<html xmlns:v="urn:schemas-microsoft-com:vml" xmlns:o="urn:schemas-microsoft-com:office:office" xmlns:w="urn:schemas-microsoft-com:office:word" xmlns:m="http://schemas.microsoft.com/office/2004/12/omml" xmlns="http://www.w3.org/TR/REC-html40"> <head> <meta http-equiv="Content-Type" content="text/html; charset=utf-8"> <meta name="Generator" content="Microsoft Word 15 (filtered medium)"> <style></style> </head> <body lang="EN-US" link="blue" vlink="purple"> <div class="WordSection1"> Thanks Alex<o:p></o:p> <o:p> </o:p> _____________________________________<o:p></o:p> Monk Liu|GPU Virtualization Team |AMD<o:p></o:p> <img width="80" height="80" style="width:.8333in;height:.8333in" id="Picture_x0020_1" src="cid:image001.png@01D5A443.CD3BA560" alt="sig-cloud-gpu"><o:p></o:p> <o:p> </o:p> From: Alex Deucher <alexdeucher@gmail.com> Sent: Tuesday, November 26, 2019 4:34 AM To: Liu, Monk <Monk.Liu@amd.com> Cc: Koenig, Christian <Christian.Koenig@amd.com>; amd-gfx@lists.freedesktop.org Subject: Re: why we unreserve MQD of KIQ ?<o:p></o:p> <o:p> </o:p> <div> <div> <o:p> </o:p> </div> <o:p> </o:p> <div> <div> On Mon, Nov 25, 2019 at 4:05 AM Liu, Monk <<a href="mailto:Monk.Liu@amd.com">Monk.Liu@amd.com</a>> wrote:<o:p></o:p> </div> <blockquote style="border:none;border-left:solid #CCCCCC 1.0pt;padding:0in 0in 0in 6.0pt;margin-left:4.8pt;margin-right:0in"> <div> <div> commit a9a8a788e5e946a9835a1365256fc4ce9e96ba2c<o:p></o:p> Author: Rex Zhu <<a href="mailto:Rex.Zhu@amd.com" target="_blank">Rex.Zhu@amd.com</a>><o:p></o:p> Date: Wed Aug 22 18:54:45 2018 +0800<o:p></o:p> <o:p></o:p> drm/amdgpu: Change kiq ring initialize sequence on gfx9<o:p></o:p> <o:p></o:p> 1. initialize kiq before initialize gfx ring.<o:p></o:p> 2. set kiq ring ready immediately when kiq initialize<o:p></o:p> successfully.<o:p></o:p> 3. split function gfx_v9_0_kiq_resume into two functions.<o:p></o:p> gfx_v9_0_kiq_resume is for kiq initialize.<o:p></o:p> gfx_v9_0_kcq_resume is for kcq initialize.<o:p></o:p> <o:p></o:p> Reviewed-by: Alex Deucher <<a href="mailto:alexander.deucher@amd.com" target="_blank">alexander.deucher@amd.com</a>><o:p></o:p> Signed-off-by: Rex Zhu <<a href="mailto:Rex.Zhu@amd.com" target="_blank">Rex.Zhu@amd.com</a>><o:p></o:p> Signed-off-by: Alex Deucher <<a href="mailto:alexander.deucher@amd.com" target="_blank">alexander.deucher@amd.com</a>><o:p></o:p> <o:p></o:p> diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c<o:p></o:p> index 21e66f8..3594704a 100644<o:p></o:p> --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c<o:p></o:p> +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c<o:p></o:p> @@ -2684,7 +2684,6 @@ static int gfx_v9_0_kiq_kcq_enable(struct amdgpu_device *adev)<o:p></o:p> queue_mask |= (1ull << i);<o:p></o:p> }<o:p></o:p> <o:p></o:p> - kiq_ring->ready = true;<o:p></o:p> r = amdgpu_ring_alloc(kiq_ring, (7 * adev->gfx.num_compute_rings) + 8);<o:p></o:p> if (r) {<o:p></o:p> DRM_ERROR("Failed to lock KIQ (%d).\n", r);<o:p></o:p> @@ -3091,26 +3090,33 @@ static int gfx_v9_0_kcq_init_queue(struct amdgpu_ring *ring)<o:p></o:p> <o:p></o:p> static int gfx_v9_0_kiq_resume(struct amdgpu_device *adev)<o:p></o:p> {<o:p></o:p> - struct amdgpu_ring *ring = NULL;<o:p></o:p> - int r = 0, i;<o:p></o:p> -<o:p></o:p> - gfx_v9_0_cp_compute_enable(adev, true);<o:p></o:p> + struct amdgpu_ring *ring;<o:p></o:p> + int r;<o:p></o:p> <o:p></o:p> ring = &adev->gfx.kiq.ring;<o:p></o:p> <o:p></o:p> r = amdgpu_bo_reserve(ring->mqd_obj, false);<o:p></o:p> if (unlikely(r != 0))<o:p></o:p> - goto done;<o:p></o:p> + return r;<o:p></o:p> <o:p></o:p> r = amdgpu_bo_kmap(ring->mqd_obj, (void **)&ring->mqd_ptr);<o:p></o:p> - if (!r) {<o:p></o:p> - r = gfx_v9_0_kiq_init_queue(ring);<o:p></o:p> - amdgpu_bo_kunmap(ring->mqd_obj);<o:p></o:p> - ring->mqd_ptr = NULL;<o:p></o:p> - }<o:p></o:p> + if (unlikely(r != 0))<o:p></o:p> + return r;<o:p></o:p> +<o:p></o:p> + gfx_v9_0_kiq_init_queue(ring);<o:p></o:p> + amdgpu_bo_kunmap(ring->mqd_obj); //not invoked before change<o:p></o:p> + ring->mqd_ptr = NULL; //not invoked before change<o:p></o:p> amdgpu_bo_unreserve(ring->mqd_obj); //not invoked before change<o:p></o:p> - if (r)<o:p></o:p> - goto done;<o:p></o:p> + ring->ready = true;<o:p></o:p> + return 0;<o:p></o:p> +}<o:p></o:p> +<o:p></o:p> +static int gfx_v9_0_kcq_resume(struct amdgpu_device *adev)<o:p></o:p> +{<o:p></o:p> <o:p></o:p> Hi guys<o:p></o:p> <o:p></o:p> I found in last year, Rex submitted a change which introduced additional “amdgpu_bo_kunmap() and amdgpu_bo_unreserve()” <o:p></o:p> During the kiq_resume() routine, <o:p></o:p> <o:p></o:p> I’m wondering why we need this change and I’m also suspecting it has potential regression:<o:p></o:p> See that the KIQ’s mqd object is unreserved, so it now available in LRU which means TTM can evict it by will, <o:p></o:p> <o:p></o:p> But the MQD of KIQ is supposed to be pinned otherwise incorrect memory access would introduced by a world switch <o:p></o:p> Since MEC always consider KIQ’ MQD is pinned <o:p></o:p> <o:p></o:p> <a href="mailto:Christian.Koenig@amd.com" target="_blank">@Koenig, Christian</a> you know about that change’s background ?<o:p></o:p> <o:p></o:p> </div> </div> </blockquote> <div> <o:p> </o:p> </div> <div> I think the unreserve here is just to match the reserve at the top of this function for the kmapping. It should still be pinned because we call amdgpu_bo_create_kernel() to allocate it.<o:p></o:p> </div> <div> <o:p> </o:p> </div> <div> Alex<o:p></o:p> </div> <div> <o:p></o:p> </div> <blockquote style="border:none;border-left:solid #CCCCCC 1.0pt;padding:0in 0in 0in 6.0pt;margin-left:4.8pt;margin-right:0in"> <div> <div> _____________________________________<o:p></o:p> Monk Liu|GPU Virtualization Team |AMD<o:p></o:p> <img border="0" width="80" height="80" style="width:.8333in;height:.8333in" id="gmail-m_-8449523087627823094Picture_x0020_1" src="cid:image001.png@01D5A443.CD3BA560" alt="sig-cloud-gpu"><o:p></o:p> <o:p></o:p> </div> </div> _______________________________________________ amd-gfx mailing list <a href="mailto:amd-gfx@lists.freedesktop.org" target="_blank">amd-gfx@lists.freedesktop.org</a> <a href="https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flists.freedesktop.org%2Fmailman%2Flistinfo%2Famd-gfx&data=02%7C01%7CMonk.Liu%40amd.com%7C89db24b863e343d560d708d771e6e595%7C3dd8961fe4884e608e11a82d994e183d%7C0%7C0%7C637103108837716631&sdata=AvkmJmyFovEOff7%2FBowMktww%2BltLEfYlgBUwmnPHM2I%3D&reserved=0" target="_blank">https://lists.freedesktop.org/mailman/listinfo/amd-gfx</a><o:p></o:p> </blockquote> </div> </div> </div> </body> </html>