It’s been a long time since I posted anything here. Moving to a new city has slowed my development down significantly, but things are getting back to normal again.

Recently there has been a fair bit of activity, and I’ll summarise what’s been going on in this post. There’s a couple new features, some changes to Pedigree itself, and some bug fixes. In-depth discussion of each, where necessary, follows the list.

• ZombieQueue (commit d5314bf7)
• Using > 4 GB of RAM now works, with the x86_64 port (commit bb6b954b)
• MemoryPool (commit 64454975)
• UdpLogger (versatile log callbacks) (commit ff5fddba)

ZombieQueue

ZombieQueue solves a problem that has been lying around for quite some time. Things such as signals require that a Process object is deleted as part of it being killed, so it is no longer schedulable. However, this was dealt with originally by a “delete this;” line. Any form of “delete this” can be dangerous when the next line is not a return statement. Even when the next line is a return statement, a reschedule out of the current context and into another, which allocates memory, may overwrite the existing object… making “this” invalid. That said, it did suffice for quite some time.

The idea of ZombieQueue is that you queue these deletions, allowing the thread to be rescheduled (or the call that needs deletion to return) and the object to be deleted at a later time, when it’s safe to do so. Eventually ZombieQueue might also have a “delay” parameter, which ensures that it never deletes an object until after a specific time period has passed. This could easily be used for things such as TIME_WAIT handling in the TCP stack.

ZombieQueue has already been implemented into Process, Pipe, and Socket, all of which needed to delete themselves somehow. Now they call into ZombieQueue, which is able to safely delete the object.

x86_64 port using > 4 GB of RAM

A TODO all through the physical memory manager for quite some time has been to handle regions above 4 GB in the x86_64 port. Until this change was made, no ranges of memory above 4 GB were actually accessible. In fact, using the x86_64 port on a system with over 4 GB of memory actually caused anything from a panic to a triple fault and reboot.

This has been rectified now. I’m still not sure it’ll handle a full 64-bit physical address space yet, but it will handle a massive amount of memory nonetheless.

MemoryPool

This one is something I’m surprised hasn’t been implemented yet. The idea of MemoryPool is to take buffer allocation out of the kernel heap, where it consumes far too much memory – and sometimes even consumed the full size of the heap.

MemoryPool takes a block of memory, outside of the heap, and splits it up into buffers of a specified size. These buffers are distributed to modules and applications where necessary, and returned to the pool when they are no longer needed. Buffers in, for example, the network stack, are around 2048 bytes (rounded up to the next power of 2 from 1500). Rather than have every incoming packet cause an allocation on the kernel heap of, on average, 800 bytes, this moves the processing into a pool of memory dedicated to the job.

A key feature to MemoryPool is its blocking nature. If an allocation cannot be satisfied by the current pool of memory (ie, all buffers are allocated out), it will block until a buffer becomes available. Compared to the heap, this means a constant memory usage is enforced. Effectively, the pool provides a strict contract on its memory consumption: a pool will never resize to fit a new allocation.

This does bring up a minor problem: some code can’t block – an interrupt handler for example. For this situation, a non-blocking allocation method is available, which merely returns NULL if no buffers are available.

UdpLogger

It has long been possible to install custom Log callbacks which take log entries and write them to a destination. Callbacks that already exist include the Serial logger, and a callback is used for the loading screen log dump. For testing on real machines, however, neither of these suffice. None of my test machines have a serial port, for example. They are all networked via Ethernet, however.

UdpLogger is able to be installed after the route table is installed, so it’s only really useful for debugging if you can actually boot to a shell on the test machine. It dumps all log entries to a given IP address and port. A test machine can then dump all its debugging output to a development machine running netcat.

Here it is, in action:

Log::installCallback has now also been updated to dump all the old log entries to the new callback, so every logging destination should have the same data once they are all installed.

These are just a couple of exciting new features and fixes that have happened in the past couple of weeks. At this stage we have not planned a second release – we’re still working through the bugs from Foster and finally getting around to fixing all those TODOs scattered in the code.

At the moment git master is being kept “stable” – that is, it builds and runs to the bash shell (apart from x86_64) – so feel free to grab the latest code and have a play!

Well, the Pedigree team is currently in bug fixing mode as we prepare for our first release, codenamed “Foster” and as such, there’s not been a lot to talk about innovation-wise in the kernel. So I’ve decided instead to chat about a feature of the kernel, which was implemented recently.

For our “live CDs” we realised we would need a read/write file system, in memory, with a set of applications and files to allow people to try Pedigree out without making changes to their computers. This meant that we needed a way to mount a disk image as a usable disk, which is a feature we lacked at that stage in Pedigree. So I spent a couple of hours whipping up a module to support such a feature. However, this module required a couple of tricky design decisions: Cache, and write support.

Cache is a must-have in such a module, as without it each read from the disk image ends up in a read to its parent hardware – not so great if a slow CD drive needs to seek! However, the cache must also be big enough to hold at least one sector of disk data (preferably more). I eventually settled upon a 4096-byte buffer cache, which holds two full CD sectors, and eight hard disk sectors. Initial reads will miss the cache and read straight from the hardware into the cache. Further reads come from that buffer in cache. For an image formatted as FAT, for example, this improves read times from the file allocation table and root directory (high-use areas) significantly.

Writing to the virtual disk makes things slightly more complicated. I decided upon an implementation, which gave the option for the virtual disk to be write-through, or totally in RAM. These options need a bit of explanation. A write-through virtual disk will place writes into cache, and write to the real disk image itself. This is an ideal setup for something like Linux-style loopback disks. However, consider the “live CD” situation: we can’t write back to a CD! So the second option writes only into the cache without affecting the original file. This means all changes are kept as long as the system is running, and do not persist. Implementing these two write options generalised the module – a massive bonus.

However, all is not well at this stage. A module to provide a disk is one part of the battle. At this stage, all we have is an abstraction of a disk – no partitions, no file systems – it’s useless for normal usage. At the time of implementation, Pedigree had no way to dynamically detect and mount such disks. This simply cannot do for a modern operating system where storage devices are hardly static – USB mass storage devices come and go, hot-pluggable hard disks exist, and so on.

I didn’t want to have my loopback disk module talk directly to the partition driver though. Exposing the internals of the partitioner to other modules makes changing the partitioner’s interface awfully complicated, and creates an explicit dependency on a specific module.

I decided instead to implement what I call Pedigree’s service manager. This kernel feature sits between different parts of the operating system and provides a standardised interface to other modules. Each service provides the following types of functionality (at the time of writing):

• Write: Send data to the module
• Read: Read data from the module
• Touch: Inform the module of new state
• Probe: Probe the module for a specific state or piece of information

Each service decides which features it provides, so it is possible for a service to provide only read as a function. The service manager takes these potential features and provides a generic interface for drivers and modules to talk to named services. In effect, this idea of services is a method of inter-process communication using named destinations. Therefore, with this new service manager, I was able to modify the partitioner to add support for the touch service. There is no need for a partitioner to support read, write, or probe, as the only notification to be sent to the partitioner is to inform it of a new disk.

With a quick modification to the loopback disk code, I was able to inform the partitioner of the presence of the new disk with error handling and no direct partitioner-specific functionality used:

// Chat to the partition service and let it pick up that we're around now
ServiceFeatures *pFeatures = ServiceManager::instance().enumerateOperations(String("partition"));
Service         *pService  = ServiceManager::instance().getService(String("partition"));
NOTICE("Asking if the partition provider supports touch");
if(pFeatures->provides(ServiceFeatures::touch))
{
    NOTICE("It does, attempting to inform the partitioner of our presence...");
    if(pService)
    {
        if(pService->serve(ServiceFeatures::touch, reinterpret_cast<void *>(this), sizeof(FileDisk)))
            NOTICE("Successful.");
        else
            ERROR("Failed.");
    }
    else
        ERROR("FileDisk: Couldn't tell the partition service about the new disk presence");
}
else
    ERROR("FileDisk: Partition service doesn't appear to support touch");

This feature has already been added to other areas of the kernel (mainly talking to the partitioner), but has the potential to even be expanded to call applications that the user runs. This means it would be theoretically possible to replace the partitioner at runtime, or replace a component of the network stack to provide a different level of service. That means that Pedigree can be modular and flexible, even though it uses the conventionally rigid “monolithic kernel” design. Now that’s something to write home about!