nouveau GSP fixes

nouveau GSP fixes

[Dave Airlie]

This is a collection of nouveau debug prints, memory leak, a very
annoying race condition causing system hangs with prime scenarios,
and a fix from Lyude to get the panel on my laptop working.

I'd like to get these into 6.7,

Dave.

[PATCH 01/11] nouveau/gsp: add three notifier callbacks that we see in normal operation (v2)

[Dave Airlie]

Add NULL callbacks for some things GSP calls that we don't handle, but know about
so we avoid the logging.

v2: Timur suggested allowing null fn.

Signed-off-by: Dave Airlie <airlied@redhat.com>
---
 drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c b/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c
index 44fb86841c05..7f831f41b598 100644
--- a/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c
+++ b/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c
@@ -298,7 +298,8 @@ r535_gsp_msg_recv(struct nvkm_gsp *gsp, int fn, u32 repc)
 		struct nvkm_gsp_msgq_ntfy *ntfy = &gsp->msgq.ntfy[i];
 
 		if (ntfy->fn == msg->function) {
-			ntfy->func(ntfy->priv, ntfy->fn, msg->data, msg->length - sizeof(*msg));
+			if (ntfy->func)
+				ntfy->func(ntfy->priv, ntfy->fn, msg->data, msg->length - sizeof(*msg));
 			break;
 		}
 	}
@@ -2186,7 +2187,9 @@ r535_gsp_oneinit(struct nvkm_gsp *gsp)
 	r535_gsp_msg_ntfy_add(gsp, NV_VGPU_MSG_EVENT_MMU_FAULT_QUEUED,
 			      r535_gsp_msg_mmu_fault_queued, gsp);
 	r535_gsp_msg_ntfy_add(gsp, NV_VGPU_MSG_EVENT_OS_ERROR_LOG, r535_gsp_msg_os_error_log, gsp);
-
+	r535_gsp_msg_ntfy_add(gsp, NV_VGPU_MSG_EVENT_PERF_BRIDGELESS_INFO_UPDATE, NULL, NULL);
+	r535_gsp_msg_ntfy_add(gsp, NV_VGPU_MSG_EVENT_UCODE_LIBOS_PRINT, NULL, NULL);
+	r535_gsp_msg_ntfy_add(gsp, NV_VGPU_MSG_EVENT_GSP_SEND_USER_SHARED_DATA, NULL, NULL);
 	ret = r535_gsp_rm_boot_ctor(gsp);
 	if (ret)
 		return ret;
-- 
2.43.0

[PATCH 02/11] nouveau/gsp: drop some acpi related debug

[Dave Airlie]

These were leftover debug, if we need to bring them back do so
for debugging later.

Signed-off-by: Dave Airlie <airlied@redhat.com>
---
 drivers/gpu/drm/nouveau/nvkm/engine/disp/r535.c | 7 -------
 drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c  | 9 ---------
 2 files changed, 16 deletions(-)

diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/r535.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/r535.c
index 298035070b3a..1c8c4cca0957 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/r535.c
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/r535.c
@@ -1465,8 +1465,6 @@ r535_disp_oneinit(struct nvkm_disp *disp)
 				bool nvhg = acpi_check_dsm(handle, &NVHG_DSM_GUID, NVHG_DSM_REV,
 						           1ULL << 0x00000014);
 
-				printk(KERN_ERR "bl: nbci:%d nvhg:%d\n", nbci, nvhg);
-
 				if (nbci || nvhg) {
 					union acpi_object argv4 = {
 						.buffer.type    = ACPI_TYPE_BUFFER,
@@ -1479,9 +1477,6 @@ r535_disp_oneinit(struct nvkm_disp *disp)
 					if (!obj) {
 						acpi_handle_info(handle, "failed to evaluate _DSM\n");
 					} else {
-						printk(KERN_ERR "bl: obj type %d\n", obj->type);
-						printk(KERN_ERR "bl: obj len %d\n", obj->package.count);
-
 						for (int i = 0; i < obj->package.count; i++) {
 							union acpi_object *elt = &obj->package.elements[i];
 							u32 size;
@@ -1491,12 +1486,10 @@ r535_disp_oneinit(struct nvkm_disp *disp)
 							else
 								size = 4;
 
-							printk(KERN_ERR "elt %03d: type %d size %d\n", i, elt->type, size);
 							memcpy(&ctrl->backLightData[ctrl->backLightDataSize], &elt->integer.value, size);
 							ctrl->backLightDataSize += size;
 						}
 
-						printk(KERN_ERR "bl: data size %d\n", ctrl->backLightDataSize);
 						ctrl->status = 0;
 						ACPI_FREE(obj);
 					}
diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c b/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c
index 7f831f41b598..365dda6c002a 100644
--- a/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c
+++ b/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c
@@ -1100,16 +1100,12 @@ r535_gsp_acpi_caps(acpi_handle handle, CAPS_METHOD_DATA *caps)
 	if (!obj)
 		return;
 
-	printk(KERN_ERR "nvop: obj type %d\n", obj->type);
-	printk(KERN_ERR "nvop: obj len %d\n", obj->buffer.length);
-
 	if (WARN_ON(obj->type != ACPI_TYPE_BUFFER) ||
 	    WARN_ON(obj->buffer.length != 4))
 		return;
 
 	caps->status = 0;
 	caps->optimusCaps = *(u32 *)obj->buffer.pointer;
-	printk(KERN_ERR "nvop: caps %08x\n", caps->optimusCaps);
 
 	ACPI_FREE(obj);
 
@@ -1136,9 +1132,6 @@ r535_gsp_acpi_jt(acpi_handle handle, JT_METHOD_DATA *jt)
 	if (!obj)
 		return;
 
-	printk(KERN_ERR "jt: obj type %d\n", obj->type);
-	printk(KERN_ERR "jt: obj len %d\n", obj->buffer.length);
-
 	if (WARN_ON(obj->type != ACPI_TYPE_BUFFER) ||
 	    WARN_ON(obj->buffer.length != 4))
 		return;
@@ -1147,7 +1140,6 @@ r535_gsp_acpi_jt(acpi_handle handle, JT_METHOD_DATA *jt)
 	jt->jtCaps = *(u32 *)obj->buffer.pointer;
 	jt->jtRevId = (jt->jtCaps & 0xfff00000) >> 20;
 	jt->bSBIOSCaps = 0;
-	printk(KERN_ERR "jt: caps %08x rev:%04x\n", jt->jtCaps, jt->jtRevId);
 
 	ACPI_FREE(obj);
 
@@ -1233,7 +1225,6 @@ r535_gsp_acpi_dod(acpi_handle handle, DOD_METHOD_DATA *dod)
 		dod->acpiIdListLen += sizeof(dod->acpiIdList[0]);
 	}
 
-	printk(KERN_ERR "_DOD: ok! len:%d\n", dod->acpiIdListLen);
 	dod->status = 0;
 }
 #endif
-- 
2.43.0

[PATCH 03/11] nouveau: fix disp disabling with GSP

[Dave Airlie]

This func ptr here is normally static allocation, but gsp r535
uses a dynamic pointer, so we need to handle that better.

This fixes a crash with GSP when you use config=disp=0 to avoid
disp problems.

Signed-off-by: Dave Airlie <airlied@redhat.com>
---
 drivers/gpu/drm/nouveau/nvkm/engine/disp/base.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/base.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/base.c
index 457ec5db794d..b24eb1e560bc 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/base.c
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/base.c
@@ -209,7 +209,7 @@ nvkm_disp_dtor(struct nvkm_engine *engine)
 		nvkm_head_del(&head);
 	}
 
-	if (disp->func->dtor)
+	if (disp->func && disp->func->dtor)
 		disp->func->dtor(disp);
 
 	return data;
@@ -243,8 +243,10 @@ nvkm_disp_new_(const struct nvkm_disp_func *func, struct nvkm_device *device,
 	spin_lock_init(&disp->client.lock);
 
 	ret = nvkm_engine_ctor(&nvkm_disp, device, type, inst, true, &disp->engine);
-	if (ret)
+	if (ret) {
+		disp->func = NULL;
 		return ret;
+	}
 
 	if (func->super) {
 		disp->super.wq = create_singlethread_workqueue("nvkm-disp");
-- 
2.43.0

[PATCH 04/11] nouveau/gsp: free acpi object after use

[Dave Airlie]

This fixes a memory leak for the acpi dod object.

Signed-off-by: Dave Airlie <airlied@redhat.com>
---
 drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c b/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c
index 365dda6c002a..1a6d7c89660d 100644
--- a/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c
+++ b/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c
@@ -1226,6 +1226,7 @@ r535_gsp_acpi_dod(acpi_handle handle, DOD_METHOD_DATA *dod)
 	}
 
 	dod->status = 0;
+	kfree(output.pointer);
 }
 #endif
 
-- 
2.43.0

[PATCH 05/11] nouveau/gsp: free userd allocation.

[Dave Airlie]

This was being leaked.

Signed-off-by: Dave Airlie <airlied@redhat.com>
---
 drivers/gpu/drm/nouveau/nvkm/engine/fifo/r535.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/r535.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/r535.c
index d088e636edc3..b903785056b5 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/r535.c
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/r535.c
@@ -242,6 +242,7 @@ r535_chan_id_put(struct nvkm_chan *chan)
 				nvkm_memory_unref(&userd->mem);
 				nvkm_chid_put(runl->chid, userd->chid, &chan->cgrp->lock);
 				list_del(&userd->head);
+				kfree(userd);
 			}
 
 			break;
-- 
2.43.0

[PATCH 06/11] drm/nouveau/gsp: Fix ACPI MXDM/MXDS method invocations

[Dave Airlie]

From: Lyude Paul <lyude@redhat.com>

Currently we get an error from ACPI because both of these arguments expect
a single argument, and we don't provide one. I'm not totally clear on what
that argument does, but we're able to find the missing value from
_acpiCacheMethodData() in src/kernel/platform/acpi_common.c in nvidia's
driver. So, let's add that - which doesn't get eDP displays to power on
quite yet, but gets rid of the argument warning at least.

Signed-off-by: Lyude Paul <lyude@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
---
 drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c b/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c
index 1a6d7c89660d..774ca47b019f 100644
--- a/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c
+++ b/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c
@@ -1150,6 +1150,8 @@ static void
 r535_gsp_acpi_mux_id(acpi_handle handle, u32 id, MUX_METHOD_DATA_ELEMENT *mode,
 						 MUX_METHOD_DATA_ELEMENT *part)
 {
+	union acpi_object mux_arg = { ACPI_TYPE_INTEGER };
+	struct acpi_object_list input = { 1, &mux_arg };
 	acpi_handle iter = NULL, handle_mux = NULL;
 	acpi_status status;
 	unsigned long long value;
@@ -1172,14 +1174,18 @@ r535_gsp_acpi_mux_id(acpi_handle handle, u32 id, MUX_METHOD_DATA_ELEMENT *mode,
 	if (!handle_mux)
 		return;
 
-	status = acpi_evaluate_integer(handle_mux, "MXDM", NULL, &value);
+	/* I -think- 0 means "acquire" according to nvidia's driver source */
+	input.pointer->integer.type = ACPI_TYPE_INTEGER;
+	input.pointer->integer.value = 0;
+
+	status = acpi_evaluate_integer(handle_mux, "MXDM", &input, &value);
 	if (ACPI_SUCCESS(status)) {
 		mode->acpiId = id;
 		mode->mode   = value;
 		mode->status = 0;
 	}
 
-	status = acpi_evaluate_integer(handle_mux, "MXDS", NULL, &value);
+	status = acpi_evaluate_integer(handle_mux, "MXDS", &input, &value);
 	if (ACPI_SUCCESS(status)) {
 		part->acpiId = id;
 		part->mode   = value;
-- 
2.43.0

[PATCH 07/11] nouveau/gsp: convert gsp errors to generic errors

[Dave Airlie]

This should let the upper layers retry as needed on EAGAIN.

There may be other values we will care about in the future, but
this covers our present needs.

Signed-off-by: Dave Airlie <airlied@redhat.com>
---
 .../gpu/drm/nouveau/nvkm/subdev/gsp/r535.c    | 26 +++++++++++++++----
 1 file changed, 21 insertions(+), 5 deletions(-)

diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c b/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c
index 774ca47b019f..54c1fbccc013 100644
--- a/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c
+++ b/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c
@@ -70,6 +70,20 @@ struct r535_gsp_msg {
 
 #define GSP_MSG_HDR_SIZE offsetof(struct r535_gsp_msg, data)
 
+static int
+r535_rpc_status_to_errno(uint32_t rpc_status)
+{
+       switch (rpc_status) {
+       case 0x55: /* NV_ERR_NOT_READY */
+       case 0x66: /* NV_ERR_TIMEOUT_RETRY */
+              return -EAGAIN;
+       case 0x51: /* NV_ERR_NO_MEMORY */
+               return -ENOMEM;
+       default:
+               return -EINVAL;
+       }
+}
+
 static void *
 r535_gsp_msgq_wait(struct nvkm_gsp *gsp, u32 repc, u32 *prepc, int *ptime)
 {
@@ -584,8 +598,9 @@ r535_gsp_rpc_rm_alloc_push(struct nvkm_gsp_object *object, void *argv, u32 repc)
 		return rpc;
 
 	if (rpc->status) {
-		nvkm_error(&gsp->subdev, "RM_ALLOC: 0x%x\n", rpc->status);
-		ret = ERR_PTR(-EINVAL);
+		ret = ERR_PTR(r535_rpc_status_to_errno(rpc->status));
+		if (ret != -EAGAIN)
+			nvkm_error(&gsp->subdev, "RM_ALLOC: 0x%x\n", rpc->status);
 	} else {
 		ret = repc ? rpc->params : NULL;
 	}
@@ -639,9 +654,10 @@ r535_gsp_rpc_rm_ctrl_push(struct nvkm_gsp_object *object, void *argv, u32 repc)
 		return rpc;
 
 	if (rpc->status) {
-		nvkm_error(&gsp->subdev, "cli:0x%08x obj:0x%08x ctrl cmd:0x%08x failed: 0x%08x\n",
-			   object->client->object.handle, object->handle, rpc->cmd, rpc->status);
-		ret = ERR_PTR(-EINVAL);
+		ret = ERR_PTR(r535_rpc_status_to_errno(rpc->status));
+		if (ret != -EAGAIN)
+			nvkm_error(&gsp->subdev, "cli:0x%08x obj:0x%08x ctrl cmd:0x%08x failed: 0x%08x\n",
+				   object->client->object.handle, object->handle, rpc->cmd, rpc->status);
 	} else {
 		ret = repc ? rpc->params : NULL;
 	}
-- 
2.43.0

[PATCH 08/11] nouveau/gsp: don't free ctrl messages on errors

[Dave Airlie]

It looks like for some messages the upper layers need to get access to the
results of the message so we can interpret it.

Rework the ctrl push interface to not free things and cleanup properly
whereever it errors out.

Requested-by: Lyude
Signed-off-by: Dave Airlie <airlied@redhat.com>
---
 .../gpu/drm/nouveau/include/nvkm/subdev/gsp.h |  17 +--
 .../gpu/drm/nouveau/nvkm/engine/disp/r535.c   | 108 +++++++++++-------
 .../gpu/drm/nouveau/nvkm/subdev/gsp/r535.c    |  36 +++---
 3 files changed, 100 insertions(+), 61 deletions(-)

diff --git a/drivers/gpu/drm/nouveau/include/nvkm/subdev/gsp.h b/drivers/gpu/drm/nouveau/include/nvkm/subdev/gsp.h
index 2fa0445d8928..d1437c08645f 100644
--- a/drivers/gpu/drm/nouveau/include/nvkm/subdev/gsp.h
+++ b/drivers/gpu/drm/nouveau/include/nvkm/subdev/gsp.h
@@ -187,7 +187,7 @@ struct nvkm_gsp {
 		void (*rpc_done)(struct nvkm_gsp *gsp, void *repv);
 
 		void *(*rm_ctrl_get)(struct nvkm_gsp_object *, u32 cmd, u32 argc);
-		void *(*rm_ctrl_push)(struct nvkm_gsp_object *, void *argv, u32 repc);
+		int (*rm_ctrl_push)(struct nvkm_gsp_object *, void **argv, u32 repc);
 		void (*rm_ctrl_done)(struct nvkm_gsp_object *, void *repv);
 
 		void *(*rm_alloc_get)(struct nvkm_gsp_object *, u32 oclass, u32 argc);
@@ -265,7 +265,7 @@ nvkm_gsp_rm_ctrl_get(struct nvkm_gsp_object *object, u32 cmd, u32 argc)
 	return object->client->gsp->rm->rm_ctrl_get(object, cmd, argc);
 }
 
-static inline void *
+static inline int
 nvkm_gsp_rm_ctrl_push(struct nvkm_gsp_object *object, void *argv, u32 repc)
 {
 	return object->client->gsp->rm->rm_ctrl_push(object, argv, repc);
@@ -275,21 +275,24 @@ static inline void *
 nvkm_gsp_rm_ctrl_rd(struct nvkm_gsp_object *object, u32 cmd, u32 repc)
 {
 	void *argv = nvkm_gsp_rm_ctrl_get(object, cmd, repc);
+	int ret;
 
 	if (IS_ERR(argv))
 		return argv;
 
-	return nvkm_gsp_rm_ctrl_push(object, argv, repc);
+	ret = nvkm_gsp_rm_ctrl_push(object, &argv, repc);
+	if (ret)
+		return ERR_PTR(ret);
+	return argv;
 }
 
 static inline int
 nvkm_gsp_rm_ctrl_wr(struct nvkm_gsp_object *object, void *argv)
 {
-	void *repv = nvkm_gsp_rm_ctrl_push(object, argv, 0);
-
-	if (IS_ERR(repv))
-		return PTR_ERR(repv);
+	int ret = nvkm_gsp_rm_ctrl_push(object, &argv, 0);
 
+	if (ret)
+		return ret;
 	return 0;
 }
 
diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/r535.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/r535.c
index 1c8c4cca0957..1b4f988df7ed 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/r535.c
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/r535.c
@@ -282,7 +282,7 @@ r535_sor_bl_get(struct nvkm_ior *sor)
 {
 	struct nvkm_disp *disp = sor->disp;
 	NV0073_CTRL_SPECIFIC_BACKLIGHT_BRIGHTNESS_PARAMS *ctrl;
-	int lvl;
+	int ret, lvl;
 
 	ctrl = nvkm_gsp_rm_ctrl_get(&disp->rm.objcom,
 				    NV0073_CTRL_CMD_SPECIFIC_GET_BACKLIGHT_BRIGHTNESS,
@@ -292,9 +292,11 @@ r535_sor_bl_get(struct nvkm_ior *sor)
 
 	ctrl->displayId = BIT(sor->asy.outp->index);
 
-	ctrl = nvkm_gsp_rm_ctrl_push(&disp->rm.objcom, ctrl, sizeof(*ctrl));
-	if (IS_ERR(ctrl))
-		return PTR_ERR(ctrl);
+	ret = nvkm_gsp_rm_ctrl_push(&disp->rm.objcom, &ctrl, sizeof(*ctrl));
+	if (ret) {
+		nvkm_gsp_rm_ctrl_done(&disp->rm.objcom, ctrl);
+		return ret;
+	}
 
 	lvl = ctrl->brightness;
 	nvkm_gsp_rm_ctrl_done(&disp->rm.objcom, ctrl);
@@ -649,9 +651,11 @@ r535_conn_new(struct nvkm_disp *disp, u32 id)
 	ctrl->subDeviceInstance = 0;
 	ctrl->displayId = BIT(id);
 
-	ctrl = nvkm_gsp_rm_ctrl_push(&disp->rm.objcom, ctrl, sizeof(*ctrl));
-	if (IS_ERR(ctrl))
-		return (void *)ctrl;
+	ret = nvkm_gsp_rm_ctrl_push(&disp->rm.objcom, &ctrl, sizeof(*ctrl));
+	if (ret) {
+		nvkm_gsp_rm_ctrl_done(&disp->rm.objcom, ctrl);
+		return ERR_PTR(ret);
+	}
 
 	list_for_each_entry(conn, &disp->conns, head) {
 		if (conn->index == ctrl->data[0].index) {
@@ -686,7 +690,7 @@ r535_outp_acquire(struct nvkm_outp *outp, bool hda)
 	struct nvkm_disp *disp = outp->disp;
 	struct nvkm_ior *ior;
 	NV0073_CTRL_DFP_ASSIGN_SOR_PARAMS *ctrl;
-	int or;
+	int ret, or;
 
 	ctrl = nvkm_gsp_rm_ctrl_get(&disp->rm.objcom,
 				    NV0073_CTRL_CMD_DFP_ASSIGN_SOR, sizeof(*ctrl));
@@ -699,9 +703,11 @@ r535_outp_acquire(struct nvkm_outp *outp, bool hda)
 	if (hda)
 		ctrl->flags |= NVDEF(NV0073_CTRL, DFP_ASSIGN_SOR_FLAGS, AUDIO, OPTIMAL);
 
-	ctrl = nvkm_gsp_rm_ctrl_push(&disp->rm.objcom, ctrl, sizeof(*ctrl));
-	if (IS_ERR(ctrl))
-		return PTR_ERR(ctrl);
+	ret = nvkm_gsp_rm_ctrl_push(&disp->rm.objcom, &ctrl, sizeof(*ctrl));
+	if (ret) {
+		nvkm_gsp_rm_ctrl_done(&disp->rm.objcom, ctrl);
+		return ret;
+	}
 
 	for (or = 0; or < ARRAY_SIZE(ctrl->sorAssignListWithTag); or++) {
 		if (ctrl->sorAssignListWithTag[or].displayMask & BIT(outp->index)) {
@@ -727,6 +733,7 @@ static int
 r535_disp_head_displayid(struct nvkm_disp *disp, int head, u32 *displayid)
 {
 	NV0073_CTRL_SYSTEM_GET_ACTIVE_PARAMS *ctrl;
+	int ret;
 
 	ctrl = nvkm_gsp_rm_ctrl_get(&disp->rm.objcom,
 				    NV0073_CTRL_CMD_SYSTEM_GET_ACTIVE, sizeof(*ctrl));
@@ -736,9 +743,11 @@ r535_disp_head_displayid(struct nvkm_disp *disp, int head, u32 *displayid)
 	ctrl->subDeviceInstance = 0;
 	ctrl->head = head;
 
-	ctrl = nvkm_gsp_rm_ctrl_push(&disp->rm.objcom, ctrl, sizeof(*ctrl));
-	if (IS_ERR(ctrl))
-		return PTR_ERR(ctrl);
+	ret = nvkm_gsp_rm_ctrl_push(&disp->rm.objcom, &ctrl, sizeof(*ctrl));
+	if (ret) {
+		nvkm_gsp_rm_ctrl_done(&disp->rm.objcom, ctrl);
+		return ret;
+	}
 
 	*displayid = ctrl->displayId;
 	nvkm_gsp_rm_ctrl_done(&disp->rm.objcom, ctrl);
@@ -772,9 +781,11 @@ r535_outp_inherit(struct nvkm_outp *outp)
 			ctrl->subDeviceInstance = 0;
 			ctrl->displayId = displayid;
 
-			ctrl = nvkm_gsp_rm_ctrl_push(&disp->rm.objcom, ctrl, sizeof(*ctrl));
-			if (IS_ERR(ctrl))
+			ret = nvkm_gsp_rm_ctrl_push(&disp->rm.objcom, &ctrl, sizeof(*ctrl));
+			if (ret) {
+				nvkm_gsp_rm_ctrl_done(&disp->rm.objcom, ctrl);
 				return NULL;
+			}
 
 			id = ctrl->index;
 			proto = ctrl->protocol;
@@ -825,6 +836,7 @@ r535_outp_dfp_get_info(struct nvkm_outp *outp)
 {
 	NV0073_CTRL_DFP_GET_INFO_PARAMS *ctrl;
 	struct nvkm_disp *disp = outp->disp;
+	int ret;
 
 	ctrl = nvkm_gsp_rm_ctrl_get(&disp->rm.objcom, NV0073_CTRL_CMD_DFP_GET_INFO, sizeof(*ctrl));
 	if (IS_ERR(ctrl))
@@ -832,9 +844,11 @@ r535_outp_dfp_get_info(struct nvkm_outp *outp)
 
 	ctrl->displayId = BIT(outp->index);
 
-	ctrl = nvkm_gsp_rm_ctrl_push(&disp->rm.objcom, ctrl, sizeof(*ctrl));
-	if (IS_ERR(ctrl))
-		return PTR_ERR(ctrl);
+	ret = nvkm_gsp_rm_ctrl_push(&disp->rm.objcom, &ctrl, sizeof(*ctrl));
+	if (ret) {
+		nvkm_gsp_rm_ctrl_done(&disp->rm.objcom, ctrl);
+		return ret;
+	}
 
 	nvkm_debug(&disp->engine.subdev, "DFP %08x: flags:%08x flags2:%08x\n",
 		   ctrl->displayId, ctrl->flags, ctrl->flags2);
@@ -858,9 +872,11 @@ r535_outp_detect(struct nvkm_outp *outp)
 	ctrl->subDeviceInstance = 0;
 	ctrl->displayMask = BIT(outp->index);
 
-	ctrl = nvkm_gsp_rm_ctrl_push(&disp->rm.objcom, ctrl, sizeof(*ctrl));
-	if (IS_ERR(ctrl))
-		return PTR_ERR(ctrl);
+	ret = nvkm_gsp_rm_ctrl_push(&disp->rm.objcom, &ctrl, sizeof(*ctrl));
+	if (ret) {
+		nvkm_gsp_rm_ctrl_done(&disp->rm.objcom, ctrl);
+		return ret;
+	}
 
 	if (ctrl->displayMask & BIT(outp->index)) {
 		ret = r535_outp_dfp_get_info(outp);
@@ -895,6 +911,7 @@ r535_dp_mst_id_get(struct nvkm_outp *outp, u32 *pid)
 {
 	NV0073_CTRL_CMD_DP_TOPOLOGY_ALLOCATE_DISPLAYID_PARAMS *ctrl;
 	struct nvkm_disp *disp = outp->disp;
+	int ret;
 
 	ctrl = nvkm_gsp_rm_ctrl_get(&disp->rm.objcom,
 				    NV0073_CTRL_CMD_DP_TOPOLOGY_ALLOCATE_DISPLAYID,
@@ -904,9 +921,11 @@ r535_dp_mst_id_get(struct nvkm_outp *outp, u32 *pid)
 
 	ctrl->subDeviceInstance = 0;
 	ctrl->displayId = BIT(outp->index);
-	ctrl = nvkm_gsp_rm_ctrl_push(&disp->rm.objcom, ctrl, sizeof(*ctrl));
-	if (IS_ERR(ctrl))
-		return PTR_ERR(ctrl);
+	ret = nvkm_gsp_rm_ctrl_push(&disp->rm.objcom, &ctrl, sizeof(*ctrl));
+	if (ret) {
+		nvkm_gsp_rm_ctrl_done(&disp->rm.objcom, ctrl);
+		return ret;
+	}
 
 	*pid = ctrl->displayIdAssigned;
 	nvkm_gsp_rm_ctrl_done(&disp->rm.objcom, ctrl);
@@ -964,9 +983,11 @@ r535_dp_train_target(struct nvkm_outp *outp, u8 target, bool mst, u8 link_nr, u8
 	    !(outp->dp.dpcd[DPCD_RC03] & DPCD_RC03_TPS4_SUPPORTED))
 	    ctrl->cmd |= NVDEF(NV0073_CTRL, DP_CMD, POST_LT_ADJ_REQ_GRANTED, YES);
 
-	ctrl = nvkm_gsp_rm_ctrl_push(&disp->rm.objcom, ctrl, sizeof(*ctrl));
-	if (IS_ERR(ctrl))
-		return PTR_ERR(ctrl);
+	ret = nvkm_gsp_rm_ctrl_push(&disp->rm.objcom, &ctrl, sizeof(*ctrl));
+	if (ret) {
+		nvkm_gsp_rm_ctrl_done(&disp->rm.objcom, ctrl);
+		return ret;
+	}
 
 	ret = ctrl->err ? -EIO : 0;
 	nvkm_gsp_rm_ctrl_done(&disp->rm.objcom, ctrl);
@@ -1036,9 +1057,11 @@ r535_dp_aux_xfer(struct nvkm_outp *outp, u8 type, u32 addr, u8 *data, u8 *psize)
 	ctrl->size = !ctrl->bAddrOnly ? (size - 1) : 0;
 	memcpy(ctrl->data, data, size);
 
-	ctrl = nvkm_gsp_rm_ctrl_push(&disp->rm.objcom, ctrl, sizeof(*ctrl));
-	if (IS_ERR(ctrl))
+	ret = nvkm_gsp_rm_ctrl_push(&disp->rm.objcom, &ctrl, sizeof(*ctrl));
+	if (ret) {
+		nvkm_gsp_rm_ctrl_done(&disp->rm.objcom, ctrl);
 		return PTR_ERR(ctrl);
+	}
 
 	memcpy(data, ctrl->data, size);
 	*psize = ctrl->size;
@@ -1111,10 +1134,13 @@ r535_tmds_edid_get(struct nvkm_outp *outp, u8 *data, u16 *psize)
 	ctrl->subDeviceInstance = 0;
 	ctrl->displayId = BIT(outp->index);
 
-	ctrl = nvkm_gsp_rm_ctrl_push(&disp->rm.objcom, ctrl, sizeof(*ctrl));
-	if (IS_ERR(ctrl))
-		return PTR_ERR(ctrl);
+	ret = nvkm_gsp_rm_ctrl_push(&disp->rm.objcom, &ctrl, sizeof(*ctrl));
+	if (ret) {
+		nvkm_gsp_rm_ctrl_done(&disp->rm.objcom, ctrl);
+		return ret;
+	}
 
+	ret = -E2BIG;
 	if (ctrl->bufferSize <= *psize) {
 		memcpy(data, ctrl->edidBuffer, ctrl->bufferSize);
 		*psize = ctrl->bufferSize;
@@ -1153,9 +1179,11 @@ r535_outp_new(struct nvkm_disp *disp, u32 id)
 	ctrl->subDeviceInstance = 0;
 	ctrl->displayId = BIT(id);
 
-	ctrl = nvkm_gsp_rm_ctrl_push(&disp->rm.objcom, ctrl, sizeof(*ctrl));
-	if (IS_ERR(ctrl))
-		return PTR_ERR(ctrl);
+	ret = nvkm_gsp_rm_ctrl_push(&disp->rm.objcom, &ctrl, sizeof(*ctrl));
+	if (ret) {
+		nvkm_gsp_rm_ctrl_done(&disp->rm.objcom, ctrl);
+		return ret;
+	}
 
 	switch (ctrl->type) {
 	case NV0073_CTRL_SPECIFIC_OR_TYPE_NONE:
@@ -1229,9 +1257,11 @@ r535_outp_new(struct nvkm_disp *disp, u32 id)
 
 		ctrl->sorIndex = ~0;
 
-		ctrl = nvkm_gsp_rm_ctrl_push(&disp->rm.objcom, ctrl, sizeof(*ctrl));
-		if (IS_ERR(ctrl))
-			return PTR_ERR(ctrl);
+		ret = nvkm_gsp_rm_ctrl_push(&disp->rm.objcom, &ctrl, sizeof(*ctrl));
+		if (ret) {
+			nvkm_gsp_rm_ctrl_done(&disp->rm.objcom, ctrl);
+			return ret;
+		}
 
 		switch (NVVAL_GET(ctrl->maxLinkRate, NV0073_CTRL_CMD, DP_GET_CAPS, MAX_LINK_RATE)) {
 		case NV0073_CTRL_CMD_DP_GET_CAPS_MAX_LINK_RATE_1_62:
diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c b/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c
index 54c1fbccc013..e2810fd1a36f 100644
--- a/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c
+++ b/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c
@@ -599,13 +599,13 @@ r535_gsp_rpc_rm_alloc_push(struct nvkm_gsp_object *object, void *argv, u32 repc)
 
 	if (rpc->status) {
 		ret = ERR_PTR(r535_rpc_status_to_errno(rpc->status));
-		if (ret != -EAGAIN)
+		if (PTR_ERR(ret) != -EAGAIN)
 			nvkm_error(&gsp->subdev, "RM_ALLOC: 0x%x\n", rpc->status);
 	} else {
 		ret = repc ? rpc->params : NULL;
 	}
 
-	if (IS_ERR_OR_NULL(ret))
+	if (ret)
 		nvkm_gsp_rpc_done(gsp, rpc);
 
 	return ret;
@@ -639,30 +639,34 @@ r535_gsp_rpc_rm_ctrl_done(struct nvkm_gsp_object *object, void *repv)
 {
 	rpc_gsp_rm_control_v03_00 *rpc = container_of(repv, typeof(*rpc), params);
 
+	if (!repv)
+		return;
 	nvkm_gsp_rpc_done(object->client->gsp, rpc);
 }
 
-static void *
-r535_gsp_rpc_rm_ctrl_push(struct nvkm_gsp_object *object, void *argv, u32 repc)
+static int
+r535_gsp_rpc_rm_ctrl_push(struct nvkm_gsp_object *object, void **argv, u32 repc)
 {
-	rpc_gsp_rm_control_v03_00 *rpc = container_of(argv, typeof(*rpc), params);
+	rpc_gsp_rm_control_v03_00 *rpc = container_of((*argv), typeof(*rpc), params);
 	struct nvkm_gsp *gsp = object->client->gsp;
-	void *ret;
+	int ret = 0;
 
 	rpc = nvkm_gsp_rpc_push(gsp, rpc, true, repc);
-	if (IS_ERR_OR_NULL(rpc))
-		return rpc;
+	if (IS_ERR_OR_NULL(rpc)) {
+		*argv = NULL;
+		return PTR_ERR(rpc);
+	}
 
 	if (rpc->status) {
-		ret = ERR_PTR(r535_rpc_status_to_errno(rpc->status));
+		ret = r535_rpc_status_to_errno(rpc->status);
 		if (ret != -EAGAIN)
 			nvkm_error(&gsp->subdev, "cli:0x%08x obj:0x%08x ctrl cmd:0x%08x failed: 0x%08x\n",
 				   object->client->object.handle, object->handle, rpc->cmd, rpc->status);
-	} else {
-		ret = repc ? rpc->params : NULL;
 	}
 
-	if (IS_ERR_OR_NULL(ret))
+	if (repc)
+		*argv = rpc->params;
+	else
 		nvkm_gsp_rpc_done(gsp, rpc);
 
 	return ret;
@@ -860,9 +864,11 @@ r535_gsp_intr_get_table(struct nvkm_gsp *gsp)
 	if (IS_ERR(ctrl))
 		return PTR_ERR(ctrl);
 
-	ctrl = nvkm_gsp_rm_ctrl_push(&gsp->internal.device.subdevice, ctrl, sizeof(*ctrl));
-	if (WARN_ON(IS_ERR(ctrl)))
-		return PTR_ERR(ctrl);
+	ret = nvkm_gsp_rm_ctrl_push(&gsp->internal.device.subdevice, &ctrl, sizeof(*ctrl));
+	if (WARN_ON(ret)) {
+		nvkm_gsp_rm_ctrl_done(&gsp->internal.device.subdevice, ctrl);
+		return ret;
+	}
 
 	for (unsigned i = 0; i < ctrl->tableLen; i++) {
 		enum nvkm_subdev_type type;
-- 
2.43.0

[PATCH 09/11] nouveau/gsp: always free the alloc messages on r535

[Dave Airlie]

Fixes a memory leak seen with kmemleak.

Signed-off-by: Dave Airlie <airlied@redhat.com>
---
 drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c b/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c
index e2810fd1a36f..cafb82826473 100644
--- a/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c
+++ b/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c
@@ -605,8 +605,7 @@ r535_gsp_rpc_rm_alloc_push(struct nvkm_gsp_object *object, void *argv, u32 repc)
 		ret = repc ? rpc->params : NULL;
 	}
 
-	if (ret)
-		nvkm_gsp_rpc_done(gsp, rpc);
+	nvkm_gsp_rpc_done(gsp, rpc);
 
 	return ret;
 }
-- 
2.43.0

[PATCH 10/11] nouveau: push event block/allowing out of the fence context

[Dave Airlie]

There is a deadlock between the irq and fctx locks,
the irq handling takes irq then fctx lock
the fence signalling takes fctx then irq lock

This splits the fence signalling path so the code that hits
the irq lock is done in a separate work queue.

This seems to fix crashes/hangs when using nouveau gsp with
i915 primary GPU.

Signed-off-by: Dave Airlie <airlied@redhat.com>
---
 drivers/gpu/drm/nouveau/nouveau_fence.c | 28 ++++++++++++++++++++-----
 drivers/gpu/drm/nouveau/nouveau_fence.h |  5 ++++-
 2 files changed, 27 insertions(+), 6 deletions(-)

diff --git a/drivers/gpu/drm/nouveau/nouveau_fence.c b/drivers/gpu/drm/nouveau/nouveau_fence.c
index ca762ea55413..5057d976fa57 100644
--- a/drivers/gpu/drm/nouveau/nouveau_fence.c
+++ b/drivers/gpu/drm/nouveau/nouveau_fence.c
@@ -62,7 +62,7 @@ nouveau_fence_signal(struct nouveau_fence *fence)
 	if (test_bit(DMA_FENCE_FLAG_USER_BITS, &fence->base.flags)) {
 		struct nouveau_fence_chan *fctx = nouveau_fctx(fence);
 
-		if (!--fctx->notify_ref)
+		if (atomic_dec_and_test(&fctx->notify_ref))
 			drop = 1;
 	}
 
@@ -103,6 +103,7 @@ nouveau_fence_context_kill(struct nouveau_fence_chan *fctx, int error)
 void
 nouveau_fence_context_del(struct nouveau_fence_chan *fctx)
 {
+	cancel_work_sync(&fctx->allow_block_work);
 	nouveau_fence_context_kill(fctx, 0);
 	nvif_event_dtor(&fctx->event);
 	fctx->dead = 1;
@@ -167,6 +168,18 @@ nouveau_fence_wait_uevent_handler(struct nvif_event *event, void *repv, u32 repc
 	return ret;
 }
 
+static void
+nouveau_fence_work_allow_block(struct work_struct *work)
+{
+	struct nouveau_fence_chan *fctx = container_of(work, struct nouveau_fence_chan,
+						       allow_block_work);
+
+	if (atomic_read(&fctx->notify_ref) == 0)
+		nvif_event_block(&fctx->event);
+	else
+		nvif_event_allow(&fctx->event);
+}
+
 void
 nouveau_fence_context_new(struct nouveau_channel *chan, struct nouveau_fence_chan *fctx)
 {
@@ -178,6 +191,7 @@ nouveau_fence_context_new(struct nouveau_channel *chan, struct nouveau_fence_cha
 	} args;
 	int ret;
 
+	INIT_WORK(&fctx->allow_block_work, nouveau_fence_work_allow_block);
 	INIT_LIST_HEAD(&fctx->flip);
 	INIT_LIST_HEAD(&fctx->pending);
 	spin_lock_init(&fctx->lock);
@@ -521,15 +535,19 @@ static bool nouveau_fence_enable_signaling(struct dma_fence *f)
 	struct nouveau_fence *fence = from_fence(f);
 	struct nouveau_fence_chan *fctx = nouveau_fctx(fence);
 	bool ret;
+	bool do_work;
 
-	if (!fctx->notify_ref++)
-		nvif_event_allow(&fctx->event);
+	if (atomic_inc_return(&fctx->notify_ref) == 0)
+		do_work = true;
 
 	ret = nouveau_fence_no_signaling(f);
 	if (ret)
 		set_bit(DMA_FENCE_FLAG_USER_BITS, &fence->base.flags);
-	else if (!--fctx->notify_ref)
-		nvif_event_block(&fctx->event);
+	else if (atomic_dec_and_test(&fctx->notify_ref))
+		do_work = true;
+
+	if (do_work)
+		schedule_work(&fctx->allow_block_work);
 
 	return ret;
 }
diff --git a/drivers/gpu/drm/nouveau/nouveau_fence.h b/drivers/gpu/drm/nouveau/nouveau_fence.h
index 64d33ae7f356..28f5cf013b89 100644
--- a/drivers/gpu/drm/nouveau/nouveau_fence.h
+++ b/drivers/gpu/drm/nouveau/nouveau_fence.h
@@ -3,6 +3,7 @@
 #define __NOUVEAU_FENCE_H__
 
 #include <linux/dma-fence.h>
+#include <linux/workqueue.h>
 #include <nvif/event.h>
 
 struct nouveau_drm;
@@ -45,7 +46,9 @@ struct nouveau_fence_chan {
 	char name[32];
 
 	struct nvif_event event;
-	int notify_ref, dead, killed;
+	struct work_struct allow_block_work;
+	atomic_t notify_ref;
+	int dead, killed;
 };
 
 struct nouveau_fence_priv {
-- 
2.43.0

[PATCH 11/11] drm/nouveau/dp: Honor GSP link training retry timeouts

[Dave Airlie]

From: Lyude Paul <lyude@redhat.com>

Turns out that one of the ways that Nvidia's driver handles the pre-LT
timeout for eDP panels is by providing a retry timeout in their link
training callbacks that we're expected to wait for. Up until now we didn't
pay any attention to this parameter.

So, start honoring the timeout if link training fails - and retry up to 3
times. The "3 times" bit comes from OpenRM's link training code.

[airlied: this fixes the panel on one of my laptops]
Signed-off-by: Lyude Paul <lyude@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
---
 .../gpu/drm/nouveau/nvkm/engine/disp/r535.c   | 62 ++++++++++++-------
 1 file changed, 40 insertions(+), 22 deletions(-)

diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/r535.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/r535.c
index 1b4f988df7ed..b287ab19a51f 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/r535.c
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/r535.c
@@ -957,40 +957,58 @@ r535_dp_train_target(struct nvkm_outp *outp, u8 target, bool mst, u8 link_nr, u8
 {
 	struct nvkm_disp *disp = outp->disp;
 	NV0073_CTRL_DP_CTRL_PARAMS *ctrl;
-	int ret;
-
-	ctrl = nvkm_gsp_rm_ctrl_get(&disp->rm.objcom, NV0073_CTRL_CMD_DP_CTRL, sizeof(*ctrl));
-	if (IS_ERR(ctrl))
-		return PTR_ERR(ctrl);
+	int ret, retries;
+	u32 cmd, data;
 
-	ctrl->subDeviceInstance = 0;
-	ctrl->displayId = BIT(outp->index);
-	ctrl->cmd = NVDEF(NV0073_CTRL, DP_CMD, SET_LANE_COUNT, TRUE) |
-		    NVDEF(NV0073_CTRL, DP_CMD, SET_LINK_BW, TRUE) |
-		    NVDEF(NV0073_CTRL, DP_CMD, TRAIN_PHY_REPEATER, YES);
-	ctrl->data = NVVAL(NV0073_CTRL, DP_DATA, SET_LANE_COUNT, link_nr) |
-		     NVVAL(NV0073_CTRL, DP_DATA, SET_LINK_BW, link_bw) |
-		     NVVAL(NV0073_CTRL, DP_DATA, TARGET, target);
+	cmd = NVDEF(NV0073_CTRL, DP_CMD, SET_LANE_COUNT, TRUE) |
+	      NVDEF(NV0073_CTRL, DP_CMD, SET_LINK_BW, TRUE) |
+	      NVDEF(NV0073_CTRL, DP_CMD, TRAIN_PHY_REPEATER, YES);
+	data = NVVAL(NV0073_CTRL, DP_DATA, SET_LANE_COUNT, link_nr) |
+	       NVVAL(NV0073_CTRL, DP_DATA, SET_LINK_BW, link_bw) |
+	       NVVAL(NV0073_CTRL, DP_DATA, TARGET, target);
 
 	if (mst)
-		ctrl->cmd |= NVDEF(NV0073_CTRL, DP_CMD, SET_FORMAT_MODE, MULTI_STREAM);
+		cmd |= NVDEF(NV0073_CTRL, DP_CMD, SET_FORMAT_MODE, MULTI_STREAM);
 
 	if (outp->dp.dpcd[DPCD_RC02] & DPCD_RC02_ENHANCED_FRAME_CAP)
-		ctrl->cmd |= NVDEF(NV0073_CTRL, DP_CMD, SET_ENHANCED_FRAMING, TRUE);
+		cmd |= NVDEF(NV0073_CTRL, DP_CMD, SET_ENHANCED_FRAMING, TRUE);
 
 	if (target == 0 &&
 	     (outp->dp.dpcd[DPCD_RC02] & 0x20) &&
 	    !(outp->dp.dpcd[DPCD_RC03] & DPCD_RC03_TPS4_SUPPORTED))
-	    ctrl->cmd |= NVDEF(NV0073_CTRL, DP_CMD, POST_LT_ADJ_REQ_GRANTED, YES);
+	    cmd |= NVDEF(NV0073_CTRL, DP_CMD, POST_LT_ADJ_REQ_GRANTED, YES);
 
-	ret = nvkm_gsp_rm_ctrl_push(&disp->rm.objcom, &ctrl, sizeof(*ctrl));
-	if (ret) {
-		nvkm_gsp_rm_ctrl_done(&disp->rm.objcom, ctrl);
-		return ret;
+	/* We should retry up to 3 times, but only if GSP asks politely */
+	for (retries = 0; retries < 3; ++retries) {
+		ctrl = nvkm_gsp_rm_ctrl_get(&disp->rm.objcom, NV0073_CTRL_CMD_DP_CTRL,
+					    sizeof(*ctrl));
+		if (IS_ERR(ctrl))
+			return PTR_ERR(ctrl);
+
+		ctrl->subDeviceInstance = 0;
+		ctrl->displayId = BIT(outp->index);
+		ctrl->retryTimeMs = 0;
+		ctrl->cmd = cmd;
+		ctrl->data = data;
+
+		ret = nvkm_gsp_rm_ctrl_push(&disp->rm.objcom, &ctrl, sizeof(*ctrl));
+		if (ret == -EAGAIN && ctrl->retryTimeMs) {
+			/* Device (likely an eDP panel) isn't ready yet, wait for the time specified
+			 * by GSP before retrying again */
+			nvkm_debug(&disp->engine.subdev,
+				   "Waiting %dms for GSP LT panel delay before retrying\n",
+				   ctrl->retryTimeMs);
+			msleep(ctrl->retryTimeMs);
+			nvkm_gsp_rm_ctrl_done(&disp->rm.objcom, ctrl);
+		} else {
+			/* GSP didn't say to retry, or we were successful */
+			if (ctrl->err)
+				ret = -EIO;
+			nvkm_gsp_rm_ctrl_done(&disp->rm.objcom, ctrl);
+			break;
+		}
 	}
 
-	ret = ctrl->err ? -EIO : 0;
-	nvkm_gsp_rm_ctrl_done(&disp->rm.objcom, ctrl);
 	return ret;
 }
 
-- 
2.43.0

Re: [PATCH 07/11] nouveau/gsp: convert gsp errors to generic errors

[Timur Tabi]

[-- Attachment #1: Type: text/plain, Size: 712 bytes --]

On Thu, Dec 21, 2023, 10:33 PM Dave Airlie <airlied@gmail.com> wrote:

> This should let the upper layers retry as needed on EAGAIN.
>
> There may be other values we will care about in the future, but
> this covers our present needs.
>
> Signed-off-by: Dave Airlie <airlied@redhat.com>
>
> +static int
> +r535_rpc_status_to_errno(uint32_t rpc_status)
> +{
> +       switch (rpc_status) {
> +       case 0x55: /* NV_ERR_NOT_READY */
> +       case 0x66: /* NV_ERR_TIMEOUT_RETRY */
> +              return -EAGAIN;
> +       case 0x51: /* NV_ERR_NO_MEMORY */
> +               return -ENOMEM;
> +       default:
> +               return -EINVAL;
> +       }
>

Shouldn't you also have this:

case 0:
    return 0;

[-- Attachment #2: Type: text/html, Size: 1287 bytes --]

Re: [PATCH 08/11] nouveau/gsp: don't free ctrl messages on errors

[Dan Carpenter]

Hi Dave,

kernel test robot noticed the following build warnings:

https://git-scm.com/docs/git-format-patch#_base_tree_information]

url:    https://github.com/intel-lab-lkp/linux/commits/Dave-Airlie/nouveau-gsp-drop-some-acpi-related-debug/20231222-180432
base:   git://anongit.freedesktop.org/drm/drm-misc drm-misc-next
patch link:    https://lore.kernel.org/r/20231222043308.3090089-9-airlied%40gmail.com
patch subject: [PATCH 08/11] nouveau/gsp: don't free ctrl messages on errors
config: powerpc-randconfig-r071-20231226 (https://download.01.org/0day-ci/archive/20231227/202312271917.55xuDMdc-lkp@intel.com/config)
compiler: clang version 18.0.0git (https://github.com/llvm/llvm-project d3ef86708241a3bee902615c190dead1638c4e09)

If you fix the issue in a separate patch/commit (i.e. not just a new version of
the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <lkp@intel.com>
| Reported-by: Dan Carpenter <dan.carpenter@linaro.org>
| Closes: https://lore.kernel.org/r/202312271917.55xuDMdc-lkp@intel.com/

New smatch warnings:
drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c:659 r535_gsp_rpc_rm_ctrl_push() warn: passing zero to 'PTR_ERR'
drivers/gpu/drm/nouveau/nvkm/engine/disp/r535.c:1063 r535_dp_aux_xfer() warn: passing a valid pointer to 'PTR_ERR'

Old smatch warnings:
drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c:1887 nvkm_gsp_radix3_sg() error: uninitialized symbol 'addr'.

vim +/PTR_ERR +659 drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c

af265ee961627a Dave Airlie 2023-12-22  649  static int
af265ee961627a Dave Airlie 2023-12-22  650  r535_gsp_rpc_rm_ctrl_push(struct nvkm_gsp_object *object, void **argv, u32 repc)
4cf2c83eb3a4c4 Ben Skeggs  2023-09-19  651  {
af265ee961627a Dave Airlie 2023-12-22  652  	rpc_gsp_rm_control_v03_00 *rpc = container_of((*argv), typeof(*rpc), params);
4cf2c83eb3a4c4 Ben Skeggs  2023-09-19  653  	struct nvkm_gsp *gsp = object->client->gsp;
af265ee961627a Dave Airlie 2023-12-22  654  	int ret = 0;
4cf2c83eb3a4c4 Ben Skeggs  2023-09-19  655  
4cf2c83eb3a4c4 Ben Skeggs  2023-09-19  656  	rpc = nvkm_gsp_rpc_push(gsp, rpc, true, repc);
af265ee961627a Dave Airlie 2023-12-22  657  	if (IS_ERR_OR_NULL(rpc)) {
af265ee961627a Dave Airlie 2023-12-22  658  		*argv = NULL;
af265ee961627a Dave Airlie 2023-12-22 @659  		return PTR_ERR(rpc);

If nvkm_gsp_rpc_push() returns NULL (probably a failure) then this
returns PTR_ERR(NULL) which is zero/success.

af265ee961627a Dave Airlie 2023-12-22  660  	}
4cf2c83eb3a4c4 Ben Skeggs  2023-09-19  661  
4cf2c83eb3a4c4 Ben Skeggs  2023-09-19  662  	if (rpc->status) {
af265ee961627a Dave Airlie 2023-12-22  663  		ret = r535_rpc_status_to_errno(rpc->status);
555bb9c29a45be Dave Airlie 2023-12-22  664  		if (ret != -EAGAIN)
4cf2c83eb3a4c4 Ben Skeggs  2023-09-19  665  			nvkm_error(&gsp->subdev, "cli:0x%08x obj:0x%08x ctrl cmd:0x%08x failed: 0x%08x\n",
4cf2c83eb3a4c4 Ben Skeggs  2023-09-19  666  				   object->client->object.handle, object->handle, rpc->cmd, rpc->status);
4cf2c83eb3a4c4 Ben Skeggs  2023-09-19  667  	}
4cf2c83eb3a4c4 Ben Skeggs  2023-09-19  668  
af265ee961627a Dave Airlie 2023-12-22  669  	if (repc)
af265ee961627a Dave Airlie 2023-12-22  670  		*argv = rpc->params;
af265ee961627a Dave Airlie 2023-12-22  671  	else
4cf2c83eb3a4c4 Ben Skeggs  2023-09-19  672  		nvkm_gsp_rpc_done(gsp, rpc);
4cf2c83eb3a4c4 Ben Skeggs  2023-09-19  673  
4cf2c83eb3a4c4 Ben Skeggs  2023-09-19  674  	return ret;
4cf2c83eb3a4c4 Ben Skeggs  2023-09-19  675  }

-- 
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki

Re: [PATCH 08/11] nouveau/gsp: don't free ctrl messages on errors

[Dave Airlie]

On Thu, 4 Jan 2024 at 00:47, Dan Carpenter <dan.carpenter@linaro.org> wrote:
>
> Hi Dave,
>
> kernel test robot noticed the following build warnings:
>
> https://git-scm.com/docs/git-format-patch#_base_tree_information]
>
> url:    https://github.com/intel-lab-lkp/linux/commits/Dave-Airlie/nouveau-gsp-drop-some-acpi-related-debug/20231222-180432
> base:   git://anongit.freedesktop.org/drm/drm-misc drm-misc-next
> patch link:    https://lore.kernel.org/r/20231222043308.3090089-9-airlied%40gmail.com
> patch subject: [PATCH 08/11] nouveau/gsp: don't free ctrl messages on errors
> config: powerpc-randconfig-r071-20231226 (https://download.01.org/0day-ci/archive/20231227/202312271917.55xuDMdc-lkp@intel.com/config)
> compiler: clang version 18.0.0git (https://github.com/llvm/llvm-project d3ef86708241a3bee902615c190dead1638c4e09)
>
> If you fix the issue in a separate patch/commit (i.e. not just a new version of
> the same patch/commit), kindly add following tags
> | Reported-by: kernel test robot <lkp@intel.com>
> | Reported-by: Dan Carpenter <dan.carpenter@linaro.org>
> | Closes: https://lore.kernel.org/r/202312271917.55xuDMdc-lkp@intel.com/

This is a false positive, I think the code is operating like I'd
expect, we maybe could restructure it to avoid this warning?

The idea is you send an rpc msg, if there's a reply you get a reply,
if no reply you get NULL and if an error you get an error.

So in the case you get an error or NULL you just want to return 0 for
the NULL as it's successful, and error otherwise.

Would using PTR_ERR_OR_ZERO make smatch happy? (even if it's not
really what we want).

Dave.
>
> New smatch warnings:
> drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c:659 r535_gsp_rpc_rm_ctrl_push() warn: passing zero to 'PTR_ERR'
> drivers/gpu/drm/nouveau/nvkm/engine/disp/r535.c:1063 r535_dp_aux_xfer() warn: passing a valid pointer to 'PTR_ERR'
>
> Old smatch warnings:
> drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c:1887 nvkm_gsp_radix3_sg() error: uninitialized symbol 'addr'.
>
> vim +/PTR_ERR +659 drivers/gpu/drm/nouveau/nvkm/subdev/gsp/r535.c
>
> af265ee961627a Dave Airlie 2023-12-22  649  static int
> af265ee961627a Dave Airlie 2023-12-22  650  r535_gsp_rpc_rm_ctrl_push(struct nvkm_gsp_object *object, void **argv, u32 repc)
> 4cf2c83eb3a4c4 Ben Skeggs  2023-09-19  651  {
> af265ee961627a Dave Airlie 2023-12-22  652      rpc_gsp_rm_control_v03_00 *rpc = container_of((*argv), typeof(*rpc), params);
> 4cf2c83eb3a4c4 Ben Skeggs  2023-09-19  653      struct nvkm_gsp *gsp = object->client->gsp;
> af265ee961627a Dave Airlie 2023-12-22  654      int ret = 0;
> 4cf2c83eb3a4c4 Ben Skeggs  2023-09-19  655
> 4cf2c83eb3a4c4 Ben Skeggs  2023-09-19  656      rpc = nvkm_gsp_rpc_push(gsp, rpc, true, repc);
> af265ee961627a Dave Airlie 2023-12-22  657      if (IS_ERR_OR_NULL(rpc)) {
> af265ee961627a Dave Airlie 2023-12-22  658              *argv = NULL;
> af265ee961627a Dave Airlie 2023-12-22 @659              return PTR_ERR(rpc);
>
> If nvkm_gsp_rpc_push() returns NULL (probably a failure) then this
> returns PTR_ERR(NULL) which is zero/success.
>
> af265ee961627a Dave Airlie 2023-12-22  660      }
> 4cf2c83eb3a4c4 Ben Skeggs  2023-09-19  661
> 4cf2c83eb3a4c4 Ben Skeggs  2023-09-19  662      if (rpc->status) {
> af265ee961627a Dave Airlie 2023-12-22  663              ret = r535_rpc_status_to_errno(rpc->status);
> 555bb9c29a45be Dave Airlie 2023-12-22  664              if (ret != -EAGAIN)
> 4cf2c83eb3a4c4 Ben Skeggs  2023-09-19  665                      nvkm_error(&gsp->subdev, "cli:0x%08x obj:0x%08x ctrl cmd:0x%08x failed: 0x%08x\n",
> 4cf2c83eb3a4c4 Ben Skeggs  2023-09-19  666                                 object->client->object.handle, object->handle, rpc->cmd, rpc->status);
> 4cf2c83eb3a4c4 Ben Skeggs  2023-09-19  667      }
> 4cf2c83eb3a4c4 Ben Skeggs  2023-09-19  668
> af265ee961627a Dave Airlie 2023-12-22  669      if (repc)
> af265ee961627a Dave Airlie 2023-12-22  670              *argv = rpc->params;
> af265ee961627a Dave Airlie 2023-12-22  671      else
> 4cf2c83eb3a4c4 Ben Skeggs  2023-09-19  672              nvkm_gsp_rpc_done(gsp, rpc);
> 4cf2c83eb3a4c4 Ben Skeggs  2023-09-19  673
> 4cf2c83eb3a4c4 Ben Skeggs  2023-09-19  674      return ret;
> 4cf2c83eb3a4c4 Ben Skeggs  2023-09-19  675  }
>
> --
> 0-DAY CI Kernel Test Service
> https://github.com/intel/lkp-tests/wiki
>

Re: [PATCH 08/11] nouveau/gsp: don't free ctrl messages on errors

[Dan Carpenter]

On Thu, Jan 04, 2024 at 10:41:50AM +1000, Dave Airlie wrote:
> On Thu, 4 Jan 2024 at 00:47, Dan Carpenter <dan.carpenter@linaro.org> wrote:
> >
> > Hi Dave,
> >
> > kernel test robot noticed the following build warnings:
> >
> > https://git-scm.com/docs/git-format-patch#_base_tree_information]
> >
> > url:    https://github.com/intel-lab-lkp/linux/commits/Dave-Airlie/nouveau-gsp-drop-some-acpi-related-debug/20231222-180432
> > base:   git://anongit.freedesktop.org/drm/drm-misc drm-misc-next
> > patch link:    https://lore.kernel.org/r/20231222043308.3090089-9-airlied%40gmail.com
> > patch subject: [PATCH 08/11] nouveau/gsp: don't free ctrl messages on errors
> > config: powerpc-randconfig-r071-20231226 (https://download.01.org/0day-ci/archive/20231227/202312271917.55xuDMdc-lkp@intel.com/config)
> > compiler: clang version 18.0.0git (https://github.com/llvm/llvm-project d3ef86708241a3bee902615c190dead1638c4e09)
> >
> > If you fix the issue in a separate patch/commit (i.e. not just a new version of
> > the same patch/commit), kindly add following tags
> > | Reported-by: kernel test robot <lkp@intel.com>
> > | Reported-by: Dan Carpenter <dan.carpenter@linaro.org>
> > | Closes: https://lore.kernel.org/r/202312271917.55xuDMdc-lkp@intel.com/
> 
> This is a false positive, I think the code is operating like I'd
> expect, we maybe could restructure it to avoid this warning?
> 
> The idea is you send an rpc msg, if there's a reply you get a reply,
> if no reply you get NULL and if an error you get an error.
> 
> So in the case you get an error or NULL you just want to return 0 for
> the NULL as it's successful, and error otherwise.
> 
> Would using PTR_ERR_OR_ZERO make smatch happy? (even if it's not
> really what we want).

Hm...  You're using the API correctly.  Linus has complained about this
warning before but in new code over 90% of the warnings are correct.
It's a high quality warning.

I looked around for an explanation to see what the NULL meant but
couldn't find it documented in the code.  The NULL vs error pointer comes
from a function pointer and it's not always clear where the
documentation should be with a function pointer.  So perhaps I missed it.

Let's not use PTR_ERR_OR_ZERO.  Perhaps I should introduce a
PTR_ERR_OR_NULL() macro to silence this warning.  But most of the code
which does this correctly is in fs/ and they probably are like Linus and
would be surprised to learn that people get it wrong...

regards,
dan carpenter