[PATCH] mm/migrate_device: fix double unlock and remove dead code

[PATCH] mm/migrate_device: fix double unlock and remove dead code

[Sunny Patel]

Fix two bugs in device migration paths:

1) migrate_vma_collect_huge_pmd() calls spin_unlock after
   softleaf_entry_wait_on_locked(), which already releases the ptl.

2) migrate_vma_insert_huge_pmd_page() has a dead else-if branch and this
   branch is always unreachable.

Signed-off-by: Sunny Patel <nueralspacetech@gmail.com>
---
 mm/migrate_device.c | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)

diff --git a/mm/migrate_device.c b/mm/migrate_device.c
index 8079676c8f1f..0e005c26ee88 100644
--- a/mm/migrate_device.c
+++ b/mm/migrate_device.c
@@ -177,7 +177,6 @@ static int migrate_vma_collect_huge_pmd(pmd_t *pmdp, unsigned long start,
 
 		if (softleaf_is_migration(entry)) {
 			softleaf_entry_wait_on_locked(entry, ptl);
-			spin_unlock(ptl);
 			return -EAGAIN;
 		}
 
@@ -869,8 +868,7 @@ static int migrate_vma_insert_huge_pmd_page(struct migrate_vma *migrate,
 		if (!is_huge_zero_pmd(*pmdp))
 			goto unlock_abort;
 		flush = true;
-	} else if (!pmd_none(*pmdp))
-		goto unlock_abort;
+	}
 
 	add_mm_counter(vma->vm_mm, MM_ANONPAGES, HPAGE_PMD_NR);
 	folio_add_new_anon_rmap(folio, vma, addr, RMAP_EXCLUSIVE);
-- 
2.43.0

Re: [PATCH] mm/migrate_device: fix double unlock and remove dead code

[David Hildenbrand (Arm)]

On 4/13/26 15:09, Sunny Patel wrote:
> Fix two bugs in device migration paths:
> 
> 1) migrate_vma_collect_huge_pmd() calls spin_unlock after
>    softleaf_entry_wait_on_locked(), which already releases the ptl.
> 
> 2) migrate_vma_insert_huge_pmd_page() has a dead else-if branch and this
>    branch is always unreachable.
> 

Can you move 1) into a separate patch, add a Fixes: tag an CC stable?

I think it is

Fixes: a30b48bf1b24 ("mm/migrate_device: implement THP migration of zone
device pages")

2) will then be a pure cleanup patch.

Thanks!

> Signed-off-by: Sunny Patel <nueralspacetech@gmail.com>
> ---
>  mm/migrate_device.c | 4 +---
>  1 file changed, 1 insertion(+), 3 deletions(-)
> 
> diff --git a/mm/migrate_device.c b/mm/migrate_device.c
> index 8079676c8f1f..0e005c26ee88 100644
> --- a/mm/migrate_device.c
> +++ b/mm/migrate_device.c
> @@ -177,7 +177,6 @@ static int migrate_vma_collect_huge_pmd(pmd_t *pmdp, unsigned long start,
>  
>  		if (softleaf_is_migration(entry)) {
>  			softleaf_entry_wait_on_locked(entry, ptl);
> -			spin_unlock(ptl);


Yes, that looks correct to me.

>  			return -EAGAIN;
>  		}
>  
> @@ -869,8 +868,7 @@ static int migrate_vma_insert_huge_pmd_page(struct migrate_vma *migrate,
>  		if (!is_huge_zero_pmd(*pmdp))
>  			goto unlock_abort;
>  		flush = true;
> -	} else if (!pmd_none(*pmdp))
> -		goto unlock_abort;
> +	}

Huh, how did that happen. I hope that it's not a typo and we wanted to
check for something else.

-- 
Cheers,

David

Re: [PATCH] mm/migrate_device: fix double unlock and remove dead code

[Zi Yan]

On 13 Apr 2026, at 15:38, David Hildenbrand (Arm) wrote:

> On 4/13/26 15:09, Sunny Patel wrote:
>> Fix two bugs in device migration paths:
>>
>> 1) migrate_vma_collect_huge_pmd() calls spin_unlock after
>>    softleaf_entry_wait_on_locked(), which already releases the ptl.
>>
>> 2) migrate_vma_insert_huge_pmd_page() has a dead else-if branch and this
>>    branch is always unreachable.
>>
>
> Can you move 1) into a separate patch, add a Fixes: tag an CC stable?
>
> I think it is
>
> Fixes: a30b48bf1b24 ("mm/migrate_device: implement THP migration of zone
> device pages")
>
> 2) will then be a pure cleanup patch.
>
> Thanks!
>
>> Signed-off-by: Sunny Patel <nueralspacetech@gmail.com>
>> ---
>>  mm/migrate_device.c | 4 +---
>>  1 file changed, 1 insertion(+), 3 deletions(-)
>>
>> diff --git a/mm/migrate_device.c b/mm/migrate_device.c
>> index 8079676c8f1f..0e005c26ee88 100644
>> --- a/mm/migrate_device.c
>> +++ b/mm/migrate_device.c
>> @@ -177,7 +177,6 @@ static int migrate_vma_collect_huge_pmd(pmd_t *pmdp, unsigned long start,
>>
>>  		if (softleaf_is_migration(entry)) {
>>  			softleaf_entry_wait_on_locked(entry, ptl);
>> -			spin_unlock(ptl);
>
>
> Yes, that looks correct to me.
>
>>  			return -EAGAIN;
>>  		}
>>
>> @@ -869,8 +868,7 @@ static int migrate_vma_insert_huge_pmd_page(struct migrate_vma *migrate,
>>  		if (!is_huge_zero_pmd(*pmdp))
>>  			goto unlock_abort;
>>  		flush = true;
>> -	} else if (!pmd_none(*pmdp))
>> -		goto unlock_abort;
>> +	}
>
> Huh, how did that happen. I hope that it's not a typo and we wanted to
> check for something else.
>


Looking at the function and trying to figure this out, but find
VM_WARN_ON_FOLIO(!folio, folio) at the top, where folio is from page_folio(page).
It is either a nop or a zero dereferencing. It should be removed.

VM_WARN_ON_ONCE(!pmd_none(*pmdp) && !is_huge_zero_pmd(*pmdp)) can be removed
too, since the above ifs takes !pmd_none(*pmdp) && !is_huge_zero_pmd(*pmdp)
to unlock_abort.

Back to the above code:

        if (!pmd_none(*pmdp)) {
                if (!is_huge_zero_pmd(*pmdp))
                        goto unlock_abort;
                flush = true;
        } else if (!pmd_none(*pmdp))
                goto unlock_abort;

It seems to me that the first if should be removed, since if pmdp is
not pmd_none(), others filled pmd entry before us, so no further
action should be taken, otherwise, the function will overwrite
some valid pmd entry.

OK, look at migrate_vma_insert_page(), which does PTE level work,
the above code might be intended to do:

        if (pmd_present(*pmdp)) {
                if (!is_huge_zero_pmd(*pmdp))
                        goto unlock_abort;
                flush = true;
        } else if (!pmd_none(*pmdp))
                goto unlock_abort;



Best Regards,
Yan, Zi

Re: [PATCH] mm/migrate_device: fix double unlock and remove dead code

[David Hildenbrand (Arm)]

On 4/13/26 22:03, Zi Yan wrote:
> On 13 Apr 2026, at 15:38, David Hildenbrand (Arm) wrote:
> 
>> On 4/13/26 15:09, Sunny Patel wrote:
>>> Fix two bugs in device migration paths:
>>>
>>> 1) migrate_vma_collect_huge_pmd() calls spin_unlock after
>>>    softleaf_entry_wait_on_locked(), which already releases the ptl.
>>>
>>> 2) migrate_vma_insert_huge_pmd_page() has a dead else-if branch and this
>>>    branch is always unreachable.
>>>
>>
>> Can you move 1) into a separate patch, add a Fixes: tag an CC stable?
>>
>> I think it is
>>
>> Fixes: a30b48bf1b24 ("mm/migrate_device: implement THP migration of zone
>> device pages")
>>
>> 2) will then be a pure cleanup patch.
>>
>> Thanks!
>>
>>> Signed-off-by: Sunny Patel <nueralspacetech@gmail.com>
>>> ---
>>>  mm/migrate_device.c | 4 +---
>>>  1 file changed, 1 insertion(+), 3 deletions(-)
>>>
>>> diff --git a/mm/migrate_device.c b/mm/migrate_device.c
>>> index 8079676c8f1f..0e005c26ee88 100644
>>> --- a/mm/migrate_device.c
>>> +++ b/mm/migrate_device.c
>>> @@ -177,7 +177,6 @@ static int migrate_vma_collect_huge_pmd(pmd_t *pmdp, unsigned long start,
>>>
>>>  		if (softleaf_is_migration(entry)) {
>>>  			softleaf_entry_wait_on_locked(entry, ptl);
>>> -			spin_unlock(ptl);
>>
>>
>> Yes, that looks correct to me.
>>
>>>  			return -EAGAIN;
>>>  		}
>>>
>>> @@ -869,8 +868,7 @@ static int migrate_vma_insert_huge_pmd_page(struct migrate_vma *migrate,
>>>  		if (!is_huge_zero_pmd(*pmdp))
>>>  			goto unlock_abort;
>>>  		flush = true;
>>> -	} else if (!pmd_none(*pmdp))
>>> -		goto unlock_abort;
>>> +	}
>>
>> Huh, how did that happen. I hope that it's not a typo and we wanted to
>> check for something else.
>>
> 
> 
> Looking at the function and trying to figure this out, but find
> VM_WARN_ON_FOLIO(!folio, folio) at the top, where folio is from page_folio(page).
> It is either a nop or a zero dereferencing. It should be removed.

Heh, VM_WARN_ON_FOLIO(!folio, folio) itself is completely odd. Dump NULL
if NULL.

Agreed that this should be removed. Or replaced by a

VM_WARN_ON_ONCE(page);

> 
> VM_WARN_ON_ONCE(!pmd_none(*pmdp) && !is_huge_zero_pmd(*pmdp)) can be removed
> too, since the above ifs takes !pmd_none(*pmdp) && !is_huge_zero_pmd(*pmdp)
> to unlock_abort.

Right, and if that's an unexpected case, we should warn on that path
instead. Like

	if (WARN_ON_ONCE(!is_huge_zero_pmd(*pmdp)))
		goto unlock_abort;

> 
> Back to the above code:
> 
>         if (!pmd_none(*pmdp)) {
>                 if (!is_huge_zero_pmd(*pmdp))
>                         goto unlock_abort;
>                 flush = true;
>         } else if (!pmd_none(*pmdp))
>                 goto unlock_abort;
> 
> It seems to me that the first if should be removed, since if pmdp is
> not pmd_none(), others filled pmd entry before us, so no further
> action should be taken, otherwise, the function will overwrite
> some valid pmd entry.
> 
> OK, look at migrate_vma_insert_page(), which does PTE level work,
> the above code might be intended to do:
> 
>         if (pmd_present(*pmdp)) {
>                 if (!is_huge_zero_pmd(*pmdp))
>                         goto unlock_abort;
>                 flush = true;
>         } else if (!pmd_none(*pmdp))
>                 goto unlock_abort;

Agreed, that makes more sense. Thanks!

-- 
Cheers,

David

Re: [PATCH] mm/migrate_device: fix double unlock and remove dead code

[Sunny Patel]

Please refer to below new cleanup Patch as discussed.

https://lore.kernel.org/linux-mm/20260414141344.29949-1-nueralspacetech@gmail.com/T/#u

Thanks,
Sunny Patel

Re: [PATCH] mm/migrate_device: fix double unlock and remove dead code

[Sunny Patel]

On Mon, 13 Apr 2026 15:09:00 +0000, David Hildenbrand wrote:
> Can you move 1) into a separate patch, add a Fixes: tag and CC stable?
> I think it is
> Fixes: a30b48bf1b24 ("mm/migrate_device: implement THP migration of
> zone device pages")

Agreed, and please refer below link for new patch
https://lore.kernel.org/linux-mm/20260413211559.20969-1-nueralspacetech@gmail.com/T/#u

> 2) will then be a pure cleanup patch.

Will send cleanup patch shortly.

Thanks,
Sunny Patel

Re: [PATCH] mm/migrate_device: fix double unlock and remove dead code

[Matthew Brost]

On Mon, Apr 13, 2026 at 06:39:23PM +0530, Sunny Patel wrote:
> Fix two bugs in device migration paths:
> 
> 1) migrate_vma_collect_huge_pmd() calls spin_unlock after
>    softleaf_entry_wait_on_locked(), which already releases the ptl.
> 
> 2) migrate_vma_insert_huge_pmd_page() has a dead else-if branch and this
>    branch is always unreachable.
> 
> Signed-off-by: Sunny Patel <nueralspacetech@gmail.com>
> ---
>  mm/migrate_device.c | 4 +---
>  1 file changed, 1 insertion(+), 3 deletions(-)
> 
> diff --git a/mm/migrate_device.c b/mm/migrate_device.c
> index 8079676c8f1f..0e005c26ee88 100644
> --- a/mm/migrate_device.c
> +++ b/mm/migrate_device.c
> @@ -177,7 +177,6 @@ static int migrate_vma_collect_huge_pmd(pmd_t *pmdp, unsigned long start,
>  
>  		if (softleaf_is_migration(entry)) {
>  			softleaf_entry_wait_on_locked(entry, ptl);
> -			spin_unlock(ptl);
>  			return -EAGAIN;
>  		}

This entire if statement is dead code, since we bail out and collect a
skip on !softleaf_is_device_private immediately above. I thought we
already had a patch from Davidlohr Bueso 'mm/migrate_device: remove dead
migration entry check in migrate_vma_collect_huge_pmd()' that removed
this whole if statement. I even have an email from Andrew pulling this
patch, but it looks like the code is still present in 7.0.

I found this link via a google search which contains the above patch
too:

https://kernel.googlesource.com/pub/scm/linux/kernel/git/akpm/25-new/+/f405e7bc5d93a541cf11663b2424baf0a03b496f%5E%21/

Look like this change got lost somehow?

Matt

>  
> @@ -869,8 +868,7 @@ static int migrate_vma_insert_huge_pmd_page(struct migrate_vma *migrate,
>  		if (!is_huge_zero_pmd(*pmdp))
>  			goto unlock_abort;
>  		flush = true;
> -	} else if (!pmd_none(*pmdp))
> -		goto unlock_abort;
> +	}
>  
>  	add_mm_counter(vma->vm_mm, MM_ANONPAGES, HPAGE_PMD_NR);
>  	folio_add_new_anon_rmap(folio, vma, addr, RMAP_EXCLUSIVE);
> -- 
> 2.43.0
> 

Re: [PATCH] mm/migrate_device: fix double unlock and remove dead code

[David Hildenbrand (Arm)]

On 4/14/26 01:30, Matthew Brost wrote:
> On Mon, Apr 13, 2026 at 06:39:23PM +0530, Sunny Patel wrote:
>> Fix two bugs in device migration paths:
>>
>> 1) migrate_vma_collect_huge_pmd() calls spin_unlock after
>>    softleaf_entry_wait_on_locked(), which already releases the ptl.
>>
>> 2) migrate_vma_insert_huge_pmd_page() has a dead else-if branch and this
>>    branch is always unreachable.
>>
>> Signed-off-by: Sunny Patel <nueralspacetech@gmail.com>
>> ---
>>  mm/migrate_device.c | 4 +---
>>  1 file changed, 1 insertion(+), 3 deletions(-)
>>
>> diff --git a/mm/migrate_device.c b/mm/migrate_device.c
>> index 8079676c8f1f..0e005c26ee88 100644
>> --- a/mm/migrate_device.c
>> +++ b/mm/migrate_device.c
>> @@ -177,7 +177,6 @@ static int migrate_vma_collect_huge_pmd(pmd_t *pmdp, unsigned long start,
>>  
>>  		if (softleaf_is_migration(entry)) {
>>  			softleaf_entry_wait_on_locked(entry, ptl);
>> -			spin_unlock(ptl);
>>  			return -EAGAIN;
>>  		}
> 
> This entire if statement is dead code, since we bail out and collect a
> skip on !softleaf_is_device_private immediately above. I thought we
> already had a patch from Davidlohr Bueso 'mm/migrate_device: remove dead
> migration entry check in migrate_vma_collect_huge_pmd()' that removed
> this whole if statement. I even have an email from Andrew pulling this
> patch, but it looks like the code is still present in 7.0.
> 
> I found this link via a google search which contains the above patch
> too:
> 
> https://kernel.googlesource.com/pub/scm/linux/kernel/git/akpm/25-new/+/f405e7bc5d93a541cf11663b2424baf0a03b496f%5E%21/
> 
> Look like this change got lost somehow?

Oh, even better, then we don't even need a Fixes:

-- 
Cheers,

David

Re: [PATCH] mm/migrate_device: fix double unlock and remove dead code

[kernel test robot]

BCC: lkp@intel.com
CC: oe-kbuild-all@lists.linux.dev
In-Reply-To: <20260413130927.13263-1-nueralspacetech@gmail.com>
References: <20260413130927.13263-1-nueralspacetech@gmail.com>
TO: Sunny Patel <nueralspacetech@gmail.com>
TO: Andrew Morton <akpm@linux-foundation.org>
CC: Linux Memory Management List <linux-mm@kvack.org>
TO: David Hildenbrand <david@kernel.org>
CC: Zi Yan <ziy@nvidia.com>
CC: Matthew Brost <matthew.brost@intel.com>
CC: Joshua Hahn <joshua.hahnjy@gmail.com>
CC: Rakie Kim <rakie.kim@sk.com>
CC: Byungchul Park <byungchul@sk.com>
CC: Gregory Price <gourry@gourry.net>
CC: Ying Huang <ying.huang@linux.alibaba.com>
CC: Alistair Popple <apopple@nvidia.com>
CC: linux-kernel@vger.kernel.org
CC: Sunny Patel <nueralspacetech@gmail.com>

Hi Sunny,

kernel test robot noticed the following build warnings:

[auto build test WARNING on akpm-mm/mm-everything]

url:    https://github.com/intel-lab-lkp/linux/commits/Sunny-Patel/mm-migrate_device-fix-double-unlock-and-remove-dead-code/20260414-011221
base:   https://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm.git mm-everything
patch link:    https://lore.kernel.org/r/20260413130927.13263-1-nueralspacetech%40gmail.com
patch subject: [PATCH] mm/migrate_device: fix double unlock and remove dead code
:::::: branch date: 27 hours ago
:::::: commit date: 27 hours ago
config: x86_64-randconfig-161-20260414 (https://download.01.org/0day-ci/archive/20260415/202604150314.JzBIroDK-lkp@intel.com/config)
compiler: clang version 20.1.8 (https://github.com/llvm/llvm-project 87f0227cb60147a26a1eeb4fb06e3b505e9c7261)
rustc: rustc 1.88.0 (6b00bc388 2025-06-23)
smatch: v0.5.0-9007-gcf3ea02b

If you fix the issue in a separate patch/commit (i.e. not just a new version of
the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <lkp@intel.com>
| Reported-by: Dan Carpenter <error27@gmail.com>
| Closes: https://lore.kernel.org/r/202604150314.JzBIroDK-lkp@intel.com/

smatch warnings:
mm/migrate_device.c:243 migrate_vma_collect_huge_pmd() warn: inconsistent returns 'ptl'.

vim +/ptl +243 mm/migrate_device.c

022a12deda53c98 Balbir Singh            2025-10-01  123  
a30b48bf1b244f1 Balbir Singh            2025-10-01  124  /** migrate_vma_collect_huge_pmd - collect THP pages without splitting the
a30b48bf1b244f1 Balbir Singh            2025-10-01  125   * folio for device private pages.
a30b48bf1b244f1 Balbir Singh            2025-10-01  126   * @pmdp: pointer to pmd entry
a30b48bf1b244f1 Balbir Singh            2025-10-01  127   * @start: start address of the range for migration
a30b48bf1b244f1 Balbir Singh            2025-10-01  128   * @end: end address of the range for migration
a30b48bf1b244f1 Balbir Singh            2025-10-01  129   * @walk: mm_walk callback structure
a30b48bf1b244f1 Balbir Singh            2025-10-01  130   * @fault_folio: folio associated with the fault if any
a30b48bf1b244f1 Balbir Singh            2025-10-01  131   *
a30b48bf1b244f1 Balbir Singh            2025-10-01  132   * Collect the huge pmd entry at @pmdp for migration and set the
a30b48bf1b244f1 Balbir Singh            2025-10-01  133   * MIGRATE_PFN_COMPOUND flag in the migrate src entry to indicate that
a30b48bf1b244f1 Balbir Singh            2025-10-01  134   * migration will occur at HPAGE_PMD granularity
a30b48bf1b244f1 Balbir Singh            2025-10-01  135   */
a30b48bf1b244f1 Balbir Singh            2025-10-01  136  static int migrate_vma_collect_huge_pmd(pmd_t *pmdp, unsigned long start,
a30b48bf1b244f1 Balbir Singh            2025-10-01  137  					unsigned long end, struct mm_walk *walk,
a30b48bf1b244f1 Balbir Singh            2025-10-01  138  					struct folio *fault_folio)
76cbbead253ddca Christoph Hellwig       2022-02-16  139  {
a30b48bf1b244f1 Balbir Singh            2025-10-01  140  	struct mm_struct *mm = walk->mm;
a30b48bf1b244f1 Balbir Singh            2025-10-01  141  	struct folio *folio;
76cbbead253ddca Christoph Hellwig       2022-02-16  142  	struct migrate_vma *migrate = walk->private;
76cbbead253ddca Christoph Hellwig       2022-02-16  143  	spinlock_t *ptl;
a30b48bf1b244f1 Balbir Singh            2025-10-01  144  	int ret;
a30b48bf1b244f1 Balbir Singh            2025-10-01  145  	unsigned long write = 0;
76cbbead253ddca Christoph Hellwig       2022-02-16  146  
a30b48bf1b244f1 Balbir Singh            2025-10-01  147  	ptl = pmd_lock(mm, pmdp);
a30b48bf1b244f1 Balbir Singh            2025-10-01  148  	if (pmd_none(*pmdp)) {
a30b48bf1b244f1 Balbir Singh            2025-10-01  149  		spin_unlock(ptl);
76cbbead253ddca Christoph Hellwig       2022-02-16  150  		return migrate_vma_collect_hole(start, end, -1, walk);
a30b48bf1b244f1 Balbir Singh            2025-10-01  151  	}
76cbbead253ddca Christoph Hellwig       2022-02-16  152  
76cbbead253ddca Christoph Hellwig       2022-02-16  153  	if (pmd_trans_huge(*pmdp)) {
a30b48bf1b244f1 Balbir Singh            2025-10-01  154  		if (!(migrate->flags & MIGRATE_VMA_SELECT_SYSTEM)) {
76cbbead253ddca Christoph Hellwig       2022-02-16  155  			spin_unlock(ptl);
a30b48bf1b244f1 Balbir Singh            2025-10-01  156  			return migrate_vma_collect_skip(start, end, walk);
76cbbead253ddca Christoph Hellwig       2022-02-16  157  		}
76cbbead253ddca Christoph Hellwig       2022-02-16  158  
b002a7b0a58acea Matthew Wilcox (Oracle  2024-03-26  159) 		folio = pmd_folio(*pmdp);
b002a7b0a58acea Matthew Wilcox (Oracle  2024-03-26  160) 		if (is_huge_zero_folio(folio)) {
76cbbead253ddca Christoph Hellwig       2022-02-16  161  			spin_unlock(ptl);
a30b48bf1b244f1 Balbir Singh            2025-10-01  162  			return migrate_vma_collect_hole(start, end, -1, walk);
a30b48bf1b244f1 Balbir Singh            2025-10-01  163  		}
a30b48bf1b244f1 Balbir Singh            2025-10-01  164  		if (pmd_write(*pmdp))
a30b48bf1b244f1 Balbir Singh            2025-10-01  165  			write = MIGRATE_PFN_WRITE;
a30b48bf1b244f1 Balbir Singh            2025-10-01  166  	} else if (!pmd_present(*pmdp)) {
0ac881efe164685 Lorenzo Stoakes         2025-11-10  167  		const softleaf_t entry = softleaf_from_pmd(*pmdp);
a30b48bf1b244f1 Balbir Singh            2025-10-01  168  
0ac881efe164685 Lorenzo Stoakes         2025-11-10  169  		folio = softleaf_to_folio(entry);
0ac881efe164685 Lorenzo Stoakes         2025-11-10  170  
0ac881efe164685 Lorenzo Stoakes         2025-11-10  171  		if (!softleaf_is_device_private(entry) ||
a30b48bf1b244f1 Balbir Singh            2025-10-01  172  			!(migrate->flags & MIGRATE_VMA_SELECT_DEVICE_PRIVATE) ||
a30b48bf1b244f1 Balbir Singh            2025-10-01  173  			(folio->pgmap->owner != migrate->pgmap_owner)) {
a30b48bf1b244f1 Balbir Singh            2025-10-01  174  			spin_unlock(ptl);
a30b48bf1b244f1 Balbir Singh            2025-10-01  175  			return migrate_vma_collect_skip(start, end, walk);
a30b48bf1b244f1 Balbir Singh            2025-10-01  176  		}
a30b48bf1b244f1 Balbir Singh            2025-10-01  177  
0ac881efe164685 Lorenzo Stoakes         2025-11-10  178  		if (softleaf_is_migration(entry)) {
b570f37a2ce480b Thomas Hellström        2026-02-10  179  			softleaf_entry_wait_on_locked(entry, ptl);
a30b48bf1b244f1 Balbir Singh            2025-10-01  180  			return -EAGAIN;
a30b48bf1b244f1 Balbir Singh            2025-10-01  181  		}
a30b48bf1b244f1 Balbir Singh            2025-10-01  182  
0ac881efe164685 Lorenzo Stoakes         2025-11-10  183  		if (softleaf_is_device_private_write(entry))
a30b48bf1b244f1 Balbir Singh            2025-10-01  184  			write = MIGRATE_PFN_WRITE;
76cbbead253ddca Christoph Hellwig       2022-02-16  185  	} else {
a30b48bf1b244f1 Balbir Singh            2025-10-01  186  		spin_unlock(ptl);
a30b48bf1b244f1 Balbir Singh            2025-10-01  187  		return -EAGAIN;
a30b48bf1b244f1 Balbir Singh            2025-10-01  188  	}
76cbbead253ddca Christoph Hellwig       2022-02-16  189  
b002a7b0a58acea Matthew Wilcox (Oracle  2024-03-26  190) 	folio_get(folio);
a30b48bf1b244f1 Balbir Singh            2025-10-01  191  	if (folio != fault_folio && unlikely(!folio_trylock(folio))) {
a30b48bf1b244f1 Balbir Singh            2025-10-01  192  		spin_unlock(ptl);
a30b48bf1b244f1 Balbir Singh            2025-10-01  193  		folio_put(folio);
a30b48bf1b244f1 Balbir Singh            2025-10-01  194  		return migrate_vma_collect_skip(start, end, walk);
a30b48bf1b244f1 Balbir Singh            2025-10-01  195  	}
a30b48bf1b244f1 Balbir Singh            2025-10-01  196  
a30b48bf1b244f1 Balbir Singh            2025-10-01  197  	if (thp_migration_supported() &&
a30b48bf1b244f1 Balbir Singh            2025-10-01  198  		(migrate->flags & MIGRATE_VMA_SELECT_COMPOUND) &&
a30b48bf1b244f1 Balbir Singh            2025-10-01  199  		(IS_ALIGNED(start, HPAGE_PMD_SIZE) &&
a30b48bf1b244f1 Balbir Singh            2025-10-01  200  		 IS_ALIGNED(end, HPAGE_PMD_SIZE))) {
a30b48bf1b244f1 Balbir Singh            2025-10-01  201  
a30b48bf1b244f1 Balbir Singh            2025-10-01  202  		struct page_vma_mapped_walk pvmw = {
a30b48bf1b244f1 Balbir Singh            2025-10-01  203  			.ptl = ptl,
a30b48bf1b244f1 Balbir Singh            2025-10-01  204  			.address = start,
a30b48bf1b244f1 Balbir Singh            2025-10-01  205  			.pmd = pmdp,
a30b48bf1b244f1 Balbir Singh            2025-10-01  206  			.vma = walk->vma,
a30b48bf1b244f1 Balbir Singh            2025-10-01  207  		};
a30b48bf1b244f1 Balbir Singh            2025-10-01  208  
a30b48bf1b244f1 Balbir Singh            2025-10-01  209  		unsigned long pfn = page_to_pfn(folio_page(folio, 0));
a30b48bf1b244f1 Balbir Singh            2025-10-01  210  
a30b48bf1b244f1 Balbir Singh            2025-10-01  211  		migrate->src[migrate->npages] = migrate_pfn(pfn) | write
a30b48bf1b244f1 Balbir Singh            2025-10-01  212  						| MIGRATE_PFN_MIGRATE
a30b48bf1b244f1 Balbir Singh            2025-10-01  213  						| MIGRATE_PFN_COMPOUND;
a30b48bf1b244f1 Balbir Singh            2025-10-01  214  		migrate->dst[migrate->npages++] = 0;
a30b48bf1b244f1 Balbir Singh            2025-10-01  215  		migrate->cpages++;
a30b48bf1b244f1 Balbir Singh            2025-10-01  216  		ret = set_pmd_migration_entry(&pvmw, folio_page(folio, 0));
a30b48bf1b244f1 Balbir Singh            2025-10-01  217  		if (ret) {
a30b48bf1b244f1 Balbir Singh            2025-10-01  218  			migrate->npages--;
a30b48bf1b244f1 Balbir Singh            2025-10-01  219  			migrate->cpages--;
a30b48bf1b244f1 Balbir Singh            2025-10-01  220  			migrate->src[migrate->npages] = 0;
a30b48bf1b244f1 Balbir Singh            2025-10-01  221  			migrate->dst[migrate->npages] = 0;
a30b48bf1b244f1 Balbir Singh            2025-10-01  222  			goto fallback;
a30b48bf1b244f1 Balbir Singh            2025-10-01  223  		}
a30b48bf1b244f1 Balbir Singh            2025-10-01  224  		migrate_vma_collect_skip(start + PAGE_SIZE, end, walk);
a30b48bf1b244f1 Balbir Singh            2025-10-01  225  		spin_unlock(ptl);
a30b48bf1b244f1 Balbir Singh            2025-10-01  226  		return 0;
a30b48bf1b244f1 Balbir Singh            2025-10-01  227  	}
a30b48bf1b244f1 Balbir Singh            2025-10-01  228  
a30b48bf1b244f1 Balbir Singh            2025-10-01  229  fallback:
76cbbead253ddca Christoph Hellwig       2022-02-16  230  	spin_unlock(ptl);
a30b48bf1b244f1 Balbir Singh            2025-10-01  231  	if (!folio_test_large(folio))
a30b48bf1b244f1 Balbir Singh            2025-10-01  232  		goto done;
b002a7b0a58acea Matthew Wilcox (Oracle  2024-03-26  233) 	ret = split_folio(folio);
1afaeb8293c9add Matthew Brost           2025-03-05  234  	if (fault_folio != folio)
b002a7b0a58acea Matthew Wilcox (Oracle  2024-03-26  235) 		folio_unlock(folio);
b002a7b0a58acea Matthew Wilcox (Oracle  2024-03-26  236) 	folio_put(folio);
76cbbead253ddca Christoph Hellwig       2022-02-16  237  	if (ret)
a30b48bf1b244f1 Balbir Singh            2025-10-01  238  		return migrate_vma_collect_skip(start, end, walk);
a30b48bf1b244f1 Balbir Singh            2025-10-01  239  	if (pmd_none(pmdp_get_lockless(pmdp)))
a30b48bf1b244f1 Balbir Singh            2025-10-01  240  		return migrate_vma_collect_hole(start, end, -1, walk);
a30b48bf1b244f1 Balbir Singh            2025-10-01  241  
a30b48bf1b244f1 Balbir Singh            2025-10-01  242  done:
a30b48bf1b244f1 Balbir Singh            2025-10-01 @243  	return -ENOENT;
76cbbead253ddca Christoph Hellwig       2022-02-16  244  }
a30b48bf1b244f1 Balbir Singh            2025-10-01  245  

-- 
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki