Re: [RFC][PATCH 4/5] Documentation: add node files to sysfs ABI

[Nishanth Aravamudan <nacc@us.ibm.com>]

On 20.04.2008 [23:06:48 -0700], Christoph Lameter wrote:
> On Sat, 19 Apr 2008, Greg KH wrote:
> 
> > > That violation is replicated in /proc/meminfo /proc/vmstat etc etc.
> > 
> > Those are /proc files, not sysfs files :)
> 
> Hmmm.. Maybe we need to have /proc/node<x>/meminfo etc that replicates
> the /proc content for each node? Otherwise this cannot be symmetric
> because the different mount points have different requirements on how
> the output should look like.

But the memory info has nothing to do with process specific information,
which is what "new" /proc files should contain (or maybe I'm
mis-remembering).

The current location (/sys/devices/system/node) reflects that memory is
tied to system devices called "nodes"; I'm not entirely convinced we'd
want to change that?  Especially, as Greg noted, it's easier to obtain
the information we want off a sysdev, rather than the raw kobject.

While I understand the desire to maintain sanity for sysfs files,
perhaps the meminfo files (and numastat, etc) are just special, in that
they only make sense as a collective (the snapshot mentioned earlier in
this thread) -- to get a view of the component (memory, NUMA statistics,
etc) as a whole.

In that sense, perhaps the sysfs notion should be extended to "One
logical value per file", where logical is defined as the minimum atomic
information needed by the user [1]? Or perhaps sysfs just isn't the best
place for this information, I don't know. I don't believe I am the
person to make that call.

Thanks,
Nish

[1] That would allow files like available_clocksource not to seem like
violators of the sysfs rule:

$ sudo cat /sys/devices/system/clocksource/clocksource0/available_clocksource
hpet acpi_pm pit jiffies tsc

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>