Re: [PATCH v2 3/4] mm: prevent KSM from completely breaking VMA merging

[David Hildenbrand]

On 21.05.25 20:20, Lorenzo Stoakes wrote:
> If a user wishes to enable KSM mergeability for an entire process and all
> fork/exec'd processes that come after it, they use the prctl()
> PR_SET_MEMORY_MERGE operation.
>
> This defaults all newly mapped VMAs to have the VM_MERGEABLE VMA flag set
> (in order to indicate they are KSM mergeable), as well as setting this flag
> for all existing VMAs.
>
> However it also entirely and completely breaks VMA merging for the process
> and all forked (and fork/exec'd) processes.
>
> This is because when a new mapping is proposed, the flags specified will
> never have VM_MERGEABLE set. However all adjacent VMAs will already have
> VM_MERGEABLE set, rendering VMAs unmergeable by default.
>
> To work around this, we try to set the VM_MERGEABLE flag prior to
> attempting a merge. In the case of brk() this can always be done.
>
> However on mmap() things are more complicated - while KSM is not supported
> for file-backed mappings, it is supported for MAP_PRIVATE file-backed
> mappings.
>
> And these mappings may have deprecated .mmap() callbacks specified which
> could, in theory, adjust flags and thus KSM eligiblity.
>
> This is unlikely to cause an issue on merge, as any adjacent file-backed
> mappings would already have the same post-.mmap() callback attributes, and
> thus would naturally not be merged.
>
> But for the purposes of establishing a VMA as KSM-eligible (as well as
> initially scanning the VMA), this is potentially very problematic.
>
> So we check to determine whether this at all possible. If not, we set
> VM_MERGEABLE prior to the merge attempt on mmap(), otherwise we retain the
> previous behaviour.
>
> When .mmap_prepare() is more widely used, we can remove this precaution.
>
> While this doesn't quite cover all cases, it covers a great many (all
> anonymous memory, for instance), meaning we should already see a
> significant improvement in VMA mergeability.
>
> Since, when it comes to file-backed mappings (other than shmem) we are
> really only interested in MAP_PRIVATE mappings which have an available anon
> page by default. Therefore, the VM_SPECIAL restriction makes less sense for
> KSM.
>
> In a future series we therefore intend to remove this limitation, which
> ought to simplify this implementation. However it makes sense to defer
> doing so until a later stage so we can first address this mergeability
> issue.
>
> Signed-off-by: Lorenzo Stoakes <lorenzo.stoakes@xxxxxxxxxx>
> Fixes: d7597f59d1d3 ("mm: add new api to enable ksm per process") # please no backport!
> Reviewed-by: Chengming Zhou <chengming.zhou@xxxxxxxxx>


Acked-by: David Hildenbrand <david@xxxxxxxxxx>


--
Cheers,

David / dhildenb