Re: [RFC PATCH] mm: net: disable kswapd for high-order network buffer allocation

[Barry Song <21cnbao@gmail.com>]

On Wed, Oct 15, 2025 at 2:39 PM Eric Dumazet <edumazet@google.com> wrote:

> > >
> > > Tell them they are wrong.
> >
> > Well, we checked Qualcomm and MTK, and it seems both set these values
> > relatively high. In other words, all the AOSP products we examined also
> > use high values for these settings. Nobody is using tcp_wmem[0]=4096.
> >
>
> The (fine and safe) default should be PAGE_SIZE.
>
> Perhaps they are dealing with systems with PAGE_SIZE=65536, but then
> the skb_page_frag_refill() would be a non issue there, because it would
> only allocate order-0 pages.

I am 100% sure that all of them handle PAGE_SIZE=4096. Google is working on
16KB page size for Android, but it is not ready yet(Please correct me
if 16KB has been
ready, Suren).

>
> > We’ll need some time to understand why these are configured this way in
> > AOSP hardware.
> >
> > >
> > > >
> > > > It might be worth exploring these settings further, but I can’t quite see
> > > > their connection to high-order allocations, since high-order allocations are
> > > > kernel macros.
> > > >
> > > > #define SKB_FRAG_PAGE_ORDER     get_order(32768)
> > > > #define PAGE_FRAG_CACHE_MAX_SIZE        __ALIGN_MASK(32768, ~PAGE_MASK)
> > > > #define PAGE_FRAG_CACHE_MAX_ORDER       get_order(PAGE_FRAG_CACHE_MAX_SIZE)
> > > >
> > > > Is there anything I’m missing?
> > >
> > > What is your question exactly ? You read these macros just fine. What
> > > is your point ?
> >
> > My question is whether these settings influence how often high-order
> > allocations occur. In other words, would lowering these values make
> > high-order allocations less frequent? If so, why?
>
> Because almost all of the buffers stored in TCP write queues are using
> order-3 pages
> on arches with 4K pages.
>
> I am a bit confused because you posted a patch changing skb_page_frag_refill()
> without realizing its first user is TCP.
>
> Look for sk_page_frag_refill() in tcp_sendmsg_locked()

Sure. Let me review the code further. The problem was observed on the MM
side, causing over-reclamation and phone heating, while the source of the
allocations lies in network activity. I am not a network expert and may be
missing many network details, so I am raising this RFC to both lists to see
if the network and MM folks can discuss together to find a solution.

As you can see, the discussion has absolutely forked into two branches. :-)

Thanks
Barry