[pve-devel] [PATCH pve-kernel bookworm-6.14 1/1] backport fix for passthrough for devices without PCI power management

Dominik Csapak d.csapak at proxmox.com
Tue Jul 22 11:56:58 CEST 2025


fixes broken start for virtual functions:

 kvm: -device vfio-pci,host=0000:07:10.0,id=hostpci0,bus=pci.0,addr=0x10: vfio 0000:07:10.0: error getting device from group 23: Permission denied
 Verify all devices in group 23 are bound to vfio-<bus> or pci-stub and not already in use

and for nvidia vgpu:
 kvm: -device vfio-pci,host=0000:21:00.4,id=hostpci0,bus=pci.0,addr=0x10: vfio 0000:21:00.4: error getting device info: Input/output error

Signed-off-by: Dominik Csapak <d.csapak at proxmox.com>
---
this can probably be cherry picked for trixie

 ...me-PM-on-devices-that-don-t-support-.patch | 57 +++++++++++++++++++
 1 file changed, 57 insertions(+)
 create mode 100644 patches/kernel/0014-PCI-Set-up-runtime-PM-on-devices-that-don-t-support-.patch

diff --git a/patches/kernel/0014-PCI-Set-up-runtime-PM-on-devices-that-don-t-support-.patch b/patches/kernel/0014-PCI-Set-up-runtime-PM-on-devices-that-don-t-support-.patch
new file mode 100644
index 0000000..3f271ef
--- /dev/null
+++ b/patches/kernel/0014-PCI-Set-up-runtime-PM-on-devices-that-don-t-support-.patch
@@ -0,0 +1,57 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Mario Limonciello <mario.limonciello at amd.com>
+Date: Wed, 11 Jun 2025 18:31:16 -0500
+Subject: [PATCH] PCI: Set up runtime PM on devices that don't support PCI PM
+
+commit 4d4c10f763d7 ("PCI: Explicitly put devices into D0 when
+initializing") intended to put PCI devices into D0, but in doing so
+unintentionally changed runtime PM initialization not to occur on
+devices that don't support PCI PM.  This caused a regression in vfio-pci
+due to an imbalance with it's use.
+
+Adjust the logic in pci_pm_init() so that even if PCI PM isn't supported
+runtime PM is still initialized.
+
+Cc: Alex Williamson <alex.williamson at redhat.com>
+Reported-by: Giovanni Cabiddu <giovanni.cabiddu at intel.com>
+Closes: https://lore.kernel.org/linux-pci/20250424043232.1848107-1-superm1@kernel.org/T/#m7e8929d6421690dc8bd6dc639d86c2b4db27cbc4
+Reported-by: Nicolas Dichtel <nicolas.dichtel at 6wind.com>
+Closes: https://lore.kernel.org/linux-pci/20250424043232.1848107-1-superm1@kernel.org/T/#m40d277dcdb9be64a1609a82412d1aa906263e201
+Tested-by: Giovanni Cabiddu <giovanni.cabiddu at intel.com>
+Fixes: 4d4c10f763d7 ("PCI: Explicitly put devices into D0 when initializing")
+Signed-off-by: Mario Limonciello <mario.limonciello at amd.com>
+(backported from https://lore.kernel.org/all/20250611233117.61810-1-superm1@kernel.org/)
+Signed-off-by: Dominik Csapak <d.csapak at proxmox.com>
+---
+ drivers/pci/pci.c | 5 +++--
+ 1 file changed, 3 insertions(+), 2 deletions(-)
+
+diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
+index 1358f4c17209..5d9208977665 100644
+--- a/drivers/pci/pci.c
++++ b/drivers/pci/pci.c
+@@ -3217,14 +3217,14 @@ void pci_pm_init(struct pci_dev *dev)
+ 	/* find PCI PM capability in list */
+ 	pm = pci_find_capability(dev, PCI_CAP_ID_PM);
+ 	if (!pm)
+-		return;
++		goto poweron;
+ 	/* Check device's ability to generate PME# */
+ 	pci_read_config_word(dev, pm + PCI_PM_PMC, &pmc);
+ 
+ 	if ((pmc & PCI_PM_CAP_VER_MASK) > 3) {
+ 		pci_err(dev, "unsupported PM cap regs version (%u)\n",
+ 			pmc & PCI_PM_CAP_VER_MASK);
+-		return;
++		goto poweron;
+ 	}
+ 
+ 	dev->pm_cap = pm;
+@@ -3269,6 +3269,7 @@ void pci_pm_init(struct pci_dev *dev)
+ 	pci_read_config_word(dev, PCI_STATUS, &status);
+ 	if (status & PCI_STATUS_IMM_READY)
+ 		dev->imm_ready = 1;
++poweron:
+ 	pci_pm_power_up_and_verify_state(dev);
+ 	pm_runtime_forbid(&dev->dev);
+ 	pm_runtime_set_active(&dev->dev);
-- 
2.39.5





More information about the pve-devel mailing list