mm: Be allowed to alloc CDM node memory for MPOL_BIND

euler inclusion
category: feature
bugzilla: 11082
CVE: NA
-----------------

CDM nodes should not be part of mems_allowed, However,
It must be allowed to alloc from CDM node, when mpol->mode was MPOL_BIND.
Signed-off-by: NLijun Fang <fanglijun3@huawei.com>
Reviewed-by: Nzhong jiang <zhongjiang@huawei.com>
Signed-off-by: NYang Yingliang <yangyingliang@huawei.com>

include/linux/mm.h

@@ -204,10 +204,6 @@ extern unsigned int kobjsize(const void *objp);
 #define VM_NORESERVE	0x00200000	/* should the VM suppress accounting */
 #define VM_HUGETLB	0x00400000	/* Huge TLB Page VM */
 
-#ifdef CONFIG_COHERENT_DEVICE
-#define VM_CDM		0x00800000	/* Contains coherent device memory */
-#endif
-
 #define VM_SYNC		0x00800000	/* Synchronous page faults */
 #define VM_ARCH_1	0x01000000	/* Architecture-specific flag */
 #define VM_WIPEONFORK	0x02000000	/* Wipe VMA contents in child. */
@@ -224,6 +220,10 @@ extern unsigned int kobjsize(const void *objp);
 #define VM_NOHUGEPAGE	0x40000000	/* MADV_NOHUGEPAGE marked this vma */
 #define VM_MERGEABLE	0x80000000	/* KSM may merge identical pages */
 
+#ifdef CONFIG_COHERENT_DEVICE
+#define VM_CDM		0x100000000	/* Contains coherent device memory */
+#endif
+
 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
 #define VM_HIGH_ARCH_BIT_0	32	/* bit only usable on 64-bit architectures */
 #define VM_HIGH_ARCH_BIT_1	33	/* bit only usable on 64-bit architectures */

mm/hugetlb.c

@@ -881,13 +881,17 @@ static struct page *dequeue_huge_page_node_exact(struct hstate *h, int nid)
 }
 
 static struct page *dequeue_huge_page_nodemask(struct hstate *h, gfp_t gfp_mask, int nid,
-		nodemask_t *nmask)
+		nodemask_t *nmask, struct mempolicy *mpol)
 {
 	unsigned int cpuset_mems_cookie;
 	struct zonelist *zonelist;
 	struct zone *zone;
 	struct zoneref *z;
 	int node = -1;
+	bool mbind_cdmnode = false;
+
+	if (is_cdm_node(nid) && mpol != NULL && mpol->mode == MPOL_BIND)
+		mbind_cdmnode = true;
 
 	zonelist = node_zonelist(nid, gfp_mask);
 
@@ -896,7 +900,8 @@ static struct page *dequeue_huge_page_nodemask(struct hstate *h, gfp_t gfp_mask,
 	for_each_zone_zonelist_nodemask(zone, z, zonelist, gfp_zone(gfp_mask), nmask) {
 		struct page *page;
 
-		if (!cpuset_zone_allowed(zone, gfp_mask))
+		if (!cpuset_zone_allowed(zone, gfp_mask) &&
+		    mbind_cdmnode == false)
 			continue;
 		/*
 		 * no need to ask again on the same node. Pool is node rather than
@@ -951,7 +956,7 @@ static struct page *dequeue_huge_page_vma(struct hstate *h,
 
 	gfp_mask = htlb_alloc_mask(h);
 	nid = huge_node(vma, address, gfp_mask, &mpol, &nodemask);
-	page = dequeue_huge_page_nodemask(h, gfp_mask, nid, nodemask);
+	page = dequeue_huge_page_nodemask(h, gfp_mask, nid, nodemask, mpol);
 	if (page && !avoid_reserve && vma_has_reserves(vma, chg)) {
 		SetPagePrivate(page);
 		h->resv_huge_pages--;
@@ -1638,7 +1643,7 @@ struct page *alloc_huge_page_node(struct hstate *h, int nid)
 
 	spin_lock(&hugetlb_lock);
 	if (h->free_huge_pages - h->resv_huge_pages > 0)
-		page = dequeue_huge_page_nodemask(h, gfp_mask, nid, NULL);
+		page = dequeue_huge_page_nodemask(h, gfp_mask, nid, NULL, NULL);
 	spin_unlock(&hugetlb_lock);
 
 	if (!page)
@@ -1657,7 +1662,8 @@ struct page *alloc_huge_page_nodemask(struct hstate *h, int preferred_nid,
 	if (h->free_huge_pages - h->resv_huge_pages > 0) {
 		struct page *page;
 
-		page = dequeue_huge_page_nodemask(h, gfp_mask, preferred_nid, nmask);
+		page = dequeue_huge_page_nodemask(h, gfp_mask, preferred_nid,
+						  nmask, NULL);
 		if (page) {
 			spin_unlock(&hugetlb_lock);
 			return page;

mm/internal.h

@@ -485,6 +485,10 @@ unsigned long reclaim_clean_pages_from_list(struct zone *zone,
 #define ALLOC_CPUSET		0x40 /* check for correct cpuset */
 #define ALLOC_CMA		0x80 /* allow allocations from CMA areas */
 
+#ifdef CONFIG_COHERENT_DEVICE
+#define ALLOC_CDM		0x100
+#endif
+
 enum ttu_flags;
 struct tlbflush_unmap_batch;
 

mm/mempolicy.c

@@ -248,6 +248,9 @@ static int mpol_set_nodemask(struct mempolicy *pol,
 	nodes_and(nsc->mask1,
 		  cpuset_current_mems_allowed, node_states[N_MEMORY]);
 
+#ifdef CONFIG_COHERENT_DEVICE
+	nodes_or(nsc->mask1, cdmmask, nsc->mask1);
+#endif
 	VM_BUG_ON(!nodes);
 	if (pol->mode == MPOL_PREFERRED && nodes_empty(*nodes))
 		nodes = NULL;	/* explicit local allocation */
@@ -1776,7 +1779,8 @@ static nodemask_t *policy_nodemask(gfp_t gfp, struct mempolicy *policy)
 	/* Lower zones don't get a nodemask applied for MPOL_BIND */
 	if (unlikely(policy->mode == MPOL_BIND) &&
 			apply_policy_zone(policy, gfp_zone(gfp)) &&
-			cpuset_nodemask_valid_mems_allowed(&policy->v.nodes))
+			(cpuset_nodemask_valid_mems_allowed(&policy->v.nodes) ||
+			nodemask_has_cdm(policy->v.nodes)))
 		return &policy->v.nodes;
 
 	return NULL;

mm/page_alloc.c

@@ -3301,7 +3301,11 @@ get_page_from_freelist(gfp_t gfp_mask, unsigned int order, int alloc_flags,
 
 		if (cpusets_enabled() &&
 			(alloc_flags & ALLOC_CPUSET) &&
-			!__cpuset_zone_allowed(zone, gfp_mask))
+			!__cpuset_zone_allowed(zone, gfp_mask)
+#ifdef CONFIG_COHERENT_DEVICE
+			&& !(alloc_flags & ALLOC_CDM)
+#endif
+		)
 				continue;
 		/*
 		 * When allocating a page cache page for writing, we
@@ -4343,8 +4347,12 @@ static inline bool prepare_alloc_pages(gfp_t gfp_mask, unsigned int order,
 		*alloc_mask |= __GFP_HARDWALL;
 		if (!ac->nodemask)
 			ac->nodemask = &cpuset_current_mems_allowed;
-		else
+		else {
 			*alloc_flags |= ALLOC_CPUSET;
+#ifdef CONFIG_COHERENT_DEVICE
+			*alloc_flags |= ALLOC_CDM;
+#endif
+		}
 	}
 
 	fs_reclaim_acquire(gfp_mask);