| Message ID | 1360085908-6050-1-git-send-email-herton.krzesinski@canonical.com |
|---|---|
| State | New |
| Headers | show |
On 02/05/2013 09:38 AM, Herton Ronaldo Krzesinski wrote: > BugLink: http://bugs.launchpad.net/bugs/1015339 > Signed-off-by: Herton Ronaldo Krzesinski <herton.krzesinski@canonical.com> > --- > debian.master/d-i/modules/nic-modules | 2 ++ > 1 file changed, 2 insertions(+) > > diff --git a/debian.master/d-i/modules/nic-modules b/debian.master/d-i/modules/nic-modules > index dca7b7b..563f4c6 100644 > --- a/debian.master/d-i/modules/nic-modules > +++ b/debian.master/d-i/modules/nic-modules > @@ -93,6 +93,8 @@ lasi_82596 ? > lne390 ? > lp486e ? > mace ? > +mlx4_core ? > +mlx4_en ? > mv643xx_eth ? > myri_sbus ? > natsemi ? >
On 05.02.2013 18:38, Herton Ronaldo Krzesinski wrote: > BugLink: http://bugs.launchpad.net/bugs/1015339 > Signed-off-by: Herton Ronaldo Krzesinski <herton.krzesinski@canonical.com> > --- > debian.master/d-i/modules/nic-modules | 2 ++ > 1 file changed, 2 insertions(+) > > diff --git a/debian.master/d-i/modules/nic-modules b/debian.master/d-i/modules/nic-modules > index dca7b7b..563f4c6 100644 > --- a/debian.master/d-i/modules/nic-modules > +++ b/debian.master/d-i/modules/nic-modules > @@ -93,6 +93,8 @@ lasi_82596 ? > lne390 ? > lp486e ? > mace ? > +mlx4_core ? > +mlx4_en ? > mv643xx_eth ? > myri_sbus ? > natsemi ? > When talking to Scott yesterday, there also seems a mlx4_ib (infiniband) and ports can be in one or the other mode. While the core driver has all the pci ids and potentially can find out which mode the ports are in, it does not cause either of the network parts to be loaded. Sounds a bit like non-ideal design and some discussion yesterday (on #ubuntu-devel) sounded like this could become another request sooner or later... -Stefan
On Wed, Feb 06, 2013 at 09:12:01AM +0100, Stefan Bader wrote: > On 05.02.2013 18:38, Herton Ronaldo Krzesinski wrote: > > BugLink: http://bugs.launchpad.net/bugs/1015339 > > Signed-off-by: Herton Ronaldo Krzesinski <herton.krzesinski@canonical.com> > > --- > > debian.master/d-i/modules/nic-modules | 2 ++ > > 1 file changed, 2 insertions(+) > > > > diff --git a/debian.master/d-i/modules/nic-modules b/debian.master/d-i/modules/nic-modules > > index dca7b7b..563f4c6 100644 > > --- a/debian.master/d-i/modules/nic-modules > > +++ b/debian.master/d-i/modules/nic-modules > > @@ -93,6 +93,8 @@ lasi_82596 ? > > lne390 ? > > lp486e ? > > mace ? > > +mlx4_core ? > > +mlx4_en ? > > mv643xx_eth ? > > myri_sbus ? > > natsemi ? > > > > When talking to Scott yesterday, there also seems a mlx4_ib (infiniband) and > ports can be in one or the other mode. While the core driver has all the pci ids > and potentially can find out which mode the ports are in, it does not cause > either of the network parts to be loaded. Sounds a bit like non-ideal design and > some discussion yesterday (on #ubuntu-devel) sounded like this could become > another request sooner or later... The way I see it, is that mlx4_en covers the ethernet capabilities of the hardware, and mlx4_ib only the infiniband part, I couldn't find anything to the contrary saying that mlx4_ib needs to be loaded with mlx4_en, but I also don't know anything about this hardware and could be wrong here. Also we don't include or provide udebs for infiniband drivers as far as I looked. So for the installer, looks loading only mlx4_en solves the ethernet part and suffices to be loaded alone, and about infiniband we don't need or support that yet inside it. But would be good if someone with the hardware or knowledge about it could confirm. > > -Stefan >
On 06.02.2013 15:07, Herton Ronaldo Krzesinski wrote: > On Wed, Feb 06, 2013 at 09:12:01AM +0100, Stefan Bader wrote: >> On 05.02.2013 18:38, Herton Ronaldo Krzesinski wrote: >>> BugLink: http://bugs.launchpad.net/bugs/1015339 >>> Signed-off-by: Herton Ronaldo Krzesinski <herton.krzesinski@canonical.com> >>> --- >>> debian.master/d-i/modules/nic-modules | 2 ++ >>> 1 file changed, 2 insertions(+) >>> >>> diff --git a/debian.master/d-i/modules/nic-modules b/debian.master/d-i/modules/nic-modules >>> index dca7b7b..563f4c6 100644 >>> --- a/debian.master/d-i/modules/nic-modules >>> +++ b/debian.master/d-i/modules/nic-modules >>> @@ -93,6 +93,8 @@ lasi_82596 ? >>> lne390 ? >>> lp486e ? >>> mace ? >>> +mlx4_core ? >>> +mlx4_en ? >>> mv643xx_eth ? >>> myri_sbus ? >>> natsemi ? >>> >> >> When talking to Scott yesterday, there also seems a mlx4_ib (infiniband) and >> ports can be in one or the other mode. While the core driver has all the pci ids >> and potentially can find out which mode the ports are in, it does not cause >> either of the network parts to be loaded. Sounds a bit like non-ideal design and >> some discussion yesterday (on #ubuntu-devel) sounded like this could become >> another request sooner or later... > > The way I see it, is that mlx4_en covers the ethernet capabilities of > the hardware, and mlx4_ib only the infiniband part, I couldn't find > anything to the contrary saying that mlx4_ib needs to be loaded with > mlx4_en, but I also don't know anything about this hardware and could be > wrong here. Also we don't include or provide udebs for infiniband > drivers as far as I looked. From the documentation I saw, it looks like those cards can have multiple ports and each port can be configured to as either infiniband or ethernet (through some management software). Not sure really whether we want to support infiniband for installation (could require a lot more different setup). So the ib module may not be as important. The other part of the problem was that only the core module has the pci ids and therefor gets autoloaded. The mlx4_en module has to be manually loaded. This may be done via udev rule or installer option but there is a certain expectation that things should just work. And that is where we may be expected to do "something"... -Stefan > > So for the installer, looks loading only mlx4_en solves the ethernet > part and suffices to be loaded alone, and about infiniband we don't need > or support that yet inside it. But would be good if someone with the > hardware or knowledge about it could confirm. > >> >> -Stefan >> >
diff --git a/debian.master/d-i/modules/nic-modules b/debian.master/d-i/modules/nic-modules index dca7b7b..563f4c6 100644 --- a/debian.master/d-i/modules/nic-modules +++ b/debian.master/d-i/modules/nic-modules @@ -93,6 +93,8 @@ lasi_82596 ? lne390 ? lp486e ? mace ? +mlx4_core ? +mlx4_en ? mv643xx_eth ? myri_sbus ? natsemi ?
BugLink: http://bugs.launchpad.net/bugs/1015339 Signed-off-by: Herton Ronaldo Krzesinski <herton.krzesinski@canonical.com> --- debian.master/d-i/modules/nic-modules | 2 ++ 1 file changed, 2 insertions(+)