system-corennnn

Fork

init.rc: fix for "add_tid_to_cgroup failed to write" warnings

I still don't understand why only 0660 works.

ueventd: load modules in a forked process

Loading some modules (e.g., ath3k) will be blocked to wait for
firmware. The forking is necessary, otherwise ueventd has no
chance to load the firmware.

allow to load modules from standard module path

init.rc: let some services run as root

The surfaceflinger needs root permission to control vt.

Run the console shell as root to make debugging easier.

libutils: fix incorrect calculation in utf8_length() method

The first character of utf-8 could be larger than 128. If use signed char
variable to hold it, it would be treated as negative. That may result in
some unexpected errors.

For example, without this patch, suppose the code is 0xE88888, then
first_char is 0xE8 and converted to int32_t type (0xFFFFFFE8) and
masked with (~to_ignore_mask). The result utf32 is FFF08208
which is incorrect.

Change-Id: I72b355f380865bc375251eb287fc225fd585a115

Set correct permission for pppd

The pppd needs setuid root to setup ppp interface.

init: avoid virtual console be set blanked

init: change firmware_dirs for Android-x86

ueventd.rc: remove redundant definitions

Change-Id: Icb4f051abf941052aca9752ca2949386738d94e5

init.rc: use relative path for symlink

Usually it's better to use relative path for symlink. For example,
to use Android in a chroot environment, the absolute symlink will
fail outside the chroot environment. Using relative path guarantees
the link will work both inside and outside chroot.

Change-Id: I2f654a05189b9a666aed1fc7bdfb5a50357aff78

UPSTREAM_PENDING newfs_msdos: Handle alignment failure

When trying to create a new FAT32 file system using alignment, the resulted
MBR structure is corrupted, because the total sectors reported does not
correspond to the actual available sectors. This is due to the reserved
sectors field which is updated during the alignment.

When fsck_msdos will try to check this file system, it will fail because it
cannot match the reported sectors with the actual available sectors.
Due to the failed check, the file system cannot be mounted by vold.

I/fsck_msdos( 2180): FAT size too small, 242822 entries won't fit into 1897 sectors
I/fsck_msdos( 2180): fsck_msdos terminated by exit(8)
E/Vold ( 2180): Filesystem check failed (unknown exit code 8)
E/Vold ( 2180): /dev/block/vold/179:49 failed FS checks (I/O error)
D/Vold ( 2180): Volume sdcard1 state changing 3 (Checking) -> 1 (Idle-Unmounted)

In order to fix this issue, we will save the initial bpb structure and restore
it if the alignment fails.

Change-Id: I58d5078551c689b28c9070585bb330c45bab9c62
Signed-off-by: Irina Patru <irina.patru@intel.com>
Reviewed-on: https://android.intel.com/256992
Reviewed-by: Dolca, Robert <robert.dolca@intel.com>
Reviewed-by: cactus <cactus@intel.com>
Reviewed-by: Popa, Valentin <valentin.popa@intel.com>

devices: Increase the uevent socket buffer size

Depending on the timing, hald may start while ueventd is not done yet
with its initial coldboot. As hald also independently starts its own
coldboot, ueventd netlink socket buffer could get filled quickly and
thus ueventd could lose some uevents.
This is fixed by increasing the ueventd netlink socket buffer size.

Change-Id: Ie92f77c14099421740b3984f1cc23f813287b78f
Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>
Reviewed-on: https://android.intel.com/219843
Reviewed-by: Escande, Thierry <thierry.escande@intel.com>
Reviewed-by: cactus <cactus@intel.com>
Reviewed-by: Ortiz, Samuel <samuel.ortiz@intel.com>
Tested-by: Ortiz, Samuel <samuel.ortiz@intel.com>
Tested-by: cactus <cactus@intel.com>
Reviewed-by: Porlan, Patrick <patrick.porlan@intel.com>
Reviewed-by: Bianti, Sebastien <sebastien.bianti@intel.com>
Reviewed-by: Mahalingam, Ganesh <ganesh.mahalingam@intel.com>
Tested-by: Mahalingam, Ganesh <ganesh.mahalingam@intel.com>

init: Add 'readprops' command

This reads the contents of a text file and sets system properties
based on its contents.

Change-Id: Ic03de7ee9afb8a12439d0dff343560011e152beb
For: AXIA-102
Signed-off-by: Andrew Boie <andrew.p.boie@intel.com>

adb: reserve a wakelock when clients are connected over socket

USB OTG connections already have a wakelock reserved, but if a network
socket connection is made, the device can unexpectedly suspend
in the middle of a session. The lock is released when the client
disconnects.

Change-Id: Id5ac01699fab9350955c9814042259542fc6a694
Signed-off-by: Andrew Boie <andrew.p.boie@intel.com>

add modprobe-like capability and automatic ueventd loading

Author: Jianxun Zhang <jianxun.zhang@intel.com>
Author: Daniel Leung <daniel.leung@intel.com>
Author: Andrew Boie <andrew.p.boie@intel.com>

- insmod_by_dep() added to libcutils; loads a module into kernel.
Modules the target module depends on will be loaded first. Loading
will be stopped when an error occurs.

- rmmod_by_dep() added to libcutils; removes a module from kernel.
It also tries to remove other modules the target module depends
on until an error occurs.

- Implement wildcard matching for ueventd rules.
The PCI and USB addresses for devices can change from devices
from devices for a particular class of peripheral, for example,
bluetooth. The ueventd rules created with these addresses are
then device-specific.

This changes the way ueventd rules with wildcard are handled.
Instead of matching just the prefix with a trailing wildcard,
now these rules can have wildcard anywhere in the rule.
The wildcard matching is implemented using fnmatch(), where
its matching is simliar to shell pathname expansion. It suits
this particular usage model well.

For example, instead of creating a rule to match:

/sys/devices/pci0000:00/0000:00:1d.0/usb2/2-1/2-1.3/2-1.3:1.0/bluetooth/hci0/rfkill*

, this would suffice:

/sys/devices/*/bluetooth/hci0/rfkill*

- Let ueventd auto-load kernel modules. Implements the functionality
for ueventd to auto-load kernel modules when uevents are triggered.
Since /system may not be mounted when uevents are fired,
a deferred loading mechanism is implemented. Once mapping of
module and alias is available, these modules are then loaded.
Modules can also be blacklisted so they will not be loaded
automatically. One example would be the Wifi driver, as
Android's has to control its loading and unloading.

- add 'probemod' builtin command. This command accepts the name of a
kernel module plus a set of command line arguments. The module will
be loaded, along with all its dependencies, using the libcutils
insmod_by_dep() API.

- Drivers in kernel can request modules by launching a program in
user space, the program's path by default is "/sbin/modprobe".
Because Android system has no modprobe and ueventd is the only
program handling the module aliases so far, This patch provides a
cheap approach to handle kernel's requests in ueventd executable.

- Add new builtin init command "coldboot". The main purpose is to
provide an approach in init.*.rc files to fire uevents for devices
under the path which is passed as the the argument. This should be
called after /system is mounted so any queued events that need to
load a module can be fired.

libdiskconfig: Enable Mac OS X host build

Change-Id: Iacc7d5463364d0cd15ca9fabaa58ad05857d9e78
Signed-off-by: Andrew P. Boie <andrew.p.boie@intel.com>

diskconfig: Fix disk_layout.conf size parsing

If the user was appending 'K' to their lengths (which is valid
but redundant) it was being treated as megabytes instead of
kilobytes.

Change-Id: I979ec4d01f1342de1f3209b946201ca87bad32a7

Use progressive offset in buffer for successive writes

The code for copying from one raw file to another looped
back and tried again with a suitably smaller number of bytes
if the entire buffer was not written on the first try.
However, it failed to advance an offset in the buffer past
the bytes that were written. This could conceivably happen
under rare circumstances, leading to a corrupted disk image.

This fix defines an offset into the buffer, advancing it by
the number of bytes written on each successive retry.

Signed-off-by: Paul Drews <paul.drews@intel.com>

Change-Id: I76b8e0b3c0d3b0c32bd2114d0867ac7489808279

libdiskconfig: HACK Don't do space checks on zero-size images

If an image file is of zero size, assume we just want to expand it
to fit the available images.

Change-Id: If8f22eb12d3a5179960d4359307bb177e3323e8a
Signed-off-by: Andrew Boie <andrew.p.boie@intel.com>

sdcard: Pass the umask to sdcardfs correctly

* Google version of the kernel code expects unsigned int rather than
octal. Don't confuse it.

Change-Id: I40f060dc5212ec587e612dc275001e1a660cce67

merge in nyc-mr1-release history after reset to nyc-mr1-dev

sdcard: Fix sdcardFS check

This check is inverted, breaking FUSE.

Change-Id: I81a710f464bce73f49fb1af6050a198b5f443f52

sdcard: Add support for sdcardfs!

* sdcardfs is a kernel filesystem developed by Samsung and recently
seen in AOSP's common kernel branches. It entirely eliminates
the need to use FUSE for sdcard emulation and moves all the
wrapping logic to kernel. This vastly improves performance for
use cases which read/write to emulated as well as removable
cards.

Change-Id: I3e394a99e0a4b5beacaeddc0d4dd4855ff0cf082

DO NOT MERGE: debuggerd: verify that traced threads belong to the right process.

Fix two races in debuggerd's PTRACE_ATTACH logic:
1. The target thread in a crash dump request could exit between the
/proc/<pid>/task/<tid> check and the PTRACE_ATTACH.
2. Sibling threads could exit between listing /proc/<pid>/task and the
PTRACE_ATTACH.

Backport of NYC change I4dfe1ea30e2c211d2389321bd66e3684dd757591
Bug: http://b/29555636
Change-Id: I6c6efcf82a49bca140d761b2d1de04215ba4d252

adb: use asocket's close function when closing.

close_all_sockets was assuming that all registered local sockets used
local_socket_close as their close function. However, this is not true
for JDWP sockets.

Bug: http://b/28347842
Change-Id: I40a1174845cd33f15f30ce70828a7081cd5a087e
(cherry picked from commit 53eb31d87cb84a4212f4850bf745646e1fb12814)
(cherry picked from commit 014b01706cc64dc9c2ad94a96f62e07c058d0b5d)

adb: switch the socket list mutex to a recursive_mutex.

sockets.cpp was branching on whether a socket close function was
local_socket_close in order to avoid a potential deadlock if the socket
list lock was held while closing a peer socket.

Bug: http://b/28347842
Change-Id: I5e56f17fa54275284787f0f1dc150d1960256ab3
(functionally a cherrypick of 903b749f + 9b587dec, with windows disabled)

libutils/Unicode.cpp: Correct length computation and add checks for utf16->utf8

Inconsistent behaviour between utf16_to_utf8 and utf16_to_utf8_length
is causing a heap overflow.

Correcting the length computation and adding bound checks to the
conversion functions.

Test: ran libutils_tests
Bug: 29250543
Change-Id: I6115e3357141ed245c63c6eb25fc0fd0a9a7a2bb
(cherry picked from commit c4966a363e46d2e1074d1a365e232af0dcedd6a1)

Merge "logd: klogd: deal with htc modified printk" into nyc-mr1-dev

Merge "Fix permissions for /data/preloads" into nyc-mr1-dev

merge in nyc-mr1-release history after reset to nyc-mr1-dev