| Message ID | 20190508084447.18191-2-liwang@redhat.com |
|---|---|
| State | Accepted |
| Headers | show |
| Series | [v2,1/2] numa: fix numa test error with non-continuous nodes | expand |
On Wed, May 08, 2019 at 04:44:47PM +0800, Li Wang wrote: > For some ppc64le systems, it has non-continuous numa nodes in > hardware configuration. So we're hitting the below warnings while > running set_mempolicy tests on that. To fix this issue, let's just > remove restrictions on numa node-id in get_mempolicy(). > > Error Log > --------- > tst_test.c:1096: INFO: Timeout per run is 0h 50m 00s > tst_numa.c:190: INFO: Found 2 NUMA memory nodes > set_mempolicy01.c:63: PASS: set_mempolicy(MPOL_BIND) node 0 > tst_numa.c:26: INFO: Node 0 allocated 16 pages > tst_numa.c:26: INFO: Node 8 allocated 0 pages > set_mempolicy01.c:82: PASS: child: Node 0 allocated 16 > set_mempolicy01.c:63: PASS: set_mempolicy(MPOL_BIND) node 8 > tst_numa.c:92: WARN: get_mempolicy(...) returned invalid node 8 > tst_numa.c:92: WARN: get_mempolicy(...) returned invalid node 8 > tst_numa.c:92: WARN: get_mempolicy(...) returned invalid node 8 > ... > tst_numa.c:26: INFO: Node 0 allocated 0 pages > tst_numa.c:26: INFO: Node 8 allocated 0 pages > set_mempolicy01.c:86: FAIL: child: Node 8 allocated 0, expected 16 > > lscpu > ----- > Architecture: ppc64le > ... > CPU(s): 128 > Core(s) per socket: 16 > Socket(s): 2 > NUMA node(s): 2 > Model name: POWER9, altivec supported > ... > NUMA node0 CPU(s): 0-63 > NUMA node8 CPU(s): 64-127 > > Signed-off-by: Li Wang <liwang@redhat.com> > Cc: Cyril Hrubis <chrubis@suse.cz> > --- > libs/libltpnuma/tst_numa.c | 5 +++-- > 1 file changed, 3 insertions(+), 2 deletions(-) > > diff --git a/libs/libltpnuma/tst_numa.c b/libs/libltpnuma/tst_numa.c > index 0ba6daf39..56c8640ff 100644 > --- a/libs/libltpnuma/tst_numa.c > +++ b/libs/libltpnuma/tst_numa.c > @@ -88,8 +88,9 @@ void tst_nodemap_count_pages(struct tst_nodemap *nodes, > if (ret < 0) > tst_brk(TBROK | TERRNO, "get_mempolicy() failed"); > > - if (node < 0 || (unsigned int)node >= nodes->cnt) { > - tst_res(TWARN, "get_mempolicy(...) returned invalid node %i\n", node); > + if (node < 0) { It is better to have check for powerpc and remove restriction as it will still hold good for other arch. -- Bala > + tst_res(TWARN, > + "get_mempolicy(...) returned invalid node %i\n", node); > continue; > } > > -- > 2.20.1 > > > -- > Mailing list info: https://lists.linux.it/listinfo/ltp
On Thu, May 9, 2019 at 2:45 PM Balamuruhan S <bala24@linux.vnet.ibm.com> wrote: > On Wed, May 08, 2019 at 04:44:47PM +0800, Li Wang wrote: > > For some ppc64le systems, it has non-continuous numa nodes in > > hardware configuration. So we're hitting the below warnings while > > running set_mempolicy tests on that. To fix this issue, let's just > > remove restrictions on numa node-id in get_mempolicy(). > > > > Error Log > > --------- > > tst_test.c:1096: INFO: Timeout per run is 0h 50m 00s > > tst_numa.c:190: INFO: Found 2 NUMA memory nodes > > set_mempolicy01.c:63: PASS: set_mempolicy(MPOL_BIND) node 0 > > tst_numa.c:26: INFO: Node 0 allocated 16 pages > > tst_numa.c:26: INFO: Node 8 allocated 0 pages > > set_mempolicy01.c:82: PASS: child: Node 0 allocated 16 > > set_mempolicy01.c:63: PASS: set_mempolicy(MPOL_BIND) node 8 > > tst_numa.c:92: WARN: get_mempolicy(...) returned invalid node 8 > > tst_numa.c:92: WARN: get_mempolicy(...) returned invalid node 8 > > tst_numa.c:92: WARN: get_mempolicy(...) returned invalid node 8 > > ... > > tst_numa.c:26: INFO: Node 0 allocated 0 pages > > tst_numa.c:26: INFO: Node 8 allocated 0 pages > > set_mempolicy01.c:86: FAIL: child: Node 8 allocated 0, expected 16 > > > > lscpu > > ----- > > Architecture: ppc64le > > ... > > CPU(s): 128 > > Core(s) per socket: 16 > > Socket(s): 2 > > NUMA node(s): 2 > > Model name: POWER9, altivec supported > > ... > > NUMA node0 CPU(s): 0-63 > > NUMA node8 CPU(s): 64-127 > > > > Signed-off-by: Li Wang <liwang@redhat.com> > > Cc: Cyril Hrubis <chrubis@suse.cz> > > --- > > libs/libltpnuma/tst_numa.c | 5 +++-- > > 1 file changed, 3 insertions(+), 2 deletions(-) > > > > diff --git a/libs/libltpnuma/tst_numa.c b/libs/libltpnuma/tst_numa.c > > index 0ba6daf39..56c8640ff 100644 > > --- a/libs/libltpnuma/tst_numa.c > > +++ b/libs/libltpnuma/tst_numa.c > > @@ -88,8 +88,9 @@ void tst_nodemap_count_pages(struct tst_nodemap *nodes, > > if (ret < 0) > > tst_brk(TBROK | TERRNO, "get_mempolicy() failed"); > > > > - if (node < 0 || (unsigned int)node >= nodes->cnt) { > > - tst_res(TWARN, "get_mempolicy(...) returned > invalid node %i\n", node); > > + if (node < 0) { > > It is better to have check for powerpc and remove restriction as it will > still hold good for other arch. > If the non-continuous numa nodes only appear in ppc machine, that should be necessary to check arch before removing the restriction. then the patch probably like: --- a/libs/libltpnuma/tst_numa.c +++ b/libs/libltpnuma/tst_numa.c @@ -88,7 +88,11 @@ void tst_nodemap_count_pages(struct tst_nodemap *nodes, if (ret < 0) tst_brk(TBROK | TERRNO, "get_mempolicy() failed"); +#if defined (__powerpc64__) || (__powerpc__) + if (node < 0) { +#elif if (node < 0 || (unsigned int)node >= nodes->cnt) { +#endif tst_res(TWARN, "get_mempolicy(...) returned invalid node %i\n", node); continue; }
On Thu, May 09, 2019 at 04:00:40PM +0800, Li Wang wrote: > On Thu, May 9, 2019 at 2:45 PM Balamuruhan S <bala24@linux.vnet.ibm.com> > wrote: > > > On Wed, May 08, 2019 at 04:44:47PM +0800, Li Wang wrote: > > > For some ppc64le systems, it has non-continuous numa nodes in > > > hardware configuration. So we're hitting the below warnings while > > > running set_mempolicy tests on that. To fix this issue, let's just > > > remove restrictions on numa node-id in get_mempolicy(). > > > > > > Error Log > > > --------- > > > tst_test.c:1096: INFO: Timeout per run is 0h 50m 00s > > > tst_numa.c:190: INFO: Found 2 NUMA memory nodes > > > set_mempolicy01.c:63: PASS: set_mempolicy(MPOL_BIND) node 0 > > > tst_numa.c:26: INFO: Node 0 allocated 16 pages > > > tst_numa.c:26: INFO: Node 8 allocated 0 pages > > > set_mempolicy01.c:82: PASS: child: Node 0 allocated 16 > > > set_mempolicy01.c:63: PASS: set_mempolicy(MPOL_BIND) node 8 > > > tst_numa.c:92: WARN: get_mempolicy(...) returned invalid node 8 > > > tst_numa.c:92: WARN: get_mempolicy(...) returned invalid node 8 > > > tst_numa.c:92: WARN: get_mempolicy(...) returned invalid node 8 > > > ... > > > tst_numa.c:26: INFO: Node 0 allocated 0 pages > > > tst_numa.c:26: INFO: Node 8 allocated 0 pages > > > set_mempolicy01.c:86: FAIL: child: Node 8 allocated 0, expected 16 > > > > > > lscpu > > > ----- > > > Architecture: ppc64le > > > ... > > > CPU(s): 128 > > > Core(s) per socket: 16 > > > Socket(s): 2 > > > NUMA node(s): 2 > > > Model name: POWER9, altivec supported > > > ... > > > NUMA node0 CPU(s): 0-63 > > > NUMA node8 CPU(s): 64-127 > > > > > > Signed-off-by: Li Wang <liwang@redhat.com> > > > Cc: Cyril Hrubis <chrubis@suse.cz> > > > --- > > > libs/libltpnuma/tst_numa.c | 5 +++-- > > > 1 file changed, 3 insertions(+), 2 deletions(-) > > > > > > diff --git a/libs/libltpnuma/tst_numa.c b/libs/libltpnuma/tst_numa.c > > > index 0ba6daf39..56c8640ff 100644 > > > --- a/libs/libltpnuma/tst_numa.c > > > +++ b/libs/libltpnuma/tst_numa.c > > > @@ -88,8 +88,9 @@ void tst_nodemap_count_pages(struct tst_nodemap *nodes, > > > if (ret < 0) > > > tst_brk(TBROK | TERRNO, "get_mempolicy() failed"); > > > > > > - if (node < 0 || (unsigned int)node >= nodes->cnt) { > > > - tst_res(TWARN, "get_mempolicy(...) returned > > invalid node %i\n", node); > > > + if (node < 0) { > > > > It is better to have check for powerpc and remove restriction as it will > > still hold good for other arch. > > > > If the non-continuous numa nodes only appear in ppc machine, that should be > necessary to check arch before removing the restriction. > > then the patch probably like: > > --- a/libs/libltpnuma/tst_numa.c > +++ b/libs/libltpnuma/tst_numa.c > @@ -88,7 +88,11 @@ void tst_nodemap_count_pages(struct tst_nodemap *nodes, > if (ret < 0) > tst_brk(TBROK | TERRNO, "get_mempolicy() failed"); > > +#if defined (__powerpc64__) || (__powerpc__) > + if (node < 0) { > +#elif > if (node < 0 || (unsigned int)node >= nodes->cnt) { > +#endif Looks good to me. Thanks! -- Bala > tst_res(TWARN, "get_mempolicy(...) returned invalid > node %i\n", node); > continue; > } > > > -- > Regards, > Li Wang
----- Original Message ----- > On Wed, May 08, 2019 at 04:44:47PM +0800, Li Wang wrote: > > For some ppc64le systems, it has non-continuous numa nodes in > > hardware configuration. So we're hitting the below warnings while > > running set_mempolicy tests on that. To fix this issue, let's just > > remove restrictions on numa node-id in get_mempolicy(). > > > > Error Log > > --------- > > tst_test.c:1096: INFO: Timeout per run is 0h 50m 00s > > tst_numa.c:190: INFO: Found 2 NUMA memory nodes > > set_mempolicy01.c:63: PASS: set_mempolicy(MPOL_BIND) node 0 > > tst_numa.c:26: INFO: Node 0 allocated 16 pages > > tst_numa.c:26: INFO: Node 8 allocated 0 pages > > set_mempolicy01.c:82: PASS: child: Node 0 allocated 16 > > set_mempolicy01.c:63: PASS: set_mempolicy(MPOL_BIND) node 8 > > tst_numa.c:92: WARN: get_mempolicy(...) returned invalid node 8 > > tst_numa.c:92: WARN: get_mempolicy(...) returned invalid node 8 > > tst_numa.c:92: WARN: get_mempolicy(...) returned invalid node 8 > > ... > > tst_numa.c:26: INFO: Node 0 allocated 0 pages > > tst_numa.c:26: INFO: Node 8 allocated 0 pages > > set_mempolicy01.c:86: FAIL: child: Node 8 allocated 0, expected 16 > > > > lscpu > > ----- > > Architecture: ppc64le > > ... > > CPU(s): 128 > > Core(s) per socket: 16 > > Socket(s): 2 > > NUMA node(s): 2 > > Model name: POWER9, altivec supported > > ... > > NUMA node0 CPU(s): 0-63 > > NUMA node8 CPU(s): 64-127 > > > > Signed-off-by: Li Wang <liwang@redhat.com> > > Cc: Cyril Hrubis <chrubis@suse.cz> > > --- > > libs/libltpnuma/tst_numa.c | 5 +++-- > > 1 file changed, 3 insertions(+), 2 deletions(-) > > > > diff --git a/libs/libltpnuma/tst_numa.c b/libs/libltpnuma/tst_numa.c > > index 0ba6daf39..56c8640ff 100644 > > --- a/libs/libltpnuma/tst_numa.c > > +++ b/libs/libltpnuma/tst_numa.c > > @@ -88,8 +88,9 @@ void tst_nodemap_count_pages(struct tst_nodemap *nodes, > > if (ret < 0) > > tst_brk(TBROK | TERRNO, "get_mempolicy() failed"); > > > > - if (node < 0 || (unsigned int)node >= nodes->cnt) { > > - tst_res(TWARN, "get_mempolicy(...) returned invalid node %i\n", node); > > + if (node < 0) { > > It is better to have check for powerpc and remove restriction as it will > still hold good for other arch. Unless memory gets offlined or test harness restricts LTP to run only on some nodes (via numactl). Here's example of former: # uname -m x86_64 # numactl -H available: 8 nodes (0-7) node 0 cpus: 0 4 8 12 node 0 size: 4061 MB node 0 free: 3437 MB node 1 cpus: 16 20 24 28 node 1 size: 0 MB node 1 free: 0 MB node 2 cpus: 1 5 9 13 node 2 size: 4096 MB node 2 free: 3871 MB node 3 cpus: 17 21 25 29 node 3 size: 0 MB node 3 free: 0 MB node 4 cpus: 2 6 10 14 node 4 size: 4096 MB node 4 free: 3568 MB node 5 cpus: 18 22 26 30 node 5 size: 0 MB node 5 free: 0 MB node 6 cpus: 19 23 27 31 node 6 size: 0 MB node 6 free: 0 MB node 7 cpus: 3 7 11 15 node 7 size: 4095 MB node 7 free: 3802 MB tst_test.c:1096: INFO: Timeout per run is 0h 05m 00s tst_numa.c:190: INFO: Found 4 NUMA memory nodes ... tst_numa.c:92: WARN: get_mempolicy(...) returned invalid node 4 tst_numa.c:92: WARN: get_mempolicy(...) returned invalid node 7 > > -- Bala > > > + tst_res(TWARN, > > + "get_mempolicy(...) returned invalid node %i\n", node); > > continue; > > } > > > > -- > > 2.20.1 > > > > > > -- > > Mailing list info: https://lists.linux.it/listinfo/ltp > > > -- > Mailing list info: https://lists.linux.it/listinfo/ltp >
Hi! > It is better to have check for powerpc and remove restriction as it will > still hold good for other arch. I would like to avoid adding arch specific ifdefs to the code, so for the release I've commited the original fix and we can add better checks later on. As I wrote checking that the numa id is <= than maximal used id would be much better and would work in all cases.
diff --git a/libs/libltpnuma/tst_numa.c b/libs/libltpnuma/tst_numa.c index 0ba6daf39..56c8640ff 100644 --- a/libs/libltpnuma/tst_numa.c +++ b/libs/libltpnuma/tst_numa.c @@ -88,8 +88,9 @@ void tst_nodemap_count_pages(struct tst_nodemap *nodes, if (ret < 0) tst_brk(TBROK | TERRNO, "get_mempolicy() failed"); - if (node < 0 || (unsigned int)node >= nodes->cnt) { - tst_res(TWARN, "get_mempolicy(...) returned invalid node %i\n", node); + if (node < 0) { + tst_res(TWARN, + "get_mempolicy(...) returned invalid node %i\n", node); continue; }
For some ppc64le systems, it has non-continuous numa nodes in hardware configuration. So we're hitting the below warnings while running set_mempolicy tests on that. To fix this issue, let's just remove restrictions on numa node-id in get_mempolicy(). Error Log --------- tst_test.c:1096: INFO: Timeout per run is 0h 50m 00s tst_numa.c:190: INFO: Found 2 NUMA memory nodes set_mempolicy01.c:63: PASS: set_mempolicy(MPOL_BIND) node 0 tst_numa.c:26: INFO: Node 0 allocated 16 pages tst_numa.c:26: INFO: Node 8 allocated 0 pages set_mempolicy01.c:82: PASS: child: Node 0 allocated 16 set_mempolicy01.c:63: PASS: set_mempolicy(MPOL_BIND) node 8 tst_numa.c:92: WARN: get_mempolicy(...) returned invalid node 8 tst_numa.c:92: WARN: get_mempolicy(...) returned invalid node 8 tst_numa.c:92: WARN: get_mempolicy(...) returned invalid node 8 ... tst_numa.c:26: INFO: Node 0 allocated 0 pages tst_numa.c:26: INFO: Node 8 allocated 0 pages set_mempolicy01.c:86: FAIL: child: Node 8 allocated 0, expected 16 lscpu ----- Architecture: ppc64le ... CPU(s): 128 Core(s) per socket: 16 Socket(s): 2 NUMA node(s): 2 Model name: POWER9, altivec supported ... NUMA node0 CPU(s): 0-63 NUMA node8 CPU(s): 64-127 Signed-off-by: Li Wang <liwang@redhat.com> Cc: Cyril Hrubis <chrubis@suse.cz> --- libs/libltpnuma/tst_numa.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-)