[LSF/MM/BPF TOPIC] Improving MGLRU

[LSF/MM/BPF TOPIC] Improving MGLRU

[Kairui Song]

Hi All,

Apologies I forgot to add the proper tag in the previous email so
resending this.

MGLRU has been introduced in the mainline for years, but we still have two LRUs
today. There are many reasons MGLRU is still not the only LRU implementation in
the kernel.

And I've been looking at a few major issues here:

1. Page flag usage: MGLRU uses many more flags (3+ more) than Active/Inactive
LRU.
2. Regressions: MGLRU might cause regression, even though in many workloads it
outperforms Active/Inactive by a lot.
3. Metrics: MGLRU makes some metrics work differently, for example: PSI,
/proc/meminfo.
4. Some reclaim behavior is less controllable.

And other issues too.
And I think there isn't a simple solution, but it can definitely be solved. I
would like to propose a session to discuss a few ideas on how to solve this, and
perhaps we can finally only have one LRU in the kernel. So I'd like topropose a
session to discuss some ideas about improving MGLRU and making it the only LRU.

Some parts are just ideas, so far I have a working series [2] following the
LFU and metric unification idea below, solving 2) and 3) above, and
providing some very basic infrastructures for 1). Would try to send that as
RFC for easier review and merge once it's stable enough soon, before LSF/MM/BPF.

So far, I already observed a 30% reduction of refault of total folios in
some workloads, including Tpcc and YCSB, and several critical regressions
compared to Active / Inactive are gone, PG_workingset and PG_referenced are
gone, yet things like PSI are more accurate (see below), and still stay
bitwise compatible with Active / Inactive LRU. If it went smoothly,
we might be able to unify and have only one LRU.

Following topic and ideas are the key points:

1. Flags usage: which is solvable, and the hard part is mostly about
   implementation details: MGLRU uses (at least) 3 extra flags for the gen
   number, and we are expecting it to use more gen flags to support more than 4
   gen. These flags can be moved to the tail of the LRU pointer after carefully
   modifying the kernel's convention on LRU operations. That would allow us to
   use up to 6 bits for the gen number and support up to 63 gens. The lower bit
   of both pointers can be packed together for CAS on gen numbers. Reducing
   flag usage by 3. Previously, Yu also suggested moving flags like PG_active to
   the LRU pointer tail, which could also be a way.

   struct folio {
       /* ... */
       union {
               struct list_head lru;
   +           struct lru_gen_list_head lru_gen;

   So whenever the folio is on lruvec, `lru_gen_list_head` is used instead of
   `lru`, which contains encoded info. We might be able to move all LRU-related
   flags there.

   Ordinary folio lists are still just fine, since when the folio is isolated,
   `lru` is still there. But places like folio split, will need to
check if that's
   a lruvec folio, or folio on an ordinary list.

   This part is just an idea yet. But might make us able to have up to 63 gens
   in upstream and enable build for every config.

2. Regressions: Currently regression is a more major problem for us.
   From our perspective, almost all regressions are caused by an under- or
   overprotected file cache. MGLRU's PID protection either gets too aggressive
   or too passive or just have a too long latency. To fix that, I'd propose a
   LFU-like design and relax the PID's aggressiveness to make it much more
   proactive and effective for file folios. The idea is always use 3 bits in
   the page flags to count the referenced time (which would also replace
   PG_workingset and PG_referenced). Initial tests showed a 30% reduction of
   refaults, and many regressions are gone. A flow chart of how the MGLRU idea
   might work:

   ========== MGLFU Tiering ==========
   Access  3 bit    lru_gen  lru_gen |(R - PG_referenced | W - PG_workingset)
   Count   L|W|R    refs     tier    |(L - LRU_GEN_REFS)
   0       0|0|0    0        0       | - Readahead & Cache
   1       0|0|1    1        0       | - LRU_REFS_REFERENCED
   ----- WORKINGSET / PROMOTE --- <--+ - <move out of min_seq>
   2       0|1|0    2        0       | - LRU_REFS_WORKINGSET
   3       0|1|1    3        1       | - Frequently used
   4       1|0|0*   4        2       |
   5       1|0|1*   5        2       |
   6       1|1|0*   6        3       |
   7       1|1|1*   7        3       | - LRU_REFS_MAX
   ---------- PROMOTION ----------> --+ - <promote to next gen>

   Once a folio has an access count > LRU_REFS_WORKINGSET, it never goes lower
   than that. Folios that hit LRU_REFS_MAX will be promoted to next gen on
   access, and remove the force protection of folios on eviction. This provides
   a more proactive protection.

   And this might also give other frameworks like DAMON a nicer interface to
   interact with MGLRU, since the referenced count can promote every folio and
   count accesses in a more reasonable and unified way for MGLRU now.

   NOTE: Still changing this design according to test results, e.g. maybe
   we should optionally still use 4 bits, so the final solution might not
   be the same.

   Another potential improvement on the regression issue is implementing the
   refault distance as I once proposed [1], which can have a huge gain for some
   workloads with heavy file folio usage. Maybe we can have both.

3. Metrics: The key here is about the meaning of page flags, including
   PG_workingset and PG_referenced. These two flags are set/cleared very
   differently for MGLRU compared to Active / Inactive LRU, but many other
   components are still using them as metrics for Active / Inactive LRU. Hence,
   I would propose to introduce a different mechanism to unify and replace these
   two flags: Using the 3 bits in the page flags field reserved for LFU-like
   tracking above, to determine the folio status.

   Then following the above LFU-like idea, and using helpers like:

   static inline bool folio_is_referenced(const struct folio *folio)
   {
    return folio_lru_refs(folio) >= LRU_REFS_REFERENCED;
   }

   static inline bool folio_is_workingset(const struct folio *folio)
   {
    return folio_lru_refs(folio) >= LRU_REFS_WORKINGSET;
   }

   static inline bool folio_is_referenced_by_bit(struct folio *folio)
   {    /* For compatibility */
    return !!(READ_ONCE(*folio_flags(folio, 0)) & BIT(LRU_REFS_PGOFF));
   }

   static inline void folio_mark_workingset_by_bit(struct folio *folio)
   {    /* For compatibility */
    set_mask_bits(folio_flags(folio, 0), BIT(LRU_REFS_PGOFF + 1),
BIT(LRU_REFS_PGOFF + 1));
   }

   To tell if a folio belongs to a working set or is referenced. The definition
   of workingset will be simplified as follows: a set referenced more than twice
   for MGLRU, and decoupled from MGLRU's tiering.

4. MGLRU's swappiness is kind of useless in some situations compared to
   Active / Inactive LRU, since its force protects the youngest two gen, so
   quite often we can only reclaim one type of folios. To workaround that, the
   user usually runs force aging before reclaim. So, can we just remove the
   force protection of the youngest two gens?

5. Async aging and aging optimization are also required to make the above ideas
   work better.

6. Other issues and discussion on whether the above improvements will help
   solve them or make them worse. e.g.

   For eBPF extension, using eBPF to determine which gen a folio should be
   landed given the shadow and after we have more than 4 gens, might be very
   helpful and enough for many workload customizations.

   Can we just ignore the shadow for anon folios? MGLRU basically activates
   anon folios unconditionally, especially if we combined with the LFU like
   idea above we might only want to track the 3 bit count, and get rid of
   the extra bit usage in the shadow. The eviction performance might be even
   better, and other components like swap table [3] will have more bits to use
   for better performance and more features.

The goal is:

- Reduce MGLRU's page flag usage to be identical or less compared to Active /
  Inactive LRU.
- Eliminate regressions.
- Unify or improve the metrics.
- Provides more extensibility.

Link: https://lwn.net/Articles/945266/ [1]
Link: https://github.com/ryncsn/linux/tree/improving-mglru [2]
Link: https://lore.kernel.org/linux-mm/20260218-swap-table-p3-v3-5-f4e34be021a7@tencent.com/
[3]

Re: [LSF/MM/BPF TOPIC] Improving MGLRU

[Johannes Weiner]

On Fri, Feb 20, 2026 at 01:25:33AM +0800, Kairui Song wrote:
> Hi All,
> 
> Apologies I forgot to add the proper tag in the previous email so
> resending this.
> 
> MGLRU has been introduced in the mainline for years, but we still have two LRUs
> today. There are many reasons MGLRU is still not the only LRU implementation in
> the kernel.
> 
> And I've been looking at a few major issues here:
> 
> 1. Page flag usage: MGLRU uses many more flags (3+ more) than Active/Inactive
> LRU.
> 2. Regressions: MGLRU might cause regression, even though in many workloads it
> outperforms Active/Inactive by a lot.
> 3. Metrics: MGLRU makes some metrics work differently, for example: PSI,
> /proc/meminfo.
> 4. Some reclaim behavior is less controllable.

I would be very interested in discussing this topic as well.

> 2. Regressions: Currently regression is a more major problem for us.
>    From our perspective, almost all regressions are caused by an under- or
>    overprotected file cache. MGLRU's PID protection either gets too aggressive
>    or too passive or just have a too long latency. To fix that, I'd propose a
>    LFU-like design and relax the PID's aggressiveness to make it much more
>    proactive and effective for file folios. The idea is always use 3 bits in
>    the page flags to count the referenced time (which would also replace
>    PG_workingset and PG_referenced). Initial tests showed a 30% reduction of
>    refaults, and many regressions are gone. A flow chart of how the MGLRU idea
>    might work:

Are you referring to refaults on the page cache side, or swapins?

Last time we evaluated MGLRU on Meta workloads, we noticed that it
tends to do better with zswap, but worse with disk swap. It seemed to
just prefer reclaiming anon, period.

For the balancing between anon and file to work well in all
situations, it needs to have a notion of backend speed and factor in
the respective cost of misses on each side.

> 4. MGLRU's swappiness is kind of useless in some situations compared to
>    Active / Inactive LRU, since its force protects the youngest two gen, so
>    quite often we can only reclaim one type of folios. To workaround that, the
>    user usually runs force aging before reclaim. So, can we just remove the
>    force protection of the youngest two gens?

[...]

> 6. Other issues and discussion on whether the above improvements will help
>    solve them or make them worse. e.g.

[...]

>    Can we just ignore the shadow for anon folios? MGLRU basically activates
>    anon folios unconditionally, especially if we combined with the LFU like
>    idea above we might only want to track the 3 bit count, and get rid of
>    the extra bit usage in the shadow. The eviction performance might be even
>    better, and other components like swap table [3] will have more bits to use
>    for better performance and more features.

On the face of it, both of these sounds problematic to me. Why are
anon pages special cased?

The cost of reclaiming a page is:

    reuse frequency * cost of a miss

The *type* of the page is not all that meaningful for workload
performance. The wait time is qualitatively the same.

If you assume every refaulting anon is hot, it'll fall apart when the
anon set is huge and has little locality.

Re: [LSF/MM/BPF TOPIC] Improving MGLRU

[Kairui Song]

On Sat, Feb 21, 2026 at 2:24 AM Johannes Weiner <hannes@cmpxchg.org> wrote:
>
> On Fri, Feb 20, 2026 at 01:25:33AM +0800, Kairui Song wrote:
> > Hi All,
> >
> > Apologies I forgot to add the proper tag in the previous email so
> > resending this.
> >
> > MGLRU has been introduced in the mainline for years, but we still have two LRUs
> > today. There are many reasons MGLRU is still not the only LRU implementation in
> > the kernel.
> >
> > And I've been looking at a few major issues here:
> >
> > 1. Page flag usage: MGLRU uses many more flags (3+ more) than Active/Inactive
> > LRU.
> > 2. Regressions: MGLRU might cause regression, even though in many workloads it
> > outperforms Active/Inactive by a lot.
> > 3. Metrics: MGLRU makes some metrics work differently, for example: PSI,
> > /proc/meminfo.
> > 4. Some reclaim behavior is less controllable.
>
> I would be very interested in discussing this topic as well.

Thanks, glad to hear that!

>
> > 2. Regressions: Currently regression is a more major problem for us.
> >    From our perspective, almost all regressions are caused by an under- or
> >    overprotected file cache. MGLRU's PID protection either gets too aggressive
> >    or too passive or just have a too long latency. To fix that, I'd propose a
> >    LFU-like design and relax the PID's aggressiveness to make it much more
> >    proactive and effective for file folios. The idea is always use 3 bits in
> >    the page flags to count the referenced time (which would also replace
> >    PG_workingset and PG_referenced). Initial tests showed a 30% reduction of
> >    refaults, and many regressions are gone. A flow chart of how the MGLRU idea
> >    might work:
>
> Are you referring to refaults on the page cache side, or swapins?
>
> Last time we evaluated MGLRU on Meta workloads, we noticed that it
> tends to do better with zswap, but worse with disk swap. It seemed to
> just prefer reclaiming anon, period.
>
> For the balancing between anon and file to work well in all
> situations, it needs to have a notion of backend speed and factor in
> the respective cost of misses on each side.

A bit more than that. When there is no swap, MGLRU still performs
worse in some workloads like MongoDB. From what I've noticed that's
because the PID protection is a bit too passive, and there is a force
protection in sort_folio which sometimes seems too aggressive.
Active/Inactive LRU will just move a folio to head if it's accessed
twice while in RAM, but MGLRU won't do so, as result hotter file
folios are evicted equally as the colder one until the PID gets
triggered, or still gets protected even if it hasn't been used for a
while. And by the time PID finally gets triggered, the workload might
has changed. This is fixable using the approach I mentioned though,
and it seems to be better than the Active/Inactive in all our known
cases after that, whether that is a good fix worth discussion.

I also notice Ridong has a series to apply a "heat" based reclaim,
which also looks interesting.

> >    Can we just ignore the shadow for anon folios? MGLRU basically activates
> >    anon folios unconditionally, especially if we combined with the LFU like
> >    idea above we might only want to track the 3 bit count, and get rid of
> >    the extra bit usage in the shadow. The eviction performance might be even
> >    better, and other components like swap table [3] will have more bits to use
> >    for better performance and more features.
>
> On the face of it, both of these sounds problematic to me. Why are
> anon pages special cased?
>
> The cost of reclaiming a page is:
>
>     reuse frequency * cost of a miss
>
> The *type* of the page is not all that meaningful for workload
> performance. The wait time is qualitatively the same.
>
> If you assume every refaulting anon is hot, it'll fall apart when the
> anon set is huge and has little locality.

Sorry I didn't make it clear. For MGLRU currently it already ignored
the shadow distance for re-activation. And yeah, basically all anons
are activated on fault, which turns out to be quite nice? None MGLRU
users considered that as a problem and in fact the performance looks
good.

Of course we can restore the old behavior to test the folio
against some distance (gen distance or eviction distance), or push it
further to only keep the reference bit (not completely ignore the
shadow, just only keep the reference bits, if the LFU + PID still
works well without the distance), and gain more performance and bits
to use.

BTW I tried to restore the refault distance behavior for both anon and
file folios sometime ago:
https://lwn.net/Articles/945266/

For file folios it indeed looked better, anon folios seems unchanged.
But later tests showed that it doesn't apply to all cases, and I think
something better can be used as suggested in this topic.

Re: [LSF/MM/BPF TOPIC] Improving MGLRU

[Kalesh Singh]

On Thu, Feb 19, 2026 at 9:26 AM Kairui Song <ryncsn@gmail.com> wrote:
>
> Hi All,
>
> Apologies I forgot to add the proper tag in the previous email so
> resending this.
>
> MGLRU has been introduced in the mainline for years, but we still have two LRUs
> today. There are many reasons MGLRU is still not the only LRU implementation in
> the kernel.
>
> And I've been looking at a few major issues here:
>
> 1. Page flag usage: MGLRU uses many more flags (3+ more) than Active/Inactive
> LRU.
> 2. Regressions: MGLRU might cause regression, even though in many workloads it
> outperforms Active/Inactive by a lot.
> 3. Metrics: MGLRU makes some metrics work differently, for example: PSI,
> /proc/meminfo.
> 4. Some reclaim behavior is less controllable.
>
> And other issues too.
> And I think there isn't a simple solution, but it can definitely be solved. I
> would like to propose a session to discuss a few ideas on how to solve this, and
> perhaps we can finally only have one LRU in the kernel. So I'd like topropose a
> session to discuss some ideas about improving MGLRU and making it the only LRU.
>
> Some parts are just ideas, so far I have a working series [2] following the
> LFU and metric unification idea below, solving 2) and 3) above, and
> providing some very basic infrastructures for 1). Would try to send that as
> RFC for easier review and merge once it's stable enough soon, before LSF/MM/BPF.
>
> So far, I already observed a 30% reduction of refault of total folios in
> some workloads, including Tpcc and YCSB, and several critical regressions
> compared to Active / Inactive are gone, PG_workingset and PG_referenced are
> gone, yet things like PSI are more accurate (see below), and still stay
> bitwise compatible with Active / Inactive LRU. If it went smoothly,
> we might be able to unify and have only one LRU.
>
> Following topic and ideas are the key points:
>
> 1. Flags usage: which is solvable, and the hard part is mostly about
>    implementation details: MGLRU uses (at least) 3 extra flags for the gen
>    number, and we are expecting it to use more gen flags to support more than 4
>    gen. These flags can be moved to the tail of the LRU pointer after carefully
>    modifying the kernel's convention on LRU operations. That would allow us to
>    use up to 6 bits for the gen number and support up to 63 gens. The lower bit
>    of both pointers can be packed together for CAS on gen numbers. Reducing
>    flag usage by 3. Previously, Yu also suggested moving flags like PG_active to
>    the LRU pointer tail, which could also be a way.
>
>    struct folio {
>        /* ... */
>        union {
>                struct list_head lru;
>    +           struct lru_gen_list_head lru_gen;
>
>    So whenever the folio is on lruvec, `lru_gen_list_head` is used instead of
>    `lru`, which contains encoded info. We might be able to move all LRU-related
>    flags there.
>
>    Ordinary folio lists are still just fine, since when the folio is isolated,
>    `lru` is still there. But places like folio split, will need to
> check if that's
>    a lruvec folio, or folio on an ordinary list.
>
>    This part is just an idea yet. But might make us able to have up to 63 gens
>    in upstream and enable build for every config.
>
> 2. Regressions: Currently regression is a more major problem for us.
>    From our perspective, almost all regressions are caused by an under- or
>    overprotected file cache. MGLRU's PID protection either gets too aggressive
>    or too passive or just have a too long latency. To fix that, I'd propose a
>    LFU-like design and relax the PID's aggressiveness to make it much more
>    proactive and effective for file folios. The idea is always use 3 bits in
>    the page flags to count the referenced time (which would also replace
>    PG_workingset and PG_referenced). Initial tests showed a 30% reduction of
>    refaults, and many regressions are gone. A flow chart of how the MGLRU idea
>    might work:
>
>    ========== MGLFU Tiering ==========
>    Access  3 bit    lru_gen  lru_gen |(R - PG_referenced | W - PG_workingset)
>    Count   L|W|R    refs     tier    |(L - LRU_GEN_REFS)
>    0       0|0|0    0        0       | - Readahead & Cache
>    1       0|0|1    1        0       | - LRU_REFS_REFERENCED
>    ----- WORKINGSET / PROMOTE --- <--+ - <move out of min_seq>
>    2       0|1|0    2        0       | - LRU_REFS_WORKINGSET
>    3       0|1|1    3        1       | - Frequently used
>    4       1|0|0*   4        2       |
>    5       1|0|1*   5        2       |
>    6       1|1|0*   6        3       |
>    7       1|1|1*   7        3       | - LRU_REFS_MAX
>    ---------- PROMOTION ----------> --+ - <promote to next gen>
>
>    Once a folio has an access count > LRU_REFS_WORKINGSET, it never goes lower
>    than that. Folios that hit LRU_REFS_MAX will be promoted to next gen on
>    access, and remove the force protection of folios on eviction. This provides
>    a more proactive protection.
>
>    And this might also give other frameworks like DAMON a nicer interface to
>    interact with MGLRU, since the referenced count can promote every folio and
>    count accesses in a more reasonable and unified way for MGLRU now.
>
>    NOTE: Still changing this design according to test results, e.g. maybe
>    we should optionally still use 4 bits, so the final solution might not
>    be the same.
>
>    Another potential improvement on the regression issue is implementing the
>    refault distance as I once proposed [1], which can have a huge gain for some
>    workloads with heavy file folio usage. Maybe we can have both.
>
> 3. Metrics: The key here is about the meaning of page flags, including
>    PG_workingset and PG_referenced. These two flags are set/cleared very
>    differently for MGLRU compared to Active / Inactive LRU, but many other
>    components are still using them as metrics for Active / Inactive LRU. Hence,
>    I would propose to introduce a different mechanism to unify and replace these
>    two flags: Using the 3 bits in the page flags field reserved for LFU-like
>    tracking above, to determine the folio status.
>
>    Then following the above LFU-like idea, and using helpers like:
>
>    static inline bool folio_is_referenced(const struct folio *folio)
>    {
>     return folio_lru_refs(folio) >= LRU_REFS_REFERENCED;
>    }
>
>    static inline bool folio_is_workingset(const struct folio *folio)
>    {
>     return folio_lru_refs(folio) >= LRU_REFS_WORKINGSET;
>    }
>
>    static inline bool folio_is_referenced_by_bit(struct folio *folio)
>    {    /* For compatibility */
>     return !!(READ_ONCE(*folio_flags(folio, 0)) & BIT(LRU_REFS_PGOFF));
>    }
>
>    static inline void folio_mark_workingset_by_bit(struct folio *folio)
>    {    /* For compatibility */
>     set_mask_bits(folio_flags(folio, 0), BIT(LRU_REFS_PGOFF + 1),
> BIT(LRU_REFS_PGOFF + 1));
>    }
>
>    To tell if a folio belongs to a working set or is referenced. The definition
>    of workingset will be simplified as follows: a set referenced more than twice
>    for MGLRU, and decoupled from MGLRU's tiering.
>
> 4. MGLRU's swappiness is kind of useless in some situations compared to
>    Active / Inactive LRU, since its force protects the youngest two gen, so
>    quite often we can only reclaim one type of folios. To workaround that, the
>    user usually runs force aging before reclaim. So, can we just remove the
>    force protection of the youngest two gens?
>
> 5. Async aging and aging optimization are also required to make the above ideas
>    work better.
>
> 6. Other issues and discussion on whether the above improvements will help
>    solve them or make them worse. e.g.
>
>    For eBPF extension, using eBPF to determine which gen a folio should be
>    landed given the shadow and after we have more than 4 gens, might be very
>    helpful and enough for many workload customizations.
>
>    Can we just ignore the shadow for anon folios? MGLRU basically activates
>    anon folios unconditionally, especially if we combined with the LFU like
>    idea above we might only want to track the 3 bit count, and get rid of
>    the extra bit usage in the shadow. The eviction performance might be even
>    better, and other components like swap table [3] will have more bits to use
>    for better performance and more features.
>
> The goal is:
>
> - Reduce MGLRU's page flag usage to be identical or less compared to Active /
>   Inactive LRU.
> - Eliminate regressions.
> - Unify or improve the metrics.
> - Provides more extensibility.

Hi Kairui,

I would be very interested in joining this discussion at LSF/MM.

We use MGLRU on Android. While the reduced CPU usage leads to power
improvements for mobile devices, we've run into a few notable issues
as well.

Off the top of my head:

1. Direct Reclaim Latency: We've observed that direct reclaim tail
latencies can sometimes be significantly higher with MGLRU.

2. PSI and OOM Response: Tying directly into your point about metrics,
the PSI memory pressure generated by MGLRU is consistently 30% to 40%
lower than the Active/Inactive LRU on Android workloads. Because
user-space OOM daemons like lmkd rely heavily on these metrics, this
causes them to be less quick to react to actual memory pressure.

3. Misleading Conventional LRU Metrics: We've noticed patterns in
standard memory tracking where nr_active and nr_inactive show sharp
vertical cliffs and rises. Since MGLRU derives these metrics by
mapping the two youngest generations to "active" and the two oldest to
"inactive," every time a new generation is created (incrementing the
seq counter), the second youngest generation (before the increment) is
suddenly recategorized as inactive (after the increment). Because the
newly created generation is empty, this manifests as a massive,
instantaneous drop in active pages and a corresponding spike in
inactive pages.

I'd love to participate and discuss how we might tackle these
regressions and metrics.

Thanks,
Kalesh

>
> Link: https://lwn.net/Articles/945266/ [1]
> Link: https://github.com/ryncsn/linux/tree/improving-mglru [2]
> Link: https://lore.kernel.org/linux-mm/20260218-swap-table-p3-v3-5-f4e34be021a7@tencent.com/
> [3]
>

Re: [LSF/MM/BPF TOPIC] Improving MGLRU

[Kairui Song]

On Thu, Feb 26, 2026 at 9:55 AM Kalesh Singh <kaleshsingh@google.com> wrote:
>
> On Thu, Feb 19, 2026 at 9:26 AM Kairui Song <ryncsn@gmail.com> wrote:
> >
> > Hi All,
> >
> > Apologies I forgot to add the proper tag in the previous email so
> > resending this.
> >
> > MGLRU has been introduced in the mainline for years, but we still have two LRUs
> > today. There are many reasons MGLRU is still not the only LRU implementation in
> > the kernel.
> Hi Kairui,
>
> I would be very interested in joining this discussion at LSF/MM.
>
> We use MGLRU on Android. While the reduced CPU usage leads to power
> improvements for mobile devices, we've run into a few notable issues
> as well.

Hi Kelash,

Glad to discuss this with you.

>
> Off the top of my head:
>
> 1. Direct Reclaim Latency: We've observed that direct reclaim tail
> latencies can sometimes be significantly higher with MGLRU.
>
> 2. PSI and OOM Response: Tying directly into your point about metrics,
> the PSI memory pressure generated by MGLRU is consistently 30% to 40%
> lower than the Active/Inactive LRU on Android workloads. Because
> user-space OOM daemons like lmkd rely heavily on these metrics, this
> causes them to be less quick to react to actual memory pressure.

Yes, this is one of the main issues for us too. Per our observation
one cause for that is MGLRU's usage of flags like PG_workingset is
different from active / inactive LRU, and flags like the PG_workingset
flags are bound with tiering now, so changing that requires some
redesign of how tiering works too. Which is one of the motivations
behind the LFU like tiering design I mentioned. That should make
things like PSI or readahead stable again.

> 3. Misleading Conventional LRU Metrics: We've noticed patterns in
> standard memory tracking where nr_active and nr_inactive show sharp
> vertical cliffs and rises. Since MGLRU derives these metrics by
> mapping the two youngest generations to "active" and the two oldest to
> "inactive," every time a new generation is created (incrementing the
> seq counter), the second youngest generation (before the increment) is
> suddenly recategorized as inactive (after the increment). Because the
> newly created generation is empty, this manifests as a massive,
> instantaneous drop in active pages and a corresponding spike in
> inactive pages.

That's also a major problem for things like K8s. The cliffs and rises
confuses the cloud scheduler. Our solution is also based on that new
tiering design, and counting the number of folios in different tiers
instead of gens will greatly improve the usability of nr_active /
nr_inactive. Whether this is a good design can be discussed.

>
> I'd love to participate and discuss how we might tackle these
> regressions and metrics.

Looking forward to that!

I also noticed Zicheng has another proposal, I've discussed with him
too previously about some ideas, hopefully we will make some progress
on this.

RE: [LSF/MM/BPF TOPIC] Improving MGLRU

[wangzicheng]

> -----Original Message-----
> From: Kairui Song <ryncsn@gmail.com>
> Sent: Thursday, February 26, 2026 11:07 AM
> To: Kalesh Singh <kaleshsingh@google.com>; wangzicheng
> <wangzicheng@honor.com>
> Cc: lsf-pc@lists.linux-foundation.org; Axel Rasmussen
> <axelrasmussen@google.com>; Yuanchu Xie <yuanchu@google.com>; Wei
> Xu <weixugc@google.com>; linux-mm <linux-mm@kvack.org>; android-mm
> <android-mm@google.com>; Suren Baghdasaryan <surenb@google.com>;
> T.J. Mercier <tjmercier@google.com>; Barry Song <21cnbao@gmail.com>
> Subject: Re: [LSF/MM/BPF TOPIC] Improving MGLRU
> 
> On Thu, Feb 26, 2026 at 9:55 AM Kalesh Singh <kaleshsingh@google.com>
> wrote:
> >
> > On Thu, Feb 19, 2026 at 9:26 AM Kairui Song <ryncsn@gmail.com> wrote:
> > >
> > > Hi All,
> > >
> > > Apologies I forgot to add the proper tag in the previous email so
> > > resending this.
> > >
> > > MGLRU has been introduced in the mainline for years, but we still have
> two LRUs
> > > today. There are many reasons MGLRU is still not the only LRU
> implementation in
> > > the kernel.
> > Hi Kairui,
> >
> > I would be very interested in joining this discussion at LSF/MM.
> >
> > We use MGLRU on Android. While the reduced CPU usage leads to power
> > improvements for mobile devices, we've run into a few notable issues
> > as well.
> 
> Hi Kelash,
> 
> Glad to discuss this with you.
> 
> >
> > Off the top of my head:
> >
> > 1. Direct Reclaim Latency: We've observed that direct reclaim tail
> > latencies can sometimes be significantly higher with MGLRU.
> >
> > 2. PSI and OOM Response: Tying directly into your point about metrics,
> > the PSI memory pressure generated by MGLRU is consistently 30% to 40%
> > lower than the Active/Inactive LRU on Android workloads. Because
> > user-space OOM daemons like lmkd rely heavily on these metrics, this
> > causes them to be less quick to react to actual memory pressure.
> 
> Yes, this is one of the main issues for us too. Per our observation
> one cause for that is MGLRU's usage of flags like PG_workingset is
> different from active / inactive LRU, and flags like the PG_workingset
> flags are bound with tiering now, so changing that requires some
> redesign of how tiering works too. Which is one of the motivations
> behind the LFU like tiering design I mentioned. That should make
> things like PSI or readahead stable again.
> 
> > 3. Misleading Conventional LRU Metrics: We've noticed patterns in
> > standard memory tracking where nr_active and nr_inactive show sharp
> > vertical cliffs and rises. Since MGLRU derives these metrics by
> > mapping the two youngest generations to "active" and the two oldest to
> > "inactive," every time a new generation is created (incrementing the
> > seq counter), the second youngest generation (before the increment) is
> > suddenly recategorized as inactive (after the increment). Because the
> > newly created generation is empty, this manifests as a massive,
> > instantaneous drop in active pages and a corresponding spike in
> > inactive pages.
> 
> That's also a major problem for things like K8s. The cliffs and rises
> confuses the cloud scheduler. Our solution is also based on that new
> tiering design, and counting the number of folios in different tiers
> instead of gens will greatly improve the usability of nr_active /
> nr_inactive. Whether this is a good design can be discussed.
> 
> >
> > I'd love to participate and discuss how we might tackle these
> > regressions and metrics.
> 
> Looking forward to that!
> 
> I also noticed Zicheng has another proposal, I've discussed with him
> too previously about some ideas, hopefully we will make some progress
> on this.

Hi Kairui, hi Kalesh,

Yes, we’re interested in this work.

We see file pages being under-protected in smartphone workload, and an LFU-like
approach sounds promising to better promote and protect hot file pages.
Kairui has shared the patches; we’ll backport them to our tree and report back
once we have results from our workloads.

Best,
Zicheng

Re: [LSF/MM/BPF TOPIC] Improving MGLRU

[Matthew Wilcox]

On Fri, Feb 20, 2026 at 01:25:33AM +0800, Kairui Song wrote:
> MGLRU has been introduced in the mainline for years, but we still have two LRUs
> today. There are many reasons MGLRU is still not the only LRU implementation in
> the kernel.

To my mind, the biggest problem with MGLRU is that Google dumped it on us
and ran away.  Commit 44958000bada claimed that it was now maintained and
added three people as maintainers.  In the six months since that commit,
none of those three people have any commits in mm/!  This is a shameful
state of affairs.

I say rip it out.

Re: [LSF/MM/BPF] Improving MGLRU

[Barry Song]

> 4. MGLRU's swappiness is kind of useless in some situations compared to
>   Active / Inactive LRU, since its force protects the youngest two gen, so
>   quite often we can only reclaim one type of folios. To workaround that, the
>   user usually runs force aging before reclaim. So, can we just remove the
>   force protection of the youngest two gens?

I guess not—MGLRU needs at least two generations to function,
similar to active and inactive lists, meaning it requires two lists.

You Zhao mentioned this in commit ec1c86b25f4b:
"This protocol, AKA second chance, requires a minimum of two
generations, hence MIN_NR_GENS."

But I do feel the issue is that anon and file folios currently share
the same generations. This may make anon and file be reclaimed more
fairly, but isn’t swappiness meant to allow some imbalance? Sharing
generations causes them to keep catching up with each other. We might
consider providing separate generations for them.

Thanks
Barry

Re: [LSF/MM/BPF] Improving MGLRU

[Barry Song]

>> MGLRU has been introduced in the mainline for years, but we still have two LRUs
>> today. There are many reasons MGLRU is still not the only LRU implementation in
>> the kernel.

> To my mind, the biggest problem with MGLRU is that Google dumped it on us
> and ran away.  Commit 44958000bada claimed that it was now maintained and
> added three people as maintainers.  In the six months since that commit,
> none of those three people have any commits in mm/!  This is a shameful
> state of affairs.
> 
> I say rip it out.

Hi Matthew,  
Can we keep it for now? Kairui, Zicheng, and I are working on it.

From what I’ve seen, it performs much better than the active/inactive  
approach after applying a few vendor hooks on Android, such as forced  
aging and avoiding direct activation of read-ahead folios during page  
faults, among others. To be honest, performance was worse than  
active/inactive without those hooks, which are still not in mainline.  

It just needs more work. MGLRU has many strong design aspects, including  
using more generations to differentiate cold from hot, the look-around  
mechanism to reduce scanning overhead by leveraging cache locality,  
and data structure designs that minimize lock holding.

Best regards
Barry

Re: [LSF/MM/BPF TOPIC] Improving MGLRU

[David Rientjes]

On Fri, 20 Feb 2026, Kairui Song wrote:

> Hi All,
> 
> Apologies I forgot to add the proper tag in the previous email so
> resending this.
> 
> MGLRU has been introduced in the mainline for years, but we still have two LRUs
> today. There are many reasons MGLRU is still not the only LRU implementation in
> the kernel.
> 
> And I've been looking at a few major issues here:
> 
> 1. Page flag usage: MGLRU uses many more flags (3+ more) than Active/Inactive
> LRU.
> 2. Regressions: MGLRU might cause regression, even though in many workloads it
> outperforms Active/Inactive by a lot.
> 3. Metrics: MGLRU makes some metrics work differently, for example: PSI,
> /proc/meminfo.
> 4. Some reclaim behavior is less controllable.
> 

I think this would be a very useful topic to discuss and I really like how 
this was framed in the context of what needs to be addressed so that MGLRU 
can be on a path to becoming the default implementation and we can 
eliminate two separate implementations.  Yes, MGLRU can form the basis of 
several extensions that are possible, like working set reporting, but its 
existence in the kernel shouldn't be based on future shiny features alone; 
I think priority number one should be ensuring that these issues, as well 
as others, are properly addressed with the goal of having a single unified 
implementation in the kernel that does not regress for end users.

One topic we can add here is oom handling with MGLRU, so adding in Michal 
and Shakeel.  MGLRU has working set protection to avoid thrashing by 
configuring min_ttl_ms in sysfs.  That can end up being very useful, and 
would probably be even more useful if there was a per-memcg version of it, 
but it doesn't work well for NUMA.  That's because we get a new oom kill 
context that is triggered from kswapd threads when aging is done, not by 
direct allocators like we're used to:

4167)	/*
4168)	 * The main goal is to OOM kill if every generation from all memcgs is
4169)	 * younger than min_ttl. However, another possibility is all memcgs are
4170)	 * either too small or below min.
4171)	 */
4172)	if (!reclaimable && mutex_trylock(&oom_lock)) {
4173)		struct oom_control oc = {
4174)			.gfp_mask = sc->gfp_mask,
4175)		};
4176) 
4177)		out_of_memory(&oc);
4178) 
4179)		mutex_unlock(&oom_lock);
4180)	}

That obviously just calls into the oom killer without any context about 
*which* node we're trying to free memory on.  The worst case scenario is 
that we oom kill every process on a single node without ever freeing 
memory for kswapd's node.

So I doubt that anybody is using this to actively defend against thrashing 
today, at least on NUMA systems.

One way to address this would be to consider resident memory on the nodes 
included in oc->nodemask when making oom kill decisions and then 
initialize an empty nodemaks here, sets pgdat->node_id, and passes it in.  
But it should be part of a larger discussion about how we handle targeted 
oom killing on specific NUMA nodes that would be applicable for cpusets, 
mempolicies, etc.

For cpusets, for example, we only look at the eligibility of a thread to 
allocate on a node, not the amount of anticipated freeing from that node 
on oom kill.  We could trivially do the same thing here for MGLRU, but it 
would kinda suck to go around oom killing processes that only have a 
single page on your target node.  (But, hey, better than the status quo 
today here!)

So we should talk about node-targeted oom killing and how that would make 
sense so that we can wire it up here if min_ttl_ms is to be used for 
MGLRU, at least for NUMA systems.  It's a tricky problem in oom contexts 
to be able to get at the information, per thread, that you want to 
consider to determine eligiblity but perhaps even more of a tricky problem 
when you have that information about what heursitic you use to compare 
processes with lots of memory on the system vs lots of memory on the node.

Has this been considered before?  For kswapd induced oom killing like this 
to work, it would have to be solved.

Re: [LSF/MM/BPF TOPIC] Improving MGLRU

[Vernon Yang]

On Fri, Feb 20, 2026 at 01:25:33AM +0800, Kairui Song wrote:
> Hi All,
>
> Apologies I forgot to add the proper tag in the previous email so
> resending this.
>
> MGLRU has been introduced in the mainline for years, but we still have two LRUs
> today. There are many reasons MGLRU is still not the only LRU implementation in
> the kernel.
>
> And I've been looking at a few major issues here:
>
> 1. Page flag usage: MGLRU uses many more flags (3+ more) than Active/Inactive
> LRU.
> 2. Regressions: MGLRU might cause regression, even though in many workloads it
> outperforms Active/Inactive by a lot.
> 3. Metrics: MGLRU makes some metrics work differently, for example: PSI,
> /proc/meminfo.
> 4. Some reclaim behavior is less controllable.
>
> And other issues too.
> And I think there isn't a simple solution, but it can definitely be solved. I
> would like to propose a session to discuss a few ideas on how to solve this, and
> perhaps we can finally only have one LRU in the kernel. So I'd like topropose a
> session to discuss some ideas about improving MGLRU and making it the only LRU.
>
> Some parts are just ideas, so far I have a working series [2] following the
> LFU and metric unification idea below, solving 2) and 3) above, and
> providing some very basic infrastructures for 1). Would try to send that as
> RFC for easier review and merge once it's stable enough soon, before LSF/MM/BPF.
>
> So far, I already observed a 30% reduction of refault of total folios in
> some workloads, including Tpcc and YCSB, and several critical regressions
> compared to Active / Inactive are gone, PG_workingset and PG_referenced are
> gone, yet things like PSI are more accurate (see below), and still stay
> bitwise compatible with Active / Inactive LRU. If it went smoothly,
> we might be able to unify and have only one LRU.
>
> Following topic and ideas are the key points:
>
> 1. Flags usage: which is solvable, and the hard part is mostly about
>    implementation details: MGLRU uses (at least) 3 extra flags for the gen
>    number, and we are expecting it to use more gen flags to support more than 4
>    gen. These flags can be moved to the tail of the LRU pointer after carefully
>    modifying the kernel's convention on LRU operations. That would allow us to
>    use up to 6 bits for the gen number and support up to 63 gens. The lower bit
>    of both pointers can be packed together for CAS on gen numbers. Reducing
>    flag usage by 3. Previously, Yu also suggested moving flags like PG_active to
>    the LRU pointer tail, which could also be a way.
>
>    struct folio {
>        /* ... */
>        union {
>                struct list_head lru;
>    +           struct lru_gen_list_head lru_gen;
>
>    So whenever the folio is on lruvec, `lru_gen_list_head` is used instead of
>    `lru`, which contains encoded info. We might be able to move all LRU-related
>    flags there.
>
>    Ordinary folio lists are still just fine, since when the folio is isolated,
>    `lru` is still there. But places like folio split, will need to
> check if that's
>    a lruvec folio, or folio on an ordinary list.
>
>    This part is just an idea yet. But might make us able to have up to 63 gens
>    in upstream and enable build for every config.
>
> 2. Regressions: Currently regression is a more major problem for us.
>    From our perspective, almost all regressions are caused by an under- or
>    overprotected file cache. MGLRU's PID protection either gets too aggressive
>    or too passive or just have a too long latency. To fix that, I'd propose a
>    LFU-like design and relax the PID's aggressiveness to make it much more
>    proactive and effective for file folios. The idea is always use 3 bits in
>    the page flags to count the referenced time (which would also replace
>    PG_workingset and PG_referenced). Initial tests showed a 30% reduction of
>    refaults, and many regressions are gone. A flow chart of how the MGLRU idea
>    might work:
>
>    ========== MGLFU Tiering ==========
>    Access  3 bit    lru_gen  lru_gen |(R - PG_referenced | W - PG_workingset)
>    Count   L|W|R    refs     tier    |(L - LRU_GEN_REFS)
>    0       0|0|0    0        0       | - Readahead & Cache
>    1       0|0|1    1        0       | - LRU_REFS_REFERENCED
>    ----- WORKINGSET / PROMOTE --- <--+ - <move out of min_seq>
>    2       0|1|0    2        0       | - LRU_REFS_WORKINGSET
>    3       0|1|1    3        1       | - Frequently used
>    4       1|0|0*   4        2       |
>    5       1|0|1*   5        2       |
>    6       1|1|0*   6        3       |
>    7       1|1|1*   7        3       | - LRU_REFS_MAX
>    ---------- PROMOTION ----------> --+ - <promote to next gen>
>
>    Once a folio has an access count > LRU_REFS_WORKINGSET, it never goes lower
>    than that. Folios that hit LRU_REFS_MAX will be promoted to next gen on
>    access, and remove the force protection of folios on eviction. This provides
>    a more proactive protection.
>
>    And this might also give other frameworks like DAMON a nicer interface to
>    interact with MGLRU, since the referenced count can promote every folio and
>    count accesses in a more reasonable and unified way for MGLRU now.
>
>    NOTE: Still changing this design according to test results, e.g. maybe
>    we should optionally still use 4 bits, so the final solution might not
>    be the same.
>
>    Another potential improvement on the regression issue is implementing the
>    refault distance as I once proposed [1], which can have a huge gain for some
>    workloads with heavy file folio usage. Maybe we can have both.
>
> 3. Metrics: The key here is about the meaning of page flags, including
>    PG_workingset and PG_referenced. These two flags are set/cleared very
>    differently for MGLRU compared to Active / Inactive LRU, but many other
>    components are still using them as metrics for Active / Inactive LRU. Hence,
>    I would propose to introduce a different mechanism to unify and replace these
>    two flags: Using the 3 bits in the page flags field reserved for LFU-like
>    tracking above, to determine the folio status.
>
>    Then following the above LFU-like idea, and using helpers like:
>
>    static inline bool folio_is_referenced(const struct folio *folio)
>    {
>     return folio_lru_refs(folio) >= LRU_REFS_REFERENCED;
>    }
>
>    static inline bool folio_is_workingset(const struct folio *folio)
>    {
>     return folio_lru_refs(folio) >= LRU_REFS_WORKINGSET;
>    }
>
>    static inline bool folio_is_referenced_by_bit(struct folio *folio)
>    {    /* For compatibility */
>     return !!(READ_ONCE(*folio_flags(folio, 0)) & BIT(LRU_REFS_PGOFF));
>    }
>
>    static inline void folio_mark_workingset_by_bit(struct folio *folio)
>    {    /* For compatibility */
>     set_mask_bits(folio_flags(folio, 0), BIT(LRU_REFS_PGOFF + 1),
> BIT(LRU_REFS_PGOFF + 1));
>    }
>
>    To tell if a folio belongs to a working set or is referenced. The definition
>    of workingset will be simplified as follows: a set referenced more than twice
>    for MGLRU, and decoupled from MGLRU's tiering.
>
> 4. MGLRU's swappiness is kind of useless in some situations compared to
>    Active / Inactive LRU, since its force protects the youngest two gen, so
>    quite often we can only reclaim one type of folios. To workaround that, the
>    user usually runs force aging before reclaim. So, can we just remove the
>    force protection of the youngest two gens?

Hi Kairui,

I would be very interested in discussing this topic as well.

In Linux desktop distributions, when the system rapidly enters low
memory state, it is almost impossible to enter S4, the success rate
only is 10%. When analyzing this issue, it was identified as the
inability to reclaim memory. Further investigation revealed that:

1. This phenomenon does not occur with Active/Inactive LRU, it only
   exists with MGLRU.
2. If force aging is performed before entering S4, the success rate
   exceeds 90%.

Detailed memory information is as follows.

MemFree:          269944 kB
Active:          4095536 kB
Inactive:        2831960 kB
Active(anon):    2667952 kB
Inactive(anon):   247208 kB
Active(file):    1427584 kB
Inactive(file):  2584752 kB

Since its force protects the youngest two gen, when wanting to reclaim
memory larger than the "Inactive" size, the MGLRU hard to reclaim enough
memory. e.g. hibernation mode call shrink_all_memory(3G).

We addressed this issue by implementing a retry mechanism similar to
memory.reclaim, the success rate of s4 has increased from 10% to 100%.

If we could directly remove the force protection of the youngest two
generations, this issue would also be resolved, and the solution would
be more universally applicable.

--
Cheers,
Vernon

> 5. Async aging and aging optimization are also required to make the above ideas
>    work better.
>
> 6. Other issues and discussion on whether the above improvements will help
>    solve them or make them worse. e.g.
>
>    For eBPF extension, using eBPF to determine which gen a folio should be
>    landed given the shadow and after we have more than 4 gens, might be very
>    helpful and enough for many workload customizations.
>
>    Can we just ignore the shadow for anon folios? MGLRU basically activates
>    anon folios unconditionally, especially if we combined with the LFU like
>    idea above we might only want to track the 3 bit count, and get rid of
>    the extra bit usage in the shadow. The eviction performance might be even
>    better, and other components like swap table [3] will have more bits to use
>    for better performance and more features.
>
> The goal is:
>
> - Reduce MGLRU's page flag usage to be identical or less compared to Active /
>   Inactive LRU.
> - Eliminate regressions.
> - Unify or improve the metrics.
> - Provides more extensibility.
>
> Link: https://lwn.net/Articles/945266/ [1]
> Link: https://github.com/ryncsn/linux/tree/improving-mglru [2]
> Link: https://lore.kernel.org/linux-mm/20260218-swap-table-p3-v3-5-f4e34be021a7@tencent.com/
> [3]
>