Messages in this thread | | | Date | Wed, 22 Sep 2010 16:16:28 -0700 | From | "H. Peter Anvin" <> | Subject | Re: [PATCH] x86: tidy e820 output |
| |
On 09/22/2010 04:12 PM, David Rientjes wrote: > On Wed, 22 Sep 2010, H. Peter Anvin wrote: > >>> This patch is going to break our userspace parsing scripts if you change >>> the output format. Admittedly, we're probably one of the few users who >>> actually parses this output, but we do have reasons to do it for our >>> firmware. If there was some improvement being introduced here, we'd >>> happily handle multiple regexs (we constantly add new patterns when new >>> kernels are released), but I'm not seeing how this is better. >> >> Kernel messages are not an ABI or API. >> >> The user space API for this stuff is /sys/firmware/memmap. >> > > I'm referring to using [start, start + addr - 1] in the output, like > /sys/firmware/memmap does, as opposed to [start, start + addr]. Is it not > valuable to include the actual e820 map in some way, especially when you > have your own BIOS? We've always used this output since it isn't > available later.
"The actual e820 map" contains (start, length, type) -- the end bracket is not part of it at all. Either way, /sys/firmware/memmap does provide the memory map as provided by the firmware through whatever means.
> Another use of this information that people may already be using it for is > ensuring the memmap= on the command line is parsed correctly. > > So, again, I'm looking for the benefit here in this patch and it's not > immediately apparent to me.
Consistency with other resources displayed seems like a major win to me.
-hpa
| |