From 1e560864159d002b453da42bd2c13a1805515a20 Mon Sep 17 00:00:00 2001 From: Johan Hovold Date: Tue, 30 Jan 2024 11:02:43 +0100 Subject: PCI/ASPM: Fix deadlock when enabling ASPM A last minute revert in 6.7-final introduced a potential deadlock when enabling ASPM during probe of Qualcomm PCIe controllers as reported by lockdep: ============================================ WARNING: possible recursive locking detected 6.7.0 #40 Not tainted -------------------------------------------- kworker/u16:5/90 is trying to acquire lock: ffffacfa78ced000 (pci_bus_sem){++++}-{3:3}, at: pcie_aspm_pm_state_change+0x58/0xdc but task is already holding lock: ffffacfa78ced000 (pci_bus_sem){++++}-{3:3}, at: pci_walk_bus+0x34/0xbc other info that might help us debug this: Possible unsafe locking scenario: CPU0 ---- lock(pci_bus_sem); lock(pci_bus_sem); *** DEADLOCK *** Call trace: print_deadlock_bug+0x25c/0x348 __lock_acquire+0x10a4/0x2064 lock_acquire+0x1e8/0x318 down_read+0x60/0x184 pcie_aspm_pm_state_change+0x58/0xdc pci_set_full_power_state+0xa8/0x114 pci_set_power_state+0xc4/0x120 qcom_pcie_enable_aspm+0x1c/0x3c [pcie_qcom] pci_walk_bus+0x64/0xbc qcom_pcie_host_post_init_2_7_0+0x28/0x34 [pcie_qcom] The deadlock can easily be reproduced on machines like the Lenovo ThinkPad X13s by adding a delay to increase the race window during asynchronous probe where another thread can take a write lock. Add a new pci_set_power_state_locked() and associated helper functions that can be called with the PCI bus semaphore held to avoid taking the read lock twice. Link: https://lore.kernel.org/r/ZZu0qx2cmn7IwTyQ@hovoldconsulting.com Link: https://lore.kernel.org/r/20240130100243.11011-1-johan+linaro@kernel.org Fixes: f93e71aea6c6 ("Revert "PCI/ASPM: Remove pcie_aspm_pm_state_change()"") Signed-off-by: Johan Hovold Signed-off-by: Bjorn Helgaas Cc: # 6.7 --- drivers/pci/pci.c | 78 ++++++++++++++++++++++++++++++++++++------------------- 1 file changed, 52 insertions(+), 26 deletions(-) (limited to 'drivers/pci/pci.c') diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c index d8f11a078924..9ab9b1008d8b 100644 --- a/drivers/pci/pci.c +++ b/drivers/pci/pci.c @@ -1354,6 +1354,7 @@ end: /** * pci_set_full_power_state - Put a PCI device into D0 and update its state * @dev: PCI device to power up + * @locked: whether pci_bus_sem is held * * Call pci_power_up() to put @dev into D0, read from its PCI_PM_CTRL register * to confirm the state change, restore its BARs if they might be lost and @@ -1363,7 +1364,7 @@ end: * to D0, it is more efficient to use pci_power_up() directly instead of this * function. */ -static int pci_set_full_power_state(struct pci_dev *dev) +static int pci_set_full_power_state(struct pci_dev *dev, bool locked) { u16 pmcsr; int ret; @@ -1399,7 +1400,7 @@ static int pci_set_full_power_state(struct pci_dev *dev) } if (dev->bus->self) - pcie_aspm_pm_state_change(dev->bus->self); + pcie_aspm_pm_state_change(dev->bus->self, locked); return 0; } @@ -1428,10 +1429,22 @@ void pci_bus_set_current_state(struct pci_bus *bus, pci_power_t state) pci_walk_bus(bus, __pci_dev_set_current_state, &state); } +static void __pci_bus_set_current_state(struct pci_bus *bus, pci_power_t state, bool locked) +{ + if (!bus) + return; + + if (locked) + pci_walk_bus_locked(bus, __pci_dev_set_current_state, &state); + else + pci_walk_bus(bus, __pci_dev_set_current_state, &state); +} + /** * pci_set_low_power_state - Put a PCI device into a low-power state. * @dev: PCI device to handle. * @state: PCI power state (D1, D2, D3hot) to put the device into. + * @locked: whether pci_bus_sem is held * * Use the device's PCI_PM_CTRL register to put it into a low-power state. * @@ -1442,7 +1455,7 @@ void pci_bus_set_current_state(struct pci_bus *bus, pci_power_t state) * 0 if device already is in the requested state. * 0 if device's power state has been successfully changed. */ -static int pci_set_low_power_state(struct pci_dev *dev, pci_power_t state) +static int pci_set_low_power_state(struct pci_dev *dev, pci_power_t state, bool locked) { u16 pmcsr; @@ -1496,29 +1509,12 @@ static int pci_set_low_power_state(struct pci_dev *dev, pci_power_t state) pci_power_name(state)); if (dev->bus->self) - pcie_aspm_pm_state_change(dev->bus->self); + pcie_aspm_pm_state_change(dev->bus->self, locked); return 0; } -/** - * pci_set_power_state - Set the power state of a PCI device - * @dev: PCI device to handle. - * @state: PCI power state (D0, D1, D2, D3hot) to put the device into. - * - * Transition a device to a new power state, using the platform firmware and/or - * the device's PCI PM registers. - * - * RETURN VALUE: - * -EINVAL if the requested state is invalid. - * -EIO if device does not support PCI PM or its PM capabilities register has a - * wrong version, or device doesn't support the requested state. - * 0 if the transition is to D1 or D2 but D1 and D2 are not supported. - * 0 if device already is in the requested state. - * 0 if the transition is to D3 but D3 is not supported. - * 0 if device's power state has been successfully changed. - */ -int pci_set_power_state(struct pci_dev *dev, pci_power_t state) +static int __pci_set_power_state(struct pci_dev *dev, pci_power_t state, bool locked) { int error; @@ -1542,7 +1538,7 @@ int pci_set_power_state(struct pci_dev *dev, pci_power_t state) return 0; if (state == PCI_D0) - return pci_set_full_power_state(dev); + return pci_set_full_power_state(dev, locked); /* * This device is quirked not to be put into D3, so don't put it in @@ -1556,16 +1552,16 @@ int pci_set_power_state(struct pci_dev *dev, pci_power_t state) * To put the device in D3cold, put it into D3hot in the native * way, then put it into D3cold using platform ops. */ - error = pci_set_low_power_state(dev, PCI_D3hot); + error = pci_set_low_power_state(dev, PCI_D3hot, locked); if (pci_platform_power_transition(dev, PCI_D3cold)) return error; /* Powering off a bridge may power off the whole hierarchy */ if (dev->current_state == PCI_D3cold) - pci_bus_set_current_state(dev->subordinate, PCI_D3cold); + __pci_bus_set_current_state(dev->subordinate, PCI_D3cold, locked); } else { - error = pci_set_low_power_state(dev, state); + error = pci_set_low_power_state(dev, state, locked); if (pci_platform_power_transition(dev, state)) return error; @@ -1573,8 +1569,38 @@ int pci_set_power_state(struct pci_dev *dev, pci_power_t state) return 0; } + +/** + * pci_set_power_state - Set the power state of a PCI device + * @dev: PCI device to handle. + * @state: PCI power state (D0, D1, D2, D3hot) to put the device into. + * + * Transition a device to a new power state, using the platform firmware and/or + * the device's PCI PM registers. + * + * RETURN VALUE: + * -EINVAL if the requested state is invalid. + * -EIO if device does not support PCI PM or its PM capabilities register has a + * wrong version, or device doesn't support the requested state. + * 0 if the transition is to D1 or D2 but D1 and D2 are not supported. + * 0 if device already is in the requested state. + * 0 if the transition is to D3 but D3 is not supported. + * 0 if device's power state has been successfully changed. + */ +int pci_set_power_state(struct pci_dev *dev, pci_power_t state) +{ + return __pci_set_power_state(dev, state, false); +} EXPORT_SYMBOL(pci_set_power_state); +int pci_set_power_state_locked(struct pci_dev *dev, pci_power_t state) +{ + lockdep_assert_held(&pci_bus_sem); + + return __pci_set_power_state(dev, state, true); +} +EXPORT_SYMBOL(pci_set_power_state_locked); + #define PCI_EXP_SAVE_REGS 7 static struct pci_cap_saved_state *_pci_find_saved_cap(struct pci_dev *pci_dev, -- cgit v1.2.3-70-g09d2 From 41044d5360685e78a869d40a168491a70cdb7e73 Mon Sep 17 00:00:00 2001 From: Alex Williamson Date: Tue, 23 Jan 2024 11:55:31 -0700 Subject: PCI: Fix active state requirement in PME polling The commit noted in fixes added a bogus requirement that runtime PM managed devices need to be in the RPM_ACTIVE state for PME polling. In fact, only devices in low power states should be polled. However there's still a requirement that the device config space must be accessible, which has implications for both the current state of the polled device and the parent bridge, when present. It's not sufficient to assume the bridge remains in D0 and cases have been observed where the bridge passes the D0 test, but the PM state indicates RPM_SUSPENDING and config space of the polled device becomes inaccessible during pci_pme_wakeup(). Therefore, since the bridge is already effectively required to be in the RPM_ACTIVE state, formalize this in the code and elevate the PM usage count to maintain the state while polling the subordinate device. This resolves a regression reported in the bugzilla below where a Thunderbolt/USB4 hierarchy fails to scan for an attached NVMe endpoint downstream of a bridge in a D3hot power state. Link: https://lore.kernel.org/r/20240123185548.1040096-1-alex.williamson@redhat.com Fixes: d3fcd7360338 ("PCI: Fix runtime PM race with PME polling") Reported-by: Sanath S Closes: https://bugzilla.kernel.org/show_bug.cgi?id=218360 Signed-off-by: Alex Williamson Signed-off-by: Bjorn Helgaas Tested-by: Sanath S Reviewed-by: Rafael J. Wysocki Cc: Lukas Wunner Cc: Mika Westerberg --- drivers/pci/pci.c | 37 ++++++++++++++++++++++--------------- 1 file changed, 22 insertions(+), 15 deletions(-) (limited to 'drivers/pci/pci.c') diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c index d8f11a078924..0a80156fe812 100644 --- a/drivers/pci/pci.c +++ b/drivers/pci/pci.c @@ -2496,29 +2496,36 @@ static void pci_pme_list_scan(struct work_struct *work) if (pdev->pme_poll) { struct pci_dev *bridge = pdev->bus->self; struct device *dev = &pdev->dev; - int pm_status; + struct device *bdev = bridge ? &bridge->dev : NULL; + int bref = 0; /* - * If bridge is in low power state, the - * configuration space of subordinate devices - * may be not accessible + * If we have a bridge, it should be in an active/D0 + * state or the configuration space of subordinate + * devices may not be accessible or stable over the + * course of the call. */ - if (bridge && bridge->current_state != PCI_D0) - continue; + if (bdev) { + bref = pm_runtime_get_if_active(bdev, true); + if (!bref) + continue; + + if (bridge->current_state != PCI_D0) + goto put_bridge; + } /* - * If the device is in a low power state it - * should not be polled either. + * The device itself should be suspended but config + * space must be accessible, therefore it cannot be in + * D3cold. */ - pm_status = pm_runtime_get_if_active(dev, true); - if (!pm_status) - continue; - - if (pdev->current_state != PCI_D3cold) + if (pm_runtime_suspended(dev) && + pdev->current_state != PCI_D3cold) pci_pme_wakeup(pdev, NULL); - if (pm_status > 0) - pm_runtime_put(dev); +put_bridge: + if (bref > 0) + pm_runtime_put(bdev); } else { list_del(&pme_dev->list); kfree(pme_dev); -- cgit v1.2.3-70-g09d2