[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <3cb5f213-3c06-46c5-a314-ce9fb5b1d175@tuxon.dev>
Date: Sat, 26 Jul 2025 15:24:44 +0300
From: "claudiu beznea (tuxon)" <claudiu.beznea@...on.dev>
To: Vineeth Karumanchi <vineeth.karumanchi@....com>,
nicolas.ferre@...rochip.com, andrew+netdev@...n.ch, davem@...emloft.net,
edumazet@...gle.com, kuba@...nel.org, pabeni@...hat.com
Cc: git@....com, netdev@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH net-next 2/6] net: macb: Integrate ENST timing parameters
and hardware unit conversion
Hi, Vineeth,
On 7/22/25 18:41, Vineeth Karumanchi wrote:
> Add Enhanced Network Scheduling and Timing (ENST) support to
> queue infrastructure with speed-dependent timing calculations for
> precise gate control.
>
> Hardware timing unit conversion:
> - Timing values programmed as hardware units based on link speed
> - Conversion formula: time_bytes = time_ns / divisor
> - Speed-specific divisors:
> * 1 Gbps: divisor = 8
> * 100 Mbps: divisor = 80
> * 10 Mbps: divisor = 800
>
> Infrastructure changes:
> - Extend macb_queue structure with ENST timing control registers
> - Add queue_enst_configs structure for per-entry TC configuration storage
> - Map ENST register offsets into existing queue management framework
> - Define ENST_NS_TO_HW_UNITS() macro for automatic speed-based conversion
>
> This enables hardware-native timing programming while abstracting the
> speed-dependent conversions
>
> Signed-off-by: Vineeth Karumanchi <vineeth.karumanchi@....com>
> ---
> drivers/net/ethernet/cadence/macb.h | 32 ++++++++++++++++++++++++
> drivers/net/ethernet/cadence/macb_main.c | 6 +++++
> 2 files changed, 38 insertions(+)
>
> diff --git a/drivers/net/ethernet/cadence/macb.h b/drivers/net/ethernet/cadence/macb.h
> index e456ac65d6c6..ef3995564c5c 100644
> --- a/drivers/net/ethernet/cadence/macb.h
> +++ b/drivers/net/ethernet/cadence/macb.h
> @@ -857,6 +857,16 @@
>
> #define MACB_READ_NSR(bp) macb_readl(bp, NSR)
>
> +/* ENST macros*/
> +#define ENST_NS_TO_HW_UNITS(ns, speed_mbps) \
> + DIV_ROUND_UP((ns) * (speed_mbps), (ENST_TIME_GRANULARITY_NS * 1000))
> +
> +#define ENST_MAX_HW_INTERVAL(speed_mbps) \
> + DIV_ROUND_UP(GENMASK(GEM_ON_TIME_SIZE - 1, 0) * ENST_TIME_GRANULARITY_NS * 1000,\
> + (speed_mbps))
> +
> +#define ENST_MAX_START_TIME_SEC GENMASK(GEM_START_TIME_SEC_SIZE - 1, 0)
These are not used in this patch.
> +
> /* struct macb_dma_desc - Hardware DMA descriptor
> * @addr: DMA address of data buffer
> * @ctrl: Control and status bits
> @@ -1262,6 +1272,11 @@ struct macb_queue {
> unsigned int RBQP;
> unsigned int RBQPH;
>
> + /* ENST register offsets for this queue */
> + unsigned int ENST_START_TIME;
> + unsigned int ENST_ON_TIME;
> + unsigned int ENST_OFF_TIME;
> +
> /* Lock to protect tx_head and tx_tail */
> spinlock_t tx_ptr_lock;
> unsigned int tx_head, tx_tail;
> @@ -1450,4 +1465,21 @@ struct macb_platform_data {
> struct clk *hclk;
> };
>
> +/**
> + * struct queue_enst_configs - Configuration for Enhanced Scheduled Traffic (ENST) queue
> + * @queue_id: Identifier for the queue
> + * @start_time_mask: Bitmask representing the start time for the queue
> + * @on_time_bytes: "on" time nsec expressed in bytes
> + * @off_time_bytes: "off" time nsec expressed in bytes
> + *
> + * This structure holds the configuration parameters for an ENST queue,
> + * used to control time-based transmission scheduling in the MACB driver.
> + */
> +struct queue_enst_configs {
s/queue_enst_config/macb_queue_enst_config
> + u8 queue_id;
Could you please move this above to avoid any padding?
> + u32 start_time_mask;
> + u32 on_time_bytes;
> + u32 off_time_bytes;
> +};
> +
This structure is not used in this patch. Can you please introduce it in the
patch it is used?
> #endif /* _MACB_H */
> diff --git a/drivers/net/ethernet/cadence/macb_main.c b/drivers/net/ethernet/cadence/macb_main.c
> index ce95fad8cedd..ff87d3e1d8a0 100644
> --- a/drivers/net/ethernet/cadence/macb_main.c
> +++ b/drivers/net/ethernet/cadence/macb_main.c
> @@ -4305,6 +4305,9 @@ static int macb_init(struct platform_device *pdev)
> queue->TBQP = GEM_TBQP(hw_q - 1);
> queue->RBQP = GEM_RBQP(hw_q - 1);
> queue->RBQS = GEM_RBQS(hw_q - 1);
> + queue->ENST_START_TIME = GEM_ENST_START_TIME(hw_q);
> + queue->ENST_ON_TIME = GEM_ENST_ON_TIME(hw_q);
> + queue->ENST_OFF_TIME = GEM_ENST_OFF_TIME(hw_q);
You can drop these lines here and move it outside of the if (hw_q) {} else {} block.
> #ifdef CONFIG_ARCH_DMA_ADDR_T_64BIT
> if (bp->hw_dma_cap & HW_DMA_CAP_64B) {
> queue->TBQPH = GEM_TBQPH(hw_q - 1);
> @@ -4319,6 +4322,9 @@ static int macb_init(struct platform_device *pdev)
> queue->IMR = MACB_IMR;
> queue->TBQP = MACB_TBQP;
> queue->RBQP = MACB_RBQP;
> + queue->ENST_START_TIME = GEM_ENST_START_TIME(0);
> + queue->ENST_ON_TIME = GEM_ENST_ON_TIME(0);
> + queue->ENST_OFF_TIME = GEM_ENST_OFF_TIME(0);
With the above suggested change, these lines could be dropped.
On the other hand these lines are used in patch 3. So I would prefer to have
them introduced there.
Thank you,
Claudiu
> #ifdef CONFIG_ARCH_DMA_ADDR_T_64BIT
> if (bp->hw_dma_cap & HW_DMA_CAP_64B) {
> queue->TBQPH = MACB_TBQPH;
Powered by blists - more mailing lists