[PATCH 1/2] admin-guide/memory-hotplug.rst: remove locking internal part from admin-guide

[PATCH 1/2] admin-guide/memory-hotplug.rst: remove locking internal part from admin-guide

[Wei Yang]

Locking Internal section exists in core-api documentation, which is more
suitable for this.

This patch removes the duplication part here.

Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
---
 Documentation/admin-guide/mm/memory-hotplug.rst | 40 -------------------------
 1 file changed, 40 deletions(-)

diff --git a/Documentation/admin-guide/mm/memory-hotplug.rst b/Documentation/admin-guide/mm/memory-hotplug.rst
index 5c4432c96c4b..241f4ce1e387 100644
--- a/Documentation/admin-guide/mm/memory-hotplug.rst
+++ b/Documentation/admin-guide/mm/memory-hotplug.rst
@@ -392,46 +392,6 @@ Need more implementation yet....
  - Notification completion of remove works by OS to firmware.
  - Guard from remove if not yet.
 
-
-Locking Internals
-=================
-
-When adding/removing memory that uses memory block devices (i.e. ordinary RAM),
-the device_hotplug_lock should be held to:
-
-- synchronize against online/offline requests (e.g. via sysfs). This way, memory
-  block devices can only be accessed (.online/.state attributes) by user
-  space once memory has been fully added. And when removing memory, we
-  know nobody is in critical sections.
-- synchronize against CPU hotplug and similar (e.g. relevant for ACPI and PPC)
-
-Especially, there is a possible lock inversion that is avoided using
-device_hotplug_lock when adding memory and user space tries to online that
-memory faster than expected:
-
-- device_online() will first take the device_lock(), followed by
-  mem_hotplug_lock
-- add_memory_resource() will first take the mem_hotplug_lock, followed by
-  the device_lock() (while creating the devices, during bus_add_device()).
-
-As the device is visible to user space before taking the device_lock(), this
-can result in a lock inversion.
-
-onlining/offlining of memory should be done via device_online()/
-device_offline() - to make sure it is properly synchronized to actions
-via sysfs. Holding device_hotplug_lock is advised (to e.g. protect online_type)
-
-When adding/removing/onlining/offlining memory or adding/removing
-heterogeneous/device memory, we should always hold the mem_hotplug_lock in
-write mode to serialise memory hotplug (e.g. access to global/zone
-variables).
-
-In addition, mem_hotplug_lock (in contrast to device_hotplug_lock) in read
-mode allows for a quite efficient get_online_mems/put_online_mems
-implementation, so code accessing memory can protect from that memory
-vanishing.
-
-
 Future Work
 ===========
 
-- 
2.15.1

[PATCH 2/2] core-api/memory-hotplug.rst: divide Locking Internal section by different locks

[Wei Yang]

Currently locking for memory hotplug is a little complicated.

Generally speaking, we leverage the two global lock:

  * device_hotplug_lock
  * mem_hotplug_lock

to serialise the process.

While for the long term, we are willing to have more fine-grained lock
to provide higher scalability.

This patch divides Locking Internal section based on these two global
locks to help readers to understand it. Also it adds some new finding to
enrich it.

[David: words arrangement]

Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
---
 Documentation/core-api/memory-hotplug.rst | 27 ++++++++++++++++++++++++---
 1 file changed, 24 insertions(+), 3 deletions(-)

diff --git a/Documentation/core-api/memory-hotplug.rst b/Documentation/core-api/memory-hotplug.rst
index de7467e48067..95662b283328 100644
--- a/Documentation/core-api/memory-hotplug.rst
+++ b/Documentation/core-api/memory-hotplug.rst
@@ -89,6 +89,20 @@ NOTIFY_STOP stops further processing of the notification queue.
 Locking Internals
 =================
 
+There are three locks involved in memory-hotplug, two global lock and one local
+lock:
+
+- device_hotplug_lock
+- mem_hotplug_lock
+- device_lock
+
+Currently, they are twisted together for all kinds of reasons. The following
+part is divided into device_hotplug_lock and mem_hotplug_lock parts
+respectively to describe those tricky situations.
+
+device_hotplug_lock
+---------------------
+
 When adding/removing memory that uses memory block devices (i.e. ordinary RAM),
 the device_hotplug_lock should be held to:
 
@@ -111,13 +125,20 @@ As the device is visible to user space before taking the device_lock(), this
 can result in a lock inversion.
 
 onlining/offlining of memory should be done via device_online()/
-device_offline() - to make sure it is properly synchronized to actions
-via sysfs. Holding device_hotplug_lock is advised (to e.g. protect online_type)
+device_offline() - to make sure it is properly synchronized to actions via
+sysfs. Even mem_hotplug_lock is used to protect the process, because of the
+lock inversion described above, holding device_hotplug_lock is still advised
+(to e.g. protect online_type)
+
+mem_hotplug_lock
+---------------------
 
 When adding/removing/onlining/offlining memory or adding/removing
 heterogeneous/device memory, we should always hold the mem_hotplug_lock in
 write mode to serialise memory hotplug (e.g. access to global/zone
-variables).
+variables). Currently, we take advantage of this to serialise sparsemem's
+mem_section handling in sparse_add_one_section() and
+sparse_remove_one_section().
 
 In addition, mem_hotplug_lock (in contrast to device_hotplug_lock) in read
 mode allows for a quite efficient get_online_mems/put_online_mems
-- 
2.15.1

Re: [PATCH 2/2] core-api/memory-hotplug.rst: divide Locking Internal section by different locks

[David Hildenbrand]

On 05.12.18 03:34, Wei Yang wrote:
> Currently locking for memory hotplug is a little complicated.
> 
> Generally speaking, we leverage the two global lock:
> 
>   * device_hotplug_lock
>   * mem_hotplug_lock
> 
> to serialise the process.
> 
> While for the long term, we are willing to have more fine-grained lock
> to provide higher scalability.
> 
> This patch divides Locking Internal section based on these two global
> locks to help readers to understand it. Also it adds some new finding to
> enrich it.
> 
> [David: words arrangement]
> 
> Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
> ---
>  Documentation/core-api/memory-hotplug.rst | 27 ++++++++++++++++++++++++---
>  1 file changed, 24 insertions(+), 3 deletions(-)
> 
> diff --git a/Documentation/core-api/memory-hotplug.rst b/Documentation/core-api/memory-hotplug.rst
> index de7467e48067..95662b283328 100644
> --- a/Documentation/core-api/memory-hotplug.rst
> +++ b/Documentation/core-api/memory-hotplug.rst
> @@ -89,6 +89,20 @@ NOTIFY_STOP stops further processing of the notification queue.
>  Locking Internals
>  =================
>  
> +There are three locks involved in memory-hotplug, two global lock and one local
> +lock:
> +
> +- device_hotplug_lock
> +- mem_hotplug_lock
> +- device_lock
> +

Do we really only ever use these three and not anything else when
adding/removing/onlining/offlining memory?

(I am thinking e.g. about pgdat_resize_lock)

If so, you should phrase that maybe more generally Or add more details :)

"In addition to fine grained locks like pgdat_resize_lock, there are
three locks involved ..."


> +Currently, they are twisted together for all kinds of reasons. The following
> +part is divided into device_hotplug_lock and mem_hotplug_lock parts
> +respectively to describe those tricky situations.
> +
> +device_hotplug_lock
> +---------------------
> +
>  When adding/removing memory that uses memory block devices (i.e. ordinary RAM),
>  the device_hotplug_lock should be held to:
>  
> @@ -111,13 +125,20 @@ As the device is visible to user space before taking the device_lock(), this
>  can result in a lock inversion.
>  
>  onlining/offlining of memory should be done via device_online()/
> -device_offline() - to make sure it is properly synchronized to actions
> -via sysfs. Holding device_hotplug_lock is advised (to e.g. protect online_type)
> +device_offline() - to make sure it is properly synchronized to actions via
> +sysfs. Even mem_hotplug_lock is used to protect the process, because of the
> +lock inversion described above, holding device_hotplug_lock is still advised
> +(to e.g. protect online_type)
> +
> +mem_hotplug_lock
> +---------------------
>  
>  When adding/removing/onlining/offlining memory or adding/removing
>  heterogeneous/device memory, we should always hold the mem_hotplug_lock in
>  write mode to serialise memory hotplug (e.g. access to global/zone
> -variables).
> +variables). Currently, we take advantage of this to serialise sparsemem's
> +mem_section handling in sparse_add_one_section() and
> +sparse_remove_one_section().
>  
>  In addition, mem_hotplug_lock (in contrast to device_hotplug_lock) in read
>  mode allows for a quite efficient get_online_mems/put_online_mems
> 


-- 

Thanks,

David / dhildenb

Re: [PATCH 2/2] core-api/memory-hotplug.rst: divide Locking Internal section by different locks

[Wei Yang]

On Wed, Dec 05, 2018 at 09:08:47AM +0100, David Hildenbrand wrote:
>On 05.12.18 03:34, Wei Yang wrote:
>> Currently locking for memory hotplug is a little complicated.
>> 
>> Generally speaking, we leverage the two global lock:
>> 
>>   * device_hotplug_lock
>>   * mem_hotplug_lock
>> 
>> to serialise the process.
>> 
>> While for the long term, we are willing to have more fine-grained lock
>> to provide higher scalability.
>> 
>> This patch divides Locking Internal section based on these two global
>> locks to help readers to understand it. Also it adds some new finding to
>> enrich it.
>> 
>> [David: words arrangement]
>> 
>> Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
>> ---
>>  Documentation/core-api/memory-hotplug.rst | 27 ++++++++++++++++++++++++---
>>  1 file changed, 24 insertions(+), 3 deletions(-)
>> 
>> diff --git a/Documentation/core-api/memory-hotplug.rst b/Documentation/core-api/memory-hotplug.rst
>> index de7467e48067..95662b283328 100644
>> --- a/Documentation/core-api/memory-hotplug.rst
>> +++ b/Documentation/core-api/memory-hotplug.rst
>> @@ -89,6 +89,20 @@ NOTIFY_STOP stops further processing of the notification queue.
>>  Locking Internals
>>  =================
>>  
>> +There are three locks involved in memory-hotplug, two global lock and one local
>> +lock:
>> +
>> +- device_hotplug_lock
>> +- mem_hotplug_lock
>> +- device_lock
>> +
>
>Do we really only ever use these three and not anything else when
>adding/removing/onlining/offlining memory?
>
>(I am thinking e.g. about pgdat_resize_lock)

Yes there are more than those three, pgdat_resize_lock is one of them.

>
>If so, you should phrase that maybe more generally Or add more details :)

Yep, while I don't get a whole picture about the pgdat_resize_lock. The
usage of this lock scatter in many places.

>
>"In addition to fine grained locks like pgdat_resize_lock, there are
>three locks involved ..."
>

Sounds better :-)

-- 
Wei Yang
Help you, Help me

Re: [PATCH 2/2] core-api/memory-hotplug.rst: divide Locking Internal section by different locks

[Mike Rapoport]

On Wed, Dec 05, 2018 at 10:34:26AM +0800, Wei Yang wrote:
> Currently locking for memory hotplug is a little complicated.
> 
> Generally speaking, we leverage the two global lock:
> 
>   * device_hotplug_lock
>   * mem_hotplug_lock
> 
> to serialise the process.
> 
> While for the long term, we are willing to have more fine-grained lock
> to provide higher scalability.
> 
> This patch divides Locking Internal section based on these two global
> locks to help readers to understand it. Also it adds some new finding to
> enrich it.
> 
> [David: words arrangement]
> 
> Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
> ---
>  Documentation/core-api/memory-hotplug.rst | 27 ++++++++++++++++++++++++---
>  1 file changed, 24 insertions(+), 3 deletions(-)
> 
> diff --git a/Documentation/core-api/memory-hotplug.rst b/Documentation/core-api/memory-hotplug.rst
> index de7467e48067..95662b283328 100644
> --- a/Documentation/core-api/memory-hotplug.rst
> +++ b/Documentation/core-api/memory-hotplug.rst
> @@ -89,6 +89,20 @@ NOTIFY_STOP stops further processing of the notification queue.
>  Locking Internals
>  =================
> 
> +There are three locks involved in memory-hotplug, two global lock and one local

typo:                                                          ^locks

> +lock:
> +
> +- device_hotplug_lock
> +- mem_hotplug_lock
> +- device_lock
> +
> +Currently, they are twisted together for all kinds of reasons. The following
> +part is divided into device_hotplug_lock and mem_hotplug_lock parts
> +respectively to describe those tricky situations.
> +
> +device_hotplug_lock
> +---------------------
> +
>  When adding/removing memory that uses memory block devices (i.e. ordinary RAM),
>  the device_hotplug_lock should be held to:
> 
> @@ -111,13 +125,20 @@ As the device is visible to user space before taking the device_lock(), this
>  can result in a lock inversion.
> 
>  onlining/offlining of memory should be done via device_online()/
> -device_offline() - to make sure it is properly synchronized to actions
> -via sysfs. Holding device_hotplug_lock is advised (to e.g. protect online_type)
> +device_offline() - to make sure it is properly synchronized to actions via
> +sysfs. Even mem_hotplug_lock is used to protect the process, because of the

I think it should be "Even if mem_hotplug_lock ..."

> +lock inversion described above, holding device_hotplug_lock is still advised
> +(to e.g. protect online_type)
> +
> +mem_hotplug_lock
> +---------------------
> 
>  When adding/removing/onlining/offlining memory or adding/removing
>  heterogeneous/device memory, we should always hold the mem_hotplug_lock in
>  write mode to serialise memory hotplug (e.g. access to global/zone
> -variables).
> +variables). Currently, we take advantage of this to serialise sparsemem's
> +mem_section handling in sparse_add_one_section() and
> +sparse_remove_one_section().
> 
>  In addition, mem_hotplug_lock (in contrast to device_hotplug_lock) in read
>  mode allows for a quite efficient get_online_mems/put_online_mems
> -- 
> 2.15.1
> 

-- 
Sincerely yours,
Mike.

Re: [PATCH 2/2] core-api/memory-hotplug.rst: divide Locking Internal section by different locks

[Wei Yang]

On Wed, Dec 05, 2018 at 10:40:45AM +0200, Mike Rapoport wrote:
>On Wed, Dec 05, 2018 at 10:34:26AM +0800, Wei Yang wrote:
>> Currently locking for memory hotplug is a little complicated.
>> 
>> Generally speaking, we leverage the two global lock:
>> 
>>   * device_hotplug_lock
>>   * mem_hotplug_lock
>> 
>> to serialise the process.
>> 
>> While for the long term, we are willing to have more fine-grained lock
>> to provide higher scalability.
>> 
>> This patch divides Locking Internal section based on these two global
>> locks to help readers to understand it. Also it adds some new finding to
>> enrich it.
>> 
>> [David: words arrangement]
>> 
>> Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
>> ---
>>  Documentation/core-api/memory-hotplug.rst | 27 ++++++++++++++++++++++++---
>>  1 file changed, 24 insertions(+), 3 deletions(-)
>> 
>> diff --git a/Documentation/core-api/memory-hotplug.rst b/Documentation/core-api/memory-hotplug.rst
>> index de7467e48067..95662b283328 100644
>> --- a/Documentation/core-api/memory-hotplug.rst
>> +++ b/Documentation/core-api/memory-hotplug.rst
>> @@ -89,6 +89,20 @@ NOTIFY_STOP stops further processing of the notification queue.
>>  Locking Internals
>>  =================
>> 
>> +There are three locks involved in memory-hotplug, two global lock and one local
>
>typo:                                                          ^locks
>

Thanks :-)

>> +lock:
>> +
>> +- device_hotplug_lock
>> +- mem_hotplug_lock
>> +- device_lock
>> +
>> +Currently, they are twisted together for all kinds of reasons. The following
>> +part is divided into device_hotplug_lock and mem_hotplug_lock parts
>> +respectively to describe those tricky situations.
>> +
>> +device_hotplug_lock
>> +---------------------
>> +
>>  When adding/removing memory that uses memory block devices (i.e. ordinary RAM),
>>  the device_hotplug_lock should be held to:
>> 
>> @@ -111,13 +125,20 @@ As the device is visible to user space before taking the device_lock(), this
>>  can result in a lock inversion.
>> 
>>  onlining/offlining of memory should be done via device_online()/
>> -device_offline() - to make sure it is properly synchronized to actions
>> -via sysfs. Holding device_hotplug_lock is advised (to e.g. protect online_type)
>> +device_offline() - to make sure it is properly synchronized to actions via
>> +sysfs. Even mem_hotplug_lock is used to protect the process, because of the
>
>I think it should be "Even if mem_hotplug_lock ..."
>

Ah, my poor English, will fix it in next version. :-)

>> +lock inversion described above, holding device_hotplug_lock is still advised
>> +(to e.g. protect online_type)
>> +
>> +mem_hotplug_lock
>> +---------------------
>> 
>>  When adding/removing/onlining/offlining memory or adding/removing
>>  heterogeneous/device memory, we should always hold the mem_hotplug_lock in
>>  write mode to serialise memory hotplug (e.g. access to global/zone
>> -variables).
>> +variables). Currently, we take advantage of this to serialise sparsemem's
>> +mem_section handling in sparse_add_one_section() and
>> +sparse_remove_one_section().
>> 
>>  In addition, mem_hotplug_lock (in contrast to device_hotplug_lock) in read
>>  mode allows for a quite efficient get_online_mems/put_online_mems
>> -- 
>> 2.15.1
>> 
>
>-- 
>Sincerely yours,
>Mike.

-- 
Wei Yang
Help you, Help me

Re: [PATCH 2/2] core-api/memory-hotplug.rst: divide Locking Internal section by different locks

[Michal Hocko]

On Wed 05-12-18 10:34:26, Wei Yang wrote:
> Currently locking for memory hotplug is a little complicated.
> 
> Generally speaking, we leverage the two global lock:
> 
>   * device_hotplug_lock
>   * mem_hotplug_lock
> 
> to serialise the process.
> 
> While for the long term, we are willing to have more fine-grained lock
> to provide higher scalability.
> 
> This patch divides Locking Internal section based on these two global
> locks to help readers to understand it. Also it adds some new finding to
> enrich it.
> 
> [David: words arrangement]
> 
> Signed-off-by: Wei Yang <richard.weiyang@gmail.com>

For a love of mine I cannot find the locking description by Oscar. Maybe
it never existed and I just made it up ;) But if it is not imaginary
then my recollection is that it was much more comprehensive. If not then
even this is a good start.

> ---
>  Documentation/core-api/memory-hotplug.rst | 27 ++++++++++++++++++++++++---
>  1 file changed, 24 insertions(+), 3 deletions(-)
> 
> diff --git a/Documentation/core-api/memory-hotplug.rst b/Documentation/core-api/memory-hotplug.rst
> index de7467e48067..95662b283328 100644
> --- a/Documentation/core-api/memory-hotplug.rst
> +++ b/Documentation/core-api/memory-hotplug.rst
> @@ -89,6 +89,20 @@ NOTIFY_STOP stops further processing of the notification queue.
>  Locking Internals
>  =================
>  
> +There are three locks involved in memory-hotplug, two global lock and one local
> +lock:
> +
> +- device_hotplug_lock
> +- mem_hotplug_lock
> +- device_lock
> +
> +Currently, they are twisted together for all kinds of reasons. The following
> +part is divided into device_hotplug_lock and mem_hotplug_lock parts
> +respectively to describe those tricky situations.
> +
> +device_hotplug_lock
> +---------------------
> +
>  When adding/removing memory that uses memory block devices (i.e. ordinary RAM),
>  the device_hotplug_lock should be held to:
>  
> @@ -111,13 +125,20 @@ As the device is visible to user space before taking the device_lock(), this
>  can result in a lock inversion.
>  
>  onlining/offlining of memory should be done via device_online()/
> -device_offline() - to make sure it is properly synchronized to actions
> -via sysfs. Holding device_hotplug_lock is advised (to e.g. protect online_type)
> +device_offline() - to make sure it is properly synchronized to actions via
> +sysfs. Even mem_hotplug_lock is used to protect the process, because of the
> +lock inversion described above, holding device_hotplug_lock is still advised
> +(to e.g. protect online_type)
> +
> +mem_hotplug_lock
> +---------------------
>  
>  When adding/removing/onlining/offlining memory or adding/removing
>  heterogeneous/device memory, we should always hold the mem_hotplug_lock in
>  write mode to serialise memory hotplug (e.g. access to global/zone
> -variables).
> +variables). Currently, we take advantage of this to serialise sparsemem's
> +mem_section handling in sparse_add_one_section() and
> +sparse_remove_one_section().
>  
>  In addition, mem_hotplug_lock (in contrast to device_hotplug_lock) in read
>  mode allows for a quite efficient get_online_mems/put_online_mems
> -- 
> 2.15.1
> 

-- 
Michal Hocko
SUSE Labs

Re: [PATCH 2/2] core-api/memory-hotplug.rst: divide Locking Internal section by different locks

[Wei Yang]

On Wed, Dec 05, 2018 at 01:13:10PM +0100, Michal Hocko wrote:
>On Wed 05-12-18 10:34:26, Wei Yang wrote:
>> Currently locking for memory hotplug is a little complicated.
>> 
>> Generally speaking, we leverage the two global lock:
>> 
>>   * device_hotplug_lock
>>   * mem_hotplug_lock
>> 
>> to serialise the process.
>> 
>> While for the long term, we are willing to have more fine-grained lock
>> to provide higher scalability.
>> 
>> This patch divides Locking Internal section based on these two global
>> locks to help readers to understand it. Also it adds some new finding to
>> enrich it.
>> 
>> [David: words arrangement]
>> 
>> Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
>
>For a love of mine I cannot find the locking description by Oscar. Maybe
>it never existed and I just made it up ;) But if it is not imaginary
>then my recollection is that it was much more comprehensive. If not then
>even this is a good start.

Thanks.

If Oscar has already has some work on it, this could be a complement to
his work :-)

>
>> ---
>>  Documentation/core-api/memory-hotplug.rst | 27 ++++++++++++++++++++++++---
>>  1 file changed, 24 insertions(+), 3 deletions(-)
>> 
>> diff --git a/Documentation/core-api/memory-hotplug.rst b/Documentation/core-api/memory-hotplug.rst
>> index de7467e48067..95662b283328 100644
>> --- a/Documentation/core-api/memory-hotplug.rst
>> +++ b/Documentation/core-api/memory-hotplug.rst
>> @@ -89,6 +89,20 @@ NOTIFY_STOP stops further processing of the notification queue.
>>  Locking Internals
>>  =================
>>  
>> +There are three locks involved in memory-hotplug, two global lock and one local
>> +lock:
>> +
>> +- device_hotplug_lock
>> +- mem_hotplug_lock
>> +- device_lock
>> +
>> +Currently, they are twisted together for all kinds of reasons. The following
>> +part is divided into device_hotplug_lock and mem_hotplug_lock parts
>> +respectively to describe those tricky situations.
>> +
>> +device_hotplug_lock
>> +---------------------
>> +
>>  When adding/removing memory that uses memory block devices (i.e. ordinary RAM),
>>  the device_hotplug_lock should be held to:
>>  
>> @@ -111,13 +125,20 @@ As the device is visible to user space before taking the device_lock(), this
>>  can result in a lock inversion.
>>  
>>  onlining/offlining of memory should be done via device_online()/
>> -device_offline() - to make sure it is properly synchronized to actions
>> -via sysfs. Holding device_hotplug_lock is advised (to e.g. protect online_type)
>> +device_offline() - to make sure it is properly synchronized to actions via
>> +sysfs. Even mem_hotplug_lock is used to protect the process, because of the
>> +lock inversion described above, holding device_hotplug_lock is still advised
>> +(to e.g. protect online_type)
>> +
>> +mem_hotplug_lock
>> +---------------------
>>  
>>  When adding/removing/onlining/offlining memory or adding/removing
>>  heterogeneous/device memory, we should always hold the mem_hotplug_lock in
>>  write mode to serialise memory hotplug (e.g. access to global/zone
>> -variables).
>> +variables). Currently, we take advantage of this to serialise sparsemem's
>> +mem_section handling in sparse_add_one_section() and
>> +sparse_remove_one_section().
>>  
>>  In addition, mem_hotplug_lock (in contrast to device_hotplug_lock) in read
>>  mode allows for a quite efficient get_online_mems/put_online_mems
>> -- 
>> 2.15.1
>> 
>
>-- 
>Michal Hocko
>SUSE Labs

-- 
Wei Yang
Help you, Help me

Re: [PATCH 1/2] admin-guide/memory-hotplug.rst: remove locking internal part from admin-guide

[David Hildenbrand]

On 05.12.18 03:34, Wei Yang wrote:
> Locking Internal section exists in core-api documentation, which is more
> suitable for this.
> 
> This patch removes the duplication part here.
> 
> Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
> ---
>  Documentation/admin-guide/mm/memory-hotplug.rst | 40 -------------------------
>  1 file changed, 40 deletions(-)
> 
> diff --git a/Documentation/admin-guide/mm/memory-hotplug.rst b/Documentation/admin-guide/mm/memory-hotplug.rst
> index 5c4432c96c4b..241f4ce1e387 100644
> --- a/Documentation/admin-guide/mm/memory-hotplug.rst
> +++ b/Documentation/admin-guide/mm/memory-hotplug.rst
> @@ -392,46 +392,6 @@ Need more implementation yet....
>   - Notification completion of remove works by OS to firmware.
>   - Guard from remove if not yet.
>  
> -
> -Locking Internals
> -=================
> -
> -When adding/removing memory that uses memory block devices (i.e. ordinary RAM),
> -the device_hotplug_lock should be held to:
> -
> -- synchronize against online/offline requests (e.g. via sysfs). This way, memory
> -  block devices can only be accessed (.online/.state attributes) by user
> -  space once memory has been fully added. And when removing memory, we
> -  know nobody is in critical sections.
> -- synchronize against CPU hotplug and similar (e.g. relevant for ACPI and PPC)
> -
> -Especially, there is a possible lock inversion that is avoided using
> -device_hotplug_lock when adding memory and user space tries to online that
> -memory faster than expected:
> -
> -- device_online() will first take the device_lock(), followed by
> -  mem_hotplug_lock
> -- add_memory_resource() will first take the mem_hotplug_lock, followed by
> -  the device_lock() (while creating the devices, during bus_add_device()).
> -
> -As the device is visible to user space before taking the device_lock(), this
> -can result in a lock inversion.
> -
> -onlining/offlining of memory should be done via device_online()/
> -device_offline() - to make sure it is properly synchronized to actions
> -via sysfs. Holding device_hotplug_lock is advised (to e.g. protect online_type)
> -
> -When adding/removing/onlining/offlining memory or adding/removing
> -heterogeneous/device memory, we should always hold the mem_hotplug_lock in
> -write mode to serialise memory hotplug (e.g. access to global/zone
> -variables).
> -
> -In addition, mem_hotplug_lock (in contrast to device_hotplug_lock) in read
> -mode allows for a quite efficient get_online_mems/put_online_mems
> -implementation, so code accessing memory can protect from that memory
> -vanishing.
> -
> -
>  Future Work
>  ===========
>  
> 

I reported this yesterday to Jonathan and Mike

https://lkml.org/lkml/2018/12/3/340


Anyhow

Reviewed-by: David Hildenbrand <david@redhat.com>

-- 

Thanks,

David / dhildenb

Re: [PATCH 1/2] admin-guide/memory-hotplug.rst: remove locking internal part from admin-guide

[Mike Rapoport]

On Wed, Dec 05, 2018 at 09:03:24AM +0100, David Hildenbrand wrote:
> On 05.12.18 03:34, Wei Yang wrote:
> > Locking Internal section exists in core-api documentation, which is more
> > suitable for this.
> > 
> > This patch removes the duplication part here.
> > 
> > Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
> > ---
> >  Documentation/admin-guide/mm/memory-hotplug.rst | 40 -------------------------
> >  1 file changed, 40 deletions(-)
> > 
> > diff --git a/Documentation/admin-guide/mm/memory-hotplug.rst b/Documentation/admin-guide/mm/memory-hotplug.rst
> > index 5c4432c96c4b..241f4ce1e387 100644
> > --- a/Documentation/admin-guide/mm/memory-hotplug.rst
> > +++ b/Documentation/admin-guide/mm/memory-hotplug.rst
> > @@ -392,46 +392,6 @@ Need more implementation yet....
> >   - Notification completion of remove works by OS to firmware.
> >   - Guard from remove if not yet.

[ ... ]

> >  Future Work
> >  ===========
> >  
> > 
> 
> I reported this yesterday to Jonathan and Mike
> 
> https://lkml.org/lkml/2018/12/3/340

Somehow I've missed it...
 
> Anyhow
> 
> Reviewed-by: David Hildenbrand <david@redhat.com>

Acked-by: Mike Rapoport <rppt@linux.ibm.com>

> 
> -- 
> 
> Thanks,
> 
> David / dhildenb
> 

-- 
Sincerely yours,
Mike.

Re: [PATCH 1/2] admin-guide/memory-hotplug.rst: remove locking internal part from admin-guide

[Wei Yang]

On Wed, Dec 05, 2018 at 10:30:13AM +0200, Mike Rapoport wrote:
>On Wed, Dec 05, 2018 at 09:03:24AM +0100, David Hildenbrand wrote:
>> On 05.12.18 03:34, Wei Yang wrote:
>> > Locking Internal section exists in core-api documentation, which is more
>> > suitable for this.
>> > 
>> > This patch removes the duplication part here.
>> > 
>> > Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
>> > ---
>> >  Documentation/admin-guide/mm/memory-hotplug.rst | 40 -------------------------
>> >  1 file changed, 40 deletions(-)
>> > 
>> > diff --git a/Documentation/admin-guide/mm/memory-hotplug.rst b/Documentation/admin-guide/mm/memory-hotplug.rst
>> > index 5c4432c96c4b..241f4ce1e387 100644
>> > --- a/Documentation/admin-guide/mm/memory-hotplug.rst
>> > +++ b/Documentation/admin-guide/mm/memory-hotplug.rst
>> > @@ -392,46 +392,6 @@ Need more implementation yet....
>> >   - Notification completion of remove works by OS to firmware.
>> >   - Guard from remove if not yet.
>
>[ ... ]
>
>> >  Future Work
>> >  ===========
>> >  
>> > 
>> 
>> I reported this yesterday to Jonathan and Mike
>> 
>> https://lkml.org/lkml/2018/12/3/340
>
>Somehow I've missed it...
> 
>> Anyhow
>> 
>> Reviewed-by: David Hildenbrand <david@redhat.com>
>
>Acked-by: Mike Rapoport <rppt@linux.ibm.com>
>

Thanks :-)

>> 
>> -- 
>> 
>> Thanks,
>> 
>> David / dhildenb
>> 
>
>-- 
>Sincerely yours,
>Mike.

-- 
Wei Yang
Help you, Help me

Re: [PATCH 1/2] admin-guide/memory-hotplug.rst: remove locking internal part from admin-guide

[Wei Yang]

On Wed, Dec 05, 2018 at 09:03:24AM +0100, David Hildenbrand wrote:
>On 05.12.18 03:34, Wei Yang wrote:
>> Locking Internal section exists in core-api documentation, which is more
>> suitable for this.
>> 
>> This patch removes the duplication part here.
>> 
>> Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
>> ---
>>  Documentation/admin-guide/mm/memory-hotplug.rst | 40 -------------------------
>>  1 file changed, 40 deletions(-)
>> 
>> diff --git a/Documentation/admin-guide/mm/memory-hotplug.rst b/Documentation/admin-guide/mm/memory-hotplug.rst
>> index 5c4432c96c4b..241f4ce1e387 100644
>> --- a/Documentation/admin-guide/mm/memory-hotplug.rst
>> +++ b/Documentation/admin-guide/mm/memory-hotplug.rst
>> @@ -392,46 +392,6 @@ Need more implementation yet....
>>   - Notification completion of remove works by OS to firmware.
>>   - Guard from remove if not yet.
>>  
>> -
>> -Locking Internals
>> -=================
>> -
>> -When adding/removing memory that uses memory block devices (i.e. ordinary RAM),
>> -the device_hotplug_lock should be held to:
>> -
>> -- synchronize against online/offline requests (e.g. via sysfs). This way, memory
>> -  block devices can only be accessed (.online/.state attributes) by user
>> -  space once memory has been fully added. And when removing memory, we
>> -  know nobody is in critical sections.
>> -- synchronize against CPU hotplug and similar (e.g. relevant for ACPI and PPC)
>> -
>> -Especially, there is a possible lock inversion that is avoided using
>> -device_hotplug_lock when adding memory and user space tries to online that
>> -memory faster than expected:
>> -
>> -- device_online() will first take the device_lock(), followed by
>> -  mem_hotplug_lock
>> -- add_memory_resource() will first take the mem_hotplug_lock, followed by
>> -  the device_lock() (while creating the devices, during bus_add_device()).
>> -
>> -As the device is visible to user space before taking the device_lock(), this
>> -can result in a lock inversion.
>> -
>> -onlining/offlining of memory should be done via device_online()/
>> -device_offline() - to make sure it is properly synchronized to actions
>> -via sysfs. Holding device_hotplug_lock is advised (to e.g. protect online_type)
>> -
>> -When adding/removing/onlining/offlining memory or adding/removing
>> -heterogeneous/device memory, we should always hold the mem_hotplug_lock in
>> -write mode to serialise memory hotplug (e.g. access to global/zone
>> -variables).
>> -
>> -In addition, mem_hotplug_lock (in contrast to device_hotplug_lock) in read
>> -mode allows for a quite efficient get_online_mems/put_online_mems
>> -implementation, so code accessing memory can protect from that memory
>> -vanishing.
>> -
>> -
>>  Future Work
>>  ===========
>>  
>> 
>
>I reported this yesterday to Jonathan and Mike
>
>https://lkml.org/lkml/2018/12/3/340
>

Ah, Thanks :-)

>
>Anyhow
>
>Reviewed-by: David Hildenbrand <david@redhat.com>
>
>-- 
>
>Thanks,
>
>David / dhildenb

-- 
Wei Yang
Help you, Help me

Re: [PATCH 1/2] admin-guide/memory-hotplug.rst: remove locking internal part from admin-guide

[Michal Hocko]

On Wed 05-12-18 10:34:25, Wei Yang wrote:
> Locking Internal section exists in core-api documentation, which is more
> suitable for this.
> 
> This patch removes the duplication part here.
> 
> Signed-off-by: Wei Yang <richard.weiyang@gmail.com>

Yes this doesn't really make any sense in an admin guide. It is a pure
implementation detail nobody should be relying on.
-- 
Michal Hocko
SUSE Labs

[PATCH v2 1/2] admin-guide/memory-hotplug.rst: remove locking internal part from admin-guide

[Wei Yang]

Locking Internal section exists in core-api documentation, which is more
suitable for this.

This patch removes the duplication part here.

Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Acked-by: Mike Rapoport <rppt@linux.ibm.com>
---
 Documentation/admin-guide/mm/memory-hotplug.rst | 40 -------------------------
 1 file changed, 40 deletions(-)

diff --git a/Documentation/admin-guide/mm/memory-hotplug.rst b/Documentation/admin-guide/mm/memory-hotplug.rst
index 5c4432c96c4b..241f4ce1e387 100644
--- a/Documentation/admin-guide/mm/memory-hotplug.rst
+++ b/Documentation/admin-guide/mm/memory-hotplug.rst
@@ -392,46 +392,6 @@ Need more implementation yet....
  - Notification completion of remove works by OS to firmware.
  - Guard from remove if not yet.
 
-
-Locking Internals
-=================
-
-When adding/removing memory that uses memory block devices (i.e. ordinary RAM),
-the device_hotplug_lock should be held to:
-
-- synchronize against online/offline requests (e.g. via sysfs). This way, memory
-  block devices can only be accessed (.online/.state attributes) by user
-  space once memory has been fully added. And when removing memory, we
-  know nobody is in critical sections.
-- synchronize against CPU hotplug and similar (e.g. relevant for ACPI and PPC)
-
-Especially, there is a possible lock inversion that is avoided using
-device_hotplug_lock when adding memory and user space tries to online that
-memory faster than expected:
-
-- device_online() will first take the device_lock(), followed by
-  mem_hotplug_lock
-- add_memory_resource() will first take the mem_hotplug_lock, followed by
-  the device_lock() (while creating the devices, during bus_add_device()).
-
-As the device is visible to user space before taking the device_lock(), this
-can result in a lock inversion.
-
-onlining/offlining of memory should be done via device_online()/
-device_offline() - to make sure it is properly synchronized to actions
-via sysfs. Holding device_hotplug_lock is advised (to e.g. protect online_type)
-
-When adding/removing/onlining/offlining memory or adding/removing
-heterogeneous/device memory, we should always hold the mem_hotplug_lock in
-write mode to serialise memory hotplug (e.g. access to global/zone
-variables).
-
-In addition, mem_hotplug_lock (in contrast to device_hotplug_lock) in read
-mode allows for a quite efficient get_online_mems/put_online_mems
-implementation, so code accessing memory can protect from that memory
-vanishing.
-
-
 Future Work
 ===========
 
-- 
2.15.1

[PATCH v2 2/2] core-api/memory-hotplug.rst: divide Locking Internal section by different locks

[Wei Yang]

Currently locking for memory hotplug is a little complicated.

Generally speaking, we leverage the two global lock:

  * device_hotplug_lock
  * mem_hotplug_lock

to serialise the process.

While for the long term, we are willing to have more fine-grained lock
to provide higher scalability.

This patch divides Locking Internal section based on these two global
locks to help readers to understand it. Also it adds some new finding to
enrich it.

[David: words arrangement]

Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
---
v2: adjustment based on David and Mike comment
---
 Documentation/core-api/memory-hotplug.rst | 27 ++++++++++++++++++++++++---
 1 file changed, 24 insertions(+), 3 deletions(-)

diff --git a/Documentation/core-api/memory-hotplug.rst b/Documentation/core-api/memory-hotplug.rst
index de7467e48067..51d477ad4b80 100644
--- a/Documentation/core-api/memory-hotplug.rst
+++ b/Documentation/core-api/memory-hotplug.rst
@@ -89,6 +89,20 @@ NOTIFY_STOP stops further processing of the notification queue.
 Locking Internals
 =================
 
+In addition to fine grained locks like pgdat_resize_lock, there are three locks
+involved
+
+- device_hotplug_lock
+- mem_hotplug_lock
+- device_lock
+
+Currently, they are twisted together for all kinds of reasons. The following
+part is divided into device_hotplug_lock and mem_hotplug_lock parts
+respectively to describe those tricky situations.
+
+device_hotplug_lock
+---------------------
+
 When adding/removing memory that uses memory block devices (i.e. ordinary RAM),
 the device_hotplug_lock should be held to:
 
@@ -111,13 +125,20 @@ As the device is visible to user space before taking the device_lock(), this
 can result in a lock inversion.
 
 onlining/offlining of memory should be done via device_online()/
-device_offline() - to make sure it is properly synchronized to actions
-via sysfs. Holding device_hotplug_lock is advised (to e.g. protect online_type)
+device_offline() - to make sure it is properly synchronized to actions via
+sysfs. Even if mem_hotplug_lock is used to protect the process, because of the
+lock inversion described above, holding device_hotplug_lock is still advised
+(to e.g. protect online_type)
+
+mem_hotplug_lock
+---------------------
 
 When adding/removing/onlining/offlining memory or adding/removing
 heterogeneous/device memory, we should always hold the mem_hotplug_lock in
 write mode to serialise memory hotplug (e.g. access to global/zone
-variables).
+variables). Currently, we take advantage of this to serialise sparsemem's
+mem_section handling in sparse_add_one_section() and
+sparse_remove_one_section().
 
 In addition, mem_hotplug_lock (in contrast to device_hotplug_lock) in read
 mode allows for a quite efficient get_online_mems/put_online_mems
-- 
2.15.1

Re: [PATCH v2 2/2] core-api/memory-hotplug.rst: divide Locking Internal section by different locks

[Mike Rapoport]

On Thu, Dec 06, 2018 at 08:26:22AM +0800, Wei Yang wrote:
> Currently locking for memory hotplug is a little complicated.
> 
> Generally speaking, we leverage the two global lock:
> 
>   * device_hotplug_lock
>   * mem_hotplug_lock
> 
> to serialise the process.
> 
> While for the long term, we are willing to have more fine-grained lock
> to provide higher scalability.
> 
> This patch divides Locking Internal section based on these two global
> locks to help readers to understand it. Also it adds some new finding to
> enrich it.
> 
> [David: words arrangement]
> 
> Signed-off-by: Wei Yang <richard.weiyang@gmail.com>

Reviewd-by: Mike Rapoport <rppt@linux.ibm.com>

> ---
> v2: adjustment based on David and Mike comment
> ---
>  Documentation/core-api/memory-hotplug.rst | 27 ++++++++++++++++++++++++---
>  1 file changed, 24 insertions(+), 3 deletions(-)
> 
> diff --git a/Documentation/core-api/memory-hotplug.rst b/Documentation/core-api/memory-hotplug.rst
> index de7467e48067..51d477ad4b80 100644
> --- a/Documentation/core-api/memory-hotplug.rst
> +++ b/Documentation/core-api/memory-hotplug.rst
> @@ -89,6 +89,20 @@ NOTIFY_STOP stops further processing of the notification queue.
>  Locking Internals
>  =================
>  
> +In addition to fine grained locks like pgdat_resize_lock, there are three locks
> +involved
> +
> +- device_hotplug_lock
> +- mem_hotplug_lock
> +- device_lock
> +
> +Currently, they are twisted together for all kinds of reasons. The following
> +part is divided into device_hotplug_lock and mem_hotplug_lock parts
> +respectively to describe those tricky situations.
> +
> +device_hotplug_lock
> +---------------------
> +
>  When adding/removing memory that uses memory block devices (i.e. ordinary RAM),
>  the device_hotplug_lock should be held to:
>  
> @@ -111,13 +125,20 @@ As the device is visible to user space before taking the device_lock(), this
>  can result in a lock inversion.
>  
>  onlining/offlining of memory should be done via device_online()/
> -device_offline() - to make sure it is properly synchronized to actions
> -via sysfs. Holding device_hotplug_lock is advised (to e.g. protect online_type)
> +device_offline() - to make sure it is properly synchronized to actions via
> +sysfs. Even if mem_hotplug_lock is used to protect the process, because of the
> +lock inversion described above, holding device_hotplug_lock is still advised
> +(to e.g. protect online_type)
> +
> +mem_hotplug_lock
> +---------------------
>  
>  When adding/removing/onlining/offlining memory or adding/removing
>  heterogeneous/device memory, we should always hold the mem_hotplug_lock in
>  write mode to serialise memory hotplug (e.g. access to global/zone
> -variables).
> +variables). Currently, we take advantage of this to serialise sparsemem's
> +mem_section handling in sparse_add_one_section() and
> +sparse_remove_one_section().
>  
>  In addition, mem_hotplug_lock (in contrast to device_hotplug_lock) in read
>  mode allows for a quite efficient get_online_mems/put_online_mems
> -- 
> 2.15.1
> 

-- 
Sincerely yours,
Mike.

Re: [PATCH v2 2/2] core-api/memory-hotplug.rst: divide Locking Internal section by different locks

[David Hildenbrand]

On 06.12.18 01:26, Wei Yang wrote:
> Currently locking for memory hotplug is a little complicated.
> 
> Generally speaking, we leverage the two global lock:
> 
>   * device_hotplug_lock
>   * mem_hotplug_lock
> 
> to serialise the process.
> 
> While for the long term, we are willing to have more fine-grained lock
> to provide higher scalability.
> 
> This patch divides Locking Internal section based on these two global
> locks to help readers to understand it. Also it adds some new finding to
> enrich it.
> 
> [David: words arrangement]
> 
> Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
> ---
> v2: adjustment based on David and Mike comment
> ---
>  Documentation/core-api/memory-hotplug.rst | 27 ++++++++++++++++++++++++---
>  1 file changed, 24 insertions(+), 3 deletions(-)
> 
> diff --git a/Documentation/core-api/memory-hotplug.rst b/Documentation/core-api/memory-hotplug.rst
> index de7467e48067..51d477ad4b80 100644
> --- a/Documentation/core-api/memory-hotplug.rst
> +++ b/Documentation/core-api/memory-hotplug.rst
> @@ -89,6 +89,20 @@ NOTIFY_STOP stops further processing of the notification queue.
>  Locking Internals
>  =================
>  
> +In addition to fine grained locks like pgdat_resize_lock, there are three locks
> +involved
> +
> +- device_hotplug_lock
> +- mem_hotplug_lock
> +- device_lock
> +
> +Currently, they are twisted together for all kinds of reasons. The following
> +part is divided into device_hotplug_lock and mem_hotplug_lock parts
> +respectively to describe those tricky situations.
> +
> +device_hotplug_lock
> +---------------------
> +
>  When adding/removing memory that uses memory block devices (i.e. ordinary RAM),
>  the device_hotplug_lock should be held to:
>  
> @@ -111,13 +125,20 @@ As the device is visible to user space before taking the device_lock(), this
>  can result in a lock inversion.
>  
>  onlining/offlining of memory should be done via device_online()/
> -device_offline() - to make sure it is properly synchronized to actions
> -via sysfs. Holding device_hotplug_lock is advised (to e.g. protect online_type)
> +device_offline() - to make sure it is properly synchronized to actions via
> +sysfs. Even if mem_hotplug_lock is used to protect the process, because of the
> +lock inversion described above, holding device_hotplug_lock is still advised
> +(to e.g. protect online_type)
> +
> +mem_hotplug_lock
> +---------------------
>  
>  When adding/removing/onlining/offlining memory or adding/removing
>  heterogeneous/device memory, we should always hold the mem_hotplug_lock in
>  write mode to serialise memory hotplug (e.g. access to global/zone
> -variables).
> +variables). Currently, we take advantage of this to serialise sparsemem's
> +mem_section handling in sparse_add_one_section() and
> +sparse_remove_one_section().
>  
>  In addition, mem_hotplug_lock (in contrast to device_hotplug_lock) in read
>  mode allows for a quite efficient get_online_mems/put_online_mems
> 

Reviewed-by: David Hildenbrand <david@redhat.com>

-- 

Thanks,

David / dhildenb