Re: Mapping PADDR into virtual memory problem.

[Rob Gardner <rob.gardner@oracle.com>]

On 07/13/2016 03:25 PM, joe moriarty wrote:
> I'm stuck trying to get this to work and I don't know why what I am 
> doing doesn't work.  I'm hoping maybe someone on here has a 
> suggestion.  The piece of code below is suppose to take a PADDR and 
> memory map it to a kernel address.  The output I get from the code 
> below is shown here:
>
> PCI: Enabling device: (0002:08:00.0), cmd 3
> vm_area->phys_addr=0x0008080100040000
> vaddr=0x0000000136180000
> nr_pages2
> page protection =0x9C08080100040690
> I/O address location mapped
> ioread32 RegBase[0] = 0x00000000
> vread RegBase[0] = 0x00000000
> PCI: Enabling device: (0002:09:00.0), cmd 3
> vm_area->phys_addr=0x0008080110040000
> vaddr=0x0000000136200000
> nr_pages2
> page protection =0x9C08080110040690
> I/O address location mapped
> ioread32 RegBase[0] = 0x00000000
> vread RegBase[0] = 0x00000000
> PCI: Enabling device: (0002:0a:00.0), cmd 3
> vm_area->phys_addr=0x0008080120040000
> vaddr=0x0000000136280000
> nr_pages2
> page protection =0x9C08080120040690
> I/O address location mapped
> ioread32 RegBase[0] = 0x00000000
> vread RegBase[0] = 0x00000000
> PCI: Enabling device: (0002:0b:00.0), cmd 3
> vm_area->phys_addr=0x0008080130040000
> vaddr=0x0000000136300000
> nr_pages2
> page protection =0x9C08080130040690
> I/O address location mapped
> ioread32 RegBase[0] = 0x00000000
> vread RegBase[0] = 0x00000000
>
> I know I should be getting 0x00EA0408 from RegBase[0].
> Do you see anything obvious that is wrong?
>
>  void __iomem *ioremap_prefetch(unsigned long offset, unsigned long size)
> {
>         struct vm_struct *vm_area = NULL;
>         unsigned long vaddr;
>         int error = 0, tmp_val;
>         pgprot_t prot = __pgprot(pgprot_val(PAGE_SHARED) | 
> (_PAGE_PADDR_4V &
>                         offset));


I don't understand why offset is put into prot. And maybe just use 
PAGE_KERNEL here?



>         void __iomem *ret_addr;
>
>         /*
>          * Get Contiguous Kernel Virtual Memory Area.  Note: 
> get_vm_area will
>          * page align the area to use.
>          */
>         vm_area = get_vm_area(size, VM_IOREMAP | VM_NO_GUARD);
>         if (!vm_area) {
>                 printk("Unable to allocate Contiguous Virtual Memory 
> for Mapping
> \n");
>                 error = 1;
>                 goto done;
>         }
>         vm_area->phys_addr = (phys_addr_t)offset;
>         vm_area->nr_pages = size / PAGE_SIZE;
>         vaddr = (unsigned long)vm_area->addr;
>         printk("vm_area->phys_addr=0x%016llX\n", (unsigned long 
> long)vm_area->ph
> ys_addr);
>         printk("vaddr=0x%016llX\n", (unsigned long long)vaddr);
>         printk("nr_pages=%d\n", vm_area->nr_pages);
>         printk("page protection =0x%08lX\n", (unsigned 
> long)pgprot_val(prot));
>         if(ioremap_page_range((unsigned long)vm_area->addr,
>                               (unsigned long)vm_area->addr + size,
>                               vm_area->phys_addr, prot)) {
>                 printk("Unable to Map Page Range.\n");
>                 error = 1;
>                 goto done;
>         }
>         else {
>                 printk("I/O address location mapped\n");
>                 ret_addr = vm_area->addr;
>                 printk("ioread32 RegBase[0] = 0x%08X\n", 
> ioread32(ret_addr));


ioread32 I think assumes the address you are reading from is physical.


>                 vread((char *)&tmp_val, ret_addr, 4);


vread explicitly returns zeros if the vm area was created with VM_IOREMAP.


Since you seem to have a virtual address now that maps to your desired 
physical address, why can't you simply dereference that virtual address?

ie,

   x = *(int *)ret_addr;



Rob






>                 printk("vread RegBase[0] = 0x%08X\n", tmp_val);
>         }
>
> done:
>         if(error) {
>                 /*
>                  * Cleanup After oneself
>                  */
>                 free_vm_area(vm_area);
>                 ret_addr = 0;
>         }
>         return (ret_addr);
>         //return ((void __iomem *)offset);
> }
>
> Thanks In Advance,
> Joe
> -- 
> To unsubscribe from this list: send the line "unsubscribe sparclinux" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html