// SPDX-License-Identifier: GPL-2.0 /* * Enable PCIe link L0s/L1 state and Clock Power Management * * Copyright (C) 2007 Intel * Copyright (C) Zhang Yanmin (yanmin.zhang@intel.com) * Copyright (C) Shaohua Li (shaohua.li@intel.com) */ #include #include #include #include #include #include #include #include #include #include #include #include "../pci.h" #ifdef MODULE_PARAM_PREFIX #undef MODULE_PARAM_PREFIX #endif #define MODULE_PARAM_PREFIX "pcie_aspm." /* Note: those are not register definitions */ #define ASPM_STATE_L0S_UP (1) /* Upstream direction L0s state */ #define ASPM_STATE_L0S_DW (2) /* Downstream direction L0s state */ #define ASPM_STATE_L1 (4) /* L1 state */ #define ASPM_STATE_L1_1 (8) /* ASPM L1.1 state */ #define ASPM_STATE_L1_2 (0x10) /* ASPM L1.2 state */ #define ASPM_STATE_L1_1_PCIPM (0x20) /* PCI PM L1.1 state */ #define ASPM_STATE_L1_2_PCIPM (0x40) /* PCI PM L1.2 state */ #define ASPM_STATE_L1_SS_PCIPM (ASPM_STATE_L1_1_PCIPM | ASPM_STATE_L1_2_PCIPM) #define ASPM_STATE_L1_2_MASK (ASPM_STATE_L1_2 | ASPM_STATE_L1_2_PCIPM) #define ASPM_STATE_L1SS (ASPM_STATE_L1_1 | ASPM_STATE_L1_1_PCIPM |\ ASPM_STATE_L1_2_MASK) #define ASPM_STATE_L0S (ASPM_STATE_L0S_UP | ASPM_STATE_L0S_DW) #define ASPM_STATE_ALL (ASPM_STATE_L0S | ASPM_STATE_L1 | \ ASPM_STATE_L1SS) struct aspm_latency { u32 l0s; /* L0s latency (nsec) */ u32 l1; /* L1 latency (nsec) */ }; struct pcie_link_state { struct pci_dev *pdev; /* Upstream component of the Link */ struct pci_dev *downstream; /* Downstream component, function 0 */ struct pcie_link_state *root; /* pointer to the root port link */ struct pcie_link_state *parent; /* pointer to the parent Link state */ struct list_head sibling; /* node in link_list */ /* ASPM state */ u32 aspm_support:7; /* Supported ASPM state */ u32 aspm_enabled:7; /* Enabled ASPM state */ u32 aspm_capable:7; /* Capable ASPM state with latency */ u32 aspm_default:7; /* Default ASPM state by BIOS */ u32 aspm_disable:7; /* Disabled ASPM state */ /* Clock PM state */ u32 clkpm_capable:1; /* Clock PM capable? */ u32 clkpm_enabled:1; /* Current Clock PM state */ u32 clkpm_default:1; /* Default Clock PM state by BIOS */ u32 clkpm_disable:1; /* Clock PM disabled */ /* Exit latencies */ struct aspm_latency latency_up; /* Upstream direction exit latency */ struct aspm_latency latency_dw; /* Downstream direction exit latency */ /* * Endpoint acceptable latencies. A pcie downstream port only * has one slot under it, so at most there are 8 functions. */ struct aspm_latency acceptable[8]; }; static int aspm_disabled, aspm_force; static bool aspm_support_enabled = true; static DEFINE_MUTEX(aspm_lock); static LIST_HEAD(link_list); #define POLICY_DEFAULT 0 /* BIOS default setting */ #define POLICY_PERFORMANCE 1 /* high performance */ #define POLICY_POWERSAVE 2 /* high power saving */ #define POLICY_POWER_SUPERSAVE 3 /* possibly even more power saving */ #ifdef CONFIG_PCIEASPM_PERFORMANCE static int aspm_policy = POLICY_PERFORMANCE; #elif defined CONFIG_PCIEASPM_POWERSAVE static int aspm_policy = POLICY_POWERSAVE; #elif defined CONFIG_PCIEASPM_POWER_SUPERSAVE static int aspm_policy = POLICY_POWER_SUPERSAVE; #else static int aspm_policy; #endif static const char *policy_str[] = { [POLICY_DEFAULT] = "default", [POLICY_PERFORMANCE] = "performance", [POLICY_POWERSAVE] = "powersave", [POLICY_POWER_SUPERSAVE] = "powersupersave" }; #define LINK_RETRAIN_TIMEOUT HZ static int policy_to_aspm_state(struct pcie_link_state *link) { switch (aspm_policy) { case POLICY_PERFORMANCE: /* Disable ASPM and Clock PM */ return 0; case POLICY_POWERSAVE: /* Enable ASPM L0s/L1 */ return (ASPM_STATE_L0S | ASPM_STATE_L1); case POLICY_POWER_SUPERSAVE: /* Enable Everything */ return ASPM_STATE_ALL; case POLICY_DEFAULT: return link->aspm_default; } return 0; } static int policy_to_clkpm_state(struct pcie_link_state *link) { switch (aspm_policy) { case POLICY_PERFORMANCE: /* Disable ASPM and Clock PM */ return 0; case POLICY_POWERSAVE: case POLICY_POWER_SUPERSAVE: /* Enable Clock PM */ return 1; case POLICY_DEFAULT: return link->clkpm_default; } return 0; } static void pcie_set_clkpm_nocheck(struct pcie_link_state *link, int enable) { struct pci_dev *child; struct pci_bus *linkbus = link->pdev->subordinate; u32 val = enable ? PCI_EXP_LNKCTL_CLKREQ_EN : 0; list_for_each_entry(child, &linkbus->devices, bus_list) pcie_capability_clear_and_set_word(child, PCI_EXP_LNKCTL, PCI_EXP_LNKCTL_CLKREQ_EN, val); link->clkpm_enabled = !!enable; } static void pcie_set_clkpm(struct pcie_link_state *link, int enable) { /* * Don't enable Clock PM if the link is not Clock PM capable * or Clock PM is disabled */ if (!link->clkpm_capable || link->clkpm_disable) enable = 0; /* Need nothing if the specified equals to current state */ if (link->clkpm_enabled == enable) return; pcie_set_clkpm_nocheck(link, enable); } static void pcie_clkpm_cap_init(struct pcie_link_state *link, int blacklist) { int capable = 1, enabled = 1; u32 reg32; u16 reg16; struct pci_dev *child; struct pci_bus *linkbus = link->pdev->subordinate; /* All functions should have the same cap and state, take the worst */ list_for_each_entry(child, &linkbus->devices, bus_list) { pcie_capability_read_dword(child, PCI_EXP_LNKCAP, ®32); if (!(reg32 & PCI_EXP_LNKCAP_CLKPM)) { capable = 0; enabled = 0; break; } pcie_capability_read_word(child, PCI_EXP_LNKCTL, ®16); if (!(reg16 & PCI_EXP_LNKCTL_CLKREQ_EN)) enabled = 0; } link->clkpm_enabled = enabled; link->clkpm_default = enabled; link->clkpm_capable = capable; link->clkpm_disable = blacklist ? 1 : 0; } static int pcie_wait_for_retrain(struct pci_dev *pdev) { unsigned long end_jiffies; u16 reg16; /* Wait for Link Training to be cleared by hardware */ end_jiffies = jiffies + LINK_RETRAIN_TIMEOUT; do { pcie_capability_read_word(pdev, PCI_EXP_LNKSTA, ®16); if (!(reg16 & PCI_EXP_LNKSTA_LT)) return 0; msleep(1); } while (time_before(jiffies, end_jiffies)); return -ETIMEDOUT; } static int pcie_retrain_link(struct pcie_link_state *link) { struct pci_dev *parent = link->pdev; int rc; u16 reg16; /* * Ensure the updated LNKCTL parameters are used during link * training by checking that there is no ongoing link training to * avoid LTSSM race as recommended in Implementation Note at the * end of PCIe r6.0.1 sec 7.5.3.7. */ rc = pcie_wait_for_retrain(parent); if (rc) return rc; pcie_capability_read_word(parent, PCI_EXP_LNKCTL, ®16); reg16 |= PCI_EXP_LNKCTL_RL; pcie_capability_write_word(parent, PCI_EXP_LNKCTL, reg16); if (parent->clear_retrain_link) { /* * Due to an erratum in some devices the Retrain Link bit * needs to be cleared again manually to allow the link * training to succeed. */ reg16 &= ~PCI_EXP_LNKCTL_RL; pcie_capability_write_word(parent, PCI_EXP_LNKCTL, reg16); } return pcie_wait_for_retrain(parent); } /* * pcie_aspm_configure_common_clock: check if the 2 ends of a link * could use common clock. If they are, configure them to use the * common clock. That will reduce the ASPM state exit latency. */ static void pcie_aspm_configure_common_clock(struct pcie_link_state *link) { int same_clock = 1; u16 reg16, ccc, parent_old_ccc, child_old_ccc[8]; struct pci_dev *child, *parent = link->pdev; struct pci_bus *linkbus = parent->subordinate; /* * All functions of a slot should have the same Slot Clock * Configuration, so just check one function */ child = list_entry(linkbus->devices.next, struct pci_dev, bus_list); BUG_ON(!pci_is_pcie(child)); /* Check downstream component if bit Slot Clock Configuration is 1 */ pcie_capability_read_word(child, PCI_EXP_LNKSTA, ®16); if (!(reg16 & PCI_EXP_LNKSTA_SLC)) same_clock = 0; /* Check upstream component if bit Slot Clock Configuration is 1 */ pcie_capability_read_word(parent, PCI_EXP_LNKSTA, ®16); if (!(reg16 & PCI_EXP_LNKSTA_SLC)) same_clock = 0; /* Port might be already in common clock mode */ pcie_capability_read_word(parent, PCI_EXP_LNKCTL, ®16); parent_old_ccc = reg16 & PCI_EXP_LNKCTL_CCC; if (same_clock && (reg16 & PCI_EXP_LNKCTL_CCC)) { bool consistent = true; list_for_each_entry(child, &linkbus->devices, bus_list) { pcie_capability_read_word(child, PCI_EXP_LNKCTL, ®16); if (!(reg16 & PCI_EXP_LNKCTL_CCC)) { consistent = false; break; } } if (consistent) return; pci_info(parent, "ASPM: current common clock configuration is inconsistent, reconfiguring\n"); } ccc = same_clock ? PCI_EXP_LNKCTL_CCC : 0; /* Configure downstream component, all functions */ list_for_each_entry(child, &linkbus->devices, bus_list) { pcie_capability_read_word(child, PCI_EXP_LNKCTL, ®16); child_old_ccc[PCI_FUNC(child->devfn)] = reg16 & PCI_EXP_LNKCTL_CCC; pcie_capability_clear_and_set_word(child, PCI_EXP_LNKCTL, PCI_EXP_LNKCTL_CCC, ccc); } /* Configure upstream component */ pcie_capability_clear_and_set_word(parent, PCI_EXP_LNKCTL, PCI_EXP_LNKCTL_CCC, ccc); if (pcie_retrain_link(link)) { /* Training failed. Restore common clock configurations */ pci_err(parent, "ASPM: Could not configure common clock\n"); list_for_each_entry(child, &linkbus->devices, bus_list) pcie_capability_clear_and_set_word(child, PCI_EXP_LNKCTL, PCI_EXP_LNKCTL_CCC, child_old_ccc[PCI_FUNC(child->devfn)]); pcie_capability_clear_and_set_word(parent, PCI_EXP_LNKCTL, PCI_EXP_LNKCTL_CCC, parent_old_ccc); } } /* Convert L0s latency encoding to ns */ static u32 calc_l0s_latency(u32 lnkcap) { u32 encoding = (lnkcap & PCI_EXP_LNKCAP_L0SEL) >> 12; if (encoding == 0x7) return (5 * 1000); /* > 4us */ return (64 << encoding); } /* Convert L0s acceptable latency encoding to ns */ static u32 calc_l0s_acceptable(u32 encoding) { if (encoding == 0x7) return -1U; return (64 << encoding); } /* Convert L1 latency encoding to ns */ static u32 calc_l1_latency(u32 lnkcap) { u32 encoding = (lnkcap & PCI_EXP_LNKCAP_L1EL) >> 15; if (encoding == 0x7) return (65 * 1000); /* > 64us */ return (1000 << encoding); } /* Convert L1 acceptable latency encoding to ns */ static u32 calc_l1_acceptable(u32 encoding) { if (encoding == 0x7) return -1U; return (1000 << encoding); } /* Convert L1SS T_pwr encoding to usec */ static u32 calc_l1ss_pwron(struct pci_dev *pdev, u32 scale, u32 val) { switch (scale) { case 0: return val * 2; case 1: return val * 10; case 2: return val * 100; } pci_err(pdev, "%s: Invalid T_PwrOn scale: %u\n", __func__, scale); return 0; } static void encode_l12_threshold(u32 threshold_us, u32 *scale, u32 *value) { u32 threshold_ns = threshold_us * 1000; /* See PCIe r3.1, sec 7.33.3 and sec 6.18 */ if (threshold_ns < 32) { *scale = 0; *value = threshold_ns; } else if (threshold_ns < 1024) { *scale = 1; *value = threshold_ns >> 5; } else if (threshold_ns < 32768) { *scale = 2; *value = threshold_ns >> 10; } else if (threshold_ns < 1048576) { *scale = 3; *value = threshold_ns >> 15; } else if (threshold_ns < 33554432) { *scale = 4; *value = threshold_ns >> 20; } else { *scale = 5; *value = threshold_ns >> 25; } } static void pcie_aspm_check_latency(struct pci_dev *endpoint) { u32 latency, l1_switch_latency = 0; struct aspm_latency *acceptable; struct pcie_link_state *link; /* Device not in D0 doesn't need latency check */ if ((endpoint->current_state != PCI_D0) && (endpoint->current_state != PCI_UNKNOWN)) return; link = endpoint->bus->self->link_state; acceptable = &link->acceptable[PCI_FUNC(endpoint->devfn)]; while (link) { /* Check upstream direction L0s latency */ if ((link->aspm_capable & ASPM_STATE_L0S_UP) && (link->latency_up.l0s > acceptable->l0s)) link->aspm_capable &= ~ASPM_STATE_L0S_UP; /* Check downstream direction L0s latency */ if ((link->aspm_capable & ASPM_STATE_L0S_DW) && (link->latency_dw.l0s > acceptable->l0s)) link->aspm_capable &= ~ASPM_STATE_L0S_DW; /* * Check L1 latency. * Every switch on the path to root complex need 1 * more microsecond for L1. Spec doesn't mention L0s. * * The exit latencies for L1 substates are not advertised * by a device. Since the spec also doesn't mention a way * to determine max latencies introduced by enabling L1 * substates on the components, it is not clear how to do * a L1 substate exit latency check. We assume that the * L1 exit latencies advertised by a device include L1 * substate latencies (and hence do not do any check). */ latency = max_t(u32, link->latency_up.l1, link->latency_dw.l1); if ((link->aspm_capable & ASPM_STATE_L1) && (latency + l1_switch_latency > acceptable->l1)) link->aspm_capable &= ~ASPM_STATE_L1; l1_switch_latency += 1000; link = link->parent; } } /* * The L1 PM substate capability is only implemented in function 0 in a * multi function device. */ static struct pci_dev *pci_function_0(struct pci_bus *linkbus) { struct pci_dev *child; list_for_each_entry(child, &linkbus->devices, bus_list) if (PCI_FUNC(child->devfn) == 0) return child; return NULL; } static void pci_clear_and_set_dword(struct pci_dev *pdev, int pos, u32 clear, u32 set) { u32 val; pci_read_config_dword(pdev, pos, &val); val &= ~clear; val |= set; pci_write_config_dword(pdev, pos, val); } /* Calculate L1.2 PM substate timing parameters */ static void aspm_calc_l1ss_info(struct pcie_link_state *link, u32 parent_l1ss_cap, u32 child_l1ss_cap) { struct pci_dev *child = link->downstream, *parent = link->pdev; u32 val1, val2, scale1, scale2; u32 t_common_mode, t_power_on, l1_2_threshold, scale, value; u32 ctl1 = 0, ctl2 = 0; u32 pctl1, pctl2, cctl1, cctl2; u32 pl1_2_enables, cl1_2_enables; if (!(link->aspm_support & ASPM_STATE_L1_2_MASK)) return; /* Choose the greater of the two Port Common_Mode_Restore_Times */ val1 = (parent_l1ss_cap & PCI_L1SS_CAP_CM_RESTORE_TIME) >> 8; val2 = (child_l1ss_cap & PCI_L1SS_CAP_CM_RESTORE_TIME) >> 8; t_common_mode = max(val1, val2); /* Choose the greater of the two Port T_POWER_ON times */ val1 = (parent_l1ss_cap & PCI_L1SS_CAP_P_PWR_ON_VALUE) >> 19; scale1 = (parent_l1ss_cap & PCI_L1SS_CAP_P_PWR_ON_SCALE) >> 16; val2 = (child_l1ss_cap & PCI_L1SS_CAP_P_PWR_ON_VALUE) >> 19; scale2 = (child_l1ss_cap & PCI_L1SS_CAP_P_PWR_ON_SCALE) >> 16; if (calc_l1ss_pwron(parent, scale1, val1) > calc_l1ss_pwron(child, scale2, val2)) { ctl2 |= scale1 | (val1 << 3); t_power_on = calc_l1ss_pwron(parent, scale1, val1); } else { ctl2 |= scale2 | (val2 << 3); t_power_on = calc_l1ss_pwron(child, scale2, val2); } /* * Set LTR_L1.2_THRESHOLD to the time required to transition the * Link from L0 to L1.2 and back to L0 so we enter L1.2 only if * downstream devices report (via LTR) that they can tolerate at * least that much latency. * * Based on PCIe r3.1, sec 5.5.3.3.1, Figures 5-16 and 5-17, and * Table 5-11. T(POWER_OFF) is at most 2us and T(L1.2) is at * least 4us. */ l1_2_threshold = 2 + 4 + t_common_mode + t_power_on; encode_l12_threshold(l1_2_threshold, &scale, &value); ctl1 |= t_common_mode << 8 | scale << 29 | value << 16; pci_read_config_dword(parent, parent->l1ss + PCI_L1SS_CTL1, &pctl1); pci_read_config_dword(parent, parent->l1ss + PCI_L1SS_CTL2, &pctl2); pci_read_config_dword(child, child->l1ss + PCI_L1SS_CTL1, &cctl1); pci_read_config_dword(child, child->l1ss + PCI_L1SS_CTL2, &cctl2); if (ctl1 == pctl1 && ctl1 == cctl1 && ctl2 == pctl2 && ctl2 == cctl2) return; /* Disable L1.2 while updating. See PCIe r5.0, sec 5.5.4, 7.8.3.3 */ pl1_2_enables = pctl1 & PCI_L1SS_CTL1_L1_2_MASK; cl1_2_enables = cctl1 & PCI_L1SS_CTL1_L1_2_MASK; if (pl1_2_enables || cl1_2_enables) { pci_clear_and_set_dword(child, child->l1ss + PCI_L1SS_CTL1, PCI_L1SS_CTL1_L1_2_MASK, 0); pci_clear_and_set_dword(parent, parent->l1ss + PCI_L1SS_CTL1, PCI_L1SS_CTL1_L1_2_MASK, 0); } /* Program T_POWER_ON times in both ports */ pci_write_config_dword(parent, parent->l1ss + PCI_L1SS_CTL2, ctl2); pci_write_config_dword(child, child->l1ss + PCI_L1SS_CTL2, ctl2); /* Program Common_Mode_Restore_Time in upstream device */ pci_clear_and_set_dword(parent, parent->l1ss + PCI_L1SS_CTL1, PCI_L1SS_CTL1_CM_RESTORE_TIME, ctl1); /* Program LTR_L1.2_THRESHOLD time in both ports */ pci_clear_and_set_dword(parent, parent->l1ss + PCI_L1SS_CTL1, PCI_L1SS_CTL1_LTR_L12_TH_VALUE | PCI_L1SS_CTL1_LTR_L12_TH_SCALE, ctl1); pci_clear_and_set_dword(child, child->l1ss + PCI_L1SS_CTL1, PCI_L1SS_CTL1_LTR_L12_TH_VALUE | PCI_L1SS_CTL1_LTR_L12_TH_SCALE, ctl1); if (pl1_2_enables || cl1_2_enables) { pci_clear_and_set_dword(parent, parent->l1ss + PCI_L1SS_CTL1, 0, pl1_2_enables); pci_clear_and_set_dword(child, child->l1ss + PCI_L1SS_CTL1, 0, cl1_2_enables); } } static void pcie_aspm_cap_init(struct pcie_link_state *link, int blacklist) { struct pci_dev *child = link->downstream, *parent = link->pdev; u32 parent_lnkcap, child_lnkcap; u16 parent_lnkctl, child_lnkctl; u32 parent_l1ss_cap, child_l1ss_cap; u32 parent_l1ss_ctl1 = 0, child_l1ss_ctl1 = 0; struct pci_bus *linkbus = parent->subordinate; if (blacklist) { /* Set enabled/disable so that we will disable ASPM later */ link->aspm_enabled = ASPM_STATE_ALL; link->aspm_disable = ASPM_STATE_ALL; return; } /* * If ASPM not supported, don't mess with the clocks and link, * bail out now. */ pcie_capability_read_dword(parent, PCI_EXP_LNKCAP, &parent_lnkcap); pcie_capability_read_dword(child, PCI_EXP_LNKCAP, &child_lnkcap); if (!(parent_lnkcap & child_lnkcap & PCI_EXP_LNKCAP_ASPMS)) return; /* Configure common clock before checking latencies */ pcie_aspm_configure_common_clock(link); /* * Re-read upstream/downstream components' register state after * clock configuration. L0s & L1 exit latencies in the otherwise * read-only Link Capabilities may change depending on common clock * configuration (PCIe r5.0, sec 7.5.3.6). */ pcie_capability_read_dword(parent, PCI_EXP_LNKCAP, &parent_lnkcap); pcie_capability_read_dword(child, PCI_EXP_LNKCAP, &child_lnkcap); pcie_capability_read_word(parent, PCI_EXP_LNKCTL, &parent_lnkctl); pcie_capability_read_word(child, PCI_EXP_LNKCTL, &child_lnkctl); /* * Setup L0s state * * Note that we must not enable L0s in either direction on a * given link unless components on both sides of the link each * support L0s. */ if (parent_lnkcap & child_lnkcap & PCI_EXP_LNKCAP_ASPM_L0S) link->aspm_support |= ASPM_STATE_L0S; if (child_lnkctl & PCI_EXP_LNKCTL_ASPM_L0S) link->aspm_enabled |= ASPM_STATE_L0S_UP; if (parent_lnkctl & PCI_EXP_LNKCTL_ASPM_L0S) link->aspm_enabled |= ASPM_STATE_L0S_DW; link->latency_up.l0s = calc_l0s_latency(parent_lnkcap); link->latency_dw.l0s = calc_l0s_latency(child_lnkcap); /* Setup L1 state */ if (parent_lnkcap & child_lnkcap & PCI_EXP_LNKCAP_ASPM_L1) link->aspm_support |= ASPM_STATE_L1; if (parent_lnkctl & child_lnkctl & PCI_EXP_LNKCTL_ASPM_L1) link->aspm_enabled |= ASPM_STATE_L1; link->latency_up.l1 = calc_l1_latency(parent_lnkcap); link->latency_dw.l1 = calc_l1_latency(child_lnkcap); /* Setup L1 substate */ pci_read_config_dword(parent, parent->l1ss + PCI_L1SS_CAP, &parent_l1ss_cap); pci_read_config_dword(child, child->l1ss + PCI_L1SS_CAP, &child_l1ss_cap); if (!(parent_l1ss_cap & PCI_L1SS_CAP_L1_PM_SS)) parent_l1ss_cap = 0; if (!(child_l1ss_cap & PCI_L1SS_CAP_L1_PM_SS)) child_l1ss_cap = 0; /* * If we don't have LTR for the entire path from the Root Complex * to this device, we can't use ASPM L1.2 because it relies on the * LTR_L1.2_THRESHOLD. See PCIe r4.0, secs 5.5.4, 6.18. */ if (!child->ltr_path) child_l1ss_cap &= ~PCI_L1SS_CAP_ASPM_L1_2; if (parent_l1ss_cap & child_l1ss_cap & PCI_L1SS_CAP_ASPM_L1_1) link->aspm_support |= ASPM_STATE_L1_1; if (parent_l1ss_cap & child_l1ss_cap & PCI_L1SS_CAP_ASPM_L1_2) link->aspm_support |= ASPM_STATE_L1_2; if (parent_l1ss_cap & child_l1ss_cap & PCI_L1SS_CAP_PCIPM_L1_1) link->aspm_support |= ASPM_STATE_L1_1_PCIPM; if (parent_l1ss_cap & child_l1ss_cap & PCI_L1SS_CAP_PCIPM_L1_2) link->aspm_support |= ASPM_STATE_L1_2_PCIPM; if (parent_l1ss_cap) pci_read_config_dword(parent, parent->l1ss + PCI_L1SS_CTL1, &parent_l1ss_ctl1); if (child_l1ss_cap) pci_read_config_dword(child, child->l1ss + PCI_L1SS_CTL1, &child_l1ss_ctl1); if (parent_l1ss_ctl1 & child_l1ss_ctl1 & PCI_L1SS_CTL1_ASPM_L1_1) link->aspm_enabled |= ASPM_STATE_L1_1; if (parent_l1ss_ctl1 & child_l1ss_ctl1 & PCI_L1SS_CTL1_ASPM_L1_2) link->aspm_enabled |= ASPM_STATE_L1_2; if (parent_l1ss_ctl1 & child_l1ss_ctl1 & PCI_L1SS_CTL1_PCIPM_L1_1) link->aspm_enabled |= ASPM_STATE_L1_1_PCIPM; if (parent_l1ss_ctl1 & child_l1ss_ctl1 & PCI_L1SS_CTL1_PCIPM_L1_2) link->aspm_enabled |= ASPM_STATE_L1_2_PCIPM; if (link->aspm_support & ASPM_STATE_L1SS) aspm_calc_l1ss_info(link, parent_l1ss_cap, child_l1ss_cap); /* Save default state */ link->aspm_default = link->aspm_enabled; /* Setup initial capable state. Will be updated later */ link->aspm_capable = link->aspm_support; /* Get and check endpoint acceptable latencies */ list_for_each_entry(child, &linkbus->devices, bus_list) { u32 reg32, encoding; struct aspm_latency *acceptable = &link->acceptable[PCI_FUNC(child->devfn)]; if (pci_pcie_type(child) != PCI_EXP_TYPE_ENDPOINT && pci_pcie_type(child) != PCI_EXP_TYPE_LEG_END) continue; pcie_capability_read_dword(child, PCI_EXP_DEVCAP, ®32); /* Calculate endpoint L0s acceptable latency */ encoding = (reg32 & PCI_EXP_DEVCAP_L0S) >> 6; acceptable->l0s = calc_l0s_acceptable(encoding); /* Calculate endpoint L1 acceptable latency */ encoding = (reg32 & PCI_EXP_DEVCAP_L1) >> 9; acceptable->l1 = calc_l1_acceptable(encoding); pcie_aspm_check_latency(child); } } /* Configure the ASPM L1 substates */ static void pcie_config_aspm_l1ss(struct pcie_link_state *link, u32 state) { u32 val, enable_req; struct pci_dev *child = link->downstream, *parent = link->pdev; enable_req = (link->aspm_enabled ^ state) & state; /* * Here are the rules specified in the PCIe spec for enabling L1SS: * - When enabling L1.x, enable bit at parent first, then at child * - When disabling L1.x, disable bit at child first, then at parent * - When enabling ASPM L1.x, need to disable L1 * (at child followed by parent). * - The ASPM/PCIPM L1.2 must be disabled while programming timing * parameters * * To keep it simple, disable all L1SS bits first, and later enable * what is needed. */ /* Disable all L1 substates */ pci_clear_and_set_dword(child, child->l1ss + PCI_L1SS_CTL1, PCI_L1SS_CTL1_L1SS_MASK, 0); pci_clear_and_set_dword(parent, parent->l1ss + PCI_L1SS_CTL1, PCI_L1SS_CTL1_L1SS_MASK, 0); /* * If needed, disable L1, and it gets enabled later * in pcie_config_aspm_link(). */ if (enable_req & (ASPM_STATE_L1_1 | ASPM_STATE_L1_2)) { pcie_capability_clear_and_set_word(child, PCI_EXP_LNKCTL, PCI_EXP_LNKCTL_ASPM_L1, 0); pcie_capability_clear_and_set_word(parent, PCI_EXP_LNKCTL, PCI_EXP_LNKCTL_ASPM_L1, 0); } val = 0; if (state & ASPM_STATE_L1_1) val |= PCI_L1SS_CTL1_ASPM_L1_1; if (state & ASPM_STATE_L1_2) val |= PCI_L1SS_CTL1_ASPM_L1_2; if (state & ASPM_STATE_L1_1_PCIPM) val |= PCI_L1SS_CTL1_PCIPM_L1_1; if (state & ASPM_STATE_L1_2_PCIPM) val |= PCI_L1SS_CTL1_PCIPM_L1_2; /* Enable what we need to enable */ pci_clear_and_set_dword(parent, parent->l1ss + PCI_L1SS_CTL1, PCI_L1SS_CTL1_L1SS_MASK, val); pci_clear_and_set_dword(child, child->l1ss + PCI_L1SS_CTL1, PCI_L1SS_CTL1_L1SS_MASK, val); } static void pcie_config_aspm_dev(struct pci_dev *pdev, u32 val) { pcie_capability_clear_and_set_word(pdev, PCI_EXP_LNKCTL, PCI_EXP_LNKCTL_ASPMC, val); } static void pcie_config_aspm_link(struct pcie_link_state *link, u32 state) { u32 upstream = 0, dwstream = 0; struct pci_dev *child = link->downstream, *parent = link->pdev; struct pci_bus *linkbus = parent->subordinate; /* Enable only the states that were not explicitly disabled */ state &= (link->aspm_capable & ~link->aspm_disable); /* Can't enable any substates if L1 is not enabled */ if (!(state & ASPM_STATE_L1)) state &= ~ASPM_STATE_L1SS; /* Spec says both ports must be in D0 before enabling PCI PM substates*/ if (parent->current_state != PCI_D0 || child->current_state != PCI_D0) { state &= ~ASPM_STATE_L1_SS_PCIPM; state |= (link->aspm_enabled & ASPM_STATE_L1_SS_PCIPM); } /* Nothing to do if the link is already in the requested state */ if (link->aspm_enabled == state) return; /* Convert ASPM state to upstream/downstream ASPM register state */ if (state & ASPM_STATE_L0S_UP) dwstream |= PCI_EXP_LNKCTL_ASPM_L0S; if (state & ASPM_STATE_L0S_DW) upstream |= PCI_EXP_LNKCTL_ASPM_L0S; if (state & ASPM_STATE_L1) { upstream |= PCI_EXP_LNKCTL_ASPM_L1; dwstream |= PCI_EXP_LNKCTL_ASPM_L1; } if (link->aspm_capable & ASPM_STATE_L1SS) pcie_config_aspm_l1ss(link, state); /* * Spec 2.0 suggests all functions should be configured the * same setting for ASPM. Enabling ASPM L1 should be done in * upstream component first and then downstream, and vice * versa for disabling ASPM L1. Spec doesn't mention L0S. */ if (state & ASPM_STATE_L1) pcie_config_aspm_dev(parent, upstream); list_for_each_entry(child, &linkbus->devices, bus_list) pcie_config_aspm_dev(child, dwstream); if (!(state & ASPM_STATE_L1)) pcie_config_aspm_dev(parent, upstream); link->aspm_enabled = state; } static void pcie_config_aspm_path(struct pcie_link_state *link) { while (link) { pcie_config_aspm_link(link, policy_to_aspm_state(link)); link = link->parent; } } static void free_link_state(struct pcie_link_state *link) { link->pdev->link_state = NULL; kfree(link); } static int pcie_aspm_sanity_check(struct pci_dev *pdev) { struct pci_dev *child; u32 reg32; /* * Some functions in a slot might not all be PCIe functions, * very strange. Disable ASPM for the whole slot */ list_for_each_entry(child, &pdev->subordinate->devices, bus_list) { if (!pci_is_pcie(child)) return -EINVAL; /* * If ASPM is disabled then we're not going to change * the BIOS state. It's safe to continue even if it's a * pre-1.1 device */ if (aspm_disabled) continue; /* * Disable ASPM for pre-1.1 PCIe device, we follow MS to use * RBER bit to determine if a function is 1.1 version device */ pcie_capability_read_dword(child, PCI_EXP_DEVCAP, ®32); if (!(reg32 & PCI_EXP_DEVCAP_RBER) && !aspm_force) { pci_info(child, "disabling ASPM on pre-1.1 PCIe device. You can enable it with 'pcie_aspm=force'\n"); return -EINVAL; } } return 0; } static struct pcie_link_state *alloc_pcie_link_state(struct pci_dev *pdev) { struct pcie_link_state *link; link = kzalloc(sizeof(*link), GFP_KERNEL); if (!link) return NULL; INIT_LIST_HEAD(&link->sibling); link->pdev = pdev; link->downstream = pci_function_0(pdev->subordinate); /* * Root Ports and PCI/PCI-X to PCIe Bridges are roots of PCIe * hierarchies. Note that some PCIe host implementations omit * the root ports entirely, in which case a downstream port on * a switch may become the root of the link state chain for all * its subordinate endpoints. */ if (pci_pcie_type(pdev) == PCI_EXP_TYPE_ROOT_PORT || pci_pcie_type(pdev) == PCI_EXP_TYPE_PCIE_BRIDGE || !pdev->bus->parent->self) { link->root = link; } else { struct pcie_link_state *parent; parent = pdev->bus->parent->self->link_state; if (!parent) { kfree(link); return NULL; } link->parent = parent; link->root = link->parent->root; } list_add(&link->sibling, &link_list); pdev->link_state = link; return link; } static void pcie_aspm_update_sysfs_visibility(struct pci_dev *pdev) { struct pci_dev *child; list_for_each_entry(child, &pdev->subordinate->devices, bus_list) sysfs_update_group(&child->dev.kobj, &aspm_ctrl_attr_group); } /* * pcie_aspm_init_link_state: Initiate PCI express link state. * It is called after the pcie and its children devices are scanned. * @pdev: the root port or switch downstream port */ void pcie_aspm_init_link_state(struct pci_dev *pdev) { struct pcie_link_state *link; int blacklist = !!pcie_aspm_sanity_check(pdev); if (!aspm_support_enabled) return; if (pdev->link_state) return; /* * We allocate pcie_link_state for the component on the upstream * end of a Link, so there's nothing to do unless this device is * downstream port. */ if (!pcie_downstream_port(pdev)) return; /* VIA has a strange chipset, root port is under a bridge */ if (pci_pcie_type(pdev) == PCI_EXP_TYPE_ROOT_PORT && pdev->bus->self) return; down_read(&pci_bus_sem); if (list_empty(&pdev->subordinate->devices)) goto out; mutex_lock(&aspm_lock); link = alloc_pcie_link_state(pdev); if (!link) goto unlock; /* * Setup initial ASPM state. Note that we need to configure * upstream links also because capable state of them can be * update through pcie_aspm_cap_init(). */ pcie_aspm_cap_init(link, blacklist); /* Setup initial Clock PM state */ pcie_clkpm_cap_init(link, blacklist); /* * At this stage drivers haven't had an opportunity to change the * link policy setting. Enabling ASPM on broken hardware can cripple * it even before the driver has had a chance to disable ASPM, so * default to a safe level right now. If we're enabling ASPM beyond * the BIOS's expectation, we'll do so once pci_enable_device() is * called. */ if (aspm_policy != POLICY_POWERSAVE && aspm_policy != POLICY_POWER_SUPERSAVE) { pcie_config_aspm_path(link); pcie_set_clkpm(link, policy_to_clkpm_state(link)); } pcie_aspm_update_sysfs_visibility(pdev); unlock: mutex_unlock(&aspm_lock); out: up_read(&pci_bus_sem); } /* Recheck latencies and update aspm_capable for links under the root */ static void pcie_update_aspm_capable(struct pcie_link_state *root) { struct pcie_link_state *link; BUG_ON(root->parent); list_for_each_entry(link, &link_list, sibling) { if (link->root != root) continue; link->aspm_capable = link->aspm_support; } list_for_each_entry(link, &link_list, sibling) { struct pci_dev *child; struct pci_bus *linkbus = link->pdev->subordinate; if (link->root != root) continue; list_for_each_entry(child, &linkbus->devices, bus_list) { if ((pci_pcie_type(child) != PCI_EXP_TYPE_ENDPOINT) && (pci_pcie_type(child) != PCI_EXP_TYPE_LEG_END)) continue; pcie_aspm_check_latency(child); } } } /* @pdev: the endpoint device */ void pcie_aspm_exit_link_state(struct pci_dev *pdev) { struct pci_dev *parent = pdev->bus->self; struct pcie_link_state *link, *root, *parent_link; if (!parent || !parent->link_state) return; down_read(&pci_bus_sem); mutex_lock(&aspm_lock); link = parent->link_state; root = link->root; parent_link = link->parent; /* * link->downstream is a pointer to the pci_dev of function 0. If * we remove that function, the pci_dev is about to be deallocated, * so we can't use link->downstream again. Free the link state to * avoid this. * * If we're removing a non-0 function, it's possible we could * retain the link state, but PCIe r6.0, sec 7.5.3.7, recommends * programming the same ASPM Control value for all functions of * multi-function devices, so disable ASPM for all of them. */ pcie_config_aspm_link(link, 0); list_del(&link->sibling); free_link_state(link); /* Recheck latencies and configure upstream links */ if (parent_link) { pcie_update_aspm_capable(root); pcie_config_aspm_path(parent_link); } mutex_unlock(&aspm_lock); up_read(&pci_bus_sem); } /* @pdev: the root port or switch downstream port */ void pcie_aspm_pm_state_change(struct pci_dev *pdev) { struct pcie_link_state *link = pdev->link_state; if (aspm_disabled || !link) return; /* * Devices changed PM state, we should recheck if latency * meets all functions' requirement */ down_read(&pci_bus_sem); mutex_lock(&aspm_lock); pcie_update_aspm_capable(link->root); pcie_config_aspm_path(link); mutex_unlock(&aspm_lock); up_read(&pci_bus_sem); } void pcie_aspm_powersave_config_link(struct pci_dev *pdev) { struct pcie_link_state *link = pdev->link_state; if (aspm_disabled || !link) return; if (aspm_policy != POLICY_POWERSAVE && aspm_policy != POLICY_POWER_SUPERSAVE) return; down_read(&pci_bus_sem); mutex_lock(&aspm_lock); pcie_config_aspm_path(link); pcie_set_clkpm(link, policy_to_clkpm_state(link)); mutex_unlock(&aspm_lock); up_read(&pci_bus_sem); } static struct pcie_link_state *pcie_aspm_get_link(struct pci_dev *pdev) { struct pci_dev *bridge; if (!pci_is_pcie(pdev)) return NULL; bridge = pci_upstream_bridge(pdev); if (!bridge || !pci_is_pcie(bridge)) return NULL; return bridge->link_state; } static int __pci_disable_link_state(struct pci_dev *pdev, int state, bool sem) { struct pcie_link_state *link = pcie_aspm_get_link(pdev); if (!link) return -EINVAL; /* * A driver requested that ASPM be disabled on this device, but * if we don't have permission to manage ASPM (e.g., on ACPI * systems we have to observe the FADT ACPI_FADT_NO_ASPM bit and * the _OSC method), we can't honor that request. Windows has * a similar mechanism using "PciASPMOptOut", which is also * ignored in this situation. */ if (aspm_disabled) { pci_warn(pdev, "can't disable ASPM; OS doesn't have ASPM control\n"); return -EPERM; } if (sem) down_read(&pci_bus_sem); mutex_lock(&aspm_lock); if (state & PCIE_LINK_STATE_L0S) link->aspm_disable |= ASPM_STATE_L0S; if (state & PCIE_LINK_STATE_L1) /* L1 PM substates require L1 */ link->aspm_disable |= ASPM_STATE_L1 | ASPM_STATE_L1SS; if (state & PCIE_LINK_STATE_L1_1) link->aspm_disable |= ASPM_STATE_L1_1; if (state & PCIE_LINK_STATE_L1_2) link->aspm_disable |= ASPM_STATE_L1_2; if (state & PCIE_LINK_STATE_L1_1_PCIPM) link->aspm_disable |= ASPM_STATE_L1_1_PCIPM; if (state & PCIE_LINK_STATE_L1_2_PCIPM) link->aspm_disable |= ASPM_STATE_L1_2_PCIPM; pcie_config_aspm_link(link, policy_to_aspm_state(link)); if (state & PCIE_LINK_STATE_CLKPM) link->clkpm_disable = 1; pcie_set_clkpm(link, policy_to_clkpm_state(link)); mutex_unlock(&aspm_lock); if (sem) up_read(&pci_bus_sem); return 0; } int pci_disable_link_state_locked(struct pci_dev *pdev, int state) { return __pci_disable_link_state(pdev, state, false); } EXPORT_SYMBOL(pci_disable_link_state_locked); /** * pci_disable_link_state - Disable device's link state, so the link will * never enter specific states. Note that if the BIOS didn't grant ASPM * control to the OS, this does nothing because we can't touch the LNKCTL * register. Returns 0 or a negative errno. * * @pdev: PCI device * @state: ASPM link state to disable */ int pci_disable_link_state(struct pci_dev *pdev, int state) { return __pci_disable_link_state(pdev, state, true); } EXPORT_SYMBOL(pci_disable_link_state); static int pcie_aspm_set_policy(const char *val, const struct kernel_param *kp) { int i; struct pcie_link_state *link; if (aspm_disabled) return -EPERM; i = sysfs_match_string(policy_str, val); if (i < 0) return i; if (i == aspm_policy) return 0; down_read(&pci_bus_sem); mutex_lock(&aspm_lock); aspm_policy = i; list_for_each_entry(link, &link_list, sibling) { pcie_config_aspm_link(link, policy_to_aspm_state(link)); pcie_set_clkpm(link, policy_to_clkpm_state(link)); } mutex_unlock(&aspm_lock); up_read(&pci_bus_sem); return 0; } static int pcie_aspm_get_policy(char *buffer, const struct kernel_param *kp) { int i, cnt = 0; for (i = 0; i < ARRAY_SIZE(policy_str); i++) if (i == aspm_policy) cnt += sprintf(buffer + cnt, "[%s] ", policy_str[i]); else cnt += sprintf(buffer + cnt, "%s ", policy_str[i]); cnt += sprintf(buffer + cnt, "\n"); return cnt; } module_param_call(policy, pcie_aspm_set_policy, pcie_aspm_get_policy, NULL, 0644); /** * pcie_aspm_enabled - Check if PCIe ASPM has been enabled for a device. * @pdev: Target device. * * Relies on the upstream bridge's link_state being valid. The link_state * is deallocated only when the last child of the bridge (i.e., @pdev or a * sibling) is removed, and the caller should be holding a reference to * @pdev, so this should be safe. */ bool pcie_aspm_enabled(struct pci_dev *pdev) { struct pcie_link_state *link = pcie_aspm_get_link(pdev); if (!link) return false; return link->aspm_enabled; } EXPORT_SYMBOL_GPL(pcie_aspm_enabled); static ssize_t aspm_attr_show_common(struct device *dev, struct device_attribute *attr, char *buf, u8 state) { struct pci_dev *pdev = to_pci_dev(dev); struct pcie_link_state *link = pcie_aspm_get_link(pdev); return sprintf(buf, "%d\n", (link->aspm_enabled & state) ? 1 : 0); } static ssize_t aspm_attr_store_common(struct device *dev, struct device_attribute *attr, const char *buf, size_t len, u8 state) { struct pci_dev *pdev = to_pci_dev(dev); struct pcie_link_state *link = pcie_aspm_get_link(pdev); bool state_enable; if (strtobool(buf, &state_enable) < 0) return -EINVAL; down_read(&pci_bus_sem); mutex_lock(&aspm_lock); if (state_enable) { link->aspm_disable &= ~state; /* need to enable L1 for substates */ if (state & ASPM_STATE_L1SS) link->aspm_disable &= ~ASPM_STATE_L1; } else { link->aspm_disable |= state; if (state & ASPM_STATE_L1) link->aspm_disable |= ASPM_STATE_L1SS; } pcie_config_aspm_link(link, policy_to_aspm_state(link)); mutex_unlock(&aspm_lock); up_read(&pci_bus_sem); return len; } #define ASPM_ATTR(_f, _s) \ static ssize_t _f##_show(struct device *dev, \ struct device_attribute *attr, char *buf) \ { return aspm_attr_show_common(dev, attr, buf, ASPM_STATE_##_s); } \ \ static ssize_t _f##_store(struct device *dev, \ struct device_attribute *attr, \ const char *buf, size_t len) \ { return aspm_attr_store_common(dev, attr, buf, len, ASPM_STATE_##_s); } ASPM_ATTR(l0s_aspm, L0S) ASPM_ATTR(l1_aspm, L1) ASPM_ATTR(l1_1_aspm, L1_1) ASPM_ATTR(l1_2_aspm, L1_2) ASPM_ATTR(l1_1_pcipm, L1_1_PCIPM) ASPM_ATTR(l1_2_pcipm, L1_2_PCIPM) static ssize_t clkpm_show(struct device *dev, struct device_attribute *attr, char *buf) { struct pci_dev *pdev = to_pci_dev(dev); struct pcie_link_state *link = pcie_aspm_get_link(pdev); return sprintf(buf, "%d\n", link->clkpm_enabled); } static ssize_t clkpm_store(struct device *dev, struct device_attribute *attr, const char *buf, size_t len) { struct pci_dev *pdev = to_pci_dev(dev); struct pcie_link_state *link = pcie_aspm_get_link(pdev); bool state_enable; if (strtobool(buf, &state_enable) < 0) return -EINVAL; down_read(&pci_bus_sem); mutex_lock(&aspm_lock); link->clkpm_disable = !state_enable; pcie_set_clkpm(link, policy_to_clkpm_state(link)); mutex_unlock(&aspm_lock); up_read(&pci_bus_sem); return len; } static DEVICE_ATTR_RW(clkpm); static DEVICE_ATTR_RW(l0s_aspm); static DEVICE_ATTR_RW(l1_aspm); static DEVICE_ATTR_RW(l1_1_aspm); static DEVICE_ATTR_RW(l1_2_aspm); static DEVICE_ATTR_RW(l1_1_pcipm); static DEVICE_ATTR_RW(l1_2_pcipm); static struct attribute *aspm_ctrl_attrs[] = { &dev_attr_clkpm.attr, &dev_attr_l0s_aspm.attr, &dev_attr_l1_aspm.attr, &dev_attr_l1_1_aspm.attr, &dev_attr_l1_2_aspm.attr, &dev_attr_l1_1_pcipm.attr, &dev_attr_l1_2_pcipm.attr, NULL }; static umode_t aspm_ctrl_attrs_are_visible(struct kobject *kobj, struct attribute *a, int n) { struct device *dev = kobj_to_dev(kobj); struct pci_dev *pdev = to_pci_dev(dev); struct pcie_link_state *link = pcie_aspm_get_link(pdev); static const u8 aspm_state_map[] = { ASPM_STATE_L0S, ASPM_STATE_L1, ASPM_STATE_L1_1, ASPM_STATE_L1_2, ASPM_STATE_L1_1_PCIPM, ASPM_STATE_L1_2_PCIPM, }; if (aspm_disabled || !link) return 0; if (n == 0) return link->clkpm_capable ? a->mode : 0; return link->aspm_capable & aspm_state_map[n - 1] ? a->mode : 0; } const struct attribute_group aspm_ctrl_attr_group = { .name = "link", .attrs = aspm_ctrl_attrs, .is_visible = aspm_ctrl_attrs_are_visible, }; static int __init pcie_aspm_disable(char *str) { if (!strcmp(str, "off")) { aspm_policy = POLICY_DEFAULT; aspm_disabled = 1; aspm_support_enabled = false; printk(KERN_INFO "PCIe ASPM is disabled\n"); } else if (!strcmp(str, "force")) { aspm_force = 1; printk(KERN_INFO "PCIe ASPM is forcibly enabled\n"); } return 1; } __setup("pcie_aspm=", pcie_aspm_disable); void pcie_no_aspm(void) { /* * Disabling ASPM is intended to prevent the kernel from modifying * existing hardware state, not to clear existing state. To that end: * (a) set policy to POLICY_DEFAULT in order to avoid changing state * (b) prevent userspace from changing policy */ if (!aspm_force) { aspm_policy = POLICY_DEFAULT; aspm_disabled = 1; } } bool pcie_aspm_support_enabled(void) { return aspm_support_enabled; } EXPORT_SYMBOL(pcie_aspm_support_enabled);