Instead of generating symbols for everything in gl.xml, read the symbol
list from src/GL/gl.symbols and only generate the functions listed
there.
That way, when new functions get added to gl.xml, we won't add any more
symbols in libGL.so.
To do that, if the target parameter to gen_gldispatch_mapi.py is a
filename, then it will read that file for symbols instead of calling
genCommon.getExportNamesFromRoots.
Also added a new genCommon.readSymbolsFile function, which is mostly
copied from symbols-check.py.
Change the ARMv7 dispatch stubs to use ARM instead of Thumb. There's no
difference in size (since either one fits in the 128-byte alignment that
we use), but using ARM means that they'll compile and work on an ARMv6
build as well.
In __glDispatchMakeCurrent, it allocates a
__GLdispatchThreadStatePrivate struct, which would normally get freed in
__glDispatchLoseCurrent.
However, if the library gets unloaded while another thread still has a
current context, then __glDispatchLoseCurrent never gets called, and so
that memory leaks.
Add the __GLdispatchThreadStatePrivate to a linked list so that it can
free any remaining structs in __glDispatchFini.
Fixes https://gitlab.freedesktop.org/glvnd/libglvnd/-/issues/230.
In the EGL destructor, DestroyThreadState gets called by
__eglCurrentTeardown to free the threadState, but the threadState does
not get changed to NULL when this happens.
The function eglReleaseThread calls __eglDestroyCurrentThreadAPIState,
which does set the threadState to NULL, right before calling
DestroyThreadState. Both eglReleaseThread and
__eglDestroyCurrentThreadAPIState check to make sure the threadState is
NULL before continuing.
If eglReleaseThread gets externally called after destruction, the
threadState is not NULL, so DestroyThreadState will be called to free
the threadState that was already freed by the destructor, causing a
double free.
This change is to add the same setting threadState to NULL in
__eglCurrentTeardown as which already __eglDestroyCurrentThreadAPIState,
to prevent this double free situation.
When Branch Target Identifier (BTI) is enabled on aarch64,
any software which run libglvnd will fail with SIGILL, Illegal instruction.
This is because some assembler code misses the BTI landing pads.
See: https://developer.arm.com/documentation/102433/0100/Jump-oriented-programming
"hint #34" is the same thing as "BTI C" landing pad, but keep
compatibility with systems without BTI enabled.
Updated the EGL headers and XML file to the Khronos repository, from commit
0997d5398d9de08ac012e7230d6546eb342c2939.
Signed-off-by: Alex Richardson <Alexander.Richardson@cl.cam.ac.uk>
There are KHR, EXT, and NV versions of eglQueryDisplayAttrib, so add
support for all of them.
Define separate eglQueryDisplayAttribEXT, eglQueryDisplayAttribKHR, and
eglQueryDisplayAttribNV functions. They all call into the same common
function, with the only difference being the name passed to an error
callback.
When loading a vendor library, check for the vendor's
eglQueryDisplayAttrib using all three suffixes, and use whichever one is
available.
Add a dispatch stub for eglQueryDisplayAttribEXT, which updates the device
hashtable if the application queries EGL_DEVICE_EXT.
It's possible that eglQueryDisplayAttribEXT could return an EGLDeviceEXT handle
that wasn't returned by eglQueryDevicesEXT, so the dispatch stub needs to add
the handle to the device hashtable.
Export a new __EGLapiExports::setVendorForDevice function, which allows
dispatch stubs to update the device hashtable when something returns an
EGLDeviceEXT handle.
Bumped the ABI version number to 1.2.
It's possible for new devices to become available while a program is running,
so eglQueryDevicesEXT might need to return a different list.
Change the EGLDeviceEXT hashtable so that each entry is allocated and added
separately, rathat than allocating every entry in a single malloc.
Added a new __eglAddDevice function, which adds an entry to the device
hashtable.
When the application calls eglQueryDevicesEXT, always call through to each
vendor library, and add (or update) each EGLDeviceEXT to the hashtable.
There's still no provision for removing an entry from the device hashtable. The
semantics of hot-unplugging a device aren't defined yet, and there's no way to
ensure that libglvnd's mapping will get updated if a device disappears
asynchonously, so it's up to the vendor library to deal with the old handle.
If x11 support is enabled, then the meson and configure scripts will set
a macro named ENABLE_EGL_X11 instead of USE_X11.
USE_X11 will also select the Xlib typedef of EGLNativeDisplayType in
eglplatform.h, and libglvnd does not need or want those.
Enabling or disabling X11 support for EGL only affects platform
detection in eglGetDisplay. The rest of libEGL is supposed to treat
EGLNativeDisplayType as an opaque void* pointer.
Also delete files and folders that are only useful in the original
repository. Updating uthash to this version allows compiling with
CHERI LLVM where the current version of uthash (1.9.8) triggers a
warning that our build infrastructure includes in -Werror by default:
```
../../libglvnd/src/util/winsys_dispatch.c:154:9: error: cast from provenance-free integer type to pointer type will give pointer that can not be dereferenced [-Werror,-Wcheri-capability-misuse]
../../libglvnd/include/lkdhash.h:86:5: note: expanded from macro 'LKDHASH_TEARDOWN'
LKDHASH_TEARDOWN_2(_lh, _param, cur ## _ht, \
^
../../libglvnd/include/lkdhash.h:55:9: note: expanded from macro 'LKDHASH_TEARDOWN_2'
HASH_DEL(_LH(_lockedhash), _cur); \
^
../../libglvnd/src/util/uthash/src/uthash.h:271:5: note: expanded from macro 'HASH_DEL'
HASH_DELETE(hh,head,delptr)
^
../../libglvnd/src/util/uthash/src/uthash.h:239:14: note: expanded from macro 'HASH_DELETE'
((UT_hash_handle*)((ptrdiff_t)_hd_hh_del->next + \
^
```
Signed-off-by: Alex Richardson <Alexander.Richardson@cl.cam.ac.uk>
I missed this in 396c10bdd7 since I was
building with -Dglx=disabled. This fixes the build with glx if the X11
libraries are installed into a custom prefix.
Signed-off-by: Alex Richardson <Alexander.Richardson@cl.cam.ac.uk>
This fixes the build for me since I have libX11 installed to a custom
prefix, so the default include path does not include libX11. This fixes
lots of instances of the following error:
../../libglvnd/include/EGL/eglplatform.h:109:10: fatal error: 'X11/Xlib.h' file not found
Signed-off-by: Alex Richardson <Alexander.Richardson@cl.cam.ac.uk>
Remove the "If only executable code is distributed..." paragraph from
the license text. Everything now uses a normal MIT license.
The only code from Khronos that's included in libglvnd is the EGL/GL
header and XML files, which do not contain that paragraph.
Fixes https://gitlab.freedesktop.org/glvnd/libglvnd/-/issues/221
LLVM currently ignores inline assembly references to symbols during
dead code and data elimination while linking with link-time optimizations
enabled.
This manifests in a very large amount of unresolved references to
entrypointFunctions when building libglvnd using clang with LTO enabled.
This issue fixes#220 by marking the entrypointFunctions array as used,
using a compiler attribute.
Signed-off-by: Charlotte Delenk <darkkirb@darkkirb.de>
It is not portable to use initial-exec TLS in dlopened libraries. glibc
and FreeBSD allocate extra memory for extra initial-exec variables
specifically for libGL, but other libcs including musl do not.
Since TLS entry asm assumes IE TLS, use TSD asm in other cases. Update
autoconf to match meson logic: enable ELF TLS if it is supported,
regardless of which type of asm is being used.
Previously, relative ICD paths were treated as being relative to the
current working directory, but this doesn't seem useful in practice:
EGL vendors cannot predict what the current working directory of EGL
applications is going to be.
Instead, interpret the path as being relative to the realpath() of the
JSON manifest describing the ICD.
Signed-off-by: Simon McVittie <smcv@collabora.com>
Resolves: https://gitlab.freedesktop.org/glvnd/libglvnd/-/issues/190
When libGLX.so is unloaded, call XESetCloseDisplay to remove the display close
callback for any displays that still exist.
Otherwise, if libGLX.so is loaded using dlopen, and then unloaded while a
display is still open, then XCloseDisplay would try to call the (now dangling)
callback.
Added a --disable-entrypoint-tracking configure option and an
'entrypoint-patching' meson option to disable libGLdispatch's entrypoint
patching at build time.
If entrypoint patching is disabled, then it #ifdef's out the mprotect call, and
acts as if mprotect had failed, which causes libGLdispatch to skip trying to
perform any patching.
Fixes https://gitlab.freedesktop.org/glvnd/libglvnd/-/issues/217
In OnDisplayClosed, unlock the __glXDisplayInfoHash lock before locking
glxContextHashLock.
In CommonMakeCurrent, it holds glxContextHashLock while it tries to take the
__glXDisplayInfoHash lock, so if CommonMakeCurrent and OnDisplayClosed run at
the same time, they could deadlock.
To avoid that, OnDisplayClosed will set a new inTeardown flag in the
__GLXdisplayInfoHash, then it will unlock the display hash before calling
__glXDisplayClosed.
After __glXDisplayClosed is finished, OnDisplayClosed will lock
__glXDisplayInfoHash again long enough to remove the display from the
hashtable.
The inTeardown flag tells __glXLookupDisplay to return NULL, since after that
point, nothing should be trying to look up the display. In principle, we could
just remove the display from the hashtable up front, but then if something did
try to look up the display, then __glXLookupDisplay would try to create a new
__GLXdisplayInfo for it.
Fixes https://gitlab.freedesktop.org/glvnd/libglvnd/-/issues/213
Currently, in InitDeviceListInternal if any egl vendor fails to list
its devices for any reason the construction of the device list is
abandoned. That means that even if we have one vendor successfully
serving the application, the failure of another vendor library will
break any api calls related to egl devices.
Instead, if a vendor fails queryDevicesEXT the device mapping logic
should proceed as if it listed no devices. If the relevant device
belongs to the failed vendor then the dispatch will still fail with
EGL_BAD_DEVICE.
Add .arm and .thumb directives before and after the ARMv7 GLX dispatch stubs.
The function addresses that get passed around don't take Thumb into account, so
if they're compiled as Thumb, then they'd still get executed as ARM, which
would cause them to crash a SIGILL.
Intel Control-flow Enforcement Technology (CET):
https://software.intel.com/en-us/articles/intel-sdm
contains shadow stack (SHSTK) and indirect branch tracking (IBT). When
CET is enabled, ELF object files must be marked with .note.gnu.property
section. Also when IBT is enabled, all indirect branch targets must
start with ENDBR instruction which is NOP on non-CET processors.
This fixes:
https://gitlab.freedesktop.org/glvnd/libglvnd/issues/202
Félix Baylac Jacqué
Kyle Brenneman