On 26.11.2015 18:08, Vladimir Davydov wrote:
> mmap_region workflow regarding locked pages was changed during rebase
> from RH6 to RH7. The patch porting locked pages accounting did not allow
> for that.
>
> Presumably, this should fix UB_LOCKEDPAGES leak on CT stop:
>
> Ub 101 helds 8 in lockedpages on put
> UB: leaked beancounter 101 (ffff880404e95000)
> Tainting kernel with flag 0xa
>
> https://jira.sw.ru/browse/PSBM-39435
>
> Signed-off-by: Vladimir Davydov <vdavy...@virtuozzo.com>
Reviewed-by: Kirill Tkhai <ktk...@odin.com>
> ---
> mm/mmap.c | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/mm/mmap.c b/mm/mmap.c
> index 8796ed643a6f..b981d80ab100 100644
> --- a/mm/mmap.c
> +++ b/mm/mmap.c
> @@ -1654,8 +1654,10 @@ unsigned long mmap_region(struct file *file, unsigned
> long addr,
> if (!((vm_flags & VM_SPECIAL) || is_vm_hugetlb_page(vma) ||
> vma == get_gate_vma(current->mm)))
> mm->locked_vm += (len >> PAGE_SHIFT);
> - else
> + else {
> vma->vm_flags &= ~VM_LOCKED;
> + ub_locked_uncharge(mm, len);
> + }
> }
>
> if (file)
>
_______________________________________________
Devel mailing list
Devel@openvz.org
https://lists.openvz.org/mailman/listinfo/devel
