Re: [RFC PATCH 2/2] net: macb: Allocate valid memory for TX and RX BD prefetch

From: Claudiu Beznea
Date: Wed Jul 04 2018 - 04:23:06 EST

Next message: Daniel Lezcano: "Re: [PATCH 0/2] Allwinner A64 timer workaround"
Previous message: Geert Uytterhoeven: "Re: [PATCH v2 3/3] m68k: switch to MEMBLOCK + NO_BOOTMEM"
Next in thread: Harini Katakam: "Re: [RFC PATCH 2/2] net: macb: Allocate valid memory for TX and RX BD prefetch"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

Hi Harini,

Few comments below.

Thank you,
Claudiu Beznea

On 29.06.2018 18:31, Harini Katakam wrote:
> GEM version in ZynqMP and most versions greater than r1p07 supports
> TX and RX BD prefetch. The number of BDs that can be prefetched is a
> HW configurable parameter. For ZynqMP, this parameter is 4.
>
> When GEM DMA is accessing the last BD in the ring, even before the
> BD is processed and the WRAP bit is noticed, it will have prefetched
> BDs outside the BD ring. These will not be processed but it is
> necessary to have accessible memory after the last BD. Especially
> in cases where SMMU is used, memory locations immediately after the
> last BD may not have translation tables triggering HRESP errors. Hence
> always allocate extra BDs to accommodate for prefetch.
> The value of tx/rx bd prefetch for any given SoC version is:
> 2 ^ (corresponding field in design config 10 register).
> (value of this field >= 1)
>
> Added a capability flag so that older IP versions that do not have
> DCFG10 or this prefetch capability are not affected.
>
> Signed-off-by: Harini Katakam <harini.katakam@xxxxxxxxxx>
> ---
> drivers/net/ethernet/cadence/macb.h | 11 +++++++++++
> drivers/net/ethernet/cadence/macb_main.c | 31 +++++++++++++++++++++++++------
> 2 files changed, 36 insertions(+), 6 deletions(-)
>
> diff --git a/drivers/net/ethernet/cadence/macb.h b/drivers/net/ethernet/cadence/macb.h
> index 8665982..b267a7b 100644
> --- a/drivers/net/ethernet/cadence/macb.h
> +++ b/drivers/net/ethernet/cadence/macb.h
> @@ -166,6 +166,7 @@
> #define GEM_DCFG6 0x0294 /* Design Config 6 */
> #define GEM_DCFG7 0x0298 /* Design Config 7 */
> #define GEM_DCFG8 0x029C /* Design Config 8 */
> +#define GEM_DCFG10 0x02A4 /* Design Config 10 */
>
> #define GEM_TXBDCTRL 0x04cc /* TX Buffer Descriptor control register */
> #define GEM_RXBDCTRL 0x04d0 /* RX Buffer Descriptor control register */
> @@ -490,6 +491,12 @@
> #define GEM_SCR2CMP_OFFSET 0
> #define GEM_SCR2CMP_SIZE 8
>
> +/* Bitfields in DCFG10 */
> +#define GEM_TXBD_RDBUFF_OFFSET 12
> +#define GEM_TXBD_RDBUFF_SIZE 4
> +#define GEM_RXBD_RDBUFF_OFFSET 8
> +#define GEM_RXBD_RDBUFF_SIZE 4
> +
> /* Bitfields in TISUBN */
> #define GEM_SUBNSINCR_OFFSET 0
> #define GEM_SUBNSINCR_SIZE 16
> @@ -635,6 +642,7 @@
> #define MACB_CAPS_USRIO_DISABLED 0x00000010
> #define MACB_CAPS_JUMBO 0x00000020
> #define MACB_CAPS_GEM_HAS_PTP 0x00000040
> +#define MACB_CAPS_BD_PREFETCH 0x00000080

Rename it to MACB_CAPS_BD_RD_PREFETCH, since it is about read prefetch.

> #define MACB_CAPS_FIFO_MODE 0x10000000
> #define MACB_CAPS_GIGABIT_MODE_AVAILABLE 0x20000000
> #define MACB_CAPS_SG_DISABLED 0x40000000
> @@ -1203,6 +1211,9 @@ struct macb {
> unsigned int max_tuples;
>
> struct tasklet_struct hresp_err_tasklet;
> +
> + int rx_bd_prefetch;
> + int tx_bd_prefetch;

Since it is about read prefetch I would say to rename these fields properly
to describe this:
int rx_bd_rd_prefetch;
int tx_bd_rd_prefetch;

or something similar as you did with macros.

> };
>
> #ifdef CONFIG_MACB_USE_HWSTAMP
> diff --git a/drivers/net/ethernet/cadence/macb_main.c b/drivers/net/ethernet/cadence/macb_main.c
> index e56ffa9..a7612f6 100644
> --- a/drivers/net/ethernet/cadence/macb_main.c
> +++ b/drivers/net/ethernet/cadence/macb_main.c
> @@ -1811,6 +1811,7 @@ static void macb_free_consistent(struct macb *bp)
> {
> struct macb_queue *queue;
> unsigned int q;
> + int size;
>
> bp->macbgem_ops.mog_free_rx_buffers(bp);
>
> @@ -1818,12 +1819,16 @@ static void macb_free_consistent(struct macb *bp)
> kfree(queue->tx_skb);
> queue->tx_skb = NULL;
> if (queue->tx_ring) {
> - dma_free_coherent(&bp->pdev->dev, TX_RING_BYTES(bp),
> + size = TX_RING_BYTES(bp) +
> + (macb_dma_desc_get_size(bp) * bp->tx_bd_prefetch);
> + dma_free_coherent(&bp->pdev->dev, size,
> queue->tx_ring, queue->tx_ring_dma);
> queue->tx_ring = NULL;
> }
> if (queue->rx_ring) {
> - dma_free_coherent(&bp->pdev->dev, RX_RING_BYTES(bp),
> + size = RX_RING_BYTES(bp) +
> + (macb_dma_desc_get_size(bp) * bp->rx_bd_prefetch);
> + dma_free_coherent(&bp->pdev->dev, size,
> queue->rx_ring, queue->rx_ring_dma);
> queue->rx_ring = NULL;
> }
> @@ -1873,7 +1878,8 @@ static int macb_alloc_consistent(struct macb *bp)
> int size;
>
> for (q = 0, queue = bp->queues; q < bp->num_queues; ++q, ++queue) {
> - size = TX_RING_BYTES(bp);
> + size = TX_RING_BYTES(bp) +
> + (macb_dma_desc_get_size(bp) * bp->tx_bd_prefetch);
> queue->tx_ring = dma_alloc_coherent(&bp->pdev->dev, size,
> &queue->tx_ring_dma,
> GFP_KERNEL);
> @@ -1889,7 +1895,8 @@ static int macb_alloc_consistent(struct macb *bp)
> if (!queue->tx_skb)
> goto out_err;
>
> - size = RX_RING_BYTES(bp);
> + size = RX_RING_BYTES(bp) +
> + (macb_dma_desc_get_size(bp) * bp->rx_bd_prefetch);
> queue->rx_ring = dma_alloc_coherent(&bp->pdev->dev, size,
> &queue->rx_ring_dma, GFP_KERNEL);
> if (!queue->rx_ring)
> @@ -3794,7 +3801,7 @@ static const struct macb_config np4_config = {
> static const struct macb_config zynqmp_config = {
> .caps = MACB_CAPS_GIGABIT_MODE_AVAILABLE |
> MACB_CAPS_JUMBO |
> - MACB_CAPS_GEM_HAS_PTP,
> + MACB_CAPS_GEM_HAS_PTP | MACB_CAPS_BD_PREFETCH,
> .dma_burst_length = 16,
> .clk_init = macb_clk_init,
> .init = macb_init,
> @@ -3855,7 +3862,7 @@ static int macb_probe(struct platform_device *pdev)
> void __iomem *mem;
> const char *mac;
> struct macb *bp;
> - int err;
> + int err, buff;

I would use "val" instead of "buff"

>
> regs = platform_get_resource(pdev, IORESOURCE_MEM, 0);
> mem = devm_ioremap_resource(&pdev->dev, regs);
> @@ -3944,6 +3951,18 @@ static int macb_probe(struct platform_device *pdev)
> else
> dev->max_mtu = ETH_DATA_LEN;
>
> + bp->rx_bd_prefetch = 0;
> + bp->tx_bd_prefetch = 0;

No need for zero init since alloc_etherdev_mq() will allocate with
__GFP_ZERO flag (actually, kvzalloc() will be called)

> + if (bp->caps & MACB_CAPS_BD_PREFETCH) {
> + buff = GEM_BFEXT(RXBD_RDBUFF, gem_readl(bp, DCFG10));
> + if (buff)
> + bp->rx_bd_prefetch = 2 << (buff - 1);
> +
> + buff = GEM_BFEXT(TXBD_RDBUFF, gem_readl(bp, DCFG10));
> + if (buff)
> + bp->tx_bd_prefetch = 2 << (buff - 1);
> + }
> +
> mac = of_get_mac_address(np);
> if (mac) {
> ether_addr_copy(bp->dev->dev_addr, mac);
>

With these you can add:
Reviewed-by: Claudiu Beznea <claudiu.beznea@xxxxxxxxxxxxx>

Next message: Daniel Lezcano: "Re: [PATCH 0/2] Allwinner A64 timer workaround"
Previous message: Geert Uytterhoeven: "Re: [PATCH v2 3/3] m68k: switch to MEMBLOCK + NO_BOOTMEM"
Next in thread: Harini Katakam: "Re: [RFC PATCH 2/2] net: macb: Allocate valid memory for TX and RX BD prefetch"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]