| Message ID | 3de0792f5088f00d135c835df6c19e63ae95f5d2.1638026251.git.christophe.jaillet@wanadoo.fr (mailing list archive) |
|---|---|
| State | Accepted |
| Delegated to: | Netdev Maintainers |
| Headers | show |
| Series | net: spider_net: Use non-atomic bitmap API when applicable | expand |
Hi Christophe, On 11/27/21 7:18 AM, Christophe JAILLET wrote: > No concurrent access is possible when a bitmap is local to a function. > So prefer the non-atomic functions to save a few cycles. > - replace a 'for' loop by an equivalent non-atomic 'bitmap_fill()' call > - use '__set_bit()' > > While at it, clear the 'bitmask' bitmap only when needed. > > Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr> > --- > This patch is *not* compile tested. I don't have the needed cross compiling > tool chain. > --- > drivers/net/ethernet/toshiba/spider_net.c | 12 ++++++------ > 1 file changed, 6 insertions(+), 6 deletions(-) As I mentioned, my tdd-builder Docker image has a gcc-powerpc-linux-gnu cross compiler that can be used to build a ppc64 kernel: https://hub.docker.com/r/glevand/tdd-builder I also have a few helper scripts to run the container and cross compile a kernel: https://github.com/glevand/tdd--docker/blob/master/builder/run-builder.sh https://github.com/glevand/tdd-project/blob/master/scripts/build-linux-kernel.sh I applied your patch to v5.16-rc3 and no spider_net warnings or errors were seen when building with ppc64_defconfig. Thanks for your contribution. Acked-by: Geoff Levand <geoff@infradead.org>
diff --git a/drivers/net/ethernet/toshiba/spider_net.c b/drivers/net/ethernet/toshiba/spider_net.c index f50f9a43d3ea..f47b8358669d 100644 --- a/drivers/net/ethernet/toshiba/spider_net.c +++ b/drivers/net/ethernet/toshiba/spider_net.c @@ -595,24 +595,24 @@ spider_net_set_multi(struct net_device *netdev) int i; u32 reg; struct spider_net_card *card = netdev_priv(netdev); - DECLARE_BITMAP(bitmask, SPIDER_NET_MULTICAST_HASHES) = {}; + DECLARE_BITMAP(bitmask, SPIDER_NET_MULTICAST_HASHES); spider_net_set_promisc(card); if (netdev->flags & IFF_ALLMULTI) { - for (i = 0; i < SPIDER_NET_MULTICAST_HASHES; i++) { - set_bit(i, bitmask); - } + bitmap_fill(bitmask, SPIDER_NET_MULTICAST_HASHES); goto write_hash; } + bitmap_zero(bitmask, SPIDER_NET_MULTICAST_HASHES); + /* well, we know, what the broadcast hash value is: it's xfd hash = spider_net_get_multicast_hash(netdev, netdev->broadcast); */ - set_bit(0xfd, bitmask); + __set_bit(0xfd, bitmask); netdev_for_each_mc_addr(ha, netdev) { hash = spider_net_get_multicast_hash(netdev, ha->addr); - set_bit(hash, bitmask); + __set_bit(hash, bitmask); } write_hash:
No concurrent access is possible when a bitmap is local to a function. So prefer the non-atomic functions to save a few cycles. - replace a 'for' loop by an equivalent non-atomic 'bitmap_fill()' call - use '__set_bit()' While at it, clear the 'bitmask' bitmap only when needed. Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr> --- This patch is *not* compile tested. I don't have the needed cross compiling tool chain. --- drivers/net/ethernet/toshiba/spider_net.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-)