X86 platform drivers
 help / color / mirror / Atom feed
* [PATCH v5 00/12] Crashlog Type1 Version2 support
@ 2025-06-27 20:43 Michael J. Ruhl
  2025-06-27 20:43 ` [PATCH v5 01/12] platform/x86/intel/pmt: fix a crashlog NULL pointer access Michael J. Ruhl
                   ` (11 more replies)
  0 siblings, 12 replies; 18+ messages in thread
From: Michael J. Ruhl @ 2025-06-27 20:43 UTC (permalink / raw)
  To: platform-driver-x86, intel-xe, hdegoede, ilpo.jarvinen,
	lucas.demarchi, rodrigo.vivi, thomas.hellstrom, airlied, simona,
	david.e.box
  Cc: Michael J. Ruhl

The Intel BMG GPU device supports the crashlog feature, which was
exposed in an Xe driver patch (drm/xe/vsec: Support BMG devices),
however the version of crashlog used by the BMG GPU does not have
a supporing PMT driver.

Update the PMT crashlog driver to support the BMG crashlog feature.

v2:
 - fix a misconfig for the crashlog DVSEC info in the xe driver
 - address review comments
v3:
 - re-order bug fix patches for stable
 - added re-order trigger logic patch
 - added helper patch to address repeated code patterns
 - address review comments
v4:
 - added pcidev to intel_pmt_entry to address null issue
 - dropped endpoint update patches
 - patch cleanup and address review comments
v5:
 - renamed helpers to avoid namespace issues
 - separate mutex cleanup from gaurd usage
 - refactor base paramters to a separate patch (from version struct)
 - add r/b reviewed patches

Michael J. Ruhl (12):
  platform/x86/intel/pmt: fix a crashlog NULL pointer access
  drm/xe: Correct BMG VSEC header sizing
  platform/x86/intel/pmt: white space cleanup
  platform/x86/intel/pmt: mutex clean up
  platform/x86/intel/pmt: use guard(mutex)
  platform/x86/intel/pmt: re-order trigger logic
  platform/x86/intel/pmt: correct types
  platform/x86/intel/pmt: decouple sysfs and namespace
  platform/x86/intel/pmt: add register access helpers
  platform/x86/intel/pmt: refactor base parameter
  platform/x86/intel/pmt: use a version struct
  platform/x86/intel/pmt: support BMG crashlog

 drivers/gpu/drm/xe/xe_vsec.c              |  20 +-
 drivers/platform/x86/intel/pmt/class.c    |  15 +-
 drivers/platform/x86/intel/pmt/class.h    |   3 +-
 drivers/platform/x86/intel/pmt/crashlog.c | 468 ++++++++++++++++++----
 4 files changed, 394 insertions(+), 112 deletions(-)

-- 
2.49.0


^ permalink raw reply	[flat|nested] 18+ messages in thread

* [PATCH v5 01/12] platform/x86/intel/pmt: fix a crashlog NULL pointer access
  2025-06-27 20:43 [PATCH v5 00/12] Crashlog Type1 Version2 support Michael J. Ruhl
@ 2025-06-27 20:43 ` Michael J. Ruhl
  2025-06-30  9:29   ` Ilpo Järvinen
  2025-06-27 20:43 ` [PATCH v5 02/12] drm/xe: Correct BMG VSEC header sizing Michael J. Ruhl
                   ` (10 subsequent siblings)
  11 siblings, 1 reply; 18+ messages in thread
From: Michael J. Ruhl @ 2025-06-27 20:43 UTC (permalink / raw)
  To: platform-driver-x86, intel-xe, hdegoede, ilpo.jarvinen,
	lucas.demarchi, rodrigo.vivi, thomas.hellstrom, airlied, simona,
	david.e.box
  Cc: Michael J. Ruhl, Tejas Upadhyay, stable

Usage of the intel_pmt_read() for binary sysfs, requires a pcidev. The
current use of the endpoint value is only valid for telemetry endpoint
usage.

Without the ep, the crashlog usage causes the following NULL pointer
exception:

BUG: kernel NULL pointer dereference, address: 0000000000000000
Oops: Oops: 0000 [#1] SMP NOPTI
RIP: 0010:intel_pmt_read+0x3b/0x70 [pmt_class]
Code:
Call Trace:
 <TASK>
 ? sysfs_kf_bin_read+0xc0/0xe0
 kernfs_fop_read_iter+0xac/0x1a0
 vfs_read+0x26d/0x350
 ksys_read+0x6b/0xe0
 __x64_sys_read+0x1d/0x30
 x64_sys_call+0x1bc8/0x1d70
 do_syscall_64+0x6d/0x110

Augment struct intel_pmt_entry with a pointer to the pcidev to avoid
the NULL pointer exception.

Reviewed-by: Tejas Upadhyay <tejas.upadhyay@intel.com>
Fixes: 416eeb2e1fc7 ("platform/x86/intel/pmt: telemetry: Export API to read telemetry")
Cc: <stable@vger.kernel.org>
Signed-off-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
---
 drivers/platform/x86/intel/pmt/class.c | 3 ++-
 drivers/platform/x86/intel/pmt/class.h | 1 +
 2 files changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/platform/x86/intel/pmt/class.c b/drivers/platform/x86/intel/pmt/class.c
index 7233b654bbad..d046e8752173 100644
--- a/drivers/platform/x86/intel/pmt/class.c
+++ b/drivers/platform/x86/intel/pmt/class.c
@@ -97,7 +97,7 @@ intel_pmt_read(struct file *filp, struct kobject *kobj,
 	if (count > entry->size - off)
 		count = entry->size - off;
 
-	count = pmt_telem_read_mmio(entry->ep->pcidev, entry->cb, entry->header.guid, buf,
+	count = pmt_telem_read_mmio(entry->pcidev, entry->cb, entry->header.guid, buf,
 				    entry->base, off, count);
 
 	return count;
@@ -252,6 +252,7 @@ static int intel_pmt_populate_entry(struct intel_pmt_entry *entry,
 		return -EINVAL;
 	}
 
+	entry->pcidev = pci_dev;
 	entry->guid = header->guid;
 	entry->size = header->size;
 	entry->cb = ivdev->priv_data;
diff --git a/drivers/platform/x86/intel/pmt/class.h b/drivers/platform/x86/intel/pmt/class.h
index b2006d57779d..f6ce80c4e051 100644
--- a/drivers/platform/x86/intel/pmt/class.h
+++ b/drivers/platform/x86/intel/pmt/class.h
@@ -39,6 +39,7 @@ struct intel_pmt_header {
 
 struct intel_pmt_entry {
 	struct telem_endpoint	*ep;
+	struct pci_dev		*pcidev;
 	struct intel_pmt_header	header;
 	struct bin_attribute	pmt_bin_attr;
 	struct kobject		*kobj;
-- 
2.49.0


^ permalink raw reply related	[flat|nested] 18+ messages in thread

* [PATCH v5 02/12] drm/xe: Correct BMG VSEC header sizing
  2025-06-27 20:43 [PATCH v5 00/12] Crashlog Type1 Version2 support Michael J. Ruhl
  2025-06-27 20:43 ` [PATCH v5 01/12] platform/x86/intel/pmt: fix a crashlog NULL pointer access Michael J. Ruhl
@ 2025-06-27 20:43 ` Michael J. Ruhl
  2025-06-30 19:57   ` Rodrigo Vivi
  2025-06-27 20:43 ` [PATCH v5 03/12] platform/x86/intel/pmt: white space cleanup Michael J. Ruhl
                   ` (9 subsequent siblings)
  11 siblings, 1 reply; 18+ messages in thread
From: Michael J. Ruhl @ 2025-06-27 20:43 UTC (permalink / raw)
  To: platform-driver-x86, intel-xe, hdegoede, ilpo.jarvinen,
	lucas.demarchi, rodrigo.vivi, thomas.hellstrom, airlied, simona,
	david.e.box
  Cc: Michael J. Ruhl

The intel_vsec_header information for the crashlog feature is
incorrect.

Update the VSEC header with correct sizing and count.

Since the crashlog entries are "merged" (num_entries = 2), the
separate capabilities entries must be merged as well.

Fixes: 0c45e76fcc62 ("drm/xe/vsec: Support BMG devices")
Signed-off-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
---
 drivers/gpu/drm/xe/xe_vsec.c | 20 +++++---------------
 1 file changed, 5 insertions(+), 15 deletions(-)

diff --git a/drivers/gpu/drm/xe/xe_vsec.c b/drivers/gpu/drm/xe/xe_vsec.c
index 3e573b0b7ebd..67238fc57a4d 100644
--- a/drivers/gpu/drm/xe/xe_vsec.c
+++ b/drivers/gpu/drm/xe/xe_vsec.c
@@ -32,28 +32,18 @@ static struct intel_vsec_header bmg_telemetry = {
 	.offset = BMG_DISCOVERY_OFFSET,
 };
 
-static struct intel_vsec_header bmg_punit_crashlog = {
-	.length = 0x10,
+static struct intel_vsec_header bmg_crashlog = {
+	.length = 0x18,
 	.id = VSEC_ID_CRASHLOG,
-	.num_entries = 1,
-	.entry_size = 4,
+	.num_entries = 2,
+	.entry_size = 6,
 	.tbir = 0,
 	.offset = BMG_DISCOVERY_OFFSET + 0x60,
 };
 
-static struct intel_vsec_header bmg_oobmsm_crashlog = {
-	.length = 0x10,
-	.id = VSEC_ID_CRASHLOG,
-	.num_entries = 1,
-	.entry_size = 4,
-	.tbir = 0,
-	.offset = BMG_DISCOVERY_OFFSET + 0x78,
-};
-
 static struct intel_vsec_header *bmg_capabilities[] = {
 	&bmg_telemetry,
-	&bmg_punit_crashlog,
-	&bmg_oobmsm_crashlog,
+	&bmg_crashlog,
 	NULL
 };
 
-- 
2.49.0


^ permalink raw reply related	[flat|nested] 18+ messages in thread

* [PATCH v5 03/12] platform/x86/intel/pmt: white space cleanup
  2025-06-27 20:43 [PATCH v5 00/12] Crashlog Type1 Version2 support Michael J. Ruhl
  2025-06-27 20:43 ` [PATCH v5 01/12] platform/x86/intel/pmt: fix a crashlog NULL pointer access Michael J. Ruhl
  2025-06-27 20:43 ` [PATCH v5 02/12] drm/xe: Correct BMG VSEC header sizing Michael J. Ruhl
@ 2025-06-27 20:43 ` Michael J. Ruhl
  2025-06-27 20:43 ` [PATCH v5 04/12] platform/x86/intel/pmt: mutex clean up Michael J. Ruhl
                   ` (8 subsequent siblings)
  11 siblings, 0 replies; 18+ messages in thread
From: Michael J. Ruhl @ 2025-06-27 20:43 UTC (permalink / raw)
  To: platform-driver-x86, intel-xe, hdegoede, ilpo.jarvinen,
	lucas.demarchi, rodrigo.vivi, thomas.hellstrom, airlied, simona,
	david.e.box
  Cc: Michael J. Ruhl

Noticed two white space issues; cleaned them.

Reviewed-by: David E. Box <david.e.box@linux.intel.com>
Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Signed-off-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
---
 drivers/platform/x86/intel/pmt/crashlog.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/platform/x86/intel/pmt/crashlog.c b/drivers/platform/x86/intel/pmt/crashlog.c
index 6a9eb3c4b313..d40c8e212733 100644
--- a/drivers/platform/x86/intel/pmt/crashlog.c
+++ b/drivers/platform/x86/intel/pmt/crashlog.c
@@ -143,7 +143,7 @@ enable_show(struct device *dev, struct device_attribute *attr, char *buf)
 
 static ssize_t
 enable_store(struct device *dev, struct device_attribute *attr,
-	    const char *buf, size_t count)
+	     const char *buf, size_t count)
 {
 	struct crashlog_entry *entry;
 	bool enabled;
@@ -177,7 +177,7 @@ trigger_show(struct device *dev, struct device_attribute *attr, char *buf)
 
 static ssize_t
 trigger_store(struct device *dev, struct device_attribute *attr,
-	    const char *buf, size_t count)
+	      const char *buf, size_t count)
 {
 	struct crashlog_entry *entry;
 	bool trigger;
-- 
2.49.0


^ permalink raw reply related	[flat|nested] 18+ messages in thread

* [PATCH v5 04/12] platform/x86/intel/pmt: mutex clean up
  2025-06-27 20:43 [PATCH v5 00/12] Crashlog Type1 Version2 support Michael J. Ruhl
                   ` (2 preceding siblings ...)
  2025-06-27 20:43 ` [PATCH v5 03/12] platform/x86/intel/pmt: white space cleanup Michael J. Ruhl
@ 2025-06-27 20:43 ` Michael J. Ruhl
  2025-06-27 20:43 ` [PATCH v5 05/12] platform/x86/intel/pmt: use guard(mutex) Michael J. Ruhl
                   ` (7 subsequent siblings)
  11 siblings, 0 replies; 18+ messages in thread
From: Michael J. Ruhl @ 2025-06-27 20:43 UTC (permalink / raw)
  To: platform-driver-x86, intel-xe, hdegoede, ilpo.jarvinen,
	lucas.demarchi, rodrigo.vivi, thomas.hellstrom, airlied, simona,
	david.e.box
  Cc: Michael J. Ruhl

The header file for mutex usage and mutex_remove() cleanup code is
absent from this module.

Add the header file and mutex_revove().

Signed-off-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
---
 drivers/platform/x86/intel/pmt/crashlog.c | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/drivers/platform/x86/intel/pmt/crashlog.c b/drivers/platform/x86/intel/pmt/crashlog.c
index d40c8e212733..6e32fc1f8f1d 100644
--- a/drivers/platform/x86/intel/pmt/crashlog.c
+++ b/drivers/platform/x86/intel/pmt/crashlog.c
@@ -12,6 +12,7 @@
 #include <linux/intel_vsec.h>
 #include <linux/kernel.h>
 #include <linux/module.h>
+#include <linux/mutex.h>
 #include <linux/pci.h>
 #include <linux/slab.h>
 #include <linux/uaccess.h>
@@ -262,8 +263,12 @@ static void pmt_crashlog_remove(struct auxiliary_device *auxdev)
 	struct pmt_crashlog_priv *priv = auxiliary_get_drvdata(auxdev);
 	int i;
 
-	for (i = 0; i < priv->num_entries; i++)
-		intel_pmt_dev_destroy(&priv->entry[i].entry, &pmt_crashlog_ns);
+	for (i = 0; i < priv->num_entries; i++) {
+		struct crashlog_entry *crashlog = &priv->entry[i];
+
+		intel_pmt_dev_destroy(&crashlog->entry, &pmt_crashlog_ns);
+		mutex_destroy(&crashlog->control_mutex);
+	}
 }
 
 static int pmt_crashlog_probe(struct auxiliary_device *auxdev,
-- 
2.49.0


^ permalink raw reply related	[flat|nested] 18+ messages in thread

* [PATCH v5 05/12] platform/x86/intel/pmt: use guard(mutex)
  2025-06-27 20:43 [PATCH v5 00/12] Crashlog Type1 Version2 support Michael J. Ruhl
                   ` (3 preceding siblings ...)
  2025-06-27 20:43 ` [PATCH v5 04/12] platform/x86/intel/pmt: mutex clean up Michael J. Ruhl
@ 2025-06-27 20:43 ` Michael J. Ruhl
  2025-06-27 20:43 ` [PATCH v5 06/12] platform/x86/intel/pmt: re-order trigger logic Michael J. Ruhl
                   ` (6 subsequent siblings)
  11 siblings, 0 replies; 18+ messages in thread
From: Michael J. Ruhl @ 2025-06-27 20:43 UTC (permalink / raw)
  To: platform-driver-x86, intel-xe, hdegoede, ilpo.jarvinen,
	lucas.demarchi, rodrigo.vivi, thomas.hellstrom, airlied, simona,
	david.e.box
  Cc: Michael J. Ruhl

Update the mutex paths to use the new guard() mechanism.

With the removal of goto, do some minor cleanup of the current logic
path.

Reviewed-by: David E. Box <david.e.box@linux.intel.com>
Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Signed-off-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
---
 drivers/platform/x86/intel/pmt/crashlog.c | 33 +++++++++++------------
 1 file changed, 16 insertions(+), 17 deletions(-)

diff --git a/drivers/platform/x86/intel/pmt/crashlog.c b/drivers/platform/x86/intel/pmt/crashlog.c
index 6e32fc1f8f1d..c3ca95854aba 100644
--- a/drivers/platform/x86/intel/pmt/crashlog.c
+++ b/drivers/platform/x86/intel/pmt/crashlog.c
@@ -9,6 +9,7 @@
  */
 
 #include <linux/auxiliary_bus.h>
+#include <linux/cleanup.h>
 #include <linux/intel_vsec.h>
 #include <linux/kernel.h>
 #include <linux/module.h>
@@ -156,9 +157,9 @@ enable_store(struct device *dev, struct device_attribute *attr,
 	if (result)
 		return result;
 
-	mutex_lock(&entry->control_mutex);
+	guard(mutex)(&entry->control_mutex);
+
 	pmt_crashlog_set_disable(&entry->entry, !enabled);
-	mutex_unlock(&entry->control_mutex);
 
 	return count;
 }
@@ -190,26 +191,24 @@ trigger_store(struct device *dev, struct device_attribute *attr,
 	if (result)
 		return result;
 
-	mutex_lock(&entry->control_mutex);
+	guard(mutex)(&entry->control_mutex);
 
 	if (!trigger) {
 		pmt_crashlog_set_clear(&entry->entry);
-	} else if (pmt_crashlog_complete(&entry->entry)) {
-		/* we cannot trigger a new crash if one is still pending */
-		result = -EEXIST;
-		goto err;
-	} else if (pmt_crashlog_disabled(&entry->entry)) {
-		/* if device is currently disabled, return busy */
-		result = -EBUSY;
-		goto err;
-	} else {
-		pmt_crashlog_set_execute(&entry->entry);
+		return count;
 	}
 
-	result = count;
-err:
-	mutex_unlock(&entry->control_mutex);
-	return result;
+	/* we cannot trigger a new crash if one is still pending */
+	if (pmt_crashlog_complete(&entry->entry))
+		return -EEXIST;
+
+	/* if device is currently disabled, return busy */
+	if (pmt_crashlog_disabled(&entry->entry))
+		return -EBUSY;
+
+	pmt_crashlog_set_execute(&entry->entry);
+
+	return count;
 }
 static DEVICE_ATTR_RW(trigger);
 
-- 
2.49.0


^ permalink raw reply related	[flat|nested] 18+ messages in thread

* [PATCH v5 06/12] platform/x86/intel/pmt: re-order trigger logic
  2025-06-27 20:43 [PATCH v5 00/12] Crashlog Type1 Version2 support Michael J. Ruhl
                   ` (4 preceding siblings ...)
  2025-06-27 20:43 ` [PATCH v5 05/12] platform/x86/intel/pmt: use guard(mutex) Michael J. Ruhl
@ 2025-06-27 20:43 ` Michael J. Ruhl
  2025-06-27 20:43 ` [PATCH v5 07/12] platform/x86/intel/pmt: correct types Michael J. Ruhl
                   ` (5 subsequent siblings)
  11 siblings, 0 replies; 18+ messages in thread
From: Michael J. Ruhl @ 2025-06-27 20:43 UTC (permalink / raw)
  To: platform-driver-x86, intel-xe, hdegoede, ilpo.jarvinen,
	lucas.demarchi, rodrigo.vivi, thomas.hellstrom, airlied, simona,
	david.e.box
  Cc: Michael J. Ruhl

Setting the clear bit or checking the complete bit before checking to
see if crashlog is disabled seems incorrect.

Check disable before accessing any other bits.

Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Reviewed-by: David E. Box <david.e.box@linux.intel.com>
Signed-off-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
---
 drivers/platform/x86/intel/pmt/crashlog.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/drivers/platform/x86/intel/pmt/crashlog.c b/drivers/platform/x86/intel/pmt/crashlog.c
index c3ca95854aba..440d2045e90d 100644
--- a/drivers/platform/x86/intel/pmt/crashlog.c
+++ b/drivers/platform/x86/intel/pmt/crashlog.c
@@ -193,6 +193,10 @@ trigger_store(struct device *dev, struct device_attribute *attr,
 
 	guard(mutex)(&entry->control_mutex);
 
+	/* if device is currently disabled, return busy */
+	if (pmt_crashlog_disabled(&entry->entry))
+		return -EBUSY;
+
 	if (!trigger) {
 		pmt_crashlog_set_clear(&entry->entry);
 		return count;
@@ -202,10 +206,6 @@ trigger_store(struct device *dev, struct device_attribute *attr,
 	if (pmt_crashlog_complete(&entry->entry))
 		return -EEXIST;
 
-	/* if device is currently disabled, return busy */
-	if (pmt_crashlog_disabled(&entry->entry))
-		return -EBUSY;
-
 	pmt_crashlog_set_execute(&entry->entry);
 
 	return count;
-- 
2.49.0


^ permalink raw reply related	[flat|nested] 18+ messages in thread

* [PATCH v5 07/12] platform/x86/intel/pmt: correct types
  2025-06-27 20:43 [PATCH v5 00/12] Crashlog Type1 Version2 support Michael J. Ruhl
                   ` (5 preceding siblings ...)
  2025-06-27 20:43 ` [PATCH v5 06/12] platform/x86/intel/pmt: re-order trigger logic Michael J. Ruhl
@ 2025-06-27 20:43 ` Michael J. Ruhl
  2025-06-27 20:43 ` [PATCH v5 08/12] platform/x86/intel/pmt: decouple sysfs and namespace Michael J. Ruhl
                   ` (4 subsequent siblings)
  11 siblings, 0 replies; 18+ messages in thread
From: Michael J. Ruhl @ 2025-06-27 20:43 UTC (permalink / raw)
  To: platform-driver-x86, intel-xe, hdegoede, ilpo.jarvinen,
	lucas.demarchi, rodrigo.vivi, thomas.hellstrom, airlied, simona,
	david.e.box
  Cc: Michael J. Ruhl

A couple of auto variables do not match the return types of some of
the functions.

Update the mismatched types to match.

Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Reviewed-by: David E. Box <david.e.box@linux.intel.com>
Signed-off-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
---
 drivers/platform/x86/intel/pmt/crashlog.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/platform/x86/intel/pmt/crashlog.c b/drivers/platform/x86/intel/pmt/crashlog.c
index 440d2045e90d..881f4abdae14 100644
--- a/drivers/platform/x86/intel/pmt/crashlog.c
+++ b/drivers/platform/x86/intel/pmt/crashlog.c
@@ -138,7 +138,7 @@ static ssize_t
 enable_show(struct device *dev, struct device_attribute *attr, char *buf)
 {
 	struct intel_pmt_entry *entry = dev_get_drvdata(dev);
-	int enabled = !pmt_crashlog_disabled(entry);
+	bool enabled = !pmt_crashlog_disabled(entry);
 
 	return sprintf(buf, "%d\n", enabled);
 }
@@ -169,7 +169,7 @@ static ssize_t
 trigger_show(struct device *dev, struct device_attribute *attr, char *buf)
 {
 	struct intel_pmt_entry *entry;
-	int trigger;
+	bool trigger;
 
 	entry = dev_get_drvdata(dev);
 	trigger = pmt_crashlog_complete(entry);
-- 
2.49.0


^ permalink raw reply related	[flat|nested] 18+ messages in thread

* [PATCH v5 08/12] platform/x86/intel/pmt: decouple sysfs and namespace
  2025-06-27 20:43 [PATCH v5 00/12] Crashlog Type1 Version2 support Michael J. Ruhl
                   ` (6 preceding siblings ...)
  2025-06-27 20:43 ` [PATCH v5 07/12] platform/x86/intel/pmt: correct types Michael J. Ruhl
@ 2025-06-27 20:43 ` Michael J. Ruhl
  2025-06-27 20:43 ` [PATCH v5 09/12] platform/x86/intel/pmt: add register access helpers Michael J. Ruhl
                   ` (3 subsequent siblings)
  11 siblings, 0 replies; 18+ messages in thread
From: Michael J. Ruhl @ 2025-06-27 20:43 UTC (permalink / raw)
  To: platform-driver-x86, intel-xe, hdegoede, ilpo.jarvinen,
	lucas.demarchi, rodrigo.vivi, thomas.hellstrom, airlied, simona,
	david.e.box
  Cc: Michael J. Ruhl

The PMT namespace includes the crashlog sysfs attribute information.
Other crashlog version/types may need different sysfs attributes.
Coupling the attributes with the namespace blocks this usage.

Decouple sysfs attributes from the name space and add them to the
specific entry.

Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Signed-off-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
---
 drivers/platform/x86/intel/pmt/class.c    | 12 ++++++------
 drivers/platform/x86/intel/pmt/class.h    |  2 +-
 drivers/platform/x86/intel/pmt/crashlog.c |  3 ++-
 3 files changed, 9 insertions(+), 8 deletions(-)

diff --git a/drivers/platform/x86/intel/pmt/class.c b/drivers/platform/x86/intel/pmt/class.c
index d046e8752173..3b6bf2f14dcb 100644
--- a/drivers/platform/x86/intel/pmt/class.c
+++ b/drivers/platform/x86/intel/pmt/class.c
@@ -285,8 +285,8 @@ static int intel_pmt_dev_register(struct intel_pmt_entry *entry,
 
 	entry->kobj = &dev->kobj;
 
-	if (ns->attr_grp) {
-		ret = sysfs_create_group(entry->kobj, ns->attr_grp);
+	if (entry->attr_grp) {
+		ret = sysfs_create_group(entry->kobj, entry->attr_grp);
 		if (ret)
 			goto fail_sysfs_create_group;
 	}
@@ -327,8 +327,8 @@ static int intel_pmt_dev_register(struct intel_pmt_entry *entry,
 fail_add_endpoint:
 	sysfs_remove_bin_file(entry->kobj, &entry->pmt_bin_attr);
 fail_ioremap:
-	if (ns->attr_grp)
-		sysfs_remove_group(entry->kobj, ns->attr_grp);
+	if (entry->attr_grp)
+		sysfs_remove_group(entry->kobj, entry->attr_grp);
 fail_sysfs_create_group:
 	device_unregister(dev);
 fail_dev_create:
@@ -370,8 +370,8 @@ void intel_pmt_dev_destroy(struct intel_pmt_entry *entry,
 	if (entry->size)
 		sysfs_remove_bin_file(entry->kobj, &entry->pmt_bin_attr);
 
-	if (ns->attr_grp)
-		sysfs_remove_group(entry->kobj, ns->attr_grp);
+	if (entry->attr_grp)
+		sysfs_remove_group(entry->kobj, entry->attr_grp);
 
 	device_unregister(dev);
 	xa_erase(ns->xa, entry->devid);
diff --git a/drivers/platform/x86/intel/pmt/class.h b/drivers/platform/x86/intel/pmt/class.h
index f6ce80c4e051..d5d86b8a2d15 100644
--- a/drivers/platform/x86/intel/pmt/class.h
+++ b/drivers/platform/x86/intel/pmt/class.h
@@ -42,6 +42,7 @@ struct intel_pmt_entry {
 	struct pci_dev		*pcidev;
 	struct intel_pmt_header	header;
 	struct bin_attribute	pmt_bin_attr;
+	const struct attribute_group *attr_grp;
 	struct kobject		*kobj;
 	void __iomem		*disc_table;
 	void __iomem		*base;
@@ -55,7 +56,6 @@ struct intel_pmt_entry {
 struct intel_pmt_namespace {
 	const char *name;
 	struct xarray *xa;
-	const struct attribute_group *attr_grp;
 	int (*pmt_header_decode)(struct intel_pmt_entry *entry,
 				 struct device *dev);
 	int (*pmt_add_endpoint)(struct intel_vsec_device *ivdev,
diff --git a/drivers/platform/x86/intel/pmt/crashlog.c b/drivers/platform/x86/intel/pmt/crashlog.c
index 881f4abdae14..23b3971da40a 100644
--- a/drivers/platform/x86/intel/pmt/crashlog.c
+++ b/drivers/platform/x86/intel/pmt/crashlog.c
@@ -243,6 +243,8 @@ static int pmt_crashlog_header_decode(struct intel_pmt_entry *entry,
 	/* Size is measured in DWORDS, but accessor returns bytes */
 	header->size = GET_SIZE(readl(disc_table + SIZE_OFFSET));
 
+	entry->attr_grp = &pmt_crashlog_group;
+
 	return 0;
 }
 
@@ -250,7 +252,6 @@ static DEFINE_XARRAY_ALLOC(crashlog_array);
 static struct intel_pmt_namespace pmt_crashlog_ns = {
 	.name = "crashlog",
 	.xa = &crashlog_array,
-	.attr_grp = &pmt_crashlog_group,
 	.pmt_header_decode = pmt_crashlog_header_decode,
 };
 
-- 
2.49.0


^ permalink raw reply related	[flat|nested] 18+ messages in thread

* [PATCH v5 09/12] platform/x86/intel/pmt: add register access helpers
  2025-06-27 20:43 [PATCH v5 00/12] Crashlog Type1 Version2 support Michael J. Ruhl
                   ` (7 preceding siblings ...)
  2025-06-27 20:43 ` [PATCH v5 08/12] platform/x86/intel/pmt: decouple sysfs and namespace Michael J. Ruhl
@ 2025-06-27 20:43 ` Michael J. Ruhl
  2025-07-02 15:50   ` Ruhl, Michael J
  2025-06-27 20:43 ` [PATCH v5 10/12] platform/x86/intel/pmt: refactor base parameter Michael J. Ruhl
                   ` (2 subsequent siblings)
  11 siblings, 1 reply; 18+ messages in thread
From: Michael J. Ruhl @ 2025-06-27 20:43 UTC (permalink / raw)
  To: platform-driver-x86, intel-xe, hdegoede, ilpo.jarvinen,
	lucas.demarchi, rodrigo.vivi, thomas.hellstrom, airlied, simona,
	david.e.box
  Cc: Michael J. Ruhl

The control register is used in a read/modify/write pattern.
The status register is used in a read/check bit pattern.

Add helpers to eliminate common code.

Signed-off-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
---
 drivers/platform/x86/intel/pmt/crashlog.c | 60 ++++++++++++-----------
 1 file changed, 31 insertions(+), 29 deletions(-)

diff --git a/drivers/platform/x86/intel/pmt/crashlog.c b/drivers/platform/x86/intel/pmt/crashlog.c
index 23b3971da40a..adaca7ce1ba5 100644
--- a/drivers/platform/x86/intel/pmt/crashlog.c
+++ b/drivers/platform/x86/intel/pmt/crashlog.c
@@ -64,20 +64,42 @@ struct pmt_crashlog_priv {
 /*
  * I/O
  */
-static bool pmt_crashlog_complete(struct intel_pmt_entry *entry)
+#define CRASHLOG_SET_BIT	true
+#define CRASHLOG_CLEAR_BIT	false
+
+/* read/modify/write */
+static void pmt_crashlog_rmw(struct intel_pmt_entry *entry, u32 bit, bool set)
 {
-	u32 control = readl(entry->disc_table + CONTROL_OFFSET);
+	u32 reg = readl(entry->disc_table + CONTROL_OFFSET);
+
+	reg &= ~CRASHLOG_FLAG_TRIGGER_MASK;
+
+	if (set)
+		reg |= bit;
+	else
+		reg &= bit;
+
+	writel(reg, entry->disc_table + CONTROL_OFFSET);
+}
+
+/* read/check */
+static bool pmt_crashlog_rc(struct intel_pmt_entry *entry, u32 bit)
+{
+	u32 reg = readl(entry->disc_table + CONTROL_OFFSET);
+
+	return !!(reg & bit);
+}
 
+static bool pmt_crashlog_complete(struct intel_pmt_entry *entry)
+{
 	/* return current value of the crashlog complete flag */
-	return !!(control & CRASHLOG_FLAG_TRIGGER_COMPLETE);
+	return pmt_crashlog_rc(entry, CRASHLOG_FLAG_TRIGGER_COMPLETE);
 }
 
 static bool pmt_crashlog_disabled(struct intel_pmt_entry *entry)
 {
-	u32 control = readl(entry->disc_table + CONTROL_OFFSET);
-
 	/* return current value of the crashlog disabled flag */
-	return !!(control & CRASHLOG_FLAG_DISABLE);
+	return pmt_crashlog_rc(entry, CRASHLOG_FLAG_DISABLE);
 }
 
 static bool pmt_crashlog_supported(struct intel_pmt_entry *entry)
@@ -98,37 +120,17 @@ static bool pmt_crashlog_supported(struct intel_pmt_entry *entry)
 static void pmt_crashlog_set_disable(struct intel_pmt_entry *entry,
 				     bool disable)
 {
-	u32 control = readl(entry->disc_table + CONTROL_OFFSET);
-
-	/* clear trigger bits so we are only modifying disable flag */
-	control &= ~CRASHLOG_FLAG_TRIGGER_MASK;
-
-	if (disable)
-		control |= CRASHLOG_FLAG_DISABLE;
-	else
-		control &= ~CRASHLOG_FLAG_DISABLE;
-
-	writel(control, entry->disc_table + CONTROL_OFFSET);
+	pmt_crashlog_rmw(entry, CRASHLOG_FLAG_DISABLE, disable);
 }
 
 static void pmt_crashlog_set_clear(struct intel_pmt_entry *entry)
 {
-	u32 control = readl(entry->disc_table + CONTROL_OFFSET);
-
-	control &= ~CRASHLOG_FLAG_TRIGGER_MASK;
-	control |= CRASHLOG_FLAG_TRIGGER_CLEAR;
-
-	writel(control, entry->disc_table + CONTROL_OFFSET);
+	pmt_crashlog_rmw(entry, CRASHLOG_FLAG_TRIGGER_CLEAR, CRASHLOG_SET_BIT);
 }
 
 static void pmt_crashlog_set_execute(struct intel_pmt_entry *entry)
 {
-	u32 control = readl(entry->disc_table + CONTROL_OFFSET);
-
-	control &= ~CRASHLOG_FLAG_TRIGGER_MASK;
-	control |= CRASHLOG_FLAG_TRIGGER_EXECUTE;
-
-	writel(control, entry->disc_table + CONTROL_OFFSET);
+	pmt_crashlog_rmw(entry, CRASHLOG_FLAG_TRIGGER_EXECUTE, CRASHLOG_SET_BIT);
 }
 
 /*
-- 
2.49.0


^ permalink raw reply related	[flat|nested] 18+ messages in thread

* [PATCH v5 10/12] platform/x86/intel/pmt: refactor base parameter
  2025-06-27 20:43 [PATCH v5 00/12] Crashlog Type1 Version2 support Michael J. Ruhl
                   ` (8 preceding siblings ...)
  2025-06-27 20:43 ` [PATCH v5 09/12] platform/x86/intel/pmt: add register access helpers Michael J. Ruhl
@ 2025-06-27 20:43 ` Michael J. Ruhl
  2025-06-27 20:43 ` [PATCH v5 11/12] platform/x86/intel/pmt: use a version struct Michael J. Ruhl
  2025-06-27 20:43 ` [PATCH v5 12/12] platform/x86/intel/pmt: support BMG crashlog Michael J. Ruhl
  11 siblings, 0 replies; 18+ messages in thread
From: Michael J. Ruhl @ 2025-06-27 20:43 UTC (permalink / raw)
  To: platform-driver-x86, intel-xe, hdegoede, ilpo.jarvinen,
	lucas.demarchi, rodrigo.vivi, thomas.hellstrom, airlied, simona,
	david.e.box
  Cc: Michael J. Ruhl

For the crashlog driver, struct crashlog_entry is the parent of
struct intel_pmt_entry. To support multiple crashlog versions, most
accesses will be to the struct crashlog_entry.

- Refactor to use struct crashlog_entry in place of
  struct intel_pmt_entry
- Rename some usages (auto-variables) from entry to crashlog

Signed-off-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
---
 drivers/platform/x86/intel/pmt/crashlog.c | 59 ++++++++++++-----------
 1 file changed, 30 insertions(+), 29 deletions(-)

diff --git a/drivers/platform/x86/intel/pmt/crashlog.c b/drivers/platform/x86/intel/pmt/crashlog.c
index adaca7ce1ba5..7975abb3c21b 100644
--- a/drivers/platform/x86/intel/pmt/crashlog.c
+++ b/drivers/platform/x86/intel/pmt/crashlog.c
@@ -68,8 +68,9 @@ struct pmt_crashlog_priv {
 #define CRASHLOG_CLEAR_BIT	false
 
 /* read/modify/write */
-static void pmt_crashlog_rmw(struct intel_pmt_entry *entry, u32 bit, bool set)
+static void pmt_crashlog_rmw(struct crashlog_entry *crashlog, u32 bit, bool set)
 {
+	struct intel_pmt_entry *entry = &crashlog->entry;
 	u32 reg = readl(entry->disc_table + CONTROL_OFFSET);
 
 	reg &= ~CRASHLOG_FLAG_TRIGGER_MASK;
@@ -83,23 +84,24 @@ static void pmt_crashlog_rmw(struct intel_pmt_entry *entry, u32 bit, bool set)
 }
 
 /* read/check */
-static bool pmt_crashlog_rc(struct intel_pmt_entry *entry, u32 bit)
+static bool pmt_crashlog_rc(struct crashlog_entry *crashlog, u32 bit)
 {
+	struct intel_pmt_entry *entry = &crashlog->entry;
 	u32 reg = readl(entry->disc_table + CONTROL_OFFSET);
 
 	return !!(reg & bit);
 }
 
-static bool pmt_crashlog_complete(struct intel_pmt_entry *entry)
+static bool pmt_crashlog_complete(struct crashlog_entry *crashlog)
 {
 	/* return current value of the crashlog complete flag */
-	return pmt_crashlog_rc(entry, CRASHLOG_FLAG_TRIGGER_COMPLETE);
+	return pmt_crashlog_rc(crashlog, CRASHLOG_FLAG_TRIGGER_COMPLETE);
 }
 
-static bool pmt_crashlog_disabled(struct intel_pmt_entry *entry)
+static bool pmt_crashlog_disabled(struct crashlog_entry *crashlog)
 {
 	/* return current value of the crashlog disabled flag */
-	return pmt_crashlog_rc(entry, CRASHLOG_FLAG_DISABLE);
+	return pmt_crashlog_rc(crashlog, CRASHLOG_FLAG_DISABLE);
 }
 
 static bool pmt_crashlog_supported(struct intel_pmt_entry *entry)
@@ -117,20 +119,19 @@ static bool pmt_crashlog_supported(struct intel_pmt_entry *entry)
 	return crash_type == CRASH_TYPE_OOBMSM && version == 0;
 }
 
-static void pmt_crashlog_set_disable(struct intel_pmt_entry *entry,
-				     bool disable)
+static void pmt_crashlog_set_disable(struct crashlog_entry *crashlog, bool disable)
 {
-	pmt_crashlog_rmw(entry, CRASHLOG_FLAG_DISABLE, disable);
+	pmt_crashlog_rmw(crashlog, CRASHLOG_FLAG_DISABLE, disable);
 }
 
-static void pmt_crashlog_set_clear(struct intel_pmt_entry *entry)
+static void pmt_crashlog_set_clear(struct crashlog_entry *crashlog)
 {
-	pmt_crashlog_rmw(entry, CRASHLOG_FLAG_TRIGGER_CLEAR, CRASHLOG_SET_BIT);
+	pmt_crashlog_rmw(crashlog, CRASHLOG_FLAG_TRIGGER_CLEAR, CRASHLOG_SET_BIT);
 }
 
-static void pmt_crashlog_set_execute(struct intel_pmt_entry *entry)
+static void pmt_crashlog_set_execute(struct crashlog_entry *crashlog)
 {
-	pmt_crashlog_rmw(entry, CRASHLOG_FLAG_TRIGGER_EXECUTE, CRASHLOG_SET_BIT);
+	pmt_crashlog_rmw(crashlog, CRASHLOG_FLAG_TRIGGER_EXECUTE, CRASHLOG_SET_BIT);
 }
 
 /*
@@ -139,8 +140,8 @@ static void pmt_crashlog_set_execute(struct intel_pmt_entry *entry)
 static ssize_t
 enable_show(struct device *dev, struct device_attribute *attr, char *buf)
 {
-	struct intel_pmt_entry *entry = dev_get_drvdata(dev);
-	bool enabled = !pmt_crashlog_disabled(entry);
+	struct crashlog_entry *crashlog = dev_get_drvdata(dev);
+	bool enabled = !pmt_crashlog_disabled(crashlog);
 
 	return sprintf(buf, "%d\n", enabled);
 }
@@ -149,19 +150,19 @@ static ssize_t
 enable_store(struct device *dev, struct device_attribute *attr,
 	     const char *buf, size_t count)
 {
-	struct crashlog_entry *entry;
+	struct crashlog_entry *crashlog;
 	bool enabled;
 	int result;
 
-	entry = dev_get_drvdata(dev);
+	crashlog = dev_get_drvdata(dev);
 
 	result = kstrtobool(buf, &enabled);
 	if (result)
 		return result;
 
-	guard(mutex)(&entry->control_mutex);
+	guard(mutex)(&crashlog->control_mutex);
 
-	pmt_crashlog_set_disable(&entry->entry, !enabled);
+	pmt_crashlog_set_disable(crashlog, !enabled);
 
 	return count;
 }
@@ -170,11 +171,11 @@ static DEVICE_ATTR_RW(enable);
 static ssize_t
 trigger_show(struct device *dev, struct device_attribute *attr, char *buf)
 {
-	struct intel_pmt_entry *entry;
+	struct crashlog_entry *crashlog;
 	bool trigger;
 
-	entry = dev_get_drvdata(dev);
-	trigger = pmt_crashlog_complete(entry);
+	crashlog = dev_get_drvdata(dev);
+	trigger = pmt_crashlog_complete(crashlog);
 
 	return sprintf(buf, "%d\n", trigger);
 }
@@ -183,32 +184,32 @@ static ssize_t
 trigger_store(struct device *dev, struct device_attribute *attr,
 	      const char *buf, size_t count)
 {
-	struct crashlog_entry *entry;
+	struct crashlog_entry *crashlog;
 	bool trigger;
 	int result;
 
-	entry = dev_get_drvdata(dev);
+	crashlog = dev_get_drvdata(dev);
 
 	result = kstrtobool(buf, &trigger);
 	if (result)
 		return result;
 
-	guard(mutex)(&entry->control_mutex);
+	guard(mutex)(&crashlog->control_mutex);
 
 	/* if device is currently disabled, return busy */
-	if (pmt_crashlog_disabled(&entry->entry))
+	if (pmt_crashlog_disabled(crashlog))
 		return -EBUSY;
 
 	if (!trigger) {
-		pmt_crashlog_set_clear(&entry->entry);
+		pmt_crashlog_set_clear(crashlog);
 		return count;
 	}
 
 	/* we cannot trigger a new crash if one is still pending */
-	if (pmt_crashlog_complete(&entry->entry))
+	if (pmt_crashlog_complete(crashlog))
 		return -EEXIST;
 
-	pmt_crashlog_set_execute(&entry->entry);
+	pmt_crashlog_set_execute(crashlog);
 
 	return count;
 }
-- 
2.49.0


^ permalink raw reply related	[flat|nested] 18+ messages in thread

* [PATCH v5 11/12] platform/x86/intel/pmt: use a version struct
  2025-06-27 20:43 [PATCH v5 00/12] Crashlog Type1 Version2 support Michael J. Ruhl
                   ` (9 preceding siblings ...)
  2025-06-27 20:43 ` [PATCH v5 10/12] platform/x86/intel/pmt: refactor base parameter Michael J. Ruhl
@ 2025-06-27 20:43 ` Michael J. Ruhl
  2025-06-27 20:43 ` [PATCH v5 12/12] platform/x86/intel/pmt: support BMG crashlog Michael J. Ruhl
  11 siblings, 0 replies; 18+ messages in thread
From: Michael J. Ruhl @ 2025-06-27 20:43 UTC (permalink / raw)
  To: platform-driver-x86, intel-xe, hdegoede, ilpo.jarvinen,
	lucas.demarchi, rodrigo.vivi, thomas.hellstrom, airlied, simona,
	david.e.box
  Cc: Michael J. Ruhl

In preparation for supporting multiple crashlog versions, use a struct
to keep bit offset info for the status and control bits.

Signed-off-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
---
 drivers/platform/x86/intel/pmt/crashlog.c | 92 ++++++++++++++++-------
 1 file changed, 66 insertions(+), 26 deletions(-)

diff --git a/drivers/platform/x86/intel/pmt/crashlog.c b/drivers/platform/x86/intel/pmt/crashlog.c
index 7975abb3c21b..d109f307f110 100644
--- a/drivers/platform/x86/intel/pmt/crashlog.c
+++ b/drivers/platform/x86/intel/pmt/crashlog.c
@@ -24,21 +24,6 @@
 /* Crashlog discovery header types */
 #define CRASH_TYPE_OOBMSM	1
 
-/* Control Flags */
-#define CRASHLOG_FLAG_DISABLE		BIT(28)
-
-/*
- * Bits 29 and 30 control the state of bit 31.
- *
- * Bit 29 will clear bit 31, if set, allowing a new crashlog to be captured.
- * Bit 30 will immediately trigger a crashlog to be generated, setting bit 31.
- * Bit 31 is the read-only status with a 1 indicating log is complete.
- */
-#define CRASHLOG_FLAG_TRIGGER_CLEAR	BIT(29)
-#define CRASHLOG_FLAG_TRIGGER_EXECUTE	BIT(30)
-#define CRASHLOG_FLAG_TRIGGER_COMPLETE	BIT(31)
-#define CRASHLOG_FLAG_TRIGGER_MASK	GENMASK(31, 28)
-
 /* Crashlog Discovery Header */
 #define CONTROL_OFFSET		0x0
 #define GUID_OFFSET		0x4
@@ -50,10 +35,63 @@
 /* size is in bytes */
 #define GET_SIZE(v)		((v) * sizeof(u32))
 
+/*
+ * Type 1 Version 0
+ * status and control registers are combined.
+ *
+ * Bits 29 and 30 control the state of bit 31.
+ * Bit 29 will clear bit 31, if set, allowing a new crashlog to be captured.
+ * Bit 30 will immediately trigger a crashlog to be generated, setting bit 31.
+ * Bit 31 is the read-only status with a 1 indicating log is complete.
+ */
+#define TYPE1_VER0_STATUS_OFFSET	0x00
+#define TYPE1_VER0_CONTROL_OFFSET	0x00
+
+#define TYPE1_VER0_DISABLE		BIT(28)
+#define TYPE1_VER0_CLEAR		BIT(29)
+#define TYPE1_VER0_EXECUTE		BIT(30)
+#define TYPE1_VER0_COMPLETE		BIT(31)
+#define TYPE1_VER0_TRIGGER_MASK		GENMASK(31, 28)
+
+/* After offset, order alphabetically, not bit ordered */
+struct crashlog_status {
+	u32 offset;
+	u32 cleared;
+	u32 complete;
+	u32 disabled;
+};
+
+struct crashlog_control {
+	u32 offset;
+	u32 trigger_mask;
+	u32 clear;
+	u32 disable;
+	u32 manual;
+};
+
+struct crashlog_info {
+	struct crashlog_status status;
+	struct crashlog_control control;
+};
+
+static const struct crashlog_info crashlog_type1_ver0 = {
+	.status.offset = TYPE1_VER0_STATUS_OFFSET,
+	.status.cleared = TYPE1_VER0_CLEAR,
+	.status.complete = TYPE1_VER0_COMPLETE,
+	.status.disabled = TYPE1_VER0_DISABLE,
+
+	.control.offset = TYPE1_VER0_CONTROL_OFFSET,
+	.control.trigger_mask = TYPE1_VER0_TRIGGER_MASK,
+	.control.clear = TYPE1_VER0_CLEAR,
+	.control.disable = TYPE1_VER0_DISABLE,
+	.control.manual = TYPE1_VER0_EXECUTE,
+};
+
 struct crashlog_entry {
 	/* entry must be first member of struct */
 	struct intel_pmt_entry		entry;
 	struct mutex			control_mutex;
+	const struct crashlog_info	*info;
 };
 
 struct pmt_crashlog_priv {
@@ -70,24 +108,25 @@ struct pmt_crashlog_priv {
 /* read/modify/write */
 static void pmt_crashlog_rmw(struct crashlog_entry *crashlog, u32 bit, bool set)
 {
+	const struct crashlog_control *control = &crashlog->info->control;
 	struct intel_pmt_entry *entry = &crashlog->entry;
-	u32 reg = readl(entry->disc_table + CONTROL_OFFSET);
+	u32 reg = readl(entry->disc_table + control->offset);
 
-	reg &= ~CRASHLOG_FLAG_TRIGGER_MASK;
+	reg &= ~control->trigger_mask;
 
 	if (set)
 		reg |= bit;
 	else
 		reg &= bit;
 
-	writel(reg, entry->disc_table + CONTROL_OFFSET);
+	writel(reg, entry->disc_table + control->offset);
 }
 
 /* read/check */
 static bool pmt_crashlog_rc(struct crashlog_entry *crashlog, u32 bit)
 {
-	struct intel_pmt_entry *entry = &crashlog->entry;
-	u32 reg = readl(entry->disc_table + CONTROL_OFFSET);
+	const struct crashlog_status *status = &crashlog->info->status;
+	u32 reg = readl(crashlog->entry.disc_table + status->offset);
 
 	return !!(reg & bit);
 }
@@ -95,13 +134,13 @@ static bool pmt_crashlog_rc(struct crashlog_entry *crashlog, u32 bit)
 static bool pmt_crashlog_complete(struct crashlog_entry *crashlog)
 {
 	/* return current value of the crashlog complete flag */
-	return pmt_crashlog_rc(crashlog, CRASHLOG_FLAG_TRIGGER_COMPLETE);
+	return pmt_crashlog_rc(crashlog, crashlog->info->status.complete);
 }
 
 static bool pmt_crashlog_disabled(struct crashlog_entry *crashlog)
 {
 	/* return current value of the crashlog disabled flag */
-	return pmt_crashlog_rc(crashlog, CRASHLOG_FLAG_DISABLE);
+	return pmt_crashlog_rc(crashlog, crashlog->info->status.disabled);
 }
 
 static bool pmt_crashlog_supported(struct intel_pmt_entry *entry)
@@ -121,17 +160,17 @@ static bool pmt_crashlog_supported(struct intel_pmt_entry *entry)
 
 static void pmt_crashlog_set_disable(struct crashlog_entry *crashlog, bool disable)
 {
-	pmt_crashlog_rmw(crashlog, CRASHLOG_FLAG_DISABLE, disable);
+	pmt_crashlog_rmw(crashlog, crashlog->info->control.disable, disable);
 }
 
 static void pmt_crashlog_set_clear(struct crashlog_entry *crashlog)
 {
-	pmt_crashlog_rmw(crashlog, CRASHLOG_FLAG_TRIGGER_CLEAR, CRASHLOG_SET_BIT);
+	pmt_crashlog_rmw(crashlog, crashlog->info->control.clear, CRASHLOG_SET_BIT);
 }
 
 static void pmt_crashlog_set_execute(struct crashlog_entry *crashlog)
 {
-	pmt_crashlog_rmw(crashlog, CRASHLOG_FLAG_TRIGGER_EXECUTE, CRASHLOG_SET_BIT);
+	pmt_crashlog_rmw(crashlog, crashlog->info->control.manual, CRASHLOG_SET_BIT);
 }
 
 /*
@@ -235,9 +274,10 @@ static int pmt_crashlog_header_decode(struct intel_pmt_entry *entry,
 	if (!pmt_crashlog_supported(entry))
 		return 1;
 
-	/* initialize control mutex */
+	/* initialize the crashlog struct */
 	crashlog = container_of(entry, struct crashlog_entry, entry);
 	mutex_init(&crashlog->control_mutex);
+	crashlog->info = &crashlog_type1_ver0;
 
 	header->access_type = GET_ACCESS(readl(disc_table));
 	header->guid = readl(disc_table + GUID_OFFSET);
-- 
2.49.0


^ permalink raw reply related	[flat|nested] 18+ messages in thread

* [PATCH v5 12/12] platform/x86/intel/pmt: support BMG crashlog
  2025-06-27 20:43 [PATCH v5 00/12] Crashlog Type1 Version2 support Michael J. Ruhl
                   ` (10 preceding siblings ...)
  2025-06-27 20:43 ` [PATCH v5 11/12] platform/x86/intel/pmt: use a version struct Michael J. Ruhl
@ 2025-06-27 20:43 ` Michael J. Ruhl
  11 siblings, 0 replies; 18+ messages in thread
From: Michael J. Ruhl @ 2025-06-27 20:43 UTC (permalink / raw)
  To: platform-driver-x86, intel-xe, hdegoede, ilpo.jarvinen,
	lucas.demarchi, rodrigo.vivi, thomas.hellstrom, airlied, simona,
	david.e.box
  Cc: Michael J. Ruhl

The Battlemage GPU has the type 1 version 2 crashlog feature.

Update the crashlog driver to support this crashlog version.

Signed-off-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
---
 drivers/platform/x86/intel/pmt/crashlog.c | 268 ++++++++++++++++++++--
 1 file changed, 255 insertions(+), 13 deletions(-)

diff --git a/drivers/platform/x86/intel/pmt/crashlog.c b/drivers/platform/x86/intel/pmt/crashlog.c
index d109f307f110..6dfd6ce07599 100644
--- a/drivers/platform/x86/intel/pmt/crashlog.c
+++ b/drivers/platform/x86/intel/pmt/crashlog.c
@@ -53,20 +53,52 @@
 #define TYPE1_VER0_COMPLETE		BIT(31)
 #define TYPE1_VER0_TRIGGER_MASK		GENMASK(31, 28)
 
+/*
+ * Type 1 Version 2
+ * status and control are two different registers
+ */
+#define TYPE1_VER2_STATUS_OFFSET	0x00
+#define TYPE1_VER2_CONTROL_OFFSET	0x14
+
+/* status register */
+#define TYPE1_VER2_CLEAR_SUPPORT	BIT(20)
+#define TYPE1_VER2_REARMED		BIT(25)
+#define TYPE1_VER2_ERROR		BIT(26)
+#define TYPE1_VER2_CONSUMED		BIT(27)
+#define TYPE1_VER2_DISABLED		BIT(28)
+#define TYPE1_VER2_CLEARED		BIT(29)
+#define TYPE1_VER2_IN_PROGRESS		BIT(30)
+#define TYPE1_VER2_COMPLETE		BIT(31)
+
+/* control register */
+#define TYPE1_VER2_CONSUME		BIT(25)
+#define TYPE1_VER2_REARM		BIT(28)
+#define TYPE1_VER2_EXECUTE		BIT(29)
+#define TYPE1_VER2_CLEAR		BIT(30)
+#define TYPE1_VER2_DISABLE		BIT(31)
+#define TYPE1_VER2_TRIGGER_MASK		(TYPE1_VER2_EXECUTE | TYPE1_VER2_CLEAR | TYPE1_VER2_DISABLE)
+
 /* After offset, order alphabetically, not bit ordered */
 struct crashlog_status {
 	u32 offset;
+	u32 clear_supported;
 	u32 cleared;
 	u32 complete;
+	u32 consumed;
 	u32 disabled;
+	u32 error;
+	u32 in_progress;
+	u32 rearmed;
 };
 
 struct crashlog_control {
 	u32 offset;
 	u32 trigger_mask;
 	u32 clear;
+	u32 consume;
 	u32 disable;
 	u32 manual;
+	u32 rearm;
 };
 
 struct crashlog_info {
@@ -87,6 +119,26 @@ static const struct crashlog_info crashlog_type1_ver0 = {
 	.control.manual = TYPE1_VER0_EXECUTE,
 };
 
+const struct crashlog_info crashlog_type1_ver2 = {
+	.status.offset = TYPE1_VER2_STATUS_OFFSET,
+	.status.clear_supported = TYPE1_VER2_CLEAR_SUPPORT,
+	.status.cleared = TYPE1_VER2_CLEARED,
+	.status.complete = TYPE1_VER2_COMPLETE,
+	.status.consumed = TYPE1_VER2_CONSUMED,
+	.status.disabled = TYPE1_VER2_DISABLED,
+	.status.error = TYPE1_VER2_ERROR,
+	.status.in_progress = TYPE1_VER2_IN_PROGRESS,
+	.status.rearmed = TYPE1_VER2_REARMED,
+
+	.control.offset = TYPE1_VER2_CONTROL_OFFSET,
+	.control.trigger_mask = TYPE1_VER2_TRIGGER_MASK,
+	.control.clear = TYPE1_VER2_CLEAR,
+	.control.consume = TYPE1_VER2_CONSUME,
+	.control.disable = TYPE1_VER2_DISABLE,
+	.control.manual = TYPE1_VER2_EXECUTE,
+	.control.rearm = TYPE1_VER2_REARM,
+};
+
 struct crashlog_entry {
 	/* entry must be first member of struct */
 	struct intel_pmt_entry		entry;
@@ -143,19 +195,23 @@ static bool pmt_crashlog_disabled(struct crashlog_entry *crashlog)
 	return pmt_crashlog_rc(crashlog, crashlog->info->status.disabled);
 }
 
-static bool pmt_crashlog_supported(struct intel_pmt_entry *entry)
+static bool pmt_crashlog_supported(struct intel_pmt_entry *entry, u32 *crash_type, u32 *version)
 {
 	u32 discovery_header = readl(entry->disc_table + CONTROL_OFFSET);
-	u32 crash_type, version;
 
-	crash_type = GET_TYPE(discovery_header);
-	version = GET_VERSION(discovery_header);
+	*crash_type = GET_TYPE(discovery_header);
+	*version = GET_VERSION(discovery_header);
 
 	/*
-	 * Currently we only recognize OOBMSM version 0 devices.
-	 * We can ignore all other crashlog devices in the system.
+	 * Currently we only recognize OOBMSM (type 1) and version 0 or 2
+	 * devices.
+	 *
+	 * Ignore all other crashlog devices in the system.
 	 */
-	return crash_type == CRASH_TYPE_OOBMSM && version == 0;
+	if (*crash_type == CRASH_TYPE_OOBMSM && (*version == 0 || *version == 2))
+		return true;
+
+	return false;
 }
 
 static void pmt_crashlog_set_disable(struct crashlog_entry *crashlog, bool disable)
@@ -173,9 +229,118 @@ static void pmt_crashlog_set_execute(struct crashlog_entry *crashlog)
 	pmt_crashlog_rmw(crashlog, crashlog->info->control.manual, CRASHLOG_SET_BIT);
 }
 
+static bool pmt_crashlog_cleared(struct crashlog_entry *crashlog)
+{
+	/* return current value of the crashlog cleared flag */
+	return pmt_crashlog_rc(crashlog, crashlog->info->status.cleared);
+}
+
+static bool pmt_crashlog_consumed(struct crashlog_entry *crashlog)
+{
+	/* return current value of the crashlog consumedflag */
+	return pmt_crashlog_rc(crashlog, crashlog->info->status.consumed);
+}
+
+static void pmt_crashlog_set_consumed(struct crashlog_entry *crashlog)
+{
+	pmt_crashlog_rmw(crashlog, crashlog->info->control.consume, CRASHLOG_SET_BIT);
+}
+
+static bool pmt_crashlog_error(struct crashlog_entry *crashlog)
+{
+	/* return current value of the crashlog error flag */
+	return pmt_crashlog_rc(crashlog, crashlog->info->status.error);
+}
+
+static bool pmt_crashlog_rearm(struct crashlog_entry *crashlog)
+{
+	/* return current value of the crashlog reamed flag */
+	return pmt_crashlog_rc(crashlog, crashlog->info->status.rearmed);
+}
+
+static void pmt_crashlog_set_rearm(struct crashlog_entry *crashlog)
+{
+	pmt_crashlog_rmw(crashlog, crashlog->info->control.rearm, CRASHLOG_SET_BIT);
+}
+
 /*
  * sysfs
  */
+static ssize_t
+clear_show(struct device *dev, struct device_attribute *attr, char *buf)
+{
+	struct crashlog_entry *crashlog = dev_get_drvdata(dev);
+	bool cleared = pmt_crashlog_cleared(crashlog);
+
+	return sysfs_emit(buf, "%d\n", cleared);
+}
+
+static ssize_t
+clear_store(struct device *dev, struct device_attribute *attr,
+	    const char *buf, size_t count)
+{
+	struct crashlog_entry *crashlog;
+	bool clear;
+	int result;
+
+	crashlog = dev_get_drvdata(dev);
+
+	result = kstrtobool(buf, &clear);
+	if (result)
+		return result;
+
+	/* set bit only */
+	if (!clear)
+		return -EINVAL;
+
+	guard(mutex)(&crashlog->control_mutex);
+
+	pmt_crashlog_set_clear(crashlog);
+
+	return count;
+}
+static DEVICE_ATTR_RW(clear);
+
+static ssize_t
+consumed_show(struct device *dev, struct device_attribute *attr, char *buf)
+{
+	struct crashlog_entry *crashlog = dev_get_drvdata(dev);
+	bool consumed = pmt_crashlog_consumed(crashlog);
+
+	return sysfs_emit(buf, "%d\n", consumed);
+}
+
+static ssize_t consumed_store(struct device *dev, struct device_attribute *attr,
+			      const char *buf, size_t count)
+{
+	struct crashlog_entry *crashlog;
+	bool consumed;
+	int result;
+
+	crashlog = dev_get_drvdata(dev);
+
+	result = kstrtobool(buf, &consumed);
+	if (result)
+		return result;
+
+	/* set bit only */
+	if (!consumed)
+		return -EINVAL;
+
+	guard(mutex)(&crashlog->control_mutex);
+
+	if (pmt_crashlog_disabled(crashlog))
+		return -EBUSY;
+
+	if (!pmt_crashlog_complete(crashlog))
+		return -EEXIST;
+
+	pmt_crashlog_set_consumed(crashlog);
+
+	return count;
+}
+static DEVICE_ATTR_RW(consumed);
+
 static ssize_t
 enable_show(struct device *dev, struct device_attribute *attr, char *buf)
 {
@@ -207,6 +372,50 @@ enable_store(struct device *dev, struct device_attribute *attr,
 }
 static DEVICE_ATTR_RW(enable);
 
+static ssize_t
+error_show(struct device *dev, struct device_attribute *attr, char *buf)
+{
+	struct crashlog_entry *crashlog = dev_get_drvdata(dev);
+	bool error = pmt_crashlog_error(crashlog);
+
+	return sysfs_emit(buf, "%d\n", error);
+}
+static DEVICE_ATTR_RO(error);
+
+static ssize_t
+rearm_show(struct device *dev, struct device_attribute *attr, char *buf)
+{
+	struct crashlog_entry *crashlog = dev_get_drvdata(dev);
+	int rearmed = pmt_crashlog_rearm(crashlog);
+
+	return sysfs_emit(buf, "%d\n", rearmed);
+}
+
+static ssize_t rearm_store(struct device *dev, struct device_attribute *attr,
+			   const char *buf, size_t count)
+{
+	struct crashlog_entry *crashlog;
+	bool rearm;
+	int result;
+
+	crashlog = dev_get_drvdata(dev);
+
+	result = kstrtobool(buf, &rearm);
+	if (result)
+		return result;
+
+	/* set only */
+	if (!rearm)
+		return -EINVAL;
+
+	guard(mutex)(&crashlog->control_mutex);
+
+	pmt_crashlog_set_rearm(crashlog);
+
+	return count;
+}
+static DEVICE_ATTR_RW(rearm);
+
 static ssize_t
 trigger_show(struct device *dev, struct device_attribute *attr, char *buf)
 {
@@ -254,30 +463,63 @@ trigger_store(struct device *dev, struct device_attribute *attr,
 }
 static DEVICE_ATTR_RW(trigger);
 
-static struct attribute *pmt_crashlog_attrs[] = {
+static struct attribute *pmt_crashlog_type1_ver0_attrs[] = {
 	&dev_attr_enable.attr,
 	&dev_attr_trigger.attr,
 	NULL
 };
 
-static const struct attribute_group pmt_crashlog_group = {
-	.attrs	= pmt_crashlog_attrs,
+static struct attribute *pmt_crashlog_type1_ver2_attrs[] = {
+	&dev_attr_clear.attr,
+	&dev_attr_consumed.attr,
+	&dev_attr_enable.attr,
+	&dev_attr_error.attr,
+	&dev_attr_rearm.attr,
+	&dev_attr_trigger.attr,
+	NULL
+};
+
+static const struct attribute_group pmt_crashlog_type1_ver0_group = {
+	.attrs	= pmt_crashlog_type1_ver0_attrs,
 };
 
+static const struct attribute_group pmt_crashlog_type1_ver2_group = {
+	.attrs = pmt_crashlog_type1_ver2_attrs,
+};
+
+static const struct crashlog_info *select_crashlog_info(u32 type, u32 version)
+{
+	if (version == 0)
+		return &crashlog_type1_ver0;
+
+	return &crashlog_type1_ver2;
+}
+
+static const struct attribute_group *select_sysfs_grp(u32 type, u32 version)
+{
+	if (version == 0)
+		return &pmt_crashlog_type1_ver2_group;
+
+	return &pmt_crashlog_type1_ver2_group;
+}
+
 static int pmt_crashlog_header_decode(struct intel_pmt_entry *entry,
 				      struct device *dev)
 {
 	void __iomem *disc_table = entry->disc_table;
 	struct intel_pmt_header *header = &entry->header;
 	struct crashlog_entry *crashlog;
+	u32 version;
+	u32 type;
 
-	if (!pmt_crashlog_supported(entry))
+	if (!pmt_crashlog_supported(entry, &type, &version))
 		return 1;
 
 	/* initialize the crashlog struct */
 	crashlog = container_of(entry, struct crashlog_entry, entry);
 	mutex_init(&crashlog->control_mutex);
-	crashlog->info = &crashlog_type1_ver0;
+
+	crashlog->info = select_crashlog_info(type, version);
 
 	header->access_type = GET_ACCESS(readl(disc_table));
 	header->guid = readl(disc_table + GUID_OFFSET);
@@ -286,7 +528,7 @@ static int pmt_crashlog_header_decode(struct intel_pmt_entry *entry,
 	/* Size is measured in DWORDS, but accessor returns bytes */
 	header->size = GET_SIZE(readl(disc_table + SIZE_OFFSET));
 
-	entry->attr_grp = &pmt_crashlog_group;
+	entry->attr_grp = select_sysfs_grp(type, version);
 
 	return 0;
 }
-- 
2.49.0


^ permalink raw reply related	[flat|nested] 18+ messages in thread

* Re: [PATCH v5 01/12] platform/x86/intel/pmt: fix a crashlog NULL pointer access
  2025-06-27 20:43 ` [PATCH v5 01/12] platform/x86/intel/pmt: fix a crashlog NULL pointer access Michael J. Ruhl
@ 2025-06-30  9:29   ` Ilpo Järvinen
  2025-06-30 15:00     ` Ruhl, Michael J
  0 siblings, 1 reply; 18+ messages in thread
From: Ilpo Järvinen @ 2025-06-30  9:29 UTC (permalink / raw)
  To: Michael J. Ruhl
  Cc: platform-driver-x86, intel-xe, Hans de Goede, lucas.demarchi,
	rodrigo.vivi, thomas.hellstrom, airlied, simona, david.e.box,
	Tejas Upadhyay, stable

On Fri, 27 Jun 2025, Michael J. Ruhl wrote:

> Usage of the intel_pmt_read() for binary sysfs, requires a pcidev. The
> current use of the endpoint value is only valid for telemetry endpoint
> usage.
> 
> Without the ep, the crashlog usage causes the following NULL pointer
> exception:
> 
> BUG: kernel NULL pointer dereference, address: 0000000000000000
> Oops: Oops: 0000 [#1] SMP NOPTI
> RIP: 0010:intel_pmt_read+0x3b/0x70 [pmt_class]
> Code:
> Call Trace:
>  <TASK>
>  ? sysfs_kf_bin_read+0xc0/0xe0
>  kernfs_fop_read_iter+0xac/0x1a0
>  vfs_read+0x26d/0x350
>  ksys_read+0x6b/0xe0
>  __x64_sys_read+0x1d/0x30
>  x64_sys_call+0x1bc8/0x1d70
>  do_syscall_64+0x6d/0x110

Can you confirm, if this was possible to trigger only after this series 
has been applied, not with the current mainline code?

-- 
 i.

> Augment struct intel_pmt_entry with a pointer to the pcidev to avoid
> the NULL pointer exception.
> 
> Reviewed-by: Tejas Upadhyay <tejas.upadhyay@intel.com>
> Fixes: 416eeb2e1fc7 ("platform/x86/intel/pmt: telemetry: Export API to read telemetry")
> Cc: <stable@vger.kernel.org>
> Signed-off-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
> ---
>  drivers/platform/x86/intel/pmt/class.c | 3 ++-
>  drivers/platform/x86/intel/pmt/class.h | 1 +
>  2 files changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/platform/x86/intel/pmt/class.c b/drivers/platform/x86/intel/pmt/class.c
> index 7233b654bbad..d046e8752173 100644
> --- a/drivers/platform/x86/intel/pmt/class.c
> +++ b/drivers/platform/x86/intel/pmt/class.c
> @@ -97,7 +97,7 @@ intel_pmt_read(struct file *filp, struct kobject *kobj,
>  	if (count > entry->size - off)
>  		count = entry->size - off;
>  
> -	count = pmt_telem_read_mmio(entry->ep->pcidev, entry->cb, entry->header.guid, buf,
> +	count = pmt_telem_read_mmio(entry->pcidev, entry->cb, entry->header.guid, buf,
>  				    entry->base, off, count);
>  
>  	return count;
> @@ -252,6 +252,7 @@ static int intel_pmt_populate_entry(struct intel_pmt_entry *entry,
>  		return -EINVAL;
>  	}
>  
> +	entry->pcidev = pci_dev;
>  	entry->guid = header->guid;
>  	entry->size = header->size;
>  	entry->cb = ivdev->priv_data;
> diff --git a/drivers/platform/x86/intel/pmt/class.h b/drivers/platform/x86/intel/pmt/class.h
> index b2006d57779d..f6ce80c4e051 100644
> --- a/drivers/platform/x86/intel/pmt/class.h
> +++ b/drivers/platform/x86/intel/pmt/class.h
> @@ -39,6 +39,7 @@ struct intel_pmt_header {
>  
>  struct intel_pmt_entry {
>  	struct telem_endpoint	*ep;
> +	struct pci_dev		*pcidev;
>  	struct intel_pmt_header	header;
>  	struct bin_attribute	pmt_bin_attr;
>  	struct kobject		*kobj;
> 

^ permalink raw reply	[flat|nested] 18+ messages in thread

* RE: [PATCH v5 01/12] platform/x86/intel/pmt: fix a crashlog NULL pointer access
  2025-06-30  9:29   ` Ilpo Järvinen
@ 2025-06-30 15:00     ` Ruhl, Michael J
  0 siblings, 0 replies; 18+ messages in thread
From: Ruhl, Michael J @ 2025-06-30 15:00 UTC (permalink / raw)
  To: Ilpo Järvinen
  Cc: platform-driver-x86@vger.kernel.org,
	intel-xe@lists.freedesktop.org, Hans de Goede, De Marchi, Lucas,
	Vivi, Rodrigo, thomas.hellstrom@linux.intel.com,
	airlied@gmail.com, simona@ffwll.ch, david.e.box@linux.intel.com,
	Upadhyay, Tejas, stable@vger.kernel.org

>-----Original Message-----
>From: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
>Sent: Monday, June 30, 2025 5:29 AM
>To: Ruhl, Michael J <michael.j.ruhl@intel.com>
>Cc: platform-driver-x86@vger.kernel.org; intel-xe@lists.freedesktop.org; Hans
>de Goede <hdegoede@redhat.com>; De Marchi, Lucas
><lucas.demarchi@intel.com>; Vivi, Rodrigo <rodrigo.vivi@intel.com>;
>thomas.hellstrom@linux.intel.com; airlied@gmail.com; simona@ffwll.ch;
>david.e.box@linux.intel.com; Upadhyay, Tejas <tejas.upadhyay@intel.com>;
>stable@vger.kernel.org
>Subject: Re: [PATCH v5 01/12] platform/x86/intel/pmt: fix a crashlog NULL
>pointer access
>
>On Fri, 27 Jun 2025, Michael J. Ruhl wrote:
>
>> Usage of the intel_pmt_read() for binary sysfs, requires a pcidev. The
>> current use of the endpoint value is only valid for telemetry endpoint
>> usage.
>>
>> Without the ep, the crashlog usage causes the following NULL pointer
>> exception:
>>
>> BUG: kernel NULL pointer dereference, address: 0000000000000000
>> Oops: Oops: 0000 [#1] SMP NOPTI
>> RIP: 0010:intel_pmt_read+0x3b/0x70 [pmt_class]
>> Code:
>> Call Trace:
>>  <TASK>
>>  ? sysfs_kf_bin_read+0xc0/0xe0
>>  kernfs_fop_read_iter+0xac/0x1a0
>>  vfs_read+0x26d/0x350
>>  ksys_read+0x6b/0xe0
>>  __x64_sys_read+0x1d/0x30
>>  x64_sys_call+0x1bc8/0x1d70
>>  do_syscall_64+0x6d/0x110
>
>Can you confirm, if this was possible to trigger only after this series
>has been applied, not with the current mainline code?

Hi Ilpo,

I somehow got the wrong fixes patch.  It should be:

Fixes: 045a513040cc ("platform/x86/intel/pmt: Use PMT callbacks")

The issue occurs before my patches are applied.

Does this answer your question?

M

>--
> i.
>
>> Augment struct intel_pmt_entry with a pointer to the pcidev to avoid
>> the NULL pointer exception.
>>
>> Reviewed-by: Tejas Upadhyay <tejas.upadhyay@intel.com>
>> Fixes: 416eeb2e1fc7 ("platform/x86/intel/pmt: telemetry: Export API to read
>telemetry")
>> Cc: <stable@vger.kernel.org>
>> Signed-off-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
>> ---
>>  drivers/platform/x86/intel/pmt/class.c | 3 ++-
>>  drivers/platform/x86/intel/pmt/class.h | 1 +
>>  2 files changed, 3 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/platform/x86/intel/pmt/class.c
>b/drivers/platform/x86/intel/pmt/class.c
>> index 7233b654bbad..d046e8752173 100644
>> --- a/drivers/platform/x86/intel/pmt/class.c
>> +++ b/drivers/platform/x86/intel/pmt/class.c
>> @@ -97,7 +97,7 @@ intel_pmt_read(struct file *filp, struct kobject *kobj,
>>  	if (count > entry->size - off)
>>  		count = entry->size - off;
>>
>> -	count = pmt_telem_read_mmio(entry->ep->pcidev, entry->cb, entry-
>>header.guid, buf,
>> +	count = pmt_telem_read_mmio(entry->pcidev, entry->cb, entry-
>>header.guid, buf,
>>  				    entry->base, off, count);
>>
>>  	return count;
>> @@ -252,6 +252,7 @@ static int intel_pmt_populate_entry(struct
>intel_pmt_entry *entry,
>>  		return -EINVAL;
>>  	}
>>
>> +	entry->pcidev = pci_dev;
>>  	entry->guid = header->guid;
>>  	entry->size = header->size;
>>  	entry->cb = ivdev->priv_data;
>> diff --git a/drivers/platform/x86/intel/pmt/class.h
>b/drivers/platform/x86/intel/pmt/class.h
>> index b2006d57779d..f6ce80c4e051 100644
>> --- a/drivers/platform/x86/intel/pmt/class.h
>> +++ b/drivers/platform/x86/intel/pmt/class.h
>> @@ -39,6 +39,7 @@ struct intel_pmt_header {
>>
>>  struct intel_pmt_entry {
>>  	struct telem_endpoint	*ep;
>> +	struct pci_dev		*pcidev;
>>  	struct intel_pmt_header	header;
>>  	struct bin_attribute	pmt_bin_attr;
>>  	struct kobject		*kobj;
>>

^ permalink raw reply	[flat|nested] 18+ messages in thread

* Re: [PATCH v5 02/12] drm/xe: Correct BMG VSEC header sizing
  2025-06-27 20:43 ` [PATCH v5 02/12] drm/xe: Correct BMG VSEC header sizing Michael J. Ruhl
@ 2025-06-30 19:57   ` Rodrigo Vivi
  2025-06-30 22:00     ` Ruhl, Michael J
  0 siblings, 1 reply; 18+ messages in thread
From: Rodrigo Vivi @ 2025-06-30 19:57 UTC (permalink / raw)
  To: Michael J. Ruhl
  Cc: platform-driver-x86, intel-xe, hdegoede, ilpo.jarvinen,
	lucas.demarchi, thomas.hellstrom, airlied, simona, david.e.box

On Fri, Jun 27, 2025 at 04:43:11PM -0400, Michael J. Ruhl wrote:
> The intel_vsec_header information for the crashlog feature is
> incorrect.
> 
> Update the VSEC header with correct sizing and count.
> 
> Since the crashlog entries are "merged" (num_entries = 2), the
> separate capabilities entries must be merged as well.

if you share some doc I might be able to help with reviews on this.
But for now,

Acked-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

to get this Xe patch merged with the rest of the series in the
PMT subsystem trees.

> 
> Fixes: 0c45e76fcc62 ("drm/xe/vsec: Support BMG devices")
> Signed-off-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
> ---
>  drivers/gpu/drm/xe/xe_vsec.c | 20 +++++---------------
>  1 file changed, 5 insertions(+), 15 deletions(-)
> 
> diff --git a/drivers/gpu/drm/xe/xe_vsec.c b/drivers/gpu/drm/xe/xe_vsec.c
> index 3e573b0b7ebd..67238fc57a4d 100644
> --- a/drivers/gpu/drm/xe/xe_vsec.c
> +++ b/drivers/gpu/drm/xe/xe_vsec.c
> @@ -32,28 +32,18 @@ static struct intel_vsec_header bmg_telemetry = {
>  	.offset = BMG_DISCOVERY_OFFSET,
>  };
>  
> -static struct intel_vsec_header bmg_punit_crashlog = {
> -	.length = 0x10,
> +static struct intel_vsec_header bmg_crashlog = {
> +	.length = 0x18,
>  	.id = VSEC_ID_CRASHLOG,
> -	.num_entries = 1,
> -	.entry_size = 4,
> +	.num_entries = 2,
> +	.entry_size = 6,
>  	.tbir = 0,
>  	.offset = BMG_DISCOVERY_OFFSET + 0x60,
>  };
>  
> -static struct intel_vsec_header bmg_oobmsm_crashlog = {
> -	.length = 0x10,
> -	.id = VSEC_ID_CRASHLOG,
> -	.num_entries = 1,
> -	.entry_size = 4,
> -	.tbir = 0,
> -	.offset = BMG_DISCOVERY_OFFSET + 0x78,
> -};
> -
>  static struct intel_vsec_header *bmg_capabilities[] = {
>  	&bmg_telemetry,
> -	&bmg_punit_crashlog,
> -	&bmg_oobmsm_crashlog,
> +	&bmg_crashlog,
>  	NULL
>  };
>  
> -- 
> 2.49.0
> 

^ permalink raw reply	[flat|nested] 18+ messages in thread

* RE: [PATCH v5 02/12] drm/xe: Correct BMG VSEC header sizing
  2025-06-30 19:57   ` Rodrigo Vivi
@ 2025-06-30 22:00     ` Ruhl, Michael J
  0 siblings, 0 replies; 18+ messages in thread
From: Ruhl, Michael J @ 2025-06-30 22:00 UTC (permalink / raw)
  To: Vivi, Rodrigo
  Cc: platform-driver-x86@vger.kernel.org,
	intel-xe@lists.freedesktop.org, hdegoede@redhat.com,
	ilpo.jarvinen@linux.intel.com, De Marchi, Lucas,
	thomas.hellstrom@linux.intel.com, airlied@gmail.com,
	simona@ffwll.ch, david.e.box@linux.intel.com

>-----Original Message-----
>From: Vivi, Rodrigo <rodrigo.vivi@intel.com>
>Sent: Monday, June 30, 2025 3:58 PM
>To: Ruhl, Michael J <michael.j.ruhl@intel.com>
>Cc: platform-driver-x86@vger.kernel.org; intel-xe@lists.freedesktop.org;
>hdegoede@redhat.com; ilpo.jarvinen@linux.intel.com; De Marchi, Lucas
><lucas.demarchi@intel.com>; thomas.hellstrom@linux.intel.com;
>airlied@gmail.com; simona@ffwll.ch; david.e.box@linux.intel.com
>Subject: Re: [PATCH v5 02/12] drm/xe: Correct BMG VSEC header sizing
>
>On Fri, Jun 27, 2025 at 04:43:11PM -0400, Michael J. Ruhl wrote:
>> The intel_vsec_header information for the crashlog feature is
>> incorrect.
>>
>> Update the VSEC header with correct sizing and count.
>>
>> Since the crashlog entries are "merged" (num_entries = 2), the
>> separate capabilities entries must be merged as well.
>
>if you share some doc I might be able to help with reviews on this.
>But for now,
>
>Acked-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
>
>to get this Xe patch merged with the rest of the series in the
>PMT subsystem trees.

Hi Rodrigo,

The data structure is matching the data structure defined by section 7.3.5 of this doc:

https://www.intel.com/content/www/us/en/content-details/710389/intel-platform-monitoring-technology-intel-pmt-technical-specification.html

Originally I was thinking that the data structure was just the basic PMT discovery struct (Section 7.1.1),
a misunderstanding on my part, but this is the Crashlog discovery/control structure.

The two entries are consecutive, so I can say 2 entries and size them correctly.

Thanks!

Mike

>>
>> Fixes: 0c45e76fcc62 ("drm/xe/vsec: Support BMG devices")
>> Signed-off-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
>> ---
>>  drivers/gpu/drm/xe/xe_vsec.c | 20 +++++---------------
>>  1 file changed, 5 insertions(+), 15 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/xe/xe_vsec.c b/drivers/gpu/drm/xe/xe_vsec.c
>> index 3e573b0b7ebd..67238fc57a4d 100644
>> --- a/drivers/gpu/drm/xe/xe_vsec.c
>> +++ b/drivers/gpu/drm/xe/xe_vsec.c
>> @@ -32,28 +32,18 @@ static struct intel_vsec_header bmg_telemetry = {
>>  	.offset = BMG_DISCOVERY_OFFSET,
>>  };
>>
>> -static struct intel_vsec_header bmg_punit_crashlog = {
>> -	.length = 0x10,
>> +static struct intel_vsec_header bmg_crashlog = {
>> +	.length = 0x18,
>>  	.id = VSEC_ID_CRASHLOG,
>> -	.num_entries = 1,
>> -	.entry_size = 4,
>> +	.num_entries = 2,
>> +	.entry_size = 6,
>>  	.tbir = 0,
>>  	.offset = BMG_DISCOVERY_OFFSET + 0x60,
>>  };
>>
>> -static struct intel_vsec_header bmg_oobmsm_crashlog = {
>> -	.length = 0x10,
>> -	.id = VSEC_ID_CRASHLOG,
>> -	.num_entries = 1,
>> -	.entry_size = 4,
>> -	.tbir = 0,
>> -	.offset = BMG_DISCOVERY_OFFSET + 0x78,
>> -};
>> -
>>  static struct intel_vsec_header *bmg_capabilities[] = {
>>  	&bmg_telemetry,
>> -	&bmg_punit_crashlog,
>> -	&bmg_oobmsm_crashlog,
>> +	&bmg_crashlog,
>>  	NULL
>>  };
>>
>> --
>> 2.49.0
>>

^ permalink raw reply	[flat|nested] 18+ messages in thread

* RE: [PATCH v5 09/12] platform/x86/intel/pmt: add register access helpers
  2025-06-27 20:43 ` [PATCH v5 09/12] platform/x86/intel/pmt: add register access helpers Michael J. Ruhl
@ 2025-07-02 15:50   ` Ruhl, Michael J
  0 siblings, 0 replies; 18+ messages in thread
From: Ruhl, Michael J @ 2025-07-02 15:50 UTC (permalink / raw)
  To: platform-driver-x86@vger.kernel.org,
	intel-xe@lists.freedesktop.org, hdegoede@redhat.com,
	ilpo.jarvinen@linux.intel.com, De Marchi, Lucas, Vivi, Rodrigo,
	thomas.hellstrom@linux.intel.com, airlied@gmail.com,
	simona@ffwll.ch, david.e.box@linux.intel.com

>-----Original Message-----
>From: Ruhl, Michael J <michael.j.ruhl@intel.com>
>Sent: Friday, June 27, 2025 4:43 PM
>To: platform-driver-x86@vger.kernel.org; intel-xe@lists.freedesktop.org;
>hdegoede@redhat.com; ilpo.jarvinen@linux.intel.com; De Marchi, Lucas
><lucas.demarchi@intel.com>; Vivi, Rodrigo <rodrigo.vivi@intel.com>;
>thomas.hellstrom@linux.intel.com; airlied@gmail.com; simona@ffwll.ch;
>david.e.box@linux.intel.com
>Cc: Ruhl, Michael J <michael.j.ruhl@intel.com>
>Subject: [PATCH v5 09/12] platform/x86/intel/pmt: add register access helpers
>
>The control register is used in a read/modify/write pattern.
>The status register is used in a read/check bit pattern.
>
>Add helpers to eliminate common code.
>
>Signed-off-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
>---
> drivers/platform/x86/intel/pmt/crashlog.c | 60 ++++++++++++-----------
> 1 file changed, 31 insertions(+), 29 deletions(-)
>
>diff --git a/drivers/platform/x86/intel/pmt/crashlog.c
>b/drivers/platform/x86/intel/pmt/crashlog.c
>index 23b3971da40a..adaca7ce1ba5 100644
>--- a/drivers/platform/x86/intel/pmt/crashlog.c
>+++ b/drivers/platform/x86/intel/pmt/crashlog.c
>@@ -64,20 +64,42 @@ struct pmt_crashlog_priv {
> /*
>  * I/O
>  */
>-static bool pmt_crashlog_complete(struct intel_pmt_entry *entry)
>+#define CRASHLOG_SET_BIT	true
>+#define CRASHLOG_CLEAR_BIT	false
>+
>+/* read/modify/write */
>+static void pmt_crashlog_rmw(struct intel_pmt_entry *entry, u32 bit, bool
>set)
> {
>-	u32 control = readl(entry->disc_table + CONTROL_OFFSET);
>+	u32 reg = readl(entry->disc_table + CONTROL_OFFSET);
>+
>+	reg &= ~CRASHLOG_FLAG_TRIGGER_MASK;
>+
>+	if (set)
>+		reg |= bit;
>+	else
>+		reg &= bit;

This should be:

reg &= ~bit;

(updating patch set).

M

>+
>+	writel(reg, entry->disc_table + CONTROL_OFFSET);
>+}
>+
>+/* read/check */
>+static bool pmt_crashlog_rc(struct intel_pmt_entry *entry, u32 bit)
>+{
>+	u32 reg = readl(entry->disc_table + CONTROL_OFFSET);
>+
>+	return !!(reg & bit);
>+}
>
>+static bool pmt_crashlog_complete(struct intel_pmt_entry *entry)
>+{
> 	/* return current value of the crashlog complete flag */
>-	return !!(control & CRASHLOG_FLAG_TRIGGER_COMPLETE);
>+	return pmt_crashlog_rc(entry,
>CRASHLOG_FLAG_TRIGGER_COMPLETE);
> }
>
> static bool pmt_crashlog_disabled(struct intel_pmt_entry *entry)
> {
>-	u32 control = readl(entry->disc_table + CONTROL_OFFSET);
>-
> 	/* return current value of the crashlog disabled flag */
>-	return !!(control & CRASHLOG_FLAG_DISABLE);
>+	return pmt_crashlog_rc(entry, CRASHLOG_FLAG_DISABLE);
> }
>
> static bool pmt_crashlog_supported(struct intel_pmt_entry *entry)
>@@ -98,37 +120,17 @@ static bool pmt_crashlog_supported(struct
>intel_pmt_entry *entry)
> static void pmt_crashlog_set_disable(struct intel_pmt_entry *entry,
> 				     bool disable)
> {
>-	u32 control = readl(entry->disc_table + CONTROL_OFFSET);
>-
>-	/* clear trigger bits so we are only modifying disable flag */
>-	control &= ~CRASHLOG_FLAG_TRIGGER_MASK;
>-
>-	if (disable)
>-		control |= CRASHLOG_FLAG_DISABLE;
>-	else
>-		control &= ~CRASHLOG_FLAG_DISABLE;
>-
>-	writel(control, entry->disc_table + CONTROL_OFFSET);
>+	pmt_crashlog_rmw(entry, CRASHLOG_FLAG_DISABLE, disable);
> }
>
> static void pmt_crashlog_set_clear(struct intel_pmt_entry *entry)
> {
>-	u32 control = readl(entry->disc_table + CONTROL_OFFSET);
>-
>-	control &= ~CRASHLOG_FLAG_TRIGGER_MASK;
>-	control |= CRASHLOG_FLAG_TRIGGER_CLEAR;
>-
>-	writel(control, entry->disc_table + CONTROL_OFFSET);
>+	pmt_crashlog_rmw(entry, CRASHLOG_FLAG_TRIGGER_CLEAR,
>CRASHLOG_SET_BIT);
> }
>
> static void pmt_crashlog_set_execute(struct intel_pmt_entry *entry)
> {
>-	u32 control = readl(entry->disc_table + CONTROL_OFFSET);
>-
>-	control &= ~CRASHLOG_FLAG_TRIGGER_MASK;
>-	control |= CRASHLOG_FLAG_TRIGGER_EXECUTE;
>-
>-	writel(control, entry->disc_table + CONTROL_OFFSET);
>+	pmt_crashlog_rmw(entry, CRASHLOG_FLAG_TRIGGER_EXECUTE,
>CRASHLOG_SET_BIT);
> }
>
> /*
>--
>2.49.0


^ permalink raw reply	[flat|nested] 18+ messages in thread

end of thread, other threads:[~2025-07-02 15:51 UTC | newest]

Thread overview: 18+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2025-06-27 20:43 [PATCH v5 00/12] Crashlog Type1 Version2 support Michael J. Ruhl
2025-06-27 20:43 ` [PATCH v5 01/12] platform/x86/intel/pmt: fix a crashlog NULL pointer access Michael J. Ruhl
2025-06-30  9:29   ` Ilpo Järvinen
2025-06-30 15:00     ` Ruhl, Michael J
2025-06-27 20:43 ` [PATCH v5 02/12] drm/xe: Correct BMG VSEC header sizing Michael J. Ruhl
2025-06-30 19:57   ` Rodrigo Vivi
2025-06-30 22:00     ` Ruhl, Michael J
2025-06-27 20:43 ` [PATCH v5 03/12] platform/x86/intel/pmt: white space cleanup Michael J. Ruhl
2025-06-27 20:43 ` [PATCH v5 04/12] platform/x86/intel/pmt: mutex clean up Michael J. Ruhl
2025-06-27 20:43 ` [PATCH v5 05/12] platform/x86/intel/pmt: use guard(mutex) Michael J. Ruhl
2025-06-27 20:43 ` [PATCH v5 06/12] platform/x86/intel/pmt: re-order trigger logic Michael J. Ruhl
2025-06-27 20:43 ` [PATCH v5 07/12] platform/x86/intel/pmt: correct types Michael J. Ruhl
2025-06-27 20:43 ` [PATCH v5 08/12] platform/x86/intel/pmt: decouple sysfs and namespace Michael J. Ruhl
2025-06-27 20:43 ` [PATCH v5 09/12] platform/x86/intel/pmt: add register access helpers Michael J. Ruhl
2025-07-02 15:50   ` Ruhl, Michael J
2025-06-27 20:43 ` [PATCH v5 10/12] platform/x86/intel/pmt: refactor base parameter Michael J. Ruhl
2025-06-27 20:43 ` [PATCH v5 11/12] platform/x86/intel/pmt: use a version struct Michael J. Ruhl
2025-06-27 20:43 ` [PATCH v5 12/12] platform/x86/intel/pmt: support BMG crashlog Michael J. Ruhl

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox