DPDK patches and discussions
 help / color / mirror / Atom feed
* questions about pinned external buffers
@ 2025-01-09 12:11 Morten Brørup
  0 siblings, 0 replies; only message in thread
From: Morten Brørup @ 2025-01-09 12:11 UTC (permalink / raw)
  To: Shahaf Shuler, Viacheslav Ovsiienko, dev; +Cc: Olivier Matz, thomas

Pinned external buffers were introduced with this patch:
https://git.dpdk.org/dpdk/commit/lib/librte_mbuf/rte_mbuf.h?id=6ef1107ad4c6d4fcb6be627367ee0b97bb13e822

Question 1:
Please confirm that the mbuf's pinned external buffer's refcnt is supposed to be 1 when an mbuf is returned to the mempool?

If so, the below assertion should be updated for completion:

rte_mbuf_raw_free(struct rte_mbuf *m)
{
	RTE_ASSERT(!RTE_MBUF_CLONED(m) &&
-			(!RTE_MBUF_HAS_EXTBUF(m) || RTE_MBUF_HAS_PINNED_EXTBUF(m)));
+			(!RTE_MBUF_HAS_EXTBUF(m) ||
+			(RTE_MBUF_HAS_PINNED_EXTBUF(m) &&
+			rte_mbuf_ext_refcnt_read(m->shinfo) == 1)));
	__rte_mbuf_raw_sanity_check(m);
	rte_mempool_put(m->pool, m);
}

The increased performance cost should be acceptable for debug builds (i.e. with assertions enabled).

Question 2:
Could this assertion be moved to __rte_mbuf_raw_sanity_check()?

I'm working on a new rte_mbuf_raw_free_bulk() function, for use with RTE_ETH_TX_OFFLOAD_MBUF_FAST_FREE; and if the assertion is not moved to __rte_mbuf_raw_sanity_check(), it needs to be copy-pasted into the new raw_free_bulk() function too.


^ permalink raw reply	[flat|nested] only message in thread

only message in thread, other threads:[~2025-01-09 12:11 UTC | newest]

Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2025-01-09 12:11 questions about pinned external buffers Morten Brørup

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).