Message ID | 1400694356-12193-1-git-send-email-stefan.bader@canonical.com |
---|---|
State | New |
Headers | show |
Should have said in the topic, this is for utopic of course
On Wed, May 21, 2014 at 07:45:56PM +0200, Stefan Bader wrote: > Someone found virtio-scsi missing in utopic cloud-images. Turns out > there were some other virtio and one vhost module missing in utopic > which had been in the inclusion list for trusty. > Re-added those and took the liberty to add vhost stuff as well as that > smelled like virtual machine stuff of some kind... VHOST* seem to be host side accellerators for virtio: tristate "Host kernel accelerator for virtio net" It is not clear that there is any benefit in those being in -virtual if we claim it is for the guest images. As we do not include most of the hardware drivers in the virtual image it is not clear you could use that as a host? > Note: This *needs* to go with an abi bump as modules move around. NOTE NOTE NOTE ^^^^ -apw
On 22.05.2014 12:06, Andy Whitcroft wrote: > On Wed, May 21, 2014 at 07:45:56PM +0200, Stefan Bader wrote: >> Someone found virtio-scsi missing in utopic cloud-images. Turns out >> there were some other virtio and one vhost module missing in utopic >> which had been in the inclusion list for trusty. >> Re-added those and took the liberty to add vhost stuff as well as that >> smelled like virtual machine stuff of some kind... > > VHOST* seem to be host side accellerators for virtio: > > tristate "Host kernel accelerator for virtio net" > > It is not clear that there is any benefit in those being in -virtual if > we claim it is for the guest images. As we do not include most of the > hardware drivers in the virtual image it is not clear you could use that > as a host? Hm, it might be that I did not read the Kconfig help carefully enough. On the other hand in Trusty there was at least one of those included (of 3). And also for Xen at least xen-netback (which is host side) also is not in extras. I think I remember Scott telling us at some occasion that he long term wants the split in a way to have the minimal set usable for virt and server bare-metal. Not sure I remember that correctly. So under vhost potentially the two other modules are not needed. The target_core stuff which those made necessary probably is useful regardless (Ok, I would not say it is sensible to export a scsi target from within a VM but nevertheless some people may want to). Stefan > >> Note: This *needs* to go with an abi bump as modules move around. > > NOTE NOTE NOTE ^^^^ > > -apw >
On Thu, May 22, 2014 at 12:19:00PM +0200, Stefan Bader wrote: > On 22.05.2014 12:06, Andy Whitcroft wrote: > > On Wed, May 21, 2014 at 07:45:56PM +0200, Stefan Bader wrote: > >> Someone found virtio-scsi missing in utopic cloud-images. Turns out > >> there were some other virtio and one vhost module missing in utopic > >> which had been in the inclusion list for trusty. > >> Re-added those and took the liberty to add vhost stuff as well as that > >> smelled like virtual machine stuff of some kind... > > > > VHOST* seem to be host side accellerators for virtio: > > > > tristate "Host kernel accelerator for virtio net" > > > > It is not clear that there is any benefit in those being in -virtual if > > we claim it is for the guest images. As we do not include most of the > > hardware drivers in the virtual image it is not clear you could use that > > as a host? > > Hm, it might be that I did not read the Kconfig help carefully enough. On the > other hand in Trusty there was at least one of those included (of 3). And also > for Xen at least xen-netback (which is host side) also is not in extras. > I think I remember Scott telling us at some occasion that he long term wants the > split in a way to have the minimal set usable for virt and server bare-metal. > Not sure I remember that correctly. Yes I think they want a "no desktop dross" version of the kernel. When I revamped the -virtual inclusion stuff I did make it possible for it to be iterative. So we could furhter sprint -extras into such that core h/w drivers are together and everything else is together. Using the existing meta packages to group them appropriatly as we do now with -generic. > So under vhost potentially the two other modules are not needed. The target_core > stuff which those made necessary probably is useful regardless (Ok, I would not > say it is sensible to export a scsi target from within a VM but nevertheless > some people may want to). O.k. -apw
diff --git a/debian.master/control.d/generic.inclusion-list b/debian.master/control.d/generic.inclusion-list index f926ebd..1f56c9c 100644 --- a/debian.master/control.d/generic.inclusion-list +++ b/debian.master/control.d/generic.inclusion-list @@ -14,6 +14,7 @@ drivers/block/floppy.ko drivers/block/cryptoloop.ko drivers/block/rbd.ko drivers/char/hangcheck-timer.ko +drivers/char/hw_random/virtio-rng.ko drivers/char/ipmi/ipmi_msghandler.ko drivers/char/lp.ko drivers/char/nvram.ko @@ -39,6 +40,7 @@ drivers/md/* drivers/message/fusion* drivers/misc/vmw_balloon.ko drivers/misc/vmw_vmci/vmw_vmci.ko +drivers/net/caif/caif_virtio.ko drivers/net/mii.ko drivers/net/ethernet/8390/8390.ko drivers/net/ethernet/realtek/8139too.ko @@ -86,7 +88,10 @@ drivers/scsi/scsi_transport_sas.ko drivers/scsi/scsi_tgt.ko drivers/scsi/vmw_pvscsi.ko drivers/scsi/hv_storvsc.ko +drivers/scsi/virtio_scsi.ko +drivers/target/target_core*.ko drivers/usb/storage/usb-storage.ko +drivers/vhost/* drivers/video/cirrusfb.ko drivers/video/output.ko drivers/video/syscopyarea.ko