I had a silly ambition that we would avoid unnecessary ref count churn by
forcing explicit use of "copy_ref()" wherever a copy was actually needed.
This was making RefPtr a bit clunky to work with, for no real benefit.
This patch adds the missing copy construction/assignment stuff to RefPtr.
You can currently use this to detect the CPU architecture like so:
#if ARCH(I386)
...
#elif ARCH(X86_64)
...
#else
...
#endif
This will be helpful for separating out architecture-specific code blocks.
Instead of computing the path length inside the syscall handler, let the
caller do that work. This allows us to implement to new variants of open()
and creat(), called open_with_path_length() and creat_with_path_length().
These are suitable for use with e.g StringView.
This makes me wonder if the open() syscall should take characters+length
and we'd compute the length at the LibC layer instead. That way we could
also provide an optional non-POSIX open() that takes the length directly..
This allows you to do things like:
vector.insert_before_matching(value, [](auto& entry) {
return value < entry;
});
Basically it scans until it finds an element that matches the condition
callback and then inserts the new value before the matching element.
The first implementation class is DebugLogStream, which can be used like so:
dbg() << "Hello friends, I am " << m_years << " years old!";
Note that it will automatically print a newline when the object created by
dbg() goes out of scope.
This API will grow and evolve, so let's see what we end up with :^)
Instead of manually doing String::format("%d"/"%u") everywhere, let's have
a String API for this. It's just a wrapper around format() for now, but it
could be made more efficient in the future.
Solve this by adding find() overloads to HashTable and SinglyLinkedList
that take a templated functor for comparing the values.
This allows HashMap to call HashTable::find() without having to create
a temporary Entry for use as the table key. :^)
This is prep work for supporting HashMap with NonnullRefPtr<T> as values.
It's currently not possible because many HashTable functions require being
able to default-construct the value type.
Update ProcessManager, top and WSCPUMonitor to handle the new format.
Since the kernel is not allowed to use floating-point math, we now compile
the JSON classes in AK without JsonValue::Type::Double support.
To accomodate large unsigned ints, I added a JsonValue::Type::UnsignedInt.
The LibC build is a bit complicated, since the toolchain depends on it.
During the toolchain bootstrap, after we've built parts of GCC, we have
to stop and build Serenity's LibC, so that the rest of GCC can use it.
This means that during that specific LibC build, we don't yet have access
to things like std::initializer_list.
For now we solve this by defining SERENITY_LIBC_BUILD during the LibC
build and excluding the Vector/initializer_list support inside LibC.
Get rid of the ConstIterator classes for these containers and use templated
FooIterator<T, ...> and FooIterator<const T, ...> helpers.
This makes the HashTable class a lot easier to read.
This means you can now do this:
void harmonize(NonnullRefPtrVector<Voice>& voices)
{
for (auto& voice : voices) {
voice.sing(); // Look, no "->"!
}
}
Pretty dang cool :^)
This is a slot-in convenience replacement for Vector<NonnullRefPtr<T>> that
makes accessors return T& instead of NonnullRefPtr<T>&.
Since NonnullRefPtr guarantees non-nullness, this allows you to access these
vector elements using dot (.) rather than arrow (->). :^)
This avoids putting pressure on kmalloc() during backtrace symbolication.
Since we dump backtrace for every process that exits, this is actually a
decent performance improvement for things like GCC that chain a lot of
processes together.
This parser assumes that the JSON is well-formed and will choke horribly
on invalid input.
Since we're primarily interested in parsing our own output right now, this
is less of a problem. Longer-term we're gonna need something better. :^)
- Delete the default constructor instead of just making it private.
It's never valid to create an empty NonnullRefPtr.
- Add copy assignment operators. I originally omitted these to force use
of .copy_ref() at call sites, but the hassle/gain ratio is minuscule.
- Allow calling all the assignment operators in all consumable states.
This codifies that it's okay to overwrite a moved-from NonnullRefPtr.
It's kinda funny how I can make a mistake like this in Serenity and then
get so used to it by spending lots of time using this API that I start to
believe that this is how printf() always worked..
There's no need for a member char* m_characters if we always store them
in the inline buffer. So with this patch, we now do.
After that, rearrange the members a bit for ideal packing. :^)
We'll now try to detect crashes that were due to dereferencing nullptr,
uninitialized malloc() memory, or recently free()'d memory.
It's not perfect but I think it's pretty good. :^)
Also added some color to the most important parts of the crash log,
and added some more modes to /bin/crash for exercising this code.
Fixes#243.
This patch adds JsonValue, JsonObject and JsonArray. You can use them to
build up a JsonObject and then serialize it to a string via to_string().
This patch only implements encoding, no decoding yet.
The underlying data structure is a singly-linked list of Vector<T>.
We never shift any of the vector contents around, but we batch the memory
allocations into 1000-element segments.
Put together a pretty well-performing queue using a Vector and an offset.
By using the new Vector::shift_left(int) instead of Vector::take_first()
we can avoid shifting the vector contents every time and instead only
do it every so often.
Maybe this could be generalized into a separate class, I'm not sure if it's
the best algorithm though, it's just what I came up with right now. :^)
StringView character buffer is not guaranteed to be null-terminated;
in particular it will not be null-terminated when making a substring.
This means it is not correct to check whether we've reached the end
of a StringView by comparing the next character to null; instead, we
need to do an explicit length (or pointer) comparison.
Without this function, comparing a String to a const char* will instantiate
a temporary String which is obviously not great.
Also add some missing null checks to StringView::operator==(const char*).
String&& is just not very practical. Also return const String& when the
returned string is a member variable. The call site is free to make a copy
if he wants, but otherwise we can avoid the retain count churn.
This is useful when you want to ensure some little thing happens when you
exit a certain scope.
This patch makes use of it in LibC's netdb code to make sure we close the
connection to the LookupServer.
This is a small change to the existing split() functionality to support
the case of splitting a string and stopping at a certain number of
tokens. This is useful for parsing e.g. key/value pairs, where the value
may contain the delimiter you're splitting on.
We should work towards a pattern where we take StringView as function
arguments, and store String as member, to push the String construction
to the last possible moment.
Also run it across the whole tree to get everything using the One True Style.
We don't yet run this in an automated fashion as it's a little slow, but
there is a snippet to do so in makeall.sh.
This is in preparation for eventually using it in userspace.
LinearAddress.h has not been moved for the time being (as it seems to be
only used by a very small part of the code).
The scheduler expects m_select_timeout to act as a deadline. That is, it
should contain the time that a task should wake at -- but we were
directly copying the time from userspace, which meant that it always
returned virtually immediately.
At the same time, fix CEventLoop to not rely on the broken select behavior
by subtracting the current time from the time of the nearest timer.
This makes out-of-tree linking possible. And at the same time, add a
CMakeToolchain.txt file that can be used to build arbitrary cmake-using
applications on Serenity by pointing to the CMAKE_TOOLCHAIN_FILE when
running cmake:
-DCMAKE_TOOLCHAIN_FILE=~/code/serenity/Toolchain/CMakeToolchain.txt
It makes sense to keep this consistent between applications, and the
purpose of the string is not immediately obvious from an API perspective.
If we need to make it configurable later, that can come from a setter.
* Added killall command
* Fixed feedbacks of awesomekling
* Implemented pidof program and helper to parse arguments called ArgsParser.
* Fixed feedbacks in pidof implem.
Fixes#26
This way you can spam small write()s on a file without the kernel writing
to disk every single time. Flushes are included in the FS::sync() operation
and will get triggered regularly by syncd. :^)
I just discovered the hard way that clobbering FPU/MMX/SSE registers in the
kernel makes things very confusing for userspace (and other kernel threads.)
Let's banish all of those things from the kernel to keep things simple.
It's now possible to edit widget properties inline in the properties window.
We're currently relying on the basic GVariant conversion functions to do
all the "parsing" but that's not gonna be good enough.
You open the configuration for an app like so:
auto config = CConfigFile::get_for_app("MyApp");
This will then open ~/MyApp.ini and parse it for you.
Immediately start using it in Minesweeper to load the field size and mine
count from a config file.
The window frame is an object that contains a window, its title bar and
window border. This way WSWindowManager doesn't have to know about all the
different types of window borders, titlebar rects, etc.
This is useful for static locals that never need to be destroyed:
Thing& Thing::the()
{
static Eternal<Thing> the;
return the;
}
The object will be allocated in data segment memory and will never have
its destructor invoked.
Cooperate with the compiler to generate and execute the _init_array list
of constructor functions on userspace program statup. This took two days
to get working, my goodness. :^)
The scheduler now operates on threads, rather than on processes.
Each process has a main thread, and can have any number of additional
threads. The process exits when the main thread exits.
This patch doesn't actually spawn any additional threads, it merely
does all the plumbing needed to make it possible. :^)
It was depressing not being able to capture anything when passing a lambda
to qsort_r() so let's just have our own QuickSort. I was gonna have to do
this eventually anyway. :^)
It forgot to clear out the moved-from vector. It's easy to see where this bug
came from: I assumed m_impl was an OwnPtr. It would be comfy if move() on some
arbitrary T* would also null it out but alas that's not how things work.
Since we know who's holding the lock, and we're gonna have to yield anyway,
we can just ask the scheduler to donate any remaining ticks to that process.
Also add assertion in Lock that the scheduler isn't currently active.
I've been seeing occasional fuckups that I suspect might be someone called
by the scheduler trying to take a busy lock.
This only works with the userspace build of SharedGraphics so far.
It's also very slow at loading fonts, but that's easy to fix.
Let's put fonts in /res/fonts/.
GObjects can now register a timer with the GEventLoop. This will eventually
cause GTimerEvents to be dispatched to the GObject.
This needed a few supporting changes in the kernel:
- The PIT now ticks 1000 times/sec.
- select() now supports an arbitrary timeout.
- gettimeofday() now returns something in the tv_usec field.
With these changes, the clock window in guitest2 finally ticks on its own.
This required a fair bit of plumbing. The CharacterDevice::close() virtual
will now be closed by ~FileDescriptor(), allowing device implementations to
do custom cleanup at that point.
One big problem remains: if the master PTY is closed before the slave PTY,
we go into crashy land.
You're never gonna be right 100% of the time when guessing how much buffer
space you need. This avoids having to make that type of decision in a bunch
of cases. :^)
Only raw octal modes are supported right now.
This patch also changes mode_t from 32-bit to 16-bit to match the on-disk
type used by Ext2FS.
I also ran into EPERM being errno=0 which was confusing, so I inserted an
ESUCCESS in its place.
Okay, now ProcFS doesn't crash due to the crappy buffer size estimates
not really working out. This thing has dogshit performance and I will
fix that separately.
It was fun for everyone to share a single framebuffer but it was also
kinda really awful. Let's move towards having a "GraphicsBitmap" as the
backing store for each Window.
This is going to need a lot of refactoring so let's get started.
The inode cache was keeping these alive forever. Added a cute little magic
trick to Retainable that calls T::one_retain_left() when the retain count
is decremented to 1.
This container is really just there to keep a retain on the individual
PhysicalPages for each page table. A HashMap does the job with far greater
space efficiency.
Use a little template magic to have Retainable::release() call out to
T::will_be_destroyed() if such a function exists before actually calling
the destructor. This gives us full access to virtual functions in the
pre-destruction code.
The kernel now bills processes for time spent in kernelspace and userspace
separately. The accounting is forwarded to the parent process in reap().
This makes the "time" builtin in bash work.
Clang demands that the size argument to the various operator new()'s
to be exactly whatever it thinks std::size_t is.
Since we can't include STL headers, this little trick will do.
This way the scheduler doesn't need to plumb the exit status into the waiter.
We still plumb the waitee pid though, I don't love it but it can be fixed.
I feel like this concept might be useful in more places. It's very naive
right now and uses dynamically growing buffers. It should really use static
size buffers and never kmalloc().
...by adding a new class called Ext2Inode that inherits CoreInode.
The idea is that a vnode will wrap a CoreInode rather than InodeIdentifier.
Each CoreInode subclass can keep whatever caches they like.
Right now, Ext2Inode caches the list of block indices since it can be very
expensive to retrieve.
I was surprised to find that dup()'ed fds don't share the close-on-exec flag.
That means it has to be stored separately from the FileDescriptor object.
Pass the file name in a stack-allocated buffer instead of using an AK::String
when iterating directories. This dramatically reduces the amount of cycles
spent traversing the filesystem.
- Process::exec() needs to restore the original paging scope when called
on a non-current process.
- Add missing InterruptDisabler guards around g_processes access.
- Only flush the TLB when modifying the active page tables.
First of all, change sys$mmap to take a struct SC_mmap_params since our
sycsall calling convention can't handle more than 3 arguments.
This exposed a bug in Syscall::invoke() needing to use clobber lists.
It was a bit confusing to debug. :^)
sys$fork() now clones all writable regions with per-page COW bits.
The pages are then mapped read-only and we handle a PF by COWing the pages.
This is quite delightful. Obviously there's lots of work to do still,
and it needs better data structures, but the general concept works.
This was the fix:
-process.m_page_directory[0] = m_kernel_page_directory[0];
-process.m_page_directory[1] = m_kernel_page_directory[1];
+process.m_page_directory->entries[0] = m_kernel_page_directory->entries[0];
+process.m_page_directory->entries[1] = m_kernel_page_directory->entries[1];
I spent a good two hours scratching my head, not being able to figure out why
user process page directories felt they had ownership of page tables in the
kernel page directory.
It was because I was copying the entire damn kernel page directory into
the process instead of only sharing the two first PDE's. Dang!
The SpinLock was all backwards and didn't actually work. Fixing it exposed
how wrong most of the locking here is.
I need to come up with a better granularity here.
This shows some info about the MM. Right now it's just the zone count
and the number of free physical pages. Lots more can be added.
Also added "exit" to sh so we can nest shells and exit from them.
I also noticed that we were leaking all the physical pages, so fixed that.
I also added a generator cache to FileHandle. This way, multiple
reads to a generated file (i.e in a synthfs) can transparently
handle multiple calls to read() without the contents changing
between calls.
The cache is discarded at EOF (or when the FileHandle is destroyed.)
It's implemented as a separate process. How cute is that.
Tasks now have a current working directory. Spawned tasks inherit their
parent task's working directory.
Currently everyone just uses "/" as there's no way to chdir().
I added a dead-simple malloc that only allows allocations < 4096 bytes.
It just forwards the request to mmap() every time.
I also added simplified versions of opendir() and readdir().
Add a separate lock to protect the VFS. I think this might be a good idea.
I'm not sure it's a good approach though. I'll fiddle with it as I go along.
It's really fun to figure out all these things on my own.
- Turn Keyboard into a CharacterDevice (85,1) at /dev/keyboard.
- Implement MM::unmapRegionsForTask() and MM::unmapRegion()
- Save SS correctly on interrupt.
- Add a simple Spawn syscall for launching another process.
- Move a bunch of IO syscall debug output behind DEBUG_IO.
- Have ASSERT do a "cli" immediately when failing.
This makes the output look proper every time.
- Implement a bunch of syscalls in LibC.
- Add a simple shell ("sh"). All it can do now is read a line
of text from /dev/keyboard and then try launching the specified
executable by calling spawn().
There are definitely bugs in here, but we're moving on forward.
