Revert "iommu/amd_iommu: Use CONFIG_DMA_DIRECT_OPS=y and dma_direct_{alloc,free}()"

This reverts commit b468620f.

It turns out that this broke drm on AMD platforms. Quoting Gabriel C:
 "I can confirm reverting b468620f fixes
  that issue for me.

  The GPU is working fine with SME enabled.

  Now with working GPU :) I can also confirm performance is back to
  normal without doing any other workarounds"

Christan König analyzed it partially:
 "As far as I analyzed it we now get an -ENOMEM from dma_alloc_attrs()
  in drivers/gpu/drm/ttm/ttm_page_alloc_dma.c when IOMMU is enabled"

and Christoph Hellwig responded:
 "I think the prime issue is that dma_direct_alloc respects the dma
  mask. Which we don't need if actually using the iommu. This would be
  mostly harmless exept for the the SEV bit high in the address that
  makes the checks fail.

  For now I'd say revert this commit for 4.17/4.18-rc and I'll look into
  addressing these issues properly"
Reported-and-bisected-by: Gabriel C <nix.or.die@gmail.com>
Acked-by: Christoph Hellwig <hch@lst.de>
Cc: Christian König <christian.koenig@amd.com>
Cc: Michel Dänzer <michel.daenzer@amd.com>
Cc: Joerg Roedel <jroedel@suse.de>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: stable@kernel.org		# v4.17
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

drivers/iommu/Kconfig

@@ -107,7 +107,6 @@ config IOMMU_PGTABLES_L2
 # AMD IOMMU support
 config AMD_IOMMU
 	bool "AMD IOMMU support"
-	select DMA_DIRECT_OPS
 	select SWIOTLB
 	select PCI_MSI
 	select PCI_ATS

drivers/iommu/amd_iommu.c

@@ -2596,32 +2596,51 @@ static void *alloc_coherent(struct device *dev, size_t size,
 			    unsigned long attrs)
 {
 	u64 dma_mask = dev->coherent_dma_mask;
-	struct protection_domain *domain = get_domain(dev);
-	bool is_direct = false;
-	void *virt_addr;
+	struct protection_domain *domain;
+	struct dma_ops_domain *dma_dom;
+	struct page *page;
+
+	domain = get_domain(dev);
+	if (PTR_ERR(domain) == -EINVAL) {
+		page = alloc_pages(flag, get_order(size));
+		*dma_addr = page_to_phys(page);
+		return page_address(page);
+	} else if (IS_ERR(domain))
+		return NULL;
 
-	if (IS_ERR(domain)) {
-		if (PTR_ERR(domain) != -EINVAL)
+	dma_dom   = to_dma_ops_domain(domain);
+	size	  = PAGE_ALIGN(size);
+	dma_mask  = dev->coherent_dma_mask;
+	flag     &= ~(__GFP_DMA | __GFP_HIGHMEM | __GFP_DMA32);
+	flag     |= __GFP_ZERO;
+
+	page = alloc_pages(flag | __GFP_NOWARN,  get_order(size));
+	if (!page) {
+		if (!gfpflags_allow_blocking(flag))
 			return NULL;
-		is_direct = true;
-	}
 
-	virt_addr = dma_direct_alloc(dev, size, dma_addr, flag, attrs);
-	if (!virt_addr || is_direct)
-		return virt_addr;
+		page = dma_alloc_from_contiguous(dev, size >> PAGE_SHIFT,
+						 get_order(size), flag);
+		if (!page)
+			return NULL;
+	}
 
 	if (!dma_mask)
 		dma_mask = *dev->dma_mask;
 
-	*dma_addr = __map_single(dev, to_dma_ops_domain(domain),
-			virt_to_phys(virt_addr), PAGE_ALIGN(size),
-			DMA_BIDIRECTIONAL, dma_mask);
+	*dma_addr = __map_single(dev, dma_dom, page_to_phys(page),
+				 size, DMA_BIDIRECTIONAL, dma_mask);
+
 	if (*dma_addr == AMD_IOMMU_MAPPING_ERROR)
 		goto out_free;
-	return virt_addr;
+
+	return page_address(page);
 
 out_free:
-	dma_direct_free(dev, size, virt_addr, *dma_addr, attrs);
+
+	if (!dma_release_from_contiguous(dev, page, size >> PAGE_SHIFT))
+		__free_pages(page, get_order(size));
+
 	return NULL;
 }
 
@@ -2632,17 +2651,24 @@ static void free_coherent(struct device *dev, size_t size,
 			  void *virt_addr, dma_addr_t dma_addr,
 			  unsigned long attrs)
 {
-	struct protection_domain *domain = get_domain(dev);
+	struct protection_domain *domain;
+	struct dma_ops_domain *dma_dom;
+	struct page *page;
 
+	page = virt_to_page(virt_addr);
 	size = PAGE_ALIGN(size);
 
-	if (!IS_ERR(domain)) {
-		struct dma_ops_domain *dma_dom = to_dma_ops_domain(domain);
+	domain = get_domain(dev);
+	if (IS_ERR(domain))
+		goto free_mem;
 
-		__unmap_single(dma_dom, dma_addr, size, DMA_BIDIRECTIONAL);
-	}
+	dma_dom = to_dma_ops_domain(domain);
+
+	__unmap_single(dma_dom, dma_addr, size, DMA_BIDIRECTIONAL);
 
-	dma_direct_free(dev, size, virt_addr, dma_addr, attrs);
+free_mem:
+	if (!dma_release_from_contiguous(dev, page, size >> PAGE_SHIFT))
+		__free_pages(page, get_order(size));
 }
 
 /*