Re: [PATCH net-next v3 02/18] net: Display info about MSG_SPLICE_PAGES memory handling in proc

[Paolo Abeni <pabeni@redhat.com>]

On Tue, 2023-06-20 at 15:53 +0100, David Howells wrote:
> Display information about the memory handling MSG_SPLICE_PAGES does to copy
> slabbed data into page fragments.
> 
> For each CPU that has a cached folio, it displays the folio pfn, the offset
> pointer within the folio and the size of the folio.
> 
> It also displays the number of pages refurbished and the number of pages
> replaced.
> 
> Signed-off-by: David Howells <dhowells@redhat.com>
> cc: Alexander Duyck <alexander.duyck@gmail.com>
> cc: Eric Dumazet <edumazet@google.com>
> cc: "David S. Miller" <davem@davemloft.net>
> cc: David Ahern <dsahern@kernel.org>
> cc: Jakub Kicinski <kuba@kernel.org>
> cc: Paolo Abeni <pabeni@redhat.com>
> cc: Jens Axboe <axboe@kernel.dk>
> cc: Matthew Wilcox <willy@infradead.org>
> cc: Menglong Dong <imagedong@tencent.com>
> cc: netdev@vger.kernel.org
> ---
>  net/core/skbuff.c | 42 +++++++++++++++++++++++++++++++++++++++---
>  1 file changed, 39 insertions(+), 3 deletions(-)
> 
> diff --git a/net/core/skbuff.c b/net/core/skbuff.c
> index d962c93a429d..36605510a76d 100644
> --- a/net/core/skbuff.c
> +++ b/net/core/skbuff.c
> @@ -83,6 +83,7 @@
>  #include <linux/user_namespace.h>
>  #include <linux/indirect_call_wrapper.h>
>  #include <linux/textsearch.h>
> +#include <linux/proc_fs.h>
>  
>  #include "dev.h"
>  #include "sock_destructor.h"
> @@ -6758,6 +6759,7 @@ nodefer:	__kfree_skb(skb);
>  struct skb_splice_frag_cache {
>  	struct folio	*folio;
>  	void		*virt;
> +	unsigned int	fsize;
>  	unsigned int	offset;
>  	/* we maintain a pagecount bias, so that we dont dirty cache line
>  	 * containing page->_refcount every time we allocate a fragment.
> @@ -6767,6 +6769,26 @@ struct skb_splice_frag_cache {
>  };
>  
>  static DEFINE_PER_CPU(struct skb_splice_frag_cache, skb_splice_frag_cache);
> +static atomic_t skb_splice_frag_replaced, skb_splice_frag_refurbished;

(in case we don't agree to restrict this series to just remove
MSG_SENDPAGE_NOTLAST)

Have you considered percpu counters instead of the above atomics?

I think the increments are in not so unlikely code-paths, and the
contention there could possibly hurt performances.

Thanks,

Paolo