This feature is not (yet?) required for Go, so we need to put it in
'recommended'.
Bug: 127712811
Change-Id: I1276965a4f5e8ba31d27166caf41220985ceb7bf
Signed-off-by: Alistair Strachan <astrachan@google.com>
If you are using Q on ext4 userdata, you need this feature.
Bug: 119769411
Change-Id: Ie00d9593614387ffc85e56961e18523b78f353f8
Signed-off-by: Alistair Strachan <astrachan@google.com>
We want to require the r8152 usb ethernet gigabit dongle driver
on any device with a usb host port (like the USB-C port on a standard
phone).
ie. we want:
CONFIG_USB in {m,y} --> CONFIG_USB_RTL8152 in {m,y}
which can be logically reversed to
CONFIG_USB_RTL8152 = n --> CONFIG_USB = n
which is implementable with current conditional xml syntax.
Bug: 110755806
Signed-off-by: Maciej Żenczykowski <maze@google.com>
Change-Id: I0fabe8c48536fbc597751afd34418e4ff42a06c6
In practice, AFAICT, this has always been required just never documented.
Bug: 124361845
Signed-off-by: Maciej Żenczykowski <maze@google.com>
Change-Id: I6035ef8e75f9daada706a9233999368a22904c4e
These are required for APEX modules, which are required starting with
Q.
Bug: 119022885
Test: builds
Change-Id: I9f1fadf3cb842f2d86f7adb5d9beb31e3c9b8be8
Require devices launching with Q to support dynamic partitions.
Also, remove from recommended list.
Bug: 124527106
Test: builds
Change-Id: I8fbf9a671e0b58eacadbb02d204ab5935061a73d
Q userspace will use eBPF for bandwidth measurement
and does not require QTAGUID netfilter extension.
As such we are okay with either QTAGUID or the less
functional but sufficient OWNER extension (provided
the requisite eBPF config options are also enabled).
ie. we require either XT_MATCH_QTAGUID or XT_MATCH_OWNER.
Furthermore XT_MATCH_OWNER also requires:
CONFIG_BPF_SYSCALL=y
CONFIG_CGROUP_BPF=y
CONFIG_NETFILTER_XT_MATCH_BPF=y
CONFIG_NET_CLS_BPF=y
(the first 3 of which are already base requirements on P 4.9+)
Note: while this change is complex, it *widens* the
set of valid kernel configurations. Any kernel that passed
configuration requirements before this change will also pass
them after this change. However additionally this allows a
Q kernel to pass on a shipped as P but then upgraded to Q
device.
Also note that as Alistair points out this doesn't work for
shipped as O then upgraded to Q devices, as:
system/bpf/libbpf_android/BpfUtils.cpp
is still checking "first_api_level" for P, and thus will not
use eBPF and thus it would actually break without QTAGUID.
Bug: 125362907
Signed-off-by: Maciej Żenczykowski <maze@google.com>
Change-Id: Ia6ce9e857d260cc6c74dd59731555479f7d38078
I've received multiple requests to enable the use of the 'netem'
packet scheduling qdisc. I've also wanted to use it myself.
The amount of code is trivial (~1000 lines of code), but it enables
some pretty useful functionality - although mostly for debugging,
troubleshooting and testing.
One can use netem to introduce delays or packet loss and in general
reproduce a lower quality network.
Now that we've fixed iproute2 to enable support for more qdiscs,
it makes sense to also enable support for it in the kernel.
Bug: 124467469
Signed-off-by: Maciej Żenczykowski <maze@google.com>
Change-Id: I0169514a997a77d044209a5b017d5f306fb3308e
Required for overlayfs integration support for adb remount
command. A requirement for any device that supports a
readonly filesystem, which includes erofs, squashfs, ext4 dedupe,
or right sized (zero space remaining) for any filesystem type
selected for the system images.
Required for vendor overlays.
Signed-off-by: Mark Salyzyn <salyzyn@google.com>
Test: adb-remount-test.sh
Bug: 109821005
Change-Id: I3ece03886db7cc97f864497cf93ec6c6c39bccd1
