Various improvements #6
Conversation
glebm
force-pushed
the
no-root
branch
5 times, most recently
from
cd06ad8 to
8d01931
Compare
There was a problem hiding this comment.
I like the idea behind this PR: being able to build without requiring commands as root on the host.
Most of the comments are about changes that probably shouldn't be in this commit.
The only fundamental concern I have is about file ownership in the created file system: there doesn't seem to be any mechanism yet that ensures the pre-installed applications are going to be owned by the user on the device.
create_data_image.sh
Outdated
| echo "Creating data partition of ${IMAGE_SIZE} MB..." | ||
| mkdir -p images | ||
| dd if=/dev/zero of=images/data.bin bs=1M count=${IMAGE_SIZE} | ||
| MKE2FS_CONFIG=mke2fs.conf /sbin/mkfs.ext4 -b4096 -I128 -m3 \ | ||
| -E lazy_itable_init=0,lazy_journal_init=0,resize=268435456 \ | ||
| -E lazy_itable_init=0,lazy_journal_init=0,resize=268435456,root_owner=0:0 \ |
There was a problem hiding this comment.
What will the owner inside the created file system be for files other than the root directory? Some files should be owned by root, while for example the pre-installed applications should be owned by the user, since the user must be allowed to remove or upgrade them.
There was a problem hiding this comment.
On my RG350, the owner of the appears to be default (uid 1000).
There was a problem hiding this comment.
I guess that's because the user IDs on your PC are the same as on OpenDingux: 1000:1000. If you run the script as a different user it probably would not work. An easy fix is to chown 1000:1000 though, as this is guaranteed to be the "od" user's IDs.
There was a problem hiding this comment.
Can you please push the fix, since you have more understanding of what's going on here?
Note that this is now configured via mke2fs.conf.
|
Re-based on top of #8
It does work (I don't know exactly why). The files are owned by user |
I wouldn't call 30 seconds on a 16 GiB card small! |
|
I'll squash the unsigned "Review comments"-type commits once all the concerns have been addressed |
glebm
changed the title
That's a 15% difference, it's not that much. My point was more that this is executed only once after flashing, so even if the difference is x2, it still wouldn't matter much, since users would almost never experience it. My gripe with the offline resizing code is the mount_data_partition() function, which does not follow what S03mountdata.sh is doing.
|
I'd say the initial boot setup experience also matters, especially since it's not difficult for us to implement this |
glebm
force-pushed
the
no-root
branch
3 times, most recently
from
44f94bb to
d68dac0
Compare
Changed to simply call |
This results in a faster resize as resize2fs can perform an "offline" resize instead of an "online" one. Timings for me on RG350 with a stock 16G SD card are: online: 3m45s offline: 3m10s Signed-off-by: Gleb Mazovetskiy <[email protected]>
Signed-off-by: Gleb Mazovetskiy <[email protected]>
Signed-off-by: Gleb Mazovetskiy <[email protected]>
Unused because root is no longer required. Signed-off-by: Gleb Mazovetskiy <[email protected]>
Various improvements: