linux/meta-updater.git/classes, branch feat/Return-python2-dockerfile

Merge pull request #810 from liuming50/drop-prepare_ostree_rootfs-prefunc

2021-07-01T05:26:12+00:00

image_types_ostree.bbclass: drop prepare_ostree_rootfs prefunc

meta: support readonly sysroot

2021-06-30T11:53:39+00:00

Currently ostree sysroot is being mounted as RW by ostree-remount systemd service, add a variable OSTREE_SYSROOT_READONLY to control that, so the end users could choose to mount sysroot as RO. Signed-off-by: Ming Liu

image_types_ostree.bbclass: drop prepare_ostree_rootfs prefunc

2021-06-19T12:59:43+00:00

The purpose of prepare_ostree_rootfs prefunc is to ensure do_image_rootfs runs from a clean ${OSTREE_ROOTFS} and call oe.path.copyhardlinktree from ${IMAGE_ROOTFS} to ${OSTREE_ROOTFS}. We dont have to maintain a prefunc to achieve that, it could be easily done in do_image_ostree itself. But the major reason for this change is to avoid hard links from ${IMAGE_ROOTFS} to ${OSTREE_ROOTFS}, which might lead to a Pseudo abortion in some cases, which was observed when IMA/EVM is enabled in rootfs. There was a commit aiming to fix the Pseudo abortion, commit db099053: [ image_types_ostree.bbclass: add fakeroot varflag to prepare_ostree_rootfs ] but it's incomplete per later tests, there is still the case some old inodes being used in ${IMAGE_ROOTFS} when do_rootfs runs, which causes 'rm -rf ${IMAGE_ROOTFS}' fail on a Pseudo abortion error. This fixes the Pseudo abortion completely. Signed-off-by: Ming Liu

image_types_ostree.bbclass: add fakeroot varflag to prepare_ostree_rootfs

2021-04-08T18:51:28+00:00

Since the ostree rootfs is generated by fakeroot, we should make prepare_ostree_rootfs run with fakeroot as well, or else we might run into a pseudo abort issue when removing OSTREE_ROOTFS directory. Reference: https://wiki.yoctoproject.org/wiki/Pseudo_Abort Signed-off-by: Ming Liu

Merge pull request #796 from ricardosalveti/master

2021-01-29T15:09:03+00:00

image_types_ota.bbclass: use standard ext4 features

Merge pull request #797 from liuming50/fix-race-problems-of-ostreecommit-ota-tasks

2021-01-29T15:08:11+00:00

image_types_ota.bbclass: get lock before accessing OSTREE_REPO

image_types_ostree.bbclass: get lock before accessing OSTREE_REPO

2021-01-28T17:50:40+00:00

In commit d13ec585ae677affd88b9d92c6ea135cc249b2fa: [ image_types_ostree: use OSTree repo location for lockfiles ] A lockfile ${OSTREE_REPO}/ostree.lock was added to prevent concurrent access to the repository by multiple bitbake instances, but it should not only apply to ostreecommit task, but ostreepush task also need it. This fixes a following race condition error: | | ERROR: Execution of '.../temp/run.do_image_ostreecommit.34262' failed with exit code 1: | error: Writing content object: Creating temp file: No such file or directory Signed-off-by: Ming Liu

image_repo_manifest.bbclass: Improvement for builds outside the .repo directory

2021-01-19T16:54:55+00:00

The repo tool searches up the directory tree to find the .repo directory, which doesn't work in a separated build directory. Signed-off-by: Dominik Wild

image_types_ota.bbclass: use standard ext4 features

2021-01-12T01:37:28+00:00

Avoid removing 64bit and metadata_csum by default and prefer the standard ext4 features instead. 64bit enables the file system to be larger than 2^32 blocks and metadata_csum enables metadata checksumming, both of which are enabled by default on recent mke2fs releases. It is unclear why 64bit was disabled by looking at the git history, but my assumption would be that either kernel or userspace was old enough for this feature to not be supported/available. Since this option is currently used by most distros by default, it should now be safe to enable it (requires kernel >= 2.6.28). metadata_csum was disabled in 4d34fa53db to make the u-boot tooling work with the ext4 file system (when saving environment), but the correct fix should instead be a fix at the userspace tooling instead, since not every target requires u-boot. Signed-off-by: Ricardo Salveti

image_types_ota.bbclass: pass -t ext4 to mke2fs

2021-01-03T07:39:15+00:00

* it doesn't automatically pass "-t ext4", because here we're calling mke2fs as mkfs.ota-ext4 instead of mkfs.ext4 so the resulting was now ext2 instead of ext4 $ file image-raspberrypi4.rootfs.ota-ext4 image-raspberrypi4.rootfs.ota-ext4: Linux rev 1.0 ext2 filesystem data, UUID=15862fcd-6204-486b-a4c6-fdf7b1710797, volume name "otaroot" (large files) after this fix: $ file image-raspberrypi4.rootfs.ota-ext4 image-raspberrypi4.rootfs.ota-ext4: Linux rev 1.0 ext4 filesystem data, UUID=3709e65c-1d2b-4280-85d1-638f5f9c3cc0, volume name "otaroot" (extents) (64bit) (large files) (huge files) that's why metadata_csum and other features enabled by default in ext4 were now missing after https://github.com/advancedtelematic/meta-updater/pull/787 we can drop 64bit feature as well, it's also enabled in default recipe-sysroot-native/etc/mke2fs.conf [defaults] base_features = sparse_super,large_file,filetype,resize_inode,dir_index,ext_attr default_mntopts = acl,user_xattr enable_periodic_fsck = 0 blocksize = 4096 inode_size = 256 inode_ratio = 16384 [fs_types] ext3 = { features = has_journal } ext4 = { features = has_journal,extent,huge_file,flex_bg,metadata_csum,64bit,dir_nlink,extra_isize inode_size = 256 } ... Signed-off-by: Martin Jansa Signed-off-by: Martin Jansa