sizeof pointer vs sizeof struct

sizeof pointer vs sizeof struct

[Logan B]

Hello,
I was recently working through the xdp tutorial and in the
packet01-parsing lesson the sizeof a pointer to the ethernet header
struct is used, not the sizeof the struct itself[0]. I peeked and the
solution for this section also still uses the sizeof a pointer and not
the struct so this isn't part of the tutorial and I was wondering what
is going on? I don't think the verifier is re-writing these addresses,
only those for the memory access into the packet data.

#include <stdio.h>
#include <linux/if_ether.h>

int main(void)

       {

                           struct ethhdr normal = {0};
    struct ethhdr *eth_hdr_ptr;

     printf("Size of struct %lu\n",sizeof(normal)); // prints 14
    printf("Size of struct pointer %lu\n",sizeof(eth_hdr_ptr)); //
prints 8
    return 0;
}


[0]https://github.com/xdp-project/xdp-tutorial/blob/master/packet01-parsing/xdp_prog_kern.c#L34
--
Logan

Re: sizeof pointer vs sizeof struct

[Jay Vosburgh]

Logan B <mrbojangles3@gmail.com> wrote:

>Hello,
>I was recently working through the xdp tutorial and in the
>packet01-parsing lesson the sizeof a pointer to the ethernet header
>struct is used, not the sizeof the struct itself[0]. I peeked and the
>solution for this section also still uses the sizeof a pointer and not
>the struct so this isn't part of the tutorial and I was wondering what
>is going on? I don't think the verifier is re-writing these addresses,
>only those for the memory access into the packet data.

	The referenced code at [0] is:

	struct ethhdr *eth = nh->pos;
	int hdrsize = sizeof(*eth);

	"*eth" means "what eth points to," so this is indeed taking the
sizeof struct ethhdr.

	I suspect you missed the "*" in your reading of the code; in
this context, "*" is the indirection operator, per K&R 2, (The C
Programming Language, 2nd Edition), Appendix A 7.4.3.

	-J

>
>#include <stdio.h>
>#include <linux/if_ether.h>
>
>int main(void)
>
>       {
>
>                           struct ethhdr normal = {0};
>    struct ethhdr *eth_hdr_ptr;
>
>     printf("Size of struct %lu\n",sizeof(normal)); // prints 14
>    printf("Size of struct pointer %lu\n",sizeof(eth_hdr_ptr)); //
>prints 8
>    return 0;
>}
>
>
>[0]https://github.com/xdp-project/xdp-tutorial/blob/master/packet01-parsing/xdp_prog_kern.c#L34
>--
>Logan
>

---
	-Jay Vosburgh, jay.vosburgh@canonical.com

Re: sizeof pointer vs sizeof struct

[Logan B]

On Thu, Feb 29, 2024 at 9:10 PM Jay Vosburgh <jay.vosburgh@canonical.com> wrote:
>
> Logan B <mrbojangles3@gmail.com> wrote:
>
> >Hello,
> >I was recently working through the xdp tutorial and in the
> >packet01-parsing lesson the sizeof a pointer to the ethernet header
> >struct is used, not the sizeof the struct itself[0]. I peeked and the
> >solution for this section also still uses the sizeof a pointer and not
> >the struct so this isn't part of the tutorial and I was wondering what
> >is going on? I don't think the verifier is re-writing these addresses,
> >only those for the memory access into the packet data.
>
>         The referenced code at [0] is:
>
>         struct ethhdr *eth = nh->pos;
>         int hdrsize = sizeof(*eth);
>
>         "*eth" means "what eth points to," so this is indeed taking the
> sizeof struct ethhdr.
>
>         I suspect you missed the "*" in your reading of the code; in
> this context, "*" is the indirection operator, per K&R 2, (The C
> Programming Language, 2nd Edition), Appendix A 7.4.3.
>
>         -J
>
I did miss that. Thank you
Logan
> >
> >#include <stdio.h>
> >#include <linux/if_ether.h>
> >
> >int main(void)
> >
> >       {
> >
> >                           struct ethhdr normal = {0};
> >    struct ethhdr *eth_hdr_ptr;
> >
> >     printf("Size of struct %lu\n",sizeof(normal)); // prints 14
> >    printf("Size of struct pointer %lu\n",sizeof(eth_hdr_ptr)); //
> >prints 8
> >    return 0;
> >}
> >
> >
> >[0]https://github.com/xdp-project/xdp-tutorial/blob/master/packet01-parsing/xdp_prog_kern.c#L34
> >--
> >Logan
> >
>
> ---
>         -Jay Vosburgh, jay.vosburgh@canonical.com



-- 
Logan