linux/drivers/gpu/drm/i915/gt/intel_ppgtt.c

324 lines
6.7 KiB
C
Raw Permalink Normal View History

// SPDX-License-Identifier: MIT
/*
* Copyright © 2020 Intel Corporation
*/
#include <linux/slab.h>
#include "gem/i915_gem_lmem.h"
#include "i915_trace.h"
#include "intel_gt.h"
#include "intel_gtt.h"
#include "gen6_ppgtt.h"
#include "gen8_ppgtt.h"
struct i915_page_table *alloc_pt(struct i915_address_space *vm, int sz)
{
struct i915_page_table *pt;
pt = kmalloc(sizeof(*pt), I915_GFP_ALLOW_FAIL);
if (unlikely(!pt))
return ERR_PTR(-ENOMEM);
pt->base = vm->alloc_pt_dma(vm, sz);
if (IS_ERR(pt->base)) {
kfree(pt);
return ERR_PTR(-ENOMEM);
}
pt->is_compact = false;
atomic_set(&pt->used, 0);
return pt;
}
struct i915_page_directory *__alloc_pd(int count)
{
struct i915_page_directory *pd;
pd = kzalloc(sizeof(*pd), I915_GFP_ALLOW_FAIL);
if (unlikely(!pd))
return NULL;
pd->entry = kcalloc(count, sizeof(*pd->entry), I915_GFP_ALLOW_FAIL);
if (unlikely(!pd->entry)) {
kfree(pd);
return NULL;
}
spin_lock_init(&pd->lock);
return pd;
}
struct i915_page_directory *alloc_pd(struct i915_address_space *vm)
{
struct i915_page_directory *pd;
pd = __alloc_pd(I915_PDES);
if (unlikely(!pd))
return ERR_PTR(-ENOMEM);
pd->pt.base = vm->alloc_pt_dma(vm, I915_GTT_PAGE_SIZE_4K);
if (IS_ERR(pd->pt.base)) {
kfree(pd->entry);
kfree(pd);
return ERR_PTR(-ENOMEM);
}
return pd;
}
void free_px(struct i915_address_space *vm, struct i915_page_table *pt, int lvl)
{
BUILD_BUG_ON(offsetof(struct i915_page_directory, pt));
if (lvl) {
struct i915_page_directory *pd =
container_of(pt, typeof(*pd), pt);
kfree(pd->entry);
}
if (pt->base)
i915_gem_object_put(pt->base);
kfree(pt);
}
static void
write_dma_entry(struct drm_i915_gem_object * const pdma,
const unsigned short idx,
const u64 encoded_entry)
{
u64 * const vaddr = __px_vaddr(pdma);
vaddr[idx] = encoded_entry;
drm_clflush_virt_range(&vaddr[idx], sizeof(u64));
}
void
__set_pd_entry(struct i915_page_directory * const pd,
const unsigned short idx,
struct i915_page_table * const to,
u64 (*encode)(const dma_addr_t, const enum i915_cache_level))
{
/* Each thread pre-pins the pd, and we may have a thread per pde. */
GEM_BUG_ON(atomic_read(px_used(pd)) > NALLOC * I915_PDES);
atomic_inc(px_used(pd));
pd->entry[idx] = to;
write_dma_entry(px_base(pd), idx, encode(px_dma(to), I915_CACHE_LLC));
}
void
clear_pd_entry(struct i915_page_directory * const pd,
const unsigned short idx,
const struct drm_i915_gem_object * const scratch)
{
GEM_BUG_ON(atomic_read(px_used(pd)) == 0);
write_dma_entry(px_base(pd), idx, scratch->encode);
pd->entry[idx] = NULL;
atomic_dec(px_used(pd));
}
bool
release_pd_entry(struct i915_page_directory * const pd,
const unsigned short idx,
struct i915_page_table * const pt,
const struct drm_i915_gem_object * const scratch)
{
bool free = false;
if (atomic_add_unless(&pt->used, -1, 1))
return false;
spin_lock(&pd->lock);
if (atomic_dec_and_test(&pt->used)) {
clear_pd_entry(pd, idx, scratch);
free = true;
}
spin_unlock(&pd->lock);
return free;
}
int i915_ppgtt_init_hw(struct intel_gt *gt)
{
struct drm_i915_private *i915 = gt->i915;
gtt_write_workarounds(gt);
if (GRAPHICS_VER(i915) == 6)
gen6_ppgtt_enable(gt);
else if (GRAPHICS_VER(i915) == 7)
gen7_ppgtt_enable(gt);
return 0;
}
static struct i915_ppgtt *
drm/i915: Reduce the number of objects subject to memcpy recover We really only need memcpy restore for objects that affect the operability of the migrate context. That is, primarily the page-table objects of the migrate VM. Add an object flag, I915_BO_ALLOC_PM_EARLY for objects that need early restores using memcpy and a way to assign LMEM page-table object flags to be used by the vms. Restore objects without this flag with the gpu blitter and only objects carrying the flag using TTM memcpy. Initially mark the migrate, gt, gtt and vgpu vms to use this flag, and defer for a later audit which vms actually need it. Most importantly, user- allocated vms with pinned page-table objects can be restored using the blitter. Performance-wise memcpy restore is probably as fast as gpu restore if not faster, but using gpu restore will help tackling future restrictions in mappable LMEM size. v4: - Don't mark the aliasing ppgtt page table flags for early resume, but rather the ggtt page table flags as intended. (Matthew Auld) - The check for user buffer objects during early resume is pointless, since they are never marked I915_BO_ALLOC_PM_EARLY. (Matthew Auld) v5: - Mark GuC LMEM objects with I915_BO_ALLOC_PM_EARLY to have them restored before we fire up the migrate context. Cc: Matthew Brost <matthew.brost@intel.com> Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Reviewed-by: Matthew Auld <matthew.auld@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20210922062527.865433-8-thomas.hellstrom@linux.intel.com
2021-09-22 08:25:25 +02:00
__ppgtt_create(struct intel_gt *gt, unsigned long lmem_pt_obj_flags)
{
if (GRAPHICS_VER(gt->i915) < 8)
return gen6_ppgtt_create(gt);
else
drm/i915: Reduce the number of objects subject to memcpy recover We really only need memcpy restore for objects that affect the operability of the migrate context. That is, primarily the page-table objects of the migrate VM. Add an object flag, I915_BO_ALLOC_PM_EARLY for objects that need early restores using memcpy and a way to assign LMEM page-table object flags to be used by the vms. Restore objects without this flag with the gpu blitter and only objects carrying the flag using TTM memcpy. Initially mark the migrate, gt, gtt and vgpu vms to use this flag, and defer for a later audit which vms actually need it. Most importantly, user- allocated vms with pinned page-table objects can be restored using the blitter. Performance-wise memcpy restore is probably as fast as gpu restore if not faster, but using gpu restore will help tackling future restrictions in mappable LMEM size. v4: - Don't mark the aliasing ppgtt page table flags for early resume, but rather the ggtt page table flags as intended. (Matthew Auld) - The check for user buffer objects during early resume is pointless, since they are never marked I915_BO_ALLOC_PM_EARLY. (Matthew Auld) v5: - Mark GuC LMEM objects with I915_BO_ALLOC_PM_EARLY to have them restored before we fire up the migrate context. Cc: Matthew Brost <matthew.brost@intel.com> Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Reviewed-by: Matthew Auld <matthew.auld@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20210922062527.865433-8-thomas.hellstrom@linux.intel.com
2021-09-22 08:25:25 +02:00
return gen8_ppgtt_create(gt, lmem_pt_obj_flags);
}
drm/i915: Reduce the number of objects subject to memcpy recover We really only need memcpy restore for objects that affect the operability of the migrate context. That is, primarily the page-table objects of the migrate VM. Add an object flag, I915_BO_ALLOC_PM_EARLY for objects that need early restores using memcpy and a way to assign LMEM page-table object flags to be used by the vms. Restore objects without this flag with the gpu blitter and only objects carrying the flag using TTM memcpy. Initially mark the migrate, gt, gtt and vgpu vms to use this flag, and defer for a later audit which vms actually need it. Most importantly, user- allocated vms with pinned page-table objects can be restored using the blitter. Performance-wise memcpy restore is probably as fast as gpu restore if not faster, but using gpu restore will help tackling future restrictions in mappable LMEM size. v4: - Don't mark the aliasing ppgtt page table flags for early resume, but rather the ggtt page table flags as intended. (Matthew Auld) - The check for user buffer objects during early resume is pointless, since they are never marked I915_BO_ALLOC_PM_EARLY. (Matthew Auld) v5: - Mark GuC LMEM objects with I915_BO_ALLOC_PM_EARLY to have them restored before we fire up the migrate context. Cc: Matthew Brost <matthew.brost@intel.com> Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Reviewed-by: Matthew Auld <matthew.auld@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20210922062527.865433-8-thomas.hellstrom@linux.intel.com
2021-09-22 08:25:25 +02:00
struct i915_ppgtt *i915_ppgtt_create(struct intel_gt *gt,
unsigned long lmem_pt_obj_flags)
{
struct i915_ppgtt *ppgtt;
drm/i915: Reduce the number of objects subject to memcpy recover We really only need memcpy restore for objects that affect the operability of the migrate context. That is, primarily the page-table objects of the migrate VM. Add an object flag, I915_BO_ALLOC_PM_EARLY for objects that need early restores using memcpy and a way to assign LMEM page-table object flags to be used by the vms. Restore objects without this flag with the gpu blitter and only objects carrying the flag using TTM memcpy. Initially mark the migrate, gt, gtt and vgpu vms to use this flag, and defer for a later audit which vms actually need it. Most importantly, user- allocated vms with pinned page-table objects can be restored using the blitter. Performance-wise memcpy restore is probably as fast as gpu restore if not faster, but using gpu restore will help tackling future restrictions in mappable LMEM size. v4: - Don't mark the aliasing ppgtt page table flags for early resume, but rather the ggtt page table flags as intended. (Matthew Auld) - The check for user buffer objects during early resume is pointless, since they are never marked I915_BO_ALLOC_PM_EARLY. (Matthew Auld) v5: - Mark GuC LMEM objects with I915_BO_ALLOC_PM_EARLY to have them restored before we fire up the migrate context. Cc: Matthew Brost <matthew.brost@intel.com> Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Reviewed-by: Matthew Auld <matthew.auld@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20210922062527.865433-8-thomas.hellstrom@linux.intel.com
2021-09-22 08:25:25 +02:00
ppgtt = __ppgtt_create(gt, lmem_pt_obj_flags);
if (IS_ERR(ppgtt))
return ppgtt;
trace_i915_ppgtt_create(&ppgtt->vm);
return ppgtt;
}
drm/i915: Preallocate stashes for vma page-directories We need to make the DMA allocations used for page directories to be performed up front so that we can include those allocations in our memory reservation pass. The downside is that we have to assume the worst case, even before we know the final layout, and always allocate enough page directories for this object, even when there will be overlap. This unfortunately can be quite expensive, especially as we have to clear/reset the page directories and DMA pages, but it should only be required during early phases of a workload when new objects are being discovered, or after memory/eviction pressure when we need to rebind. Once we reach steady state, the objects should not be moved and we no longer need to preallocating the pages tables. It should be noted that the lifetime for the page directories DMA is more or less decoupled from individual fences as they will be shared across objects across timelines. v2: Only allocate enough PD space for the PTE we may use, we do not need to allocate PD that will be left as scratch. v3: Store the shift unto the first PD level to encapsulate the different PTE counts for gen6/gen8. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Matthew Auld <matthew.auld@intel.com> Reviewed-by: Matthew Auld <matthew.auld@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20200729164219.5737-1-chris@chris-wilson.co.uk Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com> Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
2020-07-29 17:42:17 +01:00
void ppgtt_bind_vma(struct i915_address_space *vm,
struct i915_vm_pt_stash *stash,
struct i915_vma_resource *vma_res,
drm/i915: use pat_index instead of cache_level Currently the KMD is using enum i915_cache_level to set caching policy for buffer objects. This is flaky because the PAT index which really controls the caching behavior in PTE has far more levels than what's defined in the enum. In addition, the PAT index is platform dependent, having to translate between i915_cache_level and PAT index is not reliable, and makes the code more complicated. From UMD's perspective there is also a necessity to set caching policy for performance fine tuning. It's much easier for the UMD to directly use PAT index because the behavior of each PAT index is clearly defined in Bspec. Having the abstracted i915_cache_level sitting in between would only cause more ambiguity. PAT is expected to work much like MOCS already works today, and by design userspace is expected to select the index that exactly matches the desired behavior described in the hardware specification. For these reasons this patch replaces i915_cache_level with PAT index. Also note, the cache_level is not completely removed yet, because the KMD still has the need of creating buffer objects with simple cache settings such as cached, uncached, or writethrough. For kernel objects, cache_level is used for simplicity and backward compatibility. For Pre-gen12 platforms PAT can have 1:1 mapping to i915_cache_level, so these two are interchangeable. see the use of LEGACY_CACHELEVEL. One consequence of this change is that gen8_pte_encode is no longer working for gen12 platforms due to the fact that gen12 platforms has different PAT definitions. In the meantime the mtl_pte_encode introduced specfically for MTL becomes generic for all gen12 platforms. This patch renames the MTL PTE encode function into gen12_pte_encode and apply it to all gen12. Even though this change looks unrelated, but separating them would temporarily break gen12 PTE encoding, thus squash them in one patch. Special note: this patch changes the way caching behavior is controlled in the sense that some objects are left to be managed by userspace. For such objects we need to be careful not to change the userspace settings.There are kerneldoc and comments added around obj->cache_coherent, cache_dirty, and how to bypass the checkings by i915_gem_object_has_cache_level. For full understanding, these changes need to be looked at together with the two follow-up patches, one disables the {set|get}_caching ioctl's and the other adds set_pat extension to the GEM_CREATE uAPI. Bspec: 63019 Cc: Chris Wilson <chris.p.wilson@linux.intel.com> Signed-off-by: Fei Yang <fei.yang@intel.com> Reviewed-by: Andi Shyti <andi.shyti@linux.intel.com> Reviewed-by: Matt Roper <matthew.d.roper@intel.com> Signed-off-by: Andi Shyti <andi.shyti@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20230509165200.1740-3-fei.yang@intel.com
2023-05-09 09:52:00 -07:00
unsigned int pat_index,
drm/i915: Preallocate stashes for vma page-directories We need to make the DMA allocations used for page directories to be performed up front so that we can include those allocations in our memory reservation pass. The downside is that we have to assume the worst case, even before we know the final layout, and always allocate enough page directories for this object, even when there will be overlap. This unfortunately can be quite expensive, especially as we have to clear/reset the page directories and DMA pages, but it should only be required during early phases of a workload when new objects are being discovered, or after memory/eviction pressure when we need to rebind. Once we reach steady state, the objects should not be moved and we no longer need to preallocating the pages tables. It should be noted that the lifetime for the page directories DMA is more or less decoupled from individual fences as they will be shared across objects across timelines. v2: Only allocate enough PD space for the PTE we may use, we do not need to allocate PD that will be left as scratch. v3: Store the shift unto the first PD level to encapsulate the different PTE counts for gen6/gen8. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Matthew Auld <matthew.auld@intel.com> Reviewed-by: Matthew Auld <matthew.auld@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20200729164219.5737-1-chris@chris-wilson.co.uk Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com> Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
2020-07-29 17:42:17 +01:00
u32 flags)
{
u32 pte_flags;
if (!vma_res->allocated) {
vm->allocate_va_range(vm, stash, vma_res->start,
vma_res->vma_size);
vma_res->allocated = true;
}
/* Applicable to VLV, and gen8+ */
pte_flags = 0;
if (vma_res->bi.readonly)
pte_flags |= PTE_READ_ONLY;
if (vma_res->bi.lmem)
pte_flags |= PTE_LM;
drm/i915: use pat_index instead of cache_level Currently the KMD is using enum i915_cache_level to set caching policy for buffer objects. This is flaky because the PAT index which really controls the caching behavior in PTE has far more levels than what's defined in the enum. In addition, the PAT index is platform dependent, having to translate between i915_cache_level and PAT index is not reliable, and makes the code more complicated. From UMD's perspective there is also a necessity to set caching policy for performance fine tuning. It's much easier for the UMD to directly use PAT index because the behavior of each PAT index is clearly defined in Bspec. Having the abstracted i915_cache_level sitting in between would only cause more ambiguity. PAT is expected to work much like MOCS already works today, and by design userspace is expected to select the index that exactly matches the desired behavior described in the hardware specification. For these reasons this patch replaces i915_cache_level with PAT index. Also note, the cache_level is not completely removed yet, because the KMD still has the need of creating buffer objects with simple cache settings such as cached, uncached, or writethrough. For kernel objects, cache_level is used for simplicity and backward compatibility. For Pre-gen12 platforms PAT can have 1:1 mapping to i915_cache_level, so these two are interchangeable. see the use of LEGACY_CACHELEVEL. One consequence of this change is that gen8_pte_encode is no longer working for gen12 platforms due to the fact that gen12 platforms has different PAT definitions. In the meantime the mtl_pte_encode introduced specfically for MTL becomes generic for all gen12 platforms. This patch renames the MTL PTE encode function into gen12_pte_encode and apply it to all gen12. Even though this change looks unrelated, but separating them would temporarily break gen12 PTE encoding, thus squash them in one patch. Special note: this patch changes the way caching behavior is controlled in the sense that some objects are left to be managed by userspace. For such objects we need to be careful not to change the userspace settings.There are kerneldoc and comments added around obj->cache_coherent, cache_dirty, and how to bypass the checkings by i915_gem_object_has_cache_level. For full understanding, these changes need to be looked at together with the two follow-up patches, one disables the {set|get}_caching ioctl's and the other adds set_pat extension to the GEM_CREATE uAPI. Bspec: 63019 Cc: Chris Wilson <chris.p.wilson@linux.intel.com> Signed-off-by: Fei Yang <fei.yang@intel.com> Reviewed-by: Andi Shyti <andi.shyti@linux.intel.com> Reviewed-by: Matt Roper <matthew.d.roper@intel.com> Signed-off-by: Andi Shyti <andi.shyti@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20230509165200.1740-3-fei.yang@intel.com
2023-05-09 09:52:00 -07:00
vm->insert_entries(vm, vma_res, pat_index, pte_flags);
wmb();
}
void ppgtt_unbind_vma(struct i915_address_space *vm,
struct i915_vma_resource *vma_res)
{
drm/i915/gt: Batch TLB invalidations Invalidate TLB in batches, in order to reduce performance regressions. Currently, every caller performs a full barrier around a TLB invalidation, ignoring all other invalidations that may have already removed their PTEs from the cache. As this is a synchronous operation and can be quite slow, we cause multiple threads to contend on the TLB invalidate mutex blocking userspace. We only need to invalidate the TLB once after replacing our PTE to ensure that there is no possible continued access to the physical address before releasing our pages. By tracking a seqno for each full TLB invalidate we can quickly determine if one has been performed since rewriting the PTE, and only if necessary trigger one for ourselves. That helps to reduce the performance regression introduced by TLB invalidate logic. [mchehab: rebased to not require moving the code to a separate file] Cc: stable@vger.kernel.org Fixes: 7938d61591d3 ("drm/i915: Flush TLBs before releasing backing store") Suggested-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Signed-off-by: Chris Wilson <chris.p.wilson@intel.com> Cc: Fei Yang <fei.yang@intel.com> Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org> Acked-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Reviewed-by: Andi Shyti <andi.shyti@linux.intel.com> Signed-off-by: Andi Shyti <andi.shyti@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/4e97ef5deb6739cadaaf40aa45620547e9c4ec06.1658924372.git.mchehab@kernel.org (cherry picked from commit 5d36acb7198b0e5eb88e6b701f9ad7b9448f8df9) Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
2022-07-27 14:29:55 +02:00
if (!vma_res->allocated)
return;
vm->clear_range(vm, vma_res->start, vma_res->vma_size);
vma_invalidate_tlb(vm, vma_res->tlb);
}
drm/i915: Preallocate stashes for vma page-directories We need to make the DMA allocations used for page directories to be performed up front so that we can include those allocations in our memory reservation pass. The downside is that we have to assume the worst case, even before we know the final layout, and always allocate enough page directories for this object, even when there will be overlap. This unfortunately can be quite expensive, especially as we have to clear/reset the page directories and DMA pages, but it should only be required during early phases of a workload when new objects are being discovered, or after memory/eviction pressure when we need to rebind. Once we reach steady state, the objects should not be moved and we no longer need to preallocating the pages tables. It should be noted that the lifetime for the page directories DMA is more or less decoupled from individual fences as they will be shared across objects across timelines. v2: Only allocate enough PD space for the PTE we may use, we do not need to allocate PD that will be left as scratch. v3: Store the shift unto the first PD level to encapsulate the different PTE counts for gen6/gen8. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Matthew Auld <matthew.auld@intel.com> Reviewed-by: Matthew Auld <matthew.auld@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20200729164219.5737-1-chris@chris-wilson.co.uk Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com> Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
2020-07-29 17:42:17 +01:00
static unsigned long pd_count(u64 size, int shift)
{
/* Beware later misalignment */
return (size + 2 * (BIT_ULL(shift) - 1)) >> shift;
}
int i915_vm_alloc_pt_stash(struct i915_address_space *vm,
struct i915_vm_pt_stash *stash,
u64 size)
{
unsigned long count;
int shift, n, pt_sz;
drm/i915: Preallocate stashes for vma page-directories We need to make the DMA allocations used for page directories to be performed up front so that we can include those allocations in our memory reservation pass. The downside is that we have to assume the worst case, even before we know the final layout, and always allocate enough page directories for this object, even when there will be overlap. This unfortunately can be quite expensive, especially as we have to clear/reset the page directories and DMA pages, but it should only be required during early phases of a workload when new objects are being discovered, or after memory/eviction pressure when we need to rebind. Once we reach steady state, the objects should not be moved and we no longer need to preallocating the pages tables. It should be noted that the lifetime for the page directories DMA is more or less decoupled from individual fences as they will be shared across objects across timelines. v2: Only allocate enough PD space for the PTE we may use, we do not need to allocate PD that will be left as scratch. v3: Store the shift unto the first PD level to encapsulate the different PTE counts for gen6/gen8. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Matthew Auld <matthew.auld@intel.com> Reviewed-by: Matthew Auld <matthew.auld@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20200729164219.5737-1-chris@chris-wilson.co.uk Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com> Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
2020-07-29 17:42:17 +01:00
shift = vm->pd_shift;
if (!shift)
return 0;
pt_sz = stash->pt_sz;
if (!pt_sz)
pt_sz = I915_GTT_PAGE_SIZE_4K;
else
GEM_BUG_ON(!IS_DGFX(vm->i915));
GEM_BUG_ON(!is_power_of_2(pt_sz));
drm/i915: Preallocate stashes for vma page-directories We need to make the DMA allocations used for page directories to be performed up front so that we can include those allocations in our memory reservation pass. The downside is that we have to assume the worst case, even before we know the final layout, and always allocate enough page directories for this object, even when there will be overlap. This unfortunately can be quite expensive, especially as we have to clear/reset the page directories and DMA pages, but it should only be required during early phases of a workload when new objects are being discovered, or after memory/eviction pressure when we need to rebind. Once we reach steady state, the objects should not be moved and we no longer need to preallocating the pages tables. It should be noted that the lifetime for the page directories DMA is more or less decoupled from individual fences as they will be shared across objects across timelines. v2: Only allocate enough PD space for the PTE we may use, we do not need to allocate PD that will be left as scratch. v3: Store the shift unto the first PD level to encapsulate the different PTE counts for gen6/gen8. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Matthew Auld <matthew.auld@intel.com> Reviewed-by: Matthew Auld <matthew.auld@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20200729164219.5737-1-chris@chris-wilson.co.uk Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com> Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
2020-07-29 17:42:17 +01:00
count = pd_count(size, shift);
while (count--) {
struct i915_page_table *pt;
pt = alloc_pt(vm, pt_sz);
drm/i915: Preallocate stashes for vma page-directories We need to make the DMA allocations used for page directories to be performed up front so that we can include those allocations in our memory reservation pass. The downside is that we have to assume the worst case, even before we know the final layout, and always allocate enough page directories for this object, even when there will be overlap. This unfortunately can be quite expensive, especially as we have to clear/reset the page directories and DMA pages, but it should only be required during early phases of a workload when new objects are being discovered, or after memory/eviction pressure when we need to rebind. Once we reach steady state, the objects should not be moved and we no longer need to preallocating the pages tables. It should be noted that the lifetime for the page directories DMA is more or less decoupled from individual fences as they will be shared across objects across timelines. v2: Only allocate enough PD space for the PTE we may use, we do not need to allocate PD that will be left as scratch. v3: Store the shift unto the first PD level to encapsulate the different PTE counts for gen6/gen8. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Matthew Auld <matthew.auld@intel.com> Reviewed-by: Matthew Auld <matthew.auld@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20200729164219.5737-1-chris@chris-wilson.co.uk Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com> Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
2020-07-29 17:42:17 +01:00
if (IS_ERR(pt)) {
i915_vm_free_pt_stash(vm, stash);
return PTR_ERR(pt);
}
pt->stash = stash->pt[0];
stash->pt[0] = pt;
}
for (n = 1; n < vm->top; n++) {
shift += ilog2(I915_PDES); /* Each PD holds 512 entries */
count = pd_count(size, shift);
while (count--) {
struct i915_page_directory *pd;
pd = alloc_pd(vm);
if (IS_ERR(pd)) {
i915_vm_free_pt_stash(vm, stash);
return PTR_ERR(pd);
}
pd->pt.stash = stash->pt[1];
stash->pt[1] = &pd->pt;
}
}
return 0;
}
int i915_vm_map_pt_stash(struct i915_address_space *vm,
struct i915_vm_pt_stash *stash)
{
struct i915_page_table *pt;
int n, err;
for (n = 0; n < ARRAY_SIZE(stash->pt); n++) {
for (pt = stash->pt[n]; pt; pt = pt->stash) {
err = map_pt_dma_locked(vm, pt->base);
if (err)
return err;
}
}
return 0;
}
drm/i915: Preallocate stashes for vma page-directories We need to make the DMA allocations used for page directories to be performed up front so that we can include those allocations in our memory reservation pass. The downside is that we have to assume the worst case, even before we know the final layout, and always allocate enough page directories for this object, even when there will be overlap. This unfortunately can be quite expensive, especially as we have to clear/reset the page directories and DMA pages, but it should only be required during early phases of a workload when new objects are being discovered, or after memory/eviction pressure when we need to rebind. Once we reach steady state, the objects should not be moved and we no longer need to preallocating the pages tables. It should be noted that the lifetime for the page directories DMA is more or less decoupled from individual fences as they will be shared across objects across timelines. v2: Only allocate enough PD space for the PTE we may use, we do not need to allocate PD that will be left as scratch. v3: Store the shift unto the first PD level to encapsulate the different PTE counts for gen6/gen8. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Matthew Auld <matthew.auld@intel.com> Reviewed-by: Matthew Auld <matthew.auld@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20200729164219.5737-1-chris@chris-wilson.co.uk Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com> Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
2020-07-29 17:42:17 +01:00
void i915_vm_free_pt_stash(struct i915_address_space *vm,
struct i915_vm_pt_stash *stash)
{
struct i915_page_table *pt;
int n;
for (n = 0; n < ARRAY_SIZE(stash->pt); n++) {
while ((pt = stash->pt[n])) {
stash->pt[n] = pt->stash;
free_px(vm, pt, n);
drm/i915: Preallocate stashes for vma page-directories We need to make the DMA allocations used for page directories to be performed up front so that we can include those allocations in our memory reservation pass. The downside is that we have to assume the worst case, even before we know the final layout, and always allocate enough page directories for this object, even when there will be overlap. This unfortunately can be quite expensive, especially as we have to clear/reset the page directories and DMA pages, but it should only be required during early phases of a workload when new objects are being discovered, or after memory/eviction pressure when we need to rebind. Once we reach steady state, the objects should not be moved and we no longer need to preallocating the pages tables. It should be noted that the lifetime for the page directories DMA is more or less decoupled from individual fences as they will be shared across objects across timelines. v2: Only allocate enough PD space for the PTE we may use, we do not need to allocate PD that will be left as scratch. v3: Store the shift unto the first PD level to encapsulate the different PTE counts for gen6/gen8. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Matthew Auld <matthew.auld@intel.com> Reviewed-by: Matthew Auld <matthew.auld@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20200729164219.5737-1-chris@chris-wilson.co.uk Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com> Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
2020-07-29 17:42:17 +01:00
}
}
}
drm/i915: Reduce the number of objects subject to memcpy recover We really only need memcpy restore for objects that affect the operability of the migrate context. That is, primarily the page-table objects of the migrate VM. Add an object flag, I915_BO_ALLOC_PM_EARLY for objects that need early restores using memcpy and a way to assign LMEM page-table object flags to be used by the vms. Restore objects without this flag with the gpu blitter and only objects carrying the flag using TTM memcpy. Initially mark the migrate, gt, gtt and vgpu vms to use this flag, and defer for a later audit which vms actually need it. Most importantly, user- allocated vms with pinned page-table objects can be restored using the blitter. Performance-wise memcpy restore is probably as fast as gpu restore if not faster, but using gpu restore will help tackling future restrictions in mappable LMEM size. v4: - Don't mark the aliasing ppgtt page table flags for early resume, but rather the ggtt page table flags as intended. (Matthew Auld) - The check for user buffer objects during early resume is pointless, since they are never marked I915_BO_ALLOC_PM_EARLY. (Matthew Auld) v5: - Mark GuC LMEM objects with I915_BO_ALLOC_PM_EARLY to have them restored before we fire up the migrate context. Cc: Matthew Brost <matthew.brost@intel.com> Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Reviewed-by: Matthew Auld <matthew.auld@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20210922062527.865433-8-thomas.hellstrom@linux.intel.com
2021-09-22 08:25:25 +02:00
void ppgtt_init(struct i915_ppgtt *ppgtt, struct intel_gt *gt,
unsigned long lmem_pt_obj_flags)
{
struct drm_i915_private *i915 = gt->i915;
ppgtt->vm.gt = gt;
ppgtt->vm.i915 = i915;
ppgtt->vm.dma = i915->drm.dev;
ppgtt->vm.total = BIT_ULL(RUNTIME_INFO(i915)->ppgtt_size);
drm/i915: Reduce the number of objects subject to memcpy recover We really only need memcpy restore for objects that affect the operability of the migrate context. That is, primarily the page-table objects of the migrate VM. Add an object flag, I915_BO_ALLOC_PM_EARLY for objects that need early restores using memcpy and a way to assign LMEM page-table object flags to be used by the vms. Restore objects without this flag with the gpu blitter and only objects carrying the flag using TTM memcpy. Initially mark the migrate, gt, gtt and vgpu vms to use this flag, and defer for a later audit which vms actually need it. Most importantly, user- allocated vms with pinned page-table objects can be restored using the blitter. Performance-wise memcpy restore is probably as fast as gpu restore if not faster, but using gpu restore will help tackling future restrictions in mappable LMEM size. v4: - Don't mark the aliasing ppgtt page table flags for early resume, but rather the ggtt page table flags as intended. (Matthew Auld) - The check for user buffer objects during early resume is pointless, since they are never marked I915_BO_ALLOC_PM_EARLY. (Matthew Auld) v5: - Mark GuC LMEM objects with I915_BO_ALLOC_PM_EARLY to have them restored before we fire up the migrate context. Cc: Matthew Brost <matthew.brost@intel.com> Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Reviewed-by: Matthew Auld <matthew.auld@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20210922062527.865433-8-thomas.hellstrom@linux.intel.com
2021-09-22 08:25:25 +02:00
ppgtt->vm.lmem_pt_obj_flags = lmem_pt_obj_flags;
dma_resv_init(&ppgtt->vm._resv);
i915_address_space_init(&ppgtt->vm, VM_CLASS_PPGTT);
ppgtt->vm.vma_ops.bind_vma = ppgtt_bind_vma;
ppgtt->vm.vma_ops.unbind_vma = ppgtt_unbind_vma;
}