Skip to contentSkip to conversation

perf/x86/intel/uncore: Skip discovery table for offline dies

PATCH In review Zide Chen 2 participants 4 messages Jan 16

Subsystems (2)
lore
Share & tools
lore

Zide Chen

author
#0 lore raw

PATCH v3 1/2 perf/x86/intel/uncore: Skip discovery table for offline dies

This warning can be triggered if NUMA is disabled and the system boots with fewer CPUs than the number of CPUs in die 0.

WARNING: CPU: 9 PID: 7257 at uncore.c:1157 uncore_pci_pmu_register+0x136/0x160 [intel_uncore]

Currently, the discovery table continues to be parsed even if all CPUs in the associated die are offline. This can lead to an array overflow at "pmu->boxes[die] = box" in uncore_pci_pmu_register(), which may trigger the warning above or cause other issues.

Fixes edae1f06c2cd ("perf/x86/intel/uncore: Parse uncore discovery tables")

---
V2:
- Add the Tested-by tag
- Rebase onto perf/core (base commit: a491c02c2770)

V3:
- Remove the overly cautious WARN_ON()

 arch/x86/events/intel/uncore_discovery.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arch/x86/events/intel/uncore_discovery.c +1 -1
index b46575254dbe..0e414cecb6f2 100644
@@ -366,7 +366,7 @@ static bool uncore_discovery_pci(struct uncore_discovery_domain *domain)
 				     (val & UNCORE_DISCOVERY_DVSEC2_BIR_MASK) * UNCORE_DISCOVERY_BIR_STEP;
 
 			die = get_device_die_id(dev);
-			if (die < 0)
+			if ((die < 0) || (die >= uncore_max_dies()))
 				continue;
 
 			parse_discovery_table(domain, dev, die, bar_offset, &parsed);
Signature · 1 line
2.52.0
author
#1 lore raw

PATCH v3 2/2 perf/x86/intel/uncore: Fix die ID init and look up bugs

In snbep_pci2phy_map_init(), in the nr_node_ids > 8 path, uncore_device_to_die() may return -1 when all CPUs associated with the UBOX device are offline.

Remove the WARN_ON_ONCE(die_id == -1) check for two reasons:

- The current code breaks out of the loop. This is incorrect because pci_get_device() does not guarantee iteration in domain or bus order, so additional UBOX devices may be skipped during the scan.

- Returning -EINVAL is incorrect, since marking offline buses with die_id == -1 is expected and should not be treated as an error.

Separately, when NUMA is disabled on a NUMA-capable platform, pcibus_to_node() returns NUMA_NO_NODE, causing uncore_device_to_die() to return -1 for all PCI devices. As a result, spr_update_device_location(), used on Intel SPR and EMR, ignores the corresponding PMON units and does not add them to the RB tree.

Fix this by using uncore_pcibus_to_dieid(), which retrieves topology from the UBOX GIDNIDMAP register and works regardless of whether NUMA is enabled in Linux. This requires snbep_pci2phy_map_init() to be added in spr_uncore_pci_init().

Keep uncore_device_to_die() only for the nr_node_ids > 8 case, where NUMA is expected to be enabled.

Fixes 9a7832ce3d92 ("perf/x86/intel/uncore: With > 8 nodes, get pci bus die id from NUMA info")
Fixes 65248a9a9ee1 ("perf/x86/uncore: Add a quirk for UPI on SPR")

---
V2:
- Fix the commit message to note that spr_update_device_location() is used by EMR, not GNR.
- Rewrite the commit message for clarity.
- Add a Tested-by tag.

V3: no changes

 arch/x86/events/intel/uncore.c       |  1 +
 arch/x86/events/intel/uncore_snbep.c | 13 ++++++-------
 2 files changed, 7 insertions(+), 7 deletions(-)
arch/x86/events/intel/uncore.c +1 -0
index 4684649109d9..94592de6431a 100644
@@ -67,6 +67,7 @@ int uncore_die_to_segment(int die)
 	return bus ? pci_domain_nr(bus) : -EINVAL;
 }
 
+/* Note: This API can only be used when NUMA information is available. */
 int uncore_device_to_die(struct pci_dev *dev)
 {
 	int node = pcibus_to_node(dev->bus);
arch/x86/events/intel/uncore_snbep.c +6 -7
index 7ca0429c4004..52dec34d18c4 100644
@@ -1459,13 +1459,7 @@ static int snbep_pci2phy_map_init(int devid, int nodeid_loc, int idmap_loc, bool
 			}
 
 			map->pbus_to_dieid[bus] = die_id = uncore_device_to_die(ubox_dev);
-
 			raw_spin_unlock(&pci2phy_map_lock);
-
-			if (WARN_ON_ONCE(die_id == -1)) {
-				err = -EINVAL;
-				break;
-			}
 		}
 	}
 
@@ -6420,7 +6414,7 @@ static void spr_update_device_location(int type_id)
 
 	while ((dev = pci_get_device(PCI_VENDOR_ID_INTEL, device, dev)) != NULL) {
 
-		die = uncore_device_to_die(dev);
+		die = uncore_pcibus_to_dieid(dev->bus);
 		if (die < 0)
 			continue;
 
@@ -6444,6 +6438,11 @@ static void spr_update_device_location(int type_id)
 
 int spr_uncore_pci_init(void)
 {
+	int ret = snbep_pci2phy_map_init(0x3250, SKX_CPUNODEID, SKX_GIDNIDMAP, true);
+
+	if (ret)
+		return ret;
+
 	/*
 	 * The discovery table of UPI on some SPR variant is broken,
 	 * which impacts the detection of both UPI and M3UPI uncore PMON.
Signature · 1 line
2.52.0
Reviewed-by on v3 2/2
#2 lore raw

Re: [PATCH V3 2/2] perf/x86/intel/uncore: Fix die ID init and look up bugs

Quoted 85 lines · Zide Chen

In snbep_pci2phy_map_init(), in the nr_node_ids > 8 path, uncore_device_to_die() may return -1 when all CPUs associated with the UBOX device are offline.

Remove the WARN_ON_ONCE(die_id == -1) check for two reasons:

- The current code breaks out of the loop. This is incorrect because pci_get_device() does not guarantee iteration in domain or bus order, so additional UBOX devices may be skipped during the scan.

- Returning -EINVAL is incorrect, since marking offline buses with die_id == -1 is expected and should not be treated as an error.

Separately, when NUMA is disabled on a NUMA-capable platform, pcibus_to_node() returns NUMA_NO_NODE, causing uncore_device_to_die() to return -1 for all PCI devices. As a result, spr_update_device_location(), used on Intel SPR and EMR, ignores the corresponding PMON units and does not add them to the RB tree.

Fix this by using uncore_pcibus_to_dieid(), which retrieves topology from the UBOX GIDNIDMAP register and works regardless of whether NUMA is enabled in Linux. This requires snbep_pci2phy_map_init() to be added in spr_uncore_pci_init().

Keep uncore_device_to_die() only for the nr_node_ids > 8 case, where NUMA is expected to be enabled.

Fixes 9a7832ce3d92 ("perf/x86/intel/uncore: With > 8 nodes, get pci bus die id from NUMA info")
Fixes 65248a9a9ee1 ("perf/x86/uncore: Add a quirk for UPI on SPR")

---
V2:
- Fix the commit message to note that spr_update_device_location() is used by EMR, not GNR.
- Rewrite the commit message for clarity.
- Add a Tested-by tag.

V3: no changes

 arch/x86/events/intel/uncore.c       |  1 +
 arch/x86/events/intel/uncore_snbep.c | 13 ++++++-------
 2 files changed, 7 insertions(+), 7 deletions(-)
arch/x86/events/intel/uncore.c +1 -0
index 4684649109d9..94592de6431a 100644
@@ -67,6 +67,7 @@ int uncore_die_to_segment(int die)
 	return bus ? pci_domain_nr(bus) : -EINVAL;
 }
 
+/* Note: This API can only be used when NUMA information is available. */
 int uncore_device_to_die(struct pci_dev *dev)
 {
 	int node = pcibus_to_node(dev->bus);
arch/x86/events/intel/uncore_snbep.c +1 -7
index 7ca0429c4004..52dec34d18c4 100644
@@ -1459,13 +1459,7 @@ static int snbep_pci2phy_map_init(int devid, int nodeid_loc, int idmap_loc, bool
 			}
 
 			map->pbus_to_dieid[bus] = die_id = uncore_device_to_die(ubox_dev);
-
 			raw_spin_unlock(&pci2phy_map_lock);
-
-			if (WARN_ON_ONCE(die_id == -1)) {
-				err = -EINVAL;
-				break;
-			}
 		}
 	}
 
@@ -6420,7 +6414,7 @@ static void spr_update_device_location(int type_id)
 
 	while ((dev = pci_get_device(PCI_VENDOR_ID_INTEL, device, dev)) != NULL) {
 
-		die = uncore_device_to_die(dev);
+		die = uncore_pcibus_to_dieid(dev->bus);
 		if (die < 0)
 			continue;
 
@@ -6444,6 +6438,11 @@ static void spr_update_device_location(int type_id)
 

arch/x86/events/intel/uncore_snbep.c static void spr_update_device_location(int type_id) · line 6448

Quoted 10 lines
arch/x86/events/intel/uncore_snbep.c +5 -0
 int spr_uncore_pci_init(void)
 {
+	int ret = snbep_pci2phy_map_init(0x3250, SKX_CPUNODEID, SKX_GIDNIDMAP, true);
+
+	if (ret)
+		return ret;
+
 	/*
 	 * The discovery table of UPI on some SPR variant is broken,
 	 * which impacts the detection of both UPI and M3UPI uncore PMON.
Reviewed-by on v3 1/2
#3 lore raw
In reply to Zide Chen
Quoted 30 lines · Zide Chen

This warning can be triggered if NUMA is disabled and the system boots with fewer CPUs than the number of CPUs in die 0.

WARNING: CPU: 9 PID: 7257 at uncore.c:1157 uncore_pci_pmu_register+0x136/0x160 [intel_uncore]

Currently, the discovery table continues to be parsed even if all CPUs in the associated die are offline. This can lead to an array overflow at "pmu->boxes[die] = box" in uncore_pci_pmu_register(), which may trigger the warning above or cause other issues.

Fixes edae1f06c2cd ("perf/x86/intel/uncore: Parse uncore discovery tables")

---
V2:
- Add the Tested-by tag
- Rebase onto perf/core (base commit: a491c02c2770)

V3:
- Remove the overly cautious WARN_ON()

 arch/x86/events/intel/uncore_discovery.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arch/x86/events/intel/uncore_discovery.c +0 -0
index b46575254dbe..0e414cecb6f2 100644

arch/x86/events/intel/uncore_discovery.c static bool uncore_discovery_pci(struct uncore_discovery_domain *domain) · line 372

Quoted 8 lines
arch/x86/events/intel/uncore_discovery.c +1 -1
@@ -366,7 +366,7 @@ static bool uncore_discovery_pci(struct uncore_discovery_domain *domain)
 				     (val & UNCORE_DISCOVERY_DVSEC2_BIR_MASK) * UNCORE_DISCOVERY_BIR_STEP;
 
 			die = get_device_die_id(dev);
-			if (die < 0)
+			if ((die < 0) || (die >= uncore_max_dies()))
 				continue;
 
 			parse_discovery_table(domain, dev, die, bar_offset, &parsed);