Re: 2.3.43 alpha broken with >2g of ram

[Peter Rival]

I backed out the below change (just out of curiosity) from 2.3.42+Richard's
patches, and the ES40 booted up to the point where it tries to open
/dev/console (just after freeing the initmem, having succeeded in scanning
all the SCSI cards).  I'm sure there are other problems lurking that cause it
to hang there, but who knows.  Richard - Jay had mentioned that the
translated base address on the third window is set to zero when the comments
say it should be 1GB, and I noticed the same for the fourth window which
should be at 2GB - does this mean anything?

Also, I'm going to give it a whirl on a 4GB rawhide with the core_mcpcia.c
from 2.3.43 and I'll let everyone know if that works, otherwise I'll have to
hope someone can figure out the flush_tlb_pgtables linking problem so we can
continue work with 2.3.43.  (_Still_ don't know enough about how we set up
some of these things to make a difference sometimes...most discouraging...)

 - Pete

Andrea Arcangeli wrote:
<snip>

>
> @@ -225,15 +232,8 @@
>                         max_low_pfn = end;
>         }
>
> -       /* Enforce maximum of 2GB even if there is more.  Blah.  */
> -       if (max_low_pfn > PFN_MAX)
> -               max_low_pfn = PFN_MAX;
> -       printk("max_low_pfn %ld\n", max_low_pfn);
> -
>         /* Find the end of the kernel memory.  */
>         start_pfn = PFN_UP(virt_to_phys(_end));
> -       printk("_end %p, start_pfn %ld\n", _end, start_pfn);
> -
>         bootmap_start = -1;
>
>   try_again:
>

<snip>