diff mbox series

[v2] mem: fix allocation failure on non-NUMA kernel

Message ID 20201012192803.3741-1-nick.connolly@mayadata.io (mailing list archive)
State Accepted, archived
Delegated to: David Marchand
Headers
Series [v2] mem: fix allocation failure on non-NUMA kernel |

Checks

Context Check Description
ci/checkpatch success coding style OK
ci/iol-broadcom-Performance success Performance Testing PASS
ci/iol-broadcom-Functional success Functional Testing PASS
ci/iol-testing success Testing PASS
ci/iol-intel-Performance success Performance Testing PASS
ci/travis-robot success Travis build: passed
ci/iol-mellanox-Performance success Performance Testing PASS
ci/Intel-compilation success Compilation OK

Commit Message

Nick Connolly Oct. 12, 2020, 7:28 p.m. UTC 
  Running dpdk-helloworld on Linux with lib numa present,
but no kernel support for NUMA (CONFIG_NUMA=n) causes
ret_service_init() to fail with EAL: error allocating
rte services array.

alloc_seg() calls get_mempolicy to verify that the allocation
has happened on the correct socket, but receives ENOSYS from
the kernel and fails the allocation.

The allocated socket should only be verified if check_numa() is true.

Fixes: 2a96c88be83e ("mem: ease init in a docker container")
Cc: nicolas.dichtel@6wind.com
Cc: stable@dpdk.org

Signed-off-by: Nick Connolly <nick.connolly@mayadata.io>
---
v2:
* Remove unnecessary debug message and add explanatory comment

 lib/librte_eal/linux/eal_memalloc.c | 30 ++++++++++++++++++-----------
 1 file changed, 19 insertions(+), 11 deletions(-)

Comments

Nicolas Dichtel Oct. 13, 2020, 7:59 a.m. UTC | #1 
Le 12/10/2020 à 21:28, Nick Connolly a écrit :
> Running dpdk-helloworld on Linux with lib numa present,
> but no kernel support for NUMA (CONFIG_NUMA=n) causes
> ret_service_init() to fail with EAL: error allocating
> rte services array.
> 
> alloc_seg() calls get_mempolicy to verify that the allocation
> has happened on the correct socket, but receives ENOSYS from
> the kernel and fails the allocation.
> 
> The allocated socket should only be verified if check_numa() is true.
> 
> Fixes: 2a96c88be83e ("mem: ease init in a docker container")
> Cc: nicolas.dichtel@6wind.com
> Cc: stable@dpdk.org
> 
> Signed-off-by: Nick Connolly <nick.connolly@mayadata.io>
Reviewed-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
David Marchand Oct. 13, 2020, 12:01 p.m. UTC | #2 
On Tue, Oct 13, 2020 at 10:00 AM Nicolas Dichtel
<nicolas.dichtel@6wind.com> wrote:
>
> Le 12/10/2020 à 21:28, Nick Connolly a écrit :
> > Running dpdk-helloworld on Linux with lib numa present,
> > but no kernel support for NUMA (CONFIG_NUMA=n) causes
> > ret_service_init() to fail with EAL: error allocating
> > rte services array.
> >
> > alloc_seg() calls get_mempolicy to verify that the allocation
> > has happened on the correct socket, but receives ENOSYS from
> > the kernel and fails the allocation.
> >
> > The allocated socket should only be verified if check_numa() is true.
> >
> > Fixes: 2a96c88be83e ("mem: ease init in a docker container")
> > Cc: nicolas.dichtel@6wind.com
> > Cc: stable@dpdk.org
> >
> > Signed-off-by: Nick Connolly <nick.connolly@mayadata.io>
> Reviewed-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Acked-by: Anatoly Burakov <anatoly.burakov@intel.com>

Applied, thanks Nick.
diff mbox series

Patch

diff --git a/lib/librte_eal/linux/eal_memalloc.c b/lib/librte_eal/linux/eal_memalloc.c
index 40a5c4aa1..6dc1b2bae 100644
--- a/lib/librte_eal/linux/eal_memalloc.c
+++ b/lib/librte_eal/linux/eal_memalloc.c
@@ -625,17 +625,25 @@  alloc_seg(struct rte_memseg *ms, void *addr, int socket_id,
 	}
 
 #ifdef RTE_EAL_NUMA_AWARE_HUGEPAGES
-	ret = get_mempolicy(&cur_socket_id, NULL, 0, addr,
-			    MPOL_F_NODE | MPOL_F_ADDR);
-	if (ret < 0) {
-		RTE_LOG(DEBUG, EAL, "%s(): get_mempolicy: %s\n",
-			__func__, strerror(errno));
-		goto mapped;
-	} else if (cur_socket_id != socket_id) {
-		RTE_LOG(DEBUG, EAL,
-				"%s(): allocation happened on wrong socket (wanted %d, got %d)\n",
-			__func__, socket_id, cur_socket_id);
-		goto mapped;
+	/*
+	 * If the kernel has been built without NUMA support, get_mempolicy()
+	 * will return an error. If check_numa() returns false, memory
+	 * allocation is not NUMA aware and the socket_id should not be
+	 * checked.
+	 */
+	if (check_numa()) {
+		ret = get_mempolicy(&cur_socket_id, NULL, 0, addr,
+					MPOL_F_NODE | MPOL_F_ADDR);
+		if (ret < 0) {
+			RTE_LOG(DEBUG, EAL, "%s(): get_mempolicy: %s\n",
+				__func__, strerror(errno));
+			goto mapped;
+		} else if (cur_socket_id != socket_id) {
+			RTE_LOG(DEBUG, EAL,
+					"%s(): allocation happened on wrong socket (wanted %d, got %d)\n",
+				__func__, socket_id, cur_socket_id);
+			goto mapped;
+		}
 	}
 #else
 	if (rte_socket_count() > 1)