Re: 2.6.9-rc2-mm2

Re: 2.6.9-rc2-mm2

[Andrew Morton]

Jesse Barnes <jbarnes@engr.sgi.com> wrote:
>
> On Wednesday, September 22, 2004 8:39 pm, Peter Chubb wrote:
> > >>>>> "Jesse" = Jesse Barnes <jbarnes@engr.sgi.com> writes:
> >
> > Jesse> On Wednesday, September 22, 2004 4:12 pm, Andrew Morton wrote:
> > >> - This kernel doesn't work on ia64 (instant reboot).  But neither
> > >> does 2.6.9-rc2, nor current Linus -bk.  Is it just me?
> >
> > Jesse> I certainly hope so.  Current bk works on my 2p Altix, and iirc
> > Jesse> 2.6.9-rc2 worked as well.  I'm trying 2.6.9-rc2-mm2 right now.
> > Jesse> I haven't tried generic_defconfig yet either, maybe that's it?
> >
> > It no longer works on ZX.  Don't know why.
> 
> Maybe this is another, more severe instance of the problem James reported last 
> week that was worked around by enabling CONFIG_DISCONTIGMEM.
> 

It looks like Tony is wearing the BPB.  The below patch from September 8 is
what causes my non-discontigmem virtual-mem-map ia64 box instantly reboot.
Reverting it makes things happy.


--- b/include/asm-ia64/page.h	2004-09-08 10:23:43 -07:00
+++ b/include/asm-ia64/page.h	2004-09-08 16:12:10 -07:00
@@ -86,13 +86,14 @@
 #ifndef CONFIG_DISCONTIGMEM
 # ifdef CONFIG_VIRTUAL_MEM_MAP
 extern struct page *vmem_map;
-#  define pfn_valid(pfn)       (((pfn) < max_mapnr) && ia64_pfn_valid(pfn))
-#  define page_to_pfn(page)    ((unsigned long) (page - vmem_map))
-#  define pfn_to_page(pfn)     (vmem_map + (pfn))
+#  define pfn_valid(pfn)	(((pfn) < max_mapnr) && ia64_pfn_valid(pfn))
+#  define page_to_pfn(page)	((unsigned long) (page - vmem_map))
+#  define pfn_to_page(pfn)	(vmem_map + (pfn))
+# else
+#  define pfn_valid(pfn)	(((pfn) < max_mapnr) && ia64_pfn_valid(pfn))
+#  define page_to_pfn(page)	((unsigned long) (page - mem_map))
+#  define pfn_to_page(pfn)	(mem_map + (pfn))
 # endif
-#define pfn_valid(pfn)		(((pfn) < max_mapnr) && ia64_pfn_valid(pfn))
-#define page_to_pfn(page)	((unsigned long) (page - mem_map))
-#define pfn_to_page(pfn)	(mem_map + (pfn))
 #endif /* CONFIG_DISCONTIGMEM */
 
 #define page_to_phys(page)	(page_to_pfn(page) << PAGE_SHIFT)



Process question: how is it possible that the ia64 tree could have been
this dead for this long?

RE: 2.6.9-rc2-mm2

[Luck, Tony]

>It looks like Tony is wearing the BPB.  The below patch from
                                   ^^^?
Huh?  I can tell from context that this is all my fault (to which
I agree; it is), but what does "BPB" stand for?

>Process question: how is it possible that the ia64 tree could have been
>this dead for this long?

Because we have immense confusion about which combinations of
config options (NUMA, DISCONTIG, VIRTUAL_MEM_MAP, SMP) are
supported.  Kconfig allows almost any combination of them, but on
any given week only some combinations work.  The patch that broke
things for you came in to fix a problem for Peter.

I'll take a look at this.  Can you post the .config that you
are using.

-Tony

Re: 2.6.9-rc2-mm2

[Andrew Morton]

"Luck, Tony" <tony.luck@intel.com> wrote:
>
> >It looks like Tony is wearing the BPB.  The below patch from
>                                    ^^^?
> Huh?  I can tell from context that this is all my fault (to which
> I agree; it is), but what does "BPB" stand for?

brown paper bag

> I'll take a look at this.

Thanks.

>  Can you post the .config that you are using.

http://www.zip.com.au/~akpm/linux/patches/stuff/config-ia64