Re: linux-next: Tree for Aug 7 (mm/khugepaged.c)

Re: linux-next: Tree for Aug 7 (mm/khugepaged.c)

[Randy Dunlap]

On 8/7/19 1:36 AM, Stephen Rothwell wrote:
> Hi all,
> 
> Changes since 20190806:
> 

on i386:

when CONFIG_SHMEM is not set/enabled:

../mm/khugepaged.c: In function ‘khugepaged_scan_mm_slot’:
../mm/khugepaged.c:1874:2: error: implicit declaration of function ‘khugepaged_collapse_pte_mapped_thps’; did you mean ‘collapse_pte_mapped_thp’? [-Werror=implicit-function-declaration]
  khugepaged_collapse_pte_mapped_thps(mm_slot);
  ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~


-- 
~Randy

Re: linux-next: Tree for Aug 7 (mm/khugepaged.c)

[Song Liu]

Hi Randy,

> On Aug 7, 2019, at 8:11 AM, Randy Dunlap <rdunlap@infradead.org> wrote:
> 
> On 8/7/19 1:36 AM, Stephen Rothwell wrote:
>> Hi all,
>> 
>> Changes since 20190806:
>> 
> 
> on i386:
> 
> when CONFIG_SHMEM is not set/enabled:
> 
> ../mm/khugepaged.c: In function ‘khugepaged_scan_mm_slot’:
> ../mm/khugepaged.c:1874:2: error: implicit declaration of function ‘khugepaged_collapse_pte_mapped_thps’; did you mean ‘collapse_pte_mapped_thp’? [-Werror=implicit-function-declaration]
>  khugepaged_collapse_pte_mapped_thps(mm_slot);
>  ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Thanks for the report. 

Shall I resend the patch, or shall I send fix on top of current patch?

Best,
Song

Re: linux-next: Tree for Aug 7 (mm/khugepaged.c)

[Andrew Morton]

On Wed, 7 Aug 2019 16:59:14 +0000 Song Liu <songliubraving@fb.com> wrote:

> Hi Randy,
> 
> > On Aug 7, 2019, at 8:11 AM, Randy Dunlap <rdunlap@infradead.org> wrote:
> > 
> > On 8/7/19 1:36 AM, Stephen Rothwell wrote:
> >> Hi all,
> >> 
> >> Changes since 20190806:
> >> 
> > 
> > on i386:
> > 
> > when CONFIG_SHMEM is not set/enabled:
> > 
> > ../mm/khugepaged.c: In function ‘khugepaged_scan_mm_slot’:
> > ../mm/khugepaged.c:1874:2: error: implicit declaration of function ‘khugepaged_collapse_pte_mapped_thps’; did you mean ‘collapse_pte_mapped_thp’? [-Werror=implicit-function-declaration]
> >  khugepaged_collapse_pte_mapped_thps(mm_slot);
> >  ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> 
> Thanks for the report. 
> 
> Shall I resend the patch, or shall I send fix on top of current patch?

Either is OK.  If the difference is small I will turn it into an
incremental patch so that I (and others) can see what changed.

Re: linux-next: Tree for Aug 7 (mm/khugepaged.c)

[Song Liu]

Hi Andrew,

> On Aug 7, 2019, at 1:10 PM, Andrew Morton <akpm@linux-foundation.org> wrote:
> 
> On Wed, 7 Aug 2019 16:59:14 +0000 Song Liu <songliubraving@fb.com> wrote:
> 
>> Hi Randy,
>> 
>>> On Aug 7, 2019, at 8:11 AM, Randy Dunlap <rdunlap@infradead.org> wrote:
>>> 
>>> On 8/7/19 1:36 AM, Stephen Rothwell wrote:
>>>> Hi all,
>>>> 
>>>> Changes since 20190806:
>>>> 
>>> 
>>> on i386:
>>> 
>>> when CONFIG_SHMEM is not set/enabled:
>>> 
>>> ../mm/khugepaged.c: In function ‘khugepaged_scan_mm_slot’:
>>> ../mm/khugepaged.c:1874:2: error: implicit declaration of function ‘khugepaged_collapse_pte_mapped_thps’; did you mean ‘collapse_pte_mapped_thp’? [-Werror=implicit-function-declaration]
>>> khugepaged_collapse_pte_mapped_thps(mm_slot);
>>> ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>> 
>> Thanks for the report. 
>> 
>> Shall I resend the patch, or shall I send fix on top of current patch?
> 
> Either is OK.  If the difference is small I will turn it into an
> incremental patch so that I (and others) can see what changed.

Please find the patch to fix this at the end of this email. It applies 
right on top of "khugepaged: enable collapse pmd for pte-mapped THP". 
It may conflict a little with the "Enable THP for text section of 
non-shmem files" set, which renames function khugepaged_scan_shmem(). 

Also, I found v3 of the set in linux-next. The latest is v4:

https://lkml.org/lkml/2019/8/2/1587
https://lkml.org/lkml/2019/8/2/1588
https://lkml.org/lkml/2019/8/2/1589

Thanks,
Song

======== 8< ===========

From 26715c923f6cd283a2950cfd0a7cca483a3eb406 Mon Sep 17 00:00:00 2001
From: Song Liu <songliubraving@fb.com>
Date: Wed, 7 Aug 2019 10:21:30 -0700
Subject: [PATCH] khugepaged: fix build without CONFIG_SHMEM

Without CONFIG_SHMEM, we need dummy khugepaged_collapse_pte_mapped_thps.

Signed-off-by: Song Liu <songliubraving@fb.com>
---
 mm/khugepaged.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/mm/khugepaged.c b/mm/khugepaged.c
index ba36ff5c1d82..086323112c07 100644
--- a/mm/khugepaged.c
+++ b/mm/khugepaged.c
@@ -1764,6 +1764,11 @@ static void khugepaged_scan_shmem(struct mm_struct *mm,
 {
        BUILD_BUG();
 }
+
+static int khugepaged_collapse_pte_mapped_thps(struct mm_slot *mm_slot)
+{
+       return 0;
+}
 #endif

 static unsigned int khugepaged_scan_mm_slot(unsigned int pages,
--
2.17.1

Re: linux-next: Tree for Aug 7 (mm/khugepaged.c)

[Andrew Morton]

On Wed, 7 Aug 2019 21:00:04 +0000 Song Liu <songliubraving@fb.com> wrote:

> >> 
> >> Shall I resend the patch, or shall I send fix on top of current patch?
> > 
> > Either is OK.  If the difference is small I will turn it into an
> > incremental patch so that I (and others) can see what changed.
> 
> Please find the patch to fix this at the end of this email. It applies 
> right on top of "khugepaged: enable collapse pmd for pte-mapped THP". 
> It may conflict a little with the "Enable THP for text section of 
> non-shmem files" set, which renames function khugepaged_scan_shmem(). 
> 
> Also, I found v3 of the set in linux-next. The latest is v4:
> 
> https://lkml.org/lkml/2019/8/2/1587
> https://lkml.org/lkml/2019/8/2/1588
> https://lkml.org/lkml/2019/8/2/1589

It's all a bit confusing.  I'll drop 

mm-move-memcmp_pages-and-pages_identical.patch
uprobe-use-original-page-when-all-uprobes-are-removed.patch
uprobe-use-original-page-when-all-uprobes-are-removed-v2.patch
mm-thp-introduce-foll_split_pmd.patch
mm-thp-introduce-foll_split_pmd-v11.patch
uprobe-use-foll_split_pmd-instead-of-foll_split.patch
khugepaged-enable-collapse-pmd-for-pte-mapped-thp.patch
uprobe-collapse-thp-pmd-after-removing-all-uprobes.patch

Please resolve Oleg's review comments and resend everything.

Re: linux-next: Tree for Aug 7 (mm/khugepaged.c)

[Randy Dunlap]

On 8/7/19 2:27 PM, Andrew Morton wrote:
> On Wed, 7 Aug 2019 21:00:04 +0000 Song Liu <songliubraving@fb.com> wrote:
> 
>>>>
>>>> Shall I resend the patch, or shall I send fix on top of current patch?
>>>
>>> Either is OK.  If the difference is small I will turn it into an
>>> incremental patch so that I (and others) can see what changed.
>>
>> Please find the patch to fix this at the end of this email. It applies 
>> right on top of "khugepaged: enable collapse pmd for pte-mapped THP". 
>> It may conflict a little with the "Enable THP for text section of 
>> non-shmem files" set, which renames function khugepaged_scan_shmem(). 
>>
>> Also, I found v3 of the set in linux-next. The latest is v4:
>>
>> https://lkml.org/lkml/2019/8/2/1587
>> https://lkml.org/lkml/2019/8/2/1588
>> https://lkml.org/lkml/2019/8/2/1589
> 
> It's all a bit confusing.  I'll drop 
> 
> mm-move-memcmp_pages-and-pages_identical.patch
> uprobe-use-original-page-when-all-uprobes-are-removed.patch
> uprobe-use-original-page-when-all-uprobes-are-removed-v2.patch
> mm-thp-introduce-foll_split_pmd.patch
> mm-thp-introduce-foll_split_pmd-v11.patch
> uprobe-use-foll_split_pmd-instead-of-foll_split.patch
> khugepaged-enable-collapse-pmd-for-pte-mapped-thp.patch
> uprobe-collapse-thp-pmd-after-removing-all-uprobes.patch
> 
> Please resolve Oleg's review comments and resend everything.
> 

OK, that will take care of the build error that I am still seeing
when SHMEM is not enabled:

../mm/khugepaged.c:1849:2: note: in expansion of macro ‘BUILD_BUG’
  BUILD_BUG();
  ^~~~~~~~~


-- 
~Randy

Re: linux-next: Tree for Aug 7 (mm/khugepaged.c)

[Song Liu]

> On Aug 7, 2019, at 2:30 PM, Randy Dunlap <rdunlap@infradead.org> wrote:
> 
> On 8/7/19 2:27 PM, Andrew Morton wrote:
>> On Wed, 7 Aug 2019 21:00:04 +0000 Song Liu <songliubraving@fb.com> wrote:
>> 
>>>>> 
>>>>> Shall I resend the patch, or shall I send fix on top of current patch?
>>>> 
>>>> Either is OK.  If the difference is small I will turn it into an
>>>> incremental patch so that I (and others) can see what changed.
>>> 
>>> Please find the patch to fix this at the end of this email. It applies 
>>> right on top of "khugepaged: enable collapse pmd for pte-mapped THP". 
>>> It may conflict a little with the "Enable THP for text section of 
>>> non-shmem files" set, which renames function khugepaged_scan_shmem(). 
>>> 
>>> Also, I found v3 of the set in linux-next. The latest is v4:
>>> 
>>> https://lkml.org/lkml/2019/8/2/1587
>>> https://lkml.org/lkml/2019/8/2/1588
>>> https://lkml.org/lkml/2019/8/2/1589
>> 
>> It's all a bit confusing.  I'll drop 
>> 
>> mm-move-memcmp_pages-and-pages_identical.patch
>> uprobe-use-original-page-when-all-uprobes-are-removed.patch
>> uprobe-use-original-page-when-all-uprobes-are-removed-v2.patch
>> mm-thp-introduce-foll_split_pmd.patch
>> mm-thp-introduce-foll_split_pmd-v11.patch
>> uprobe-use-foll_split_pmd-instead-of-foll_split.patch
>> khugepaged-enable-collapse-pmd-for-pte-mapped-thp.patch
>> uprobe-collapse-thp-pmd-after-removing-all-uprobes.patch
>> 
>> Please resolve Oleg's review comments and resend everything.
>> 
> 
> OK, that will take care of the build error that I am still seeing
> when SHMEM is not enabled:
> 
> ../mm/khugepaged.c:1849:2: note: in expansion of macro ‘BUILD_BUG’
>  BUILD_BUG();
>  ^~~~~~~~~

This was broken by one of my other patch. Sorry!

The following patch (on top of linux-next/master) fixes it. 

Thanks,
Song

================= 8< ==============================

From 4d6e3a3a28bf85b1debbb90c55f66c81e9ebb9ec Mon Sep 17 00:00:00 2001
From: Song Liu <songliubraving@fb.com>
Date: Wed, 7 Aug 2019 14:57:38 -0700
Subject: [PATCH] khugepaged: fix build without CONFIG_SHMEM

khugepaged_scan_file() should be fully bypassed without CONFIG_SHMEM.

Fixes: f57286140d96 ("mm,thp: add read-only THP support for (non-shmem) FS")
Signed-off-by: Song Liu <songliubraving@fb.com>
---
 mm/khugepaged.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mm/khugepaged.c b/mm/khugepaged.c
index 272fed3ed0f0..40c25ddf29e4 100644
--- a/mm/khugepaged.c
+++ b/mm/khugepaged.c
@@ -1778,7 +1778,7 @@ static unsigned int khugepaged_scan_mm_slot(unsigned int pages,
                        VM_BUG_ON(khugepaged_scan.address < hstart ||
                                  khugepaged_scan.address + HPAGE_PMD_SIZE >
                                  hend);
-                       if (vma->vm_file) {
+                       if (IS_ENABLED(CONFIG_SHMEM) && vma->vm_file) {
                                struct file *file;
                                pgoff_t pgoff = linear_page_index(vma,
                                                khugepaged_scan.address);
--
2.17.1

Re: linux-next: Tree for Aug 7 (mm/khugepaged.c)

[Song Liu]

> On Aug 7, 2019, at 2:27 PM, Andrew Morton <akpm@linux-foundation.org> wrote:
> 
> On Wed, 7 Aug 2019 21:00:04 +0000 Song Liu <songliubraving@fb.com> wrote:
> 
>>>> 
>>>> Shall I resend the patch, or shall I send fix on top of current patch?
>>> 
>>> Either is OK.  If the difference is small I will turn it into an
>>> incremental patch so that I (and others) can see what changed.
>> 
>> Please find the patch to fix this at the end of this email. It applies 
>> right on top of "khugepaged: enable collapse pmd for pte-mapped THP". 
>> It may conflict a little with the "Enable THP for text section of 
>> non-shmem files" set, which renames function khugepaged_scan_shmem(). 
>> 
>> Also, I found v3 of the set in linux-next. The latest is v4:
>> 
>> https://lkml.org/lkml/2019/8/2/1587
>> https://lkml.org/lkml/2019/8/2/1588
>> https://lkml.org/lkml/2019/8/2/1589
> 
> It's all a bit confusing.  I'll drop 

Agreed this is confusing. 

> 
> mm-move-memcmp_pages-and-pages_identical.patch
> uprobe-use-original-page-when-all-uprobes-are-removed.patch
> uprobe-use-original-page-when-all-uprobes-are-removed-v2.patch
> mm-thp-introduce-foll_split_pmd.patch
> mm-thp-introduce-foll_split_pmd-v11.patch
> uprobe-use-foll_split_pmd-instead-of-foll_split.patch
> khugepaged-enable-collapse-pmd-for-pte-mapped-thp.patch
> uprobe-collapse-thp-pmd-after-removing-all-uprobes.patch
> 
> Please resolve Oleg's review comments and resend everything.

Will do. 

Thanks,
Song

Re: linux-next: Tree for Aug 7 (mm/khugepaged.c)

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 671 bytes --]

Hi Andrew,

On Wed, 7 Aug 2019 14:27:55 -0700 Andrew Morton <akpm@linux-foundation.org> wrote:
>
> It's all a bit confusing.  I'll drop 
> 
> mm-move-memcmp_pages-and-pages_identical.patch
> uprobe-use-original-page-when-all-uprobes-are-removed.patch
> uprobe-use-original-page-when-all-uprobes-are-removed-v2.patch
> mm-thp-introduce-foll_split_pmd.patch
> mm-thp-introduce-foll_split_pmd-v11.patch
> uprobe-use-foll_split_pmd-instead-of-foll_split.patch
> khugepaged-enable-collapse-pmd-for-pte-mapped-thp.patch
> uprobe-collapse-thp-pmd-after-removing-all-uprobes.patch

I have dropped them all from linux-next today.

-- 
Cheers,
Stephen Rothwell

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: linux-next: Tree for Aug 7 (mm/khugepaged.c)

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 1434 bytes --]

Hi Song,

On Wed, 7 Aug 2019 22:11:28 +0000 Song Liu <songliubraving@fb.com> wrote:
>
> From: Song Liu <songliubraving@fb.com>
> Date: Wed, 7 Aug 2019 14:57:38 -0700
> Subject: [PATCH] khugepaged: fix build without CONFIG_SHMEM
> 
> khugepaged_scan_file() should be fully bypassed without CONFIG_SHMEM.
> 
> Fixes: f57286140d96 ("mm,thp: add read-only THP support for (non-shmem) FS")
> Signed-off-by: Song Liu <songliubraving@fb.com>
> ---
>  mm/khugepaged.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/mm/khugepaged.c b/mm/khugepaged.c
> index 272fed3ed0f0..40c25ddf29e4 100644
> --- a/mm/khugepaged.c
> +++ b/mm/khugepaged.c
> @@ -1778,7 +1778,7 @@ static unsigned int khugepaged_scan_mm_slot(unsigned int pages,
>                         VM_BUG_ON(khugepaged_scan.address < hstart ||
>                                   khugepaged_scan.address + HPAGE_PMD_SIZE >
>                                   hend);
> -                       if (vma->vm_file) {
> +                       if (IS_ENABLED(CONFIG_SHMEM) && vma->vm_file) {
>                                 struct file *file;
>                                 pgoff_t pgoff = linear_page_index(vma,
>                                                 khugepaged_scan.address);

I have applied this to linux-next today (it needed a little adjusting
after removing the other patches).

-- 
Cheers,
Stephen Rothwell

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]