Re: [dpdk-dev] Crash related to virtio NICs in DPDK 2.0.0 on Freebsd 10.1 VM

["Ouyang, Changchun" <changchun.ouyang@intel.com>]

Hi Raz,

Thanks for identifying this issue.
A comments below.

> -----Original Message-----
> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Raz Amir
> Sent: Monday, April 13, 2015 7:54 PM
> To: 'Thomas Monjalon'
> Cc: dev@dpdk.org
> Subject: Re: [dpdk-dev] Crash related to virtio NICs in DPDK 2.0.0 on Freebsd
> 10.1 VM
> 
> Thanks. I will submit a patch

When you submit another patch,
Would you pls think about this solution?
Declare a static var to keep the fd which is opened for freebsd 
Then define a deinit function for virtio device,
Inside the deinit function, close the fd which was opened in init stage.
Done.

thanks 
Changchun

> 
> -----Original Message-----
> From: Thomas Monjalon [mailto:thomas.monjalon@6wind.com]
> Sent: 13 April 2015 13:46
> To: Raz Amir
> Cc: dev@dpdk.org; david.marchand@6wind.com
> Subject: Re: Crash related to virtio NICs in DPDK 2.0.0 on Freebsd 10.1 VM
> 
> 2015-04-08 18:53, Raz Amir:
> > The issues happens also in dpdk 1.8.0, and related to patch
> > http://dpdk.org/dev/patchwork/patch/239/
> >
> > Adding Thomas and David to the thread and I will appreciate your input.
> >
> > The patch comes to solve a file descriptor leak in the bsdapp version
> > of rte_eal_iopl_init after opening the /dev/io device.
> >
> > Seems like this isn't a file descriptor leak, and it should remain
> > open - as I wrote below, I am using virtio.
> 
> Thanks for the bug report.
> It seems there was no validation for FreeBSD with virtio.
> 
> > After removing it and testing the crash was resolved.
> >
> > Any objection for removing the close(fd) that was added at dpdk 1.8.0?
> 
> No, there was a doubt because the man page was not clear.
> 	http://www.freebsd.org/cgi/man.cgi?query=io&sektion=4
> 
> In case you submit a patch, please add this line:
> Fixes: 8a312224bcde ("eal/bsd: fix fd leak")
> 
> > Are there scenarios that might be impacted by removing it?
> 
> I don't think so.
> 
> Thanks