~curtin-dev/curtin/trunk : changes

~curtin-dev/curtin/trunk » Changes from revision 554

From Revision 228 to 209

Rev	Summary	Authors	Date
228	add tox for easier testing add tox for easier testing	Scott Moser	8 years ago
227	fix bad usage of get_carryover_params which meant no params ... fix bad usage of get_carryover_params which meant no params carried over	Scott Moser	8 years ago
226	kernel command line: prefer the --- over -- to indicate copy... kernel command line: prefer the --- over -- to indicate copy to install In kernel version 3.15, the linux kernel began ignoring command line options that come after a '--' on the kernel command line. That conflicted with curtin's use of this argument, which meant "copy the following arguments over to the installed system". Now both d-i and curtin recognize the '---' as the preferred token to indicate install arguments versus those to be copied over to installed system.	Scott Moser	8 years ago
225	partition: when partitioning mbr and boot partition expected... partition: when partitioning mbr and boot partition expected 2 partitions In the case where we partitioned with a boot partition, we were not asserting existance of both partitions 1 and 2. Additionally, we were not wiping filesystem header from the second.	Scott Moser	8 years ago
224	partition: when creating mbr table, leave space for secondar... partition: when creating mbr table, leave space for secondary gpt header Even though we're using an mbr table, it makes sense to leave the very small amount of space at the end so that the user could subsequently convert the table to gpt, and not have to resize a filesystem in the last partition on the disk. This can easily be done with: sgdisk --mbrtogpt <device>	Scott Moser	8 years ago
223	tools/maas2roottar: exit failure if no mount-image-callback ... tools/maas2roottar: exit failure if no mount-image-callback available if the user does not have mount-image-callback, they'll get a reasonable error message showing that. Also cleans up some other exit cases.	Scott Moser	8 years ago
222	fix uefi install path on nvme devices fix uefi install path on nvme devices This fixes the install to a NVME device (named like 'nvme0n1') for UEFI. The issue was that the 'pt_uefi' method in helpers/common was not using the 'find_partno' but was expecting the partition to be named 'nvme0n115'	Scott Moser	8 years ago
221	detect_multipath: force rescan of block devices prior to blk... detect_multipath: force rescan of block devices prior to blkid Given /dev/sda and /dev/sdi point to the same block device, and: wipe-disk-and-partitions /dev/sda partition /dev/sda blockdev --rereadpt /dev/sda udevadm settle mkfs /dev/sda2 'blkid -c /dev/null' will only see /dev/sdi2 and read its filesystem UUID if the device had the same partition table before the wipe operation. Nothing would have caused the kernel to update its knowledge of partitions on /dev/sdi. blkid only reads data from devices (and partitions) the kernel provides data about. It does not open block devices and read partition table information itself. The change here is to re-scan all block devices (those unmounted and readwrite) before running blkid. The other change is to grow the list of cache files that are known to be read by blkid. Experimentation shows that 14.04 reads /dev/.blkid.tab even though that is not documented. The reason we're removing cache files rather than using 'blkid -c /dev/null' is that this command would have no affect on the cache files. If we're going to scan, we might as well cache.	Scott Moser	8 years ago
220	merge more multipath work / boot by root=/dev/mappings/mpath... merge more multipath work / boot by root=/dev/mappings/mpathX-partY This contains a couple different changes/fixes. The biggest one is change in how multipath boot is set up. Instead of configuring all the multipath devices with mpath names, we only do so for the disk with the root partition. Then, we boot with root=/dev/mappings/mpath0[-partX] rather than booting with UUID, which is unreliable (LP: #1429327). * support wiping full partition or disk in helpers/common * wipe full PReP partition to avoid grub complaining about it not being empty. * when we run 'multipath -F' to un-setup multipath devices we need to accept an exit code of '1' which occurs when not some multipath devices are busy (such as if they're mounted).	Scott Moser	8 years ago
219	curthooks: know kernel mapping for wily (4.2) curthooks: know kernel mapping for wily (4.2) The KERNEL_MAPPING needs to be kept up to date for lts enablement kernel versions. This just adds wily as 4.2. This is only relevant for mapping kernel major.minor to -lts-<releasename>.	Scott Moser	8 years ago
218	multipath: minor bug fix and comment clarifications multipath: minor bug fix and comment clarifications	Scott Moser	8 years ago
217	support UEFI-based ARM64 systems support UEFI-based ARM64 systems This adds logic to support installation of ARM64 systems that use UEFI.	Dann Frazier	8 years ago
216	helpers/common:install-grub fix usage of lsblk helpers/common:install-grub fix usage of lsblk This should have been fixed back with bug 1386275. install-grub used lsblk with the shorter '--out=' which is now ambiguous.	Scott Moser	8 years ago
215	multipath: Generate /etc/multipath/bindings file during inst... multipath: Generate /etc/multipath/bindings file during installation. File /etc/multipath/bindings assigns permanent names to the multipath devices. This file needs to be created during installation to avoid spontaneous device name shuffling during boot. We can't use multipath ('multipath -r') utility to generate bindings because it needs to have dm-multipath kernel module inserted. That module might not be available or loaded in the install environment. So we manually generate this file. Currently this only inserts entries into the file for disks that show up in output of 'blkid'. Those will be any disk that has a filesystem on it (root disk or as a partition).	Oleg Strikov	8 years ago
214	Detect multipath by using UUIDs not WWNs. Detect multipath by using UUIDs not WWNs. World Wide Names of the disks may not be unique which leads to dangerous false positives. Multipath detection code needs to look for multiple partitions with the same UUID instead. This approach seems to be more reliable.	Scott Moser	8 years ago
213	subp, apt_update: cleanups subp, apt_update: cleanups subp: support 'retries' as any iterable rather than requiring list log debug message on non-final command failure deps/install: do not retry the apt-get install command we were retrying apt-get install. apt_update: change function signature. retries=None rather than having the list in the method signature. Also change the backoff a little.	Scott Moser	8 years ago
212	Retry apt-get update and apt-get install command a maximum o... Retry apt-get update and apt-get install command a maximum of 3 times to make sure no transient issues prevent the deployment from succeeding.	Blake Rouse	8 years ago
211	Detect and handle multipath devices. Detect and handle multipath devices. Detect multipath by looking for multiple devices with the same scsi name (derived from manufacturer, model, and serial number). Install and configure multipath-tools-boot package if such devices exist. curtin will install to the non-multipath device, but boot will correctly utilize the multipath device.	Scott Moser	8 years ago
210	ChrootableTarget: if unmounting /dev, run udevadm settle fir... ChrootableTarget: if unmounting /dev, run udevadm settle first. When we unmount /dev, run udevadm settle (outside the chroot) first. This is more of a catch-all for anything that occurred inside the target. Alternatively we could have expected users of this to ensure that what they ran would wait appropriately.	Scott Moser	8 years ago
209	doc fix doc fix	Scott Moser	9 years ago