If userdata is default encrypted, we should mount it at boot
to avoid bringing the framework up and then down unnecessarily.
Needs matching vold changes from
https://googleplex-android-review.googlesource.com/#/c/412649/
Bug: 8769627
Change-Id: I4b8276befd832cd788e15c36edfbf8f0e18d7e6b
libselinux selinux_android_restorecon API is changing to the more
general interface with flags and dropping the older variants.
Also get rid of the old, no longer used selinux_android_setfilecon API
and rename selinux_android_setfilecon2 to it as it is the only API in use.
Change-Id: I1e71ec398ccdc24cac4ec76f1b858d0f680f4925
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
Any app can
socket_local_client_connect(s, "property_service", ANDROID_SOCKET_NAMESPACE_RESERVED, SOCK_STREAM)
and just keep it open without sending anything.
From then on, the property_service.c::handle_property_set_fd() will
just recv() forever.
This prevents any other properties (i.e. property_set("sys.powerctl", "reboot"))
from being processed.
Now, we just poll() for 2 sec before recv(...,MSG_DONTWAIT).
Bug: 12061560
Change-Id: Iffea4ebb444c7100b59c43ed87aecc5c99f9d3e8
This requires telling libselinux to use the sehandle already
obtained by init rather than re-acquiring it internally. init
retains ownership of the sehandle because it performs the
initial load, uses the sehandle for other purposes (e.g. labeling
of directories created via mkdir and labeling of socket files),
and handles the policy reload property trigger.
Change-Id: I4a380caab7f8481c33eb64fcdb16b6cabe918ebd
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
This cleans up most of the size-related problems in system/core.
There are still a few changes needed for a clean 64-bit build,
but they look like they might require changes to things like the
fastboot protocol.
Change-Id: I1560425a289fa158e13e2e3173cc3e71976f92c0
Add a service called "pre-recovery" which is normally stopped but can
be started by the system server when we want to go into recovery. It
will do any preparation needed (currently needed to handle update
packages that reside on an encrypted /data partition) and then set
sys.powerctl when it's ready to actually reboot.
Bug: 12188746
Change-Id: I894a4cb200395a0f6d7fe643ed4c2ac0a45d2052
Currently, the restorecon_recursive("/sys") call in
init.c takes approx 2 seconds on hammerhead. This change
reduces the delay to 1.2 seconds.
1) Avoid double stat call when using nftw (time savings
of 0.3 seconds)
2) Avoid the repeated calls to is_selinux_enabled() (time
savings of 0.5 seconds)
Avoid calling lsetfilecon if the file is already properly
labeled. This doesn't speed up the restorecon on /sys,
but it should help when handling files on /data.
Bug: 11640230
Change-Id: Ie212ce4f4acade208c5676d60c1f03f50e2388a4
By default ueventd creates device nodes under /dev based on the ueventd
DEVPATH. Several subsystems have special rules which are hardcoded in
devices.c. Moving forward these special rules should go in ueventd.rc.
Special rules have the syntax:
subsystem <s>
devname (uevent_devname|uevent_devpath)
[dirname <dir>]
Devices matching SUBSYSTEM=<s> will be populated under <dir>. dirname
is optional and defaults to /dev. If dirname is provided, <dir> must
start with "/".
If devname is uevent_devname, ueventd will create the device node as
<dir>/DEVNAME. DEVNAME may include intermediate subdirectories, which
ueventd will automatically create.
If devname is uevent_devpath, ueventd will use the legacy behavior of
computing DEVPATH_BASE=basepath(DEVPATH), and creating the device node
as <dir>/DEVPATH_BASE.
The new parsing code is based on init_parser.c, with small tweaks to
handle commands which don't fall under a section header.
Change-Id: I3bd1b59d7e62dfc9d289cf6ae889e237fb5bd7c5
Signed-off-by: Greg Hackmann <ghackmann@google.com>
Log an error before discarding problematic events, and add a missing
truncation check to the usb subsystem's unique codepath
Change-Id: I0d05aa287ffc63b46d1752d2a7409d35dc8caca7
Signed-off-by: Greg Hackmann <ghackmann@google.com>
To prevent clashing with forthcoming changes to uevent_parser.c
Change-Id: I2ee183261c7f43e0e4104a16a280c7ee73d7df96
Signed-off-by: Greg Hackmann <ghackmann@google.com>
Under certain conditions, poll() may raise the POLLERR
flag along with POLLIN, in which case the check for
(ufd.revents == POLLIN) results in an endless busy loop.
The following fix was applied to
hardware/libhardware_legacy/uevent/uevent.c
to fix a similar bug:
commit 3aabb260ceef10377c31c9e45fb239247f5cfeba
Author: Mathias Agopian <mathias@google.com>
Date: Mon Oct 1 14:53:18 2012 -0700
fix a typo in uevent_next_eventi
Bug: 7114973
Change-Id: I15a4c714b59aeb1d02db00517d70b5f0e5ab22c2
Applying the same fix for two more poll loops in init
and ueventd.
Change-Id: I50693f6d3c904992ac4b8a9a14a83c7106e6b9e0
Obsolete RLE 565 logo is used nowhere,
because 565 framebuffer isn't used for years.
It's not necessary to keep this thing alive anymore.
Change-Id: Ie61e168790f791230530cd3eb1c68b1f7344c9a7
Functionally equivalent to the restorecon -R toolbox command.
A use case is given by:
I48eaa2b9901ac8c978192c14493ba1058a089423
Also, fix error handling and documentation for restorecon command.
Change-Id: Ia7fbcc82645baf52c6bff0490d3492f458881cbb
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
Don't queue actions for "on property=*" if the property does not exist.
This fixes these errors on boot:
init: property 'sys.powerctl' doesn't exist while
expanding '${sys.powerctl}'
init: powerctl: cannot expand '${sys.powerctl}
Change-Id: I3bd354d73a860f856be5df2c654f940445f9efd9
The Linux kernel supports command lines up to 2048 bytes on x86,
see COMMAND_LINE_SIZE in asm/setup.h.
If any androidboot.xxxx arguments were past the 1024 byte limit
here, they were lost.
Change-Id: I6247c511a7de04109490fffa0125801d274a5501
Signed-off-by: Andrew Boie <andrew.p.boie@intel.com>
Create a new "androidboot.selinux" option, to control how userspace
handles SELinux. This kernel command line can have three options:
* disabled
* permissive
* enforcing
"disabled" completely disables userspace support for SELinux. No
policy is ever loaded, nor is the SELinux filesystem /sys/fs/selinux
ever mounted.
"permissive" loads the SELinux policy, but puts SELinux into
permissive mode. SELinux policy violations are logged, but not rejected.
"enforcing", the default, loads the SELinux policy, and places
SELinux into enforcing mode. Policy violations are rejected.
This change addresses post review comments for change
b710ed21de .
Change-Id: I912583db8e6a0e9c63380de32ad8ffc47a8a440f
When init starts up, immediately put SELinux into enforcing mode.
This is currently a no-op. We currently have everything in the
unconfined domain, so this should not break anything.
(if it does, I'll roll it back immediately)
If the kernel doesn't have SELinux support compiled in, then
don't try loading a policy and continue without SELinux protections.
Change-Id: Id0279cf82c545ea0f7090137b7566a5bc3ddd641
Add "gps." and "persist.gps." to property_perms white list.
Change-Id: I7177170676d774c87e08d1548f5264ac27cba3ff
Orig-Change-Id: I7a94caa26be4aefd2fe4f155908e2f8f09858bee
Signed-off-by: Nicolas SUET <nicolas.suet@intel.com>
Signed-off-by: jerome Pantaloni <jeromex.pantaloni@intel.com>
Signed-off-by: Jian Luo <jian.luo@intel.com>
Signed-off-by: Bruce Beare <bruce.j.beare@intel.com>
Signed-off-by: Jack Ren <jack.ren@intel.com>
Author-tracking-BZ: 118091
Add an optional argument to the socket option for specifying
a SELinux security context for the socket. Normally the socket
security context is automatically computed from the service security
context or set using the seclabel option, but this facility allows
dealing with two scenarios that cannot be addressed using the existing
mechanisms:
1) Use of logwrapper to wrap a service.
In this case, init cannot determine the service security context
as it does not directly execute it and we do not want logwrapper
to run in the same domain as the service.
2) Situations where a service has multiple sockets and we want to
label them distinctly.
Change-Id: I7ae9088c326a2140e56a8044bfb21a91505aea11
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
These properties are typically set at device provisioning time
or in the factory. They contain unit-specific data that isn't
touched by software updates or factory data reset. Only
read-only properties can be read by this mechanism.
Change-Id: Ifff9184f039072c3c0ce99f825c3075afb524514
Signed-off-by: Andrew Boie <andrew.p.boie@intel.com>
The Linux RNG may have little entropy during boot. As more and more
devices have a Hardware RNG, we mix in 512 bytes from Hardware RNG
(if present) into Linux RNG early during boot (after
wait_for_coldboot_done and before property_service_init actions in
init).
To avoid having to trust the output of Hardware RNG, we do not mix it
into the Linux RNG's primary pool or increase the Linux RNG's entropy
estimates.
Bug: 10362513
Change-Id: I80617f21710400747f5e7533e518d90ea74e2f11
This change adds a "verify" fs_mgr flag specifying that
the device in question should be verified.
Devices marked with this flag are expected to have a
footer immediately after their data containing all
the information needed to set up a verity instance.
Change-Id: I10101f2c3240228ee0932e3767fe35e673d2e720
Restarting ueventd upon policy reloads has reportedly created
stability problems for some users and could cause events to be lost.
Stop restarting ueventd and instead handle policy reloads within ueventd.
Also stops restarting installd upon policy reloads.
Change-Id: Ic7f310d69a7c420e48fbc974000cf4a5b9ab4a3b
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
We have several partitions with underscores in their names
which would not be properly linked in:
/dev/block/platform/msm_sdcc.1/by-name/
With this change more characters (_-.) are allowed in
partition name links. Also, any other character is
replaced with '_' so the resulting link names have the
same length as the partition name.
Change-Id: I746566c03db98b10326c755692362d2c10e528ae
Create a new "androidboot.selinux" option, to control how userspace
handles SELinux. This kernel command line can have three options:
* disabled
* permissive
* enforcing
"disabled" completely disables userspace support for SELinux. No
policy is ever loaded, nor is the SELinux filesystem /sys/fs/selinux
ever mounted.
"permissive" loads the SELinux policy, but puts SELinux into
permissive mode. SELinux policy violations are logged, but not rejected.
"enforcing", the default, loads the SELinux policy, and places
SELinux into enforcing mode. Policy violations are rejected.
This change addresses post review comments for change
b710ed21de .
Change-Id: I912583db8e6a0e9c63380de32ad8ffc47a8a440f
When init starts up, immediately put SELinux into enforcing mode.
This is currently a no-op. We currently have everything in the
unconfined domain, so this should not break anything.
(if it does, I'll roll it back immediately)
If the kernel doesn't have SELinux support compiled in, then
don't try loading a policy and continue without SELinux protections.
Change-Id: Id0279cf82c545ea0f7090137b7566a5bc3ddd641
* commit 'b1a9f8cf4b15a861ab998a4c5f0c69068f22c62a':
init: move initial property area allocation into bionic
property_service: make /dev/__properties__ readable
bionic's __system_property_add() now expands the property area as needed
by mapping in more pages. Rather than duplicate the mapping code, move
it inside bionic and have bionic's __system_property_area_init() set up
the first page.
Signed-off-by: Greg Hackmann <ghackmann@google.com>
(cherry picked from commit f14eef0c3c)
Change-Id: Ieb94caab1527c71f2155efe3795490b0ea215a29
Currently, system properties are passed via the environment
variable ANDROID_PROPERTY_WORKSPACE and a file descriptor passed
from parent to child. This is insecure for setuid executables,
as the environment variable can be changed by the caller.
Make the /dev/__properties__ file accessible, so an app can
get properties directly from the file, rather than relying on
environment variables.
Preserve the environment variable for compatibility with pre-existing
apps.
Bug: 8045561
(cherry picked from commit 7ece0a862c)
Change-Id: I762da21ef4075f288745efed0ec7d16c2b71303c
bionic's __system_property_add() now expands the property area as needed
by mapping in more pages. Rather than duplicate the mapping code, move
it inside bionic and have bionic's __system_property_area_init() set up
the first page.
Change-Id: If9917d5f775c1a82eb89be55b84635395145ca49
Signed-off-by: Greg Hackmann <ghackmann@google.com>
The previous patch "init: verify size of property buffers passed
to property_get" incorrectly modified one of the callers,
resulting in ro.serialno, ro.bootmode, ro.baseband, and
ro.bootloader always being set to their default values.
Bug: 9469860
(cherry picked from commit 67e3663fc9)
Change-Id: Ia7b337e1fab6e334729f47ee1269e6c736615177
Verify that the buffer passed as the value parameter to property_get
is always big enough.
(cherry picked from commit 88ac54a4e8)
Change-Id: Iacc2b42bfe4069e0bfcbb1c48474f30126a93139
Move the system property writer implementation into bionic to keep
it next to the reader implementation and allow for better testing.
(cherry picked from commit 9f5af63501)
Change-Id: Idf6100d1d0170751acd5163a22597912bff480f0
The previous patch "init: verify size of property buffers passed
to property_get" incorrectly modified one of the callers,
resulting in ro.serialno, ro.bootmode, ro.baseband, and
ro.bootloader always being set to their default values.
Bug: 9469860
Change-Id: Id45bd8dd657e8d61f4cfaf7e6b2559d2bfd05181
Move the system property writer implementation into bionic to keep
it next to the reader implementation and allow for better testing.
Change-Id: I9026e604109e30546b2849b60cab2e7e5ff00ba5
Property triggers may cause an action to be queued twice, resulting
in a loop in the action queue. Keep actions that are not on the queue
in the list_empty state (act->qlist->next == act->qlist), and only
add them to the list if they are in that state.
Bug: 8335133
Change-Id: I3a3ec18176cf19cbaa3a45220a03c7560eacfe79
This helps to ensure that when a new system image is installed,
old userdata policy isn't applied over the top of it.
Bug: 8841348
Change-Id: I135af32250aa62979763e775842ce0af3c8b6f9f
This forces a policy reload + fixcon to deal with dynamically
delivered policy changing labels on device nodes.
It's implemented as a new keyword in init.
Bug: 8702843
Change-Id: I803cf1ecf6ff8318ce25dcc5cda4f292adc9738c
Failing to set properties due to an exhausted pool leads
to difficult to debug problems. This change adds an
error log to make the situation a bit easier to handle.
Change-Id: Icc1d97a80882fb23e74428a0195f1d28dba68fa8
Move the responsibility for rebooting the system from the
reboot command to init. Init is in a better position to take
actions to bring the system down cleanly, including making sure
filesystems are mounted read-only.
The only UIDs which can perform an init triggered reboot are
root, system, and shell.
Modify the reboot command so that it calls into init to perform
the reboot. The reboot command no longer requires CAP_SYS_BOOT.
Remove the -n reboot option and code which supports it. Anyone needing
to do an unclean shutdown can just do a 'echo c > /proc/sysrq-trigger'.
Modify adb so that it calls into init to perform a shutdown.
Bug: 8646621
Change-Id: I84c0513acb549720cb0e8c9fcbda0050f5c396f5
Currently, the output of e2fsck is not saved, and we have no insight
into how many errors e2fsck is finding and fixing. Using the new
abbreviated logging feature in liblogwrap, up to the first 100 lines,
and last 4K bytes of the output of e2fsck is captured by fs_mgr, and
added to the kernel log.
Usually, the filesystem will be clean, and this will only add a few
lines to the kernel log on boot, but when things go wrong, it should
save enough to indicate what the problem is, without potentially
filling the kernel log with only e2fsck output if the filesystem is
really corrupted.
Change-Id: I9c264798e6fe721c8f818b5ce15d0975027ddbdd
This change enables labeling of dynamically created sysfs nodes
with specific SELinux security contexts.
Change-Id: If8b8d66120453123c1371ce063b6f20e8b96b6ef
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
With the old strdup() mechanism, it may return NULL if insufficient memory was
available. So we check the return value or do not use strdup to avoid null
pointer issue.
Change-Id: Id684948d6cb6c2f06327f29a2ba692f9542fce80
Signed-off-by: Hong-Mei Li <a21834@motorola.com>
Adding a new location for policy files under
/data, the new location is /data/security. The
new location is used before attempting to use
any other location.
This requires a new directory to be created by
the init script and an update to the location of
the property_contexts file for property service.
Change-Id: I955a722ac3e51fa6c1b97201b8bdef3f601cf09d
Adding a new location for policy files under
/data, the new location is /data/security. The
new location is used before attempting to use
any other location.
This requires a new directory to be created by
the init script and an update to the location of
the property_contexts file for property service.
Change-Id: I955a722ac3e51fa6c1b97201b8bdef3f601cf09d
When using device tree, platform devices may not have a
/devices/platform/ path prefix, but can be rooted in
/devices/.
Modify the platform device tracking code to store the device
path as well as the name. This way, when we create symlinks, we
can correctly skip the base platform device prefix and get to the
proper device node path.
Change-Id: I939ef8fbcb45c5c803cd9a054e40136a912efc72
Signed-off-by: Dima Zavin <dima@android.com>
This logic has been moved to platform/frameworks/base commit
8a0a929422682ba3eb6a205dc6c0638e68b909de and is no longer needed
here.
This reverts commit 01b1dee0ab.
When using device tree, platform devices may not have a
/devices/platform/ path prefix, but can be rooted in
/devices/.
Modify the platform device tracking code to store the device
path as well as the name. This way, when we create symlinks, we
can correctly skip the base platform device prefix and get to the
proper device node path.
Change-Id: I939ef8fbcb45c5c803cd9a054e40136a912efc72
Signed-off-by: Dima Zavin <dima@android.com>
Property triggers may cause an action to be queued twice, resulting
in a loop in the action queue. Keep actions that are not on the queue
in the list_empty state (act->qlist->next == act->qlist), and only
add them to the list if they are in that state.
Bug: 8335133
Change-Id: I3a3ec18176cf19cbaa3a45220a03c7560eacfe79
This reverts commit 87980b5c9f
Getting more reports of app crashes (e.g. 7990090), so given timeline for MR2, want to revert.
Change-Id: Ic2d101d2d1156b3864d5c9b88aa8c94cd5aa345f
# By Stephen Smalley
# Via Android Git Automerger (1) and others
* commit '7480c0079d8c7fcfd5b3d9decfdabfebde162781':
Label sockets consistently with the seclabel value if specified.
Currently, system properties are passed via the environment
variable ANDROID_PROPERTY_WORKSPACE and a file descriptor passed
from parent to child. This is insecure for setuid executables,
as the environment variable can be changed by the caller.
Make the /dev/__properties__ file accessible, so an app can
get properties directly from the file, rather than relying on
environment variables.
Preserve the environment variable for compatibility with pre-existing
apps.
Bug: 8045561
Change-Id: If81e2b705eca43ddd8b491871aceff59c0e00387
Currently, system properties are passed via the environment
variable ANDROID_PROPERTY_WORKSPACE and a file descriptor passed
from parent to child. This is insecure for setuid executables,
as the environment variable can be changed by the caller.
Modify system property handling so that we get the properties
from a root owned properties file, rather than using an
environment variable.
Related to bug: 8029617
Change-Id: I18610128e11c4037ed6f4dec6eba20f69fa647eb
As it turns out, no, 64K is not enough. A device provisioning
tool was generating many uevents as it manipulated the GPT,
causing events near the end of the process (including events
as a result of the final BLKRRPART ioctl()) to get lost.
It's not clear what the best value for this should be, but
increasing to 256K fixed the problem in this case.
Change-Id: I4883b34e96c89e8a6fa581bc9cd121bb021b5694
Signed-off-by: Andrew Boie <andrew.p.boie@intel.com>
Change setsebool syntax from name=value to name value.
This is to make it consistent with setprop and similar commands.
Update both the init built-in command and the toolbox command
for consistency.
Change-Id: I2c8e016ba26731c4a2ad4a49ae3b89362bf8f8a8
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
get_hardware_name() uses a fixed size buffer to slurp the contents of
/proc/cpuinfo into and with newer multicore systems this is not big
enough, leading to the inability to pick up the hardware name.
Fix this by using a dynamically allocated, exponentially growing buffer.
Change-Id: I51c6c276b6e110f462839e205a4428adc6656e75
Signed-off-by: Jon Medhurst <tixy@linaro.org>
This is necessary to ensure that the adbd socket is created in the
adbd domain rather than the init domain.
Change-Id: Id4997d7f074aeefea62b41c87b46a6609e03f527
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
In "init: harden property service" patch, dirfd() is invoked
without checking whether opendir() return successfully. It may
fail if load_persistent_properties() is invoked before userdata
partition is mounted; then dirfd(NULL) will make init crash.
This may happen if "is_charger" is true.
Change-Id: I216fb743a3c9fa050f92fcb31b62e766346d84bb
Signed-off-by: Yi-wei Zhao <gbjc64@motorola.com>
Some Android programs have problems with memory which grows
from the top down. Temporarily set ADDR_COMPAT_LAYOUT to
avoid breaking those programs.
Bug: 7188322
Change-Id: I61760500e670b4563838c63b82d4a0b6e354a86e
* commit 'e3bdaeb603ab7ac941b35b41856c6b8ebf47c6bb':
Use sigemptyset to initialise signal mask as it is more portable. Referencing sa_restorer fails on MIPS
On first boot, the directory is created with root:root ownership and
then chowned. chown clears the ISUID and ISGID bits, so we need to chmod
the directory again after chown.
Change-Id: I02dfe7a19a637678256b4e7cc09e6b5431e6f11e
This is a set of changes to the init property service
implementation to apply a SELinux check over who can
change what properties. Also included control hooks
for the 'ctl' keys.
Change-Id: I5a18809bf5536f6459a36b6bf0d622b9f5061aa0
Signed-off-by: rpcraig <rpcraig@tycho.ncsc.mil>
To support runtime policy management, add support for reloading
policy from /data/system. This can be triggered by setting the
selinux.loadpolicy property to 1, whether from init.rc after
mounting /data or from the system_server (e.g. upon invocation of
a new device admin API for provisioning policy). ueventd and
installd are restarted upon policy reloads to pick up the new
policy configurations relevant to their operation.
Change-Id: I97479aecef8cec23b32f60e09cc778cc5520b691
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
When devices are provisioned, bd address path is set in ro.bt.bdaddr_path.
On devices where this property is not present, a random bd address is
generated and saved to the property: persist.service.bdroid.bdaddr
This change allows bluetooth process to update this property
bug 6885309
Change-Id: I2e8a2471a0e890da82e6bbec6a2ef67ec8e1f8f4
Now get_android_id function returns -EINVAL if the uid/gid is not in the list
of android ids. This will allow ueventd to catch invalid ids and report the
error.
Change-Id: I943b04dd64d518891623e1ee2d561b8061af4863
Signed-off-by: Veeren Mandalia <v.mandalia@sta.samsung.com>
Automatically set the SELinux security label on directories created
by init.rc. This avoids the need to separately call restorecon on
each such directory from the init.rc file. Also restorecon /dev
and /dev/socket after initial policy load so that they are labeled
correctly before any other dev nodes or sockets are created.
Change-Id: If6af6c4887cdead949737cebdd673957e9273ead
Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
Change uevents may be triggered after new files are created on a device
in /sys, run the sys permissions fixup when they occur.
Change-Id: Iec2725c9f8a032e5124190444edaf189a766b0b2
The wait command was currently implemented but
undocumented and only used internally. This change
adds a second option toi (optionally) override the default
timeout of 5 seconds.
The wait command can be used to wait for a file to appear
during the boot process, for example a disk or other device,
and is a better alternative than an unconditional sleep.
Change-Id: Ia11d8419c0ea84deec0e1bf5adfca5bc76ff38b7
Signed-off-by: Andrew Boie <andrew.p.boie@intel.com>
This device is required by libdrm for GPUs like IvyBridge.
Change-Id: I0ac47056a9cec2100f3e6eaa5591571fe6bbc145
Signed-off-by: Lukasz Anaczkowski <lukasz.anaczkowski@intel.com>
Signed-off-by: Andrew Boie <andrew.p.boie@intel.com>
"/sbin/watchdogd <interval> <margin>" will open /dev/watchdog, try
to set the timeout to <interval>+<margin> then write to it every
<interval> seconds to reset the watchdog.
Change-Id: I15571980cdb868ec19f20e80bf8274b32107d36d
This change brings init's do_chmod, mkdir, and do_chown into line
with open's O_NOFOLLOW semantics, causing them to fail when the
last element of their target path is a symlink.
Change-Id: If00e1a25cfe17ef6f738af4bf0541abd0c1b084b
The new fs_mgr library moves much of the knowledge of what filesystems
to mount into a new fstab.<device> file, and just calls one function to
mount all the filesystems.
Change-Id: If3db37530a0676000cba3e679db27aca734227e5
The new fs_mgr library moves much of the knowledge of what filesystems
to mount into a new fstab.<device> file, and just calls one function to
mount all the filesystems.
Change-Id: If3db37530a0676000cba3e679db27aca734227e5
Modify init to set the umask to 077 when forking processes.
This helps protect against inadvertant information disclosure
in init's child processes.
ueventd: Keep umask at 000. uevent needs to be able to
create device nodes with exactly the permissions it
indicates.
Testing:
1) Do an "ls -lR /data /dev" on the device before and after
the umask change and diff the output. Verified by hand
that the permission change wouldn't cause any problems.
2) Verify that package installation works, and the permissions
are as expected, when installing a program from market and
"adb install".
Bug: 3272072
Change-Id: Ie4f7f06c0ee9da8d9b6fce25d71d8991a9bce406
Adds an SVC_RESTART state that's used for an explicit "restart" of a
running service. This retains the traditional restart behavior for
critical and oneshot services (previously altered by 7e36edd8), whereby
these services are "simply restarted" instead of counting as a crash (for a
critical serivce) or going into the disabled state (for a oneshot service).
Previously, service restarts (either an explicit "restart", or a "stop,
start" pair) exhibited a race condition whereby the new (restarting)
service process was often spawned before the old (stopping) process had
terminated. This may have resulted in the new service process failing to
acquire a limited resource (file lock, socket bind, etc.) that the old
process had not yet released.
Now, a stopping service remains in the SVC_RUNNING state until its exiting
process has been reaped by waitpid. This prevents a "stop, start" sequence
from spawning a second service process before resources held by the first
are released. This enables safe service restarts by stopping the service,
waiting for the old service process to terminate, and (only then) starting
the new service process. In the event of "restarting" an already stopped
service, the previous behavior is maintained whereby the service is simply
started.
This scenario could be special-cased by the restart command, however, we
have observed instances where services are, unintentionally, stopped and
started "too quickly," and so simultaneous processes for the same service
should never be allowed.
Note that this commit alters the behaviors for explicit restarts of
critical and oneshot services. Previously these serivces would simply be
restarted, whereas now, an explicit restart of a critical service counts as
a crash (which may result in a recovery reboot) and oneshot services go
into the disabled state.
This change brings init's do_chmod, mkdir, and do_chown into line
with open's O_NOFOLLOW semantics, causing them to fail when the
last element of their target path is a symlink.
Change-Id: If00e1a25cfe17ef6f738af4bf0541abd0c1b084b
Normally, calling open on a tty will set that tty as the process
group controlling tty if none already exists. However, if the tty
is /dev/console, the kernel will never automatically set it as the
controlling tty. Call the TIOCSCTTY manually on the fd, which will
always attempt to set it as the controlling tty.
Fixes ctrl-c on the console shell when androidboot.console is not
passed on the kernel command line and the default /dev/console is
used.
Change-Id: I449cc41b47e93ac38ad6987413bb54131e1ec0cd
Add SE Android support for init and ueventd.
init:
- Load policy at boot.
- Set the security context for service daemons and their sockets.
- New built-in commands: setcon, setenforce, restorecon, setsebool.
- New option for services: seclabel.
ueventd:
- Set the security context for device directories and nodes.
Change-Id: I98ed752cde503c94d99dfa5b5a47e3c33db16aac
Creating a root owned /data/local.prop is one of the most common
ways to root an Android device. /data/local.prop is only intended
to assist developers on debuggable devices, and is never
intended to be used on production devices.
Change-Id: Ifcfa21c2ee9914b0b54445218b4cf0fea0a98e9c
If we process the import directive inline, then the ordering of the
commands for the "on xxx" sections would be a little unexpected. The
init.rc files do not really have an implied order as to which section
appears and gets processed first. The init code itself provides that
ordering explicitly. For the user, the expectation is that if both the
current file and the imported file define a section (e.g. "on init"),
then the commands in the current file will be executed first, and then
the ones from the imported file(s).
The current implementation did not do that. It processed the import
directive inline, and thus the imported (i.e. dependent) files would
appear first in the command lists for the sections. This created
unintended side effects and the solution would have been to try and
put the import lines somewhere in the middle of the init file. This
would be difficult to notice and hard to extract the dependencies.
To solve this, we add the imports to a list for each file being parsed
and process the list after finishing parsing the file. This provides
predictable order for imports and provides a logical flow from the
user perspective: the currently parsed file gets to run its commands
before the files being imported.
Change-Id: I06dc35ff286314060e16b18923683cd2787269de
Signed-off-by: Dima Zavin <dima@android.com>
Also, clean up how we initialize the ro.xx properties and process
the kernel command line.
Change-Id: Iedda6c90e31340a189171a44b2767480403354f7
Signed-off-by: Dima Zavin <dima@android.com>
This removes the hardcoding of the file import in init and instead
allows the init.rc file to fully control what is loaded.
Change-Id: I933e5bbab57f1e8705a370d660f92c6508da94d2
Signed-off-by: Dima Zavin <dima@android.com>
Adds new property syntax in init files during init file filename
expansion during the import command:
${prop.name}
So, one can do: import /init.${ro.hardware}.usb.rc
Should convert other usages of property names to use the new function.
Change-Id: I9205d7d7a2da620bc8e6b89ac0eb554fad53ded3
Signed-off-by: Dima Zavin <dima@android.com>
The property service is still started later, but the property area
and the initial boot properties are initialized before the init.rc
file is processed. This allows init.rc files to have access to boot
properties during parsing.
Change-Id: Iae9ed1093c821831a864b39ae6bc697e62b94757
Signed-off-by: Dima Zavin <dima@android.com>
If we process the import directive inline, then the ordering of the
commands for the "on xxx" sections would be a little unexpected. The
init.rc files do not really have an implied order as to which section
appears and gets processed first. The init code itself provides that
ordering explicitly. For the user, the expectation is that if both the
current file and the imported file define a section (e.g. "on init"),
then the commands in the current file will be executed first, and then
the ones from the imported file(s).
The current implementation did not do that. It processed the import
directive inline, and thus the imported (i.e. dependent) files would
appear first in the command lists for the sections. This created
unintended side effects and the solution would have been to try and
put the import lines somewhere in the middle of the init file. This
would be difficult to notice and hard to extract the dependencies.
To solve this, we add the imports to a list for each file being parsed
and process the list after finishing parsing the file. This provides
predictable order for imports and provides a logical flow from the
user perspective: the currently parsed file gets to run its commands
before the files being imported.
Change-Id: I06dc35ff286314060e16b18923683cd2787269de
Signed-off-by: Dima Zavin <dima@android.com>
