This commit adds the libbpf_prog soong module which is heavily based on
the bpf soong module. Due to issues encountered and concerns around
compatibility and testability of mainlined bpf modules, support for
libbpf is added as a separate module.
Specifically, when changing the bpf module from
InitAndroidArchModule(... android.MultilibCommon) to MultilibFirst, the
soong apex bpfs functionality breaks. We have not identified a good
mechanism to fix this, and as we would then be creating architecture
specific bpf program binarys, additional testing and infrastructure
would be required.
It should be noted that libbpf_prog should not be used for features
which target 32bit kernel architectures at this point, as we do not
have a mechanism to identify the kernel arch at build time. I.E. 32bit
userspace could be on either 32bit or 64bit kernel arch, and we require
arch specific includes (vmlinux.h). Therefore, we default to 32bit
userspace on 64bit kernel to match the precedence set with the legacy
bpf programs.
Example use:
libbpf_prog {
name: "xyz.o",
srcs: ["xyz.c"],
arch: {
arm: {
local_include_dirs: ["vmlinux/arm64"],
},
arm64: {
local_include_dirs: ["vmlinux/arm64"],
},
x86: {
local_include_dirs: ["vmlinux/x86_64"],
},
x86_64: {
local_include_dirs: ["vmlinux/x86_64"],
},
},
}
Bug: 359646531
Test: manual
Change-Id: Ie58515d70abee061470cf4bb803228e00d496ac3
Signed-off-by: Neill Kapron <nkapron@google.com>
Add support for defaults clause in the bpf soong module, as well as
local_include_dirs property.
Test: TH
Change-Id: Iaa03e7f1f6dcf0e929d1d5b23bb79ca2a8366bef
Signed-off-by: Neill Kapron <nkapron@google.com>
Currently, all bpf programs require a cflags field which defines -Wall
and -Werror. This change enables this by default, and also enables
-Wextra. This removes boilerplate required for every bpf program, and
improves code quality.
Additionally, this change enables the default of 'btf: true' further
reducing the boilerplate required for most bpf programs in their
Android.bp file. BTF should be enabled by default, with only limited
mainlined bpf programs requiring BTF being disabled for compatibility
with older bpfloader releases.
Test: Treehugger
Change-Id: I8efd0f63115030d40c0ff7fe81d5345ff3436e5a
Signed-off-by: Neill Kapron <nkapron@google.com>
Although image variation generation logic has moved out of cc package to
the android package, the vendor and product partition variants
generation logic is still specific to cc package. Therefore, in order to
create a product or vendor variant, they have to specified in
`ExtraImageVariants`. In order to avoid such confusing behaviors and
enforce modules to specify product and vendor installation rules, this
change moves the vendor and product variant generation logic to
android.ImageInterface.
Test: m nothing --no-skip-soong-tests && diff contents of out/soong/Android-{product}.mk
Change-Id: I9e14f3739d9dea94167ee6a91e92b2f942055aba
This change modifies the interface method of
ImageInterface.SetImageVariation so that the image variation is set
directly at the caller image variation module, instead of passing the
pointer to set the image variation.
Test: m nothing
Change-Id: I8eadb5149365530243e19a8cd37eb49d335fbeef
This changes makes bpf module type uses OutputFilesProvider,
instead of current OutputFileProducer for inter-module-
communication.
Test: CI
Bug: 339477385
Bug: 342406930
Change-Id: I85d1141e9f6583cc5427756107da99f56b0c7ea1
Aconfig file dependencies are module-independent, and properly part of
ModuleBase.
Bug: b/308625757
Test: manual
Change-Id: I38c5907d1671cc69bb198345201316ae781fdc9f
Convert all of the callers of SetProvider to use the type-safe
android.SetProvider API.
Bug: 316410648
Test: builds
Change-Id: If58f4b5355264ddab2045bc3591a4eac19cd58fc
Remove the ConvertWithBp2build implementations from all the module
types, along with the related code.
Bug: 315353489
Test: m blueprint_tests
Change-Id: I212672286686a318893bc7348ddd5a5ec51e77a7
All the necessary license data is already propagated through the license
metadata file, and the entries in Android-${TARGET_PRODUCT}.mk are not
used. Remove them, which should both reduce the size and parsing cost
of the Android-${TARGET_PRODUCT}.mk file and also removes a variable
that changes when dependencies are added or removed, which will
eventually allow running kati less often.
Bug: 309006256
Test: No change to out/Android-aosp_cf_x86_64_phone.ninja
Change-Id: Idfeb6fd10d77497ca274f8e74fe8e994ce1cafe6
This no-op refactoring facilitates some upcoming functional changes for
"bp2build allowlist v2". The work requires that the bp2build conversion
mutator be changed from a TopDown mutator to a BottomUp mutator.
Refactoring all bp2build-related methods so that they use Bp2buildMutatorContext
makes it easier to make this functional change without touching tens of
files and multiple projects.
Bug: 285631638
Test: m bp2build
Change-Id: I3d1ef3064146e959c6f0dc315350fc9764bf2bd2
Following CLs will move frameworks/libs/net to packages/modules/Connectivity.
This CL is a preparation for that.
Bug: 296014682
Test: m
Change-Id: Ie6483d1e5dfbf6c4745c10d3bb031344368c8c8f
Add AndroidMkEmitAssignList to emit a line to assign the items from
the given list of string arrays.
Test: treehugger
Change-Id: Id5acbef38ea4e91349bd2461f226db352d4b8123
The first line of each section will be
```
include $(CLEAR_VARS) # <module type>
...
```
Bug: 257037252
Test: treehugger
Change-Id: Iba586155b682fe4e1e5817d8397eda8f9d9c8789
relPwd is only set if runtime.GOOS != "darwin" but is used
unconditionally. Instead, unconditionally set relPwd using the
cc.PwdPrefix() helper.
Bug: 238165437
Test: m timeInState.o
Signed-off-by: Connor O'Brien <connoro@google.com>
Change-Id: I5294bb6ea95dad60fbf64e9d958e1f1b32a02a48
Enabling BTF leads clang to generate debug line info that includes the
absolute path to the source file. For C++ modules the
-fdebug-prefix-map flag is used to convert these to relative paths. Do
the same for BPF programs when BTF is enabled.
Test: build timeInState.c; strings shows only relative path to source
file
Bug: 238165437
Signed-off-by: Connor O'Brien <connoro@google.com>
Change-Id: Ie4ebc8f9a7d115baeddde5c7ac3ae710335ddf13
Current design:
1. The bpf compiled object name is derived from the source name
(e.g. foo.c -> foo.o).
2. Full bpf program/map name are concatenated by object name + '_' +
program/map name in run-time. (e.g. obj name: x.o; program name: y_z;
full bpf program name will be x_y_z)
Issue:
x.o with map y_z and x_y.o with map z can cause naming collision in
run-time, since both result in x_y_z. This commit prevents it from
happening with a build-time check.
Bug: 236706995
Test: m
Change-Id: Ic03bfcf07a5748ed63246b71d5ae8de0405e658a
Bug: 140330870
Test: use bpf{} to install bpf program to vendor and have it
successfully loaded and installed by Android's bpfloader.
Change-Id: If0aac3098e39498f0238516ec839b6ac910cc61e
Add "btf" option that generates BTF debug info to support easier map
inspection. This is accomplished by passing the "-g" flag to clang
when compiling the BPF program.
The "-g" option also generates a number of DWARF debug sections which
are not necessary for loading BTF information, so strip these to avoid
increasing file size unnecessarily. bpfloader currently only supports
BTF info for maps, not programs, so we also strip the .BTF.ext section
containing program BTF info.
Bug: 203823368
Test: libbpf_load_test
Test: verify time_in_state.o includes .BTF section iff "btf: true" is
set
Test: verify time_in_state.o still loads if BTF is enabled
Change-Id: Ica25b253bace59d04130b0210350188399889bbe
Signed-off-by: Connor O'Brien <connoro@google.com>
Currently, the bpf module netd.o (source system/netd/bpf_progs/netd.c)
will be built to /system/etc/bpf/netd.o. In Android T, it will be moved
to mainline module com.android.tethering.
The expected behavior is:
- In T device, it uses the netd.o in mainline module.
- In pre-T devices, it uses the original netd, built from platform.
However, netd.o will be double loaded if the tethering module is
installed in Pre-T devices. Because:
1. bpf in apex is packed into /apex/MAINLINE_MODULE/etc/bpf/
2. bpf in platform is packed into /system/etc/bpf/
3. bpfloader in pre-T loads ANY bpf modules under
/apex/com.android.tethering/etc/bpf/ and /system/etc/bpf/.
We can't change the behavior of bpfloader in pre-T devices. We can't
delete the /system/etc/bpf/netd.o from pre-T devices. Both of them are
not mainline modules. So the mainlined netd.o needs to be packed into a
folder other than /apex/com.android.tethering/etc/bpf/ or
/system/etc/bpf/.
This commit adds a tag 'sub_dir' for bpf module. The installation path
of bpf modules will be:
- /system/etc/bpf/SUB_DIR/ (for platform code)
- /apex/MAINLINE_MODULE/etc/bpf/SUB_DIR/ (for mainline module)
Bug: 202086915
Test: add test in apex_test.go and build
Change-Id: Icc6619768ab006de9f86620a7df1bb2853eaba13
Delete it because there is only one user left
(system/bpfprogs/time_in_state.c). It can be specified locally.
Bug: 202086915
Test: build and boot
Change-Id: I1cb0ceb7b8e3fb63af34e11e5fece11fa038b24e
BPF headers used by both platform and mainline modules should be in
frameworks/libs/net.
Bug: 202086915
Test: build and boot
Test: cd packages/modules/Connectivity/netd; atest
Change-Id: I5a86ccd44a673a18138201f186b104f5214fd449
These packages have already been migrated to use per test build
directory so have no need for a FixtureFactory.
Bug: 183235980
Test: m nothing
Change-Id: I667d1d992caaf0f615de91f89efdae11c44986c2
Restructures the bpf package test setup code to create FixturePreparer
instances for setting up a test fixture and converts the test to use
it.
Bug: 181070625
Test: m nothing
Change-Id: I7c76ed6dc292ca92d76ef95c4167f7ca2d68e1af
Revert submission 1377717-metalics
Reason for revert: This has broken renderscript_mac target for aosp-master, see b/176909442
Reverted Changes:
I26ac54ca9:Define the standard license_kind rules.
I656486070:Export soong license data to make.
If9d661dfc:Export soong license data to make.
I97943de53:Add ability to declare licenses in soong.
Icaff40171:Rough-in license metadata support to make.
Ib8e538bd0:Add variables for notice deps, license kinds etc.
Change-Id: I51799c94a274eadab414abd80a07b5cda4584be9
Prepare for using Config when adding singletons by passing
Config to NewTestContext and NewContext instead of to ctx.Register.
This will enable a followup change to store SingletonMakeVarsProviders
registered on the Context in the Config, which is necessary to run
multiple tests in parallel without data races.
Test: all soong tests
Change-Id: Id229629a4e42ff4487d317241673837726c075fc
This reverts commit 323dc60712.
Reason for revert: Possible cause of test instability
Bug: 170513220
Test: soong tests
Change-Id: Iee168e9fbb4210569e6cffcc23e60d111403abb8
Putting t.Parallel() in each test makes them run in parallel.
Additional t.Parallel() could be added to each subtest, although
that requires making a local copy of the loop variable for
table driven tests.
Test: m checkbuild
Change-Id: I5d9869ead441093f4d7c5757f2447385333a95a4
bpf program is put to an APEX via 'bpfs' property. It is
placed under etc/bpf directory in it.
Fix: 167530625
Test: m
Change-Id: Ia36b486f0cffb619ecc4f7a318cde881abc5baf4
Test: Ran "m out/soong/.intermediates/system/netd/bpf_progs/clatd.o/android_common/obj/system/netd/bpf_progs/clatd.o"
and verified that the output .d file did NOT contain absolute paths.
Change-Id: I2d187638eb7a6db8de057b83312cba3116985ac8
This CL adds RBE support to javac, r8, and d8 rules which is only
enabled if respective environment variables are set.
Test: an aosp_crosshatch build with and without the new variables.
Change-Id: Ic82f3627944f6a5ee7b9f3228170c2709b1bfcb8
