From: Dave Airlie <[email protected]> This avoids constant need to ioremap when instobjs move at least on 64-bit systems.
This create the io mapping on first use, because creating it at init time causes a resource mapping error, because nouveau hasn't kicked simpledrm off the hardware yet, but ioremap_wc the whole BAR causes an overlap with BOOTFB/simpledrm. I think the resource system could do better here, but it's easier to just delay creating the mapping until first use. Signed-off-by: Dave Airlie <[email protected]> --- .../nouveau/nvkm/subdev/gsp/rm/r535/fbsr.c | 1 + .../drm/nouveau/nvkm/subdev/instmem/nv50.c | 38 +++++++++++++++---- .../drm/nouveau/nvkm/subdev/instmem/priv.h | 1 + 3 files changed, 32 insertions(+), 8 deletions(-) diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/rm/r535/fbsr.c b/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/rm/r535/fbsr.c index 700cea5def35..f128330f30d7 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/rm/r535/fbsr.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/rm/r535/fbsr.c @@ -297,6 +297,7 @@ r535_fbsr = { static void * r535_instmem_dtor(struct nvkm_instmem *imem) { + nv50_instmem_dtor(imem); kfree(imem->func); return imem; } diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/instmem/nv50.c b/drivers/gpu/drm/nouveau/nvkm/subdev/instmem/nv50.c index 6fbaa1e5876d..7314650f6d23 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/instmem/nv50.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/instmem/nv50.c @@ -30,10 +30,14 @@ #include <subdev/gsp.h> #include <subdev/mmu.h> +#include <linux/io-mapping.h> + struct nv50_instmem { struct nvkm_instmem base; u64 addr; + struct io_mapping iomap; + /* Mappings that can be evicted when BAR2 space has been exhausted. */ struct list_head lru; }; @@ -124,7 +128,6 @@ nv50_instobj_kmap(struct nv50_instobj *iobj, struct nvkm_vmm *vmm) struct nv50_instobj *eobj; struct nvkm_memory *memory = &iobj->base.memory; struct nvkm_subdev *subdev = &imem->base.subdev; - struct nvkm_device *device = subdev->device; struct nvkm_vma *bar = NULL, *ebar; u64 size = nvkm_memory_size(memory); void *emap; @@ -155,7 +158,7 @@ nv50_instobj_kmap(struct nv50_instobj *iobj, struct nvkm_vmm *vmm) mutex_unlock(&imem->base.mutex); if (!eobj) break; - iounmap(emap); + io_mapping_unmap(emap); nvkm_vmm_put(vmm, &ebar); } @@ -172,8 +175,7 @@ nv50_instobj_kmap(struct nv50_instobj *iobj, struct nvkm_vmm *vmm) /* Make the mapping visible to the host. */ iobj->bar = bar; - iobj->map = ioremap_wc(device->func->resource_addr(device, NVKM_BAR2_INST) + - (u32)iobj->bar->addr, size); + iobj->map = io_mapping_map_wc(&imem->iomap, (u32)iobj->bar->addr, size); if (!iobj->map) { nvkm_warn(subdev, "PRAMIN ioremap failed\n"); nvkm_vmm_put(vmm, &iobj->bar); @@ -188,6 +190,17 @@ nv50_instobj_map(struct nvkm_memory *memory, u64 offset, struct nvkm_vmm *vmm, return nvkm_memory_map(memory, offset, vmm, vma, argv, argc); } +static void +check_io_mapping(struct nv50_instmem *imem) +{ + if (imem->iomap.size == 0) { + struct nvkm_device *device = imem->base.subdev.device; + io_mapping_init_wc(&imem->iomap, + device->func->resource_addr(device, NVKM_BAR2_INST), + device->func->resource_size(device, NVKM_BAR2_INST)); + } +} + static void nv50_instobj_release(struct nvkm_memory *memory) { @@ -239,7 +252,9 @@ nv50_instobj_acquire(struct nvkm_memory *memory) /* Attempt to get a direct CPU mapping of the object. */ if ((vmm = nvkm_bar_bar2_vmm(imem->subdev.device))) { - if (!iobj->map) + + check_io_mapping(iobj->imem); + if (!iobj->map && iobj->imem->iomap.size) nv50_instobj_kmap(iobj, vmm); map = iobj->map; } @@ -277,6 +292,7 @@ nv50_instobj_boot(struct nvkm_memory *memory, struct nvkm_vmm *vmm) iobj->lru.next = NULL; } + check_io_mapping(iobj->imem); nv50_instobj_kmap(iobj, vmm); nvkm_instmem_boot(imem); mutex_unlock(&imem->mutex); @@ -330,7 +346,7 @@ nv50_instobj_dtor(struct nvkm_memory *memory) if (map) { struct nvkm_vmm *vmm = nvkm_bar_bar2_vmm(imem->subdev.device); - iounmap(map); + io_mapping_unmap(map); if (likely(vmm)) /* Can be NULL during BAR destructor. */ nvkm_vmm_put(vmm, &bar); } @@ -406,10 +422,13 @@ nv50_instmem_fini(struct nvkm_instmem *base) nv50_instmem(base)->addr = ~0ULL; } -static void * +void * nv50_instmem_dtor(struct nvkm_instmem *base) { - return nv50_instmem(base); + struct nv50_instmem *imem = nv50_instmem(base); + + io_mapping_fini(&imem->iomap); + return imem; } static const struct nvkm_instmem_func @@ -433,8 +452,11 @@ nv50_instmem_new_(const struct nvkm_instmem_func *func, if (!(imem = kzalloc_obj(*imem))) return -ENOMEM; + + nvkm_instmem_ctor(func, device, type, inst, &imem->base); INIT_LIST_HEAD(&imem->lru); + *pimem = &imem->base; return 0; } diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/instmem/priv.h b/drivers/gpu/drm/nouveau/nvkm/subdev/instmem/priv.h index 87bbdd786eaa..b423f01f5fcd 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/instmem/priv.h +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/instmem/priv.h @@ -26,6 +26,7 @@ int nv50_instobj_new(struct nvkm_instmem *, u32 size, u32 align, bool zero, struct nvkm_memory **); int nv50_instobj_wrap(struct nvkm_instmem *, struct nvkm_memory *vram, struct nvkm_memory **bar2); +void *nv50_instmem_dtor(struct nvkm_instmem *base); void nvkm_instmem_ctor(const struct nvkm_instmem_func *, struct nvkm_device *, enum nvkm_subdev_type, int, struct nvkm_instmem *); -- 2.54.0
