Home | History | Annotate | Download | only in crash_reporter
      1 # crash_reporter
      2 
      3 `crash_reporter` is a deamon running on the device that saves the call stack of
      4 crashing programs. It makes use of the
      5 [Breakpad](https://chromium.googlesource.com/breakpad/breakpad/) library.
      6 
      7 During a build, Breakpad symbol files are generated for all binaries.  They are
      8 packaged into a zip file when running `m dist`, so that a developer can upload
      9 them to the crash server.
     10 
     11 On a device, if the user has opted in to metrics and crash reporting, a
     12 Breakpad minidump is generated when an executable crashes, which is then
     13 uploaded to the crash server.
     14 
     15 On the crash server, it compares the minidump's signature to the symbol files
     16 that the developer has uploaded, and extracts and symbolizes the stack trace
     17 from the minidump.
     18 
     19 ## SELinux policies
     20 
     21 In order to correctly generate a minidump, `crash_reporter` needs to be given
     22 the proper SELinux permissions for accessing the domain of the crashing
     23 executable.  By default, `crash_reporter` has only been given access to a select
     24 number of system domains, such as `metricsd`, `weave`, and `update_engine`.  If
     25 a developer wants their executable's crashes to be caught by `crash_reporter`,
     26 they will have to set their SELinux policies in their .te file to allow
     27 `crash_reporter` access to their domain.  This can be done through a simple
     28 [macro](https://android.googlesource.com/device/generic/brillo/+/master/sepolicy/te_macros):
     29 
     30     allow_crash_reporter(domain_name)
     31 
     32 Replace *domain_name* with whatever domain is assigned to the executable in
     33 the `file_contexts` file.
     34 
     35 ## Configuration
     36 
     37 `crash_reporter` has a few different configuration options that have to be set.
     38 
     39 - Crashes are only handled and uploaded if analytics reporting is enabled,
     40   either via the weave call to set `_metrics.enableAnalyticsReporting` or by
     41   manually creating the file `/data/misc/metrics/enabled` (for testing only).
     42 - The `BRILLO_CRASH_SERVER` make variable should be set in the `product.mk`
     43   file to the URL of the crash server.  For Brillo builds, it is set
     44   automatically through the product configuration.  Setting this variable will
     45   populate the `/etc/os-release.d/crash_server` file on the device, which is
     46   read by `crash_sender`.
     47 - The `BRILLO_PRODUCT_ID` make variable should be set in the `product.mk` file
     48   to the product's ID.  For Brillo builds, it is set automatically through the
     49   product configuration.  Setting this variable will populate the
     50   `/etc/os-release.d/product_id`, which is read by `crash_sender`.
     51 
     52 ## Uploading crash reports in *eng* builds
     53 
     54 By default, crash reports are only uploaded to the server for production
     55 *user* and *userdebug* images.  In *eng* builds, with crash reporting enabled
     56 the device will generate minidumps for any crashing executables but will not
     57 send them to the crash server.  If a developer does want to force an upload,
     58 they can do so by issuing the command `SECONDS_SEND_SPREAD=5 FORCE_OFFICIAL=1
     59 crash_sender` from an ADB shell.  This will send the report to the server, with
     60 the *image_type* field set to *force-official* so that these reports can be
     61 differentiated from normal reports.
     62