From 3e70c5d6ea510e38f612d07fa0fd7487277b7087 Mon Sep 17 00:00:00 2001 From: Nicolas Iooss Date: Mon, 26 Dec 2016 14:52:23 +0100 Subject: [PATCH 01/26] drm/i915/gvt: verify functions types in new_mmio_info() The current prototype of new_mmio_info() uses void* for parameters read and write, which are functions with precise calling conventions (argument types and return type). Write down these conventions in new_mmio_info() definition. This has been reported by the following warnings when clang is used to build the kernel: drivers/gpu/drm/i915/gvt/handlers.c:124:21: error: pointer type mismatch ('void *' and 'int (*)(struct intel_vgpu *, unsigned int, void *, unsigned int)') [-Werror,-Wpointer-type-mismatch] info->read = read ? read : intel_vgpu_default_mmio_read; ^ ~~~~ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~ drivers/gpu/drm/i915/gvt/handlers.c:125:23: error: pointer type mismatch ('void *' and 'int (*)(struct intel_vgpu *, unsigned int, void *, unsigned int)') [-Werror,-Wpointer-type-mismatch] info->write = write ? write : intel_vgpu_default_mmio_write; ^ ~~~~~ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ This allows the compiler to detect that sbi_ctl_mmio_write() returns a "bool" value instead of an expected "int" one. Fix this. Signed-off-by: Nicolas Iooss Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/handlers.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/handlers.c b/drivers/gpu/drm/i915/gvt/handlers.c index 522809710312..052e57124c0a 100644 --- a/drivers/gpu/drm/i915/gvt/handlers.c +++ b/drivers/gpu/drm/i915/gvt/handlers.c @@ -93,7 +93,8 @@ static void write_vreg(struct intel_vgpu *vgpu, unsigned int offset, static int new_mmio_info(struct intel_gvt *gvt, u32 offset, u32 flags, u32 size, u32 addr_mask, u32 ro_mask, u32 device, - void *read, void *write) + int (*read)(struct intel_vgpu *, unsigned int, void *, unsigned int), + int (*write)(struct intel_vgpu *, unsigned int, void *, unsigned int)) { struct intel_gvt_mmio_info *info, *p; u32 start, end, i; @@ -974,7 +975,7 @@ static int sbi_data_mmio_read(struct intel_vgpu *vgpu, unsigned int offset, return 0; } -static bool sbi_ctl_mmio_write(struct intel_vgpu *vgpu, unsigned int offset, +static int sbi_ctl_mmio_write(struct intel_vgpu *vgpu, unsigned int offset, void *p_data, unsigned int bytes) { u32 data; From a12010534d0984f91bc5bdcf9e27bd55e20d82da Mon Sep 17 00:00:00 2001 From: Changbin Du Date: Tue, 27 Dec 2016 13:24:52 +0800 Subject: [PATCH 02/26] drm/i915/gvt: fix error handing of tlb_control emulation Return ealier for a invalid access, else it would false set tlb flag for RCS. Signed-off-by: Changbin Du Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/handlers.c | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/handlers.c b/drivers/gpu/drm/i915/gvt/handlers.c index 052e57124c0a..e06d5f37bb92 100644 --- a/drivers/gpu/drm/i915/gvt/handlers.c +++ b/drivers/gpu/drm/i915/gvt/handlers.c @@ -1367,7 +1367,6 @@ static int ring_mode_mmio_write(struct intel_vgpu *vgpu, unsigned int offset, static int gvt_reg_tlb_control_handler(struct intel_vgpu *vgpu, unsigned int offset, void *p_data, unsigned int bytes) { - int rc = 0; unsigned int id = 0; write_vreg(vgpu, offset, p_data, bytes); @@ -1390,12 +1389,11 @@ static int gvt_reg_tlb_control_handler(struct intel_vgpu *vgpu, id = VECS; break; default: - rc = -EINVAL; - break; + return -EINVAL; } set_bit(id, (void *)vgpu->tlb_handle_pending); - return rc; + return 0; } static int ring_reset_ctl_write(struct intel_vgpu *vgpu, From 39762ad437f1149b904e6baeaf28824da34a89c1 Mon Sep 17 00:00:00 2001 From: Changbin Du Date: Tue, 27 Dec 2016 13:25:06 +0800 Subject: [PATCH 03/26] drm/i915/gvt: fix return value in mul_force_wake_write All mmio handlers should return a negetive value for failure, not 1. Signed-off-by: Changbin Du Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/handlers.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/gvt/handlers.c b/drivers/gpu/drm/i915/gvt/handlers.c index e06d5f37bb92..8cbaf1c83720 100644 --- a/drivers/gpu/drm/i915/gvt/handlers.c +++ b/drivers/gpu/drm/i915/gvt/handlers.c @@ -220,7 +220,7 @@ static int mul_force_wake_write(struct intel_vgpu *vgpu, default: /*should not hit here*/ gvt_err("invalid forcewake offset 0x%x\n", offset); - return 1; + return -EINVAL; } } else { ack_reg_offset = FORCEWAKE_ACK_HSW_REG; From 905a5035ebe79e89712cda0bed1887c73aa8e9bb Mon Sep 17 00:00:00 2001 From: Changbin Du Date: Fri, 30 Dec 2016 14:10:53 +0800 Subject: [PATCH 04/26] drm/i915/gvt: always use readq and writeq The readq and writeq are already offered by drm_os_linux.h. So we can use them directly whithout dectecting their presence. This patch removed the duplicated code. Signed-off-by: Changbin Du Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/gtt.c | 15 ++------------- 1 file changed, 2 insertions(+), 13 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/gtt.c b/drivers/gpu/drm/i915/gvt/gtt.c index 6c5fdf5b2ce2..a32e59de0eff 100644 --- a/drivers/gpu/drm/i915/gvt/gtt.c +++ b/drivers/gpu/drm/i915/gvt/gtt.c @@ -240,15 +240,8 @@ static inline int get_pse_type(int type) static u64 read_pte64(struct drm_i915_private *dev_priv, unsigned long index) { void __iomem *addr = (gen8_pte_t __iomem *)dev_priv->ggtt.gsm + index; - u64 pte; -#ifdef readq - pte = readq(addr); -#else - pte = ioread32(addr); - pte |= (u64)ioread32(addr + 4) << 32; -#endif - return pte; + return readq(addr); } static void write_pte64(struct drm_i915_private *dev_priv, @@ -256,12 +249,8 @@ static void write_pte64(struct drm_i915_private *dev_priv, { void __iomem *addr = (gen8_pte_t __iomem *)dev_priv->ggtt.gsm + index; -#ifdef writeq writeq(pte, addr); -#else - iowrite32((u32)pte, addr); - iowrite32(pte >> 32, addr + 4); -#endif + I915_WRITE(GFX_FLSH_CNTL_GEN6, GFX_FLSH_CNTL_EN); POSTING_READ(GFX_FLSH_CNTL_GEN6); } From 901a14b721feef1b37cfe6362ee103e135133677 Mon Sep 17 00:00:00 2001 From: Pei Zhang Date: Wed, 4 Jan 2017 22:32:23 +0800 Subject: [PATCH 05/26] drm/i915/gvt: print correct value for untracked mmio In function intel_vgpu_emulate_mmio_read, the untracked mmio register is dumped through kernel log, but the register value is not correct. This patch fixes this issue. V2: fix the fromat warning from checkpatch.pl. Signed-off-by: Pei Zhang Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/mmio.c | 31 ++++++++++++++++--------------- 1 file changed, 16 insertions(+), 15 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/mmio.c b/drivers/gpu/drm/i915/gvt/mmio.c index 09c9450a1946..e60701397ac2 100644 --- a/drivers/gpu/drm/i915/gvt/mmio.c +++ b/drivers/gpu/drm/i915/gvt/mmio.c @@ -125,25 +125,12 @@ int intel_vgpu_emulate_mmio_read(struct intel_vgpu *vgpu, uint64_t pa, if (WARN_ON(!reg_is_mmio(gvt, offset + bytes - 1))) goto err; - mmio = intel_gvt_find_mmio_info(gvt, rounddown(offset, 4)); - if (!mmio && !vgpu->mmio.disable_warn_untrack) { - gvt_err("vgpu%d: read untracked MMIO %x len %d val %x\n", - vgpu->id, offset, bytes, *(u32 *)p_data); - - if (offset == 0x206c) { - gvt_err("------------------------------------------\n"); - gvt_err("vgpu%d: likely triggers a gfx reset\n", - vgpu->id); - gvt_err("------------------------------------------\n"); - vgpu->mmio.disable_warn_untrack = true; - } - } - if (!intel_gvt_mmio_is_unalign(gvt, offset)) { if (WARN_ON(!IS_ALIGNED(offset, bytes))) goto err; } + mmio = intel_gvt_find_mmio_info(gvt, rounddown(offset, 4)); if (mmio) { if (!intel_gvt_mmio_is_unalign(gvt, mmio->offset)) { if (WARN_ON(offset + bytes > mmio->offset + mmio->size)) @@ -152,9 +139,23 @@ int intel_vgpu_emulate_mmio_read(struct intel_vgpu *vgpu, uint64_t pa, goto err; } ret = mmio->read(vgpu, offset, p_data, bytes); - } else + } else { ret = intel_vgpu_default_mmio_read(vgpu, offset, p_data, bytes); + if (!vgpu->mmio.disable_warn_untrack) { + gvt_err("vgpu%d: read untracked MMIO %x(%dB) val %x\n", + vgpu->id, offset, bytes, *(u32 *)p_data); + + if (offset == 0x206c) { + gvt_err("------------------------------------------\n"); + gvt_err("vgpu%d: likely triggers a gfx reset\n", + vgpu->id); + gvt_err("------------------------------------------\n"); + vgpu->mmio.disable_warn_untrack = true; + } + } + } + if (ret) goto err; From 888530b57f88f2bc856f181479df732c9622fa22 Mon Sep 17 00:00:00 2001 From: Zhenyu Wang Date: Thu, 5 Jan 2017 10:26:13 +0800 Subject: [PATCH 06/26] drm/i915/gvt: adjust high memory size for default vGPU type Previous high mem size initialized for vGPU type was too small which caused failure for some VMs. This trys to take minimal value of 384MB for each VM and enlarge default high mem size to make guest driver happy. Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/vgpu.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/vgpu.c b/drivers/gpu/drm/i915/gvt/vgpu.c index 536d2b9d5777..398abb98dd0a 100644 --- a/drivers/gpu/drm/i915/gvt/vgpu.c +++ b/drivers/gpu/drm/i915/gvt/vgpu.c @@ -177,7 +177,7 @@ int intel_gvt_init_vgpu_types(struct intel_gvt *gvt) if (low_avail / min_low == 0) break; gvt->types[i].low_gm_size = min_low; - gvt->types[i].high_gm_size = 3 * gvt->types[i].low_gm_size; + gvt->types[i].high_gm_size = max((min_low<<3), MB_TO_BYTES(384U)); gvt->types[i].fence = 4; gvt->types[i].max_instance = low_avail / min_low; gvt->types[i].avail_instance = gvt->types[i].max_instance; @@ -217,7 +217,7 @@ static void intel_gvt_update_vgpu_types(struct intel_gvt *gvt) */ low_gm_avail = MB_TO_BYTES(256) - HOST_LOW_GM_SIZE - gvt->gm.vgpu_allocated_low_gm_size; - high_gm_avail = MB_TO_BYTES(256) * 3 - HOST_HIGH_GM_SIZE - + high_gm_avail = MB_TO_BYTES(256) * 8UL - HOST_HIGH_GM_SIZE - gvt->gm.vgpu_allocated_high_gm_size; fence_avail = gvt_fence_sz(gvt) - HOST_FENCE - gvt->fence.vgpu_allocated_fence_num; From 2fcdb66364ee467d69228a3d2ea074498c177211 Mon Sep 17 00:00:00 2001 From: Zhenyu Wang Date: Thu, 5 Jan 2017 10:26:24 +0800 Subject: [PATCH 07/26] drm/i915/gvt: remove duplicated definition Remove duplicated definition for resource size in aperture_gm.c which are already defined in gvt.h. Need only one to take effect. Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/aperture_gm.c | 7 ------- 1 file changed, 7 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/aperture_gm.c b/drivers/gpu/drm/i915/gvt/aperture_gm.c index 0d41ebc4aea6..65200313515c 100644 --- a/drivers/gpu/drm/i915/gvt/aperture_gm.c +++ b/drivers/gpu/drm/i915/gvt/aperture_gm.c @@ -37,13 +37,6 @@ #include "i915_drv.h" #include "gvt.h" -#define MB_TO_BYTES(mb) ((mb) << 20ULL) -#define BYTES_TO_MB(b) ((b) >> 20ULL) - -#define HOST_LOW_GM_SIZE MB_TO_BYTES(128) -#define HOST_HIGH_GM_SIZE MB_TO_BYTES(384) -#define HOST_FENCE 4 - static int alloc_gm(struct intel_vgpu *vgpu, bool high_gm) { struct intel_gvt *gvt = vgpu->gvt; From 2e51ef32b0d66fcd5fe45c437cf7c6aef8350746 Mon Sep 17 00:00:00 2001 From: Changbin Du Date: Thu, 5 Jan 2017 13:28:05 +0800 Subject: [PATCH 08/26] drm/i915/gvt: fix use after free for workload In the function workload_thread(), we invoke complete_current_workload() to cleanup the just processed workload (workload will be freed there). So we cannot access workload->req after that. This patch move complete_current_workload() afterward. Signed-off-by: Changbin Du Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/scheduler.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c index 4db242250235..c694dd039f3b 100644 --- a/drivers/gpu/drm/i915/gvt/scheduler.c +++ b/drivers/gpu/drm/i915/gvt/scheduler.c @@ -459,11 +459,11 @@ complete: gvt_dbg_sched("will complete workload %p\n, status: %d\n", workload, workload->status); - complete_current_workload(gvt, ring_id); - if (workload->req) i915_gem_request_put(fetch_and_zero(&workload->req)); + complete_current_workload(gvt, ring_id); + if (need_force_wake) intel_uncore_forcewake_put(gvt->dev_priv, FORCEWAKE_ALL); From 440a9b9fae37dfd7e4c7d76db34fada57f9afd92 Mon Sep 17 00:00:00 2001 From: Changbin Du Date: Thu, 5 Jan 2017 16:49:03 +0800 Subject: [PATCH 09/26] drm/i915/gvt: dec vgpu->running_workload_num after the workload is really done The vgpu->running_workload_num is used to determine whether a vgpu has any workload running or not. So we should make sure the workload is really done before we dec running_workload_num. Function complete_current_workload is not the right place to do it, since this function is still processing the workload. This patch move the dec op afterward. v2: move dec op before wake_up(&scheduler->workload_complete_wq) (Min He) Signed-off-by: Changbin Du Reviewed-by: Min He Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/scheduler.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c index c694dd039f3b..e91885dffeff 100644 --- a/drivers/gpu/drm/i915/gvt/scheduler.c +++ b/drivers/gpu/drm/i915/gvt/scheduler.c @@ -350,13 +350,15 @@ static void complete_current_workload(struct intel_gvt *gvt, int ring_id) { struct intel_gvt_workload_scheduler *scheduler = &gvt->scheduler; struct intel_vgpu_workload *workload; + struct intel_vgpu *vgpu; int event; mutex_lock(&gvt->lock); workload = scheduler->current_workload[ring_id]; + vgpu = workload->vgpu; - if (!workload->status && !workload->vgpu->resetting) { + if (!workload->status && !vgpu->resetting) { wait_event(workload->shadow_ctx_status_wq, !atomic_read(&workload->shadow_ctx_active)); @@ -364,8 +366,7 @@ static void complete_current_workload(struct intel_gvt *gvt, int ring_id) for_each_set_bit(event, workload->pending_events, INTEL_GVT_EVENT_MAX) - intel_vgpu_trigger_virtual_event(workload->vgpu, - event); + intel_vgpu_trigger_virtual_event(vgpu, event); } gvt_dbg_sched("ring id %d complete workload %p status %d\n", @@ -373,11 +374,10 @@ static void complete_current_workload(struct intel_gvt *gvt, int ring_id) scheduler->current_workload[ring_id] = NULL; - atomic_dec(&workload->vgpu->running_workload_num); - list_del_init(&workload->list); workload->complete(workload); + atomic_dec(&vgpu->running_workload_num); wake_up(&scheduler->workload_complete_wq); mutex_unlock(&gvt->lock); } From 59c0573dfbd5f66e3aa54c2ce0bebcb0953d4db4 Mon Sep 17 00:00:00 2001 From: Jike Song Date: Fri, 6 Jan 2017 15:16:21 +0800 Subject: [PATCH 10/26] drm/i915/gvt: init/destroy vgpu_idr properly An idr should be initialized before use and destroyed afterwards. Signed-off-by: Jike Song Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/gvt.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/gvt/gvt.c b/drivers/gpu/drm/i915/gvt/gvt.c index 398877c3d2fd..e6bf5c533fbe 100644 --- a/drivers/gpu/drm/i915/gvt/gvt.c +++ b/drivers/gpu/drm/i915/gvt/gvt.c @@ -201,6 +201,8 @@ void intel_gvt_clean_device(struct drm_i915_private *dev_priv) intel_gvt_hypervisor_host_exit(&dev_priv->drm.pdev->dev, gvt); intel_gvt_clean_vgpu_types(gvt); + idr_destroy(&gvt->vgpu_idr); + kfree(dev_priv->gvt); dev_priv->gvt = NULL; } @@ -237,6 +239,8 @@ int intel_gvt_init_device(struct drm_i915_private *dev_priv) gvt_dbg_core("init gvt device\n"); + idr_init(&gvt->vgpu_idr); + mutex_init(&gvt->lock); gvt->dev_priv = dev_priv; @@ -244,7 +248,7 @@ int intel_gvt_init_device(struct drm_i915_private *dev_priv) ret = intel_gvt_setup_mmio_info(gvt); if (ret) - return ret; + goto out_clean_idr; ret = intel_gvt_load_firmware(gvt); if (ret) @@ -313,6 +317,8 @@ out_free_firmware: intel_gvt_free_firmware(gvt); out_clean_mmio_info: intel_gvt_clean_mmio_info(gvt); +out_clean_idr: + idr_destroy(&gvt->vgpu_idr); kfree(gvt); return ret; } From 4e5378918b5b96e6b93fcadf1ab84a8486ca60a1 Mon Sep 17 00:00:00 2001 From: Jike Song Date: Fri, 6 Jan 2017 15:16:22 +0800 Subject: [PATCH 11/26] drm/i915/gvt: destroy the allocated idr on vgpu creating failures Once idr_alloc gets called data is allocated within the idr list, if any error occurs afterwards, we should undo that by idr_remove on the error path. Signed-off-by: Jike Song Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/vgpu.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/gvt/vgpu.c b/drivers/gpu/drm/i915/gvt/vgpu.c index 398abb98dd0a..f0e86123e45b 100644 --- a/drivers/gpu/drm/i915/gvt/vgpu.c +++ b/drivers/gpu/drm/i915/gvt/vgpu.c @@ -304,7 +304,7 @@ static struct intel_vgpu *__intel_gvt_create_vgpu(struct intel_gvt *gvt, ret = setup_vgpu_mmio(vgpu); if (ret) - goto out_free_vgpu; + goto out_clean_idr; ret = intel_vgpu_alloc_resource(vgpu, param); if (ret) @@ -355,6 +355,8 @@ out_clean_vgpu_resource: intel_vgpu_free_resource(vgpu); out_clean_vgpu_mmio: clean_vgpu_mmio(vgpu); +out_clean_idr: + idr_remove(&gvt->vgpu_idr, vgpu->id); out_free_vgpu: vfree(vgpu); mutex_unlock(&gvt->lock); From 03551e971f6e52c8dedd5741bf48631e65675759 Mon Sep 17 00:00:00 2001 From: Jike Song Date: Fri, 6 Jan 2017 15:16:23 +0800 Subject: [PATCH 12/26] drm/i915/gvt: cleanup opregion memory allocation code According to the spec, ACPI OpRegion must be placed at a physical address below 4G. That is, for a vGPU it must be associated with a GPA below 4G, but on host side, it doesn't matter where the backing pages actually are. So when allocating pages from host, the GFP_DMA32 flag is unnecessary. Also the allocation is from a sleepable context, so GFP_ATOMIC is also unnecessary. This patch also removes INTEL_GVT_OPREGION_PORDER and use get_order() instead. Signed-off-by: Jike Song Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/opregion.c | 8 ++++---- drivers/gpu/drm/i915/gvt/reg.h | 3 +-- 2 files changed, 5 insertions(+), 6 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/opregion.c b/drivers/gpu/drm/i915/gvt/opregion.c index 81cd921770c6..d9fb41ab7119 100644 --- a/drivers/gpu/drm/i915/gvt/opregion.c +++ b/drivers/gpu/drm/i915/gvt/opregion.c @@ -36,9 +36,9 @@ static int init_vgpu_opregion(struct intel_vgpu *vgpu, u32 gpa) vgpu->id)) return -EINVAL; - vgpu_opregion(vgpu)->va = (void *)__get_free_pages(GFP_ATOMIC | - GFP_DMA32 | __GFP_ZERO, - INTEL_GVT_OPREGION_PORDER); + vgpu_opregion(vgpu)->va = (void *)__get_free_pages(GFP_KERNEL | + __GFP_ZERO, + get_order(INTEL_GVT_OPREGION_SIZE)); if (!vgpu_opregion(vgpu)->va) return -ENOMEM; @@ -97,7 +97,7 @@ void intel_vgpu_clean_opregion(struct intel_vgpu *vgpu) if (intel_gvt_host.hypervisor_type == INTEL_GVT_HYPERVISOR_XEN) { map_vgpu_opregion(vgpu, false); free_pages((unsigned long)vgpu_opregion(vgpu)->va, - INTEL_GVT_OPREGION_PORDER); + get_order(INTEL_GVT_OPREGION_SIZE)); vgpu_opregion(vgpu)->va = NULL; } diff --git a/drivers/gpu/drm/i915/gvt/reg.h b/drivers/gpu/drm/i915/gvt/reg.h index 0dfe789d8f02..fbd023a16f18 100644 --- a/drivers/gpu/drm/i915/gvt/reg.h +++ b/drivers/gpu/drm/i915/gvt/reg.h @@ -50,8 +50,7 @@ #define INTEL_GVT_OPREGION_PARM 0x204 #define INTEL_GVT_OPREGION_PAGES 2 -#define INTEL_GVT_OPREGION_PORDER 1 -#define INTEL_GVT_OPREGION_SIZE (2 * 4096) +#define INTEL_GVT_OPREGION_SIZE (INTEL_GVT_OPREGION_PAGES * PAGE_SIZE) #define VGT_SPRSTRIDE(pipe) _PIPE(pipe, _SPRA_STRIDE, _PLANE_STRIDE_2_B) From 5753394b64a07dd502cb288a5fd52e71fb01fc5d Mon Sep 17 00:00:00 2001 From: Jike Song Date: Fri, 6 Jan 2017 15:16:20 +0800 Subject: [PATCH 13/26] drm/i915/gvt/kvmgt: return meaningful error for vgpu creating failure The vgpu_create() routine we called returns meaningful errors to indicate failures, so we'd better to pass it to our caller, the mdev framework, whereby the sysfs is able to tell userspace what happened. Signed-off-by: Jike Song Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/kvmgt.c | 14 ++++++++++---- 1 file changed, 10 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/kvmgt.c b/drivers/gpu/drm/i915/gvt/kvmgt.c index faaae07ae487..0c9234a87a20 100644 --- a/drivers/gpu/drm/i915/gvt/kvmgt.c +++ b/drivers/gpu/drm/i915/gvt/kvmgt.c @@ -398,6 +398,7 @@ static int intel_vgpu_create(struct kobject *kobj, struct mdev_device *mdev) struct intel_vgpu_type *type; struct device *pdev; void *gvt; + int ret; pdev = mdev_parent_dev(mdev); gvt = kdev_to_i915(pdev)->gvt; @@ -406,13 +407,15 @@ static int intel_vgpu_create(struct kobject *kobj, struct mdev_device *mdev) if (!type) { gvt_err("failed to find type %s to create\n", kobject_name(kobj)); - return -EINVAL; + ret = -EINVAL; + goto out; } vgpu = intel_gvt_ops->vgpu_create(gvt, type); if (IS_ERR_OR_NULL(vgpu)) { - gvt_err("create intel vgpu failed\n"); - return -EINVAL; + ret = vgpu == NULL ? -EFAULT : PTR_ERR(vgpu); + gvt_err("failed to create intel vgpu: %d\n", ret); + goto out; } INIT_WORK(&vgpu->vdev.release_work, intel_vgpu_release_work); @@ -422,7 +425,10 @@ static int intel_vgpu_create(struct kobject *kobj, struct mdev_device *mdev) gvt_dbg_core("intel_vgpu_create succeeded for mdev: %s\n", dev_name(mdev_dev(mdev))); - return 0; + ret = 0; + +out: + return ret; } static int intel_vgpu_remove(struct mdev_device *mdev) From 9631739f8196ec80b5d9bf955f79b711490c0205 Mon Sep 17 00:00:00 2001 From: Jike Song Date: Mon, 9 Jan 2017 15:38:38 +0800 Subject: [PATCH 14/26] drm/i915/gvt: cleanup GFP flags In gvt, almost all memory allocations are in sleepable contexts. It's fault-prone to use GFP_ATOMIC everywhere. Replace it with GFP_KERNEL wherever possible. Signed-off-by: Jike Song Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/gtt.c | 39 +++++++++++++--------------------- 1 file changed, 15 insertions(+), 24 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/gtt.c b/drivers/gpu/drm/i915/gvt/gtt.c index a32e59de0eff..3cf0df0bb391 100644 --- a/drivers/gpu/drm/i915/gvt/gtt.c +++ b/drivers/gpu/drm/i915/gvt/gtt.c @@ -1369,8 +1369,7 @@ static int gen8_mm_alloc_page_table(struct intel_vgpu_mm *mm) info->gtt_entry_size; mem = kzalloc(mm->has_shadow_page_table ? mm->page_table_entry_size * 2 - : mm->page_table_entry_size, - GFP_ATOMIC); + : mm->page_table_entry_size, GFP_KERNEL); if (!mem) return -ENOMEM; mm->virtual_page_table = mem; @@ -1521,7 +1520,7 @@ struct intel_vgpu_mm *intel_vgpu_create_mm(struct intel_vgpu *vgpu, struct intel_vgpu_mm *mm; int ret; - mm = kzalloc(sizeof(*mm), GFP_ATOMIC); + mm = kzalloc(sizeof(*mm), GFP_KERNEL); if (!mm) { ret = -ENOMEM; goto fail; @@ -1875,30 +1874,27 @@ static int alloc_scratch_pages(struct intel_vgpu *vgpu, struct intel_gvt_gtt_pte_ops *ops = vgpu->gvt->gtt.pte_ops; int page_entry_num = GTT_PAGE_SIZE >> vgpu->gvt->device_info.gtt_entry_size_shift; - struct page *scratch_pt; + void *scratch_pt; unsigned long mfn; int i; - void *p; if (WARN_ON(type < GTT_TYPE_PPGTT_PTE_PT || type >= GTT_TYPE_MAX)) return -EINVAL; - scratch_pt = alloc_page(GFP_KERNEL | GFP_ATOMIC | __GFP_ZERO); + scratch_pt = (void *)get_zeroed_page(GFP_KERNEL); if (!scratch_pt) { gvt_err("fail to allocate scratch page\n"); return -ENOMEM; } - p = kmap_atomic(scratch_pt); - mfn = intel_gvt_hypervisor_virt_to_mfn(p); + mfn = intel_gvt_hypervisor_virt_to_mfn(scratch_pt); if (mfn == INTEL_GVT_INVALID_ADDR) { - gvt_err("fail to translate vaddr:0x%llx\n", (u64)p); - kunmap_atomic(p); - __free_page(scratch_pt); + gvt_err("fail to translate vaddr:0x%lx\n", (unsigned long)scratch_pt); + free_page((unsigned long)scratch_pt); return -EFAULT; } gtt->scratch_pt[type].page_mfn = mfn; - gtt->scratch_pt[type].page = scratch_pt; + gtt->scratch_pt[type].page = virt_to_page(scratch_pt); gvt_dbg_mm("vgpu%d create scratch_pt: type %d mfn=0x%lx\n", vgpu->id, type, mfn); @@ -1907,7 +1903,7 @@ static int alloc_scratch_pages(struct intel_vgpu *vgpu, * scratch_pt[type] indicate the scratch pt/scratch page used by the * 'type' pt. * e.g. scratch_pt[GTT_TYPE_PPGTT_PDE_PT] is used by - * GTT_TYPE_PPGTT_PDE_PT level pt, that means this scatch_pt it self + * GTT_TYPE_PPGTT_PDE_PT level pt, that means this scratch_pt it self * is GTT_TYPE_PPGTT_PTE_PT, and full filled by scratch page mfn. */ if (type > GTT_TYPE_PPGTT_PTE_PT && type < GTT_TYPE_MAX) { @@ -1925,11 +1921,9 @@ static int alloc_scratch_pages(struct intel_vgpu *vgpu, se.val64 |= PPAT_CACHED_INDEX; for (i = 0; i < page_entry_num; i++) - ops->set_entry(p, &se, i, false, 0, vgpu); + ops->set_entry(scratch_pt, &se, i, false, 0, vgpu); } - kunmap_atomic(p); - return 0; } @@ -2197,7 +2191,7 @@ int intel_vgpu_g2v_destroy_ppgtt_mm(struct intel_vgpu *vgpu, int intel_gvt_init_gtt(struct intel_gvt *gvt) { int ret; - void *page_addr; + void *page; gvt_dbg_core("init gtt\n"); @@ -2210,17 +2204,14 @@ int intel_gvt_init_gtt(struct intel_gvt *gvt) return -ENODEV; } - gvt->gtt.scratch_ggtt_page = - alloc_page(GFP_KERNEL | GFP_ATOMIC | __GFP_ZERO); - if (!gvt->gtt.scratch_ggtt_page) { + page = (void *)get_zeroed_page(GFP_KERNEL); + if (!page) { gvt_err("fail to allocate scratch ggtt page\n"); return -ENOMEM; } + gvt->gtt.scratch_ggtt_page = virt_to_page(page); - page_addr = page_address(gvt->gtt.scratch_ggtt_page); - - gvt->gtt.scratch_ggtt_mfn = - intel_gvt_hypervisor_virt_to_mfn(page_addr); + gvt->gtt.scratch_ggtt_mfn = intel_gvt_hypervisor_virt_to_mfn(page); if (gvt->gtt.scratch_ggtt_mfn == INTEL_GVT_INVALID_ADDR) { gvt_err("fail to translate scratch ggtt page\n"); __free_page(gvt->gtt.scratch_ggtt_page); From e88893fea17996018b2d68a22e677ea04f3baadf Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Thu, 5 Jan 2017 15:59:40 +0000 Subject: [PATCH 15/26] drm/i915: Clear ret before unbinding in i915_gem_evict_something() Missed when rebasing patches, I failed to set ret to zero before starting the unbind loop (which depends upon ret being zero). Reported-by: Matthew Auld Fixes: 9332f3b1b99a ("drm/i915: Combine loops within i915_gem_evict_something") Signed-off-by: Chris Wilson Cc: Matthew Auld Link: http://patchwork.freedesktop.org/patch/msgid/20170105155940.10033-1-chris@chris-wilson.co.uk Reviewed-by: Matthew Auld Cc: # v4.9+ (cherry picked from commit 121dfbb2a2ef1c5f49e15c38ccc47ff0beb59446) Signed-off-by: Jani Nikula --- drivers/gpu/drm/i915/i915_gem_evict.c | 1 + 1 file changed, 1 insertion(+) diff --git a/drivers/gpu/drm/i915/i915_gem_evict.c b/drivers/gpu/drm/i915/i915_gem_evict.c index bd08814b015c..d534a316a16e 100644 --- a/drivers/gpu/drm/i915/i915_gem_evict.c +++ b/drivers/gpu/drm/i915/i915_gem_evict.c @@ -199,6 +199,7 @@ found: } /* Unbinding will emit any required flushes */ + ret = 0; while (!list_empty(&eviction_list)) { vma = list_first_entry(&eviction_list, struct i915_vma, From e4621b73b6b472fe2b434b4f0f76b8f33ee26a73 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Fri, 6 Jan 2017 15:22:38 +0000 Subject: [PATCH 16/26] drm/i915: Fix phys pwrite for struct_mutex-less operation Since commit fe115628d567 ("drm/i915: Implement pwrite without struct-mutex") the lowlevel pwrite calls are now called without the protection of struct_mutex, but pwrite_phys was still asserting that it held the struct_mutex and later tried to drop and relock it. Fixes: fe115628d567 ("drm/i915: Implement pwrite without struct-mutex") Signed-off-by: Chris Wilson Cc: Joonas Lahtinen Cc: Link: http://patchwork.freedesktop.org/patch/msgid/20170106152240.5793-1-chris@chris-wilson.co.uk Reviewed-by: Tvrtko Ursulin (cherry picked from commit 10466d2a59b23aa6d5ecd5310296c8cdb6458dac) Signed-off-by: Jani Nikula --- drivers/gpu/drm/i915/i915_gem.c | 34 ++++----------------------------- 1 file changed, 4 insertions(+), 30 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 3dd7fc662859..4b23a7814713 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -595,47 +595,21 @@ i915_gem_phys_pwrite(struct drm_i915_gem_object *obj, struct drm_i915_gem_pwrite *args, struct drm_file *file) { - struct drm_device *dev = obj->base.dev; void *vaddr = obj->phys_handle->vaddr + args->offset; char __user *user_data = u64_to_user_ptr(args->data_ptr); - int ret; /* We manually control the domain here and pretend that it * remains coherent i.e. in the GTT domain, like shmem_pwrite. */ - lockdep_assert_held(&obj->base.dev->struct_mutex); - ret = i915_gem_object_wait(obj, - I915_WAIT_INTERRUPTIBLE | - I915_WAIT_LOCKED | - I915_WAIT_ALL, - MAX_SCHEDULE_TIMEOUT, - to_rps_client(file)); - if (ret) - return ret; - intel_fb_obj_invalidate(obj, ORIGIN_CPU); - if (__copy_from_user_inatomic_nocache(vaddr, user_data, args->size)) { - unsigned long unwritten; - - /* The physical object once assigned is fixed for the lifetime - * of the obj, so we can safely drop the lock and continue - * to access vaddr. - */ - mutex_unlock(&dev->struct_mutex); - unwritten = copy_from_user(vaddr, user_data, args->size); - mutex_lock(&dev->struct_mutex); - if (unwritten) { - ret = -EFAULT; - goto out; - } - } + if (copy_from_user(vaddr, user_data, args->size)) + return -EFAULT; drm_clflush_virt_range(vaddr, args->size); - i915_gem_chipset_flush(to_i915(dev)); + i915_gem_chipset_flush(to_i915(obj->base.dev)); -out: intel_fb_obj_flush(obj, false, ORIGIN_CPU); - return ret; + return 0; } void *i915_gem_object_alloc(struct drm_device *dev) From d22a48bf7302ef064295749fa79cd47093c5a000 Mon Sep 17 00:00:00 2001 From: Changbin Du Date: Fri, 13 Jan 2017 11:15:56 +0800 Subject: [PATCH 17/26] drm/i915/gvt: introudce intel_vgpu_reset_resource() to reset vgpu resource state This patch introudces a new function intel_vgpu_reset_resource() to reset allocated vgpu resources by intel_vgpu_alloc_resource(). So far we only need clear the fence registers. The function _clear_vgpu_fence() will reset both virtual and physical fence registers to 0. Signed-off-by: Changbin Du Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/aperture_gm.c | 29 ++++++++++++++++++++++++-- drivers/gpu/drm/i915/gvt/gvt.h | 1 + 2 files changed, 28 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/aperture_gm.c b/drivers/gpu/drm/i915/gvt/aperture_gm.c index 65200313515c..f7bce8603958 100644 --- a/drivers/gpu/drm/i915/gvt/aperture_gm.c +++ b/drivers/gpu/drm/i915/gvt/aperture_gm.c @@ -158,6 +158,14 @@ void intel_vgpu_write_fence(struct intel_vgpu *vgpu, POSTING_READ(fence_reg_lo); } +static void _clear_vgpu_fence(struct intel_vgpu *vgpu) +{ + int i; + + for (i = 0; i < vgpu_fence_sz(vgpu); i++) + intel_vgpu_write_fence(vgpu, i, 0); +} + static void free_vgpu_fence(struct intel_vgpu *vgpu) { struct intel_gvt *gvt = vgpu->gvt; @@ -171,9 +179,9 @@ static void free_vgpu_fence(struct intel_vgpu *vgpu) intel_runtime_pm_get(dev_priv); mutex_lock(&dev_priv->drm.struct_mutex); + _clear_vgpu_fence(vgpu); for (i = 0; i < vgpu_fence_sz(vgpu); i++) { reg = vgpu->fence.regs[i]; - intel_vgpu_write_fence(vgpu, i, 0); list_add_tail(®->link, &dev_priv->mm.fence_list); } @@ -201,13 +209,14 @@ static int alloc_vgpu_fence(struct intel_vgpu *vgpu) continue; list_del(pos); vgpu->fence.regs[i] = reg; - intel_vgpu_write_fence(vgpu, i, 0); if (++i == vgpu_fence_sz(vgpu)) break; } if (i != vgpu_fence_sz(vgpu)) goto out_free_fence; + _clear_vgpu_fence(vgpu); + mutex_unlock(&dev_priv->drm.struct_mutex); intel_runtime_pm_put(dev_priv); return 0; @@ -306,6 +315,22 @@ void intel_vgpu_free_resource(struct intel_vgpu *vgpu) free_resource(vgpu); } +/** + * intel_vgpu_reset_resource - reset resource state owned by a vGPU + * @vgpu: a vGPU + * + * This function is used to reset resource state owned by a vGPU. + * + */ +void intel_vgpu_reset_resource(struct intel_vgpu *vgpu) +{ + struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv; + + intel_runtime_pm_get(dev_priv); + _clear_vgpu_fence(vgpu); + intel_runtime_pm_put(dev_priv); +} + /** * intel_alloc_vgpu_resource - allocate HW resource for a vGPU * @vgpu: vGPU diff --git a/drivers/gpu/drm/i915/gvt/gvt.h b/drivers/gpu/drm/i915/gvt/gvt.h index 0af17016f33f..455281d07d75 100644 --- a/drivers/gpu/drm/i915/gvt/gvt.h +++ b/drivers/gpu/drm/i915/gvt/gvt.h @@ -323,6 +323,7 @@ struct intel_vgpu_creation_params { int intel_vgpu_alloc_resource(struct intel_vgpu *vgpu, struct intel_vgpu_creation_params *param); +void intel_vgpu_reset_resource(struct intel_vgpu *vgpu); void intel_vgpu_free_resource(struct intel_vgpu *vgpu); void intel_vgpu_write_fence(struct intel_vgpu *vgpu, u32 fence, u64 value); From b611581b375ce28536ab50be9cd507bb6092fb1e Mon Sep 17 00:00:00 2001 From: Changbin Du Date: Fri, 13 Jan 2017 11:15:57 +0800 Subject: [PATCH 18/26] drm/i915/gvt: introuduce intel_vgpu_reset_gtt() to reset gtt This patch introduces a new function intel_vgpu_reset_gtt() to reset the all GTT related status, including GGTT, PPGTT, scratch page. This function can free all shadowed PPGTT, clear all GGTT entry, and clear scratch page to all zero. After this, we can ensure no gtt related information can be leakaged from one VM to anothor one when assign vgpu instance across different VMs (not simultaneously). Signed-off-by: Changbin Du Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/gtt.c | 27 +++++++++++++++++++++++++++ drivers/gpu/drm/i915/gvt/gtt.h | 1 + 2 files changed, 28 insertions(+) diff --git a/drivers/gpu/drm/i915/gvt/gtt.c b/drivers/gpu/drm/i915/gvt/gtt.c index 3cf0df0bb391..47dec4acf7ff 100644 --- a/drivers/gpu/drm/i915/gvt/gtt.c +++ b/drivers/gpu/drm/i915/gvt/gtt.c @@ -2277,3 +2277,30 @@ void intel_vgpu_reset_ggtt(struct intel_vgpu *vgpu) for (offset = 0; offset < num_entries; offset++) ops->set_entry(NULL, &e, index + offset, false, 0, vgpu); } + +/** + * intel_vgpu_reset_gtt - reset the all GTT related status + * @vgpu: a vGPU + * @dmlr: true for vGPU Device Model Level Reset, false for GT Reset + * + * This function is called from vfio core to reset reset all + * GTT related status, including GGTT, PPGTT, scratch page. + * + */ +void intel_vgpu_reset_gtt(struct intel_vgpu *vgpu, bool dmlr) +{ + int i; + + ppgtt_free_all_shadow_page(vgpu); + if (!dmlr) + return; + + intel_vgpu_reset_ggtt(vgpu); + + /* clear scratch page for security */ + for (i = GTT_TYPE_PPGTT_PTE_PT; i < GTT_TYPE_MAX; i++) { + if (vgpu->gtt.scratch_pt[i].page != NULL) + memset(page_address(vgpu->gtt.scratch_pt[i].page), + 0, PAGE_SIZE); + } +} diff --git a/drivers/gpu/drm/i915/gvt/gtt.h b/drivers/gpu/drm/i915/gvt/gtt.h index b315ab3593ec..f88eb5e89bea 100644 --- a/drivers/gpu/drm/i915/gvt/gtt.h +++ b/drivers/gpu/drm/i915/gvt/gtt.h @@ -208,6 +208,7 @@ extern void intel_vgpu_clean_gtt(struct intel_vgpu *vgpu); void intel_vgpu_reset_ggtt(struct intel_vgpu *vgpu); extern int intel_gvt_init_gtt(struct intel_gvt *gvt); +extern void intel_vgpu_reset_gtt(struct intel_vgpu *vgpu, bool dmlr); extern void intel_gvt_clean_gtt(struct intel_gvt *gvt); extern struct intel_vgpu_mm *intel_gvt_find_ppgtt_mm(struct intel_vgpu *vgpu, From 536fc234074b09adae1763d8fb5b2d947847ad1d Mon Sep 17 00:00:00 2001 From: Changbin Du Date: Fri, 13 Jan 2017 11:15:58 +0800 Subject: [PATCH 19/26] drm/i915/gvt: move cfg space inititation function to cfg_space.c Move the configuration space inititation function setup_vgpu_cfg_space() in vgpu.c to dedicated source file cfg_space.c, and rename the function as intel_vgpu_init_cfg_space(). Signed-off-by: Changbin Du Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/cfg_space.c | 49 ++++++++++++++++++++++++++++ drivers/gpu/drm/i915/gvt/gvt.h | 2 ++ drivers/gpu/drm/i915/gvt/vgpu.c | 44 +------------------------ 3 files changed, 52 insertions(+), 43 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/cfg_space.c b/drivers/gpu/drm/i915/gvt/cfg_space.c index 711c31c8d8b4..82f449d59884 100644 --- a/drivers/gpu/drm/i915/gvt/cfg_space.c +++ b/drivers/gpu/drm/i915/gvt/cfg_space.c @@ -282,3 +282,52 @@ int intel_vgpu_emulate_cfg_write(struct intel_vgpu *vgpu, unsigned int offset, } return 0; } + +/** + * intel_vgpu_init_cfg_space - init vGPU configuration space when create vGPU + * + * @vgpu: a vGPU + * @primary: is the vGPU presented as primary + * + */ +void intel_vgpu_init_cfg_space(struct intel_vgpu *vgpu, + bool primary) +{ + struct intel_gvt *gvt = vgpu->gvt; + const struct intel_gvt_device_info *info = &gvt->device_info; + u16 *gmch_ctl; + int i; + + memcpy(vgpu_cfg_space(vgpu), gvt->firmware.cfg_space, + info->cfg_space_size); + + if (!primary) { + vgpu_cfg_space(vgpu)[PCI_CLASS_DEVICE] = + INTEL_GVT_PCI_CLASS_VGA_OTHER; + vgpu_cfg_space(vgpu)[PCI_CLASS_PROG] = + INTEL_GVT_PCI_CLASS_VGA_OTHER; + } + + /* Show guest that there isn't any stolen memory.*/ + gmch_ctl = (u16 *)(vgpu_cfg_space(vgpu) + INTEL_GVT_PCI_GMCH_CONTROL); + *gmch_ctl &= ~(BDW_GMCH_GMS_MASK << BDW_GMCH_GMS_SHIFT); + + intel_vgpu_write_pci_bar(vgpu, PCI_BASE_ADDRESS_2, + gvt_aperture_pa_base(gvt), true); + + vgpu_cfg_space(vgpu)[PCI_COMMAND] &= ~(PCI_COMMAND_IO + | PCI_COMMAND_MEMORY + | PCI_COMMAND_MASTER); + /* + * Clear the bar upper 32bit and let guest to assign the new value + */ + memset(vgpu_cfg_space(vgpu) + PCI_BASE_ADDRESS_1, 0, 4); + memset(vgpu_cfg_space(vgpu) + PCI_BASE_ADDRESS_3, 0, 4); + memset(vgpu_cfg_space(vgpu) + INTEL_GVT_PCI_OPREGION, 0, 4); + + for (i = 0; i < INTEL_GVT_MAX_BAR_NUM; i++) { + vgpu->cfg_space.bar[i].size = pci_resource_len( + gvt->dev_priv->drm.pdev, i * 2); + vgpu->cfg_space.bar[i].tracked = false; + } +} diff --git a/drivers/gpu/drm/i915/gvt/gvt.h b/drivers/gpu/drm/i915/gvt/gvt.h index 455281d07d75..ab505552d313 100644 --- a/drivers/gpu/drm/i915/gvt/gvt.h +++ b/drivers/gpu/drm/i915/gvt/gvt.h @@ -412,6 +412,8 @@ int intel_gvt_ggtt_index_g2h(struct intel_vgpu *vgpu, unsigned long g_index, int intel_gvt_ggtt_h2g_index(struct intel_vgpu *vgpu, unsigned long h_index, unsigned long *g_index); +void intel_vgpu_init_cfg_space(struct intel_vgpu *vgpu, + bool primary); int intel_vgpu_emulate_cfg_read(struct intel_vgpu *vgpu, unsigned int offset, void *p_data, unsigned int bytes); diff --git a/drivers/gpu/drm/i915/gvt/vgpu.c b/drivers/gpu/drm/i915/gvt/vgpu.c index f0e86123e45b..35c274177da8 100644 --- a/drivers/gpu/drm/i915/gvt/vgpu.c +++ b/drivers/gpu/drm/i915/gvt/vgpu.c @@ -66,48 +66,6 @@ int setup_vgpu_mmio(struct intel_vgpu *vgpu) return 0; } -static void setup_vgpu_cfg_space(struct intel_vgpu *vgpu, - struct intel_vgpu_creation_params *param) -{ - struct intel_gvt *gvt = vgpu->gvt; - const struct intel_gvt_device_info *info = &gvt->device_info; - u16 *gmch_ctl; - int i; - - memcpy(vgpu_cfg_space(vgpu), gvt->firmware.cfg_space, - info->cfg_space_size); - - if (!param->primary) { - vgpu_cfg_space(vgpu)[PCI_CLASS_DEVICE] = - INTEL_GVT_PCI_CLASS_VGA_OTHER; - vgpu_cfg_space(vgpu)[PCI_CLASS_PROG] = - INTEL_GVT_PCI_CLASS_VGA_OTHER; - } - - /* Show guest that there isn't any stolen memory.*/ - gmch_ctl = (u16 *)(vgpu_cfg_space(vgpu) + INTEL_GVT_PCI_GMCH_CONTROL); - *gmch_ctl &= ~(BDW_GMCH_GMS_MASK << BDW_GMCH_GMS_SHIFT); - - intel_vgpu_write_pci_bar(vgpu, PCI_BASE_ADDRESS_2, - gvt_aperture_pa_base(gvt), true); - - vgpu_cfg_space(vgpu)[PCI_COMMAND] &= ~(PCI_COMMAND_IO - | PCI_COMMAND_MEMORY - | PCI_COMMAND_MASTER); - /* - * Clear the bar upper 32bit and let guest to assign the new value - */ - memset(vgpu_cfg_space(vgpu) + PCI_BASE_ADDRESS_1, 0, 4); - memset(vgpu_cfg_space(vgpu) + PCI_BASE_ADDRESS_3, 0, 4); - memset(vgpu_cfg_space(vgpu) + INTEL_GVT_PCI_OPREGION, 0, 4); - - for (i = 0; i < INTEL_GVT_MAX_BAR_NUM; i++) { - vgpu->cfg_space.bar[i].size = pci_resource_len( - gvt->dev_priv->drm.pdev, i * 2); - vgpu->cfg_space.bar[i].tracked = false; - } -} - void populate_pvinfo_page(struct intel_vgpu *vgpu) { /* setup the ballooning information */ @@ -300,7 +258,7 @@ static struct intel_vgpu *__intel_gvt_create_vgpu(struct intel_gvt *gvt, vgpu->gvt = gvt; bitmap_zero(vgpu->tlb_handle_pending, I915_NUM_ENGINES); - setup_vgpu_cfg_space(vgpu, param); + intel_vgpu_init_cfg_space(vgpu, param->primary); ret = setup_vgpu_mmio(vgpu); if (ret) From c64ff6c774413fdbffd7f0f3ef5b04127d461cf4 Mon Sep 17 00:00:00 2001 From: Changbin Du Date: Fri, 13 Jan 2017 11:15:59 +0800 Subject: [PATCH 20/26] drm/i915/gvt: introduce intel_vgpu_reset_cfg_space to reset configuration space This patch introduces a new function intel_vgpu_reset_cfg_space() to reset vGPU configuration space. This function will unmap gttmmio and aperture if they are mapped before. Then entire cfg space will be restored to default values. Currently we only do such reset when vGPU is not owned by any VM so we simply restore entire cfg space to default value, not following the PCIe FLR spec that some fields should remain unchanged. Signed-off-by: Changbin Du Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/cfg_space.c | 25 +++++++++++++++++++++++++ drivers/gpu/drm/i915/gvt/gvt.h | 2 ++ 2 files changed, 27 insertions(+) diff --git a/drivers/gpu/drm/i915/gvt/cfg_space.c b/drivers/gpu/drm/i915/gvt/cfg_space.c index 82f449d59884..4a6a2ed65732 100644 --- a/drivers/gpu/drm/i915/gvt/cfg_space.c +++ b/drivers/gpu/drm/i915/gvt/cfg_space.c @@ -331,3 +331,28 @@ void intel_vgpu_init_cfg_space(struct intel_vgpu *vgpu, vgpu->cfg_space.bar[i].tracked = false; } } + +/** + * intel_vgpu_reset_cfg_space - reset vGPU configuration space + * + * @vgpu: a vGPU + * + */ +void intel_vgpu_reset_cfg_space(struct intel_vgpu *vgpu) +{ + u8 cmd = vgpu_cfg_space(vgpu)[PCI_COMMAND]; + bool primary = vgpu_cfg_space(vgpu)[PCI_CLASS_DEVICE] != + INTEL_GVT_PCI_CLASS_VGA_OTHER; + + if (cmd & PCI_COMMAND_MEMORY) { + trap_gttmmio(vgpu, false); + map_aperture(vgpu, false); + } + + /** + * Currently we only do such reset when vGPU is not + * owned by any VM, so we simply restore entire cfg + * space to default value. + */ + intel_vgpu_init_cfg_space(vgpu, primary); +} diff --git a/drivers/gpu/drm/i915/gvt/gvt.h b/drivers/gpu/drm/i915/gvt/gvt.h index ab505552d313..9c97d7d36449 100644 --- a/drivers/gpu/drm/i915/gvt/gvt.h +++ b/drivers/gpu/drm/i915/gvt/gvt.h @@ -414,6 +414,8 @@ int intel_gvt_ggtt_h2g_index(struct intel_vgpu *vgpu, unsigned long h_index, void intel_vgpu_init_cfg_space(struct intel_vgpu *vgpu, bool primary); +void intel_vgpu_reset_cfg_space(struct intel_vgpu *vgpu); + int intel_vgpu_emulate_cfg_read(struct intel_vgpu *vgpu, unsigned int offset, void *p_data, unsigned int bytes); From cdcc43479c9b929940a1955d2e7bae696d2b9496 Mon Sep 17 00:00:00 2001 From: Changbin Du Date: Fri, 13 Jan 2017 11:16:00 +0800 Subject: [PATCH 21/26] drm/i915/gvt: move mmio init/clean function to mmio.c Move the mmio space inititation function setup_vgpu_mmio() and cleanup function clean_vgpu_mmio() in vgpu.c to dedicated source file mmio.c, and rename them as intel_vgpu_init_mmio() and intel_vgpu_clean_mmio() respectively. Signed-off-by: Changbin Du Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/gvt.h | 1 - drivers/gpu/drm/i915/gvt/handlers.c | 2 +- drivers/gpu/drm/i915/gvt/mmio.c | 42 +++++++++++++++++++++++++++++ drivers/gpu/drm/i915/gvt/mmio.h | 3 +++ drivers/gpu/drm/i915/gvt/vgpu.c | 37 +++---------------------- 5 files changed, 49 insertions(+), 36 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/gvt.h b/drivers/gpu/drm/i915/gvt/gvt.h index 9c97d7d36449..9832e7f2a662 100644 --- a/drivers/gpu/drm/i915/gvt/gvt.h +++ b/drivers/gpu/drm/i915/gvt/gvt.h @@ -429,7 +429,6 @@ void intel_vgpu_clean_opregion(struct intel_vgpu *vgpu); int intel_vgpu_init_opregion(struct intel_vgpu *vgpu, u32 gpa); int intel_vgpu_emulate_opregion_request(struct intel_vgpu *vgpu, u32 swsci); -int setup_vgpu_mmio(struct intel_vgpu *vgpu); void populate_pvinfo_page(struct intel_vgpu *vgpu); struct intel_gvt_ops { diff --git a/drivers/gpu/drm/i915/gvt/handlers.c b/drivers/gpu/drm/i915/gvt/handlers.c index 8cbaf1c83720..fd96898db71d 100644 --- a/drivers/gpu/drm/i915/gvt/handlers.c +++ b/drivers/gpu/drm/i915/gvt/handlers.c @@ -257,7 +257,7 @@ static int handle_device_reset(struct intel_vgpu *vgpu, unsigned int offset, mutex_unlock(&vgpu->gvt->lock); intel_vgpu_clean_gtt(vgpu); mutex_lock(&vgpu->gvt->lock); - setup_vgpu_mmio(vgpu); + intel_vgpu_init_mmio(vgpu); populate_pvinfo_page(vgpu); intel_vgpu_init_gtt(vgpu); } diff --git a/drivers/gpu/drm/i915/gvt/mmio.c b/drivers/gpu/drm/i915/gvt/mmio.c index e60701397ac2..f7da735b7919 100644 --- a/drivers/gpu/drm/i915/gvt/mmio.c +++ b/drivers/gpu/drm/i915/gvt/mmio.c @@ -303,3 +303,45 @@ err: mutex_unlock(&gvt->lock); return ret; } + +/** + * intel_vgpu_init_mmio - init MMIO space + * @vgpu: a vGPU + * + * Returns: + * Zero on success, negative error code if failed + */ +int intel_vgpu_init_mmio(struct intel_vgpu *vgpu) +{ + const struct intel_gvt_device_info *info = &vgpu->gvt->device_info; + + if (vgpu->mmio.vreg) + memset(vgpu->mmio.vreg, 0, info->mmio_size * 2); + else { + vgpu->mmio.vreg = vzalloc(info->mmio_size * 2); + if (!vgpu->mmio.vreg) + return -ENOMEM; + } + vgpu->mmio.sreg = vgpu->mmio.vreg + info->mmio_size; + + memcpy(vgpu->mmio.vreg, vgpu->gvt->firmware.mmio, info->mmio_size); + memcpy(vgpu->mmio.sreg, vgpu->gvt->firmware.mmio, info->mmio_size); + + vgpu_vreg(vgpu, GEN6_GT_THREAD_STATUS_REG) = 0; + + /* set the bit 0:2(Core C-State ) to C0 */ + vgpu_vreg(vgpu, GEN6_GT_CORE_STATUS) = 0; + + return 0; +} + +/** + * intel_vgpu_clean_mmio - clean MMIO space + * @vgpu: a vGPU + * + */ +void intel_vgpu_clean_mmio(struct intel_vgpu *vgpu) +{ + vfree(vgpu->mmio.vreg); + vgpu->mmio.vreg = vgpu->mmio.sreg = NULL; +} diff --git a/drivers/gpu/drm/i915/gvt/mmio.h b/drivers/gpu/drm/i915/gvt/mmio.h index 87d5b5e366a3..1b74ac9650d5 100644 --- a/drivers/gpu/drm/i915/gvt/mmio.h +++ b/drivers/gpu/drm/i915/gvt/mmio.h @@ -86,6 +86,9 @@ struct intel_gvt_mmio_info *intel_gvt_find_mmio_info(struct intel_gvt *gvt, *offset; \ }) +int intel_vgpu_init_mmio(struct intel_vgpu *vgpu); +void intel_vgpu_clean_mmio(struct intel_vgpu *vgpu); + int intel_vgpu_gpa_to_mmio_offset(struct intel_vgpu *vgpu, u64 gpa); int intel_vgpu_emulate_mmio_read(struct intel_vgpu *vgpu, u64 pa, diff --git a/drivers/gpu/drm/i915/gvt/vgpu.c b/drivers/gpu/drm/i915/gvt/vgpu.c index 35c274177da8..958c3ed1c6c6 100644 --- a/drivers/gpu/drm/i915/gvt/vgpu.c +++ b/drivers/gpu/drm/i915/gvt/vgpu.c @@ -35,37 +35,6 @@ #include "gvt.h" #include "i915_pvinfo.h" -static void clean_vgpu_mmio(struct intel_vgpu *vgpu) -{ - vfree(vgpu->mmio.vreg); - vgpu->mmio.vreg = vgpu->mmio.sreg = NULL; -} - -int setup_vgpu_mmio(struct intel_vgpu *vgpu) -{ - struct intel_gvt *gvt = vgpu->gvt; - const struct intel_gvt_device_info *info = &gvt->device_info; - - if (vgpu->mmio.vreg) - memset(vgpu->mmio.vreg, 0, info->mmio_size * 2); - else { - vgpu->mmio.vreg = vzalloc(info->mmio_size * 2); - if (!vgpu->mmio.vreg) - return -ENOMEM; - } - - vgpu->mmio.sreg = vgpu->mmio.vreg + info->mmio_size; - - memcpy(vgpu->mmio.vreg, gvt->firmware.mmio, info->mmio_size); - memcpy(vgpu->mmio.sreg, gvt->firmware.mmio, info->mmio_size); - - vgpu_vreg(vgpu, GEN6_GT_THREAD_STATUS_REG) = 0; - - /* set the bit 0:2(Core C-State ) to C0 */ - vgpu_vreg(vgpu, GEN6_GT_CORE_STATUS) = 0; - return 0; -} - void populate_pvinfo_page(struct intel_vgpu *vgpu) { /* setup the ballooning information */ @@ -226,7 +195,7 @@ void intel_gvt_destroy_vgpu(struct intel_vgpu *vgpu) intel_vgpu_clean_gtt(vgpu); intel_gvt_hypervisor_detach_vgpu(vgpu); intel_vgpu_free_resource(vgpu); - clean_vgpu_mmio(vgpu); + intel_vgpu_clean_mmio(vgpu); vfree(vgpu); intel_gvt_update_vgpu_types(gvt); @@ -260,7 +229,7 @@ static struct intel_vgpu *__intel_gvt_create_vgpu(struct intel_gvt *gvt, intel_vgpu_init_cfg_space(vgpu, param->primary); - ret = setup_vgpu_mmio(vgpu); + ret = intel_vgpu_init_mmio(vgpu); if (ret) goto out_clean_idr; @@ -312,7 +281,7 @@ out_detach_hypervisor_vgpu: out_clean_vgpu_resource: intel_vgpu_free_resource(vgpu); out_clean_vgpu_mmio: - clean_vgpu_mmio(vgpu); + intel_vgpu_clean_mmio(vgpu); out_clean_idr: idr_remove(&gvt->vgpu_idr, vgpu->id); out_free_vgpu: From 97d58f7dd0ff12e5fddeffb40aed845daa628149 Mon Sep 17 00:00:00 2001 From: Changbin Du Date: Fri, 13 Jan 2017 11:16:01 +0800 Subject: [PATCH 22/26] drm/i915/gvt: introduce intel_vgpu_reset_mmio() to reset mmio space This patch introduces a new function intel_vgpu_reset_mmio() to reset vGPU MMIO space (virtual registers of the vGPU). The default values are loaded as firmware during gvt inititiation. Signed-off-by: Changbin Du Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/mmio.c | 39 +++++++++++++++++++++------------ drivers/gpu/drm/i915/gvt/mmio.h | 1 + 2 files changed, 26 insertions(+), 14 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/mmio.c b/drivers/gpu/drm/i915/gvt/mmio.c index f7da735b7919..4df078bc5d04 100644 --- a/drivers/gpu/drm/i915/gvt/mmio.c +++ b/drivers/gpu/drm/i915/gvt/mmio.c @@ -304,6 +304,26 @@ err: return ret; } + +/** + * intel_vgpu_reset_mmio - reset virtual MMIO space + * @vgpu: a vGPU + * + */ +void intel_vgpu_reset_mmio(struct intel_vgpu *vgpu) +{ + struct intel_gvt *gvt = vgpu->gvt; + const struct intel_gvt_device_info *info = &gvt->device_info; + + memcpy(vgpu->mmio.vreg, gvt->firmware.mmio, info->mmio_size); + memcpy(vgpu->mmio.sreg, gvt->firmware.mmio, info->mmio_size); + + vgpu_vreg(vgpu, GEN6_GT_THREAD_STATUS_REG) = 0; + + /* set the bit 0:2(Core C-State ) to C0 */ + vgpu_vreg(vgpu, GEN6_GT_CORE_STATUS) = 0; +} + /** * intel_vgpu_init_mmio - init MMIO space * @vgpu: a vGPU @@ -315,22 +335,13 @@ int intel_vgpu_init_mmio(struct intel_vgpu *vgpu) { const struct intel_gvt_device_info *info = &vgpu->gvt->device_info; - if (vgpu->mmio.vreg) - memset(vgpu->mmio.vreg, 0, info->mmio_size * 2); - else { - vgpu->mmio.vreg = vzalloc(info->mmio_size * 2); - if (!vgpu->mmio.vreg) - return -ENOMEM; - } + vgpu->mmio.vreg = vzalloc(info->mmio_size * 2); + if (!vgpu->mmio.vreg) + return -ENOMEM; + vgpu->mmio.sreg = vgpu->mmio.vreg + info->mmio_size; - memcpy(vgpu->mmio.vreg, vgpu->gvt->firmware.mmio, info->mmio_size); - memcpy(vgpu->mmio.sreg, vgpu->gvt->firmware.mmio, info->mmio_size); - - vgpu_vreg(vgpu, GEN6_GT_THREAD_STATUS_REG) = 0; - - /* set the bit 0:2(Core C-State ) to C0 */ - vgpu_vreg(vgpu, GEN6_GT_CORE_STATUS) = 0; + intel_vgpu_reset_mmio(vgpu); return 0; } diff --git a/drivers/gpu/drm/i915/gvt/mmio.h b/drivers/gpu/drm/i915/gvt/mmio.h index 1b74ac9650d5..3bc620f56f35 100644 --- a/drivers/gpu/drm/i915/gvt/mmio.h +++ b/drivers/gpu/drm/i915/gvt/mmio.h @@ -87,6 +87,7 @@ struct intel_gvt_mmio_info *intel_gvt_find_mmio_info(struct intel_gvt *gvt, }) int intel_vgpu_init_mmio(struct intel_vgpu *vgpu); +void intel_vgpu_reset_mmio(struct intel_vgpu *vgpu); void intel_vgpu_clean_mmio(struct intel_vgpu *vgpu); int intel_vgpu_gpa_to_mmio_offset(struct intel_vgpu *vgpu, u64 gpa); From cfe65f4037cedb911a840ebcf6dafc5b69e535b4 Mon Sep 17 00:00:00 2001 From: Changbin Du Date: Fri, 13 Jan 2017 11:16:02 +0800 Subject: [PATCH 23/26] drm/i915/gvt: fix vGPU instance reuse issues by vGPU reset function Our function tests found several issues related to reusing vGPU instance. They are qemu reboot failure, guest tdr after reboot, host hang when reboot guest. All these issues are caused by dirty status inherited from last VM. This patch fix all these issues by resetting a virtual GPU before VM use it. The reset logical is put into a low level function _intel_gvt_reset_vgpu(), which supports Device Model Level Reset, Full GT Reset and Per-Engine Reset. vGPU Device Model Level Reset (DMLR) simulates the PCI reset to reset the whole vGPU to default state as when it is created, including GTT, execlist, scratch pages, cfg space, mmio space, pvinfo page, scheduler and fence registers. The ultimate goal of vGPU DMLR is that reuse a vGPU instance by different virtual machines. When we reassign a vGPU to a virtual machine we must issue such reset first. Full GT Reset and Per-Engine GT Reset are soft reset flow for GPU engines (Render, Blitter, Video, Video Enhancement). It is defined by GPU Spec. Unlike the FLR, GT reset only reset particular resource of a vGPU per the reset request. Guest driver can issue a GT reset by programming the virtual GDRST register to reset specific virtual GPU engine or all engines. Since vGPU DMLR and GT reset can share some code so we implement both these two into one single function intel_gvt_reset_vgpu_locked(). The parameter dmlr is to identify if we will do FLR or GT reset. The parameter engine_mask is to specific the engines that need to be resetted. If value ALL_ENGINES is given for engine_mask, it means the caller requests a full gt reset that we will reset all virtual GPU engines. Signed-off-by: Changbin Du Reviewed-by: Jike Song Reviewed-by: Kevin Tian Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/gvt.h | 2 + drivers/gpu/drm/i915/gvt/vgpu.c | 73 ++++++++++++++++++++++++++++++++- 2 files changed, 74 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/gvt/gvt.h b/drivers/gpu/drm/i915/gvt/gvt.h index 9832e7f2a662..e227caf5859e 100644 --- a/drivers/gpu/drm/i915/gvt/gvt.h +++ b/drivers/gpu/drm/i915/gvt/gvt.h @@ -376,6 +376,8 @@ void intel_gvt_clean_vgpu_types(struct intel_gvt *gvt); struct intel_vgpu *intel_gvt_create_vgpu(struct intel_gvt *gvt, struct intel_vgpu_type *type); void intel_gvt_destroy_vgpu(struct intel_vgpu *vgpu); +void intel_gvt_reset_vgpu_locked(struct intel_vgpu *vgpu, bool dmlr, + unsigned int engine_mask); void intel_gvt_reset_vgpu(struct intel_vgpu *vgpu); diff --git a/drivers/gpu/drm/i915/gvt/vgpu.c b/drivers/gpu/drm/i915/gvt/vgpu.c index 958c3ed1c6c6..7295bc8e12fb 100644 --- a/drivers/gpu/drm/i915/gvt/vgpu.c +++ b/drivers/gpu/drm/i915/gvt/vgpu.c @@ -327,7 +327,75 @@ struct intel_vgpu *intel_gvt_create_vgpu(struct intel_gvt *gvt, } /** - * intel_gvt_reset_vgpu - reset a virtual GPU + * intel_gvt_reset_vgpu_locked - reset a virtual GPU by DMLR or GT reset + * @vgpu: virtual GPU + * @dmlr: vGPU Device Model Level Reset or GT Reset + * @engine_mask: engines to reset for GT reset + * + * This function is called when user wants to reset a virtual GPU through + * device model reset or GT reset. The caller should hold the gvt lock. + * + * vGPU Device Model Level Reset (DMLR) simulates the PCI level reset to reset + * the whole vGPU to default state as when it is created. This vGPU function + * is required both for functionary and security concerns.The ultimate goal + * of vGPU FLR is that reuse a vGPU instance by virtual machines. When we + * assign a vGPU to a virtual machine we must isse such reset first. + * + * Full GT Reset and Per-Engine GT Reset are soft reset flow for GPU engines + * (Render, Blitter, Video, Video Enhancement). It is defined by GPU Spec. + * Unlike the FLR, GT reset only reset particular resource of a vGPU per + * the reset request. Guest driver can issue a GT reset by programming the + * virtual GDRST register to reset specific virtual GPU engine or all + * engines. + * + * The parameter dev_level is to identify if we will do DMLR or GT reset. + * The parameter engine_mask is to specific the engines that need to be + * resetted. If value ALL_ENGINES is given for engine_mask, it means + * the caller requests a full GT reset that we will reset all virtual + * GPU engines. For FLR, engine_mask is ignored. + */ +void intel_gvt_reset_vgpu_locked(struct intel_vgpu *vgpu, bool dmlr, + unsigned int engine_mask) +{ + struct intel_gvt *gvt = vgpu->gvt; + struct intel_gvt_workload_scheduler *scheduler = &gvt->scheduler; + + gvt_dbg_core("------------------------------------------\n"); + gvt_dbg_core("resseting vgpu%d, dmlr %d, engine_mask %08x\n", + vgpu->id, dmlr, engine_mask); + vgpu->resetting = true; + + intel_vgpu_stop_schedule(vgpu); + /* + * The current_vgpu will set to NULL after stopping the + * scheduler when the reset is triggered by current vgpu. + */ + if (scheduler->current_vgpu == NULL) { + mutex_unlock(&gvt->lock); + intel_gvt_wait_vgpu_idle(vgpu); + mutex_lock(&gvt->lock); + } + + intel_vgpu_reset_execlist(vgpu, dmlr ? ALL_ENGINES : engine_mask); + + /* full GPU reset or device model level reset */ + if (engine_mask == ALL_ENGINES || dmlr) { + intel_vgpu_reset_gtt(vgpu, dmlr); + intel_vgpu_reset_resource(vgpu); + intel_vgpu_reset_mmio(vgpu); + populate_pvinfo_page(vgpu); + + if (dmlr) + intel_vgpu_reset_cfg_space(vgpu); + } + + vgpu->resetting = false; + gvt_dbg_core("reset vgpu%d done\n", vgpu->id); + gvt_dbg_core("------------------------------------------\n"); +} + +/** + * intel_gvt_reset_vgpu - reset a virtual GPU (Function Level) * @vgpu: virtual GPU * * This function is called when user wants to reset a virtual GPU. @@ -335,4 +403,7 @@ struct intel_vgpu *intel_gvt_create_vgpu(struct intel_gvt *gvt, */ void intel_gvt_reset_vgpu(struct intel_vgpu *vgpu) { + mutex_lock(&vgpu->gvt->lock); + intel_gvt_reset_vgpu_locked(vgpu, true, 0); + mutex_unlock(&vgpu->gvt->lock); } From c34eaa8d0f9d9ae26a4a6af7bc3aca57310cf483 Mon Sep 17 00:00:00 2001 From: Changbin Du Date: Fri, 13 Jan 2017 11:16:03 +0800 Subject: [PATCH 24/26] drm/i915/gvt: rewrite gt reset handler using new function intel_gvt_reset_vgpu_locked GT reset and FLR share some operations and they are both implemented in our new function intel_gvt_reset_vgpu_locked(). This patch rewrite the gt reset handler using this new function. Besides, this new implementation fixed the old issue in GT reset. The old implementation reset GGTT entries which is illegal. We only clear GGTT entries at PCI level reset. Signed-off-by: Changbin Du Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/handlers.c | 90 ++++++++++------------------- 1 file changed, 29 insertions(+), 61 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/handlers.c b/drivers/gpu/drm/i915/gvt/handlers.c index fd96898db71d..ab2ea157da4c 100644 --- a/drivers/gpu/drm/i915/gvt/handlers.c +++ b/drivers/gpu/drm/i915/gvt/handlers.c @@ -231,77 +231,45 @@ static int mul_force_wake_write(struct intel_vgpu *vgpu, return 0; } -static int handle_device_reset(struct intel_vgpu *vgpu, unsigned int offset, - void *p_data, unsigned int bytes, unsigned long bitmap) -{ - struct intel_gvt_workload_scheduler *scheduler = - &vgpu->gvt->scheduler; - - vgpu->resetting = true; - - intel_vgpu_stop_schedule(vgpu); - /* - * The current_vgpu will set to NULL after stopping the - * scheduler when the reset is triggered by current vgpu. - */ - if (scheduler->current_vgpu == NULL) { - mutex_unlock(&vgpu->gvt->lock); - intel_gvt_wait_vgpu_idle(vgpu); - mutex_lock(&vgpu->gvt->lock); - } - - intel_vgpu_reset_execlist(vgpu, bitmap); - - /* full GPU reset */ - if (bitmap == 0xff) { - mutex_unlock(&vgpu->gvt->lock); - intel_vgpu_clean_gtt(vgpu); - mutex_lock(&vgpu->gvt->lock); - intel_vgpu_init_mmio(vgpu); - populate_pvinfo_page(vgpu); - intel_vgpu_init_gtt(vgpu); - } - - vgpu->resetting = false; - - return 0; -} - static int gdrst_mmio_write(struct intel_vgpu *vgpu, unsigned int offset, - void *p_data, unsigned int bytes) + void *p_data, unsigned int bytes) { + unsigned int engine_mask = 0; u32 data; - u64 bitmap = 0; write_vreg(vgpu, offset, p_data, bytes); data = vgpu_vreg(vgpu, offset); if (data & GEN6_GRDOM_FULL) { gvt_dbg_mmio("vgpu%d: request full GPU reset\n", vgpu->id); - bitmap = 0xff; + engine_mask = ALL_ENGINES; + } else { + if (data & GEN6_GRDOM_RENDER) { + gvt_dbg_mmio("vgpu%d: request RCS reset\n", vgpu->id); + engine_mask |= (1 << RCS); + } + if (data & GEN6_GRDOM_MEDIA) { + gvt_dbg_mmio("vgpu%d: request VCS reset\n", vgpu->id); + engine_mask |= (1 << VCS); + } + if (data & GEN6_GRDOM_BLT) { + gvt_dbg_mmio("vgpu%d: request BCS Reset\n", vgpu->id); + engine_mask |= (1 << BCS); + } + if (data & GEN6_GRDOM_VECS) { + gvt_dbg_mmio("vgpu%d: request VECS Reset\n", vgpu->id); + engine_mask |= (1 << VECS); + } + if (data & GEN8_GRDOM_MEDIA2) { + gvt_dbg_mmio("vgpu%d: request VCS2 Reset\n", vgpu->id); + if (HAS_BSD2(vgpu->gvt->dev_priv)) + engine_mask |= (1 << VCS2); + } } - if (data & GEN6_GRDOM_RENDER) { - gvt_dbg_mmio("vgpu%d: request RCS reset\n", vgpu->id); - bitmap |= (1 << RCS); - } - if (data & GEN6_GRDOM_MEDIA) { - gvt_dbg_mmio("vgpu%d: request VCS reset\n", vgpu->id); - bitmap |= (1 << VCS); - } - if (data & GEN6_GRDOM_BLT) { - gvt_dbg_mmio("vgpu%d: request BCS Reset\n", vgpu->id); - bitmap |= (1 << BCS); - } - if (data & GEN6_GRDOM_VECS) { - gvt_dbg_mmio("vgpu%d: request VECS Reset\n", vgpu->id); - bitmap |= (1 << VECS); - } - if (data & GEN8_GRDOM_MEDIA2) { - gvt_dbg_mmio("vgpu%d: request VCS2 Reset\n", vgpu->id); - if (HAS_BSD2(vgpu->gvt->dev_priv)) - bitmap |= (1 << VCS2); - } - return handle_device_reset(vgpu, offset, p_data, bytes, bitmap); + + intel_gvt_reset_vgpu_locked(vgpu, false, engine_mask); + + return 0; } static int gmbus_mmio_read(struct intel_vgpu *vgpu, unsigned int offset, From 4fc020d864647ea3ae8cb8f17d63e48e87ebd0bf Mon Sep 17 00:00:00 2001 From: Francisco Jerez Date: Thu, 12 Jan 2017 12:44:54 +0200 Subject: [PATCH 25/26] drm/i915: Remove WaDisableLSQCROPERFforOCL KBL workaround. The WaDisableLSQCROPERFforOCL workaround has the side effect of disabling an L3SQ optimization that has huge performance implications and is unlikely to be necessary for the correct functioning of usual graphic workloads. Userspace is free to re-enable the workaround on demand, and is generally in a better position to determine whether the workaround is necessary than the DRM is (e.g. only during the execution of compute kernels that rely on both L3 fences and HDC R/W requests). The same workaround seems to apply to BDW (at least to production stepping G1) and SKL as well (the internal workaround database claims that it does for all steppings, while the BSpec workaround table only mentions pre-production steppings), but the DRM doesn't do anything beyond whitelisting the L3SQCREG4 register so userspace can enable it when it sees fit. Do the same on KBL platforms. Improves performance of the GFXBench4 gl_manhattan31 benchmark by 60%, and gl_4 (AKA car chase) by 14% on a KBL GT2 running Mesa master -- This is followed by a regression of 35% and 10% respectively for the same benchmarks and platform caused by my recent patch series switching userspace to use the dataport constant cache instead of the sampler to implement uniform pull constant loads, which caused us to hit more heavily the L3 cache (and on platforms other than KBL had the opposite effect of improving performance of the same two benchmarks). The overall effect on KBL of this change combined with the recent userspace change is respectively 4.6% and 2.6%. SynMark2 OglShMapPcf was affected by the constant cache changes (though it improved as it did on other platforms rather than regressing), but is not significantly affected by this patch (with statistical significance of 5% and sample size 20). v2: Drop some more code to avoid unused variable warning. Fixes: 738fa1b3123f ("drm/i915/kbl: Add WaDisableLSQCROPERFforOCL") Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=99256 Signed-off-by: Francisco Jerez Cc: Matthew Auld Cc: Eero Tamminen Cc: Jani Nikula Cc: Mika Kuoppala Cc: beignet@lists.freedesktop.org Cc: # v4.7+ Reviewed-by: Mika Kuoppala [Removed double Fixes tag] Signed-off-by: Mika Kuoppala Link: http://patchwork.freedesktop.org/patch/msgid/1484217894-20505-1-git-send-email-mika.kuoppala@intel.com (cherry picked from commit 8726f2faa371514fba2f594d799db95203dfeee0) Signed-off-by: Jani Nikula --- drivers/gpu/drm/i915/intel_lrc.c | 10 ---------- drivers/gpu/drm/i915/intel_ringbuffer.c | 8 -------- 2 files changed, 18 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c index d4961fa20c73..beabc17e7c8a 100644 --- a/drivers/gpu/drm/i915/intel_lrc.c +++ b/drivers/gpu/drm/i915/intel_lrc.c @@ -979,18 +979,8 @@ static inline int gen8_emit_flush_coherentl3_wa(struct intel_engine_cs *engine, uint32_t *batch, uint32_t index) { - struct drm_i915_private *dev_priv = engine->i915; uint32_t l3sqc4_flush = (0x40400000 | GEN8_LQSC_FLUSH_COHERENT_LINES); - /* - * WaDisableLSQCROPERFforOCL:kbl - * This WA is implemented in skl_init_clock_gating() but since - * this batch updates GEN8_L3SQCREG4 with default value we need to - * set this bit here to retain the WA during flush. - */ - if (IS_KBL_REVID(dev_priv, 0, KBL_REVID_E0)) - l3sqc4_flush |= GEN8_LQSC_RO_PERF_DIS; - wa_ctx_emit(batch, index, (MI_STORE_REGISTER_MEM_GEN8 | MI_SRM_LRM_GLOBAL_GTT)); wa_ctx_emit_reg(batch, index, GEN8_L3SQCREG4); diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c index aeb637dc1fdf..91cb4c422ad5 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.c +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c @@ -1095,14 +1095,6 @@ static int kbl_init_workarounds(struct intel_engine_cs *engine) WA_SET_BIT_MASKED(HDC_CHICKEN0, HDC_FENCE_DEST_SLM_DISABLE); - /* GEN8_L3SQCREG4 has a dependency with WA batch so any new changes - * involving this register should also be added to WA batch as required. - */ - if (IS_KBL_REVID(dev_priv, 0, KBL_REVID_E0)) - /* WaDisableLSQCROPERFforOCL:kbl */ - I915_WRITE(GEN8_L3SQCREG4, I915_READ(GEN8_L3SQCREG4) | - GEN8_LQSC_RO_PERF_DIS); - /* WaToEnableHwFixForPushConstHWBug:kbl */ if (IS_KBL_REVID(dev_priv, KBL_REVID_C0, REVID_FOREVER)) WA_SET_BIT_MASKED(COMMON_SLICE_CHICKEN2, From 3bfdfdcbce2796ce75bf2d85fd8471858d702e5d Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Mon, 7 Nov 2016 22:20:54 +0200 Subject: [PATCH 26/26] drm/i915: Ignore bogus plane coordinates on SKL when the plane is not visible MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit When the plane is invisible we may have all sorts of bogus stuff in the coordinates, which we must ignore or else we might fail the plane update. This started to happen on SKL when I moved the plane offset computation to happen in the check phase. Previously we happily ignored it all since we never called the update_plane hook with an invisible plane. Cc: Sivakumar Thulasimani Cc: drm-intel-fixes@lists.freedesktop.org Fixes: b63a16f6cd89 ("drm/i915: Compute display surface offset in the plane check hook for SKL+") Signed-off-by: Ville Syrjälä Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=98258 Testcase: igt/pm_rpm/legacy-planes Testcase: igt/pm_rpm/universal-planes Reviewed-by: Matt Roper Signed-off-by: Matt Roper Link: http://patchwork.freedesktop.org/patch/msgid/1478550057-24864-3-git-send-email-ville.syrjala@linux.intel.com (cherry picked from commit a5e4c7d0aa6784d8abe95c3ceef0da9656d17468) Signed-off-by: Jani Nikula --- drivers/gpu/drm/i915/intel_display.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 3dc8724df400..8d702cf1a616 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -2967,6 +2967,9 @@ int skl_check_plane_surface(struct intel_plane_state *plane_state) unsigned int rotation = plane_state->base.rotation; int ret; + if (!plane_state->base.visible) + return 0; + /* Rotate src coordinates to match rotated GTT view */ if (drm_rotation_90_or_270(rotation)) drm_rect_rotate(&plane_state->base.src,