Skip to contentSkip to conversation

ACPI: resource: Add quirk for Dell PMC memory region overlap

PATCH In review Manuel Quintero Fonseca 3 participants 6 messages May 11 – May 23 1 message not in the archive ACPI

lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

#1 lore raw

PATCH v2 1/1 ACPI: resource: Add quirk for Dell PMC memory region overlap

On Dell OptiPlex 3080 systems, specifically during the initialization of the intel_pmc_core driver, the kernel reports a resource sanity check conflict. The issue was identified via dmesg logs showing:

resource sanity check: requesting [mem 0xfdffe800-0xfe0007ff], which spans more than PNP0C02:03 [mem 0xfdb00000-0xfdffffff]

And subsequently:

resource sanity check: requesting [mem 0xfdffe800-0xfe0007ff], which spans more than Reserved [mem 0xfe000000-0xfe010fff]

Investigation of /proc/iomem and ACPI tables on this specific model revealed that the BIOS defines reserved regions (as a PNP0C02 resource and an E820 reserved entry) that overlap with the Power Management Controller (PMC) register space.

This overlap is harmless for system stability but produces persistent warnings in dmesg. This patch introduces a quirk to:

1. Filter out the conflicting fixed memory resources in drivers/acpi/resource.c.
2. Skip the iomem sanity check for the specific problematic range in kernel/resource.c.

This ensures a clean boot log on these Dell units and allows the Intel PMC driver to bind correctly without resource conflicts.

---
v2:
- Add DMI matching to restrict the quirk to Dell systems.
- Include <linux/dmi.h> to fix compilation errors.
- Expand commit description with specific hardware (Dell OptiPlex 3080).
- Include dmesg logs showing the resource conflict.
- Reformat description to 75 characters per line.

 drivers/acpi/resource.c | 32 ++++++++++++++++++++++++++++++++
 kernel/resource.c       |  9 +++++++++
 2 files changed, 41 insertions(+)
drivers/acpi/resource.c +32 -0
index bc8050d8a6f5..464195e30080 100644
@@ -50,6 +50,20 @@ static inline bool is_gsi(struct acpi_resource_extended_irq *ext_irq)
 }
 #endif
 
+/*
+ * List of Dell systems requiring memory resource quirks due to BIOS
+ * reservations overlapping with Intel PMC register space.
+ */
+static const struct dmi_system_id dell_memory_quirks[] = {
+	{
+		.ident = "Dell Inc.",
+		.matches = {
+			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
+		},
+	},
+	{ }
+};
+
 static bool acpi_dev_resource_len_valid(u64 start, u64 end, u64 len, bool io)
 {
 	u64 reslen = end - start + 1;
@@ -110,6 +124,24 @@ bool acpi_dev_resource_memory(struct acpi_resource *ares, struct resource *res)
 	struct acpi_resource_memory32 *memory32;
 	struct acpi_resource_fixed_memory32 *fixed_memory32;
 
+	/*
+	 * Quirk for Dell systems: ignore BIOS memory reservations that overlap
+	 * with Intel PMC resources. Restricted via DMI to avoid impact on
+	 * other manufacturers.
+	 */
+	if (dmi_check_system(dell_memory_quirks)) {
+		u32 addr = 0;
+
+		if (ares->type == ACPI_RESOURCE_TYPE_FIXED_MEMORY32)
+			addr = ares->data.fixed_memory32.address;
+		else if (ares->type == ACPI_RESOURCE_TYPE_MEMORY32)
+			addr = ares->data.memory32.minimum;
+
+		if (addr == 0xfdb00000 ||
+		    (addr >= 0xfe000000 && addr <= 0xfe000fff))
+			return false;
+	}
+
 	switch (ares->type) {
 	case ACPI_RESOURCE_TYPE_MEMORY24:
 		memory24 = &ares->data.memory24;
kernel/resource.c +9 -0
index bb966699da31..16c82d9ea14f 100644
@@ -30,6 +30,7 @@
 #include <linux/string.h>
 #include <linux/vmalloc.h>
 #include <asm/io.h>
+#include <linux/dmi.h>
 
 
 struct resource ioport_resource = {
@@ -1857,6 +1858,14 @@ int iomem_map_sanity_check(resource_size_t addr, unsigned long size)
 		 */
 		if (p->flags & IORESOURCE_BUSY)
 			continue;
+		/*
+		 * Only apply this quirk on Dell systems to avoid false
+		 * positives elsewhere.
+		 */
+		if (dmi_match(DMI_SYS_VENDOR, "Dell Inc.")) {
+			if (p->start == 0xfe000000 && p->end == 0xfe010fff)
+				continue;
+		}
 
 		pr_warn("resource sanity check: requesting [mem %pa-%pa], which spans more than %s %pR\n",
 			&addr, &end, p->name, p);
Signature · 1 line
2.47.3
#2 lore raw
Quoted 26 lines · Manuel Quintero Fonseca

On Dell OptiPlex 3080 systems, specifically during the initialization of the intel_pmc_core driver, the kernel reports a resource sanity check conflict. The issue was identified via dmesg logs showing:

resource sanity check: requesting [mem 0xfdffe800-0xfe0007ff], which spans more than PNP0C02:03 [mem 0xfdb00000-0xfdffffff]

And subsequently:

resource sanity check: requesting [mem 0xfdffe800-0xfe0007ff], which spans more than Reserved [mem 0xfe000000-0xfe010fff]

Investigation of /proc/iomem and ACPI tables on this specific model revealed that the BIOS defines reserved regions (as a PNP0C02 resource and an E820 reserved entry) that overlap with the Power Management Controller (PMC) register space.

This overlap is harmless for system stability but produces persistent warnings in dmesg. This patch introduces a quirk to:

1. Filter out the conflicting fixed memory resources in drivers/acpi/resource.c.
2. Skip the iomem sanity check for the specific problematic range in kernel/resource.c.

Quoted 4 lines

This ensures a clean boot log on these Dell units and allows the Intel PMC driver to bind correctly without resource conflicts.

I'm not a big fan of this approach TBH. Adding quirks specifically to avoid printing spurious messages is a bit of a stretch IMV.

Have you considered downgrading the log level of the messages in question to "debug" instead?

Quoted 93 lines

---
v2:
- Add DMI matching to restrict the quirk to Dell systems.
- Include <linux/dmi.h> to fix compilation errors.
- Expand commit description with specific hardware (Dell OptiPlex 3080).
- Include dmesg logs showing the resource conflict.
- Reformat description to 75 characters per line.

 drivers/acpi/resource.c | 32 ++++++++++++++++++++++++++++++++
 kernel/resource.c       |  9 +++++++++
 2 files changed, 41 insertions(+)
drivers/acpi/resource.c +32 -0
index bc8050d8a6f5..464195e30080 100644
@@ -50,6 +50,20 @@ static inline bool is_gsi(struct acpi_resource_extended_irq *ext_irq)
 }
 #endif
 
+/*
+ * List of Dell systems requiring memory resource quirks due to BIOS
+ * reservations overlapping with Intel PMC register space.
+ */
+static const struct dmi_system_id dell_memory_quirks[] = {
+       {
+               .ident = "Dell Inc.",
+               .matches = {
+                       DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
+               },
+       },
+       { }
+};
+
 static bool acpi_dev_resource_len_valid(u64 start, u64 end, u64 len, bool io)
 {
        u64 reslen = end - start + 1;
@@ -110,6 +124,24 @@ bool acpi_dev_resource_memory(struct acpi_resource *ares, struct resource *res)
        struct acpi_resource_memory32 *memory32;
        struct acpi_resource_fixed_memory32 *fixed_memory32;
 
+       /*
+        * Quirk for Dell systems: ignore BIOS memory reservations that overlap
+        * with Intel PMC resources. Restricted via DMI to avoid impact on
+        * other manufacturers.
+        */
+       if (dmi_check_system(dell_memory_quirks)) {
+               u32 addr = 0;
+
+               if (ares->type == ACPI_RESOURCE_TYPE_FIXED_MEMORY32)
+                       addr = ares->data.fixed_memory32.address;
+               else if (ares->type == ACPI_RESOURCE_TYPE_MEMORY32)
+                       addr = ares->data.memory32.minimum;
+
+               if (addr == 0xfdb00000 ||
+                   (addr >= 0xfe000000 && addr <= 0xfe000fff))
+                       return false;
+       }
+
        switch (ares->type) {
        case ACPI_RESOURCE_TYPE_MEMORY24:
                memory24 = &ares->data.memory24;
kernel/resource.c +9 -1
index bb966699da31..16c82d9ea14f 100644
@@ -30,6 +30,7 @@
 #include <linux/string.h>
 #include <linux/vmalloc.h>
 #include <asm/io.h>
+#include <linux/dmi.h>
 
 
 struct resource ioport_resource = {
@@ -1857,6 +1858,14 @@ int iomem_map_sanity_check(resource_size_t addr, unsigned long size)
                 */
                if (p->flags & IORESOURCE_BUSY)
                        continue;
+               /*
+                * Only apply this quirk on Dell systems to avoid false
+                * positives elsewhere.
+                */
+               if (dmi_match(DMI_SYS_VENDOR, "Dell Inc.")) {
+                       if (p->start == 0xfe000000 && p->end == 0xfe010fff)
+                               continue;
+               }
 
                pr_warn("resource sanity check: requesting [mem %pa-%pa], which spans more than %s %pR\n",
                        &addr, &end, p->name, p);
--

2.47.3

author
#3 lore raw
Quoted 6 lines

I'm not a big fan of this approach TBH. Adding quirks specifically to avoid printing spurious messages is a bit of a stretch IMV.

Have you considered downgrading the log level of the messages in question to "debug" instead?

Hello Rafael,

Thank you for your feedback.

You're right, adding a quirk just to hide a spurious warning is not the best approach. I agree with your suggestion.

Instead, I will resubmit the patch changing the log level of the message from dev_warn() to dev_dbg(). That way, the message will still be available for developers when dynamic debugging is enabled, but regular users won't see it in their dmesg.

I'll prepare a v3 patch following this simpler, cleaner approach.

Thanks again for the guidance.

Best regards,

#4 lore raw

PATCH v3 1/1 resource: Downgrade "resource sanity check" warning to debug level

In reply to Rafael J. Wysocki

The "resource sanity check" warning currently triggers on certain systems, causing unnecessary noise in the kernel logs for users. Following the maintainer's suggestion, this patch downgrades the log level from pr_warn to pr_debug.

This change reduces log clutter while keeping the diagnostic information available for debugging purposes if needed.

---
v3:
- Downgrade 'resource sanity check' warning to debug level to reduce log noise, as suggested by the maintainer.
- Drop the previous DMI-based quirk approach as it is no longer needed.

 kernel/resource.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
kernel/resource.c +1 -1
index d02a53fb95d8..e676ea99c93c 100644
@@ -1859,7 +1859,7 @@ int iomem_map_sanity_check(resource_size_t addr, unsigned long size)
 		if (p->flags & IORESOURCE_BUSY)
 			continue;
 
-		pr_warn("resource sanity check: requesting [mem %pa-%pa], which spans more than %s %pR\n",
+		pr_debug("resource sanity check: requesting [mem %pa-%pa], which spans more than %s %pR\n",
 			&addr, &end, p->name, p);
 		err = -1;
 		break;
Signature · 1 line
2.47.3
#5 lore raw

Re: [PATCH v3] resource: Downgrade "resource sanity check" warning to debug level

Quoted 4 lines · On Fri, 22 May 2026 17:01:31 -0700 Manuel Quintero Fonseca <[email protected]>…

The "resource sanity check" warning currently triggers on certain systems, causing unnecessary noise in the kernel logs for users. Following the maintainer's suggestion, this patch downgrades the log level from pr_warn to pr_debug.

um, who is "the maintainer"?

hp2:/usr/src/25> grep kernel/resource.c MAINTAINERS hp2:/usr/src/25>

Quoted 8 lines

This change reduces log clutter while keeping the diagnostic information available for debugging purposes if needed.

...

kernel/resource.c int iomem_map_sanity_check(resource_size_t addr, unsigned long size) · line 1865

Quoted 8 lines
kernel/resource.c +1 -1
@@ -1859,7 +1859,7 @@ int iomem_map_sanity_check(resource_size_t addr, unsigned long size)
 		if (p->flags & IORESOURCE_BUSY)
 			continue;
 
-		pr_warn("resource sanity check: requesting [mem %pa-%pa], which spans more than %s %pR\n",
+		pr_debug("resource sanity check: requesting [mem %pa-%pa], which spans more than %s %pR\n",
 			&addr, &end, p->name, p);
 		err = -1;
 		break;

lgtm, thanks.

#6 lore raw

Re: [PATCH v3] resource: Downgrade "resource sanity check" warning to debug level

In reply to Manuel Quintero Fonseca
Quoted 3 lines · Manuel Quintero Fonseca

The "resource sanity check" warning currently triggers on certain systems, causing unnecessary noise in the kernel logs for users.

The fact that it triggers at all is not a problem. In principle, it may help to diagnose a functional issue. However, it sometimes doesn't mean that there is a functional issue, which is a problem.

I would say something like

The "resource sanity check" warning need not mean that there is a
functional issue
and in the cases when there isn't one, it is just useless and confusing noise.

Quoted 2 lines

Following the maintainer's suggestion, this patch downgrades the log level from pr_warn to pr_debug.

Instead of the first part of the sentence above, you may use a Suggested-by tag (pointing to the person whose suggestion is followed) and the "this patch" part is redundant.

I would just say

For this reason, downgrade the log level of it from "warn" to "debug".

Quoted 2 lines

This change reduces log clutter while keeping the diagnostic information available for debugging purposes if needed.

The word "change" can be dropped from the above sentence.

Quoted 1 line

It would be nice to add a Link: tag pointing to the discussion on the previous version of the patch.

Quoted 14 lines

---
v3:
- Downgrade 'resource sanity check' warning to debug level to reduce log noise, as suggested by the maintainer.
- Drop the previous DMI-based quirk approach as it is no longer needed.

 kernel/resource.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
kernel/resource.c +0 -0
index d02a53fb95d8..e676ea99c93c 100644

kernel/resource.c int iomem_map_sanity_check(resource_size_t addr, unsigned long size)

Quoted 9 lines
kernel/resource.c +1 -2
@@ -1859,7 +1859,7 @@ int iomem_map_sanity_check(resource_size_t addr, unsigned long size)
                if (p->flags & IORESOURCE_BUSY)
                        continue;
 
-               pr_warn("resource sanity check: requesting [mem %pa-%pa], which spans more than %s %pR\n",
+               pr_debug("resource sanity check: requesting [mem %pa-%pa], which spans more than %s %pR\n",
                        &addr, &end, p->name, p);
                err = -1;
                break;
--

And the code change looks good to me.