Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
From:   Shijith Thotton <sthotton@marvell.com>
To:     Arnaud Ebalard <arno@natisbad.org>,
        Herbert Xu <herbert@gondor.apana.org.au>,
        Boris Brezillon <bbrezillon@kernel.org>
CC:     Srujana Challa <schalla@marvell.com>,
        <linux-crypto@vger.kernel.org>, <jerinj@marvell.com>,
        <sgoutham@marvell.com>, Shijith Thotton <sthotton@marvell.com>,
        "David S. Miller" <davem@davemloft.net>,
        open list <linux-kernel@vger.kernel.org>
Subject: [PATCH] crypto: octeontx2: increase CPT HW instruction queue length
Date:   Tue, 25 Jan 2022 23:56:24 +0530
Message-ID: <2088adbb37511e84b963fbcd6eff16c62610e2d2.1643134404.git.sthotton@marvell.com>
In-Reply-To: <ab2269cb3ef3049ed0ab73f28be29f6669a06e36.1643134480.git.sthotton@marvell.com>
References: <ab2269cb3ef3049ed0ab73f28be29f6669a06e36.1643134480.git.sthotton@marvell.com>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Content-Type: text/plain
Precedence: bulk

From: Srujana Challa <schalla@marvell.com>

LDWB is getting incorrectly used in HW when
CPT_AF_LF()_PTR_CTL[IQB_LDWB]=1 and CPT instruction queue has less than
320 free entries. So, increase HW instruction queue size by 320 and give
320 entries less for SW/NIX RX as a SW workaround.

Signed-off-by: Srujana Challa <schalla@marvell.com>
Signed-off-by: Shijith Thotton <sthotton@marvell.com>
---
 drivers/crypto/marvell/octeontx2/otx2_cptlf.h | 19 +++++++++++++++----
 1 file changed, 15 insertions(+), 4 deletions(-)

diff --git a/drivers/crypto/marvell/octeontx2/otx2_cptlf.h b/drivers/crypto/marvell/octeontx2/otx2_cptlf.h
index b691b6c1d5c4..4fcaf61a70e3 100644
--- a/drivers/crypto/marvell/octeontx2/otx2_cptlf.h
+++ b/drivers/crypto/marvell/octeontx2/otx2_cptlf.h
@@ -26,12 +26,22 @@
  */
 #define OTX2_CPT_INST_QLEN_MSGS	((OTX2_CPT_SIZE_DIV40 - 1) * 40)
 
+/*
+ * LDWB is getting incorrectly used when IQB_LDWB = 1 and CPT instruction
+ * queue has less than 320 free entries. So, increase HW instruction queue
+ * size by 320 and give 320 entries less for SW/NIX RX as a workaround.
+ */
+#define OTX2_CPT_INST_QLEN_EXTRA_BYTES  (320 * OTX2_CPT_INST_SIZE)
+#define OTX2_CPT_EXTRA_SIZE_DIV40       (320/40)
+
 /* CPT instruction queue length in bytes */
-#define OTX2_CPT_INST_QLEN_BYTES (OTX2_CPT_SIZE_DIV40 * 40 * \
-				  OTX2_CPT_INST_SIZE)
+#define OTX2_CPT_INST_QLEN_BYTES                                               \
+		((OTX2_CPT_SIZE_DIV40 * 40 * OTX2_CPT_INST_SIZE) +             \
+		OTX2_CPT_INST_QLEN_EXTRA_BYTES)
 
 /* CPT instruction group queue length in bytes */
-#define OTX2_CPT_INST_GRP_QLEN_BYTES (OTX2_CPT_SIZE_DIV40 * 16)
+#define OTX2_CPT_INST_GRP_QLEN_BYTES                                           \
+		((OTX2_CPT_SIZE_DIV40 + OTX2_CPT_EXTRA_SIZE_DIV40) * 16)
 
 /* CPT FC length in bytes */
 #define OTX2_CPT_Q_FC_LEN 128
@@ -179,7 +189,8 @@ static inline void otx2_cptlf_do_set_iqueue_size(struct otx2_cptlf_info *lf)
 {
 	union otx2_cptx_lf_q_size lf_q_size = { .u = 0x0 };
 
-	lf_q_size.s.size_div40 = OTX2_CPT_SIZE_DIV40;
+	lf_q_size.s.size_div40 = OTX2_CPT_SIZE_DIV40 +
+				 OTX2_CPT_EXTRA_SIZE_DIV40;
 	otx2_cpt_write64(lf->lfs->reg_base, BLKADDR_CPT0, lf->slot,
 			 OTX2_CPT_LF_Q_SIZE, lf_q_size.u);
 }
-- 
2.25.1