* The nvidiafb driver is requiring too much vmalloc space and other problems.
@ 2005-03-30 22:59 Miles Lane
2005-03-31 23:10 ` Benjamin Herrenschmidt
0 siblings, 1 reply; 3+ messages in thread
From: Miles Lane @ 2005-03-30 22:59 UTC (permalink / raw)
To: linux-fbdev-devel, Randy.Dunlap, Andrew Morton,
Antonino A. Daplas
Hello,
With guidance from Randy Dunlap, it has been determined that
the nvidiafb driver will fail to operate correctly unless more than
128M of memory is provided by using "vmalloc=". The default
kernel configuration provides for 128M, but Randy found that:
>I started looking at vmalloc() and what it calls (which is
> __get_vm_area). _get_vm_area() always allocates one extra
> page (called a "guard page") between all vmalloc allocations,
> so even though 128 MB is the default amount and the amount
> that nvidiafb wants to use, the kernel wants to allocate
> 128 MB + PAGE_SIZE (4 KB on x86; are you on x86?), so even
> if nvidiafb is the only caller, the vmalloc() call will fail.
Booting with vmalloc=256M works, but Andrew thinks that
nvidiafb should not be requiring 128M in the first place.
Regarding the problems with i2c bit-banging drivers, I don't
know if the Oops I have been getting is related to the vmalloc
issue somehow. Comments? I'll try to repro the crash with
vmalloc=256M set.
Lastly, with vmalloc=256M, my machine crashes whenever
I try to switch from XFree86 to a framebuffer console. I cannot
see the oops. Unfortunately, the only machine I could use for
debugging using a serial connection is a Powerbook, which has
no serial port.
I could try setting up dumping the stack over the network,
but I have never tried to do this before and do not know
whether I can do the dump to a Powerbook running OS/X.
Any suggestions?
Thanks,
Miles
-------------------------------------------------------
This SF.net email is sponsored by Demarc:
A global provider of Threat Management Solutions.
Download our HomeAdmin security software for free today!
http://www.demarc.com/info/Sentarus/hamr30
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: The nvidiafb driver is requiring too much vmalloc space and other problems.
2005-03-30 22:59 The nvidiafb driver is requiring too much vmalloc space and other problems Miles Lane
@ 2005-03-31 23:10 ` Benjamin Herrenschmidt
2005-03-31 23:18 ` Randy.Dunlap
0 siblings, 1 reply; 3+ messages in thread
From: Benjamin Herrenschmidt @ 2005-03-31 23:10 UTC (permalink / raw)
To: Linux Fbdev development list
Cc: Randy.Dunlap, Andrew Morton, Antonino A. Daplas
On Wed, 2005-03-30 at 14:59 -0800, Miles Lane wrote:
> Hello,
>
> With guidance from Randy Dunlap, it has been determined that
> the nvidiafb driver will fail to operate correctly unless more than
> 128M of memory is provided by using "vmalloc=". The default
> kernel configuration provides for 128M, but Randy found that:
.../...
it's a problem with ioremap of the framebuffer ? (which uses get_vm_area
as well) or is there an actual vmalloc in there ?
In radeonfb, I had to limit the amount of vram actually ioremap'd by the
kernel because of that.
Ben.
-------------------------------------------------------
This SF.net email is sponsored by Demarc:
A global provider of Threat Management Solutions.
Download our HomeAdmin security software for free today!
http://www.demarc.com/info/Sentarus/hamr30
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: The nvidiafb driver is requiring too much vmalloc space and other problems.
2005-03-31 23:10 ` Benjamin Herrenschmidt
@ 2005-03-31 23:18 ` Randy.Dunlap
0 siblings, 0 replies; 3+ messages in thread
From: Randy.Dunlap @ 2005-03-31 23:18 UTC (permalink / raw)
To: Benjamin Herrenschmidt
Cc: Linux Fbdev development list, Andrew Morton, Antonino A. Daplas
Benjamin Herrenschmidt wrote:
> On Wed, 2005-03-30 at 14:59 -0800, Miles Lane wrote:
>
>>Hello,
>>
>>With guidance from Randy Dunlap, it has been determined that
>>the nvidiafb driver will fail to operate correctly unless more than
>>128M of memory is provided by using "vmalloc=". The default
>>kernel configuration provides for 128M, but Randy found that:
>
>
> .../...
>
> it's a problem with ioremap of the framebuffer ? (which uses get_vm_area
> as well) or is there an actual vmalloc in there ?
It's ioremap of the frame buffer -- 128 MB being requested and it
fails if the default 128 MB kernel vmalloc area size is used.
> In radeonfb, I had to limit the amount of vram actually ioremap'd by the
> kernel because of that.
I guess that Tony will know what to do when he sees this....
--
~Randy
-------------------------------------------------------
This SF.net email is sponsored by Demarc:
A global provider of Threat Management Solutions.
Download our HomeAdmin security software for free today!
http://www.demarc.com/info/Sentarus/hamr30
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2005-03-31 23:18 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2005-03-30 22:59 The nvidiafb driver is requiring too much vmalloc space and other problems Miles Lane
2005-03-31 23:10 ` Benjamin Herrenschmidt
2005-03-31 23:18 ` Randy.Dunlap
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).