PD#SWPL-45884
BUG:187441275
Problem:
ueventd: coldboot took too much time
Solution:
1.The parallel restorecon dirs can be configured in ueventd.rc
2.Add 'parallel_restorecon_dir' keywords
parallel_restorecon_dir <directory>
Verify:
redi
Change-Id: Id5b13f18d36afb894891697f21ac63e78b3fe130
Signed-off-by: yuehu mi <yuehu.mi@amlogic.com>
ueventd.rc scripts belong in the /etc/ directory of their given
partition, not the root of the partition. This can cause problems,
especially since Android.bp cannot write to the root directly, forcing
vendors to use Android.mk for these files. Note that
/system/etc/ueventd.rc moved long ago.
Test: Tree-hugger
Change-Id: I2dcaafc3c3f687f76ab6bc38af979c8b43346db0
Vendors have an interest in importing ueventd files based on certain
property values. Instead of baking this logic in the ueventd binary,
add the import option from the init parser to the ueventd parser, to
allow vendors to expand as needed.
Test: imported files are parsed
Change-Id: I674987fd48f3218e4703528c6d905b1afb5fb366
Now we have more things e.g. loading kernel modules, initialize
selinux. And before all these, system cannot make further progress. It
should be beneficial to boost this critical peroid in init.
Benchmark on a Pixel device shows this saves 100+ms on early boot
ToT release + This CL (P15538587)
D/BaseBootTest: init_stage_second_START_TIME_avg : 1563.4
D/BaseBootTest: ueventd_Coldboot_avg : 219.0
D/BaseBootTest: action_init_/system/etc/init/hw/init.rc:114_START_TIME_avg : 2103.7
ToT release (6654154)
D/BaseBootTest: init_stage_second_START_TIME_avg : 1639.0
D/BaseBootTest: ueventd_Coldboot_avg : 238.2
D/BaseBootTest: action_init_/system/etc/init/hw/init.rc:114_START_TIME_avg : 2226.0
Bug: 143857500
Bug: 147997403
Bug: 160271169
Bug: 160696502
Test: Boottime
Signed-off-by: Wei Wang <wvw@google.com>
Change-Id: I21c9e051f4ae3e953d991c031f151b2779702548
1) There's no reason or way to support boot_clock for waiting for
property changes, since the underlying futex_wait uses
CLOCK_MONOTONIC. We probably wouldn't want boot_clock even if it
did, since it doesn't make sense to consider the time a device was
suspending in the timeout for waiting for a property to change.
2) The init tokenizer has been essentially unchanged for a decade,
there's no motivation to 'fix' it to not require a trailing
newline.
3) The ueventd TODO regarding moving vendor specific ueventd.rc
entries out of rootdir has been fixed.
Test: n/a
Change-Id: I3b68e3d2f25cbd539f9f8ff526669b8af04d833d
There is no reason for these scripts to continue to exist in /, when
they are better suited for /system/etc. There are problems keeping
them at / as well, particularly that they cannot be updated with
overlayfs.
Bug: 131087886
Bug: 140313207
Test: build/boot
Merged-In: I043d9a02ba588ca37ceba2c4e28ed631792b2586
Change-Id: I043d9a02ba588ca37ceba2c4e28ed631792b2586
Userspace may want to load a different firmware than the one that the
kernel requests in some cases, therefore this change adds the ability
to ueventd to run an external handler that will determine the name of
the file that should actually be loaded.
Bug: 138352500
Test: unit tests
Change-Id: Ic5da37268fd78109f83ae52d1b903bf7322a5ee5
5aa6197d5f added the ability to
parallelize restorecon to speed up boot for devices that have not
completely moved to genfscon. This parallel restorecon happens after
the parallel ueventd handling.
This causes a performance regression for devices that have moved to
genfscon, since previously, the restorecon() was done in the main
ueventd thread in parallel with the uevent handlers.
I also tried to run the fully parallelized restorecon in parallel with
the uevent handlers, but that did not make any change to the cold boot
time, likely due to the additional overhead of parallelizing the work.
Bug: 140458170
Test: blueline coldboot time returns to pre-regression time.
Change-Id: I3cd6a869cc9b62792466813d94ad6c69834e854e
generate subdir for /sys and /sys/devices, handle restorecon in parallel.
This reduces coldboot time on our target about 300ms.
Change-Id: I9c3d0e97aacff0ca127880d936dfd5fcc2aee125
Modprobe functionality is required both within first stage init and also
as a standalone binary. Create a library for this using and extending
the logic in modalias_handler.cpp.
First stage init will attempt to load modules from /lib/modules.
Bug: 129780532
Change-Id: Ie3582358fd839c2f64e1b386b30ed551a86aef5d
Ueventd can't set properties currently, but this is an artificial
limitation, since ueventd communicates to init that it has finished
cold boot via a file, and init polls this file instead of returning to
the epoll loop, where properties are handled.
This change replaces that file with a property and thus frees ueventd
to be able to set properties.
Bug: 62301678
Test: boot, check that properties are set
Change-Id: I985688e9299456efcb2dfeef9b92668991aa9c05
This change factors out functions that handle selabels from
selinux.h/cpp into selabel.h/cpp. This allows util.cpp to be used by
the upcoming native zygote without a bunch of define flags that are
required for selinux.cpp.
Bug: 133443795
Test: Build and boot cuttlefish.
Change-Id: Ie238a96c6407c6698a605dd8803c1727abfaae7b
Some configurations won't allow ueventd to have CAP_NET_ADMIN, so the
new default size of 16M is not possible for those. Those
configurations also won't need such a large buffer size, so this
change allows devices to customize the SO_RCVBUF(FORCE) size for the
uevent socket.
This is done by adding the line 'uevent_socket_rcvbuf_size <size>' to
your device's ueventd.rc file. <size> is specified as a byte count,
for example '16M' is 16MiB.
The last parsed uevent_socket_rcvbuf_size line is the one that is
used.
Bug: 120485624
Test: boot sailfish
Test: ueventd unit tests
Change-Id: If8123b92ca8a9b089ad50318caada2f21bc94707
'Critical' services have rebooted into bootloader, like all other
catastrophic init crashes, for years now. Update the text to match.
Test: n/a
Change-Id: Icfc41bf3e383958f14ecfaab9ca187e2c3dc7fd9
Some devices have modules.alias and modules.dep for modprobe and other
purposes but do not want to opt into ueventd auto loading their
modules. Therefore we add a flag that can be added to ueventd
configuration files to opt into this behavior.
Bug: 111916071
Bug: 112048758
Test: check that modules are loaded with this opt-in
Test: check that modules are not loaded without this opt-in
Change-Id: Ifb281b273059b4671eea1ca5bc726c9e79f3adfb
For some platforms it is not known at build time what devices
will be attached at runtime. Building into the kernel or pre-loading
at init all the modules that could be needed would unnecessary bloat
the kernel. The solution is dynamic kernel module loading.
The kernel will generate uevents when devices are added, userspace
should monitor for these events and load the compatible modules.
The init process already monitors for uevents, add here the ability
to respond to modalias events and preform the correct action.
Adding this to init is preferred over an external program as we
can read and process the module alias and dependency files once,
instead of for each module needing to be loaded.
Test: Run on Beagle-X15, check all needed modules are loaded (lsmod)
Change-Id: I1b57d9aeb0a9770f309207183dc4bc2b7b905f14
Signed-off-by: Andrew F. Davis <afd@ti.com>
Reviewed-by: Sam Protsenko <semen.protsenko@linaro.org>
Clean up a few mistakes in logging initialization
1) Only init needs to clear stdout/stderr/stdin, so remove this from
ueventd, watchdogd, and subcontext init
2) Only init should reboot due to FATAL errors. This was true even
before this change due to getpid() checks, but there's no reason to
not just use the DefaultAborter for other processes.
3) It's probably a mistake for FATAL logs in init to try to gracefully
shutdown the system, so simply call RebootSystem() here.
4) Lastly, remove log.cpp since it's not actually shared code anymore
Test: build
Change-Id: Ic8c323393dc7ee98ed6bb9691361b51d0d915267
Since some vendors will have firmware in mount points in
/mnt/vendor/..., we extend the ueventd script language to allow
specifying the firmware directories.
Also, move the existing 4 directories to ueventd.rc as a primary user
of this mechanism.
Bug: 111337229
Test: boot sailfish; firmwares load
Change-Id: I0854b0b786ad761e40d2332312c637610432fce2
During uevent processing, some "by-name" symlinks will be created.
/dev/block/<type>/<device>/by-name/<partition>
<type> can be: platform, pci or vbd.
<device> might be: soc.0/f9824900.sdhci, soc.0/f9824900.sdhci, etc.
<partition> might be: system, vendor, system_a, system_b, etc.
e.g., on a non-A/B device:
/dev/block/platform/soc.0/f9824900.sdhci/by-name/system
/dev/block/platform/soc.0/f9824900.sdhci/by-name/vendor
On a A/B device:
/dev/block/platform/soc/1da4000.ufshc/by-name/system_a
/dev/block/platform/soc/1da4000.ufshc/by-name/system_b
/dev/block/platform/soc/1da4000.ufshc/by-name/vendor_a
/dev/block/platform/soc/1da4000.ufshc/by-name/vendor_b
However, those symlinks are "device-specific".
This change adds the "generic" symlinks in ueventd, in addition to
the existing symlinks, when the possible "boot devices" are specified
in device tree. e.g.,
&firmware_android {
compatible = "android,firmware";
boot_devices ="soc/1da4000.ufshc,soc.0/f9824900.sdhci";
}
The following symlinks will then be created on the aforementioned non-A/B
and A/B devices, respectively.
/dev/block/by-name/system
/dev/block/by-name/vendor
/dev/block/by-name/system_a
/dev/block/by-name/system_b
/dev/block/by-name/vendor_a
/dev/block/by-name/vendor_b
Note that both <type> and <device> are skipped in the newly create symlinks.
It assumes there is no more than one devices with the same <partition>,
which is the assumption of current first stage mount flow.
Finally, when 'boot_devices' in DT is absent, it fallbacks to extract
'boot_devices' from fstab settings. e.g., using 'soc/1da4000.ufshc',
'soc.0/f9824900.sdhci' for a fstab with the following content:
/dev/block/platform/soc/1da4000.ufshc/by-name/system
/dev/block/platform/soc.0/f9824900.sdhci/by-name/vendor
Bug: 78613232
Test: adb shell ls /dev/block/by-name
Change-Id: Iec920b5a72409b6a2bdbeeb290f0a3acd2046b5d
This change splits out the selinux initialization and supporting
functionality into selinux.cpp and splits the security related
initialization of the rng, etc to security.cpp. It also provides
additional documentation for SEPolicy loading as this has been
requested by some teams.
It additionally cleans up sehandle and sehandle_prop. The former is
static within selinux.cpp and new wrapper functions are created around
selabel_lookup*() to better serve the users. The latter is moved to
property_service.cpp as it is isolated to that file for its usage.
Test: boot bullhead
Merged-In: Idc95d493cebc681fbe686b5160502f36af149f60
Change-Id: Idc95d493cebc681fbe686b5160502f36af149f60
(cherry picked from commit 9afb86b25d8675927cb37c86119a7ecf19f74819)
ueventd may be asked to handle firmware during the time critical
coldboot process. If we double fork to avoid needing to reap the
firmware handler, then we may add significant delay to this process,
as the first child may not get scheduled quickly enough for waitpid()
to complete without delay.
Bug: 63081260
Test: boot bullhead and sailfish, check that firmwares are loaded,
no zombie ueventd processes remain, and no new errors are shown
Change-Id: I2bac3b1fbc3a58557a00326e491c104656db27ae
1) property_set() takes const std::string& for both of its arguments,
so stop using .c_str() with its parameters
2) Simplify a few places where StringPrintf() is used to concatenate strings
3) Use std::to_string() instead of StringPrintf() where it's better suited
Test: Boot bullhead
Test: init unit tests
Change-Id: I68ebda0e469f6230c8f9ad3c8d5f9444e0c4fdfd
First stage mount in init currently attempts to regenerate uevents for
specific devices to create the corresponding dev nodes. However, this
is racy as first stage mount happens early in the boot process and
it's possible that some of these devices have not yet been created by
the kernel.
To fix this issue, init will poll on the uevent socket for up to 10
seconds waiting for the kernel to create the required device. It will
return false and panic if this 10 second timeout passes.
Note that the same uevent socket is used in the uevent regeneration
and the polling code, so there is no race if the device is created
after the uevent regeneration and before polling starts; the first
poll will pick up the device.
Bug: 62681642
Bug: 62682821
Test: Boot bullhead
Test: Boot sailfish
Test: Boot hikey + hotplug/unplug sdcard
Change-Id: I4a6ff043eb7115b729ca4954ebc6c9e000132993
In order to create symlinks for USB and block devices, the path for
their parent platform device must be known.
Previously, ueventd would save each platform device that it encounters
to a list and query this list when creating the symlinks. That,
however, is racy because the uevent socket does not differentiate
uevents from RegenerateUevents() and uevents sent by the kernel when
probing a device first the first time. The below scenario is the
faulty case:
1) Kernel probes parent platform device for a block device
2) ueventd calls RegenerateUevents() and starts processing uevents
3) Kernel probes block device and sends its uevents
4) ueventd picks up the block device uevent during its uevent processing,
without yet regenerating the platform device uevent, causing improper
symlinks to be created.
This change stops storing the platform devices in a list, and instead
traverses up the directory structure for each USB or block device
until it reaches a platform device, defined as one whose subsystem is
the platform bus. This fixes the race and simplifies the ueventd
code.
Bug: 62436493
Bug: 62681642
Test: Boot bullhead
Test: Boot sailfish
Test: Init unit tests
Test: Boot hikey + hotplug/unplug sdcard
Change-Id: I21636355d8e434f30e0cba568598a6cf139e67f9
ueventd already does restorecon() for /sys/{block,class,devices}, so
instead of duplicating this effort with init, move the restorecon()
that init does for all of /sys to ueventd.
Bug: 62420036
Change-Id: I6125f8ff5316a0cf45872d1100d089d71802958f
Merged-In: I6125f8ff5316a0cf45872d1100d089d71802958f
Test: Boot sailfish, bullhead
fork() subprocesses to handle uevents in parallel.
This reduces coldboot time on bullhead from ~446ms to ~230ms.
This reduces coldboot time on sailfish from ~690ms to ~360ms.
This reduces coldboot time on ryu from ~187ms to ~122ms.
Bug: 33785894
Test: boot bullhead x40, observe no major differences in /dev and /sys
Test: boot sailfish x40, observe no major differences in /dev and /sys
Test: boot ryu x40, observe no major differences in /dev and /sys
Test: boottime tests on bullhead and sailfish
Test: init unit tests
Change-Id: Ie2f63e000b8af78d187477d31fe109f20304d749
devices.cpp handles too many things for creating one class. This
change breaks it up into various files and classes.
* Parsing is moved to ueventd_parser.cpp
* Reading from the uevent socket and Cold booting is moved to a
UeventListener class, in uevent_listener.cpp
* Firmware handling is moved to firmware_handler.cpp
* The remaining contents form a DeviceHandler class within devices.cpp
Bug: 33785894
Test: boot bullhead x40, observe no major differences in /dev and /sys
Test: boot sailfish x40, observe no major differences in /dev and /sys
Test: init unit tests
Change-Id: I846a2e5995fbb344c7a8e349065c18a934fa6aba
An forward declaration for a function that was removed in later
patchsets was merged anyway. Since it's not used anywhere, this CL
removes it.
Test: build bullhead
Change-Id: Ie7223f7d23305b71438cb063b64a574ce32f96bb
Previously init_parser.cpp was made generic and capable of parsing any
number of differently named 'sections' or prefixed lines. We now use
these capabilities to do the parsing for ueventd.
Bug: 36250207
Bug: 33785894
Test: boot bullhead and ensure the right /dev nodes exist
with the right permissions set
Test: verify no boot time difference
Change-Id: I698ca962d414f8135af32f6c9cd778841b2b8b53
Create classes for dev and sys permissions and store these permissions
in std::vector instead of the C list.
Test: boot bullhead
Test: init unit tests
Change-Id: I874039a3db29b4c70149506da8e407123ab7eca2
Remove includes of "log.h" that really want <android-base/logging.h>
Fix header include order
Remove headers included in .cpp files that their associated .h already includes
Remove some unused headers
Test: boot bullhead
Change-Id: I2b415adfe86a5c8bbe4fb1ebc53c7b0ee2253824
