[RFC v1 2/4] x86: mm: simplify enabling direct_gbpages

["Luis R. Rodriguez" <mcgrof@do-not-panic.com>]

From: "Luis R. Rodriguez" <mcgrof@suse.com>

direct_gbpages can be force enabled as an early parameter
but not really have taken effect when DEBUG_PAGEALLOC
or KMEMCHECK is enabled. You can also enable direct_gbpages
right now if you have an x86_64 architecture but your CPU
doesn't really have support for this feature. In both cases
PG_LEVEL_1G won't actually be enabled but direct_gbpages is used
in other areas under the assumptions that PG_LEVEL_1G
was set. Fix this by putting together all requirements
which make this feature sensible to enable under, and only
enable both finally flipping on PG_LEVEL_1G and leaving
PG_LEVEL_1G set when this is true.

We only enable this feature then to be possible on sensible
builds defined by the new ENABLE_DIRECT_GBPAGES. If the
CPU has support for it you can either enable this by using
the DIRECT_GBPAGES option or using the early kernel parameter.
If a platform had support for this you can always force disable
it as well.

Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Tony Lindgren <tony@atomide.com>
Cc: linux-kernel@vger.kernel.org
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: x86@kernel.org
Cc: Juergen Gross <jgross@suse.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Toshi Kani <toshi.kani@hp.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Jan Beulich <JBeulich@suse.com>
Cc: Pavel Machek <pavel@ucw.cz>
Cc: Xishi Qiu <qiuxishi@huawei.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: David Vrabel <david.vrabel@citrix.com>
Cc: Borislav Petkov <bp@suse.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Dexuan Cui <decui@microsoft.com>
Signed-off-by: Luis R. Rodriguez <mcgrof@suse.com>
---
 arch/x86/Kconfig       | 18 +++++++++++++-----
 arch/x86/mm/init.c     | 17 +++++++++--------
 arch/x86/mm/pageattr.c |  2 --
 3 files changed, 22 insertions(+), 15 deletions(-)

diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
index fb8e8cd..f3fd260 100644
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -1300,14 +1300,22 @@ config ARCH_DMA_ADDR_T_64BIT
 	def_bool y
 	depends on X86_64 || HIGHMEM64G
 
+config ENABLE_DIRECT_GBPAGES
+	def_bool y
+	depends on X86_64 && !DEBUG_PAGEALLOC && !KMEMCHECK
+
 config DIRECT_GBPAGES
 	bool "Enable 1GB pages for kernel pagetables" if EXPERT
 	default y
-	depends on X86_64
-	---help---
-	  Allow the kernel linear mapping to use 1GB pages on CPUs that
-	  support it. This can improve the kernel's performance a tiny bit by
-	  reducing TLB pressure. If in doubt, say "Y".
+	depends on ENABLE_DIRECT_GBPAGES
+	---help---
+	  Enable by default the kernel linear mapping to use 1GB pages on CPUs
+	  that support it. This can improve the kernel's performance a tiny bit
+	  by reducing TLB pressure. If in doubt, say "Y". If you've disabled
+	  option but your platform is capable of handling support for this
+	  you can use the gbpages kernel parameter. Likewise if you've enabled
+	  this but you'd like to force disable this option you can use the
+	  nogbpages kernel parameter.
 
 # Common NUMA Features
 config NUMA
diff --git a/arch/x86/mm/init.c b/arch/x86/mm/init.c
index b880d06..8d375ba 100644
--- a/arch/x86/mm/init.c
+++ b/arch/x86/mm/init.c
@@ -131,16 +131,21 @@ void  __init early_alloc_pgt_buf(void)
 
 int after_bootmem;
 
+static int page_size_mask;
+
 int direct_gbpages = IS_ENABLED(CONFIG_DIRECT_GBPAGES);
 
 static void __init init_gbpages(void)
 {
-#ifdef CONFIG_X86_64
-	if (direct_gbpages && cpu_has_gbpages)
+	if (!IS_ENABLED(CONFIG_ENABLE_DIRECT_GBPAGES)) {
+		direct_gbpages = 0;
+		return;
+	}
+	if (direct_gbpages && cpu_has_gbpages) {
 		printk(KERN_INFO "Using GB pages for direct mapping\n");
-	else
+		page_size_mask |= 1 << PG_LEVEL_1G;
+	} else
 		direct_gbpages = 0;
-#endif
 }
 
 struct map_range {
@@ -149,8 +154,6 @@ struct map_range {
 	unsigned page_size_mask;
 };
 
-static int page_size_mask;
-
 static void __init probe_page_size_mask(void)
 {
 	init_gbpages();
@@ -161,8 +164,6 @@ static void __init probe_page_size_mask(void)
 	 * This will simplify cpa(), which otherwise needs to support splitting
 	 * large pages into small in interrupt context, etc.
 	 */
-	if (direct_gbpages)
-		page_size_mask |= 1 << PG_LEVEL_1G;
 	if (cpu_has_pse)
 		page_size_mask |= 1 << PG_LEVEL_2M;
 #endif
diff --git a/arch/x86/mm/pageattr.c b/arch/x86/mm/pageattr.c
index 536ea2f..070b7c2 100644
--- a/arch/x86/mm/pageattr.c
+++ b/arch/x86/mm/pageattr.c
@@ -81,11 +81,9 @@ void arch_report_meminfo(struct seq_file *m)
 	seq_printf(m, "DirectMap4M:    %8lu kB\n",
 			direct_pages_count[PG_LEVEL_2M] << 12);
 #endif
-#ifdef CONFIG_X86_64
 	if (direct_gbpages)
 		seq_printf(m, "DirectMap1G:    %8lu kB\n",
 			direct_pages_count[PG_LEVEL_1G] << 20);
-#endif
 }
 #else
 static inline void split_page_count(int level) { }
-- 
2.2.2