This involves refactoring VirtIOConsole into VirtIOConsole and
VirtIOConsolePort. VirtIOConsole is the VirtIODevice, it owns multiple
VirtIOConsolePorts as well as two control queues. Each
VirtIOConsolePort is a CharacterDevice.
By making sure the PhysicalPage instance is fully destructed the
allocators will have a chance to reclaim the PhysicalPageEntry for
free-list purposes. Just pass them the physical address of the page
that was freed, which is enough to lookup the PhysicalPageEntry later.
By moving the PhysicalPage classes out of the kernel heap into a static
array, one for each physical page, we can avoid the added overhead and
easily find them by indexing into an array.
This also wraps the PhysicalPage into a PhysicalPageEntry, which allows
us to re-use each slot with information where to find the next free
page.
We already use PAE for the NX bit, but this changes the PhysicalAddress
structure to be able to hold 64 bit physical addresses. This allows us
to use all the available physical memory.
Userland faulted on the very first instruction before because the
PML4T/PDPT/etc. weren't marked as user-accessible. For some reason
x86 doesn't care about that.
Also, we need to provide an appropriate userspace stack segment
selector to iretq.
This commit converts naked `new`s to `AK::try_make` and `AK::try_create`
wherever possible. If the called constructor is private, this can not be
done, so we instead now use the standard-defined and compiler-agnostic
`new (nothrow)`.
This adds just enough stubs to make the kernel compile on x86_64. Obviously
it won't do anything useful - in fact it won't even attempt to boot because
Multiboot doesn't support ELF64 binaries - but it gets those compiler errors
out of the way so more progress can be made getting all the missing
functionality in place.
This avoids allocations in the VMObject constructor. The number of
inline elements was determined empirically and covers most common cases
including LibC malloc.
Now that Region::name() has been changed to return a StringView we
can't rely on keeping a copy of the region's name past the region's
destruction just by holding a copy of the StringView.
- Make Region::create_kernel_only OOM safe.
- Make Region::create_user_accessible mostly OOM safe, there are still
some tendrils to untangle before it and be completely fixed.
Propagate allocation failure of m_shared_committed_cow_pages,
and uncommit previously committed COW pages on failure.
This method needs a closer look in terms of error handling, as we
will eventually need to rollback all changes on allocation failure.
Alternatively we could allocate the anonymous object much earlier
and only initialize it once the other steps have succeeded.
Replace the AK::String used for Region::m_name with a KString.
This seems beneficial across the board, but as a specific data point,
it reduces time spent in sys$set_mmap_name() by ~50% on test-js. :^)
Spinlocks guard short regions, with hopefully no other locks being taken
in the process. Violating constraints usually had detrimental effects on
platform stability as well as performance and scalability. Allocating
memory takes it own locks, and can in some cases even allocate new
regions, and thus violates these tenants.
Move the AnonymousVMObject creation outside of the spinlock as
creation does not modify any shared state.
Problem:
- `static` variables consume memory and sometimes are less
optimizable.
- `static const` variables can be `constexpr`, usually.
- `static` function-local variables require an initialization check
every time the function is run.
Solution:
- If a global `static` variable is only used in a single function then
move it into the function and make it non-`static` and `constexpr`.
- Make all global `static` variables `constexpr` instead of `const`.
- Change function-local `static const[expr]` variables to be just
`constexpr`.
By constraining two implementations, the compiler will select the best
fitting one. All this will require is duplicating the implementation and
simplifying for the `void` case.
This constraining also informs both the caller and compiler by passing
the callback parameter types as part of the constraint
(e.g.: `IterationFunction<int>`).
Some `for_each` functions in LibELF only take functions which return
`void`. This is a minimal correctness check, as it removes one way for a
function to incompletely do something.
There seems to be a possible idiom where inside a lambda, a `return;` is
the same as `continue;` in a for-loop.
AnonymousVMObject::create_with_physical_page(s) can't be NonnullRefPtr
as it allocates internally. Fixing the API then surfaced an issue in
ScatterGatherList, where the code was attempting to create an
AnonymousVMObject in the constructor which will not be observable
during OOM.
Fix all of these issues and start propagating errors at the callers
of the AnonymousVMObject and ScatterGatherList APis.
Currently, when passing buffers into VirtIOQueues, we use scatter-gather
lists, which contain an internal vector of buffers. This vector is
allocated, filled and the destroy whenever we try to provide buffers
into a virtqueue, which would happen a lot in performance cricital code
(the main transport mechanism for certain paravirtualized devices).
This commit moves it over to using VirtIOQueueChains and building the
chain in place in the VirtIOQueue. Also included are a bunch of fixups
for the VirtIO Console device, making it use an internal VM::RingBuffer
instead.
We want to move this out of the AHCI subsystem into the VM system,
since other parts of the kernel may need to perform scatter-gather IO.
We rename the current VM::ScatterGatherList impl that's used in the
virtio subsystem to VM::ScatterGatherRefList, since its distinguishing
feature from the AHCI scatter-gather list is that it doesn't own its
buffers.
The error handling in all these cases was still using the old style
negative values to indicate errors. We have a nicer solution for this
now with KResultOr<T>. This change switches the interface and then all
implementers to use the new style.
We had some inconsistencies before:
- Sometimes "The", sometimes "the"
- Sometimes trailing ".", sometimes no trailing "."
I picked the most common one (lowecase "the", trailing ".") and applied
it to all copyright headers.
By using the exact same string everywhere we can ensure nothing gets
missed during a global search (and replace), and that these
inconsistencies are not spread any further (as copyright headers are
commonly copied to new files).
SPDX License Identifiers are a more compact / standardized
way of representing file license information.
See: https://spdx.dev/resources/use/#identifiers
This was done with the `ambr` search and replace tool.
ambr --no-parent-ignore --key-from-file --rep-from-file key.txt rep.txt *
This allows converting a single virtual buffer into its non-physically
contiguous parts, this is especially useful for DMA-based devices that
support scatter/gather-like functionality, as it eliminates the need
to clone outgoing buffers into one physically contiguous buffer.
This should provide some speed up, as currently searches for regions
containing a given address were performed in O(n) complexity, while
this container allows us to do those in O(logn).
This does not affect functionality right now, but it means that the
regions vector will now never have any overlapping regions, which will
allow the use of balance binary search trees instead of a vector in the
future. (since they require keys to be exclusive)
Alot of code is shared between i386/i686/x86 and x86_64
and a lot probably will be used for compatability modes.
So we start by moving the headers into one Directory.
We will probalby be able to move some cpp files aswell.
Increase type-safety moving the MemoryManager APIs which take a
Region::Access to actually use that type instead of a `u8`.
Eventually the actually m_access can be moved there as well, but
I hit some weird bug where it wasn't using the correct operators
in `set_access_bit(..)` even though it's declared (and tested).
Something to fix-up later.
As it turns out, Dr. POSIX doesn't require that post-mmap() changes
to a file are reflected in the memory mappings. So we don't actually
have to care about the file size changing (or the contents.)
IIUC, as long as all the MAP_SHARED mappings that refer to the same
inode are in sync, we're good.
This means that VMObjects don't need resizing capabilities. I'm sure
there are ways we can take advantage of this fact.
Add Bitmap::view() and forward most of the calls to BitmapView since
the code was identical.
Bitmap is now primarily concerned with its dynamically allocated
backing store and BitmapView deals with the rest.
Instead of keeping AnonymousVMObject::m_cow_map in an OwnPtr<Bitmap>,
just make the Bitmap a regular value member. This increases the size
of the VMObject by 8 bytes, but removes some of the kmalloc/kfree spam
incurred by sys$fork().
Since we know for sure that the virtual memory regions in the new
process being created are not being used on any CPU, there's no need
to do TLB flushes for every mapped page.
This may seem like a no-op change, however it shrinks down the Kernel by a bit:
.text -432
.unmap_after_init -60
.data -480
.debug_info -673
.debug_aranges 8
.debug_ranges -232
.debug_line -558
.debug_str -308
.debug_frame -40
With '= default', the compiler can do more inlining, hence the savings.
I intentionally omitted some opportunities for '= default', because they
would increase the Kernel size.
We were only 448 KiB away from filling up the old slot size we reserve
for the kernel above the 3 GiB mark. This expands the slot to 16 MiB,
which allows us to continue booting the kernel until somebody takes
the time to improve our loader.
(...and ASSERT_NOT_REACHED => VERIFY_NOT_REACHED)
Since all of these checks are done in release builds as well,
let's rename them to VERIFY to prevent confusion, as everyone is
used to assertions being compiled out in release.
We can introduce a new ASSERT macro that is specifically for debug
checks, but I'm doing this wholesale conversion first since we've
accumulated thousands of these already, and it's not immediately
obvious which ones are suitable for ASSERT.
There's no real system here, I just added it to various functions
that I don't believe we ever want to call after initialization
has finished.
With these changes, we're able to unmap 60 KiB of kernel text
after init. :^)
You can now declare functions with UNMAP_AFTER_INIT and they'll get
segregated into a separate kernel section that gets completely
unmapped at the end of initialization.
This can be used for anything we don't need to call once we've booted
into userspace.
There are two nice things about this mechanism:
- It allows us to free up entire pages of memory for other use.
(Note that this patch does not actually make use of the freed
pages yet, but in the future we totally could!)
- It allows us to get rid of obviously dangerous gadgets like
write-to-CR0 and write-to-CR4 which are very useful for an attacker
trying to disable SMAP/SMEP/etc.
I've also made sure to include a helpful panic message in case you
hit a kernel crash because of this protection. :^)
You can now use the READONLY_AFTER_INIT macro when declaring a variable
and we will put it in a special ".ro_after_init" section in the kernel.
Data in that section remains writable during the boot and init process,
and is then marked read-only just before launching the SystemServer.
This is based on an idea from the Linux kernel. :^)
If we try to align a number above 0xfffff000 to the next multiple of
the page size (4 KiB), it would wrap around to 0. This is most likely
never what we want, so let's assert if that happens.
Now that we no longer need to support the signal trampolines being
user-accessible inside the kernel memory range, we can get rid of the
"kernel" and "user-accessible" flags on Region and simply use the
address of the region to determine whether it's kernel or user.
This also tightens the page table mapping code, since it can now set
user-accessibility based solely on the virtual address of a page.
This patch adds Space, a class representing a process's address space.
- Each Process has a Space.
- The Space owns the PageDirectory and all Regions in the Process.
This allows us to reorganize sys$execve() so that it constructs and
populates a new Space fully before committing to it.
Previously, we would construct the new address space while still
running in the old one, and encountering an error meant we had to do
tedious and error-prone rollback.
Those problems are now gone, replaced by what's hopefully a set of much
smaller problems and missing cleanups. :^)
This patch adds sys$msyscall() which is loosely based on an OpenBSD
mechanism for preventing syscalls from non-blessed memory regions.
It works similarly to pledge and unveil, you can call it as many
times as you like, and when you're finished, you call it with a null
pointer and it will stop accepting new regions from then on.
If a syscall later happens and doesn't originate from one of the
previously blessed regions, the kernel will simply crash the process.
The random address proposals were not checked with the size so it was
increasely likely to try to allocate outside of available space with
larger and larger sizes.
Now they will be ignored instead of triggering a Kernel assertion
failure.
This is a continuation of: c8e7baf4b8
This patch adds enforcement of two new rules:
- Memory that was previously writable cannot become executable
- Memory that was previously executable cannot become writable
Unfortunately we have to make an exception for text relocations in the
dynamic loader. Since those necessitate writing into a private copy
of library code, we allow programs to transition from RW to RX under
very specific conditions. See the implementation of sys$mprotect()'s
should_make_executable_exception_for_dynamic_loader() for details.
This can be used to request random VM placement instead of the highly
predictable regular mmap(nullptr, ...) VM allocation strategy.
It will soon be used to implement ASLR in the dynamic loader. :^)
We need to make sure other processors can grab the MM lock while we
wait, so release it when we might block. Reading the page from
disk may also block, so release it during that time as well.
This eliminates the window between calling Processor::current and
the member function where a thread could be moved to another
processor. This is generally not as big of a concern as with
Processor::current_thread, but also slightly more light weight.
If we find ourselves with a user-accessible, non-shared Region backed by
a SharedInodeVMObject, that's pretty bad news, so let's just panic the
kernel instead of getting abused.
There might be a better place for this kind of check, so I've added a
FIXME about putting more thought into that.
This was exploitable since the shared flag determines whether inode
permission checks are applied in sys$mprotect().
The bug was pretty hard to spot due to default arguments being used
instead. This patch removes the default arguments to make explicit
at each call site what's being done.
This was done with the help of several scripts, I dump them here to
easily find them later:
awk '/#ifdef/ { print "#cmakedefine01 "$2 }' AK/Debug.h.in
for debug_macro in $(awk '/#ifdef/ { print $2 }' AK/Debug.h.in)
do
find . \( -name '*.cpp' -o -name '*.h' -o -name '*.in' \) -not -path './Toolchain/*' -not -path './Build/*' -exec sed -i -E 's/#ifdef '$debug_macro'/#if '$debug_macro'/' {} \;
done
# Remember to remove WRAPPER_GERNERATOR_DEBUG from the list.
awk '/#cmake/ { print "set("$2" ON)" }' AK/Debug.h.in
The kernel ignored the first 8 MiB of RAM while parsing the memory map
because the kmalloc heaps and the super physical pages lived here. Move
all that stuff inside the .bss segment so that those memory regions are
accounted for, otherwise we risk overwriting boot modules placed next
to the kernel.
This adds support for FUTEX_WAKE_OP, FUTEX_WAIT_BITSET, FUTEX_WAKE_BITSET,
FUTEX_REQUEUE, and FUTEX_CMP_REQUEUE, as well well as global and private
futex and absolute/relative timeouts against the appropriate clock. This
also changes the implementation so that kernel resources are only used when
a thread is blocked on a futex.
Global futexes are implemented as offsets in VMObjects, so that different
processes can share a futex against the same VMObject despite potentially
being mapped at different virtual addresses.
Problem:
- Many constructors are defined as `{}` rather than using the ` =
default` compiler-provided constructor.
- Some types provide an implicit conversion operator from `nullptr_t`
instead of requiring the caller to default construct. This violates
the C++ Core Guidelines suggestion to declare single-argument
constructors explicit
(https://isocpp.github.io/CppCoreGuidelines/CppCoreGuidelines#c46-by-default-declare-single-argument-constructors-explicit).
Solution:
- Change default constructors to use the compiler-provided default
constructor.
- Remove implicit conversion operators from `nullptr_t` and change
usage to enforce type consistency without conversion.
These changes are arbitrarily divided into multiple commits to make it
easier to find potentially introduced bugs with git bisect.Everything:
The modifications in this commit were automatically made using the
following command:
find . -name '*.cpp' -exec sed -i -E 's/dbg\(\) << ("[^"{]*");/dbgln\(\1\);/' {} \;
If a TLB flush request is broadcast to other processors and the addresses
to flush are user mode addresses, we can ignore such a request on the
target processor if the page directory currently in use doesn't match
the addresses to be flushed. We still need to broadcast to all processors
in that case because the other processors may switch to that same page
directory at any time.
If we remap pages (e.g. lazy allocation) inside a VMObject that is
shared among more than one region, broadcast it to any other region
that may be mapping the same page.