[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <0004e3d5-0f62-49dc-b51f-5a302006c303@ti.com>
Date: Thu, 29 Feb 2024 14:57:48 +0530
From: Siddharth Vadapalli <s-vadapalli@...com>
To: Andrew Lunn <andrew@...n.ch>
CC: Siddharth Vadapalli <s-vadapalli@...com>, Jiri Pirko <jiri@...nulli.us>,
<davem@...emloft.net>, <edumazet@...gle.com>, <kuba@...nel.org>,
<pabeni@...hat.com>, <rogerq@...nel.org>, <vladimir.oltean@....com>,
<hkallweit1@...il.com>, <dan.carpenter@...aro.org>, <horms@...nel.org>,
<yuehaibing@...wei.com>, <netdev@...r.kernel.org>,
<linux-kernel@...r.kernel.org>, <linux-arm-kernel@...ts.infradead.org>,
<srk@...com>
Subject: Re: [PATCH net-next] net: ethernet: ti: am65-cpsw: Add priv-flag for
Switch VLAN Aware mode
On Wed, Feb 28, 2024 at 02:36:55PM +0100, Andrew Lunn wrote:
> > What if there is no kernel behavior associated with it? How can it be mimicked
> > then?
>
> Simple. Implement the feature in software in the kernel for
> everybody. Then offload it to your hardware.
>
> Your hardware is an accelerator. You use it to accelerate what linux
> can already do. If Linux does not have the feature your accelerator
> has, that accelerator feature goes unused.
Is it acceptable to have a macro in the Ethernet Driver to conditionally
disable/enable the feature (via setting the corresponding bit in the
register)?
The current implementation is:
/* Control register */
writel(AM65_CPSW_CTL_P0_ENABLE | AM65_CPSW_CTL_P0_TX_CRC_REMOVE |
AM65_CPSW_CTL_VLAN_AWARE | AM65_CPSW_CTL_P0_RX_PAD,
common->cpsw_base + AM65_CPSW_REG_CTL);
which sets the "AM65_CPSW_CTL_VLAN_AWARE" bit by default.
Could it be changed to:
#define TI_K3_CPSW_VLAN_AWARE 1
....
/* Control register */
val = AM65_CPSW_CTL_P0_ENABLE | AM65_CPSW_CTL_P0_TX_CRC_REMOVE |
AM65_CPSW_CTL_P0_RX_PAD;
#ifdef TI_K3_CPSW_VLAN_AWARE
val |= AM65_CPSW_CTL_VLAN_AWARE;
#endif
writel(val, common->cpsw_base + AM65_CPSW_REG_CTL);
Since no additional configuration is necessary to disable/enable the
functionality except clearing/setting a bit in a register, I am unsure of
the implementation for the offloading part being suggested. Please let me
know if the above implementation is an acceptable alternative.
Regards,
Siddharth.
Powered by blists - more mailing lists