Re: remap_page_range problem on 2.3.x

Re: remap_page_range problem on 2.3.x

[Andy Henroid]

--- Jeff Garzik <jgarzik@mandrakesoft.com> wrote:
> Andy Henroid wrote:
> > 
> >                        Name: mmtest.tar.gz
> >    mmtest.tar.gz       Type: Unix Tape Archive
> (application/x-tar)
> >                    Encoding: base64
> >                 Description: mmtest.tar.gz
> 
> Are these the correct test files?
> 
> rum:~/tmp/mmtest> grep -i remap *
> rum:~/tmp/mmtest> 

Yes, the remap_page_range is done indirectly
through mmap call to the /dev/mem driver.

> I think you'll need to do something like
> 
> init():
> 	dsdt = get_free_pages(...)
> 
> chrdev mmap() op:
> 	remap_page_range(dsdt, ...)

OK, yes I bet that would work.  I just don't see
any good reason why the remap_page_range doesn't
appear to be doing the right thing for a piece
of kmalloced memory.

> If you are going to present data via /proc, you
> might as well simply dump the raw data out to
> whoever is reading /proc/driver/acpi/dsdt...

It's actually a bit less wasteful, for systems
with large DSDTs, to do the mmap.  But, right,
this is another possible work-around to my
current problem.

-Andy

__________________________________________________
Do You Yahoo!?
Talk to your friends online with Yahoo! Messenger.
http://im.yahoo.com
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux.eu.org/Linux-MM/

Re: remap_page_range problem on 2.3.x

[Thomas Sailer]

Andy Henroid wrote:

> Yes, the remap_page_range is done indirectly
> through mmap call to the /dev/mem driver.

I think I saw this too some time ago.

I tried to duplicate the kernel's view of the
address space above 0xc0000000 at an offset in
a usermode program. I ended up being able to
read the correct data from /dev/mem, but reading
from an mmaped page from /dev/mem returned completely
bogus data (all zero).

My code is in usb.in.tum.de, module usbstress,
files uhcidump.c and kmem*

Tom
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux.eu.org/Linux-MM/

Re: remap_page_range problem on 2.3.x

[Andy Henroid]

--- Thomas Sailer <sailer@ife.ee.ethz.ch> wrote:
> Andy Henroid wrote:
> 
> > Yes, the remap_page_range is done indirectly
> > through mmap call to the /dev/mem driver.
> 
> I think I saw this too some time ago.
> 
> I tried to duplicate the kernel's view of the
> address space above 0xc0000000 at an offset in
> a usermode program. I ended up being able to
> read the correct data from /dev/mem, but reading
> from an mmaped page from /dev/mem returned
> completely bogus data (all zero).

Yes, the read through /dev/mem work for me too.
It looks like remap_page_range is doing the right
thing and it certainly works for other remapping.
I wonder what the problem is?

Oh well, low priority compared to the other things
that need to get fixed as we move into pre2.4

Thanks,
Andy

__________________________________________________
Do You Yahoo!?
Talk to your friends online with Yahoo! Messenger.
http://im.yahoo.com
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux.eu.org/Linux-MM/

remap_page_range problem on 2.3.x

[Andy Henroid]

[-- Attachment #1: Type: text/plain, Size: 305 bytes --]

Attached is a tiny kernel module and app to reproduce
the problem on 2.3.49.  Or am I doing something stupid
here?

Thank you,
Andy Henroid
andy_henroid@yahoo.com






__________________________________________________
Do You Yahoo!?
Talk to your friends online with Yahoo! Messenger.
http://im.yahoo.com

[-- Attachment #2: mmtest.tar.gz --]
[-- Type: application/x-gzip-compressed, Size: 1326 bytes --]

Re: remap_page_range problem on 2.3.x

[Jeff Garzik]

Andy Henroid wrote:
> 
>                        Name: mmtest.tar.gz
>    mmtest.tar.gz       Type: Unix Tape Archive (application/x-tar)
>                    Encoding: base64
>                 Description: mmtest.tar.gz

Are these the correct test files?

rum:~/tmp/mmtest> grep -i remap *
rum:~/tmp/mmtest> 

I think you'll need to do something like

init():
	dsdt = get_free_pages(...)

chrdev mmap() op:
	remap_page_range(dsdt, ...)

If you are going to present data via /proc, you might as well simply
dump the raw data out to whoever is reading /proc/driver/acpi/dsdt...

-- 
Jeff Garzik              | My to-do list is a function
Building 1024            | which approaches infinity.
MandrakeSoft, Inc.       |
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux.eu.org/Linux-MM/