From: Nithin Dabilpuram <ndabilpuram@marvell.com>
To: Nithin Dabilpuram <ndabilpuram@marvell.com>,
Kiran Kumar K <kirankumark@marvell.com>,
Sunil Kumar Kori <skori@marvell.com>,
Satha Rao <skoteshwar@marvell.com>
Cc: <jerinj@marvell.com>, <dev@dpdk.org>
Subject: [PATCH 02/15] net/cnxk: fix data len for first seg with multi seg pkt
Date: Fri, 3 Mar 2023 13:40:00 +0530 [thread overview]
Message-ID: <20230303081013.589868-2-ndabilpuram@marvell.com> (raw)
In-Reply-To: <20230303081013.589868-1-ndabilpuram@marvell.com>
When coming from vector routine, the first seg length is
set to same as pkt len assuming it is a single segment pkt.
Read just the same in cn10k_nix_prepare_mseg() that is called
in case of mbuf-fast-free offload disabled.
In CN9K, clear other data len fields to avoid using stale data.
Fixes: 8520bce63379 ("net/cnxk: rework no-fast-free offload")
Fixes: ae2c2cb60635 ("net/cnxk: avoid command copy from Tx queue")
Signed-off-by: Nithin Dabilpuram <ndabilpuram@marvell.com>
---
drivers/net/cnxk/cn10k_tx.h | 24 +++++++++++++++++++-----
drivers/net/cnxk/cn9k_tx.h | 1 +
2 files changed, 20 insertions(+), 5 deletions(-)
diff --git a/drivers/net/cnxk/cn10k_tx.h b/drivers/net/cnxk/cn10k_tx.h
index d0f7bc22a4..a72a803e10 100644
--- a/drivers/net/cnxk/cn10k_tx.h
+++ b/drivers/net/cnxk/cn10k_tx.h
@@ -1012,9 +1012,13 @@ cn10k_nix_prepare_mseg(struct cn10k_eth_txq *txq,
ol_flags = m->ol_flags;
/* Start from second segment, first segment is already there */
+ dlen = m->data_len;
is_sg2 = 0;
l_sg.u = sg->u;
- len -= l_sg.u & 0xFFFF;
+ /* Clear l_sg.u first seg length that might be stale from vector path */
+ l_sg.u &= ~0xFFFFUL;
+ l_sg.u |= dlen;
+ len -= dlen;
nb_segs = m->nb_segs - 1;
m_next = m->next;
slist = &cmd[3 + off + 1];
@@ -1940,7 +1944,7 @@ cn10k_nix_xmit_pkts_vector(void *tx_queue, uint64_t *ws,
uint64x2_t xtmp128, ytmp128;
uint64x2_t xmask01, xmask23;
uintptr_t c_laddr = laddr;
- uint8_t lnum, shift, loff;
+ uint8_t lnum, shift, loff = 0;
rte_iova_t c_io_addr;
uint64_t sa_base;
union wdata {
@@ -2059,10 +2063,20 @@ cn10k_nix_xmit_pkts_vector(void *tx_queue, uint64_t *ws,
!!(flags & NIX_TX_OFFLOAD_TSTAMP_F);
}
- /* Check if there are enough LMTLINES for this loop */
- if (lnum + 4 > 32) {
+ /* Check if there are enough LMTLINES for this loop.
+ * Consider previous line to be partial.
+ */
+ if (lnum + 4 >= 32) {
uint8_t ldwords_con = 0, lneeded = 0;
- for (j = 0; j < NIX_DESCS_PER_LOOP; j++) {
+
+ if ((loff >> 4) + segdw[0] > 8) {
+ lneeded += 1;
+ ldwords_con = segdw[0];
+ } else {
+ ldwords_con = (loff >> 4) + segdw[0];
+ }
+
+ for (j = 1; j < NIX_DESCS_PER_LOOP; j++) {
ldwords_con += segdw[j];
if (ldwords_con > 8) {
lneeded += 1;
diff --git a/drivers/net/cnxk/cn9k_tx.h b/drivers/net/cnxk/cn9k_tx.h
index 52661a624c..e956c1ad2a 100644
--- a/drivers/net/cnxk/cn9k_tx.h
+++ b/drivers/net/cnxk/cn9k_tx.h
@@ -461,6 +461,7 @@ cn9k_nix_prepare_mseg(struct cn9k_eth_txq *txq,
/* Start from second segment, first segment is already there */
i = 1;
sg_u = sg->u;
+ sg_u &= 0xFC0000000000FFFF;
nb_segs = m->nb_segs - 1;
m_next = m->next;
slist = &cmd[3 + off + 1];
--
2.25.1
next prev parent reply other threads:[~2023-03-03 8:11 UTC|newest]
Thread overview: 16+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-03-03 8:09 [PATCH 01/15] net/cnxk: resolve sefgault caused during transmit completion Nithin Dabilpuram
2023-03-03 8:10 ` Nithin Dabilpuram [this message]
2023-03-03 8:10 ` [PATCH 03/15] net/cnxk: release LBK bpid for after freeing resources Nithin Dabilpuram
2023-03-03 8:10 ` [PATCH 04/15] common/cnxk: add separate inline dev stats API Nithin Dabilpuram
2023-03-03 8:10 ` [PATCH 05/15] common/cnxk: distribute SQ's to sdp channels Nithin Dabilpuram
2023-03-03 8:10 ` [PATCH 06/15] common/cnxk: remove flow control config at queue setup Nithin Dabilpuram
2023-03-03 8:10 ` [PATCH 07/15] common/cnxk: enable 10K B0 support for inline IPsec Nithin Dabilpuram
2023-03-03 8:10 ` [PATCH 08/15] net/cnxk: check flow control config per queue on dev start Nithin Dabilpuram
2023-03-03 8:10 ` [PATCH 09/15] net/cnxk: don't allow PFC configuration on started port Nithin Dabilpuram
2023-03-03 8:10 ` [PATCH 10/15] net/cnxk: aura handle for fastpath Rx queues Nithin Dabilpuram
2023-03-03 8:10 ` [PATCH 11/15] common/cnxk: support of per NIX LF meta aura Nithin Dabilpuram
2023-03-03 8:10 ` [PATCH 12/15] common/cnxk: enable one to one SQ QINT mapping Nithin Dabilpuram
2023-03-03 8:10 ` [PATCH 13/15] common/cnxk: add RSS error messages on mbox failure Nithin Dabilpuram
2023-03-03 8:10 ` [PATCH 14/15] common/cnxk: add memory clobber to steor and ldeor Nithin Dabilpuram
2023-03-03 8:10 ` [PATCH 15/15] common/cnxk: enable SDP channel backpressure to TL4 Nithin Dabilpuram
2023-03-06 9:55 ` Jerin Jacob
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20230303081013.589868-2-ndabilpuram@marvell.com \
--to=ndabilpuram@marvell.com \
--cc=dev@dpdk.org \
--cc=jerinj@marvell.com \
--cc=kirankumark@marvell.com \
--cc=skori@marvell.com \
--cc=skoteshwar@marvell.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
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).