ANDROID: integrate GCMA with CMA using dt-bindings

This patch introduces a new "guarantee" property for shared-dma-pool.
With the property, admin can create specific memory pool as
GCMA-based-CMA if they care about allocation success rate and latency
without reserving a dedicated memory area.

Bug: 236887352
Bug: 270089503
Bug: 391879760
Change-Id: I3a019f889f7ffa4d2255609c19b15078b9c5b6a4
Signed-off-by: Minchan Kim <minchan@google.com>
Signed-off-by: Suren Baghdasaryan <surenb@google.com>

arch/powerpc/kernel/fadump.c

@@ -95,7 +95,7 @@ void __init fadump_cma_init(void)
 	base = fw_dump.reserve_dump_area_start;
 	size = fw_dump.boot_memory_size;
 
-	rc = cma_init_reserved_mem(base, size, 0, "fadump_cma", &fadump_cma);
+	rc = cma_init_reserved_mem(base, size, 0, "fadump_cma", &fadump_cma, false);
 	if (rc) {
 		pr_err("Failed to init cma area for firmware-assisted dump,%d\n", rc);
 		/*

include/linux/cma.h

@@ -43,7 +43,8 @@ static inline int __init cma_declare_contiguous(phys_addr_t base,
 extern int cma_init_reserved_mem(phys_addr_t base, phys_addr_t size,
 					unsigned int order_per_bit,
 					const char *name,
-					struct cma **res_cma);
+					struct cma **res_cma,
+					bool gcma);
 extern struct page *__cma_alloc(struct cma *cma, unsigned long count, unsigned int align,
 				gfp_t gfp_mask);
 extern struct page *cma_alloc(struct cma *cma, unsigned long count, unsigned int align,

kernel/dma/contiguous.c

@@ -465,6 +465,7 @@ static int __init rmem_cma_setup(struct reserved_mem *rmem)
 	unsigned long node = rmem->fdt_node;
 	bool default_cma = of_get_flat_dt_prop(node, "linux,cma-default", NULL);
 	struct cma *cma;
+	bool gcma;
 	int err;
 
 	if (size_cmdline != -1 && default_cma) {
@@ -482,7 +483,15 @@ static int __init rmem_cma_setup(struct reserved_mem *rmem)
 		return -EINVAL;
 	}
 
-	err = cma_init_reserved_mem(rmem->base, rmem->size, 0, rmem->name, &cma);
+	gcma = !!of_get_flat_dt_prop(node, "guarantee", NULL);
+#ifndef CONFIG_GCMA
+	if (gcma) {
+		pr_err("Reserved memory: unable to setup GCMA region, GCMA is not enabled\n");
+		return -EINVAL;
+	}
+#endif
+	err = cma_init_reserved_mem(rmem->base, rmem->size, 0, rmem->name,
+				    &cma, gcma);
 	if (err) {
 		pr_err("Reserved memory: unable to setup CMA region\n");
 		return err;

mm/cma.c

@@ -29,6 +29,7 @@
 #include <linux/kmemleak.h>
 #include <linux/sched.h>
 #include <linux/jiffies.h>
+#include <linux/gcma.h>
 #define CREATE_TRACE_POINTS
 #include <trace/events/cma.h>
 #undef CREATE_TRACE_POINTS
@@ -126,9 +127,14 @@ static void __init cma_activate_area(struct cma *cma)
 			goto not_in_zone;
 	}
 
-	for (pfn = base_pfn; pfn < base_pfn + cma->count;
-	     pfn += pageblock_nr_pages)
-		init_cma_reserved_pageblock(pfn_to_page(pfn));
+	if (cma->gcma) {
+		register_gcma_area(cma->name, PFN_PHYS(base_pfn),
+				   cma->count << PAGE_SHIFT);
+	} else {
+		for (pfn = base_pfn; pfn < base_pfn + cma->count;
+		     pfn += pageblock_nr_pages)
+			init_cma_reserved_pageblock(pfn_to_page(pfn));
+	}
 
 	spin_lock_init(&cma->lock);
 
@@ -186,7 +192,7 @@ void __init cma_reserve_pages_on_error(struct cma *cma)
 int __init cma_init_reserved_mem(phys_addr_t base, phys_addr_t size,
 				 unsigned int order_per_bit,
 				 const char *name,
-				 struct cma **res_cma)
+				 struct cma **res_cma, bool gcma)
 {
 	struct cma *cma;
 
@@ -212,11 +218,13 @@ int __init cma_init_reserved_mem(phys_addr_t base, phys_addr_t size,
 	if (name)
 		snprintf(cma->name, CMA_MAX_NAME, name);
 	else
-		snprintf(cma->name, CMA_MAX_NAME,  "cma%d\n", cma_area_count);
+		snprintf(cma->name, CMA_MAX_NAME,
+			 gcma ? "gcma%d\n" : "cma%d\n", cma_area_count);
 
 	cma->base_pfn = PFN_DOWN(base);
 	cma->count = size >> PAGE_SHIFT;
 	cma->order_per_bit = order_per_bit;
+	cma->gcma = gcma;
 	*res_cma = cma;
 	cma_area_count++;
 	totalcma_pages += cma->count;
@@ -379,7 +387,8 @@ int __init cma_declare_contiguous_nid(phys_addr_t base,
 		base = addr;
 	}
 
-	ret = cma_init_reserved_mem(base, size, order_per_bit, name, res_cma);
+	ret = cma_init_reserved_mem(base, size, order_per_bit, name, res_cma,
+				    false);
 	if (ret)
 		goto free_mem;
 
@@ -509,7 +518,12 @@ struct page *__cma_alloc(struct cma *cma, unsigned long count,
 
 		pfn = cma->base_pfn + (bitmap_no << cma->order_per_bit);
 		mutex_lock(&cma_mutex);
-		ret = alloc_contig_range(pfn, pfn + count, MIGRATE_CMA, gfp);
+		if (cma->gcma) {
+			gcma_alloc_range(pfn, pfn + count - 1);
+			ret = 0;
+		} else {
+			ret = alloc_contig_range(pfn, pfn + count, MIGRATE_CMA, gfp);
+		}
 		mutex_unlock(&cma_mutex);
 		if (ret == 0) {
 			page = pfn_to_page(pfn);
@@ -632,7 +646,10 @@ bool cma_release(struct cma *cma, const struct page *pages,
 
 	VM_BUG_ON(pfn + count > cma->base_pfn + cma->count);
 
-	free_contig_range(pfn, count);
+	if (cma->gcma)
+		gcma_free_range(pfn, pfn + count - 1);
+	else
+		free_contig_range(pfn, count);
 	cma_clear_bitmap(cma, pfn, count);
 	cma_sysfs_account_release_pages(cma, count);
 	trace_cma_release(cma->name, pfn, pages, count);

mm/cma.h

@@ -23,6 +23,7 @@ struct cma {
 	struct debugfs_u32_array dfs_bitmap;
 #endif
 	char name[CMA_MAX_NAME];
+	bool gcma;
 #ifdef CONFIG_CMA_SYSFS
 	/* the number of CMA page successful allocations */
 	atomic64_t nr_pages_succeeded;

mm/cma_sysfs.c

@@ -62,6 +62,15 @@ static ssize_t release_pages_success_show(struct kobject *kobj,
 }
 CMA_ATTR_RO(release_pages_success);
 
+static ssize_t gcma_show(struct kobject *kobj,
+			 struct kobj_attribute *attr, char *buf)
+{
+	struct cma *cma = cma_from_kobj(kobj);
+
+	return sysfs_emit(buf, "%d\n", cma->gcma);
+}
+CMA_ATTR_RO(gcma);
+
 static void cma_kobj_release(struct kobject *kobj)
 {
 	struct cma *cma = cma_from_kobj(kobj);
@@ -75,6 +84,7 @@ static struct attribute *cma_attrs[] = {
 	&alloc_pages_success_attr.attr,
 	&alloc_pages_fail_attr.attr,
 	&release_pages_success_attr.attr,
+	&gcma_attr.attr,
 	NULL,
 };
 ATTRIBUTE_GROUPS(cma);