net/enic: fix build with GCC 7.5
Checks
Commit Message
Build error:
../drivers/net/enic/enic_fm_flow.c: In function 'enic_fm_flow_parse':
../drivers/net/enic/enic_fm_flow.c:1467:24:
error: 'dev' may be used uninitialized in this function
[-Werror=maybe-uninitialized]
struct rte_eth_dev *dev;
^~~
../drivers/net/enic/enic_fm_flow.c:1580:24:
error: 'dev' may be used uninitialized in this function
[-Werror=maybe-uninitialized]
struct rte_eth_dev *dev;
^~~
../drivers/net/enic/enic_fm_flow.c:1599:24:
error: 'dev' may be used uninitialized in this function
[-Werror=maybe-uninitialized]
struct rte_eth_dev *dev;
^~~
Build error looks like false positive, but to silence the compiler
initializing the pointer with NULL.
Fixes: 7968917ccf64 ("net/enic: support meta flow actions to overrule destinations")
Reported-by: David Marchand <david.marchand@redhat.com>
Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
---
Cc: andrew.rybchenko@oktetlabs.ru
I am not sure about the solution and I don't have environment to verify,
sending this patch to verify the solution in CI and trigger discussion
for fix.
The patch is still in next-net, when a proper fix is found, it can be
squashed in next-net.
---
drivers/net/enic/enic_fm_flow.c | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
Comments
On Fri, Oct 15, 2021 at 12:28 PM Ferruh Yigit <ferruh.yigit@intel.com> wrote:
>
> Build error:
> ../drivers/net/enic/enic_fm_flow.c: In function 'enic_fm_flow_parse':
> ../drivers/net/enic/enic_fm_flow.c:1467:24:
> error: 'dev' may be used uninitialized in this function
> [-Werror=maybe-uninitialized]
> struct rte_eth_dev *dev;
> ^~~
> ../drivers/net/enic/enic_fm_flow.c:1580:24:
> error: 'dev' may be used uninitialized in this function
> [-Werror=maybe-uninitialized]
> struct rte_eth_dev *dev;
> ^~~
> ../drivers/net/enic/enic_fm_flow.c:1599:24:
> error: 'dev' may be used uninitialized in this function
> [-Werror=maybe-uninitialized]
> struct rte_eth_dev *dev;
> ^~~
>
> Build error looks like false positive, but to silence the compiler
> initializing the pointer with NULL.
enic_fm_check_transfer_dst() contains branches where dev is not set
and those branches return rte_flow_error_set return value.
dev is dereferenced later based on this return value == 0.
So the compiler probably thinks that rte_flow_error_set may return 0.
rte_flow_error_set is outside of compiler "view" at the moment it
compiles enic_fm_flow.c, so the compiler making the assumption this
function can return 0 is being prudent from my pov.
>
> Fixes: 7968917ccf64 ("net/enic: support meta flow actions to overrule destinations")
>
> Reported-by: David Marchand <david.marchand@redhat.com>
> Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
Your fix looks good in any case.
Reviewed-by: David Marchand <david.marchand@redhat.com>
On 10/15/2021 12:16 PM, David Marchand wrote:
> On Fri, Oct 15, 2021 at 12:28 PM Ferruh Yigit <ferruh.yigit@intel.com> wrote:
>>
>> Build error:
>> ../drivers/net/enic/enic_fm_flow.c: In function 'enic_fm_flow_parse':
>> ../drivers/net/enic/enic_fm_flow.c:1467:24:
>> error: 'dev' may be used uninitialized in this function
>> [-Werror=maybe-uninitialized]
>> struct rte_eth_dev *dev;
>> ^~~
>> ../drivers/net/enic/enic_fm_flow.c:1580:24:
>> error: 'dev' may be used uninitialized in this function
>> [-Werror=maybe-uninitialized]
>> struct rte_eth_dev *dev;
>> ^~~
>> ../drivers/net/enic/enic_fm_flow.c:1599:24:
>> error: 'dev' may be used uninitialized in this function
>> [-Werror=maybe-uninitialized]
>> struct rte_eth_dev *dev;
>> ^~~
>>
>> Build error looks like false positive, but to silence the compiler
>> initializing the pointer with NULL.
>
> enic_fm_check_transfer_dst() contains branches where dev is not set
> and those branches return rte_flow_error_set return value.
> dev is dereferenced later based on this return value == 0.
>
> So the compiler probably thinks that rte_flow_error_set may return 0.
> rte_flow_error_set is outside of compiler "view" at the moment it
> compiles enic_fm_flow.c, so the compiler making the assumption this
> function can return 0 is being prudent from my pov.
>
>
>>
>> Fixes: 7968917ccf64 ("net/enic: support meta flow actions to overrule destinations")
>>
>> Reported-by: David Marchand <david.marchand@redhat.com>
>> Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
>
> Your fix looks good in any case.
>
> Reviewed-by: David Marchand <david.marchand@redhat.com>
>
Applied to dpdk-next-net/main, thanks.
On 10/18/2021 12:16 PM, Ferruh Yigit wrote:
> On 10/15/2021 12:16 PM, David Marchand wrote:
>> On Fri, Oct 15, 2021 at 12:28 PM Ferruh Yigit <ferruh.yigit@intel.com> wrote:
>>>
>>> Build error:
>>> ../drivers/net/enic/enic_fm_flow.c: In function 'enic_fm_flow_parse':
>>> ../drivers/net/enic/enic_fm_flow.c:1467:24:
>>> error: 'dev' may be used uninitialized in this function
>>> [-Werror=maybe-uninitialized]
>>> struct rte_eth_dev *dev;
>>> ^~~
>>> ../drivers/net/enic/enic_fm_flow.c:1580:24:
>>> error: 'dev' may be used uninitialized in this function
>>> [-Werror=maybe-uninitialized]
>>> struct rte_eth_dev *dev;
>>> ^~~
>>> ../drivers/net/enic/enic_fm_flow.c:1599:24:
>>> error: 'dev' may be used uninitialized in this function
>>> [-Werror=maybe-uninitialized]
>>> struct rte_eth_dev *dev;
>>> ^~~
>>>
>>> Build error looks like false positive, but to silence the compiler
>>> initializing the pointer with NULL.
>>
>> enic_fm_check_transfer_dst() contains branches where dev is not set
>> and those branches return rte_flow_error_set return value.
>> dev is dereferenced later based on this return value == 0.
>>
>> So the compiler probably thinks that rte_flow_error_set may return 0.
>> rte_flow_error_set is outside of compiler "view" at the moment it
>> compiles enic_fm_flow.c, so the compiler making the assumption this
>> function can return 0 is being prudent from my pov.
>>
>>
>>>
>>> Fixes: 7968917ccf64 ("net/enic: support meta flow actions to overrule destinations")
>>>
>>> Reported-by: David Marchand <david.marchand@redhat.com>
>>> Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
>>
>> Your fix looks good in any case.
>>
>> Reviewed-by: David Marchand <david.marchand@redhat.com>
>>
>
> Applied to dpdk-next-net/main, thanks.
>
Bugzilla ID: 812
Updated in next-net.
@@ -1464,7 +1464,7 @@ enic_fm_copy_action(struct enic_flowman *fm,
}
case RTE_FLOW_ACTION_TYPE_PORT_ID: {
const struct rte_flow_action_port_id *port;
- struct rte_eth_dev *dev;
+ struct rte_eth_dev *dev = NULL;
if (!ingress && (overlap & PORT_ID)) {
ENICPMD_LOG(DEBUG, "cannot have multiple egress PORT_ID actions");
@@ -1577,7 +1577,7 @@ enic_fm_copy_action(struct enic_flowman *fm,
}
case RTE_FLOW_ACTION_TYPE_PORT_REPRESENTOR: {
const struct rte_flow_action_ethdev *ethdev;
- struct rte_eth_dev *dev;
+ struct rte_eth_dev *dev = NULL;
ethdev = actions->conf;
ret = enic_fm_check_transfer_dst(enic, ethdev->port_id,
@@ -1596,7 +1596,7 @@ enic_fm_copy_action(struct enic_flowman *fm,
}
case RTE_FLOW_ACTION_TYPE_REPRESENTED_PORT: {
const struct rte_flow_action_ethdev *ethdev;
- struct rte_eth_dev *dev;
+ struct rte_eth_dev *dev = NULL;
if (overlap & PORT_ID) {
ENICPMD_LOG(DEBUG, "cannot have multiple egress PORT_ID actions");