Re: [PATCH 3/6] drm/amdgpu: Rework coredump to use memory dynamically

From: Christian König
Date: Wed Jul 12 2023 - 07:37:10 EST


Am 12.07.23 um 12:56 schrieb Lucas Stach:
Sorry, accidentally hit sent on the previous mail.

Am Mittwoch, dem 12.07.2023 um 12:39 +0200 schrieb Christian König:
Am 12.07.23 um 10:59 schrieb Lucas Stach:
Am Mittwoch, dem 12.07.2023 um 10:37 +0200 schrieb Christian König:
Am 11.07.23 um 23:34 schrieb André Almeida:
Instead of storing coredump information inside amdgpu_device struct,
move if to a proper separated struct and allocate it dynamically. This
will make it easier to further expand the logged information.
Verry big NAK to this. The problem is that memory allocation isn't
allowed during a GPU reset.

What you could do is to allocate the memory with GFP_ATOMIC or similar,
but for a large structure that might not be possible.

I'm still not fully clear on what the rules are here. In etnaviv we do
devcoredump allocation in the GPU reset path with __GFP_NOWARN |
__GFP_NORETRY, which means the allocation will kick memory reclaim if
necessary, but will just give up if no memory could be made available
easily. This satisfies the forward progress guarantee in the absence of
successful memory allocation, which is the most important property in
this path, I think.

However, I'm not sure if the reclaim could lead to locking issues or
something like that with the more complex use-cases with MMU notifiers
and stuff like that. Christian, do you have any experience or
information that would be good to share in this regard?
Yeah, very good question.

__GFP_NORETRY isn't sufficient as far as I know. Reclaim must be
completely suppressed to be able to allocate in a GPU reset handler.

Daniel added lockdep annotation to some of the dma-fence signaling paths
and this yielded quite a bunch of potential deadlocks.

It's not even that reclaim itself waits for dma_fences (that can happen,
but is quite unlikely), but rather that reclaim needs locks and under
those locks we then wait for dma_fences.

We should probably add a define somewhere which documents that
(GFP_ATOMIC | __NO_WARN) should be used in the GPU reset handlers when
allocating memory for coredumps.

Hm, if the problem is the direct reclaim path where we might recurse on
a lock through those indirect dependencies then we should document this
somewhere. kswapd reclaim should be fine as far as I can see, as we'll
guarantee progress without waiting for the background reclaim.

I don't think it's appropriate to dip into the atomic allocation
reserves for a best-effort thing like writing the devcoredump,

Yeah, that was also my concern the last time we discussed this.

so I think this should be GFP_NOWAIT, which will also avoid the direct
reclaim path.

Yeah, good idea. I wasn't aware that this existed.

Regards,
Christian.


Regards,
Lucas

Regards,
Christian.

Regards,
Lucas

Regards,
Christian.

Signed-off-by: André Almeida <andrealmeid@xxxxxxxxxx>
---
drivers/gpu/drm/amd/amdgpu/amdgpu.h | 14 +++--
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 65 ++++++++++++++--------
2 files changed, 51 insertions(+), 28 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
index dbe062a087c5..e1cc83a89d46 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
@@ -1068,11 +1068,6 @@ struct amdgpu_device {
uint32_t *reset_dump_reg_list;
uint32_t *reset_dump_reg_value;
int num_regs;
-#ifdef CONFIG_DEV_COREDUMP
- struct amdgpu_task_info reset_task_info;
- bool reset_vram_lost;
- struct timespec64 reset_time;
-#endif
bool scpm_enabled;
uint32_t scpm_status;
@@ -1085,6 +1080,15 @@ struct amdgpu_device {
uint32_t aid_mask;
};
+#ifdef CONFIG_DEV_COREDUMP
+struct amdgpu_coredump_info {
+ struct amdgpu_device *adev;
+ struct amdgpu_task_info reset_task_info;
+ struct timespec64 reset_time;
+ bool reset_vram_lost;
+};
+#endif
+
static inline struct amdgpu_device *drm_to_adev(struct drm_device *ddev)
{
return container_of(ddev, struct amdgpu_device, ddev);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
index e25f085ee886..23b9784e9787 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
@@ -4963,12 +4963,17 @@ static int amdgpu_reset_reg_dumps(struct amdgpu_device *adev)
return 0;
}
-#ifdef CONFIG_DEV_COREDUMP
+#ifndef CONFIG_DEV_COREDUMP
+static void amdgpu_coredump(struct amdgpu_device *adev, bool vram_lost,
+ struct amdgpu_reset_context *reset_context)
+{
+}
+#else
static ssize_t amdgpu_devcoredump_read(char *buffer, loff_t offset,
size_t count, void *data, size_t datalen)
{
struct drm_printer p;
- struct amdgpu_device *adev = data;
+ struct amdgpu_coredump_info *coredump = data;
struct drm_print_iterator iter;
int i;
@@ -4982,21 +4987,21 @@ static ssize_t amdgpu_devcoredump_read(char *buffer, loff_t offset,
drm_printf(&p, "**** AMDGPU Device Coredump ****\n");
drm_printf(&p, "kernel: " UTS_RELEASE "\n");
drm_printf(&p, "module: " KBUILD_MODNAME "\n");
- drm_printf(&p, "time: %lld.%09ld\n", adev->reset_time.tv_sec, adev->reset_time.tv_nsec);
- if (adev->reset_task_info.pid)
+ drm_printf(&p, "time: %lld.%09ld\n", coredump->reset_time.tv_sec, coredump->reset_time.tv_nsec);
+ if (coredump->reset_task_info.pid)
drm_printf(&p, "process_name: %s PID: %d\n",
- adev->reset_task_info.process_name,
- adev->reset_task_info.pid);
+ coredump->reset_task_info.process_name,
+ coredump->reset_task_info.pid);
- if (adev->reset_vram_lost)
+ if (coredump->reset_vram_lost)
drm_printf(&p, "VRAM is lost due to GPU reset!\n");
- if (adev->num_regs) {
+ if (coredump->adev->num_regs) {
drm_printf(&p, "AMDGPU register dumps:\nOffset: Value:\n");
- for (i = 0; i < adev->num_regs; i++)
+ for (i = 0; i < coredump->adev->num_regs; i++)
drm_printf(&p, "0x%08x: 0x%08x\n",
- adev->reset_dump_reg_list[i],
- adev->reset_dump_reg_value[i]);
+ coredump->adev->reset_dump_reg_list[i],
+ coredump->adev->reset_dump_reg_value[i]);
}
return count - iter.remain;
@@ -5004,14 +5009,34 @@ static ssize_t amdgpu_devcoredump_read(char *buffer, loff_t offset,
static void amdgpu_devcoredump_free(void *data)
{
+ kfree(data);
}
-static void amdgpu_reset_capture_coredumpm(struct amdgpu_device *adev)
+static void amdgpu_coredump(struct amdgpu_device *adev, bool vram_lost,
+ struct amdgpu_reset_context *reset_context)
{
+ struct amdgpu_coredump_info *coredump;
struct drm_device *dev = adev_to_drm(adev);
- ktime_get_ts64(&adev->reset_time);
- dev_coredumpm(dev->dev, THIS_MODULE, adev, 0, GFP_KERNEL,
+ coredump = kmalloc(sizeof(*coredump), GFP_KERNEL);
+
+ if (!coredump) {
+ DRM_ERROR("%s: failed to allocate memory for coredump\n", __func__);
+ return;
+ }
+
+ memset(coredump, 0, sizeof(*coredump));
+
+ coredump->reset_vram_lost = vram_lost;
+
+ if (reset_context->job && reset_context->job->vm)
+ coredump->reset_task_info = reset_context->job->vm->task_info;
+
+ coredump->adev = adev;
+
+ ktime_get_ts64(&coredump->reset_time);
+
+ dev_coredumpm(dev->dev, THIS_MODULE, coredump, 0, GFP_KERNEL,
amdgpu_devcoredump_read, amdgpu_devcoredump_free);
}
#endif
@@ -5119,15 +5144,9 @@ int amdgpu_do_asic_reset(struct list_head *device_list_handle,
goto out;
vram_lost = amdgpu_device_check_vram_lost(tmp_adev);
-#ifdef CONFIG_DEV_COREDUMP
- tmp_adev->reset_vram_lost = vram_lost;
- memset(&tmp_adev->reset_task_info, 0,
- sizeof(tmp_adev->reset_task_info));
- if (reset_context->job && reset_context->job->vm)
- tmp_adev->reset_task_info =
- reset_context->job->vm->task_info;
- amdgpu_reset_capture_coredumpm(tmp_adev);
-#endif
+
+ amdgpu_coredump(tmp_adev, vram_lost, reset_context);
+
if (vram_lost) {
DRM_INFO("VRAM is lost due to GPU reset!\n");
amdgpu_inc_vram_lost(tmp_adev);