[3/3] sfc: selftest: fix struct packing

  From: Arnd Bergmann <arnd@arndb.de>

Three of the sfc drivers define a packed loopback_payload structure with an
ethernet header followed by an IP header. However, the kernel definition
of iphdr specifies that this is 4-byte aligned, causing a W=1 warning:

net/ethernet/sfc/siena/selftest.c:46:15: error: field ip within 'struct efx_loopback_payload' is less aligned than 'struct iphdr' and is usually due to 'struct efx_loopback_payload' being packed, which can lead to unaligned accesses [-Werror,-Wunaligned-access]
        struct iphdr ip;

As the iphdr packing is not easily changed without breaking other code,
change the three structures to use a local definition instead.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
---
 drivers/net/ethernet/sfc/falcon/selftest.c | 21 ++++++++++++++++++++-
 drivers/net/ethernet/sfc/selftest.c        | 21 ++++++++++++++++++++-
 drivers/net/ethernet/sfc/siena/selftest.c  | 21 ++++++++++++++++++++-
 3 files changed, 60 insertions(+), 3 deletions(-)
  

On 19/06/2023 10:12, Arnd Bergmann wrote:
> From: Arnd Bergmann <arnd@arndb.de>
> 
> Three of the sfc drivers define a packed loopback_payload structure with an
> ethernet header followed by an IP header. However, the kernel definition
> of iphdr specifies that this is 4-byte aligned, causing a W=1 warning:
> 
> net/ethernet/sfc/siena/selftest.c:46:15: error: field ip within 'struct efx_loopback_payload' is less aligned than 'struct iphdr' and is usually due to 'struct efx_loopback_payload' being packed, which can lead to unaligned accesses [-Werror,-Wunaligned-access]
>         struct iphdr ip;
> 
> As the iphdr packing is not easily changed without breaking other code,
> change the three structures to use a local definition instead.
> 
> Signed-off-by: Arnd Bergmann <arnd@arndb.de>

Duplicating the definition isn't the prettiest thing in the world; it'd
 do for a quick fix if needed but I assume W=1 warnings aren't blocking
 anyone, so maybe defer this one for now and I'll follow up soon with a
 rewrite that fixes this more cleanly?  My idea is to drop the __packed
 from the containing struct, make efx_begin_loopback() copy the layers
 separately, and efx_loopback_rx_packet() similarly do something less
 direct than casting the packet data to the struct.

But I don't insist on it; if you want this fix in immediately then I'm
 okay with that too.

> ---
>  drivers/net/ethernet/sfc/falcon/selftest.c | 21 ++++++++++++++++++++-
>  drivers/net/ethernet/sfc/selftest.c        | 21 ++++++++++++++++++++-
>  drivers/net/ethernet/sfc/siena/selftest.c  | 21 ++++++++++++++++++++-
>  3 files changed, 60 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/net/ethernet/sfc/falcon/selftest.c b/drivers/net/ethernet/sfc/falcon/selftest.c
> index 6a454ac6f8763..fb7fcd27a33a5 100644
> --- a/drivers/net/ethernet/sfc/falcon/selftest.c
> +++ b/drivers/net/ethernet/sfc/falcon/selftest.c
> @@ -40,7 +40,26 @@
>   */
>  struct ef4_loopback_payload {
>  	struct ethhdr header;
> -	struct iphdr ip;
> +	struct {
> +#if defined(__LITTLE_ENDIAN_BITFIELD)
> +		__u8	ihl:4,
> +			version:4;
> +#elif defined (__BIG_ENDIAN_BITFIELD)
> +		__u8	version:4,
> +			ihl:4;
> +#else
> +#error	"Please fix <asm/byteorder.h>"
> +#endif
> +		__u8	tos;
> +		__be16	tot_len;
> +		__be16	id;
> +		__be16	frag_off;
> +		__u8	ttl;
> +		__u8	protocol;
> +		__sum16	check;
> +		__be32	saddr;
> +		__be32	daddr;
> +	} __packed ip; /* unaligned struct iphdr */
>  	struct udphdr udp;
>  	__be16 iteration;
>  	char msg[64];
> diff --git a/drivers/net/ethernet/sfc/selftest.c b/drivers/net/ethernet/sfc/selftest.c
> index 3c5227afd4977..440a57953779c 100644
> --- a/drivers/net/ethernet/sfc/selftest.c
> +++ b/drivers/net/ethernet/sfc/selftest.c
> @@ -43,7 +43,26 @@
>   */
>  struct efx_loopback_payload {
>  	struct ethhdr header;
> -	struct iphdr ip;
> +	struct {
> +#if defined(__LITTLE_ENDIAN_BITFIELD)
> +		__u8	ihl:4,
> +			version:4;
> +#elif defined (__BIG_ENDIAN_BITFIELD)
> +		__u8	version:4,
> +			ihl:4;
> +#else
> +#error	"Please fix <asm/byteorder.h>"
> +#endif
> +		__u8	tos;
> +		__be16	tot_len;
> +		__be16	id;
> +		__be16	frag_off;
> +		__u8	ttl;
> +		__u8	protocol;
> +		__sum16	check;
> +		__be32	saddr;
> +		__be32	daddr;
> +	} __packed ip; /* unaligned struct iphdr */
>  	struct udphdr udp;
>  	__be16 iteration;
>  	char msg[64];
> diff --git a/drivers/net/ethernet/sfc/siena/selftest.c b/drivers/net/ethernet/sfc/siena/selftest.c
> index 07715a3d6beab..b8a8b0495f661 100644
> --- a/drivers/net/ethernet/sfc/siena/selftest.c
> +++ b/drivers/net/ethernet/sfc/siena/selftest.c
> @@ -43,7 +43,26 @@
>   */
>  struct efx_loopback_payload {
>  	struct ethhdr header;
> -	struct iphdr ip;
> +	struct {
> +#if defined(__LITTLE_ENDIAN_BITFIELD)
> +		__u8	ihl:4,
> +			version:4;
> +#elif defined (__BIG_ENDIAN_BITFIELD)
> +		__u8	version:4,
> +			ihl:4;
> +#else
> +#error	"Please fix <asm/byteorder.h>"
> +#endif
> +		__u8	tos;
> +		__be16	tot_len;
> +		__be16	id;
> +		__be16	frag_off;
> +		__u8	ttl;
> +		__u8	protocol;
> +		__sum16	check;
> +		__be32	saddr;
> +		__be32	daddr;
> +	} __packed ip; /* unaligned struct iphdr */
>  	struct udphdr udp;
>  	__be16 iteration;
>  	char msg[64];
>
  

On Mon, Jun 19, 2023, at 12:25, Edward Cree wrote:
> On 19/06/2023 10:12, Arnd Bergmann wrote:
>> From: Arnd Bergmann <arnd@arndb.de>
>> 
>> Three of the sfc drivers define a packed loopback_payload structure with an
>> ethernet header followed by an IP header. However, the kernel definition
>> of iphdr specifies that this is 4-byte aligned, causing a W=1 warning:
>> 
>> net/ethernet/sfc/siena/selftest.c:46:15: error: field ip within 'struct efx_loopback_payload' is less aligned than 'struct iphdr' and is usually due to 'struct efx_loopback_payload' being packed, which can lead to unaligned accesses [-Werror,-Wunaligned-access]
>>         struct iphdr ip;
>> 
>> As the iphdr packing is not easily changed without breaking other code,
>> change the three structures to use a local definition instead.
>> 
>> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
>
> Duplicating the definition isn't the prettiest thing in the world; it'd
>  do for a quick fix if needed but I assume W=1 warnings aren't blocking
>  anyone, so maybe defer this one for now and I'll follow up soon with a
>  rewrite that fixes this more cleanly?  My idea is to drop the __packed
>  from the containing struct, make efx_begin_loopback() copy the layers
>  separately, and efx_loopback_rx_packet() similarly do something less
>  direct than casting the packet data to the struct.
>
> But I don't insist on it; if you want this fix in immediately then I'm
>  okay with that too.
>
>> ---
>>  drivers/net/ethernet/sfc/falcon/selftest.c | 21 ++++++++++++++++++++-
>>  drivers/net/ethernet/sfc/selftest.c        | 21 ++++++++++++++++++++-
>>  drivers/net/ethernet/sfc/siena/selftest.c  | 21 ++++++++++++++++++++-
>>  3 files changed, 60 insertions(+), 3 deletions(-)
>> 
>> diff --git a/drivers/net/ethernet/sfc/falcon/selftest.c b/drivers/net/ethernet/sfc/falcon/selftest.c
>> index 6a454ac6f8763..fb7fcd27a33a5 100644
>> --- a/drivers/net/ethernet/sfc/falcon/selftest.c
>> +++ b/drivers/net/ethernet/sfc/falcon/selftest.c
>> @@ -40,7 +40,26 @@
>>   */
>>  struct ef4_loopback_payload {
>>  	struct ethhdr header;
>> -	struct iphdr ip;
>> +	struct {
>> +#if defined(__LITTLE_ENDIAN_BITFIELD)
>> +		__u8	ihl:4,
>> +			version:4;
>> +#elif defined (__BIG_ENDIAN_BITFIELD)
>> +		__u8	version:4,
>> +			ihl:4;
>> +#else
>> +#error	"Please fix <asm/byteorder.h>"
>> +#endif
>> +		__u8	tos;
>> +		__be16	tot_len;
>> +		__be16	id;
>> +		__be16	frag_off;
>> +		__u8	ttl;
>> +		__u8	protocol;
>> +		__sum16	check;
>> +		__be32	saddr;
>> +		__be32	daddr;
>> +	} __packed ip; /* unaligned struct iphdr */
>>  	struct udphdr udp;
>>  	__be16 iteration;
>>  	char msg[64];
>> diff --git a/drivers/net/ethernet/sfc/selftest.c b/drivers/net/ethernet/sfc/selftest.c
>> index 3c5227afd4977..440a57953779c 100644
>> --- a/drivers/net/ethernet/sfc/selftest.c
>> +++ b/drivers/net/ethernet/sfc/selftest.c
>> @@ -43,7 +43,26 @@
>>   */
>>  struct efx_loopback_payload {
>>  	struct ethhdr header;
>> -	struct iphdr ip;
>> +	struct {
>> +#if defined(__LITTLE_ENDIAN_BITFIELD)
>> +		__u8	ihl:4,
>> +			version:4;
>> +#elif defined (__BIG_ENDIAN_BITFIELD)
>> +		__u8	version:4,
>> +			ihl:4;
>> +#else
>> +#error	"Please fix <asm/byteorder.h>"
>> +#endif
>> +		__u8	tos;
>> +		__be16	tot_len;
>> +		__be16	id;
>> +		__be16	frag_off;
>> +		__u8	ttl;
>> +		__u8	protocol;
>> +		__sum16	check;
>> +		__be32	saddr;
>> +		__be32	daddr;
>> +	} __packed ip; /* unaligned struct iphdr */
>>  	struct udphdr udp;
>>  	__be16 iteration;
>>  	char msg[64];
>> diff --git a/drivers/net/ethernet/sfc/siena/selftest.c b/drivers/net/ethernet/sfc/siena/selftest.c
>> index 07715a3d6beab..b8a8b0495f661 100644
>> --- a/drivers/net/ethernet/sfc/siena/selftest.c
>> +++ b/drivers/net/ethernet/sfc/siena/selftest.c
>> @@ -43,7 +43,26 @@
>>   */
>>  struct efx_loopback_payload {
>>  	struct ethhdr header;
>> -	struct iphdr ip;
>> +	struct {
>> +#if defined(__LITTLE_ENDIAN_BITFIELD)
>> +		__u8	ihl:4,
>> +			version:4;
>> +#elif defined (__BIG_ENDIAN_BITFIELD)
>> +		__u8	version:4,
>> +			ihl:4;
>> +#else
>> +#error	"Please fix <asm/byteorder.h>"
>> +#endif
>> +		__u8	tos;
>> +		__be16	tot_len;
>> +		__be16	id;
>> +		__be16	frag_off;
>> +		__u8	ttl;
>> +		__u8	protocol;
>> +		__sum16	check;
>> +		__be32	saddr;
>> +		__be32	daddr;
>> +	} __packed ip; /* unaligned struct iphdr */
>>  	struct udphdr udp;
>>  	__be16 iteration;
>>  	char msg[64];
>>
  

On Mon, Jun 19, 2023, at 12:25, Edward Cree wrote:
> On 19/06/2023 10:12, Arnd Bergmann wrote:
>> From: Arnd Bergmann <arnd@arndb.de>
>> 
>> Three of the sfc drivers define a packed loopback_payload structure with an
>> ethernet header followed by an IP header. However, the kernel definition
>> of iphdr specifies that this is 4-byte aligned, causing a W=1 warning:
>> 
>> net/ethernet/sfc/siena/selftest.c:46:15: error: field ip within 'struct efx_loopback_payload' is less aligned than 'struct iphdr' and is usually due to 'struct efx_loopback_payload' being packed, which can lead to unaligned accesses [-Werror,-Wunaligned-access]
>>         struct iphdr ip;
>> 
>> As the iphdr packing is not easily changed without breaking other code,
>> change the three structures to use a local definition instead.
>> 
>> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
>
> Duplicating the definition isn't the prettiest thing in the world; it'd
>  do for a quick fix if needed but I assume W=1 warnings aren't blocking
>  anyone, so maybe defer this one for now and I'll follow up soon with a
>  rewrite that fixes this more cleanly?  My idea is to drop the __packed
>  from the containing struct, make efx_begin_loopback() copy the layers
>  separately, and efx_loopback_rx_packet() similarly do something less
>  direct than casting the packet data to the struct.
>
> But I don't insist on it; if you want this fix in immediately then I'm
>  okay with that too.

It's not urgent, if you can come up with a nicer solution, that is
probably better than applying my patch now. I have a patch [1] that
addresses -Wunaligned-access for all x86 and arm randconfig builds,
and this currently affects 23 drivers. Most of the changes don't
have changelog texts yet, and some need a more detailed analysis to
come up with a correct patch. I'd probably aim for linux-6.6 or
later to get them all done, at which point we could move the warning
from W=1 to the default set.

     Arnd

[1] https://pastebin.com/g6D9NTS4
  

...
> Duplicating the definition isn't the prettiest thing in the world; it'd
>  do for a quick fix if needed but I assume W=1 warnings aren't blocking
>  anyone, so maybe defer this one for now and I'll follow up soon with a
>  rewrite that fixes this more cleanly?  My idea is to drop the __packed
>  from the containing struct, make efx_begin_loopback() copy the layers
>  separately, and efx_loopback_rx_packet() similarly do something less
>  direct than casting the packet data to the struct.

Maybe you can get away with adding a 16bit pad before the ethernet
header so that the IP header is actually aligned.

(Then fight all the stuff that stops you doing a memcpy()
that runs into a second field of a structure.)

Failing that maybe a single shared copy of the misaligned
IP header.

I also suspect you could just add __packed to the two 32bit
address fields.
That would generate better code on systems that care.

	David

-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
Registration No: 1397386 (Wales)
  

On 23/06/2023 11:52, David Laight wrote:
> Maybe you can get away with adding a 16bit pad before the ethernet
> header so that the IP header is actually aligned.

That's what I ended up doing, because my original idea was
 overcomplicated and turned out super ugly.
See https://lore.kernel.org/netdev/6f87fdf5-1844-4633-b4fe-6b247bc6ab49@app.fastmail.com/T/

> (Then fight all the stuff that stops you doing a memcpy()
> that runs into a second field of a structure.)

Yeah, I don't know how you're meant to annotate that stuff.
I guess I'll have to wait until Kees shouts at me and tells
 me what to do :S

-ed