Re: [dpdk-dev] [PATCH] net/af_xdp: enable support for unaligned umem chunks

["Loftus, Ciara" <ciara.loftus@intel.com>]

> 
> Hi, Ciara
> 
> Thanks for the patch, the performance number is quite impressive.
> 
> On 08/29, Ciara Loftus wrote:
> >This patch enables the unaligned chunks feature for AF_XDP which allows
> >chunks to be placed at arbitrary places in the umem, as opposed to them
> >being required to be aligned to 2k. This allows for DPDK application
> >mempools to be mapped directly into the umem and in turn enable zero
> >copy transfer between umem and the PMD.
> >
> >This patch replaces the zero copy via external mbuf mechanism
> >introduced in commit e9ff8bb71943 ("net/af_xdp: enable zero copy by
> external mbuf").
> >The pmd_zero copy vdev argument is also removed as now the PMD will
> >auto-detect presence of the unaligned chunks feature and enable it if
> >so and otherwise fall back to copy mode if not detected.
> >
> >When enabled, this feature significantly improves single-core
> >performance of the PMD.
> >
> >Signed-off-by: Ciara Loftus <ciara.loftus@intel.com>
> >Signed-off-by: Kevin Laatz <kevin.laatz@intel.com>
> >---
> > doc/guides/nics/af_xdp.rst             |   1 -
> > doc/guides/rel_notes/release_19_11.rst |   9 +
> > drivers/net/af_xdp/rte_eth_af_xdp.c    | 304 ++++++++++++++++++-------
> > 3 files changed, 231 insertions(+), 83 deletions(-)
> >
> >diff --git a/doc/guides/nics/af_xdp.rst b/doc/guides/nics/af_xdp.rst
> >index ec46f08f0..48dd788ac 100644
> >--- a/doc/guides/nics/af_xdp.rst
> >+++ b/doc/guides/nics/af_xdp.rst
> >@@ -35,7 +35,6 @@ The following options can be provided to set up an
> af_xdp port in DPDK.
> > *   ``iface`` - name of the Kernel interface to attach to (required);
> > *   ``start_queue`` - starting netdev queue id (optional, default 0);
> > *   ``queue_count`` - total netdev queue number (optional, default 1);
> >-*   ``pmd_zero_copy`` - enable zero copy or not (optional, default 0);
> >
> > Prerequisites
> > -------------
> >diff --git a/doc/guides/rel_notes/release_19_11.rst
> >b/doc/guides/rel_notes/release_19_11.rst
> >index 8490d897c..28a8e5372 100644
> >--- a/doc/guides/rel_notes/release_19_11.rst
> >+++ b/doc/guides/rel_notes/release_19_11.rst
> >@@ -56,6 +56,13 @@ New Features
> >      Also, make sure to start the actual text at the margin.
> >
> =========================================================
> >
> >+* **Updated the AF_XDP PMD.**
> >+
> >+  Updated the AF_XDP PMD. The new features include:
> >+
> >+  * Enabled zero copy between application mempools and UMEM by
> enabling the
> >+    XDP_UMEM_UNALIGNED_CHUNKS UMEM flag.
> >+
> 
> Better to document the kernel dependency in the af_xdp.rst.

Will do.

> 
> >
> > Removed Items
> > -------------
> >@@ -69,6 +76,8 @@ Removed Items
> >    Also, make sure to start the actual text at the margin.
> >
> =========================================================
> >
> >+* Removed AF_XDP pmd_zero copy vdev argument. Support is now auto-
> detected.
> >+
> >
> > API Changes
> > -----------
> >diff --git a/drivers/net/af_xdp/rte_eth_af_xdp.c
> >b/drivers/net/af_xdp/rte_eth_af_xdp.c
> >index 41ed5b2af..7956d5778 100644
> >--- a/drivers/net/af_xdp/rte_eth_af_xdp.c
> >+++ b/drivers/net/af_xdp/rte_eth_af_xdp.c
> [snip]
> > reserve_fill_queue(struct xsk_umem_info *umem, uint16_t reserve_size)
> >{
> > 	struct xsk_ring_prod *fq = &umem->fq;
> >-	void *addrs[reserve_size];
> > 	uint32_t idx;
> > 	uint16_t i;
> >+#if defined(XDP_UMEM_UNALIGNED_CHUNK_FLAG)
> >+
> >+	if (unlikely(!xsk_ring_prod__reserve(fq, reserve_size, &idx))) {
> >+		AF_XDP_LOG(DEBUG, "Failed to reserve enough fq
> descs.\n");
> >+		return -1;
> >+	}
> >+
> >+	for (i = 0; i < reserve_size; i++) {
> >+		struct rte_mbuf *mbuf;
> >+		__u64 *fq_addr;
> >+		uint64_t addr;
> >+
> >+		mbuf = rte_pktmbuf_alloc(umem->mb_pool);
> >+		if (unlikely(mbuf == NULL))
> >+			break;
> 
> If this rare case happens, not all of the reserved slots of fq will be filled with
> proper mbuf addr, then we just call xsk_ring_prod__submit(fq,
> reserve_size) to let kernel receive packets on these addrs, something
> unexpected may happen.

Good catch. I'll fix this in the v2.

Thanks!
Ciara

> 
> Thanks,
> Xiaolong
> 
> >+