Re: [PATCH v8 00/50] linux: towards virtio-1 guest support

[Michael S. Tsirkin]

On Mon, Dec 01, 2014 at 06:02:42PM +0200, Michael S. Tsirkin wrote:
> David, assuming patches are acceptable, I'd like them all to be merged through
> virtio or vhost trees to avoid dependency headaches.  Could you please ack
> doing that for net related patches, especially these:
> 	  tun: move internal flag defines out of uapi
> 	  tun: drop most type defines
> 	  tun: add VNET_LE flag
> 	  tun: TUN_VNET_LE support, fix sparse warnings for virtio headers
> 	  macvtap: TUN_VNET_HDR support
> 	  vhost/scsi: partial virtio 1.0 support
> 	  af_packet: virtio 1.0 stubs
> Alternatively, I can prepare a tree with virtio changes and merge
> that into your net-next - if Rusty approves, or keeps
> silence :)

David, Rusty - ping.
Are both of you ok withmerging this through the vhost tree?

> 
> Could you comment on this please?
> 
> Up to half a hundred patches now. This is getting ridiculous.
> 
> Based on patches by Cornelia and others, but with an API that should allow
> better static checking of code, slightly more concervative changes in vring and
> drivers, and compatibility for existing drivers so that this series can be
> applied gradually.
> At the end of series, all drivers are converted, and compatibility flags are
> dropped.
> 
> All drivers have been converted now, with the exception
> of the balloon driver which we plan to rewrite
> rather than convert.
> All drivers now pass sparse without warnings.
> 
> net and blk patches have been tested on s390.
> Other patches pass sparse so they are most likely
> ok too, but haven't been through testing yet -
> they can be dropped from patchset if necessary.
> 
> Please review, and consider for 3.19
> 
> changes since v7:
> 	all drivers have been converted now.
> 	new patches:
> 	  virtio_console: virtio 1.0 support
> 	  virtio_balloon: add legacy_only flag
> 	  virtio: make VIRTIO_F_VERSION_1 a transport bit
> 	  virtio: drop VIRTIO_F_VERSION_1 from drivers
> 	vhost: make features 64 bit
> 		fix bug noted by Sergei Shtylyov,
> 		bugfix suggested by Ben Hutchings
> 	virtio_net: disable mac write for virtio 1.0
> 		moved earlier in the series
> 	vhost/net: force len for TX to host endian
> 		commit log reworded
> 	vhost: virtio 1.0 endian-ness support
> 		simplify error printing code sligthly
> 	vhost/net: suppress compiler warning
> 		smashed into code introducing the warning,
> 		added a comment
> 
> 
> changes since v6:
> 	tun: move internal flag defines out of uapi
> 		document reasons for patch split
> 	macvtap: TUN_VNET_LE support
> 		fix patch subject
> 	virtio: add low-level APIs for feature bits
> 		fix comment in code s/set/clear/
> 		document reasons for names selected
> 	KVM: s390: enable virtio-ccw revision 1
> 		fix up comment
> 	virtio: use u32, not bitmap for features
> 		fix whitespace
> 	virtio_ccw: add support for 64 bit features
> 		tweak comment in code
> 
> 	included reviewed-by tags by Cornelia David and Jason
> 
> Changes since v5:
> 	rewritten feature bit handling, based on suggestion by David:
> 	changes are now smaller and easier to review.
> 	new patch:
> 	  virtio: add low-level APIs for feature bits
> 	rewritten patches:
> 	  virtio: use u32, not bitmap for features
> 	  virtio: add support for 64 bit features.
> 	
> Changes since v4:
> 	I have dropped 64 bit feature negotiation from
> 	most transports (except ccw).
> 	For symmetry, I split out ccw bits:
> 		  virtio: add support for 64 bit features.
> 			adds core and device support for 64 bit features
> 		  virtio_ccw: add support for 64 bit features.
> 			actually enables 64 bit features in ccw.
> 	detail below.
> 
> 	this is still really a cosmetic change: just reordering/splitting
> 	patches, and dropping code changes that aren't needed.
> 
> 	new patches:
> 	  mic_virtio: robust feature array size calculation
> 	  virtio: assert 32 bit features in transports
> 
> 	addressed comments by Cornelia:
> 		      virtio: disable virtio 1.0 in transports:
> 			patch dropped
> 		      virtio: memory access APIs
> 			tweaked comments as suggested by Cornelia
> 		      virtio_config: endian conversion for v1.0
> 			dropped reviewed-by tags -
> 				they applied to a very old version of patch
> 		      virtio: set FEATURES_OK
> 			replaced printk by dev_err
> 		      virtio: add legacy feature table support
> 			comment changes suggested by Cornelia
> 		      virtio_blk: v1.0 support
> 			dropped reviewed-by tags -
> 				they applied to a very old version of patch
> 		      virtio_net: enable v1.0 support
> 			only write mac for legacy devices
> 		      vhost: add memory access wrappers
> 			move chunk to vhost: make features 64 bit
> 			better patch description
> 		      vhost: make features 64 bit
> 			moved this patch earlier in series
> 		      vhost/net: force len for TX to host endian
> 			better patch description
> 		      virtio_scsi: v1.0 support
> 			fix bug: use high level, not low level wrappers
> 			
> 
> 	included reviewed-by tags for patches:
> 		KVM: s390 allow virtio_ccw status writes to fail
> 				added reviewed-by tag by Cornelia
> 		virtio_scsi: v1.0 support
> 		virtio_scsi: move to uapi
> 		virtio_scsi: export to userspace
> 		vhost/scsi: partial virtio 1.0 support
> 				added reviewed-by tag by Paolo
> 
> Changes since v3:
> 	- move virtio 1.0 feature bit out of transport bits:
> 	  for now it's managed by drivers
> 	- disable virtio 1.0 feature bit explicitly - for transports
> 	  that don't support it
> 
> Cornelia Huck (3):
>   virtio: allow transports to get avail/used addresses
>   KVM: s390: virtio-ccw revision 1 SET_VQ
>   KVM: s390: enable virtio-ccw revision 1
> 
> Michael S. Tsirkin (46):
>   virtio: add low-level APIs for feature bits
>   virtio: use u32, not bitmap for features
>   mic_virtio: robust feature array size calculation
>   virtio: add support for 64 bit features.
>   virtio: assert 32 bit features in transports
>   virtio_ccw: add support for 64 bit features.
>   virtio: add virtio 1.0 feature bit
>   virtio: memory access APIs
>   virtio_ring: switch to new memory access APIs
>   virtio_config: endian conversion for v1.0
>   virtio: set FEATURES_OK
>   virtio: simplify feature bit handling
>   virtio: add legacy feature table support
>   virtio_net: v1.0 endianness
>   virtio_blk: v1.0 support
>   KVM: s390 allow virtio_ccw status writes to fail
>   virtio_blk: make serial attribute static
>   virtio_blk: fix race at module removal
>   virtio_net: pass vi around
>   virtio_net: get rid of virtio_net_hdr/skb_vnet_hdr
>   virtio_net: stricter short buffer length checks
>   virtio_net: bigger header when VERSION_1 is set
>   virtio_net: disable mac write for virtio 1.0
>   virtio_net: enable v1.0 support
>   vhost: make features 64 bit
>   vhost: add memory access wrappers
>   vhost/net: force len for TX to host endian
>   vhost: switch to __get/__put_user exclusively
>   vhost: virtio 1.0 endian-ness support
>   vhost/net: virtio 1.0 byte swap
>   vhost/net: larger header for virtio 1.0
>   vhost/net: enable virtio 1.0
>   tun: move internal flag defines out of uapi
>   tun: drop most type defines
>   tun: add VNET_LE flag
>   tun: TUN_VNET_LE support, fix sparse warnings for virtio headers
>   macvtap: TUN_VNET_LE support
>   virtio_scsi: v1.0 support
>   virtio_scsi: move to uapi
>   virtio_scsi: export to userspace
>   vhost/scsi: partial virtio 1.0 support
>   af_packet: virtio 1.0 stubs
>   virtio_console: virtio 1.0 support
>   virtio_balloon: add legacy_only flag
>   virtio: make VIRTIO_F_VERSION_1 a transport bit
>   virtio: drop VIRTIO_F_VERSION_1 from drivers
> 
> Thomas Huth (1):
>   KVM: s390: Set virtio-ccw transport revision
> 
>  drivers/vhost/vhost.h                  |  37 ++++++-
>  include/linux/virtio.h                 |  12 ++-
>  include/linux/virtio_byteorder.h       |  59 +++++++++++
>  include/linux/virtio_config.h          | 100 +++++++++++++++++--
>  include/uapi/linux/if_tun.h            |  17 +---
>  include/uapi/linux/virtio_blk.h        |  15 +--
>  include/uapi/linux/virtio_config.h     |   9 +-
>  include/uapi/linux/virtio_console.h    |   7 +-
>  include/uapi/linux/virtio_net.h        |  15 +--
>  include/uapi/linux/virtio_ring.h       |  45 ++++-----
>  include/{ => uapi}/linux/virtio_scsi.h | 106 ++++++++++----------
>  include/uapi/linux/virtio_types.h      |  46 +++++++++
>  tools/virtio/linux/virtio.h            |  22 +----
>  tools/virtio/linux/virtio_config.h     |   2 +-
>  drivers/block/virtio_blk.c             |  74 ++++++++------
>  drivers/char/virtio_console.c          |  30 +++---
>  drivers/lguest/lguest_device.c         |  13 ++-
>  drivers/misc/mic/card/mic_virtio.c     |  10 +-
>  drivers/net/macvtap.c                  |  68 ++++++++-----
>  drivers/net/tun.c                      | 168 ++++++++++++++------------------
>  drivers/net/virtio_net.c               | 161 +++++++++++++++---------------
>  drivers/remoteproc/remoteproc_virtio.c |   7 +-
>  drivers/s390/kvm/kvm_virtio.c          |   7 +-
>  drivers/s390/kvm/virtio_ccw.c          | 172 +++++++++++++++++++++++++++------
>  drivers/scsi/virtio_scsi.c             |  50 ++++++----
>  drivers/vhost/net.c                    |  31 +++---
>  drivers/vhost/scsi.c                   |  22 +++--
>  drivers/vhost/vhost.c                  |  93 +++++++++++-------
>  drivers/virtio/virtio.c                |  78 +++++++++++----
>  drivers/virtio/virtio_balloon.c        |   1 +
>  drivers/virtio/virtio_mmio.c           |  13 ++-
>  drivers/virtio/virtio_pci.c            |   8 +-
>  drivers/virtio/virtio_ring.c           | 109 ++++++++++++---------
>  net/packet/af_packet.c                 |  35 ++++---
>  tools/virtio/virtio_test.c             |   5 +-
>  tools/virtio/vringh_test.c             |  16 +--
>  include/uapi/linux/Kbuild              |   2 +
>  37 files changed, 1061 insertions(+), 604 deletions(-)
>  create mode 100644 include/linux/virtio_byteorder.h
>  rename include/{ => uapi}/linux/virtio_scsi.h (73%)
>  create mode 100644 include/uapi/linux/virtio_types.h
> 
> -- 
> MST
> 
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/