In a syscall microbench this improves performance by ~19% on my
Snapdragon 888.
Going from ~9.6 million syscalls per second to ~11.5 million.
Macbook Pro is less effective here due to high syscall overhead due to
VM. Going form 7.2M/s to 7.6M/s, ~6% improvement
We can also inline some 32-bit syscalls but that will need some more
work which isn't done yet. Even though the op in the JIT supports it.
Ran in to this when removing syscall nodes. If an IR op has side-effects
then this generic helper can not remove them.
First time this was encountered and it was confusing
These aren't strictly necessary for enum flags and through discussion in
\#1363, would lead to an awkward to use overload.
If these are ever needed, they can be added back at a later date.
Only a partial fix for #1330, still needs preemption disabled to work.
On x86-64 hosts the Linux kernel resides in the top bit of VA which
isn't mapped in to userspace.
This means that userspace will never receive pointers living with that
top bit set unless you're running a 57bit VA host.
This results in userspace pointers never needing to do the sign
extending pointer canonicalization. But additionally some applications
actually don't understand the pointer canonicalization.
This results in bugs like: https://github.com/golang/go/issues/49405
Now if you're running on a 57bit VA host, this will end up behaving like
FEX but it seems like no one in golang land has really messed with 57bit
VA yet.
In AArch64, when configured with a 48bit VA, the userspace gets the full
48bit VA space and on EL mode switch has the full address range change
to the kernel's 48bit VA.
This means that we will /very/ likely allocate pointers in the high
48bit space since Linux currently allocates top-down.
So behave more like x86-64, hide the top 128TB of memory space from the
guest before boot.
Testing: Took the M1Max 15ms to 21ms allocate the top 128TB.
We've run out of the range of 32 bits already and will need to use
another flag in upcoming changes, so we need to expand our flags to be
64-bit.
While we're at it, we can use a dedicated type alias for the instruction
flags to make the interface changeable from one spot in the future.
There's zero need for these to have an independent size field and it was
just confusing.
For stores it was always set to zero and for loads it was just
duplicated.
In addition this allows introspection of the store op without casting
the op, which can be useful in edge cases
These only perform bit arithmetic, so we can allow them to be used in
constexpr contexts.
This allows clang to collapse quite a bit of code for the table
initializing functions. For example, in InitializeVEXTables(),
with these as inline (but not constexpr) functions, clang will
individually put all of the table entries onto the stack.
With these as constexpr functions, clang will be able to deduce that it
can construct the tables at compile time and reduces the amount of
generated code quite a bit.
There were two issues:
1. The OrderedNode *After overload created the wrong iterator
2. Despite being named After, they were actually inclusive
Kept ReplaceAllUsesWithRange with the current inclusive behaviour and
adjusted the argument name to match
In the case of running inside of a container then we need to redirect
where we look for configuration.
Currently we only care about pressure vessel so we just redirect some
options to check inside of `/run/host/`
This will resolve an issue where installed thunks wouldn't be found.
We were using the glibc context structure layout which doesn't match
what the kernel is doing.
Switches over to allocating fpstate independentally of the ucontext_t,
then pointing to it from uc_mcontext how we're supposed to.
glibc uses the __fpregs_mem region for other purposes.
This also fixes xmm state being stored on the 32-bit side, and also
fixes SIGALRM and SIGVTALRM siginfo overwriting structure data.
On 32-bit we can't just memcpy the siginfo_t over because the sizes
don't match. Throw a message instead.
This isn't quite a 100% clean sweep of IWYU.
There are some false positives where clang fails.
Additionally there are still a few missed in the frontend side of things
that I didn't get to
This is working towards getting the stack frame for guest signals being
pushed over to the Frontend.
Still some more work to do but this is the first step that can be split
up.
Hangover currently abuses the syscall op for thunking purposes.
Very likely this will be changed in the future but for now make sure to
support that use case.
This was using the implicit thread TLS object. All users of this
have access to the thread object directly.
Use that instead. Fixes a subtle bug were the frontend could be trying
to do a callback and TLS sections weren't correctly set.
This always returns success and for easier state management, just return
our parent thread.
Frontend needs full visibility of this state anyway for thread
management.
Puts the visibility of the main layer, application layers, and
environment in to FEXCore instead of FEX.
These layers aren't specific to FEX/FEXLoader and should live in
FEXCore.
Only the EmptyMapper remains in FEX, which should eventually move over
to FEXConfig since that is the only user.
This is very tricky to handle and it has a bunch of rough edges.
One of the major problems that we can't workaround is that if we receive a
clone flag that pthreads can't support with THREAD, then we are required to fall down
the pthreads code path.
This is because threads going down the clone path will break TLS and we don't have
a way to work around it currently.
So this adds a clone path, a clone3 path, and keeps the legacy path as well.
Which makes this fairly convoluted but it gets pressure-vessel working on x86-64 host.
It's a bit tricky to setup but it does work.
Still some work necessary to get pressure-vessel working on AArch64 host, but I'm working on that.
These were causing static initializer construction all over the codebase.
Scope of these variables has also been changed over to the module instead of the header as well
This information is only ever going to be offline. Will be useful for multiple reasons.
1) Searching for split lock usage in applications, which can be a programming bug.
a) This isn't visible on AMD systems and on Intel is a fairly new linux feature
2) Having more information about when an application breaks.
3) Useful for some minor profiling for devs looking for statistical data
A decent amount of this was broken.
To the point that the ABI was incorrect even, causing 32-bit signals to completely crash in most instances.
This is enough to get my 32-bit signal handler test application to work and basic SIGUSR behaviour
vsyscall is a legacy interface to expose three syscalls in a faster fashion.
Most applications should have moved to vDSO but this isn't entirely the case.
golang specifically still used vsyscall in the case that vdso isn't available.
Which that isn't completely safe since vsyscall can be completely disabled.
This is fixed in a newer version of their runtime but we must support legacy behaviour.
This makes it so FEX checks if the incoming RIP is a "special" region and adjusts where it
loads instructions from accordingly. Lying to the guest application about the instruction stream
while maintaining the illusion that its RIP is in the correct location.
This was fairly straightforward since we already decoupled the instruction stream from the RIP location.
This fixes golang execution as long as you also disable their preempt signaling.
environment variable to do so is `GODEBUG=asyncpreemptoff=1`
Prior to the introduction of fmt, there wasn't a nice and easy way to
format std::string_view, but now that we have it in place, we can modify
relevant log calls to make use of it and get rid of the need to
construct a string around the view.
While we're in the same area, we can make the rest of the modified file
use fmt where appropriate.
Only the frontends need to deal with ELF files specifically.
The backend doesn't need to be aware of them at all.
Since the ELF handling is the frontend's responsibility, move all the code to the frontend.
I saw a red herring that I thought the high cpu usage in steamwebhelper could come from signal handlers.
This turned out to not be the case, but now I've got this implemented.
Installs the few signal handlers that we need upfront but for everything that isn't a mandatory signal
we instead now wait until the guest also installs that signal handler.
This fixes#1107
Ryan Houdek