Messages in this thread | | | Subject | Re: [PATCH] net: dsa: properly disconnect the slave PHYs | From | John Crispin <> | Date | Tue, 18 Oct 2016 14:54:30 +0200 |
| |
On 18/10/2016 14:29, Andrew Lunn wrote: > On Tue, Oct 18, 2016 at 02:12:40PM +0200, John Crispin wrote: >> The shutdown code only stopped the PHYs but does not diconnect them >> properly. This could lead to null pointer deref related kernel oopses >> during reboot. Fix this by calling phy_disconnect() after the PHYs are >> stopped. > > Humm, i don't follow this. > > The phy is disconnected in dsa_slave_destroy(). Why is that not > sufficient? > > Also, after calling dsa_slave_close(), dsa_slave_open() can be > called. But with your change, the phy has gone, so we are going to > have trouble. > > Andrew >
Hi Andrew
i am testing on v4.4 which did not have a phy_disconnect() call. this seems to have been fixed by cda5c15b so please ignore this patch
John
>> >> Signed-off-by: John Crispin <john@phrozen.org> >> --- >> net/dsa/slave.c | 4 +++- >> 1 file changed, 3 insertions(+), 1 deletion(-) >> >> diff --git a/net/dsa/slave.c b/net/dsa/slave.c >> index 68714a5..725d9f7 100644 >> --- a/net/dsa/slave.c >> +++ b/net/dsa/slave.c >> @@ -154,8 +154,10 @@ static int dsa_slave_close(struct net_device *dev) >> struct net_device *master = p->parent->dst->master_netdev; >> struct dsa_switch *ds = p->parent; >> >> - if (p->phy) >> + if (p->phy) { >> phy_stop(p->phy); >> + phy_disconnect(p->phy); >> + } >> >> dev_mc_unsync(master, dev); >> dev_uc_unsync(master, dev); >> -- >> 1.7.10.4 >>
| |