答复: [PATCH] mm/memblock.c: fix index adjustment error in __next_mem_range_rev()

[<zijun_hu@htc.com>]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1.1: Type: text/plain; charset="gb2312", Size: 4366 bytes --]

Hi Tejun,



I am sorry, I don't take any test for the patch attached in previous mail, and it can't fix the bug completely, please ignore it

I provide a new patch attached in this mail which pass test and can fix the issue described below



__next_mem_range_rev() defined in mm/memblock.c doesn't achieve desired purpose if parameter type_b ==NULL

This new patch can fix the issue and get the last reversed region contained in type_a rightly



The new patch is descripted as follows:

From 0e242eda7696f176a9a2e585a1db01f0575b39c9 Mon Sep 17 00:00:00 2001

From: zijun_hu <zijun_hu@htc.com<mailto:zijun_hu@htc.com>>

Date: Mon, 25 Jul 2016 15:06:57 +0800

Subject: [PATCH] mm/memblock.c: fix index adjustment error in

__next_mem_range_rev()



fix region index adjustment error when parameter type_b of

__next_mem_range_rev() == NULL



Signed-off-by: zijun_hu <zijun_hu@htc.com<mailto:zijun_hu@htc.com>>

---

mm/memblock.c | 8 ++++++--

1 file changed, 6 insertions(+), 2 deletions(-)



diff --git a/mm/memblock.c b/mm/memblock.c index ac12489..cc5aeab 100644

--- a/mm/memblock.c

+++ b/mm/memblock.c

@@ -991,7 +991,11 @@ void __init_memblock __next_mem_range_rev(u64 *idx, int nid, ulong flags,

      if (*idx == (u64)ULLONG_MAX) {

            idx_a = type_a->cnt - 1;

-             idx_b = type_b->cnt;

+            /* in order to get the last reversed region rightly */

+            if (type_b != NULL)

+                   idx_b = type_b->cnt;

+            else

+                   idx_b = 0;

     }

      for (; idx_a >= 0; idx_a--) {

@@ -1024,7 +1028,7 @@ void __init_memblock __next_mem_range_rev(u64 *idx, int nid, ulong flags,

                          *out_end = m_end;

                   if (out_nid)

                          *out_nid = m_nid;

-                    idx_a++;

+                   idx_a--;

                   *idx = (u32)idx_a | (u64)idx_b << 32;

                   return;

            }

--

1.9.1



Zijun Hu

Tel: +86 021 3813 0008 Ext: 8556



F1 Building, 299 Kang Wei Road, Pudong New Area, Shanghai 201315, China htc.com







-----ÓʼþÔ­¼þ-----

·¢¼þÈË: Tejun Heo [mailto:htejun@gmail.com]<mailto:[mailto:htejun@gmail.com]> ´ú±í Tejun Heo

·¢ËÍʱ¼ä: 2016Äê7ÔÂ26ÈÕ 2:52

ÊÕ¼þÈË: Zijun Hu(ºú×ÔÜŠ)

³­ËÍ: akpm@linux-foundation.org<mailto:akpm@linux-foundation.org>; kuleshovmail@gmail.com<mailto:kuleshovmail@gmail.com>; ard.biesheuvel@linaro.org<mailto:ard.biesheuvel@linaro.org>; tangchen@cn.fujitsu.com<mailto:tangchen@cn.fujitsu.com>; weiyang@linux.vnet.ibm.com<mailto:weiyang@linux.vnet.ibm.com>; dev@g0hl1n.net<mailto:dev@g0hl1n.net>; david@gibson.dropbear.id.au<mailto:david@gibson.dropbear.id.au>; linux-mm@kvack.org<mailto:linux-mm@kvack.org>; linux-kernel@vger.kernel.org<mailto:linux-kernel@vger.kernel.org>

Ö÷Ìâ: Re: [PATCH] mm/memblock.c: fix index adjustment error in __next_mem_range_rev()



On Mon, Jul 25, 2016 at 07:34:12AM +0000, zijun_hu@htc.com<mailto:zijun_hu@htc.com> wrote:

> Hi All,

>          There is a bug in mm/memblock.c

>          Could you review and phase-in this patch?

>          Thanks a lot

>

> From 3abf1822d30f77f126bd7a3c09bb243d9c17a029 Mon Sep 17 00:00:00 2001

> From: zijun_hu <zijun_hu@htc.com<mailto:zijun_hu@htc.com>>

> Date: Mon, 25 Jul 2016 15:06:57 +0800

> Subject: [PATCH] mm/memblock.c: fix index adjustment error in

> __next_mem_range_rev()

>

> fix region index adjustment error when parameter type_b of

> __next_mem_range_rev() == NULL

>

> Signed-off-by: zijun_hu <zijun_hu@htc.com<mailto:zijun_hu@htc.com>>

> ---

> mm/memblock.c | 2 +-

> 1 file changed, 1 insertion(+), 1 deletion(-)

>

> diff --git a/mm/memblock.c b/mm/memblock.c index ac12489..b14973e

> 100644

> --- a/mm/memblock.c

> +++ b/mm/memblock.c

> @@ -1024,7 +1024,7 @@ void __init_memblock __next_mem_range_rev(u64 *idx, int nid, ulong flags,

>                                  *out_end = m_end;

>                         if (out_nid)

>                                  *out_nid = m_nid;

> -                         idx_a++;

> +                        idx_a--;



Looks good to me.  Do you happen to have a test case for this bug?



Thanks.



--

tejun



[-- Attachment #1.2: Type: text/html, Size: 19279 bytes --]

[-- Attachment #2: 0001-mm-memblock.c-fix-index-adjustment-error-in.patch --]
[-- Type: application/octet-stream, Size: 1163 bytes --]

From 0e242eda7696f176a9a2e585a1db01f0575b39c9 Mon Sep 17 00:00:00 2001
From: zijun_hu <zijun_hu@htc.com>
Date: Mon, 25 Jul 2016 15:06:57 +0800
Subject: [PATCH] mm/memblock.c: fix index adjustment error in
 __next_mem_range_rev()

fix region index adjustment error when parameter type_b of
__next_mem_range_rev() == NULL

Signed-off-by: zijun_hu <zijun_hu@htc.com>
---
 mm/memblock.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/mm/memblock.c b/mm/memblock.c
index ac12489..cc5aeab 100644
--- a/mm/memblock.c
+++ b/mm/memblock.c
@@ -991,7 +991,11 @@ void __init_memblock __next_mem_range_rev(u64 *idx, int nid, ulong flags,
 
 	if (*idx == (u64)ULLONG_MAX) {
 		idx_a = type_a->cnt - 1;
-		idx_b = type_b->cnt;
+		/* in order to get the last reversed region rightly */
+		if (type_b != NULL)
+			idx_b = type_b->cnt;
+		else
+			idx_b = 0;
 	}
 
 	for (; idx_a >= 0; idx_a--) {
@@ -1024,7 +1028,7 @@ void __init_memblock __next_mem_range_rev(u64 *idx, int nid, ulong flags,
 				*out_end = m_end;
 			if (out_nid)
 				*out_nid = m_nid;
-			idx_a++;
+			idx_a--;
 			*idx = (u32)idx_a | (u64)idx_b << 32;
 			return;
 		}
-- 
1.9.1