Re: [PATCH hotfix 6.12 4/8] mm: resolve faulty mmap_region() error path behaviour

[Vlastimil Babka <vbabka@suse.cz>]

On 10/22/24 22:40, Lorenzo Stoakes wrote:
> The mmap_region() function is somewhat terrifying, with spaghetti-like
> control flow and numerous means by which issues can arise and incomplete
> state, memory leaks and other unpleasantness can occur.
> 
> A large amount of the complexity arises from trying to handle errors late
> in the process of mapping a VMA, which forms the basis of recently observed
> issues with resource leaks and observable inconsistent state.
> 
> Taking advantage of previous patches in this series we move a number of
> checks earlier in the code, simplifying things by moving the core of the
> logic into a static internal function __mmap_region().
> 
> Doing this allows us to perform a number of checks up front before we do
> any real work, and allows us to unwind the writable unmap check
> unconditionally as required and to perform a CONFIG_DEBUG_VM_MAPLE_TREE
> validation unconditionally also.
> 
> We move a number of things here:
> 
> 1. We preallocate memory for the iterator before we call the file-backed
>    memory hook, allowing us to exit early and avoid having to perform
>    complicated and error-prone close/free logic. We carefully free
>    iterator state on both success and error paths.
> 
> 2. The enclosing mmap_region() function handles the mapping_map_writable()
>    logic early. Previously the logic had the mapping_map_writable() at the
>    point of mapping a newly allocated file-backed VMA, and a matching
>    mapping_unmap_writable() on success and error paths.
> 
>    We now do this unconditionally if this is a file-backed, shared writable
>    mapping. If a driver changes the flags to eliminate VM_MAYWRITE, however
>    doing so does not invalidate the seal check we just performed, and we in
>    any case always decrement the counter in the wrapper.
> 
>    We perform a debug assert to ensure a driver does not attempt to do the
>    opposite.
> 
> 3. We also move arch_validate_flags() up into the mmap_region()
>    function. This is only relevant on arm64 and sparc64, and the check is
>    only meaningful for SPARC with ADI enabled. We explicitly add a warning
>    for this arch if a driver invalidates this check, though the code ought
>    eventually to be fixed to eliminate the need for this.
> 
> With all of these measures in place, we no longer need to explicitly close
> the VMA on error paths, as we place all checks which might fail prior to a
> call to any driver mmap hook.
> 
> This eliminates an entire class of errors, makes the code easier to reason
> about and more robust.
> 
> Reported-by: Jann Horn <jannh@google.com>
> Fixes: deb0f6562884 ("mm/mmap: undo ->mmap() when arch_validate_flags() fails")
> Cc: stable <stable@kernel.org>
> Signed-off-by: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>

Reviewed-by: Vlastimil Babka <vbabka@suse.cz>

some nits below

> ---
>  mm/mmap.c | 120 ++++++++++++++++++++++++++++++------------------------
>  1 file changed, 66 insertions(+), 54 deletions(-)
> 
> diff --git a/mm/mmap.c b/mm/mmap.c
> index 66edf0ebba94..7d02b47a1895 100644
> --- a/mm/mmap.c
> +++ b/mm/mmap.c
> @@ -1361,20 +1361,18 @@ int do_munmap(struct mm_struct *mm, unsigned long start, size_t len,
>  	return do_vmi_munmap(&vmi, mm, start, len, uf, false);
>  }
> 
> -unsigned long mmap_region(struct file *file, unsigned long addr,
> +static unsigned long __mmap_region(struct file *file, unsigned long addr,
>  		unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
>  		struct list_head *uf)
>  {
>  	struct mm_struct *mm = current->mm;
>  	struct vm_area_struct *vma = NULL;
>  	pgoff_t pglen = PHYS_PFN(len);
> -	struct vm_area_struct *merge;
>  	unsigned long charged = 0;
>  	struct vma_munmap_struct vms;
>  	struct ma_state mas_detach;
>  	struct maple_tree mt_detach;
>  	unsigned long end = addr + len;
> -	bool writable_file_mapping = false;
>  	int error;
>  	VMA_ITERATOR(vmi, mm, addr);
>  	VMG_STATE(vmg, mm, &vmi, addr, end, vm_flags, pgoff);
> @@ -1448,28 +1446,26 @@ unsigned long mmap_region(struct file *file, unsigned long addr,
>  	vm_flags_init(vma, vm_flags);
>  	vma->vm_page_prot = vm_get_page_prot(vm_flags);
> 
> +	if (vma_iter_prealloc(&vmi, vma)) {
> +		error = -ENOMEM;
> +		goto free_vma;
> +	}
> +
>  	if (file) {
>  		vma->vm_file = get_file(file);
>  		error = mmap_file(file, vma);
>  		if (error)
> -			goto unmap_and_free_vma;
> -
> -		if (vma_is_shared_maywrite(vma)) {
> -			error = mapping_map_writable(file->f_mapping);
> -			if (error)
> -				goto close_and_free_vma;
> -
> -			writable_file_mapping = true;
> -		}
> +			goto unmap_and_free_file_vma;
> 
> +		/* Drivers cannot alter the address of the VMA. */
> +		WARN_ON_ONCE(addr != vma->vm_start);
>  		/*
> -		 * Expansion is handled above, merging is handled below.
> -		 * Drivers should not alter the address of the VMA.
> +		 * Drivers should not permit writability when previously it was
> +		 * disallowed.
>  		 */
> -		if (WARN_ON((addr != vma->vm_start))) {
> -			error = -EINVAL;
> -			goto close_and_free_vma;
> -		}
> +		VM_WARN_ON_ONCE(vm_flags != vma->vm_flags &&
> +				!(vm_flags & VM_MAYWRITE) &&
> +				(vma->vm_flags & VM_MAYWRITE));
> 
>  		vma_iter_config(&vmi, addr, end);

I wonder if this one could be removed, earlier above we did the same config
and neither parameters changed? But it was true before this patch as well,
and maybe it's further refactored away later in the series, just noting.

>  		/*
> @@ -1477,6 +1473,8 @@ unsigned long mmap_region(struct file *file, unsigned long addr,
>  		 * vma again as we may succeed this time.
>  		 */
>  		if (unlikely(vm_flags != vma->vm_flags && vmg.prev)) {
> +			struct vm_area_struct *merge;
> +
>  			vmg.flags = vma->vm_flags;
>  			/* If this fails, state is reset ready for a reattempt. */
>  			merge = vma_merge_new_range(&vmg);
> @@ -1491,10 +1489,11 @@ unsigned long mmap_region(struct file *file, unsigned long addr,
>  				 */
>  				fput(vma->vm_file);
>  				vm_area_free(vma);
> +				vma_iter_free(&vmi);

If we merged successfully, I think this is not necessary? But doesn't hurt?

>  				vma = merge;
>  				/* Update vm_flags to pick up the change. */
>  				vm_flags = vma->vm_flags;
> -				goto unmap_writable;
> +				goto file_expanded;
>  			}
>  			vma_iter_config(&vmi, addr, end);
>  		}
> @@ -1503,26 +1502,15 @@ unsigned long mmap_region(struct file *file, unsigned long addr,
>  	} else if (vm_flags & VM_SHARED) {
>  		error = shmem_zero_setup(vma);
>  		if (error)
> -			goto free_vma;
> +			goto free_iter_vma;
>  	} else {
>  		vma_set_anonymous(vma);
>  	}
> 
> -	if (map_deny_write_exec(vma->vm_flags, vma->vm_flags)) {
> -		error = -EACCES;
> -		goto close_and_free_vma;
> -	}
> -
> -	/* Allow architectures to sanity-check the vm_flags */
> -	if (!arch_validate_flags(vma->vm_flags)) {
> -		error = -EINVAL;
> -		goto close_and_free_vma;
> -	}
> -
> -	if (vma_iter_prealloc(&vmi, vma)) {
> -		error = -ENOMEM;
> -		goto close_and_free_vma;
> -	}
> +#ifdef CONFIG_SPARC64
> +	/* TODO: Fix SPARC ADI! */
> +	WARN_ON_ONCE(!arch_validate_flags(vm_flags));
> +#endif
> 
>  	/* Lock the VMA since it is modified after insertion into VMA tree */
>  	vma_start_write(vma);
> @@ -1536,10 +1524,7 @@ unsigned long mmap_region(struct file *file, unsigned long addr,
>  	 */
>  	khugepaged_enter_vma(vma, vma->vm_flags);
> 
> -	/* Once vma denies write, undo our temporary denial count */
> -unmap_writable:
> -	if (writable_file_mapping)
> -		mapping_unmap_writable(file->f_mapping);
> +file_expanded:
>  	file = vma->vm_file;
>  	ksm_add_vma(vma);
>  expanded:
> @@ -1572,23 +1557,17 @@ unsigned long mmap_region(struct file *file, unsigned long addr,
> 
>  	vma_set_page_prot(vma);
> 
> -	validate_mm(mm);
>  	return addr;
> 
> -close_and_free_vma:
> -	vma_close(vma);
> -
> -	if (file || vma->vm_file) {
> -unmap_and_free_vma:
> -		fput(vma->vm_file);
> -		vma->vm_file = NULL;
> +unmap_and_free_file_vma:
> +	fput(vma->vm_file);
> +	vma->vm_file = NULL;
> 
> -		vma_iter_set(&vmi, vma->vm_end);
> -		/* Undo any partial mapping done by a device driver. */
> -		unmap_region(&vmi.mas, vma, vmg.prev, vmg.next);
> -	}
> -	if (writable_file_mapping)
> -		mapping_unmap_writable(file->f_mapping);
> +	vma_iter_set(&vmi, vma->vm_end);
> +	/* Undo any partial mapping done by a device driver. */
> +	unmap_region(&vmi.mas, vma, vmg.prev, vmg.next);
> +free_iter_vma:
> +	vma_iter_free(&vmi);
>  free_vma:
>  	vm_area_free(vma);
>  unacct_error:
> @@ -1598,10 +1577,43 @@ unsigned long mmap_region(struct file *file, unsigned long addr,
>  abort_munmap:
>  	vms_abort_munmap_vmas(&vms, &mas_detach);
>  gather_failed:
> -	validate_mm(mm);
>  	return error;
>  }
> 
> +unsigned long mmap_region(struct file *file, unsigned long addr,
> +			  unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
> +			  struct list_head *uf)
> +{
> +	unsigned long ret;
> +	bool writable_file_mapping = false;
> +
> +	/* Allow architectures to sanity-check the vm_flags. */
> +	if (!arch_validate_flags(vm_flags))
> +		return -EINVAL;
> +
> +	/* Check to see if MDWE is applicable. */
> +	if (map_deny_write_exec(vm_flags, vm_flags))
> +		return -EACCES;

The two checks above used to be in the opposite order. Can we keep that just
to be sure we don't change user observable behavior unnecessarily?

> +	/* Map writable and ensure this isn't a sealed memfd. */
> +	if (file && is_shared_maywrite(vm_flags)) {
> +		int error = mapping_map_writable(file->f_mapping);
> +
> +		if (error)
> +			return error;
> +		writable_file_mapping = true;
> +	}
> +
> +	ret = __mmap_region(file, addr, len, vm_flags, pgoff, uf);
> +
> +	/* Clear our write mapping regardless of error. */
> +	if (writable_file_mapping)
> +		mapping_unmap_writable(file->f_mapping);
> +
> +	validate_mm(current->mm);
> +	return ret;
> +}
> +
>  static int __vm_munmap(unsigned long start, size_t len, bool unlock)
>  {
>  	int ret;
> --
> 2.47.0