[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <000601d2c333$abb526d0$031f7470$@vip.163.com>
Date: Tue, 2 May 2017 19:03:02 +0800
From: "Gao Feng" <gfree.wind@....163.com>
To: "'Xin Long'" <lucien.xin@...il.com>
Cc: "'davem'" <davem@...emloft.net>, <jarod@...hat.com>,
"'Stephen Hemminger'" <stephen@...workplumber.org>,
<dsa@...ulusnetworks.com>, "'network dev'" <netdev@...r.kernel.org>
Subject: RE: [PATCH net v3] driver: veth: Fix one possbile memleak when fail to register_netdevice
> From: Xin Long [mailto:lucien.xin@...il.com]
> Sent: Tuesday, May 2, 2017 3:56 PM
> On Sat, Apr 29, 2017 at 11:51 AM, <gfree.wind@...mail.com> wrote:
> > From: Gao Feng <gfree.wind@...mail.com>
[...]
> > -static void veth_dev_free(struct net_device *dev)
> > +static void veth_destructor_free(struct net_device *dev)
> > {
> > free_percpu(dev->vstats);
> > +}
> not sure why you needed to add this function.
> to use free_percpu() directly may be clearer.
Because both of ndo_uninit and destructor need to perform same free statements.
It is good at maintain the codes with the common function.
>
> > +
> > +static void veth_dev_uninit(struct net_device *dev) {
> call free_percpu() here, no need to check dev->reg_state.
> free_percpu will just return if dev->vstats is NULL.
It would break the original design if don't check the reg_state.
The original logic is that free the resources in the destructor, not in ndo_init.
BTW, because I send multiple patches too fast today, the email server blocks my account.
So I have to reply you with a different email account. Sorry.
Best Regards
Feng
>
[...]
Powered by blists - more mailing lists