diff mbox series

[3/4] cxl/pmem: Export dirty shutdown count via sysfs

Message ID 20250219021420.214991-4-dave@stgolabs.net
State Superseded
Headers show
Series cxl: Dirty shutdown followups | expand

Commit Message

Davidlohr Bueso Feb. 19, 2025, 2:14 a.m. UTC
Similar to how the acpi_nfit driver exports Optane dirty shutdown count,
introduce:

  /sys/bus/cxl/devices/nvdimm-bridge0/ndbusX/nmemY/cxl/dirty_shutdown

Under the conditions that 1) dirty shutdown can be set, 2) Device GPF
DVSEC exists, and 3) the count itself can be retrieved.

Suggested-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Davidlohr Bueso <dave@stgolabs.net>
---
 Documentation/ABI/testing/sysfs-bus-cxl       | 12 +++
 Documentation/driver-api/cxl/maturity-map.rst |  2 +-
 drivers/cxl/core/mbox.c                       | 21 +++++
 drivers/cxl/cxl.h                             |  1 +
 drivers/cxl/cxlmem.h                          | 15 ++++
 drivers/cxl/pmem.c                            | 77 +++++++++++++++++--
 6 files changed, 119 insertions(+), 9 deletions(-)

Comments

Davidlohr Bueso Feb. 19, 2025, 2:34 a.m. UTC | #1
On Tue, 18 Feb 2025, Davidlohr Bueso wrote:

>diff --git a/drivers/cxl/cxlmem.h b/drivers/cxl/cxlmem.h
>index 6d60030139df..03ad3c8ba88d 100644
>--- a/drivers/cxl/cxlmem.h
>+++ b/drivers/cxl/cxlmem.h
>@@ -866,4 +879,6 @@ struct cxl_hdm {
> struct seq_file;
> struct dentry *cxl_debugfs_create_dir(const char *dir);
> void cxl_dpa_debug(struct seq_file *file, struct cxl_dev_state *cxlds);
>+
>+int cxl_gpf_device(struct cxl_dev_state *cxlds);

Bleh this is a leftover of v1, it slipped through. I will send a v3.

Thanks,
Davidlohr
diff mbox series

Patch

diff --git a/Documentation/ABI/testing/sysfs-bus-cxl b/Documentation/ABI/testing/sysfs-bus-cxl
index 3f5627a1210a..a7491d214098 100644
--- a/Documentation/ABI/testing/sysfs-bus-cxl
+++ b/Documentation/ABI/testing/sysfs-bus-cxl
@@ -586,3 +586,15 @@  Description:
 		See Documentation/ABI/stable/sysfs-devices-node. access0 provides
 		the number to the closest initiator and access1 provides the
 		number to the closest CPU.
+
+
+What:		/sys/bus/cxl/devices/nvdimm-bridge0/ndbusX/nmemY/cxl/dirty_shutdown
+Date:		Feb, 2025
+KernelVersion:	v6.15
+Contact:	linux-cxl@vger.kernel.org
+Description:
+		(RO) The device dirty shutdown count value, which is the number
+		of times the device could have incurred in potential data loss.
+		The count is persistent across power loss and wraps back to 0
+		upon overflow. If this file is not present, the device does not
+		have the necessary support for dirty tracking.
diff --git a/Documentation/driver-api/cxl/maturity-map.rst b/Documentation/driver-api/cxl/maturity-map.rst
index 99dd2c841e69..a2288f9df658 100644
--- a/Documentation/driver-api/cxl/maturity-map.rst
+++ b/Documentation/driver-api/cxl/maturity-map.rst
@@ -130,7 +130,7 @@  Mailbox commands
 * [0] Switch CCI
 * [3] Timestamp
 * [1] PMEM labels
-* [1] PMEM GPF / Dirty Shutdown
+* [3] PMEM GPF / Dirty Shutdown
 * [0] Scan Media
 
 PMU
diff --git a/drivers/cxl/core/mbox.c b/drivers/cxl/core/mbox.c
index 86d13f4a1c18..f1009a265f9d 100644
--- a/drivers/cxl/core/mbox.c
+++ b/drivers/cxl/core/mbox.c
@@ -1281,6 +1281,27 @@  int cxl_mem_dpa_fetch(struct cxl_memdev_state *mds, struct cxl_dpa_info *info)
 }
 EXPORT_SYMBOL_NS_GPL(cxl_mem_dpa_fetch, "CXL");
 
+int cxl_get_dirty_count(struct cxl_memdev_state *mds, u32 *count)
+{
+	int rc;
+	struct cxl_mailbox *cxl_mbox = &mds->cxlds.cxl_mbox;
+	struct cxl_mbox_cmd mbox_cmd;
+	struct cxl_mbox_get_health_info_out hi;
+
+	mbox_cmd = (struct cxl_mbox_cmd) {
+		.opcode = CXL_MBOX_OP_GET_HEALTH_INFO,
+		.size_out = sizeof(hi),
+		.payload_out = &hi,
+	};
+
+	rc = cxl_internal_send_cmd(cxl_mbox, &mbox_cmd);
+	if (!rc)
+		*count = le32_to_cpu(hi.dirty_shutdown_cnt);
+
+	return rc;
+}
+EXPORT_SYMBOL_NS_GPL(cxl_get_dirty_count, "CXL");
+
 int cxl_arm_dirty_shutdown(struct cxl_memdev_state *mds)
 {
 	struct cxl_mailbox *cxl_mbox = &mds->cxlds.cxl_mbox;
diff --git a/drivers/cxl/cxl.h b/drivers/cxl/cxl.h
index acbbba41356d..4dbf1cc60047 100644
--- a/drivers/cxl/cxl.h
+++ b/drivers/cxl/cxl.h
@@ -542,6 +542,7 @@  struct cxl_nvdimm {
 	struct device dev;
 	struct cxl_memdev *cxlmd;
 	u8 dev_id[CXL_DEV_ID_LEN]; /* for nvdimm, string of 'serial' */
+	u64 dirty_shutdowns;
 };
 
 struct cxl_pmem_region_mapping {
diff --git a/drivers/cxl/cxlmem.h b/drivers/cxl/cxlmem.h
index 6d60030139df..03ad3c8ba88d 100644
--- a/drivers/cxl/cxlmem.h
+++ b/drivers/cxl/cxlmem.h
@@ -681,6 +681,18 @@  struct cxl_mbox_set_partition_info {
 
 #define  CXL_SET_PARTITION_IMMEDIATE_FLAG	BIT(0)
 
+/* Get Health Info Output Payload CXL 3.2 Spec 8.2.10.9.3.1 Table 8-148 */
+struct cxl_mbox_get_health_info_out {
+	u8 health_status;
+	u8 media_status;
+	u8 additional_status;
+	u8 life_used;
+	__le16 device_temperature;
+	__le32 dirty_shutdown_cnt;
+	__le32 corrected_volatile_error_cnt;
+	__le32 corrected_persistent_error_cnt;
+} __packed;
+
 /* Set Shutdown State Input Payload CXL 3.2 Spec 8.2.10.9.3.5 Table 8-152 */
 struct cxl_mbox_set_shutdown_state_in {
 	u8 state;
@@ -822,6 +834,7 @@  void cxl_event_trace_record(const struct cxl_memdev *cxlmd,
 			    enum cxl_event_log_type type,
 			    enum cxl_event_type event_type,
 			    const uuid_t *uuid, union cxl_event *evt);
+int cxl_get_dirty_count(struct cxl_memdev_state *mds, u32 *count);
 int cxl_arm_dirty_shutdown(struct cxl_memdev_state *mds);
 int cxl_set_timestamp(struct cxl_memdev_state *mds);
 int cxl_poison_state_init(struct cxl_memdev_state *mds);
@@ -866,4 +879,6 @@  struct cxl_hdm {
 struct seq_file;
 struct dentry *cxl_debugfs_create_dir(const char *dir);
 void cxl_dpa_debug(struct seq_file *file, struct cxl_dev_state *cxlds);
+
+int cxl_gpf_device(struct cxl_dev_state *cxlds);
 #endif /* __CXL_MEM_H__ */
diff --git a/drivers/cxl/pmem.c b/drivers/cxl/pmem.c
index 6b284962592f..aee1afe9d287 100644
--- a/drivers/cxl/pmem.c
+++ b/drivers/cxl/pmem.c
@@ -42,15 +42,44 @@  static ssize_t id_show(struct device *dev, struct device_attribute *attr, char *
 }
 static DEVICE_ATTR_RO(id);
 
+static ssize_t dirty_shutdown_show(struct device *dev,
+				   struct device_attribute *attr, char *buf)
+{
+	struct nvdimm *nvdimm = to_nvdimm(dev);
+	struct cxl_nvdimm *cxl_nvd = nvdimm_provider_data(nvdimm);
+
+	return sysfs_emit(buf, "%lld\n", cxl_nvd->dirty_shutdowns);
+}
+static DEVICE_ATTR_RO(dirty_shutdown);
+
 static struct attribute *cxl_dimm_attributes[] = {
 	&dev_attr_id.attr,
 	&dev_attr_provider.attr,
+	&dev_attr_dirty_shutdown.attr,
 	NULL
 };
 
+#define CXL_INVALID_DIRTY_SHUTDOWN_COUNT -1
+static umode_t cxl_dimm_visible(struct kobject *kobj,
+				struct attribute *a, int n)
+{
+	if (a == &dev_attr_dirty_shutdown.attr) {
+		struct device *dev = kobj_to_dev(kobj);
+		struct nvdimm *nvdimm = to_nvdimm(dev);
+		struct cxl_nvdimm *cxl_nvd = nvdimm_provider_data(nvdimm);
+
+		if (cxl_nvd->dirty_shutdowns ==
+		    CXL_INVALID_DIRTY_SHUTDOWN_COUNT)
+			return 0;
+	}
+
+	return a->mode;
+}
+
 static const struct attribute_group cxl_dimm_attribute_group = {
 	.name = "cxl",
 	.attrs = cxl_dimm_attributes,
+	.is_visible = cxl_dimm_visible
 };
 
 static const struct attribute_group *cxl_dimm_attribute_groups[] = {
@@ -58,6 +87,38 @@  static const struct attribute_group *cxl_dimm_attribute_groups[] = {
 	NULL
 };
 
+static void cxl_nvdimm_setup_dirty_tracking(struct cxl_nvdimm *cxl_nvd)
+{
+	u32 count;
+	struct cxl_memdev *cxlmd = cxl_nvd->cxlmd;
+	struct cxl_dev_state *cxlds = cxlmd->cxlds;
+	struct cxl_memdev_state *mds = to_cxl_memdev_state(cxlds);
+	struct device *dev = &cxl_nvd->dev;
+
+	/*
+	 * Dirty tracking is enabled and exposed to the user, only when:
+	 *   - dirty shutdown on the device can be set, and,
+	 *   - the device has a Device GPF DVSEC (albeit unused), and,
+	 *   - the Get Health Info cmd can retrieve the device's dirty count.
+	 */
+	cxl_nvd->dirty_shutdowns = CXL_INVALID_DIRTY_SHUTDOWN_COUNT;
+
+	if (cxl_arm_dirty_shutdown(mds)) {
+		dev_warn(dev, "GPF: could not set dirty shutdown state\n");
+		return;
+	}
+
+	if (cxl_gpf_get_dvsec(cxlds->dev, false) <= 0)
+		return;
+
+	if (cxl_get_dirty_count(mds, &count)) {
+		dev_warn(dev, "GPF: could not retrieve dirty count\n");
+		return;
+	}
+
+	cxl_nvd->dirty_shutdowns = count;
+}
+
 static int cxl_nvdimm_probe(struct device *dev)
 {
 	struct cxl_nvdimm *cxl_nvd = to_cxl_nvdimm(dev);
@@ -78,20 +139,20 @@  static int cxl_nvdimm_probe(struct device *dev)
 	set_bit(ND_CMD_GET_CONFIG_SIZE, &cmd_mask);
 	set_bit(ND_CMD_GET_CONFIG_DATA, &cmd_mask);
 	set_bit(ND_CMD_SET_CONFIG_DATA, &cmd_mask);
-	nvdimm = __nvdimm_create(cxl_nvb->nvdimm_bus, cxl_nvd,
-				 cxl_dimm_attribute_groups, flags,
-				 cmd_mask, 0, NULL, cxl_nvd->dev_id,
-				 cxl_security_ops, NULL);
-	if (!nvdimm)
-		return -ENOMEM;
 
 	/*
 	 * Set dirty shutdown now, with the expectation that the device
 	 * clear it upon a successful GPF flow. The exception to this
 	 * is upon Viral detection, per CXL 3.2 section 12.4.2.
 	 */
-	if (cxl_arm_dirty_shutdown(mds))
-		dev_warn(dev, "GPF: could not dirty shutdown state\n");
+	cxl_nvdimm_setup_dirty_tracking(cxl_nvd);
+
+	nvdimm = __nvdimm_create(cxl_nvb->nvdimm_bus, cxl_nvd,
+				 cxl_dimm_attribute_groups, flags,
+				 cmd_mask, 0, NULL, cxl_nvd->dev_id,
+				 cxl_security_ops, NULL);
+	if (!nvdimm)
+		return -ENOMEM;
 
 	dev_set_drvdata(dev, nvdimm);
 	return devm_add_action_or_reset(dev, unregister_nvdimm, nvdimm);