On Wed, 2009-08-19 at 01:00 -0700, Eric Cheng wrote: > On Tue, Aug 18, 2009 at 10:39:02AM -0400, Sebastien Roy wrote: > > > 1148: > > > this assert doesn't seem valid all the time. (e.g. if the daemon dies > > > and couldn't handle the upcall) > > > > What upcall are you referring to? dls_devnet_hold() does not fail if > > dlmgmtd is unable to handle upcalls. It can only fail if the linkid > > doesn't exist in i_dls_devnet_id_hash, or if the dls_devnet_t is > > condemned. Because i_dls_devnet_create_iptun() succeeded, the entry > > must be there and isn't condemned. > > > > I was talking about dls_devnet_hold()->dls_devnet_hold_common()-> > dls_mgmt_get_phydev()->i_dls_mgmt_get_linkattr()->i_dls_mgmt_upcall().
Yes, I thought so, but dls_devnet_hold_common() doesn't fail if dls_mgmt_get_phydev() fails. > but anyway I think it's very unlikely that this could happen since > i_dls_devnet_create_iptun() would fail first if the dlmgmtd disappeared. > > I'm fine with your other responses. Okay, thanks. -Seb
