Re: [PATCH RFC net-next v2 7/7] net: skbuff: always try to recycle PP pages directly when in softirq
From: Jakub Kicinski <kuba@kernel.org>
Date: 2023-07-19 00:40:49
Also in:
lkml
On Fri, 14 Jul 2023 19:08:52 +0200 Alexander Lobakin wrote:
Suggested-by: Jakub Kicinski <kuba@kernel.org> # in_softirq()
I thought I said something along the lines as "if this is safe you can as well" which falls short of a suggestion, cause I don't think it is safe :)
quoted hunk ↗ jump to hunk
diff --git a/net/core/skbuff.c b/net/core/skbuff.c index fc1470aab5cf..1c22fd33be6c 100644 --- a/net/core/skbuff.c +++ b/net/core/skbuff.c@@ -902,7 +902,7 @@ bool page_pool_return_skb_page(struct page *page, bool napi_safe) * in the same context as the consumer would run, so there's * no possible race. */ - if (napi_safe) { + if (napi_safe || in_softirq()) { const struct napi_struct *napi = READ_ONCE(pp->p.napi); allow_direct = napi &&
What if we got here from netpoll? napi budget was 0, so napi_safe is false, but in_softirq() can be true or false. XDP SKB is a toy, I really don't think 3-4% in XDP SKB warrants the risk here.