Re: [v6 PATCH 4/5] iommu: Support mm PASID 1:n with sva domains

[Jason Gunthorpe <jgg@nvidia.com>]

On Wed, Oct 11, 2023 at 02:51:31PM +0800, Tina Zhang wrote:

> diff --git a/kernel/fork.c b/kernel/fork.c
> index 3b6d20dfb9a8..985403a7a747 100644
> --- a/kernel/fork.c
> +++ b/kernel/fork.c
> @@ -1277,7 +1277,6 @@ static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p,
>  	mm_init_cpumask(mm);
>  	mm_init_aio(mm);
>  	mm_init_owner(mm, p);
> -	mm_pasid_init(mm);
>  	RCU_INIT_POINTER(mm->exe_file, NULL);
>  	mmu_notifier_subscriptions_init(mm);
>  	init_tlb_flush_pending(mm);

Nicolin debugged his crash report last night and sent me the details.

This hunk is the cause of the bug that Nicolin reported.

The dup_mm() flow does:

static struct mm_struct *dup_mm(struct task_struct *tsk,
				struct mm_struct *oldmm)
{
	struct mm_struct *mm;
	int err;

	mm = allocate_mm();
	if (!mm)
		goto fail_nomem;

	memcpy(mm, oldmm, sizeof(*mm));

	if (!mm_init(mm, tsk, mm->user_ns))
		goto fail_nomem;

It is essential that mm_pasid_init() zero the new pointer otherwise,
due to the memcpy, after a fork two mm structs will point to the same
thing and one will UAF/doube free.

Keep mm_pasid_init() and add zeroing the new pointer to it.

Jason