2.6.10-rc1-mm2: intelfb/AGP unknown symbols

2.6.10-rc1-mm2: intelfb/AGP unknown symbols

[Adrian Bunk]

On Fri, Oct 29, 2004 at 01:49:30AM -0700, Andrew Morton wrote:
>...
> Changes since 2.6.10-rc1-mm1:
> 
>  linus.patch
>...
>  bk-agpgart.patch
>...


The removal of 3 "unneeded exports" in bk-agpgart.patch conflicts with 
code adding usage of them in Linus' tree:


<--  snip  -->

...
if [ -r System.map ]; then /sbin/depmod -ae -F System.map  2.6.10-rc1-mm2; fi
WARNING: /lib/modules/2.6.10-rc1-mm2/kernel/drivers/video/intelfb/intelfb.ko needs unknown symbol agp_backend_acquire
WARNING: /lib/modules/2.6.10-rc1-mm2/kernel/drivers/video/intelfb/intelfb.ko needs unknown symbol agp_copy_info
WARNING: /lib/modules/2.6.10-rc1-mm2/kernel/drivers/video/intelfb/intelfb.ko needs unknown symbol agp_backend_release
WARNING: /lib/modules/2.6.10-rc1-mm2/kernel/drivers/video/i810/i810fb.ko needs unknown symbol agp_backend_acquire
WARNING: /lib/modules/2.6.10-rc1-mm2/kernel/drivers/video/i810/i810fb.ko needs unknown symbol agp_backend_release

<--  snip  -->


cu
Adrian

-- 

       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed

Re: 2.6.10-rc1-mm2: intelfb/AGP unknown symbols

[Arjan van de Ven]

> The removal of 3 "unneeded exports" in bk-agpgart.patch conflicts with 
> code adding usage of them in Linus' tree:

that makes me really really curious why the fb driver calls into the backend and not just the agp frontend layer like the rest of the world does...

Re: Re: 2.6.10-rc1-mm2: intelfb/AGP unknown symbols

[Antonino A. Daplas]

On Saturday 30 October 2004 16:28, Arjan van de Ven wrote:
> > The removal of 3 "unneeded exports" in bk-agpgart.patch conflicts with
> > code adding usage of them in Linus' tree:
>
> that makes me really really curious why the fb driver calls into the
> backend and not just the agp frontend layer like the rest of the world
> does...
>

Because all functions in the frontend are marked static and are accessible
only via ioctl.

Anyway, I think the drivers can make do without the
agp_backend_acquire/release() functions, since all they do is
increment/decrement  a use count field.  I don't know about agp_copy_info()
but it might be possible to get the agp information from pci_dev structure.
This part I'm not sure.

What's wrong with exporting the symbols back again?

Tony




-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click

Re: [Linux-fbdev-devel] Re: 2.6.10-rc1-mm2: intelfb/AGP unknown symbols

[Arjan van de Ven]

> 
> What's wrong with exporting the symbols back again?

if they are the right api to use; nothing. If they aren't (and what you
describe somehow suggests they aren't) it sounds better to make the
frontend usable for the intelfb driver instead...

Re: [Linux-fbdev-devel] Re: 2.6.10-rc1-mm2: intelfb/AGP unknown symbols

[Antonino A. Daplas]

On Saturday 30 October 2004 19:34, Arjan van de Ven wrote:
> > What's wrong with exporting the symbols back again?
>
> if they are the right api to use; nothing. If they aren't (and what you
> describe somehow suggests they aren't) it sounds better to make the
> frontend usable for the intelfb driver instead...
>

I think the functions are the right API to use for clients within the
kernel.  The frontend is directed more for userspace clients. 

The old interface was to do an inter_module_get/put, but this is to
be deprecated. And all it does is to provide all the backend functions
to the requestor.

Either  a new interface is provided by agpgart, otherwise, not just intelfb
and i810fb will be affected, but also DRM (which currently uses
inter_module_get/put("drm_agp")).

If I remember correctly, the DRI people also have a new patch that removes
inter_module_get/put and they did it by calling the backend functions directly.

Tony

Re: 2.6.10-rc1-mm2: intelfb/AGP unknown symbols

[Christoph Hellwig]

On Sat, Oct 30, 2004 at 10:28:41AM +0200, Arjan van de Ven wrote:
> 
> > The removal of 3 "unneeded exports" in bk-agpgart.patch conflicts with 
> > code adding usage of them in Linus' tree:
> 
> that makes me really really curious why the fb driver calls into the backend and not just the agp frontend layer like the rest of the world does...

In agpgart context the "fronend" is the character device for userland.
these functions should be exported because they are the driver API.  The
only reason they weren't used previously is because of the inter_module_*
braindamage to hide it.