| Message ID | 1657852318-33812-1-git-send-email-liyonglong@chinatelecom.cn (mailing list archive) |
|---|---|
| State | Superseded |
| Delegated to: | Netdev Maintainers |
| Headers | show |
| Series | [v2] net: sort queues in xps maps | expand |
On Fri, 15 Jul 2022 10:31:58 +0800 Yonglong Li wrote: > It is more reasonable that pacekts in the same stream be hashed to the same > tx queue when all tx queue bind with the same CPUs. > > --- > v1 -> v2: > Jakub suggestion: factor out the second loop in __netif_set_xps_queue() - > starting from the "add tx-queue to CPU/rx-queue maps" comment into a helper > --- Please put the changelog under the --- after your s-o-b. git am will cut off the commit message at the first --- it finds which means right now it'll cut off before the changelog, and also discard your s-o-b. > Signed-off-by: Yonglong Li <liyonglong@chinatelecom.cn> > --- > net/core/dev.c | 49 ++++++++++++++++++++++++++++++++----------------- > 1 file changed, 32 insertions(+), 17 deletions(-) > > diff --git a/net/core/dev.c b/net/core/dev.c > index 978ed06..b9ae5d5 100644 > --- a/net/core/dev.c > +++ b/net/core/dev.c > @@ -150,6 +150,7 @@ > #include <linux/pm_runtime.h> > #include <linux/prandom.h> > #include <linux/once_lite.h> > +#include <linux/sort.h> > > #include "dev.h" > #include "net-sysfs.h" > @@ -199,6 +200,11 @@ static int call_netdevice_notifiers_extack(unsigned long val, > > static DECLARE_RWSEM(devnet_rename_sem); > > +static int cmp_u16(const void *a, const void *b) > +{ > + return *(u16 *)a - *(u16 *)b; > +} > + > static inline void dev_base_seq_inc(struct net *net) > { > while (++net->dev_base_seq == 0) > @@ -2537,6 +2543,31 @@ static void xps_copy_dev_maps(struct xps_dev_maps *dev_maps, > } > } > > +static void update_xps_map(struct xps_map *map, int cpu, u16 index, > + bool *skip_tc, int *numa_node_id, > + enum xps_map_type type) > +{ > + int pos = 0; > + > + *skip_tc = true; Maybe let's keep the skip_tc in the main function, output arguments are not universally liked in the kernel, and it's just a single line. Otherwise looks good, thank you!
diff --git a/net/core/dev.c b/net/core/dev.c index 978ed06..b9ae5d5 100644 --- a/net/core/dev.c +++ b/net/core/dev.c @@ -150,6 +150,7 @@ #include <linux/pm_runtime.h> #include <linux/prandom.h> #include <linux/once_lite.h> +#include <linux/sort.h> #include "dev.h" #include "net-sysfs.h" @@ -199,6 +200,11 @@ static int call_netdevice_notifiers_extack(unsigned long val, static DECLARE_RWSEM(devnet_rename_sem); +static int cmp_u16(const void *a, const void *b) +{ + return *(u16 *)a - *(u16 *)b; +} + static inline void dev_base_seq_inc(struct net *net) { while (++net->dev_base_seq == 0) @@ -2537,6 +2543,31 @@ static void xps_copy_dev_maps(struct xps_dev_maps *dev_maps, } } +static void update_xps_map(struct xps_map *map, int cpu, u16 index, + bool *skip_tc, int *numa_node_id, + enum xps_map_type type) +{ + int pos = 0; + + *skip_tc = true; + + while ((pos < map->len) && (map->queues[pos] != index)) + pos++; + + if (pos == map->len) + map->queues[map->len++] = index; + + sort(map->queues, map->len, sizeof(u16), cmp_u16, NULL); +#ifdef CONFIG_NUMA + if (type == XPS_CPUS) { + if (*numa_node_id == -2) + *numa_node_id = cpu_to_node(cpu); + else if (*numa_node_id != cpu_to_node(cpu)) + *numa_node_id = -1; + } +#endif +} + /* Must be called under cpus_read_lock */ int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, u16 index, enum xps_map_type type) @@ -2629,24 +2660,8 @@ int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, if (netif_attr_test_mask(j, mask, nr_ids) && netif_attr_test_online(j, online_mask, nr_ids)) { /* add tx-queue to CPU/rx-queue maps */ - int pos = 0; - - skip_tc = true; - map = xmap_dereference(new_dev_maps->attr_map[tci]); - while ((pos < map->len) && (map->queues[pos] != index)) - pos++; - - if (pos == map->len) - map->queues[map->len++] = index; -#ifdef CONFIG_NUMA - if (type == XPS_CPUS) { - if (numa_node_id == -2) - numa_node_id = cpu_to_node(j); - else if (numa_node_id != cpu_to_node(j)) - numa_node_id = -1; - } -#endif + update_xps_map(map, j, index, &skip_tc, &numa_node_id, type); } if (copy)
in the following case that set xps of each tx-queue with same cpu mask, packets in the same tcp stream may be hash to different tx queue. Because the order of queues in each xps map is not the same. first set each tx-queue with different cpu mask $ echo 0 > /sys/class/net/eth0/queues/tx-0/xps_cpus $ echo 1 > /sys/class/net/eth0/queues/tx-1/xps_cpus $ echo 2 > /sys/class/net/eth0/queues/tx-2/xps_cpus $ echo 4 > /sys/class/net/eth0/queues/tx-3/xps_cpus and then set each tx-queue with same cpu mask $ echo f > /sys/class/net/eth0/queues/tx-0/xps_cpus $ echo f > /sys/class/net/eth0/queues/tx-1/xps_cpus $ echo f > /sys/class/net/eth0/queues/tx-2/xps_cpus $ echo f > /sys/class/net/eth0/queues/tx-3/xps_cpus at this point the order of each map queues is differnet, It will cause packets in the same stream be hashed to diffetent tx queue: attr_map[0].queues = [0,1,2,3] attr_map[1].queues = [1,0,2,3] attr_map[2].queues = [2,0,1,3] attr_map[3].queues = [3,0,1,2] It is more reasonable that pacekts in the same stream be hashed to the same tx queue when all tx queue bind with the same CPUs. --- v1 -> v2: Jakub suggestion: factor out the second loop in __netif_set_xps_queue() - starting from the "add tx-queue to CPU/rx-queue maps" comment into a helper --- Signed-off-by: Yonglong Li <liyonglong@chinatelecom.cn> --- net/core/dev.c | 49 ++++++++++++++++++++++++++++++++----------------- 1 file changed, 32 insertions(+), 17 deletions(-)