lkml.org 
[lkml]   [2022]   [May]   [23]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
SubjectRe: [PATCH net] net/smc: fix listen processing for SMC-Rv2
From
On 23/05/2022 14:12, liuyacan@corp.netease.com wrote:
>>> From: liuyacan <liuyacan@corp.netease.com>
>>>
>>> In the process of checking whether RDMAv2 is available, the current
>>> implementation first sets ini->smcrv2.ib_dev_v2, and then allocates
>>> smc buf desc, but the latter may fail. Unfortunately, the caller
>>> will only check the former. In this case, a NULL pointer reference
>>> will occur in smc_clc_send_confirm_accept() when accessing
>>> conn->rmb_desc.
>>>
>>> This patch does two things:
>>> 1. Use the return code to determine whether V2 is available.
>>> 2. If the return code is NODEV, continue to check whether V1 is
>>> available.
>>>
>>> Fixes: e49300a6bf62 ("net/smc: add listen processing for SMC-Rv2")
>>> Signed-off-by: liuyacan <liuyacan@corp.netease.com>
>>> ---
>>
>> I am not happy with this patch. You are right that this is a problem,
>> but the fix should be much simpler: set ini->smcrv2.ib_dev_v2 = NULL in
>> smc_find_rdma_v2_device_serv() after the not_found label, just like it is
>> done in a similar way for the ISM device in smc_find_ism_v1_device_serv().
>>
>> Your patch changes many more things, and beside that you eliminated the calls
>> to smc_find_ism_store_rc() completely, which is not correct.
>>
>> Since your patch was already applied (btw. 3:20 hours after you submitted it),
>> please revert it and resend. Thank you.
>
> I also have considered this way, one question is that do we need to do more roll
> back work before V1 check?
>
> Specifically, In smc_find_rdma_v2_device_serv(), there are the following steps:
>
> 1. smc_listen_rdma_init()
> 1.1 smc_conn_create()
> 1.2 smc_buf_create() --> may fail
> 2. smc_listen_rdma_reg() --> may fail
>
> When later steps fail, Do we need to roll back previous steps?

That is a good question and I think that is a different problem for another patch.
smc_listen_rdma_init() maybe should call smc_conn_abort() similar to what smc_listen_ism_init()
does in this situation. And when smc_listen_rdma_reg() fails ... hmm we need to think about this.

We will also discuss this here in our team.

\
 
 \ /
  Last update: 2022-05-23 14:44    [W:0.177 / U:0.180 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site