[PATCH v3] vhost: compilation fix for GCC-12

[Maxime Coquelin]

Hi Amit,

On 10/6/22 09:22, Amit Prakash Shukla wrote:
> Ping.

I posted a patch yesterday that fixes this issue without having to
change the runtime behavior:

http://patches.dpdk.org/project/dpdk/patch/20221005203524.89336-1-maxime.coquelin@redhat.com/

Could you test it and provide ack if OK on your side?

Thanks,
Maxime

>> -----Original Message-----
>> From: Amit Prakash Shukla <amitprakashs at marvell.com>
>> Sent: Friday, September 2, 2022 8:36 PM
>> To: Maxime Coquelin <maxime.coquelin at redhat.com>; Chenbo Xia
>> <chenbo.xia at intel.com>
>> Cc: dev at dpdk.org; Jerin Jacob Kollanukkaran <jerinj at marvell.com>;
>> stable at dpdk.org; Ruifeng.Wang at arm.com; Amit Prakash Shukla
>> <amitprakashs at marvell.com>
>> Subject: [PATCH v3] vhost: compilation fix for GCC-12
>>
>> GCC-12 complains about the possible use of un-initialized array. At compile
>> time it seems like it is not able to evaluate the size as it involves run-time
>> variable and at compile time it seems like gcc assumes value of "size" variable
>> to be zero which makes gcc-12 to jump the while loop.
>> "size = pkt->pkt_len + sizeof(struct virtio_net_hdr_mrg_rxbuf);"
>>
>> As part of the fix, "while (){}" is replaced by "do {} while()" which make the
>> compiler to generate a code in which buf_vec will never be used un-
>> initialized.
>>
>> ../lib/vhost/virtio_net.c:941:35: error:
>> 		'buf_vec[0].buf_len' may be used uninitialized
>> 		[-Werror=maybe-uninitialized]
>>    941 |         buf_len = buf_vec[vec_idx].buf_len;
>>        |                   ~~~~~~~~~~~~~~~~^~~~~~~~
>> ../lib/vhost/virtio_net.c: In function 'virtio_dev_rx_packed':
>> ../lib/vhost/virtio_net.c:1285:27: note: 'buf_vec' declared here
>>   1285 |         struct buf_vector buf_vec[BUF_VECTOR_MAX];
>>        |                           ^~~~~~~
>> cc1: all warnings being treated as errors
>>
>> Fixes: 93520085efda ("vhost: add packed ring single enqueue")
>> Cc: stable at dpdk.org
>>
>> Signed-off-by: Amit Prakash Shukla <amitprakashs at marvell.com>
>> ---
>> v2:
>> - Changes for code review suggestion
>>
>> v3:
>> - Added a description
>>
>>   lib/vhost/virtio_net.c | 8 ++++----
>>   1 file changed, 4 insertions(+), 4 deletions(-)
>>
>> diff --git a/lib/vhost/virtio_net.c b/lib/vhost/virtio_net.c index
>> b3d954aab4..9b77d3d10f 100644
>> --- a/lib/vhost/virtio_net.c
>> +++ b/lib/vhost/virtio_net.c
>> @@ -1069,7 +1069,7 @@ vhost_enqueue_single_packed(struct virtio_net
>> *dev,
>>   	else
>>   		max_tries = 1;
>>
>> -	while (size > 0) {
>> +	do {
>>   		/*
>>   		 * if we tried all available ring items, and still
>>   		 * can't get enough buf, it means something abnormal @@ -
>> 1097,7 +1097,7 @@ vhost_enqueue_single_packed(struct virtio_net *dev,
>>   		avail_idx += desc_count;
>>   		if (avail_idx >= vq->size)
>>   			avail_idx -= vq->size;
>> -	}
>> +	} while (size > 0);
>>
>>   	if (mbuf_to_desc(dev, vq, pkt, buf_vec, nr_vec, num_buffers, false)
>> < 0)
>>   		return -1;
>> @@ -1574,7 +1574,7 @@ vhost_enqueue_async_packed(struct virtio_net
>> *dev,
>>   	else
>>   		max_tries = 1;
>>
>> -	while (size > 0) {
>> +	do {
>>   		/*
>>   		 * if we tried all available ring items, and still
>>   		 * can't get enough buf, it means something abnormal @@ -
>> 1601,7 +1601,7 @@ vhost_enqueue_async_packed(struct virtio_net *dev,
>>   		avail_idx += desc_count;
>>   		if (avail_idx >= vq->size)
>>   			avail_idx -= vq->size;
>> -	}
>> +	} while (size > 0);
>>
>>   	if (unlikely(mbuf_to_desc(dev, vq, pkt, buf_vec, nr_vec,
>> *nr_buffers, true) < 0))
>>   		return -1;
>> --
>> 2.25.1
>