Skip to contentSkip to conversation

perf/x86/intel/uncore: Fix NULL pointer dereference issue in upi_fill_topology()

PATCH Superseded [email protected] 2 participants 5 messages Nov 15, 2023 – Nov 21, 2023

Subsystems (2)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 perf/x86/intel/uncore: Fix NULL pointer dereference issue in upi_fill_topology()

From: Alexander Antonov <[email protected]>

The NULL dereference happens inside upi_fill_topology() procedure in case of disabling one of the sockets on the system.

For example, if you disable the 2nd socket on a 4-socket system then uncore_max_dies() returns 3 and inside pmu_alloc_topology() memory will be allocated only for 3 sockets and stored in type->topology. In discover_upi_topology() memory is accessed by socket id from CPUNODEID registers which contain physical ids (from 0 to 3) and on the line:

    upi = &type->topology[nid][idx];

out-of-bound access will happen and the 'upi' pointer will be passed to upi_fill_topology() where it will be dereferenced.

To avoid this issue update the code to convert physical socket id to logical socket id in discover_upi_topology() before accessing memory.

Fixes f680b6e6062e ("perf/x86/intel/uncore: Enable UPI topology discovery for Icelake Server")
---
 arch/x86/events/intel/uncore_snbep.c | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)
arch/x86/events/intel/uncore_snbep.c +8 -2
index 8250f0f59c2b..49bc27ab26ad 100644
@@ -5596,7 +5596,7 @@ static int discover_upi_topology(struct intel_uncore_type *type, int ubox_did, i
 	struct pci_dev *ubox = NULL;
 	struct pci_dev *dev = NULL;
 	u32 nid, gid;
-	int i, idx, ret = -EPERM;
+	int i, idx, lgc_pkg, ret = -EPERM;
 	struct intel_uncore_topology *upi;
 	unsigned int devfn;
 
@@ -5614,8 +5614,13 @@ static int discover_upi_topology(struct intel_uncore_type *type, int ubox_did, i
 		for (i = 0; i < 8; i++) {
 			if (nid != GIDNIDMAP(gid, i))
 				continue;
+			lgc_pkg = topology_phys_to_logical_pkg(i);
+			if (lgc_pkg < 0) {
+				ret = -EPERM;
+				goto err;
+			}
 			for (idx = 0; idx < type->num_boxes; idx++) {
-				upi = &type->topology[nid][idx];
+				upi = &type->topology[lgc_pkg][idx];
 				devfn = PCI_DEVFN(dev_link0 + idx, ICX_UPI_REGS_ADDR_FUNCTION);
 				dev = pci_get_domain_bus_and_slot(pci_domain_nr(ubox->bus),
 								  ubox->bus->number,
@@ -5626,6 +5631,7 @@ static int discover_upi_topology(struct intel_uncore_type *type, int ubox_did, i
 						goto err;
 				}
 			}
+			break;
 		}
 	}
 err:

base-commit: 9bacdd8996c77c42ca004440be610692275ff9d0

Signature · 1 line
2.25.1
#1 lore raw
Quoted 41 lines · On 2023-11-15 10:13 a.m., [email protected] wrote:

From: Alexander Antonov <[email protected]>

The NULL dereference happens inside upi_fill_topology() procedure in case of disabling one of the sockets on the system.

For example, if you disable the 2nd socket on a 4-socket system then uncore_max_dies() returns 3 and inside pmu_alloc_topology() memory will be allocated only for 3 sockets and stored in type->topology. In discover_upi_topology() memory is accessed by socket id from CPUNODEID registers which contain physical ids (from 0 to 3) and on the line:

    upi = &type->topology[nid][idx];

out-of-bound access will happen and the 'upi' pointer will be passed to upi_fill_topology() where it will be dereferenced.

To avoid this issue update the code to convert physical socket id to logical socket id in discover_upi_topology() before accessing memory.

Fixes f680b6e6062e ("perf/x86/intel/uncore: Enable UPI topology discovery for Icelake Server")
---
 arch/x86/events/intel/uncore_snbep.c | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)
arch/x86/events/intel/uncore_snbep.c +1 -1
index 8250f0f59c2b..49bc27ab26ad 100644
@@ -5596,7 +5596,7 @@ static int discover_upi_topology(struct intel_uncore_type *type, int ubox_did, i
 	struct pci_dev *ubox = NULL;
 	struct pci_dev *dev = NULL;
 	u32 nid, gid;
-	int i, idx, ret = -EPERM;
+	int i, idx, lgc_pkg, ret = -EPERM;
 	struct intel_uncore_topology *upi;
 	unsigned int devfn;
 

arch/x86/events/intel/uncore_snbep.c static int discover_upi_topology(struct intel_uncore_type *type, int ubox_did, i · line 5621

Quoted 8 lines
arch/x86/events/intel/uncore_snbep.c +5 -0
@@ -5614,8 +5614,13 @@ static int discover_upi_topology(struct intel_uncore_type *type, int ubox_did, i
 		for (i = 0; i < 8; i++) {
 			if (nid != GIDNIDMAP(gid, i))
 				continue;
+			lgc_pkg = topology_phys_to_logical_pkg(i);
+			if (lgc_pkg < 0) {
+				ret = -EPERM;
+				goto err;
+			}

In the snbep_pci2phy_map_init(), there are similar codes to find the logical die id. Can we factor a common function for both of them?

Thanks,
Kan

Quoted 16 lines
arch/x86/events/intel/uncore_snbep.c +2 -1
 			for (idx = 0; idx < type->num_boxes; idx++) {
-				upi = &type->topology[nid][idx];
+				upi = &type->topology[lgc_pkg][idx];
 				devfn = PCI_DEVFN(dev_link0 + idx, ICX_UPI_REGS_ADDR_FUNCTION);
 				dev = pci_get_domain_bus_and_slot(pci_domain_nr(ubox->bus),
 								  ubox->bus->number,
@@ -5626,6 +5631,7 @@ static int discover_upi_topology(struct intel_uncore_type *type, int ubox_did, i
 						goto err;
 				}
 			}
+			break;
 		}
 	}
 err:

base-commit: 9bacdd8996c77c42ca004440be610692275ff9d0

#2 lore raw
Quoted 54 lines · Kan
Quoted 41 lines · On 2023-11-15 10:13 a.m., [email protected] wrote:

From: Alexander Antonov <[email protected]>

The NULL dereference happens inside upi_fill_topology() procedure in case of disabling one of the sockets on the system.

For example, if you disable the 2nd socket on a 4-socket system then uncore_max_dies() returns 3 and inside pmu_alloc_topology() memory will be allocated only for 3 sockets and stored in type->topology. In discover_upi_topology() memory is accessed by socket id from CPUNODEID registers which contain physical ids (from 0 to 3) and on the line:

    upi = &type->topology[nid][idx];

out-of-bound access will happen and the 'upi' pointer will be passed to upi_fill_topology() where it will be dereferenced.

To avoid this issue update the code to convert physical socket id to logical socket id in discover_upi_topology() before accessing memory.

Fixes f680b6e6062e ("perf/x86/intel/uncore: Enable UPI topology discovery for Icelake Server")
---
  arch/x86/events/intel/uncore_snbep.c | 10 ++++++++--
  1 file changed, 8 insertions(+), 2 deletions(-)
arch/x86/events/intel/uncore_snbep.c +1 -1
index 8250f0f59c2b..49bc27ab26ad 100644
@@ -5596,7 +5596,7 @@ static int discover_upi_topology(struct intel_uncore_type *type, int ubox_did, i
  	struct pci_dev *ubox = NULL;
  	struct pci_dev *dev = NULL;
  	u32 nid, gid;
-	int i, idx, ret = -EPERM;
+	int i, idx, lgc_pkg, ret = -EPERM;
  	struct intel_uncore_topology *upi;
  	unsigned int devfn;
 

arch/x86/events/intel/uncore_snbep.c static int discover_upi_topology(struct intel_uncore_type *type, int ubox_did, i · line 5621

Quoted 8 lines
arch/x86/events/intel/uncore_snbep.c +5 -0
@@ -5614,8 +5614,13 @@ static int discover_upi_topology(struct intel_uncore_type *type, int ubox_did, i
  		for (i = 0; i < 8; i++) {
  			if (nid != GIDNIDMAP(gid, i))
  				continue;
+			lgc_pkg = topology_phys_to_logical_pkg(i);
+			if (lgc_pkg < 0) {
+				ret = -EPERM;
+				goto err;
+			}

In the snbep_pci2phy_map_init(), there are similar codes to find the logical die id. Can we factor a common function for both of them?

Quoted 2 lines

Thanks,
Kan

Hi Kan,

Thank you for your comment.
Yes, I think we can factor out the common loop where GIDNIDMAP is being checked. But inside snbep_pci2phy_map_init() we have a bit different procedure which also does the following:

if (topology_max_die_per_package() > 1)
    die_id = i;

I think that having this code, at least, in our case could bring us to the same issue which we are trying to fix. But of course we could parametrize this checking.

What do you think?

Thanks,
Alexander

Quoted 17 lines
Quoted 17 lines
arch/x86/events/intel/uncore_snbep.c +2 -1
  			for (idx = 0; idx < type->num_boxes; idx++) {
-				upi = &type->topology[nid][idx];
+				upi = &type->topology[lgc_pkg][idx];
  				devfn = PCI_DEVFN(dev_link0 + idx, ICX_UPI_REGS_ADDR_FUNCTION);
  				dev = pci_get_domain_bus_and_slot(pci_domain_nr(ubox->bus),
  								  ubox->bus->number,
@@ -5626,6 +5631,7 @@ static int discover_upi_topology(struct intel_uncore_type *type, int ubox_did, i
  						goto err;
  				}
  			}
+			break;
  		}
  	}
  err:

base-commit: 9bacdd8996c77c42ca004440be610692275ff9d0

#3 lore raw
Quoted 70 lines · Alexander Antonov
Quoted 58 lines · Kan
Quoted 45 lines · On 2023-11-15 10:13 a.m., [email protected] wrote:

From: Alexander Antonov <[email protected]>

The NULL dereference happens inside upi_fill_topology() procedure in case of disabling one of the sockets on the system.

For example, if you disable the 2nd socket on a 4-socket system then uncore_max_dies() returns 3 and inside pmu_alloc_topology() memory will be allocated only for 3 sockets and stored in type->topology. In discover_upi_topology() memory is accessed by socket id from CPUNODEID
registers which contain physical ids (from 0 to 3) and on the line:

     upi = &type->topology[nid][idx];

out-of-bound access will happen and the 'upi' pointer will be passed to upi_fill_topology() where it will be dereferenced.

To avoid this issue update the code to convert physical socket id to logical socket id in discover_upi_topology() before accessing memory.

Fixes f680b6e6062e ("perf/x86/intel/uncore: Enable UPI topology

discovery for Icelake Server")

---
  arch/x86/events/intel/uncore_snbep.c | 10 ++++++++--

  1 file changed, 8 insertions(+), 2 deletions(-)
arch/x86/events/intel/uncore_snbep.c +0 -0
index 8250f0f59c2b..49bc27ab26ad 100644
@@ -5596,7 +5596,7 @@ static int discover_upi_topology(struct

intel_uncore_type *type, int ubox_did, i
      struct pci_dev *ubox = NULL;
      struct pci_dev *dev = NULL;
      u32 nid, gid;
-    int i, idx, ret = -EPERM;
+    int i, idx, lgc_pkg, ret = -EPERM;
      struct intel_uncore_topology *upi;
      unsigned int devfn;
  @@ -5614,8 +5614,13 @@ static int discover_upi_topology(struct intel_uncore_type *type, int ubox_did, i

Quoted 8 lines

          for (i = 0; i < 8; i++) {
              if (nid != GIDNIDMAP(gid, i)) continue;
+            lgc_pkg = topology_phys_to_logical_pkg(i); +            if (lgc_pkg < 0) {
+                ret = -EPERM;
+                goto err;
+            }

In the snbep_pci2phy_map_init(), there are similar codes to find the logical die id. Can we factor a common function for both of them?

Quoted 2 lines

Thanks,
Kan

Hi Kan,

Thank you for your comment.
Yes, I think we can factor out the common loop where GIDNIDMAP is being checked.
But inside snbep_pci2phy_map_init() we have a bit different procedure which also does the following:

Quoted 6 lines

if (topology_max_die_per_package() > 1)
    die_id = i;

I think that having this code, at least, in our case could bring us to the same issue which we are trying to fix. But of course we could parametrize this checking.

The topology_max_die_per_package() > 1 means there are more that 1 die in a socket. AFAIK, it only happens on the Cascade Lake AP.

Did you observe it in the ICX?

Thanks,
Kan

Quoted 25 lines

What do you think?

Thanks,
Alexander

Quoted 20 lines
Quoted 20 lines

              for (idx = 0; idx < type->num_boxes; idx++) { -                upi = &type->topology[nid][idx]; +                upi = &type->topology[lgc_pkg][idx]; devfn = PCI_DEVFN(dev_link0 + idx, ICX_UPI_REGS_ADDR_FUNCTION);
                  dev = pci_get_domain_bus_and_slot(pci_domain_nr(ubox->bus), ubox->bus->number,

arch/x86/events/intel/uncore_snbep.c +0 -0
@@ -5626,6 +5631,7 @@ static int discover_upi_topology(struct

intel_uncore_type *type, int ubox_did, i
                          goto err;
                  }
              }
+            break;
          }
      }
  err:

base-commit: 9bacdd8996c77c42ca004440be610692275ff9d0

#4 lore raw
Quoted 76 lines · Kan
Quoted 68 lines · Alexander Antonov
Quoted 57 lines · Kan
Quoted 45 lines · On 2023-11-15 10:13 a.m., [email protected] wrote:

From: Alexander Antonov <[email protected]>

The NULL dereference happens inside upi_fill_topology() procedure in case of disabling one of the sockets on the system.

For example, if you disable the 2nd socket on a 4-socket system then uncore_max_dies() returns 3 and inside pmu_alloc_topology() memory will be allocated only for 3 sockets and stored in type->topology. In discover_upi_topology() memory is accessed by socket id from CPUNODEID
registers which contain physical ids (from 0 to 3) and on the line:

     upi = &type->topology[nid][idx];

out-of-bound access will happen and the 'upi' pointer will be passed to upi_fill_topology() where it will be dereferenced.

To avoid this issue update the code to convert physical socket id to logical socket id in discover_upi_topology() before accessing memory.

Fixes f680b6e6062e ("perf/x86/intel/uncore: Enable UPI topology

discovery for Icelake Server")

---
   arch/x86/events/intel/uncore_snbep.c | 10 ++++++++--
   1 file changed, 8 insertions(+), 2 deletions(-)
arch/x86/events/intel/uncore_snbep.c +0 -0
index 8250f0f59c2b..49bc27ab26ad 100644
@@ -5596,7 +5596,7 @@ static int discover_upi_topology(struct

intel_uncore_type *type, int ubox_did, i
      struct pci_dev *ubox = NULL;
      struct pci_dev *dev = NULL;
      u32 nid, gid;
-    int i, idx, ret = -EPERM;
+    int i, idx, lgc_pkg, ret = -EPERM;
      struct intel_uncore_topology *upi;
      unsigned int devfn;
  @@ -5614,8 +5614,13 @@ static int discover_upi_topology(struct
intel_uncore_type *type, int ubox_did, i

Quoted 8 lines

          for (i = 0; i < 8; i++) {
              if (nid != GIDNIDMAP(gid, i)) continue;
+            lgc_pkg = topology_phys_to_logical_pkg(i); +            if (lgc_pkg < 0) {
+                ret = -EPERM;
+                goto err;
+            }

In the snbep_pci2phy_map_init(), there are similar codes to find the logical die id. Can we factor a common function for both of them?

Quoted 2 lines

Thanks,
Kan

Hi Kan,

Thank you for your comment.
Yes, I think we can factor out the common loop where GIDNIDMAP is being checked.
But inside snbep_pci2phy_map_init() we have a bit different procedure which also does the following:

Quoted 6 lines

if (topology_max_die_per_package() > 1)
    die_id = i;

I think that having this code, at least, in our case could bring us to the same issue which we are trying to fix. But of course we could parametrize this checking.

Quoted 7 lines

The topology_max_die_per_package() > 1 means there are more that 1 die in a socket. AFAIK, it only happens on the Cascade Lake AP.

Did you observe it in the ICX?

Thanks,
Kan

No, I didn't observe it on ICX. Seems for now we have it only on CLX-AP

Thanks,
Alexander

Quoted 24 lines
Quoted 24 lines

What do you think?

Thanks,
Alexander

Quoted 19 lines
Quoted 19 lines

              for (idx = 0; idx < type->num_boxes; idx++) {
-                upi = &type->topology[nid][idx]; +                upi = &type->topology[lgc_pkg][idx];
                  devfn = PCI_DEVFN(dev_link0 + idx,
ICX_UPI_REGS_ADDR_FUNCTION);
                  dev =
pci_get_domain_bus_and_slot(pci_domain_nr(ubox->bus),
                                    ubox->bus->number,

arch/x86/events/intel/uncore_snbep.c +0 -0
@@ -5626,6 +5631,7 @@ static int discover_upi_topology(struct

intel_uncore_type *type, int ubox_did, i
                          goto err;
                  }
              }
+            break;
          }
      }
  err:

base-commit: 9bacdd8996c77c42ca004440be610692275ff9d0