<html><head> <meta http-equiv="Content-Type" content="text/html; charset=utf-8"> </head> <body> <div class="moz-cite-prefix">On 2024-09-11 02:54, Christian König wrote: </div> <blockquote type="cite" cite="mid:a0c252aa-e232-4d61-8dad-3782fe864ba5@amd.com"> Yeah, I completely agree with Xiaogang. The PASID is an identifier of an address space. And the idea of the KFD was that we can just use the same address space and with it the page tables for multiple execution devices, e.g. CPUs, GPUs etc... That idea turned out to be a bad one because it clashes with some use cases (e.g. native context virtualization). The better approach is to see the CPU and GPU processes as separate things which just share the same underlying data. Opening the KFD node multiple times currently results in the same KFD process being used. We should probably consider changing that. </blockquote> It is one KFD process binding to one app process, with count to support multiple open/close of for the same process. The IOMMU most likely uses Linux process pid, not from kfd process->pasid. The KFD process->pasid is passed to F/W to map queues, flush TLB. The reason to replace vm->pasid with KFD process->pasid is to find vm from fault pasid, then for compute vm, find the kfd process from pasid. I can see a bug in amdgpu_vm_handle_fault, only for compute vm, to force update PTE no-retry-fault to the incorrect VM for multiple partitions. This patch will fix this bug but we can have a simple fix. Regards, Philip <blockquote type="cite" cite="mid:a0c252aa-e232-4d61-8dad-3782fe864ba5@amd.com"> Regards, Christian. <div class="moz-cite-prefix">Am 11.09.24 um 01:59 schrieb Chen, Xiaogang: </div> <blockquote type="cite" cite="mid:be8eb0ef-2368-47c2-b4bb-d24f8cb0165d@amd.com"> You want have 1:1 mapping between vm and pasid so can query vm from pasid. I think there is a basic existing issue that we cannot have vm and pasid 1:1 correspondence. PASIDs are global address space identifiers that can be shared between the GPU, an IOMMU and the driver. One app should have one pasid that iommu uses to decide which page table to use when device access system resource. But one app can open render/kfd node multiple times even for one gpu. That said one app could have multiple GPU vms . I think we did not have this issue because app usually open a rent node or kfd node only once. With one adev has multiple partitions there are multiple vms on one adev, so have this issue. But the root cause is not from multiple partitions and solution is not to introduce multiple pasids. I think we should have one pasid for one app and use different way to get vm from pasid. Regards Xiaogang <div class="moz-cite-prefix">On 9/10/2024 3:47 PM, Kim, Jonathan wrote: </div> <blockquote type="cite" cite="mid:CY8PR12MB743552BE87CE07964CC89558859A2@CY8PR12MB7435.namprd12.prod.outlook.com"> <meta name="Generator" content="Microsoft Word 15 (filtered medium)"> <style>@font-face {font-family:"Cambria Math"; panose-1:2 4 5 3 5 4 6 3 2 4;}@font-face {font-family:Calibri; panose-1:2 15 5 2 2 2 4 3 2 4;}@font-face {font-family:"Malgun Gothic"; panose-1:2 11 5 3 2 0 0 2 0 4;}@font-face {font-family:Aptos;}@font-face {font-family:"\@Malgun Gothic";}p.MsoNormal, li.MsoNormal, div.MsoNormal {margin:0in; font-size:12.0pt; font-family:"Aptos",sans-serif;}a:link, span.MsoHyperlink {mso-style-priority:99; color:blue; text-decoration:underline;}span.EmailStyle19 {mso-style-type:personal-reply; font-family:"Arial",sans-serif; color:windowtext;}.MsoChpDefault {mso-style-type:export-only; font-size:10.0pt; mso-ligatures:none;}div.WordSection1 {page:WordSection1;}</style> [Public] <div> <table width="100%" cellspacing="0" cellpadding="0" border="0" align="left"> <tbody> <tr>  <td style="background:#ffb900;padding:5pt 2pt 5pt 2pt"> </td> <td cellpadding="7px 6px 7px 15px" style="background:#fff8e5;padding:5pt 4pt 5pt 12pt;word-wrap:break-word" width="100%"> <div style="color:#222222;">Caution: This message originated from an External Source. Use proper caution when opening attachments, clicking links, or responding. </div> </td> </tr> </tbody> </table> <div> [Public] <div> <div class="WordSection1"> KMS open still set per pasid-vm bindings per adev (socket) so I don’t see how the per-partition overwrite PASID issue is primarily a KFD concern.<o:p></o:p> Are you saying the KFD process devices holds a shadow copy of the correct VM during page restore during fault?<o:p></o:p> Doesn’t it acquire the wrong VM object on process init in the first place?<o:p></o:p> Even if it were the case the KFD had a separate VM reference, the underlying IRQ fault handling is still broken.<o:p></o:p> We probably don’t want to bandage over something to fix one symptom.<o:p></o:p> <o:p> </o:p> Jon<o:p></o:p> <o:p> </o:p> <o:p> </o:p> <div style="border:none;border-left:solid blue 1.5pt;padding:0in 0in 0in 4.0pt"> <div> <div style="border:none;border-top:solid #E1E1E1 1.0pt;padding:3.0pt 0in 0in 0in"> From: Yang, Philip <a class="moz-txt-link-rfc2396E" href="mailto:Philip.Yang@amd.com" moz-do-not-send="true"><Philip.Yang@amd.com></a> Sent: Tuesday, September 10, 2024 11:24 AM To: Koenig, Christian <a class="moz-txt-link-rfc2396E" href="mailto:Christian.Koenig@amd.com" moz-do-not-send="true"><Christian.Koenig@amd.com></a>; Kim, Jonathan <a class="moz-txt-link-rfc2396E" href="mailto:Jonathan.Kim@amd.com" moz-do-not-send="true"><Jonathan.Kim@amd.com></a>; <a class="moz-txt-link-abbreviated moz-txt-link-freetext" href="mailto:amd-gfx@lists.freedesktop.org" moz-do-not-send="true">amd-gfx@lists.freedesktop.org</a> Cc: Kuehling, Felix <a class="moz-txt-link-rfc2396E" href="mailto:Felix.Kuehling@amd.com" moz-do-not-send="true"><Felix.Kuehling@amd.com></a>; Deucher, Alexander <a class="moz-txt-link-rfc2396E" href="mailto:Alexander.Deucher@amd.com" moz-do-not-send="true"><Alexander.Deucher@amd.com></a>; Joshi, Mukul <a class="moz-txt-link-rfc2396E" href="mailto:Mukul.Joshi@amd.com" moz-do-not-send="true"><Mukul.Joshi@amd.com></a> Subject: Re: [PATCH] drm/amdkfd: fix vm-pasid lookup for multiple partitions<o:p></o:p> </div> </div> <o:p> </o:p> <o:p> </o:p> <div> On 2024-09-09 14:46, Christian König wrote:<o:p></o:p> </div> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> Am 09.09.24 um 18:02 schrieb Kim, Jonathan: <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> [Public] <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> -----Original Message----- From: Christian König <a href="mailto:ckoenig.leichtzumerken@gmail.com" moz-do-not-send="true"><ckoenig.leichtzumerken@gmail.com></a> Sent: Thursday, September 5, 2024 10:24 AM To: Kim, Jonathan <a href="mailto:Jonathan.Kim@amd.com" moz-do-not-send="true"><Jonathan.Kim@amd.com></a>; <a href="mailto:amd-gfx@lists.freedesktop.org" moz-do-not-send="true" class="moz-txt-link-freetext">amd-gfx@lists.freedesktop.org</a> Cc: Kuehling, Felix <a href="mailto:Felix.Kuehling@amd.com" moz-do-not-send="true"><Felix.Kuehling@amd.com></a>; Deucher, Alexander <a href="mailto:Alexander.Deucher@amd.com" moz-do-not-send="true"><Alexander.Deucher@amd.com></a>; Joshi, Mukul <a href="mailto:Mukul.Joshi@amd.com" moz-do-not-send="true"><Mukul.Joshi@amd.com></a> Subject: Re: [PATCH] drm/amdkfd: fix vm-pasid lookup for multiple partitions Caution: This message originated from an External Source. Use proper caution when opening attachments, clicking links, or responding. Am 19.08.24 um 19:59 schrieb Jonathan Kim: <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> Currently multiple partitions will incorrectly overwrite the VM lookup table since the table is indexed by PASID and multiple partitions can register different VM objects on the same PASID. <o:p></o:p> </blockquote> That's a rather bad idea. Why do we have the same PASID for different VM objects in the first place? <o:p></o:p> </blockquote> Alex can probably elaborate on the KGD side, but from what I can see, the KMS driver open call has always assigned a new VM object per PASID on an open call. The KFD acquires and replaces the KGD PASID-VMID registration on its own compute process open/creation call. If this is the bad_idea you're referring to, then someone else will have to chime in. I don't have much history on this unfortunately. <o:p></o:p> </blockquote> Yeah, Felix and I designed that. <o:p></o:p> </blockquote> app opens drm node to create vm for each partition, with different vm->pasid for each vm, issue is from kfd_ioctl_acquire_vm -> kfd_process_device_init_vm -> amdgpu_amdkfd_gpuvm_set_vm_pasid, to replace all vm->pasid with kfd process->pasid, which is from open kfd node. This ends up to store only one vm to adev->vm_manager.pasids with KFD process pasid, so we cannot retrieve correct vm from adev->vm_manager.pasids on mGPUs or multiple partitions.<o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> That aside, the current problem is, is that all KFD device structures are logical partitions and register their PASID-VM binding using this concept of a device. <o:p></o:p> </blockquote> As far as I can see that is the fundamental problem. This needs to be fixed instead. <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> On the KGD side however, the registration table is maintained in the adev struct, which is a physical socket. So there's a mismatch in understanding of what a device is between the KFD & KGD with regard to the look up table that results in bad bindings. Adding a per-partition dimension to the existing lookup table resolves issues where seeing, for example, with memory violation interception and XNACK i.e bad bindings result in wrong vm object found to set no-retry flags on memory violations. <o:p></o:p> </blockquote> </blockquote> svm_range_restore_pages retry fault recover uses fault pasid to get kfd process, and use the fault node_id to get pdd->vm, maybe you can use this way to fix the debugger issue.<o:p></o:p> Regards,<o:p></o:p> Philip<o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> Yeah that is pretty much a no-go. The PASID and how it is used is defined by the PCIe specifications. If we now start to assign multiple VMs to the same PASID then we are violating the PCIe specification. The problems you see are most likely just the tip of the iceberg here. Regards, Christian. <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> Jon <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> Regards, Christian. <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> This results in loading the wrong VM object on PASID query. To correct this, setup the lookup table to be per-partition-per-PASID instead. Signed-off-by: Jonathan Kim <a href="mailto:jonathan.kim@amd.com" moz-do-not-send="true"><jonathan.kim@amd.com></a> --- drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c | 12 ++++ drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_job.c | 4 +- drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c | 7 ++- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 55 +++++++++++------ <o:p></o:p> </blockquote> -- <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 11 +++- drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c | 5 +- drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c | 5 +- drivers/gpu/drm/amd/amdgpu/gmc_v12_0.c | 5 +- drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c | 3 +- drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c | 5 +- drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 16 ++---- drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c | 2 +- drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c | 4 +- drivers/gpu/drm/amd/amdkfd/kfd_events.c | 3 +- .../gpu/drm/amd/amdkfd/kfd_int_process_v10.c | 8 +-- .../gpu/drm/amd/amdkfd/kfd_int_process_v9.c | 8 +-- drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c | 3 +- 18 files changed, 92 insertions(+), 65 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c <o:p></o:p> </blockquote> b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> index c272461d70a9..28db789610e1 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c @@ -887,3 +887,15 @@ int amdgpu_amdkfd_unmap_hiq(struct <o:p></o:p> </blockquote> amdgpu_device *adev, u32 doorbell_off, <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> return r; } + +int amdgpu_amdkfd_node_id_to_xcc_id(struct amdgpu_device *adev, <o:p></o:p> </blockquote> uint32_t node_id) <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> +{ + if (adev->gfx.funcs->ih_node_to_logical_xcc) { + int xcc_id = adev->gfx.funcs->ih_node_to_logical_xcc(adev, node_id); + + if (xcc_id >= 0) + return xcc_id; + } + + return 0; +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h <o:p></o:p> </blockquote> b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> index 4ed49265c764..bf8bb45d8ab6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h @@ -356,6 +356,7 @@ void amdgpu_amdkfd_unreserve_mem_limit(struct <o:p></o:p> </blockquote> amdgpu_device *adev, <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> uint64_t size, u32 alloc_flag, int8_t xcp_id); u64 amdgpu_amdkfd_xcp_memory_size(struct amdgpu_device *adev, int <o:p></o:p> </blockquote> xcp_id); <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> +int amdgpu_amdkfd_node_id_to_xcc_id(struct amdgpu_device *adev, <o:p></o:p> </blockquote> uint32_t node_id); <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> #define KFD_XCP_MEM_ID(adev, xcp_id) \ ((adev)->xcp_mgr && (xcp_id) >= 0 ?\ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c <o:p></o:p> </blockquote> b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> index c6a1783fc9ef..bf9f8802e18d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c @@ -37,7 +37,7 @@ static enum drm_gpu_sched_stat <o:p></o:p> </blockquote> amdgpu_job_timedout(struct drm_sched_job *s_job) <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> struct amdgpu_job *job = to_amdgpu_job(s_job); struct amdgpu_task_info *ti; struct amdgpu_device *adev = ring->adev; - int idx; + int idx, xcp_id = !job->vm ? 0 : job->vm->xcp_id; int r; if (!drm_dev_enter(adev_to_drm(adev), &idx)) { @@ -62,7 +62,7 @@ static enum drm_gpu_sched_stat <o:p></o:p> </blockquote> amdgpu_job_timedout(struct drm_sched_job *s_job) <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> job->base.sched->name, atomic_read(&ring->fence_drv.last_seq), ring->fence_drv.sync_seq); - ti = amdgpu_vm_get_task_info_pasid(ring->adev, job->pasid); + ti = amdgpu_vm_get_task_info_pasid(ring->adev, job->pasid, xcp_id); if (ti) { dev_err(adev->dev, "Process information: process %s pid %d thread %s pid %d\n", diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c <o:p></o:p> </blockquote> b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> index d9fde38f6ee2..e413bf4a3e84 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c @@ -1275,17 +1275,20 @@ int amdgpu_info_ioctl(struct drm_device *dev, <o:p></o:p> </blockquote> void *data, struct drm_file *filp) <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> struct amdgpu_vm *vm = &fpriv->vm; struct drm_amdgpu_info_gpuvm_fault gpuvm_fault; unsigned long flags; + int i; if (!vm) return -EINVAL; memset(&gpuvm_fault, 0, sizeof(gpuvm_fault)); - xa_lock_irqsave(&adev->vm_manager.pasids, flags); + for (i = 0; i < adev->xcp_mgr->num_xcps; i++) + xa_lock_irqsave(&adev->vm_manager.pasids[i], flags); gpuvm_fault.addr = vm->fault_info.addr; gpuvm_fault.status = vm->fault_info.status; gpuvm_fault.vmhub = vm->fault_info.vmhub; - xa_unlock_irqrestore(&adev->vm_manager.pasids, flags); + for (i = 0; i < adev->xcp_mgr->num_xcps; i++) + xa_unlock_irqrestore(&adev->vm_manager.pasids[i], flags); return copy_to_user(out, &gpuvm_fault, min((size_t)size, sizeof(gpuvm_fault))) ? -EFAULT : 0; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c <o:p></o:p> </blockquote> b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> index bcb729094521..f43e1c15f423 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -146,7 +146,7 @@ int amdgpu_vm_set_pasid(struct amdgpu_device <o:p></o:p> </blockquote> *adev, struct amdgpu_vm *vm, <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> return 0; if (vm->pasid) { - r = xa_err(xa_erase_irq(&adev->vm_manager.pasids, vm->pasid)); + r = xa_err(xa_erase_irq(&adev->vm_manager.pasids[vm->xcp_id], <o:p> </o:p> </blockquote> vm->pasid)); <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> if (r < 0) return r; @@ -154,7 +154,7 @@ int amdgpu_vm_set_pasid(struct amdgpu_device <o:p></o:p> </blockquote> *adev, struct amdgpu_vm *vm, <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> } if (pasid) { - r = xa_err(xa_store_irq(&adev->vm_manager.pasids, pasid, vm, + r = xa_err(xa_store_irq(&adev->vm_manager.pasids[vm->xcp_id], <o:p> </o:p> </blockquote> pasid, vm, <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> GFP_KERNEL)); if (r < 0) return r; @@ -2288,14 +2288,14 @@ static void <o:p></o:p> </blockquote> amdgpu_vm_destroy_task_info(struct kref *kref) <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> } static inline struct amdgpu_vm * -amdgpu_vm_get_vm_from_pasid(struct amdgpu_device *adev, u32 pasid) +amdgpu_vm_get_vm_from_pasid(struct amdgpu_device *adev, u32 pasid, <o:p></o:p> </blockquote> u32 xcp_id) <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> { struct amdgpu_vm *vm; unsigned long flags; - xa_lock_irqsave(&adev->vm_manager.pasids, flags); - vm = xa_load(&adev->vm_manager.pasids, pasid); - xa_unlock_irqrestore(&adev->vm_manager.pasids, flags); + xa_lock_irqsave(&adev->vm_manager.pasids[xcp_id], flags); + vm = xa_load(&adev->vm_manager.pasids[xcp_id], pasid); + xa_unlock_irqrestore(&adev->vm_manager.pasids[xcp_id], flags); return vm; } @@ -2343,10 +2343,10 @@ amdgpu_vm_get_task_info_vm(struct <o:p></o:p> </blockquote> amdgpu_vm *vm) <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> * referenced down with amdgpu_vm_put_task_info. */ struct amdgpu_task_info * -amdgpu_vm_get_task_info_pasid(struct amdgpu_device *adev, u32 pasid) +amdgpu_vm_get_task_info_pasid(struct amdgpu_device *adev, u32 pasid, <o:p></o:p> </blockquote> u32 xcp_id) <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> { return amdgpu_vm_get_task_info_vm( - amdgpu_vm_get_vm_from_pasid(adev, pasid)); + amdgpu_vm_get_vm_from_pasid(adev, pasid, xcp_id)); } static int amdgpu_vm_create_task_info(struct amdgpu_vm *vm) @@ -2481,6 +2481,8 @@ int amdgpu_vm_init(struct amdgpu_device <o:p></o:p> </blockquote> *adev, struct amdgpu_vm *vm, <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> amdgpu_bo_unreserve(vm->root.bo); amdgpu_bo_unref(&root_bo); + vm->xcp_id = xcp_id < 0 ? 0 : xcp_id; + return 0; error_free_root: @@ -2695,8 +2697,8 @@ void amdgpu_vm_manager_init(struct <o:p></o:p> </blockquote> amdgpu_device *adev) <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> #else adev->vm_manager.vm_update_mode = 0; #endif - - xa_init_flags(&adev->vm_manager.pasids, XA_FLAGS_LOCK_IRQ); + for (i = 0; i < MAX_XCP; i++) + xa_init_flags(&(adev->vm_manager.pasids[i]), <o:p></o:p> </blockquote> XA_FLAGS_LOCK_IRQ); <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> } /** @@ -2708,10 +2710,15 @@ void amdgpu_vm_manager_init(struct <o:p></o:p> </blockquote> amdgpu_device *adev) <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> */ void amdgpu_vm_manager_fini(struct amdgpu_device *adev) { - WARN_ON(!xa_empty(&adev->vm_manager.pasids)); - xa_destroy(&adev->vm_manager.pasids); + int i; + + for (i = 0; i < MAX_XCP; i++) { + WARN_ON(!xa_empty(&adev->vm_manager.pasids[i])); + xa_destroy(&adev->vm_manager.pasids[i]); + } amdgpu_vmid_mgr_fini(adev); + } /** @@ -2778,17 +2785,18 @@ bool amdgpu_vm_handle_fault(struct <o:p></o:p> </blockquote> amdgpu_device *adev, u32 pasid, <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> unsigned long irqflags; uint64_t value, flags; struct amdgpu_vm *vm; - int r; + int r, xcp_id; - xa_lock_irqsave(&adev->vm_manager.pasids, irqflags); - vm = xa_load(&adev->vm_manager.pasids, pasid); + xcp_id = amdgpu_amdkfd_node_id_to_xcc_id(adev, node_id)/adev- gfx.num_xcc_per_xcp; + xa_lock_irqsave(&adev->vm_manager.pasids[xcp_id], irqflags); + vm = xa_load(&adev->vm_manager.pasids[xcp_id], pasid); if (vm) { root = amdgpu_bo_ref(vm->root.bo); is_compute_context = vm->is_compute_context; } else { root = NULL; } - xa_unlock_irqrestore(&adev->vm_manager.pasids, irqflags); + xa_unlock_irqrestore(&adev->vm_manager.pasids[xcp_id], irqflags); if (!root) return false; @@ -2806,11 +2814,11 @@ bool amdgpu_vm_handle_fault(struct <o:p></o:p> </blockquote> amdgpu_device *adev, u32 pasid, <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> goto error_unref; /* Double check that the VM still exists */ - xa_lock_irqsave(&adev->vm_manager.pasids, irqflags); - vm = xa_load(&adev->vm_manager.pasids, pasid); + xa_lock_irqsave(&adev->vm_manager.pasids[xcp_id], irqflags); + vm = xa_load(&adev->vm_manager.pasids[xcp_id], pasid); if (vm && vm->root.bo != root) vm = NULL; - xa_unlock_irqrestore(&adev->vm_manager.pasids, irqflags); + xa_unlock_irqrestore(&adev->vm_manager.pasids[xcp_id], irqflags); if (!vm) goto error_unlock; @@ -2968,14 +2976,15 @@ void amdgpu_vm_update_fault_cache(struct <o:p></o:p> </blockquote> amdgpu_device *adev, <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> unsigned int pasid, uint64_t addr, uint32_t status, - unsigned int vmhub) + unsigned int vmhub, + uint32_t xcp_id) { struct amdgpu_vm *vm; unsigned long flags; - xa_lock_irqsave(&adev->vm_manager.pasids, flags); + xa_lock_irqsave(&adev->vm_manager.pasids[xcp_id], flags); - vm = xa_load(&adev->vm_manager.pasids, pasid); + vm = xa_load(&adev->vm_manager.pasids[xcp_id], pasid); /* Don't update the fault cache if status is 0. In the multiple * fault case, subsequent faults will return a 0 status which is * useless for userspace and replaces the useful fault status, so @@ -3008,7 +3017,7 @@ void amdgpu_vm_update_fault_cache(struct <o:p></o:p> </blockquote> amdgpu_device *adev, <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> WARN_ONCE(1, "Invalid vmhub %u\n", vmhub); } } - xa_unlock_irqrestore(&adev->vm_manager.pasids, flags); + xa_unlock_irqrestore(&adev->vm_manager.pasids[xcp_id], flags); } /** diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h <o:p></o:p> </blockquote> b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> index 046949c4b695..1499f5f731e9 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h @@ -35,6 +35,7 @@ #include "amdgpu_sync.h" #include "amdgpu_ring.h" #include "amdgpu_ids.h" +#include "amdgpu_xcp.h" struct drm_exec; @@ -418,6 +419,9 @@ struct amdgpu_vm { /* cached fault info */ struct amdgpu_vm_fault_info fault_info; + + /* XCP ID */ + int xcp_id; }; struct amdgpu_vm_manager { @@ -456,7 +460,7 @@ struct amdgpu_vm_manager { /* PASID to VM mapping, will be used in interrupt context to * look up VM of a page fault */ - struct xarray pasids; + struct xarray pasids[MAX_XCP]; /* Global registration of recent page fault information */ struct amdgpu_vm_fault_info fault_info; }; @@ -550,7 +554,7 @@ bool amdgpu_vm_need_pipeline_sync(struct <o:p></o:p> </blockquote> amdgpu_ring *ring, <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> void amdgpu_vm_check_compute_bug(struct amdgpu_device *adev); struct amdgpu_task_info * -amdgpu_vm_get_task_info_pasid(struct amdgpu_device *adev, u32 pasid); +amdgpu_vm_get_task_info_pasid(struct amdgpu_device *adev, u32 pasid, <o:p></o:p> </blockquote> u32 xcp_id); <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> struct amdgpu_task_info * amdgpu_vm_get_task_info_vm(struct amdgpu_vm *vm); @@ -649,7 +653,8 @@ void amdgpu_vm_update_fault_cache(struct <o:p></o:p> </blockquote> amdgpu_device *adev, <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> unsigned int pasid, uint64_t addr, uint32_t status, - unsigned int vmhub); + unsigned int vmhub, + uint32_t xcp_id); void amdgpu_vm_tlb_fence_create(struct amdgpu_device *adev, struct amdgpu_vm *vm, struct dma_fence **fence); diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c <o:p></o:p> </blockquote> b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> index f0ceab3ce5bf..24b042febf5c 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c @@ -151,7 +151,8 @@ static int gmc_v10_0_process_interrupt(struct <o:p></o:p> </blockquote> amdgpu_device *adev, <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> WREG32_P(hub->vm_l2_pro_fault_cntl, 1, ~1); amdgpu_vm_update_fault_cache(adev, entry->pasid, addr, status, - entry->vmid_src ? AMDGPU_MMHUB0(0) : <o:p> </o:p> </blockquote> AMDGPU_GFXHUB(0)); <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> + entry->vmid_src ? AMDGPU_MMHUB0(0) : <o:p></o:p> </blockquote> AMDGPU_GFXHUB(0), <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> + 0); } if (!printk_ratelimit()) @@ -161,7 +162,7 @@ static int gmc_v10_0_process_interrupt(struct <o:p></o:p> </blockquote> amdgpu_device *adev, <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> "[%s] page fault (src_id:%u ring:%u vmid:%u pasid:%u)\n", entry->vmid_src ? "mmhub" : "gfxhub", entry->src_id, entry->ring_id, entry->vmid, entry->pasid); - task_info = amdgpu_vm_get_task_info_pasid(adev, entry->pasid); + task_info = amdgpu_vm_get_task_info_pasid(adev, entry->pasid, 0); if (task_info) { dev_err(adev->dev, " in process %s pid %d thread %s pid %d\n", diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c <o:p></o:p> </blockquote> b/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> index 2797fd84432b..3507046d33e6 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c @@ -122,7 +122,8 @@ static int gmc_v11_0_process_interrupt(struct <o:p></o:p> </blockquote> amdgpu_device *adev, <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> WREG32_P(hub->vm_l2_pro_fault_cntl, 1, ~1); amdgpu_vm_update_fault_cache(adev, entry->pasid, addr, status, - entry->vmid_src ? AMDGPU_MMHUB0(0) : <o:p> </o:p> </blockquote> AMDGPU_GFXHUB(0)); <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> + entry->vmid_src ? AMDGPU_MMHUB0(0) : <o:p></o:p> </blockquote> AMDGPU_GFXHUB(0), <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> + 0); } if (printk_ratelimit()) { @@ -132,7 +133,7 @@ static int gmc_v11_0_process_interrupt(struct <o:p></o:p> </blockquote> amdgpu_device *adev, <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> "[%s] page fault (src_id:%u ring:%u vmid:%u pasid:%u)\n", entry->vmid_src ? "mmhub" : "gfxhub", entry->src_id, entry->ring_id, entry->vmid, entry->pasid); - task_info = amdgpu_vm_get_task_info_pasid(adev, entry->pasid); + task_info = amdgpu_vm_get_task_info_pasid(adev, entry->pasid, 0); if (task_info) { dev_err(adev->dev, " in process %s pid %d thread %s pid %d)\n", diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v12_0.c <o:p></o:p> </blockquote> b/drivers/gpu/drm/amd/amdgpu/gmc_v12_0.c <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> index 60acf676000b..9844564c6c74 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v12_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v12_0.c @@ -115,7 +115,8 @@ static int gmc_v12_0_process_interrupt(struct <o:p></o:p> </blockquote> amdgpu_device *adev, <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> WREG32_P(hub->vm_l2_pro_fault_cntl, 1, ~1); amdgpu_vm_update_fault_cache(adev, entry->pasid, addr, status, - entry->vmid_src ? AMDGPU_MMHUB0(0) : <o:p> </o:p> </blockquote> AMDGPU_GFXHUB(0)); <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> + entry->vmid_src ? AMDGPU_MMHUB0(0) : <o:p></o:p> </blockquote> AMDGPU_GFXHUB(0), <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> + 0); } if (printk_ratelimit()) { @@ -125,7 +126,7 @@ static int gmc_v12_0_process_interrupt(struct <o:p></o:p> </blockquote> amdgpu_device *adev, <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> "[%s] page fault (src_id:%u ring:%u vmid:%u pasid:%u)\n", entry->vmid_src ? "mmhub" : "gfxhub", entry->src_id, entry->ring_id, entry->vmid, entry->pasid); - task_info = amdgpu_vm_get_task_info_pasid(adev, entry->pasid); + task_info = amdgpu_vm_get_task_info_pasid(adev, entry->pasid, 0); if (task_info) { dev_err(adev->dev, " in process %s pid %d thread %s pid %d)\n", diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c <o:p></o:p> </blockquote> b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> index 994432fb57ea..2cdb0cbb7c4d 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c @@ -1268,7 +1268,8 @@ static int gmc_v7_0_process_interrupt(struct <o:p></o:p> </blockquote> amdgpu_device *adev, <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> return 0; amdgpu_vm_update_fault_cache(adev, entry->pasid, - ((u64)addr) << AMDGPU_GPU_PAGE_SHIFT, status, <o:p> </o:p> </blockquote> AMDGPU_GFXHUB(0)); <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> + ((u64)addr) << AMDGPU_GPU_PAGE_SHIFT, status, + AMDGPU_GFXHUB(0), 0); if (amdgpu_vm_fault_stop == AMDGPU_VM_FAULT_STOP_FIRST) gmc_v7_0_set_fault_enable_default(adev, false); diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c <o:p></o:p> </blockquote> b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> index 86488c052f82..6855caeb7f74 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c @@ -1437,7 +1437,8 @@ static int gmc_v8_0_process_interrupt(struct <o:p></o:p> </blockquote> amdgpu_device *adev, <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> return 0; amdgpu_vm_update_fault_cache(adev, entry->pasid, - ((u64)addr) << AMDGPU_GPU_PAGE_SHIFT, status, <o:p> </o:p> </blockquote> AMDGPU_GFXHUB(0)); <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> + ((u64)addr) << AMDGPU_GPU_PAGE_SHIFT, status, + AMDGPU_GFXHUB(0), 0); if (amdgpu_vm_fault_stop == AMDGPU_VM_FAULT_STOP_FIRST) gmc_v8_0_set_fault_enable_default(adev, false); @@ -1448,7 +1449,7 @@ static int gmc_v8_0_process_interrupt(struct <o:p></o:p> </blockquote> amdgpu_device *adev, <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> dev_err(adev->dev, "GPU fault detected: %d 0x%08x\n", entry->src_id, entry->src_data[0]); - task_info = amdgpu_vm_get_task_info_pasid(adev, entry->pasid); + task_info = amdgpu_vm_get_task_info_pasid(adev, entry->pasid, 0); if (task_info) { dev_err(adev->dev, " for process %s pid %d thread %s pid %d\n", task_info->process_name, task_info->tgid, diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c <o:p></o:p> </blockquote> b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> index b73136d390cc..e183e08b2c02 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c @@ -556,10 +556,12 @@ static int gmc_v9_0_process_interrupt(struct <o:p></o:p> </blockquote> amdgpu_device *adev, <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> unsigned int vmhub; u64 addr; uint32_t cam_index = 0; - int ret, xcc_id = 0; - uint32_t node_id; + int ret; + uint32_t node_id, xcc_id, xcp_id; node_id = entry->node_id; + xcc_id = amdgpu_amdkfd_node_id_to_xcc_id(adev, node_id); + xcp_id = xcc_id/adev->gfx.num_xcc_per_xcp; addr = (u64)entry->src_data[0] << 12; addr |= ((u64)entry->src_data[1] & 0xf) << 44; @@ -572,12 +574,6 @@ static int gmc_v9_0_process_interrupt(struct <o:p></o:p> </blockquote> amdgpu_device *adev, <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> vmhub = AMDGPU_MMHUB1(0); } else { hub_name = "gfxhub0"; - if (adev->gfx.funcs->ih_node_to_logical_xcc) { - xcc_id = adev->gfx.funcs->ih_node_to_logical_xcc(adev, - node_id); - if (xcc_id < 0) - xcc_id = 0; - } vmhub = xcc_id; } hub = &adev->vmhub[vmhub]; @@ -631,7 +627,7 @@ static int gmc_v9_0_process_interrupt(struct <o:p></o:p> </blockquote> amdgpu_device *adev, <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> retry_fault ? "retry" : "no-retry", entry->src_id, entry->ring_id, entry->vmid, entry->pasid); - task_info = amdgpu_vm_get_task_info_pasid(adev, entry->pasid); + task_info = amdgpu_vm_get_task_info_pasid(adev, entry->pasid, <o:p></o:p> </blockquote> xcp_id); <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> if (task_info) { dev_err(adev->dev, " for process %s pid %d thread %s pid %d)\n", @@ -675,7 +671,7 @@ static int gmc_v9_0_process_interrupt(struct <o:p></o:p> </blockquote> amdgpu_device *adev, <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> if (!amdgpu_sriov_vf(adev)) WREG32_P(hub->vm_l2_pro_fault_cntl, 1, ~1); - amdgpu_vm_update_fault_cache(adev, entry->pasid, addr, status, <o:p></o:p> </blockquote> vmhub); <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> + amdgpu_vm_update_fault_cache(adev, entry->pasid, addr, status, <o:p></o:p> </blockquote> vmhub, xcp_id); <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> dev_err(adev->dev, "VM_L2_PROTECTION_FAULT_STATUS:0x%08X\n", diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c <o:p></o:p> </blockquote> b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> index 23ef4eb36b40..1ac4224bbe5b 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c @@ -2182,7 +2182,7 @@ static int sdma_v4_0_print_iv_entry(struct <o:p></o:p> </blockquote> amdgpu_device *adev, <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> instance, addr, entry->src_id, entry->ring_id, entry->vmid, entry->pasid); - task_info = amdgpu_vm_get_task_info_pasid(adev, entry->pasid); + task_info = amdgpu_vm_get_task_info_pasid(adev, entry->pasid, 0); if (task_info) { dev_dbg_ratelimited(adev->dev, " for process %s pid %d thread %s pid %d\n", diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c <o:p></o:p> </blockquote> b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> index 57f16c09abfc..c8b5c0302ca7 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c @@ -1683,6 +1683,8 @@ static int sdma_v4_4_2_print_iv_entry(struct <o:p></o:p> </blockquote> amdgpu_device *adev, <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> int instance; struct amdgpu_task_info *task_info; u64 addr; + uint32_t xcc_id = amdgpu_amdkfd_node_id_to_xcc_id(adev, entry- node_id); + uint32_t xcp_id = xcc_id/adev->gfx.num_xcc_per_xcp; instance = sdma_v4_4_2_irq_id_to_seq(adev, entry->client_id); if (instance < 0 || instance >= adev->sdma.num_instances) { @@ -1698,7 +1700,7 @@ static int sdma_v4_4_2_print_iv_entry(struct <o:p></o:p> </blockquote> amdgpu_device *adev, <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> instance, addr, entry->src_id, entry->ring_id, entry->vmid, entry->pasid); - task_info = amdgpu_vm_get_task_info_pasid(adev, entry->pasid); + task_info = amdgpu_vm_get_task_info_pasid(adev, entry->pasid, <o:p></o:p> </blockquote> xcp_id); <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> if (task_info) { dev_dbg_ratelimited(adev->dev, " for process %s pid %d thread %s <o:p> </o:p> </blockquote> pid %d\n", <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> task_info->process_name, task_info->tgid, diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_events.c <o:p></o:p> </blockquote> b/drivers/gpu/drm/amd/amdkfd/kfd_events.c <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> index ea3792249209..c098fbaf0e1c 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_events.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_events.c @@ -1262,8 +1262,9 @@ void kfd_signal_reset_event(struct kfd_node <o:p></o:p> </blockquote> *dev) <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> if (dev->dqm->detect_hang_count) { struct amdgpu_task_info *ti; + uint32_t xcp_id = dev->xcp ? dev->xcp->id : 0; - ti = amdgpu_vm_get_task_info_pasid(dev->adev, p->pasid); + ti = amdgpu_vm_get_task_info_pasid(dev->adev, p->pasid, <o:p> </o:p> </blockquote> xcp_id); <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> if (ti) { dev_err(dev->adev->dev, "Queues reset on process %s tid %d thread %s pid %d\n", diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v10.c <o:p></o:p> </blockquote> b/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v10.c <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> index 8e0d0356e810..d7cbf9525698 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v10.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v10.c @@ -377,12 +377,8 @@ static void event_interrupt_wq_v10(struct <o:p></o:p> </blockquote> kfd_node *dev, <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> struct kfd_hsa_memory_exception_data exception_data; /* gfxhub */ - if (!vmid_type && dev->adev->gfx.funcs->ih_node_to_logical_xcc) { - hub_inst = dev->adev->gfx.funcs->ih_node_to_logical_xcc(dev- adev, - node_id); - if (hub_inst < 0) - hub_inst = 0; - } + if (!vmid_type) + hub_inst = amdgpu_amdkfd_node_id_to_xcc_id(dev->adev, <o:p> </o:p> </blockquote> node_id); <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> /* mmhub */ if (vmid_type && client_id == SOC15_IH_CLIENTID_VMC) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c <o:p></o:p> </blockquote> b/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> index a9c3580be8c9..4708b8c811a5 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c @@ -437,12 +437,8 @@ static void event_interrupt_wq_v9(struct <o:p></o:p> </blockquote> kfd_node *dev, <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> struct kfd_hsa_memory_exception_data exception_data; /* gfxhub */ - if (!vmid_type && dev->adev->gfx.funcs->ih_node_to_logical_xcc) { - hub_inst = dev->adev->gfx.funcs->ih_node_to_logical_xcc(dev- adev, - node_id); - if (hub_inst < 0) - hub_inst = 0; - } + if (!vmid_type) + hub_inst = amdgpu_amdkfd_node_id_to_xcc_id(dev->adev, <o:p> </o:p> </blockquote> node_id); <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> /* mmhub */ if (vmid_type && client_id == SOC15_IH_CLIENTID_VMC) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c <o:p></o:p> </blockquote> b/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> index ea6a8e43bd5b..b5f2f5b1069c 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c @@ -251,8 +251,9 @@ void <o:p></o:p> </blockquote> kfd_smi_event_update_thermal_throttling(struct kfd_node *dev, <o:p></o:p> <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt"> void kfd_smi_event_update_vmfault(struct kfd_node *dev, uint16_t pasid) { struct amdgpu_task_info *task_info; + uint32_t xcp_id = dev->xcp ? dev->xcp->id : 0; - task_info = amdgpu_vm_get_task_info_pasid(dev->adev, pasid); + task_info = amdgpu_vm_get_task_info_pasid(dev->adev, pasid, xcp_id); if (task_info) { /* Report VM faults from user applications, not retry from kernel */ if (task_info->pid) <o:p></o:p> </blockquote> </blockquote> </blockquote> <o:p> </o:p> </blockquote> </div> </div> </div> </div> </div> </blockquote> </blockquote> </blockquote> </body> </html>