----- Original Message -----
> Some drivers probe PHY addresses beyond the first one, and
> also (unfortunately) don't check for MDIC_ERROR on reads. It appears
> that the driver assumes that the data reads will return zero in this
> case (invalid PHY address). Anecdotally, hardware is believed to do
> this; but qemu wasn't, and instead was reusing the previous value of the
> MDIC register. This screwed up driver init.
>
> To fix, don't reuse the current value of the MDIC register when an
> invalid PHY address is referenced; let the value be set to precisely
> MDIC_ERROR|MDIC_READY.
>
> Signed-off-by: Richard Tollerton <rich.toller...@ni.com>
> Signed-off-by: Jeff Westfahl <jeff.westf...@ni.com>
> ---
Looks correct.
Reviewed-by: Jason Wang <jasow...@redhat.com>
> hw/net/e1000.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/hw/net/e1000.c b/hw/net/e1000.c
> index e33a4da..44ae3a8 100644
> --- a/hw/net/e1000.c
> +++ b/hw/net/e1000.c
> @@ -457,7 +457,7 @@ set_mdic(E1000State *s, int index, uint32_t val)
> uint32_t addr = ((val & E1000_MDIC_REG_MASK) >> E1000_MDIC_REG_SHIFT);
>
> if ((val & E1000_MDIC_PHY_MASK) >> E1000_MDIC_PHY_SHIFT != 1) // phy #
> - val = s->mac_reg[MDIC] | E1000_MDIC_ERROR;
> + val = E1000_MDIC_ERROR;
> else if (val & E1000_MDIC_OP_READ) {
> DBGOUT(MDIC, "MDIC read reg 0x%x\n", addr);
> if (!(phy_regcap[addr] & PHY_R)) {
> --
> 2.1.3
>
>
>
