| Both sides previous revision Previous revision Next revision | Previous revision |
| diagnose_joysticks_issues [2022/02/20 01:39] – minor cleanup (WIP), added section about connection issues atari | diagnose_joysticks_issues [2025/07/25 18:13] (current) – Precisions on evtest lbrpdx |
|---|
| ====== Input problems ====== | ====== Input problems ====== |
| |
| ==== ROMs take a long time to load ==== | <WRAP center round todo> |
| | Under construction. |
| | </WRAP> |
| |
| This is caused by the X-Input settings due XBOX ONE and some 8bitdo Pads. It took 30-45s to load a ROM, this is independent from core. To fix this 8bitdo should be setted to Android Mode. You can read more [[https://retropie.org.uk/forum/topic/25820/|about here]]. | ===== ROMs take a long time to load ===== |
| |
| ==== Bluetooth Connection ==== | Some controllers (including Xbox One and 8-bitdo ones) can cause certain cores to take a long time to load a ROM, upwards of thirty seconds. In these situations, it can be worked around by setting the controller to D-input mode (may be referred to as "Android mode" by some manuals). Read more about it in [[https://retropie.org.uk/forum/topic/25820/|this Retropie forum post]]. |
| |
| This is a phenomena causes with some issues and like input lag (especially if you connect more than one BT controller) | ===== Bluetooth connection ===== |
| |
| - Try reducing distance from your controller to Raspberry (helps sometimes) | * Reduce the distance from the controller to the Batocera machine. |
| - Try to disable internal WiFi - this helps to extent signal strength, you can use ''rfkill wifi'' to disable/enable Wifi (several users report PS3 controller connections problems are vanished since) | * Disable the internal Wi-Fi - this helps to extent signal strength, you can use ''rfkill wifi'' to disable/enable Wi-Fi. |
| - Try forcing a lower baud rate in ''/etc/init.d/S31emulationstation'' or ''/etc/init.d/S32bluetooth''. Smaller values of 115200 or 230400 have been known to alleviate input lag on a Pi 3B. | * Force a lower baud rate in ''/etc/init.d/S31emulationstation'' or ''/etc/init.d/S32bluetooth''. Smaller values of 115200 or 230400 have been known to alleviate input lag on a Pi 3B. |
| - Buy a new BT adapter and put it to your Raspberry. Add ''dtoverlay=pi3-disable-bt'' in ''/boot/config.txt'' to disable internal BT module (ultima ratio, speeds up communication, extents range and speed depending on your BT adapter model) | * Buy a new BT adapter and put it to your Raspberry. Add ''dtoverlay=pi3-disable-bt'' in ''/boot/config.txt'' to disable internal BT module. |
| |
| ==== No controller detected even when connected via USB ==== | ===== No controller detected even when connected via USB ===== |
| |
| Few things to try: | Few things to try: |
| * Use a different USB port | * Use a different USB port |
| * Ensure controller itself is even functional | * Ensure controller itself is even functional |
| | * If using a third-party controller, switch the input mode on it (refer to your controller's manual) |
| | * D-input typically works best, followed by X-input, then "Mac" and Android modes |
| * Disable IOMMU setting in BIOS | * Disable IOMMU setting in BIOS |
| * Ensure there's no other obvious settings in the BIOS that could be disabling USB functionality | * Ensure there's no other obvious settings in the BIOS that could be disabling USB functionality |
| | * Add ''hid.ignore_special_drivers=1'' to the ''APPEND'' line in ''/boot/EFI/batocera/syslinux.cfg'' (for Batocera **v39** or higher) or ''/boot/EFI/BOOT/syslinux.cfg'' (for Batocera **v38** or lower) in the [[:edit_boot_partition|boot partition]]. The line should look similar to: ''%%APPEND label=BATOCERA console=tty3 quiet loglevel=0 vt.global_cursor_default=0 mitigations=off hid.ignore_special_drivers=1%%'' [[https://forum.batocera.org/d/7360-set-hidignore-special-drivers1-on-x86-architecture/8|Detailed forum post.]] |
| * Clean the contacts with rubbing alcohol (when everything is turned off and disconnected from power), let it dry, and try again | * Clean the contacts with rubbing alcohol (when everything is turned off and disconnected from power), let it dry, and try again |
| * If using a third-party controller, switch the input mode on it (refer to your controller's manual) | |
| * D-input typically works best, followed by X-input, then "Mac" and Android modes | |
| * Scream | * Scream |
| |
| ==== Controller connected and detected, but no inputs ==== | If all else fails, please provide a [[:report_issue|detailed issue report]] so that the devs can look into it. [[https://github.com/batocera-linux/batocera.linux/issues/5696|Here's a real example.]] |
| |
| Try configuring the controller by manually editing the ''/userdata/system/configs/emulationstation/es_input.cfg'' file. | ===== Controller connected and detected, but no inputs ===== |
| |
| ===== Diagnose Joysticks Issues ===== | Try configuring the controller [[#configuring_es_input_manually|by manually editing]] the ''/userdata/system/configs/emulationstation/es_input.cfg'' file. |
| |
| Physical joystick <--> linux driver <--> linux events stack <-A-> SDL2 library <-B-> EmulationStation | ===== Controller connected, detected and making inputs, but they are unexpected/ghost inputs ===== |
| |
| From your joystick to the batocera interface (EmulationStation), there are several technical steps. | Ensure that all of the controller's inputs are in a neutral position when plugging in/connected the controller. |
| In case of issues, there are 2 mains tools, evtest and sdl2-jstest | |
| |
| | If that still gives unexpected ghost inputs during mapping, try to repair the mapping process by renaming ''/userdata/system/.sdl2'' to ''/userdata/system/.sdl2_bad'' and immediately restarting ES (''[Alt]'' + ''[F4]'' or restart via SSH). If this fixes the issue, it could be because of Batocera's neutral position detection method failing, please report such issues to the Github and the steps to reproduce. |
| |
| evtest acts on point A to analyse linux events while sdl2-jstest acts on point B to analyse events going to EmulationStation. | <WRAP center round tip> |
| | Run ''%%cat /userdata/system/.sdl2/<GUID><name of controller>.cache%%'' to grab the neutral values that have been detected. |
| | </WRAP> |
| |
| ==== evtests ==== | ===== When using "Emulate cursor with joystick", the cursor is moving by itself ===== |
| | |
| | You can increase the deadzone for pad2key's mouse function by doing the following: |
| | |
| | - [[:modify_the_system_while_it_s_running|Learn how to use Batocera overlays.]] |
| | - Download the following file and save it to ''/etc/evmapy.json'': <file - evmapy.json> |
| | { "mouse_config": { "every": 0.005, "deadzone": 0.15 } } |
| | |
| | </file> |
| | - Launch a game using the feature and test the cursor. If it's still drifting, increase the "deadzone" value and try again. Higher values decrease the accuracy of pad2key's mouse emulation, so change this conservatively. |
| | - When a suitable value has been found, run ''batocera-save-overlay''. |
| | |
| | ===== Diagnose joysticks issues ===== |
| | |
| | Physical joystick <--> Linux driver <--> Linux events stack <-**A**-> SDL2 library <-**B**-> EmulationStation |
| | |
| | From your joystick to Batocera's menu interface (EmulationStation), there are several technical steps. There are two tools to analyze two separate parts of the sequence: ''evtest'' and ''sdl2-jstest''. |
| | |
| | ''evtest'' acts on point **A** in the above diagram, used to debug very low-level issues (like a button's state change not being registered by the controller or an incorrect amount of inputs being registered). ''sdl2-jstest'' acts on point **B** in the above diagram, used to debug high-level issues (like EmulationStation not reacting correctly when button X is pressed, certain emulator not "seeing" a controller's inputs which otherwise work fine in other programs, etc.). |
| | |
| | ==== evtest ==== |
| | |
| | The Linux kernel uses ''events'' to handle mouses, keyboards, joysticks, and other peripherals. |
| | |
| | SDL2 is a popular library used to help to code programs compatible with multiple OSs (Linux, Windows, MacOs). It converts Linux events to a common way to define events. On Linux, events are define in ''/usr/include/linux/input-event-codes.h''. For example, the button "yellow" if your joystick has one is : |
| |
| events in the linux kernel are the way to handle mouses, keyboards, joysticks, etc... | |
| sdl2 is a library to help to code program compatible with linux, windows, mac os x, ... it converts linux events to a common way to define events. On linux, events are define in /usr/include/linux/input-event-codes.h. For example, the button "yellow" if your joystick has one is : | |
| #define KEY_YELLOW 0x190 (1*16*16 + 9*16 + 0 = 400) | #define KEY_YELLOW 0x190 (1*16*16 + 9*16 + 0 = 400) |
| On Windows, possibly that this button is not defined, or at least, for sure, with an other value, while on Linux, the value is arbitrary. Thus, SDL doesn't use the code. Instead, SDL uses button numbers, thus, this button may be "BUTTON 1". | |
| | |
| | On Windows, it is possible that this button is not defined, or at least defined with another value, while on Linux, the value is arbitrary. As a result SDL doesn't use the corresponding code but instead button numbers. For exampplem this button may be referred to as "BUTTON 1". |
| |
| Let's see an example : | Let's see an example : |
| - Connect via ssh to batocera | - Connect to Batocera through SSH |
| - run evtest to check that linux sees your joysticks : | - run ''evtest'' to check that Linux sees your controller. |
| In my case, i've only 1 joystick plugged, it displays : | |
| | |
| | In my case, with only one controller plugged, it displays: |
| |
| # evtest | # evtest |
| ==== sdl2-jstest ==== | ==== sdl2-jstest ==== |
| |
| - run sdl2-jstest --list to check what emulationstation sees when you press buttons | Run ''%%sdl2-jstest --list%%'' to check what EmulationStation sees when you press buttons |
| In my case, it displays : | In my case, it displays : |
| # export DISPLAY=:0.0 | # export DISPLAY=:0.0 |
| SDL_JOYBUTTONUP: joystick: 0 button: 2 state: 0 code:306 | SDL_JOYBUTTONUP: joystick: 0 button: 2 state: 0 code:306 |
| ... | ... |
| | |
| | You can see a graphical representation of your controller by running: |
| | |
| | <code> |
| | sdl2-jstest --test # |
| | </code> |
| | |
| | where ''#'' is the number of the joystick you want to test. |
| | |
| | {{:sdl2-jstest_-t_0.png?500|A very old joystick indeed!}} |
| | |
| | ===== Configuring es_input manually ===== |
| | |
| | Although Batocera should be able to handle the mapping of nearly any controller, there comes a time where it is only feasibly possible to map a controller manually, such as when the controller is sending multiple signals at once for a single button press or when Batocera thinks the gyroscope is an input for all buttons. The ''/userdata/system/configs/emulationstation/es_input.cfg'' file can be edited manually. |
| | |
| | We need to know the codes the controller's inputs are sending. Open up two SSH sessions to Batocera. Yes, you can do that, just open the program twice. On one session: |
| | - Type ''sdl2-jstest -l'' and press ''[Enter]''. Take note of the number of the joystick you intend to remap. |
| | - Type ''%%sdl2-jstest --test #%%'' and press ''[Enter]'', where # is the joystick number. |
| | |
| | On the other session: |
| | - Type ''evtest'' and press ''[Enter]''. |
| | - Type in the number of the input device you intend to remap (this may be different from the one in the other session). |
| | |
| | The controller's inputs will now output values on both SSH sessions. You now have all the input information you need to make a manual remap. |
| | |
| | ==== es_input syntax ==== |
| | |
| | The syntax for each input line mapped: |
| | |
| | <code> |
| | <input name="<name>" type="<type>" id="<#>" value="<#>" code="<#>" /> |
| | </code> |
| | |
| | The two leading spaces are tabs, not spaces. This is important. |
| | |
| | Explanation of the attributes: |
| | * **name** is the virtual input on the [[https://docs.libretro.com/guides/input-and-controls/|Batocera Retropad]], the input that is ultimately sent to the emulator. This is most similar to a SNES controller, however ''pageup'' and ''pagedown'' represent ''[L1]'' and ''[R1]'' respectively. Refer to the already provided configs in ''es_input.cfg'' for examples. |
| | * **type** is the type of input being scanned for on the controller. This can be either ''button'' or ''axis''. It does not have to match up with the kind of input specified in **name**. |
| | * **id** is the number affected in the ''sdl2-jstest'' session when the button/axis is manipulated. This behaves differently for hats (usually D-pads), as they determine their direction pressed from value alone (when there is only one hat, it is always 0). |
| | * **value** is the value that the control is "pressed" on in the ''evtest'' session. Typically boolean 0 or 1 for buttons, decimal -1 to 1 for axis. |
| | * **code** is the code that the control is "pressed" on in the ''evtest'' session. This should be unique to every input. |
| | |
| | So for instance a button input line might look like this: |
| | |
| | <code> |
| | <input name="a" type="button" id="0" value="1" code="304" /> |
| | </code> |
| | |
| | The "a" is the input that is sent to the emulator, typically bound to {{:wiki:east.png?nolink&20|East button (A SNES)}} on the controller itself. "button" is its type. An id of "0" is the class it belongs to; typically all the face buttons on a controller will belong to the same id class (so "b" would also have an id of "0"). The value of "1" is achieved when the button is depressed, so this is what the emulator is looking for when reading inputs, all other values will fail to register this as the input (analog controls work differently, example below). And the code "304" is the unique part that associates the configured input with that physically button; this should always be a unique value per input on the controller. |
| | |
| | An axis might look like this: |
| | |
| | <code> |
| | <input name="joystick1left" type="axis" id="0" value="-1" code="0" /> |
| | </code> |
| | |
| | The "joystick1left" input represents the first analog stick ("1", usually the left stick) being pushed to the left (usually a negative axis value). It is of the "axis" type input. Belongs to the id class of "0". The value of "-1" indicates the direction the axis must be pushed toward to represent the input name (in this example, pushing left on the left analog stick). And the code "0" is unique to the stick. |
| | |
| | Batocera's configgen allows for "logical deduction" in regard to the axis inputs; if "joystick1left" is achieved at axis value "-1", then it assumes that "joystick1right" must be the opposite at axis value "1" without explicitly needing to be told. So theoretically, only two lines are needed for a typical analog stick: |
| | |
| | <code> |
| | <input name="joystick1left" type="axis" id="0" value="-1" code="0" /> |
| | <input name="joystick1up" type="axis" id="0" value="-1" code="1" /> |
| | </code> |
| | |
| | And this would cover all directions the stick could be pushed in. |
| | |
| | There is no logical deduction logic for any buttons, they must all be manually specified. |