Skip to contentSkip to conversation

mm/cma: Fix potential memory loss on cma_declare_contiguous_nid

PATCH Applied Levi Yun 2 participants 2 messages Jan 18, 2023

Subsystems (2)
lore
Share & tools
lore

Levi Yun

author
#0 lore raw

PATCH 1/1 mm/cma: Fix potential memory loss on cma_declare_contiguous_nid

Suppose, memblock_alloc_ranged_nid with highmem_start is success when cma_declare_contiguous_nid is called with !fixed in the system 32-bit system with PHYS_ADDR_T_64BIT enabled with memblock.bottom_up == false.

But, Next trial to memblock_alloc_range_nid to allocate in [SIZE_4G, limits) nullifies former success allocated addr and it retries to memblock_alloc_ragne_nid again.

In this situation, first success address area is lost.

By changing order of allocation (SIZE_4G, high_memory and base) and checking whether allocated prevents potential memory lost.

---
 mm/cma.c | 24 ++++++++++++------------
 1 file changed, 12 insertions(+), 12 deletions(-)
mm/cma.c +12 -12
index 4a978e09547a..49d9b08d9a47 100644
@@ -321,18 +321,6 @@ int __init cma_declare_contiguous_nid(phys_addr_t base,
 	} else {
 		phys_addr_t addr = 0;
 
-		/*
-		 * All pages in the reserved area must come from the same zone.
-		 * If the requested region crosses the low/high memory boundary,
-		 * try allocating from high memory first and fall back to low
-		 * memory in case of failure.
-		 */
-		if (base < highmem_start && limit > highmem_start) {
-			addr = memblock_alloc_range_nid(size, alignment,
-					highmem_start, limit, nid, true);
-			limit = highmem_start;
-		}
-
 		/*
 		 * If there is enough memory, try a bottom-up allocation first.
 		 * It will place the new cma area close to the start of the node
@@ -350,6 +338,18 @@ int __init cma_declare_contiguous_nid(phys_addr_t base,
 		}
 #endif
 
+		/*
+		 * All pages in the reserved area must come from the same zone.
+		 * If the requested region crosses the low/high memory boundary,
+		 * try allocating from high memory first and fall back to low
+		 * memory in case of failure.
+		 */
+		if (!addr && base < highmem_start && limit > highmem_start) {
+			addr = memblock_alloc_range_nid(size, alignment,
+					highmem_start, limit, nid, true);
+			limit = highmem_start;
+		}
+
 		if (!addr) {
 			addr = memblock_alloc_range_nid(size, alignment, base,
 					limit, nid, true);

--
2.35.1

maintainer
#1 lore raw
Quoted 5 lines · On Wed, 18 Jan 2023 17:05:23 +0900 Levi Yun <[email protected]> wrote:

Suppose, memblock_alloc_ranged_nid with highmem_start is success when cma_declare_contiguous_nid is called with !fixed in the system 32-bit system with PHYS_ADDR_T_64BIT enabled with memblock.bottom_up == false.

Quoted 9 lines

But, Next trial to memblock_alloc_range_nid to allocate in [SIZE_4G, limits) nullifies former success allocated addr and it retries to memblock_alloc_ragne_nid again.

In this situation, first success address area is lost.

By changing order of allocation (SIZE_4G, high_memory and base) and checking whether allocated prevents potential memory lost.

The changelog is hard to follow. I redid it as below - please review and correct if needed?

From: Levi Yun <[email protected]>
Subject: mm/cma: fix potential memory loss on cma_declare_contiguous_nid
Date: Wed, 18 Jan 2023 17:05:23 +0900

Suppose memblock_alloc_range_nid() with highmem_start succeeds when cma_declare_contiguous_nid is called with !fixed on a 32-bit system with PHYS_ADDR_T_64BIT enabled with memblock.bottom_up == false.

But the next trial to memblock_alloc_range_nid() to allocate in [SIZE_4G, limits) nullifies former successfully allocated addr and it retries memblock_alloc_ragne_nid().

In this situation, the first successfully allocated address area is lost.

Change the order of allocation (SIZE_4G, high_memory and base) and check whether the allocated succeeded to prevent potential memory loss.

---

 mm/cma.c |   24 ++++++++++++------------
 1 file changed, 12 insertions(+), 12 deletions(-)
mm/cma.c renamed from mm/cma.c~mm-cma-fix-potential-memory-loss-on-cma_declare_contiguous_nid +12 -12
@@ -331,18 +331,6 @@ int __init cma_declare_contiguous_nid(ph
 		phys_addr_t addr = 0;
 
 		/*
-		 * All pages in the reserved area must come from the same zone.
-		 * If the requested region crosses the low/high memory boundary,
-		 * try allocating from high memory first and fall back to low
-		 * memory in case of failure.
-		 */
-		if (base < highmem_start && limit > highmem_start) {
-			addr = memblock_alloc_range_nid(size, alignment,
-					highmem_start, limit, nid, true);
-			limit = highmem_start;
-		}
-
-		/*
 		 * If there is enough memory, try a bottom-up allocation first.
 		 * It will place the new cma area close to the start of the node
 		 * and guarantee that the compaction is moving pages out of the
@@ -359,6 +347,18 @@ int __init cma_declare_contiguous_nid(ph
 		}
 #endif
 
+		/*
+		 * All pages in the reserved area must come from the same zone.
+		 * If the requested region crosses the low/high memory boundary,
+		 * try allocating from high memory first and fall back to low
+		 * memory in case of failure.
+		 */
+		if (!addr && base < highmem_start && limit > highmem_start) {
+			addr = memblock_alloc_range_nid(size, alignment,
+					highmem_start, limit, nid, true);
+			limit = highmem_start;
+		}
+
 		if (!addr) {
 			addr = memblock_alloc_range_nid(size, alignment, base,
 					limit, nid, true);

_