0
Fork 0
mirror of https://github.com/Droid-VM/crosvm.git synced 2026-11-17 06:58:03 +00:00

Commit graph

  • a569df4ba6 simplefb: no cache walk on a tick that reads nothing macOS lateautumn233 2026-11-16 01:20:53 -05:00
  • 075cf8164c virtio-media msm_vidc: ask the driver while it works -- its encoder never wakes a poller, its decoder not for every buffer wip/virtio-mmio wip/hackintosh HuJK 2026-11-15 20:30:40 +08:00
  • ba887a2d32 virtio-media encoder: backend= as the decoder's, with Qualcomm's msm_vidc encoder as the first direct driver HuJK 2026-11-15 19:50:24 +08:00
  • 9748c89871 virtio-media decoder: a direct decoder turns the udmabuf module's msm_prime on, so msm_vidc takes guest buffers HuJK 2026-11-15 14:39:09 +08:00
  • 7aae3735f1 virtio-media msm_vidc: tear a session's queues down before close, and log each visible-rectangle query HuJK 2026-11-14 19:41:38 +08:00
  • 012b908497 virtio-media msm_vidc: the visible rectangle from COMPOSE, then CROP, then the stream's own size HuJK 2026-11-13 23:15:49 +08:00
  • d3fa114842 virtio-media: the VMM opens /dev/udmabuf for a decoder helper that may decode direct HuJK 2026-11-13 23:15:49 +08:00
  • ee955fc899 virtio-media decoder: decode order is offered only for a format the direct driver takes HuJK 2026-11-13 23:15:49 +08:00
  • a88e7c1168 virtio-snd: a prepared or stopped stream parks its endpoint as an unfed one does HuJK 2026-11-13 22:33:59 +08:00
  • 13abbc1dac android_audio: an endpoint's performance= attribute, power saving when it names none HuJK 2026-11-13 22:33:59 +08:00
  • edddd58cc8 virtio-media decoder: backend= chooses who decodes -- MediaCodec, or a direct driver with MediaCodec behind it HuJK 2026-11-13 17:22:26 +08:00
  • 746434efba android_audio: the output ring holds the endpoint's bursts, and a period waits as long as they take HuJK 2026-11-13 17:22:26 +08:00
  • 42fe262640 third_party/libdatachannel: may go into the com.android.virt APEX, which carries crosvm HuJK 2026-11-13 17:22:08 +08:00
  • 266818b7ca Merge the macOS branch (Droid-VM/crosvm macOS 14084603): the msm_vidc decoder, KMS and WebRTC display outputs, PCI MSI over GIC MBI SPIs, xHCI MSI-X, the AAudio callback output HuJK 2026-11-13 15:55:10 +08:00
  • 14084603f2 fix(audio): wait for AAudio playback to start before accepting PCM lateautumn233 2026-11-12 16:25:56 +08:00
  • a62d621d90 fix(gpu): size Android cursor surfaces from guest resources lateautumn233 2026-11-12 04:07:58 +08:00
  • 938e2e70d9 fix(xhci): provide three interrupters for Apple USB host startup lateautumn233 2026-11-12 01:54:07 +08:00
  • 9c63a289df feat(usb): add single-vector MSI-X to virtual xHCI lateautumn233 2026-11-12 01:20:19 +08:00
  • 7df3cf7a75 fix(pci): advertise macOS-compatible GIC MSI frame type lateautumn233 2026-11-11 21:29:27 +08:00
  • 83b3a15e68 virtio-console on virtio-mmio (--virtio-mmio-console), and 32 virtio-mmio slots HuJK 2026-11-11 20:36:44 +08:00
  • 98ed82e55f input: trackpad haptics (trackpad[...,haptics=true]): a multi-finger trackpad that declares the HID Haptics waveforms as EV_FF and hands what the guest sends it back to the socket HuJK 2026-11-11 20:02:14 +08:00
  • 2f3fcf217f aarch64: --mbi-spis gives PCI devices MSI/MSI-X on Gunyah through GIC MBI SPIs lateautumn233 2026-11-10 17:50:46 -05:00
  • 6a48b341bd virtio-mmio: --virtio-mmio-edge-irq signals the devices on edge SPIs lateautumn233 2026-11-10 17:50:31 -05:00
  • 927158207c virtio: a ready queue takes its MSI-X vector until the device is activated lateautumn233 2026-11-10 17:50:31 -05:00
  • f82707d2cb input: a vibrator (vibrator[path=PATH,name=N]: a device that declares FF_RUMBLE and hands what the guest sends it to the socket) HuJK 2026-11-10 19:58:21 +08:00
  • cee8d210ab input: the motion sensors' magnetometer (motion[magnetometer=true]: the magnetic field and the tilt-compensated heading) HuJK 2026-11-10 19:12:18 +08:00
  • 91724fac23 virtio-console: a named port (--serial ...,name=NAME) is announced to the guest, and takes input only while the guest has it open HuJK 2026-11-10 18:37:57 +08:00
  • 23ed5ac721 input: a multi-finger trackpad (trackpad with contacts=N) HuJK 2026-11-10 01:41:09 +08:00
  • 1ee4dc8278 virtio-snd: a playback period a little late is waited for, not replaced by filler and dropped lateautumn233 2026-11-07 02:24:49 -05:00
  • ab2a3b1cab
    Merge 558d54e67b into 849e05f12d Xirren 2026-11-07 12:48:55 +08:00
  • 558d54e67b crosvm: emulate optical disc MMC commands in the SCSI cdrom xirren 2026-11-07 00:03:05 +08:00
  • d7fdd3079b virtio-snd: AAudio output played from a data callback over a ring, late periods dropped to stay on time lateautumn233 2026-11-06 02:31:33 -05:00
  • b7960bae64 virtio-media: a decoder straight through msm_vidc into guest buffers (vidc=true) lateautumn233 2026-11-04 10:28:35 -05:00
  • fe24182b62 virtio-media: MediaCodec decoders within a pixel budget, asked to decode as fast as they can lateautumn233 2026-11-04 10:28:35 -05:00
  • 9da9b17d14 virtio-media: a one-entry guest buffer is seen through the GuestMemory's own mapping lateautumn233 2026-11-03 00:57:09 -05:00
  • 9107c0a96c crosvm: --pre-alloc may be given more than once, its keys merged lateautumn233 2026-11-02 12:10:42 -05:00
  • ff4e1dba35 gpu_display: a WebRTC display output backend (--webrtc-server): one screen served to browsers, hardware-encoded lateautumn233 2026-11-02 11:44:00 -05:00
  • b5e309f959 kms: turn the udmabuf module's msm_prime on, so guest scanouts are PRIME-imported instead of CPU-copied lateautumn233 2026-11-02 11:44:00 -05:00
  • 8b754f6068 virtio-gpu: half-float RGBA scanouts (a macOS guest's 'RGhA' framebuffer, converted by the display's blit), and an unshown simplefb screen's frames go to the GPU screen's window until a guest driver sets a scanout lateautumn233 2026-10-27 02:41:30 -04:00
  • fc26bd8d9b feat(kms): support externally managed display takeover lateautumn233 2026-10-20 23:27:01 +08:00
  • 81dc75e32d fix(input): preserve held touch reports for Windows gestures lateautumn233 2026-10-15 22:52:59 +08:00
  • 273bdc941c fix(kms): use monotonic time for watchdog log freshness lateautumn233 2026-10-15 22:50:24 +08:00
  • 6538a0336e fix(kms): clear inherited SDE scaler state on plane submission lateautumn233 2026-10-15 21:17:00 +08:00
  • 58ea7327da fix(kms): reject takeover while the Android panel is dark lateautumn233 2026-10-15 18:47:31 +08:00
  • 9f45c5ff3a fix(kms): rebuild Android framework when returning display ownership lateautumn233 2026-10-15 18:21:29 +08:00
  • 21af2bc32f fix(input): normalize oversized evdev position axes for Windows HID lateautumn233 2026-10-15 15:18:59 +08:00
  • 66a80741c2 feat(kms): display Windows scanout on the Android panel lateautumn233 2026-10-15 15:09:37 +08:00
  • c000dc773a fix(kms): guard Android takeover with a watchdog-aware sentinel lateautumn233 2026-10-15 14:57:40 +08:00
  • 3ce07defba fix(input): allow evdev devices without a serial number lateautumn233 2026-10-15 14:30:59 +08:00
  • c9e24ed54b input: a stylus device (pen tablet: tip and eraser, pressure, hover distance, tilt, barrel buttons) HuJK 2026-11-04 01:22:50 +08:00
  • d661920599 crosvm: raise the open files limit to the system's ceiling at startup HuJK 2026-11-03 17:57:29 +08:00
  • 47cd593ee3 base: a worker thread takes the process's CPUs, not those of the thread that started it HuJK 2026-11-01 20:25:36 +08:00
  • 1a5592aa98 virtcpufreq_v2: no throttling or clamp ceiling without the worker's own pCPU HuJK 2026-10-31 23:07:29 +08:00
  • d5c13c494c --virtio-mmio-media: put the --virtio-media devices on virtio-mmio, after the sound cards HuJK 2026-10-29 01:11:10 +08:00
  • 8d587f2669 power_monitor: a sysfs backend, so the goldfish battery shows the host's charge without ChromeOS powerd HuJK 2026-10-28 23:00:38 +08:00
  • cc88041e54 --virtio-mmio-snd: put the --virtio-snd cards on virtio-mmio, after the GPU HuJK 2026-10-28 23:00:38 +08:00
  • 8871a821f2 vhost-user frontend: work on virtio-mmio, and honour VIRTQ_AVAIL_F_NO_INTERRUPT on its shared interrupt HuJK 2026-10-28 23:00:38 +08:00
  • 0f2db62ae8 fix(gpu): isolate renderer fences across device resets lateautumn233 2026-10-14 01:50:22 +08:00
  • ac467c6141 feat(gpu): enhance pixel clock management in EDID lateautumn233 2026-09-21 11:34:08 +08:00
  • 8daec4ee17 feat(gpu): advertise a fixed host3d VRAM budget to the guest KMD lateautumn233 2026-09-19 23:48:25 +08:00
  • 0a1e44dbf2 feat(gpu): add diagnostic logging for display copy operation lateautumn233 2026-09-15 00:48:02 +08:00
  • 41a7315868 feat(gpu): enhance logging for resource management lateautumn233 2026-09-11 11:29:58 +08:00
  • 93e158d2d1 feat(gpu): add boot-shared DRM blob pool support lateautumn233 2026-09-07 19:22:00 +08:00
  • eab7518739 media: fill the decoder device's profile and level menus from the codec store HuJK 2026-09-19 21:07:42 +08:00
  • 5a759166bd android_audio: count paced frames in u64, so a long stream cannot panic on the counter HuJK 2026-09-19 19:17:48 +08:00
  • 8f9ffc3f92 media: put the QTI AVC decoder in decode order so the reorder cannot deadlock a client (D91) HuJK 2026-09-18 13:58:00 +08:00
  • cf915ccda9 Revert "media: set KEY_LOW_LATENCY on the AV1 decoder so it emits in decode order (VA2j)" HuJK 2026-09-17 20:39:30 +08:00
  • 48f940529c Revert "media: select the .low_latency AV1 decoder variant so it emits in decode order (VA2k2)" HuJK 2026-09-17 20:39:19 +08:00
  • 4b3e250bcd media: select the .low_latency AV1 decoder variant so it emits in decode order (VA2k2) HuJK 2026-09-16 22:28:43 +08:00
  • 1b295b236a media: set KEY_LOW_LATENCY on the AV1 decoder so it emits in decode order (VA2j) HuJK 2026-09-16 21:49:54 +08:00
  • 6c5ad51eaf media: write the repacked NV12 at the lent CAPTURE buffer's stride (VA2g) HuJK 2026-09-16 18:59:12 +08:00
  • b6881e82c1 media: the encoder learns num-output-slots and reports the CAPTURE floor, and staging is bounded to one codec window (D78 residual) HuJK 2026-09-13 13:18:44 +08:00
  • 03d58fb847 media: drop the D64 back-pressure gate; the hold cost 40% on a plain decode and prevented no loss (D64) HuJK 2026-09-13 12:58:30 +08:00
  • c115c78986 media: the encoder stages its input and drains it into CAPTURE on STREAMOFF(OUTPUT) (D78) HuJK 2026-09-13 10:23:12 +08:00
  • 546aec0afb media: hold decoded outputs back from the codec instead of dropping the head GOP under a stalled CAPTURE queue, and learn the CAPTURE minimum per format (D64, D77) HuJK 2026-09-13 10:10:04 +08:00
  • 99603893b8 media: the pool accounting line is rate-limited to one a second per lease, not just one a step (D74) HuJK 2026-09-13 08:13:54 +08:00
  • cfcc25ce05 media: the SOURCE_CHANGE hold is one-shot per pending-format window, so a one-buffer client is not rate-limited (D64) HuJK 2026-09-13 07:25:14 +08:00
  • 92ff0e1d10 media: the decoder announces the codec's real output-slot count as the CAPTURE minimum (D69) HuJK 2026-09-13 05:11:04 +08:00
  • b31d310e11 media: the pool accounting line steps at 32 MiB instead of flooding on every release (D65) HuJK 2026-09-13 05:05:44 +08:00
  • 58e814bc68 vhost-user media: import MappedPool, which R8-11 moved into start() (D62) HuJK 2026-09-06 08:07:29 +08:00
  • c55fce3b2c media: the decoder backend keeps its staged bitstream across a reinit STREAMOFF(OUTPUT) (D53) HuJK 2026-09-06 08:02:50 +08:00
  • 83aef3e134 media: the pool's nits -- a bounded send or no device, honest error lines, no orphans, no false EOF (R8-8..R8-11) HuJK 2026-09-06 07:10:41 +08:00
  • 607a763b26 media: a helper introduces itself to the pool server by its real card string (R8-6) HuJK 2026-09-06 07:08:50 +08:00
  • 2acb3f3d53 media: pool requests carry ids, and one lock owns the whole round trip (R8-4, R8-5) HuJK 2026-09-06 07:07:03 +08:00
  • 9280d7f8a4 media: a helper-side reset returns its pool space through Drop and ReleaseAll (R8-3) HuJK 2026-09-06 07:04:23 +08:00
  • b88bcca57d media: a failed buffer build gives its remote reservation straight back (R8-2) HuJK 2026-09-06 07:01:34 +08:00
  • 769b53f480 media: the pool server sweeps a lease on EOF alone, and parks through anything less (R8-1) HuJK 2026-09-06 07:00:32 +08:00
  • 17f1429a87 media: the VMM allocates the media_host pool for its helpers over a tube (M8) HuJK 2026-09-06 06:31:02 +08:00
  • f4a2ea84ef media: the decoder holds InputBufferDone on a grace timer, drops the recycle release, and spells AV1 AV01 HuJK 2026-09-06 06:03:52 +08:00
  • de77e3bdf5 media: a device helper is started at the VMM's own log level, not a hardcoded info (D57) HuJK 2026-09-06 05:51:31 +08:00
  • ceffed6686 media: each helper gets its own slice of the media_host pool, so their buffers stop aliasing (D49) HuJK 2026-09-06 04:04:57 +08:00
  • 646444b073 media: the decoder backend bounds the SOURCE_CHANGE hold so a one-buffer client cannot deadlock (D48) HuJK 2026-09-06 04:00:43 +08:00
  • e5f54f5574 camera: a control set that reaches no running stream is announced, not dropped in silence (D52) HuJK 2026-09-06 03:57:42 +08:00
  • a5dc5fe340 media: the decoder backend stages bitstream so OUTPUT drains, and holds InputBufferDone behind the first SOURCE_CHANGE HuJK 2026-09-06 01:47:30 +08:00
  • 5b85383067 camera: a mode the guest sets mid-stream is judged by the request the camera ran, not by a count of frames (D47, D39) HuJK 2026-09-06 01:28:42 +08:00
  • e0897116ef camera: read the three doubtful modes back out of the capture result, and stop crying stall at a long exposure (D36, D38) HuJK 2026-09-05 22:23:01 +08:00
  • 263729f292 media: the encoder backend takes no stale EOS, and publishes the rectangle the codec honours HuJK 2026-09-05 17:01:23 +08:00
  • e44d7accca media: the decoder backend marks a resolution change's LAST buffer and takes no stale EOS HuJK 2026-09-05 17:01:23 +08:00
  • b60d066aa4 android_codec: release_output ends the output borrow, and a bottom-up layout is refused, not indexed HuJK 2026-09-05 17:01:23 +08:00
  • 2f5922ed16 media: the codec arms hand the backend's own outcome to _exit HuJK 2026-09-05 15:42:24 +08:00
  • c60f6055ea media: kind=encoder, in the helper only, over the MediaCodec NDK HuJK 2026-09-05 15:39:00 +08:00
  • 13ad8f878e media: the MediaCodec encoder backend for the fork's video encoder device HuJK 2026-09-05 15:39:00 +08:00
  • 59e9684d62 media: a stale codec index is skipped, and a phone without codecs still gets its device HuJK 2026-09-05 15:39:00 +08:00
  • a72caf6cd4 android_codec: codec_probe takes hex levels, keeps its banners off stdout, and leaves through _exit HuJK 2026-09-05 15:39:00 +08:00
  • 25a803358e android_codec: an event carries the generation it was posted under, and the encoder's keys HuJK 2026-09-05 15:39:00 +08:00
  • 77d40ef70a media: the camera backend maps the V4L2 controls onto the Camera2 request (design 7.1) HuJK 2026-09-05 11:13:13 +08:00
  • 51a720652f camera: camera_probe controls, and what the capture results said HuJK 2026-09-05 11:13:13 +08:00
  • ffd2558570 android_camera: request updates, capture results, and the control characteristics HuJK 2026-09-05 11:13:13 +08:00
  • c45bb85f6a media: announce a size change only after the frames before it, and build the VMM again HuJK 2026-09-05 10:39:42 +08:00
  • db4a516314 media: kind=decoder, in the helper only, over the MediaCodec NDK HuJK 2026-09-05 10:33:07 +08:00
  • afe7a0f90c media: the MediaCodec decoder backend for the fork's video decoder device HuJK 2026-09-05 10:32:37 +08:00
  • 7b567e8a71 android_codec: let a consumer wake its own descriptor, and repack NV12 a row at a time HuJK 2026-09-05 10:32:14 +08:00
  • eb5709482d camera: camera_probe list prints every size and every frame duration HuJK 2026-09-05 10:08:18 +08:00
  • dd7cac3976 media: a guest import answers how long it is HuJK 2026-09-05 09:31:16 +08:00
  • 963ebe1d86 camera: free nothing the NDK still points at, and let a guest see dropped frames HuJK 2026-09-05 09:30:28 +08:00
  • 67c798d457 media: bound the two waits that could park a camera worker or the VMM HuJK 2026-09-05 09:27:11 +08:00
  • 423ddccebf android_codec: the MediaCodec NDK crate and codec_probe (M6, design 7.2) HuJK 2026-09-05 09:10:19 +08:00
  • 9d2dc33fd8 media: name a helper when it is launched, and refuse a kind before launching one HuJK 2026-09-05 09:04:52 +08:00
  • 3ca46c9f81 media: kind=camera, in the helper only, over the Camera2 NDK HuJK 2026-09-05 08:20:56 +08:00
  • 74a6b3b823 vhost-user media: the device is built on the worker thread HuJK 2026-09-05 08:20:56 +08:00
  • 79d77b8d38 android_camera: what the camera device needs from the NDK crate HuJK 2026-09-05 08:20:56 +08:00
  • e38d6f4908 media: a send_event waiting for the guest ends on the worker's kill event HuJK 2026-09-04 14:32:52 +08:00
  • d4da830477 linux: --virtio-media uid= runs the device in a crosvm device media helper HuJK 2026-09-04 14:10:58 +08:00
  • 622347b252 vhost-user: a virtio-media backend, pool-mode only, for the app-uid helper HuJK 2026-09-04 14:10:49 +08:00
  • 60db8a0bbe media: the worker, the queues and the guest mapper are shared with a helper HuJK 2026-09-04 14:10:49 +08:00
  • 4876843fcd media: the media_host pool is its own module, with a test for the lease sweep HuJK 2026-09-04 13:23:49 +08:00
  • a5a03f6e1f cmdline: --virtio-media says the gunyah pool rule covers driver_owned_queues=all HuJK 2026-09-04 10:58:26 +08:00
  • b3707847e6 media: one media_host offset space for the whole VM HuJK 2026-09-04 10:58:16 +08:00
  • 1a1858a7ad media: OUTPUT buffers are mapped read-only into the host HuJK 2026-09-04 10:57:59 +08:00
  • f22a827160 config: the virtio_media default is gated on media, not on audio HuJK 2026-09-04 10:47:03 +08:00
  • b7dcce6fda media: serve MMAP buffers from the media_host pool, and add --virtio-media HuJK 2026-09-04 08:41:56 +08:00
  • 77edc026b6 media: resolve guest-owned buffers through find_region, and import them as dma-bufs HuJK 2026-09-04 08:41:55 +08:00
  • 86cbededbf pools: media_host and media_guest, the two virtio-media buffer pools HuJK 2026-09-04 07:54:32 +08:00
  • 18bac45927 virtio-gpu: half-float RGBA scanouts (a macOS guest's 'RGhA' framebuffer, converted by the display's blit), and an unshown simplefb screen's frames go to the GPU screen's window until a guest driver sets a scanout backup/macOS-1027 lateautumn233 2026-10-27 02:41:30 -04:00
  • cdf58224b3 droidvm-gpu: up to 64 IBs per submit (macOS's guest sends that many; a guest built with 16 still fits) lateautumn233 2026-10-27 02:41:30 -04:00
  • 78f31540e8 virtio-gpu: a resource on the CPU copy tries the display import again every 120 presents HuJK 2026-10-27 10:40:28 +08:00
  • 806372101d fix(gpu): isolate renderer fences across device resets wip/vpu lateautumn233 2026-10-14 01:50:22 +08:00
  • 83620e2358 feat(gpu): enhance pixel clock management in EDID lateautumn233 2026-09-21 11:34:08 +08:00
  • 1d8d8fe2d8 feat(gpu): advertise a fixed host3d VRAM budget to the guest KMD lateautumn233 2026-09-19 23:48:25 +08:00
  • 496aa629d4 feat(gpu): add diagnostic logging for display copy operation lateautumn233 2026-09-15 00:48:02 +08:00
  • 606e0f6a61 feat(gpu): enhance logging for resource management lateautumn233 2026-09-11 11:29:58 +08:00
  • 86d3ef77cd feat(gpu): add boot-shared DRM blob pool support lateautumn233 2026-09-07 19:22:00 +08:00
  • 13031c18dc media: fill the decoder device's profile and level menus from the codec store HuJK 2026-09-19 21:07:42 +08:00
  • 48a0b183b2 android_audio: count paced frames in u64, so a long stream cannot panic on the counter HuJK 2026-09-19 19:17:48 +08:00
  • bb030ec316 media: put the QTI AVC decoder in decode order so the reorder cannot deadlock a client (D91) HuJK 2026-09-18 13:58:00 +08:00
  • 031e7a1dcd Revert "media: set KEY_LOW_LATENCY on the AV1 decoder so it emits in decode order (VA2j)" HuJK 2026-09-17 20:39:30 +08:00
  • f14075dbf1 Revert "media: select the .low_latency AV1 decoder variant so it emits in decode order (VA2k2)" HuJK 2026-09-17 20:39:19 +08:00
  • a807d89bb4 media: select the .low_latency AV1 decoder variant so it emits in decode order (VA2k2) HuJK 2026-09-16 22:28:43 +08:00
  • 5ed82fe3fc media: set KEY_LOW_LATENCY on the AV1 decoder so it emits in decode order (VA2j) HuJK 2026-09-16 21:49:54 +08:00
  • eb3a6def1c media: write the repacked NV12 at the lent CAPTURE buffer's stride (VA2g) HuJK 2026-09-16 18:59:12 +08:00