[dpdk-dev] Segmentation Fault on printf()

[dpdk-dev] Segmentation Fault on printf()

[sabu kurian]

Hello friends,

I'm trying to print the ether_type for a packet that I captured from a port
on my machine. Suppose 'm' holds the packet. 'm' is of type 'struct
rte_mbuf'. Intels API reference for DPDK says 'ether_type' is of uint16_t.
I used the following code to retrieve ether_type.

void * eth_type;
struct ether_hdr *eth;

eth = rte_pktmbuf_mtod(m, struct ether_hdr *);
eth_type = &eth->ether_type;

printf("\n Type is %" PRIu16 , *((uint16_t *)eth_type));

rte_pktmbuf_free(m);

I get a 'Segmentation fault' when the printf() statement gets executed.
Where am I going wrong ?

Thanks in advance

Re: [dpdk-dev] Segmentation Fault on printf()

[Chris Wright]

* sabu kurian (sabu2kurian@gmail.com) wrote:
> Hello friends,
> 
> I'm trying to print the ether_type for a packet that I captured from a port
> on my machine. Suppose 'm' holds the packet. 'm' is of type 'struct
> rte_mbuf'. Intels API reference for DPDK says 'ether_type' is of uint16_t.
> I used the following code to retrieve ether_type.
> 
> void * eth_type;
> struct ether_hdr *eth;
> 
> eth = rte_pktmbuf_mtod(m, struct ether_hdr *);
> eth_type = &eth->ether_type;
> 
> printf("\n Type is %" PRIu16 , *((uint16_t *)eth_type));

Looks ok, albeit slightly overly complicated.

struct ether_hdr *eth;
uint16_t eth_type;

eth = rte_pktmbuf_mtod(m, struct ether_hdr *);
eth_type = eth->ether_type;
printf("\n Type is %" PRIu16 , eth_type));

This would remove all the extra casting.  Perhaps the mbuf is the problem
here.  And if so, the above snippet would segfault on eth->ether_type
showing you that mbuf is invalid.

You could try to (above mtod):

rte_pktmbuf_dump(m, sizeof(struct ether_hdr));

as that will show key contents of mbuf and packet data (and do some
basic validation along the way).

thanks,
-chris

Re: [dpdk-dev] Segmentation Fault on printf()

[sabu kurian]

Hello chris,

Thanks for your reply. I tried dumping the contents of mbuf as you have
suggested.
This is what I get:

dump mbuf at 0x0x6c7620, phys=7f34d91eba28, buf_len=59488
pkt_len=0, ol_flags=0, nb_segs=0, in_port=0

and soon after that I get a segmentation fault. Do you have any idea on
what could be causing it ?

Regards



On Fri, Mar 14, 2014 at 9:40 PM, Chris Wright <chrisw@redhat.com> wrote:

> * sabu kurian (sabu2kurian@gmail.com) wrote:
> > Hello friends,
> >
> > I'm trying to print the ether_type for a packet that I captured from a
> port
> > on my machine. Suppose 'm' holds the packet. 'm' is of type 'struct
> > rte_mbuf'. Intels API reference for DPDK says 'ether_type' is of
> uint16_t.
> > I used the following code to retrieve ether_type.
> >
> > void * eth_type;
> > struct ether_hdr *eth;
> >
> > eth = rte_pktmbuf_mtod(m, struct ether_hdr *);
> > eth_type = &eth->ether_type;
> >
> > printf("\n Type is %" PRIu16 , *((uint16_t *)eth_type));
>
> Looks ok, albeit slightly overly complicated.
>
> struct ether_hdr *eth;
> uint16_t eth_type;
>
> eth = rte_pktmbuf_mtod(m, struct ether_hdr *);
> eth_type = eth->ether_type;
> printf("\n Type is %" PRIu16 , eth_type));
>
> This would remove all the extra casting.  Perhaps the mbuf is the problem
> here.  And if so, the above snippet would segfault on eth->ether_type
> showing you that mbuf is invalid.
>
> You could try to (above mtod):
>
> rte_pktmbuf_dump(m, sizeof(struct ether_hdr));
>
> as that will show key contents of mbuf and packet data (and do some
> basic validation along the way).
>
> thanks,
> -chris
>

Re: [dpdk-dev] Segmentation Fault on printf()

[Daniel Kaminsky]

Sabu,

I think that in this case the packet length is zero which will result in
nothing from mtod.

Daniel


On Sun, Mar 16, 2014 at 3:03 PM, sabu kurian <sabu2kurian@gmail.com> wrote:

> Hello chris,
>
> Thanks for your reply. I tried dumping the contents of mbuf as you have
> suggested.
> This is what I get:
>
> dump mbuf at 0x0x6c7620, phys=7f34d91eba28, buf_len=59488
> pkt_len=0, ol_flags=0, nb_segs=0, in_port=0
>
> and soon after that I get a segmentation fault. Do you have any idea on
> what could be causing it ?
>
> Regards
>
>
>
> On Fri, Mar 14, 2014 at 9:40 PM, Chris Wright <chrisw@redhat.com> wrote:
>
> > * sabu kurian (sabu2kurian@gmail.com) wrote:
> > > Hello friends,
> > >
> > > I'm trying to print the ether_type for a packet that I captured from a
> > port
> > > on my machine. Suppose 'm' holds the packet. 'm' is of type 'struct
> > > rte_mbuf'. Intels API reference for DPDK says 'ether_type' is of
> > uint16_t.
> > > I used the following code to retrieve ether_type.
> > >
> > > void * eth_type;
> > > struct ether_hdr *eth;
> > >
> > > eth = rte_pktmbuf_mtod(m, struct ether_hdr *);
> > > eth_type = &eth->ether_type;
> > >
> > > printf("\n Type is %" PRIu16 , *((uint16_t *)eth_type));
> >
> > Looks ok, albeit slightly overly complicated.
> >
> > struct ether_hdr *eth;
> > uint16_t eth_type;
> >
> > eth = rte_pktmbuf_mtod(m, struct ether_hdr *);
> > eth_type = eth->ether_type;
> > printf("\n Type is %" PRIu16 , eth_type));
> >
> > This would remove all the extra casting.  Perhaps the mbuf is the problem
> > here.  And if so, the above snippet would segfault on eth->ether_type
> > showing you that mbuf is invalid.
> >
> > You could try to (above mtod):
> >
> > rte_pktmbuf_dump(m, sizeof(struct ether_hdr));
> >
> > as that will show key contents of mbuf and packet data (and do some
> > basic validation along the way).
> >
> > thanks,
> > -chris
> >
>