2024-02-09 06:53:06

by Anshuman Khandual

Subject: [PATCH] mm/hugetlb: Ensure adequate CMA areas available for hugetlb_cma[]

HugeTLB CMA area array is being created for possible MAX_NUMNODES without
ensuring corresponding MAX_CMA_AREAS support in CMA. Let's just warn for
such scenarios indicating need for CONFIG_CMA_AREAS adjustment.

Cc: Muchun Song <[email protected]>
Cc: Andrew Morton <[email protected]>
Cc: [email protected]
Cc: [email protected]
Signed-off-by: Anshuman Khandual <[email protected]>
---
This applies on v6.8-rc3

mm/hugetlb.c | 7 +++++++
1 file changed, 7 insertions(+)

diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index a188b7c60f4f..fda8ced69832 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -7750,6 +7750,13 @@ void __init hugetlb_cma_reserve(int order)
}

reserved = 0;
+
+ /*
+ * There needs to be enough MAX_CMA_AREAS to accommodate
+ * MAX_NUMNODES heap areas being created here. Otherwise
+ * adjust CONFIG_CMA_AREAS as required.
+ */
+ VM_WARN_ON(MAX_CMA_AREAS < MAX_NUMNODES);
for_each_online_node(nid) {
int res;
char name[CMA_MAX_NAME];
--
2.25.1

2024-02-09 22:18:45

by Andrew Morton

Subject: Re: [PATCH] mm/hugetlb: Ensure adequate CMA areas available for hugetlb_cma[]

On Fri, 9 Feb 2024 12:20:36 +0530 Anshuman Khandual <[email protected]> wrote:

> HugeTLB CMA area array is being created for possible MAX_NUMNODES without
> ensuring corresponding MAX_CMA_AREAS support in CMA. Let's just warn for
> such scenarios indicating need for CONFIG_CMA_AREAS adjustment.
>
> ...
>
> --- a/mm/hugetlb.c
> +++ b/mm/hugetlb.c
> @@ -7750,6 +7750,13 @@ void __init hugetlb_cma_reserve(int order)
> }
>
> reserved = 0;
> +
> + /*
> + * There needs to be enough MAX_CMA_AREAS to accommodate
> + * MAX_NUMNODES heap areas being created here. Otherwise
> + * adjust CONFIG_CMA_AREAS as required.
> + */
> + VM_WARN_ON(MAX_CMA_AREAS < MAX_NUMNODES);

Could this simply be fixed up in Kconfig logic?

And I think this could be detected at compile-time? BUILD_BUG_ON()?

> for_each_online_node(nid) {
> int res;
> char name[CMA_MAX_NAME];
> --
> 2.25.1

2024-02-12 02:07:14

by Anshuman Khandual

Subject: Re: [PATCH] mm/hugetlb: Ensure adequate CMA areas available for hugetlb_cma[]

On 2/10/24 03:46, Andrew Morton wrote:
> On Fri, 9 Feb 2024 12:20:36 +0530 Anshuman Khandual <[email protected]> wrote:
>
>> HugeTLB CMA area array is being created for possible MAX_NUMNODES without
>> ensuring corresponding MAX_CMA_AREAS support in CMA. Let's just warn for
>> such scenarios indicating need for CONFIG_CMA_AREAS adjustment.
>>
>> ...
>>
>> --- a/mm/hugetlb.c
>> +++ b/mm/hugetlb.c
>> @@ -7750,6 +7750,13 @@ void __init hugetlb_cma_reserve(int order)
>> }
>>
>> reserved = 0;
>> +
>> + /*
>> + * There needs to be enough MAX_CMA_AREAS to accommodate
>> + * MAX_NUMNODES heap areas being created here. Otherwise
>> + * adjust CONFIG_CMA_AREAS as required.
>> + */
>> + VM_WARN_ON(MAX_CMA_AREAS < MAX_NUMNODES);
>
> Could this simply be fixed up in Kconfig logic?

CMA_AREAS should default as (1 << NODES_SHIFT) ? But the system admin might want
to create more heap areas for other purposes as well. The idea here is to ensure
MAX_CMA_AREAS is at least MAX_NUMNODES if HugeTLB support is enabled. Do we have
some other methods ?

>
> And I think this could be detected at compile-time? BUILD_BUG_ON()?

Right, was thinking about this at first. Makes sense, will change here, seems to
be the right location for a build check as well.

>
>> for_each_online_node(nid) {
>> int res;
>> char name[CMA_MAX_NAME];
>> --
>> 2.25.1
>