Re: [PATCH] x86: add missing free_memtype if get_vm_area_caller failed

[Ingo Molnar <mingo@elte.hu>]

* Xiaotian Feng <dfeng@redhat.com> wrote:

> When __ioremap_caller goes into get_vm_area, kernel already reserved
> memtype. so if get_vm_area fails, we need to free it before return.
> 
> Signed-off-by: Xiaotian Feng <dfeng@redhat.com>
> Cc: Thomas Gleixner <tglx@linutronix.de>
> Cc: Ingo Molnar <mingo@redhat.com>
> Cc: H. Peter Anvin <hpa@zytor.com>
> Cc: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
> Cc: Suresh Siddha <suresh.b.siddha@intel.com>
> ---
>  arch/x86/mm/ioremap.c |    4 +++-
>  1 files changed, 3 insertions(+), 1 deletions(-)
> 
> diff --git a/arch/x86/mm/ioremap.c b/arch/x86/mm/ioremap.c
> index 334e63c..7859f77 100644
> --- a/arch/x86/mm/ioremap.c
> +++ b/arch/x86/mm/ioremap.c
> @@ -196,8 +196,10 @@ static void __iomem *__ioremap_caller(resource_size_t phys_addr,
>  	 * Ok, go for it..
>  	 */
>  	area = get_vm_area_caller(size, VM_IOREMAP, caller);
> -	if (!area)
> +	if (!area) {
> +		free_memtype(phys_addr, phys_addr + size);
>  		return NULL;
> +	}
>  	area->phys_addr = phys_addr;
>  	vaddr = (unsigned long) area->addr;

Nice one!

Mind structuring this fix in a bit different way please?

The main cause of this bug is the following unclean resource-teardown 
pattern in __ioremap_caller():

	area = get_vm_area_caller(size, VM_IOREMAP, caller);
	if (!area)
		return NULL;
	area->phys_addr = phys_addr;
	vaddr = (unsigned long) area->addr;

	if (kernel_map_sync_memtype(phys_addr, size, prot_val)) {
		free_memtype(phys_addr, phys_addr + size);
		free_vm_area(area);
		return NULL;
	}

	if (ioremap_page_range(vaddr, vaddr + size, phys_addr, prot)) {
		free_memtype(phys_addr, phys_addr + size);
		free_vm_area(area);
		return NULL;
	}

see how repetitive it is, and how the return sequence is duplicated?

The way we do this is to add an error path to the tail of the function:

err_free_area:
	free_vm_area(area);
err_free_memtype:
	free_memtype(phys_addr, phys_addr + size);
	return NULL;

and changing the failure branches to:

	if (!area)
		goto err_free_memtype;
	...

	if (ioremap_page_range(vaddr, vaddr + size, phys_addr, prot))
		goto err_free_area;

etc.

Ok?
	
	Ingo