Once again, QEMU creates threads while running its constructors, which
is a recipe for disaster if we switch out the stack guard while that is
already running in the background.
To solve that, move initialization to our LibC initialization stage,
which is before any actual external initialization code runs.
`sigsuspend` was previously implemented using a poll on an empty set of
file descriptors. However, this broke quite a few assumptions in
`SelectBlocker`, as it verifies at least one file descriptor to be
ready after waking up and as it relies on being notified by the file
descriptor.
A bare-bones `sigsuspend` may also be implemented by relying on any of
the `sigwait` functions, but as `sigsuspend` features several (currently
unimplemented) restrictions on how returns work, it is a syscall on its
own.
This commit has no behavior changes.
In particular, this does not fix any of the wrong uses of the previous
default parameter (which used to be 'false', meaning "only replace the
first occurence in the string"). It simply replaces the default uses by
String::replace(..., ReplaceMode::FirstOnly), leaving them incorrect.
The POSIX documentation for `endgrent` only mentions that it "closes
the group database", not that it clears the backing storage for return
values. This means that applications might make use of the returned
values even after closing the group database itself. This includes our
own implementations for `getgrnam` and `getgrgid`.
The specification also states that "the storage areas might be
overwritten by a subsequent call to `getgrgid`, `getgrnam`, or
`getgrent`". This implies that `getgrgid` and `getgrnam` aren't meant
to have their own static storage and instead rely on the storage of
`getgrent`.
newlib has an extra character slot at the beginning to enable some
macro tricks that cause a warning when someone passes a type that's not
"int" into a ctype function. Our deviation from this causes issues for
LLVM.
This is a LibC function that POSIX defines to help userspace programs
to get suboptions. An example of a suboption is the token "pixclk" from
a Shell command running "edid-decode --gtf w=1024,h=768,pixclk=48".
The function should be run in a while loop to acquire all suboptions
until the last one.
* Always return 0 if `WNOHANG` is specified and no waitable child is
found, even if `wstatus` is null.
* Do not return 0 if the child is continued. Treat it the same way as
all the other states.
Refer to the RETURN VALUE section of the POSIX spec:
https://pubs.opengroup.org/onlinepubs/9699919799/functions/wait.html
Implement futimes() in terms of utimensat(). Now, utimensat() strays
from POSIX compliance because it also accepts a combination of a file
descriptor of a regular file and an empty path. utimensat() then uses
this file descriptor instead of the path to update the last access
and/or modification time of a file. That being said, its prior behavior
remains intact.
With the new behavior of utimensat(), `path` must point to a valid
string; given a null pointer instead of an empty string, utimensat()
sets `errno` to `EFAULT` and returns a failure.
Some ports linked against posix_memalign, but didn't use it, and others
used it if it was Available. So I decided to implement posix_memalign.
My implementation adds almost no overhead to regular mallocs. However,
if an alignment is specified, it will use the smallest ChunkedBlock, for
which aligned chunks exist, and simply use one of the chunks that is
aligned. If it cannot use a ChunkedBlock, for size or alignment reasons,
it will use a BigAllocationBlock, and return a pointer to the first
aligned address past the start of the block. This implementation
supports alignments up to 32768, due to the limitations of the
BigAllocationBlock technique.
Most of the string.h and wchar.h functions are implemented quite naively
at the moment, and GCC's pattern recognition pass might realize what we
are trying to do, and transform them into libcalls. This is usually a
useful optimization, but not when we're implementing the functions
themselves :^)
Relevant discussion from the GCC Bugzilla:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102725
This prevents the infamous recursive `strlen`.
A more proper fix would be writing these functions in assembly. That
would likely give a small performance boost as well ;)
The DisplayConnector class is meant to replace the FramebufferDevice
class. The advantage of this class over the FramebufferDevice class is:
1. It removes the mmap interface entirely. This interface is unsafe, as
multiple processes could try to use it, and when switching to and from
text console mode, there's no "good" way to revoke a memory mapping from
this interface, let alone when there are multiple processes that call
this interface. Therefore, in the DisplayConnector class there's no
implementation for this method at all.
2. The class uses a new real-world structure called ModeSetting, which
takes into account the fact that real hardware requires more than width,
height and pitch settings to mode-set the display resolution.
3. The class assumes all instances should supply some sort of EDID,
so it facilitates such mechanism to do so. Even if a given driver does
not know what is the actual EDID, it will ask to create default-generic
EDID blob.
3. This class shifts the responsibilies of switching between console
mode and graphical mode from a GraphicsAdapter to the DisplayConnector
class, so when doing the switch, the GraphicsManagement code actually
asks each DisplayConnector object to do the switch and doesn't rely on
the GraphicsAdapter objects at all.
This option sets -fprofile-instr-generate -fcoverage-mapping for Clang
builds only on almost all of Userland. Loader and LibTimeZone are
exempt. This can be used for generating code coverage reports, or even
PGO in the future.
This commit addresses the following shortcomings of our current, simple
and elegant memset function:
- REP STOSB/STOSQ has considerable startup overhead, it's impractical to
use for smaller sizes.
- Up until very recently, AMD CPUs didn't have support for "Enhanced REP
MOVSB/STOSB", so it performed pretty poorly on them.
With this commit applied, I could measure a ~5% decrease in `test-js`'s
runtime when I used qemu's TCG backend. The implementation is based on
the following article from Microsoft:
https://msrc-blog.microsoft.com/2021/01/11/building-faster-amd64-memset-routines
Two versions of the routine are implemented: one that uses the ERMS
extension mentioned above, and one that performs plain SSE stores. The
version appropriate for the CPU is selected at load time using an IFUNC.
IFUNC is a GNU extension to the ELF standard that allows a function to
have multiple implementations. A resolver function has to be called at
load time to choose the right one to use. The PLT will contain the entry
to the resolved function, so branching and more indirect jumps can be
avoided at run-time.
This mechanism is usually used when a routine can be made faster using
CPU features that are available in only some models, and a fallback
implementation has to exist for others.
We will use this feature to have two separate memset implementations for
CPUs with and without ERMS (Enhanced REP MOVSB/STOSB) support.
