diff mbox series

[v3,1/2] mm/filemap: use xas_try_split() in __filemap_add_folio()

Message ID 20250226210854.2045816-2-ziy@nvidia.com (mailing list archive)
State New
Headers show
Series Minimize xa_node allocation during xarry split | expand

Commit Message

Zi Yan Feb. 26, 2025, 9:08 p.m. UTC 
During __filemap_add_folio(), a shadow entry is covering n slots and a
folio covers m slots with m < n is to be added.  Instead of splitting all
n slots, only the m slots covered by the folio need to be split and the
remaining n-m shadow entries can be retained with orders ranging from m to
n-1.  This method only requires

	(n/XA_CHUNK_SHIFT) - (m/XA_CHUNK_SHIFT)

new xa_nodes instead of
	(n % XA_CHUNK_SHIFT) * ((n/XA_CHUNK_SHIFT) - (m/XA_CHUNK_SHIFT))

new xa_nodes, compared to the original xas_split_alloc() + xas_split()
one.  For example, to insert an order-0 folio when an order-9 shadow entry
is present (assuming XA_CHUNK_SHIFT is 6), 1 xa_node is needed instead of
8.

xas_try_split_min_order() is introduced to reduce the number of calls to
xas_try_split() during split.

Signed-off-by: Zi Yan <ziy@nvidia.com>
Cc: Baolin Wang <baolin.wang@linux.alibaba.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Kairui Song <kasong@tencent.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Mattew Wilcox <willy@infradead.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Kefeng Wang <wangkefeng.wang@huawei.com>
Cc: Kirill A. Shuemov <kirill.shutemov@linux.intel.com>
Cc: Ryan Roberts <ryan.roberts@arm.com>
Cc: Yang Shi <yang@os.amperecomputing.com>
Cc: Yu Zhao <yuzhao@google.com>
---
 include/linux/xarray.h |  7 +++++++
 lib/xarray.c           | 25 +++++++++++++++++++++++
 mm/filemap.c           | 45 +++++++++++++++++-------------------------
 3 files changed, 50 insertions(+), 27 deletions(-)

Comments

SeongJae Park March 8, 2025, 6:14 p.m. UTC | #1 
Hello,

On Wed, 26 Feb 2025 16:08:53 -0500 Zi Yan <ziy@nvidia.com> wrote:

> During __filemap_add_folio(), a shadow entry is covering n slots and a
> folio covers m slots with m < n is to be added.  Instead of splitting all
> n slots, only the m slots covered by the folio need to be split and the
> remaining n-m shadow entries can be retained with orders ranging from m to
> n-1.  This method only requires
> 
> 	(n/XA_CHUNK_SHIFT) - (m/XA_CHUNK_SHIFT)
> 
> new xa_nodes instead of
> 	(n % XA_CHUNK_SHIFT) * ((n/XA_CHUNK_SHIFT) - (m/XA_CHUNK_SHIFT))
> 
> new xa_nodes, compared to the original xas_split_alloc() + xas_split()
> one.  For example, to insert an order-0 folio when an order-9 shadow entry
> is present (assuming XA_CHUNK_SHIFT is 6), 1 xa_node is needed instead of
> 8.
> 
> xas_try_split_min_order() is introduced to reduce the number of calls to
> xas_try_split() during split.
> 
> Signed-off-by: Zi Yan <ziy@nvidia.com>
> Cc: Baolin Wang <baolin.wang@linux.alibaba.com>
> Cc: Hugh Dickins <hughd@google.com>
> Cc: Kairui Song <kasong@tencent.com>
> Cc: Miaohe Lin <linmiaohe@huawei.com>
> Cc: Mattew Wilcox <willy@infradead.org>
> Cc: David Hildenbrand <david@redhat.com>
> Cc: John Hubbard <jhubbard@nvidia.com>
> Cc: Kefeng Wang <wangkefeng.wang@huawei.com>
> Cc: Kirill A. Shuemov <kirill.shutemov@linux.intel.com>
> Cc: Ryan Roberts <ryan.roberts@arm.com>
> Cc: Yang Shi <yang@os.amperecomputing.com>
> Cc: Yu Zhao <yuzhao@google.com>
> ---
>  include/linux/xarray.h |  7 +++++++
>  lib/xarray.c           | 25 +++++++++++++++++++++++
>  mm/filemap.c           | 45 +++++++++++++++++-------------------------
>  3 files changed, 50 insertions(+), 27 deletions(-)
> 
> diff --git a/include/linux/xarray.h b/include/linux/xarray.h
> index 4010195201c9..78eede109b1a 100644
> --- a/include/linux/xarray.h
> +++ b/include/linux/xarray.h
> @@ -1556,6 +1556,7 @@ int xas_get_order(struct xa_state *xas);
>  void xas_split(struct xa_state *, void *entry, unsigned int order);
>  void xas_split_alloc(struct xa_state *, void *entry, unsigned int order, gfp_t);
>  void xas_try_split(struct xa_state *xas, void *entry, unsigned int order);
> +unsigned int xas_try_split_min_order(unsigned int order);
>  #else
>  static inline int xa_get_order(struct xarray *xa, unsigned long index)
>  {
> @@ -1582,6 +1583,12 @@ static inline void xas_try_split(struct xa_state *xas, void *entry,
>  		unsigned int order)
>  {
>  }
> +
> +static inline unsigned int xas_try_split_min_order(unsigned int order)
> +{
> +	return 0;
> +}
> +
>  #endif
>  
>  /**
> diff --git a/lib/xarray.c b/lib/xarray.c
> index bc197c96d171..8067182d3e43 100644
> --- a/lib/xarray.c
> +++ b/lib/xarray.c
> @@ -1133,6 +1133,28 @@ void xas_split(struct xa_state *xas, void *entry, unsigned int order)
>  }
>  EXPORT_SYMBOL_GPL(xas_split);
>  
> +/**
> + * xas_try_split_min_order() - Minimal split order xas_try_split() can accept
> + * @order: Current entry order.
> + *
> + * xas_try_split() can split a multi-index entry to smaller than @order - 1 if
> + * no new xa_node is needed. This function provides the minimal order
> + * xas_try_split() supports.
> + *
> + * Return: the minimal order xas_try_split() supports
> + *
> + * Context: Any context.
> + *
> + */
> +unsigned int xas_try_split_min_order(unsigned int order)
> +{
> +	if (order % XA_CHUNK_SHIFT == 0)
> +		return order == 0 ? 0 : order - 1;
> +
> +	return order - (order % XA_CHUNK_SHIFT);
> +}
> +EXPORT_SYMBOL_GPL(xas_try_split_min_order);
> +

I found this makes build fails when CONFIG_XARRAY_MULTI is unset, like below.

    /linux/lib/xarray.c:1251:14: error: redefinition of ‘xas_try_split_min_order’
     1251 | unsigned int xas_try_split_min_order(unsigned int order)
          |              ^~~~~~~~~~~~~~~~~~~~~~~
    In file included from /linux/lib/xarray.c:13:
    /linux/include/linux/xarray.h:1587:28: note: previous definition of ‘xas_try_split_min_order’ with type ‘unsigned int(unsigned int)’
     1587 | static inline unsigned int xas_try_split_min_order(unsigned int order)
          |                            ^~~~~~~~~~~~~~~~~~~~~~~

I think we should have the definition only when CONFIG_XARRAY_MULTI?


Thanks,
SJ

[...]
Zi Yan March 8, 2025, 6:32 p.m. UTC | #2 
On 8 Mar 2025, at 13:14, SeongJae Park wrote:

> Hello,
>
> On Wed, 26 Feb 2025 16:08:53 -0500 Zi Yan <ziy@nvidia.com> wrote:
>
>> During __filemap_add_folio(), a shadow entry is covering n slots and a
>> folio covers m slots with m < n is to be added.  Instead of splitting all
>> n slots, only the m slots covered by the folio need to be split and the
>> remaining n-m shadow entries can be retained with orders ranging from m to
>> n-1.  This method only requires
>>
>> 	(n/XA_CHUNK_SHIFT) - (m/XA_CHUNK_SHIFT)
>>
>> new xa_nodes instead of
>> 	(n % XA_CHUNK_SHIFT) * ((n/XA_CHUNK_SHIFT) - (m/XA_CHUNK_SHIFT))
>>
>> new xa_nodes, compared to the original xas_split_alloc() + xas_split()
>> one.  For example, to insert an order-0 folio when an order-9 shadow entry
>> is present (assuming XA_CHUNK_SHIFT is 6), 1 xa_node is needed instead of
>> 8.
>>
>> xas_try_split_min_order() is introduced to reduce the number of calls to
>> xas_try_split() during split.
>>
>> Signed-off-by: Zi Yan <ziy@nvidia.com>
>> Cc: Baolin Wang <baolin.wang@linux.alibaba.com>
>> Cc: Hugh Dickins <hughd@google.com>
>> Cc: Kairui Song <kasong@tencent.com>
>> Cc: Miaohe Lin <linmiaohe@huawei.com>
>> Cc: Mattew Wilcox <willy@infradead.org>
>> Cc: David Hildenbrand <david@redhat.com>
>> Cc: John Hubbard <jhubbard@nvidia.com>
>> Cc: Kefeng Wang <wangkefeng.wang@huawei.com>
>> Cc: Kirill A. Shuemov <kirill.shutemov@linux.intel.com>
>> Cc: Ryan Roberts <ryan.roberts@arm.com>
>> Cc: Yang Shi <yang@os.amperecomputing.com>
>> Cc: Yu Zhao <yuzhao@google.com>
>> ---
>>  include/linux/xarray.h |  7 +++++++
>>  lib/xarray.c           | 25 +++++++++++++++++++++++
>>  mm/filemap.c           | 45 +++++++++++++++++-------------------------
>>  3 files changed, 50 insertions(+), 27 deletions(-)
>>
>> diff --git a/include/linux/xarray.h b/include/linux/xarray.h
>> index 4010195201c9..78eede109b1a 100644
>> --- a/include/linux/xarray.h
>> +++ b/include/linux/xarray.h
>> @@ -1556,6 +1556,7 @@ int xas_get_order(struct xa_state *xas);
>>  void xas_split(struct xa_state *, void *entry, unsigned int order);
>>  void xas_split_alloc(struct xa_state *, void *entry, unsigned int order, gfp_t);
>>  void xas_try_split(struct xa_state *xas, void *entry, unsigned int order);
>> +unsigned int xas_try_split_min_order(unsigned int order);
>>  #else
>>  static inline int xa_get_order(struct xarray *xa, unsigned long index)
>>  {
>> @@ -1582,6 +1583,12 @@ static inline void xas_try_split(struct xa_state *xas, void *entry,
>>  		unsigned int order)
>>  {
>>  }
>> +
>> +static inline unsigned int xas_try_split_min_order(unsigned int order)
>> +{
>> +	return 0;
>> +}
>> +
>>  #endif
>>
>>  /**
>> diff --git a/lib/xarray.c b/lib/xarray.c
>> index bc197c96d171..8067182d3e43 100644
>> --- a/lib/xarray.c
>> +++ b/lib/xarray.c
>> @@ -1133,6 +1133,28 @@ void xas_split(struct xa_state *xas, void *entry, unsigned int order)
>>  }
>>  EXPORT_SYMBOL_GPL(xas_split);
>>
>> +/**
>> + * xas_try_split_min_order() - Minimal split order xas_try_split() can accept
>> + * @order: Current entry order.
>> + *
>> + * xas_try_split() can split a multi-index entry to smaller than @order - 1 if
>> + * no new xa_node is needed. This function provides the minimal order
>> + * xas_try_split() supports.
>> + *
>> + * Return: the minimal order xas_try_split() supports
>> + *
>> + * Context: Any context.
>> + *
>> + */
>> +unsigned int xas_try_split_min_order(unsigned int order)
>> +{
>> +	if (order % XA_CHUNK_SHIFT == 0)
>> +		return order == 0 ? 0 : order - 1;
>> +
>> +	return order - (order % XA_CHUNK_SHIFT);
>> +}
>> +EXPORT_SYMBOL_GPL(xas_try_split_min_order);
>> +
>
> I found this makes build fails when CONFIG_XARRAY_MULTI is unset, like below.
>
>     /linux/lib/xarray.c:1251:14: error: redefinition of ‘xas_try_split_min_order’
>      1251 | unsigned int xas_try_split_min_order(unsigned int order)
>           |              ^~~~~~~~~~~~~~~~~~~~~~~
>     In file included from /linux/lib/xarray.c:13:
>     /linux/include/linux/xarray.h:1587:28: note: previous definition of ‘xas_try_split_min_order’ with type ‘unsigned int(unsigned int)’
>      1587 | static inline unsigned int xas_try_split_min_order(unsigned int order)
>           |                            ^~~~~~~~~~~~~~~~~~~~~~~
>
> I think we should have the definition only when CONFIG_XARRAY_MULTI?

I think it might be a merge issue, since my original patch[1] places
xas_try_split_min_order() above xas_try_split(), both of which are
in #ifdef CONFIG_XARRAY_MULTI #endif. But mm-everything-2025-03-08-00-43
seems to move xas_try_split_min_order() below xas_try_split() and
out of CONFIG_XARRAY_MULTI guard.

[1] https://lore.kernel.org/linux-mm/20250226210854.2045816-2-ziy@nvidia.com/

--
Best Regards,
Yan, Zi
Zi Yan March 8, 2025, 6:36 p.m. UTC | #3 
On 8 Mar 2025, at 13:32, Zi Yan wrote:

> On 8 Mar 2025, at 13:14, SeongJae Park wrote:
>
>> Hello,
>>
>> On Wed, 26 Feb 2025 16:08:53 -0500 Zi Yan <ziy@nvidia.com> wrote:
>>
>>> During __filemap_add_folio(), a shadow entry is covering n slots and a
>>> folio covers m slots with m < n is to be added.  Instead of splitting all
>>> n slots, only the m slots covered by the folio need to be split and the
>>> remaining n-m shadow entries can be retained with orders ranging from m to
>>> n-1.  This method only requires
>>>
>>> 	(n/XA_CHUNK_SHIFT) - (m/XA_CHUNK_SHIFT)
>>>
>>> new xa_nodes instead of
>>> 	(n % XA_CHUNK_SHIFT) * ((n/XA_CHUNK_SHIFT) - (m/XA_CHUNK_SHIFT))
>>>
>>> new xa_nodes, compared to the original xas_split_alloc() + xas_split()
>>> one.  For example, to insert an order-0 folio when an order-9 shadow entry
>>> is present (assuming XA_CHUNK_SHIFT is 6), 1 xa_node is needed instead of
>>> 8.
>>>
>>> xas_try_split_min_order() is introduced to reduce the number of calls to
>>> xas_try_split() during split.
>>>
>>> Signed-off-by: Zi Yan <ziy@nvidia.com>
>>> Cc: Baolin Wang <baolin.wang@linux.alibaba.com>
>>> Cc: Hugh Dickins <hughd@google.com>
>>> Cc: Kairui Song <kasong@tencent.com>
>>> Cc: Miaohe Lin <linmiaohe@huawei.com>
>>> Cc: Mattew Wilcox <willy@infradead.org>
>>> Cc: David Hildenbrand <david@redhat.com>
>>> Cc: John Hubbard <jhubbard@nvidia.com>
>>> Cc: Kefeng Wang <wangkefeng.wang@huawei.com>
>>> Cc: Kirill A. Shuemov <kirill.shutemov@linux.intel.com>
>>> Cc: Ryan Roberts <ryan.roberts@arm.com>
>>> Cc: Yang Shi <yang@os.amperecomputing.com>
>>> Cc: Yu Zhao <yuzhao@google.com>
>>> ---
>>>  include/linux/xarray.h |  7 +++++++
>>>  lib/xarray.c           | 25 +++++++++++++++++++++++
>>>  mm/filemap.c           | 45 +++++++++++++++++-------------------------
>>>  3 files changed, 50 insertions(+), 27 deletions(-)
>>>
>>> diff --git a/include/linux/xarray.h b/include/linux/xarray.h
>>> index 4010195201c9..78eede109b1a 100644
>>> --- a/include/linux/xarray.h
>>> +++ b/include/linux/xarray.h
>>> @@ -1556,6 +1556,7 @@ int xas_get_order(struct xa_state *xas);
>>>  void xas_split(struct xa_state *, void *entry, unsigned int order);
>>>  void xas_split_alloc(struct xa_state *, void *entry, unsigned int order, gfp_t);
>>>  void xas_try_split(struct xa_state *xas, void *entry, unsigned int order);
>>> +unsigned int xas_try_split_min_order(unsigned int order);
>>>  #else
>>>  static inline int xa_get_order(struct xarray *xa, unsigned long index)
>>>  {
>>> @@ -1582,6 +1583,12 @@ static inline void xas_try_split(struct xa_state *xas, void *entry,
>>>  		unsigned int order)
>>>  {
>>>  }
>>> +
>>> +static inline unsigned int xas_try_split_min_order(unsigned int order)
>>> +{
>>> +	return 0;
>>> +}
>>> +
>>>  #endif
>>>
>>>  /**
>>> diff --git a/lib/xarray.c b/lib/xarray.c
>>> index bc197c96d171..8067182d3e43 100644
>>> --- a/lib/xarray.c
>>> +++ b/lib/xarray.c
>>> @@ -1133,6 +1133,28 @@ void xas_split(struct xa_state *xas, void *entry, unsigned int order)
>>>  }
>>>  EXPORT_SYMBOL_GPL(xas_split);
>>>
>>> +/**
>>> + * xas_try_split_min_order() - Minimal split order xas_try_split() can accept
>>> + * @order: Current entry order.
>>> + *
>>> + * xas_try_split() can split a multi-index entry to smaller than @order - 1 if
>>> + * no new xa_node is needed. This function provides the minimal order
>>> + * xas_try_split() supports.
>>> + *
>>> + * Return: the minimal order xas_try_split() supports
>>> + *
>>> + * Context: Any context.
>>> + *
>>> + */
>>> +unsigned int xas_try_split_min_order(unsigned int order)
>>> +{
>>> +	if (order % XA_CHUNK_SHIFT == 0)
>>> +		return order == 0 ? 0 : order - 1;
>>> +
>>> +	return order - (order % XA_CHUNK_SHIFT);
>>> +}
>>> +EXPORT_SYMBOL_GPL(xas_try_split_min_order);
>>> +
>>
>> I found this makes build fails when CONFIG_XARRAY_MULTI is unset, like below.
>>
>>     /linux/lib/xarray.c:1251:14: error: redefinition of ‘xas_try_split_min_order’
>>      1251 | unsigned int xas_try_split_min_order(unsigned int order)
>>           |              ^~~~~~~~~~~~~~~~~~~~~~~
>>     In file included from /linux/lib/xarray.c:13:
>>     /linux/include/linux/xarray.h:1587:28: note: previous definition of ‘xas_try_split_min_order’ with type ‘unsigned int(unsigned int)’
>>      1587 | static inline unsigned int xas_try_split_min_order(unsigned int order)
>>           |                            ^~~~~~~~~~~~~~~~~~~~~~~
>>
>> I think we should have the definition only when CONFIG_XARRAY_MULTI?
>
> I think it might be a merge issue, since my original patch[1] places
> xas_try_split_min_order() above xas_try_split(), both of which are
> in #ifdef CONFIG_XARRAY_MULTI #endif. But mm-everything-2025-03-08-00-43
> seems to move xas_try_split_min_order() below xas_try_split() and
> out of CONFIG_XARRAY_MULTI guard.
>
> [1] https://lore.kernel.org/linux-mm/20250226210854.2045816-2-ziy@nvidia.com/

In addition, the new comment for xas_try_split() is added to xas_split() comment.
See https://web.git.kernel.org/pub/scm/linux/kernel/git/akpm/mm.git/tree/lib/xarray.c?h=mm-everything-2025-03-08-00-43#n1084

Something went wrong when this patch was applied.

--
Best Regards,
Yan, Zi
SeongJae Park March 8, 2025, 9:34 p.m. UTC | #4 
On Sat, 08 Mar 2025 13:32:02 -0500 Zi Yan <ziy@nvidia.com> wrote:

> On 8 Mar 2025, at 13:14, SeongJae Park wrote:
> 
> > Hello,
> >
> > On Wed, 26 Feb 2025 16:08:53 -0500 Zi Yan <ziy@nvidia.com> wrote:
[...]
> >> diff --git a/include/linux/xarray.h b/include/linux/xarray.h
> >> index 4010195201c9..78eede109b1a 100644
> >> --- a/include/linux/xarray.h
> >> +++ b/include/linux/xarray.h
> >> @@ -1556,6 +1556,7 @@ int xas_get_order(struct xa_state *xas);
> >>  void xas_split(struct xa_state *, void *entry, unsigned int order);
> >>  void xas_split_alloc(struct xa_state *, void *entry, unsigned int order, gfp_t);
> >>  void xas_try_split(struct xa_state *xas, void *entry, unsigned int order);
> >> +unsigned int xas_try_split_min_order(unsigned int order);
> >>  #else
> >>  static inline int xa_get_order(struct xarray *xa, unsigned long index)
> >>  {
> >> @@ -1582,6 +1583,12 @@ static inline void xas_try_split(struct xa_state *xas, void *entry,
> >>  		unsigned int order)
> >>  {
> >>  }
> >> +
> >> +static inline unsigned int xas_try_split_min_order(unsigned int order)
> >> +{
> >> +	return 0;
> >> +}
> >> +
> >>  #endif
> >>
> >>  /**
> >> diff --git a/lib/xarray.c b/lib/xarray.c
> >> index bc197c96d171..8067182d3e43 100644
> >> --- a/lib/xarray.c
> >> +++ b/lib/xarray.c
> >> @@ -1133,6 +1133,28 @@ void xas_split(struct xa_state *xas, void *entry, unsigned int order)
> >>  }
> >>  EXPORT_SYMBOL_GPL(xas_split);
> >>
> >> +/**
> >> + * xas_try_split_min_order() - Minimal split order xas_try_split() can accept
> >> + * @order: Current entry order.
> >> + *
> >> + * xas_try_split() can split a multi-index entry to smaller than @order - 1 if
> >> + * no new xa_node is needed. This function provides the minimal order
> >> + * xas_try_split() supports.
> >> + *
> >> + * Return: the minimal order xas_try_split() supports
> >> + *
> >> + * Context: Any context.
> >> + *
> >> + */
> >> +unsigned int xas_try_split_min_order(unsigned int order)
> >> +{
> >> +	if (order % XA_CHUNK_SHIFT = 0)
> >> +		return order = 0 ? 0 : order - 1;
> >> +
> >> +	return order - (order % XA_CHUNK_SHIFT);
> >> +}
> >> +EXPORT_SYMBOL_GPL(xas_try_split_min_order);
> >> +
> >
> > I found this makes build fails when CONFIG_XARRAY_MULTI is unset, like below.
> >
> >     /linux/lib/xarray.c:1251:14: error: redefinition of ‘xas_try_split_min_order’
> >      1251 | unsigned int xas_try_split_min_order(unsigned int order)
> >           |              ^~~~~~~~~~~~~~~~~~~~~~~
> >     In file included from /linux/lib/xarray.c:13:
> >     /linux/include/linux/xarray.h:1587:28: note: previous definition of ‘xas_try_split_min_order’ with type ‘unsigned int(unsigned int)’
> >      1587 | static inline unsigned int xas_try_split_min_order(unsigned int order)
> >           |                            ^~~~~~~~~~~~~~~~~~~~~~~
> >
> > I think we should have the definition only when CONFIG_XARRAY_MULTI?
> 
> I think it might be a merge issue, since my original patch[1] places
> xas_try_split_min_order() above xas_try_split(), both of which are
> in #ifdef CONFIG_XARRAY_MULTI #endif. But mm-everything-2025-03-08-00-43
> seems to move xas_try_split_min_order() below xas_try_split() and
> out of CONFIG_XARRAY_MULTI guard.

You're right.  I was testing this on the mm-unstable tree, more specifically,
commit 2f0c87542d97.

I confirmed the build failure goes away after moving the definition to the
original place.

> 
> [1] https://lore.kernel.org/linux-mm/20250226210854.2045816-2-ziy@nvidia.com/
> 
> --
> Best Regards,
> Yan, Zi


Thanks,
SJ
diff mbox series

Patch

diff --git a/include/linux/xarray.h b/include/linux/xarray.h
index 4010195201c9..78eede109b1a 100644
--- a/include/linux/xarray.h
+++ b/include/linux/xarray.h
@@ -1556,6 +1556,7 @@  int xas_get_order(struct xa_state *xas);
 void xas_split(struct xa_state *, void *entry, unsigned int order);
 void xas_split_alloc(struct xa_state *, void *entry, unsigned int order, gfp_t);
 void xas_try_split(struct xa_state *xas, void *entry, unsigned int order);
+unsigned int xas_try_split_min_order(unsigned int order);
 #else
 static inline int xa_get_order(struct xarray *xa, unsigned long index)
 {
@@ -1582,6 +1583,12 @@  static inline void xas_try_split(struct xa_state *xas, void *entry,
 		unsigned int order)
 {
 }
+
+static inline unsigned int xas_try_split_min_order(unsigned int order)
+{
+	return 0;
+}
+
 #endif
 
 /**
diff --git a/lib/xarray.c b/lib/xarray.c
index bc197c96d171..8067182d3e43 100644
--- a/lib/xarray.c
+++ b/lib/xarray.c
@@ -1133,6 +1133,28 @@  void xas_split(struct xa_state *xas, void *entry, unsigned int order)
 }
 EXPORT_SYMBOL_GPL(xas_split);
 
+/**
+ * xas_try_split_min_order() - Minimal split order xas_try_split() can accept
+ * @order: Current entry order.
+ *
+ * xas_try_split() can split a multi-index entry to smaller than @order - 1 if
+ * no new xa_node is needed. This function provides the minimal order
+ * xas_try_split() supports.
+ *
+ * Return: the minimal order xas_try_split() supports
+ *
+ * Context: Any context.
+ *
+ */
+unsigned int xas_try_split_min_order(unsigned int order)
+{
+	if (order % XA_CHUNK_SHIFT == 0)
+		return order == 0 ? 0 : order - 1;
+
+	return order - (order % XA_CHUNK_SHIFT);
+}
+EXPORT_SYMBOL_GPL(xas_try_split_min_order);
+
 /**
  * xas_try_split() - Try to split a multi-index entry.
  * @xas: XArray operation state.
@@ -1144,6 +1166,9 @@  EXPORT_SYMBOL_GPL(xas_split);
  * needed, the function will use GFP_NOWAIT to get one if xas->xa_alloc is
  * NULL. If more new xa_node are needed, the function gives EINVAL error.
  *
+ * NOTE: use xas_try_split_min_order() to get next split order instead of
+ * @order - 1 if you want to minmize xas_try_split() calls.
+ *
  * Context: Any context.  The caller should hold the xa_lock.
  */
 void xas_try_split(struct xa_state *xas, void *entry, unsigned int order)
diff --git a/mm/filemap.c b/mm/filemap.c
index 2b860b59a521..cfb49ed659a1 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -857,11 +857,10 @@  EXPORT_SYMBOL_GPL(replace_page_cache_folio);
 noinline int __filemap_add_folio(struct address_space *mapping,
 		struct folio *folio, pgoff_t index, gfp_t gfp, void **shadowp)
 {
-	XA_STATE(xas, &mapping->i_pages, index);
-	void *alloced_shadow = NULL;
-	int alloced_order = 0;
+	XA_STATE_ORDER(xas, &mapping->i_pages, index, folio_order(folio));
 	bool huge;
 	long nr;
+	unsigned int forder = folio_order(folio);
 
 	VM_BUG_ON_FOLIO(!folio_test_locked(folio), folio);
 	VM_BUG_ON_FOLIO(folio_test_swapbacked(folio), folio);
@@ -870,7 +869,6 @@  noinline int __filemap_add_folio(struct address_space *mapping,
 	mapping_set_update(&xas, mapping);
 
 	VM_BUG_ON_FOLIO(index & (folio_nr_pages(folio) - 1), folio);
-	xas_set_order(&xas, index, folio_order(folio));
 	huge = folio_test_hugetlb(folio);
 	nr = folio_nr_pages(folio);
 
@@ -880,7 +878,7 @@  noinline int __filemap_add_folio(struct address_space *mapping,
 	folio->index = xas.xa_index;
 
 	for (;;) {
-		int order = -1, split_order = 0;
+		int order = -1;
 		void *entry, *old = NULL;
 
 		xas_lock_irq(&xas);
@@ -898,21 +896,25 @@  noinline int __filemap_add_folio(struct address_space *mapping,
 				order = xas_get_order(&xas);
 		}
 
-		/* entry may have changed before we re-acquire the lock */
-		if (alloced_order && (old != alloced_shadow || order != alloced_order)) {
-			xas_destroy(&xas);
-			alloced_order = 0;
-		}
-
 		if (old) {
-			if (order > 0 && order > folio_order(folio)) {
+			if (order > 0 && order > forder) {
+				unsigned int split_order = max(forder,
+						xas_try_split_min_order(order));
+
 				/* How to handle large swap entries? */
 				BUG_ON(shmem_mapping(mapping));
-				if (!alloced_order) {
-					split_order = order;
-					goto unlock;
+
+				while (order > forder) {
+					xas_set_order(&xas, index, split_order);
+					xas_try_split(&xas, old, order);
+					if (xas_error(&xas))
+						goto unlock;
+					order = split_order;
+					split_order =
+						max(xas_try_split_min_order(
+							    split_order),
+						    forder);
 				}
-				xas_split(&xas, old, order);
 				xas_reset(&xas);
 			}
 			if (shadowp)
@@ -936,17 +938,6 @@  noinline int __filemap_add_folio(struct address_space *mapping,
 unlock:
 		xas_unlock_irq(&xas);
 
-		/* split needed, alloc here and retry. */
-		if (split_order) {
-			xas_split_alloc(&xas, old, split_order, gfp);
-			if (xas_error(&xas))
-				goto error;
-			alloced_shadow = old;
-			alloced_order = split_order;
-			xas_reset(&xas);
-			continue;
-		}
-
 		if (!xas_nomem(&xas, gfp))
 			break;
 	}