10815 2016-12-15 16:51:04 +0000 wic does not obey ROOTFS_SIZE / IMAGE_ROOTFS_EXTRA_SPACE 2017-01-17 09:55:21 +0000 1 1 7 Build System, Metadata & Runtime OE-Core Scripts and Tools unspecified x86 Multiple RESOLVED FIXED Medium normal 2.3 M2 10816 1 kergoth eduard.bartosh eduard.bartosh richard.purdie ross.burton --- --- No (bug/feature does not impact docs) oldest_to_newest 69164 0 kergoth 2016-12-15 16:51:04 +0000 I just attempted to build an image for a meta-fsl-arm machine using a wks based on the wks files in that layer, and then added dev-pkgs to the image, and the build failed. This is a case where the size determined by `du` isn't sufficient to hold the actual files in the rootfs (it doesn't scale well as size increases). Our distro bumps IMAGE_ROOTFS_EXTRA_SPACE specifically to deal with this, but the --extra-size in the .wks was much smaller than this, so the build failed. It seems that there is logic in imager/direct.py to fall back to ROOTFS_SIZE if --size= isn't specified for the rootfs partition, but it does so only when part.rootfs_dir is set. part.rootfs_dir isn't passed explicitly when building with the wic image fstype, as it relies upon use of IMAGE_ROOTFS, and the fallback from rootfs_dir to IMAGE_ROOTFS doesn't happen until part.prepare() runs the rootfs source plugin's preparation method. 69165 1 kergoth 2016-12-15 16:52:03 +0000 To sum up: wic doesn't seem to be obeying ROOTFS_SIZE/IMAGE_ROOTFS_EXTRA_SPACE when the rootfs partition command in the ks file doesn't pass --size=. 69171 2 kergoth 2016-12-15 18:02:39 +0000 Correction, the analysis of the cause was incorrect. It's when --rootfs-dir= isn't passed on the partition command of the ks file, the rootfs dir passed to wic itself isn't relevant. The logic which falls back to ROOTFS_SIZE, due to where it's done, seems to go based on rootfs_dir as that's the only way for it to know if the partition in question is a rootfs, but it'd be better to just do the fallback in prepare_rootfs(), as we know that's done for a rootfs, and ensures that the source prep method is run after part.size is set. Testing that now. 69172 3 kergoth 2016-12-15 18:09:58 +0000 Confirmed. This patch changes the behavior so it obeys the overhead factor and extra space as defined in the metadata when there's no --size= or --rootfs-dir= in the ks: diff --git i/scripts/lib/wic/partition.py w/scripts/lib/wic/partition.py index f383533..55ec4c0 100644 --- i/scripts/lib/wic/partition.py +++ w/scripts/lib/wic/partition.py @@ -28,7 +28,7 @@ import os import tempfile from wic.utils.oe.misc import msger, parse_sourceparams -from wic.utils.oe.misc import exec_cmd, exec_native_cmd +from wic.utils.oe.misc import exec_cmd, exec_native_cmd, get_bitbake_var from wic.plugin import pluginmgr partition_methods = { @@ -190,6 +190,16 @@ class Partition(): if os.path.isfile(rootfs): os.remove(rootfs) + # Get rootfs size from bitbake variable if it's not set in .ks file + if not self.size: + # Bitbake variable ROOTFS_SIZE is calculated in + # Image._get_rootfs_size method from meta/lib/oe/image.py + # using IMAGE_ROOTFS_SIZE, IMAGE_ROOTFS_ALIGNMENT, + # IMAGE_OVERHEAD_FACTOR and IMAGE_ROOTFS_EXTRA_SPACE + rsize_bb = get_bitbake_var('ROOTFS_SIZE') + if rsize_bb: + self.size = int(round(float(rsize_bb))) + for prefix in ("ext", "btrfs", "vfat", "squashfs"): if self.fstype.startswith(prefix): method = getattr(self, "prepare_rootfs_" + prefix) Log fragment from before the change (note the 8192, that's from --extra-size 8M in the ks): Debug: Requested partition size for /: 0 Debug: current_blocks: 558876 Debug: Added 8192 extra blocks to / to get to 567068 total blocks Log fragment from after the change: Debug: Requested partition size for /: 770048 Debug: current_blocks: 558876 Debug: Requested blocks 770048, current_blocks 558876 Debug: Added 211172 extra blocks to / to get to 770048 total blocks 69173 4 eduard.bartosh 2016-12-15 18:12:34 +0000 The patch looks ok to me. Can you send it for review to the mailing list? 69174 5 eduard.bartosh 2016-12-15 18:14:18 +0000 assigning to Christopher as he did all work. Please reassign back to me if anything is still required from me. 69175 6 kergoth 2016-12-15 18:16:15 +0000 One downside to that particular patch is, the overhead-factor as specified in the ks applies in addition to any defined overhead factor in the metadata in this case. Thoughts on that? 69176 7 kergoth 2016-12-15 18:17:24 +0000 Adding Ed to CC to get his thoughts on comment #6 :) 69177 8 eduard.bartosh 2016-12-15 18:21:49 +0000 I think it's enough to warn user if overhead-factor is specified in .wks file , but --size is not and we're using bitbake variables to determine rootfs size. Does it make sense? 69183 9 kergoth 2016-12-16 05:06:20 +0000 Patch submitted to oe-core. 69202 10 eduard.bartosh 2016-12-18 11:39:31 +0000 My fix for the bug 10816 uncovered test regression brought by your changes. Here is a full test log: $ oe-selftest -r wic 2016-12-17 13:50:16,979 - selftest - INFO - Running bitbake -e to get BBPATH 2016-12-17 13:50:19,120 - selftest - INFO - Checking that everything is in order before running the tests 2016-12-17 13:50:21,269 - selftest - INFO - Running bitbake -p 2016-12-17 13:50:24,173 - selftest - INFO - test runner init'ed like unittest 2016-12-17 13:50:26,174 - selftest - INFO - Loading tests from: oeqa.selftest.wic 2016-12-17 13:51:28,596 - selftest - INFO - Adding: "include selftest.inc" in local.conf 2016-12-17 13:51:28,596 - selftest - INFO - Adding: "include bblayers.inc" in bblayers.conf 2016-12-17 13:51:28 - test_bmap (oeqa.selftest.wic.Wic) Test generation of .bmap file ... ok 2016-12-17 13:51:52 - test_build_artifacts (oeqa.selftest.wic.Wic) Test wic create directdisk providing all artifacts. ... ok 2016-12-17 13:52:16 - test_build_image_name (oeqa.selftest.wic.Wic) Test wic create directdisk --image-name core-image-minimal ... ok 2016-12-17 13:52:22 - test_compress_bzip2 (oeqa.selftest.wic.Wic) Test compressing an image with bzip2 ... ok 2016-12-17 13:52:30 - test_compress_gzip (oeqa.selftest.wic.Wic) Test compressing an image with gzip ... ok 2016-12-17 13:52:37 - test_compress_xz (oeqa.selftest.wic.Wic) Test compressing an image with xz ... ok 2016-12-17 13:52:49 - test_createhelp (oeqa.selftest.wic.Wic) Test wic create --help ... ok 2016-12-17 13:52:49 - test_directdisk_bootloader_config (oeqa.selftest.wic.Wic) Test creation of directdisk-bootloader-config image ... ok 2016-12-17 13:52:55 - test_gpt_image (oeqa.selftest.wic.Wic) Test creation of core-image-minimal with gpt table and UUID boot ... ok 2016-12-17 13:53:03 - test_help (oeqa.selftest.wic.Wic) Test wic --help ... ok 2016-12-17 13:53:03 - test_help_kickstart (oeqa.selftest.wic.Wic) Test wic help kickstart ... ok 2016-12-17 13:53:03 - test_help_overview (oeqa.selftest.wic.Wic) Test wic help overview ... ok 2016-12-17 13:53:04 - test_help_plugins (oeqa.selftest.wic.Wic) Test wic help plugins ... ok 2016-12-17 13:53:06 - test_image_env (oeqa.selftest.wic.Wic) Test generation of <image>.env files. ... ok 2016-12-17 13:53:30 - test_iso_image (oeqa.selftest.wic.Wic) Test creation of hybrid iso image with legacy and EFI boot ... ok 2016-12-17 13:53:42 - test_listhelp (oeqa.selftest.wic.Wic) Test wic list --help ... ok 2016-12-17 13:53:42 - test_mkefidisk (oeqa.selftest.wic.Wic) Test creation of mkefidisk image ... ok 2016-12-17 13:53:49 - test_mkgummidisk (oeqa.selftest.wic.Wic) Test creation of mkgummidisk image ... ok 2016-12-17 13:53:55 - test_no_command (oeqa.selftest.wic.Wic) Test wic without command ... ok 2016-12-17 13:53:55 - test_qemu (oeqa.selftest.wic.Wic) Test wic-image-minimal under qemu ... Loading cache: 0% | Loading cache: 100% |#################################################################################################################################################| Time: 0:00:00 NOTE: rootfs file: /home/ed/git/yocto/poky/build/tmp/deploy/images/qemux86-64/wic-image-minimal-qemux86-64.wic NOTE: Qemu log file: /home/ed/git/yocto/poky/build/tmp/work/qemux86_64-poky-linux/wic-image-minimal/1.0-r0/testimage/qemu_boot_log.20161217115406 NOTE: SSH log file: /home/ed/git/yocto/poky/build/tmp/work/qemux86_64-poky-linux/wic-image-minimal/1.0-r0/testimage/ssh_target_log.20161217115406 Created listening socket for qemu serial console on: 127.0.0.1:60614 Created listening socket for qemu serial console on: 127.0.0.1:44022 Not using kvm for runqemu launchcmd=runqemu snapshot tcpserial=44022 qemux86-64 /home/ed/git/yocto/poky/build/tmp/deploy/images/qemux86-64/wic-image-minimal-qemux86-64.wic nographic bootparams="console=tty1 console=ttyS0,115200n8 printk.time=1" qemuparams="-serial tcp:127.0.0.1:60614" runqemu started, pid is 3379 waiting at most 60 seconds for qemu pid qemu started - qemu procces pid is 3391 Starting logging thread Starting thread event loop Connection request received Setting connection established event Output from runqemu: runqemu - INFO - MACHINE: qemux86-64 runqemu - INFO - DEPLOY_DIR_IMAGE: /home/ed/git/yocto/poky/build/tmp/deploy/images/qemux86-64 runqemu - INFO - CONFFILE: /home/ed/git/yocto/poky/build/tmp/deploy/images/qemux86-64/wic-image-minimal-qemux86-64.qemuboot.conf runqemu - INFO - Continuing with the following parameters: stty: standard input: Inappropriate ioctl for device runqemu - INFO - Running /bin/ip link... runqemu - INFO - Acquiring lockfile /tmp/qemu-tap-locks/tap0.lock... runqemu - INFO - Using preconfigured tap device tap0 runqemu - INFO - If this is not intended, touch /tmp/qemu-tap-locks/tap0.skip to make runqemu skip tap0. runqemu - WARNING - Can't detect drive type /home/ed/git/yocto/poky/build/tmp/deploy/images/qemux86-64/wic-image-minimal-qemux86-64.wic runqemu - WARNING - Tring to use virtio block drive runqemu - INFO - Running ldd /home/ed/git/yocto/poky/build/tmp/sysroots/x86_64-linux/usr/bin/qemu-system-x86_64... runqemu - INFO - Running /home/ed/git/yocto/poky/build/tmp/sysroots/x86_64-linux/usr/bin/qemu-system-x86_64 -device virtio-net-pci,netdev=net0,mac=52:54:00:12:34:02 -netdev tap,id=net0,ifname=tap0,script=no,downscript=no -nographic -serial tcp:127.0.0.1:60614 -cpu core2duo -m 256 -serial tcp:127.0.0.1:44022 -drive if=virtio,file=/home/ed/git/yocto/poky/build/tmp/deploy/images/qemux86-64/wic-image-minimal-qemux86-64.wic,format=raw -no-reboot -vga vmware -show-cursor -usb -usbdevice tablet -device virtio-rng-pci -snapshot QEMU 2.7.0 monitor - type 'help' for more information (qemu) warning: TCG doesn't support requested feature: CPUID.01H:EDX.vme [bit 1] Waiting at most 1000 seconds for login banner Connection from 127.0.0.1:44079 Reached login banner Logged as root in serial console Stopping logging thread Stop event received Tearing down logging thread Sending SIGTERM to runqemu ok 2016-12-17 13:54:27 - test_qemux86_directdisk (oeqa.selftest.wic.Wic) Test creation of qemux-86-directdisk image ... ok 2016-12-17 13:54:33 - test_rootfs_artifacts (oeqa.selftest.wic.Wic) Test usage of rootfs plugin with rootfs paths ... FAIL 2016-12-17 13:55:00 - test_rootfs_indirect_recipes (oeqa.selftest.wic.Wic) Test usage of rootfs plugin with rootfs recipes ... FAIL 2016-12-17 13:55:10 - test_sdimage_bootpart (oeqa.selftest.wic.Wic) Test creation of sdimage-bootpart image ... ok 2016-12-17 13:55:15 - test_systemd_bootdisk (oeqa.selftest.wic.Wic) Test creation of systemd-bootdisk image ... ok 2016-12-17 13:55:22 - test_unsupported_subcommand (oeqa.selftest.wic.Wic) Test unsupported subcommand ... ok 2016-12-17 13:55:23 - test_wic_image_type (oeqa.selftest.wic.Wic) Test building wic images by bitbake ... ok 2016-12-17 13:55:37 - test_wrong_compressor (oeqa.selftest.wic.Wic) Test how wic breaks if wrong compressor is provided ... ok ====================================================================== FAIL: test_rootfs_artifacts (oeqa.selftest.wic.Wic) Test usage of rootfs plugin with rootfs paths ---------------------------------------------------------------------- Traceback (most recent call last): File "/home/ed/git/yocto/poky/meta/lib/oeqa/utils/decorators.py", line 109, in wrapped_f return func(*args, **kwargs) File "/home/ed/git/yocto/poky/meta/lib/oeqa/selftest/wic.py", line 186, in test_rootfs_artifacts % bbvars).status File "/home/ed/git/yocto/poky/meta/lib/oeqa/utils/commands.py", line 114, in runCmd raise AssertionError("Command '%s' returned non-zero exit status %d:\n%s" % (command, result.status, result.output)) AssertionError: Command 'wic create directdisk-multi-rootfs -b /home/ed/git/yocto/poky/build/tmp/sysroots/qemux86-64/usr/share -k /home/ed/git/yocto/poky/build/tmp/deploy/images/qemux86-64 -n /home/ed/git/yocto/poky/build/tmp/sysroots/x86_64-linux --rootfs-dir rootfs1=/home/ed/git/yocto/poky/build/tmp/work/qemux86_64-poky-linux/core-image-minimal/1.0-r0/rootfs --rootfs-dir rootfs2=/home/ed/git/yocto/poky/build/tmp/work/qemux86_64-poky-linux/core-image-minimal/1.0-r0/rootfs' returned non-zero exit status 1: Error: _exec_cmd: bitbake -e rootfs1 returned '1' instead of 0 output: Loading cache...done. ERROR: Nothing PROVIDES 'rootfs1' Loaded 1319 entries from dependency cache. Summary: There was 1 ERROR message shown, returning a non-zero exit code. Creating image(s)... ====================================================================== FAIL: test_rootfs_indirect_recipes (oeqa.selftest.wic.Wic) Test usage of rootfs plugin with rootfs recipes ---------------------------------------------------------------------- Traceback (most recent call last): File "/home/ed/git/yocto/poky/meta/lib/oeqa/utils/decorators.py", line 109, in wrapped_f return func(*args, **kwargs) File "/home/ed/git/yocto/poky/meta/lib/oeqa/selftest/wic.py", line 170, in test_rootfs_indirect_recipes % wks).status) File "/home/ed/git/yocto/poky/meta/lib/oeqa/utils/commands.py", line 114, in runCmd raise AssertionError("Command '%s' returned non-zero exit status %d:\n%s" % (command, result.status, result.output)) AssertionError: Command 'wic create directdisk-multi-rootfs --image-name core-image-minimal --rootfs rootfs1=core-image-minimal --rootfs rootfs2=core-image-minimal' returned non-zero exit status 1: Error: _exec_cmd: bitbake -e rootfs1 returned '1' instead of 0 output: Loading cache...done. ERROR: Nothing PROVIDES 'rootfs1' Loaded 1319 entries from dependency cache. Summary: There was 1 ERROR message shown, returning a non-zero exit code. Checking basic build environment... Done. Creating image(s)... ---------------------------------------------------------------------- Ran 28 tests in 249.657s FAILED (failures=2) 2016-12-17 13:55:38,254 - selftest - INFO - Finished 2016-12-17 13:55:38,255 - selftest - INFO - Removing the include from local.conf 2016-12-17 13:55:38,256 - selftest - INFO - Removing the include from bblayers.conf Wic seems to break on this line of your code: rsize_bb = get_bitbake_var('ROOTFS_SIZE', image_name) You can reproduce it by using directdisk-multi-rootfs: wic create directdisk-multi-rootfs --image-name core-image-minimal --rootfs rootfs1=core-image-minimal --rootfs rootfs2=core-image-minimal $ wic create directdisk-multi-rootfs --image-name core-image-minimal --rootfs rootfs1=core-image-minimal --rootfs rootfs2=core-image-minimal Checking basic build environment... Done. >>> bitbake -e core-image-minimal Creating image(s)... >>> bitbake -e rootfs1 Error: _exec_cmd: bitbake -e rootfs1 returned '1' instead of 0 output: Loading cache...done. ERROR: Nothing PROVIDES 'rootfs1' Loaded 1319 entries from dependency cache. Summary: There was 1 ERROR message shown, returning a non-zero exit code. 69216 11 kergoth 2016-12-19 15:31:28 +0000 My guess would be it's partly due to the fact that we aren't passing the image name to get_bitbake_var, but of course ROOTFS_SIZE is emitted into the wicenvs by the bbclass, it isn't a normal bitbake variable, so I'd expect it to fail in cases where it's run externally without wicenvs written, unless we adjust both parts of the code using ROOTFS_SIZE to use the actual bitbake vars directly and then also pass in the image name. Thoughts? 69217 12 eduard.bartosh 2016-12-19 15:37:25 +0000 The problem here as I can see is that rootfs1/rootfs2 is used as an image name. In multi-rootfs mode it's not the case. wic is called with --rootfs rootfs1=core-image-minimal --rootfs2=core-image-minimal in this mode. This is not handled correctly in the code. 69651 13 eduard.bartosh 2017-01-10 14:30:36 +0000 I'll take it as it blocks my bug. 69656 14 kergoth 2017-01-10 15:26:20 +0000 Sorry about my inability to follow up on the regression, I've been fighting fires 99% of the time at work recently, not much time to dive into anything else. 69664 15 eduard.bartosh 2017-01-10 18:18:08 +0000 No problem. The issue should be fixed by this patch: http://lists.openembedded.org/pipermail/openembedded-core/2017-January/130950.html 69850 16 eduard.bartosh 2017-01-17 09:55:21 +0000 The patch has been merged to master.