Merge tag 'misc-habanalabs-fixes-2019-06-06' of git://people.freedesktop.org/~gabbayo/linux into char-misc-linus

Oded writes:

This tag contains the following fixes:

- Fix the code that checks whether we can use 2MB page size when mapping
  memory in the ASIC's MMU. The current code had a "hole" which happened
  in architectures other then x86-64.

- Fix the debugfs interface to read/write from/to the device using device
  virtual addresses. There was a bug in the translation regarding
  addresses that were mapped using 2MB page size.

- Fix a bug in the debug/profiling code, where the code didn't read the
  full address but only the lower 32-bits of the address.

* tag 'misc-habanalabs-fixes-2019-06-06' of git://people.freedesktop.org/~gabbayo/linux:
  habanalabs: Read upper bits of trace buffer from RWPHI
  habanalabs: Fix virtual address access via debugfs for 2MB pages
  habanalabs: fix bug in checking huge page optimization

3 files changed, 16 insertions, 9 deletions

diff --git a/drivers/misc/habanalabs/debugfs.c b/drivers/misc/habanalabs/debugfs.c
index 0ce5621c1324..ba418aaa404c 100644
--- a/drivers/misc/habanalabs/debugfs.c
+++ b/drivers/misc/habanalabs/debugfs.c
@@ -500,6 +500,7 @@ static int device_va_to_pa(struct hl_device *hdev, u64 virt_addr,
 {
 	struct hl_ctx *ctx = hdev->user_ctx;
 	u64 hop_addr, hop_pte_addr, hop_pte;
+	u64 offset_mask = HOP4_MASK | OFFSET_MASK;
 	int rc = 0;
 
 	if (!ctx) {
@@ -542,12 +543,14 @@ static int device_va_to_pa(struct hl_device *hdev, u64 virt_addr,
 			goto not_mapped;
 		hop_pte_addr = get_hop4_pte_addr(ctx, hop_addr, virt_addr);
 		hop_pte = hdev->asic_funcs->read_pte(hdev, hop_pte_addr);
+
+		offset_mask = OFFSET_MASK;
 	}
 
 	if (!(hop_pte & PAGE_PRESENT_MASK))
 		goto not_mapped;
 
-	*phys_addr = (hop_pte & PTE_PHYS_ADDR_MASK) | (virt_addr & OFFSET_MASK);
+	*phys_addr = (hop_pte & ~offset_mask) | (virt_addr & offset_mask);
 
 	goto out;
 
diff --git a/drivers/misc/habanalabs/goya/goya_coresight.c b/drivers/misc/habanalabs/goya/goya_coresight.c
index 39f62ce72660..d7ec7ad84cc6 100644
--- a/drivers/misc/habanalabs/goya/goya_coresight.c
+++ b/drivers/misc/habanalabs/goya/goya_coresight.c
@@ -425,8 +425,18 @@ static int goya_config_etr(struct hl_device *hdev,
 		WREG32(base_reg + 0x28, 0);
 		WREG32(base_reg + 0x304, 0);
 
-		if (params->output_size >= sizeof(u32))
-			*(u32 *) params->output = RREG32(base_reg + 0x18);
+		if (params->output_size >= sizeof(u64)) {
+			u32 rwp, rwphi;
+
+			/*
+			 * The trace buffer address is 40 bits wide. The end of
+			 * the buffer is set in the RWP register (lower 32
+			 * bits), and in the RWPHI register (upper 8 bits).
+			 */
+			rwp = RREG32(base_reg + 0x18);
+			rwphi = RREG32(base_reg + 0x3c) & 0xff;
+			*(u64 *) params->output = ((u64) rwphi << 32) | rwp;
+		}
 	}
 
 	return 0;
diff --git a/drivers/misc/habanalabs/memory.c b/drivers/misc/habanalabs/memory.c
index d67d24c13efd..693877e37fd8 100644
--- a/drivers/misc/habanalabs/memory.c
+++ b/drivers/misc/habanalabs/memory.c
@@ -675,11 +675,6 @@ static int init_phys_pg_pack_from_userptr(struct hl_ctx *ctx,
 
 		total_npages += npages;
 
-		if (first) {
-			first = false;
-			dma_addr &= PAGE_MASK_2MB;
-		}
-
 		if ((npages % PGS_IN_2MB_PAGE) ||
 					(dma_addr & (PAGE_SIZE_2MB - 1)))
 			is_huge_page_opt = false;
@@ -704,7 +699,6 @@ static int init_phys_pg_pack_from_userptr(struct hl_ctx *ctx,
 	phys_pg_pack->total_size = total_npages * page_size;
 
 	j = 0;
-	first = true;
 	for_each_sg(userptr->sgt->sgl, sg, userptr->sgt->nents, i) {
 		npages = get_sg_info(sg, &dma_addr);