<html><head> <meta http-equiv="Content-Type" content="text/html; charset=utf-8"> </head> <body> <div class="moz-cite-prefix">On 8/20/2024 9:31 PM, Alex Deucher wrote: </div> <blockquote type="cite" cite="mid:CADnq5_P4XQ_v_fJP=p8-5dVTBNC5UxUfEU2gYkwkbzQbqte62g@mail.gmail.com"> <pre class="moz-quote-pre" wrap="">On Tue, Aug 20, 2024 at 11:31 AM Khatri, Sunil <a class="moz-txt-link-rfc2396E" href="mailto:sunil.khatri@amd.com"><sunil.khatri@amd.com></a> wrote: </pre> <blockquote type="cite"> <pre class="moz-quote-pre" wrap=""> On 8/20/2024 1:00 PM, Huang, Trigger wrote: </pre> <blockquote type="cite"> <pre class="moz-quote-pre" wrap="">[AMD Official Use Only - AMD Internal Distribution Only] </pre> <blockquote type="cite"> <pre class="moz-quote-pre" wrap="">-----Original Message----- From: Khatri, Sunil <a class="moz-txt-link-rfc2396E" href="mailto:Sunil.Khatri@amd.com"><Sunil.Khatri@amd.com></a> Sent: Monday, August 19, 2024 6:31 PM To: Huang, Trigger <a class="moz-txt-link-rfc2396E" href="mailto:Trigger.Huang@amd.com"><Trigger.Huang@amd.com></a>; <a class="moz-txt-link-abbreviated" href="mailto:amd-gfx@lists.freedesktop.org">amd-gfx@lists.freedesktop.org</a> Cc: Deucher, Alexander <a class="moz-txt-link-rfc2396E" href="mailto:Alexander.Deucher@amd.com"><Alexander.Deucher@amd.com></a> Subject: Re: [PATCH 2/2] drm/amdgpu: Do core dump immediately when job tmo On 8/19/2024 3:23 PM, <a class="moz-txt-link-abbreviated" href="mailto:Trigger.Huang@amd.com">Trigger.Huang@amd.com</a> wrote: </pre> <blockquote type="cite"> <pre class="moz-quote-pre" wrap="">From: Trigger Huang <a class="moz-txt-link-rfc2396E" href="mailto:Trigger.Huang@amd.com"><Trigger.Huang@amd.com></a> Do the coredump immediately after a job timeout to get a closer representation of GPU's error status. V2: This will skip printing vram_lost as the GPU reset is not happened yet (Alex) V3: Unconditionally call the core dump as we care about all the reset functions(soft-recovery and queue reset and full adapter reset, Alex) Signed-off-by: Trigger Huang <a class="moz-txt-link-rfc2396E" href="mailto:Trigger.Huang@amd.com"><Trigger.Huang@amd.com></a> --- drivers/gpu/drm/amd/amdgpu/amdgpu_job.c | 62 </pre> </blockquote> <pre class="moz-quote-pre" wrap="">+++++++++++++++++++++++++ </pre> <blockquote type="cite"> <pre class="moz-quote-pre" wrap=""> 1 file changed, 62 insertions(+) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c index c6a1783fc9ef..ebbb1434073e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c @@ -30,6 +30,61 @@ #include "amdgpu.h" #include "amdgpu_trace.h" #include "amdgpu_reset.h" +#include "amdgpu_dev_coredump.h" +#include "amdgpu_xgmi.h" + +static void amdgpu_job_do_core_dump(struct amdgpu_device *adev, + struct amdgpu_job *job) +{ + int i; + + dev_info(adev->dev, "Dumping IP State\n"); + for (i = 0; i < adev->num_ip_blocks; i++) { + if (adev->ip_blocks[i].version->funcs->dump_ip_state) + adev->ip_blocks[i].version->funcs + ->dump_ip_state((void *)adev); + dev_info(adev->dev, "Dumping IP State Completed\n"); + } + + amdgpu_coredump(adev, true, false, job); } + +static void amdgpu_job_core_dump(struct amdgpu_device *adev, + struct amdgpu_job *job) +{ + struct list_head device_list, *device_list_handle = NULL; + struct amdgpu_device *tmp_adev = NULL; + struct amdgpu_hive_info *hive = NULL; + + if (!amdgpu_sriov_vf(adev)) + hive = amdgpu_get_xgmi_hive(adev); + if (hive) + mutex_lock(&hive->hive_lock); + /* + * Reuse the logic in amdgpu_device_gpu_recover() to build list of + * devices for code dump + */ + INIT_LIST_HEAD(&device_list); + if (!amdgpu_sriov_vf(adev) && (adev- gmc.xgmi.num_physical_nodes > 1) && hive) { + list_for_each_entry(tmp_adev, &hive->device_list, </pre> </blockquote> <pre class="moz-quote-pre" wrap="">gmc.xgmi.head) </pre> <blockquote type="cite"> <pre class="moz-quote-pre" wrap="">+ list_add_tail(&tmp_adev->reset_list, &device_list); + if (!list_is_first(&adev->reset_list, &device_list)) + list_rotate_to_front(&adev->reset_list, &device_list); + device_list_handle = &device_list; + } else { + list_add_tail(&adev->reset_list, &device_list); + device_list_handle = &device_list; + } + + /* Do the coredump for each device */ + list_for_each_entry(tmp_adev, device_list_handle, reset_list) + amdgpu_job_do_core_dump(tmp_adev, job); + + if (hive) { + mutex_unlock(&hive->hive_lock); + amdgpu_put_xgmi_hive(hive); + } +} static enum drm_gpu_sched_stat amdgpu_job_timedout(struct </pre> </blockquote> <pre class="moz-quote-pre" wrap="">drm_sched_job *s_job) </pre> <blockquote type="cite"> <pre class="moz-quote-pre" wrap=""> { @@ -48,6 +103,7 @@ static enum drm_gpu_sched_stat </pre> </blockquote> <pre class="moz-quote-pre" wrap="">amdgpu_job_timedout(struct drm_sched_job *s_job) </pre> <blockquote type="cite"> <pre class="moz-quote-pre" wrap=""> return DRM_GPU_SCHED_STAT_ENODEV; } + amdgpu_job_core_dump(adev, job);</pre> </blockquote> </blockquote> </blockquote> </blockquote> </blockquote> Let's call this after adev->job_hang = true; regards Sunil <blockquote type="cite" cite="mid:CADnq5_P4XQ_v_fJP=p8-5dVTBNC5UxUfEU2gYkwkbzQbqte62g@mail.gmail.com"> <blockquote type="cite"> <blockquote type="cite"> <blockquote type="cite"> <blockquote type="cite"> <pre class="moz-quote-pre" wrap=""> </pre> </blockquote> <pre class="moz-quote-pre" wrap="">The philosophy is hang and recovery is to let the HW and software try to recover. Here we try to do a soft recovery first and i think we should wait for seft recovery and if fails then we do dump and thats exactly we are doing here. </pre> </blockquote> <pre class="moz-quote-pre" wrap="">Hi Sunil , thanks for the suggestion, and that's reasonable. But my concern is that after soft recovery happened, the GPU's status may change(take gfx 9 for example, it will try to kill the current hang wave) Actually, in most cases, a real shader hang cannot be resolved through soft recovery, and at that moment, we need to get a very close dump/snapshot/representation of GPU's current error status. Just like the scandump, when we trying to do a scandump for a shader hang, we will disable gpu_recovery, and no soft recovery/per-queue reset/HW reset will happen before the scandump, right? On most products, there are no scandump interfaces, so core dump is even more important for debugging GPU hang issue. Regards, Trigger </pre> <blockquote type="cite"> <pre class="moz-quote-pre" wrap="">Also we need to make sure that the tasks which are already in queue are put on hold and the their sync points are signalled before we dump. check once what all steps are taken before we dump in the current implementation. </pre> </blockquote> <pre class="moz-quote-pre" wrap="">Do you mean sometimes like: drm_sched_wqueue_stop(&ring->sched); amdgpu_fence_driver_force_completion(ring); // Since there is no GPU reset happened, is it reasonable to call it here? amdgpu_job_core_dump(adev, job); </pre> </blockquote> <pre class="moz-quote-pre" wrap="">Before hard reset we do all those things. But i guess we may not need that in case we capturing before soft reset itself but you need to test it, to see the dump values are true or not. Also apart from hardware state we dump a lot of other information like ring buffers and in case jobs are still submitting we might be not able to get the right data as the ring might be in use and being consumed/filled up that time and that's why scheduler stop helps. But in case soft reset is successful we do not want to do that. So here is what i think but Alex please suggest if it make sense. If recovery is disabled : Capture ip dump before soft reset. (Give close register state but ring buffer need to be seen as it is in use as scheduler is running) if recovery is enabled : capture ip dump (Current implementation make sure to disable drm sched and fence time out) </pre> </blockquote> <pre class="moz-quote-pre" wrap=""> I think Trigger's proposed logic makes sense. I don't see a reason not to dump earlier if we can. I don't really see what it buys us to have different behavior depending on whether or not recovery is enabled. If per queue resets are successful, then we'd miss the dump in that case. If we only dump in job_timedout, then we'd miss the dump when there is a reset due to something the kernel driver or KFD has done.</pre> </blockquote> I am totally in for dumping the ip registers as early as possible and its a good idea to dump before soft reset. I am little worried if we need to stop scheduling any further jobs and process the ones already scheduled. If that's not a problem it looks good to me.  With one minor change above patch looks fine to me.<span style="font-size:12.0pt;font-family:"Aptos",sans-serif; mso-ascii-theme-font:minor-latin;mso-fareast-font-family:Aptos;mso-fareast-theme-font: minor-latin;mso-hansi-theme-font:minor-latin;mso-bidi-font-family:"Times New Roman"; mso-bidi-theme-font:minor-bidi;mso-ansi-language:EN-US;mso-fareast-language: EN-US;mso-bidi-language:AR-SA"> Just make your try it and see the time it takes between "Dumping IP State" and "Dumping IP State Completed" i.e reading registers. <span style="font-size:12.0pt;font-family:"Aptos",sans-serif; mso-ascii-theme-font:minor-latin;mso-fareast-font-family:Aptos;mso-fareast-theme-font: minor-latin;mso-hansi-theme-font:minor-latin;mso-bidi-font-family:"Times New Roman"; mso-bidi-theme-font:minor-bidi;mso-ansi-language:EN-US;mso-fareast-language: EN-US;mso-bidi-language:AR-SA">Acked-by: Sunil Khatri <a class="moz-txt-link-rfc2396E" href="mailto:sunil.khatri@amd.com"><sunil.khatri@amd.com></a> Regards Sunil K <blockquote type="cite" cite="mid:CADnq5_P4XQ_v_fJP=p8-5dVTBNC5UxUfEU2gYkwkbzQbqte62g@mail.gmail.com"> <pre class="moz-quote-pre" wrap="">I think what we want is for job_timedout or the similar logic on the KFD side to call the coredump code when we detect a stuck queue. We'll need to keep the codedump code in gpu_recover to cover the cases where we need to reset due to something outside of the user submission paths. Alex </pre> <blockquote type="cite"> <pre class="moz-quote-pre" wrap=""> function ptr print ip state could be called to capture dump when its needed in both above cases. Right now print is called when dump is actually dumped which is when data file which is generated in devcoredump is read. Regards Sunil Khatri </pre> <blockquote type="cite"> <pre class="moz-quote-pre" wrap="">Regards, Trigger </pre> <blockquote type="cite"> <pre class="moz-quote-pre" wrap="">Regards Sunil khatri </pre> <blockquote type="cite"> <pre class="moz-quote-pre" wrap=""> adev->job_hang = true; @@ -101,6 +157,12 @@ static enum drm_gpu_sched_stat </pre> </blockquote> <pre class="moz-quote-pre" wrap="">amdgpu_job_timedout(struct drm_sched_job *s_job) </pre> <blockquote type="cite"> <pre class="moz-quote-pre" wrap=""> reset_context.src = AMDGPU_RESET_SRC_JOB; clear_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags); + /* + * To avoid an unnecessary extra coredump, as we have </pre> </blockquote> <pre class="moz-quote-pre" wrap="">already </pre> <blockquote type="cite"> <pre class="moz-quote-pre" wrap="">+ * got the very close representation of GPU's error status + */ + set_bit(AMDGPU_SKIP_COREDUMP, &reset_context.flags); + r = amdgpu_device_gpu_recover(ring->adev, job, </pre> </blockquote> <pre class="moz-quote-pre" wrap="">&reset_context); </pre> <blockquote type="cite"> <pre class="moz-quote-pre" wrap=""> if (r) dev_err(adev->dev, "GPU Recovery Failed: %d\n", r); </pre> </blockquote> </blockquote> </blockquote> </blockquote> </blockquote> </body> </html>