<html> <head> <meta content="text/html; charset=utf-8" http-equiv="Content-Type"> </head> <body text="#000000" bgcolor="#FFFFFF"> <div class="moz-cite-prefix">Am 17.05.2017 um 10:46 schrieb zhoucm1: </div> <blockquote cite="mid:591C0DFB.8030604@amd.com" type="cite"> <meta content="text/html; charset=utf-8" http-equiv="Content-Type"> <div class="moz-cite-prefix">On 2017年05月17日 16:40, Christian König wrote: </div> <blockquote cite="mid:7a302ebe-1de1-734f-fb21-aadcc7904d37@vodafone.de" type="cite">Am 17.05.2017 um 10:01 schrieb Michel Dänzer: <blockquote type="cite">On 17/05/17 04:13 PM, zhoucm1 wrote: <blockquote type="cite">On 2017年05月17日 14:57, Michel Dänzer wrote: <blockquote type="cite">On 17/05/17 01:28 PM, zhoucm1 wrote: <blockquote type="cite">On 2017年05月17日 11:15, Michel Dänzer wrote: <blockquote type="cite">On 17/05/17 12:04 PM, zhoucm1 wrote: <blockquote type="cite">On 2017年05月17日 09:18, Michel Dänzer wrote: <blockquote type="cite">On 16/05/17 06:25 PM, Chunming Zhou wrote: <blockquote type="cite">Change-Id: I8eb6d7f558da05510e429d3bf1d48c8cec6c1977 Signed-off-by: Chunming Zhou <a moz-do-not-send="true" class="moz-txt-link-rfc2396E" href="mailto:David1.Zhou@amd.com"><David1.Zhou@amd.com></a> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index bca1fb5..f3e7525 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -2547,6 +2547,9 @@ int amdgpu_vm_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) case AMDGPU_VM_OP_UNRESERVE_VMID: amdgpu_vm_free_reserved_vmid(adev, &fpriv->vm, AMDGPU_GFXHUB); break; + case AMDGPU_VM_OP_RESET: + fpriv->vram_lost_counter = atomic_read(&adev->vram_lost_counter); + break; </blockquote> How do you envision the UMDs using this? I can mostly think of them calling this ioctl when a context is created or destroyed. But that would also allow any other remaining contexts using the same DRM file descriptor to use all ioctls again. So, I think there needs to be a vram_lost_counter in struct amdgpu_ctx instead of in struct amdgpu_fpriv. </blockquote> struct amdgpu_fpriv for vram_lost_counter is proper place, especially for ioctl return value. if you need to reset ctx one by one, we can mark all contexts of that vm, and then reset by userspace. </blockquote> I'm not following. With vram_lost_counter in amdgpu_fpriv, if any context calls this ioctl, all other contexts using the same file descriptor will also be considered safe again, right? </blockquote> Yes, but it really depends on userspace requirement, if you need to reset ctx one by one, we can mark all contexts of that vm to guilty, and then reset one context by userspace. </blockquote> Still not sure what you mean by that. E.g. what do you mean by "guilty"? I thought that refers to the context which caused a hang. But it seems like you're using it to refer to any context which hasn't reacted yet to VRAM contents being lost. </blockquote> When vram is lost, we treat all contexts need to reset. </blockquote> Essentially, your patches only track VRAM contents being lost per file descriptor, not per context. I'm not sure (rather skeptical) that this is suitable for OpenGL UMDs, since state is usually tracked per context. Marek / Nicolai? </blockquote> Oh, yeah that's a good point. The problem with tracking it per context is that Vulkan also wants the ENODEV on the amdgpu_gem_va_ioct() and amdgpu_info_ioctl() which are context less. But thinking more about this blocking those two doesn't make much sense. The VM content can be restored and why should be disallow reading GPU info? </blockquote> I can re-paste the Vulkan APIs requiring ENODEV: " The Vulkan APIs listed below could return VK_ERROR_DEVICE_LOST according to the spec. <o:p></o:p> I tries to provide a list of u/k interfaces that could be called for each vk API. </blockquote> Well those are the Vulkan requirements, but that doesn't necessary mean we must follow that on the kernel side. Keep in mind that Vulkan can't made any requirements towards the kernel driver. IIRC we already have a query Vulkan can use to figure out if a GPU reset happened or not. So they could use that instead. Regards, Christian. <blockquote cite="mid:591C0DFB.8030604@amd.com" type="cite"> <o:p></o:p> <o:p> </o:p> vkCreateDevice<o:p></o:p> - amdgpu_device_initialize.<o:p></o:p> - amdgpu_query_gpu_info<o:p></o:p> <o:p> </o:p> vkQueueSubmit<o:p></o:p> - amdgpu_cs_submit<o:p></o:p> <o:p> </o:p> vkWaitForFences<o:p></o:p> amdgpu_cs_wait_fences<o:p></o:p> <o:p> </o:p> vkGetEventStatus<o:p></o:p> vkQueueWaitIdle<o:p></o:p> vkDeviceWaitIdle<o:p></o:p> vkGetQueryPoolResults<o:p></o:p> amdgpu_cs_query_Fence_status<o:p></o:p> <o:p> </o:p> vkQueueBindSparse<o:p></o:p> amdgpu_bo_va_op<o:p></o:p> amdgpu_bo_va_op_raw<o:p></o:p> <o:p> </o:p> vkCreateSwapchainKHR<o:p></o:p> vkAcquireNextImageKHR<o:p></o:p> vkQueuePresentKHR<o:p></o:p> Not related with u/k interface.<o:p></o:p> <o:p> </o:p> Besides those listed above, I think amdgpu_cs_signal_Sem/amdgpu_cs_wait_sem should respond to gpu reset as well." <blockquote cite="mid:7a302ebe-1de1-734f-fb21-aadcc7904d37@vodafone.de" type="cite"> Christian. </blockquote> <fieldset class="mimeAttachmentHeader"></fieldset> <pre wrap="">_______________________________________________ amd-gfx mailing list <a class="moz-txt-link-abbreviated" href="mailto:amd-gfx@lists.freedesktop.org">amd-gfx@lists.freedesktop.org</a> <a class="moz-txt-link-freetext" href="https://lists.freedesktop.org/mailman/listinfo/amd-gfx">https://lists.freedesktop.org/mailman/listinfo/amd-gfx</a> </pre> </blockquote> </body> </html>