Re: [PATCH 00/19] Wangxun fixes and supports

[Ferruh Yigit <ferruh.yigit@amd.com>]

On 7/12/2024 1:47 PM, Ferruh Yigit wrote:
> On 7/12/2024 10:32 AM, David Marchand wrote:
>> On Thu, Jul 11, 2024 at 1:53 PM David Marchand
>> <david.marchand@redhat.com> wrote:
>>> On Sat, Jul 6, 2024 at 5:39 AM Ferruh Yigit <ferruh.yigit@amd.com> wrote:
>>>>>>> Fix some bugs in txgbe/ngbe, and support new features in ngbe driver.
>>>>>>>
>>>>>>> Jiawen Wu (19):
>>>>>>>   net/txgbe: fix to parse tunnel packets
>>>>>>>   net/txgbe: fix flow filters in VT mode
>>>>>>>   net/txgbe: fix Tx hang on queue disable
>>>>>>>   net/txgbe: restrict the configuration of VLAN strip offload
>>>>>>>   net/txgbe: reconfigure more MAC Rx registers
>>>>>>>   net/txgbe: fix VF promiscuous and allmulticast
>>>>>>>   net/ngbe: special config for YT8531SH-CA PHY
>>>>>>>   net/ngbe: keep PHY power down while device probing
>>>>>>>   net/ngbe: add WOL and NCSI capability
>>>>>>>   net/txgbe: fix hotplug remove
>>>>>>>   net/ngbe: fix hotplug remove
>>>>>>>   net/txgbe: correct valid MTU range
>>>>>>>   net/ngbe: correct valid MTU range
>>>>>>>   net/txgbe: fix memory leak
>>>>>>>   net/ngbe: fix memory leak
>>>>>>>   net/txgbe: fix Rx interrupt
>>>>>>>   net/ngbe: support Rx interrupt
>>>>>>>   net/txgbe: disable LLDP by default
>>>>>>>   net/ngbe: disable LLDP by default
>>>>>>
>>>>>
>>>>
>>>> Series applied to dpdk-next-net/main, thanks.
>>>
>>> Compilation is broken for some ARM targets with next-net.
>>> It was detected by the CI:
>>> http://mails.dpdk.org/archives/test-report/2024-June/700549.html
>>>
>>> Jiawen, please send a fix.
>>
>> Ferruh, Thomas,
>>
>> We have many build failure reports because of this series.
>> What is the best approach while waiting for a fix?
>> Can we drop patch 35c4da655a95 ("net/ngbe: disable LLDP by default")
>> (which seems to be the cause) from next-net?
>>
> 
> Hi David,
> 
> I can check if the error is trivial, and if so send a patch for it.
> 

Error is:
In function 'ngbe_host_interface_command',
inlined from 'ngbe_host_interface_command' at
../drivers/net/ngbe/base/ngbe_mng.c:86:1,
inlined from 'ngbe_hic_get_lldp' at
../drivers/net/ngbe/base/ngbe_mng.c:394:8:
../drivers/net/ngbe/base/ngbe_mng.c:127:36: error: array subscript 2 is
outside array bounds of 'struct ngbe_hic_write_lldp[1]'
[-Werror=array-bounds]
127 |                         buffer[bi] = rd32a(hw, NGBE_MNGMBX, bi);
../drivers/net/ngbe/base/ngbe_mng.c: In function 'ngbe_hic_get_lldp':
../drivers/net/ngbe/base/ngbe_mng.c:385:36: note: at offset 8 into
object 'buffer' of size 8
385 |         struct ngbe_hic_write_lldp buffer;
|                                    ^~~~~~


Problematic case only happens when 'resp->cmd == 0x30', but for
'get_lldp', cmd is 'FW_LLDP_GET_CMD' which is 0xF5, so practically this
is not a real issue.

To build the build log, we have options:
1. Extend "struct ngbe_hic_read_pcie" with 'u32 pad4'
2. Remove 'resp->cmd == 0x30' exception in
'ngbe_host_interface_command()' as that command seems not supported at all.

But I can't be sure on possible side impacts of above changes.

Another option is dropping this patch from the next-net and a fixed
version can be merged back.