[RFC] [PATCH] vmemmap fixes to use smaller pages

Geoff Levand geoffrey.levand at am.sony.com
Thu May 1 05:06:39 EST 2008


Benjamin Herrenschmidt wrote:
> This patch changes vmemmap to use a different region (region 0xf) of the
> address space whose page size can be dynamically configured at boot.
> 
> The problem with the current approach of always using 16M pages is that
> it's not well suited to machines that have small amounts of memory such
> as small partitions on pseries, or PS3's.
> 
> In fact, on the PS3, failure to allocate the 16M page backing vmmemmap
> tends to prevent hotplugging the HV's "additional" memory, thus limiting
> the available memory even more, from my experience down to something
> like 80M total, which makes it really not very useable.
> 
> The logic used by my match to choose the vmemmap page size is:
> 
>  - If 16M pages are available and there's 1G or more RAM at boot, use that size.
>  - Else if 64K pages are available, use that
>  - Else use 4K pages
> 
> I've tested on a POWER6 (16M pages) and on an iSeries POWER3 (4K pages)
> and it seems to work fine.
> 
> However, when attempting to test on a PS3, it didn't boot.
> 
> In fact, it doesn't boot without my patch with current upstream. 


Yes, this is a know problem I am working on, related to recent
changes in bootmem.  Errors with: 'sparse_early_usemap_alloc: allocation failed'.


I tried
> booting 2.6.25 with a ps3_defconfig and that doesn't work neither
> (though at least when doing the later, I do get a black screen & no
> sync, like of ps3fb failed monitor detection, while with current
> upstream, I just get the last kexec messages and nothing happens).


This should work.  You are the first to report a problem with 
2.6.25.  Could you double check your build, and if you still have
trouble, put your vmlinux somewhere I can get it?


> Since the PS3 boot failures are impossible to debug unless your email is
> @sony* and you have the special magic tools, I'll let Geoff try the
> patch out.


OK, I'll try it with the upstream kernel from last week and report
within the next day or so.


-Geoff




More information about the Linuxppc-dev mailing list