netdev - Re: 答复: [PATCH] net: ethtool: not call vzalloc for zero sized memory request

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]

Date:   Thu, 28 Mar 2019 11:25:54 +0100
From:   Michal Kubecek <mkubecek@...e.cz>
To:     "Li,Rongqing" <lirongqing@...du.com>
Cc:     "netdev@...r.kernel.org" <netdev@...r.kernel.org>
Subject: Re: 答复: [PATCH] net:
 ethtool: not call vzalloc for zero sized memory request

On Thu, Mar 28, 2019 at 09:51:56AM +0000, Li,Rongqing wrote:
> > -----邮件原件-----
> > 发件人: Michal Kubecek [mailto:mkubecek@...e.cz]
> > 发送时间: 2019年3月28日 17:09
> > 收件人: Li,Rongqing <lirongqing@...du.com>
> > 抄送: netdev@...r.kernel.org
> > 主题: Re: [PATCH] net: ethtool: not call vzalloc for zero sized memory request
> > 
> > On Thu, Mar 28, 2019 at 02:01:09PM +0800, Li RongQing wrote:
> > > NULL or ZERO_SIZE_PTR will be returned for zero sized memory request,
> > > and derefencing them will lead to a segfault
> > >
> > > so it is unnecessory to call vzalloc for zero sized memory request and
> > > not call __ethtool_get_strings which always uses the allocated memory
> > >
> > > this also fixes a possible memory leak if phy_ethtool_get_stats
> > > returns error, memory should be freed before exit
> > >
> > > Signed-off-by: Li RongQing <lirongqing@...du.com>
> > > Reviewed-by: Wang Li <wangli39@...du.com>
> > > ---
> > >  net/core/ethtool.c | 37 ++++++++++++++++++++++++++-----------
> > >  1 file changed, 26 insertions(+), 11 deletions(-)
> > >
> > > diff --git a/net/core/ethtool.c b/net/core/ethtool.c index
> > > b1eb32419732..3e971a36e37c 100644
> > > --- a/net/core/ethtool.c
> > > +++ b/net/core/ethtool.c
> > ...
> > > @@ -1897,9 +1902,14 @@ static int ethtool_get_stats(struct net_device
> > *dev, void __user *useraddr)
> > >  		return -EFAULT;
> > >
> > >  	stats.n_stats = n_stats;
> > > -	data = vzalloc(array_size(n_stats, sizeof(u64)));
> > > -	if (n_stats && !data)
> > > -		return -ENOMEM;
> > > +
> > > +	if (n_stats) {
> > > +		data = vzalloc(array_size(n_stats, sizeof(u64)));
> > > +		if (!data)
> > > +			return -ENOMEM;
> > > +	} else {
> > > +		data = NULL;
> > > +	}
> > >
> > >  	ops->get_ethtool_stats(dev, &stats, data);
> > >
> > 
> > You avoid the vzalloc() call here but you still pass null data pointer to device's
> > get_ethtool_stats() handler which seems to contradict what the commit
> > message says. Is it really what you want? (The same applies to
> > ethtool_get_phy_stats() below.)
> > 
> 
> 
> I keep it deliberately
> 
> ops->get_ethtool_stats(dev, &stats, data) have three parameter, 
> if n_stats is 0 [we assume the returning 0 of ops->get_sset_count is correct, or 
> get_sset_count should be fixed], data is NULL,  get_ethtool_stats () maybe still 
> store the data into its seconds parameter, stats, even if I did not find which drivers
> is like that

stats is a local variable declared as

	struct ethtool_stats stats;

where struct ethtool_stats which looks like

struct ethtool_stats {
	__u32	cmd;
	__u32	n_stats;
	__u64	data[0];
};

so that storing anything to stats.data would result in stack corruption.
This is why ethtool_ops::get_ethtool_stats() has the third parameter
telling it where to put the statistics.

There isn't much point touching cmd and n_stats should be set to the
same value as we got from ethtool_ops::get_sset_counts earlier, i.e.
zero in our case.

Michal