Add the support for Split Header feature in the RX path and enable it in
XGMAC cores.
This does not impact neither beneficts bandwidth but it does reduces CPU
usage because without the feature all the entire packet is memcpy'ed,
while that with the feature only the header is.
With Split Header disabled 'perf stat -d' gives:
86870.624945 task-clock (msec) # 0.429 CPUs utilized
1073352 context-switches # 0.012 M/sec
1 cpu-migrations # 0.000 K/sec
213 page-faults # 0.002 K/sec
327113872376 cycles # 3.766 GHz (62.53%)
56618161216 instructions # 0.17 insn per cycle (75.06%)
10742205071 branches # 123.658 M/sec (75.36%)
584309242 branch-misses # 5.44% of all branches (75.19%)
17594787965 L1-dcache-loads # 202.540 M/sec (74.88%)
4003773131 L1-dcache-load-misses # 22.76% of all L1-dcache hits (74.89%)
1313301468 LLC-loads # 15.118 M/sec (49.75%)
355906510 LLC-load-misses # 27.10% of all LL-cache hits (49.92%)
With Split Header enabled 'perf stat -d' gives:
49324.456539 task-clock (msec) # 0.245 CPUs utilized
2542387 context-switches # 0.052 M/sec
1 cpu-migrations # 0.000 K/sec
213 page-faults # 0.004 K/sec
177092791469 cycles # 3.590 GHz (62.30%)
68555756017 instructions # 0.39 insn per cycle (75.16%)
12697019382 branches # 257.418 M/sec (74.81%)
442081897 branch-misses # 3.48% of all branches (74.79%)
20337958358 L1-dcache-loads # 412.330 M/sec (75.46%)
3820210140 L1-dcache-load-misses # 18.78% of all L1-dcache hits (75.35%)
1257719198 LLC-loads # 25.499 M/sec (49.73%)
685543923 LLC-load-misses # 54.51% of all LL-cache hits (49.86%)
Changes from v2:
- Reword commit message (Jakub)
Changes from v1:
- Add performance info (David)
- Add misssing dma_sync_single_for_device()
Signed-off-by: Jose Abreu <joabreu@synopsys.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
unsigned int addr64;
unsigned int rssen;
unsigned int vlhash;
+ unsigned int sphen;
};
/* GMAC TX FIFO is 8K, Rx FIFO is 16K */
#define XGMAC_CONFIG_ARPEN BIT(31)
#define XGMAC_CONFIG_GPSL GENMASK(29, 16)
#define XGMAC_CONFIG_GPSL_SHIFT 16
+#define XGMAC_CONFIG_HDSMS GENMASK(14, 12)
+#define XGMAC_CONFIG_HDSMS_SHIFT 12
+#define XGMAC_CONFIG_HDSMS_256 (0x2 << XGMAC_CONFIG_HDSMS_SHIFT)
#define XGMAC_CONFIG_S2KP BIT(11)
#define XGMAC_CONFIG_LM BIT(10)
#define XGMAC_CONFIG_IPC BIT(9)
#define XGMAC_HW_FEATURE1 0x00000120
#define XGMAC_HWFEAT_RSSEN BIT(20)
#define XGMAC_HWFEAT_TSOEN BIT(18)
+#define XGMAC_HWFEAT_SPHEN BIT(17)
#define XGMAC_HWFEAT_ADDR64 GENMASK(15, 14)
#define XGMAC_HWFEAT_TXFIFOSIZE GENMASK(10, 6)
#define XGMAC_HWFEAT_RXFIFOSIZE GENMASK(4, 0)
#define XGMAC_TCEIE BIT(0)
#define XGMAC_DMA_ECC_INT_STATUS 0x0000306c
#define XGMAC_DMA_CH_CONTROL(x) (0x00003100 + (0x80 * (x)))
+#define XGMAC_SPH BIT(24)
#define XGMAC_PBLx8 BIT(16)
#define XGMAC_DMA_CH_TX_CONTROL(x) (0x00003104 + (0x80 * (x)))
#define XGMAC_TxPBL GENMASK(21, 16)
#define XGMAC_TDES3_CIC_SHIFT 16
#define XGMAC_TDES3_TPL GENMASK(17, 0)
#define XGMAC_TDES3_FL GENMASK(14, 0)
+#define XGMAC_RDES2_HL GENMASK(9, 0)
#define XGMAC_RDES3_OWN BIT(31)
#define XGMAC_RDES3_CTXT BIT(30)
#define XGMAC_RDES3_IOC BIT(30)
if (unlikely(rdes3 & XGMAC_RDES3_OWN))
return dma_own;
+ if (unlikely(rdes3 & XGMAC_RDES3_CTXT))
+ return discard_frame;
if (likely(!(rdes3 & XGMAC_RDES3_LD)))
return rx_not_ls;
if (unlikely((rdes3 & XGMAC_RDES3_ES) && (rdes3 & XGMAC_RDES3_LD)))
static void dwxgmac2_set_rx_owner(struct dma_desc *p, int disable_rx_ic)
{
- p->des3 = cpu_to_le32(XGMAC_RDES3_OWN);
+ p->des3 |= cpu_to_le32(XGMAC_RDES3_OWN);
if (!disable_rx_ic)
p->des3 |= cpu_to_le32(XGMAC_RDES3_IOC);
return -EINVAL;
}
+static int dwxgmac2_get_rx_header_len(struct dma_desc *p, unsigned int *len)
+{
+ *len = le32_to_cpu(p->des2) & XGMAC_RDES2_HL;
+ return 0;
+}
+
+static void dwxgmac2_set_sec_addr(struct dma_desc *p, dma_addr_t addr)
+{
+ p->des2 = cpu_to_le32(lower_32_bits(addr));
+ p->des3 = cpu_to_le32(upper_32_bits(addr));
+}
+
const struct stmmac_desc_ops dwxgmac210_desc_ops = {
.tx_status = dwxgmac2_get_tx_status,
.rx_status = dwxgmac2_get_rx_status,
.set_addr = dwxgmac2_set_addr,
.clear = dwxgmac2_clear,
.get_rx_hash = dwxgmac2_get_rx_hash,
+ .get_rx_header_len = dwxgmac2_get_rx_header_len,
+ .set_sec_addr = dwxgmac2_set_sec_addr,
};
hw_cap = readl(ioaddr + XGMAC_HW_FEATURE1);
dma_cap->rssen = (hw_cap & XGMAC_HWFEAT_RSSEN) >> 20;
dma_cap->tsoen = (hw_cap & XGMAC_HWFEAT_TSOEN) >> 18;
+ dma_cap->sphen = (hw_cap & XGMAC_HWFEAT_SPHEN) >> 17;
dma_cap->addr64 = (hw_cap & XGMAC_HWFEAT_ADDR64) >> 14;
switch (dma_cap->addr64) {
writel(value, ioaddr + XGMAC_DMA_CH_RX_CONTROL(chan));
}
+static void dwxgmac2_enable_sph(void __iomem *ioaddr, bool en, u32 chan)
+{
+ u32 value = readl(ioaddr + XGMAC_RX_CONFIG);
+
+ value &= ~XGMAC_CONFIG_HDSMS;
+ value |= XGMAC_CONFIG_HDSMS_256; /* Segment max 256 bytes */
+ writel(value, ioaddr + XGMAC_RX_CONFIG);
+
+ value = readl(ioaddr + XGMAC_DMA_CH_CONTROL(chan));
+ if (en)
+ value |= XGMAC_SPH;
+ else
+ value &= ~XGMAC_SPH;
+ writel(value, ioaddr + XGMAC_DMA_CH_CONTROL(chan));
+}
+
const struct stmmac_dma_ops dwxgmac210_dma_ops = {
.reset = dwxgmac2_dma_reset,
.init = dwxgmac2_dma_init,
.enable_tso = dwxgmac2_enable_tso,
.qmode = dwxgmac2_qmode,
.set_bfsize = dwxgmac2_set_bfsize,
+ .enable_sph = dwxgmac2_enable_sph,
};
/* RSS */
int (*get_rx_hash)(struct dma_desc *p, u32 *hash,
enum pkt_hash_types *type);
+ int (*get_rx_header_len)(struct dma_desc *p, unsigned int *len);
+ void (*set_sec_addr)(struct dma_desc *p, dma_addr_t addr);
};
#define stmmac_init_rx_desc(__priv, __args...) \
stmmac_do_void_callback(__priv, desc, clear, __args)
#define stmmac_get_rx_hash(__priv, __args...) \
stmmac_do_callback(__priv, desc, get_rx_hash, __args)
+#define stmmac_get_rx_header_len(__priv, __args...) \
+ stmmac_do_callback(__priv, desc, get_rx_header_len, __args)
+#define stmmac_set_desc_sec_addr(__priv, __args...) \
+ stmmac_do_void_callback(__priv, desc, set_sec_addr, __args)
struct stmmac_dma_cfg;
struct dma_features;
void (*enable_tso)(void __iomem *ioaddr, bool en, u32 chan);
void (*qmode)(void __iomem *ioaddr, u32 channel, u8 qmode);
void (*set_bfsize)(void __iomem *ioaddr, int bfsize, u32 chan);
+ void (*enable_sph)(void __iomem *ioaddr, bool en, u32 chan);
};
#define stmmac_reset(__priv, __args...) \
stmmac_do_void_callback(__priv, dma, qmode, __args)
#define stmmac_set_dma_bfsize(__priv, __args...) \
stmmac_do_void_callback(__priv, dma, set_bfsize, __args)
+#define stmmac_enable_sph(__priv, __args...) \
+ stmmac_do_void_callback(__priv, dma, enable_sph, __args)
struct mac_device_info;
struct net_device;
struct stmmac_rx_buffer {
struct page *page;
+ struct page *sec_page;
dma_addr_t addr;
+ dma_addr_t sec_addr;
};
struct stmmac_rx_queue {
int hwts_tx_en;
bool tx_path_in_lpi_mode;
bool tso;
+ int sph;
unsigned int dma_buf_sz;
unsigned int rx_copybreak;
if (!buf->page)
return -ENOMEM;
+ if (priv->sph) {
+ buf->sec_page = page_pool_dev_alloc_pages(rx_q->page_pool);
+ if (!buf->sec_page)
+ return -ENOMEM;
+
+ buf->sec_addr = page_pool_get_dma_addr(buf->sec_page);
+ stmmac_set_desc_sec_addr(priv, p, buf->sec_addr);
+ } else {
+ buf->sec_page = NULL;
+ }
+
buf->addr = page_pool_get_dma_addr(buf->page);
stmmac_set_desc_addr(priv, p, buf->addr);
if (priv->dma_buf_sz == BUF_SIZE_16KiB)
if (buf->page)
page_pool_put_page(rx_q->page_pool, buf->page, false);
buf->page = NULL;
+
+ if (buf->sec_page)
+ page_pool_put_page(rx_q->page_pool, buf->sec_page, false);
+ buf->sec_page = NULL;
}
/**
stmmac_enable_tso(priv, priv->ioaddr, 1, chan);
}
+ /* Enable Split Header */
+ if (priv->sph && priv->hw->rx_csum) {
+ for (chan = 0; chan < rx_cnt; chan++)
+ stmmac_enable_sph(priv, priv->ioaddr, 1, chan);
+ }
+
/* Start the ball rolling... */
stmmac_start_all_dma(priv);
break;
}
+ if (priv->sph && !buf->sec_page) {
+ buf->sec_page = page_pool_dev_alloc_pages(rx_q->page_pool);
+ if (!buf->sec_page)
+ break;
+
+ buf->sec_addr = page_pool_get_dma_addr(buf->sec_page);
+
+ dma_sync_single_for_device(priv->device, buf->sec_addr,
+ len, DMA_FROM_DEVICE);
+ }
+
buf->addr = page_pool_get_dma_addr(buf->page);
/* Sync whole allocation to device. This will invalidate old
DMA_FROM_DEVICE);
stmmac_set_desc_addr(priv, p, buf->addr);
+ stmmac_set_desc_sec_addr(priv, p, buf->sec_addr);
stmmac_refill_desc3(priv, rx_q, p);
rx_q->rx_count_frames++;
stmmac_display_ring(priv, rx_head, DMA_RX_SIZE, true);
}
while (count < limit) {
+ unsigned int hlen = 0, prev_len = 0;
enum pkt_hash_types hash_type;
struct stmmac_rx_buffer *buf;
- unsigned int prev_len = 0;
struct dma_desc *np, *p;
+ unsigned int sec_len;
int entry;
u32 hash;
break;
read_again:
+ sec_len = 0;
entry = next_entry;
buf = &rx_q->buf_pool[entry];
np = rx_q->dma_rx + next_entry;
prefetch(np);
+ prefetch(page_address(buf->page));
if (priv->extend_desc)
stmmac_rx_extended_status(priv, &priv->dev->stats,
}
if (!skb) {
+ int ret = stmmac_get_rx_header_len(priv, p, &hlen);
+
+ if (priv->sph && !ret && (hlen > 0)) {
+ sec_len = len;
+ if (!(status & rx_not_ls))
+ sec_len = sec_len - hlen;
+ len = hlen;
+
+ prefetch(page_address(buf->sec_page));
+ }
+
skb = napi_alloc_skb(&ch->rx_napi, len);
if (!skb) {
priv->dev->stats.rx_dropped++;
buf->page = NULL;
}
+ if (sec_len > 0) {
+ dma_sync_single_for_cpu(priv->device, buf->sec_addr,
+ sec_len, DMA_FROM_DEVICE);
+ skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags,
+ buf->sec_page, 0, sec_len,
+ priv->dma_buf_sz);
+
+ len += sec_len;
+
+ /* Data payload appended into SKB */
+ page_pool_release_page(rx_q->page_pool, buf->sec_page);
+ buf->sec_page = NULL;
+ }
+
if (likely(status & rx_not_ls))
goto read_again;
netdev_features_t features)
{
struct stmmac_priv *priv = netdev_priv(netdev);
+ bool sph_en;
+ u32 chan;
/* Keep the COE Type in case of csum is supporting */
if (features & NETIF_F_RXCSUM)
*/
stmmac_rx_ipc(priv, priv->hw);
+ sph_en = (priv->hw->rx_csum > 0) && priv->sph;
+ for (chan = 0; chan < priv->plat->rx_queues_to_use; chan++)
+ stmmac_enable_sph(priv, priv->ioaddr, sph_en, chan);
+
return 0;
}
dev_info(priv->device, "TSO feature enabled\n");
}
+ if (priv->dma_cap.sphen) {
+ ndev->hw_features |= NETIF_F_GRO;
+ priv->sph = true;
+ dev_info(priv->device, "SPH feature enabled\n");
+ }
+
if (priv->dma_cap.addr64) {
ret = dma_set_mask_and_coherent(device,
DMA_BIT_MASK(priv->dma_cap.addr64));