- Sep 22, 2024
-
-
Dmitry Osipenko authored
Support passing guest fences to host for waiting by using new VIRTIO_GPU_CMD_SUBMIT_3D command field 'num_in_fences' that replaces previousy unused padding field. The in-fence IDs are prepended to the cmdbuf data. The new corresponding VIRTIO_GPU_F_FENCE_PASSING feature is added as well, it must be supported by guest in order to activate fence passing on host. Fence passing allows virglrenderer contexts to share fences on host instead of doing it on guest side, removing significant overhead of host-guest roundtrips. In particular this improves performance of GL/VK interoperability by more than two times for applications that are bound by the inter-context synchronization overhead. This functionality will also allow to pass GPU fence directly to a kernel display driver for syncing framebuffer flips with completion of a GPU job using explicit fencing. Signed-off-by:Dmitry Osipenko <dmitry.osipenko@collabora.com>
-
Dmitry Osipenko authored
Add support for DRM native contexts to VirtIO-GPU that are be enabled using new GL device option "drm=true". Unlike Virgl and Venus contexts that operate on application level, DRM native contexts work on a kernel UAPI level. The lower level access results in a simple and lightweight context implementations that yield better GPU performance. With native context, host GPU is presented as native one on guest, requiring little modification to existing userspace GPU drivers. Signed-off-by: -
Dmitry Osipenko authored
Support asynchronous fencing feature of virglrenderer. It allows Qemu to handle fence as soon as it's signalled instead of periodically polling fence status. This feature is required for enabling DRM context support in Qemu. DRM context doesn't support legacy fencing. Signed-off-by: -
Dmitry Osipenko authored
Print out error messages when virgl fence creation fails to aid debugging of the fencing bugs. Signed-off-by:
-
- Sep 16, 2024
-
-
If EGL is used, we can rely on dmabuf to import textures without doing copies. To get this working on X11, we use the existing SDL hint: SDL_HINT_VIDEO_X11_FORCE_EGL (because dmabuf can't be used with GLX). Signed-off-by:
Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com> Signed-off-by:
-
Dmitry Osipenko authored
SDL API changes GL context to a newly created GL context, which differs from other GL providers that don't switch context. Change SDL backend to restore the original GL context. This allows Qemu's virtio-gpu to support new virglrenderer async fence notifications feature for GL context type, otherwise it's impossible to switch to a created context from a different thread that performs async fence waiting because context remains set to the main thread. Signed-off-by: -
Request Venus when initializing VirGL and if venus=true flag is set for virtio-gpu-gl device. Signed-off-by:
Antonio Caggiano <antonio.caggiano@collabora.com> Signed-off-by:
Huang Rui <ray.huang@amd.com> Signed-off-by:
-
virtio_gpu_virgl_get_num_capsets will return "num_capsets", but we can't assume that capset_index 1 is always VIRGL2 once we'll support more capsets, like Venus and DRM capsets. Register capsets dynamically to avoid that problem. Reviewed-by:
Manos Pitsidianakis <manos.pitsidianakis@linaro.org> Signed-off-by:
-
Support BLOB resources creation, mapping, unmapping and set-scanout by calling the new stable virglrenderer 0.10 interface. Only enabled when available and via the blob config. E.g. -device virtio-vga-gl,blob=true Signed-off-by:
Xenia Ragiadakou <xenia.ragiadakou@amd.com> Signed-off-by:
-
Dmitry Osipenko authored
Check whether command processing has been finished; otherwise, stop processing commands and retry the command again next time. This allows us to support asynchronous execution of non-fenced commands needed for unmapping host blobs safely. Suggested-by:
Akihiko Odaki <akihiko.odaki@daynix.com> Signed-off-by:
-
In a preparation to adding host blobs support to virtio-gpu, add virgl resource management that allows to retrieve resource based on its ID and virgl resource wrapper on top of simple resource that will be contain fields specific to virgl. Signed-off-by:
Antonio Caggiano <quic_acaggian@quicinc.com> Signed-off-by:
-
Dmitry Osipenko authored
The udmabuf usage is mandatory when virgl is disabled and blobs feature enabled in the Qemu machine configuration. If virgl and blobs are enabled, then udmabuf requirement is optional. Since udmabuf isn't widely supported by a popular Linux distros today, let's relax the udmabuf requirement for blobs=on,virgl=on. Now, a full-featured virtio-gpu acceleration is available to Qemu users without a need to have udmabuf available in the system. Reviewed-by:
Marc-André Lureau <marcandre.lureau@redhat.com> Signed-off-by:
-
Patch "virtio-gpu: CONTEXT_INIT feature" has added the context_init feature flags. Expose this feature and support creating virglrenderer context with flags using context_id if libvirglrenderer is new enough. Originally-by:
-
Dmitry Osipenko authored
New virglrerenderer features were stabilized with release of v1.0.0. Presence of symbols in virglrenderer.h doesn't guarantee ABI compatibility with pre-release development versions of libvirglerender. Use virglrenderer version to decide reliably which virgl features are available. Reviewed-by:
Alex Bennée <alex.bennee@linaro.org> Signed-off-by:
-
Dmitry Osipenko authored
Even though GL GPU doesn't support hotplugging today, free virgl resources when GL device is unrealized. For consistency. Signed-off-by: -
Dmitry Osipenko authored
virtio_gpu_virgl_init() may fail, leading to a further Qemu crash because Qemu assumes it never fails. Check virtio_gpu_virgl_init() return code and don't execute virtio commands on error. Failed virtio_gpu_virgl_init() will result in a timed out virtio commands for a guest OS. Signed-off-by: -
Dmitry Osipenko authored
Move print_stats timer to VirtIOGPUGL for consistency with cmdq_resume_bh and fence_poll that are used only by GL device. Signed-off-by: -
Dmitry Osipenko authored
Move fence_poll timer to VirtIOGPUGL for consistency with cmdq_resume_bh that are used only by GL device. Signed-off-by: -
Dmitry Osipenko authored
Replace printf's used for tracking of in-flight fence inc/dec events with tracing, for consistency with the rest of virtio-gpu code that uses tracing. Suggested-by:
Marc-André Lureau <marcandre.lureau@gmail.com> Signed-off-by:
-
- Sep 15, 2024
-
-
https://github.com/philmd/qemuPeter Maydell authored
Misc HW & UI patches - Remove deprecated SH4 SHIX machine TC58128 NAND EEPROM (Phil) - Remove deprecated CRIS target (Phil) - Remove deprecated RISC-V 'any' CPU type (Phil) - Add fifo8_peek_buf() to correctly handle FIFO wraparound (Mark) - Minor cleanups in Designware PCIe, PL011 and loongson IPI models (Phil) - Fixes in TI TMP105 temperature (Guenter) - Convert Sun ESCC and ADB mouses to QemuInputHandler (Mark) - Prevent heap overflow in VIRTIO sound device (Volker) - Cleanups around g_assert_not_reached() call (Pierrick) - Add Clément as VT-d reviewer (Clément) - Prevent stuck modifier keys and unexpected text input on Windows (Volker) - Explicitly set SDL2 swap interval when OpenGL is enabled (Gert) # -----BEGIN PGP SIGNATURE----- # # iQIzBAABCAAdFiEE+qvnXhKRciHc/Wuy4+MsLN6twN4FAmbkoOUACgkQ4+MsLN6t # wN56SBAA1nYl5YTI0BwvToKysoNIJ5aTgDEbT87XYTatKYyWYlN5W1yrAeMcuwFO # Rw8edtn2YY89ZMG/LPcY1h6+nI2qVMpuOzNP3ve4D+eM6AMUMX0bzeqXndUind7c # 1kTcV7Wctfk34zzs5B9GIU0WswSSbL3FeETS47ySNg6J7GbVWIu41dh52oeg8XWa # Zfw0FZDt7hSAbsUckBIC9/Nbh0hucxBnQevQLkVb6u8O0yX3wV2OZRIC7+NkvHkH # ZLPT8lg197PitH0NouUJI5oMv5Ty3PXGHtrIAKvts+fGBpWL+XzEtQmT8RzqgxTl # 9Z6C/PvfMHNtA7BE66D4iLOKBPpV0RCUDOAGsfcWy5GuklzeUy03DxZ/2xO8XERm # TP0EP0nt2YddnELX7H65b78gJkPvnhME/MtA6Y6R7uxAA8gawZVWZQ1oDGUBNGDI # zJ62Cu4nYPwpGiewwb+ZTkjeiaYddJsPNsE8f8d4XZCpTwpIM/oYzJapxedBwjrg # a4eAWiy7xIvvGPxWN7IQPosGYcyO6zhbI+iAbxp1xmWsX0TPgLUcJtK9+pklqWS7 # 9ucrvkq5XRZSJMaGF0LZuZH7Qx6us7m0rik5wG96d8qrIXRpi8kXfWxI17SWQkGa # cG91u+FrKmfBr+yD7Q1gVbaYzkD+X1hPkQmSVmnlNS+5axrnSYQ= # =sPwb # -----END PGP SIGNATURE----- # gpg: Signature made Fri 13 Sep 2024 21:30:29 BST # gpg: using RSA key FAABE75E12917221DCFD6BB2E3E32C2CDEADC0DE # gpg: Good signature from "Philippe Mathieu-Daudé (F4BUG) <f4bug@amsat.org>" [full] # Primary key fingerprint: FAAB E75E 1291 7221 DCFD 6BB2 E3E3 2C2C DEAD C0DE * tag 'hw-misc-20240913' of https://github.com/philmd/qemu : (60 commits) ui: remove break after g_assert_not_reached() ui/sdl2: set swap interval explicitly when OpenGL is enabled ui/sdl2: ignore GUI keys in SDL_TEXTINPUT handler ui/sdl2: release all modifiers system: replace assert(0) with g_assert_not_reached() hw/pci-host: remove break after g_assert_not_reached() hw/misc: remove break after g_assert_not_reached() hw/gpio: remove break after g_assert_not_reached() hw/watchdog: replace assert(0) with g_assert_not_reached() hw/core: replace assert(0) with g_assert_not_reached() hw/char: replace assert(0) with g_assert_not_reached() hw/input/adb-mouse: convert to use QemuInputHandler hw/char/escc: convert Sun mouse to use QemuInputHandler hw/sensor/tmp105: Lower 4 bit of limit registers are always 0 hw/sensor/tmp105: OS (one-shot) bit in config register always returns 0 hw/sensor/tmp105: Pass 'oneshot' argument to tmp105_alarm_update() hw/sensor/tmp105: Use registerfields API hw/sensor/tmp105: Coding style fixes tests/unit: Comment FIFO8 tests tests/unit: Expand test_fifo8_peek_buf_wrap() coverage ... Signed-off-by:
Peter Maydell <peter.maydell@linaro.org>
-
- Sep 13, 2024
-
-
Pierrick Bouvier authored
Use of assert(false) can trip spurious control flow warnings from some versions of GCC (i.e. using -fsanitize=thread with gcc-12): error: control reaches end of non-void function [-Werror=return-type] default: g_assert_not_reached(); break; | ^^^^^ Solve that by removing the unreachable 'break' statement, unifying the code base on g_assert_not_reached() instead. Signed-off-by:Pierrick Bouvier <pierrick.bouvier@linaro.org> Reviewed-by:
Richard Henderson <richard.henderson@linaro.org> Reviewed-by:
Philippe Mathieu-Daudé <philmd@linaro.org> Message-ID: <20240910221606.1817478-37-pierrick.bouvier@linaro.org> [PMD: Add description suggested by Eric Blake] Signed-off-by:
-
Gert Wollny authored
Before 176e3783 (ui/sdl2: OpenGL window context) SDL_CreateRenderer was called unconditionally setting the swap interval to 0. Since SDL_CreateRenderer is now no longer called when OpenGL is enabled, the swap interval is no longer set explicitly and vsync handling depends on the environment settings which may lead to a performance regression with virgl as reported in https://gitlab.com/qemu-project/qemu/-/issues/2565 Restore the old vsync handling by explicitly calling SDL_GL_SetSwapInterval if OpenGL is enabled. Fixes: 176e3783 (ui/sdl2: OpenGL window context) Closes: https://gitlab.com/qemu-project/qemu/-/issues/2565 Signed-off-by:
Gert Wollny <gert.wollny@collabora.com> Acked-by:
-
Volker Rümelin authored
Ignore GUI keys for SDL_TEXTINPUT events, just like GUI keys are ignored for SDL_KEYDOWN events. This prevents unintended text input in a text console when hiding the text console with the GUI keys. The SDL_TEXTINPUT event always comes after the SDL_KEYDOWN event. See https://github.com/libsdl-org/SDL/issues/1659 . Tested-by:
Howard Spoelstra <hsp.cat7@gmail.com> Signed-off-by:
Volker Rümelin <vr_qemu@t-online.de> Tested-by:
Bernhard Beschow <shentey@gmail.com> Message-ID: <20240909061552.6122-3-vr_qemu@t-online.de> Signed-off-by:
-
Volker Rümelin authored
Each virtual console in the SDL2 frontend has a key state map. When switching windows with GUI keys we have to release all pressed modifier keys in the currently active window, because after the switch the now inactive window no longer receives the key release events. To reproduce the issue open a text editor in the SDL UI and then press Ctrl-Alt-2 to open a Compat Monitor Console. Close the console with the mouse. Try to enter text in the text editor and notice that the modifier keys Ctrl and Alt are stuck and need to be pressed once to be released. Tested-by:
-
Pierrick Bouvier authored
Use of assert(false) can trip spurious control flow warnings from some versions of GCC (i.e. using -fsanitize=thread with gcc-12): error: control reaches end of non-void function [-Werror=return-type] default: assert(0); | } | ^ Solve that by unifying the code base on g_assert_not_reached() instead. Signed-off-by: -
Pierrick Bouvier authored
Use of assert(false) can trip spurious control flow warnings from some versions of GCC (i.e. using -fsanitize=thread with gcc-12): error: control reaches end of non-void function [-Werror=return-type] default: g_assert_not_reached(); break; | ^^^^^ Solve that by removing the unreachable 'break' statement, unifying the code base on g_assert_not_reached() instead. Signed-off-by: -
Pierrick Bouvier authored
Use of assert(false) can trip spurious control flow warnings from some versions of GCC (i.e. using -fsanitize=thread with gcc-12): error: control reaches end of non-void function [-Werror=return-type] default: g_assert_not_reached(); break; | ^^^^^ Solve that by removing the unreachable 'break' statement, unifying the code base on g_assert_not_reached() instead. Signed-off-by: -
Pierrick Bouvier authored
Use of assert(false) can trip spurious control flow warnings from some versions of GCC (i.e. using -fsanitize=thread with gcc-12): error: control reaches end of non-void function [-Werror=return-type] default: g_assert_not_reached(); break; | ^^^^^ Solve that by removing the unreachable 'break' statement, unifying the code base on g_assert_not_reached() instead. Signed-off-by: -
Pierrick Bouvier authored
Use of assert(false) can trip spurious control flow warnings from some versions of GCC (i.e. using -fsanitize=thread with gcc-12): error: control reaches end of non-void function [-Werror=return-type] default: assert(0); | } | ^ Solve that by unifying the code base on g_assert_not_reached() instead. Signed-off-by:Richard W.M. Jones <rjones@redhat.com> Message-ID: <20240910221606.1817478-8-pierrick.bouvier@linaro.org> [PMD: Add description suggested by Eric Blake] Signed-off-by:
-
Pierrick Bouvier authored
Use of assert(false) can trip spurious control flow warnings from some versions of GCC (i.e. using -fsanitize=thread with gcc-12): error: control reaches end of non-void function [-Werror=return-type] default: assert(0); | } | ^ Solve that by unifying the code base on g_assert_not_reached() instead. Signed-off-by: -
Pierrick Bouvier authored
Use of assert(false) can trip spurious control flow warnings from some versions of GCC (i.e. using -fsanitize=thread with gcc-12): error: control reaches end of non-void function [-Werror=return-type] default: assert(0); | } | ^ Solve that by unifying the code base on g_assert_not_reached() instead. Signed-off-by: -
Mark Cave-Ayland authored
Update the ADB mouse implementation to use QemuInputHandler instead of the legacy qemu_add_mouse_event_handler() function. Signed-off-by:
Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk> Reviewed-by:
-
Mark Cave-Ayland authored
Update the Sun mouse implementation to use QemuInputHandler instead of the legacy qemu_add_mouse_event_handler() function. Note that this conversion adds extra sunmouse_* members to ESCCChannelState but they are not added to the migration stream (similar to the Sun keyboard members). If this were desired in future, the Sun devices should be split into separate devices and added to the migration stream there instead. Signed-off-by:
Carl Hauser <chauser@pullman.com> Message-ID: <20240904102301.175706-1-mark.cave-ayland@ilande.co.uk> Signed-off-by:
-
Guenter Roeck authored
Per datasheet, "HIGH AND LOW LIMIT REGISTERS", the lower 4 bit of the limit registers are unused and always report 0. The lower 4 bit should not be used for temperature comparisons, so mask the unused bits before storing the limits. Signed-off-by:
Guenter Roeck <linux@roeck-us.net> Reviewed-by:
Cédric Le Goater <clg@kaod.org> Message-Id: <20240906154911.86803-6-philmd@linaro.org> [PMD: Update tests/qtest/ files] Signed-off-by:
-
Philippe Mathieu-Daudé authored
Per datasheet, "ONE-SHOT (OS)", the OS bit always returns 0 when reading the configuration register. Clear the ONE_SHOT bit in the WRITE path. Now than the READ path is simpler, we can also simplify tmp105_alarm_update(). Signed-off-by:
-
Philippe Mathieu-Daudé authored
The next commit will clear the ONE_SHOT bit in the WRITE path (to keep the READ path trivial). As a preliminary step, pass the 'oneshot' value as argument to tmp105_alarm_update(). No logical change intended. Signed-off-by:
-
Philippe Mathieu-Daudé authored
To improve readability, use the registerfields API. Define the register bits with FIELD(), and use the FIELD_EX8() and FIELD_DP8() macros. Remove the abbreviations in comments. Signed-off-by:
-
Guenter Roeck authored
Coding style asks for no space between variable and "++". The next patch in this series will change one of those assignments. Instead of changing just one with that patch, change all of them for consistency. While at it, also fix other coding style problems reported by checkpatch. Signed-off-by:
-
Philippe Mathieu-Daudé authored
Add comments describing how the FIFO evolves during each test. Signed-off-by:
-
Philippe Mathieu-Daudé authored
Test fifo8_peek_buf() can fill a buffer with wrapped data. Signed-off-by:
-
