[PATCH] mm/migrate: initialize pud_entry in migrate_vma()

[PATCH] mm/migrate: initialize pud_entry in migrate_vma()

[Ralph Campbell]

When CONFIG_MIGRATE_VMA_HELPER is enabled, migrate_vma() calls
migrate_vma_collect() which initializes a struct mm_walk but
didn't initialize mm_walk.pud_entry. (Found by code inspection)
Use a C structure initialization to make sure it is set to NULL.

Fixes: 8763cb45ab967 ("mm/migrate: new memory migration helper for use with
device memory")
Cc: stable@vger.kernel.org
Signed-off-by: Ralph Campbell <rcampbell@nvidia.com>
Cc: "Jérôme Glisse" <jglisse@redhat.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
---
 mm/migrate.c | 17 +++++++----------
 1 file changed, 7 insertions(+), 10 deletions(-)

diff --git a/mm/migrate.c b/mm/migrate.c
index 515718392b24..a42858d8e00b 100644
--- a/mm/migrate.c
+++ b/mm/migrate.c
@@ -2340,16 +2340,13 @@ static int migrate_vma_collect_pmd(pmd_t *pmdp,
 static void migrate_vma_collect(struct migrate_vma *migrate)
 {
 	struct mmu_notifier_range range;
-	struct mm_walk mm_walk;
-
-	mm_walk.pmd_entry = migrate_vma_collect_pmd;
-	mm_walk.pte_entry = NULL;
-	mm_walk.pte_hole = migrate_vma_collect_hole;
-	mm_walk.hugetlb_entry = NULL;
-	mm_walk.test_walk = NULL;
-	mm_walk.vma = migrate->vma;
-	mm_walk.mm = migrate->vma->vm_mm;
-	mm_walk.private = migrate;
+	struct mm_walk mm_walk = {
+		.pmd_entry = migrate_vma_collect_pmd,
+		.pte_hole = migrate_vma_collect_hole,
+		.vma = migrate->vma,
+		.mm = migrate->vma->vm_mm,
+		.private = migrate,
+	};
 
 	mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, NULL, mm_walk.mm,
 				migrate->start,
-- 
2.20.1

Re: [PATCH] mm/migrate: initialize pud_entry in migrate_vma()

[John Hubbard]

On 7/19/19 4:32 PM, Ralph Campbell wrote:
> When CONFIG_MIGRATE_VMA_HELPER is enabled, migrate_vma() calls
> migrate_vma_collect() which initializes a struct mm_walk but
> didn't initialize mm_walk.pud_entry. (Found by code inspection)
> Use a C structure initialization to make sure it is set to NULL.
> 
> Fixes: 8763cb45ab967 ("mm/migrate: new memory migration helper for use with
> device memory")
> Cc: stable@vger.kernel.org
> Signed-off-by: Ralph Campbell <rcampbell@nvidia.com>
> Cc: "Jérôme Glisse" <jglisse@redhat.com>
> Cc: Andrew Morton <akpm@linux-foundation.org>
> ---
>  mm/migrate.c | 17 +++++++----------
>  1 file changed, 7 insertions(+), 10 deletions(-)
> 
> diff --git a/mm/migrate.c b/mm/migrate.c
> index 515718392b24..a42858d8e00b 100644
> --- a/mm/migrate.c
> +++ b/mm/migrate.c
> @@ -2340,16 +2340,13 @@ static int migrate_vma_collect_pmd(pmd_t *pmdp,
>  static void migrate_vma_collect(struct migrate_vma *migrate)
>  {
>  	struct mmu_notifier_range range;
> -	struct mm_walk mm_walk;
> -
> -	mm_walk.pmd_entry = migrate_vma_collect_pmd;
> -	mm_walk.pte_entry = NULL;
> -	mm_walk.pte_hole = migrate_vma_collect_hole;
> -	mm_walk.hugetlb_entry = NULL;
> -	mm_walk.test_walk = NULL;
> -	mm_walk.vma = migrate->vma;
> -	mm_walk.mm = migrate->vma->vm_mm;
> -	mm_walk.private = migrate;
> +	struct mm_walk mm_walk = {
> +		.pmd_entry = migrate_vma_collect_pmd,
> +		.pte_hole = migrate_vma_collect_hole,
> +		.vma = migrate->vma,
> +		.mm = migrate->vma->vm_mm,
> +		.private = migrate,
> +	};

Neatly done.

>  
>  	mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, NULL, mm_walk.mm,
>  				migrate->start,
> 

Reviewed-by: John Hubbard <jhubbard@nvidia.com>

thanks,
-- 
John Hubbard
NVIDIA

Re: [PATCH] mm/migrate: initialize pud_entry in migrate_vma()

[Sasha Levin]

Hi,

[This is an automated email]

This commit has been processed because it contains a "Fixes:" tag,
fixing commit: 8763cb45ab96 mm/migrate: new memory migration helper for use with device memory.

The bot has tested the following trees: v5.2.1, v5.1.18, v4.19.59, v4.14.133.

v5.2.1: Build OK!
v5.1.18: Build OK!
v4.19.59: Failed to apply! Possible dependencies:
    41b4deeaa123 ("RDMA/umem: Make ib_umem_odp into a sub structure of ib_umem")
    597ecc5a0954 ("RDMA/umem: Get rid of struct ib_umem.odp_data")
    5d6527a784f7 ("mm/mmu_notifier: use structure for invalidate_range_start/end callback")
    ac46d4f3c432 ("mm/mmu_notifier: use structure for invalidate_range_start/end calls v2")
    b5231b019d76 ("RDMA/umem: Use ib_umem_odp in all function signatures connected to ODP")
    c9990ab39b6e ("RDMA/umem: Move all the ODP related stuff out of ucontext and into per_mm")
    d4b4dd1b9706 ("RDMA/umem: Do not use current->tgid to track the mm_struct")

v4.14.133: Failed to apply! Possible dependencies:
    155494dbbbf4 ("drm/amdgpu: Update kgd2kfd_shared_resources for dGPU support")
    179c02fe90a4 ("drm/tve200: Add new driver for TVE200")
    1b0c0f9dc5ca ("drm/amdgpu: move userptr BOs to CPU domain during CS v2")
    1b1f42d8fde4 ("drm: move amd_gpu_scheduler into common location")
    1ed3d2567c80 ("drm/amdgpu: keep the MMU lock until the update ends v4")
    3fe89771cb0a ("drm/amdgpu: stop reserving the BO in the MMU callback v3")
    4c660c8fbbf7 ("drm/amdgpu: Add submit IB function for KFD")
    528e083d85bd ("drm/amdgpu: rename rmn to amn in the MMU notifier code (v2)")
    60de1c1740f3 ("drm/amdgpu: use a rw_semaphore for MMU notifiers")
    8cce58fe698a ("drm/amd: add new interface to query cu info")
    93065ac753e4 ("mm, oom: distinguish blockable mode for mmu notifiers")
    9f0a0b41ffcc ("drm/amdgpu: Add support for reporting VRAM usage")
    a216ab09955d ("drm/amdgpu: fix userptr put_page handling")
    a46a2cd103a8 ("drm/amdgpu: Add GPUVM memory management functions for KFD")
    ac46d4f3c432 ("mm/mmu_notifier: use structure for invalidate_range_start/end calls v2")
    b72cf4fca2bb ("drm/amdgpu: move taking mmap_sem into get_user_pages v2")
    ca666a3c298f ("drm/amdgpu: stop using BO status for user pages")
    d8d019ccffb8 ("drm/amdgpu: Add KFD eviction fence")
    e52482dec836 ("drm/amdgpu: Add MMU notifier type for KFD userptr")
    ebdebf428ae6 ("drm/amdgpu: add amdgpu interface to query cu info")


NOTE: The patch will not be queued to stable trees until it is upstream.

How should we proceed with this patch?

--
Thanks,
Sasha

Re: [PATCH] mm/migrate: initialize pud_entry in migrate_vma()

[Vlastimil Babka]

On 7/20/19 1:32 AM, Ralph Campbell wrote:
> When CONFIG_MIGRATE_VMA_HELPER is enabled, migrate_vma() calls
> migrate_vma_collect() which initializes a struct mm_walk but
> didn't initialize mm_walk.pud_entry. (Found by code inspection)
> Use a C structure initialization to make sure it is set to NULL.
> 
> Fixes: 8763cb45ab967 ("mm/migrate: new memory migration helper for use with
> device memory")
> Cc: stable@vger.kernel.org
> Signed-off-by: Ralph Campbell <rcampbell@nvidia.com>
> Cc: "Jérôme Glisse" <jglisse@redhat.com>
> Cc: Andrew Morton <akpm@linux-foundation.org>

So this bug can manifest by some garbage address on stack being called, right? I
wonder, how comes it didn't actually happen yet?

> ---
>  mm/migrate.c | 17 +++++++----------
>  1 file changed, 7 insertions(+), 10 deletions(-)
> 
> diff --git a/mm/migrate.c b/mm/migrate.c
> index 515718392b24..a42858d8e00b 100644
> --- a/mm/migrate.c
> +++ b/mm/migrate.c
> @@ -2340,16 +2340,13 @@ static int migrate_vma_collect_pmd(pmd_t *pmdp,
>  static void migrate_vma_collect(struct migrate_vma *migrate)
>  {
>  	struct mmu_notifier_range range;
> -	struct mm_walk mm_walk;
> -
> -	mm_walk.pmd_entry = migrate_vma_collect_pmd;
> -	mm_walk.pte_entry = NULL;
> -	mm_walk.pte_hole = migrate_vma_collect_hole;
> -	mm_walk.hugetlb_entry = NULL;
> -	mm_walk.test_walk = NULL;
> -	mm_walk.vma = migrate->vma;
> -	mm_walk.mm = migrate->vma->vm_mm;
> -	mm_walk.private = migrate;
> +	struct mm_walk mm_walk = {
> +		.pmd_entry = migrate_vma_collect_pmd,
> +		.pte_hole = migrate_vma_collect_hole,
> +		.vma = migrate->vma,
> +		.mm = migrate->vma->vm_mm,
> +		.private = migrate,
> +	};
>  
>  	mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, NULL, mm_walk.mm,
>  				migrate->start,
> 

Re: [PATCH] mm/migrate: initialize pud_entry in migrate_vma()

[Ralph Campbell]

On 8/26/19 8:11 AM, Vlastimil Babka wrote:
> On 7/20/19 1:32 AM, Ralph Campbell wrote:
>> When CONFIG_MIGRATE_VMA_HELPER is enabled, migrate_vma() calls
>> migrate_vma_collect() which initializes a struct mm_walk but
>> didn't initialize mm_walk.pud_entry. (Found by code inspection)
>> Use a C structure initialization to make sure it is set to NULL.
>>
>> Fixes: 8763cb45ab967 ("mm/migrate: new memory migration helper for use with
>> device memory")
>> Cc: stable@vger.kernel.org
>> Signed-off-by: Ralph Campbell <rcampbell@nvidia.com>
>> Cc: "Jérôme Glisse" <jglisse@redhat.com>
>> Cc: Andrew Morton <akpm@linux-foundation.org>
> 
> So this bug can manifest by some garbage address on stack being called, right? I
> wonder, how comes it didn't actually happen yet?

Right.
Probably because HMM isn't widely being used in production yet.

> 
>> ---
>>   mm/migrate.c | 17 +++++++----------
>>   1 file changed, 7 insertions(+), 10 deletions(-)
>>
>> diff --git a/mm/migrate.c b/mm/migrate.c
>> index 515718392b24..a42858d8e00b 100644
>> --- a/mm/migrate.c
>> +++ b/mm/migrate.c
>> @@ -2340,16 +2340,13 @@ static int migrate_vma_collect_pmd(pmd_t *pmdp,
>>   static void migrate_vma_collect(struct migrate_vma *migrate)
>>   {
>>   	struct mmu_notifier_range range;
>> -	struct mm_walk mm_walk;
>> -
>> -	mm_walk.pmd_entry = migrate_vma_collect_pmd;
>> -	mm_walk.pte_entry = NULL;
>> -	mm_walk.pte_hole = migrate_vma_collect_hole;
>> -	mm_walk.hugetlb_entry = NULL;
>> -	mm_walk.test_walk = NULL;
>> -	mm_walk.vma = migrate->vma;
>> -	mm_walk.mm = migrate->vma->vm_mm;
>> -	mm_walk.private = migrate;
>> +	struct mm_walk mm_walk = {
>> +		.pmd_entry = migrate_vma_collect_pmd,
>> +		.pte_hole = migrate_vma_collect_hole,
>> +		.vma = migrate->vma,
>> +		.mm = migrate->vma->vm_mm,
>> +		.private = migrate,
>> +	};
>>   
>>   	mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, NULL, mm_walk.mm,
>>   				migrate->start,
>>
>