svn commit: r325841 - in head/sys: conf dev/mlx4 dev/mlx4/mlx4_core dev/mlx4/mlx4_en dev/mlx4/mlx4_ib modules/mlx4

[Hans Petter Selasky]

On 12/18/17 23:50, Colin Percival wrote:
> On 12/18/17 00:15, Hans Petter Selasky wrote:
>> On 12/18/17 05:29, Colin Percival wrote:
>>> Also, it breaks some work I have in progress for instrumenting SYSINITs.
>>> Would you mind moving the DEFINE_MUTEX line to occur immediately prior to
>>> the set_port_type function, rather than being placed inside it?
>>
>> I'll have a look at this later today. Your point is valid!
> 
> On further examination, it looks like DEFINE_MUTEX is something used in Linux
> kernel code, and the way it works there does allow it to be used inside a
> function.  Is it possible to change the linuxkpi code to make it safe?  (It
> looks like our mutex initialization is considerably more complicated than
> what Linux does, so maybe not...?)

Hi,

Can you explain what you mean by safe?

DEFINE_MUTEX() is simply a wrapper for another SYSINIT. It has nothing 
to do with the function it is placed inside.

> 
> I have a feeling that we probably don't want to end up in a position of
> "every time we import code from Linux, we need to grep for DEFINE_MUTEX
> and hoist all of them out of functions".
> 

Can we continue this discussion here:

https://reviews.freebsd.org/D13530

--HPS