[PATCH net-next v2 3/4] net: ethernet: mtk_eth_soc: add support for in-SoC SRAM

From: Daniel Golle
Date: Mon Aug 21 2023 - 19:30:35 EST


MT7981, MT7986 and MT7988 come with in-SoC SRAM dedicated for Ethernet
DMA rings. Support using the SRAM without breaking existing device tree
bindings, ie. only new SoC starting from MT7988 will have the SRAM
declared as additional resource in device tree. For MT7981 and MT7986
an offset on top of the main I/O base is used.

Signed-off-by: Daniel Golle <daniel@xxxxxxxxxxxxxx>
---
drivers/net/ethernet/mediatek/mtk_eth_soc.c | 88 ++++++++++++++++-----
drivers/net/ethernet/mediatek/mtk_eth_soc.h | 13 ++-
2 files changed, 79 insertions(+), 22 deletions(-)

diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.c b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
index 2482f47313085..ec6a251a0f026 100644
--- a/drivers/net/ethernet/mediatek/mtk_eth_soc.c
+++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
@@ -1135,10 +1135,13 @@ static int mtk_init_fq_dma(struct mtk_eth *eth)
dma_addr_t dma_addr;
int i;

- eth->scratch_ring = dma_alloc_coherent(eth->dma_dev,
- cnt * soc->txrx.txd_size,
- &eth->phy_scratch_ring,
- GFP_KERNEL);
+ if (MTK_HAS_CAPS(eth->soc->caps, MTK_SRAM))
+ eth->scratch_ring = eth->sram_base;
+ else
+ eth->scratch_ring = dma_alloc_coherent(eth->dma_dev,
+ cnt * soc->txrx.txd_size,
+ &eth->phy_scratch_ring,
+ GFP_KERNEL);
if (unlikely(!eth->scratch_ring))
return -ENOMEM;

@@ -2446,8 +2449,14 @@ static int mtk_tx_alloc(struct mtk_eth *eth)
if (!ring->buf)
goto no_tx_mem;

- ring->dma = dma_alloc_coherent(eth->dma_dev, ring_size * sz,
- &ring->phys, GFP_KERNEL);
+ if (MTK_HAS_CAPS(soc->caps, MTK_SRAM)) {
+ ring->dma = eth->sram_base + ring_size * sz;
+ ring->phys = eth->phy_scratch_ring + ring_size * (dma_addr_t)sz;
+ } else {
+ ring->dma = dma_alloc_coherent(eth->dma_dev, ring_size * sz,
+ &ring->phys, GFP_KERNEL);
+ }
+
if (!ring->dma)
goto no_tx_mem;

@@ -2546,8 +2555,7 @@ static void mtk_tx_clean(struct mtk_eth *eth)
kfree(ring->buf);
ring->buf = NULL;
}
-
- if (ring->dma) {
+ if (!MTK_HAS_CAPS(soc->caps, MTK_SRAM) && ring->dma) {
dma_free_coherent(eth->dma_dev,
ring->dma_size * soc->txrx.txd_size,
ring->dma, ring->phys);
@@ -2566,9 +2574,14 @@ static int mtk_rx_alloc(struct mtk_eth *eth, int ring_no, int rx_flag)
{
const struct mtk_reg_map *reg_map = eth->soc->reg_map;
struct mtk_rx_ring *ring;
- int rx_data_len, rx_dma_size;
+ int rx_data_len, rx_dma_size, tx_ring_size;
int i;

+ if (MTK_HAS_CAPS(eth->soc->caps, MTK_QDMA))
+ tx_ring_size = MTK_QDMA_RING_SIZE;
+ else
+ tx_ring_size = MTK_DMA_SIZE;
+
if (rx_flag == MTK_RX_FLAGS_QDMA) {
if (ring_no)
return -EINVAL;
@@ -2603,9 +2616,20 @@ static int mtk_rx_alloc(struct mtk_eth *eth, int ring_no, int rx_flag)
ring->page_pool = pp;
}

- ring->dma = dma_alloc_coherent(eth->dma_dev,
- rx_dma_size * eth->soc->txrx.rxd_size,
- &ring->phys, GFP_KERNEL);
+ if (!MTK_HAS_CAPS(eth->soc->caps, MTK_SRAM) ||
+ rx_flag != MTK_RX_FLAGS_NORMAL) {
+ ring->dma = dma_alloc_coherent(eth->dma_dev,
+ rx_dma_size * eth->soc->txrx.rxd_size,
+ &ring->phys, GFP_KERNEL);
+ } else {
+ struct mtk_tx_ring *tx_ring = &eth->tx_ring;
+
+ ring->dma = tx_ring->dma + tx_ring_size *
+ eth->soc->txrx.txd_size * (ring_no + 1);
+ ring->phys = tx_ring->phys + tx_ring_size *
+ eth->soc->txrx.txd_size * (ring_no + 1);
+ }
+
if (!ring->dma)
return -ENOMEM;

@@ -2690,7 +2714,7 @@ static int mtk_rx_alloc(struct mtk_eth *eth, int ring_no, int rx_flag)
return 0;
}

-static void mtk_rx_clean(struct mtk_eth *eth, struct mtk_rx_ring *ring)
+static void mtk_rx_clean(struct mtk_eth *eth, struct mtk_rx_ring *ring, bool in_sram)
{
int i;

@@ -2713,7 +2737,7 @@ static void mtk_rx_clean(struct mtk_eth *eth, struct mtk_rx_ring *ring)
ring->data = NULL;
}

- if (ring->dma) {
+ if (!in_sram && ring->dma) {
dma_free_coherent(eth->dma_dev,
ring->dma_size * eth->soc->txrx.rxd_size,
ring->dma, ring->phys);
@@ -3073,7 +3097,7 @@ static void mtk_dma_free(struct mtk_eth *eth)
for (i = 0; i < MTK_MAX_DEVS; i++)
if (eth->netdev[i])
netdev_reset_queue(eth->netdev[i]);
- if (eth->scratch_ring) {
+ if (!MTK_HAS_CAPS(soc->caps, MTK_SRAM) && eth->scratch_ring) {
dma_free_coherent(eth->dma_dev,
MTK_QDMA_RING_SIZE * soc->txrx.txd_size,
eth->scratch_ring, eth->phy_scratch_ring);
@@ -3081,13 +3105,13 @@ static void mtk_dma_free(struct mtk_eth *eth)
eth->phy_scratch_ring = 0;
}
mtk_tx_clean(eth);
- mtk_rx_clean(eth, &eth->rx_ring[0]);
- mtk_rx_clean(eth, &eth->rx_ring_qdma);
+ mtk_rx_clean(eth, &eth->rx_ring[0], MTK_HAS_CAPS(soc->caps, MTK_SRAM));
+ mtk_rx_clean(eth, &eth->rx_ring_qdma, false);

if (eth->hwlro) {
mtk_hwlro_rx_uninit(eth);
for (i = 1; i < MTK_MAX_RX_RING_NUM; i++)
- mtk_rx_clean(eth, &eth->rx_ring[i]);
+ mtk_rx_clean(eth, &eth->rx_ring[i], false);
}

kfree(eth->scratch_head);
@@ -4676,7 +4700,7 @@ static int mtk_sgmii_init(struct mtk_eth *eth)

static int mtk_probe(struct platform_device *pdev)
{
- struct resource *res = NULL;
+ struct resource *res = NULL, *res_sram;
struct device_node *mac_np;
struct mtk_eth *eth;
int err, i;
@@ -4696,6 +4720,20 @@ static int mtk_probe(struct platform_device *pdev)
if (MTK_HAS_CAPS(eth->soc->caps, MTK_SOC_MT7628))
eth->ip_align = NET_IP_ALIGN;

+ if (MTK_HAS_CAPS(eth->soc->caps, MTK_SRAM)) {
+ /* SRAM is actual memory and supports transparent access just like DRAM.
+ * Hence we don't require __iomem being set and don't need to use accessor
+ * functions to read from or write to SRAM.
+ */
+ if (mtk_is_netsys_v3_or_greater(eth)) {
+ eth->sram_base = (void __force *)devm_platform_ioremap_resource(pdev, 1);
+ if (IS_ERR(eth->sram_base))
+ return PTR_ERR(eth->sram_base);
+ } else {
+ eth->sram_base = (void __force *)eth->base + MTK_ETH_SRAM_OFFSET;
+ }
+ }
+
spin_lock_init(&eth->page_lock);
spin_lock_init(&eth->tx_irq_lock);
spin_lock_init(&eth->rx_irq_lock);
@@ -4759,6 +4797,18 @@ static int mtk_probe(struct platform_device *pdev)
err = -EINVAL;
goto err_destroy_sgmii;
}
+ if (MTK_HAS_CAPS(eth->soc->caps, MTK_SRAM)) {
+ if (mtk_is_netsys_v3_or_greater(eth)) {
+ res_sram = platform_get_resource(pdev, IORESOURCE_MEM, 1);
+ if (!res_sram) {
+ err = -EINVAL;
+ goto err_destroy_sgmii;
+ }
+ eth->phy_scratch_ring = res_sram->start;
+ } else {
+ eth->phy_scratch_ring = res->start + MTK_ETH_SRAM_OFFSET;
+ }
+ }
}

if (eth->soc->offload_version) {
diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.h b/drivers/net/ethernet/mediatek/mtk_eth_soc.h
index cf9381a3d68b7..7c180aedcc0cd 100644
--- a/drivers/net/ethernet/mediatek/mtk_eth_soc.h
+++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.h
@@ -139,6 +139,9 @@
#define MTK_GDMA_MAC_ADRH(x) ({ typeof(x) _x = (x); (_x == MTK_GMAC3_ID) ? \
0x54C : 0x50C + (_x * 0x1000); })

+/* Internal SRAM offset */
+#define MTK_ETH_SRAM_OFFSET 0x40000
+
/* FE global misc reg*/
#define MTK_FE_GLO_MISC 0x124

@@ -938,6 +941,7 @@ enum mkt_eth_capabilities {
MTK_RSTCTRL_PPE1_BIT,
MTK_RSTCTRL_PPE2_BIT,
MTK_U3_COPHY_V2_BIT,
+ MTK_SRAM_BIT,

/* MUX BITS*/
MTK_ETH_MUX_GDM1_TO_GMAC1_ESW_BIT,
@@ -973,6 +977,7 @@ enum mkt_eth_capabilities {
#define MTK_RSTCTRL_PPE1 BIT_ULL(MTK_RSTCTRL_PPE1_BIT)
#define MTK_RSTCTRL_PPE2 BIT_ULL(MTK_RSTCTRL_PPE2_BIT)
#define MTK_U3_COPHY_V2 BIT_ULL(MTK_U3_COPHY_V2_BIT)
+#define MTK_SRAM BIT_ULL(MTK_SRAM_BIT)

#define MTK_ETH_MUX_GDM1_TO_GMAC1_ESW \
BIT_ULL(MTK_ETH_MUX_GDM1_TO_GMAC1_ESW_BIT)
@@ -1048,14 +1053,14 @@ enum mkt_eth_capabilities {
#define MT7981_CAPS (MTK_GMAC1_SGMII | MTK_GMAC2_SGMII | MTK_GMAC2_GEPHY | \
MTK_MUX_GMAC12_TO_GEPHY_SGMII | MTK_QDMA | \
MTK_MUX_U3_GMAC2_TO_QPHY | MTK_U3_COPHY_V2 | \
- MTK_RSTCTRL_PPE1)
+ MTK_RSTCTRL_PPE1 | MTK_SRAM)

#define MT7986_CAPS (MTK_GMAC1_SGMII | MTK_GMAC2_SGMII | \
MTK_MUX_GMAC12_TO_GEPHY_SGMII | MTK_QDMA | \
- MTK_RSTCTRL_PPE1)
+ MTK_RSTCTRL_PPE1 | MTK_SRAM)

#define MT7988_CAPS (MTK_GDM1_ESW | MTK_QDMA | MTK_RSTCTRL_PPE1 | \
- MTK_RSTCTRL_PPE2)
+ MTK_RSTCTRL_PPE2 | MTK_SRAM)

struct mtk_tx_dma_desc_info {
dma_addr_t addr;
@@ -1170,6 +1175,7 @@ struct mtk_soc_data {
* @dev: The device pointer
* @dev: The device pointer used for dma mapping/alloc
* @base: The mapped register i/o base
+ * @sram_base: The mapped SRAM base
* @page_lock: Make sure that register operations are atomic
* @tx_irq__lock: Make sure that IRQ register operations are atomic
* @rx_irq__lock: Make sure that IRQ register operations are atomic
@@ -1215,6 +1221,7 @@ struct mtk_eth {
struct device *dev;
struct device *dma_dev;
void __iomem *base;
+ void *sram_base;
spinlock_t page_lock;
spinlock_t tx_irq_lock;
spinlock_t rx_irq_lock;
--
2.41.0