Re: [RFC PATCH v2 1/3] meminfo_extra: introduce meminfo extra

[Greg KH]

On Tue, Mar 24, 2020 at 08:37:38PM +0900, Jaewon Kim wrote:
> 
> 
> On 2020ë 03ì 24ì 19:11, Greg KH wrote:
> > On Tue, Mar 24, 2020 at 06:11:17PM +0900, Jaewon Kim wrote:
> >> On 2020ë 03ì 23ì 18:53, Greg KH wrote:
> >>>> +int register_meminfo_extra(atomic_long_t *val, int shift, const char *name)
> >>>> +{
> >>>> +	struct meminfo_extra *meminfo, *memtemp;
> >>>> +	int len;
> >>>> +	int error = 0;
> >>>> +
> >>>> +	meminfo = kzalloc(sizeof(*meminfo), GFP_KERNEL);
> >>>> +	if (!meminfo) {
> >>>> +		error = -ENOMEM;
> >>>> +		goto out;
> >>>> +	}
> >>>> +
> >>>> +	meminfo->val = val;
> >>>> +	meminfo->shift_for_page = shift;
> >>>> +	strncpy(meminfo->name, name, NAME_SIZE);
> >>>> +	len = strlen(meminfo->name);
> >>>> +	meminfo->name[len] = ':';
> >>>> +	strncpy(meminfo->name_pad, meminfo->name, NAME_BUF_SIZE);
> >>>> +	while (++len < NAME_BUF_SIZE - 1)
> >>>> +		meminfo->name_pad[len] = ' ';
> >>>> +
> >>>> +	spin_lock(&meminfo_lock);
> >>>> +	list_for_each_entry_rcu(memtemp, &meminfo_head, list) {
> >>>> +		if (memtemp->val == val) {
> >>>> +			error = -EINVAL;
> >>>> +			break;
> >>>> +		}
> >>>> +	}
> >>>> +	if (!error)
> >>>> +		list_add_tail_rcu(&meminfo->list, &meminfo_head);
> >>>> +	spin_unlock(&meminfo_lock);
> >>> If you have a lock, why are you needing rcu?
> >> I think _rcu should be removed out of list_for_each_entry_rcu.
> >> But I'm confused about what you meant.
> >> I used rcu_read_lock on __meminfo_extra,
> >> and I think spin_lock is also needed for addition and deletion to handle multiple modifiers.
> > If that's the case, then that's fine, it just didn't seem like that was
> > needed.  Or I might have been reading your rcu logic incorrectly...
> >
> >>>> +	if (error)
> >>>> +		kfree(meminfo);
> >>>> +out:
> >>>> +
> >>>> +	return error;
> >>>> +}
> >>>> +EXPORT_SYMBOL(register_meminfo_extra);
> >>> EXPORT_SYMBOL_GPL()?  I have to ask :)
> >> I can use EXPORT_SYMBOL_GPL.
> >>> thanks,
> >>>
> >>> greg k-h
> >>>
> >>>
> >> Hello
> >> Thank you for your comment.
> >>
> >> By the way there was not resolved discussion on v1 patch as I mentioned on cover page.
> >> I'd like to hear your opinion on this /proc/meminfo_extra node.
> > I think it is the propagation of an old and obsolete interface that you
> > will have to support for the next 20+ years and yet not actually be
> > useful :)
> >
> >> Do you think this is meaningful or cannot co-exist with other future
> >> sysfs based API.
> > What sysfs-based API?
> Please refer to mail thread on v1 patch set - https://lkml.org/lkml/fancy/2020/3/10/2102
> especially discussion with Leon Romanovsky on https://lkml.org/lkml/fancy/2020/3/16/140

I really do not understand what you are referring to here, sorry.   I do
not see any sysfs-based code in that thread.

And try to use lore.kernel.org, lkml.org doesn't always work and we have
no control over that :(

> > I still don't know _why_ you want this.  The ION stuff is not needed as
> > that code is about to be deleted, so who else wants this?  What is the
> > use-case for it that is so desperately needed that parsing
> > yet-another-proc file is going to solve the problem?
> In my Android device, there are graphic driver memory, zsmalloc memory except ION.

Ok, so what does Android have to do with this?

> I don't know other cases in other platform.
> Not desperately needed but I think we need one userspace knob to see overall hidden huge memory.

Why?  Who wants that?  What would userspace do with that?  And what
exactly do you want to show?

Is this just a debugging thing?  Then use debugfs for that, not proc.
Isn't that what the DRM developers are starting to do?

> Additionally I'd like to see all those hidden memory in OutOfMemory log.

How is anything hidden, can't you see it in the slab information?

> This is useful to get clue to find memory hogger.
> i.e.) show_mem on oom
> <6>[  420.856428]  Mem-Info:
> <6>[  420.856433]  IonSystemHeap:32813kB ZsPages:44114kB GraphicDriver::13091kB
> <6>[  420.856450]  active_anon:957205 inactive_anon:159383 isolated_anon:0

So what does this show you?  That someone is takign a ton of ION memory
for some unknown use?  What can you do with that?  What would you do
with that?

And memory is almost never assigned to a "driver", it is assigned to a
"device" that uses it.  Drivers can handle multiple devices at the same
time, so why would you break this down by drivers?  Are you assuming
that a driver only talks to one piece of hardware?

I think you need a much better use case for all of this other than
"wouldn't it be nice to see some numbers", as that isn't going to help
anyone out in the end.

thanks,

greg k-h