Re: [PATCHv4] mm: Don't offset memmap for flatmem

From: Tony Luck <tony.luck@gmail.com>
To: Laura Abbott <labbott@fedoraproject.org>
Cc: Srinivas Kandagatla <srinivas.kandagatla@linaro.org>,
	Vlastimil Babka <vbabka@suse.cz>,
	Bjorn Andersson <bjorn.andersson@sonymobile.com>,
	Laura Abbott <laura@labbott.name>,
	Santosh Shilimkar <ssantosh@kernel.org>,
	Russell King <rmk@arm.linux.org.uk>,
	Kevin Hilman <khilman@linaro.org>, Arnd Bergman <arnd@arndb.de>,
	Stephen Boyd <sboyd@codeaurora.org>,
	Andy Gross <agross@codeaurora.org>, Mel Gorman <mgorman@suse.de>,
	Steven Rostedt <rostedt@goodmis.org>,
	"linux-arm-kernel@lists.infradead.org" 
	<linux-arm-kernel@lists.infradead.org>,
	Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
	"linux-mm@kvack.org" <linux-mm@kvack.org>,
	Andrew Morton <akpm@linux-foundation.org>,
	Laura Abbott <lauraa@codeaurora.org>
Subject: Re: [PATCHv4] mm: Don't offset memmap for flatmem
Date: Mon, 9 Nov 2015 15:20:58 -0800	[thread overview]
Message-ID: <CA+8MBbLGdYfQRPnVmT=te1y3C7PhCcXqbDGXb7LtqvCWTA+vDQ@mail.gmail.com> (raw)
In-Reply-To: <1444253335-5811-1-git-send-email-labbott@fedoraproject.org>

[-- Attachment #1: Type: text/plain, Size: 866 bytes --]

> @@ -4984,9 +4987,9 @@ static void __init_refok alloc_node_mem_map(struct pglist_data *pgdat)
>          */
>         if (pgdat == NODE_DATA(0)) {
>                 mem_map = NODE_DATA(0)->node_mem_map;
> -#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
> +#if defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) || defined(CONFIG_FLATMEM)
>                 if (page_to_pfn(mem_map) != pgdat->node_start_pfn)
> -                       mem_map -= (pgdat->node_start_pfn - ARCH_PFN_OFFSET);
> +                       mem_map -= offset;
>  #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
>         }
>  #endif

This piece breaks ia64.  See the comment earlier in the function
that "ia64 gets its own node_mem_map" ... so we skip the initialization
of offset ... and arrive down here and just subtract "0" from mem_map.

Attached patch fixes ia64 ... does ARM still work if this is applied?

-Tony

[-- Attachment #2: fixia64.patch --]
[-- Type: text/x-patch, Size: 1226 bytes --]

diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 208e4c7e771b..c4705095a516 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -5266,6 +5266,7 @@ static void __paginginit free_area_init_core(struct pglist_data *pgdat)
 
 static void __init_refok alloc_node_mem_map(struct pglist_data *pgdat)
 {
+	unsigned long __maybe_unused start = 0;
 	unsigned long __maybe_unused offset = 0;
 
 	/* Skip empty nodes */
@@ -5273,9 +5274,11 @@ static void __init_refok alloc_node_mem_map(struct pglist_data *pgdat)
 		return;
 
 #ifdef CONFIG_FLAT_NODE_MEM_MAP
+	start = pgdat->node_start_pfn & ~(MAX_ORDER_NR_PAGES - 1);
+	offset = pgdat->node_start_pfn - start;
 	/* ia64 gets its own node_mem_map, before this, without bootmem */
 	if (!pgdat->node_mem_map) {
-		unsigned long size, start, end;
+		unsigned long size, end;
 		struct page *map;
 
 		/*
@@ -5284,7 +5287,6 @@ static void __init_refok alloc_node_mem_map(struct pglist_data *pgdat)
 		 * for the buddy allocator to function correctly.
 		 */
 		start = pgdat->node_start_pfn & ~(MAX_ORDER_NR_PAGES - 1);
-		offset = pgdat->node_start_pfn - start;
 		end = pgdat_end_pfn(pgdat);
 		end = ALIGN(end, MAX_ORDER_NR_PAGES);
 		size =  (end - start) * sizeof(struct page);
