On 04/04/2018 02:48, David Rientjes wrote:
> On Wed, 28 Mar 2018, Laurent Dufour wrote:
>> On 26/03/2018 00:10, David Rientjes wrote:
>>> On Wed, 21 Mar 2018, Laurent Dufour wrote:
>>>> I found the root cause of this lockdep warning.
>>>> In mmap_region(), unmap_region() may be called while vma_link() has not
>>>> called. This happens during the error path if call_mmap() failed.
>>>> The only to fix that particular case is to call
>>>> seqcount_init(&vma->vm_sequence) when initializing the vma in
>>> Ack, although that would require a fixup to dup_mmap() as well.
>> You're right, I'll fix that too.
> I also think the following is needed:
> diff --git a/fs/exec.c b/fs/exec.c
> --- a/fs/exec.c
> +++ b/fs/exec.c
> @@ -312,6 +312,10 @@ static int __bprm_mm_init(struct linux_binprm *bprm)
> vma->vm_flags = VM_SOFTDIRTY | VM_STACK_FLAGS |
> vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
> +#ifdef CONFIG_SPECULATIVE_PAGE_FAULT
> + seqcount_init(&vma->vm_sequence);
> + atomic_set(&vma->vm_ref_count, 0);
> err = insert_vm_struct(mm, vma);
> if (err)
No, this not needed because the vma is allocated with kmem_cache_zalloc() so
vm_ref_count is 0, and insert_vm_struc() will later call
__vma_link_rb() which will call seqcount_init().
Furhtermore, in case of error, the vma structure is freed without calling
get_vma() so there is risk of lockdep warning.