From: Greg KH <gregkh@linuxfoundation.org>
To: Qi Zheng <zhengqi.arch@bytedance.com>
Cc: akpm@linux-foundation.org, rafael@kernel.org, linux-mm@kvack.org,
linux-kernel@vger.kernel.org, songmuchun@bytedance.com
Subject: Re: [PATCH 1/2] mm: add MemAvailable to per-node meminfo
Date: Thu, 16 Dec 2021 16:37:06 +0100 [thread overview]
Message-ID: <YbtdIpJal3keEWf8@kroah.com> (raw)
In-Reply-To: <34cc2f26-9fba-f7c6-5dae-d21498ccbe2d@bytedance.com>
On Thu, Dec 16, 2021 at 11:31:36PM +0800, Qi Zheng wrote:
>
>
> On 12/16/21 9:16 PM, Greg KH wrote:
> > On Thu, Dec 16, 2021 at 08:46:54PM +0800, Qi Zheng wrote:
> > > In /proc/meminfo, we can show the sum of all the available memory
> > > as "MemAvailable". Add the same counter also to per-node meminfo
> > > under /sys.
> > >
> > > With this counter, some processes that bind nodes can make some
> > > decisions by reading the "MemAvailable" of the corresponding nodes
> > > directly.
> > >
> > > Signed-off-by: Qi Zheng <zhengqi.arch@bytedance.com>
> > > ---
> > > drivers/base/node.c | 4 ++++
> > > include/linux/mm.h | 1 +
> > > include/linux/mmzone.h | 5 +++++
> > > mm/page_alloc.c | 49 +++++++++++++++++++++++++++++++++++++++++++++++++
> > > 4 files changed, 59 insertions(+)
> > >
> > > diff --git a/drivers/base/node.c b/drivers/base/node.c
> > > index 87acc47e8951..deb2a7965ae4 100644
> > > --- a/drivers/base/node.c
> > > +++ b/drivers/base/node.c
> > > @@ -375,8 +375,10 @@ static ssize_t node_read_meminfo(struct device *dev,
> > > struct sysinfo i;
> > > unsigned long sreclaimable, sunreclaimable;
> > > unsigned long swapcached = 0;
> > > + long available;
> > > si_meminfo_node(&i, nid);
> > > + available = si_mem_available_node(&i, nid);
> > > sreclaimable = node_page_state_pages(pgdat, NR_SLAB_RECLAIMABLE_B);
> > > sunreclaimable = node_page_state_pages(pgdat, NR_SLAB_UNRECLAIMABLE_B);
> > > #ifdef CONFIG_SWAP
> > > @@ -386,6 +388,7 @@ static ssize_t node_read_meminfo(struct device *dev,
> > > "Node %d MemTotal: %8lu kB\n"
> > > "Node %d MemFree: %8lu kB\n"
> > > "Node %d MemUsed: %8lu kB\n"
> > > + "Node %d MemAvailable: %8lu kB\n"
> >
> > You just changed a user/kernel api without documenting it anywhere, or
> > ensuring that you did not just break anything.
>
> Hi greg k-h,
>
> The MemAvailable has long existed in the /proc/meminfo, it's meaning
> has been described in the Documentation/filesystems/proc.rst. Since
> the semantics of per-node MemAvailable has not changed, so I did not
> add a new document description.
This is not a proc file, it is in sysfs.
And it violates all of the sysfs rules, and has all of the problems that
proc files have. So the worst of both worlds :(
> > Also, this api is crazy, and not ok, please never add anything new to
> > it, it is broken as-is.
>
> The consideration of adding per-node MemAvailable is that some processes
> that bind nodes need this information to do some decisions.
>
> Now their approach is to read other information in per-node meminfo
> and /proc/sys/vm/watermark_scale_factor, and then approximate this
> value. With this counter, they can directly read
> /sys/devices/system/node/node*/meminfo to get the MemAvailable
> information of each node.
>
> And MemTotal, MemFree and SReclaimable(etc.) all have corresponding
> per-node versions, so I think that adding per-node MemAvailable might
> also make sense. :)
Please no, I do not want new things added to this file, as you might
break parsers of this file.
Also, again, you did not document this at all in Documentation/ABI/ so
for that reason alone it is not ok.
thanks,
greg k-h
next prev parent reply other threads:[~2021-12-16 15:37 UTC|newest]
Thread overview: 12+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-12-16 12:46 [PATCH 0/2] add MemAvailable to per-node meminfo Qi Zheng
2021-12-16 12:46 ` [PATCH 1/2] mm: " Qi Zheng
2021-12-16 13:16 ` Greg KH
2021-12-16 15:31 ` Qi Zheng
2021-12-16 15:37 ` Greg KH [this message]
2021-12-16 15:43 ` Qi Zheng
2021-12-16 12:46 ` [PATCH 2/2] mm: reimplement si_mem_available() Qi Zheng
2021-12-16 13:17 ` Greg KH
2021-12-16 15:39 ` Qi Zheng
2021-12-16 15:06 ` kernel test robot
2021-12-16 18:21 ` kernel test robot
2021-12-16 21:05 ` kernel test robot
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=YbtdIpJal3keEWf8@kroah.com \
--to=gregkh@linuxfoundation.org \
--cc=akpm@linux-foundation.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=rafael@kernel.org \
--cc=songmuchun@bytedance.com \
--cc=zhengqi.arch@bytedance.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).