For applying update from SD card, we used to use a thread to serve the
file with fuse. Since accessing through fuse involves going from kernel
to userspace to kernel, it may run into deadlock (e.g. for mmap_sem)
when a page fault occurs. Switch to using a process instead.
Bug: 23783099
Bug: 26313124
Change-Id: Iac0f55b1bdb078cadb520cfe1133e70fbb26eadd
Fork a logger process and send over the log lines through a pipe.
Prepend a timestamp to each line for debugging purpose. Timestamps are
relative to the start of the logger.
Example lines with the change in this CL:
[ 445.948393] Verifying update package...
[ 446.279139] I:comment is 1738 bytes; signature 1720 bytes from end
[ 449.463652] I:whole-file signature verified against RSA key 0
[ 449.463704] I:verify_file returned 0
Change-Id: I139d02ed8f2e944c1618c91d5cc43282efd50b99
When calling ScreenRecoveryUI::ShowFile(), the only thing that gets
inadequately logged is the progress bar. Replace the call to
ScreenRecoveryUI::Print() with ScreenRecoveryUI::PrintOnScreenOnly() for
the progress bar, so we can avoid calling redirect_stdio().
Change-Id: I4d7c5d5b39bebe0d5880a99d7a72cee4f0b8f325
Malloc might fail when replacing package path. In this case, print a
clear error message in the logs and let the OTA fails.
Change-Id: I7209d95edc025e3ee1b4478f5e04f6e852d97205
Signed-off-by: Jeremy Compostella <jeremy.compostella@intel.com>
Signed-off-by: Gaelle Nassiet <gaellex.nassiet@intel.com>
When system images contain the root directory, there is no entry of
"/system" in the fstab. Change it to look for "/" instead if
ro.build.system_root_image is true. We actually mount the partition
to /system_root instead, and create a symlink to /system_root/system
for /system. This allows "adb shell" to work properly.
Bug: 22855115
Change-Id: Ibac493a5a9320c98ee3b60bd2cc635b925f5454a
This makes it easier to go back and forth without losing current output.
Also make the display more like regular more(1).
Bug: http://b/20834540
Change-Id: Icc5703e9c8a378cc7072d8ebb79e34451267ee1b
Currently we are keeping one copy of the kernel log (LAST_KMSG_FILE).
This CL changes to keep up to KEEP_LOG_COUNT copies for kernel logs.
Bug: http://b/18092237
Change-Id: I1bf5e230de3efd6a48a5b2ae5a34241cb4d9ca90
For userdebug and eng builds, turn on the text display automatically
if no command is specified.
Bug: http://b/17489952
Change-Id: I3d42ba2848b968da12164ddfda915ca69dcecba1
This makes it easier for us to deal with arbitrary information at the
top, and means that headers added by specific commands don't overwrite
the default ones.
Add the fingerprint back, but broken up so it fits even on sprout's
display.
Change-Id: Id71da79ab1aa455a611d72756a3100a97ceb4c1c
Currently we are keeping one copy of the kernel log (LAST_KMSG_FILE).
This CL changes to append it to the recovery log.
Bug: 18092237
Change-Id: I06ad5629016846927153064f1663753a90296f79
Currently fugu has a custom subclass to handle this. The default code
supports devices with trackballs but not all shipping Nexus devices?
That's just silly.
Change-Id: Id2779c91284899a26b4bb1af41e7033aa889df10
Rather than add code to wrap menu items, let's just put output the
recovery version to the log. It'll be visible at the bottom of the
screen and automatically wrap.
Change-Id: I158fe2d85bc56b195e00619fba455321743923bd
The original attempt missed the fact that Print is a member function,
so the first argument is the implicit 'this'.
Change-Id: I963b668c5432804c767f0a2e3ef7dea5978a1218
Currently it rotates the log files every time it boots into the recovery
mode. We lose useful logs after ten times. This CL changes the rotation
condition so that it will rotate only if it performs some actual
operations that modify the flash (installs, wipes, sideloads and etc).
Bug: 19695622
Change-Id: Ie708ad955ef31aa500b6590c65faa72391705940
It also changes the return type of wipe_data/wipe_cache to bool, so the
caller can get the status accordingly.
Change-Id: I3022dcdadd6504dac757a52c2932d1176ffd1918
This now gets done at the framework level.
Doing it here breaks the signature on the partition.
This reverts commit ee19387905.
Bug: 19967123
Change-Id: I2a977cb0f0ba94defa1bf9091219398ddc1d3528
(cherry picked from commit 037444642b)
When the command file contains "--sideload" (as a result of 'adb reboot
sideload'), it goes into sideload mode directly. Text display will be
turned on by default. It waits for user interaction upon finishing.
When the command file contains "--sideload_auto_reboot", it enters
sideload mode silently. And it will reboot after the installation
regardless of its result, which is designed for automated testing purpose.
Change-Id: Ifdf173351221c7bbf635cfd32463b48e1fff5740
It's surprising that only one of the wipe options asks for confirmation.
This change makes it easier to add confirmation to any action.
I've also removed the version information from all but the main menu,
because I find I'm not really reading the red text because there's
so much of it all the time.
(Given that fingerprints are long and menu items aren't wrapped, we
might want to go with an actual "About" menu item instead.)
Change-Id: I7d809fdd53f9af32efc78bee618f98a69883fffe
Add support for the wipe command when using sideload within the
recovery. All the support for this command is in place, only the
execution of the actual wipe command itself was missing.
Change-Id: Ia9cdfc912bfb9f558fa89b9f0ed54e843ede41f2
Signed-off-by: Christian Poetzsch <christian.potzsch@imgtec.com>
At the moment, this is the only difference in the sprout recovery
image. That's silly. Let's just improve the error handling slightly
and always have this option present.
Also make the obscure "<3e>" less unclear.
Also use "power button" as the default text rather than "enter button",
because it's been years since anyone had one of those. (Longer term we
should let subclassers tell us the keycode and we translate it to the
correct string.)
Also move the two "Reboot" options together, put "Power off" at the
bottom (and use that terminology, like the real UI, rather than
"Power down"), and use capitals throughout.
Finally, add a README.md with some useful instructions.
Change-Id: I94fb19f73d79c54fed2dda30cefb884426641b5c
I think everything left now is here to stay (services.c might get
massaged in to libadbd if it gets refactored).
Bug: 17626262
Change-Id: I01faf8b277a601a40e3a0f4c3b8206c97f1d2ce6
The recovery system behaves a little bit differently on userdebug or
eng builds by presenting error reports to the user in the ui.
This is controlled by checking the build fingerprint for the string
:userdebug/ or :eng/. But with AOSP version numbers most AOSP
builds blows the 92 char limit of ro.build.fingerprint and therefore
the property is not set, so this condition will always be evaluated
to false, for most builds.
Instead of depending on the flaky ro.build.fingerprint this change
uses ro.debuggable.
Change-Id: I74bc00c655ac596aaf4b488ecea58f0a8de9c26b
Create a new recovery UI option to allow the user to view
/cache/recovery/last_log for their device. This gives enhanced
debugging information which may be necessary when a failed
OTA occurs.
Bug: 18094012
Change-Id: Ic3228de96e9bfc2a0141c7aab4ce392a38140cf3
We need to wipe the challenges on this partition
if OEM unlock is enabled, as this is a signal that
the user has opted out of factory reset protection.
go/factory-reset
Bug: 16633064
Change-Id: Icb8f1433bf99ca57813f5b72d5a3dd15fa94a263
Make a fuse filesystem that sits on top of the selected package file
on the sdcard, so we can verify that the file contents don't change
while being read and avoid copying the file to /tmp (that is, RAM)
before verifying and installing it.
Change-Id: Ifd982aa68bfe469eda5f839042648654bf7386a1
Drop support for sideloading OTA packages of the cache partition (a
half-solution that's long since been deprecated by "adb sideload").
Refactor the code to sideload OTA packages from SD cards: remove the
installation code from the file browser.
Change-Id: Id0dff6b27c4a5837546f174f50e2e1d0379c43db
Implement a new method of sideloading over ADB that does not require
the entire package to be held in RAM (useful for low-RAM devices and
devices using block OTA where we'd rather have more RAM available for
binary patching).
We communicate with the host using a new adb service called
"sideload-host", which makes the host act as a server, sending us
different parts of the package file on request.
We create a FUSE filesystem that creates a virtual file
"/sideload/package.zip" that is backed by the ADB connection -- users
see a normal file, but when they read from the file we're actually
fetching the data from the adb host. This file is then passed to the
verification and installation systems like any other.
To prevent a malicious adb host implementation from serving different
data to the verification and installation phases of sideloading, the
FUSE filesystem verifies that the contents of the file don't change
between reads -- every time we fetch a block from the host we compare
its hash to the previous hash for that block (if it was read before)
and cause the read to fail if it changes.
One necessary change is that the minadbd started by recovery in
sideload mode no longer drops its root privileges (they're needed to
mount the FUSE filesystem). We rely on SELinux enforcement to
restrict the set of things that can be accessed.
Change-Id: Ida7dbd3b04c1d4e27a2779d88c1da0c7c81fb114
