| Message ID | 20191029165134.28567-1-matthew.auld@intel.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | drm/i915/lmem: add the fake lmem region | expand |
Quoting Matthew Auld (2019-10-29 16:51:34) > diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c > index 21273b516dbe..db1736d95651 100644 > --- a/drivers/gpu/drm/i915/i915_drv.c > +++ b/drivers/gpu/drm/i915/i915_drv.c > @@ -1483,6 +1483,21 @@ int i915_driver_probe(struct pci_dev *pdev, const struct pci_device_id *ent) > if (!i915_modparams.nuclear_pageflip && match_info->gen < 5) > dev_priv->drm.driver_features &= ~DRIVER_ATOMIC; > > + /* > + * Check if we support fake LMEM -- for now we only unleash this for > + * the live selftests. > + */ > + if (IS_ENABLED(CONFIG_DRM_I915_UNSTABLE)) { The pattern I have in mind for unstable config options was to add config DRM_I915_UNSTABLE_FAKE_LMEM bool "Enable the experimental fake lmem" depends on DRM_I915_UNSTABLE default n help Convert some system memory into a fake local memory region for testing. So each is isolated and less likely to cross-contanimate. > + if (INTEL_GEN(dev_priv) >= 9 && i915_selftest.live && Probably want i915_selftest.live < 0 so that we only enable it for selftest-and-exit rather than inline selftests that keep the module loaded afterwards i915_selftest.live=0 => no tests i915_selftest.live=-1 => test and exit i915_selftest.live=1 => test and run userspace > + i915_modparams.fake_lmem_start) { > + mkwrite_device_info(dev_priv)->memory_regions = > + REGION_SMEM | REGION_LMEM | REGION_STOLEN; > + mkwrite_device_info(dev_priv)->is_dgfx = true; > + GEM_BUG_ON(!HAS_LMEM(dev_priv)); > + GEM_BUG_ON(!IS_DGFX(dev_priv)); > + } > + } > + > +struct intel_memory_region * > +intel_setup_fake_lmem(struct drm_i915_private *i915) > +{ > + struct pci_dev *pdev = i915->drm.pdev; > + struct intel_memory_region *mem; > + resource_size_t mappable_end; > + resource_size_t io_start; > + resource_size_t start; > + > + GEM_BUG_ON(i915_ggtt_has_aperture(&i915->ggtt)); > + GEM_BUG_ON(!i915_modparams.fake_lmem_start); > + > + /* Your mappable aperture belongs to me now! */ > + mappable_end = pci_resource_len(pdev, 2); > + io_start = pci_resource_start(pdev, 2), > + start = i915_modparams.fake_lmem_start; > + > + mem = intel_memory_region_create(i915, > + start, > + mappable_end, > + PAGE_SIZE, > + io_start, > + &intel_region_lmem_ops); > + if (!IS_ERR(mem)) { > + DRM_INFO("Intel graphics fake LMEM: %pR\n", &mem->region); > + DRM_INFO("Intel graphics fake LMEM IO start: %llx\n", > + (u64)mem->io_start); > + DRM_INFO("Intel graphics fake LMEM size: %llx\n", > + (u64)resource_size(&mem->region)); Ok, as this is unstable and therefore dev-centric I'll let you off having clear user information. resource_size_t => %pa Remember to make BAT happy! -Chris
On Tue, 29 Oct 2019 at 16:51, Matthew Auld <matthew.auld@intel.com> wrote: > > Intended for upstream testing so that we can still exercise the LMEM > plumbing and !i915_ggtt_has_aperture paths. Smoke tested on Skull Canyon > device. This works by allocating an intel_memory_region for a reserved > portion of system memory, which we treat like LMEM. For the LMEMBAR we > steal the aperture and 1:1 it map to the stolen region. > > To enable simply set the i915 modparam fake_lmem_start= on the kernel > cmdline with the start of reserved region(see memmap=). The size of the > region we can use is determined by the size of the mappable aperture, so > the size of reserved region should be >= mappable_end. For now we only > enable for the selftests. Depends on CONFIG_DRM_I915_UNSTABLE being > enabled. > > eg. memmap=2G$16G i915.fake_lmem_start=0x400000000 Hi Arek, Would you be able to update the fi-skl-lmem kernel cmd line with s/i915_fake_lmem_start/i915.fake_lmem_start?
On Wed, Oct 30, 2019 at 10:22:37PM +0000, Matthew Auld wrote: > On Tue, 29 Oct 2019 at 16:51, Matthew Auld <matthew.auld@intel.com> wrote: > > > > Intended for upstream testing so that we can still exercise the LMEM > > plumbing and !i915_ggtt_has_aperture paths. Smoke tested on Skull Canyon > > device. This works by allocating an intel_memory_region for a reserved > > portion of system memory, which we treat like LMEM. For the LMEMBAR we > > steal the aperture and 1:1 it map to the stolen region. > > > > To enable simply set the i915 modparam fake_lmem_start= on the kernel > > cmdline with the start of reserved region(see memmap=). The size of the > > region we can use is determined by the size of the mappable aperture, so > > the size of reserved region should be >= mappable_end. For now we only > > enable for the selftests. Depends on CONFIG_DRM_I915_UNSTABLE being > > enabled. > > > > eg. memmap=2G$16G i915.fake_lmem_start=0x400000000 > > Hi Arek, > > Would you be able to update the fi-skl-lmem kernel cmd line with > s/i915_fake_lmem_start/i915.fake_lmem_start? done
Quoting Arkadiusz Hiler (2019-10-31 12:40:35) > On Wed, Oct 30, 2019 at 10:22:37PM +0000, Matthew Auld wrote: > > On Tue, 29 Oct 2019 at 16:51, Matthew Auld <matthew.auld@intel.com> wrote: > > > > > > Intended for upstream testing so that we can still exercise the LMEM > > > plumbing and !i915_ggtt_has_aperture paths. Smoke tested on Skull Canyon > > > device. This works by allocating an intel_memory_region for a reserved > > > portion of system memory, which we treat like LMEM. For the LMEMBAR we > > > steal the aperture and 1:1 it map to the stolen region. > > > > > > To enable simply set the i915 modparam fake_lmem_start= on the kernel > > > cmdline with the start of reserved region(see memmap=). The size of the > > > region we can use is determined by the size of the mappable aperture, so > > > the size of reserved region should be >= mappable_end. For now we only > > > enable for the selftests. Depends on CONFIG_DRM_I915_UNSTABLE being > > > enabled. > > > > > > eg. memmap=2G$16G i915.fake_lmem_start=0x400000000 > > > > Hi Arek, > > > > Would you be able to update the fi-skl-lmem kernel cmd line with > > s/i915_fake_lmem_start/i915.fake_lmem_start? > > done <6>[ 497.897456] [drm] Intel graphics fake LMEM: [mem 0x100000000-0x13fffffff] <6>[ 497.897459] [drm] Intel graphics fake LMEM IO start: 40000000 <6>[ 497.897461] [drm] Intel graphics fake LMEM size: 40000000 All present. -Chris
On Thu, 31 Oct 2019 at 20:40, Chris Wilson <chris@chris-wilson.co.uk> wrote: > > Quoting Arkadiusz Hiler (2019-10-31 12:40:35) > > On Wed, Oct 30, 2019 at 10:22:37PM +0000, Matthew Auld wrote: > > > On Tue, 29 Oct 2019 at 16:51, Matthew Auld <matthew.auld@intel.com> wrote: > > > > > > > > Intended for upstream testing so that we can still exercise the LMEM > > > > plumbing and !i915_ggtt_has_aperture paths. Smoke tested on Skull Canyon > > > > device. This works by allocating an intel_memory_region for a reserved > > > > portion of system memory, which we treat like LMEM. For the LMEMBAR we > > > > steal the aperture and 1:1 it map to the stolen region. > > > > > > > > To enable simply set the i915 modparam fake_lmem_start= on the kernel > > > > cmdline with the start of reserved region(see memmap=). The size of the > > > > region we can use is determined by the size of the mappable aperture, so > > > > the size of reserved region should be >= mappable_end. For now we only > > > > enable for the selftests. Depends on CONFIG_DRM_I915_UNSTABLE being > > > > enabled. > > > > > > > > eg. memmap=2G$16G i915.fake_lmem_start=0x400000000 > > > > > > Hi Arek, > > > > > > Would you be able to update the fi-skl-lmem kernel cmd line with > > > s/i915_fake_lmem_start/i915.fake_lmem_start? > > > > done > > <6>[ 497.897456] [drm] Intel graphics fake LMEM: [mem 0x100000000-0x13fffffff] > <6>[ 497.897459] [drm] Intel graphics fake LMEM IO start: 40000000 > <6>[ 497.897461] [drm] Intel graphics fake LMEM size: 40000000 > > All present. Arek, Could we enable DRM_I915_UNSTABLE_FAKE_LMEM in CI? That should give us some coverage of the fake local-memory setup on fi-skl-lmem.
On Tue, Nov 05, 2019 at 01:05:20PM +0000, Matthew Auld wrote: > On Thu, 31 Oct 2019 at 20:40, Chris Wilson <chris@chris-wilson.co.uk> wrote: > > > > Quoting Arkadiusz Hiler (2019-10-31 12:40:35) > > > On Wed, Oct 30, 2019 at 10:22:37PM +0000, Matthew Auld wrote: > > > > On Tue, 29 Oct 2019 at 16:51, Matthew Auld <matthew.auld@intel.com> wrote: > > > > > > > > > > Intended for upstream testing so that we can still exercise the LMEM > > > > > plumbing and !i915_ggtt_has_aperture paths. Smoke tested on Skull Canyon > > > > > device. This works by allocating an intel_memory_region for a reserved > > > > > portion of system memory, which we treat like LMEM. For the LMEMBAR we > > > > > steal the aperture and 1:1 it map to the stolen region. > > > > > > > > > > To enable simply set the i915 modparam fake_lmem_start= on the kernel > > > > > cmdline with the start of reserved region(see memmap=). The size of the > > > > > region we can use is determined by the size of the mappable aperture, so > > > > > the size of reserved region should be >= mappable_end. For now we only > > > > > enable for the selftests. Depends on CONFIG_DRM_I915_UNSTABLE being > > > > > enabled. > > > > > > > > > > eg. memmap=2G$16G i915.fake_lmem_start=0x400000000 > > > > > > > > Hi Arek, > > > > > > > > Would you be able to update the fi-skl-lmem kernel cmd line with > > > > s/i915_fake_lmem_start/i915.fake_lmem_start? > > > > > > done > > > > <6>[ 497.897456] [drm] Intel graphics fake LMEM: [mem 0x100000000-0x13fffffff] > > <6>[ 497.897459] [drm] Intel graphics fake LMEM IO start: 40000000 > > <6>[ 497.897461] [drm] Intel graphics fake LMEM size: 40000000 > > > > All present. > > Arek, > > Could we enable DRM_I915_UNSTABLE_FAKE_LMEM in CI? That should give us > some coverage of the fake local-memory setup on fi-skl-lmem. Hey, config DRM_I915_UNSTABLE bool "Enable unstable API for early prototype development" depends on EXPERT depends on STAGING depends on BROKEN # should never be enabled by distros! config DRM_I915_UNSTABLE_FAKE_LMEM bool "Enable the experimental fake lmem" depends on DRM_I915_UNSTABLE AFAIU because of that dependency on CONFIG_BROKEN we cannot just enable it through .config - we have to edit the value in init/Kconfig[0]. Please push that change to core-for-CI (or other branch that gets integrated into drm-tip) and then just send a merge request to the kernel configs the CI is using[1]. [0]: https://lkml.org/lkml/2006/1/6/248 [1]: https://gitlab.freedesktop.org/gfx-ci/i915-infra/tree/master/kconfig
Quoting Arkadiusz Hiler (2019-11-06 08:30:37) > On Tue, Nov 05, 2019 at 01:05:20PM +0000, Matthew Auld wrote: > > On Thu, 31 Oct 2019 at 20:40, Chris Wilson <chris@chris-wilson.co.uk> wrote: > > > > > > Quoting Arkadiusz Hiler (2019-10-31 12:40:35) > > > > On Wed, Oct 30, 2019 at 10:22:37PM +0000, Matthew Auld wrote: > > > > > On Tue, 29 Oct 2019 at 16:51, Matthew Auld <matthew.auld@intel.com> wrote: > > > > > > > > > > > > Intended for upstream testing so that we can still exercise the LMEM > > > > > > plumbing and !i915_ggtt_has_aperture paths. Smoke tested on Skull Canyon > > > > > > device. This works by allocating an intel_memory_region for a reserved > > > > > > portion of system memory, which we treat like LMEM. For the LMEMBAR we > > > > > > steal the aperture and 1:1 it map to the stolen region. > > > > > > > > > > > > To enable simply set the i915 modparam fake_lmem_start= on the kernel > > > > > > cmdline with the start of reserved region(see memmap=). The size of the > > > > > > region we can use is determined by the size of the mappable aperture, so > > > > > > the size of reserved region should be >= mappable_end. For now we only > > > > > > enable for the selftests. Depends on CONFIG_DRM_I915_UNSTABLE being > > > > > > enabled. > > > > > > > > > > > > eg. memmap=2G$16G i915.fake_lmem_start=0x400000000 > > > > > > > > > > Hi Arek, > > > > > > > > > > Would you be able to update the fi-skl-lmem kernel cmd line with > > > > > s/i915_fake_lmem_start/i915.fake_lmem_start? > > > > > > > > done > > > > > > <6>[ 497.897456] [drm] Intel graphics fake LMEM: [mem 0x100000000-0x13fffffff] > > > <6>[ 497.897459] [drm] Intel graphics fake LMEM IO start: 40000000 > > > <6>[ 497.897461] [drm] Intel graphics fake LMEM size: 40000000 > > > > > > All present. > > > > Arek, > > > > Could we enable DRM_I915_UNSTABLE_FAKE_LMEM in CI? That should give us > > some coverage of the fake local-memory setup on fi-skl-lmem. > > Hey, > > config DRM_I915_UNSTABLE > bool "Enable unstable API for early prototype development" > depends on EXPERT > depends on STAGING > depends on BROKEN # should never be enabled by distros! > > config DRM_I915_UNSTABLE_FAKE_LMEM > bool "Enable the experimental fake lmem" > depends on DRM_I915_UNSTABLE > > AFAIU because of that dependency on CONFIG_BROKEN we cannot just enable > it through .config - we have to edit the value in init/Kconfig[0]. Before the revert last night, CONFIG_BROKEN was already enabled in CI. It's now enabled again. The above output was from CI setting up lmem without extra hackery. -Chris
On Wed, Nov 06, 2019 at 11:17:29AM +0000, Chris Wilson wrote: > Quoting Arkadiusz Hiler (2019-11-06 08:30:37) > > On Tue, Nov 05, 2019 at 01:05:20PM +0000, Matthew Auld wrote: > > > On Thu, 31 Oct 2019 at 20:40, Chris Wilson <chris@chris-wilson.co.uk> wrote: > > > > > > > > Quoting Arkadiusz Hiler (2019-10-31 12:40:35) > > > > > On Wed, Oct 30, 2019 at 10:22:37PM +0000, Matthew Auld wrote: > > > > > > On Tue, 29 Oct 2019 at 16:51, Matthew Auld <matthew.auld@intel.com> wrote: > > > > > > > > > > > > > > Intended for upstream testing so that we can still exercise the LMEM > > > > > > > plumbing and !i915_ggtt_has_aperture paths. Smoke tested on Skull Canyon > > > > > > > device. This works by allocating an intel_memory_region for a reserved > > > > > > > portion of system memory, which we treat like LMEM. For the LMEMBAR we > > > > > > > steal the aperture and 1:1 it map to the stolen region. > > > > > > > > > > > > > > To enable simply set the i915 modparam fake_lmem_start= on the kernel > > > > > > > cmdline with the start of reserved region(see memmap=). The size of the > > > > > > > region we can use is determined by the size of the mappable aperture, so > > > > > > > the size of reserved region should be >= mappable_end. For now we only > > > > > > > enable for the selftests. Depends on CONFIG_DRM_I915_UNSTABLE being > > > > > > > enabled. > > > > > > > > > > > > > > eg. memmap=2G$16G i915.fake_lmem_start=0x400000000 > > > > > > > > > > > > Hi Arek, > > > > > > > > > > > > Would you be able to update the fi-skl-lmem kernel cmd line with > > > > > > s/i915_fake_lmem_start/i915.fake_lmem_start? > > > > > > > > > > done > > > > > > > > <6>[ 497.897456] [drm] Intel graphics fake LMEM: [mem 0x100000000-0x13fffffff] > > > > <6>[ 497.897459] [drm] Intel graphics fake LMEM IO start: 40000000 > > > > <6>[ 497.897461] [drm] Intel graphics fake LMEM size: 40000000 > > > > > > > > All present. > > > > > > Arek, > > > > > > Could we enable DRM_I915_UNSTABLE_FAKE_LMEM in CI? That should give us > > > some coverage of the fake local-memory setup on fi-skl-lmem. > > > > Hey, > > > > config DRM_I915_UNSTABLE > > bool "Enable unstable API for early prototype development" > > depends on EXPERT > > depends on STAGING > > depends on BROKEN # should never be enabled by distros! > > > > config DRM_I915_UNSTABLE_FAKE_LMEM > > bool "Enable the experimental fake lmem" > > depends on DRM_I915_UNSTABLE > > > > AFAIU because of that dependency on CONFIG_BROKEN we cannot just enable > > it through .config - we have to edit the value in init/Kconfig[0]. > > Before the revert last night, CONFIG_BROKEN was already enabled in > CI. It's now enabled again. The above output was from CI setting up lmem > without extra hackery. > -Chris CI_DRM_7269 should have the DRM_I915_UNSTABLE_FAKE_LMEM enabled.
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_lmem.c b/drivers/gpu/drm/i915/gem/i915_gem_lmem.c index 926f6c940e0d..0e2bf6b7e143 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_lmem.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_lmem.c @@ -24,6 +24,7 @@ i915_gem_object_lmem_io_map_page(struct drm_i915_gem_object *obj, resource_size_t offset; offset = i915_gem_object_get_dma_address(obj, n); + offset -= obj->mm.region->region.start; return io_mapping_map_wc(&obj->mm.region->iomap, offset, PAGE_SIZE); } @@ -35,6 +36,7 @@ i915_gem_object_lmem_io_map_page_atomic(struct drm_i915_gem_object *obj, resource_size_t offset; offset = i915_gem_object_get_dma_address(obj, n); + offset -= obj->mm.region->region.start; return io_mapping_map_atomic_wc(&obj->mm.region->iomap, offset); } @@ -49,6 +51,7 @@ i915_gem_object_lmem_io_map(struct drm_i915_gem_object *obj, GEM_BUG_ON(!i915_gem_object_is_contiguous(obj)); offset = i915_gem_object_get_dma_address(obj, n); + offset -= obj->mm.region->region.start; return io_mapping_map_wc(&obj->mm.region->iomap, offset, size); } diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 21273b516dbe..db1736d95651 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -1483,6 +1483,21 @@ int i915_driver_probe(struct pci_dev *pdev, const struct pci_device_id *ent) if (!i915_modparams.nuclear_pageflip && match_info->gen < 5) dev_priv->drm.driver_features &= ~DRIVER_ATOMIC; + /* + * Check if we support fake LMEM -- for now we only unleash this for + * the live selftests. + */ + if (IS_ENABLED(CONFIG_DRM_I915_UNSTABLE)) { + if (INTEL_GEN(dev_priv) >= 9 && i915_selftest.live && + i915_modparams.fake_lmem_start) { + mkwrite_device_info(dev_priv)->memory_regions = + REGION_SMEM | REGION_LMEM | REGION_STOLEN; + mkwrite_device_info(dev_priv)->is_dgfx = true; + GEM_BUG_ON(!HAS_LMEM(dev_priv)); + GEM_BUG_ON(!IS_DGFX(dev_priv)); + } + } + ret = pci_enable_device(pdev); if (ret) goto out_fini; diff --git a/drivers/gpu/drm/i915/i915_params.c b/drivers/gpu/drm/i915/i915_params.c index 3fa79adb2c1c..9db3437fbd14 100644 --- a/drivers/gpu/drm/i915/i915_params.c +++ b/drivers/gpu/drm/i915/i915_params.c @@ -179,6 +179,11 @@ i915_param_named(enable_gvt, bool, 0400, "Enable support for Intel GVT-g graphics virtualization host support(default:false)"); #endif +#if IS_ENABLED(CONFIG_DRM_I915_UNSTABLE) +i915_param_named_unsafe(fake_lmem_start, ulong, 0600, + "Fake LMEM start offset (default: 0)"); +#endif + static __always_inline void _print_param(struct drm_printer *p, const char *name, const char *type, diff --git a/drivers/gpu/drm/i915/i915_params.h b/drivers/gpu/drm/i915/i915_params.h index a276317ad74b..31b88f297fbc 100644 --- a/drivers/gpu/drm/i915/i915_params.h +++ b/drivers/gpu/drm/i915/i915_params.h @@ -66,6 +66,7 @@ struct drm_printer; param(int, fastboot, -1) \ param(int, enable_dpcd_backlight, 0) \ param(char *, force_probe, CONFIG_DRM_I915_FORCE_PROBE) \ + param(unsigned long, fake_lmem_start, 0) \ /* leave bools at the end to not create holes */ \ param(bool, alpha_support, IS_ENABLED(CONFIG_DRM_I915_ALPHA_SUPPORT)) \ param(bool, enable_hangcheck, true) \ diff --git a/drivers/gpu/drm/i915/intel_memory_region.c b/drivers/gpu/drm/i915/intel_memory_region.c index a60f77ff58d4..baaeaecc64af 100644 --- a/drivers/gpu/drm/i915/intel_memory_region.c +++ b/drivers/gpu/drm/i915/intel_memory_region.c @@ -228,6 +228,9 @@ int intel_memory_regions_hw_probe(struct drm_i915_private *i915) case INTEL_MEMORY_STOLEN: mem = i915_gem_stolen_setup(i915); break; + case INTEL_MEMORY_LOCAL: + mem = intel_setup_fake_lmem(i915); + break; } if (IS_ERR(mem)) { diff --git a/drivers/gpu/drm/i915/intel_memory_region.h b/drivers/gpu/drm/i915/intel_memory_region.h index 19920c256ede..238722009677 100644 --- a/drivers/gpu/drm/i915/intel_memory_region.h +++ b/drivers/gpu/drm/i915/intel_memory_region.h @@ -10,6 +10,7 @@ #include <linux/ioport.h> #include <linux/mutex.h> #include <linux/io-mapping.h> +#include <drm/drm_mm.h> #include "i915_buddy.h" @@ -71,6 +72,9 @@ struct intel_memory_region { struct io_mapping iomap; struct resource region; + /* For fake LMEM */ + struct drm_mm_node fake_mappable; + struct i915_buddy_mm mm; struct mutex mm_lock; @@ -83,6 +87,8 @@ struct intel_memory_region { unsigned int instance; unsigned int id; + dma_addr_t remap_addr; + struct { struct mutex lock; /* Protects access to objects */ struct list_head list; diff --git a/drivers/gpu/drm/i915/intel_region_lmem.c b/drivers/gpu/drm/i915/intel_region_lmem.c index 9a351af45ce6..583118095635 100644 --- a/drivers/gpu/drm/i915/intel_region_lmem.c +++ b/drivers/gpu/drm/i915/intel_region_lmem.c @@ -9,9 +9,62 @@ #include "gem/i915_gem_region.h" #include "intel_region_lmem.h" +static int init_fake_lmem_bar(struct intel_memory_region *mem) +{ + struct drm_i915_private *i915 = mem->i915; + struct i915_ggtt *ggtt = &i915->ggtt; + unsigned long n; + int ret; + + /* We want to 1:1 map the mappable aperture to our reserved region */ + + mem->fake_mappable.start = 0; + mem->fake_mappable.size = resource_size(&mem->region); + mem->fake_mappable.color = I915_COLOR_UNEVICTABLE; + + ret = drm_mm_reserve_node(&ggtt->vm.mm, &mem->fake_mappable); + if (ret) + return ret; + + mem->remap_addr = dma_map_resource(&i915->drm.pdev->dev, + mem->region.start, + mem->fake_mappable.size, + PCI_DMA_BIDIRECTIONAL, + DMA_ATTR_FORCE_CONTIGUOUS); + if (dma_mapping_error(&i915->drm.pdev->dev, mem->remap_addr)) { + drm_mm_remove_node(&mem->fake_mappable); + return -EINVAL; + } + + for (n = 0; n < mem->fake_mappable.size >> PAGE_SHIFT; ++n) { + ggtt->vm.insert_page(&ggtt->vm, + mem->remap_addr + (n << PAGE_SHIFT), + n << PAGE_SHIFT, + I915_CACHE_NONE, 0); + } + + mem->region = (struct resource)DEFINE_RES_MEM(mem->remap_addr, + mem->fake_mappable.size); + + return 0; +} + +static void release_fake_lmem_bar(struct intel_memory_region *mem) +{ + if (drm_mm_node_allocated(&mem->fake_mappable)) + drm_mm_remove_node(&mem->fake_mappable); + + dma_unmap_resource(&mem->i915->drm.pdev->dev, + mem->remap_addr, + mem->fake_mappable.size, + PCI_DMA_BIDIRECTIONAL, + DMA_ATTR_FORCE_CONTIGUOUS); +} + static void region_lmem_release(struct intel_memory_region *mem) { + release_fake_lmem_bar(mem); io_mapping_fini(&mem->iomap); intel_memory_region_release_buddy(mem); } @@ -21,6 +74,11 @@ region_lmem_init(struct intel_memory_region *mem) { int ret; + if (i915_modparams.fake_lmem_start) { + ret = init_fake_lmem_bar(mem); + GEM_BUG_ON(ret); + } + if (!io_mapping_init_wc(&mem->iomap, mem->io_start, resource_size(&mem->region))) @@ -38,3 +96,37 @@ const struct intel_memory_region_ops intel_region_lmem_ops = { .release = region_lmem_release, .create_object = __i915_gem_lmem_object_create, }; + +struct intel_memory_region * +intel_setup_fake_lmem(struct drm_i915_private *i915) +{ + struct pci_dev *pdev = i915->drm.pdev; + struct intel_memory_region *mem; + resource_size_t mappable_end; + resource_size_t io_start; + resource_size_t start; + + GEM_BUG_ON(i915_ggtt_has_aperture(&i915->ggtt)); + GEM_BUG_ON(!i915_modparams.fake_lmem_start); + + /* Your mappable aperture belongs to me now! */ + mappable_end = pci_resource_len(pdev, 2); + io_start = pci_resource_start(pdev, 2), + start = i915_modparams.fake_lmem_start; + + mem = intel_memory_region_create(i915, + start, + mappable_end, + PAGE_SIZE, + io_start, + &intel_region_lmem_ops); + if (!IS_ERR(mem)) { + DRM_INFO("Intel graphics fake LMEM: %pR\n", &mem->region); + DRM_INFO("Intel graphics fake LMEM IO start: %llx\n", + (u64)mem->io_start); + DRM_INFO("Intel graphics fake LMEM size: %llx\n", + (u64)resource_size(&mem->region)); + } + + return mem; +} diff --git a/drivers/gpu/drm/i915/intel_region_lmem.h b/drivers/gpu/drm/i915/intel_region_lmem.h index ed2a3bab6443..213def7c7b8a 100644 --- a/drivers/gpu/drm/i915/intel_region_lmem.h +++ b/drivers/gpu/drm/i915/intel_region_lmem.h @@ -6,6 +6,11 @@ #ifndef __INTEL_REGION_LMEM_H #define __INTEL_REGION_LMEM_H +struct drm_i915_private; + extern const struct intel_memory_region_ops intel_region_lmem_ops; +struct intel_memory_region * +intel_setup_fake_lmem(struct drm_i915_private *i915); + #endif /* !__INTEL_REGION_LMEM_H */
Intended for upstream testing so that we can still exercise the LMEM plumbing and !i915_ggtt_has_aperture paths. Smoke tested on Skull Canyon device. This works by allocating an intel_memory_region for a reserved portion of system memory, which we treat like LMEM. For the LMEMBAR we steal the aperture and 1:1 it map to the stolen region. To enable simply set the i915 modparam fake_lmem_start= on the kernel cmdline with the start of reserved region(see memmap=). The size of the region we can use is determined by the size of the mappable aperture, so the size of reserved region should be >= mappable_end. For now we only enable for the selftests. Depends on CONFIG_DRM_I915_UNSTABLE being enabled. eg. memmap=2G$16G i915.fake_lmem_start=0x400000000 v2: make fake_lmem_start an i915 modparam Signed-off-by: Matthew Auld <matthew.auld@intel.com> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Abdiel Janulgue <abdiel.janulgue@linux.intel.com> Cc: Arkadiusz Hiler <arkadiusz.hiler@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> --- drivers/gpu/drm/i915/gem/i915_gem_lmem.c | 3 + drivers/gpu/drm/i915/i915_drv.c | 15 ++++ drivers/gpu/drm/i915/i915_params.c | 5 ++ drivers/gpu/drm/i915/i915_params.h | 1 + drivers/gpu/drm/i915/intel_memory_region.c | 3 + drivers/gpu/drm/i915/intel_memory_region.h | 6 ++ drivers/gpu/drm/i915/intel_region_lmem.c | 92 ++++++++++++++++++++++ drivers/gpu/drm/i915/intel_region_lmem.h | 5 ++ 8 files changed, 130 insertions(+)