Due to GKI, the kernel UTS release string will not always (if ever)
match the vendor's UTS release string that is used to create the
initramfs file structure -- /lib/modules/<vendor uname>. This causes
module load failures when `-d DIR` is omitted. To fix this, we can
include all of the versions under /lib/modules that match the kernel's
major and minor version instead of directly using the value of uname().
In addition, we can also support modules being loaded directly from
/lib/modules.
Test: verify GKI kernel + initramfs with different UTS strings
Test: verify GKI kernel + initramfs with modules directly in /lib/modules
Fixes: 8320778425 ("toolbox/modprobe: Fallback to /lib/modules/<uname> ")
Bug: 282917063
Bug: 254835242
Merged-In: I5368f5cff139ba3165323a6a91066be38bfa0736
Change-Id: I5368f5cff139ba3165323a6a91066be38bfa0736
Make the module directory optional by reading the kernel release
version. This path is where the kernel installs modules by default.
Similar behaviour can be found in several modprobe implementations.
Bug: 254835242
Change-Id: I61707636705e5b4d9bd8ccf6351e7057eae6bcf5
Remove the function EnableBlocklist() and add a constructor argument to
enable/disable the use of modules.blocklist. In all cases, the
enabling/disabling of the blocklist happens immediately after creating
the Modprobe object. So this simplies libmodprobe.
Additionally, the use of the blocklist by libmodprobe should be enabled
by default unless explicitly disabled during creation of the Modprobe
object. Currently, only modprobe(8) defaults to not using the blocklist
and includes the argument -b BLOCKLIST for enabling it. That
functionality remains.
This refactor allows us to use the blocklist during first stage init.
However, additional logic is needed to not return an error for the
blocked non-aliased modules during first stage init; otherwise, the
error would result in an init crash leading to a device reboot. So fixup
LoadListedModules() to allow blocking modules without returning an
error.
Bug: 182582036
Test: boot test on pixel 5 with a module in modules.blocklist
Change-Id: I394b5aa98fa98821011982cfe693749010c381f7
None of them are necessary, and it's more intention-revealing to say
`c++2a` or whatever anyway.
Test: treehugger
Change-Id: Ie1df26499d160d6fc757d17fcb0121997bda14f9
There shouldn't be much activity in here, but getprop/setprop and
modprobe probably aren't going anywhere...
Test: treehugger
Change-Id: If5726dcb62a47549e10767c7a984614bcff8ea1f
There is a desire to ensure that modprobe as a service can log to
kmesg to help triage issues, so add support for the -s or --syslog
flag to do so.
SideEffects:
- help goes to stdout instead of stderr.
- verbose flag once, sets DEBUG, twice, sets VERBOSE minimum.
- quiet flag sets WARNING minimum.
Bug: 159424228
Bug: 151950334
Test: use modprobe as a service to load modules, check logs
Change-Id: I884995f364b0fc604861797eb90d7225a372f864
Add a means to specify the modules to load from a file, rather than
on the argument list by adding an optional argument to the --all flag.
This allows modprobe to be designated as a standalone service to load
a long series of modules in the background and be specified
separately. The specified (module.load) file contains a newline
separate list of module names, and supports line comments using '#'
since this file may be maintained by a human or scripting that
requires tagging for regions of the file.
Bug: 159424228
Bug: 151950334
Test: use modprobe as a service to load modules
Change-Id: Id32641c7244e65848fca3a4a82c8d08b2042bf2f
Add support for long options, and fit existing options to upstream
behaviors and extensions. Fix some missing std::endl and
android::base::Join() usage.
Bug: 159424228
Bug: 151950334
Test: manually test long options work
Change-Id: Id792d87d4407628e706aeccecb6e2bce22bcad10
Some system services (e.g. SoundTriggerMiddleware) assume that
whenever they start, the audio HAL is in its default (reset) state.
init.rc scripts tie the lifetimes of the audio HAL and system
processes, but when using stop/start this is not the case, and this
may cause spurious crashes in this case.
stop/start is apparently being relied on in some test infrastructure,
so this change is needed to avoid causing problems for those systems
or false detection of problems, which are not actually present in
production.
Bug: 154029444
Test: Manually verify that the audio HAL is stopped on
'adb shell stop', restarted on 'adb shell start' and that the
system boots correctly after.
Change-Id: I39878b978b47a169c4fe446c43d7347809d15e06
Merged-In: I39878b978b47a169c4fe446c43d7347809d15e06
Some system services (e.g. SoundTriggerMiddleware) assume that
whenever they start, the audio HAL is in its default (reset) state.
init.rc scripts tie the lifetimes of the audio HAL and system
processes, but when using stop/start this is not the case, and this
may cause spurious crashes in this case.
stop/start is apparently being relied on in some test infrastructure,
so this change is needed to avoid causing problems for those systems
or false detection of problems, which are not actually present in
production.
Bug: 154029444
Test: Manually verify that the audio HAL is stopped on
'adb shell stop', restarted on 'adb shell start' and that the
system boots correctly after.
Change-Id: I39878b978b47a169c4fe446c43d7347809d15e06
iorapd depends on pm in system_server, so turn it off too when we
turn off system_server with the 'stop' command.
Bug: 153964298
Test: adb shell stop
Test: adb shell start
Change-Id: I08390472b1e1b969d0380843efa1bbd35b194250
Move these from toybox as they're Android specific and more easily
maintained here.
A few changes during the conversion:
* Report error in setprop if value is not a UTF8 string; this would
previously fail only after attempting to set the property
* Inform users to check dmesg for further failure messages for all
three programs
* Only start/stop zygote_secondary if the device is not single arch
Bug: 117321744
Bug: 133901248
Test: setprop works, start and stop work on single and multi-arch.
Change-Id: Id2194cf2b65221bde38aff91f0e86b33edb37f42
Like the regular fgetln(), grep_fgetln() doesn't NUL-terminate the
string, which regexec() doesn't like. ASan just gained the ability to
intercept regexec(), which is why we didn't find this previously.
Bug: http://b/129089665
Test: adb shell grep -R /system -e "abc"
Test: toybox grep tests
Change-Id: Id707cea66a873b83bd763a3dcdf726ac7d062ce0
No significant changes, but this does get rid of our local modifications.
Bug: http://b/129089665
Test: builds
Change-Id: Ie6e3cc2198c302fc998fe6fcf027661e5dca88f0
Make "r" package available for /vendor, so that we can easily add:
PRODUCT_PACKAGES_DEBUG += r.vendor
in device/google/<device>/device.mk
Change-Id: I46c23a85c95444083b8e4b68961c541d9af9b216
So that the getevent output can be redirected and still be responsive
when events occur.
A good example is when using 'adb shell getevent' command, this latter
would only display the events when the buffer is full. With this patch
as soon as an event is received/printed, it is flushed to stdout.
Change-Id: I77eee96ed1aa59bf8d82f17eab666976ce966def
Signed-off-by: Gary Bisson <gbisson@essential.com>
This reverts commit 90a018a87a.
Not a clean revert because things have changed slightly, and I'm sticking
with the egrep/fgrep shell scripts for now.
Bug: http://b/111849261
Bug: https://bugs.exim.org/show_bug.cgi?id=2294
Test: manual
Change-Id: I72ae637c84f0eb1c2b5291db73ebff1628d54110
adbd has been built as a static executable since the same binary was
copied to the recovery partition where shared library is not supported.
However, since we now support shared library in the recovery partition,
adbd is built as a dynamic executable.
In addition, the dependency from adbd to libdebuggerd_handler is removed
as debuggerd is handled by the dynamic linker.
A few more modules in /system/core are marked as recovery_available:
true as they are transitive dependencies of the dynamic linker.
This change also includes ld.config.recovery.txt which is the linker
config file for the recovery mode. It is installed to /etc/ld.config.txt
and contains linker namespace config for the dynamic binaries under
/sbin.
Bug: 63673171
Test: `adb reboot recovery; adb devices` shows the device ID
Test: Select 'mount /system' in the recovery mode, then `adb shell`.
$ lsof -p `pidof adbd` shows that libm.so, libc.so, etc. are loaded from
the /lib directory.
Change-Id: I363d5a787863f1677ee40afb5d5841321ddaae77
For potential use with build/make/tools/fat16copy.py
Bug: 63035155
Test: mmma system/core/toolbox; newfs_msdos
Change-Id: I8ad50b8c181fc4ec21e191ec9fea8e9aa11d2d9b
Just the symlink for now. If this sticks, I'll come back to remove the code.
Bug: N/A
Test: `toolbox dd --help`
Change-Id: I9b967e9246f42db0d1f48d9d147c538d57fb3bc8
Add a non-toybox version of getprop, so that we can interface with the
new C++ PropertyInfoAreaFile class to return property context
information.
Bug: 36001741
Test: Compared toolbox getprop results with toybox getprop
Change-Id: I4e90aa5b843cb5cfcbe85f05f23ae8e22729b26e
This drops the *TOOLS automated logic, though since there are only
three tools now, that seems less important. It would have been possible
to implement that logic in a Go plugin, but that seems like a lot of
overkill to fix a few duplicated entries.
See build/soong/README.md for more information.
Test: mmma system/core/toolbox
Test: List of installed files is the same
Test: input.h-labels.h and tools.h are the same
Change-Id: I9753083c3f689e689e40001b998b39ad20509598
