Merge git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net-next

Pull networking updates from David Miller:
 "Another merge window, another pull full of stuff:

   1) Support alternative names for network devices, from Jiri Pirko.

   2) Introduce per-netns netdev notifiers, also from Jiri Pirko.

   3) Support MSG_PEEK in vsock/virtio, from Matias Ezequiel Vara
      Larsen.

   4) Allow compiling out the TLS TOE code, from Jakub Kicinski.

   5) Add several new tracepoints to the kTLS code, also from Jakub.

   6) Support set channels ethtool callback in ena driver, from Sameeh
      Jubran.

   7) New SCTP events SCTP_ADDR_ADDED, SCTP_ADDR_REMOVED,
      SCTP_ADDR_MADE_PRIM, and SCTP_SEND_FAILED_EVENT. From Xin Long.

   8) Add XDP support to mvneta driver, from Lorenzo Bianconi.

   9) Lots of netfilter hw offload fixes, cleanups and enhancements,
      from Pablo Neira Ayuso.

  10) PTP support for aquantia chips, from Egor Pomozov.

  11) Add UDP segmentation offload support to igb, ixgbe, and i40e. From
      Josh Hunt.

  12) Add smart nagle to tipc, from Jon Maloy.

  13) Support L2 field rewrite by TC offloads in bnxt_en, from Venkat
      Duvvuru.

  14) Add a flow mask cache to OVS, from Tonghao Zhang.

  15) Add XDP support to ice driver, from Maciej Fijalkowski.

  16) Add AF_XDP support to ice driver, from Krzysztof Kazimierczak.

  17) Support UDP GSO offload in atlantic driver, from Igor Russkikh.

  18) Support it in stmmac driver too, from Jose Abreu.

  19) Support TIPC encryption and auth, from Tuong Lien.

  20) Introduce BPF trampolines, from Alexei Starovoitov.

  21) Make page_pool API more numa friendly, from Saeed Mahameed.

  22) Introduce route hints to ipv4 and ipv6, from Paolo Abeni.

  23) Add UDP segmentation offload to cxgb4, Rahul Lakkireddy"

* git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net-next: (1857 commits)
  libbpf: Fix usage of u32 in userspace code
  mm: Implement no-MMU variant of vmalloc_user_node_flags
  slip: Fix use-after-free Read in slip_open
  net: dsa: sja1105: fix sja1105_parse_rgmii_delays()
  macvlan: schedule bc_work even if error
  enetc: add support Credit Based Shaper(CBS) for hardware offload
  net: phy: add helpers phy_(un)lock_mdio_bus
  mdio_bus: don't use managed reset-controller
  ax88179_178a: add ethtool_op_get_ts_info()
  mlxsw: spectrum_router: Fix use of uninitialized adjacency index
  mlxsw: spectrum_router: After underlay moves, demote conflicting tunnels
  bpf: Simplify __bpf_arch_text_poke poke type handling
  bpf: Introduce BPF_TRACE_x helper for the tracing tests
  bpf: Add bpf_jit_blinding_enabled for !CONFIG_BPF_JIT
  bpf, testing: Add various tail call test cases
  bpf, x86: Emit patchable direct jump as tail call
  bpf: Constant map key tracking for prog array pokes
  bpf: Add poke dependency tracking for prog array maps
  bpf: Add initial poke descriptor table for jit images
  bpf: Move owner type, jited info into array auxiliary data
  ...

1 files changed, 213 insertions, 23 deletions

diff --git a/drivers/net/wireless/realtek/rtw88/pci.c b/drivers/net/wireless/realtek/rtw88/pci.c
index d90928be663b..a58e8276a41a 100644
--- a/drivers/net/wireless/realtek/rtw88/pci.c
+++ b/drivers/net/wireless/realtek/rtw88/pci.c
@@ -9,6 +9,7 @@
 #include "tx.h"
 #include "rx.h"
 #include "fw.h"
+#include "ps.h"
 #include "debug.h"
 
 static bool rtw_disable_msi;
@@ -457,9 +458,9 @@ static void rtw_pci_reset_buf_desc(struct rtw_dev *rtwdev)
 	/* reset read/write point */
 	rtw_write32(rtwdev, RTK_PCI_TXBD_RWPTR_CLR, 0xffffffff);
 
-	/* rest H2C Queue index */
-	rtw_write32_set(rtwdev, RTK_PCI_TXBD_H2CQ_CSR, BIT_CLR_H2CQ_HOST_IDX);
-	rtw_write32_set(rtwdev, RTK_PCI_TXBD_H2CQ_CSR, BIT_CLR_H2CQ_HW_IDX);
+	/* reset H2C Queue index in a single write */
+	rtw_write32_set(rtwdev, RTK_PCI_TXBD_H2CQ_CSR,
+			BIT_CLR_H2CQ_HOST_IDX | BIT_CLR_H2CQ_HW_IDX);
 }
 
 static void rtw_pci_reset_trx_ring(struct rtw_dev *rtwdev)
@@ -536,6 +537,69 @@ static void rtw_pci_stop(struct rtw_dev *rtwdev)
 	spin_unlock_irqrestore(&rtwpci->irq_lock, flags);
 }
 
+static void rtw_pci_deep_ps_enter(struct rtw_dev *rtwdev)
+{
+	struct rtw_pci *rtwpci = (struct rtw_pci *)rtwdev->priv;
+	struct rtw_pci_tx_ring *tx_ring;
+	bool tx_empty = true;
+	u8 queue;
+
+	lockdep_assert_held(&rtwpci->irq_lock);
+
+	/* Deep PS state is not allowed to TX-DMA */
+	for (queue = 0; queue < RTK_MAX_TX_QUEUE_NUM; queue++) {
+		/* BCN queue is rsvd page, does not have DMA interrupt
+		 * H2C queue is managed by firmware
+		 */
+		if (queue == RTW_TX_QUEUE_BCN ||
+		    queue == RTW_TX_QUEUE_H2C)
+			continue;
+
+		tx_ring = &rtwpci->tx_rings[queue];
+
+		/* check if there is any skb DMAing */
+		if (skb_queue_len(&tx_ring->queue)) {
+			tx_empty = false;
+			break;
+		}
+	}
+
+	if (!tx_empty) {
+		rtw_dbg(rtwdev, RTW_DBG_PS,
+			"TX path not empty, cannot enter deep power save state\n");
+		return;
+	}
+
+	set_bit(RTW_FLAG_LEISURE_PS_DEEP, rtwdev->flags);
+	rtw_power_mode_change(rtwdev, true);
+}
+
+static void rtw_pci_deep_ps_leave(struct rtw_dev *rtwdev)
+{
+	struct rtw_pci *rtwpci = (struct rtw_pci *)rtwdev->priv;
+
+	lockdep_assert_held(&rtwpci->irq_lock);
+
+	if (test_and_clear_bit(RTW_FLAG_LEISURE_PS_DEEP, rtwdev->flags))
+		rtw_power_mode_change(rtwdev, false);
+}
+
+static void rtw_pci_deep_ps(struct rtw_dev *rtwdev, bool enter)
+{
+	struct rtw_pci *rtwpci = (struct rtw_pci *)rtwdev->priv;
+	unsigned long flags;
+
+	spin_lock_irqsave(&rtwpci->irq_lock, flags);
+
+	if (enter && !test_bit(RTW_FLAG_LEISURE_PS_DEEP, rtwdev->flags))
+		rtw_pci_deep_ps_enter(rtwdev);
+
+	if (!enter && test_bit(RTW_FLAG_LEISURE_PS_DEEP, rtwdev->flags))
+		rtw_pci_deep_ps_leave(rtwdev);
+
+	spin_unlock_irqrestore(&rtwpci->irq_lock, flags);
+}
+
 static u8 ac_to_hwq[] = {
 	[IEEE80211_AC_VO] = RTW_TX_QUEUE_VO,
 	[IEEE80211_AC_VI] = RTW_TX_QUEUE_VI,
@@ -616,6 +680,7 @@ static int rtw_pci_xmit(struct rtw_dev *rtwdev,
 	u8 *pkt_desc;
 	struct rtw_pci_tx_buffer_desc *buf_desc;
 	u32 bd_idx;
+	unsigned long flags;
 
 	ring = &rtwpci->tx_rings[queue];
 
@@ -651,6 +716,10 @@ static int rtw_pci_xmit(struct rtw_dev *rtwdev,
 	tx_data = rtw_pci_get_tx_data(skb);
 	tx_data->dma = dma;
 	tx_data->sn = pkt_info->sn;
+
+	spin_lock_irqsave(&rtwpci->irq_lock, flags);
+
+	rtw_pci_deep_ps_leave(rtwdev);
 	skb_queue_tail(&ring->queue, skb);
 
 	/* kick off tx queue */
@@ -666,6 +735,7 @@ static int rtw_pci_xmit(struct rtw_dev *rtwdev,
 		reg_bcn_work |= BIT_PCI_BCNQ_FLAG;
 		rtw_write8(rtwdev, RTK_PCI_TXBD_BCN_WORK, reg_bcn_work);
 	}
+	spin_unlock_irqrestore(&rtwpci->irq_lock, flags);
 
 	return 0;
 }
@@ -990,23 +1060,49 @@ static void rtw_pci_io_unmapping(struct rtw_dev *rtwdev,
 static void rtw_dbi_write8(struct rtw_dev *rtwdev, u16 addr, u8 data)
 {
 	u16 write_addr;
-	u16 remainder = addr & 0x3;
+	u16 remainder = addr & ~(BITS_DBI_WREN | BITS_DBI_ADDR_MASK);
 	u8 flag;
-	u8 cnt = 20;
+	u8 cnt;
 
-	write_addr = ((addr & 0x0ffc) | (BIT(0) << (remainder + 12)));
+	write_addr = addr & BITS_DBI_ADDR_MASK;
+	write_addr |= u16_encode_bits(BIT(remainder), BITS_DBI_WREN);
 	rtw_write8(rtwdev, REG_DBI_WDATA_V1 + remainder, data);
 	rtw_write16(rtwdev, REG_DBI_FLAG_V1, write_addr);
-	rtw_write8(rtwdev, REG_DBI_FLAG_V1 + 2, 0x01);
+	rtw_write8(rtwdev, REG_DBI_FLAG_V1 + 2, BIT_DBI_WFLAG >> 16);
+
+	for (cnt = 0; cnt < RTW_PCI_WR_RETRY_CNT; cnt++) {
+		flag = rtw_read8(rtwdev, REG_DBI_FLAG_V1 + 2);
+		if (flag == 0)
+			return;
 
-	flag = rtw_read8(rtwdev, REG_DBI_FLAG_V1 + 2);
-	while (flag && (cnt != 0)) {
 		udelay(10);
+	}
+
+	WARN(flag, "failed to write to DBI register, addr=0x%04x\n", addr);
+}
+
+static int rtw_dbi_read8(struct rtw_dev *rtwdev, u16 addr, u8 *value)
+{
+	u16 read_addr = addr & BITS_DBI_ADDR_MASK;
+	u8 flag;
+	u8 cnt;
+
+	rtw_write16(rtwdev, REG_DBI_FLAG_V1, read_addr);
+	rtw_write8(rtwdev, REG_DBI_FLAG_V1 + 2, BIT_DBI_RFLAG >> 16);
+
+	for (cnt = 0; cnt < RTW_PCI_WR_RETRY_CNT; cnt++) {
 		flag = rtw_read8(rtwdev, REG_DBI_FLAG_V1 + 2);
-		cnt--;
+		if (flag == 0) {
+			read_addr = REG_DBI_RDATA_V1 + (addr & 3);
+			*value = rtw_read8(rtwdev, read_addr);
+			return 0;
+		}
+
+		udelay(10);
 	}
 
-	WARN(flag, "DBI write fail\n");
+	WARN(1, "failed to read DBI register, addr=0x%04x\n", addr);
+	return -EIO;
 }
 
 static void rtw_mdio_write(struct rtw_dev *rtwdev, u8 addr, u16 data, bool g1)
@@ -1017,23 +1113,113 @@ static void rtw_mdio_write(struct rtw_dev *rtwdev, u8 addr, u16 data, bool g1)
 
 	rtw_write16(rtwdev, REG_MDIO_V1, data);
 
-	page = addr < 0x20 ? 0 : 1;
-	page += g1 ? 0 : 2;
-	rtw_write8(rtwdev, REG_PCIE_MIX_CFG, addr & 0x1f);
+	page = addr < RTW_PCI_MDIO_PG_SZ ? 0 : 1;
+	page += g1 ? RTW_PCI_MDIO_PG_OFFS_G1 : RTW_PCI_MDIO_PG_OFFS_G2;
+	rtw_write8(rtwdev, REG_PCIE_MIX_CFG, addr & BITS_MDIO_ADDR_MASK);
 	rtw_write8(rtwdev, REG_PCIE_MIX_CFG + 3, page);
-
 	rtw_write32_mask(rtwdev, REG_PCIE_MIX_CFG, BIT_MDIO_WFLAG_V1, 1);
-	wflag = rtw_read32_mask(rtwdev, REG_PCIE_MIX_CFG, BIT_MDIO_WFLAG_V1);
 
-	cnt = 20;
-	while (wflag && (cnt != 0)) {
-		udelay(10);
+	for (cnt = 0; cnt < RTW_PCI_WR_RETRY_CNT; cnt++) {
 		wflag = rtw_read32_mask(rtwdev, REG_PCIE_MIX_CFG,
 					BIT_MDIO_WFLAG_V1);
-		cnt--;
+		if (wflag == 0)
+			return;
+
+		udelay(10);
+	}
+
+	WARN(wflag, "failed to write to MDIO register, addr=0x%02x\n", addr);
+}
+
+static void rtw_pci_clkreq_set(struct rtw_dev *rtwdev, bool enable)
+{
+	u8 value;
+	int ret;
+
+	ret = rtw_dbi_read8(rtwdev, RTK_PCIE_LINK_CFG, &value);
+	if (ret) {
+		rtw_err(rtwdev, "failed to read CLKREQ_L1, ret=%d", ret);
+		return;
 	}
 
-	WARN(wflag, "MDIO write fail\n");
+	if (enable)
+		value |= BIT_CLKREQ_SW_EN;
+	else
+		value &= ~BIT_CLKREQ_SW_EN;
+
+	rtw_dbi_write8(rtwdev, RTK_PCIE_LINK_CFG, value);
+}
+
+static void rtw_pci_aspm_set(struct rtw_dev *rtwdev, bool enable)
+{
+	u8 value;
+	int ret;
+
+	ret = rtw_dbi_read8(rtwdev, RTK_PCIE_LINK_CFG, &value);
+	if (ret) {
+		rtw_err(rtwdev, "failed to read ASPM, ret=%d", ret);
+		return;
+	}
+
+	if (enable)
+		value |= BIT_L1_SW_EN;
+	else
+		value &= ~BIT_L1_SW_EN;
+
+	rtw_dbi_write8(rtwdev, RTK_PCIE_LINK_CFG, value);
+}
+
+static void rtw_pci_link_ps(struct rtw_dev *rtwdev, bool enter)
+{
+	struct rtw_pci *rtwpci = (struct rtw_pci *)rtwdev->priv;
+
+	/* Like CLKREQ, ASPM is also implemented by two HW modules, and can
+	 * only be enabled when host supports it.
+	 *
+	 * And ASPM mechanism should be enabled when driver/firmware enters
+	 * power save mode, without having heavy traffic. Because we've
+	 * experienced some inter-operability issues that the link tends
+	 * to enter L1 state on the fly even when driver is having high
+	 * throughput. This is probably because the ASPM behavior slightly
+	 * varies from different SOC.
+	 */
+	if (rtwpci->link_ctrl & PCI_EXP_LNKCTL_ASPM_L1)
+		rtw_pci_aspm_set(rtwdev, enter);
+}
+
+static void rtw_pci_link_cfg(struct rtw_dev *rtwdev)
+{
+	struct rtw_pci *rtwpci = (struct rtw_pci *)rtwdev->priv;
+	struct pci_dev *pdev = rtwpci->pdev;
+	u16 link_ctrl;
+	int ret;
+
+	/* Though there is standard PCIE configuration space to set the
+	 * link control register, but by Realtek's design, driver should
+	 * check if host supports CLKREQ/ASPM to enable the HW module.
+	 *
+	 * These functions are implemented by two HW modules associated,
+	 * one is responsible to access PCIE configuration space to
+	 * follow the host settings, and another is in charge of doing
+	 * CLKREQ/ASPM mechanisms, it is default disabled. Because sometimes
+	 * the host does not support it, and due to some reasons or wrong
+	 * settings (ex. CLKREQ# not Bi-Direction), it could lead to device
+	 * loss if HW misbehaves on the link.
+	 *
+	 * Hence it's designed that driver should first check the PCIE
+	 * configuration space is sync'ed and enabled, then driver can turn
+	 * on the other module that is actually working on the mechanism.
+	 */
+	ret = pcie_capability_read_word(pdev, PCI_EXP_LNKCTL, &link_ctrl);
+	if (ret) {
+		rtw_err(rtwdev, "failed to read PCI cap, ret=%d\n", ret);
+		return;
+	}
+
+	if (link_ctrl & PCI_EXP_LNKCTL_CLKREQ_EN)
+		rtw_pci_clkreq_set(rtwdev, true);
+
+	rtwpci->link_ctrl = link_ctrl;
 }
 
 static void rtw_pci_phy_cfg(struct rtw_dev *rtwdev)
@@ -1074,6 +1260,8 @@ static void rtw_pci_phy_cfg(struct rtw_dev *rtwdev)
 		else
 			rtw_dbi_write8(rtwdev, offset, value);
 	}
+
+	rtw_pci_link_cfg(rtwdev);
 }
 
 static int rtw_pci_claim(struct rtw_dev *rtwdev, struct pci_dev *pdev)
@@ -1120,8 +1308,6 @@ static int rtw_pci_setup_resource(struct rtw_dev *rtwdev, struct pci_dev *pdev)
 		goto err_io_unmap;
 	}
 
-	rtw_pci_phy_cfg(rtwdev);
-
 	return 0;
 
 err_io_unmap:
@@ -1142,6 +1328,8 @@ static struct rtw_hci_ops rtw_pci_ops = {
 	.setup = rtw_pci_setup,
 	.start = rtw_pci_start,
 	.stop = rtw_pci_stop,
+	.deep_ps = rtw_pci_deep_ps,
+	.link_ps = rtw_pci_link_ps,
 
 	.read8 = rtw_pci_read8,
 	.read16 = rtw_pci_read16,
@@ -1233,6 +1421,8 @@ static int rtw_pci_probe(struct pci_dev *pdev,
 		goto err_destroy_pci;
 	}
 
+	rtw_pci_phy_cfg(rtwdev);
+
 	ret = rtw_register_hw(rtwdev, hw);
 	if (ret) {
 		rtw_err(rtwdev, "failed to register hw\n");