drivers/bus: set device NUMA node to unknown by default

Message ID 20220929120512.480-1-olivier.matz@6wind.com (mailing list archive)
State Changes Requested, archived
Delegated to: David Marchand
Headers
Series drivers/bus: set device NUMA node to unknown by default |

Checks

Context Check Description
ci/checkpatch success coding style OK
ci/Intel-compilation success Compilation OK
ci/github-robot: build success github build: passed
ci/iol-mellanox-Performance success Performance Testing PASS
ci/iol-aarch64-unit-testing success Testing PASS
ci/iol-aarch64-compile-testing success Testing PASS
ci/iol-intel-Functional success Functional Testing PASS
ci/iol-x86_64-unit-testing success Testing PASS
ci/iol-intel-Performance success Performance Testing PASS
ci/iol-x86_64-compile-testing success Testing PASS
ci/intel-Testing success Testing PASS

Commit Message

Olivier Matz Sept. 29, 2022, 12:05 p.m. UTC
  The dev->device.numa_node field is set by each bus driver for
every device it manages to indicate on which NUMA node this device lies.

When this information is unknown, the assigned value is not consistent
across the bus drivers.

Set the default value to SOCKET_ID_ANY (-1) by all bus drivers
when the NUMA information is unavailable.

Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
---
 doc/guides/rel_notes/deprecation.rst     |  7 -------
 doc/guides/rel_notes/release_22_11.rst   |  6 ++++++
 drivers/bus/auxiliary/auxiliary_common.c |  8 ++------
 drivers/bus/auxiliary/linux/auxiliary.c  | 13 +++++--------
 drivers/bus/dpaa/dpaa_bus.c              |  1 +
 drivers/bus/fslmc/fslmc_bus.c            |  1 +
 drivers/bus/pci/bsd/pci.c                |  2 +-
 drivers/bus/pci/linux/pci.c              | 16 ++++++----------
 drivers/bus/pci/pci_common.c             |  8 ++------
 drivers/bus/pci/windows/pci.c            |  1 -
 drivers/bus/vmbus/linux/vmbus_bus.c      |  1 -
 drivers/bus/vmbus/vmbus_common.c         |  8 ++------
 12 files changed, 26 insertions(+), 46 deletions(-)
  

Comments

David Marchand Sept. 30, 2022, 7:10 a.m. UTC | #1
On Thu, Sep 29, 2022 at 2:05 PM Olivier Matz <olivier.matz@6wind.com> wrote:
>
> The dev->device.numa_node field is set by each bus driver for
> every device it manages to indicate on which NUMA node this device lies.
>
> When this information is unknown, the assigned value is not consistent
> across the bus drivers.
>
> Set the default value to SOCKET_ID_ANY (-1) by all bus drivers
> when the NUMA information is unavailable.
>
> Signed-off-by: Olivier Matz <olivier.matz@6wind.com>

-  The dma/idxd driver, which has its own bus, seems ok, though I
would align its code for readability.

@@ -322,7 +322,7 @@ dsa_scan(void)

        while ((wq = readdir(dev_dir)) != NULL) {
                struct rte_dsa_device *dev;
-               int numa_node = -1;
+               int numa_node = SOCKET_ID_ANY;

                if (strncmp(wq->d_name, "wq", 2) != 0)
                        continue;


- There is an impact on upper classes of devices.

For ethdev, a port inherits the numa_node value from the rte_device object.
Yet, rte_eth_dev_socket_id() is described as:
 * @return
 *   The NUMA socket ID to which the Ethernet device is connected or
 *   a default of zero if the socket could not be determined.
 *   -1 is returned is the port_id value is out of range.
  
Olivier Matz Sept. 30, 2022, 8:11 a.m. UTC | #2
On Fri, Sep 30, 2022 at 09:10:55AM +0200, David Marchand wrote:
> On Thu, Sep 29, 2022 at 2:05 PM Olivier Matz <olivier.matz@6wind.com> wrote:
> >
> > The dev->device.numa_node field is set by each bus driver for
> > every device it manages to indicate on which NUMA node this device lies.
> >
> > When this information is unknown, the assigned value is not consistent
> > across the bus drivers.
> >
> > Set the default value to SOCKET_ID_ANY (-1) by all bus drivers
> > when the NUMA information is unavailable.
> >
> > Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
> 
> -  The dma/idxd driver, which has its own bus, seems ok, though I
> would align its code for readability.
> 
> @@ -322,7 +322,7 @@ dsa_scan(void)
> 
>         while ((wq = readdir(dev_dir)) != NULL) {
>                 struct rte_dsa_device *dev;
> -               int numa_node = -1;
> +               int numa_node = SOCKET_ID_ANY;
> 
>                 if (strncmp(wq->d_name, "wq", 2) != 0)
>                         continue;
> 
> 
> - There is an impact on upper classes of devices.
> 
> For ethdev, a port inherits the numa_node value from the rte_device object.
> Yet, rte_eth_dev_socket_id() is described as:
>  * @return
>  *   The NUMA socket ID to which the Ethernet device is connected or
>  *   a default of zero if the socket could not be determined.
>  *   -1 is returned is the port_id value is out of range.

Good catches, I'll fix them in v2.

Thanks,
Olivier
  

Patch

diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index be231ff8d2..75645f0a9a 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -36,13 +36,6 @@  Deprecation Notices
   ``__atomic_thread_fence`` must be used for patches that need to be merged in
   20.08 onwards. This change will not introduce any performance degradation.
 
-* bus: The ``dev->device.numa_node`` field is set by each bus driver for
-  every device it manages to indicate on which NUMA node this device lies.
-  When this information is unknown, the assigned value is not consistent
-  across the bus drivers.
-  In DPDK 22.11, the default value will be set to -1 by all bus drivers
-  when the NUMA information is unavailable.
-
 * kni: The KNI kernel module and library are not recommended for use by new
   applications - other technologies such as virtio-user are recommended instead.
   Following the DPDK technical board
diff --git a/doc/guides/rel_notes/release_22_11.rst b/doc/guides/rel_notes/release_22_11.rst
index 0231959874..5989e4a04f 100644
--- a/doc/guides/rel_notes/release_22_11.rst
+++ b/doc/guides/rel_notes/release_22_11.rst
@@ -266,6 +266,12 @@  ABI Changes
 * eventdev: Added ``weight`` and ``affinity`` fields
   to ``rte_event_queue_conf`` structure.
 
+* bus: Changed the device numa node to -1 when NUMA information is unavailable.
+  The ``dev->device.numa_node`` field is set by each bus driver for
+  every device it manages to indicate on which NUMA node this device lies.
+  When this information is unknown, the assigned value was not consistent
+  across the bus drivers.
+
 
 Known Issues
 ------------
diff --git a/drivers/bus/auxiliary/auxiliary_common.c b/drivers/bus/auxiliary/auxiliary_common.c
index 259ff152c4..6bb1fe7c96 100644
--- a/drivers/bus/auxiliary/auxiliary_common.c
+++ b/drivers/bus/auxiliary/auxiliary_common.c
@@ -105,12 +105,8 @@  rte_auxiliary_probe_one_driver(struct rte_auxiliary_driver *drv,
 		return -1;
 	}
 
-	if (dev->device.numa_node < 0) {
-		if (rte_socket_count() > 1)
-			AUXILIARY_LOG(INFO, "Device %s is not NUMA-aware, defaulting socket to 0",
-					dev->name);
-		dev->device.numa_node = 0;
-	}
+	if (dev->device.numa_node < 0 && rte_socket_count() > 1)
+		RTE_LOG(INFO, EAL, "Device %s is not NUMA-aware\n", dev->name);
 
 	iova_mode = rte_eal_iova_mode();
 	if ((drv->drv_flags & RTE_AUXILIARY_DRV_NEED_IOVA_AS_VA) > 0 &&
diff --git a/drivers/bus/auxiliary/linux/auxiliary.c b/drivers/bus/auxiliary/linux/auxiliary.c
index d4c564cd78..02fc9285dc 100644
--- a/drivers/bus/auxiliary/linux/auxiliary.c
+++ b/drivers/bus/auxiliary/linux/auxiliary.c
@@ -40,14 +40,11 @@  auxiliary_scan_one(const char *dirname, const char *name)
 	/* Get NUMA node, default to 0 if not present */
 	snprintf(filename, sizeof(filename), "%s/%s/numa_node",
 		 dirname, name);
-	if (access(filename, F_OK) != -1) {
-		if (eal_parse_sysfs_value(filename, &tmp) == 0)
-			dev->device.numa_node = tmp;
-		else
-			dev->device.numa_node = -1;
-	} else {
-		dev->device.numa_node = 0;
-	}
+	if (access(filename, F_OK) == 0 &&
+	    eal_parse_sysfs_value(filename, &tmp) == 0)
+		dev->device.numa_node = tmp;
+	else
+		dev->device.numa_node = SOCKET_ID_ANY;
 
 	auxiliary_on_scan(dev);
 
diff --git a/drivers/bus/dpaa/dpaa_bus.c b/drivers/bus/dpaa/dpaa_bus.c
index 682427ba2c..447b222a76 100644
--- a/drivers/bus/dpaa/dpaa_bus.c
+++ b/drivers/bus/dpaa/dpaa_bus.c
@@ -179,6 +179,7 @@  dpaa_create_device_list(void)
 		}
 
 		dev->device.bus = &rte_dpaa_bus.bus;
+		dev->device.numa_node = SOCKET_ID_ANY;
 
 		/* Allocate interrupt handle instance */
 		dev->intr_handle =
diff --git a/drivers/bus/fslmc/fslmc_bus.c b/drivers/bus/fslmc/fslmc_bus.c
index 8503004e3d..57bfb5111a 100644
--- a/drivers/bus/fslmc/fslmc_bus.c
+++ b/drivers/bus/fslmc/fslmc_bus.c
@@ -156,6 +156,7 @@  scan_one_fslmc_device(char *dev_name)
 	}
 
 	dev->device.bus = &rte_fslmc_bus.bus;
+	dev->device.numa_node = SOCKET_ID_ANY;
 
 	/* Allocate interrupt instance */
 	dev->intr_handle =
diff --git a/drivers/bus/pci/bsd/pci.c b/drivers/bus/pci/bsd/pci.c
index 844d145fed..7459d15c7e 100644
--- a/drivers/bus/pci/bsd/pci.c
+++ b/drivers/bus/pci/bsd/pci.c
@@ -246,7 +246,7 @@  pci_scan_one(int dev_pci_fd, struct pci_conf *conf)
 	dev->max_vfs = 0;
 
 	/* FreeBSD has no NUMA support (yet) */
-	dev->device.numa_node = 0;
+	dev->device.numa_node = SOCKET_ID_ANY;
 
 	pci_common_set(dev);
 
diff --git a/drivers/bus/pci/linux/pci.c b/drivers/bus/pci/linux/pci.c
index c8703d52f3..ade17079fc 100644
--- a/drivers/bus/pci/linux/pci.c
+++ b/drivers/bus/pci/linux/pci.c
@@ -283,17 +283,13 @@  pci_scan_one(const char *dirname, const struct rte_pci_addr *addr)
 	}
 
 	/* get numa node, default to 0 if not present */
-	snprintf(filename, sizeof(filename), "%s/numa_node",
-		 dirname);
+	snprintf(filename, sizeof(filename), "%s/numa_node", dirname);
 
-	if (access(filename, F_OK) != -1) {
-		if (eal_parse_sysfs_value(filename, &tmp) == 0)
-			dev->device.numa_node = tmp;
-		else
-			dev->device.numa_node = -1;
-	} else {
-		dev->device.numa_node = 0;
-	}
+	if (access(filename, F_OK) == 0 &&
+	    eal_parse_sysfs_value(filename, &tmp) == 0)
+		dev->device.numa_node = tmp;
+	else
+		dev->device.numa_node = SOCKET_ID_ANY;
 
 	pci_common_set(dev);
 
diff --git a/drivers/bus/pci/pci_common.c b/drivers/bus/pci/pci_common.c
index 5ea72bcf23..a59c5b4286 100644
--- a/drivers/bus/pci/pci_common.c
+++ b/drivers/bus/pci/pci_common.c
@@ -235,12 +235,8 @@  rte_pci_probe_one_driver(struct rte_pci_driver *dr,
 		return 1;
 	}
 
-	if (dev->device.numa_node < 0) {
-		if (rte_socket_count() > 1)
-			RTE_LOG(INFO, EAL, "Device %s is not NUMA-aware, defaulting socket to 0\n",
-					dev->name);
-		dev->device.numa_node = 0;
-	}
+	if (dev->device.numa_node < 0 && rte_socket_count() > 1)
+		RTE_LOG(INFO, EAL, "Device %s is not NUMA-aware\n", dev->name);
 
 	already_probed = rte_dev_is_probed(&dev->device);
 	if (already_probed && !(dr->drv_flags & RTE_PCI_DRV_PROBE_AGAIN)) {
diff --git a/drivers/bus/pci/windows/pci.c b/drivers/bus/pci/windows/pci.c
index 3f7a8b9432..5cf05ce1a0 100644
--- a/drivers/bus/pci/windows/pci.c
+++ b/drivers/bus/pci/windows/pci.c
@@ -249,7 +249,6 @@  get_device_resource_info(HDEVINFO dev_info,
 		DWORD error = GetLastError();
 		if (error == ERROR_NOT_FOUND) {
 			/* On older CPUs, NUMA is not bound to PCIe locality. */
-			dev->device.numa_node = 0;
 			return ERROR_SUCCESS;
 		}
 		RTE_LOG_WIN32_ERR("SetupDiGetDevicePropertyW"
diff --git a/drivers/bus/vmbus/linux/vmbus_bus.c b/drivers/bus/vmbus/linux/vmbus_bus.c
index f502783f7a..01d8111b85 100644
--- a/drivers/bus/vmbus/linux/vmbus_bus.c
+++ b/drivers/bus/vmbus/linux/vmbus_bus.c
@@ -293,7 +293,6 @@  vmbus_scan_one(const char *name)
 			goto error;
 		dev->device.numa_node = tmp;
 	} else {
-		/* if no NUMA support, set default to 0 */
 		dev->device.numa_node = SOCKET_ID_ANY;
 	}
 
diff --git a/drivers/bus/vmbus/vmbus_common.c b/drivers/bus/vmbus/vmbus_common.c
index 03b39c82b7..8d32d66504 100644
--- a/drivers/bus/vmbus/vmbus_common.c
+++ b/drivers/bus/vmbus/vmbus_common.c
@@ -111,12 +111,8 @@  vmbus_probe_one_driver(struct rte_vmbus_driver *dr,
 	/* reference driver structure */
 	dev->driver = dr;
 
-	if (dev->device.numa_node < 0) {
-		if (rte_socket_count() > 1)
-			VMBUS_LOG(INFO, "Device %s is not NUMA-aware, defaulting socket to 0",
-					guid);
-		dev->device.numa_node = 0;
-	}
+	if (dev->device.numa_node < 0 && rte_socket_count() > 1)
+		VMBUS_LOG(INFO, "Device %s is not NUMA-aware", guid);
 
 	/* call the driver probe() function */
 	VMBUS_LOG(INFO, "  probe driver: %s", dr->driver.name);