Re: [RFC][PATCH] hugetlb: add information and interface in sysfs [Was Re: [RFC][PATCH 4/5] Documentation: add node files to sysfs ABI]

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

On 29.04.2008 [11:26:13 -0700], Greg KH wrote:
> On Tue, Apr 29, 2008 at 11:14:15AM -0700, Nishanth Aravamudan wrote:
> > On 29.04.2008 [10:22:43 -0700], Greg KH wrote:
> > > On Tue, Apr 29, 2008 at 10:11:15AM -0700, Nishanth Aravamudan wrote:
> > > > +struct hstate_attribute {
> > > > +	struct attribute attr;
> > > > +	ssize_t (*show)(struct hstate *h, char *buf);
> > > > +	ssize_t (*store)(struct hstate *h, const char *buf, size_t count);
> > > > +};
> > > 
> > > Do you need your own attribute type with show and store?  Can't you just
> > > use the "default" kobject attributes?
> > 
> > Hrm, I don't know? Probably. Like I said, I was using the
> > /sys/kernel/slab code as my reference. Can you explain this more? Or
> > just point me to the source/documentation I should read for info.
> 
> Documentation/kobject.txt, with sample examples in samples/kobject/ for
> you to copy and use.

Great thanks!

> > Are you referring to kobj_attr_show/kobj_attr_store? Should I just be
> > using kobj_sysfs_ops, then, most likely?
> 
> See the above examples for more details.

Will do -- I think we'll need our own store, at least, though, because
of locking issues? And I'm guessing if we provide our own store, we're
going to need to provide our own show?

> > > Also, you have no release function for your kobject to be cleaned up,
> > > that's a major bug.
> > 
> > Well, these kobjects never go away? They will be statically initialized
> > at boot-time and then stick around until the kernel goes away. Looking
> > at /sys/kernel/slab's code, again, the release() function there does a
> > kfree() on the containing kmem_cache, but for hugetlb, the hstates are
> > static... If we do move to dynamic allocations ever (or allow adding
> > hugepage sizes at run-time somehow), then perhaps we'll need a release
> > method then?
> 
> Yes you will.  Please always create one, what happens when you want to
> clean them up at shut-down time...

Again, I'm not sure what you want me to clean-up? The examples in
samples/ are freeing dynamically allocated objects containing the
kobject in question -- but /sys/kernel/hugepages only dynamically
allocates the kobject itself... Although, I guess I should free the name
string since I used kasprintf()...

Thanks,
Nish

-- 
Nishanth Aravamudan <nacc@us.ibm.com>
IBM Linux Technology Center

--
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>