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

From: Harini Katakam
Date: Wed Jul 04 2018 - 04:45:37 EST


Hi Claudiu,

On Wed, Jul 4, 2018 at 1:52 PM, Claudiu Beznea
<Claudiu.Beznea@xxxxxxxxxxxxx> wrote:
> 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>

Thanks for the review. I'll send a v2 with these changes.

Regards,
Harini