Re: port-amd64/54988: possible memory leaks/swap problems

[mlh%goathill.org@localhost (MLH)]

Lars Reichardt wrote:
> The following reply was made to PR port-amd64/54988; it has been noted by GNATS.
> 
> From: Lars Reichardt <lars%paradoxon.info@localhost>
> To: MLH <mlh%goathill.org@localhost>, gnats-bugs%netbsd.org@localhost
> Cc: port-amd64-maintainer%netbsd.org@localhost, gnats-admin%netbsd.goathill.org@localhost,
>  netbsd-bugs%netbsd.org@localhost
> Subject: Re: port-amd64/54988: possible memory leaks/swap problems
> Date: Tue, 10 Mar 2020 18:29:43 +0100
> 
>  On 10.03.2020 17:54, MLH wrote:
>  > mlh wrote:
>  >> MLH wrote:
>  >>> The following reply was made to PR port-amd64/54988; it has been noted by GNATS.
>  >>>   
>  >>>   Any idea on what is preventing the memory from being released and
>  >>>   why it just locks up?
>  > ...
>  >> though there still appear
>  >> to be memory alloc/dealloc and swap issues because I have to reboot
>  >> at least once a day when phymem is exhausted and the system becomes
>  >> unusable due to severe swapping. This is when basically nothing is
>  >> still running and disc cache is only taking up a tiny bit of phymem
>  >> and only a tiny bit of swap appears to be in use.
...
>  > So the system is down to less than 20M of phymem and never recovers
>  > much more than maybe 10M of it. Have to reboot to reclaim a workable
>  > system, so I have to reboot 2-3 times a day to do much of any work.
>  >
>  what does vmstat -mvW show when memory gets low the whole output?
>  
>  I'm interested especially in the kva pools.
>  
>  Lars

Ok. I will check that next time. As of NetBSD 9.99.51 Sun Mar 22,
the system just grinds to a halt when it runs out of phymem. Still
seems that almost no swap is being used but at least it doesn't
crash. Yesterday it got down to less than 1M of phymem and took me
about an hour to log in from my phone over wifi in order to do a
clean shutdown as I couldn't get a console up. I had managed to
shut X down but was getting out of memory errors. Even on shutdown
many things couldn't cleanly shut down due to out of memory errors.

I still suspect something in radeondrmkms is leaking memory.