Re: Userspace woes with 5.1.5 due to TIPC

[Ying Xue]

On 5/31/19 4:47 AM, Mihai Moldovan wrote:
> * On 5/30/19 9:51 PM, Jon Maloy wrote:
>> Make sure the following three commits are present in TIPC *after* the offending commit:
>>
>> commit 532b0f7ece4c "tipc: fix modprobe tipc failed after switch order of device registration"
> 
> This *is* the offending commit, as far as I understand. Merely rebased in
> linux-stable, and hence having a different SHA, but mentioning the original SHA
> (i.e., 532b0f7ece4c) in its commit message.
> 
> 
>> Since that patch one was flawed it had to be reverted:
>> commit 5593530e5694  ""Revert tipc: fix modprobe tipc failed after switch order of device registration"
>>
>> It was then replaced with this one: 
>> commit 526f5b851a96 "tipc: fix modprobe tipc failed after switch order of device registration"
> 
> Okay, these two are not part of 5.1.5. I've backported them (and only these two)
> to 5.1.5 and the issue(s) seem to be gone. Definitely something that should be
> backported to/included in 5.1.6.
> 
> 
> Thanks for pointing all that out! Unfortunately I didn't add anything useful but
> noise, since you obviously already knew, that this commit was broken. I'd urge
> Greg to release a new stable version including the fixes soon, if possible,
> though, for not being able to start/use userspace browsers sounds like a pretty
> bad regression to me.
> 

Not only commit 526f5b851a96 )("tipc: fix modprobe tipc failed after
switch order of device registration") has to be reverted, but also I
found commit 7e27e8d6130c ("tipc: switch order of device registration to
fix a crash") introduced a serious regression which makes tipc internal
topology service server failed to be created.

Today I will fix it with the following approaches:
1. Revert commit 7e27e8d6130c ("tipc: switch order of device
registration to fix a crash")
2. Use another method to solve the problem that commit 7e27e8d6130c
tries to fix.

Thanks,
Ying

> 
> 
> Mihai
>