On 2018/05/24 15:06, [email protected] wrote:
> Thanks.
>  
> You think it's a bug in vmxnet3 (vm tools ?) or something like running out of 
> buffers that
> probably could be fixed by assigning more of them ?
>  
> Gesendet: Donnerstag, 24. Mai 2018 um 13:56 Uhr
> Von: "Stuart Henderson" <[email protected]>
> An: keenonkites <[email protected]>
> Cc: [email protected]
> Betreff: Re: panic: vmxnet3_getbuf: buffer has mbuf
> On 2018/05/24 02:03, keenonkites wrote:
> > Stuart,
> > did you get any further ?
> >
> > We're having the same problem here with OpenBSD (6.0 and OpenBSD 6.2) when
> > taking ESXi snapshots of an OpenBSD machine after some uptime (freshly
> > rebooted machines do not crash during snapshot).
> >
> >
> >
> > Any solution/workaround to prevent the crash would be welcomed
> 
> No idea on a fix. Workaround: change the vmx file to use
> 
> ethernet0.virtualDev = "e1000"
> 
> and OpenBSD config to use em(4).
>  

"OpenBSD kernel crashes under normal use" -> to my eyes, that's a bug.
But it might be that it's triggered by a VMware bug, too.

Pragmatically:

- it's stable with e1000

- it's going to be a pig to fix: something that takes days of uptime
to trigger makes debugging very hard

- the people who know their way around the kernel well enough to
debug this likely have hundreds of to-dos that will be a bigger
benefit to more people ...

So I have taken the view that switching to e1000 is the sane option
here.

Reply via email to