Question about ioremap and io_remap_page_range

Question about ioremap and io_remap_page_range

[舒国强]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain; charset="GB2312", Size: 952 bytes --]


 Here is some rather basic questions I want ask ,any reply or comment please
 CC to my emailbox,thank you very much.
 
 When I work with kernel 2.4.2 in Intel X86 , I use 
		
     VIRT_ADDR = ioremap(BUS_ADDR); to map a section of PCI memory, and
     X_ADDR = virt_to_phys(VIRT_ADDR);

  I think in x86 platform X_ADDR should equel with BUS_ADDR, but it turns to be
 NOT, can you explain ?

  In X86 platform ,Can I use return value of ioremap() as a memory pointer?


  I use io_remap_page_range(BUS_ADDR,,) in mmap() function,but the result is 
 that the memory I map is READ ONLY from user space,when I try to write to it,
 a "do_sw_pg. bogus page(XXXXXXXX)" appears. Can you explain to me?


  I know these questions are childish,but I need the answer urgently,thanks!


 George Shu
 

ý:.žË›±Êâmçë¢kaŠÉb²ßìzwm…ébïîžË›±Êâmébžìÿ‘êçz_âžØ^n‡r¡ö¦zË\x1aëh™¨è­Ú&£ûàz¿äz¹Þ—ú+€Ê+zf£¢·hšˆ§~†­†Ûiÿÿïêÿ‘êçz_è®\x0fæj:+v‰¨þ)ߣømšSåy«\x1e­æ¶\x17…\x01\x06­†ÛiÿÿðÃ\x0fí»\x1fè®\x0få’i\x7f

Re: Question about ioremap and io_remap_page_range

[Ingo Molnar]

On Thu, 27 Sep 2001, [GB2312] Êæ¹úÇ¿ wrote:

>  Here is some rather basic questions I want ask ,any reply or comment
>  please CC to my emailbox,thank you very much.
>
>  When I work with kernel 2.4.2 in Intel X86 , I use
>
>      VIRT_ADDR = ioremap(BUS_ADDR); to map a section of PCI memory, and
>      X_ADDR = virt_to_phys(VIRT_ADDR);

i'd suggest to read Documentation/DMA-mapping.txt in any recent kernel
source, the bus_to_virt()/virt_to_phys() interface is obsolete and has
been replaced by the pci_alloc_*/pci_map_*/pci_free_*() dynamic
DMA-mapping API. You can find some nice examples of usage in
drivers/net/*.c.

	Ingo

Re: Question about ioremap and io_remap_page_range

[Armin Schindler]

On Thu, 27 Sep 2001, Ingo Molnar wrote:

>
> On Thu, 27 Sep 2001, [GB2312] Êæ¹úÇ¿ wrote:
>
> >  Here is some rather basic questions I want ask ,any reply or comment
> >  please CC to my emailbox,thank you very much.
> >
> >  When I work with kernel 2.4.2 in Intel X86 , I use
> >
> >      VIRT_ADDR = ioremap(BUS_ADDR); to map a section of PCI memory, and
> >      X_ADDR = virt_to_phys(VIRT_ADDR);
>
> i'd suggest to read Documentation/DMA-mapping.txt in any recent kernel
> source, the bus_to_virt()/virt_to_phys() interface is obsolete and has
> been replaced by the pci_alloc_*/pci_map_*/pci_free_*() dynamic
> DMA-mapping API. You can find some nice examples of usage in
> drivers/net/*.c.
>
> 	Ingo

virt_to_phys() is obsolete ? What should be used if I need the phys address
of a memory page ? (e.g. for mmap() remapping)

Armin

Re: Question about ioremap and io_remap_page_range

[David S. Miller]

   From: Armin Schindler <mac@melware.de>
   Date: Thu, 27 Sep 2001 10:03:26 +0200 (MEST)
   
   virt_to_phys() is obsolete ? What should be used if I need the phys address
   of a memory page ? (e.g. for mmap() remapping)
   
Ingo really is talking about virt_to_bus/bus_to_virt.

Franks a lot,
David S. Miller
davem@redhat.com

Re: Question about ioremap and io_remap_page_range

[Russell King]

On Thu, Sep 27, 2001 at 10:03:26AM +0200, Armin Schindler wrote:
> virt_to_phys() is obsolete ? What should be used if I need the phys address
> of a memory page ? (e.g. for mmap() remapping)

virt_to_phys is only valid on the kernel direct mapped RAM area - memory
returned from kmalloc and non-highmem memory returned from get_free_page.

Since vmalloc and ioremap do not return a pointer into the direct mapped
RAM area, but its own mapping, it is not valid to use virt_to_phys on
these returned values, even through it might be a virtual address.

--
Russell King (rmk@arm.linux.org.uk)                The developer of ARM Linux
             http://www.arm.linux.org.uk/personal/aboutme.html

Re: Question about ioremap and io_remap_page_range

[Jonathan Corbet]

> >      VIRT_ADDR = ioremap(BUS_ADDR); to map a section of PCI memory, and
> >      X_ADDR = virt_to_phys(VIRT_ADDR);
> 
> i'd suggest to read Documentation/DMA-mapping.txt in any recent kernel
> source, the bus_to_virt()/virt_to_phys() interface is obsolete and has
> been replaced by the pci_alloc_*/pci_map_*/pci_free_*() dynamic
> DMA-mapping API.

...or, of course, _Linux_Device_Drivers_, Second Edition, available online
at http://www.xml.com/ldd/chapter/book/index.html.  The DMA chapter covers
this API as well.

</self-promotion>

jon

Jonathan Corbet
Executive editor, LWN.net
corbet@lwn.net

Re: Question about ioremap and io_remap_page_range

[Jonathan Lundell]

At 8:33 AM -0600 2001-09-27, Jonathan Corbet wrote:
>  > >      VIRT_ADDR = ioremap(BUS_ADDR); to map a section of PCI memory, and
>>  >      X_ADDR = virt_to_phys(VIRT_ADDR);
>>
>>  i'd suggest to read Documentation/DMA-mapping.txt in any recent kernel
>>  source, the bus_to_virt()/virt_to_phys() interface is obsolete and has
>>  been replaced by the pci_alloc_*/pci_map_*/pci_free_*() dynamic
>>  DMA-mapping API.
>
>...or, of course, _Linux_Device_Drivers_, Second Edition, available online
>at http://www.xml.com/ldd/chapter/book/index.html.  The DMA chapter covers
>this API as well.

That chapter says, in small part,

>Based on this discussion, you might also want to map addresses 
>returned by ioremap to user space. This mapping is easily 
>accomplished because you can use remap_page_range directly, without 
>implementing methods for virtual memory areas. In other words, 
>remap_page_range is already usable for building new page tables that 
>map I/O memory to user space

Are you suggesting that ioremap() objects can be passed to 
remap_page_range()? How can that be valid?


Also, there's an html conversion error:

>      if (offset >= _ _pa(high_memory) || (filp->f_flags & O_SYNC))

  Looks like   got translated to  
-- 
/Jonathan Lundell.