[RFC net-next] ethtool: add a stricter length check

[RFC net-next] ethtool: add a stricter length check

[Jakub Kicinski]

There has been a few errors in the ethtool reply size calculations,
most of those are hard to trigger during basic testing because of
skb size rounding up and netdev names being shorter than max.
Add a more precise check.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
Michal, WDYT?

 net/ethtool/netlink.c | 11 ++++++++---
 1 file changed, 8 insertions(+), 3 deletions(-)

diff --git a/net/ethtool/netlink.c b/net/ethtool/netlink.c
index 88d8a0243f35..3f9a1a96b4df 100644
--- a/net/ethtool/netlink.c
+++ b/net/ethtool/netlink.c
@@ -315,9 +315,9 @@ static int ethnl_default_doit(struct sk_buff *skb, struct genl_info *info)
 	struct ethnl_req_info *req_info = NULL;
 	const u8 cmd = info->genlhdr->cmd;
 	const struct ethnl_request_ops *ops;
+	int hdr_len, reply_len;
 	struct sk_buff *rskb;
 	void *reply_payload;
-	int reply_len;
 	int ret;
 
 	ops = ethnl_default_requests[cmd];
@@ -346,15 +346,20 @@ static int ethnl_default_doit(struct sk_buff *skb, struct genl_info *info)
 	ret = ops->reply_size(req_info, reply_data);
 	if (ret < 0)
 		goto err_cleanup;
-	reply_len = ret + ethnl_reply_header_size();
+	reply_len = ret;
 	ret = -ENOMEM;
-	rskb = ethnl_reply_init(reply_len, req_info->dev, ops->reply_cmd,
+	rskb = ethnl_reply_init(reply_len + ethnl_reply_header_size(),
+				req_info->dev, ops->reply_cmd,
 				ops->hdr_attr, info, &reply_payload);
 	if (!rskb)
 		goto err_cleanup;
+	hdr_len = rskb->len;
 	ret = ops->fill_reply(rskb, req_info, reply_data);
 	if (ret < 0)
 		goto err_msg;
+	WARN(rskb->len - hdr_len > reply_len,
+	     "ethnl cmd %d: calculated reply length %d, but consumed %d\n",
+	     cmd, reply_len, rskb->len - hdr_len);
 	if (ops->cleanup_data)
 		ops->cleanup_data(reply_data);
 
-- 
2.31.1

Re: [RFC net-next] ethtool: add a stricter length check

[Michal Kubecek]

On Fri, Jun 11, 2021 at 08:11:35PM -0700, Jakub Kicinski wrote:
> There has been a few errors in the ethtool reply size calculations,
> most of those are hard to trigger during basic testing because of
> skb size rounding up and netdev names being shorter than max.
> Add a more precise check.
> 
> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
> ---
> Michal, WDYT?

It's definitely an improvement and I agree with it. I only have two
minor comments below.

> 
>  net/ethtool/netlink.c | 11 ++++++++---
>  1 file changed, 8 insertions(+), 3 deletions(-)
> 
> diff --git a/net/ethtool/netlink.c b/net/ethtool/netlink.c
> index 88d8a0243f35..3f9a1a96b4df 100644
> --- a/net/ethtool/netlink.c
> +++ b/net/ethtool/netlink.c
> @@ -315,9 +315,9 @@ static int ethnl_default_doit(struct sk_buff *skb, struct genl_info *info)
>  	struct ethnl_req_info *req_info = NULL;
>  	const u8 cmd = info->genlhdr->cmd;
>  	const struct ethnl_request_ops *ops;
> +	int hdr_len, reply_len;
>  	struct sk_buff *rskb;
>  	void *reply_payload;
> -	int reply_len;
>  	int ret;
>  
>  	ops = ethnl_default_requests[cmd];
> @@ -346,15 +346,20 @@ static int ethnl_default_doit(struct sk_buff *skb, struct genl_info *info)
>  	ret = ops->reply_size(req_info, reply_data);
>  	if (ret < 0)
>  		goto err_cleanup;
> -	reply_len = ret + ethnl_reply_header_size();
> +	reply_len = ret;
>  	ret = -ENOMEM;
> -	rskb = ethnl_reply_init(reply_len, req_info->dev, ops->reply_cmd,
> +	rskb = ethnl_reply_init(reply_len + ethnl_reply_header_size(),
> +				req_info->dev, ops->reply_cmd,
>  				ops->hdr_attr, info, &reply_payload);
>  	if (!rskb)
>  		goto err_cleanup;
> +	hdr_len = rskb->len;
>  	ret = ops->fill_reply(rskb, req_info, reply_data);
>  	if (ret < 0)
>  		goto err_msg;
> +	WARN(rskb->len - hdr_len > reply_len,
> +	     "ethnl cmd %d: calculated reply length %d, but consumed %d\n",
> +	     cmd, reply_len, rskb->len - hdr_len);
>  	if (ops->cleanup_data)
>  		ops->cleanup_data(reply_data);

We may want WARN_ONCE or ratelimited here, if there is bug in reply
length estimate for a request not requiring admin privileges, the
warning might be invoked by a regular user at will.

Also the patch changes the meaning of reply_len which is also used in
the original warning after err_msg label. But it's probably not a big
deal, it's not obvious what exactly "payload" means there so that anyone
trying to investigate the problem has to start by checking what exactly
the value reported means.

Michal

Re: [RFC net-next] ethtool: add a stricter length check

[Jakub Kicinski]

On Wed, 16 Jun 2021 01:10:33 +0200 Michal Kubecek wrote:
> > @@ -346,15 +346,20 @@ static int ethnl_default_doit(struct sk_buff *skb, struct genl_info *info)
> >  	ret = ops->reply_size(req_info, reply_data);
> >  	if (ret < 0)
> >  		goto err_cleanup;
> > -	reply_len = ret + ethnl_reply_header_size();
> > +	reply_len = ret;
> >  	ret = -ENOMEM;
> > -	rskb = ethnl_reply_init(reply_len, req_info->dev, ops->reply_cmd,
> > +	rskb = ethnl_reply_init(reply_len + ethnl_reply_header_size(),
> > +				req_info->dev, ops->reply_cmd,
> >  				ops->hdr_attr, info, &reply_payload);
> >  	if (!rskb)
> >  		goto err_cleanup;
> > +	hdr_len = rskb->len;
> >  	ret = ops->fill_reply(rskb, req_info, reply_data);
> >  	if (ret < 0)
> >  		goto err_msg;
> > +	WARN(rskb->len - hdr_len > reply_len,
> > +	     "ethnl cmd %d: calculated reply length %d, but consumed %d\n",
> > +	     cmd, reply_len, rskb->len - hdr_len);
> >  	if (ops->cleanup_data)
> >  		ops->cleanup_data(reply_data);  
> 
> We may want WARN_ONCE or ratelimited here, if there is bug in reply
> length estimate for a request not requiring admin privileges, the
> warning might be invoked by a regular user at will.

Ah, good point!

> Also the patch changes the meaning of reply_len which is also used in
> the original warning after err_msg label. But it's probably not a big
> deal, it's not obvious what exactly "payload" means there so that anyone
> trying to investigate the problem has to start by checking what exactly
> the value reported means.

I'll add a note to this effect to the commit message.

Thanks!