This is needed to make the monitoring/recovery shell on tty2 work,
as the default console is tty0, which will just print to whichever
virtual console is active at the moment, making the shell unusable.
We spawn a shell:
- When Bash is first built, on tty2. This shell uses the old Bash,
so interactive mode needs to be emulated using redirection. Thus,
entering commands needs to be done using Enter followed by Ctrl+D,
and certain redirection features are unavailable.
- After moving the system to disk, on tty2. Old Bash, same limitations.
- After 2nd Bash is built, on tty3. This is a fully functional shell.
This is disabled in chroot-like bootstrap modes, or when -i is not set.
Before the simplify refactor, the convention was to name the first
or default build script "<package name>.sh". The refactor changed
this default to "pass1.sh", with "pass2.sh" and so on used for
subsequent passes. This wasn't an issue when building usig scripts
generated by script-generator (which explicitly declares the build
script to use), but broke running "build <package name>" without
an explicit script name from the post-bootstrap bash prompt.
By deriving the script name from the current revision/pass number,
this commit restores that functionality.
An unfortunate blemish is that we number build scripts from 1, but
repo tarballs from 0 - it would be nice to standardize on a single
indexing convention.
The Linux kernel won't autocreate /dev/shm inside devtmpfs.
Without it, semaphores won't work properly in Python.
Previously, /dev/shm was set up in populate_device_nodes, but we
no longer run that after booting into Linux.
The sysrq shutdown trigger takes some time to fully shut down the system,
during which init is expected to continue running. Since after.sh is the
last step in our init, if it quits before shutdown is complete, Linux will
panic with "Attempted to kill init".
Add an infinite loop after shutdown is issued via sysrq to prevent this.
This was removed as part of the simplify refactor, severely slowing
down qemu and bare-metal builds. Restoring it brings us back to the
same build times that we saw before the refactor.
Perform variable substitution at runtime, rather than at generation
time. This way, if bootstrap.cfg changes after update_env, the new
values there take effect immediately.
* Enable additional hardware drivers in Linux for better bare metal
display, network & input device compatibility
* Disable ATA-over-Ethernet support, because it spams the network
with unnecessary packets, is basically useless for bootstrapping,
and may even be a security risk
* Increase Fiwix initrd size to 1280MB to fit a larger Linux build
* Make the Fiwix kexec size configurable the same way as initrd
(and reduce from 280MB to 256MB which is sufficient in my testing)
* Use a more conservative memory map for Fiwix & Linux
* Boot Linux with consoleblank=0 on bare metal, so the build won't
go blind after 5 minutes
* Support kexec-fiwix with interrupts disabled (will be useful later
when builder-hex0 is updated)
* Use -S32 -H64 --force to trick sfdisk into allowing MiB-aligned
partitions on a drive that reports a CHS geometry
* Explicitly create partition at sector 2097152 (=1GiB)
* Force mkfs.ext4 to overwrite any existing filesystem it might find
* Wait up to 2 minutes for the disk to become readable (especially
USB drives often show up with a delay)
With this, finalize_fhs.sh can be rerun as needed, e.g. when rebooting.
Also, the preferred nameserver will persist after DHCP.
Thanks to devtmpfs, we no longer need to manage /dev once Linux is up.
These are implemented as dummy jumps with a script that just exits
with success. Since this script will be sourced, rather than called,
this causes the bootstrap process to exit at that point.
The breakpoints are conditional on INTERNAL_CI (we check for "pass1"
because that's the only pass when script-generator runs, but the
resulting effect is that each pass only bootstraps its own part of
the manifest).
Linux's KBUILD doesn't follow SOURCE_DATE_EPOCH, but rather it
uses its own variable, KBUILD_BUILD_TIMESTAMP.
While we are at it, also update kexec-linux's checksum, which
didn't match either. (This one was reproducible, it was just
out of date.)
- This idea originates from very early in the project and was, at the
time, a very easy way to categorise things.
- Now, it doesn't really make much sense - it is fairly arbitary, often
occuring when there is a change in kernel, but not from builder-hex0
to fiwix, and sysb is in reality completely unnecessary.
- In short, the sys* stuff is a bit of a mess that makes the project
more difficult to understand.
- This puts everything down into one folder and has a manifest file that
is used to generate the build scripts on the fly rather than using
coded scripts.
- This is created in the "seed" stage.
stage0-posix -- (calls) --> seed -- (generates) --> main steps
Alongside this change there are a variety of other smaller fixups to the
general structure of the live-bootstrap rootfs.
- Creating a rootfs has become much simpler and is defined as code in
go.sh. The new structure, for an about-to-be booted system, is
/
-- /steps (direct copy of steps/)
-- /distfiles (direct copy of distfiles/)
-- all files from seed/*
-- all files from seed/stage0-posix/*
- There is no longer such a thing as /usr/include/musl, this didn't
really make any sense, as musl is the final libc used. Rather, to
separate musl and mes, we have /usr/include/mes, which is much easier
to work with.
- This also makes mes easier to blow away later.
- A few things that weren't properly in packages have been changed;
checksum-transcriber, simple-patch, kexec-fiwix have all been given
fully qualified package names.
- Highly breaking change, scripts now exist in their package directory
but NOT WITH THE packagename.sh. Rather, they use pass1.sh, pass2.sh,
etc. This avoids manual definition of passes.
- Ditto with patches; default directory is patches, but then any patch
series specific to a pass are named patches-passX.
fosslinux
Gábor Stefanik
Eduardo Sánchez Muñoz
Andrius Štikonas
rick-masters
rick-masters