ESP8266 based Irrigation ControllerLast edited on Apr 8, 2016


In the past couple of months, I've taken interest in the popular ESP8266 wifi enable uC. I've also wanted for a long time to start designing PCBs, so I decided to mix both interests and build an irrigation controller. I am no expert in PCB design nor any electronics for that matter. I'm a software guy. So this whole PCB design is done in a very amateur way and some of the experts out there might find many flaws in my design. But the important thing is that it does work as expected and most of all: I had a lot of fun and learned a lot of things while doing this. That project may look like a failure to some, but it is definitely a victory for me since I ended up learning things. And next time, I'll learn even more until I get good at it. That's the process I went through to learn about software design. I have to fail once before taking on the theory.


An irrigation controller is a pretty simple device. It's just a couple of relays, a microcontroller and a power supply. For my project, I decided to use 4 relays (to control 4 valves). The uC I chose is a esp8266, more specifically, the ESP-03 board. This brings wifi capabilities to my project. So the esp8266 connects to my network and with a multicast message, it advertises its presence. My New Home Automation system then picks it up and connects to it. From that point, it can control all 4 valves.

Part list

  • My custom built PCB
  • esp-03
  • 4 G5LE-1A4 relays
  • 1 LM2596
  • capacitors (see diagram for values)
  • resistors (see diagram for values)
  • diodes (see diagram for values)
  • uf.l SMD connector
  • 2.4ghz adhesive antenna
  • 4 MPS2907A transistors (PNP)

Power supply

The ESP8266 needs a 3.3 VDC input and my valves need 24VAC. So I decided to supply my board with 24VAC, supply the valves (through the relays) directly and convert to 3.3vdc before the ESP8266 input. I had very limited experience in designing a power supply before this project. I still do. But I figured that using a diode and a capacitor, I could easily convert VAC to VDC. I then used a switched regulator to convert the high VDC down to 3.3v. I used the LM2596T regulator because I already had some pre-built power supplies that had that on it and found that they were working pretty well. The choice of capacitor to use was confusing. Looking online, I found many people using different values than what the datasheet suggests. So I experimented with a breadboard until I found a combination that worked (I know....). So bascially: 24VAC gets converted to something like 33VDC after going through 1 diode and a capacitor to smooth it out. Then Through a LM2596 to step down to 3.3V.


I wasn't expecting to use an external antenna at first. But when trying my board for the first time, I saw that it couldn't connect to wifi. When I designed the PCB, I added a trace to a SMD uf.l connector. But at that time, I had no connector nor any antenna on hand. So I was wondering if that trace could be the problem. So I cut it. But it still wasn't working. So I finally bought a connector and an antenna, soldered a wire on the board because the trace had been cut. And then everything worked out great. So yeah, you need an external antenna. The small ceramic antenna on the esp-03 won't do the job. Possibly because I put it too close to the power supply on my board? Or because I have no ground plane? I have no idea, that's something I'm gonna have to investigate.


The choice of PNP transistors was because the ESP8266 GPIO are set high on boot. So I wanted my relays to stay opened during that time. I keep all GPIOs high when valves must be shut off, and pull the GPIO down when a valve needs to be opened. I did a lot of searching about where to put the load (my relay) in the circuit. Do I put it after the collector of before the emitter? Apperently it's better to put it after the collector. I still don't understand why though. So I still have some reading to do. But it works.


I've already talked about the basics of the ESP8266 in my Led strip controller using ESP8266. article. It is using my custom made protocole "DWN protocol" to speak with my home automation software. So after my home automation software is connected to the irrigation controller, it sends 2-bytes commands to activate a valve or to turn it off. The firmware takes care of a couple of things like:

  • Hard limit of 30 minutes "on" time for every valve. In case you forget to shut it off. If longer period is needed, then the controlling software will need to take care of it.
  • Only 1 valve at a time can be opened. If the controlling software opens a valve, the irrigation controller will shut off any other opened valves.
  • No valve sequence support. For example: V1 on for 10min, then V3 for 5min and then V2 for 1min. This needs to be done by the controlling software.

The sources:


So far the biggest flaw I found is that I mixed up the RX/TX pins of the FTDI header. In eagleCAD, the part I was using for the header had to have its RX connected to the uC's RX pin. I didn't know that, I thought I had to reverse them. If I had looked at the FTDI pinout, I would have noticed it. Lesson learned. But it's not a big deal, I made an adapter that reverses both pins.

Also, like I said earlier, the wifi doesn't work without an external antenna. No big deal, I can use an external antenna. But I added that SMD footprint like 5 min before sending to the fab. I'm pretty glad I did that. The lesson learned here: anticipate for errors: put extra stuff in case you need it. That's obviously not good thinking for production but for a hobbyist like me, it's a good thing to do. Noticed the "spare" capacitor foot print on the board? hehehe.


I've seen a lot of amateur projects like mine on sites like Hackaday and a lot of people always make the same comments so I figured I'd answer those questions immediately.

  • Why did you do it this way instead of that way? Because I felt like doing it that way.
  • Why not use this and that instead? Because that's not what I felt like playing with.
  • Isn't this a fire hazard? I don't know! Is it?!?!?! Please tell me at youjustbuiltadeathtrapyoulunatic@dumaisnet.ca
  • Why not use an Insteon EZFlora? I did. But I wanted to build an irrigation controller. So now my EZFlora is useless

Led strip controller using ESP8266.Last edited on Mar 2, 2016

I recently installed an RGB led string under my kitchen cabinet. I wanted to way to control it using any web browser on a cell phone, tablet or desktop computer. There are many ways I could have achieved this but this time, I wanted to do it all by myself and design my first PCB. This is how I did it.


The ESP8266 is small cheap wifi microcontroller. It is a 32bit microcontroller with and external 1024kb or flash memory. There are tons of information out there about this chip so I won't dig too much into the basics.

When I discovered this board, I realized the potential it could bring to my home automation projects. I can now have multiple devices that talk to my central software (New Home Automation system) wirelessly.


The ESP8266 usually comes with a pre-loaded "AT" firmware. If it isn't, you can download it for free and upload it, using an FTDI cable, to your chip. The AT firmware is a simple firmware that lets you send "AT" style commands, like in the old days with modems, to control the chip. There's an AT command to connect to an access point, another to start a TCP server, another to send a packet, etc.

A lot of people use the AT firmware. This allows you to connect a another microcontroller to the ESP8266 using the UART pins and send commands to the device. But you can also write your own firmware and take advantage of the the 1024kb of flash on the chip. You have a few GPIOs and 2 UARTs. You don't need an external microcontroller.


To upload a firmware, whether it is the AT firmware or your own, you will need the flash tool. To build your firmware, you will need a complete toolchain. The esptool can be found at https://github.com/themadinventor/esptool/. Once you have downloaded the file, just copy it somewhere like in /usr/bin so that it is in your exec path. To upload a firmware, you must obviously obtain one or build one. The firmware is separated in a couple of file. If you are uploading the AT firmware, this should work (change /dev/ttyUSB8 for the serial port where your FTDI cable is attached):

esptool.py --port /dev/ttyUSB8 --baud 9600 write_flash 0x00000 boot_v1.1.bin 0x01000 v0.20/user1.bin 0x7c000 esp_init_data_default.bin 0x7e000 blank.bin


Keep in mind that the chip must be reset while GPIO0 is pulled low in order to enter the bootloader to accept a firmware update. To create your own firmware, you will need to download the toolchain. The following should get you going:

git clone --recursive https://github.com/pfalcon/esp-open-sdk
mv xtensa-lx106-elf /opt/

Now add /opt/xtensa-lx106-elf/bin in your PATH. Next, you will need a Makefile. I've created the following:

all: firmware

SRC = main.c
CC = xtensa-lx106-elf-gcc
CFLAGS = -I. -mlongcalls
LDLIBS = -nostdlib -Wl,--start-group -lmain -lnet80211 -lwpa -llwip -lpp -lphy -Wl,--end-group -lgcc
LDFLAGS = -Teagle.app.v6.ld

    $(CC) -c $(CFLAGS) $< -o $@

firmware: $(OBJECTS)
    $(CC) $(LDFLAGS) $(LDLIBS) $(OBJECTS) -o firmware.elf
    esptool.py elf2image firmware.elf

install: firmware
    esptool.py --port $(TTY) --baud 115200 write_flash 0 firmware.elf-0x00000.bin 0x40000 firmware.elf-0x40000.bin

    rm *.bin
    rm *.elf
    rm *.o

Again, remember to reset the chip with GPIO0 pulled low before invoking 'make install'. Here is a basic source file to build a firmware

#include "ets_sys.h"
#include "gpio.h"

void ICACHE_FLASH_ATTR user_init()

    gpio_output_set(BIT5, 0, BIT5, 0);

Notice how the entry function is called user_init() instead of main(int argc, char** argv). This is because you are not running your firmware bare-metal on the board. There is an OS running in there and it will invoke user_init(). This is an event-based system. You cannot make an infinite loop like you would on another microcontroller. Instead, you let the OS run and hook your app to some events with callback functions. There is a way to create a timer and give the OS a pointer to your timer callback function. You can also give the OS a function pointer to a message handler and post messages to your app in order to receive messages in your message handler. If your code takes too much time inside a callback, the OS watchdog will reset your chip. So you cannot make infinit loops nor complex functions that do massive computations. You have to build your architecture around a single-threaded event-based system. Kindof like a javascript application.

The SDK, included in the toolchain, is pretty easy to use. For examle, to connect to a wireless access point, all you need to do is:

void ICACHE_FLASH_ATTR init_wifi(char ssid[32], char password[64])
    struct station_config station_config;
    wifi_set_opmode(1); // station mode. will connect to AP
    os_memcpy(&station_config.ssid, ssid, 32);
    os_memcpy(&station_config.password, password, 64);

void ICACHE_FLASH_ATTR user_init()

If you want to create an infinite loop, as I said previously, you cannot do this within a callback since it will block the os thread and the watchdog will reset your board. Instead, you need to post your self a message so that the OS will call your function everytime. For example:

#define TASK_QUEUE_SIZE 64
os_event_t    main_loop_task_queue[TASK_QUEUE_SIZE];

// handle a message
static void ICACHE_FLASH_ATTR main_loop(os_event_t *events)
    // do something with the message

    // now post a message again to our self, creating an infinit loop
    system_os_post(MAIN_LOOP_TASK_PRIORITY, 0, 0);

void ICACHE_FLASH_ATTR user_init()
    system_os_task(main_loop, MAIN_LOOP_TASK_PRIORITY, main_loop_task_queue, TASK_QUEUE_SIZE);

    // post a message to ourself
    system_os_post(MAIN_LOOP_TASK_PRIORITY, 0, 0);

The Board

I'm not going to show my board diagram, because frankly, I'm a little ashamed of the work I did. This is my first PCB design and I'm still learning basic electronics. Bascially, I made a PCB with 3 mosfets, a step-down voltage regulator (12V to 3.3V) and a ESP-03 chip.


The chip supports software PWM for all GPIOs but I never was able to make it work properly. So I ended up writing my own software PWM. It's pretty straight forward, I just setup a timer and toggle GPIOs according to a duty cycle set by the user application. You can get my source code here:

DWN protocol

For each of my new Wifi devices to talk to my home automation server, I had to invent a new protocol. I could have used MQTT but I felt like my idea was even more simple. I created what I call the DWN protocol (DHAS Wifi Nodes).

The protocol is very simple. The wifi node broadcasts a JSON message containing its name and unique ID every 5 seconds on a multicast address at port 242. Whenever my home automation server sees such an advertisement, it will detect the source IP address of the broadcasting device and connect to the device's TCP server listening on port 242. Once the wifi device has a TCP client connected, it continue to send advertisements every 5 seconds and also sends heartbeats to the connected client. The DWN protocol does not define the format of payloads being exchanged between the home automation server and the wifi device. The DWN protocol is just a protocol that allows wifi nodes discovery.

Using such a system, my home automation can detect any wifi devices that gets turned on in my house and it will connect to it. Since the devices broadcast their name (ie: "IRRIGATION_CONTROLLER"), the server loads the appropriate "driver" or "module" for that device and starts communicating with them. This allows me to have a plug-and-play architecture. If I want to add a second led controller device, I just build one, turn it on and the home automation server automatically detects it and starts using it.

Future projects

This was my led controller board. I am planning on making a sprinkler controller with a esp-03 board with 4 relays, using the DWM protocol with my home automation server. I will also make a small device with a ESP-01 board with 2 input GPIOs to monitor my alarm system's relays. I currently monitor my alarm system's relay with an arduino setup as an ACM device plugged in the USB of my computer. I'd rather go wifi now. I would also like to build a device that would stay connected in my car. Onces I park my car in the garage, because of the DWN protocol, my home automation server will know when my car is at home and when it isn't. So as you can see, these esp8266 chips can be used for many things in the home automation world.

Openvswitch on slackwareLast edited on Dec 23, 2015


I want to create two VM on two separate hosts (one per host) that reside on the same network. I also want the VMs to reside in a different network than the host's network. To do this with openvswitch is really easy.

  • Make sure openvswitch and tun/tap are compiled in kernel
  • install openvswitch tools
  • create bridge on each VMs
  • create masquerading rules to share internet connection with VM network

Installing openvswitch

Openvswitch is already part of the kernel. It needs to be enabled when compiling the kernel. Then the tools need to be installed. On slackware, it was just a matter of ./configure, make, make install.

Creating a bridge with GRE tunnel

To have both VMs communicate with each other, we can build a virtual switch that spans accross the two hosts. This is done with a bridge containing a GRE tunnel. Creating a bridge on each VMs is done as follow. Unless noted otherwise, these commands should be re-executed after every reboot

# create tun/tap adapters
/usr/sbin/tunctl -t tap7

# launch openvswitch server
ovsdb-server --detach --remote=punix:/usr/local/var/run/openvswitch/db.sock --remote=ptcp:6640
ovs-vsctl --no-wait init
ovs-vswitchd --pidfile --log-file --detach

# Creating the bridge. Adding tap interface and GRE tunnel in the bridge
# This is the exception: the following only needs to be done once as it will be persistant
/usr/local/bin/ovs-vsctl add-br br1
/usr/local/bin/ovs-vsctl add-port br1 tap7
/usr/local/bin/ovs-vsctl add-port br1 gre0 -- set interface gre0 type=gre options:remote_ip=**IP_ADDRESS_OF_OTHER_HOST**

# bring the bridge up
/sbin/ifconfig br1 promisc up
/sbin/ifconfig tap7 promisc up

Then repeat the same on the other host

Now, it is possible to start a VM on each host using their respective TAP7 interface. Each vm is assigned an IP address in the same subnet but different than the one of the hosts. In my case I used and Because a GRE interface is present in the bridges, both VM can communicate with each other now.

Access to the internet

Since the VMs reside on subnet but my internet gateway is at, there is no way for my VMs to communicate with the outside world. So the first thing that needs to be done, is to assign an ip address to the bridge of one of the host. This will be the gateway. So obviously, only one is needed. I chose to put it on host2. So the "br1" on host1 still has no ip address assigned to it. On host2 I did:

ifconfig br1 up

Then, masquerading needs to be done. This is not absolutely required. But if nat is not used, then the gateway of the host's network must know about the VM network and have a route to redirect traffic to the gateway. In my case, I chose to use NAT. The following commands did the trick:

# Enable forwarding. So that linux will see that
# traffic on br1 that wants to go out on a network known by another
# interface, eth0 in my case, will be forwarded on that interface.
echo 1 > /proc/sys/net/ipv4/ip_forward

# Create IP masquering rules. Assuming that eth0 is the host's interface that can
# communicate with the internet gateway (or is the gateway). Packets destined
# for external traffic will leave out eth0. So make a postrouting rule
# that will masquerade those outgoing packets. This will allow other networks
# to reach the VM network without knowing anything about it.
iptables -t nat -A POSTROUTING -o eth0 -j MASQUERADE

Of course, the default route, within the VMs, should point to At this point VMs should be able to connect to this website

Additional notes

  • port of type "patch" can be used to "connect a patch cable" between 2 vswitches. This would be used to connect two vswitches together as long as they reside on the same host. To connect two vswitches that reside on different hosts, a port of type GRE would be needed.
  • To get additional information on what type of interface can be created (patch, internal, gre, etc.): http://openvswitch.org/ovs-vswitchd.conf.db.5.pdf
  • In the example above, the ovsdb-server was started with --remote=ptcp:6640. This means that it is possible to control the vswitch from any other host using "ovs-vsctl --db=tcp: ". This is great when a virtual switch spans across several host because all host bridges can be controlled from one single point.


The best description of OpenFlow and SDN I found was this: OpenFlow is a protocol which lets you offload the control plane of all the switches to a central controller and lets a central software define the behavior of the network (also called Software Defined Networks). To me, that description was the one that opened the doors to SDN in a time where all of those terminologies was confusing to me.


Before going any further, let's first define what a flow is. An open flow switch makes forwarding/routing decisions based on its flow tables. Each flow represent a "rule" that must be matched and the action that must be taken if the rule does match. Flows are added by a separate piece of software to the switch. The switch does not "build" its own flows. These must be explicitely added to the switch by the switch administrator when first configuring the switch or dynamically by an SDN controller. If using an SDN controller, when a frame/packet/segment does not match any flows, it will be sent to the SDN controller in order to find a flow that matches.

The flows contain a "match" part. It defines fields of frame/packet/segment must be matched so that the flow applies. If the flow does apply, then the "action" part is used. Matches can be specify that a flow must match a source MAC address, a destination TCP port, or any other L2/3/4 fields. Combinations of fields can be matched and wildcards can be used.

The flows also contain an "action" part. Actions are very flexible. I won't detail all the possible actions here. An action could be to drop the packet, forward the frame, redirect to another flow table (internal to the switch) etc.

Using openflow with openvswitch

The man page for ovs-ofctl contains a section that details the flow syntax. Here is an example on how to add a flow that would drop any TCP segment with destination port 80

ovs-ofctl add-flow br0 "table=0, tcp,tp_dst=80, actions=drop"

This one would rewrite any segment going towards TCP port 80 to go to tcp port 25. This is basically port forwarding. Note that two flows need to be written. Once the server on port 25 sends a reply, it needs to be modified to use port 80 since the client is expecting to be communicating on port 80.

ovs-ofctl add-flow br0 "table=0, tcp,tp_dst=80,nw_dst=, actions=mod_tp_dst:25,normal"
ovs-ofctl add-flow br0 "table=0, tcp,tp_src=25,nw_src=, actions=mod_tp_src:80,normal"

This will add flows to the bridge on the local host. This could be done when configuring the switch initially. Obviously, more complex (and more useful) flows could be written, but this is just to get a hang of it.

It could also be useful to have the switch's flow tables be populated dynamically by some other "brain". Such a brain would be an "opendlow controller". The controller can reside on a separate host.

I downloaded pox, an openflow controller, on one of my host. I am running it on a different host than the one that is hosting the vswitch. vswitch is running on and pox is running on

git clone https://github.com/noxrepo/pox
cd pox
./pox.py forwarding.l2_learning

And then, on the vswitch host, I tell the bridge who the controller is

ovs-vsctl set-controller br0 tcp:

Using "forwarding.l2_learning" with pox creates a very basic flow that makes the switch act like a normal layer2 switch. You could write other modules for pox, but that is beyond the scope of this example. At least, now you know how to have a remote controller generate the flows dynamically. You could download another software to do it or even write your own.

By the way, using an openflow controller, we could drop the iptables usage from the above section and implement NAT using openflow only.

What this means

I wanted to learn what this was all about and I'm glad I did. I don't fully understand how this can be very usefull though. I do see the power that this brings to networking infrastructures. But for me, running a couple of computers in my basement with a relatively simple setup, it is not that usefull. I can't list a lot of uses-cases where this would change the world. I'm sure that one day, I'm going to stumble across a problem and say "Hey! That would be super-easy to do with openflow" and I'm gonna be glad I learned the basics of it.

SHA256 assembly implementationLast edited on Dec 17, 2015

Assembly implementation

For a moment now, I've been wanting to try the intel AVX instructions. So I decided to write an SHA256 function in pure x86-64 assembly. That algorithm might not be the best thing to parallelize though but it still was fun to do.

Update dec 16 2015: I have modified my code to use AVX2 instructions. I recently bought a Intel i5-6400 which supports a lot of new instructions I didn't have before. So I modified the algorithm to use bleeding-edge instructions.


The use of AVX instructions in this algorithm might not give better performances. The only reason I did this was because I wanted to play with AVX. Using AVX2 probably helps a lot more since a lot of the AVX instructions are eliminated.

In fact, I'm not entirely convinced that using AVX will benefit a lot. One thing to consider here, is that for small hashes it could be slower. The thing is that when using AVX, you are using the xmm/ymm registers. During a context switch, the OS does not automatically save/restore the state of the AVX registers. Those are lazy-saved. Meaning that, without going too much into the details, the CPU will only save/restore the AVX registers if it detects that the current thread is using them (using a dirty flag and an exception). Such a save/restore is crazy expensive. So introducing the usage of AVX registers in a thread will cost a lot for the context switch, yielding less processing time for the thread. So the thing to consider is: will the thread use the AVX instructions enough to overcome the cost of the context switch?

Process Context ID and the TLBLast edited on Nov 11, 2015

In an earlier post I said that the x86_64 architecture did not have a way to tag TLB entries. Apparently, it is possible. I don't know how I missed it. But there are caveats. I wrote another post about memory paging: Memory Paging

What the TLB does

The TLB is a cache for page translation descriptors. It stores information about virtual to physical memory mapping. When the CPU wants to access memory, it looks for a translation in the TLB. If a translation is not cached in the TLB, then this is considered a "TLB miss". The CPU then has to fetch the page descriptor from RAM using the table pointer stored in register cr3. Once the descriptor is loaded in the TLB, it stays there until the TLB gets full. When the TLB is full, the CPU purges entries to to replace them with newer ones. Access to the TLB is a lot faster than accesses to RAM. So The TLB really is just a page descriptor cache.

Benefits of TLB tagging

On the x86 architecture, when loading the cr3 register with a new PML4 address, the whole TLB gets invalidated automatically. This is because entries in the TLB might not describe pages correctly according to the new tables loaded. Entries in there are considered stale. Adress vX might point pX, but before the PML4 change, it was pointing to pY. You don't want the CPU to use those mappings.

Normally, you would change the mapping on a process change. Since the page mapping is different for each process. But the TLB is quite large. It could easily hold entries for 2 processes. So a TLB flush would be expensive for no reason. That is why Process Context Identifiers (PCID) were introduced.

First, you need to enable the PCID feature on the CPU in the cr4 register. With PCID enabled, a load to cr3 will no longer invalidate the TLB So this is dangerous if you do not maintain the PCID for each process at that point. Now, everytime you load cr3, you must change the current PCID. Each Entry added in the TLB will be tagged against the current PCID. This "tag" is invisible to us, it is only used internally in the TLB (The whole TLB is invisible to us anymway). So now, if two process access adrress vX with two different physical mapping, those two addresses can reside in the TLB without conflicting with each other. When doing a lookup in the TLB the CPU will ignore any entries tagged against a PCID that is different than the current one. So if virtual address vA exists in the TLB, and is tagged with PCID-1, then process 2 tries to use address vA, the CPU will generate a TLB-miss. Exactly what we want.

But then what about pages such as the one that contains the interrupt handler code? Every thread will want such a page mapped and the mapping would be identical. Using PCIDs would result in several TLB entries describing the same mapping but for different PCID. Those entries pollute the TLB (and waste precious cache space) since they all map to the same place. Well this is exactly why there is a "g" bit, the Global flag, in each page descriptors. The CPU will ignore the PCID for pages that are global, It would be considered an incorrect usage of the paging system if a page is global but has a differnt physical address on different threads. So the global flag is to be used carefully. I use it for kernel code and MMIO.

Advantages at a certain cost

So PCIDs are a way to avoid flushing the TLB on each cr3 load, which would become VERY expensive as the CPU would generate TLB-misses at every context switch. Now, there are no more TLB flushes but the tagging still guarantees integrity of page mapping accross threads. Can you see how important that feature is and how big the befefits are? This will considerably increase performances in your OS.

But there is one thing you must consider. When destroying a process and then, possibly, recycling the processID for a new process, you must make sure that there are no stale entries of that last process in the TLB. The INVPCID instruction is just for that. It will allow you to invalidate all TLB entries associated to a particular PCID. But then, if you are running in a multi-CPU system, things get complicated. The INVPCID instruction will only execute on the CPU executing it (obviously). But what if other CPUS have stale entries in their TLB? You then need to do a "TLB shootdown". In my OS, a TLB shootdown is done by sending an Inter-Processor Interrupt (IPI) to all CPUs. The IPI tells them to invalidate their TLB with the PCID that was shared as part of the IPI. As you can guess, this can be very costly. Sending an IPI is very expensive as all CPUs will acquire a lock, disabled their interrupts plus all the needed processing. But it would only happen everytime a new process gets created.

How to use it

First, enable the PCID feature. This is simply done by setting bit 17 in CR4. Then, everytime you load CR3, the lower 12bits are used as the PCID. So you need to guarantee a unique 12bit ID for every process. But that can be a problem. 12bits only allows 4096 processes. If you plan on supporting more than 4096 process simultaneously, then you will need to come up with some dynamic PCID scheme.

Unfortunately, my CPU does not support INVPCID. It does support PCID though. It make no sense, in my head, that the CPU would support PCID but not INVPID. So I had to work around it. I found the solution by starting a thread at forum.osdev.org. By setting bit 63 of cr3, the CPU will delete all TLB entries associated with the loaded PCID. So I came up with the following solution

// emulate_invpcid(rdx=PCID)
// will invalidate all TLB entries associated with PCID specified in RDI
// This emulation uses bit63 of cr3 to do the invalidation.
// It loads a temporary address
    push        %rax
    and         $0xFFF,%rdi
    or          $PML4TABLE,%rdi
    btsq        $63,%rdi
    mov         %cr3,%rax

    // We need to mask interrupts because getting preempted
    // while using the temporary page table address
    // things would go bad.
    mov         %rdi,%cr3
    mov         %rax,%cr3
    pop         %rax