lkml.org 
[lkml]   [2015]   [Mar]   [4]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
From
Subject[RFC v1 2/4] x86: mm: simplify enabling direct_gbpages
Date
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


\
 
 \ /
  Last update: 2015-03-05 02:41    [W:0.095 / U:0.020 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site