2022-11-06 08:34:42

by Deepak R Varma

[permalink] [raw]
Subject: [PATCH] staging: r8188eu: simplify complex pointer casting

Pointers to structures udphdr and dhcpMessage are derived by casting
adjacent pointers with size_t. Such typecast of pointer using size_t
is not preferred. The code looks complex and delicate. Simplify such
casting by utilizing generic "void *" casting.
While at this change, remove the unnecessary __be32 casting for member
variable "cookie".

Suggested-by: Joe Perches <[email protected]>
Signed-off-by: Deepak R Varma <[email protected]>
---
drivers/staging/r8188eu/core/rtw_br_ext.c | 10 ++++++----
1 file changed, 6 insertions(+), 4 deletions(-)

diff --git a/drivers/staging/r8188eu/core/rtw_br_ext.c b/drivers/staging/r8188eu/core/rtw_br_ext.c
index a23f7df373ed..e9b0906d0d74 100644
--- a/drivers/staging/r8188eu/core/rtw_br_ext.c
+++ b/drivers/staging/r8188eu/core/rtw_br_ext.c
@@ -610,13 +610,15 @@ void dhcp_flag_bcast(struct adapter *priv, struct sk_buff *skb)
struct iphdr *iph = (struct iphdr *)(skb->data + ETH_HLEN);

if (iph->protocol == IPPROTO_UDP) { /* UDP */
- struct udphdr *udph = (struct udphdr *)((size_t)iph + (iph->ihl << 2));
+ struct udphdr *udph = (void *)iph + (iph->ihl << 2);

if ((udph->source == htons(CLIENT_PORT)) &&
(udph->dest == htons(SERVER_PORT))) { /* DHCP request */
- struct dhcpMessage *dhcph =
- (struct dhcpMessage *)((size_t)udph + sizeof(struct udphdr));
- u32 cookie = be32_to_cpu((__be32)dhcph->cookie);
+ u32 cookie;
+ struct dhcpMessage *dhcph;
+
+ dhcph = (void *)udph + sizeof(struct udphdr);
+ cookie = be32_to_cpu(dhcph->cookie);

if (cookie == DHCP_MAGIC) { /* match magic word */
if (!(dhcph->flags & htons(BROADCAST_FLAG))) {
--
2.34.1





2022-11-06 09:02:57

by Philipp Hortmann

[permalink] [raw]
Subject: Re: [PATCH] staging: r8188eu: simplify complex pointer casting

On 11/6/22 09:26, Deepak R Varma wrote:
> Pointers to structures udphdr and dhcpMessage are derived by casting
> adjacent pointers with size_t. Such typecast of pointer using size_t
> is not preferred. The code looks complex and delicate. Simplify such
> casting by utilizing generic "void *" casting.
> While at this change, remove the unnecessary __be32 casting for member
> variable "cookie".
>
> Suggested-by: Joe Perches <[email protected]>
> Signed-off-by: Deepak R Varma <[email protected]>
> ---
> drivers/staging/r8188eu/core/rtw_br_ext.c | 10 ++++++----
> 1 file changed, 6 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/staging/r8188eu/core/rtw_br_ext.c b/drivers/staging/r8188eu/core/rtw_br_ext.c
> index a23f7df373ed..e9b0906d0d74 100644
> --- a/drivers/staging/r8188eu/core/rtw_br_ext.c
> +++ b/drivers/staging/r8188eu/core/rtw_br_ext.c
> @@ -610,13 +610,15 @@ void dhcp_flag_bcast(struct adapter *priv, struct sk_buff *skb)
> struct iphdr *iph = (struct iphdr *)(skb->data + ETH_HLEN);
>
> if (iph->protocol == IPPROTO_UDP) { /* UDP */
> - struct udphdr *udph = (struct udphdr *)((size_t)iph + (iph->ihl << 2));
> + struct udphdr *udph = (void *)iph + (iph->ihl << 2);
>
> if ((udph->source == htons(CLIENT_PORT)) &&
> (udph->dest == htons(SERVER_PORT))) { /* DHCP request */
> - struct dhcpMessage *dhcph =
> - (struct dhcpMessage *)((size_t)udph + sizeof(struct udphdr));
> - u32 cookie = be32_to_cpu((__be32)dhcph->cookie);
> + u32 cookie;
> + struct dhcpMessage *dhcph;
> +
> + dhcph = (void *)udph + sizeof(struct udphdr);
> + cookie = be32_to_cpu(dhcph->cookie);
>
> if (cookie == DHCP_MAGIC) { /* match magic word */
> if (!(dhcph->flags & htons(BROADCAST_FLAG))) {
> --
> 2.34.1
>
>
>
>
Tested-by: Philipp Hortmann <[email protected]> # Edimax N150

2022-11-06 16:54:32

by Greg Kroah-Hartman

[permalink] [raw]
Subject: Re: [PATCH] staging: r8188eu: simplify complex pointer casting

On Sun, Nov 06, 2022 at 01:56:01PM +0530, Deepak R Varma wrote:
> Pointers to structures udphdr and dhcpMessage are derived by casting
> adjacent pointers with size_t. Such typecast of pointer using size_t
> is not preferred. The code looks complex and delicate. Simplify such
> casting by utilizing generic "void *" casting.
> While at this change, remove the unnecessary __be32 casting for member
> variable "cookie".

Wait, why is that unecessary?

And why is that part of this change, that should be a separate one, so
that it can be reverted when it's found to be buggy :)

>
> Suggested-by: Joe Perches <[email protected]>
> Signed-off-by: Deepak R Varma <[email protected]>
> ---
> drivers/staging/r8188eu/core/rtw_br_ext.c | 10 ++++++----
> 1 file changed, 6 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/staging/r8188eu/core/rtw_br_ext.c b/drivers/staging/r8188eu/core/rtw_br_ext.c
> index a23f7df373ed..e9b0906d0d74 100644
> --- a/drivers/staging/r8188eu/core/rtw_br_ext.c
> +++ b/drivers/staging/r8188eu/core/rtw_br_ext.c
> @@ -610,13 +610,15 @@ void dhcp_flag_bcast(struct adapter *priv, struct sk_buff *skb)
> struct iphdr *iph = (struct iphdr *)(skb->data + ETH_HLEN);
>
> if (iph->protocol == IPPROTO_UDP) { /* UDP */
> - struct udphdr *udph = (struct udphdr *)((size_t)iph + (iph->ihl << 2));
> + struct udphdr *udph = (void *)iph + (iph->ihl << 2);
>
> if ((udph->source == htons(CLIENT_PORT)) &&
> (udph->dest == htons(SERVER_PORT))) { /* DHCP request */
> - struct dhcpMessage *dhcph =
> - (struct dhcpMessage *)((size_t)udph + sizeof(struct udphdr));
> - u32 cookie = be32_to_cpu((__be32)dhcph->cookie);
> + u32 cookie;
> + struct dhcpMessage *dhcph;

Why reorder these variables?

> +
> + dhcph = (void *)udph + sizeof(struct udphdr);
> + cookie = be32_to_cpu(dhcph->cookie);

The cookie change should be separate please.

thanks,

greg k-h

2022-11-06 18:22:00

by Deepak R Varma

[permalink] [raw]
Subject: Re: [PATCH] staging: r8188eu: simplify complex pointer casting

On Sun, Nov 06, 2022 at 05:43:38PM +0100, Greg Kroah-Hartman wrote:
> On Sun, Nov 06, 2022 at 01:56:01PM +0530, Deepak R Varma wrote:
> > Pointers to structures udphdr and dhcpMessage are derived by casting
> > adjacent pointers with size_t. Such typecast of pointer using size_t
> > is not preferred. The code looks complex and delicate. Simplify such
> > casting by utilizing generic "void *" casting.
> > While at this change, remove the unnecessary __be32 casting for member
> > variable "cookie".
>
> Wait, why is that unecessary?

Hello Greg,
cookie is already defined to be a __be32 type as part of the dhcpMesssade
structure declared in the same file.

15 struct dhcpMessage {
14 u_int8_t op;
1 u_int8_t file[128];
597 __be32 cookie;
1 u_int8_t options[308]; /* 312 - cookie */
2 };
3

>
> And why is that part of this change, that should be a separate one, so
> that it can be reverted when it's found to be buggy :)

Sounds good. I will submit that as a separate patch.
>
> >
> > Suggested-by: Joe Perches <[email protected]>
> > Signed-off-by: Deepak R Varma <[email protected]>
> > ---
> > drivers/staging/r8188eu/core/rtw_br_ext.c | 10 ++++++----
> > 1 file changed, 6 insertions(+), 4 deletions(-)
> >
> > diff --git a/drivers/staging/r8188eu/core/rtw_br_ext.c b/drivers/staging/r8188eu/core/rtw_br_ext.c
> > index a23f7df373ed..e9b0906d0d74 100644
> > --- a/drivers/staging/r8188eu/core/rtw_br_ext.c
> > +++ b/drivers/staging/r8188eu/core/rtw_br_ext.c
> > @@ -610,13 +610,15 @@ void dhcp_flag_bcast(struct adapter *priv, struct sk_buff *skb)
> > struct iphdr *iph = (struct iphdr *)(skb->data + ETH_HLEN);
> >
> > if (iph->protocol == IPPROTO_UDP) { /* UDP */
> > - struct udphdr *udph = (struct udphdr *)((size_t)iph + (iph->ihl << 2));
> > + struct udphdr *udph = (void *)iph + (iph->ihl << 2);
> >
> > if ((udph->source == htons(CLIENT_PORT)) &&
> > (udph->dest == htons(SERVER_PORT))) { /* DHCP request */
> > - struct dhcpMessage *dhcph =
> > - (struct dhcpMessage *)((size_t)udph + sizeof(struct udphdr));
> > - u32 cookie = be32_to_cpu((__be32)dhcph->cookie);
> > + u32 cookie;
> > + struct dhcpMessage *dhcph;
>
> Why reorder these variables?

No specific reason. Since next instruction is associated with *dhcph, I thought
it will simpler to follow. I will revise and resend.

>
> > +
> > + dhcph = (void *)udph + sizeof(struct udphdr);
> > + cookie = be32_to_cpu(dhcph->cookie);
>
> The cookie change should be separate please.

Sounds good. Will split the patch into a patch set and resubmit.
>
> thanks,
>
> greg k-h
>