> -----Original Message----- > From: David Marchand <david.march...@redhat.com> > Sent: 2023年1月3日 21:32 > To: dev@dpdk.org > Cc: Matz, Olivier <olivier.m...@6wind.com>; ferruh.yi...@amd.com; You, > KaisenX <kaisenx....@intel.com>; zhou...@loongson.cn; Burakov, Anatoly > <anatoly.bura...@intel.com> > Subject: [PATCH v3] malloc: enhance NUMA affinity heuristic > > Trying to allocate memory on the first detected numa node has less chance > to find some memory actually available rather than on the main lcore numa > node (especially when the DPDK application is started only on one numa > node). > > Signed-off-by: David Marchand <david.march...@redhat.com> > --- > Changes since v2: > - add uncommitted local change and fix compilation, > > Changes since v1: > - accomodate for configurations with main lcore running on multiples > physical cores belonging to different numa, > > --- > lib/eal/common/malloc_heap.c | 4 ++++ > 1 file changed, 4 insertions(+) > > diff --git a/lib/eal/common/malloc_heap.c b/lib/eal/common/malloc_heap.c > index d7c410b786..3ee19aee15 100644 > --- a/lib/eal/common/malloc_heap.c > +++ b/lib/eal/common/malloc_heap.c > @@ -717,6 +717,10 @@ malloc_get_numa_socket(void) > return socket_id; > } > > + socket_id = rte_lcore_to_socket_id(rte_get_main_lcore()); > + if (socket_id != (unsigned int)SOCKET_ID_ANY) > + return socket_id; > + > return rte_socket_id_by_idx(0); > } > Because the initial "socket_id" of the interrupt thread is not necessarily "SOCKET_ID_ANY"(-1), in malloc_get_numa_socket(void) function, It may be returned before the content modified by this patch, resulting in the subsequent search for memory being performed on the socket that initializes the unallocated memory. To avoid this, I will submit the V4 version to fix this problem. > -- > 2.39.0
