TP-Link Easy Smart Switches are a family of managed switches capable of supporting QoS, VLANs and Link Aggregation Groups (LAGs). They are managed through a web based interface, giving access a number of functions, including basic packets counters per-port. Unfortunately, these switches do not implement SNMP for access to these counters, nor do they appear to implement a discrete URL for direct access to this information. This project addresses this issue to produce per-port statistics from a single command line invocation with output that can be trivially parsed for formatted output, or entered into a monitoring system like Zabbix.
This project has been tested against TP-Link switch models TL-SG1016DE and TL-SG108E. It should also be compatible with the other members of this family, including the TL-SG105E and TL-SG1024DE.
*** WARNING ***
The Easy Smart Switch family has a number of unresolved vulnerabilities, including CVE-2017-17746. As described in https://seclists.org/fulldisclosure/2017/Dec/67, once a user from a given source IP address authenticates to the web-based management interface of the switch, any other user from that same source IP address is treated as authenticated.
The Python scripts in this project should be used only from a host that does not have general user access.
essstat.py is a lightweight utility is used to pull port statistics from the switch and output in a readily parsable format. Additional code will be added to parse and either output or store these statistics.
This lightweight Python application performs a quick login through the switch's web based administrative interface, and then queries the unit for the current port statistics. Credentials for accessing the unit are passed on the command line. The utility was coded with Python 3.6 and uses the Beautiful Soup library.
essstat.py [-h] [-1] [-d] -p TPpswd [-s] [-u TPuser] TPhost
TPhost IP address or hostname of switch
-h, --help show this help message and exit
-1, --1line output in a single line
-d, --debug activate debugging output
-j, --json output in JSON format
-p TPpswd, --password TPpswd
password for switch access
-s, --statsonly output post statistics only
-u TPuser, --username TPuser
username for switch access
$ essstat.py myswitch -p ChangeMe
2020-03-28 11:25:15
max_port_num=8
1;Enabled;Link Down;0,0,0,0
2;Enabled;10M Full;3568644976,0,3144940915,0
3;Enabled;1000M Full;237232286,0,66662515,0
4;Enabled;1000M Full;4019260430,0,3721138807,0
5;Enabled;1000M Full;1300360968,0,355032522,0
6;Enabled;Link Down;0,0,0,0
7;Enabled;1000M Full;2903398648,0,4293632425,5
8;Enabled;Link Down;0,0,0,0
The simplest way to accumulate data from the switches is to have essstat.py execute with the --1line option and
append the output to a CSV file. You can then pull down a copy of the CSV file and process the raw data through
this Excel workbook to produce a dynamic chart that will automatically rescale to the available data.
The first step is to setup a directory where the CSV files will accumulate the data. I chose to run all this under the zabbix user that supports the monitoring application on this host. You may choose a different user, but just make sure that the group of the directory matches the group of the user you will use.
$ ls -ald /var/log/essstat
drwxrwxr-x. 2 root zabbix 68 Mar 30 10:56 /var/log/essstat
Next, create the cron job(s) for periodic data collection. To do this, create the file /etc/cron.d/essstat and add
one schedule for each switch you will monitor. To make this reasonably self-maintaining, include the current year as
part of the CSV file specification.
*/10 * * * * zabbix /usr/local/bin/essstat.py -1 -p ChangeMe1 orange >> /var/log/essstat/essstat-orange-`date +\%G`.csv
*/10 * * * * zabbix /usr/local/bin/essstat.py -1 -p ChangeMe2 black >> /var/log/essstat/essstat-black-`date +\%G`.csv
In the above example, there are two switches being monitored, named orange and black. For each switch, data collection
will run every 10 minutes starting on each hour. The data for orange will be accumulated in the file
/var/log/essstat/essstat-orange-2020.csv during the calendar year 2020.
This macro-enabled Excel workbook is probably the best way to read and chart the port statistics. The workbook will automatically construct a query and execute a web GET operation against the monitoring server using the essstat2.cgi script. To configure the workbook for your local installation, the defined name essstatBaseURL must be modified to point to the webserver operating on your monitoring host and the name of the CGI script. To make this update in Excel 2019 on Windows, click the Excel Formulas tab, then click the Name Manager button on the ribbon. Click on the entry for essstatBaseURL and modify the entry to suit. Be sure to click the button with the green checkmark to save the modification, close the dialog and save the updated workbook. This needs to be done only once.
When using the workbook, the name of the switch and the reporting from and to date/times are specified in the parameter table at the top left of the WebData tab. Click the Update From Web button to fetch the data into the table and dynamically update the plot on the PPS Chart tab. If the switch under study has only eight ports, the extra ports will be hidden automatically.
The name of the switch and the metric plotted appears in the title of the chart. Once the metrics have been loaded into the table, the different metrics may be loaded into the chart by selecting from the choices in the dropdown cell next to the ‘Chart metriclabel. Moving between theses metrics for the same switch does *not* require doing anotherUpdate From Web` operation.
The table on the *LocalPortNames tab allows you to override the default port names shown in the chart. This table is entirely optional and defining entries for all ports on a given switch is not required (it is perfectly fine to define port name overrides for just a couple ports for a given switch). If you have multiple switches, you can add entries for all of them in a single table.
This Excel workbook prototype can be used to process a copy of the raw --1line data output from essstat.py that has been
accumulated in a CSV file. Start by copying the file to a new name, incorporating the name of the switch being
monitored. This will be the switch monitoring notebook. For example:
C:\user\me\Documents> copy essstat-TPLhost.xlsx essstat-orange.xlsx
Next, download a copy of the CSV data that has been accumulated on your monitoring host and open it
in Excel, as well as the switch monitoring workbook. At this stage, you will need to copy the data from the CSV to the RawData tab
of the switch monitoring notebook by value. To do this, go to the CSV file in Excel and select the top-left cell, A1. In Windows,
you can use the key sequence Ctrl-Shift-End to select all the data, then press Ctrl-Insert to copy all of it. Then go to the switch
monitoring workbook and select the first data cell in the RawData tab at A2 (top-left cell, below the headings). Right click
and choose the option to paste values. With the raw data in place, you should scroll down to make note of the last populated row. The
original CSV file can now be closed.
Click on the PPS Table tab to extend the analysis table and select the metric to be charted. The key is to extend the structured
table range to match the available RawData. Press Ctrl-End to locate the end of the table. Mouse over the tiny square at the
lower left corner of the cell until your mouse pointer changes to a crosshair. Click and drag down to the same row number as the last
populated row number in the RawData tab. All the formulae and ranges in PPS Table and PPS Chart will extend automatically.
There are four metrics that are being tracked for each port: Tx Good Packets, Tx Bad Packets, Rx Good Pkts, and Rx Bad Pkts. The
dropdown at cell B2 on the PPS Table tab is used to select which metric should be populated in the table and charted.
This CGI script is used to query the CSV data that has been accumulated on your monitoring host and return matching entries. The CGI is self-contained, relying only upon access to awk for execution of a simple inline script. There are three variables implmented in the CGI:
- esTPLhost: (required) The name of the switch as used to store the accumulated CSV data.
- esFrom: Return statistics starting with this date/time in format yyyy-mm-dd HH:MM:SS. Default value is for January 1st of the current year.
- esTo: Return statistics ending with this date/time in format yyyy-mm-dd HH:MM:SS. Default value is for all data to the current date/time.
Note that partial date/time specifications are allowed, since the matching entries are determined by a simple string comparison. The From and To dates are allowed to span across a year boundary (e.g. from 2019 to 2020).
To query the monitoring server for port statistics for the switch known as orange for the time range from 2/23/2020 00:00 to 3/7/2020 11:30, the URI would look like:
http://monitoring.mydomain.com/cgi-bin/essstat.cgi?esTPLhost=orange&esTo=2020-03-07%2011:30&esFrom=2020-02-23
This CGI script is designed to support operation of the essstat.xlsm Excel workbook. The script supports the same query parameters as essstat.cgi and returns data for the same metrics. However, instead of returning raw packets counts in each record, essstat2.cgi will return the average packets per second rate since the previous record. The script will calculate the actual delta time between the current and previous record to ensure the rate is accurate. The script also has handling for individual counters resetting to zero as they wrap the maximum integer size for the counter. In such a case, the packet per second rate from the previous interval will be returned for the affected statistic.
The TP-Link Easy Smart Switch has more capabilities than a completely unmanaged switch. However, the management environment is relatively closed, with only a proprietary management client (Easy Smart Configuration Utility) or a web-based management page on the switch available. Furthermore, there is no support for monitoring the switch with SNMP. This means that our only entry into the switch will be via the protocol used by the proprietary client, or by scaping the web-based management interface on the switch.
First, a little background on the UDP-base Easy Smart Configuration Protocol (ESCP) that this project does not use. The Easy Smart Configuration Utility interacts with the switch over UDP with broadcasts. The client will send a UDP broadcast from port 29809 to 29808 of a specially encoded discovery packet. Compatible switches on the network will broadcast a response from port 29808 to 29809 than includes the name, model and IP address of the switch. From this point, it is up to the client to encode a login sequence and broadcast this on the network, with the expectation that the specific target switch will receive and process the instruction. This could be to send back information in another broadcast packet, modify the configuration of the switch, or take some other action.
This design and implementation has a number of issues that should cause some concern which have been highlighted by security researchers (@chrisdcmoore in Information disclosure vulnerability in TP-Link Easy Smart switches and @chmod7850 in Vulnerability disclosure TP-Link multiples CVEs). While hacking into the ESCP would be easy enough, I really did not like the idea of literally broadcasting credentials across the network on a regular basis to grab statistics.
The apprach that this project does use, the web-based client, is problematic as well. Using a TCP unicast connections is better, but SSL is not implemented by the switch. While it is possible to reconfigure the switch to use a different administrative username, there is only one username for accessing the switch. This precludes employing role-based access with a dedicated username for reading statistics only. The credential we use to grab the statistics could also be used to access the management interface allowing resetting of counters, reconfiguring the switch or even replacing the firmware.
Worse still are the vulnerabilities reported in CVE-2017-17746. As described in https://seclists.org/fulldisclosure/2017/Dec/67, once a user from a given source IP address authenticates to the web-based management interface of the switch, any other user from that same source IP address is treated as authenticated. This condition is created by the execution of the Python scripts in this project, where other users logged into or tunneling through the same host would then have unauthenticated access to the management interface of the switch. This problem can be mitigated by running the scripts from a dedicated management host. Use of a dedicated out-of-band management LAN could offer protection as well, but these switches are unlikely to be used in such an elaborately structured environment.
Peter Smode
psmode [at] kitsnet.us