variant length array?

variant length array?

[Wenda Ni]

Hi all,

I come across the following code in a kernel module code. It defines an
array whose length is variant at runtime, depending on the actual inputs.
It seems that kernel compiler supports this, which is obvious an error in
the standard ANSI C. Do I have the correct understanding on it?

Thank you.


u32 rxe_icrc_hdr(struct rxe_pkt_info *pkt, struct sk_buff *skb)
{
         ...
         int hdr_size = sizeof(struct udphdr) +
                 (skb->protocol == htons(ETH_P_IP) ?
                 sizeof(struct iphdr) : sizeof(struct ipv6hdr));
         u8 tmp[hdr_size + RXE_BTH_BYTES];
         ...
}

Cheers,

Wenda Ni, Ph.D.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.kernelnewbies.org/pipermail/kernelnewbies/attachments/20160405/faf2fc9e/attachment.html 

variant length array?

[Robert P. J. Day]

On Tue, 5 Apr 2016, Wenda Ni wrote:

> Hi all,
>
> I come across the following code in a kernel module code. It defines
> an array whose length is variant at runtime, depending on the actual
> inputs. It seems that kernel compiler supports this, which is
> obvious an error in the standard ANSI C. Do I have the correct
> understanding on it?
>
> Thank you.
>
>
> u32 rxe_icrc_hdr(struct rxe_pkt_info *pkt, struct sk_buff *skb)
> {
> ???????? ...
> ???????? int hdr_size = sizeof(struct udphdr) +
> ???????????????? (skb->protocol == htons(ETH_P_IP) ?
> ???????????????? sizeof(struct iphdr) : sizeof(struct ipv6hdr));
> ???????? u8 tmp[hdr_size + RXE_BTH_BYTES];
> ???????? ...
> }

  pretty sure "sizeof" can be calculated at compile time so i don't
see a problem here.

rday


-- 

========================================================================
Robert P. J. Day                                 Ottawa, Ontario, CANADA
                        http://crashcourse.ca

Twitter:                                       http://twitter.com/rpjday
LinkedIn:                               http://ca.linkedin.com/in/rpjday
========================================================================

variant length array?

[Bjørn Mork]

"Robert P. J. Day" <rpjday@crashcourse.ca> writes:
> On Tue, 5 Apr 2016, Wenda Ni wrote:
>
>> Hi all,
>>
>> I come across the following code in a kernel module code. It defines
>> an array whose length is variant at runtime, depending on the actual
>> inputs. It seems that kernel compiler supports this, which is
>> obvious an error in the standard ANSI C. Do I have the correct
>> understanding on it?
>>
>> Thank you.
>>
>>
>> u32 rxe_icrc_hdr(struct rxe_pkt_info *pkt, struct sk_buff *skb)
>> {
>> ???????? ...
>> ???????? int hdr_size = sizeof(struct udphdr) +
>> ???????????????? (skb->protocol == htons(ETH_P_IP) ?
>> ???????????????? sizeof(struct iphdr) : sizeof(struct ipv6hdr));
>> ???????? u8 tmp[hdr_size + RXE_BTH_BYTES];
>> ???????? ...
>> }
>
>   pretty sure "sizeof" can be calculated at compile time so i don't
> see a problem here.

Yes, but skb->protocol is variable and sizeof(struct iphdr) !=
sizeof(struct ipv6hdr)).  Is the compiler smart enough to just use the
largest possible value?  The logic here is pretty similar to a union,
which it of course wouldn't have any problems calculating the size of.


Bj?rn

variant length array?

[Rajat Sharma]

On Tue, Apr 5, 2016 at 1:00 PM, Bj?rn Mork <bjorn@mork.no> wrote:
>
> "Robert P. J. Day" <rpjday@crashcourse.ca> writes:
> > On Tue, 5 Apr 2016, Wenda Ni wrote:
> >
> >> Hi all,
> >>
> >> I come across the following code in a kernel module code. It defines
> >> an array whose length is variant at runtime, depending on the actual
> >> inputs. It seems that kernel compiler supports this, which is
> >> obvious an error in the standard ANSI C. Do I have the correct
> >> understanding on it?
> >>
> >> Thank you.
> >>
> >>
> >> u32 rxe_icrc_hdr(struct rxe_pkt_info *pkt, struct sk_buff *skb)
> >> {
> >>          ...
> >>          int hdr_size = sizeof(struct udphdr) +
> >>                  (skb->protocol == htons(ETH_P_IP) ?
> >>                  sizeof(struct iphdr) : sizeof(struct ipv6hdr));
> >>          u8 tmp[hdr_size + RXE_BTH_BYTES];
> >>          ...
> >> }
> >
> >   pretty sure "sizeof" can be calculated at compile time so i don't
> > see a problem here.
>
> Yes, but skb->protocol is variable and sizeof(struct iphdr) !=
> sizeof(struct ipv6hdr)).  Is the compiler smart enough to just use the
> largest possible value?  The logic here is pretty similar to a union,
> which it of course wouldn't have any problems calculating the size of.
>
>
> Bj?rn
>
> _______________________________________________
> Kernelnewbies mailing list
> Kernelnewbies at kernelnewbies.org
> http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies


It is actually a variable length array on stack. BTW nothing fancy
about kernel compiler, its GCC extention.
https://gcc.gnu.org/onlinedocs/gcc/Variable-Length.html

-Rajat

variant length array?

[Robert P. J. Day]

On Tue, 5 Apr 2016, Bj?rn Mork wrote:

> "Robert P. J. Day" <rpjday@crashcourse.ca> writes:
> > On Tue, 5 Apr 2016, Wenda Ni wrote:
> >
> >> Hi all,
> >>
> >> I come across the following code in a kernel module code. It defines
> >> an array whose length is variant at runtime, depending on the actual
> >> inputs. It seems that kernel compiler supports this, which is
> >> obvious an error in the standard ANSI C. Do I have the correct
> >> understanding on it?
> >>
> >> Thank you.
> >>
> >>
> >> u32 rxe_icrc_hdr(struct rxe_pkt_info *pkt, struct sk_buff *skb)
> >> {
> >> ???????? ...
> >> ???????? int hdr_size = sizeof(struct udphdr) +
> >> ???????????????? (skb->protocol == htons(ETH_P_IP) ?
> >> ???????????????? sizeof(struct iphdr) : sizeof(struct ipv6hdr));
> >> ???????? u8 tmp[hdr_size + RXE_BTH_BYTES];
> >> ???????? ...
> >> }
> >
> >   pretty sure "sizeof" can be calculated at compile time so i don't
> > see a problem here.
>
> Yes, but skb->protocol is variable and sizeof(struct iphdr) !=
> sizeof(struct ipv6hdr)).  Is the compiler smart enough to just use the
> largest possible value?  The logic here is pretty similar to a union,
> which it of course wouldn't have any problems calculating the size of.

  ah, quite so, i didn't look closely enough.

rday

-- 

========================================================================
Robert P. J. Day                                 Ottawa, Ontario, CANADA
                        http://crashcourse.ca

Twitter:                                       http://twitter.com/rpjday
LinkedIn:                               http://ca.linkedin.com/in/rpjday
========================================================================

variant length array?

[Valdis.Kletnieks at vt.edu]

On Tue, 05 Apr 2016 15:29:25 -0400, Wenda Ni said:

> I come across the following code in a kernel module code.

> u32 rxe_icrc_hdr(struct rxe_pkt_info *pkt, struct sk_buff *skb)

Is this an out-of-tree module, or an older kernel?

The current linux-next tree has zero occurences of 'rxe_icrc_hdr'
or 'rxe_pkt_info' in it.


-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 848 bytes
Desc: not available
Url : http://lists.kernelnewbies.org/pipermail/kernelnewbies/attachments/20160405/26234844/attachment.bin 

variant length array?

[Wenda Ni]

On Tue, Apr 5, 2016 at 4:19 PM, <Valdis.Kletnieks@vt.edu> wrote:

> On Tue, 05 Apr 2016 15:29:25 -0400, Wenda Ni said:
>
> > I come across the following code in a kernel module code.
>
> > u32 rxe_icrc_hdr(struct rxe_pkt_info *pkt, struct sk_buff *skb)
>
> Is this an out-of-tree module, or an older kernel?
>
> The current linux-next tree has zero occurences of 'rxe_icrc_hdr'
> or 'rxe_pkt_info' in it.
>
>
> You are right. It is an out-of-tree module, called softRoCE,
https://github.com/SoftRoCE/rxe-dev

uname -a gives kernel version to be 4.0.0+, I do not know exactly which
kernel base they use for development.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.kernelnewbies.org/pipermail/kernelnewbies/attachments/20160405/95d80135/attachment.html 

variant length array?

[Bernd Petrovitsch]

Hi!

On Die, 2016-04-05 at 15:29 -0400, Wenda Ni wrote:
[...]
> I come across the following code in a kernel module code. It defines an
> array whose length is variant at runtime, depending on the actual inputs.
> It seems that kernel compiler supports this, which is obvious an error in
> the standard ANSI C. Do I have the correct understanding on it?

ANSI-C ist old and enforced with -std=c89 with gcc (see manual page).
VLAs are allowed since (standard) C99.

	Bernd
-- 
"What happens when you read some doc and either it doesn't answer your
question or is demonstrably wrong? In Linux, you say "Linux sucks" and
go read the code. In Windows/Oracle/etc you say "Windows sucks" and
start banging your head against the wall."    - Denis Vlasenko on lkml