[PATCH 0/4] ia64: switch to NO_BOOTMEM

[PATCH 0/4] ia64: switch to NO_BOOTMEM

[Mike Rapoport]

Hi,

These patches convert ia64 to use NO_BOOTMEM.

The first two patches are cleanups, the third patches reduces usage of
'struct bootmem_data' for easier transition and the forth patch actually
replaces bootmem with memblock + nobootmem.

I've tested the sim_defconfig with the ski simulator and build tested other
defconfigs.

Mike Rapoport (4):
  ia64: contig/paging_init: reduce code duplication
  ia64: remove unused num_dma_physpages member from 'struct early_node_data'
  ia64: use mem_data to detect nodes' minimal and maximal PFNs
  ia64: switch to NO_BOOTMEM

 arch/ia64/Kconfig        |   1 +
 arch/ia64/kernel/setup.c |  11 +++-
 arch/ia64/mm/contig.c    |  75 +++-----------------------
 arch/ia64/mm/discontig.c | 134 ++++++-----------------------------------------
 4 files changed, 33 insertions(+), 188 deletions(-)

-- 
2.7.4

[PATCH 1/4] ia64: contig/paging_init: reduce code duplication

[Mike Rapoport]

The FLATMEM version of paging_init has calls to free_area_init_nodes() in
the end of every branch of 'if' and 'ifdef' statements.

Let's call this function outside the 'ifdef' and 'if' statements instead.

Signed-off-by: Mike Rapoport <rppt@linux.vnet.ibm.com>
---
 arch/ia64/mm/contig.c | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)

diff --git a/arch/ia64/mm/contig.c b/arch/ia64/mm/contig.c
index 7d64b30..1835144 100644
--- a/arch/ia64/mm/contig.c
+++ b/arch/ia64/mm/contig.c
@@ -248,7 +248,6 @@ paging_init (void)
 	efi_memmap_walk(find_largest_hole, (u64 *)&max_gap);
 	if (max_gap < LARGE_GAP) {
 		vmem_map = (struct page *) 0;
-		free_area_init_nodes(max_zone_pfns);
 	} else {
 		unsigned long map_size;
 
@@ -266,13 +265,12 @@ paging_init (void)
 		 */
 		NODE_DATA(0)->node_mem_map = vmem_map +
 			find_min_pfn_with_active_regions();
-		free_area_init_nodes(max_zone_pfns);
 
 		printk("Virtual mem_map starts at 0x%p\n", mem_map);
 	}
 #else /* !CONFIG_VIRTUAL_MEM_MAP */
 	memblock_add_node(0, PFN_PHYS(max_low_pfn), 0);
-	free_area_init_nodes(max_zone_pfns);
 #endif /* !CONFIG_VIRTUAL_MEM_MAP */
+	free_area_init_nodes(max_zone_pfns);
 	zero_page_memmap_ptr = virt_to_page(ia64_imva(empty_zero_page));
 }
-- 
2.7.4

[PATCH 2/4] ia64: remove unused num_dma_physpages member from 'struct early_node_data'

[Mike Rapoport]

Since commit 05e0caad3b7b ("[PATCH] Have ia64 use add_active_range() and
free_area_init_nodes") the num_dma_physpages member of 'struct
early_node_data' is calculated but never used. Remove it.

Signed-off-by: Mike Rapoport <rppt@linux.vnet.ibm.com>
---
 arch/ia64/mm/discontig.c | 8 --------
 1 file changed, 8 deletions(-)

diff --git a/arch/ia64/mm/discontig.c b/arch/ia64/mm/discontig.c
index 7d9bd20..6148ea8 100644
--- a/arch/ia64/mm/discontig.c
+++ b/arch/ia64/mm/discontig.c
@@ -38,9 +38,6 @@ struct early_node_data {
 	struct ia64_node_data *node_data;
 	unsigned long pernode_addr;
 	unsigned long pernode_size;
-#ifdef CONFIG_ZONE_DMA32
-	unsigned long num_dma_physpages;
-#endif
 	unsigned long min_pfn;
 	unsigned long max_pfn;
 };
@@ -669,11 +666,6 @@ static __init int count_node_pages(unsigned long start, unsigned long len, int n
 {
 	unsigned long end = start + len;
 
-#ifdef CONFIG_ZONE_DMA32
-	if (start <= __pa(MAX_DMA_ADDRESS))
-		mem_data[node].num_dma_physpages +=
-			(min(end, __pa(MAX_DMA_ADDRESS)) - start) >>PAGE_SHIFT;
-#endif
 	start = GRANULEROUNDDOWN(start);
 	end = GRANULEROUNDUP(end);
 	mem_data[node].max_pfn = max(mem_data[node].max_pfn,
-- 
2.7.4

[PATCH 3/4] ia64: use mem_data to detect nodes' minimal and maximal PFNs

[Mike Rapoport]

When EFI memory map is traversed to determine the extents of each node, the
minimal and maximal PFNs are stored in the bootmem_data structures. The
same information ls later stored in the mem_data array of 'struct
early_node_data'.

Switch to using mem_data from the very beginning.

Signed-off-by: Mike Rapoport <rppt@linux.vnet.ibm.com>
---
 arch/ia64/mm/discontig.c | 60 +++++++++++-------------------------------------
 1 file changed, 14 insertions(+), 46 deletions(-)

diff --git a/arch/ia64/mm/discontig.c b/arch/ia64/mm/discontig.c
index 6148ea8..8e99d8e 100644
--- a/arch/ia64/mm/discontig.c
+++ b/arch/ia64/mm/discontig.c
@@ -57,33 +57,31 @@ pg_data_t *pgdat_list[MAX_NUMNODES];
 	     (((node)*PERCPU_PAGE_SIZE) & (MAX_NODE_ALIGN_OFFSET - 1)))
 
 /**
- * build_node_maps - callback to setup bootmem structs for each node
+ * build_node_maps - callback to setup mem_data structs for each node
  * @start: physical start of range
  * @len: length of range
  * @node: node where this range resides
  *
- * We allocate a struct bootmem_data for each piece of memory that we wish to
+ * Detect extents of each piece of memory that we wish to
  * treat as a virtually contiguous block (i.e. each node). Each such block
  * must start on an %IA64_GRANULE_SIZE boundary, so we round the address down
  * if necessary.  Any non-existent pages will simply be part of the virtual
- * memmap.  We also update min_low_pfn and max_low_pfn here as we receive
- * memory ranges from the caller.
+ * memmap.
  */
 static int __init build_node_maps(unsigned long start, unsigned long len,
 				  int node)
 {
 	unsigned long spfn, epfn, end = start + len;
-	struct bootmem_data *bdp = &bootmem_node_data[node];
 
 	epfn = GRANULEROUNDUP(end) >> PAGE_SHIFT;
 	spfn = GRANULEROUNDDOWN(start) >> PAGE_SHIFT;
 
-	if (!bdp->node_low_pfn) {
-		bdp->node_min_pfn = spfn;
-		bdp->node_low_pfn = epfn;
+	if (!mem_data[node].min_pfn) {
+		mem_data[node].min_pfn = spfn;
+		mem_data[node].max_pfn = epfn;
 	} else {
-		bdp->node_min_pfn = min(spfn, bdp->node_min_pfn);
-		bdp->node_low_pfn = max(epfn, bdp->node_low_pfn);
+		mem_data[node].min_pfn = min(spfn, mem_data[node].min_pfn);
+		mem_data[node].max_pfn = max(epfn, mem_data[node].max_pfn);
 	}
 
 	return 0;
@@ -323,19 +321,18 @@ static int __init find_pernode_space(unsigned long start, unsigned long len,
 {
 	unsigned long spfn, epfn;
 	unsigned long pernodesize = 0, pernode, pages, mapsize;
-	struct bootmem_data *bdp = &bootmem_node_data[node];
 
 	spfn = start >> PAGE_SHIFT;
 	epfn = (start + len) >> PAGE_SHIFT;
 
-	pages = bdp->node_low_pfn - bdp->node_min_pfn;
+	pages = mem_data[node].max_pfn - mem_data[node].min_pfn;
 	mapsize = bootmem_bootmap_pages(pages) << PAGE_SHIFT;
 
 	/*
 	 * Make sure this memory falls within this node's usable memory
 	 * since we may have thrown some away in build_maps().
 	 */
-	if (spfn < bdp->node_min_pfn || epfn > bdp->node_low_pfn)
+	if (spfn < mem_data[node].min_pfn || epfn > mem_data[node].max_pfn)
 		return 0;
 
 	/* Don't setup this node's local space twice... */
@@ -397,7 +394,7 @@ static void __init reserve_pernode_space(void)
 		bdp = pdp->bdata;
 
 		/* First the bootmem_map itself */
-		pages = bdp->node_low_pfn - bdp->node_min_pfn;
+		pages = mem_data[node].max_pfn - mem_data[node].min_pfn;
 		size = bootmem_bootmap_pages(pages) << PAGE_SHIFT;
 		base = __pa(bdp->node_bootmem_map);
 		reserve_bootmem_node(pdp, base, size, BOOTMEM_DEFAULT);
@@ -541,10 +538,8 @@ void __init find_memory(void)
 	efi_memmap_walk(find_max_min_low_pfn, NULL);
 
 	for_each_online_node(node)
-		if (bootmem_node_data[node].node_low_pfn) {
+		if (mem_data[node].min_pfn)
 			node_clear(node, memory_less_mask);
-			mem_data[node].min_pfn = ~0UL;
-		}
 
 	efi_memmap_walk(filter_memory, register_active_ranges);
 
@@ -568,8 +563,8 @@ void __init find_memory(void)
 
 		init_bootmem_node(pgdat_list[node],
 				  map>>PAGE_SHIFT,
-				  bdp->node_min_pfn,
-				  bdp->node_low_pfn);
+				  mem_data[node].min_pfn,
+				  mem_data[node].max_pfn);
 	}
 
 	efi_memmap_walk(filter_rsvd_memory, free_node_bootmem);
@@ -652,31 +647,6 @@ void call_pernode_memory(unsigned long start, unsigned long len, void *arg)
 }
 
 /**
- * count_node_pages - callback to build per-node memory info structures
- * @start: physical start of range
- * @len: length of range
- * @node: node where this range resides
- *
- * Each node has it's own number of physical pages, DMAable pages, start, and
- * end page frame number.  This routine will be called by call_pernode_memory()
- * for each piece of usable memory and will setup these values for each node.
- * Very similar to build_maps().
- */
-static __init int count_node_pages(unsigned long start, unsigned long len, int node)
-{
-	unsigned long end = start + len;
-
-	start = GRANULEROUNDDOWN(start);
-	end = GRANULEROUNDUP(end);
-	mem_data[node].max_pfn = max(mem_data[node].max_pfn,
-				     end >> PAGE_SHIFT);
-	mem_data[node].min_pfn = min(mem_data[node].min_pfn,
-				     start >> PAGE_SHIFT);
-
-	return 0;
-}
-
-/**
  * paging_init - setup page tables
  *
  * paging_init() sets up the page tables for each node of the system and frees
@@ -692,8 +662,6 @@ void __init paging_init(void)
 
 	max_dma = virt_to_phys((void *) MAX_DMA_ADDRESS) >> PAGE_SHIFT;
 
-	efi_memmap_walk(filter_rsvd_memory, count_node_pages);
-
 	sparse_memory_present_with_active_regions(MAX_NUMNODES);
 	sparse_init();
 
-- 
2.7.4

[PATCH 4/4] ia64: switch to NO_BOOTMEM

[Mike Rapoport]

Since ia64 already uses memblock to register available physical memory it
is only required to move the calls to register_active_ranges() that wrap
memblock_add_node() earlier and replace bootmem memory reservations with
memblock_reserve(). Of course, all the code that find the place to put the
bootmem bitmap is removed.

Signed-off-by: Mike Rapoport <rppt@linux.vnet.ibm.com>
---
 arch/ia64/Kconfig        |  1 +
 arch/ia64/kernel/setup.c | 11 ++++++-
 arch/ia64/mm/contig.c    | 71 ++++------------------------------------------
 arch/ia64/mm/discontig.c | 74 ++++--------------------------------------------
 4 files changed, 22 insertions(+), 135 deletions(-)

diff --git a/arch/ia64/Kconfig b/arch/ia64/Kconfig
index ff86142..107b138 100644
--- a/arch/ia64/Kconfig
+++ b/arch/ia64/Kconfig
@@ -31,6 +31,7 @@ config IA64
 	select HAVE_ARCH_TRACEHOOK
 	select HAVE_MEMBLOCK
 	select HAVE_MEMBLOCK_NODE_MAP
+	select NO_BOOTMEM
 	select HAVE_VIRT_CPU_ACCOUNTING
 	select ARCH_HAS_DMA_MARK_CLEAN
 	select ARCH_HAS_SG_CHAIN
diff --git a/arch/ia64/kernel/setup.c b/arch/ia64/kernel/setup.c
index ad43cbf..b042d0c 100644
--- a/arch/ia64/kernel/setup.c
+++ b/arch/ia64/kernel/setup.c
@@ -32,6 +32,7 @@
 #include <linux/delay.h>
 #include <linux/cpu.h>
 #include <linux/kernel.h>
+#include <linux/memblock.h>
 #include <linux/reboot.h>
 #include <linux/sched/mm.h>
 #include <linux/sched/clock.h>
@@ -383,8 +384,16 @@ reserve_memory (void)
 
 	sort_regions(rsvd_region, num_rsvd_regions);
 	num_rsvd_regions = merge_regions(rsvd_region, num_rsvd_regions);
-}
 
+	/* reserve all regions except the end of memory marker with meblock */
+	for (n = 0; n < num_rsvd_regions - 1; n++) {
+		struct rsvd_region *region = &rsvd_region[n];
+		phys_addr_t addr = __pa(region->start);
+		phys_addr_t size = region->end - region->start;
+
+		memblock_reserve(addr, size);
+	}
+}
 
 /**
  * find_initrd - get initrd parameters from the boot parameter structure
diff --git a/arch/ia64/mm/contig.c b/arch/ia64/mm/contig.c
index 1835144..e2e40bb 100644
--- a/arch/ia64/mm/contig.c
+++ b/arch/ia64/mm/contig.c
@@ -34,53 +34,6 @@ static unsigned long max_gap;
 /* physical address where the bootmem map is located */
 unsigned long bootmap_start;
 
-/**
- * find_bootmap_location - callback to find a memory area for the bootmap
- * @start: start of region
- * @end: end of region
- * @arg: unused callback data
- *
- * Find a place to put the bootmap and return its starting address in
- * bootmap_start.  This address must be page-aligned.
- */
-static int __init
-find_bootmap_location (u64 start, u64 end, void *arg)
-{
-	u64 needed = *(unsigned long *)arg;
-	u64 range_start, range_end, free_start;
-	int i;
-
-#if IGNORE_PFN0
-	if (start == PAGE_OFFSET) {
-		start += PAGE_SIZE;
-		if (start >= end)
-			return 0;
-	}
-#endif
-
-	free_start = PAGE_OFFSET;
-
-	for (i = 0; i < num_rsvd_regions; i++) {
-		range_start = max(start, free_start);
-		range_end   = min(end, rsvd_region[i].start & PAGE_MASK);
-
-		free_start = PAGE_ALIGN(rsvd_region[i].end);
-
-		if (range_end <= range_start)
-			continue; /* skip over empty range */
-
-		if (range_end - range_start >= needed) {
-			bootmap_start = __pa(range_start);
-			return -1;	/* done */
-		}
-
-		/* nothing more available in this segment */
-		if (range_end == end)
-			return 0;
-	}
-	return 0;
-}
-
 #ifdef CONFIG_SMP
 static void *cpu_data;
 /**
@@ -196,8 +149,6 @@ setup_per_cpu_areas(void)
 void __init
 find_memory (void)
 {
-	unsigned long bootmap_size;
-
 	reserve_memory();
 
 	/* first find highest page frame number */
@@ -205,21 +156,12 @@ find_memory (void)
 	max_low_pfn = 0;
 	efi_memmap_walk(find_max_min_low_pfn, NULL);
 	max_pfn = max_low_pfn;
-	/* how many bytes to cover all the pages */
-	bootmap_size = bootmem_bootmap_pages(max_pfn) << PAGE_SHIFT;
 
-	/* look for a location to hold the bootmap */
-	bootmap_start = ~0UL;
-	efi_memmap_walk(find_bootmap_location, &bootmap_size);
-	if (bootmap_start == ~0UL)
-		panic("Cannot find %ld bytes for bootmap\n", bootmap_size);
-
-	bootmap_size = init_bootmem_node(NODE_DATA(0),
-			(bootmap_start >> PAGE_SHIFT), 0, max_pfn);
-
-	/* Free all available memory, then mark bootmem-map as being in use. */
-	efi_memmap_walk(filter_rsvd_memory, free_bootmem);
-	reserve_bootmem(bootmap_start, bootmap_size, BOOTMEM_DEFAULT);
+#ifdef CONFIG_VIRTUAL_MEM_MAP
+	efi_memmap_walk(filter_memory, register_active_ranges);
+#else
+	memblock_add_node(0, PFN_PHYS(max_low_pfn), 0);
+#endif
 
 	find_initrd();
 
@@ -244,7 +186,6 @@ paging_init (void)
 	max_zone_pfns[ZONE_NORMAL] = max_low_pfn;
 
 #ifdef CONFIG_VIRTUAL_MEM_MAP
-	efi_memmap_walk(filter_memory, register_active_ranges);
 	efi_memmap_walk(find_largest_hole, (u64 *)&max_gap);
 	if (max_gap < LARGE_GAP) {
 		vmem_map = (struct page *) 0;
@@ -268,8 +209,6 @@ paging_init (void)
 
 		printk("Virtual mem_map starts at 0x%p\n", mem_map);
 	}
-#else /* !CONFIG_VIRTUAL_MEM_MAP */
-	memblock_add_node(0, PFN_PHYS(max_low_pfn), 0);
 #endif /* !CONFIG_VIRTUAL_MEM_MAP */
 	free_area_init_nodes(max_zone_pfns);
 	zero_page_memmap_ptr = virt_to_page(ia64_imva(empty_zero_page));
diff --git a/arch/ia64/mm/discontig.c b/arch/ia64/mm/discontig.c
index 8e99d8e..1928d57 100644
--- a/arch/ia64/mm/discontig.c
+++ b/arch/ia64/mm/discontig.c
@@ -20,6 +20,7 @@
 #include <linux/nmi.h>
 #include <linux/swap.h>
 #include <linux/bootmem.h>
+#include <linux/memblock.h>
 #include <linux/acpi.h>
 #include <linux/efi.h>
 #include <linux/nodemask.h>
@@ -264,7 +265,6 @@ static void __init fill_pernode(int node, unsigned long pernode,
 {
 	void *cpu_data;
 	int cpus = early_nr_cpus_node(node);
-	struct bootmem_data *bdp = &bootmem_node_data[node];
 
 	mem_data[node].pernode_addr = pernode;
 	mem_data[node].pernode_size = pernodesize;
@@ -279,8 +279,6 @@ static void __init fill_pernode(int node, unsigned long pernode,
 
 	mem_data[node].node_data = __va(pernode);
 	pernode += L1_CACHE_ALIGN(sizeof(struct ia64_node_data));
-
-	pgdat_list[node]->bdata = bdp;
 	pernode += L1_CACHE_ALIGN(sizeof(pg_data_t));
 
 	cpu_data = per_cpu_node_setup(cpu_data, node);
@@ -320,14 +318,11 @@ static int __init find_pernode_space(unsigned long start, unsigned long len,
 				     int node)
 {
 	unsigned long spfn, epfn;
-	unsigned long pernodesize = 0, pernode, pages, mapsize;
+	unsigned long pernodesize = 0, pernode;
 
 	spfn = start >> PAGE_SHIFT;
 	epfn = (start + len) >> PAGE_SHIFT;
 
-	pages = mem_data[node].max_pfn - mem_data[node].min_pfn;
-	mapsize = bootmem_bootmap_pages(pages) << PAGE_SHIFT;
-
 	/*
 	 * Make sure this memory falls within this node's usable memory
 	 * since we may have thrown some away in build_maps().
@@ -347,32 +342,13 @@ static int __init find_pernode_space(unsigned long start, unsigned long len,
 	pernode = NODEDATA_ALIGN(start, node);
 
 	/* Is this range big enough for what we want to store here? */
-	if (start + len > (pernode + pernodesize + mapsize))
+	if (start + len > (pernode + pernodesize))
 		fill_pernode(node, pernode, pernodesize);
 
 	return 0;
 }
 
 /**
- * free_node_bootmem - free bootmem allocator memory for use
- * @start: physical start of range
- * @len: length of range
- * @node: node where this range resides
- *
- * Simply calls the bootmem allocator to free the specified ranged from
- * the given pg_data_t's bdata struct.  After this function has been called
- * for all the entries in the EFI memory map, the bootmem allocator will
- * be ready to service allocation requests.
- */
-static int __init free_node_bootmem(unsigned long start, unsigned long len,
-				    int node)
-{
-	free_bootmem_node(pgdat_list[node], start, len);
-
-	return 0;
-}
-
-/**
  * reserve_pernode_space - reserve memory for per-node space
  *
  * Reserve the space used by the bootmem maps & per-node space in the boot
@@ -381,28 +357,17 @@ static int __init free_node_bootmem(unsigned long start, unsigned long len,
  */
 static void __init reserve_pernode_space(void)
 {
-	unsigned long base, size, pages;
-	struct bootmem_data *bdp;
+	unsigned long base, size;
 	int node;
 
 	for_each_online_node(node) {
-		pg_data_t *pdp = pgdat_list[node];
-
 		if (node_isset(node, memory_less_mask))
 			continue;
 
-		bdp = pdp->bdata;
-
-		/* First the bootmem_map itself */
-		pages = mem_data[node].max_pfn - mem_data[node].min_pfn;
-		size = bootmem_bootmap_pages(pages) << PAGE_SHIFT;
-		base = __pa(bdp->node_bootmem_map);
-		reserve_bootmem_node(pdp, base, size, BOOTMEM_DEFAULT);
-
 		/* Now the per-node space */
 		size = mem_data[node].pernode_size;
 		base = __pa(mem_data[node].pernode_addr);
-		reserve_bootmem_node(pdp, base, size, BOOTMEM_DEFAULT);
+		memblock_reserve(base, size);
 	}
 }
 
@@ -522,6 +487,7 @@ void __init find_memory(void)
 	int node;
 
 	reserve_memory();
+	efi_memmap_walk(filter_memory, register_active_ranges);
 
 	if (num_online_nodes() == 0) {
 		printk(KERN_ERR "node info missing!\n");
@@ -541,34 +507,6 @@ void __init find_memory(void)
 		if (mem_data[node].min_pfn)
 			node_clear(node, memory_less_mask);
 
-	efi_memmap_walk(filter_memory, register_active_ranges);
-
-	/*
-	 * Initialize the boot memory maps in reverse order since that's
-	 * what the bootmem allocator expects
-	 */
-	for (node = MAX_NUMNODES - 1; node >= 0; node--) {
-		unsigned long pernode, pernodesize, map;
-		struct bootmem_data *bdp;
-
-		if (!node_online(node))
-			continue;
-		else if (node_isset(node, memory_less_mask))
-			continue;
-
-		bdp = &bootmem_node_data[node];
-		pernode = mem_data[node].pernode_addr;
-		pernodesize = mem_data[node].pernode_size;
-		map = pernode + pernodesize;
-
-		init_bootmem_node(pgdat_list[node],
-				  map>>PAGE_SHIFT,
-				  mem_data[node].min_pfn,
-				  mem_data[node].max_pfn);
-	}
-
-	efi_memmap_walk(filter_rsvd_memory, free_node_bootmem);
-
 	reserve_pernode_space();
 	memory_less_nodes();
 	initialize_pernode_data();
-- 
2.7.4

Re: [PATCH 0/4] ia64: switch to NO_BOOTMEM

[Luck, Tony]

On Mon, Jul 23, 2018 at 08:56:54AM +0300, Mike Rapoport wrote:
> Hi,
> 
> These patches convert ia64 to use NO_BOOTMEM.
> 
> The first two patches are cleanups, the third patches reduces usage of
> 'struct bootmem_data' for easier transition and the forth patch actually
> replaces bootmem with memblock + nobootmem.
> 
> I've tested the sim_defconfig with the ski simulator and build tested other
> defconfigs.

Boots OK on my real ia64 system.

Unless somebody else sees an issue I'll push to Linus nest merge window.

-Tony