0x6},
-       {"duplicate_dllp", 0x5, 0x0},
+       {"duplicate_tlp", 0x5, 0x0},
        {"nullified_tlp", 0x5, 0x1},
 };

So sorry for the inconvenience! Should I post a patch for this?

> 	Krzysztof

RE: [PATCH v7 5/5] Add debugfs based statistical counter support in DWC

[Shradha Todi]

> -----Original Message-----
> From: Fan Ni <nifan.cxl@gmail.com>
> Sent: 04 March 2025 02:33
> To: Krzysztof Wilczyński <kw@linux.com>
> Cc: Fan Ni <nifan.cxl@gmail.com>; Shradha Todi <shradha.t@samsung.com>; linux-kernel@vger.kernel.org; linux-
> pci@vger.kernel.org; linux-arm-kernel@lists.infradead.org; linux-perf-users@vger.kernel.org; manivannan.sadhasivam@linaro.org;
> lpieralisi@kernel.org; robh@kernel.org; bhelgaas@google.com; jingoohan1@gmail.com; Jonathan.Cameron@huawei.com;
> a.manzanares@samsung.com; pankaj.dubey@samsung.com; cassel@kernel.org; 18255117159@163.com;
> xueshuai@linux.alibaba.com; renyu.zj@linux.alibaba.com; will@kernel.org; mark.rutland@arm.com
> Subject: Re: [PATCH v7 5/5] Add debugfs based statistical counter support in DWC
> 
> On Tue, Mar 04, 2025 at 04:42:28AM +0900, Krzysztof Wilczyński wrote:
> > Hello,
> >
> > [...]
> > > > +static ssize_t counter_value_read(struct file *file, char __user
> > > > +*buf, size_t count, loff_t *ppos) {
> > > > +	struct dwc_pcie_rasdes_priv *pdata = file->private_data;
> > > > +	struct dw_pcie *pci = pdata->pci;
> > > > +	struct dwc_pcie_rasdes_info *rinfo = pci->debugfs->rasdes_info;
> > > > +	char debugfs_buf[DWC_DEBUGFS_BUF_MAX];
> > > > +	ssize_t pos;
> > > > +	u32 val;
> > > > +
> > > > +	mutex_lock(&rinfo->reg_event_lock);
> > > > +	set_event_number(pdata, pci, rinfo);
> > > > +	val = dw_pcie_readl_dbi(pci, rinfo->ras_cap_offset + RAS_DES_EVENT_COUNTER_DATA_REG);
> > > > +	mutex_unlock(&rinfo->reg_event_lock);
> > > > +	pos = scnprintf(debugfs_buf, DWC_DEBUGFS_BUF_MAX, "Counter
> > > > +value: %d\n", val);
> > > > +
> > > > +	return simple_read_from_buffer(buf, count, ppos, debugfs_buf,
> > > > +pos); }
> > >
> > > Do we need to check whether the counter is enabled or not for the
> > > event before retrieving the counter value?
> >
> > I believe, we have a patch that aims to address, have a look at:
> >
> >
> > https://lore.kernel.org/linux-pci/20250225171239.19574-1-manivannan.sa
> > dhasivam@linaro.org
> 
> Maybe I missed something, that seems to fix counter_enable_read(), but here is to retrieve counter value.
> How dw_pcie_readl_dbi() can return something like "Counter Disabled"?
> 
> Fan

Hey Fan,
So the counter value will show 0 in case it is disabled so there will not be any issues as per say. We could add the
check here but I feel I have already exposed the functionality to check if a counter is enabled or disabled, (by reading the
counter_enable debugfs entry) so this could be handled in user space to only read the counter if it's enabled.

> >
> > Thank you!
> >
> > 	Krzysztof

Re: [PATCH v7 3/5] Add debugfs based silicon debug support in DWC

[Manivannan Sadhasivam]

On Wed, Mar 05, 2025 at 12:46:38AM +0900, Krzysztof Wilczyński wrote:
> Hello,
> 
> > This patch is now commit 1ff54f4cbaed9ec6 ("PCI: dwc: Add debugfs
> > based Silicon Debug support for DWC") in pci/next (next-20250304).
> > 
> > On Mon, 3 Mar 2025 at 20:47, Krzysztof Wilczyński <kw@linux.com> wrote:
> > > [...]
> > > > > +int dwc_pcie_debugfs_init(struct dw_pcie *pci)
> > > > > +{
> > > > > +   char dirname[DWC_DEBUGFS_BUF_MAX];
> > > > > +   struct device *dev = pci->dev;
> > > > > +   struct debugfs_info *debugfs;
> > > > > +   struct dentry *dir;
> > > > > +   int ret;
> > > > > +
> > > > > +   /* Create main directory for each platform driver */
> > > > > +   snprintf(dirname, DWC_DEBUGFS_BUF_MAX, "dwc_pcie_%s", dev_name(dev));
> > > > > +   dir = debugfs_create_dir(dirname, NULL);
> > > > > +   debugfs = devm_kzalloc(dev, sizeof(*debugfs), GFP_KERNEL);
> > > > > +   if (!debugfs)
> > > > > +           return -ENOMEM;
> > > > > +
> > > > > +   debugfs->debug_dir = dir;
> > > > > +   pci->debugfs = debugfs;
> > > > > +   ret = dwc_pcie_rasdes_debugfs_init(pci, dir);
> > > > > +   if (ret)
> > > > > +           dev_dbg(dev, "RASDES debugfs init failed\n");
> > > >
> > > > What will happen if ret != 0? still return 0?
> > 
> > And that is exactly what happens on Gray Hawk Single with R-Car
> > V4M: dw_pcie_find_rasdes_capability() returns NULL, causing
> > dwc_pcie_rasdes_debugfs_init() to return -ENODEV.
> 
> Thank you for testing and for catching this issue.  Much appreciated.
> 
> > > Given that callers of dwc_pcie_debugfs_init() check for errors,
> > 
> > Debugfs issues should never be propagated upstream!
> 
> Makes complete sense.  Sorry for breaking things here!
> 
> > > this probably should correctly bubble up any failure coming from
> > > dwc_pcie_rasdes_debugfs_init().
> > >
> > > I made updates to the code directly on the current branch, have a look:
> > 
> > So while applying, you changed this like:
> > 
> >             ret = dwc_pcie_rasdes_debugfs_init(pci, dir);
> >     -       if (ret)
> >     -               dev_dbg(dev, "RASDES debugfs init failed\n");
> >     +       if (ret) {
> >     +               dev_err(dev, "failed to initialize RAS DES debugfs\n");
> >     +               return ret;
> >     +       }
> > 
> >             return 0;
> > 
> > Hence this is now a fatal error, causing the probe to fail.
> 
> I removed the changed, and also move the log level to be a warning, per:
> 
>   https://web.git.kernel.org/pub/scm/linux/kernel/git/pci/pci.git/commit/?h=controller/dwc&id=c6759a967e69aba16aef0d92f43e527b112e98a5
> 
> Would this be acceptable here?
> 
> Mani, would this be acceptable to you, too?  Given that you posted the
> following recently:
> 
>   https://lore.kernel.org/linux-pci/20250303200055.GA1881771@rocinante/T/#mab9cbd5834390d259afea056eee9a73d8c3b435f
> 
> That said, perhaps moving the log level to a debug would be better served here.
> 

Even though debugfs_init() failure is not supposed to fail the probe(),
dwc_pcie_rasdes_debugfs_init() has a devm_kzalloc() and propagating that
failure would be canolically correct IMO.

So I would still opt to have my version + your previous one.

- Mani

-- 
மணிவண்ணன் சதாசிவம்

Re: [PATCH v7 3/5] Add debugfs based silicon debug support in DWC

[Krzysztof Wilczyński]

Hello,

[...]
> > > > > > +   ret = dwc_pcie_rasdes_debugfs_init(pci, dir);
> > > > > > +   if (ret)
> > > > > > +           dev_dbg(dev, "RASDES debugfs init failed\n");
> > > > >
> > > > > What will happen if ret != 0? still return 0?
> > > 
> > > And that is exactly what happens on Gray Hawk Single with R-Car
> > > V4M: dw_pcie_find_rasdes_capability() returns NULL, causing
> > > dwc_pcie_rasdes_debugfs_init() to return -ENODEV.
> > 
> > Thank you for testing and for catching this issue.  Much appreciated.
> > 
> > > > Given that callers of dwc_pcie_debugfs_init() check for errors,
> > > 
> > > Debugfs issues should never be propagated upstream!
> > 
> > Makes complete sense.  Sorry for breaking things here!
> > 
> > > > this probably should correctly bubble up any failure coming from
> > > > dwc_pcie_rasdes_debugfs_init().
> > > >
> > > > I made updates to the code directly on the current branch, have a look:
> > > 
> > > So while applying, you changed this like:
> > > 
> > >             ret = dwc_pcie_rasdes_debugfs_init(pci, dir);
> > >     -       if (ret)
> > >     -               dev_dbg(dev, "RASDES debugfs init failed\n");
> > >     +       if (ret) {
> > >     +               dev_err(dev, "failed to initialize RAS DES debugfs\n");
> > >     +               return ret;
> > >     +       }
> > > 
> > >             return 0;
> > > 
> > > Hence this is now a fatal error, causing the probe to fail.
> > 
> > I removed the changed, and also move the log level to be a warning, per:
> > 
> >   https://web.git.kernel.org/pub/scm/linux/kernel/git/pci/pci.git/commit/?h=controller/dwc&id=c6759a967e69aba16aef0d92f43e527b112e98a5
> > 
> > Would this be acceptable here?
> > 
> > Mani, would this be acceptable to you, too?  Given that you posted the
> > following recently:
> > 
> >   https://lore.kernel.org/linux-pci/20250303200055.GA1881771@rocinante/T/#mab9cbd5834390d259afea056eee9a73d8c3b435f
> > 
> > That said, perhaps moving the log level to a debug would be better served here.
> > 
> 
> Even though debugfs_init() failure is not supposed to fail the probe(),
> dwc_pcie_rasdes_debugfs_init() has a devm_kzalloc() and propagating that
> failure would be canolically correct IMO.
> 
> So I would still opt to have my version + your previous one.

Sounds good!

I combined both changes (squashed your fix for the RAS DES capability
detection) together directly on the branch.

Thank you!

	Krzysztof

Re: [PATCH v7 5/5] Add debugfs based statistical counter support in DWC

[Fan Ni]

On Tue, Mar 04, 2025 at 10:40:43PM +0530, Shradha Todi wrote:
> 
> 
> > -----Original Message-----
> > From: Fan Ni <nifan.cxl@gmail.com>
> > Sent: 04 March 2025 02:33
> > To: Krzysztof Wilczyński <kw@linux.com>
> > Cc: Fan Ni <nifan.cxl@gmail.com>; Shradha Todi <shradha.t@samsung.com>; linux-kernel@vger.kernel.org; linux-
> > pci@vger.kernel.org; linux-arm-kernel@lists.infradead.org; linux-perf-users@vger.kernel.org; manivannan.sadhasivam@linaro.org;
> > lpieralisi@kernel.org; robh@kernel.org; bhelgaas@google.com; jingoohan1@gmail.com; Jonathan.Cameron@huawei.com;
> > a.manzanares@samsung.com; pankaj.dubey@samsung.com; cassel@kernel.org; 18255117159@163.com;
> > xueshuai@linux.alibaba.com; renyu.zj@linux.alibaba.com; will@kernel.org; mark.rutland@arm.com
> > Subject: Re: [PATCH v7 5/5] Add debugfs based statistical counter support in DWC
> > 
> > On Tue, Mar 04, 2025 at 04:42:28AM +0900, Krzysztof Wilczyński wrote:
> > > Hello,
> > >
> > > [...]
> > > > > +static ssize_t counter_value_read(struct file *file, char __user
> > > > > +*buf, size_t count, loff_t *ppos) {
> > > > > +	struct dwc_pcie_rasdes_priv *pdata = file->private_data;
> > > > > +	struct dw_pcie *pci = pdata->pci;
> > > > > +	struct dwc_pcie_rasdes_info *rinfo = pci->debugfs->rasdes_info;
> > > > > +	char debugfs_buf[DWC_DEBUGFS_BUF_MAX];
> > > > > +	ssize_t pos;
> > > > > +	u32 val;
> > > > > +
> > > > > +	mutex_lock(&rinfo->reg_event_lock);
> > > > > +	set_event_number(pdata, pci, rinfo);
> > > > > +	val = dw_pcie_readl_dbi(pci, rinfo->ras_cap_offset + RAS_DES_EVENT_COUNTER_DATA_REG);
> > > > > +	mutex_unlock(&rinfo->reg_event_lock);
> > > > > +	pos = scnprintf(debugfs_buf, DWC_DEBUGFS_BUF_MAX, "Counter
> > > > > +value: %d\n", val);
> > > > > +
> > > > > +	return simple_read_from_buffer(buf, count, ppos, debugfs_buf,
> > > > > +pos); }
> > > >
> > > > Do we need to check whether the counter is enabled or not for the
> > > > event before retrieving the counter value?
> > >
> > > I believe, we have a patch that aims to address, have a look at:
> > >
> > >
> > > https://lore.kernel.org/linux-pci/20250225171239.19574-1-manivannan.sa
> > > dhasivam@linaro.org
> > 
> > Maybe I missed something, that seems to fix counter_enable_read(), but here is to retrieve counter value.
> > How dw_pcie_readl_dbi() can return something like "Counter Disabled"?
> > 
> > Fan
> 
> Hey Fan,
> So the counter value will show 0 in case it is disabled so there will not be any issues as per say. We could add the
> check here but I feel I have already exposed the functionality to check if a counter is enabled or disabled, (by reading the
> counter_enable debugfs entry) so this could be handled in user space to only read the counter if it's enabled.
Ok. 
Returning 0 when the counter is disabled makes sense to me.

Just some thought.

It seems natural to me if we make "counter_value" only visiable to
users when the counter is enabled. 

Fan
> 
> > >
> > > Thank you!
> > >
> > > 	Krzysztof
> 
> 

Re: [PATCH v7 4/5] Add debugfs based error injection support in DWC

['Krzysztof Wilczyński']

Hello,

[...]
> > > > > +		29) Generates duplicate TLPs - duplicate_dllp
> > > > > +		30) Generates Nullified TLPs - nullified_tlp
> > > >
> > > > Would the above field called "duplicate_dllp" for duplicate TLPs be
> > > > a potential typo?  Perhaps this should be called "duplicate_tlp"?
> > > >
> > >
> > > Looks like a typo. As per Synopsys documentation, there is only 'duplicate TLP'
> > > field.
> > >
> > > Good catch!
> > 
> > Updated.  Thank you!
> > 
> 
> Sorry, this was a typo. Krzysztof, we need another change for this typo.

Not a problem.  I am glad we caught this early.

[...]
> --- a/drivers/pci/controller/dwc/pcie-designware-debugfs.c
> +++ b/drivers/pci/controller/dwc/pcie-designware-debugfs.c
> @@ -113,7 +113,7 @@ static const struct dwc_pcie_err_inj err_inj_list[] = {
>         {"posted_tlp_data", 0x4, 0x4},
>         {"non_post_tlp_data", 0x4, 0x5},
>         {"cmpl_tlp_data", 0x4, 0x6},
> -       {"duplicate_dllp", 0x5, 0x0},
> +       {"duplicate_tlp", 0x5, 0x0},
>         {"nullified_tlp", 0x5, 0x1},
>  };

Oh here too.  No worries.  I missed this one. :)

> So sorry for the inconvenience! Should I post a patch for this?

No, no need.  I will update this directly on the branch.

Thank you!

	Krzysztof

Re: [PATCH v7 3/5] Add debugfs based silicon debug support in DWC

['Krzysztof Wilczyński']

Hello,

> I think we shouldn't move the log level to be a WARN. I believe many
> controllers might not support RAS DES feature in their design and giving
> a warn dump would draw unnecessary attention.

There will be no backtrack printed with neither dev_err() nor dev_warn(),
which is what we were using here.  Using dev_WARN() or the WARN() macro
directly would be an overkill in this case, indeed.

> My opinion is to silently let it fail unless the user is actually
> interested in getting the RAS DES feature up.

I think, what we have there now is fine.  We don't error on the lack of RAS
DES capability when the platform does not support it, and only return an
error following a memory allocation failure, which should ideally never
happen.

That said, have a look at the following:

  https://web.git.kernel.org/pub/scm/linux/kernel/git/pci/pci.git/log/?h=controller/dwc

This is how the code looks like at the moment.

We can still move it to dev_dbg(), so basically suppress any errors or
warnings from being printed outside of the debug log level, if you think
it would be better.

Thank you!

	Krzysztof

RE: [PATCH v7 3/5] Add debugfs based silicon debug support in DWC

[Shradha Todi]

> -----Original Message-----
> From: 'Krzysztof Wilczyński' <kw@linux.com>
> Sent: 05 March 2025 13:15
> To: Shradha Todi <shradha.t@samsung.com>
> Cc: 'Geert Uytterhoeven' <geert@linux-m68k.org>; 'Fan Ni' <nifan.cxl@gmail.com>; linux-kernel@vger.kernel.org; linux-
> pci@vger.kernel.org; linux-arm-kernel@lists.infradead.org; linux-perf-users@vger.kernel.org; manivannan.sadhasivam@linaro.org;
> lpieralisi@kernel.org; robh@kernel.org; bhelgaas@google.com; jingoohan1@gmail.com; Jonathan.Cameron@huawei.com;
> a.manzanares@samsung.com; pankaj.dubey@samsung.com; cassel@kernel.org; 18255117159@163.com;
> xueshuai@linux.alibaba.com; renyu.zj@linux.alibaba.com; will@kernel.org; mark.rutland@arm.com; 'Yoshihiro Shimoda'
> <yoshihiro.shimoda.uh@renesas.com>; 'Linux-Renesas' <linux-renesas-soc@vger.kernel.org>
> Subject: Re: [PATCH v7 3/5] Add debugfs based silicon debug support in DWC
> 
> Hello,
> 
> > I think we shouldn't move the log level to be a WARN. I believe many
> > controllers might not support RAS DES feature in their design and
> > giving a warn dump would draw unnecessary attention.
> 
> There will be no backtrack printed with neither dev_err() nor dev_warn(), which is what we were using here.  Using dev_WARN() or
> the WARN() macro directly would be an overkill in this case, indeed.
> 

Oh sorry, I thought we were talking about WARN() here.

> > My opinion is to silently let it fail unless the user is actually
> > interested in getting the RAS DES feature up.
> 
> I think, what we have there now is fine.  We don't error on the lack of RAS DES capability when the platform does not support it,
and
> only return an error following a memory allocation failure, which should ideally never happen.
> 
> That said, have a look at the following:
> 
>   https://web.git.kernel.org/pub/scm/linux/kernel/git/pci/pci.git/log/?h=controller/dwc
> 
> This is how the code looks like at the moment.
> 
> We can still move it to dev_dbg(), so basically suppress any errors or warnings from being printed outside of the debug log level,
if you
> think it would be better.
> 

No, the current version looks perfect. Thanks!

> Thank you!
> 
> 	Krzysztof

Re: [PATCH v7 3/5] Add debugfs based silicon debug support in DWC

[Bjorn Helgaas]

On Tue, Mar 04, 2025 at 10:41:54PM +0530, Manivannan Sadhasivam wrote:
> On Wed, Mar 05, 2025 at 12:46:38AM +0900, Krzysztof Wilczyński wrote:
> > > On Mon, 3 Mar 2025 at 20:47, Krzysztof Wilczyński <kw@linux.com> wrote:
> > > > [...]
> > > > > > +int dwc_pcie_debugfs_init(struct dw_pcie *pci)
> > > > > > +{
> > > > > > +   char dirname[DWC_DEBUGFS_BUF_MAX];
> > > > > > +   struct device *dev = pci->dev;
> > > > > > +   struct debugfs_info *debugfs;
> > > > > > +   struct dentry *dir;
> > > > > > +   int ret;
> > > > > > +
> > > > > > +   /* Create main directory for each platform driver */
> > > > > > +   snprintf(dirname, DWC_DEBUGFS_BUF_MAX, "dwc_pcie_%s", dev_name(dev));
> > > > > > +   dir = debugfs_create_dir(dirname, NULL);
> > > > > > +   debugfs = devm_kzalloc(dev, sizeof(*debugfs), GFP_KERNEL);
> > > > > > +   if (!debugfs)
> > > > > > +           return -ENOMEM;
> > > > > > +
> > > > > > +   debugfs->debug_dir = dir;
> > > > > > +   pci->debugfs = debugfs;
> > > > > > +   ret = dwc_pcie_rasdes_debugfs_init(pci, dir);
> > > > > > +   if (ret)
> > > > > > +           dev_dbg(dev, "RASDES debugfs init failed\n");
> > > > >
> > > > > What will happen if ret != 0? still return 0?
> > > 
> > > And that is exactly what happens on Gray Hawk Single with R-Car
> > > V4M: dw_pcie_find_rasdes_capability() returns NULL, causing
> > > dwc_pcie_rasdes_debugfs_init() to return -ENODEV.
> > > 
> > > Debugfs issues should never be propagated upstream!
> ...

> > > So while applying, you changed this like:
> > > 
> > >             ret = dwc_pcie_rasdes_debugfs_init(pci, dir);
> > >     -       if (ret)
> > >     -               dev_dbg(dev, "RASDES debugfs init failed\n");
> > >     +       if (ret) {
> > >     +               dev_err(dev, "failed to initialize RAS DES debugfs\n");
> > >     +               return ret;
> > >     +       }
> > > 
> > >             return 0;
> > > 
> > > Hence this is now a fatal error, causing the probe to fail.

> Even though debugfs_init() failure is not supposed to fail the probe(),
> dwc_pcie_rasdes_debugfs_init() has a devm_kzalloc() and propagating that
> failure would be canolically correct IMO.

I'm not sure about this.  What's the requirement to propagate
devm_kzalloc() failures?  I think devres will free any allocs that
were successful regardless.

IIUC, we resolved the Gray Hawk Single issue by changing
dwc_pcie_rasdes_debugfs_init() to return success without doing
anything when there's no RAS DES Capability.

But dwc_pcie_debugfs_init() can still return failure, and that still
causes dw_pcie_ep_init_registers() to fail, which breaks the "don't
propagate debugfs issues upstream" rule:

  int dw_pcie_ep_init_registers(struct dw_pcie_ep *ep)
  {
          ...
          ret = dwc_pcie_debugfs_init(pci);
          if (ret)
                  goto err_remove_edma;

          return 0;

  err_remove_edma:
          dw_pcie_edma_remove(pci);

          return ret;
  }

We can say that kzalloc() failure should "never" happen, and therefore
it's OK to fail the driver probe if it happens, but that doesn't seem
like a strong argument for breaking the "don't propagate debugfs
issues" rule.  And someday there may be other kinds of failures from
dwc_pcie_debugfs_init().

Bjorn

Re: [PATCH v7 3/5] Add debugfs based silicon debug support in DWC

[Manivannan Sadhasivam]

On Wed, Mar 05, 2025 at 11:38:26AM -0600, Bjorn Helgaas wrote:
> On Tue, Mar 04, 2025 at 10:41:54PM +0530, Manivannan Sadhasivam wrote:
> > On Wed, Mar 05, 2025 at 12:46:38AM +0900, Krzysztof Wilczyński wrote:
> > > > On Mon, 3 Mar 2025 at 20:47, Krzysztof Wilczyński <kw@linux.com> wrote:
> > > > > [...]
> > > > > > > +int dwc_pcie_debugfs_init(struct dw_pcie *pci)
> > > > > > > +{
> > > > > > > +   char dirname[DWC_DEBUGFS_BUF_MAX];
> > > > > > > +   struct device *dev = pci->dev;
> > > > > > > +   struct debugfs_info *debugfs;
> > > > > > > +   struct dentry *dir;
> > > > > > > +   int ret;
> > > > > > > +
> > > > > > > +   /* Create main directory for each platform driver */
> > > > > > > +   snprintf(dirname, DWC_DEBUGFS_BUF_MAX, "dwc_pcie_%s", dev_name(dev));
> > > > > > > +   dir = debugfs_create_dir(dirname, NULL);
> > > > > > > +   debugfs = devm_kzalloc(dev, sizeof(*debugfs), GFP_KERNEL);
> > > > > > > +   if (!debugfs)
> > > > > > > +           return -ENOMEM;
> > > > > > > +
> > > > > > > +   debugfs->debug_dir = dir;
> > > > > > > +   pci->debugfs = debugfs;
> > > > > > > +   ret = dwc_pcie_rasdes_debugfs_init(pci, dir);
> > > > > > > +   if (ret)
> > > > > > > +           dev_dbg(dev, "RASDES debugfs init failed\n");
> > > > > >
> > > > > > What will happen if ret != 0? still return 0?
> > > > 
> > > > And that is exactly what happens on Gray Hawk Single with R-Car
> > > > V4M: dw_pcie_find_rasdes_capability() returns NULL, causing
> > > > dwc_pcie_rasdes_debugfs_init() to return -ENODEV.
> > > > 
> > > > Debugfs issues should never be propagated upstream!
> > ...
> 
> > > > So while applying, you changed this like:
> > > > 
> > > >             ret = dwc_pcie_rasdes_debugfs_init(pci, dir);
> > > >     -       if (ret)
> > > >     -               dev_dbg(dev, "RASDES debugfs init failed\n");
> > > >     +       if (ret) {
> > > >     +               dev_err(dev, "failed to initialize RAS DES debugfs\n");
> > > >     +               return ret;
> > > >     +       }
> > > > 
> > > >             return 0;
> > > > 
> > > > Hence this is now a fatal error, causing the probe to fail.
> 
> > Even though debugfs_init() failure is not supposed to fail the probe(),
> > dwc_pcie_rasdes_debugfs_init() has a devm_kzalloc() and propagating that
> > failure would be canolically correct IMO.
> 
> I'm not sure about this.  What's the requirement to propagate
> devm_kzalloc() failures?  I think devres will free any allocs that
> were successful regardless.
> 
> IIUC, we resolved the Gray Hawk Single issue by changing
> dwc_pcie_rasdes_debugfs_init() to return success without doing
> anything when there's no RAS DES Capability.
> 
> But dwc_pcie_debugfs_init() can still return failure, and that still
> causes dw_pcie_ep_init_registers() to fail, which breaks the "don't
> propagate debugfs issues upstream" rule:
> 
>   int dw_pcie_ep_init_registers(struct dw_pcie_ep *ep)
>   {
>           ...
>           ret = dwc_pcie_debugfs_init(pci);
>           if (ret)
>                   goto err_remove_edma;
> 
>           return 0;
> 
>   err_remove_edma:
>           dw_pcie_edma_remove(pci);
> 
>           return ret;
>   }
> 
> We can say that kzalloc() failure should "never" happen, and therefore
> it's OK to fail the driver probe if it happens, but that doesn't seem
> like a strong argument for breaking the "don't propagate debugfs
> issues" rule.  And someday there may be other kinds of failures from
> dwc_pcie_debugfs_init().
> 

Fine with me. I was not too sure about propagating failure either.

- Mani

-- 
மணிவண்ணன் சதாசிவம்

Re: [PATCH v7 3/5] Add debugfs based silicon debug support in DWC

[Krzysztof Wilczyński]

Hello,

[...]
> > > Even though debugfs_init() failure is not supposed to fail the probe(),
> > > dwc_pcie_rasdes_debugfs_init() has a devm_kzalloc() and propagating that
> > > failure would be canolically correct IMO.
> > 
> > I'm not sure about this.  What's the requirement to propagate
> > devm_kzalloc() failures?  I think devres will free any allocs that
> > were successful regardless.
> > 
> > IIUC, we resolved the Gray Hawk Single issue by changing
> > dwc_pcie_rasdes_debugfs_init() to return success without doing
> > anything when there's no RAS DES Capability.
> > 
> > But dwc_pcie_debugfs_init() can still return failure, and that still
> > causes dw_pcie_ep_init_registers() to fail, which breaks the "don't
> > propagate debugfs issues upstream" rule:
> > 
> >   int dw_pcie_ep_init_registers(struct dw_pcie_ep *ep)
> >   {
> >           ...
> >           ret = dwc_pcie_debugfs_init(pci);
> >           if (ret)
> >                   goto err_remove_edma;
> > 
> >           return 0;
> > 
> >   err_remove_edma:
> >           dw_pcie_edma_remove(pci);
> > 
> >           return ret;
> >   }
> > 
> > We can say that kzalloc() failure should "never" happen, and therefore
> > it's OK to fail the driver probe if it happens, but that doesn't seem
> > like a strong argument for breaking the "don't propagate debugfs
> > issues" rule.  And someday there may be other kinds of failures from
> > dwc_pcie_debugfs_init().
> > 
> 
> Fine with me. I was not too sure about propagating failure either.

What if we do this?

diff --git i/drivers/pci/controller/dwc/pcie-designware-debugfs.c w/drivers/pci/controller/dwc/pcie-designware-debugfs.c
index 586a9d107434..fddf71f014c4 100644
--- i/drivers/pci/controller/dwc/pcie-designware-debugfs.c
+++ w/drivers/pci/controller/dwc/pcie-designware-debugfs.c
@@ -162,7 +162,7 @@ void dwc_pcie_debugfs_deinit(struct dw_pcie *pci)
 	debugfs_remove_recursive(pci->debugfs->debug_dir);
 }

-int dwc_pcie_debugfs_init(struct dw_pcie *pci)
+void dwc_pcie_debugfs_init(struct dw_pcie *pci)
 {
 	char dirname[DWC_DEBUGFS_BUF_MAX];
 	struct device *dev = pci->dev;
@@ -174,17 +174,15 @@ int dwc_pcie_debugfs_init(struct dw_pcie *pci)
 	snprintf(dirname, DWC_DEBUGFS_BUF_MAX, "dwc_pcie_%s", dev_name(dev));
 	dir = debugfs_create_dir(dirname, NULL);
 	debugfs = devm_kzalloc(dev, sizeof(*debugfs), GFP_KERNEL);
-	if (!debugfs)
-		return -ENOMEM;
+	if (!debugfs) {
+		dev_err(dev, "failed to allocate memory for debugfs\n");
+		return;
+	}

 	debugfs->debug_dir = dir;
 	pci->debugfs = debugfs;
 	err = dwc_pcie_rasdes_debugfs_init(pci, dir);
-	if (err) {
-		dev_err(dev, "failed to initialize RAS DES debugfs, err=%d\n",
-			err);
-		return err;
-	}
-
-	return 0;
+	if (err)
+		dev_warn(dev, "failed to initialize RAS DES debugfs, err=%d",
+			 err);
 }
diff --git i/drivers/pci/controller/dwc/pcie-designware-ep.c w/drivers/pci/controller/dwc/pcie-designware-ep.c
index c6e76a07c2c9..11ff292ca87d 100644
--- i/drivers/pci/controller/dwc/pcie-designware-ep.c
+++ w/drivers/pci/controller/dwc/pcie-designware-ep.c
@@ -838,9 +838,7 @@ int dw_pcie_ep_init_registers(struct dw_pcie_ep *ep)

 	dw_pcie_ep_init_non_sticky_registers(pci);

-	ret = dwc_pcie_debugfs_init(pci);
-	if (ret)
-		goto err_remove_edma;
+	dwc_pcie_debugfs_init(pci);

 	return 0;

diff --git i/drivers/pci/controller/dwc/pcie-designware-host.c w/drivers/pci/controller/dwc/pcie-designware-host.c
index 2081e8c72d12..6501fb062c70 100644
--- i/drivers/pci/controller/dwc/pcie-designware-host.c
+++ w/drivers/pci/controller/dwc/pcie-designware-host.c
@@ -548,9 +548,7 @@ int dw_pcie_host_init(struct dw_pcie_rp *pp)
 	if (pp->ops->post_init)
 		pp->ops->post_init(pp);

-	ret = dwc_pcie_debugfs_init(pci);
-	if (ret)
-		goto err_stop_link;
+	dwc_pcie_debugfs_init(pci);

 	return 0;

diff --git i/drivers/pci/controller/dwc/pcie-designware.h w/drivers/pci/controller/dwc/pcie-designware.h
index 7f9807d4e5de..dd129718fb41 100644
--- i/drivers/pci/controller/dwc/pcie-designware.h
+++ w/drivers/pci/controller/dwc/pcie-designware.h
@@ -815,12 +815,11 @@ dw_pcie_ep_get_func_from_ep(struct dw_pcie_ep *ep, u8 func_no)
 #endif

 #ifdef CONFIG_PCIE_DW_DEBUGFS
-int dwc_pcie_debugfs_init(struct dw_pcie *pci);
+void dwc_pcie_debugfs_init(struct dw_pcie *pci);
 void dwc_pcie_debugfs_deinit(struct dw_pcie *pci);
 #else
-static inline int dwc_pcie_debugfs_init(struct dw_pcie *pci)
+static inline void dwc_pcie_debugfs_init(struct dw_pcie *pci)
 {
-	return 0;
 }
 static inline void dwc_pcie_debugfs_deinit(struct dw_pcie *pci)
 {

I think this would be fine, especially given the rules around debugfs and
a quick look around Git history to see what the prefernce would be typically.

Thank you!

	Krzysztof

Re: [PATCH v7 3/5] Add debugfs based silicon debug support in DWC

[Krzysztof Wilczyński]

> > > > Even though debugfs_init() failure is not supposed to fail the probe(),
> > > > dwc_pcie_rasdes_debugfs_init() has a devm_kzalloc() and propagating that
> > > > failure would be canolically correct IMO.
> > > 
> > > I'm not sure about this.  What's the requirement to propagate
> > > devm_kzalloc() failures?  I think devres will free any allocs that
> > > were successful regardless.
> > > 
> > > IIUC, we resolved the Gray Hawk Single issue by changing
> > > dwc_pcie_rasdes_debugfs_init() to return success without doing
> > > anything when there's no RAS DES Capability.
> > > 
> > > But dwc_pcie_debugfs_init() can still return failure, and that still
> > > causes dw_pcie_ep_init_registers() to fail, which breaks the "don't
> > > propagate debugfs issues upstream" rule:
> > > 
> > >   int dw_pcie_ep_init_registers(struct dw_pcie_ep *ep)
> > >   {
> > >           ...
> > >           ret = dwc_pcie_debugfs_init(pci);
> > >           if (ret)
> > >                   goto err_remove_edma;
> > > 
> > >           return 0;
> > > 
> > >   err_remove_edma:
> > >           dw_pcie_edma_remove(pci);
> > > 
> > >           return ret;
> > >   }
> > > 
> > > We can say that kzalloc() failure should "never" happen, and therefore
> > > it's OK to fail the driver probe if it happens, but that doesn't seem
> > > like a strong argument for breaking the "don't propagate debugfs
> > > issues" rule.  And someday there may be other kinds of failures from
> > > dwc_pcie_debugfs_init().
> > > 
> > 
> > Fine with me. I was not too sure about propagating failure either.
> 
> What if we do this?
> 
> diff --git i/drivers/pci/controller/dwc/pcie-designware-debugfs.c w/drivers/pci/controller/dwc/pcie-designware-debugfs.c
> index 586a9d107434..fddf71f014c4 100644
> --- i/drivers/pci/controller/dwc/pcie-designware-debugfs.c
> +++ w/drivers/pci/controller/dwc/pcie-designware-debugfs.c
> @@ -162,7 +162,7 @@ void dwc_pcie_debugfs_deinit(struct dw_pcie *pci)
>  	debugfs_remove_recursive(pci->debugfs->debug_dir);
>  }
> 
> -int dwc_pcie_debugfs_init(struct dw_pcie *pci)
> +void dwc_pcie_debugfs_init(struct dw_pcie *pci)
>  {
>  	char dirname[DWC_DEBUGFS_BUF_MAX];
>  	struct device *dev = pci->dev;
> @@ -174,17 +174,15 @@ int dwc_pcie_debugfs_init(struct dw_pcie *pci)
>  	snprintf(dirname, DWC_DEBUGFS_BUF_MAX, "dwc_pcie_%s", dev_name(dev));
>  	dir = debugfs_create_dir(dirname, NULL);
>  	debugfs = devm_kzalloc(dev, sizeof(*debugfs), GFP_KERNEL);
> -	if (!debugfs)
> -		return -ENOMEM;
> +	if (!debugfs) {
> +		dev_err(dev, "failed to allocate memory for debugfs\n");
> +		return;
> +	}
> 
>  	debugfs->debug_dir = dir;
>  	pci->debugfs = debugfs;
>  	err = dwc_pcie_rasdes_debugfs_init(pci, dir);
> -	if (err) {
> -		dev_err(dev, "failed to initialize RAS DES debugfs, err=%d\n",
> -			err);
> -		return err;
> -	}
> -
> -	return 0;
> +	if (err)
> +		dev_warn(dev, "failed to initialize RAS DES debugfs, err=%d",
> +			 err);
>  }
> diff --git i/drivers/pci/controller/dwc/pcie-designware-ep.c w/drivers/pci/controller/dwc/pcie-designware-ep.c
> index c6e76a07c2c9..11ff292ca87d 100644
> --- i/drivers/pci/controller/dwc/pcie-designware-ep.c
> +++ w/drivers/pci/controller/dwc/pcie-designware-ep.c
> @@ -838,9 +838,7 @@ int dw_pcie_ep_init_registers(struct dw_pcie_ep *ep)
> 
>  	dw_pcie_ep_init_non_sticky_registers(pci);
> 
> -	ret = dwc_pcie_debugfs_init(pci);
> -	if (ret)
> -		goto err_remove_edma;
> +	dwc_pcie_debugfs_init(pci);
> 
>  	return 0;
> 
> diff --git i/drivers/pci/controller/dwc/pcie-designware-host.c w/drivers/pci/controller/dwc/pcie-designware-host.c
> index 2081e8c72d12..6501fb062c70 100644
> --- i/drivers/pci/controller/dwc/pcie-designware-host.c
> +++ w/drivers/pci/controller/dwc/pcie-designware-host.c
> @@ -548,9 +548,7 @@ int dw_pcie_host_init(struct dw_pcie_rp *pp)
>  	if (pp->ops->post_init)
>  		pp->ops->post_init(pp);
> 
> -	ret = dwc_pcie_debugfs_init(pci);
> -	if (ret)
> -		goto err_stop_link;
> +	dwc_pcie_debugfs_init(pci);
> 
>  	return 0;
> 
> diff --git i/drivers/pci/controller/dwc/pcie-designware.h w/drivers/pci/controller/dwc/pcie-designware.h
> index 7f9807d4e5de..dd129718fb41 100644
> --- i/drivers/pci/controller/dwc/pcie-designware.h
> +++ w/drivers/pci/controller/dwc/pcie-designware.h
> @@ -815,12 +815,11 @@ dw_pcie_ep_get_func_from_ep(struct dw_pcie_ep *ep, u8 func_no)
>  #endif
> 
>  #ifdef CONFIG_PCIE_DW_DEBUGFS
> -int dwc_pcie_debugfs_init(struct dw_pcie *pci);
> +void dwc_pcie_debugfs_init(struct dw_pcie *pci);
>  void dwc_pcie_debugfs_deinit(struct dw_pcie *pci);
>  #else
> -static inline int dwc_pcie_debugfs_init(struct dw_pcie *pci)
> +static inline void dwc_pcie_debugfs_init(struct dw_pcie *pci)
>  {
> -	return 0;
>  }
>  static inline void dwc_pcie_debugfs_deinit(struct dw_pcie *pci)
>  {
> 
> I think this would be fine, especially given the rules around debugfs and
> a quick look around Git history to see what the prefernce would be typically.

Changed dev_warn() to dev_err() per Bjorn's feedback off mailing list,
and squashed against the current code on the branch.

Thank you!

	Krzysztof

Re: [PATCH v7 3/5] Add debugfs based silicon debug support in DWC

[Geert Uytterhoeven]

Hi Krzysztof,

On Wed, 5 Mar 2025 at 20:10, Krzysztof Wilczyński <kw@linux.com> wrote:
> [...]
> > > > Even though debugfs_init() failure is not supposed to fail the probe(),
> > > > dwc_pcie_rasdes_debugfs_init() has a devm_kzalloc() and propagating that
> > > > failure would be canolically correct IMO.
> > >
> > > I'm not sure about this.  What's the requirement to propagate
> > > devm_kzalloc() failures?  I think devres will free any allocs that
> > > were successful regardless.
> > >
> > > IIUC, we resolved the Gray Hawk Single issue by changing
> > > dwc_pcie_rasdes_debugfs_init() to return success without doing
> > > anything when there's no RAS DES Capability.
> > >
> > > But dwc_pcie_debugfs_init() can still return failure, and that still
> > > causes dw_pcie_ep_init_registers() to fail, which breaks the "don't
> > > propagate debugfs issues upstream" rule:
> > >
> > >   int dw_pcie_ep_init_registers(struct dw_pcie_ep *ep)
> > >   {
> > >           ...
> > >           ret = dwc_pcie_debugfs_init(pci);
> > >           if (ret)
> > >                   goto err_remove_edma;
> > >
> > >           return 0;
> > >
> > >   err_remove_edma:
> > >           dw_pcie_edma_remove(pci);
> > >
> > >           return ret;
> > >   }
> > >
> > > We can say that kzalloc() failure should "never" happen, and therefore
> > > it's OK to fail the driver probe if it happens, but that doesn't seem
> > > like a strong argument for breaking the "don't propagate debugfs
> > > issues" rule.  And someday there may be other kinds of failures from
> > > dwc_pcie_debugfs_init().

pcie-designware-debugfs.c only does small allocations.  If any of
these fail, you have much bigger problems, and the system will die soon,
irrespective of propagating the -ENOMEM or not...

Another issue is that the caller does not handle failures correctly,
given (A) the irqdomain WARNING I got, and (B) the half-registered
PCI bus, oopsing on "lspci"...

> > Fine with me. I was not too sure about propagating failure either.
>
> What if we do this?
>
> diff --git i/drivers/pci/controller/dwc/pcie-designware-debugfs.c w/drivers/pci/controller/dwc/pcie-designware-debugfs.c
> index 586a9d107434..fddf71f014c4 100644
> --- i/drivers/pci/controller/dwc/pcie-designware-debugfs.c
> +++ w/drivers/pci/controller/dwc/pcie-designware-debugfs.c
> @@ -162,7 +162,7 @@ void dwc_pcie_debugfs_deinit(struct dw_pcie *pci)
>         debugfs_remove_recursive(pci->debugfs->debug_dir);
>  }
>
> -int dwc_pcie_debugfs_init(struct dw_pcie *pci)
> +void dwc_pcie_debugfs_init(struct dw_pcie *pci)
>  {
>         char dirname[DWC_DEBUGFS_BUF_MAX];
>         struct device *dev = pci->dev;
> @@ -174,17 +174,15 @@ int dwc_pcie_debugfs_init(struct dw_pcie *pci)
>         snprintf(dirname, DWC_DEBUGFS_BUF_MAX, "dwc_pcie_%s", dev_name(dev));
>         dir = debugfs_create_dir(dirname, NULL);
>         debugfs = devm_kzalloc(dev, sizeof(*debugfs), GFP_KERNEL);
> -       if (!debugfs)
> -               return -ENOMEM;
> +       if (!debugfs) {
> +               dev_err(dev, "failed to allocate memory for debugfs\n");

There is no need to print an error message when a memory allocation
fails, as the memory allocation core already takes care of that.
So please drop the dev_err() call.

> +               return;
> +       }
>

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

Re: [PATCH v7 3/5] Add debugfs based silicon debug support in DWC

[Krzysztof Wilczyński]

Hello,

[...]
> Another issue is that the caller does not handle failures correctly,
> given (A) the irqdomain WARNING I got, and (B) the half-registered
> PCI bus, oopsing on "lspci"...

This is something we will look into.  A more robust DesignWare core is
something we would definitely want to have.

Sorry about the issues with this...

[...]
> > -int dwc_pcie_debugfs_init(struct dw_pcie *pci)
> > +void dwc_pcie_debugfs_init(struct dw_pcie *pci)
> >  {
> >         char dirname[DWC_DEBUGFS_BUF_MAX];
> >         struct device *dev = pci->dev;
> > @@ -174,17 +174,15 @@ int dwc_pcie_debugfs_init(struct dw_pcie *pci)
> >         snprintf(dirname, DWC_DEBUGFS_BUF_MAX, "dwc_pcie_%s", dev_name(dev));
> >         dir = debugfs_create_dir(dirname, NULL);
> >         debugfs = devm_kzalloc(dev, sizeof(*debugfs), GFP_KERNEL);
> > -       if (!debugfs)
> > -               return -ENOMEM;
> > +       if (!debugfs) {
> > +               dev_err(dev, "failed to allocate memory for debugfs\n");
> 
> There is no need to print an error message when a memory allocation
> fails, as the memory allocation core already takes care of that.
> So please drop the dev_err() call.

Done.  Thank you!

	Krzysztof

RE: [PATCH v7 3/5] Add debugfs based silicon debug support in DWC

[Shradha Todi]

> -----Original Message-----
> From: Krzysztof Wilczyński <kw@linux.com>
> Sent: 06 March 2025 14:33
> To: Geert Uytterhoeven <geert@linux-m68k.org>
> Cc: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>; Bjorn Helgaas <helgaas@kernel.org>; Fan Ni
> <nifan.cxl@gmail.com>; Shradha Todi <shradha.t@samsung.com>; linux-kernel@vger.kernel.org; linux-pci@vger.kernel.org; linux-
> arm-kernel@lists.infradead.org; linux-perf-users@vger.kernel.org; lpieralisi@kernel.org; robh@kernel.org; bhelgaas@google.com;
> jingoohan1@gmail.com; Jonathan.Cameron@huawei.com; a.manzanares@samsung.com; pankaj.dubey@samsung.com;
> cassel@kernel.org; 18255117159@163.com; xueshuai@linux.alibaba.com; renyu.zj@linux.alibaba.com; will@kernel.org;
> mark.rutland@arm.com; Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>; Linux-Renesas <linux-renesas-
> soc@vger.kernel.org>
> Subject: Re: [PATCH v7 3/5] Add debugfs based silicon debug support in DWC
> 
> Hello,
> 
> [...]
> > Another issue is that the caller does not handle failures correctly,
> > given (A) the irqdomain WARNING I got, and (B) the half-registered PCI
> > bus, oopsing on "lspci"...
> 
> This is something we will look into.  A more robust DesignWare core is something we would definitely want to have.
> 

The issue here was that " pci_host_probe(bridge)" was being called right before doing the debugfs init.
So upon failure, the cleanup path should have included:
        pci_stop_root_bus(pp->bridge->bus);
        pci_remove_root_bus(pp->bridge->bus);

I missed that, which was probably causing the half-registered PCI bus. Since we are going with no error
propagation now, there is no issue anymore. Sorry for missing that.

> Sorry about the issues with this...
> 
> [...]
> > > -int dwc_pcie_debugfs_init(struct dw_pcie *pci)
> > > +void dwc_pcie_debugfs_init(struct dw_pcie *pci)
> > >  {
> > >         char dirname[DWC_DEBUGFS_BUF_MAX];
> > >         struct device *dev = pci->dev; @@ -174,17 +174,15 @@ int
> > > dwc_pcie_debugfs_init(struct dw_pcie *pci)
> > >         snprintf(dirname, DWC_DEBUGFS_BUF_MAX, "dwc_pcie_%s", dev_name(dev));
> > >         dir = debugfs_create_dir(dirname, NULL);
> > >         debugfs = devm_kzalloc(dev, sizeof(*debugfs), GFP_KERNEL);
> > > -       if (!debugfs)
> > > -               return -ENOMEM;
> > > +       if (!debugfs) {
> > > +               dev_err(dev, "failed to allocate memory for
> > > + debugfs\n");
> >
> > There is no need to print an error message when a memory allocation
> > fails, as the memory allocation core already takes care of that.
> > So please drop the dev_err() call.
> 
> Done.  Thank you!
> 
> 	Krzysztof

RE: [PATCH v7 5/5] Add debugfs based statistical counter support in DWC

[Shradha Todi]

> -----Original Message-----
> From: Fan Ni <nifan.cxl@gmail.com>
> Sent: 05 March 2025 09:56
> To: Shradha Todi <shradha.t@samsung.com>
> Cc: 'Fan Ni' <nifan.cxl@gmail.com>; 'Krzysztof Wilczyński' <kw@linux.com>; linux-kernel@vger.kernel.org; linux-pci@vger.kernel.org;
> linux-arm-kernel@lists.infradead.org; linux-perf-users@vger.kernel.org; manivannan.sadhasivam@linaro.org; lpieralisi@kernel.org;
> robh@kernel.org; bhelgaas@google.com; jingoohan1@gmail.com; Jonathan.Cameron@huawei.com; a.manzanares@samsung.com;
> pankaj.dubey@samsung.com; cassel@kernel.org; 18255117159@163.com; xueshuai@linux.alibaba.com; renyu.zj@linux.alibaba.com;
> will@kernel.org; mark.rutland@arm.com
> Subject: Re: [PATCH v7 5/5] Add debugfs based statistical counter support in DWC
> 
> On Tue, Mar 04, 2025 at 10:40:43PM +0530, Shradha Todi wrote:
> >
> >
> > > -----Original Message-----
> > > From: Fan Ni <nifan.cxl@gmail.com>
> > > Sent: 04 March 2025 02:33
> > > To: Krzysztof Wilczyński <kw@linux.com>
> > > Cc: Fan Ni <nifan.cxl@gmail.com>; Shradha Todi
> > > <shradha.t@samsung.com>; linux-kernel@vger.kernel.org; linux-
> > > pci@vger.kernel.org; linux-arm-kernel@lists.infradead.org;
> > > linux-perf-users@vger.kernel.org; manivannan.sadhasivam@linaro.org;
> > > lpieralisi@kernel.org; robh@kernel.org; bhelgaas@google.com;
> > > jingoohan1@gmail.com; Jonathan.Cameron@huawei.com;
> > > a.manzanares@samsung.com; pankaj.dubey@samsung.com;
> > > cassel@kernel.org; 18255117159@163.com; xueshuai@linux.alibaba.com;
> > > renyu.zj@linux.alibaba.com; will@kernel.org; mark.rutland@arm.com
> > > Subject: Re: [PATCH v7 5/5] Add debugfs based statistical counter
> > > support in DWC
> > >
> > > On Tue, Mar 04, 2025 at 04:42:28AM +0900, Krzysztof Wilczyński wrote:
> > > > Hello,
> > > >
> > > > [...]
> > > > > > +static ssize_t counter_value_read(struct file *file, char
> > > > > > +__user *buf, size_t count, loff_t *ppos) {
> > > > > > +	struct dwc_pcie_rasdes_priv *pdata = file->private_data;
> > > > > > +	struct dw_pcie *pci = pdata->pci;
> > > > > > +	struct dwc_pcie_rasdes_info *rinfo = pci->debugfs->rasdes_info;
> > > > > > +	char debugfs_buf[DWC_DEBUGFS_BUF_MAX];
> > > > > > +	ssize_t pos;
> > > > > > +	u32 val;
> > > > > > +
> > > > > > +	mutex_lock(&rinfo->reg_event_lock);
> > > > > > +	set_event_number(pdata, pci, rinfo);
> > > > > > +	val = dw_pcie_readl_dbi(pci, rinfo->ras_cap_offset + RAS_DES_EVENT_COUNTER_DATA_REG);
> > > > > > +	mutex_unlock(&rinfo->reg_event_lock);
> > > > > > +	pos = scnprintf(debugfs_buf, DWC_DEBUGFS_BUF_MAX, "Counter
> > > > > > +value: %d\n", val);
> > > > > > +
> > > > > > +	return simple_read_from_buffer(buf, count, ppos,
> > > > > > +debugfs_buf, pos); }
> > > > >
> > > > > Do we need to check whether the counter is enabled or not for
> > > > > the event before retrieving the counter value?
> > > >
> > > > I believe, we have a patch that aims to address, have a look at:
> > > >
> > > >
> > > > https://lore.kernel.org/linux-pci/20250225171239.19574-1-manivanna
> > > > n.sa
> > > > dhasivam@linaro.org
> > >
> > > Maybe I missed something, that seems to fix counter_enable_read(), but here is to retrieve counter value.
> > > How dw_pcie_readl_dbi() can return something like "Counter Disabled"?
> > >
> > > Fan
> >
> > Hey Fan,
> > So the counter value will show 0 in case it is disabled so there will
> > not be any issues as per say. We could add the check here but I feel I
> > have already exposed the functionality to check if a counter is enabled or disabled, (by reading the counter_enable debugfs entry)
> so this could be handled in user space to only read the counter if it's enabled.
> Ok.
> Returning 0 when the counter is disabled makes sense to me.
> 
> Just some thought.
> 
> It seems natural to me if we make "counter_value" only visiable to users when the counter is enabled.
> 

Hey Fan,

This looks like a good suggestion to me. I have implemented this and in the process of testing. Since there is no support in the
debugfs framework for conditionally hiding certain files from user, the custom implementation is a little tricky and will need some
discussion before taking in. So let's take this up as a top up patch and live with returning 0 when the counter is disabled for now?

Thanks a lot for the suggestion.
Shradha 

> Fan
> >
> > > >
> > > > Thank you!
> > > >
> > > > 	Krzysztof
> >
> >