xmidt-org/aker

Aker is an experimental MAC address blocking scheduler.

aker's Issues

A duplicate schedule time can result in an infinite loop

If a duplicate time entry is specified it can result in an infinite loop. Below is an example of how to cause the problem:

char decode_buffer_4[] = "\x85"
                            "\xa6""weekly"
                                "\x95"
                                    "\x82"
                                        "\xa4""time" /*: */ "\x0a"
                                        "\xa7""indexes"
                                            "\x93" /* [ */ "\x00" "\x01" "\x03" /* ] */
                                    "\x82"
                                        "\xa4""time" /*: */ "\x14"
                                        "\xa7""indexes"
                                            "\x92" /* [ */ "\x00" "\x04" /* ] */
                                    "\x82"
                                        "\xa4""time" /*: */ "\x1e"
                                        "\xa7""indexes" "\xc0"
                                    "\x82"
                                        "\xa4""time" /*: */ "\x28"
                                        "\xa7""indexes"
                                            "\x91" /* [ */ "\x00" /* ] */
                                    "\x82"
                                        "\xa4""time" /*: */ "\x0a"  /*<----- Duplicate time that causes the infinite loop */
                                        "\xa7""indexes"
                                            "\x93" /* [ */ "\x03" "\x05" "\xcd\x03\x8f" /* ] */
                            "\xa4""macs"
                                "\x9a"
                                    "\xb1""11:22:33:44:55:aa"
                                    "\xb1""22:33:44:55:66:bb"
                                    "\xb1""33:44:55:66:77:cc"
                                    "\xb1""44:55:66:77:88:dd"
                                    "\xb1""55:33:44:55:66:bb"
                                    "\xb1""66:44:55:66:77:cc"
                                    "\xb1""77:55:66:77:88:dd"
                                    "\xb1""88:33:44:55:66:bb"
                                    "\xb1""99:44:55:66:77:cc"
                                    "\xb1""00:55:66:77:88:dd"
                            "\xa8""absolute"
                                "\x91"
                                    "\x82"
                                        "\xa9""unix_time" /* : */ "\xce""\x5a\x0b\x4a\xa8"
                                        "\xa7""indexes"   /* : */ "\x93" /* [ */ "\x00" "\x02" "\x09" /* ] */
                            "\xab""report_rate" /* : */ "\xcd""\x0e\x10"
                            "\xa6""ignore"      /* : */ "\xae""this parameter";

Add usage help for the command line arguments.

It would be awesome to be able to get how to use the cli via -h

Add md5 calculation, storage and verification

We want to calculate the MD5 checksum and store it on disk upon receipt and storage of the file. Upon reading of the data file from disk, we want to re-check the MD5 to ensure the data was not corrupted.

Schedule - recurring times ignored while valid absolute unix times are present

Processing of recurring times will not occur until all valid absolute unix time schedules are all past.

Scheduler should imply last weekly event if there is one

schedule {
   s->mac_count: 3
       [0]: '11:22:33:44:55:66'
       [1]: '22:33:44:55:66:aa'
       [2]: '33:44:55:66:aa:BB'
   s->absolute:
       time: ????, block_count: 2 [2, 1] #absolute time of week time 25
       time: ????, block_count: 1 [2] #absolute time of week time 32
   s->weekly:
       time: 0, block_count: 0 []  <---- implied - this is wrong, we should imply the last weekly scheduled event
       time: 23, block_count: 1 [0]
       time: 30, block_count: 1 [1]
}

There should be no implied event at time 0 for the relative schedule. Instead we want to use the last time from the previous week are in place until the first event.

Unixtime to weekly time needs to be written.

Right now the actual converter does nothing, but the unixtime to weekly time needs to be finished. Of note - we need to ensure that the time changes are accounted for correctly.

Correct behavior for the time change:

• Only the most recent event in the window of time that is truncated is applied
• During the repeated windows of time, the schedule is also repeated as thou that there was not an immediate repeat.
• Ensure that the correct next event window is calculated so Aker doesn't sleep for too long.

Change current polling model to waiting on conditional variable

Right now the schedule polls every 5 seconds to check if is time to process a schedule. Instead, change to use pthread based conditional variable visible to other threads also, especially when a change to the schedule is received.

An empty list is not accepted.

If an empty list is passed for weekly or absolute the schedule is not accepted as part of the logic added to prevent the infinite loop.

Change service name from `parental-control` to `aker`

Change service name from parental-control to aker.

Rename master branch to main

Also have to change references to the branch in .travis.yml, README, and CONTRIBUTING. Double check any other markdown files as well - sometimes links have the branch name in them.

RDK logger module override func to be patched in yocto

https://github.com/Comcast/aker/blob/e5afc7830e0df776a62b8351d43ee07d82c3409e/src/main.c#L239
Add this function as patch in yocto and not here

Handling of new schedule message

1. Message owner of schedule using a message queue.
2. Replace schedule.
3. If step # 2 is success, write the new schedule msgpack binary to file.

The timezone should be vetted before returning successful.

The success or failure of setting the timezone should be captured and reported in the acceptance of a payload.

aker/src/time.c

Line 99 in 3678973

debug_error("set_unix_time_zone() error, TZ = %s\n", time_zone);

(Status not returned)

aker/src/decode.c

Line 284 in 3678973

(void ) set_unix_time_zone((*t)->time_zone);

(Status ignored)

aker/src/decode.c

Line 111 in 3678973

decode_string_type(key, val, &s);

(Status ignored)

Add support for `zoneinfo` specification.

For devices that do not have space for the entire tzdata, we need a mechanism for supplying the zoneinfo file. We need to add the following:
[ ] add an optional value to the payload named zoneinfo field that expects a binary blob
[ ] add to the command line interface an option: --overwrite-zoneinfo dir

If the --overwrite-zone directory is specified and the zoneinfo field is present in the /aker/schedule request, then the blob specified will be written as the file: ${cli-dir}/${time_zone} where ${cli-dir} is the directory name specified with the --overwrite-zone option and the ${time_zone} is the string specified in the schedule.

This behavior should be performed each time the schedule is set.

If the schedule is deleted, the file should also be deleted.

If the --overwrite-zone option is not specified, Aker should happily use what POSIX and friends gives us.

Aker should keep trying to connect to Parodus.

Current behavior: Aker tries to Parodus once and gives up if unsuccessful.
Desired behavior: Aker should try connecting to Parodus again using back off retry logic

Add `max-macs` command line argument to specifiy max blocked device macs

Instead of having a hard coded limit, it should come in via a command line option: max-macs.

If not specified or specified to 0, then there is no limit. The size of the index for each entry will be verified to be less then the limit specified.

Add a test to confirm that all arrays can be empty or null

We need to ensure that we support arrays that are:

• empty & present
• null & present
• not present

Without unintended side effects (like crashing).

Connect to firewall to provide list of device MAC addresses

CRUD response not sent

The CRUD request is not responded to immediately (or at all from what I can tell).

Add time measurement and output in log

Measure the amount of time taken to process each time slot and output into log. This will be a useful metric to have.

Add Assurance Metrics

We would like to better understand how well Aker is performing. To enable this understanding we'd like to collect the following:

• the count of the number of devices put on the block list
• the count of the number of transitions between different blocking windows
• the count of the number of times the schedule is set
• the count of the number of times a schedule failed to decode
• the count of the number of times the md5 check as failed
• the current timezone as set by the schedule
• the timezone offset based on the timezone applied to the schedule
• the time the process started

The reporting times may be added to the schedule so the cloud can coordinate the events.

Schedule Schema

Below is an example schema represented in JSON for clarity, then MsgPack for accuracy.

• ps - required - The unix time the process started running. This should not change except when the process is restarted.
• id - required - The device-id running this instance of Aker.
• rows - required - The array of rows of data. The default is to provide the last 3 data points (typically 24h worth of data).
• ts - required - The unix time of the data set in UTC time.
• off - required - The offset in seconds based on the schedule specified timezone.
• dbc - required - The device block count is the number of device ids that have been blocked since the program started.
• wtc - required - The window transition count is the number of times a transition between defined blocking windows has been encountered since the program started.
• ssc - required - The schedule set count is the number of times a schedule has been set since the program started.
• md5 - required - The md5 check failure count is the number of times the md5 did not match the file since the program started.
• tz - required - The timezone string is the most recently specified timezone from the schedule in effect.

{
    "ps": 1622763073,
    "id": "mac:112233445566",
    "rows": [
        {
            "ts": 1622761073,
            "off": 28800,
            "dbc": 5312,
            "wtc": 204,
            "ssc": 3,
            "md5": 0,
            "tz": "PST8PDT"
        }, {
            "ts": 1622674673,
            "off": 28800,
            "dbc": 4612,
            "wtc": 123,
            "ssc": 2,
            "md5": 0,
            "tz": "PST8PDT"
        }, {
            "ts": 1622588273,
            "off": 28800,
            "dbc": 2212,
            "wtc": 54,
            "ssc": 1,
            "md5": 0,
            "tz": "PST8PDT"
        }
    ]
}

Hex MsgPack:

00000000: 82a2 7073 ce60 aeac 29a4 726f 7773 9387  ..ps.`..).rows..
00000010: a274 73ce 60b9 5e71 a36f 6666 cd70 80a3  .ts.`.^q.off.p..
00000020: 6462 63cd 14c0 a377 7463 cccc a373 7363  dbc....wtc...ssc
00000030: 03a3 6d64 3500 a274 7aa7 5053 5438 5044  ..md5..tz.PST8PD
00000040: 5487 a274 73ce 60b8 0cf1 a36f 6666 cd70  T..ts.`....off.p
00000050: 80a3 6462 63cd 1204 a377 7463 7ba3 7373  ..dbc....wtc{.ss
00000060: 6302 a36d 6435 00a2 747a a750 5354 3850  c..md5..tz.PST8P
00000070: 4454 87a2 7473 ce60 b6bb 71a3 6f66 66cd  DT..ts.`..q.off.
00000080: 7080 a364 6263 cd08 a4a3 7774 6336 a373  p..dbc....wtc6.s
00000090: 7363 01a3 6d64 3500 a274 7aa7 5053 5438  sc..md5..tz.PST8
000000a0: 5044 54                                  PDT

Rename weekly-schedule and absolute-schedule

Per a request, we need to rename weekly-schedule to weekly and absolute-schedule to absolute.

Urgent - Add tests for finalize_schedule() in decode.c

Reference - https://github.com/Comcast/aker/pull/126/files#diff-9981e0124d54e77417593d84a589ae47R121

Add new endpoint for aker version

/aker/service/version/upstream should return the github repo commit hash in msgpack format like:

{
    "upstream": "8c310fcae2a35fd1dc139584ce9d5b289736e1a8"
}

Use key names for processing the input

We should use key names to process the input file, not just key:value type combos. This will make the code significantly more robust.

Binary payload is not interpreted correctly

The binary payload below is not being interpreted correctly and is being discarded:
84 A9 74 69 6D 65 5F 7A 6F 6E 65 B0 41 6D 65 72 69 63 61 2F 4E 65 77 5F 59 6F 72 6B A6 77 65 65 6B 6C 79 93 82 A4 74 69 6D 65 0A A7 69 6E 64 65 78 65 73 93 00 01 03 82 A4 74 69 6D 65 14 A7 69 6E 64 65 78 65 73 91 00 82 A4 74 69 6D 65 1E A7 69 6E 64 65 78 65 73 C0 A4 6D 61 63 73 94 B1 31 31 3A 32 32 3A 33 33 3A 34 34 3A 35 35 3A 61 61 B1 32 32 3A 33 33 3A 34 34 3A 35 35 3A 36 36 3A 62 62 B1 33 33 3A 34 34 3A 35 35 3A 36 36 3A 37 37 3A 63 63 B1 34 34 3A 35 35 3A 36 36 3A 37 37 3A 38 38 3A 64 64 A8 61 62 73 6F 6C 75 74 65 91 82 A9 75 6E 69 78 5F 74 69 6D 65 CE 59 E5 83 17 A7 69 6E 64 65 78 65 73 92 00 02

Scheduling needs to be separate from Parodus message handling

Scheduling function needs to be done

1. whether or not parodus is available to be connected to and
2. whether or not listening for parodus events.

Clean up log messages

We want to make sure that the log messages are useful and helpful when debugging problems. We need to ensure they are valuable and not too chatty.

Add time_zone support in MsgPack

• Add time_zone type string to MsgPack structure
• Add the code to actually use the new timezone data as described below (for the weekly schedule times only).

Suggested place to start:

int main (int argc, char *argv[])
{
   struct tm *mt;
   time_t mtt;
   char ftime[10];

   setenv("TZ", "America/New_York", 1);
   tzset();
   mtt = time(NULL);
   mt = localtime(&mtt);
   strftime(ftime,sizeof(ftime),"%Z %H%M",mt);

   printf("%s\n", ftime);
}

Add endpoint to fetch blocked devices at specified time

If the user makes a Request CRUD call that hits /aker/active/now, aker should return the list of mac addresses blocked at this moment.

Scheduling needs to be added in the main loop

Use file name passed in as input parameter rather than current default assumed now.