Re: [2/2,v3] myri10ge: avoid uninitialized variable use

[Arnd Bergmann]

On Thu, Feb 5, 2026, at 05:57, Jakub Kicinski wrote:
>> @@ -2238,6 +2250,8 @@ static int myri10ge_get_txrx(struct myri10ge_priv *mgp, int slice)
>>  	status = 0;
>>  	if (slice == 0 || (mgp->dev->real_num_tx_queues > 1)) {
>>  		cmd.data0 = slice;
>> +		cmd.data1 = 0;
>> +		cmd.data2 = 0;
>>  		status = myri10ge_send_cmd(mgp, MXGEFW_CMD_GET_SEND_OFFSET,
>>  					   &cmd, 0);
>>  		ss->tx.lanai = (struct mcp_kreq_ether_send __iomem *)
>>  			    (mgp->sram + cmd.data0);
>>  	}
>>  	cmd.data0 = slice;
>>  	status |= myri10ge_send_cmd(mgp, MXGEFW_CMD_GET_SMALL_RX_OFFSET,
>>  				    &cmd, 0);
>
> The initialization of cmd.data1 and cmd.data2 is inside the conditional
> block. When slice > 0 and real_num_tx_queues <= 1, does the subsequent
> myri10ge_send_cmd() call for MXGEFW_CMD_GET_SMALL_RX_OFFSET use
> uninitialized data1/data2 values?
>
> Simon Horman pointed this out in his review of v2:
> https://lore.kernel.org/all/20250622162715.GA297140@xxxxxxxxxxxxxxxx/
>
> He provided a diff showing that the second cmd.data0 = slice assignment
> also needs cmd.data1 and cmd.data2 initialization. This feedback does
> not appear to have been addressed in v3.

Indeed, I thought I had addressed this, but somehow ended up
rebasing to an older version of my original patch. Sent v4 now.

     Arnd