[PATCH] x86: Enable fast 32-bit put_user_64 for copy_to_user

[PATCH] x86: Enable fast 32-bit put_user_64 for copy_to_user

[Chris Wilson]

For fixed sized copies, copy_to_user() will utilize __put_user_size
fastpaths. However, it is missing the translation for 64bit copies on
x86/32. Testing on a Pinetrail Atom, the 64 bit put_user fastpath is
substantially faster than the generic copy_to_user() fallback.

Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: x86@kernel.org
CC: linux-kernel@vger.kernel.org
---
 arch/x86/include/asm/uaccess_32.h | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/arch/x86/include/asm/uaccess_32.h b/arch/x86/include/asm/uaccess_32.h
index 3c03a5de64d3..0ed5504c6060 100644
--- a/arch/x86/include/asm/uaccess_32.h
+++ b/arch/x86/include/asm/uaccess_32.h
@@ -59,6 +59,10 @@ __copy_to_user_inatomic(void __user *to, const void *from, unsigned long n)
 			__put_user_size(*(u32 *)from, (u32 __user *)to,
 					4, ret, 4);
 			return ret;
+		case 8:
+			__put_user_size(*(u64 *)from, (u64 __user *)to,
+					8, ret, 8);
+			return ret;
 		}
 	}
 	return __copy_to_user_ll(to, from, n);
-- 
2.1.4

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [PATCH] x86: Enable fast 32-bit put_user_64 for copy_to_user

[shuang.he]

Tested-By: Intel Graphics QA PRTS (Patch Regression Test System Contact: shuang.he@intel.com)
Task id: 6197
-------------------------------------Summary-------------------------------------
Platform          Delta          drm-intel-nightly          Series Applied
PNV                                  276/276              276/276
ILK                 -1              302/302              301/302
SNB                                  318/318              318/318
IVB                                  341/341              341/341
BYT                                  287/287              287/287
HSW                                  395/395              395/395
BDW                                  318/318              318/318
-------------------------------------Detailed-------------------------------------
Platform  Test                                drm-intel-nightly          Series Applied
*ILK  igt@gem_fenced_exec_thrash@no-spare-fences-busy-interruptible      PASS(3)      DMESG_WARN(1)PASS(1)
(dmesg patch applied)drm:i915_hangcheck_elapsed[i915]]*ERROR*Hangcheck_timer_elapsed...bsd_ring_idle@Hangcheck timer elapsed... bsd ring idle
Note: You need to pay more attention to line start with '*'
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [PATCH] x86: Enable fast 32-bit put_user_64 for copy_to_user

[Ingo Molnar]

* Chris Wilson <chris@chris-wilson.co.uk> wrote:

> For fixed sized copies, copy_to_user() will utilize __put_user_size
> fastpaths. However, it is missing the translation for 64bit copies on
> x86/32. Testing on a Pinetrail Atom, the 64 bit put_user fastpath is
> substantially faster than the generic copy_to_user() fallback.
> 
> Cc: Thomas Gleixner <tglx@linutronix.de>
> Cc: Ingo Molnar <mingo@redhat.com>
> Cc: "H. Peter Anvin" <hpa@zytor.com>
> Cc: x86@kernel.org
> CC: linux-kernel@vger.kernel.org

The patch makes sense, but your Signed-off-by line is missing.

Thanks,

	Ingo
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [PATCH] x86: Enable fast 32-bit put_user_64 for copy_to_user

[Chris Wilson]

On Thu, Apr 16, 2015 at 09:28:02AM +0200, Ingo Molnar wrote:
> 
> * Chris Wilson <chris@chris-wilson.co.uk> wrote:
> 
> > For fixed sized copies, copy_to_user() will utilize __put_user_size
> > fastpaths. However, it is missing the translation for 64bit copies on
> > x86/32. Testing on a Pinetrail Atom, the 64 bit put_user fastpath is
> > substantially faster than the generic copy_to_user() fallback.
> > 
> > Cc: Thomas Gleixner <tglx@linutronix.de>
> > Cc: Ingo Molnar <mingo@redhat.com>
> > Cc: "H. Peter Anvin" <hpa@zytor.com>
> > Cc: x86@kernel.org
> > CC: linux-kernel@vger.kernel.org
> 
> The patch makes sense, but your Signed-off-by line is missing.

Sorry, totally forgot that when rewriting the changelog.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre