Re: [PATCH] hw/net: ftgmac100: copy eth_hdr for alignment

[Philippe Mathieu-Daudé]

Hi Patrick,

On 27/2/25 16:42, Patrick Venture wrote:
> eth_hdr requires 2 byte alignment
>
> Signed-off-by: Patrick Venture <venture@google.com>
> ---
>   hw/net/ftgmac100.c | 15 ++++++++++++---
>   1 file changed, 12 insertions(+), 3 deletions(-)


> @@ -1028,6 +1032,7 @@ static ssize_t ftgmac100_receive(NetClientState *nc, 
> const uint8_t *buf,
>   {
>       FTGMAC100State *s = FTGMAC100(qemu_get_nic_opaque(nc));
>       FTGMAC100Desc bd;
> +    struct eth_header eth_hdr = {};
>       uint32_t flags = 0;
>       uint64_t addr;
>       uint32_t crc;
> @@ -1036,7 +1041,11 @@ static ssize_t ftgmac100_receive(NetClientState *nc, 
> const uint8_t *buf,
>       uint32_t buf_len;
>       size_t size = len;
>       uint32_t first = FTGMAC100_RXDES0_FRS;
> -    uint16_t proto = be16_to_cpu(PKT_GET_ETH_HDR(buf)->h_proto);

The LD/ST API deals with unaligned fields, would that help?

    uint16_t proto = lduw_be_p(&PKT_GET_ETH_HDR(buf)->h_proto);

> +    uint16_t proto;
> +
> +    memcpy(&eth_hdr, PKT_GET_ETH_HDR(buf),
> +           (sizeof(eth_hdr) > len) ? len : sizeof(eth_hdr));
> +    proto = be16_to_cpu(eth_hdr.h_proto);
>       int max_frame_size = ftgmac100_max_frame_size(s, proto);