netdev - Re: [PATCH net-next] net/fungible: Fix local_memory

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <CAOkoqZ=mPesf-PYDW-Aoq=nMrdfY-RDdJGgWYk48oja_OVxoNw@mail.gmail.com>
Date:   Wed, 9 Mar 2022 01:46:44 -0800
From:   Dimitris Michailidis <d.michailidis@...gible.com>
To:     Jakub Kicinski <kuba@...nel.org>
Cc:     "David S. Miller" <davem@...emloft.net>,
        Networking <netdev@...r.kernel.org>,
        Stephen Rothwell <sfr@...b.auug.org.au>
Subject: Re: [PATCH net-next] net/fungible: Fix local_memory_node error

On Tue, Mar 8, 2022 at 10:14 PM Jakub Kicinski <kuba@...nel.org> wrote:
>
> On Mon,  7 Mar 2022 20:53:21 -0800 Dimitris Michailidis wrote:
> > Stephen Rothwell reported the following failure on powerpc:
> >
> > ERROR: modpost: ".local_memory_node"
> > [drivers/net/ethernet/fungible/funeth/funeth.ko] undefined!
> >
> > AFAICS this is because local_memory_node() is a non-inline non-exported
> > function when CONFIG_HAVE_MEMORYLESS_NODES=y. It is also the wrong API
> > to get a CPU's memory node. Use cpu_to_mem() in the two spots it's used.
>
> Can the ids actually not match? I'm asking because nobody else is doing
> the cpu -> mem node conversions.

They can differ if CONFIG_HAVE_MEMORYLESS_NODES=y and the machine has
memoryless nodes. That config is only offered by IA64 and powerpc so I
guess there are just a few machines where they can differ. It is true
that cpu_to_mem() calls aren't common but the related call
numa_mem_id(), the special case for the local CPU, is easier to find.
For example, page pools's default preferred node is numa_mem_id()
rather than numa_node_id().

> > Fixes: ee6373ddf3a9 ("net/funeth: probing and netdev ops")
> > Fixes: db37bc177dae ("net/funeth: add the data path")
> > Reported-by: Stephen Rothwell <sfr@...b.auug.org.au>
> > Signed-off-by: Dimitris Michailidis <dmichail@...gible.com>
> > ---
> >  drivers/net/ethernet/fungible/funeth/funeth_main.c | 2 +-
> >  drivers/net/ethernet/fungible/funeth/funeth_txrx.h | 2 +-
> >  2 files changed, 2 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/net/ethernet/fungible/funeth/funeth_main.c b/drivers/net/ethernet/fungible/funeth/funeth_main.c
> > index c58b10c216ef..67dd02ed1fa3 100644
> > --- a/drivers/net/ethernet/fungible/funeth/funeth_main.c
> > +++ b/drivers/net/ethernet/fungible/funeth/funeth_main.c
> > @@ -253,7 +253,7 @@ static struct fun_irq *fun_alloc_qirq(struct funeth_priv *fp, unsigned int idx,
> >       int cpu, res;
> >
> >       cpu = cpumask_local_spread(idx, node);
> > -     node = local_memory_node(cpu_to_node(cpu));
> > +     node = cpu_to_mem(cpu);
> >
> >       irq = kzalloc_node(sizeof(*irq), GFP_KERNEL, node);
> >       if (!irq)
> > diff --git a/drivers/net/ethernet/fungible/funeth/funeth_txrx.h b/drivers/net/ethernet/fungible/funeth/funeth_txrx.h
> > index 7aed0561aeac..04c9f91b7489 100644
> > --- a/drivers/net/ethernet/fungible/funeth/funeth_txrx.h
> > +++ b/drivers/net/ethernet/fungible/funeth/funeth_txrx.h
> > @@ -239,7 +239,7 @@ static inline void fun_txq_wr_db(const struct funeth_txq *q)
> >
> >  static inline int fun_irq_node(const struct fun_irq *p)
> >  {
> > -     return local_memory_node(cpu_to_node(cpumask_first(&p->affinity_mask)));
> > +     return cpu_to_mem(cpumask_first(&p->affinity_mask));
> >  }
> >
> >  int fun_rxq_napi_poll(struct napi_struct *napi, int budget);
>