Re: kern/55664: rump race condition

[Ruslan Nikolaev <nruslan_devel%yahoo.com@localhost>]

Good point, Christos! I was also a bit concerned about this... but 
eventually ended up with this implementation to avoid too many intrusive 
changes in the code. The reason was that each sleepq gets the 1:1 
mapping to the corresponding rump conditional variable. Therefore, it 
was easier to just keep it inside sleepq. Do you have any specific 
suggestions?

On 11/1/20 12:50 PM, Christos Zoulas wrote:
> The following reply was made to PR kern/55664; it has been noted by GNATS.
>
> From: Christos Zoulas <christos%zoulas.com@localhost>
> To: gnats-bugs%netbsd.org@localhost
> Cc: kern-bug-people%netbsd.org@localhost,
>   gnats-admin%netbsd.org@localhost,
>   netbsd-bugs%netbsd.org@localhost,
>   nruslan_devel%yahoo.com@localhost
> Subject: Re: kern/55664: rump race condition
> Date: Sun, 1 Nov 2020 12:47:42 -0500
>
>   --Apple-Mail=_4CED2A92-2D99-4473-BC8B-D343C781E17E
>   Content-Transfer-Encoding: quoted-printable
>   Content-Type: text/plain;
>   	charset=us-ascii
>   
>   I just looked at it. Part 1 is fine, but the implementation of part 2 =
>   (adding an additional kcondvar_t to struct sleepq depending on #ifdef =
>   _RUMPKERN) is something that we don't do in general. We try to keep the =
>   code to have the least amount of #ifdef _RUMPKERN and the data =
>   structures be the same between the two implementations. Perhaps there is =
>   a different way to do this?
>   
>   christos
>