On 3/5/24 17:43, Andrew Lunn wrote:
3) From 2), am65_cpsw_alloc_skb() function removed and replaced by
netdev_alloc_skb_ip_align(), as used by the driver before -> res = 506
Conclusion: Here is where the loss comes from.
IOW, My am65_cpsw_alloc_skb() function is not good.
Initially, I mainly created this 'custom' am65_cpsw_alloc_skb() function
because I thought that none of XDP memory models could be used along
with netdev_alloc_skb_ip_align() function. Was I wrong ?
By creating this custom am65_cpsw_alloc_skb(), I also wanted to handle
the way headroom is reserved differently.
What is special about your device? Why would
netdev_alloc_skb_ip_align() not work?
Andrew
Nothing special about my device, I just misunderstood.
Regarding page pool, I now have better performance.
Two things were missing:
- I did not call skb_mark_for_recycle(), so pages were freed instead of
being recycled !
- In page_pool_params, that's better when I specify the "napi" parameter.
Performance improvement is not that impressive, but it's better:
505 Mbits/sec (with page pool) instead of 495 Mbits/sec (without).
There is a ~ 5 Mbits/sec loss due to additional processing in the path, for XDP stuffs.
So, the difference in favor of page pool using is ~ 15 Mbits/sec.
I'll send a v4 soon.
Julien
