drm/nouveau/mmu/gp100-: add privileged methods for fault replay/cancel

Host methods exist to do at least some of what we need, but we are not
currently pushing replay/cancels through a channel like UVM does as it's
not clear whether it's necessary in our case (UVM also updates PTEs with
the GPU).

UVM also pushes a software method for fault cancels on Pascal, seemingly
because the host methods don't appear to be sufficient.  If/when we want
to push the replay/cancel on the GPU, we can re-purpose the cancellation
code here to implement that swmthd.

Keep it simple for now, until we figure out exactly what we need here.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

8 files changed, 101 insertions, 0 deletions

diff --git a/drivers/gpu/drm/nouveau/include/nvif/if000c.h b/drivers/gpu/drm/nouveau/include/nvif/if000c.h
index 20374882ac8f..d6dd40f21eed 100644
--- a/drivers/gpu/drm/nouveau/include/nvif/if000c.h
+++ b/drivers/gpu/drm/nouveau/include/nvif/if000c.h
@@ -17,6 +17,7 @@ struct nvif_vmm_v0 {
 #define NVIF_VMM_V0_UNMAP                                                  0x04
 #define NVIF_VMM_V0_PFNMAP                                                 0x05
 #define NVIF_VMM_V0_PFNCLR                                                 0x06
+#define NVIF_VMM_V0_MTHD(i)                                         ((i) + 0x80)
 
 struct nvif_vmm_page_v0 {
 	__u8  version;
diff --git a/drivers/gpu/drm/nouveau/include/nvif/ifc00d.h b/drivers/gpu/drm/nouveau/include/nvif/ifc00d.h
index 1d9c637859f3..33ff6c791643 100644
--- a/drivers/gpu/drm/nouveau/include/nvif/ifc00d.h
+++ b/drivers/gpu/drm/nouveau/include/nvif/ifc00d.h
@@ -18,4 +18,19 @@ struct gp100_vmm_map_v0 {
 	__u8  priv;
 	__u8  kind;
 };
+
+#define GP100_VMM_VN_FAULT_REPLAY                         NVIF_VMM_V0_MTHD(0x00)
+#define GP100_VMM_VN_FAULT_CANCEL                         NVIF_VMM_V0_MTHD(0x01)
+
+struct gp100_vmm_fault_replay_vn {
+};
+
+struct gp100_vmm_fault_cancel_v0 {
+	__u8  version;
+	__u8  hub;
+	__u8  gpc;
+	__u8  client;
+	__u8  pad04[4];
+	__u64 inst;
+};
 #endif
diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/mmu/uvmm.c b/drivers/gpu/drm/nouveau/nvkm/subdev/mmu/uvmm.c
index b483cda151bf..c43b8248c682 100644
--- a/drivers/gpu/drm/nouveau/nvkm/subdev/mmu/uvmm.c
+++ b/drivers/gpu/drm/nouveau/nvkm/subdev/mmu/uvmm.c
@@ -341,6 +341,13 @@ nvkm_uvmm_mthd(struct nvkm_object *object, u32 mthd, void *argv, u32 argc)
 	case NVIF_VMM_V0_UNMAP : return nvkm_uvmm_mthd_unmap (uvmm, argv, argc);
 	case NVIF_VMM_V0_PFNMAP: return nvkm_uvmm_mthd_pfnmap(uvmm, argv, argc);
 	case NVIF_VMM_V0_PFNCLR: return nvkm_uvmm_mthd_pfnclr(uvmm, argv, argc);
+	case NVIF_VMM_V0_MTHD(0x00) ... NVIF_VMM_V0_MTHD(0x7f):
+		if (uvmm->vmm->func->mthd) {
+			return uvmm->vmm->func->mthd(uvmm->vmm,
+						     uvmm->object.client,
+						     mthd, argv, argc);
+		}
+		break;
 	default:
 		break;
 	}
diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/mmu/vmm.h b/drivers/gpu/drm/nouveau/nvkm/subdev/mmu/vmm.h
index 3c74bba03138..e85f19fdc597 100644
--- a/drivers/gpu/drm/nouveau/nvkm/subdev/mmu/vmm.h
+++ b/drivers/gpu/drm/nouveau/nvkm/subdev/mmu/vmm.h
@@ -145,6 +145,9 @@ struct nvkm_vmm_func {
 		     struct nvkm_vmm_map *);
 	void (*flush)(struct nvkm_vmm *, int depth);
 
+	int (*mthd)(struct nvkm_vmm *, struct nvkm_client *,
+		    u32 mthd, void *argv, u32 argc);
+
 	void (*invalidate_pdb)(struct nvkm_vmm *, u64 addr);
 
 	u64 page_block;
@@ -220,6 +223,7 @@ int gm200_vmm_join(struct nvkm_vmm *, struct nvkm_memory *);
 int gp100_vmm_join(struct nvkm_vmm *, struct nvkm_memory *);
 int gp100_vmm_valid(struct nvkm_vmm *, void *, u32, struct nvkm_vmm_map *);
 void gp100_vmm_flush(struct nvkm_vmm *, int);
+int gp100_vmm_mthd(struct nvkm_vmm *, struct nvkm_client *, u32, void *, u32);
 void gp100_vmm_invalidate_pdb(struct nvkm_vmm *, u64 addr);
 
 int gv100_vmm_join(struct nvkm_vmm *, struct nvkm_memory *);
diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/mmu/vmmgp100.c b/drivers/gpu/drm/nouveau/nvkm/subdev/mmu/vmmgp100.c
index 84c855130cb5..af427daf9077 100644
--- a/drivers/gpu/drm/nouveau/nvkm/subdev/mmu/vmmgp100.c
+++ b/drivers/gpu/drm/nouveau/nvkm/subdev/mmu/vmmgp100.c
@@ -21,8 +21,11 @@
  */
 #include "vmm.h"
 
+#include <core/client.h>
 #include <subdev/fb.h>
 #include <subdev/ltc.h>
+#include <subdev/timer.h>
+#include <engine/gr.h>
 
 #include <nvif/ifc00d.h>
 #include <nvif/unpack.h>
@@ -384,6 +387,73 @@ gp100_vmm_valid(struct nvkm_vmm *vmm, void *argv, u32 argc,
 	return 0;
 }
 
+static int
+gp100_vmm_fault_cancel(struct nvkm_vmm *vmm, void *argv, u32 argc)
+{
+	struct nvkm_device *device = vmm->mmu->subdev.device;
+	union {
+		struct gp100_vmm_fault_cancel_v0 v0;
+	} *args = argv;
+	int ret = -ENOSYS;
+	u32 inst, aper;
+
+	if ((ret = nvif_unpack(ret, &argv, &argc, args->v0, 0, 0, false)))
+		return ret;
+
+	/* Translate MaxwellFaultBufferA instance pointer to the same
+	 * format as the NV_GR_FECS_CURRENT_CTX register.
+	 */
+	aper = (args->v0.inst >> 8) & 3;
+	args->v0.inst >>= 12;
+	args->v0.inst |= aper << 28;
+	args->v0.inst |= 0x80000000;
+
+	if (!WARN_ON(nvkm_gr_ctxsw_pause(device))) {
+		if ((inst = nvkm_gr_ctxsw_inst(device)) == args->v0.inst) {
+			gf100_vmm_invalidate(vmm, 0x0000001b
+					     /* CANCEL_TARGETED. */ |
+					     (args->v0.hub    << 20) |
+					     (args->v0.gpc    << 15) |
+					     (args->v0.client << 9));
+		}
+		WARN_ON(nvkm_gr_ctxsw_resume(device));
+	}
+
+	return 0;
+}
+
+static int
+gp100_vmm_fault_replay(struct nvkm_vmm *vmm, void *argv, u32 argc)
+{
+	union {
+		struct gp100_vmm_fault_replay_vn vn;
+	} *args = argv;
+	int ret = -ENOSYS;
+
+	if (!(ret = nvif_unvers(ret, &argv, &argc, args->vn))) {
+		gf100_vmm_invalidate(vmm, 0x0000000b); /* REPLAY_GLOBAL. */
+	}
+
+	return ret;
+}
+
+int
+gp100_vmm_mthd(struct nvkm_vmm *vmm,
+	       struct nvkm_client *client, u32 mthd, void *argv, u32 argc)
+{
+	if (client->super) {
+		switch (mthd) {
+		case GP100_VMM_VN_FAULT_REPLAY:
+			return gp100_vmm_fault_replay(vmm, argv, argc);
+		case GP100_VMM_VN_FAULT_CANCEL:
+			return gp100_vmm_fault_cancel(vmm, argv, argc);
+		default:
+			break;
+		}
+	}
+	return -EINVAL;
+}
+
 void
 gp100_vmm_invalidate_pdb(struct nvkm_vmm *vmm, u64 addr)
 {
@@ -417,6 +487,7 @@ gp100_vmm = {
 	.aper = gf100_vmm_aper,
 	.valid = gp100_vmm_valid,
 	.flush = gp100_vmm_flush,
+	.mthd = gp100_vmm_mthd,
 	.invalidate_pdb = gp100_vmm_invalidate_pdb,
 	.page = {
 		{ 47, &gp100_vmm_desc_16[4], NVKM_VMM_PAGE_Sxxx },
diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/mmu/vmmgp10b.c b/drivers/gpu/drm/nouveau/nvkm/subdev/mmu/vmmgp10b.c
index 07b91d835e18..317a83e40def 100644
--- a/drivers/gpu/drm/nouveau/nvkm/subdev/mmu/vmmgp10b.c
+++ b/drivers/gpu/drm/nouveau/nvkm/subdev/mmu/vmmgp10b.c
@@ -28,6 +28,7 @@ gp10b_vmm = {
 	.aper = gk20a_vmm_aper,
 	.valid = gp100_vmm_valid,
 	.flush = gp100_vmm_flush,
+	.mthd = gp100_vmm_mthd,
 	.invalidate_pdb = gp100_vmm_invalidate_pdb,
 	.page = {
 		{ 47, &gp100_vmm_desc_16[4], NVKM_VMM_PAGE_Sxxx },
diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/mmu/vmmgv100.c b/drivers/gpu/drm/nouveau/nvkm/subdev/mmu/vmmgv100.c
index da5841aad70a..172d685932d0 100644
--- a/drivers/gpu/drm/nouveau/nvkm/subdev/mmu/vmmgv100.c
+++ b/drivers/gpu/drm/nouveau/nvkm/subdev/mmu/vmmgv100.c
@@ -66,6 +66,7 @@ gv100_vmm = {
 	.aper = gf100_vmm_aper,
 	.valid = gp100_vmm_valid,
 	.flush = gp100_vmm_flush,
+	.mthd = gp100_vmm_mthd,
 	.invalidate_pdb = gp100_vmm_invalidate_pdb,
 	.page = {
 		{ 47, &gp100_vmm_desc_16[4], NVKM_VMM_PAGE_Sxxx },
diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/mmu/vmmtu102.c b/drivers/gpu/drm/nouveau/nvkm/subdev/mmu/vmmtu102.c
index ddab7c6bfe7c..3bfef146dd1f 100644
--- a/drivers/gpu/drm/nouveau/nvkm/subdev/mmu/vmmtu102.c
+++ b/drivers/gpu/drm/nouveau/nvkm/subdev/mmu/vmmtu102.c
@@ -56,6 +56,7 @@ tu102_vmm = {
 	.aper = gf100_vmm_aper,
 	.valid = gp100_vmm_valid,
 	.flush = tu102_vmm_flush,
+	.mthd = gp100_vmm_mthd,
 	.page = {
 		{ 47, &gp100_vmm_desc_16[4], NVKM_VMM_PAGE_Sxxx },
 		{ 38, &gp100_vmm_desc_16[3], NVKM_VMM_PAGE_Sxxx },