[PATCH] ARM: mmu: Add debug_ll_io_init() mappings to early mappings

[sboyd@codeaurora.org (Stephen Boyd)]

On 06/14, Rob Herring wrote:
> On 06/12/2013 12:21 PM, Stephen Boyd wrote:
> > On 06/11, Rob Herring wrote:
> >> On Tue, Jun 11, 2013 at 8:30 PM, Stephen Boyd <sboyd@codeaurora.org> wrote:
> >>
> >>> diff --git a/arch/arm/mm/mmu.c b/arch/arm/mm/mmu.c
> >>> index e0d8565..04fe160 100644
> >>> --- a/arch/arm/mm/mmu.c
> >>> +++ b/arch/arm/mm/mmu.c
> >>> @@ -938,8 +938,12 @@ static void __init pci_reserve_io(void)
> >>>  #ifdef CONFIG_DEBUG_LL
> >>>  void __init debug_ll_io_init(void)
> >>>  {
> >>> +       struct vm_struct *vm;
> >>> +       struct static_vm *svm;
> >>>         struct map_desc map;
> >>>
> >>> +       svm = early_alloc_aligned(sizeof(*svm), __alignof__(*svm));
> >>> +
> >>>         debug_ll_addr(&map.pfn, &map.virtual);
> >>>         if (!map.pfn || !map.virtual)
> >>>                 return;
> >>> @@ -948,6 +952,15 @@ void __init debug_ll_io_init(void)
> >>>         map.length = PAGE_SIZE;
> >>>         map.type = MT_DEVICE;
> >>>         create_mapping(&map);
> >>> +
> >>> +       vm = &svm->vm;
> >>> +       vm->addr = (void *)map.virtual;
> >>> +       vm->size = map.length;
> >>> +       vm->phys_addr = __pfn_to_phys(map.pfn);
> >>> +       vm->flags = VM_IOREMAP | VM_ARM_STATIC_MAPPING;
> >>> +       vm->flags |= VM_ARM_MTYPE(map.type);
> >>> +       vm->caller = debug_ll_io_init;
> >>> +       add_static_vm_early(svm);
> >>
> >> Can you use vm_reserve_area_early here or perhaps just call
> >> iotable_init instead of create_mapping directly? I don't recall if
> >> there was some reason I didn't do that.
> >>
> > 
> > I had iotable_init() before but I thought it was better to have
> > the ->caller argument say debug_ll_io_init() instead of
> > iotable_init(). Shall I extract out the similar code?
> 
> Well, that's always welcome...
> 
> Another option would be use __builtin_return_address. That would change
> vmallocinfo from showing iotable_init to the caller of iotable_init.
> Arguably, knowing the caller would be better.
> 

Ok sounds fair. I'll use iotable_init() and send a follow-up
patch to make iotable_init() more informative.

-- 
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
hosted by The Linux Foundation