linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [net-next PATCH] octeontx2-af: Add NIX Errata workaround on CN10K silicon
@ 2023-02-17  5:51 Sai Krishna
  2023-02-19  9:17 ` Leon Romanovsky
  2023-02-20 10:50 ` patchwork-bot+netdevbpf
  0 siblings, 2 replies; 3+ messages in thread
From: Sai Krishna @ 2023-02-17  5:51 UTC (permalink / raw)
  To: davem, edumazet, kuba, pabeni, netdev, linux-kernel, sgoutham, gakula
  Cc: Sai Krishna

From: Geetha sowjanya <gakula@marvell.com>

This patch adds workaround for below 2 HW erratas

1. Due to improper clock gating, NIXRX may free the same
NPA buffer multiple times.. to avoid this, always enable
NIX RX conditional clock.

2. NIX FIFO does not get initialized on reset, if the SMQ
flush is triggered before the first packet is processed, it
will lead to undefined state. The workaround to perform SMQ
flush only if packet count is non-zero in MDQ.

Signed-off-by: Geetha sowjanya <gakula@marvell.com>
Signed-off-by: Sunil Kovvuri Goutham <sgoutham@marvell.com>
Signed-off-by: Sai Krishna <saikrishnag@marvell.com>
---
 .../net/ethernet/marvell/octeontx2/af/rvu.h    |  3 +++
 .../ethernet/marvell/octeontx2/af/rvu_cn10k.c  | 18 ++++++++++++++++++
 .../ethernet/marvell/octeontx2/af/rvu_nix.c    | 10 ++++++++++
 .../ethernet/marvell/octeontx2/af/rvu_reg.h    |  2 ++
 4 files changed, 33 insertions(+)

diff --git a/drivers/net/ethernet/marvell/octeontx2/af/rvu.h b/drivers/net/ethernet/marvell/octeontx2/af/rvu.h
index 5eea2b6cf6bd..389663a13d1d 100644
--- a/drivers/net/ethernet/marvell/octeontx2/af/rvu.h
+++ b/drivers/net/ethernet/marvell/octeontx2/af/rvu.h
@@ -888,6 +888,9 @@ int rvu_cpt_init(struct rvu *rvu);
 int rvu_set_channels_base(struct rvu *rvu);
 void rvu_program_channels(struct rvu *rvu);
 
+/* CN10K NIX */
+void rvu_nix_block_cn10k_init(struct rvu *rvu, struct nix_hw *nix_hw);
+
 /* CN10K RVU - LMT*/
 void rvu_reset_lmt_map_tbl(struct rvu *rvu, u16 pcifunc);
 
diff --git a/drivers/net/ethernet/marvell/octeontx2/af/rvu_cn10k.c b/drivers/net/ethernet/marvell/octeontx2/af/rvu_cn10k.c
index 7dbbc115cde4..4ad9ff025c96 100644
--- a/drivers/net/ethernet/marvell/octeontx2/af/rvu_cn10k.c
+++ b/drivers/net/ethernet/marvell/octeontx2/af/rvu_cn10k.c
@@ -538,3 +538,21 @@ void rvu_program_channels(struct rvu *rvu)
 	rvu_lbk_set_channels(rvu);
 	rvu_rpm_set_channels(rvu);
 }
+
+void rvu_nix_block_cn10k_init(struct rvu *rvu, struct nix_hw *nix_hw)
+{
+	int blkaddr = nix_hw->blkaddr;
+	u64 cfg;
+
+	/* Set AF vWQE timer interval to a LF configurable range of
+	 * 6.4us to 1.632ms.
+	 */
+	rvu_write64(rvu, blkaddr, NIX_AF_VWQE_TIMER, 0x3FULL);
+
+	/* Enable NIX RX stream and global conditional clock to
+	 * avoild multiple free of NPA buffers.
+	 */
+	cfg = rvu_read64(rvu, blkaddr, NIX_AF_CFG);
+	cfg |= BIT_ULL(1) | BIT_ULL(2);
+	rvu_write64(rvu, blkaddr, NIX_AF_CFG, cfg);
+}
diff --git a/drivers/net/ethernet/marvell/octeontx2/af/rvu_nix.c b/drivers/net/ethernet/marvell/octeontx2/af/rvu_nix.c
index 89e94569e74c..26e639e57dae 100644
--- a/drivers/net/ethernet/marvell/octeontx2/af/rvu_nix.c
+++ b/drivers/net/ethernet/marvell/octeontx2/af/rvu_nix.c
@@ -2058,6 +2058,13 @@ static int nix_smq_flush(struct rvu *rvu, int blkaddr,
 	int err, restore_tx_en = 0;
 	u64 cfg;
 
+	if (!is_rvu_otx2(rvu)) {
+		/* Skip SMQ flush if pkt count is zero */
+		cfg = rvu_read64(rvu, blkaddr, NIX_AF_MDQX_IN_MD_COUNT(smq));
+		if (!cfg)
+			return 0;
+	}
+
 	/* enable cgx tx if disabled */
 	if (is_pf_cgxmapped(rvu, pf)) {
 		rvu_get_cgx_lmac_id(rvu->pf2cgxlmac_map[pf], &cgx_id, &lmac_id);
@@ -4309,6 +4316,9 @@ static int rvu_nix_block_init(struct rvu *rvu, struct nix_hw *nix_hw)
 
 	rvu_write64(rvu, blkaddr, NIX_AF_SEB_CFG, cfg);
 
+	if (!is_rvu_otx2(rvu))
+		rvu_nix_block_cn10k_init(rvu, nix_hw);
+
 	if (is_block_implemented(hw, blkaddr)) {
 		err = nix_setup_txschq(rvu, nix_hw, blkaddr);
 		if (err)
diff --git a/drivers/net/ethernet/marvell/octeontx2/af/rvu_reg.h b/drivers/net/ethernet/marvell/octeontx2/af/rvu_reg.h
index 5437bd20c719..1729b22580ce 100644
--- a/drivers/net/ethernet/marvell/octeontx2/af/rvu_reg.h
+++ b/drivers/net/ethernet/marvell/octeontx2/af/rvu_reg.h
@@ -189,6 +189,7 @@
 #define NIX_AF_RX_CFG			(0x00D0)
 #define NIX_AF_AVG_DELAY		(0x00E0)
 #define NIX_AF_CINT_DELAY		(0x00F0)
+#define NIX_AF_VWQE_TIMER		(0x00F8)
 #define NIX_AF_RX_MCAST_BASE		(0x0100)
 #define NIX_AF_RX_MCAST_CFG		(0x0110)
 #define NIX_AF_RX_MCAST_BUF_BASE	(0x0120)
@@ -426,6 +427,7 @@
 #define NIX_AF_RX_NPC_MIRROR_DROP	(0x4730)
 #define NIX_AF_RX_ACTIVE_CYCLES_PCX(a)	(0x4800 | (a) << 16)
 #define NIX_AF_LINKX_CFG(a)		(0x4010 | (a) << 17)
+#define NIX_AF_MDQX_IN_MD_COUNT(a)	(0x14e0 | (a) << 16)
 
 #define NIX_PRIV_AF_INT_CFG		(0x8000000)
 #define NIX_PRIV_LFX_CFG		(0x8000010)
-- 
2.25.1


^ permalink raw reply related	[flat|nested] 3+ messages in thread

* Re: [net-next PATCH] octeontx2-af: Add NIX Errata workaround on CN10K silicon
  2023-02-17  5:51 [net-next PATCH] octeontx2-af: Add NIX Errata workaround on CN10K silicon Sai Krishna
@ 2023-02-19  9:17 ` Leon Romanovsky
  2023-02-20 10:50 ` patchwork-bot+netdevbpf
  1 sibling, 0 replies; 3+ messages in thread
From: Leon Romanovsky @ 2023-02-19  9:17 UTC (permalink / raw)
  To: Sai Krishna
  Cc: davem, edumazet, kuba, pabeni, netdev, linux-kernel, sgoutham, gakula

On Fri, Feb 17, 2023 at 11:21:12AM +0530, Sai Krishna wrote:
> From: Geetha sowjanya <gakula@marvell.com>
> 
> This patch adds workaround for below 2 HW erratas
> 
> 1. Due to improper clock gating, NIXRX may free the same
> NPA buffer multiple times.. to avoid this, always enable
> NIX RX conditional clock.
> 
> 2. NIX FIFO does not get initialized on reset, if the SMQ
> flush is triggered before the first packet is processed, it
> will lead to undefined state. The workaround to perform SMQ
> flush only if packet count is non-zero in MDQ.
> 
> Signed-off-by: Geetha sowjanya <gakula@marvell.com>
> Signed-off-by: Sunil Kovvuri Goutham <sgoutham@marvell.com>
> Signed-off-by: Sai Krishna <saikrishnag@marvell.com>
> ---
>  .../net/ethernet/marvell/octeontx2/af/rvu.h    |  3 +++
>  .../ethernet/marvell/octeontx2/af/rvu_cn10k.c  | 18 ++++++++++++++++++
>  .../ethernet/marvell/octeontx2/af/rvu_nix.c    | 10 ++++++++++
>  .../ethernet/marvell/octeontx2/af/rvu_reg.h    |  2 ++
>  4 files changed, 33 insertions(+)

Just curious, why aren't these erratas coded as PCI quirks?

Thanks

^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: [net-next PATCH] octeontx2-af: Add NIX Errata workaround on CN10K silicon
  2023-02-17  5:51 [net-next PATCH] octeontx2-af: Add NIX Errata workaround on CN10K silicon Sai Krishna
  2023-02-19  9:17 ` Leon Romanovsky
@ 2023-02-20 10:50 ` patchwork-bot+netdevbpf
  1 sibling, 0 replies; 3+ messages in thread
From: patchwork-bot+netdevbpf @ 2023-02-20 10:50 UTC (permalink / raw)
  To: Sai Krishna
  Cc: davem, edumazet, kuba, pabeni, netdev, linux-kernel, sgoutham, gakula

Hello:

This patch was applied to netdev/net-next.git (master)
by David S. Miller <davem@davemloft.net>:

On Fri, 17 Feb 2023 11:21:12 +0530 you wrote:
> From: Geetha sowjanya <gakula@marvell.com>
> 
> This patch adds workaround for below 2 HW erratas
> 
> 1. Due to improper clock gating, NIXRX may free the same
> NPA buffer multiple times.. to avoid this, always enable
> NIX RX conditional clock.
> 
> [...]

Here is the summary with links:
  - [net-next] octeontx2-af: Add NIX Errata workaround on CN10K silicon
    https://git.kernel.org/netdev/net-next/c/933a01ad5997

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html



^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2023-02-20 10:50 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-02-17  5:51 [net-next PATCH] octeontx2-af: Add NIX Errata workaround on CN10K silicon Sai Krishna
2023-02-19  9:17 ` Leon Romanovsky
2023-02-20 10:50 ` patchwork-bot+netdevbpf

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).