only use config/submodules/ which the build system then
uses to run git clones manually, replicating the submodules
feature. we must never use a project's own gitmodules feature,
because we can't easily control it. better to let it break first,
and then figure out what modules to add manually, so that we
have only what we need for each project.
it's done this way, because git's own submodules feature
doesn't have very good error checking in general, nor
does it have good redundancy.
with the current design, we can declare backup repositories
for each submodule.
we replicate it precisely. for example:
3rdparty/vboot
this is a coreboot submodule, and we handle that in the
coreboot trees.
however, our current design also allows you to do this even
if the upstream repository does not contain a .gitmodules file
Signed-off-by: Leah Rowe <leah@libreboot.org>
now it no longer hardcodes a check for whether the
project name is coreboot. this maintains the same
behaviour but will now work for other multi-tree
projects; in practise, the other multi-tree projects
did not use .gitmodules files anyway, but some of
them used config/submodules/ in our build system.
Signed-off-by: Leah Rowe <leah@libreboot.org>
Re-add xHCI only on haswell and broadwell machines, where
they are needed. Otherwise, keep the same GRUB code.
The xHCI patches were removed because they caused issues
on Sandybridge-based Dell Latitude laptops. See:
https://codeberg.org/libreboot/lbmk/issues/216
The issue was not reported elsewhere, including on the
Haswell/Broadwell hardware where they are needed, but the
build system could only build one version of GRUB.
The older machines do not need xHCI patches, because they
either do not have xHCI patches, or work (in GRUB) because
they're in EHCI mode when running the payload.
So, the problem is that we need the xHCI patches for GRUB
on Haswell/Broadwell hardware, but the patches break
Sandybridge hardware, and we only had the one build of GRUB.
To mitigate this problem, the build system now supports
building multiple revisions of GRUB, with different patches,
and each given coreboot target can say which GRUB tree to use
by setting this in target.cfg:
grubtree="xhci"
In the above example, the "xhci" tree would be used. Some
generic GRUB config has been moved to config/data/grub/
and config/grub/ now looks like config/coreboot/ - also,
the grub.cfg file (named "payload" in each tree) is copied
to the GRUB source tree as ".config", then added to GRUB's
memdisk in the same way, as grub.cfg.
Several other design changes had to be made because of this:
* grub.cfg in memdisk no longer automatically jumps to one
in CBFS, but now shows a menuentry for it if available
* Certain commands in script/trees are disabled for GRUB,
such as *config make commands.
* gnulib is now defined in config/submodule/grub/, instead
of config/git/grub - and this mitigates an existing bug
where downloading gnulib first would make grub no longer
possible to download in lbmk.
The coreboot option CONFIG_FINALIZE_USB_ROUTE_XHCI has been
re-enabled on: Dell OptiPlex 9020 MT, Dell OptiPlex 9020 SFF,
Lenovo ThinkPad T440p and Lenovo ThinkPad W541 - now USB should
work again in GRUB.
The GRUB payload has been re-enabled on HP EliteBook 820 G2.
This change will enable per-board GRUB optimisation in the
future. For example, we hardcode what partitions and LVMs
GRUB scans because * is slow on ICH7-based machines, due
to GRUB's design. On other machines, * is reasonably fast,
for automatically enumerating the list of devices for boot.
Use of * (and other wildcards) could enable our GRUB payload
to automatically boot more distros, with minimal fuss. This
can be done at a later date, in subsequent revisions.
Signed-off-by: Leah Rowe <leah@libreboot.org>
the reason for it is because sometimes the coreboot build
system auto-downloads submodules which we don't want.
however, we now pass UPDATED_SUBMODULES=1 in make, which
disables this behaviour in coreboot's build system.
therefore, remove this unnecessary logic.
Signed-off-by: Leah Rowe <leah@libreboot.org>
cbmk revision:
cdce8ba70b863ea3fe0ad7a4d7b27d0c5ca30421
as of date 30 May 2024
Canoeboot provides deblobbing, fully, on all sources, so
as to provide a GNU FSDG compliant coreboot distro.
Libreboot used to do this but now uses a more pragmatic
Binary Blob Reduction Policy, allowing better hardware
support in general. See:
https://libreboot.org/news/policy.html
Well! We sometimes still need to delete files in Libreboot,
but for other reasons. For example, the poorly licensed
strlcat.c file that we delete from U-Boot, in both projects.
I currently hardcode such deletions in lbmk. After this
revision, I will start using "nuke.list" files as in cbmk.
Simply patching the sources to exclude such files, in this
context, is not OK because then we are still including them
but as diffs. This is why the nuke() function exists.
Import Canoeboot's nuke technology.
Signed-off-by: Leah Rowe <leah@libreboot.org>
in each submodule configuration directory, a module.cfg
file can now be provided. in it, the user can specify
two repository links (main and backup) and a revision, like
so:
subrepo="repo link goes here"
subrepo_bkup="backup repo link goes here"
subrev="git revision id goes here"
additionally:
in the *main* project directory for the submodules,
a module.list file can be provided. example entries:
3rdparty/vboot
3rdparty/libgfxinit
if the module.list file is provided, only those submodules
will be downloaded. this can be combined with the module.cfg
files, if you wish, but it's optional. you can mix and match.
example locations:
multi-tree project:
config/submodule/coreboot/default/module.list
config/submodule/coreboot/default/vboot/module.cfg
single-tree project:
config/submodule/flashprog/module.list
config/submodule/flashprog/foo/module.cfg
*no* configuration files have been provided, in this commit,
which means that the current behaviour is maintained.
follow-up commits will absolutely configure the submodules.
this is being done to reduce the number of modules downloaded,
because we don't use most of the coreboot submodules that are
downloaded, thus wasting bandwidth and the releases are also
much bigger than necessary.
Signed-off-by: Leah Rowe <leah@libreboot.org>
again, the directory in question is simply used
in a for loop using asterisk (git_am_patches) and
the for loop simply won't iterate if either the
directory doesn't exist or it contains no items.
Signed-off-by: Leah Rowe <leah@libreboot.org>
in the function that immediately follows, it
starts two for loops that check every item in
that directory, using the asterisk wildcard.
if the directory does not exist, then the for
loop will simply break on first pass.
Signed-off-by: Leah Rowe <leah@libreboot.org>
"./update project trees" is a leftover from the
old build system design, prior to audits.
this particular call is for when xtree is defined,
which means that a given tree must rely on the given
coreboot tree defined by xtree. the "xtree" tree is
downloaded, so that its crossgcc builds can be re-used
to save time when building targets across many trees.
this is because trees often use identical crossgcc builds.
Signed-off-by: Leah Rowe <leah@libreboot.org>
for single-tree project (e.g. flashprog):
config/submodule/PROJECT/MODNAME/patches
for multi-tree project (e.g. coreboot):
config/submodule/PROJECT/TREE/MODNAME/patches
MODNAME is e.g.:
3rdparty/vboot directory in coreboot: would become vboot
(the submodule codepath is filtered to up to the final slash)
another example:
submodire src dir 3rdparty/foo/bar
MODNAME would be "bar"
Add whatever patches you like to a given submodule.
An example patch is included in this commit.
Signed-off-by: Leah Rowe <leah@libreboot.org>
the function is very small and only called once,
from fetch_project_trees()
merge it into fetch_project_trees()
Signed-off-by: Leah Rowe <leah@libreboot.org>
The build system already deletes .git in all source
directories for each given release, but does so at
the very end; it still does, but now it is deleted
one by one per project, to save space during very
large builds (release sizes vary wildly, depending
on how many trees exist for coreboot basically).
If you're building entirely in tmpfs (as I do), this
could be a problem if you have lots of .git/ directories.
This change reduces disk usage, or in the above example,
memory usage when running the build system from tmpfs.
This complements another recent change, where ROM images
are compressed per target during release builds, rather
than all at the very end of the process. It is part of a
series of optimisations, to reduce the memory and disk
usage of the build system, and to reduce I/O wastage
in general.
This change will not be the last of such changes!
Signed-off-by: Leah Rowe <leah@libreboot.org>
in shell scripts, a function named the same as a program included in
the $PATH will override that program. for example, you could make a
function called ls() and this would override the standand "ls".
in lbmk, a part of it was first trying to run the "fail" command,
deferring to "err", because some scripts call fail() which does
some minor cleanup before calling err.
in most cases, fail() is not defined, and it's possible that the user
could have a program called "fail" in their $PATH, the behaviour of
which we could not determine, and it could have disastrous effects.
lbmk error handling has been re-engineered in such a way that the
err function is defined in a variable, which defaults to err_ which
calls err_, so defined under include/err.sh.
in functions that require cleanup prior to error handling, a fail()
function is still defined, and err is overridden, thus:
err="fail"
this change has made xx_() obsolete, so now only x_ is used. the x_
function is a wrapper that can be used to run a command and exit with
non-zero status (from lbmk) if the command fails. the xx_ command
did the same thing, but called fail() which would have called err();
now everything is $err
example:
rm -f "$filename" || err "could not delete file"
this would now be:
rm -f "$filename" || $err "could not delete file"
overriding of err= must be done *after* including err.sh. for
example:
err="fail"
. "include/err.sh"
^ this is wrong. instead, one must do:
. "include/err.sh"
err="fail"
this is because err is set as a global variable under err.sh
the new error handling is much cleaner, and safer. it also reduces
the chance of mistakes such as: calling err when you meant to
call fail. this is because the standard way is now to call $err,
so you set err="fail" at the top of the script and all is well.
Signed-off-by: Leah Rowe <leah@libreboot.org>
a tree can specify:
tree_depend="treename"
this will make the other tree be downloaded. this is
used for coreboot trees, to ensure that dependency
trees are downloaded, because trees can now re-use
crossgcc from other trees.
Signed-off-by: Leah Rowe <leah@libreboot.org>
u-boot doesn't use submodules, so there's no point in
checking for it. now we can do with just one call to
the git submodule command, for simplicity
also, general code cleanup in this file (minor code
formatting improvements)
Signed-off-by: Leah Rowe <leah@libreboot.org>
the directory is checked for deletion, but it's already
checked before download, to see whether it already exists.
lbmk already exits with zero status if the directory exists,
so the check is pointless (in this function)
also, general code style/formatting cleanup
Signed-off-by: Leah Rowe <leah@libreboot.org>
the patchfail variable was only needed in the old design,
where git am was being handled inside a subshell, and
also when we did it directly in the target directory
without using a temporary directory. with the current
design, we can just call err() and ditch the tmp repo
Signed-off-by: Leah Rowe <leah@libreboot.org>
there isn't really a problem right now, but a desired
and implemented behavioural change was that patches are
to be applied *before* updating submodules. well, the
previous commit reversed this change, under certain
conditions, such that submodules were applied first.
this patch fixes it, so that patches are done first.
Signed-off-by: Leah Rowe <leah@libreboot.org>
right now, if we want to patch a project such that certain
submodules are no tdownloaded, or diffreent submodules are
downloaded, or current ones are downloaded from other
locations, we cannot do this, because we apply submodule
updates *before* applying patches.
therefore, we should change it so that they are applied
*after* installing patches.
Signed-off-by: Leah Rowe <leah@libreboot.org>
while seemingly pedantic, this does actually make code
easier to read. mostly just switching to shorthand for
variable names, where no expansions or patterns are used
Signed-off-by: Leah Rowe <leah@libreboot.org>
lbmk didn't quote certain arguments in commands, or
used ! -z instead of -n, things like that. simple fixes.
Signed-off-by: Leah Rowe <leah@libreboot.org>
most of these are probably redundant, and will never
be called, but lbmk needs to be as safe as possible
under fault conditions. fail early, fail hard.
Signed-off-by: Leah Rowe <leah@libreboot.org>
update/trees wasn't correctly returning non-zero status,
even though it was printing an error message, when git-am
failed. this is due to the way subshells work, and it was
overlooked in previous auditing.
additionally: don't directly copy trees to the destination,
instead patch/reset first, then copy only under normal
condition, just as with single-tree projects.
when running build/roms, the script would continue after
a bad git-am, without exit. this patch fixes it in the
most paranoid way possible. i'm now fairly confident that
lbmk will fail gracefully and efficiently, under error
conditions. this should prevent bad image builds.
Signed-off-by: Leah Rowe <leah@libreboot.org>
when [] is used right at the end of a function, or
certain loops/subshells, some sh implementations will
just return a non-zero exit
Signed-off-by: Leah Rowe <leah@libreboot.org>
only fetch if .gitmodules exists
in some cases, lbmk is compiling source trees that
use submodules, without having downloaded them first.
in all cases, those submodules are either optional,
or the build system auto-fetches them (or if it can,
we sometimes disable it as with grub and gnulib).
this is a nice fallback behaviour, for situations where
we forget to put submodules as dependencies under
config/git (and disable submodules in the given project).
with this change, release archives are guaranteed to
be complete, sans crossgcc downloads in coreboot; this
will be handled in a follow-up commit.
Signed-off-by: Leah Rowe <leah@libreboot.org>
in some cases, use of x_ or xx_ can be error-prone,
due to the way $@ is handled; commands requiring
quotes, or with funny file names as arguments such
as spaces in the file name, or other special
characters, can make the x/xx functions break.
in those cases, where x/xx must not be used, the
commands use || err instead
in other cases, use of x/xx is superfluous, and has
been removed in some commands.
Signed-off-by: Leah Rowe <leah@libreboot.org>
as opposed to the current 3-level structure.
recent build system simplifications have enabled
this change, thus:
./build fw coreboot -> ./build roms
./build fw grub -> ./build grub
./build fw serprog -> ./build serprog
./update project release -> ./update release
./update project trees -> ./update trees
./update vendor download -> ./vendor download
./update vendor inject -> ./vendor inject
alper criticised that the commands were too long,
so i made them shorter!
Signed-off-by: Leah Rowe <leah@libreboot.org>
Just one script.
Just one!
Well, two, but the 2nd one already existed:
logic in update/project/trees and
update/project/repo was merged into
include/git.sh and update/project/build
was renamed to update/project/trees; an -f
option was added, which calls the functions
under git.sh
so git clones are now handled by the main build
script (for handling makefiles and defconfigs)
but the logic there is a stub, where git.sh
does all the actual heavy lifting
this cuts the file count down by two, and reduces
sloccount a reasonable amount because much of
the logic already exists in the build script, when
it comes to handling targets. git.sh was adjusted
to integrate with this, rather than act standalone
Signed-off-by: Leah Rowe <leah@libreboot.org>
do not update them in project/repos - despite what
the previous commit message says, this behaviour is
error prone and should be avoided.
Signed-off-by: Leah Rowe <leah@libreboot.org>
With this change, lbmk now also updates submodules on
simple git clones, not just multi-tree clones.
This is OK, because git does not return non-zero status
when git submodule update is ran, where git submodules
are not actually defined.
Signed-off-by: Leah Rowe <leah@libreboot.org>
This is done recursively, with the following rule:
files first, then directories.
Where all patch files are applied from within the
patch directory, subdirectories (within the patch
directory) are then tried in alphanumerical order.
Then, within each subdirectory tried, the same rule
is once again applied. This is done recursively,
until every patch file is applied.
The code no longer applies *.patch, but instead any
file. Additionally, symlinks are avoided.
Signed-off-by: Leah Rowe <leah@libreboot.org>
Leah Rowe