Re: [PATCH] mm, page_alloc: actually ignore mempolicies for high priority allocations

Re: [PATCH] mm, page_alloc: actually ignore mempolicies for high priority allocations

[Andrew Morton]

On Tue, 12 Jun 2018 14:26:24 +0200 Vlastimil Babka <vbabka@suse.cz> wrote:

> The __alloc_pages_slowpath() function has for a long time contained code to
> ignore node restrictions from memory policies for high priority allocations.
> The current code that resets the zonelist iterator however does effectively
> nothing after commit 7810e6781e0f ("mm, page_alloc: do not break __GFP_THISNODE
> by zonelist reset") removed a buggy zonelist reset. Even before that commit,
> mempolicy restrictions were still not ignored, as they are passed in
> ac->nodemask which is untouched by the code.
> 
> We can either remove the code, or make it work as intended. Since
> ac->nodemask can be set from task's mempolicy via alloc_pages_current() and
> thus also alloc_pages(), it may indeed affect kernel allocations, and it makes
> sense to ignore it to allow progress for high priority allocations.
> 
> Thus, this patch resets ac->nodemask to NULL in such cases. This assumes all
> callers can handle it (i.e. there are no guarantees as in the case of
> __GFP_THISNODE) which seems to be the case. The same assumption is already
> present in check_retry_cpuset() for some time.
> 
> The expected effect is that high priority kernel allocations in the context of
> userspace tasks (e.g. OOM victims) restricted by mempolicies will have higher
> chance to succeed if they are restricted to nodes with depleted memory, while
> there are other nodes with free memory left.

We don't have any reviews or acks on ths one, perhaps because linux-mm
wasn't cc'ed.  Could people please take a look?


From: Vlastimil Babka <vbabka@suse.cz>
Subject: mm, page_alloc: actually ignore mempolicies for high priority allocations

The __alloc_pages_slowpath() function has for a long time contained code
to ignore node restrictions from memory policies for high priority
allocations.  The current code that resets the zonelist iterator however
does effectively nothing after commit 7810e6781e0f ("mm, page_alloc: do
not break __GFP_THISNODE by zonelist reset") removed a buggy zonelist
reset.  Even before that commit, mempolicy restrictions were still not
ignored, as they are passed in ac->nodemask which is untouched by the
code.

We can either remove the code, or make it work as intended.  Since
ac->nodemask can be set from task's mempolicy via alloc_pages_current()
and thus also alloc_pages(), it may indeed affect kernel allocations, and
it makes sense to ignore it to allow progress for high priority
allocations.

Thus, this patch resets ac->nodemask to NULL in such cases.  This assumes
all callers can handle it (i.e.  there are no guarantees as in the case of
__GFP_THISNODE) which seems to be the case.  The same assumption is
already present in check_retry_cpuset() for some time.

The expected effect is that high priority kernel allocations in the
context of userspace tasks (e.g.  OOM victims) restricted by mempolicies
will have higher chance to succeed if they are restricted to nodes with
depleted memory, while there are other nodes with free memory left.


Ot's not a new intention, but for the first time the code will match the
intention, AFAICS.  It was intended by commit 183f6371aac2 ("mm: ignore
mempolicies when using ALLOC_NO_WATERMARK") in v3.6 but I think it never
really worked, as mempolicy restriction was already encoded in nodemask,
not zonelist, at that time.

So originally that was for ALLOC_NO_WATERMARK only.  Then it was adjusted
by e46e7b77c909 ("mm, page_alloc: recalculate the preferred zoneref if the
context can ignore memory policies") and cd04ae1e2dc8 ("mm, oom: do not
rely on TIF_MEMDIE for memory reserves access") to the current state.  So
even GFP_ATOMIC would now ignore mempolicies after the initial attempts
fail - if the code worked as people thought it does.

Link: http://lkml.kernel.org/r/20180612122624.8045-1-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
---

 mm/page_alloc.c |    7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

--- a/mm/page_alloc.c~mm-page_alloc-actually-ignore-mempolicies-for-high-priority-allocations
+++ a/mm/page_alloc.c
@@ -4165,11 +4165,12 @@ retry:
 		alloc_flags = reserve_flags;
 
 	/*
-	 * Reset the zonelist iterators if memory policies can be ignored.
-	 * These allocations are high priority and system rather than user
-	 * orientated.
+	 * Reset the nodemask and zonelist iterators if memory policies can be
+	 * ignored. These allocations are high priority and system rather than
+	 * user oriented.
 	 */
 	if (!(alloc_flags & ALLOC_CPUSET) || reserve_flags) {
+		ac->nodemask = NULL;
 		ac->preferred_zoneref = first_zones_zonelist(ac->zonelist,
 					ac->high_zoneidx, ac->nodemask);
 	}
_

Re: [PATCH] mm, page_alloc: actually ignore mempolicies for high priority allocations

[Mel Gorman]

On Wed, Aug 15, 2018 at 03:16:52PM -0700, Andrew Morton wrote:
> From: Vlastimil Babka <vbabka@suse.cz>
> Subject: mm, page_alloc: actually ignore mempolicies for high priority allocations
> 
> The __alloc_pages_slowpath() function has for a long time contained code
> to ignore node restrictions from memory policies for high priority
> allocations.  The current code that resets the zonelist iterator however
> does effectively nothing after commit 7810e6781e0f ("mm, page_alloc: do
> not break __GFP_THISNODE by zonelist reset") removed a buggy zonelist
> reset.  Even before that commit, mempolicy restrictions were still not
> ignored, as they are passed in ac->nodemask which is untouched by the
> code.
> 
> We can either remove the code, or make it work as intended.  Since
> ac->nodemask can be set from task's mempolicy via alloc_pages_current()
> and thus also alloc_pages(), it may indeed affect kernel allocations, and
> it makes sense to ignore it to allow progress for high priority
> allocations.
> 
> Thus, this patch resets ac->nodemask to NULL in such cases.  This assumes
> all callers can handle it (i.e.  there are no guarantees as in the case of
> __GFP_THISNODE) which seems to be the case.  The same assumption is
> already present in check_retry_cpuset() for some time.
> 
> The expected effect is that high priority kernel allocations in the
> context of userspace tasks (e.g.  OOM victims) restricted by mempolicies
> will have higher chance to succeed if they are restricted to nodes with
> depleted memory, while there are other nodes with free memory left.
> 
> 
> Ot's not a new intention, but for the first time the code will match the
> intention, AFAICS.  It was intended by commit 183f6371aac2 ("mm: ignore
> mempolicies when using ALLOC_NO_WATERMARK") in v3.6 but I think it never
> really worked, as mempolicy restriction was already encoded in nodemask,
> not zonelist, at that time.
> 
> So originally that was for ALLOC_NO_WATERMARK only.  Then it was adjusted
> by e46e7b77c909 ("mm, page_alloc: recalculate the preferred zoneref if the
> context can ignore memory policies") and cd04ae1e2dc8 ("mm, oom: do not
> rely on TIF_MEMDIE for memory reserves access") to the current state.  So
> even GFP_ATOMIC would now ignore mempolicies after the initial attempts
> fail - if the code worked as people thought it does.
> 
> Link: http://lkml.kernel.org/r/20180612122624.8045-1-vbabka@suse.cz
> Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
> Cc: Mel Gorman <mgorman@techsingularity.net>
> Cc: Michal Hocko <mhocko@kernel.org>
> Cc: David Rientjes <rientjes@google.com>
> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

FWIW, I thought I acked this already.

Acked-by: Mel Gorman <mgorman@techsingularity.net>

-- 
Mel Gorman
SUSE Labs

Re: [PATCH] mm, page_alloc: actually ignore mempolicies for high priority allocations

[Michal Hocko]

On Wed 15-08-18 15:16:52, Andrew Morton wrote:
[...]
> From: Vlastimil Babka <vbabka@suse.cz>
> Subject: mm, page_alloc: actually ignore mempolicies for high priority allocations
> 
> The __alloc_pages_slowpath() function has for a long time contained code
> to ignore node restrictions from memory policies for high priority
> allocations.  The current code that resets the zonelist iterator however
> does effectively nothing after commit 7810e6781e0f ("mm, page_alloc: do
> not break __GFP_THISNODE by zonelist reset") removed a buggy zonelist
> reset.  Even before that commit, mempolicy restrictions were still not
> ignored, as they are passed in ac->nodemask which is untouched by the
> code.
> 
> We can either remove the code, or make it work as intended.  Since
> ac->nodemask can be set from task's mempolicy via alloc_pages_current()
> and thus also alloc_pages(), it may indeed affect kernel allocations, and
> it makes sense to ignore it to allow progress for high priority
> allocations.
> 
> Thus, this patch resets ac->nodemask to NULL in such cases.  This assumes
> all callers can handle it (i.e.  there are no guarantees as in the case of
> __GFP_THISNODE) which seems to be the case.  The same assumption is
> already present in check_retry_cpuset() for some time.
> 
> The expected effect is that high priority kernel allocations in the
> context of userspace tasks (e.g.  OOM victims) restricted by mempolicies
> will have higher chance to succeed if they are restricted to nodes with
> depleted memory, while there are other nodes with free memory left.
> 
> 
> Ot's not a new intention, but for the first time the code will match the
> intention, AFAICS.  It was intended by commit 183f6371aac2 ("mm: ignore
> mempolicies when using ALLOC_NO_WATERMARK") in v3.6 but I think it never
> really worked, as mempolicy restriction was already encoded in nodemask,
> not zonelist, at that time.
> 
> So originally that was for ALLOC_NO_WATERMARK only.  Then it was adjusted
> by e46e7b77c909 ("mm, page_alloc: recalculate the preferred zoneref if the
> context can ignore memory policies") and cd04ae1e2dc8 ("mm, oom: do not
> rely on TIF_MEMDIE for memory reserves access") to the current state.  So
> even GFP_ATOMIC would now ignore mempolicies after the initial attempts
> fail - if the code worked as people thought it does.
> 
> Link: http://lkml.kernel.org/r/20180612122624.8045-1-vbabka@suse.cz
> Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
> Cc: Mel Gorman <mgorman@techsingularity.net>
> Cc: Michal Hocko <mhocko@kernel.org>
> Cc: David Rientjes <rientjes@google.com>
> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

The code is quite subtle and we have a bad history of copying stuff
without rethinking whether the code still is needed. Which is sad and a
clear sign that the code is too complex. I cannot say this change
doesn't have any subtle side effects but it makes the intention clear at
least so I _think_ it is good to go. If we find some unintended side
effects we should simply rethink the whole reset zonelist thing.

That being said
Acked-by: Michal Hocko <mhocko@suse.com>

> ---
> 
>  mm/page_alloc.c |    7 ++++---
>  1 file changed, 4 insertions(+), 3 deletions(-)
> 
> --- a/mm/page_alloc.c~mm-page_alloc-actually-ignore-mempolicies-for-high-priority-allocations
> +++ a/mm/page_alloc.c
> @@ -4165,11 +4165,12 @@ retry:
>  		alloc_flags = reserve_flags;
>  
>  	/*
> -	 * Reset the zonelist iterators if memory policies can be ignored.
> -	 * These allocations are high priority and system rather than user
> -	 * orientated.
> +	 * Reset the nodemask and zonelist iterators if memory policies can be
> +	 * ignored. These allocations are high priority and system rather than
> +	 * user oriented.
>  	 */
>  	if (!(alloc_flags & ALLOC_CPUSET) || reserve_flags) {
> +		ac->nodemask = NULL;
>  		ac->preferred_zoneref = first_zones_zonelist(ac->zonelist,
>  					ac->high_zoneidx, ac->nodemask);
>  	}
> _
> 

-- 
Michal Hocko
SUSE Labs

Re: [PATCH] mm, page_alloc: actually ignore mempolicies for high priority allocations

[Tetsuo Handa]

On 2018/08/16 19:03, Michal Hocko wrote:
> The code is quite subtle and we have a bad history of copying stuff
> without rethinking whether the code still is needed. Which is sad and a
> clear sign that the code is too complex. I cannot say this change
> doesn't have any subtle side effects but it makes the intention clear at
> least so I _think_ it is good to go. If we find some unintended side
> effects we should simply rethink the whole reset zonelist thing.

Does this change affect

        /*
         * This is not a __GFP_THISNODE allocation, so a truncated nodemask in
         * the page allocator means a mempolicy is in effect.  Cpuset policy
         * is enforced in get_page_from_freelist().
         */
        if (oc->nodemask &&
            !nodes_subset(node_states[N_MEMORY], *oc->nodemask)) {
                oc->totalpages = total_swap_pages;
                for_each_node_mask(nid, *oc->nodemask)
                        oc->totalpages += node_spanned_pages(nid);
                return CONSTRAINT_MEMORY_POLICY;
        }

in constrained_alloc() called from

        /*
         * Check if there were limitations on the allocation (only relevant for
         * NUMA and memcg) that may require different handling.
         */
        constraint = constrained_alloc(oc);
        if (constraint != CONSTRAINT_MEMORY_POLICY)
                oc->nodemask = NULL;

in out_of_memory() ?

Re: [PATCH] mm, page_alloc: actually ignore mempolicies for high priority allocations

[Michal Hocko]

On Sat 18-08-18 22:02:14, Tetsuo Handa wrote:
> On 2018/08/16 19:03, Michal Hocko wrote:
> > The code is quite subtle and we have a bad history of copying stuff
> > without rethinking whether the code still is needed. Which is sad and a
> > clear sign that the code is too complex. I cannot say this change
> > doesn't have any subtle side effects but it makes the intention clear at
> > least so I _think_ it is good to go. If we find some unintended side
> > effects we should simply rethink the whole reset zonelist thing.
> 
> Does this change affect
> 
>         /*
>          * This is not a __GFP_THISNODE allocation, so a truncated nodemask in
>          * the page allocator means a mempolicy is in effect.  Cpuset policy
>          * is enforced in get_page_from_freelist().
>          */
>         if (oc->nodemask &&
>             !nodes_subset(node_states[N_MEMORY], *oc->nodemask)) {
>                 oc->totalpages = total_swap_pages;
>                 for_each_node_mask(nid, *oc->nodemask)
>                         oc->totalpages += node_spanned_pages(nid);
>                 return CONSTRAINT_MEMORY_POLICY;
>         }
> 
> in constrained_alloc() called from
> 
>         /*
>          * Check if there were limitations on the allocation (only relevant for
>          * NUMA and memcg) that may require different handling.
>          */
>         constraint = constrained_alloc(oc);
>         if (constraint != CONSTRAINT_MEMORY_POLICY)
>                 oc->nodemask = NULL;
> 
> in out_of_memory() ?

No practical difference AFAICS. We are losing the nodemask for oom
victims but their mere existance should make oom decisions void
and so the constrain shouldn't really matter.
-- 
Michal Hocko
SUSE Labs

Re: [PATCH] mm, page_alloc: actually ignore mempolicies for high priority allocations

[Tetsuo Handa]

On 2018/08/20 19:41, Michal Hocko wrote:
> On Sat 18-08-18 22:02:14, Tetsuo Handa wrote:
>> On 2018/08/16 19:03, Michal Hocko wrote:
>>> The code is quite subtle and we have a bad history of copying stuff
>>> without rethinking whether the code still is needed. Which is sad and a
>>> clear sign that the code is too complex. I cannot say this change
>>> doesn't have any subtle side effects but it makes the intention clear at
>>> least so I _think_ it is good to go. If we find some unintended side
>>> effects we should simply rethink the whole reset zonelist thing.
>>
>> Does this change affect
>>
>>         /*
>>          * This is not a __GFP_THISNODE allocation, so a truncated nodemask in
>>          * the page allocator means a mempolicy is in effect.  Cpuset policy
>>          * is enforced in get_page_from_freelist().
>>          */
>>         if (oc->nodemask &&
>>             !nodes_subset(node_states[N_MEMORY], *oc->nodemask)) {
>>                 oc->totalpages = total_swap_pages;
>>                 for_each_node_mask(nid, *oc->nodemask)
>>                         oc->totalpages += node_spanned_pages(nid);
>>                 return CONSTRAINT_MEMORY_POLICY;
>>         }
>>
>> in constrained_alloc() called from
>>
>>         /*
>>          * Check if there were limitations on the allocation (only relevant for
>>          * NUMA and memcg) that may require different handling.
>>          */
>>         constraint = constrained_alloc(oc);
>>         if (constraint != CONSTRAINT_MEMORY_POLICY)
>>                 oc->nodemask = NULL;
>>
>> in out_of_memory() ?
> 
> No practical difference AFAICS. We are losing the nodemask for oom
> victims but their mere existance should make oom decisions void
> and so the constrain shouldn't really matter.
> 

If my guess that whether oc->nodemask is NULL affects whether oom_unkillable_task()
(effectively has_intersects_mems_allowed()) returns true is correct, I worried that
OOM victims select next OOM victim from wider targets if MMF_OOM_SKIP was already set
on the OOM victim's mm. That might be an unexpected behavior...

Re: [PATCH] mm, page_alloc: actually ignore mempolicies for high priority allocations

[Michal Hocko]

On Mon 20-08-18 19:52:59, Tetsuo Handa wrote:
> On 2018/08/20 19:41, Michal Hocko wrote:
> > On Sat 18-08-18 22:02:14, Tetsuo Handa wrote:
> >> On 2018/08/16 19:03, Michal Hocko wrote:
> >>> The code is quite subtle and we have a bad history of copying stuff
> >>> without rethinking whether the code still is needed. Which is sad and a
> >>> clear sign that the code is too complex. I cannot say this change
> >>> doesn't have any subtle side effects but it makes the intention clear at
> >>> least so I _think_ it is good to go. If we find some unintended side
> >>> effects we should simply rethink the whole reset zonelist thing.
> >>
> >> Does this change affect
> >>
> >>         /*
> >>          * This is not a __GFP_THISNODE allocation, so a truncated nodemask in
> >>          * the page allocator means a mempolicy is in effect.  Cpuset policy
> >>          * is enforced in get_page_from_freelist().
> >>          */
> >>         if (oc->nodemask &&
> >>             !nodes_subset(node_states[N_MEMORY], *oc->nodemask)) {
> >>                 oc->totalpages = total_swap_pages;
> >>                 for_each_node_mask(nid, *oc->nodemask)
> >>                         oc->totalpages += node_spanned_pages(nid);
> >>                 return CONSTRAINT_MEMORY_POLICY;
> >>         }
> >>
> >> in constrained_alloc() called from
> >>
> >>         /*
> >>          * Check if there were limitations on the allocation (only relevant for
> >>          * NUMA and memcg) that may require different handling.
> >>          */
> >>         constraint = constrained_alloc(oc);
> >>         if (constraint != CONSTRAINT_MEMORY_POLICY)
> >>                 oc->nodemask = NULL;
> >>
> >> in out_of_memory() ?
> > 
> > No practical difference AFAICS. We are losing the nodemask for oom
> > victims but their mere existance should make oom decisions void
> > and so the constrain shouldn't really matter.
> > 
> 
> If my guess that whether oc->nodemask is NULL affects whether oom_unkillable_task()
> (effectively has_intersects_mems_allowed()) returns true is correct, I worried that
> OOM victims select next OOM victim from wider targets if MMF_OOM_SKIP was already set
> on the OOM victim's mm. That might be an unexpected behavior...

What I've tried to say is that the oom victim should be present also
in wider oom domain. There are potential races (e.g. the victim having
MMF_OOM_SKIP set) but I am not really sure we should be losing sleep
over those. As I've said in an earlier email. If we have an unexpected
behavior we should deal with it.

-- 
Michal Hocko
SUSE Labs