[PATCH 16/21] drm/nouveau/nvkm: move pci probe() runpm quirk from drm

Ben Skeggs bskeggs at nvidia.com
Thu Jun 13 17:00:08 UTC 2024


Signed-off-by: Ben Skeggs <bskeggs at nvidia.com>
---
 .../gpu/drm/nouveau/include/nvkm/core/pci.h   |  2 +
 drivers/gpu/drm/nouveau/nouveau_drm.c         | 61 ------------------
 drivers/gpu/drm/nouveau/nouveau_drv.h         |  2 -
 drivers/gpu/drm/nouveau/nvkm/device/pci.c     | 62 +++++++++++++++++++
 4 files changed, 64 insertions(+), 63 deletions(-)

diff --git a/drivers/gpu/drm/nouveau/include/nvkm/core/pci.h b/drivers/gpu/drm/nouveau/include/nvkm/core/pci.h
index 0797225ab038..95deea8c65ff 100644
--- a/drivers/gpu/drm/nouveau/include/nvkm/core/pci.h
+++ b/drivers/gpu/drm/nouveau/include/nvkm/core/pci.h
@@ -7,6 +7,8 @@ struct nvkm_device_pci {
 	struct nvkm_device device;
 	struct pci_dev *pdev;
 
+	u8 old_pm_cap;
+
 	struct dev_pm_domain vga_pm_domain;
 };
 
diff --git a/drivers/gpu/drm/nouveau/nouveau_drm.c b/drivers/gpu/drm/nouveau/nouveau_drm.c
index 4bcfc2291c4d..76eddf172bb5 100644
--- a/drivers/gpu/drm/nouveau/nouveau_drm.c
+++ b/drivers/gpu/drm/nouveau/nouveau_drm.c
@@ -731,63 +731,6 @@ nouveau_drm_device_new(const struct drm_driver *drm_driver, struct device *paren
 	return ret ? ERR_PTR(ret) : drm;
 }
 
-/*
- * On some Intel PCIe bridge controllers doing a
- * D0 -> D3hot -> D3cold -> D0 sequence causes Nvidia GPUs to not reappear.
- * Skipping the intermediate D3hot step seems to make it work again. This is
- * probably caused by not meeting the expectation the involved AML code has
- * when the GPU is put into D3hot state before invoking it.
- *
- * This leads to various manifestations of this issue:
- *  - AML code execution to power on the GPU hits an infinite loop (as the
- *    code waits on device memory to change).
- *  - kernel crashes, as all PCI reads return -1, which most code isn't able
- *    to handle well enough.
- *
- * In all cases dmesg will contain at least one line like this:
- * 'nouveau 0000:01:00.0: Refused to change power state, currently in D3'
- * followed by a lot of nouveau timeouts.
- *
- * In the \_SB.PCI0.PEG0.PG00._OFF code deeper down writes bit 0x80 to the not
- * documented PCI config space register 0x248 of the Intel PCIe bridge
- * controller (0x1901) in order to change the state of the PCIe link between
- * the PCIe port and the GPU. There are alternative code paths using other
- * registers, which seem to work fine (executed pre Windows 8):
- *  - 0xbc bit 0x20 (publicly available documentation claims 'reserved')
- *  - 0xb0 bit 0x10 (link disable)
- * Changing the conditions inside the firmware by poking into the relevant
- * addresses does resolve the issue, but it seemed to be ACPI private memory
- * and not any device accessible memory at all, so there is no portable way of
- * changing the conditions.
- * On a XPS 9560 that means bits [0,3] on \CPEX need to be cleared.
- *
- * The only systems where this behavior can be seen are hybrid graphics laptops
- * with a secondary Nvidia Maxwell, Pascal or Turing GPU. It's unclear whether
- * this issue only occurs in combination with listed Intel PCIe bridge
- * controllers and the mentioned GPUs or other devices as well.
- *
- * documentation on the PCIe bridge controller can be found in the
- * "7th Generation Intel® Processor Families for H Platforms Datasheet Volume 2"
- * Section "12 PCI Express* Controller (x16) Registers"
- */
-
-static void quirk_broken_nv_runpm(struct pci_dev *pdev)
-{
-	struct nouveau_drm *drm = pci_get_drvdata(pdev);
-	struct pci_dev *bridge = pci_upstream_bridge(pdev);
-
-	if (!bridge || bridge->vendor != PCI_VENDOR_ID_INTEL)
-		return;
-
-	switch (bridge->device) {
-	case 0x1901:
-		drm->old_pm_cap = pdev->pm_cap;
-		pdev->pm_cap = 0;
-		NV_INFO(drm, "Disabling PCI power management to avoid bug\n");
-		break;
-	}
-}
-
 static int nouveau_drm_probe(struct pci_dev *pdev,
 			     const struct pci_device_id *pent)
 {
@@ -822,7 +765,6 @@ static int nouveau_drm_probe(struct pci_dev *pdev,
 	else
 		drm_fbdev_ttm_setup(drm->dev, 32);
 
-	quirk_broken_nv_runpm(pdev);
 	return 0;
 
 fail_drm:
@@ -846,9 +788,6 @@ nouveau_drm_remove(struct pci_dev *pdev)
 {
 	struct nouveau_drm *drm = pci_get_drvdata(pdev);
 
-	/* revert our workaround */
-	if (drm->old_pm_cap)
-		pdev->pm_cap = drm->old_pm_cap;
 	nouveau_drm_device_remove(drm);
 
 	nvkm_device_pci_driver.remove(pdev);
diff --git a/drivers/gpu/drm/nouveau/nouveau_drv.h b/drivers/gpu/drm/nouveau/nouveau_drv.h
index b44f0d408ccc..9ca0f6ab4359 100644
--- a/drivers/gpu/drm/nouveau/nouveau_drv.h
+++ b/drivers/gpu/drm/nouveau/nouveau_drv.h
@@ -218,8 +218,6 @@ struct nouveau_drm {
 	 */
 	struct mutex clients_lock;
 
-	u8 old_pm_cap;
-
 	struct {
 		struct agp_bridge_data *bridge;
 		u32 base;
diff --git a/drivers/gpu/drm/nouveau/nvkm/device/pci.c b/drivers/gpu/drm/nouveau/nvkm/device/pci.c
index d9b8e3bc4169..d454d56a7909 100644
--- a/drivers/gpu/drm/nouveau/nvkm/device/pci.c
+++ b/drivers/gpu/drm/nouveau/nvkm/device/pci.c
@@ -1598,6 +1598,11 @@ static void *
 nvkm_device_pci_dtor(struct nvkm_device *device)
 {
 	struct nvkm_device_pci *pdev = nvkm_device_pci(device);
+
+	/* revert our workaround */
+	if (pdev->old_pm_cap)
+		pdev->pdev->pm_cap = pdev->old_pm_cap;
+
 	pci_disable_device(pdev->pdev);
 	return pdev;
 }
@@ -1624,6 +1629,62 @@ nvkm_device_pci_remove(struct pci_dev *dev)
 	nvkm_device_del(&device);
 }
 
+/*
+ * On some Intel PCIe bridge controllers doing a
+ * D0 -> D3hot -> D3cold -> D0 sequence causes Nvidia GPUs to not reappear.
+ * Skipping the intermediate D3hot step seems to make it work again. This is
+ * probably caused by not meeting the expectation the involved AML code has
+ * when the GPU is put into D3hot state before invoking it.
+ *
+ * This leads to various manifestations of this issue:
+ *  - AML code execution to power on the GPU hits an infinite loop (as the
+ *    code waits on device memory to change).
+ *  - kernel crashes, as all PCI reads return -1, which most code isn't able
+ *    to handle well enough.
+ *
+ * In all cases dmesg will contain at least one line like this:
+ * 'nouveau 0000:01:00.0: Refused to change power state, currently in D3'
+ * followed by a lot of nouveau timeouts.
+ *
+ * In the \_SB.PCI0.PEG0.PG00._OFF code deeper down writes bit 0x80 to the not
+ * documented PCI config space register 0x248 of the Intel PCIe bridge
+ * controller (0x1901) in order to change the state of the PCIe link between
+ * the PCIe port and the GPU. There are alternative code paths using other
+ * registers, which seem to work fine (executed pre Windows 8):
+ *  - 0xbc bit 0x20 (publicly available documentation claims 'reserved')
+ *  - 0xb0 bit 0x10 (link disable)
+ * Changing the conditions inside the firmware by poking into the relevant
+ * addresses does resolve the issue, but it seemed to be ACPI private memory
+ * and not any device accessible memory at all, so there is no portable way of
+ * changing the conditions.
+ * On a XPS 9560 that means bits [0,3] on \CPEX need to be cleared.
+ *
+ * The only systems where this behavior can be seen are hybrid graphics laptops
+ * with a secondary Nvidia Maxwell, Pascal or Turing GPU. It's unclear whether
+ * this issue only occurs in combination with listed Intel PCIe bridge
+ * controllers and the mentioned GPUs or other devices as well.
+ *
+ * documentation on the PCIe bridge controller can be found in the
+ * "7th Generation Intel® Processor Families for H Platforms Datasheet Volume 2"
+ * Section "12 PCI Express* Controller (x16) Registers"
+ */
+
+static void quirk_broken_nv_runpm(struct nvkm_device_pci *pdev)
+{
+	struct pci_dev *bridge = pci_upstream_bridge(pdev->pdev);
+
+	if (!bridge || bridge->vendor != PCI_VENDOR_ID_INTEL)
+		return;
+
+	switch (bridge->device) {
+	case 0x1901:
+		pdev->old_pm_cap = pdev->pdev->pm_cap;
+		pdev->pdev->pm_cap = 0;
+		nvdev_info(&pdev->device, "Disabling PCI power management to avoid bug\n");
+		break;
+	}
+}
+
 static int
 nvkm_device_pci_probe(struct pci_dev *pci_dev, const struct pci_device_id *id)
 {
@@ -1701,6 +1762,7 @@ nvkm_device_pci_probe(struct pci_dev *pci_dev, const struct pci_device_id *id)
 		pdev->device.mmu->dma_bits = 32;
 	}
 
+	quirk_broken_nv_runpm(pdev);
 done:
 	if (ret) {
 		nvkm_device_del(&device);
-- 
2.44.0



More information about the Nouveau mailing list