Skip to contentSkip to conversation

drm/nouveau/subdev/bus: Ratelimit logging for fault errors

PATCH In review Lyude Paul 2 participants 8 messages Apr 29, 2022 – May 4, 2022

Subsystems (3)
lore
Share & tools
lore

Lyude Paul

author maintainer
#0 lore raw

PATCH 1/1 drm/nouveau/subdev/bus: Ratelimit logging for fault errors

There's plenty of ways to fudge the GPU when developing on nouveau by mistake, some of which can result in nouveau seriously spamming dmesg with fault errors. This can be somewhat annoying, as it can quickly overrun the message buffer (or your terminal emulator's buffer) and get rid of actually useful feedback from the driver. While working on my new atomic only MST branch, I ran into this issue a couple of times.

So, let's fix this by adding nvkm_error_ratelimited(), and using it to ratelimit errors from faults. This should be fine for developers, since it's nearly always only the first few faults that we care about seeing. Plus, you can turn off rate limiting in the kernel if you really need to.

---
 drivers/gpu/drm/nouveau/include/nvkm/core/subdev.h |  2 ++
 drivers/gpu/drm/nouveau/nvkm/subdev/bus/gf100.c    | 14 +++++++-------
 drivers/gpu/drm/nouveau/nvkm/subdev/bus/nv31.c     |  6 +++---
 drivers/gpu/drm/nouveau/nvkm/subdev/bus/nv50.c     |  6 +++---
 4 files changed, 15 insertions(+), 13 deletions(-)
drivers/gpu/drm/nouveau/include/nvkm/core/subdev.h +2 -0
index 1665738948fb..96113c8bee8c 100644
@@ -62,4 +62,6 @@ void nvkm_subdev_intr(struct nvkm_subdev *);
 #define nvkm_debug(s,f,a...) nvkm_printk((s), DEBUG,   info, f, ##a)
 #define nvkm_trace(s,f,a...) nvkm_printk((s), TRACE,   info, f, ##a)
 #define nvkm_spam(s,f,a...)  nvkm_printk((s),  SPAM,    dbg, f, ##a)
+
+#define nvkm_error_ratelimited(s,f,a...) nvkm_printk((s), ERROR, err_ratelimited, f, ##a)
 #endif
drivers/gpu/drm/nouveau/nvkm/subdev/bus/gf100.c +7 -7
index 53a6651ac225..80b5aaceeaad 100644
@@ -35,13 +35,13 @@ gf100_bus_intr(struct nvkm_bus *bus)
 		u32 addr = nvkm_rd32(device, 0x009084);
 		u32 data = nvkm_rd32(device, 0x009088);
 
-		nvkm_error(subdev,
-			   "MMIO %s of %08x FAULT at %06x [ %s%s%s]\n",
-			   (addr & 0x00000002) ? "write" : "read", data,
-			   (addr & 0x00fffffc),
-			   (stat & 0x00000002) ? "!ENGINE " : "",
-			   (stat & 0x00000004) ? "PRIVRING " : "",
-			   (stat & 0x00000008) ? "TIMEOUT " : "");
+		nvkm_error_ratelimited(subdev,
+				       "MMIO %s of %08x FAULT at %06x [ %s%s%s]\n",
+				       (addr & 0x00000002) ? "write" : "read", data,
+				       (addr & 0x00fffffc),
+				       (stat & 0x00000002) ? "!ENGINE " : "",
+				       (stat & 0x00000004) ? "PRIVRING " : "",
+				       (stat & 0x00000008) ? "TIMEOUT " : "");
 
 		nvkm_wr32(device, 0x009084, 0x00000000);
 		nvkm_wr32(device, 0x001100, (stat & 0x0000000e));
drivers/gpu/drm/nouveau/nvkm/subdev/bus/nv31.c +3 -3
index ad8da523bb22..c75e463f3501 100644
@@ -45,9 +45,9 @@ nv31_bus_intr(struct nvkm_bus *bus)
 		u32 addr = nvkm_rd32(device, 0x009084);
 		u32 data = nvkm_rd32(device, 0x009088);
 
-		nvkm_error(subdev, "MMIO %s of %08x FAULT at %06x\n",
-			   (addr & 0x00000002) ? "write" : "read", data,
-			   (addr & 0x00fffffc));
+		nvkm_error_ratelimited(subdev, "MMIO %s of %08x FAULT at %06x\n",
+				       (addr & 0x00000002) ? "write" : "read", data,
+				       (addr & 0x00fffffc));
 
 		stat &= ~0x00000008;
 		nvkm_wr32(device, 0x001100, 0x00000008);
drivers/gpu/drm/nouveau/nvkm/subdev/bus/nv50.c +3 -3
index 3a1e45adeedc..2055d0b100d3 100644
@@ -60,9 +60,9 @@ nv50_bus_intr(struct nvkm_bus *bus)
 		u32 addr = nvkm_rd32(device, 0x009084);
 		u32 data = nvkm_rd32(device, 0x009088);
 
-		nvkm_error(subdev, "MMIO %s of %08x FAULT at %06x\n",
-			   (addr & 0x00000002) ? "write" : "read", data,
-			   (addr & 0x00fffffc));
+		nvkm_error_ratelimited(subdev, "MMIO %s of %08x FAULT at %06x\n",
+				       (addr & 0x00000002) ? "write" : "read", data,
+				       (addr & 0x00fffffc));
 
 		stat &= ~0x00000008;
 		nvkm_wr32(device, 0x001100, 0x00000008);
Signature · 1 line
2.35.1
#1 lore raw
Quoted 8 lines · Lyude Paul

There's plenty of ways to fudge the GPU when developing on nouveau by mistake, some of which can result in nouveau seriously spamming dmesg with fault errors. This can be somewhat annoying, as it can quickly overrun the message buffer (or your terminal emulator's buffer) and get rid of actually useful feedback from the driver. While working on my new atomic only MST branch, I ran into this issue a couple of times.

Quoted 5 lines

So, let's fix this by adding nvkm_error_ratelimited(), and using it to ratelimit errors from faults. This should be fine for developers, since it's nearly always only the first few faults that we care about seeing. Plus, you can turn off rate limiting in the kernel if you really need to.

good idea :)

Although I think we ultimately need a detection of "GPU state is trashed" where we cease to do anything and disable interrupts as some of those issues actually cause the kernel to get DOSed by nouveau spamming/handling interrupts. Not sure if we should go with this patch in the meantime, or if we should just go straight ahead and disable interrupts, which would also stop the spam.

Do you know if this patch here is enough to get the system to reboot normally or is it only spamming less?

Quoted 81 lines
---
 drivers/gpu/drm/nouveau/include/nvkm/core/subdev.h |  2 ++
 drivers/gpu/drm/nouveau/nvkm/subdev/bus/gf100.c    | 14 +++++++-------
 drivers/gpu/drm/nouveau/nvkm/subdev/bus/nv31.c     |  6 +++---
 drivers/gpu/drm/nouveau/nvkm/subdev/bus/nv50.c     |  6 +++---
 4 files changed, 15 insertions(+), 13 deletions(-)
drivers/gpu/drm/nouveau/include/nvkm/core/subdev.h +2 -0
index 1665738948fb..96113c8bee8c 100644
@@ -62,4 +62,6 @@ void nvkm_subdev_intr(struct nvkm_subdev *);
 #define nvkm_debug(s,f,a...) nvkm_printk((s), DEBUG,   info, f, ##a)
 #define nvkm_trace(s,f,a...) nvkm_printk((s), TRACE,   info, f, ##a)
 #define nvkm_spam(s,f,a...)  nvkm_printk((s),  SPAM,    dbg, f, ##a)
+
+#define nvkm_error_ratelimited(s,f,a...) nvkm_printk((s), ERROR, err_ratelimited, f, ##a)
 #endif
drivers/gpu/drm/nouveau/nvkm/subdev/bus/gf100.c +7 -7
index 53a6651ac225..80b5aaceeaad 100644
@@ -35,13 +35,13 @@ gf100_bus_intr(struct nvkm_bus *bus)
                u32 addr = nvkm_rd32(device, 0x009084);
                u32 data = nvkm_rd32(device, 0x009088);
 
-               nvkm_error(subdev,
-                          "MMIO %s of %08x FAULT at %06x [ %s%s%s]\n",
-                          (addr & 0x00000002) ? "write" : "read", data,
-                          (addr & 0x00fffffc),
-                          (stat & 0x00000002) ? "!ENGINE " : "",
-                          (stat & 0x00000004) ? "PRIVRING " : "",
-                          (stat & 0x00000008) ? "TIMEOUT " : "");
+               nvkm_error_ratelimited(subdev,
+                                      "MMIO %s of %08x FAULT at %06x [ %s%s%s]\n",
+                                      (addr & 0x00000002) ? "write" : "read", data,
+                                      (addr & 0x00fffffc),
+                                      (stat & 0x00000002) ? "!ENGINE " : "",
+                                      (stat & 0x00000004) ? "PRIVRING " : "",
+                                      (stat & 0x00000008) ? "TIMEOUT " : "");
 
                nvkm_wr32(device, 0x009084, 0x00000000);
                nvkm_wr32(device, 0x001100, (stat & 0x0000000e));
drivers/gpu/drm/nouveau/nvkm/subdev/bus/nv31.c +3 -3
index ad8da523bb22..c75e463f3501 100644
@@ -45,9 +45,9 @@ nv31_bus_intr(struct nvkm_bus *bus)
                u32 addr = nvkm_rd32(device, 0x009084);
                u32 data = nvkm_rd32(device, 0x009088);
 
-               nvkm_error(subdev, "MMIO %s of %08x FAULT at %06x\n",
-                          (addr & 0x00000002) ? "write" : "read", data,
-                          (addr & 0x00fffffc));
+               nvkm_error_ratelimited(subdev, "MMIO %s of %08x FAULT at %06x\n",
+                                      (addr & 0x00000002) ? "write" : "read", data,
+                                      (addr & 0x00fffffc));
 
                stat &= ~0x00000008;
                nvkm_wr32(device, 0x001100, 0x00000008);
drivers/gpu/drm/nouveau/nvkm/subdev/bus/nv50.c +3 -4
index 3a1e45adeedc..2055d0b100d3 100644
@@ -60,9 +60,9 @@ nv50_bus_intr(struct nvkm_bus *bus)
                u32 addr = nvkm_rd32(device, 0x009084);
                u32 data = nvkm_rd32(device, 0x009088);
 
-               nvkm_error(subdev, "MMIO %s of %08x FAULT at %06x\n",
-                          (addr & 0x00000002) ? "write" : "read", data,
-                          (addr & 0x00fffffc));
+               nvkm_error_ratelimited(subdev, "MMIO %s of %08x FAULT at %06x\n",
+                                      (addr & 0x00000002) ? "write" : "read", data,
+                                      (addr & 0x00fffffc));
 
                stat &= ~0x00000008;
                nvkm_wr32(device, 0x001100, 0x00000008);
--

2.35.1

author maintainer
#2 lore raw
Quoted 25 lines · Karol Herbst
Quoted 9 lines · Lyude Paul

There's plenty of ways to fudge the GPU when developing on nouveau by mistake, some of which can result in nouveau seriously spamming dmesg with fault errors. This can be somewhat annoying, as it can quickly overrun the message buffer (or your terminal emulator's buffer) and get rid of actually
useful feedback from the driver. While working on my new atomic only MST branch, I ran into this issue a couple of times.

Quoted 5 lines

So, let's fix this by adding nvkm_error_ratelimited(), and using it to ratelimit errors from faults. This should be fine for developers, since it's nearly always only the first few faults that we care about seeing. Plus, you can turn off rate limiting in the kernel if you really need to.

good idea :)

Although I think we ultimately need a detection of "GPU state is trashed" where we cease to do anything and disable interrupts as some of those issues actually cause the kernel to get DOSed by nouveau spamming/handling interrupts. Not sure if we should go with this patch in the meantime, or if we should just go straight ahead and disable interrupts, which would also stop the spam.

Quoted 2 lines

Do you know if this patch here is enough to get the system to reboot normally or is it only spamming less?

I'd have to try to come up with a setup that can hit this again, although it might not be that hard - last time I hit this it was because I was programming the wrong PBN values for MST payloads into nvidia's hardware.

So I can't remember for sure if I could reboot, but I do know I could actually log into the system easily and do stuff with ratelimiting applied so it's definitely better then the current situation.

Quoted 97 lines
Quoted 96 lines

---
 drivers/gpu/drm/nouveau/include/nvkm/core/subdev.h |  2 ++ drivers/gpu/drm/nouveau/nvkm/subdev/bus/gf100.c    | 14 +++++++------- drivers/gpu/drm/nouveau/nvkm/subdev/bus/nv31.c     |  6 +++--- drivers/gpu/drm/nouveau/nvkm/subdev/bus/nv50.c     |  6 +++---

 4 files changed, 15 insertions(+), 13 deletions(-)
drivers/gpu/drm/nouveau/include/nvkm/core/subdev.h +2 -0
index 1665738948fb..96113c8bee8c 100644
@@ -62,4 +62,6 @@ void nvkm_subdev_intr(struct nvkm_subdev *);
  #define nvkm_debug(s,f,a...) nvkm_printk((s), DEBUG,   info, f, ##a)
  #define nvkm_trace(s,f,a...) nvkm_printk((s), TRACE,   info, f, ##a)
  #define nvkm_spam(s,f,a...)  nvkm_printk((s),  SPAM,    dbg, f, ##a)
+
+#define nvkm_error_ratelimited(s,f,a...) nvkm_printk((s), ERROR,
 err_ratelimited, f, ##a)

 #endif

drivers/gpu/drm/nouveau/nvkm/subdev/bus/gf100.c +6 -7
index 53a6651ac225..80b5aaceeaad 100644
@@ -35,13 +35,13 @@ gf100_bus_intr(struct nvkm_bus *bus)
                 u32 addr = nvkm_rd32(device, 0x009084);
                 u32 data = nvkm_rd32(device, 0x009088);
 
-               nvkm_error(subdev,
-                          "MMIO %s of %08x FAULT at %06x [ %s%s%s]\n",
-                          (addr & 0x00000002) ? "write" : "read", data,
-                          (addr & 0x00fffffc),
-                          (stat & 0x00000002) ? "!ENGINE " : "",
-                          (stat & 0x00000004) ? "PRIVRING " : "",
-                          (stat & 0x00000008) ? "TIMEOUT " : "");
+               nvkm_error_ratelimited(subdev,
+                                      "MMIO %s of %08x FAULT at %06x [
 %s%s%s]\n",
+                                      (addr & 0x00000002) ? "write" :
 "read", data,
+                                      (addr & 0x00fffffc),
+                                      (stat & 0x00000002) ? "!ENGINE " :
 "",
+                                      (stat & 0x00000004) ? "PRIVRING " :

"",
+                                      (stat & 0x00000008) ? "TIMEOUT " :
"");

                nvkm_wr32(device, 0x009084, 0x00000000); nvkm_wr32(device, 0x001100, (stat & 0x0000000e));

drivers/gpu/drm/nouveau/nvkm/subdev/bus/nv31.c +3 -3
index ad8da523bb22..c75e463f3501 100644
@@ -45,9 +45,9 @@ nv31_bus_intr(struct nvkm_bus *bus)
                 u32 addr = nvkm_rd32(device, 0x009084);
                 u32 data = nvkm_rd32(device, 0x009088);
 
-               nvkm_error(subdev, "MMIO %s of %08x FAULT at %06x\n",
-                          (addr & 0x00000002) ? "write" : "read", data,
-                          (addr & 0x00fffffc));
+               nvkm_error_ratelimited(subdev, "MMIO %s of %08x FAULT at
 %06x\n",
+                                      (addr & 0x00000002) ? "write" :
 "read", data,
+                                      (addr & 0x00fffffc));
 

                stat &= ~0x00000008; nvkm_wr32(device, 0x001100, 0x00000008);

drivers/gpu/drm/nouveau/nvkm/subdev/bus/nv50.c +3 -3
index 3a1e45adeedc..2055d0b100d3 100644
@@ -60,9 +60,9 @@ nv50_bus_intr(struct nvkm_bus *bus)
                 u32 addr = nvkm_rd32(device, 0x009084);
                 u32 data = nvkm_rd32(device, 0x009088);
 
-               nvkm_error(subdev, "MMIO %s of %08x FAULT at %06x\n",
-                          (addr & 0x00000002) ? "write" : "read", data,
-                          (addr & 0x00fffffc));
+               nvkm_error_ratelimited(subdev, "MMIO %s of %08x FAULT at
 %06x\n",
+                                      (addr & 0x00000002) ? "write" :
 "read", data,
+                                      (addr & 0x00fffffc));
 

                stat &= ~0x00000008; nvkm_wr32(device, 0x001100, 0x00000008); --
2.35.1

Signature · 3 lines
Cheers,
 Lyude Paul (she/her)
 Software Engineer at Red Hat
#3 lore raw
Quoted 30 lines · Lyude Paul
Quoted 25 lines · Karol Herbst
Quoted 9 lines · Lyude Paul

There's plenty of ways to fudge the GPU when developing on nouveau by mistake, some of which can result in nouveau seriously spamming dmesg with fault errors. This can be somewhat annoying, as it can quickly overrun the message buffer (or your terminal emulator's buffer) and get rid of actually
useful feedback from the driver. While working on my new atomic only MST branch, I ran into this issue a couple of times.

Quoted 5 lines

So, let's fix this by adding nvkm_error_ratelimited(), and using it to ratelimit errors from faults. This should be fine for developers, since it's nearly always only the first few faults that we care about seeing. Plus, you can turn off rate limiting in the kernel if you really need to.

good idea :)

Although I think we ultimately need a detection of "GPU state is trashed" where we cease to do anything and disable interrupts as some of those issues actually cause the kernel to get DOSed by nouveau spamming/handling interrupts. Not sure if we should go with this patch in the meantime, or if we should just go straight ahead and disable interrupts, which would also stop the spam.

Quoted 2 lines

Do you know if this patch here is enough to get the system to reboot normally or is it only spamming less?

Quoted 8 lines

I'd have to try to come up with a setup that can hit this again, although it might not be that hard - last time I hit this it was because I was programming the wrong PBN values for MST payloads into nvidia's hardware.

So I can't remember for sure if I could reboot, but I do know I could actually log into the system easily and do stuff with ratelimiting applied so it's definitely better then the current situation.

I can play around with your patch a little, as with broken multithreading and OpenCL I can trash my entire machine within split seconds quite reliably :)

Quoted 103 lines
Quoted 91 lines
Quoted 91 lines
---
 drivers/gpu/drm/nouveau/include/nvkm/core/subdev.h |  2 ++
 drivers/gpu/drm/nouveau/nvkm/subdev/bus/gf100.c    | 14 +++++++-------
 drivers/gpu/drm/nouveau/nvkm/subdev/bus/nv31.c     |  6 +++---
 drivers/gpu/drm/nouveau/nvkm/subdev/bus/nv50.c     |  6 +++---
 4 files changed, 15 insertions(+), 13 deletions(-)
drivers/gpu/drm/nouveau/include/nvkm/core/subdev.h +2 -0
index 1665738948fb..96113c8bee8c 100644
@@ -62,4 +62,6 @@ void nvkm_subdev_intr(struct nvkm_subdev *);
 #define nvkm_debug(s,f,a...) nvkm_printk((s), DEBUG,   info, f, ##a)
 #define nvkm_trace(s,f,a...) nvkm_printk((s), TRACE,   info, f, ##a)
 #define nvkm_spam(s,f,a...)  nvkm_printk((s),  SPAM,    dbg, f, ##a)
+
+#define nvkm_error_ratelimited(s,f,a...) nvkm_printk((s), ERROR,
 err_ratelimited, f, ##a)

#endif

drivers/gpu/drm/nouveau/nvkm/subdev/bus/gf100.c +6 -7
index 53a6651ac225..80b5aaceeaad 100644
@@ -35,13 +35,13 @@ gf100_bus_intr(struct nvkm_bus *bus)
                u32 addr = nvkm_rd32(device, 0x009084);
                u32 data = nvkm_rd32(device, 0x009088);
 
-               nvkm_error(subdev,
-                          "MMIO %s of %08x FAULT at %06x [ %s%s%s]\n",
-                          (addr & 0x00000002) ? "write" : "read", data,
-                          (addr & 0x00fffffc),
-                          (stat & 0x00000002) ? "!ENGINE " : "",
-                          (stat & 0x00000004) ? "PRIVRING " : "",
-                          (stat & 0x00000008) ? "TIMEOUT " : "");
+               nvkm_error_ratelimited(subdev,
+                                      "MMIO %s of %08x FAULT at %06x [
 %s%s%s]\n",
+                                      (addr & 0x00000002) ? "write" :
 "read", data,
+                                      (addr & 0x00fffffc),
+                                      (stat & 0x00000002) ? "!ENGINE " :
 "",
+                                      (stat & 0x00000004) ? "PRIVRING " :

"",
+ (stat & 0x00000008) ? "TIMEOUT " :
"");

                nvkm_wr32(device, 0x009084, 0x00000000);
                nvkm_wr32(device, 0x001100, (stat & 0x0000000e));
drivers/gpu/drm/nouveau/nvkm/subdev/bus/nv31.c +3 -3
index ad8da523bb22..c75e463f3501 100644
@@ -45,9 +45,9 @@ nv31_bus_intr(struct nvkm_bus *bus)
                u32 addr = nvkm_rd32(device, 0x009084);
                u32 data = nvkm_rd32(device, 0x009088);
 
-               nvkm_error(subdev, "MMIO %s of %08x FAULT at %06x\n",
-                          (addr & 0x00000002) ? "write" : "read", data,
-                          (addr & 0x00fffffc));
+               nvkm_error_ratelimited(subdev, "MMIO %s of %08x FAULT at
 %06x\n",
+                                      (addr & 0x00000002) ? "write" :
 "read", data,
+                                      (addr & 0x00fffffc));
 
                stat &= ~0x00000008;
                nvkm_wr32(device, 0x001100, 0x00000008);
drivers/gpu/drm/nouveau/nvkm/subdev/bus/nv50.c +3 -3
index 3a1e45adeedc..2055d0b100d3 100644
@@ -60,9 +60,9 @@ nv50_bus_intr(struct nvkm_bus *bus)
                u32 addr = nvkm_rd32(device, 0x009084);
                u32 data = nvkm_rd32(device, 0x009088);
 
-               nvkm_error(subdev, "MMIO %s of %08x FAULT at %06x\n",
-                          (addr & 0x00000002) ? "write" : "read", data,
-                          (addr & 0x00fffffc));
+               nvkm_error_ratelimited(subdev, "MMIO %s of %08x FAULT at
 %06x\n",
+                                      (addr & 0x00000002) ? "write" :
 "read", data,
+                                      (addr & 0x00fffffc));
 
Quoted 6 lines
Quoted 5 lines
                stat &= ~0x00000008;
                nvkm_wr32(device, 0x001100, 0x00000008);

--
2.35.1

--
Cheers,
Lyude Paul (she/her)
Software Engineer at Red Hat

Reviewed-by on v1 1/1
#4 lore raw
In reply to Lyude Paul

not able to hit any error on my machine, but regardless:

I suspect there are more places where we could put it, but we can add those later.

Anyway, I think it's valuable to push it through fixes, not sure how far back we want to CC stable though.

Quoted 94 lines · Lyude Paul

There's plenty of ways to fudge the GPU when developing on nouveau by mistake, some of which can result in nouveau seriously spamming dmesg with fault errors. This can be somewhat annoying, as it can quickly overrun the message buffer (or your terminal emulator's buffer) and get rid of actually useful feedback from the driver. While working on my new atomic only MST branch, I ran into this issue a couple of times.

So, let's fix this by adding nvkm_error_ratelimited(), and using it to ratelimit errors from faults. This should be fine for developers, since it's nearly always only the first few faults that we care about seeing. Plus, you can turn off rate limiting in the kernel if you really need to.

---
 drivers/gpu/drm/nouveau/include/nvkm/core/subdev.h |  2 ++
 drivers/gpu/drm/nouveau/nvkm/subdev/bus/gf100.c    | 14 +++++++-------
 drivers/gpu/drm/nouveau/nvkm/subdev/bus/nv31.c     |  6 +++---
 drivers/gpu/drm/nouveau/nvkm/subdev/bus/nv50.c     |  6 +++---
 4 files changed, 15 insertions(+), 13 deletions(-)
drivers/gpu/drm/nouveau/include/nvkm/core/subdev.h +2 -0
index 1665738948fb..96113c8bee8c 100644
@@ -62,4 +62,6 @@ void nvkm_subdev_intr(struct nvkm_subdev *);
 #define nvkm_debug(s,f,a...) nvkm_printk((s), DEBUG,   info, f, ##a)
 #define nvkm_trace(s,f,a...) nvkm_printk((s), TRACE,   info, f, ##a)
 #define nvkm_spam(s,f,a...)  nvkm_printk((s),  SPAM,    dbg, f, ##a)
+
+#define nvkm_error_ratelimited(s,f,a...) nvkm_printk((s), ERROR, err_ratelimited, f, ##a)
 #endif
drivers/gpu/drm/nouveau/nvkm/subdev/bus/gf100.c +7 -7
index 53a6651ac225..80b5aaceeaad 100644
@@ -35,13 +35,13 @@ gf100_bus_intr(struct nvkm_bus *bus)
                u32 addr = nvkm_rd32(device, 0x009084);
                u32 data = nvkm_rd32(device, 0x009088);
 
-               nvkm_error(subdev,
-                          "MMIO %s of %08x FAULT at %06x [ %s%s%s]\n",
-                          (addr & 0x00000002) ? "write" : "read", data,
-                          (addr & 0x00fffffc),
-                          (stat & 0x00000002) ? "!ENGINE " : "",
-                          (stat & 0x00000004) ? "PRIVRING " : "",
-                          (stat & 0x00000008) ? "TIMEOUT " : "");
+               nvkm_error_ratelimited(subdev,
+                                      "MMIO %s of %08x FAULT at %06x [ %s%s%s]\n",
+                                      (addr & 0x00000002) ? "write" : "read", data,
+                                      (addr & 0x00fffffc),
+                                      (stat & 0x00000002) ? "!ENGINE " : "",
+                                      (stat & 0x00000004) ? "PRIVRING " : "",
+                                      (stat & 0x00000008) ? "TIMEOUT " : "");
 
                nvkm_wr32(device, 0x009084, 0x00000000);
                nvkm_wr32(device, 0x001100, (stat & 0x0000000e));
drivers/gpu/drm/nouveau/nvkm/subdev/bus/nv31.c +3 -3
index ad8da523bb22..c75e463f3501 100644
@@ -45,9 +45,9 @@ nv31_bus_intr(struct nvkm_bus *bus)
                u32 addr = nvkm_rd32(device, 0x009084);
                u32 data = nvkm_rd32(device, 0x009088);
 
-               nvkm_error(subdev, "MMIO %s of %08x FAULT at %06x\n",
-                          (addr & 0x00000002) ? "write" : "read", data,
-                          (addr & 0x00fffffc));
+               nvkm_error_ratelimited(subdev, "MMIO %s of %08x FAULT at %06x\n",
+                                      (addr & 0x00000002) ? "write" : "read", data,
+                                      (addr & 0x00fffffc));
 
                stat &= ~0x00000008;
                nvkm_wr32(device, 0x001100, 0x00000008);
drivers/gpu/drm/nouveau/nvkm/subdev/bus/nv50.c +3 -4
index 3a1e45adeedc..2055d0b100d3 100644
@@ -60,9 +60,9 @@ nv50_bus_intr(struct nvkm_bus *bus)
                u32 addr = nvkm_rd32(device, 0x009084);
                u32 data = nvkm_rd32(device, 0x009088);
 
-               nvkm_error(subdev, "MMIO %s of %08x FAULT at %06x\n",
-                          (addr & 0x00000002) ? "write" : "read", data,
-                          (addr & 0x00fffffc));
+               nvkm_error_ratelimited(subdev, "MMIO %s of %08x FAULT at %06x\n",
+                                      (addr & 0x00000002) ? "write" : "read", data,
+                                      (addr & 0x00fffffc));
 
                stat &= ~0x00000008;
                nvkm_wr32(device, 0x001100, 0x00000008);
--

2.35.1

#5 lore raw

Actually, there is another place we have to put this, we can spam "SCHED_ERROR" things which are triggered within multiple drivers/gpu/drm/nouveau/nvkm/engine/fifo/ files.

Quoted 106 lines · Karol Herbst

not able to hit any error on my machine, but regardless:

I suspect there are more places where we could put it, but we can add those later.

Anyway, I think it's valuable to push it through fixes, not sure how far back we want to CC stable though.

Quoted 94 lines · Lyude Paul

There's plenty of ways to fudge the GPU when developing on nouveau by mistake, some of which can result in nouveau seriously spamming dmesg with fault errors. This can be somewhat annoying, as it can quickly overrun the message buffer (or your terminal emulator's buffer) and get rid of actually useful feedback from the driver. While working on my new atomic only MST branch, I ran into this issue a couple of times.

So, let's fix this by adding nvkm_error_ratelimited(), and using it to ratelimit errors from faults. This should be fine for developers, since it's nearly always only the first few faults that we care about seeing. Plus, you can turn off rate limiting in the kernel if you really need to.

---
 drivers/gpu/drm/nouveau/include/nvkm/core/subdev.h |  2 ++
 drivers/gpu/drm/nouveau/nvkm/subdev/bus/gf100.c    | 14 +++++++-------
 drivers/gpu/drm/nouveau/nvkm/subdev/bus/nv31.c     |  6 +++---
 drivers/gpu/drm/nouveau/nvkm/subdev/bus/nv50.c     |  6 +++---
 4 files changed, 15 insertions(+), 13 deletions(-)
drivers/gpu/drm/nouveau/include/nvkm/core/subdev.h +2 -0
index 1665738948fb..96113c8bee8c 100644
@@ -62,4 +62,6 @@ void nvkm_subdev_intr(struct nvkm_subdev *);
 #define nvkm_debug(s,f,a...) nvkm_printk((s), DEBUG,   info, f, ##a)
 #define nvkm_trace(s,f,a...) nvkm_printk((s), TRACE,   info, f, ##a)
 #define nvkm_spam(s,f,a...)  nvkm_printk((s),  SPAM,    dbg, f, ##a)
+
+#define nvkm_error_ratelimited(s,f,a...) nvkm_printk((s), ERROR, err_ratelimited, f, ##a)
 #endif
drivers/gpu/drm/nouveau/nvkm/subdev/bus/gf100.c +7 -7
index 53a6651ac225..80b5aaceeaad 100644
@@ -35,13 +35,13 @@ gf100_bus_intr(struct nvkm_bus *bus)
                u32 addr = nvkm_rd32(device, 0x009084);
                u32 data = nvkm_rd32(device, 0x009088);
 
-               nvkm_error(subdev,
-                          "MMIO %s of %08x FAULT at %06x [ %s%s%s]\n",
-                          (addr & 0x00000002) ? "write" : "read", data,
-                          (addr & 0x00fffffc),
-                          (stat & 0x00000002) ? "!ENGINE " : "",
-                          (stat & 0x00000004) ? "PRIVRING " : "",
-                          (stat & 0x00000008) ? "TIMEOUT " : "");
+               nvkm_error_ratelimited(subdev,
+                                      "MMIO %s of %08x FAULT at %06x [ %s%s%s]\n",
+                                      (addr & 0x00000002) ? "write" : "read", data,
+                                      (addr & 0x00fffffc),
+                                      (stat & 0x00000002) ? "!ENGINE " : "",
+                                      (stat & 0x00000004) ? "PRIVRING " : "",
+                                      (stat & 0x00000008) ? "TIMEOUT " : "");
 
                nvkm_wr32(device, 0x009084, 0x00000000);
                nvkm_wr32(device, 0x001100, (stat & 0x0000000e));
drivers/gpu/drm/nouveau/nvkm/subdev/bus/nv31.c +3 -3
index ad8da523bb22..c75e463f3501 100644
@@ -45,9 +45,9 @@ nv31_bus_intr(struct nvkm_bus *bus)
                u32 addr = nvkm_rd32(device, 0x009084);
                u32 data = nvkm_rd32(device, 0x009088);
 
-               nvkm_error(subdev, "MMIO %s of %08x FAULT at %06x\n",
-                          (addr & 0x00000002) ? "write" : "read", data,
-                          (addr & 0x00fffffc));
+               nvkm_error_ratelimited(subdev, "MMIO %s of %08x FAULT at %06x\n",
+                                      (addr & 0x00000002) ? "write" : "read", data,
+                                      (addr & 0x00fffffc));
 
                stat &= ~0x00000008;
                nvkm_wr32(device, 0x001100, 0x00000008);
drivers/gpu/drm/nouveau/nvkm/subdev/bus/nv50.c +3 -4
index 3a1e45adeedc..2055d0b100d3 100644
@@ -60,9 +60,9 @@ nv50_bus_intr(struct nvkm_bus *bus)
                u32 addr = nvkm_rd32(device, 0x009084);
                u32 data = nvkm_rd32(device, 0x009088);
 
-               nvkm_error(subdev, "MMIO %s of %08x FAULT at %06x\n",
-                          (addr & 0x00000002) ? "write" : "read", data,
-                          (addr & 0x00fffffc));
+               nvkm_error_ratelimited(subdev, "MMIO %s of %08x FAULT at %06x\n",
+                                      (addr & 0x00000002) ? "write" : "read", data,
+                                      (addr & 0x00fffffc));
 
                stat &= ~0x00000008;
                nvkm_wr32(device, 0x001100, 0x00000008);
--

2.35.1

author maintainer
#6 lore raw
In reply to Karol Herbst
Quoted 9 lines · Karol Herbst

not able to hit any error on my machine, but regardless:

I suspect there are more places where we could put it, but we can add those later.

Anyway, I think it's valuable to push it through fixes, not sure how far back we want to CC stable though.

JFYI - dim fixes can figure this out if you just give it the commit that introduced the issue.

Anyway, will push this upstream in a moment

Quoted 112 lines · Lyude Paul
Quoted 109 lines

There's plenty of ways to fudge the GPU when developing on nouveau by mistake, some of which can result in nouveau seriously spamming dmesg with fault errors. This can be somewhat annoying, as it can quickly overrun the message buffer (or your terminal emulator's buffer) and get rid of actually
useful feedback from the driver. While working on my new atomic only MST branch, I ran into this issue a couple of times.

So, let's fix this by adding nvkm_error_ratelimited(), and using it to ratelimit errors from faults. This should be fine for developers, since it's nearly always only the first few faults that we care about seeing. Plus, you can turn off rate limiting in the kernel if you really need to.

---
 drivers/gpu/drm/nouveau/include/nvkm/core/subdev.h |  2 ++ drivers/gpu/drm/nouveau/nvkm/subdev/bus/gf100.c    | 14 +++++++------- drivers/gpu/drm/nouveau/nvkm/subdev/bus/nv31.c     |  6 +++--- drivers/gpu/drm/nouveau/nvkm/subdev/bus/nv50.c     |  6 +++---

 4 files changed, 15 insertions(+), 13 deletions(-)
drivers/gpu/drm/nouveau/include/nvkm/core/subdev.h +2 -0
index 1665738948fb..96113c8bee8c 100644
@@ -62,4 +62,6 @@ void nvkm_subdev_intr(struct nvkm_subdev *);
  #define nvkm_debug(s,f,a...) nvkm_printk((s), DEBUG,   info, f, ##a)
  #define nvkm_trace(s,f,a...) nvkm_printk((s), TRACE,   info, f, ##a)
  #define nvkm_spam(s,f,a...)  nvkm_printk((s),  SPAM,    dbg, f, ##a)
+
+#define nvkm_error_ratelimited(s,f,a...) nvkm_printk((s), ERROR,
 err_ratelimited, f, ##a)

 #endif

drivers/gpu/drm/nouveau/nvkm/subdev/bus/gf100.c +6 -7
index 53a6651ac225..80b5aaceeaad 100644
@@ -35,13 +35,13 @@ gf100_bus_intr(struct nvkm_bus *bus)
                 u32 addr = nvkm_rd32(device, 0x009084);
                 u32 data = nvkm_rd32(device, 0x009088);
 
-               nvkm_error(subdev,
-                          "MMIO %s of %08x FAULT at %06x [ %s%s%s]\n",
-                          (addr & 0x00000002) ? "write" : "read", data,
-                          (addr & 0x00fffffc),
-                          (stat & 0x00000002) ? "!ENGINE " : "",
-                          (stat & 0x00000004) ? "PRIVRING " : "",
-                          (stat & 0x00000008) ? "TIMEOUT " : "");
+               nvkm_error_ratelimited(subdev,
+                                      "MMIO %s of %08x FAULT at %06x [
 %s%s%s]\n",
+                                      (addr & 0x00000002) ? "write" :
 "read", data,
+                                      (addr & 0x00fffffc),
+                                      (stat & 0x00000002) ? "!ENGINE " :
 "",
+                                      (stat & 0x00000004) ? "PRIVRING " :

"",
+                                      (stat & 0x00000008) ? "TIMEOUT " :
"");

                nvkm_wr32(device, 0x009084, 0x00000000); nvkm_wr32(device, 0x001100, (stat & 0x0000000e));

drivers/gpu/drm/nouveau/nvkm/subdev/bus/nv31.c +3 -3
index ad8da523bb22..c75e463f3501 100644
@@ -45,9 +45,9 @@ nv31_bus_intr(struct nvkm_bus *bus)
                 u32 addr = nvkm_rd32(device, 0x009084);
                 u32 data = nvkm_rd32(device, 0x009088);
 
-               nvkm_error(subdev, "MMIO %s of %08x FAULT at %06x\n",
-                          (addr & 0x00000002) ? "write" : "read", data,
-                          (addr & 0x00fffffc));
+               nvkm_error_ratelimited(subdev, "MMIO %s of %08x FAULT at
 %06x\n",
+                                      (addr & 0x00000002) ? "write" :
 "read", data,
+                                      (addr & 0x00fffffc));
 

                stat &= ~0x00000008; nvkm_wr32(device, 0x001100, 0x00000008);

drivers/gpu/drm/nouveau/nvkm/subdev/bus/nv50.c +3 -3
index 3a1e45adeedc..2055d0b100d3 100644
@@ -60,9 +60,9 @@ nv50_bus_intr(struct nvkm_bus *bus)
                 u32 addr = nvkm_rd32(device, 0x009084);
                 u32 data = nvkm_rd32(device, 0x009088);
 
-               nvkm_error(subdev, "MMIO %s of %08x FAULT at %06x\n",
-                          (addr & 0x00000002) ? "write" : "read", data,
-                          (addr & 0x00fffffc));
+               nvkm_error_ratelimited(subdev, "MMIO %s of %08x FAULT at
 %06x\n",
+                                      (addr & 0x00000002) ? "write" :
 "read", data,
+                                      (addr & 0x00fffffc));
 

                stat &= ~0x00000008; nvkm_wr32(device, 0x001100, 0x00000008); --
2.35.1

Signature · 3 lines
Cheers,
 Lyude Paul (she/her)
 Software Engineer at Red Hat
author maintainer
#7 lore raw

Mhhh - looks like it goes far back enough I probably should just Cc it without a version tag and see how far it applies for greg. Will do that

Quoted 135 lines · Lyude Paul
Quoted 9 lines · Karol Herbst

not able to hit any error on my machine, but regardless:

I suspect there are more places where we could put it, but we can add those later.

Anyway, I think it's valuable to push it through fixes, not sure how far back we want to CC stable though.

JFYI - dim fixes can figure this out if you just give it the commit that introduced the issue.

Anyway, will push this upstream in a moment

Quoted 118 lines · Lyude Paul
Quoted 115 lines

There's plenty of ways to fudge the GPU when developing on nouveau by mistake, some of which can result in nouveau seriously spamming dmesg with
fault errors. This can be somewhat annoying, as it can quickly overrun the
message buffer (or your terminal emulator's buffer) and get rid of actually
useful feedback from the driver. While working on my new atomic only MST branch, I ran into this issue a couple of times.

So, let's fix this by adding nvkm_error_ratelimited(), and using it to ratelimit errors from faults. This should be fine for developers, since it's nearly always only the first few faults that we care about seeing. Plus, you can turn off rate limiting in the kernel if you really need to.

---
 drivers/gpu/drm/nouveau/include/nvkm/core/subdev.h |  2 ++ drivers/gpu/drm/nouveau/nvkm/subdev/bus/gf100.c    | 14 +++++++------- drivers/gpu/drm/nouveau/nvkm/subdev/bus/nv31.c     |  6 +++--- drivers/gpu/drm/nouveau/nvkm/subdev/bus/nv50.c     |  6 +++---

 4 files changed, 15 insertions(+), 13 deletions(-)
drivers/gpu/drm/nouveau/include/nvkm/core/subdev.h +2 -0
index 1665738948fb..96113c8bee8c 100644
@@ -62,4 +62,6 @@ void nvkm_subdev_intr(struct nvkm_subdev *);
  #define nvkm_debug(s,f,a...) nvkm_printk((s), DEBUG,   info, f, ##a)
  #define nvkm_trace(s,f,a...) nvkm_printk((s), TRACE,   info, f, ##a)
  #define nvkm_spam(s,f,a...)  nvkm_printk((s),  SPAM,    dbg, f, ##a)
+
+#define nvkm_error_ratelimited(s,f,a...) nvkm_printk((s), ERROR,
 err_ratelimited, f, ##a)

 #endif

drivers/gpu/drm/nouveau/nvkm/subdev/bus/gf100.c +5 -7
index 53a6651ac225..80b5aaceeaad 100644
@@ -35,13 +35,13 @@ gf100_bus_intr(struct nvkm_bus *bus)
                 u32 addr = nvkm_rd32(device, 0x009084);
                 u32 data = nvkm_rd32(device, 0x009088);
 
-               nvkm_error(subdev,
-                          "MMIO %s of %08x FAULT at %06x [ %s%s%s]\n",
-                          (addr & 0x00000002) ? "write" : "read", data,
-                          (addr & 0x00fffffc),
-                          (stat & 0x00000002) ? "!ENGINE " : "",
-                          (stat & 0x00000004) ? "PRIVRING " : "",
-                          (stat & 0x00000008) ? "TIMEOUT " : "");
+               nvkm_error_ratelimited(subdev,
+                                      "MMIO %s of %08x FAULT at %06x [
 %s%s%s]\n",
+                                      (addr & 0x00000002) ? "write" :
 "read", data,
+                                      (addr & 0x00fffffc),
+                                      (stat & 0x00000002) ? "!ENGINE "

:
"",
+                                      (stat & 0x00000004) ? "PRIVRING " :
"",
+                                      (stat & 0x00000008) ? "TIMEOUT " :
"");

                nvkm_wr32(device, 0x009084, 0x00000000); nvkm_wr32(device, 0x001100, (stat & 0x0000000e));

drivers/gpu/drm/nouveau/nvkm/subdev/bus/nv31.c +3 -3
index ad8da523bb22..c75e463f3501 100644
@@ -45,9 +45,9 @@ nv31_bus_intr(struct nvkm_bus *bus)
                 u32 addr = nvkm_rd32(device, 0x009084);
                 u32 data = nvkm_rd32(device, 0x009088);
 
-               nvkm_error(subdev, "MMIO %s of %08x FAULT at %06x\n",
-                          (addr & 0x00000002) ? "write" : "read", data,
-                          (addr & 0x00fffffc));
+               nvkm_error_ratelimited(subdev, "MMIO %s of %08x FAULT at
 %06x\n",
+                                      (addr & 0x00000002) ? "write" :
 "read", data,
+                                      (addr & 0x00fffffc));
 

                stat &= ~0x00000008; nvkm_wr32(device, 0x001100, 0x00000008);

drivers/gpu/drm/nouveau/nvkm/subdev/bus/nv50.c +3 -3
index 3a1e45adeedc..2055d0b100d3 100644
@@ -60,9 +60,9 @@ nv50_bus_intr(struct nvkm_bus *bus)
                 u32 addr = nvkm_rd32(device, 0x009084);
                 u32 data = nvkm_rd32(device, 0x009088);
 
-               nvkm_error(subdev, "MMIO %s of %08x FAULT at %06x\n",
-                          (addr & 0x00000002) ? "write" : "read", data,
-                          (addr & 0x00fffffc));
+               nvkm_error_ratelimited(subdev, "MMIO %s of %08x FAULT at
 %06x\n",
+                                      (addr & 0x00000002) ? "write" :
 "read", data,
+                                      (addr & 0x00fffffc));
 

                stat &= ~0x00000008; nvkm_wr32(device, 0x001100, 0x00000008); --
2.35.1

Signature · 3 lines
Cheers,
 Lyude Paul (she/her)
 Software Engineer at Red Hat