Re: [PATCH 2/2]: acpica/nfit: Rename not-armed bit definition

[Toshi Kani <toshi.kani@hp.com>]

On Wed, 2015-08-26 at 14:30 -0700, Dan Williams wrote:
> On Wed, Aug 26, 2015 at 2:12 PM, Toshi Kani <toshi.kani@hp.com> wrote:
> > On Wed, 2015-08-26 at 10:16 -0700, Dan Williams wrote:
> > > On Wed, Aug 26, 2015 at 9:20 AM, Toshi Kani <toshi.kani@hp.com> wrote:
> > > > ACPI 6.0 NFIT Memory Device State Flags in Table 5-129 defines
> > > > bit 3 as follows.
> > > > 
> > > >   Bit [3] set to 1 to indicate that the Memory Device is observed
> > > >   to be not armed prior to OSPM hand off. A Memory Device is
> > > >   considered armed if it is able to accept persistent writes.
> > > > 
> > > > This bit is currently defined as ACPI_NFIT_MEM_ARMED, which can be
> > > > confusing as if the Memory Device is armed when this bit is set.
> > > > 
> > > > Change the name to ACPI_NFIT_MEM_NOT_ARMED per the spec.
> > > > 
> > > > Signed-off-by: Toshi Kani <toshi.kani@hp.com>
> > > > Cc: Dan Williams <dan.j.williams@intel.com>
> > > > Cc: Bob Moore <robert.moore@intel.com>
> > > > Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
> > > > ---
> > > >  drivers/acpi/nfit.c              |    6 +++---
> > > >  drivers/acpi/nfit.h              |    2 +-
> > > >  include/acpi/actbl1.h            |    2 +-
> > > 
> > > This file "include/acpi/actbl1.h" is owned by the ACPICA project so
> > > any changes need to come through them.  But that said, I'm not sure we
> > > need friendly names at this level.
> > 
> > I think the name is misleading, but I agree with the process and this 
> > patch2
> > can be dropped.  It'd be nice if the ACPICA project can pick it up later
> > when they have a chance, though.
> > 
> > > What I usually say about sysfs name changes to be more human friendly
> > > is "sysfs is not a UI", i.e. it's not necessarily meant to be user
> > > friendly.  As long as the names for the flags are distinct then
> > > wrapping descriptive / accurate names around them is the role of
> > > libndctl and userspace management software.
> > > 
> > > Similar feedback for patch1 in the sense that I don't think we need to
> > > update the sysfs naming.  For example the API to retrieve the state of
> > > the "arm" flag in libndctl is ndctl_dimm_failed_arm().
> > 
> > I agree that we do not want to change sysfs API for friendliness, and I
> > understand that libndctl already consumes the strings...  But I think 
> > they
> > can be confusing for the long run, i.e. the flags is likely extended for
> > additional info, and more people may be looking at sysfs for the state.
> >  It'd be a lot harder to change them later.
> 
> The starting premise though is that this will be nicer for scripts
> that want to avoid the library.  Properly handling the async device
> registration semantics of the libnvdimm-sysfs interface is hard to get
> right in a script.  I'm trying my best to discourage raw use of sysfs
> for this reason.  Small fixes to the names of flags seems to miss this
> wider point.

Okay, I guess I will have to jump on the bandwagon and discourage people to
look at sysfs... ;-P

Thanks,
-Toshi