On 2023/10/25 18:08, Greg Kroah-Hartman wrote:
> On Mon, Oct 23, 2023 at 08:05:01PM +0800, Guangguan Wang wrote:
>> Hi, Greg.
>>
>> [PATCH 6.5 159/241] net/smc: support smc release version negotiation in clc handshake
>> [PATCH 6.5 160/241] net/smc: support smc v2.x features validate
>>
>> The above two patches should not backport to stable tree 6.5, which may result in unexpected
>> fallback if communication between 6.6 and 6.5(with these two patch) via SMC-R v2.1. The above
>> two patches should not exist individually without the patch 7f0620b9(net/smc: support max
>> connections per lgr negotiation) and the patch 69b888e3(net/smc: support max links per lgr
>> negotiation in clc handshake).
>>
>> The patch c68681ae46ea ("net/smc: fix smc clc failed issue when netdevice not in init_net")
>> does not rely the feature SMC-R v2.1. But I think it may have conflict here when backport
>> to stable tree 6.5:
>>
>> @@ -1201,6 +1201,7 @@ static int smc_connect_rdma_v2_prepare(struct smc_sock *smc,
>> (struct smc_clc_msg_accept_confirm_v2 *)aclc;
>> struct smc_clc_first_contact_ext *fce =
>> smc_get_clc_first_contact_ext(clc_v2, false); --conflict here
>> + struct net *net = sock_net(&smc->sk);
>>
>>
>> I think it is better to resolve the confilict rather than backport more patches.
>> The resolution of the conflict should be like:
>>
>> @@ -1201,6 +1201,7 @@ static int smc_connect_rdma_v2_prepare(struct smc_sock *smc,
>> (struct smc_clc_msg_accept_confirm_v2 *)aclc;
>> struct smc_clc_first_contact_ext *fce =
>> (struct smc_clc_first_contact_ext *)
>> (((u8 *)clc_v2) + sizeof(*clc_v2)); --replace the line smc_get_clc_first_contact_ext(clc_v2, false);
>> + struct net *net = sock_net(&smc->sk);
>
> Thanks for letting me know.
>
> I've dropped this patch entirely from the 6.5.y queue now (and the
> follow-on ones.) Can you send a backported, and tested, set of patches
> to us for inclusion if you want this fixed up in the 6.5.y tree? That
> way we make sure to get this done properly.
>
> thanks,
>
> greg k-h
I think it is more appropriate for Albert Huang, who is the author of the patch("net/smc: fix smc clc failed issue when
netdevice not in init_net"), to do this because he knows the background of the fix and how to test it.
Thanks,
Guangguan Wang
