[v11,2/2] memory tier: create CPUless memory tiers after obtaining HMAT info

Commit Message

Ho-Ren (Jack) Chuang April 5, 2024, 12:07 a.m. UTC

The current implementation treats emulated memory devices, such as
CXL1.1 type3 memory, as normal DRAM when they are emulated as normal memory
(E820_TYPE_RAM). However, these emulated devices have different
characteristics than traditional DRAM, making it important to
distinguish them. Thus, we modify the tiered memory initialization process
to introduce a delay specifically for CPUless NUMA nodes. This delay
ensures that the memory tier initialization for these nodes is deferred
until HMAT information is obtained during the boot process. Finally,
demotion tables are recalculated at the end.

* late_initcall(memory_tier_late_init);
Some device drivers may have initialized memory tiers between
`memory_tier_init()` and `memory_tier_late_init()`, potentially bringing
online memory nodes and configuring memory tiers. They should be excluded
in the late init.

* Handle cases where there is no HMAT when creating memory tiers
There is a scenario where a CPUless node does not provide HMAT information.
If no HMAT is specified, it falls back to using the default DRAM tier.

* Introduce another new lock `default_dram_perf_lock` for adist calculation
In the current implementation, iterating through CPUlist nodes requires
holding the `memory_tier_lock`. However, `mt_calc_adistance()` will end up
trying to acquire the same lock, leading to a potential deadlock.
Therefore, we propose introducing a standalone `default_dram_perf_lock` to
protect `default_dram_perf_*`. This approach not only avoids deadlock
but also prevents holding a large lock simultaneously.

* Upgrade `set_node_memory_tier` to support additional cases, including
  default DRAM, late CPUless, and hot-plugged initializations.
To cover hot-plugged memory nodes, `mt_calc_adistance()` and
`mt_find_alloc_memory_type()` are moved into `set_node_memory_tier()` to
handle cases where memtype is not initialized and where HMAT information is
available.

* Introduce `default_memory_types` for those memory types that are not
  initialized by device drivers.
Because late initialized memory and default DRAM memory need to be managed,
a default memory type is created for storing all memory types that are
not initialized by device drivers and as a fallback.

Signed-off-by: Ho-Ren (Jack) Chuang <[e-mail]>
Signed-off-by: Hao Xiang <[e-mail]>
Reviewed-by: "Huang, Ying" <[e-mail]>
---
 mm/memory-tiers.c | 94 +++++++++++++++++++++++++++++++++++------------
 1 file changed, 70 insertions(+), 24 deletions(-)

Comments

Jonathan Cameron April 5, 2024, 2:02 p.m. UTC | #1

On Fri,  5 Apr 2024 00:07:06 +0000
"Ho-Ren (Jack) Chuang" <[e-mail]> wrote:

> The current implementation treats emulated memory devices, such as
> CXL1.1 type3 memory, as normal DRAM when they are emulated as normal memory
> (E820_TYPE_RAM). However, these emulated devices have different
> characteristics than traditional DRAM, making it important to
> distinguish them. Thus, we modify the tiered memory initialization process
> to introduce a delay specifically for CPUless NUMA nodes. This delay
> ensures that the memory tier initialization for these nodes is deferred
> until HMAT information is obtained during the boot process. Finally,
> demotion tables are recalculated at the end.
> 
> * late_initcall(memory_tier_late_init);
> Some device drivers may have initialized memory tiers between
> `memory_tier_init()` and `memory_tier_late_init()`, potentially bringing
> online memory nodes and configuring memory tiers. They should be excluded
> in the late init.
> 
> * Handle cases where there is no HMAT when creating memory tiers
> There is a scenario where a CPUless node does not provide HMAT information.
> If no HMAT is specified, it falls back to using the default DRAM tier.
> 
> * Introduce another new lock `default_dram_perf_lock` for adist calculation
> In the current implementation, iterating through CPUlist nodes requires
> holding the `memory_tier_lock`. However, `mt_calc_adistance()` will end up
> trying to acquire the same lock, leading to a potential deadlock.
> Therefore, we propose introducing a standalone `default_dram_perf_lock` to
> protect `default_dram_perf_*`. This approach not only avoids deadlock
> but also prevents holding a large lock simultaneously.
> 
> * Upgrade `set_node_memory_tier` to support additional cases, including
>   default DRAM, late CPUless, and hot-plugged initializations.
> To cover hot-plugged memory nodes, `mt_calc_adistance()` and
> `mt_find_alloc_memory_type()` are moved into `set_node_memory_tier()` to
> handle cases where memtype is not initialized and where HMAT information is
> available.
> 
> * Introduce `default_memory_types` for those memory types that are not
>   initialized by device drivers.
> Because late initialized memory and default DRAM memory need to be managed,
> a default memory type is created for storing all memory types that are
> not initialized by device drivers and as a fallback.
> 
> Signed-off-by: Ho-Ren (Jack) Chuang <[e-mail]>
> Signed-off-by: Hao Xiang <[e-mail]>
> Reviewed-by: "Huang, Ying" <[e-mail]>

Hi - one remaining question. Why can't we delay init for all nodes
to either drivers or your fallback late_initcall code.
It would be nice to reduce possible code paths.

Jonathan


> ---
>  mm/memory-tiers.c | 94 +++++++++++++++++++++++++++++++++++------------
>  1 file changed, 70 insertions(+), 24 deletions(-)
> 
> diff --git a/mm/memory-tiers.c b/mm/memory-tiers.c
> index 516b144fd45a..6632102bd5c9 100644
> --- a/mm/memory-tiers.c
> +++ b/mm/memory-tiers.c



> @@ -855,7 +892,8 @@ static int __init memory_tier_init(void)
>  	 * For now we can have 4 faster memory tiers with smaller adistance
>  	 * than default DRAM tier.
>  	 */
> -	default_dram_type = alloc_memory_type(MEMTIER_ADISTANCE_DRAM);
> +	default_dram_type = mt_find_alloc_memory_type(MEMTIER_ADISTANCE_DRAM,
> +						      &default_memory_types);
>  	if (IS_ERR(default_dram_type))
>  		panic("%s() failed to allocate default DRAM tier\n", __func__);
>  
> @@ -865,6 +903,14 @@ static int __init memory_tier_init(void)
>  	 * types assigned.
>  	 */
>  	for_each_node_state(node, N_MEMORY) {
> +		if (!node_state(node, N_CPU))
> +			/*
> +			 * Defer memory tier initialization on
> +			 * CPUless numa nodes. These will be initialized
> +			 * after firmware and devices are initialized.

Could the comment also say why we can't defer them all?

(In an odd coincidence we have a similar issue for some CPU hotplug
 related bring up where review feedback was move all cases later).

> +			 */
> +			continue;
> +
>  		memtier = set_node_memory_tier(node);
>  		if (IS_ERR(memtier))
>  			/*

Ho-Ren (Jack) Chuang April 5, 2024, 10:43 p.m. UTC | #2

On Fri, Apr 5, 2024 at 7:03 AM Jonathan Cameron
<[e-mail]> wrote:
>
> On Fri,  5 Apr 2024 00:07:06 +0000
> "Ho-Ren (Jack) Chuang" <[e-mail]> wrote:
>
> > The current implementation treats emulated memory devices, such as
> > CXL1.1 type3 memory, as normal DRAM when they are emulated as normal memory
> > (E820_TYPE_RAM). However, these emulated devices have different
> > characteristics than traditional DRAM, making it important to
> > distinguish them. Thus, we modify the tiered memory initialization process
> > to introduce a delay specifically for CPUless NUMA nodes. This delay
> > ensures that the memory tier initialization for these nodes is deferred
> > until HMAT information is obtained during the boot process. Finally,
> > demotion tables are recalculated at the end.
> >
> > * late_initcall(memory_tier_late_init);
> > Some device drivers may have initialized memory tiers between
> > `memory_tier_init()` and `memory_tier_late_init()`, potentially bringing
> > online memory nodes and configuring memory tiers. They should be excluded
> > in the late init.
> >
> > * Handle cases where there is no HMAT when creating memory tiers
> > There is a scenario where a CPUless node does not provide HMAT information.
> > If no HMAT is specified, it falls back to using the default DRAM tier.
> >
> > * Introduce another new lock `default_dram_perf_lock` for adist calculation
> > In the current implementation, iterating through CPUlist nodes requires
> > holding the `memory_tier_lock`. However, `mt_calc_adistance()` will end up
> > trying to acquire the same lock, leading to a potential deadlock.
> > Therefore, we propose introducing a standalone `default_dram_perf_lock` to
> > protect `default_dram_perf_*`. This approach not only avoids deadlock
> > but also prevents holding a large lock simultaneously.
> >
> > * Upgrade `set_node_memory_tier` to support additional cases, including
> >   default DRAM, late CPUless, and hot-plugged initializations.
> > To cover hot-plugged memory nodes, `mt_calc_adistance()` and
> > `mt_find_alloc_memory_type()` are moved into `set_node_memory_tier()` to
> > handle cases where memtype is not initialized and where HMAT information is
> > available.
> >
> > * Introduce `default_memory_types` for those memory types that are not
> >   initialized by device drivers.
> > Because late initialized memory and default DRAM memory need to be managed,
> > a default memory type is created for storing all memory types that are
> > not initialized by device drivers and as a fallback.
> >
> > Signed-off-by: Ho-Ren (Jack) Chuang <[e-mail]>
> > Signed-off-by: Hao Xiang <[e-mail]>
> > Reviewed-by: "Huang, Ying" <[e-mail]>
>
> Hi - one remaining question. Why can't we delay init for all nodes
> to either drivers or your fallback late_initcall code.
> It would be nice to reduce possible code paths.

I try not to change too much of the existing code structure in
this patchset.

To me, postponing/moving all memory tier registrations to
late_initcall() is another possible action item for the next patchset.

After tier_mem(), hmat_init() is called, which requires registering
`default_dram_type` info. This is when `default_dram_type` is needed.
However, it is indeed possible to postpone the latter part,
set_node_memory_tier(), to `late_init(). So, memory_tier_init() can
indeed be split into two parts, and the latter part can be moved to
late_initcall() to be processed together.

Doing this all memory-type drivers have to call late_initcall() to
register a memory tier. I’m not sure how many they are?

What do you guys think?

>
> Jonathan
>
>
> > ---
> >  mm/memory-tiers.c | 94 +++++++++++++++++++++++++++++++++++------------
> >  1 file changed, 70 insertions(+), 24 deletions(-)
> >
> > diff --git a/mm/memory-tiers.c b/mm/memory-tiers.c
> > index 516b144fd45a..6632102bd5c9 100644
> > --- a/mm/memory-tiers.c
> > +++ b/mm/memory-tiers.c
>
>
>
> > @@ -855,7 +892,8 @@ static int __init memory_tier_init(void)
> >        * For now we can have 4 faster memory tiers with smaller adistance
> >        * than default DRAM tier.
> >        */
> > -     default_dram_type = alloc_memory_type(MEMTIER_ADISTANCE_DRAM);
> > +     default_dram_type = mt_find_alloc_memory_type(MEMTIER_ADISTANCE_DRAM,
> > +                                                   &default_memory_types);
> >       if (IS_ERR(default_dram_type))
> >               panic("%s() failed to allocate default DRAM tier\n", __func__);
> >
> > @@ -865,6 +903,14 @@ static int __init memory_tier_init(void)
> >        * types assigned.
> >        */
> >       for_each_node_state(node, N_MEMORY) {
> > +             if (!node_state(node, N_CPU))
> > +                     /*
> > +                      * Defer memory tier initialization on
> > +                      * CPUless numa nodes. These will be initialized
> > +                      * after firmware and devices are initialized.
>
> Could the comment also say why we can't defer them all?
>
> (In an odd coincidence we have a similar issue for some CPU hotplug
>  related bring up where review feedback was move all cases later).
>
> > +                      */
> > +                     continue;
> > +
> >               memtier = set_node_memory_tier(node);
> >               if (IS_ERR(memtier))
> >                       /*
>

Jonathan Cameron April 9, 2024, 4:12 p.m. UTC | #3

On Fri, 5 Apr 2024 15:43:47 -0700
"Ho-Ren (Jack) Chuang" <[e-mail]> wrote:

> On Fri, Apr 5, 2024 at 7:03 AM Jonathan Cameron
> <[e-mail]> wrote:
> >
> > On Fri,  5 Apr 2024 00:07:06 +0000
> > "Ho-Ren (Jack) Chuang" <[e-mail]> wrote:
> >  
> > > The current implementation treats emulated memory devices, such as
> > > CXL1.1 type3 memory, as normal DRAM when they are emulated as normal memory
> > > (E820_TYPE_RAM). However, these emulated devices have different
> > > characteristics than traditional DRAM, making it important to
> > > distinguish them. Thus, we modify the tiered memory initialization process
> > > to introduce a delay specifically for CPUless NUMA nodes. This delay
> > > ensures that the memory tier initialization for these nodes is deferred
> > > until HMAT information is obtained during the boot process. Finally,
> > > demotion tables are recalculated at the end.
> > >
> > > * late_initcall(memory_tier_late_init);
> > > Some device drivers may have initialized memory tiers between
> > > `memory_tier_init()` and `memory_tier_late_init()`, potentially bringing
> > > online memory nodes and configuring memory tiers. They should be excluded
> > > in the late init.
> > >
> > > * Handle cases where there is no HMAT when creating memory tiers
> > > There is a scenario where a CPUless node does not provide HMAT information.
> > > If no HMAT is specified, it falls back to using the default DRAM tier.
> > >
> > > * Introduce another new lock `default_dram_perf_lock` for adist calculation
> > > In the current implementation, iterating through CPUlist nodes requires
> > > holding the `memory_tier_lock`. However, `mt_calc_adistance()` will end up
> > > trying to acquire the same lock, leading to a potential deadlock.
> > > Therefore, we propose introducing a standalone `default_dram_perf_lock` to
> > > protect `default_dram_perf_*`. This approach not only avoids deadlock
> > > but also prevents holding a large lock simultaneously.
> > >
> > > * Upgrade `set_node_memory_tier` to support additional cases, including
> > >   default DRAM, late CPUless, and hot-plugged initializations.
> > > To cover hot-plugged memory nodes, `mt_calc_adistance()` and
> > > `mt_find_alloc_memory_type()` are moved into `set_node_memory_tier()` to
> > > handle cases where memtype is not initialized and where HMAT information is
> > > available.
> > >
> > > * Introduce `default_memory_types` for those memory types that are not
> > >   initialized by device drivers.
> > > Because late initialized memory and default DRAM memory need to be managed,
> > > a default memory type is created for storing all memory types that are
> > > not initialized by device drivers and as a fallback.
> > >
> > > Signed-off-by: Ho-Ren (Jack) Chuang <[e-mail]>
> > > Signed-off-by: Hao Xiang <[e-mail]>
> > > Reviewed-by: "Huang, Ying" <[e-mail]>  
> >
> > Hi - one remaining question. Why can't we delay init for all nodes
> > to either drivers or your fallback late_initcall code.
> > It would be nice to reduce possible code paths.  
> 
> I try not to change too much of the existing code structure in
> this patchset.
> 
> To me, postponing/moving all memory tier registrations to
> late_initcall() is another possible action item for the next patchset.
> 
> After tier_mem(), hmat_init() is called, which requires registering
> `default_dram_type` info. This is when `default_dram_type` is needed.
> However, it is indeed possible to postpone the latter part,
> set_node_memory_tier(), to `late_init(). So, memory_tier_init() can
> indeed be split into two parts, and the latter part can be moved to
> late_initcall() to be processed together.
> 
> Doing this all memory-type drivers have to call late_initcall() to
> register a memory tier. I’m not sure how many they are?
> 
> What do you guys think?

Gut feeling - if you are going to move it for some cases, move it for
all of them.  Then we only have to test once ;)

J
> 
> >
> > Jonathan
> >
> >  
> > > ---
> > >  mm/memory-tiers.c | 94 +++++++++++++++++++++++++++++++++++------------
> > >  1 file changed, 70 insertions(+), 24 deletions(-)
> > >
> > > diff --git a/mm/memory-tiers.c b/mm/memory-tiers.c
> > > index 516b144fd45a..6632102bd5c9 100644
> > > --- a/mm/memory-tiers.c
> > > +++ b/mm/memory-tiers.c  
> >
> >
> >  
> > > @@ -855,7 +892,8 @@ static int __init memory_tier_init(void)
> > >        * For now we can have 4 faster memory tiers with smaller adistance
> > >        * than default DRAM tier.
> > >        */
> > > -     default_dram_type = alloc_memory_type(MEMTIER_ADISTANCE_DRAM);
> > > +     default_dram_type = mt_find_alloc_memory_type(MEMTIER_ADISTANCE_DRAM,
> > > +                                                   &default_memory_types);
> > >       if (IS_ERR(default_dram_type))
> > >               panic("%s() failed to allocate default DRAM tier\n", __func__);
> > >
> > > @@ -865,6 +903,14 @@ static int __init memory_tier_init(void)
> > >        * types assigned.
> > >        */
> > >       for_each_node_state(node, N_MEMORY) {
> > > +             if (!node_state(node, N_CPU))
> > > +                     /*
> > > +                      * Defer memory tier initialization on
> > > +                      * CPUless numa nodes. These will be initialized
> > > +                      * after firmware and devices are initialized.  
> >
> > Could the comment also say why we can't defer them all?
> >
> > (In an odd coincidence we have a similar issue for some CPU hotplug
> >  related bring up where review feedback was move all cases later).
> >  
> > > +                      */
> > > +                     continue;
> > > +
> > >               memtier = set_node_memory_tier(node);
> > >               if (IS_ERR(memtier))
> > >                       /*  
> >  
> 
>

Ho-Ren (Jack) Chuang April 9, 2024, 7:02 p.m. UTC | #4

Hi Jonathan,

On Tue, Apr 9, 2024 at 9:12 AM Jonathan Cameron
<[e-mail]> wrote:
>
> On Fri, 5 Apr 2024 15:43:47 -0700
> "Ho-Ren (Jack) Chuang" <[e-mail]> wrote:
>
> > On Fri, Apr 5, 2024 at 7:03 AM Jonathan Cameron
> > <[e-mail]> wrote:
> > >
> > > On Fri,  5 Apr 2024 00:07:06 +0000
> > > "Ho-Ren (Jack) Chuang" <[e-mail]> wrote:
> > >
> > > > The current implementation treats emulated memory devices, such as
> > > > CXL1.1 type3 memory, as normal DRAM when they are emulated as normal memory
> > > > (E820_TYPE_RAM). However, these emulated devices have different
> > > > characteristics than traditional DRAM, making it important to
> > > > distinguish them. Thus, we modify the tiered memory initialization process
> > > > to introduce a delay specifically for CPUless NUMA nodes. This delay
> > > > ensures that the memory tier initialization for these nodes is deferred
> > > > until HMAT information is obtained during the boot process. Finally,
> > > > demotion tables are recalculated at the end.
> > > >
> > > > * late_initcall(memory_tier_late_init);
> > > > Some device drivers may have initialized memory tiers between
> > > > `memory_tier_init()` and `memory_tier_late_init()`, potentially bringing
> > > > online memory nodes and configuring memory tiers. They should be excluded
> > > > in the late init.
> > > >
> > > > * Handle cases where there is no HMAT when creating memory tiers
> > > > There is a scenario where a CPUless node does not provide HMAT information.
> > > > If no HMAT is specified, it falls back to using the default DRAM tier.
> > > >
> > > > * Introduce another new lock `default_dram_perf_lock` for adist calculation
> > > > In the current implementation, iterating through CPUlist nodes requires
> > > > holding the `memory_tier_lock`. However, `mt_calc_adistance()` will end up
> > > > trying to acquire the same lock, leading to a potential deadlock.
> > > > Therefore, we propose introducing a standalone `default_dram_perf_lock` to
> > > > protect `default_dram_perf_*`. This approach not only avoids deadlock
> > > > but also prevents holding a large lock simultaneously.
> > > >
> > > > * Upgrade `set_node_memory_tier` to support additional cases, including
> > > >   default DRAM, late CPUless, and hot-plugged initializations.
> > > > To cover hot-plugged memory nodes, `mt_calc_adistance()` and
> > > > `mt_find_alloc_memory_type()` are moved into `set_node_memory_tier()` to
> > > > handle cases where memtype is not initialized and where HMAT information is
> > > > available.
> > > >
> > > > * Introduce `default_memory_types` for those memory types that are not
> > > >   initialized by device drivers.
> > > > Because late initialized memory and default DRAM memory need to be managed,
> > > > a default memory type is created for storing all memory types that are
> > > > not initialized by device drivers and as a fallback.
> > > >
> > > > Signed-off-by: Ho-Ren (Jack) Chuang <[e-mail]>
> > > > Signed-off-by: Hao Xiang <[e-mail]>
> > > > Reviewed-by: "Huang, Ying" <[e-mail]>
> > >
> > > Hi - one remaining question. Why can't we delay init for all nodes
> > > to either drivers or your fallback late_initcall code.
> > > It would be nice to reduce possible code paths.
> >
> > I try not to change too much of the existing code structure in
> > this patchset.
> >
> > To me, postponing/moving all memory tier registrations to
> > late_initcall() is another possible action item for the next patchset.
> >
> > After tier_mem(), hmat_init() is called, which requires registering
> > `default_dram_type` info. This is when `default_dram_type` is needed.
> > However, it is indeed possible to postpone the latter part,
> > set_node_memory_tier(), to `late_init(). So, memory_tier_init() can
> > indeed be split into two parts, and the latter part can be moved to
> > late_initcall() to be processed together.
> >
> > Doing this all memory-type drivers have to call late_initcall() to
> > register a memory tier. I’m not sure how many they are?
> >
> > What do you guys think?
>
> Gut feeling - if you are going to move it for some cases, move it for
> all of them.  Then we only have to test once ;)
>
> J

Thank you for your reminder! I agree~ That's why I'm considering
changing them in the next patchset because of the amount of changes.
And also, this patchset already contains too many things.

> >
> > >
> > > Jonathan
> > >
> > >
> > > > ---
> > > >  mm/memory-tiers.c | 94 +++++++++++++++++++++++++++++++++++------------
> > > >  1 file changed, 70 insertions(+), 24 deletions(-)
> > > >
> > > > diff --git a/mm/memory-tiers.c b/mm/memory-tiers.c
> > > > index 516b144fd45a..6632102bd5c9 100644
> > > > --- a/mm/memory-tiers.c
> > > > +++ b/mm/memory-tiers.c
> > >
> > >
> > >
> > > > @@ -855,7 +892,8 @@ static int __init memory_tier_init(void)
> > > >        * For now we can have 4 faster memory tiers with smaller adistance
> > > >        * than default DRAM tier.
> > > >        */
> > > > -     default_dram_type = alloc_memory_type(MEMTIER_ADISTANCE_DRAM);
> > > > +     default_dram_type = mt_find_alloc_memory_type(MEMTIER_ADISTANCE_DRAM,
> > > > +                                                   &default_memory_types);
> > > >       if (IS_ERR(default_dram_type))
> > > >               panic("%s() failed to allocate default DRAM tier\n", __func__);
> > > >
> > > > @@ -865,6 +903,14 @@ static int __init memory_tier_init(void)
> > > >        * types assigned.
> > > >        */
> > > >       for_each_node_state(node, N_MEMORY) {
> > > > +             if (!node_state(node, N_CPU))
> > > > +                     /*
> > > > +                      * Defer memory tier initialization on
> > > > +                      * CPUless numa nodes. These will be initialized
> > > > +                      * after firmware and devices are initialized.
> > >
> > > Could the comment also say why we can't defer them all?
> > >
> > > (In an odd coincidence we have a similar issue for some CPU hotplug
> > >  related bring up where review feedback was move all cases later).
> > >
> > > > +                      */
> > > > +                     continue;
> > > > +
> > > >               memtier = set_node_memory_tier(node);
> > > >               if (IS_ERR(memtier))
> > > >                       /*
> > >
> >
> >
>

Huang, Ying April 10, 2024, 2:30 a.m. UTC | #5

"Ho-Ren (Jack) Chuang" <[e-mail]> writes:

> On Fri, Apr 5, 2024 at 7:03 AM Jonathan Cameron
> <[e-mail]> wrote:
>>
>> On Fri,  5 Apr 2024 00:07:06 +0000
>> "Ho-Ren (Jack) Chuang" <[e-mail]> wrote:
>>
>> > The current implementation treats emulated memory devices, such as
>> > CXL1.1 type3 memory, as normal DRAM when they are emulated as normal memory
>> > (E820_TYPE_RAM). However, these emulated devices have different
>> > characteristics than traditional DRAM, making it important to
>> > distinguish them. Thus, we modify the tiered memory initialization process
>> > to introduce a delay specifically for CPUless NUMA nodes. This delay
>> > ensures that the memory tier initialization for these nodes is deferred
>> > until HMAT information is obtained during the boot process. Finally,
>> > demotion tables are recalculated at the end.
>> >
>> > * late_initcall(memory_tier_late_init);
>> > Some device drivers may have initialized memory tiers between
>> > `memory_tier_init()` and `memory_tier_late_init()`, potentially bringing
>> > online memory nodes and configuring memory tiers. They should be excluded
>> > in the late init.
>> >
>> > * Handle cases where there is no HMAT when creating memory tiers
>> > There is a scenario where a CPUless node does not provide HMAT information.
>> > If no HMAT is specified, it falls back to using the default DRAM tier.
>> >
>> > * Introduce another new lock `default_dram_perf_lock` for adist calculation
>> > In the current implementation, iterating through CPUlist nodes requires
>> > holding the `memory_tier_lock`. However, `mt_calc_adistance()` will end up
>> > trying to acquire the same lock, leading to a potential deadlock.
>> > Therefore, we propose introducing a standalone `default_dram_perf_lock` to
>> > protect `default_dram_perf_*`. This approach not only avoids deadlock
>> > but also prevents holding a large lock simultaneously.
>> >
>> > * Upgrade `set_node_memory_tier` to support additional cases, including
>> >   default DRAM, late CPUless, and hot-plugged initializations.
>> > To cover hot-plugged memory nodes, `mt_calc_adistance()` and
>> > `mt_find_alloc_memory_type()` are moved into `set_node_memory_tier()` to
>> > handle cases where memtype is not initialized and where HMAT information is
>> > available.
>> >
>> > * Introduce `default_memory_types` for those memory types that are not
>> >   initialized by device drivers.
>> > Because late initialized memory and default DRAM memory need to be managed,
>> > a default memory type is created for storing all memory types that are
>> > not initialized by device drivers and as a fallback.
>> >
>> > Signed-off-by: Ho-Ren (Jack) Chuang <[e-mail]>
>> > Signed-off-by: Hao Xiang <[e-mail]>
>> > Reviewed-by: "Huang, Ying" <[e-mail]>
>>
>> Hi - one remaining question. Why can't we delay init for all nodes
>> to either drivers or your fallback late_initcall code.
>> It would be nice to reduce possible code paths.
>
> I try not to change too much of the existing code structure in
> this patchset.
>
> To me, postponing/moving all memory tier registrations to
> late_initcall() is another possible action item for the next patchset.
>
> After tier_mem(), hmat_init() is called, which requires registering
> `default_dram_type` info. This is when `default_dram_type` is needed.
> However, it is indeed possible to postpone the latter part,
> set_node_memory_tier(), to `late_init(). So, memory_tier_init() can
> indeed be split into two parts, and the latter part can be moved to
> late_initcall() to be processed together.

I don't think that it's good to move all memory_tier initialization in
drivers to late_initcall().  It's natural to keep them in
device_initcall() level.

If so, we can allocate default_dram_type in memory_tier_init(), and call
set_node_memory_tier() only in memory_tier_lateinit().  We can call
memory_tier_lateinit() in device_initcall() level too.

--
Best Regards,
Huang, Ying

> Doing this all memory-type drivers have to call late_initcall() to
> register a memory tier. I’m not sure how many they are?
>
> What do you guys think?
>
>>
>> Jonathan
>>
>>
>> > ---
>> >  mm/memory-tiers.c | 94 +++++++++++++++++++++++++++++++++++------------
>> >  1 file changed, 70 insertions(+), 24 deletions(-)
>> >
>> > diff --git a/mm/memory-tiers.c b/mm/memory-tiers.c
>> > index 516b144fd45a..6632102bd5c9 100644
>> > --- a/mm/memory-tiers.c
>> > +++ b/mm/memory-tiers.c
>>
>>
>>
>> > @@ -855,7 +892,8 @@ static int __init memory_tier_init(void)
>> >        * For now we can have 4 faster memory tiers with smaller adistance
>> >        * than default DRAM tier.
>> >        */
>> > -     default_dram_type = alloc_memory_type(MEMTIER_ADISTANCE_DRAM);
>> > +     default_dram_type = mt_find_alloc_memory_type(MEMTIER_ADISTANCE_DRAM,
>> > +                                                   &default_memory_types);
>> >       if (IS_ERR(default_dram_type))
>> >               panic("%s() failed to allocate default DRAM tier\n", __func__);
>> >
>> > @@ -865,6 +903,14 @@ static int __init memory_tier_init(void)
>> >        * types assigned.
>> >        */
>> >       for_each_node_state(node, N_MEMORY) {
>> > +             if (!node_state(node, N_CPU))
>> > +                     /*
>> > +                      * Defer memory tier initialization on
>> > +                      * CPUless numa nodes. These will be initialized
>> > +                      * after firmware and devices are initialized.
>>
>> Could the comment also say why we can't defer them all?
>>
>> (In an odd coincidence we have a similar issue for some CPU hotplug
>>  related bring up where review feedback was move all cases later).
>>
>> > +                      */
>> > +                     continue;
>> > +
>> >               memtier = set_node_memory_tier(node);
>> >               if (IS_ERR(memtier))
>> >                       /*
>>

Ho-Ren (Jack) Chuang April 10, 2024, 5:55 a.m. UTC | #6

On Tue, Apr 9, 2024 at 7:33 PM Huang, Ying <[e-mail]> wrote:
>
> "Ho-Ren (Jack) Chuang" <[e-mail]> writes:
>
> > On Fri, Apr 5, 2024 at 7:03 AM Jonathan Cameron
> > <[e-mail]> wrote:
> >>
> >> On Fri,  5 Apr 2024 00:07:06 +0000
> >> "Ho-Ren (Jack) Chuang" <[e-mail]> wrote:
> >>
> >> > The current implementation treats emulated memory devices, such as
> >> > CXL1.1 type3 memory, as normal DRAM when they are emulated as normal memory
> >> > (E820_TYPE_RAM). However, these emulated devices have different
> >> > characteristics than traditional DRAM, making it important to
> >> > distinguish them. Thus, we modify the tiered memory initialization process
> >> > to introduce a delay specifically for CPUless NUMA nodes. This delay
> >> > ensures that the memory tier initialization for these nodes is deferred
> >> > until HMAT information is obtained during the boot process. Finally,
> >> > demotion tables are recalculated at the end.
> >> >
> >> > * late_initcall(memory_tier_late_init);
> >> > Some device drivers may have initialized memory tiers between
> >> > `memory_tier_init()` and `memory_tier_late_init()`, potentially bringing
> >> > online memory nodes and configuring memory tiers. They should be excluded
> >> > in the late init.
> >> >
> >> > * Handle cases where there is no HMAT when creating memory tiers
> >> > There is a scenario where a CPUless node does not provide HMAT information.
> >> > If no HMAT is specified, it falls back to using the default DRAM tier.
> >> >
> >> > * Introduce another new lock `default_dram_perf_lock` for adist calculation
> >> > In the current implementation, iterating through CPUlist nodes requires
> >> > holding the `memory_tier_lock`. However, `mt_calc_adistance()` will end up
> >> > trying to acquire the same lock, leading to a potential deadlock.
> >> > Therefore, we propose introducing a standalone `default_dram_perf_lock` to
> >> > protect `default_dram_perf_*`. This approach not only avoids deadlock
> >> > but also prevents holding a large lock simultaneously.
> >> >
> >> > * Upgrade `set_node_memory_tier` to support additional cases, including
> >> >   default DRAM, late CPUless, and hot-plugged initializations.
> >> > To cover hot-plugged memory nodes, `mt_calc_adistance()` and
> >> > `mt_find_alloc_memory_type()` are moved into `set_node_memory_tier()` to
> >> > handle cases where memtype is not initialized and where HMAT information is
> >> > available.
> >> >
> >> > * Introduce `default_memory_types` for those memory types that are not
> >> >   initialized by device drivers.
> >> > Because late initialized memory and default DRAM memory need to be managed,
> >> > a default memory type is created for storing all memory types that are
> >> > not initialized by device drivers and as a fallback.
> >> >
> >> > Signed-off-by: Ho-Ren (Jack) Chuang <[e-mail]>
> >> > Signed-off-by: Hao Xiang <[e-mail]>
> >> > Reviewed-by: "Huang, Ying" <[e-mail]>
> >>
> >> Hi - one remaining question. Why can't we delay init for all nodes
> >> to either drivers or your fallback late_initcall code.
> >> It would be nice to reduce possible code paths.
> >
> > I try not to change too much of the existing code structure in
> > this patchset.
> >
> > To me, postponing/moving all memory tier registrations to
> > late_initcall() is another possible action item for the next patchset.
> >
> > After tier_mem(), hmat_init() is called, which requires registering
> > `default_dram_type` info. This is when `default_dram_type` is needed.
> > However, it is indeed possible to postpone the latter part,
> > set_node_memory_tier(), to `late_init(). So, memory_tier_init() can
> > indeed be split into two parts, and the latter part can be moved to
> > late_initcall() to be processed together.
>
> I don't think that it's good to move all memory_tier initialization in
> drivers to late_initcall().  It's natural to keep them in
> device_initcall() level.
>
> If so, we can allocate default_dram_type in memory_tier_init(), and call
> set_node_memory_tier() only in memory_tier_lateinit().  We can call
> memory_tier_lateinit() in device_initcall() level too.
>

It makes sense to me to leave only `default_dram_type ` and
hotplug_init() in memory_tier_init(), postponing all
set_node_memory_tier()s to memory_tier_late_init()

Would it be possible there is no device_initcall() calling
memory_tier_late_init()? If yes, I think putting memory_tier_late_init()
in late_init() is still necessary.

> --
> Best Regards,
> Huang, Ying
>
> > Doing this all memory-type drivers have to call late_initcall() to
> > register a memory tier. I’m not sure how many they are?
> >
> > What do you guys think?
> >
> >>
> >> Jonathan
> >>
> >>
> >> > ---
> >> >  mm/memory-tiers.c | 94 +++++++++++++++++++++++++++++++++++------------
> >> >  1 file changed, 70 insertions(+), 24 deletions(-)
> >> >
> >> > diff --git a/mm/memory-tiers.c b/mm/memory-tiers.c
> >> > index 516b144fd45a..6632102bd5c9 100644
> >> > --- a/mm/memory-tiers.c
> >> > +++ b/mm/memory-tiers.c
> >>
> >>
> >>
> >> > @@ -855,7 +892,8 @@ static int __init memory_tier_init(void)
> >> >        * For now we can have 4 faster memory tiers with smaller adistance
> >> >        * than default DRAM tier.
> >> >        */
> >> > -     default_dram_type = alloc_memory_type(MEMTIER_ADISTANCE_DRAM);
> >> > +     default_dram_type = mt_find_alloc_memory_type(MEMTIER_ADISTANCE_DRAM,
> >> > +                                                   &default_memory_types);
> >> >       if (IS_ERR(default_dram_type))
> >> >               panic("%s() failed to allocate default DRAM tier\n", __func__);
> >> >
> >> > @@ -865,6 +903,14 @@ static int __init memory_tier_init(void)
> >> >        * types assigned.
> >> >        */
> >> >       for_each_node_state(node, N_MEMORY) {
> >> > +             if (!node_state(node, N_CPU))
> >> > +                     /*
> >> > +                      * Defer memory tier initialization on
> >> > +                      * CPUless numa nodes. These will be initialized
> >> > +                      * after firmware and devices are initialized.
> >>
> >> Could the comment also say why we can't defer them all?
> >>
> >> (In an odd coincidence we have a similar issue for some CPU hotplug
> >>  related bring up where review feedback was move all cases later).
> >>
> >> > +                      */
> >> > +                     continue;
> >> > +
> >> >               memtier = set_node_memory_tier(node);
> >> >               if (IS_ERR(memtier))
> >> >                       /*
> >>

Jonathan Cameron April 10, 2024, 4:51 p.m. UTC | #7

On Tue, 9 Apr 2024 12:02:31 -0700
"Ho-Ren (Jack) Chuang" <[e-mail]> wrote:

> Hi Jonathan,
> 
> On Tue, Apr 9, 2024 at 9:12 AM Jonathan Cameron
> <[e-mail]> wrote:
> >
> > On Fri, 5 Apr 2024 15:43:47 -0700
> > "Ho-Ren (Jack) Chuang" <[e-mail]> wrote:
> >  
> > > On Fri, Apr 5, 2024 at 7:03 AM Jonathan Cameron
> > > <[e-mail]> wrote:  
> > > >
> > > > On Fri,  5 Apr 2024 00:07:06 +0000
> > > > "Ho-Ren (Jack) Chuang" <[e-mail]> wrote:
> > > >  
> > > > > The current implementation treats emulated memory devices, such as
> > > > > CXL1.1 type3 memory, as normal DRAM when they are emulated as normal memory
> > > > > (E820_TYPE_RAM). However, these emulated devices have different
> > > > > characteristics than traditional DRAM, making it important to
> > > > > distinguish them. Thus, we modify the tiered memory initialization process
> > > > > to introduce a delay specifically for CPUless NUMA nodes. This delay
> > > > > ensures that the memory tier initialization for these nodes is deferred
> > > > > until HMAT information is obtained during the boot process. Finally,
> > > > > demotion tables are recalculated at the end.
> > > > >
> > > > > * late_initcall(memory_tier_late_init);
> > > > > Some device drivers may have initialized memory tiers between
> > > > > `memory_tier_init()` and `memory_tier_late_init()`, potentially bringing
> > > > > online memory nodes and configuring memory tiers. They should be excluded
> > > > > in the late init.
> > > > >
> > > > > * Handle cases where there is no HMAT when creating memory tiers
> > > > > There is a scenario where a CPUless node does not provide HMAT information.
> > > > > If no HMAT is specified, it falls back to using the default DRAM tier.
> > > > >
> > > > > * Introduce another new lock `default_dram_perf_lock` for adist calculation
> > > > > In the current implementation, iterating through CPUlist nodes requires
> > > > > holding the `memory_tier_lock`. However, `mt_calc_adistance()` will end up
> > > > > trying to acquire the same lock, leading to a potential deadlock.
> > > > > Therefore, we propose introducing a standalone `default_dram_perf_lock` to
> > > > > protect `default_dram_perf_*`. This approach not only avoids deadlock
> > > > > but also prevents holding a large lock simultaneously.
> > > > >
> > > > > * Upgrade `set_node_memory_tier` to support additional cases, including
> > > > >   default DRAM, late CPUless, and hot-plugged initializations.
> > > > > To cover hot-plugged memory nodes, `mt_calc_adistance()` and
> > > > > `mt_find_alloc_memory_type()` are moved into `set_node_memory_tier()` to
> > > > > handle cases where memtype is not initialized and where HMAT information is
> > > > > available.
> > > > >
> > > > > * Introduce `default_memory_types` for those memory types that are not
> > > > >   initialized by device drivers.
> > > > > Because late initialized memory and default DRAM memory need to be managed,
> > > > > a default memory type is created for storing all memory types that are
> > > > > not initialized by device drivers and as a fallback.
> > > > >
> > > > > Signed-off-by: Ho-Ren (Jack) Chuang <[e-mail]>
> > > > > Signed-off-by: Hao Xiang <[e-mail]>
> > > > > Reviewed-by: "Huang, Ying" <[e-mail]>  
> > > >
> > > > Hi - one remaining question. Why can't we delay init for all nodes
> > > > to either drivers or your fallback late_initcall code.
> > > > It would be nice to reduce possible code paths.  
> > >
> > > I try not to change too much of the existing code structure in
> > > this patchset.
> > >
> > > To me, postponing/moving all memory tier registrations to
> > > late_initcall() is another possible action item for the next patchset.
> > >
> > > After tier_mem(), hmat_init() is called, which requires registering
> > > `default_dram_type` info. This is when `default_dram_type` is needed.
> > > However, it is indeed possible to postpone the latter part,
> > > set_node_memory_tier(), to `late_init(). So, memory_tier_init() can
> > > indeed be split into two parts, and the latter part can be moved to
> > > late_initcall() to be processed together.
> > >
> > > Doing this all memory-type drivers have to call late_initcall() to
> > > register a memory tier. I’m not sure how many they are?
> > >
> > > What do you guys think?  
> >
> > Gut feeling - if you are going to move it for some cases, move it for
> > all of them.  Then we only have to test once ;)
> >
> > J  
> 
> Thank you for your reminder! I agree~ That's why I'm considering
> changing them in the next patchset because of the amount of changes.
> And also, this patchset already contains too many things.

Makes sense.  (Interestingly we are reaching the same conclusion
for the thread that motivated suggesting bringing them all together
in the first place!)

Get things work in a clean fashion, then consider moving everything to
happen at the same time to simplify testing etc.

Jonathan

Ho-Ren (Jack) Chuang April 17, 2024, 8:53 a.m. UTC | #8

On Wed, Apr 10, 2024 at 9:51 AM Jonathan Cameron
<[e-mail]> wrote:
>
> On Tue, 9 Apr 2024 12:02:31 -0700
> "Ho-Ren (Jack) Chuang" <[e-mail]> wrote:
>
> > Hi Jonathan,
> >
> > On Tue, Apr 9, 2024 at 9:12 AM Jonathan Cameron
> > <[e-mail]> wrote:
> > >
> > > On Fri, 5 Apr 2024 15:43:47 -0700
> > > "Ho-Ren (Jack) Chuang" <[e-mail]> wrote:
> > >
> > > > On Fri, Apr 5, 2024 at 7:03 AM Jonathan Cameron
> > > > <[e-mail]> wrote:
> > > > >
> > > > > On Fri,  5 Apr 2024 00:07:06 +0000
> > > > > "Ho-Ren (Jack) Chuang" <[e-mail]> wrote:
> > > > >
> > > > > > The current implementation treats emulated memory devices, such as
> > > > > > CXL1.1 type3 memory, as normal DRAM when they are emulated as normal memory
> > > > > > (E820_TYPE_RAM). However, these emulated devices have different
> > > > > > characteristics than traditional DRAM, making it important to
> > > > > > distinguish them. Thus, we modify the tiered memory initialization process
> > > > > > to introduce a delay specifically for CPUless NUMA nodes. This delay
> > > > > > ensures that the memory tier initialization for these nodes is deferred
> > > > > > until HMAT information is obtained during the boot process. Finally,
> > > > > > demotion tables are recalculated at the end.
> > > > > >
> > > > > > * late_initcall(memory_tier_late_init);
> > > > > > Some device drivers may have initialized memory tiers between
> > > > > > `memory_tier_init()` and `memory_tier_late_init()`, potentially bringing
> > > > > > online memory nodes and configuring memory tiers. They should be excluded
> > > > > > in the late init.
> > > > > >
> > > > > > * Handle cases where there is no HMAT when creating memory tiers
> > > > > > There is a scenario where a CPUless node does not provide HMAT information.
> > > > > > If no HMAT is specified, it falls back to using the default DRAM tier.
> > > > > >
> > > > > > * Introduce another new lock `default_dram_perf_lock` for adist calculation
> > > > > > In the current implementation, iterating through CPUlist nodes requires
> > > > > > holding the `memory_tier_lock`. However, `mt_calc_adistance()` will end up
> > > > > > trying to acquire the same lock, leading to a potential deadlock.
> > > > > > Therefore, we propose introducing a standalone `default_dram_perf_lock` to
> > > > > > protect `default_dram_perf_*`. This approach not only avoids deadlock
> > > > > > but also prevents holding a large lock simultaneously.
> > > > > >
> > > > > > * Upgrade `set_node_memory_tier` to support additional cases, including
> > > > > >   default DRAM, late CPUless, and hot-plugged initializations.
> > > > > > To cover hot-plugged memory nodes, `mt_calc_adistance()` and
> > > > > > `mt_find_alloc_memory_type()` are moved into `set_node_memory_tier()` to
> > > > > > handle cases where memtype is not initialized and where HMAT information is
> > > > > > available.
> > > > > >
> > > > > > * Introduce `default_memory_types` for those memory types that are not
> > > > > >   initialized by device drivers.
> > > > > > Because late initialized memory and default DRAM memory need to be managed,
> > > > > > a default memory type is created for storing all memory types that are
> > > > > > not initialized by device drivers and as a fallback.
> > > > > >
> > > > > > Signed-off-by: Ho-Ren (Jack) Chuang <[e-mail]>
> > > > > > Signed-off-by: Hao Xiang <[e-mail]>
> > > > > > Reviewed-by: "Huang, Ying" <[e-mail]>
> > > > >
> > > > > Hi - one remaining question. Why can't we delay init for all nodes
> > > > > to either drivers or your fallback late_initcall code.
> > > > > It would be nice to reduce possible code paths.
> > > >
> > > > I try not to change too much of the existing code structure in
> > > > this patchset.
> > > >
> > > > To me, postponing/moving all memory tier registrations to
> > > > late_initcall() is another possible action item for the next patchset.
> > > >
> > > > After tier_mem(), hmat_init() is called, which requires registering
> > > > `default_dram_type` info. This is when `default_dram_type` is needed.
> > > > However, it is indeed possible to postpone the latter part,
> > > > set_node_memory_tier(), to `late_init(). So, memory_tier_init() can
> > > > indeed be split into two parts, and the latter part can be moved to
> > > > late_initcall() to be processed together.
> > > >
> > > > Doing this all memory-type drivers have to call late_initcall() to
> > > > register a memory tier. I’m not sure how many they are?
> > > >
> > > > What do you guys think?
> > >
> > > Gut feeling - if you are going to move it for some cases, move it for
> > > all of them.  Then we only have to test once ;)
> > >
> > > J
> >
> > Thank you for your reminder! I agree~ That's why I'm considering
> > changing them in the next patchset because of the amount of changes.
> > And also, this patchset already contains too many things.
>
> Makes sense.  (Interestingly we are reaching the same conclusion
> for the thread that motivated suggesting bringing them all together
> in the first place!)
>
> Get things work in a clean fashion, then consider moving everything to
> happen at the same time to simplify testing etc.
Hi Jonathan,

Thank you and I will do! Could you please take another look and see if
there are any further changes needed for this patchset? If everything
looks good to you, could you please also provide a 'Reviewed-by' for
this patch?

Per discussion, I'm going to prepare another patchset "memory tier
initialization path optimization" and will send it out once ready.

>
> Jonathan

Jonathan Cameron April 19, 2024, 2:01 p.m. UTC | #9

On Fri,  5 Apr 2024 00:07:06 +0000
"Ho-Ren (Jack) Chuang" <[e-mail]> wrote:

> The current implementation treats emulated memory devices, such as
> CXL1.1 type3 memory, as normal DRAM when they are emulated as normal memory
> (E820_TYPE_RAM). However, these emulated devices have different
> characteristics than traditional DRAM, making it important to
> distinguish them. Thus, we modify the tiered memory initialization process
> to introduce a delay specifically for CPUless NUMA nodes. This delay
> ensures that the memory tier initialization for these nodes is deferred
> until HMAT information is obtained during the boot process. Finally,
> demotion tables are recalculated at the end.
> 
> * late_initcall(memory_tier_late_init);
> Some device drivers may have initialized memory tiers between
> `memory_tier_init()` and `memory_tier_late_init()`, potentially bringing
> online memory nodes and configuring memory tiers. They should be excluded
> in the late init.
> 
> * Handle cases where there is no HMAT when creating memory tiers
> There is a scenario where a CPUless node does not provide HMAT information.
> If no HMAT is specified, it falls back to using the default DRAM tier.
> 
> * Introduce another new lock `default_dram_perf_lock` for adist calculation
> In the current implementation, iterating through CPUlist nodes requires
> holding the `memory_tier_lock`. However, `mt_calc_adistance()` will end up
> trying to acquire the same lock, leading to a potential deadlock.
> Therefore, we propose introducing a standalone `default_dram_perf_lock` to
> protect `default_dram_perf_*`. This approach not only avoids deadlock
> but also prevents holding a large lock simultaneously.
> 
> * Upgrade `set_node_memory_tier` to support additional cases, including
>   default DRAM, late CPUless, and hot-plugged initializations.
> To cover hot-plugged memory nodes, `mt_calc_adistance()` and
> `mt_find_alloc_memory_type()` are moved into `set_node_memory_tier()` to
> handle cases where memtype is not initialized and where HMAT information is
> available.
> 
> * Introduce `default_memory_types` for those memory types that are not
>   initialized by device drivers.
> Because late initialized memory and default DRAM memory need to be managed,
> a default memory type is created for storing all memory types that are
> not initialized by device drivers and as a fallback.
> 
> Signed-off-by: Ho-Ren (Jack) Chuang <[e-mail]>
> Signed-off-by: Hao Xiang <[e-mail]>
> Reviewed-by: "Huang, Ying" <[e-mail]>
Reviewed-by: Jonathan Cameron <[e-mail]>

Patch

diff --git a/mm/memory-tiers.c b/mm/memory-tiers.c
index 516b144fd45a..6632102bd5c9 100644
--- a/mm/memory-tiers.c
+++ b/mm/memory-tiers.c
@@ -36,6 +36,11 @@  struct node_memory_type_map {
 
 static DEFINE_MUTEX(memory_tier_lock);
 static LIST_HEAD(memory_tiers);
+/*
+ * The list is used to store all memory types that are not created
+ * by a device driver.
+ */
+static LIST_HEAD(default_memory_types);
 static struct node_memory_type_map node_memory_types[MAX_NUMNODES];
 struct memory_dev_type *default_dram_type;
 
@@ -108,6 +113,8 @@  static struct demotion_nodes *node_demotion __read_mostly;
 
 static BLOCKING_NOTIFIER_HEAD(mt_adistance_algorithms);
 
+/* The lock is used to protect `default_dram_perf*` info and nid. */
+static DEFINE_MUTEX(default_dram_perf_lock);
 static bool default_dram_perf_error;
 static struct access_coordinate default_dram_perf;
 static int default_dram_perf_ref_nid = NUMA_NO_NODE;
@@ -505,7 +512,8 @@  static inline void __init_node_memory_type(int node, struct memory_dev_type *mem
 static struct memory_tier *set_node_memory_tier(int node)
 {
 	struct memory_tier *memtier;
-	struct memory_dev_type *memtype;
+	struct memory_dev_type *memtype = default_dram_type;
+	int adist = MEMTIER_ADISTANCE_DRAM;
 	pg_data_t *pgdat = NODE_DATA(node);
 
 
@@ -514,7 +522,16 @@  static struct memory_tier *set_node_memory_tier(int node)
 	if (!node_state(node, N_MEMORY))
 		return ERR_PTR(-EINVAL);
 
-	__init_node_memory_type(node, default_dram_type);
+	mt_calc_adistance(node, &adist);
+	if (!node_memory_types[node].memtype) {
+		memtype = mt_find_alloc_memory_type(adist, &default_memory_types);
+		if (IS_ERR(memtype)) {
+			memtype = default_dram_type;
+			pr_info("Failed to allocate a memory type. Fall back.\n");
+		}
+	}
+
+	__init_node_memory_type(node, memtype);
 
 	memtype = node_memory_types[node].memtype;
 	node_set(node, memtype->nodes);
@@ -652,6 +669,35 @@  void mt_put_memory_types(struct list_head *memory_types)
 }
 EXPORT_SYMBOL_GPL(mt_put_memory_types);
 
+/*
+ * This is invoked via `late_initcall()` to initialize memory tiers for
+ * CPU-less memory nodes after driver initialization, which is
+ * expected to provide `adistance` algorithms.
+ */
+static int __init memory_tier_late_init(void)
+{
+	int nid;
+
+	guard(mutex)(&memory_tier_lock);
+	for_each_node_state(nid, N_MEMORY) {
+		/*
+		 * Some device drivers may have initialized memory tiers
+		 * between `memory_tier_init()` and `memory_tier_late_init()`,
+		 * potentially bringing online memory nodes and
+		 * configuring memory tiers. Exclude them here.
+		 */
+		if (node_memory_types[nid].memtype)
+			continue;
+
+		set_node_memory_tier(nid);
+	}
+
+	establish_demotion_targets();
+
+	return 0;
+}
+late_initcall(memory_tier_late_init);
+
 static void dump_hmem_attrs(struct access_coordinate *coord, const char *prefix)
 {
 	pr_info(
@@ -663,25 +709,19 @@  static void dump_hmem_attrs(struct access_coordinate *coord, const char *prefix)
 int mt_set_default_dram_perf(int nid, struct access_coordinate *perf,
 			     const char *source)
 {
-	int rc = 0;
-
-	mutex_lock(&memory_tier_lock);
-	if (default_dram_perf_error) {
-		rc = -EIO;
-		goto out;
-	}
+	guard(mutex)(&default_dram_perf_lock);
+	if (default_dram_perf_error)
+		return -EIO;
 
 	if (perf->read_latency + perf->write_latency == 0 ||
-	    perf->read_bandwidth + perf->write_bandwidth == 0) {
-		rc = -EINVAL;
-		goto out;
-	}
+	    perf->read_bandwidth + perf->write_bandwidth == 0)
+		return -EINVAL;
 
 	if (default_dram_perf_ref_nid == NUMA_NO_NODE) {
 		default_dram_perf = *perf;
 		default_dram_perf_ref_nid = nid;
 		default_dram_perf_ref_source = kstrdup(source, GFP_KERNEL);
-		goto out;
+		return 0;
 	}
 
 	/*
@@ -709,27 +749,25 @@  int mt_set_default_dram_perf(int nid, struct access_coordinate *perf,
 		pr_info(
 "  disable default DRAM node performance based abstract distance algorithm.\n");
 		default_dram_perf_error = true;
-		rc = -EINVAL;
+		return -EINVAL;
 	}
 
-out:
-	mutex_unlock(&memory_tier_lock);
-	return rc;
+	return 0;
 }
 
 int mt_perf_to_adistance(struct access_coordinate *perf, int *adist)
 {
+	guard(mutex)(&default_dram_perf_lock);
 	if (default_dram_perf_error)
 		return -EIO;
 
-	if (default_dram_perf_ref_nid == NUMA_NO_NODE)
-		return -ENOENT;
-
 	if (perf->read_latency + perf->write_latency == 0 ||
 	    perf->read_bandwidth + perf->write_bandwidth == 0)
 		return -EINVAL;
 
-	mutex_lock(&memory_tier_lock);
+	if (default_dram_perf_ref_nid == NUMA_NO_NODE)
+		return -ENOENT;
+
 	/*
 	 * The abstract distance of a memory node is in direct proportion to
 	 * its memory latency (read + write) and inversely proportional to its
@@ -742,7 +780,6 @@  int mt_perf_to_adistance(struct access_coordinate *perf, int *adist)
 		(default_dram_perf.read_latency + default_dram_perf.write_latency) *
 		(default_dram_perf.read_bandwidth + default_dram_perf.write_bandwidth) /
 		(perf->read_bandwidth + perf->write_bandwidth);
-	mutex_unlock(&memory_tier_lock);
 
 	return 0;
 }
@@ -855,7 +892,8 @@  static int __init memory_tier_init(void)
 	 * For now we can have 4 faster memory tiers with smaller adistance
 	 * than default DRAM tier.
 	 */
-	default_dram_type = alloc_memory_type(MEMTIER_ADISTANCE_DRAM);
+	default_dram_type = mt_find_alloc_memory_type(MEMTIER_ADISTANCE_DRAM,
+						      &default_memory_types);
 	if (IS_ERR(default_dram_type))
 		panic("%s() failed to allocate default DRAM tier\n", __func__);
 
@@ -865,6 +903,14 @@  static int __init memory_tier_init(void)
 	 * types assigned.
 	 */
 	for_each_node_state(node, N_MEMORY) {
+		if (!node_state(node, N_CPU))
+			/*
+			 * Defer memory tier initialization on
+			 * CPUless numa nodes. These will be initialized
+			 * after firmware and devices are initialized.
+			 */
+			continue;
+
 		memtier = set_node_memory_tier(node);
 		if (IS_ERR(memtier))
 			/*