On Wednesday, July 22, 2015 at 1:43:59 PM UTC-7, Mike Christie wrote:
>
> On 07/22/2015 10:24 AM, The Lee-Man wrote: 
> > On Tuesday, July 21, 2015 at 9:29:19 PM UTC-7, Mike Christie wrote: 
> > 
> >     On 07/21/2015 05:47 PM, leeman...@gmail.com <javascript:> wrote: 
> >     > From: Lee Duncan <ldu...@suse.com <javascript:>> 
> >     > 
> >     > This patch allows iser transport to be used for the discovery 
> >     > daemon. Otherwise, iscsid core dumps when attempting this. 
> >     > --- 
> >     >  usr/discoveryd.c | 5 ----- 
> >     >  1 file changed, 5 deletions(-) 
> >     > 
> >     > diff --git a/usr/discoveryd.c b/usr/discoveryd.c 
> >     > index 1e149771a50b..2d3ccbcd722f 100644 
> >     > --- a/usr/discoveryd.c 
> >     > +++ b/usr/discoveryd.c 
> >     > @@ -1034,11 +1034,6 @@ static void __do_st_disc_and_login(struct 
> >     discovery_rec *drec) 
> >     >          drec->u.sendtargets.reopen_max = 0; 
> >     >   
> >     >          iface_link_ifaces(&setup_ifaces); 
> >     > -        /* 
> >     > -         * disc code assumes this is not set and wants to use 
> >     > -         * the userspace IO code. 
> >     > -         */ 
> >     > -        ipc = NULL; 
> >     >   
> >     >          rc = idbm_bind_ifaces_to_nodes(discovery_sendtargets, 
> drec, 
> >     >                                          &setup_ifaces, 
> &rec_list); 
> >     > 
> > 
> >     Do you need this patch for offload support too, and does it work ok 
> now 
> >     too, or was that already working? 
> > 
> > 
> > That was already working. With this patch, offload via IB/iSER seems 
> > to be working for us. 
> > 
>
> For offload, like bnx2i, was it doing discovery through the offload 
> engine or in software for you? I thought it would crash in 
> iscsi_create_leading_conn when it references the ipc pointer here: 
>
>         conn->socket_fd = ipc->ctldev_open(); 
>
> for bnx2i. 
>
> Your patch is correct. I am just trying to figure out why I wrote that 
> "disc code assumes" comment above. It seems like my comment in the code 
> is very very wrong, because if CAP_TEXT_NEGO, like with 
> bnx2i/cxgb/be2iscsi and in newer kernels where we now set that bit iser, 
> then we want a valid ipc pointer. 
>

I have never tried running the discovery daemon with bnx2i. Regular 
discovery
through bnx2i works fine without this patch.

So I set it up just now and confirmed: using the discovery daemon for bnx2i
also gets a core dump without this patch. And regular discovery is verified
as working with or without this patch using bnx2i. 

-- 
You received this message because you are subscribed to the Google Groups 
"open-iscsi" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to open-iscsi+unsubscr...@googlegroups.com.
To post to this group, send email to open-iscsi@googlegroups.com.
Visit this group at http://groups.google.com/group/open-iscsi.
For more options, visit https://groups.google.com/d/optout.

Reply via email to