Re: [PATCH v4 1/2] mm: migration: fix the FOLL_GET failure on following huge page

[Alistair Popple]

On Monday, 15 August 2022 3:20:28 PM AEST Wang, Haiyue wrote:
> > -----Original Message-----
> > From: Alistair Popple <apopple@xxxxxxxxxx>
> > Sent: Monday, August 15, 2022 13:17
> > To: linux-mm@xxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx; Wang, Haiyue 
<haiyue.wang@xxxxxxxxx>
> > Cc: akpm@xxxxxxxxxxxxxxxxxxxx; david@xxxxxxxxxx; linmiaohe@xxxxxxxxxx; 
Huang, Ying
> > <ying.huang@xxxxxxxxx>; songmuchun@xxxxxxxxxxxxx; 
naoya.horiguchi@xxxxxxxxx; alex.sierra@xxxxxxx
> > Subject: Re: [PATCH v4 1/2] mm: migration: fix the FOLL_GET failure on 
following huge page
> > 
> > On Monday, 15 August 2022 2:40:48 PM AEST Wang, Haiyue wrote:
> > > > -----Original Message-----
> > > > From: Alistair Popple <apopple@xxxxxxxxxx>
> > > > Sent: Monday, August 15, 2022 12:29
> > > > To: linux-mm@xxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx; Wang, Haiyue
> > <haiyue.wang@xxxxxxxxx>
> > > > Cc: akpm@xxxxxxxxxxxxxxxxxxxx; david@xxxxxxxxxx; linmiaohe@xxxxxxxxxx;
> > Huang, Ying
> > > > <ying.huang@xxxxxxxxx>; songmuchun@xxxxxxxxxxxxx;
> > naoya.horiguchi@xxxxxxxxx; alex.sierra@xxxxxxx; Wang,
> > > > Haiyue <haiyue.wang@xxxxxxxxx>
> > > > Subject: Re: [PATCH v4 1/2] mm: migration: fix the FOLL_GET failure on
> > following huge page
> > > >
> > > > On Monday, 15 August 2022 11:59:08 AM AEST Haiyue Wang wrote:
> > > > > Not all huge page APIs support FOLL_GET option, so the 
__NR_move_pages
> > > > > will fail to get the page node information for huge page.
> > > >
> > > > I think you should be explicit in the commit message about which 
functions
> > do
> > > > not support FOLL_GET as it's not obvious what support needs to be 
added
> > before
> > > > this fix can be reverted.
> > >
> > > Yes, make sense, will add them in new patch.
> > 
> > Actually while you're at it I think it would be good to include a 
description
> > of the impact of this failure in the commit message. Ie. You're answer to:
> > 
> > > What are the user-visible runtime effects of this bug?
> > 
> > As it documents what should be tested if this fix does actually ever get
> > reverted.
> 
> An short example *.c code to capture the bug in commit message ?

That's probably overkill. Just being a bit more explicit about the 
circumstances in which sys_move_pages() actually fails would be good. Eg. 
something like this:

"Without this sys_move_pages() will return -ENOENT for 1GB huge page
memory map when dumping the page node information for nodes != NULL"

> > 
> > > >
> > > > Thanks.
> > > >
> > > >  - Alistair
> > > >
> > > > > This is an temporary solution to mitigate the racing fix.
> > > > >
> > > > > After supporting follow huge page by FOLL_GET is done, this fix can 
be
> > > > > reverted safely.
> > > > >
> > > > > Fixes: 4cd614841c06 ("mm: migration: fix possible 
do_pages_stat_array
> > racing
> > > > with memory offline")
> > > > > Signed-off-by: Haiyue Wang <haiyue.wang@xxxxxxxxx>
> > > > > ---
> > > > >  mm/migrate.c | 10 ++++++++--
> > > > >  1 file changed, 8 insertions(+), 2 deletions(-)
> > > > >
> > > > > diff --git a/mm/migrate.c b/mm/migrate.c
> > > > > index 6a1597c92261..581dfaad9257 100644
> > > > > --- a/mm/migrate.c
> > > > > +++ b/mm/migrate.c
> > > > > @@ -1848,6 +1848,7 @@ static void do_pages_stat_array(struct 
mm_struct
> > *mm,
> > > > unsigned long nr_pages,
> > > > >
> > > > >  	for (i = 0; i < nr_pages; i++) {
> > > > >  		unsigned long addr = (unsigned long)(*pages);
> > > > > +		unsigned int foll_flags = FOLL_DUMP;
> > > > >  		struct vm_area_struct *vma;
> > > > >  		struct page *page;
> > > > >  		int err = -EFAULT;
> > > > > @@ -1856,8 +1857,12 @@ static void do_pages_stat_array(struct 
mm_struct
> > *mm,
> > > > unsigned long nr_pages,
> > > > >  		if (!vma)
> > > > >  			goto set_status;
> > > > >
> > > > > +		/* Not all huge page follow APIs support 'FOLL_GET' 
*/
> > > > > +		if (!is_vm_hugetlb_page(vma))
> > > > > +			foll_flags |= FOLL_GET;
> > > > > +
> > > > >  		/* FOLL_DUMP to ignore special (like zero) pages */
> > > > > -		page = follow_page(vma, addr, FOLL_GET | FOLL_DUMP);
> > > > > +		page = follow_page(vma, addr, foll_flags);
> > > > >
> > > > >  		err = PTR_ERR(page);
> > > > >  		if (IS_ERR(page))
> > > > > @@ -1865,7 +1870,8 @@ static void do_pages_stat_array(struct 
mm_struct
> > *mm,
> > > > unsigned long nr_pages,
> > > > >
> > > > >  		if (page && !is_zone_device_page(page)) {
> > > > >  			err = page_to_nid(page);
> > > > > -			put_page(page);
> > > > > +			if (foll_flags & FOLL_GET)
> > > > > +				put_page(page);
> > > > >  		} else {
> > > > >  			err = -ENOENT;
> > > > >  		}
> > > > >
> > > >
> > > >
> > > >
> > >
> > >
> > 
> > 
> > 
> 
>