Merge branch 'linux-5.7' of git://github.com/skeggsb/linux into drm-next

airlied · airlied · commit 9c346968b862 · 2020-04-08T09:16:21.000+10:00
A couple of misc fixes/workarounds for some issues that are causing a lot of pain for people. Of most interest are the PCI power management and GR init WARs, which effect a rather significant number of laptop systems that are in use today. Signed-off-by: Dave Airlie <airlied@redhat.com> From: Ben Skeggs <skeggsb@gmail.com> Link: https://patchwork.freedesktop.org/patch/msgid/ <CACAvsv5Ef5YKS9EPBH3YUubzvVr++_rzjgSqV_B5nC0L2kB6-Q@mail.gmail.com
diff --git a/drivers/gpu/drm/nouveau/dispnv04/dac.c b/drivers/gpu/drm/nouveau/dispnv04/dac.c
@@ -35,7 +35,8 @@
 
 #include <subdev/bios/gpio.h>
 #include <subdev/gpio.h>
-#include <subdev/timer.h>
+
+#include <nvif/timer.h>
 
 int nv04_dac_output_offset(struct drm_encoder *encoder)
 {
diff --git a/drivers/gpu/drm/nouveau/dispnv04/hw.c b/drivers/gpu/drm/nouveau/dispnv04/hw.c
@@ -26,6 +26,7 @@
 #include "hw.h"
 
 #include <subdev/bios/pll.h>
+#include <nvif/timer.h>
 
 #define CHIPSET_NFORCE 0x01a0
 #define CHIPSET_NFORCE2 0x01f0
diff --git a/drivers/gpu/drm/nouveau/dispnv50/base507c.c b/drivers/gpu/drm/nouveau/dispnv50/base507c.c
@@ -23,6 +23,7 @@
 
 #include <nvif/cl507c.h>
 #include <nvif/event.h>
+#include <nvif/timer.h>
 
 #include <drm/drm_atomic_helper.h>
 #include <drm/drm_fourcc.h>
diff --git a/drivers/gpu/drm/nouveau/dispnv50/core507d.c b/drivers/gpu/drm/nouveau/dispnv50/core507d.c
@@ -23,6 +23,7 @@
 #include "head.h"
 
 #include <nvif/cl507d.h>
+#include <nvif/timer.h>
 
 #include "nouveau_bo.h"
 
diff --git a/drivers/gpu/drm/nouveau/dispnv50/corec37d.c b/drivers/gpu/drm/nouveau/dispnv50/corec37d.c
@@ -24,6 +24,8 @@
 
 #include <nouveau_bo.h>
 
+#include <nvif/timer.h>
+
 void
 corec37d_wndw_owner(struct nv50_core *core)
 {
diff --git a/drivers/gpu/drm/nouveau/dispnv50/curs507a.c b/drivers/gpu/drm/nouveau/dispnv50/curs507a.c
@@ -24,21 +24,36 @@
 #include "head.h"
 
 #include <nvif/cl507a.h>
+#include <nvif/timer.h>
 
 #include <drm/drm_atomic_helper.h>
 #include <drm/drm_plane_helper.h>
 
+bool
+curs507a_space(struct nv50_wndw *wndw)
+{
+	nvif_msec(&nouveau_drm(wndw->plane.dev)->client.device, 2,
+		if (nvif_rd32(&wndw->wimm.base.user, 0x0008) >= 4)
+			return true;
+	);
+	WARN_ON(1);
+	return false;
+}
+
 static void
 curs507a_update(struct nv50_wndw *wndw, u32 *interlock)
 {
-	nvif_wr32(&wndw->wimm.base.user, 0x0080, 0x00000000);
+	if (curs507a_space(wndw))
+		nvif_wr32(&wndw->wimm.base.user, 0x0080, 0x00000000);
 }
 
 static void
 curs507a_point(struct nv50_wndw *wndw, struct nv50_wndw_atom *asyw)
 {
-	nvif_wr32(&wndw->wimm.base.user, 0x0084, asyw->point.y << 16 |
-						 asyw->point.x);
+	if (curs507a_space(wndw)) {
+		nvif_wr32(&wndw->wimm.base.user, 0x0084, asyw->point.y << 16 |
+							 asyw->point.x);
+	}
 }
 
 const struct nv50_wimm_func
diff --git a/drivers/gpu/drm/nouveau/dispnv50/cursc37a.c b/drivers/gpu/drm/nouveau/dispnv50/cursc37a.c
@@ -25,14 +25,17 @@
 static void
 cursc37a_update(struct nv50_wndw *wndw, u32 *interlock)
 {
-	nvif_wr32(&wndw->wimm.base.user, 0x0200, 0x00000001);
+	if (curs507a_space(wndw))
+		nvif_wr32(&wndw->wimm.base.user, 0x0200, 0x00000001);
 }
 
 static void
 cursc37a_point(struct nv50_wndw *wndw, struct nv50_wndw_atom *asyw)
 {
-	nvif_wr32(&wndw->wimm.base.user, 0x0208, asyw->point.y << 16 |
-						 asyw->point.x);
+	if (curs507a_space(wndw)) {
+		nvif_wr32(&wndw->wimm.base.user, 0x0208, asyw->point.y << 16 |
+							 asyw->point.x);
+	}
 }
 
 static const struct nv50_wimm_func
diff --git a/drivers/gpu/drm/nouveau/dispnv50/disp.c b/drivers/gpu/drm/nouveau/dispnv50/disp.c
@@ -45,6 +45,7 @@
 #include <nvif/cl5070.h>
 #include <nvif/cl507d.h>
 #include <nvif/event.h>
+#include <nvif/timer.h>
 
 #include "nouveau_drv.h"
 #include "nouveau_dma.h"
diff --git a/drivers/gpu/drm/nouveau/dispnv50/ovly827e.c b/drivers/gpu/drm/nouveau/dispnv50/ovly827e.c
@@ -24,6 +24,8 @@
 
 #include <nouveau_bo.h>
 
+#include <nvif/timer.h>
+
 static void
 ovly827e_image_set(struct nv50_wndw *wndw, struct nv50_wndw_atom *asyw)
 {
diff --git a/drivers/gpu/drm/nouveau/dispnv50/wndw.h b/drivers/gpu/drm/nouveau/dispnv50/wndw.h
@@ -97,6 +97,7 @@ struct nv50_wimm_func {
 };
 
 extern const struct nv50_wimm_func curs507a;
+bool curs507a_space(struct nv50_wndw *);
 
 int wndwc37e_new(struct nouveau_drm *, enum drm_plane_type, int, s32,
 		 struct nv50_wndw **);
diff --git a/drivers/gpu/drm/nouveau/include/nvif/device.h b/drivers/gpu/drm/nouveau/include/nvif/device.h
@@ -23,27 +23,6 @@ int  nvif_device_init(struct nvif_object *, u32 handle, s32 oclass, void *, u32,
 void nvif_device_fini(struct nvif_device *);
 u64  nvif_device_time(struct nvif_device *);
 
-/* Delay based on GPU time (ie. PTIMER).
- *
- * Will return -ETIMEDOUT unless the loop was terminated with 'break',
- * where it will return the number of nanoseconds taken instead.
- */
-#define nvif_nsec(d,n,cond...) ({                                              \
-	struct nvif_device *_device = (d);                                     \
-	u64 _nsecs = (n), _time0 = nvif_device_time(_device);                  \
-	s64 _taken = 0;                                                        \
-                                                                               \
-	do {                                                                   \
-		cond                                                           \
-	} while (_taken = nvif_device_time(_device) - _time0, _taken < _nsecs);\
-                                                                               \
-	if (_taken >= _nsecs)                                                  \
-		_taken = -ETIMEDOUT;                                           \
-	_taken;                                                                \
-})
-#define nvif_usec(d,u,cond...) nvif_nsec((d), (u) * 1000, ##cond)
-#define nvif_msec(d,m,cond...) nvif_usec((d), (m) * 1000, ##cond)
-
 /*XXX*/
 #include <subdev/bios.h>
 #include <subdev/fb.h>
diff --git a/drivers/gpu/drm/nouveau/include/nvif/timer.h b/drivers/gpu/drm/nouveau/include/nvif/timer.h
@@ -0,0 +1,35 @@
+#ifndef __NVIF_TIMER_H__
+#define __NVIF_TIMER_H__
+#include <nvif/os.h>
+
+struct nvif_timer_wait {
+	struct nvif_device *device;
+	u64 limit;
+	u64 time0;
+	u64 time1;
+	int reads;
+};
+
+void nvif_timer_wait_init(struct nvif_device *, u64 nsec,
+			  struct nvif_timer_wait *);
+s64 nvif_timer_wait_test(struct nvif_timer_wait *);
+
+/* Delay based on GPU time (ie. PTIMER).
+ *
+ * Will return -ETIMEDOUT unless the loop was terminated with 'break',
+ * where it will return the number of nanoseconds taken instead.
+ */
+#define nvif_nsec(d,n,cond...) ({                                              \
+	struct nvif_timer_wait _wait;                                          \
+	s64 _taken = 0;                                                        \
+                                                                               \
+	nvif_timer_wait_init((d), (n), &_wait);                                \
+	do {                                                                   \
+		cond                                                           \
+	} while ((_taken = nvif_timer_wait_test(&_wait)) >= 0);                \
+                                                                               \
+	_taken;                                                                \
+})
+#define nvif_usec(d,u,cond...) nvif_nsec((d), (u) * 1000, ##cond)
+#define nvif_msec(d,m,cond...) nvif_usec((d), (m) * 1000, ##cond)
+#endif
diff --git a/drivers/gpu/drm/nouveau/include/nvif/user.h b/drivers/gpu/drm/nouveau/include/nvif/user.h
@@ -10,6 +10,7 @@ struct nvif_user {
 
 struct nvif_user_func {
 	void (*doorbell)(struct nvif_user *, u32 token);
+	u64 (*time)(struct nvif_user *);
 };
 
 int nvif_user_init(struct nvif_device *);
diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c b/drivers/gpu/drm/nouveau/nouveau_bo.c
@@ -1494,8 +1494,13 @@ nouveau_ttm_io_mem_reserve(struct ttm_bo_device *bdev, struct ttm_mem_reg *reg)
 			ret = nvif_object_map_handle(&mem->mem.object,
 						     &args, argc,
 						     &handle, &length);
-			if (ret != 1)
-				return ret ? ret : -EINVAL;
+			if (ret != 1) {
+				if (WARN_ON(ret == 0))
+					return -EINVAL;
+				if (ret == -ENOSPC)
+					return -EAGAIN;
+				return ret;
+			}
 
 			reg->bus.base = 0;
 			reg->bus.offset = handle;
diff --git a/drivers/gpu/drm/nouveau/nouveau_debugfs.c b/drivers/gpu/drm/nouveau/nouveau_debugfs.c
@@ -222,22 +222,18 @@ nouveau_drm_debugfs_init(struct drm_minor *minor)
 {
 	struct nouveau_drm *drm = nouveau_drm(minor->dev);
 	struct dentry *dentry;
-	int i, ret;
+	int i;
 
 	for (i = 0; i < ARRAY_SIZE(nouveau_debugfs_files); i++) {
-		dentry = debugfs_create_file(nouveau_debugfs_files[i].name,
-					     S_IRUGO | S_IWUSR,
-					     minor->debugfs_root, minor->dev,
-					     nouveau_debugfs_files[i].fops);
-		if (!dentry)
-			return -ENOMEM;
+		debugfs_create_file(nouveau_debugfs_files[i].name,
+				    S_IRUGO | S_IWUSR,
+				    minor->debugfs_root, minor->dev,
+				    nouveau_debugfs_files[i].fops);
 	}
 
-	ret = drm_debugfs_create_files(nouveau_debugfs_list,
-				       NOUVEAU_DEBUGFS_ENTRIES,
-				       minor->debugfs_root, minor);
-	if (ret)
-		return ret;
+	drm_debugfs_create_files(nouveau_debugfs_list,
+				 NOUVEAU_DEBUGFS_ENTRIES,
+				 minor->debugfs_root, minor);
 
 	/* Set the size of the vbios since we know it, and it's confusing to
 	 * userspace if it wants to seek() but the file has a length of 0
diff --git a/drivers/gpu/drm/nouveau/nouveau_drm.c b/drivers/gpu/drm/nouveau/nouveau_drm.c
@@ -618,6 +618,64 @@ nouveau_drm_device_fini(struct drm_device *dev)
 	kfree(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 drm_device *dev = pci_get_drvdata(pdev);
+	struct nouveau_drm *drm = nouveau_drm(dev);
+	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)
 {
@@ -699,6 +757,7 @@ static int nouveau_drm_probe(struct pci_dev *pdev,
 	if (ret)
 		goto fail_drm_dev_init;
 
+	quirk_broken_nv_runpm(pdev);
 	return 0;
 
 fail_drm_dev_init:
@@ -734,7 +793,11 @@ static void
 nouveau_drm_remove(struct pci_dev *pdev)
 {
 	struct drm_device *dev = pci_get_drvdata(pdev);
+	struct nouveau_drm *drm = nouveau_drm(dev);
 
+	/* revert our workaround */
+	if (drm->old_pm_cap)
+		pdev->pm_cap = drm->old_pm_cap;
 	nouveau_drm_device_remove(dev);
 	pci_disable_device(pdev);
 }
diff --git a/drivers/gpu/drm/nouveau/nouveau_drv.h b/drivers/gpu/drm/nouveau/nouveau_drv.h
@@ -140,6 +140,8 @@ struct nouveau_drm {
 
 	struct list_head clients;
 
+	u8 old_pm_cap;
+
 	struct {
 		struct agp_bridge_data *bridge;
 		u32 base;
diff --git a/drivers/gpu/drm/nouveau/nouveau_svm.c b/drivers/gpu/drm/nouveau/nouveau_svm.c
@@ -171,6 +171,11 @@ nouveau_svmm_bind(struct drm_device *dev, void *data,
 	mm = get_task_mm(current);
 	down_read(&mm->mmap_sem);
 
+	if (!cli->svm.svmm) {
+		up_read(&mm->mmap_sem);
+		return -EINVAL;
+	}
+
 	for (addr = args->va_start, end = args->va_start + size; addr < end;) {
 		struct vm_area_struct *vma;
 		unsigned long next;
@@ -179,6 +184,7 @@ nouveau_svmm_bind(struct drm_device *dev, void *data,
 		if (!vma)
 			break;
 
+		addr = max(addr, vma->vm_start);
 		next = min(vma->vm_end, end);
 		/* This is a best effort so we ignore errors */
 		nouveau_dmem_migrate_vma(cli->drm, vma, addr, next);
@@ -657,9 +663,6 @@ nouveau_svm_fault(struct nvif_notify *notify)
 		limit = start + (ARRAY_SIZE(args.phys) << PAGE_SHIFT);
 		if (start < svmm->unmanaged.limit)
 			limit = min_t(u64, limit, svmm->unmanaged.start);
-		else
-		if (limit > svmm->unmanaged.start)
-			start = max_t(u64, start, svmm->unmanaged.limit);
 		SVMM_DBG(svmm, "wndw %016llx-%016llx", start, limit);
 
 		mm = svmm->notifier.mm;
diff --git a/drivers/gpu/drm/nouveau/nvif/Kbuild b/drivers/gpu/drm/nouveau/nvif/Kbuild
@@ -8,6 +8,7 @@ nvif-y += nvif/fifo.o
 nvif-y += nvif/mem.o
 nvif-y += nvif/mmu.o
 nvif-y += nvif/notify.o
+nvif-y += nvif/timer.o
 nvif-y += nvif/vmm.o
 
 # Usermode classes
diff --git a/drivers/gpu/drm/nouveau/nvif/device.c b/drivers/gpu/drm/nouveau/nvif/device.c
diff --git a/drivers/gpu/drm/nouveau/nvif/timer.c b/drivers/gpu/drm/nouveau/nvif/timer.c
diff --git a/drivers/gpu/drm/nouveau/nvif/userc361.c b/drivers/gpu/drm/nouveau/nvif/userc361.c
diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf100.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf100.c

Original file line number	Diff line number	Diff line change
`@@ -35,7 +35,8 @@`
`35`	`35`
`36`	`36`	`#include <subdev/bios/gpio.h>`
`37`	`37`	`#include <subdev/gpio.h>`
`38`		`-#include <subdev/timer.h>`
	`38`	`+`
	`39`	`+#include <nvif/timer.h>`
`39`	`40`
`40`	`41`	`int nv04_dac_output_offset(struct drm_encoder *encoder)`
`41`	`42`	`{`
Original file line number	Diff line number	Diff line change
`@@ -24,6 +24,8 @@`
`24`	`24`
`25`	`25`	`#include <nouveau_bo.h>`
`26`	`26`
	`27`	`+#include <nvif/timer.h>`
	`28`	`+`
`27`	`29`	`void`
`28`	`30`	`corec37d_wndw_owner(struct nv50_core *core)`
`29`	`31`	`{`
Original file line number	Diff line number	Diff line change
`@@ -25,14 +25,17 @@`
`25`	`25`	`static void`
`26`	`26`	`cursc37a_update(struct nv50_wndw wndw, u32 interlock)`
`27`	`27`	`{`
`28`		`- nvif_wr32(&wndw->wimm.base.user, 0x0200, 0x00000001);`
	`28`	`+ if (curs507a_space(wndw))`
	`29`	`+ nvif_wr32(&wndw->wimm.base.user, 0x0200, 0x00000001);`
`29`	`30`	`}`
`30`	`31`
`31`	`32`	`static void`
`32`	`33`	`cursc37a_point(struct nv50_wndw wndw, struct nv50_wndw_atom asyw)`
`33`	`34`	`{`
`34`		`- nvif_wr32(&wndw->wimm.base.user, 0x0208, asyw->point.y << 16 \|`
`35`		`- asyw->point.x);`
	`35`	`+ if (curs507a_space(wndw)) {`
	`36`	`+ nvif_wr32(&wndw->wimm.base.user, 0x0208, asyw->point.y << 16 \|`
	`37`	`+ asyw->point.x);`
	`38`	`+ }`
`36`	`39`	`}`
`37`	`40`
`38`	`41`	`static const struct nv50_wimm_func`