)]}'
{"/COMMIT_MSG":[{"author":{"_account_id":1779,"name":"Daniel Berrange","email":"berrange@redhat.com","username":"berrange"},"change_message_id":"3d08c3f9692c61b60ab3ae4258e5965c00aeba0a","unresolved":false,"context_lines":[{"line_number":10,"context_line":"But nova doesn\u0027t support setting the controller index according"},{"line_number":11,"context_line":"to the disk devname, which may result in scsi device can\u0027t be"},{"line_number":12,"context_line":"recognized by the instance.So add the function of setting the"},{"line_number":13,"context_line":"controller index according to the dev name."},{"line_number":14,"context_line":""},{"line_number":15,"context_line":"Closes-bug: #1591827"},{"line_number":16,"context_line":""}],"source_content_type":"text/x-gerrit-commit-message","patch_set":2,"id":"dada55a8_931931da","line":13,"updated":"2016-07-20 13:33:51.000000000","message":"The real problem here, as described in the bug is that Nova is giving the CDROM a name of \u0027sdz\u0027, when there are no other devices on the SCSI bus already. \n\nThe correct approach is to fix nova\u0027s assignment of device names so that instead of hardcoding sdz for the config drive, it simply picks the first drive letter on the SCSI bus that is available, after all other drives have been assigned. This would likely be sda for aarch64.  At which point everything else should just work, and the code in this patch becomes irrelevant","commit_id":"8dbcd11aca3855308535e34c324df5cfea41f6cc"},{"author":{"_account_id":22076,"name":"Kevin Zhao","email":"kevin.zhao@linaro.org","username":"KevinZhao"},"change_message_id":"37d73eb7999cf2a2536ef29d266d50751415034d","unresolved":false,"context_lines":[{"line_number":10,"context_line":"But nova doesn\u0027t support setting the controller index according"},{"line_number":11,"context_line":"to the disk devname, which may result in scsi device can\u0027t be"},{"line_number":12,"context_line":"recognized by the instance.So add the function of setting the"},{"line_number":13,"context_line":"controller index according to the dev name."},{"line_number":14,"context_line":""},{"line_number":15,"context_line":"Closes-bug: #1591827"},{"line_number":16,"context_line":""}],"source_content_type":"text/x-gerrit-commit-message","patch_set":2,"id":"dada55a8_7fd71c4b","line":13,"in_reply_to":"dada55a8_44ffeda9","updated":"2016-07-20 14:25:06.000000000","message":"OK got it~ That make sense.Seems that I don\u0027t need to modify it with virtio-scsi controller index. Just pick the first drive letter on scsi bus.\n\nMay be you say that bug in qemu I have reported :-)\nhttps://bugs.launchpad.net/qemu/+bug/1594239\nFixed in commit f58b39d2d5b6dea1a757e1dc7d67a44eac1c4f9c.\nFixed in Qemu and libvirt now....","commit_id":"8dbcd11aca3855308535e34c324df5cfea41f6cc"},{"author":{"_account_id":1779,"name":"Daniel Berrange","email":"berrange@redhat.com","username":"berrange"},"change_message_id":"429f57a0885208e7ca552c3a41742352a4269809","unresolved":false,"context_lines":[{"line_number":10,"context_line":"But nova doesn\u0027t support setting the controller index according"},{"line_number":11,"context_line":"to the disk devname, which may result in scsi device can\u0027t be"},{"line_number":12,"context_line":"recognized by the instance.So add the function of setting the"},{"line_number":13,"context_line":"controller index according to the dev name."},{"line_number":14,"context_line":""},{"line_number":15,"context_line":"Closes-bug: #1591827"},{"line_number":16,"context_line":""}],"source_content_type":"text/x-gerrit-commit-message","patch_set":2,"id":"dada55a8_44ffeda9","line":13,"in_reply_to":"dada55a8_492b8694","updated":"2016-07-20 14:11:15.000000000","message":"If the first controller is full then libvirt will automatically create additional controllers as needed.\n\nI suspect you\u0027re just hitting a bug in libirt which was recently fixed\n\ncommit 12ec22b68befce0ae8020635e7fd71d783b1d716\nAuthor: Marc Hartmayer \u003cmhartmay@linux.vnet.ibm.com\u003e\nDate:   Mon Jun 27 16:43:47 2016 +0200\n\n    qemu: SCSI hostdev hot-plug: Fix automatic creation of SCSI controllers","commit_id":"8dbcd11aca3855308535e34c324df5cfea41f6cc"},{"author":{"_account_id":22076,"name":"Kevin Zhao","email":"kevin.zhao@linaro.org","username":"KevinZhao"},"change_message_id":"23ee64c7b4ef01f130f8734ada32654bdea9f9a0","unresolved":false,"context_lines":[{"line_number":10,"context_line":"But nova doesn\u0027t support setting the controller index according"},{"line_number":11,"context_line":"to the disk devname, which may result in scsi device can\u0027t be"},{"line_number":12,"context_line":"recognized by the instance.So add the function of setting the"},{"line_number":13,"context_line":"controller index according to the dev name."},{"line_number":14,"context_line":""},{"line_number":15,"context_line":"Closes-bug: #1591827"},{"line_number":16,"context_line":""}],"source_content_type":"text/x-gerrit-commit-message","patch_set":2,"id":"dada55a8_492b8694","line":13,"in_reply_to":"dada55a8_931931da","updated":"2016-07-20 14:07:14.000000000","message":"Hi Daniel,\n    Nova has hardcode the config drive --\u003e cdrom as the last device name sdz. With your method, we can change the cdrom dev name to the front of the devname such as sda,sdb,,,  That will work due to add controller:\n\u003ccontroller type\u003d\"scsi\" model\u003d\"virtio-scsi\"/\u003e\nThe default index for controller virtio-scsi is index\u003d0, which corresponding to sda to sdg. But for example, sdh, it needs virtio-scsi controller that has index\u003d1 and so as sdz need index\u003d3. So...\n    Actually I meet another problem by virtio-scsi mode here. When I set hw_scsi_model\u003dvirtio-scsi and hw_disk_bus\u003dscsi, then I use \"nova volume-attach\" to add the volume to an existing instanc, I have add device one by one. Adding volume as sdb, sdc,sde are all OK, they can be hotplug. When the adding device has devname \"sdh\", it can\u0027t be recognized by the instance, due to the controller index problem.\n    So I think maybe add the index calculator according to its devname and add the controller xml corresponding devname can solve the problem.\n    Thanks~","commit_id":"8dbcd11aca3855308535e34c324df5cfea41f6cc"}],"nova/virt/libvirt/blockinfo.py":[{"author":{"_account_id":1779,"name":"Daniel Berrange","email":"berrange@redhat.com","username":"berrange"},"change_message_id":"abc504088ed640e08a0bd5651d218e802210589d","unresolved":false,"context_lines":[{"line_number":606,"context_line":"        if disk_bus \u003d\u003d \"scsi\" and guestarch \u003d\u003d arch.AARCH64:"},{"line_number":607,"context_line":"            last_device \u003d False"},{"line_number":608,"context_line":"        else:"},{"line_number":609,"context_line":"            last_device \u003d True"},{"line_number":610,"context_line":""},{"line_number":611,"context_line":"        config_info \u003d get_next_disk_info(mapping,"},{"line_number":612,"context_line":"                                         disk_bus,"}],"source_content_type":"text/x-python","patch_set":4,"id":"dada55a8_12f2ed06","line":609,"updated":"2016-07-22 09:26:59.000000000","message":"No, we don\u0027t want this to be architecture or bus specific. We should just never set last_device \u003d True. In fact we should just remove the last_device parameter from the get_next_disk_info method entirely, as its broken by design.","commit_id":"01a9a31817cc5f9a014915f0463c063127d1cdac"},{"author":{"_account_id":22076,"name":"Kevin Zhao","email":"kevin.zhao@linaro.org","username":"KevinZhao"},"change_message_id":"731f15067d341051966e560796d92cfb22ec68f7","unresolved":false,"context_lines":[{"line_number":606,"context_line":"        if disk_bus \u003d\u003d \"scsi\" and guestarch \u003d\u003d arch.AARCH64:"},{"line_number":607,"context_line":"            last_device \u003d False"},{"line_number":608,"context_line":"        else:"},{"line_number":609,"context_line":"            last_device \u003d True"},{"line_number":610,"context_line":""},{"line_number":611,"context_line":"        config_info \u003d get_next_disk_info(mapping,"},{"line_number":612,"context_line":"                                         disk_bus,"}],"source_content_type":"text/x-python","patch_set":4,"id":"dada55a8_081f5a5c","line":609,"in_reply_to":"dada55a8_12f2ed06","updated":"2016-07-22 09:59:05.000000000","message":"Thanks Daniel ,I will remove the arch specify code and tweak the test case","commit_id":"01a9a31817cc5f9a014915f0463c063127d1cdac"}]}