Brenton,
Yes, the mapping is rather strange, saw that also. You might want to
look at how they implemented the __phys_to_virt and __virt_to_phys in
the 2.4 kernel as supplied by the vendor. It is a bit of a hack but
does appear to map things up into a more approprate address. I was
using these at one point in a 2.6 kernel but I didn't try that change
along with your patches.
That may fix the issue.
- Doug
--- "Breton M. Saunders" <> wrote:
> Hi Doug,
>
> Yes, the bug happens here. I was running memtest-4.0.6 to check
> things were working ok. They are not yet.
>
> It seems that as soon as the kernel out of memory killer is
> invoked,
> show_mem is called, crashing the system. The general rule seems to
> be:
> do not try to mlock more than 48mb of ram yet. I have successfully
> run
> memtest with 48mb though. I also suspect that setting up a swap file
>
> may not work at the moment.
>
> I'm not very familiar with the linux memory system yet - and as
> such,
> the patch is more of a hack at the moment than a correct fix. What I
>
> have found so far is that the high-order pages at physical offset
> 0xe0000000, 0xe1000000, 0xe4000000 and 0xe5000000 are being mapped to
>
> virtual addresses 0xa0000000, 0xa1000000, ... more by fluke than
> anything else.
>
> The calculation of a virtual address consists of adding PAGE_OFFSET
>
> (0xc0000000) to the physical address. For the low memory 0x00000000,
>
> 0x10000000 ... this works fine, mapping the kernel memory in at
> 0xC0000000, etc... For the high page memory 0xC0000000 + 0xE0000000
> =
> 0x1A0000000 - since its a 32 bit processor, the 1 is truncated off
> the
> front.
>
> If anyone knows where there is some documentation on implementing
> DISCONTIGMEM, let me know!
>
> In the meantime I am likely going to try and hard code the memory
> map
> using the DISCONTIGMEM macros defined in
> include/asm-arm/arch-ep93xx/memory.h...
>
> Cheers,
>
> -Brett
>
>
>
> Doug Chapman wrote:
> >
> > Brenton,
> >
> > Thanks for the patches! Between what you have and what I had
> already
> > dug through I was able to get this working on the TS-7260 also
> (minor
> > changes). I think my changes will also allow TS-7250 to work as
> well.
> >
> > For those interested, I am going to clean up what I have and post a
> > patch + .config for the TS-7260. I should be able to do this in a
> > couple of days and I will put it in the files section of the yahoo
> > group.
> >
> > One bug I found so far is the show_mem function triggered by
> sysrq-m
> > hits an oops. Can you try this on your TS7400:
> >
> > echo m > /proc/sysrq-trigger
> >
> > and let me know if you get an oops? Might be related to the changes
> I
> > made.
> >
> > - Doug
> >
> > --- "Breton M. Saunders" <
> > <breton.saunders%40ntlworld.com>> wrote:
> >
> > > Folks,
> > >
> > > I've put some patches together to run 2.6.19 on a TS7400 w/64MiB
> > > ram -
> > > the 8 banks by 8MiB version. The patches are assembled from M.
> > > Burian's
> > > tree, and seems to boot just fine on my TS7400.
> > >
> > > Find the patches at: http://www.mynah-software.com/arm/
> > <http://www.mynah-software.com/arm/>
> > >
> > > Have Fun,
> > >
> > > -Brett
> > >
> > >
> > >
> > >
> > > Yahoo! Groups Links
> > >
> > >
> > >
> > >
> >
> >
>
>
>
>
> Yahoo! Groups Links
>
>
>
>
Yahoo! Groups Links
<*> To visit your group on the web, go to:
http://groups.yahoo.com/group/ts-7000/
<*> Your email settings:
Individual Email | Traditional
<*> To change settings online go to:
http://groups.yahoo.com/group/ts-7000/join
(Yahoo! ID required)
<*> To change settings via email:
<*> To unsubscribe from this group, send an email to:
<*> Your use of Yahoo! Groups is subject to:
http://docs.yahoo.com/info/terms/
|