rust: samples: fix init argument #335

Signed-off-by: Hector Martin <[email protected]>

Fixes warning: [ 8.502802] ------------[ cut here ]------------ [ 8.503445] cpu_latency_qos_remove_request called for unknown object [ 8.504269] WARNING: CPU: 5 PID: 2790 at kernel/power/qos.c:322 cpu_latency_qos_remove_request+0x48/0x98 [ 8.505499] CPU: 5 PID: 2790 Comm: wireplumber Tainted: G W 6.5.0-asahi-00708-gb9b88240f7ae #2291 [ 8.506777] Hardware name: Apple MacBook Air (13-inch, M2, 2022) (DT) <snip regs> [ 8.519099] Call trace: [ 8.519402] cpu_latency_qos_remove_request+0x48/0x98 [ 8.520027] snd_pcm_ioctl+0x86c/0x182c [ 8.520519] __arm64_sys_ioctl+0xf8/0xbd0 [ 8.521020] invoke_syscall.constprop.0+0x78/0xc8 [ 8.521604] do_el0_svc+0x58/0x154 [ 8.522026] el0_svc+0x34/0xe4 [ 8.522409] el0t_64_sync_handler+0x120/0x12c [ 8.522951] el0t_64_sync+0x190/0x194 [ 8.523408] ---[ end trace 0000000000000000 ]--- Signed-off-by: Hector Martin <[email protected]>

alsamixer/etc really don't like write-only controls... Signed-off-by: Hector Martin <[email protected]>

On the speakers PCM, this sequence: 1. Open playback 2. Open sense 3. Close playback 4. Close sense would result in the sense FE being marked as clocks in use at (2), since there is a clock provider (playback FE). Then at (4) this would WARN since there is no driver any more when closing the in use clocks. If (1) and (2) are reversed this does not happen, since the sense PCM is not marked as using the clocks when there is no provider yet. So, check explicitly whether the substream FE is a clock provider in be_prepare, and skip everything if it isn't. Signed-off-by: Hector Martin <[email protected]>

For machines where we do not consider things safe yet, require the commandline argument. Without it, speakers are simply disabled, we don't refuse probe entirely. Signed-off-by: Hector Martin <[email protected]>

Signed-off-by: Hector Martin <[email protected]>

These are unintentionally aliased. Pending a solution for this, let's just use the same limit for now. Signed-off-by: Hector Martin <[email protected]>

Signed-off-by: Hector Martin <[email protected]>

The ASoC convention is that clocks are removed after codec mute, and power up/down is more about top level power management. For these chips, the "mute" state still expects a TDM clock, and yanking the clock in this state will trigger clock errors. So, do the full shutdown<->mute<->active transition on the mute operation, so the amp is in software shutdown by the time the clocks are removed. This fixes TDM clock errors when streams are stopped. Signed-off-by: Hector Martin <[email protected]>

Multiple amps can be connected to the same SDZ GPIO. Using raw GPIOs for this breaks, as there is no concept of refcounting/sharing. In order to model these platforms, introduce support for an SDZ "regulator". This allows us to represent the SDZ GPIO as a simple regulator-fixed, and then the regulator core takes care of refcounting so that all codecs are only powered down once all the driver instances are in the suspend state. Signed-off-by: Hector Martin <[email protected]>

Otherwise we can end up recursively locking the controls lock in the start/stop path, since it can be called from a control change. Signed-off-by: Hector Martin <[email protected]>

Signed-off-by: Hector Martin <[email protected]>

Since the bit is self-clearing. Signed-off-by: Hector Martin <[email protected]>

Signed-off-by: Hector Martin <[email protected]>

Saw this fail once, let's be safer. Signed-off-by: Hector Martin <[email protected]>

When a PCM is suspended, we pause the DMA. If the PCM is then closed while in this state, it does not receive the STOP trigger (as it is not running). In this case, we fail to properly terminate the DMA, calling dmaengine_synchronize() nonetheless, which is undefined behavior. Make sure we always call dmaengine_terminate_async() on PCM close, regardless of whether it has been called previously or not in the trigger callbacks. Signed-off-by: Hector Martin <[email protected]>

When we shut down the amp, we need to wait for the built-in ramp-down before we can remove the TDM clocks. There is no documented status regiter to poll, so the best we can do is a delay. Datasheet says 5.9ms for ramp-down and 1.5ms between shutdown and next startup, so let's do 6ms after mute and 2ms after shutdown. That gives us a cumulative 8ms for ramp-down and guaratees the required minimum shutdown time. Signed-off-by: Hector Martin <[email protected]>

0xf seems to fix the random overcurrent behavior. Signed-off-by: Hector Martin <[email protected]>

Instead of open-coding a fixup function for each platform, let's make it declarative. This is a lot less error-prone. Signed-off-by: Hector Martin <[email protected]>

Multiple amps can be connected to the same SDZ GPIO. Using raw GPIOs for this breaks, as there is no concept of refcounting/sharing. In order to model these platforms, introduce support for an SDZ "regulator". This allows us to represent the SDZ GPIO as a simple regulator-fixed, and then the regulator core takes care of refcounting so that all codecs are only powered down once all the driver instances are in the suspend state. This also reworks the sleep/resume logic to copy what tas2764 does, which makes more sense. Signed-off-by: Hector Martin <[email protected]>

The ISENSE/VSENSE blocks are only powered up when the amplifier transitions from shutdown to active. This means that if those controls are flipped on while the amplifier is already playing back audio, they will have no effect. Fix this by forcing a power cycle around transitions in those controls. Signed-off-by: Hector Martin <[email protected]>

Expose the bits that control the behavior of the SDOUT pin when not actively transmitting slot data. Zero-fill is useful when there is a single amp on the SDOUT bus (e.g. Apple machines with mono speakers or a single stereo pair, where L/R are on separate buses). Pull-down is useful, though not perfect, when multiple amps share a bus. It typically takes around 2 bits for the line to transition from high to low after going Hi-Z, with the pull-down. Signed-off-by: Hector Martin <[email protected]>

We don't actually support configuring the PDM input right now. Rather, this is useful as a hack. On Apple Silicon machines, amps are split between two I2S buses which are logically ANDed internally at the SoC. Odd and even slot groups are driven by amps on either bus respectively. Since the signals are ANDed, unused slot groups must be driven as zero to avoid corrupting the data from the other side. On most recent machines (TAS2764-based), this is accomplished using the "SDOUT zero mask" feature of that chip. Unfortunately, TAS2770 does not support this. It does support zeroing out *all* unused slots, which works well for machines with a single amp per I2S bus. That is all, except one. The 13" M1 MacBook Pro is the only machine using TAS2764 and two amps per I2S bus: L Bus: SPK0I SPK0V Hi-Z Hi-Z SPK2I SPK2V Hi-Z Hi-Z R Bus: Hi-Z Hi-Z SPK1I SPK2V Hi-Z Hi-Z SPK3I SPK3V To ensure uncorrupted data, we need to force all the Hi-Z periods to zero. We cannot use the "force all zero" feature, as that would cause a bus conflict between both amps. We can use the pull-down feature, but that leaves a few bits of garbage on the trailing edge of the speaker data, since the pull-down is weak. This is where the PDM transmit feature comes in. With PDM grounded and disabled (the default state), the PDM slot is transmitted as all zeroes. We can use that to force a zero 16-bit slot after the voltage data for each speaker, cleaning it up. Then the pull-down ensures the line stays low for the subsequent slot: L Bus: SPK0I SPK0V PDM0 PulDn SPK2I SPK2V PDM0 PulDn R Bus: PDM0 PulDn SPK1I SPK2V PDM0 PulDn SPK3I SPK3V Yes, this is a horrible hack, but it beats adding dummy slots that would be visible to the userspace capture side. There may be some other way to fix the logical AND behavior on the MCA side... that would make this unnecessary. ("How does Apple deal with this"? - they don't, macOS does not use IVSENSE on TAS2764 machines even though it's physically wired up, but we want to do so on Linux.) Signed-off-by: Hector Martin <[email protected]>

The scale starts at -100dB, not -128dB. Signed-off-by: Hector Martin <[email protected]>

This one already uses a gain lower than the others. It doesn't look like full scale no-DSP output with typical music is particularly dangerous here, and we probably want the headroom for DSP, so let's not do it for this one. Signed-off-by: Hector Martin <[email protected]>

This PCM triggers speakersafetyd, so hide it if it can't work. Signed-off-by: Hector Martin <[email protected]>

Still hard gated on speakersafetyd for now. Signed-off-by: Hector Martin <[email protected]>

TX launch polarity needs to be the opposite of RX capture polarity, to generate the right bit slot alignment. Signed-off-by: Hector Martin <[email protected]>

Fix fallthrough warning and remove stray spaces before tabs for indenting. Signed-off-by: Janne Grunau <[email protected]>

This might fix a udev race, and also makes it possible to switch to a more descriptive "AppleJxxx" name (but before that we need to update userspace to avoid breaking users). Signed-off-by: Hector Martin <[email protected]>

Signed-off-by: Hector Martin <[email protected]>

1 enables new models, 2 further removes safeties. Mostly so that people who set it to 1 for early access and forget don't get stuck without safety nets. Signed-off-by: Hector Martin <[email protected]>

We want the extra headroom, and speakersafetyd seems to be reliable. 3dB lower gain isn't going to buy us much safety at this point. Signed-off-by: Hector Martin <[email protected]>

Signed-off-by: Hector Martin <[email protected]>

On systems with cs42l84 headset codec "* " can't be used as control name pattern since it would match "Jack HPF Corner Frequency". Its control is not an enum and thus will always return -EINVAL. Signed-off-by: Janne Grunau <[email protected]>

Right after the early FE setup, ADMAC gets told to start the DMA. This can end up in a weird "slip" state with the channels transposed. Waiting a bit fixes this; presumably this allows the clock to stabilize. Signed-off-by: Hector Martin <[email protected]>

We are going to enable these out of band. If you are a distro packager: ** WARNING: ** ** YOU ABSOLUTELY NEED THIS PATCH IN YOUR LSP-PLUGINS PACKAGE ** lsp-plugins/lsp-dsp-lib#20 Do NOT enable speakers without that patch, on any model. It can/will result in nasty noise that could damage them. Signed-off-by: Hector Martin <[email protected]>

Sigh... hope this works. Signed-off-by: Hector Martin <[email protected]>

¯\_(ツ)_/¯ Signed-off-by: Hector Martin <[email protected]>

Signed-off-by: Sasha Finkelstein <[email protected]>

Signed-off-by: Martin Povišer <[email protected]>

Signed-off-by: Jean-Francois Bortolotti <[email protected]>

This is the SPMI counterpart to simple-mfd-i2c. It merely exposes the SPMI register address space as an MFD device, such that different aspects of a device can be managed by separate drivers. Signed-off-by: Hector Martin <[email protected]>

This driver exposes part of an SPMI MFD device as an NVMEM device. It is intended to be used with e.g. PMUs/PMICs that are used to hold power-management configuration, such as used on Apple Silicon Macs. Signed-off-by: Hector Martin <[email protected]>

Fixes probing of nvmem cells in v6.8-rc6 Signed-off-by: Janne Grunau <[email protected]>

…r_poll_disable during suspend When drivers call drm_kms_helper_poll_disable from their device suspend implementation without enabled output polling before, following warning will be reported,due to work->func not be initialized: [ 55.141361] WARNING: CPU: 3 PID: 372 at kernel/workqueue.c:3066 __flush_work+0x22f/0x240 [ 55.141382] Modules linked in: nls_iso8859_1 snd_hda_codec_generic ledtrig_audio snd_hda_intel snd_intel_dspcfg snd_intel_sdw_acpi snd_hda_codec snd_hda_core snd_hwdep snd_pcm snd_seq_midi snd_seq_midi_event snd_rawmidi snd_seq intel_rapl_msr intel_rapl_common bochs drm_vram_helper drm_ttm_helper snd_seq_device nfit ttm crct10dif_pclmul snd_timer ghash_clmulni_intel binfmt_misc sha512_ssse3 aesni_intel drm_kms_helper joydev input_leds syscopyarea crypto_simd snd cryptd sysfillrect sysimgblt mac_hid serio_raw soundcore qemu_fw_cfg sch_fq_codel msr parport_pc ppdev lp parport drm ramoops reed_solomon pstore_blk pstore_zone efi_pstore virtio_rng ip_tables x_tables autofs4 hid_generic usbhid hid ahci virtio_net i2c_i801 crc32_pclmul psmouse virtio_scsi libahci i2c_smbus lpc_ich xhci_pci net_failover virtio_blk xhci_pci_renesas failover [ 55.141430] CPU: 3 PID: 372 Comm: kworker/u16:9 Not tainted 6.2.0-rc6+ Rust-for-Linux#16 [ 55.141433] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.12.1-0-ga5cab58e9a3f-prebuilt.qemu.org 04/01/2014 [ 55.141435] Workqueue: events_unbound async_run_entry_fn [ 55.141441] RIP: 0010:__flush_work+0x22f/0x240 [ 55.141444] Code: 8b 43 28 48 8b 53 30 89 c1 e9 f9 fe ff ff 4c 89 f7 e8 b5 95 d9 00 e8 00 53 08 00 45 31 ff e9 11 ff ff ff 0f 0b e9 0a ff ff ff <0f> 0b 45 31 ff e9 00 ff ff ff e8 e2 54 d8 00 66 90 90 90 90 90 90 [ 55.141446] RSP: 0018:ff59221940833c18 EFLAGS: 00010246 [ 55.141449] RAX: 0000000000000000 RBX: 0000000000000000 RCX: ffffffff9b72bcbe [ 55.141450] RDX: 0000000000000001 RSI: 0000000000000001 RDI: ff3ea01e4265e330 [ 55.141451] RBP: ff59221940833c90 R08: 0000000000000000 R09: 8080808080808080 [ 55.141453] R10: ff3ea01e42b3caf4 R11: 000000000000000f R12: ff3ea01e4265e330 [ 55.141454] R13: 0000000000000001 R14: ff3ea01e505e5e80 R15: 0000000000000001 [ 55.141455] FS: 0000000000000000(0000) GS:ff3ea01fb7cc0000(0000) knlGS:0000000000000000 [ 55.141456] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 55.141458] CR2: 0000563543ad1546 CR3: 000000010ee82005 CR4: 0000000000771ee0 [ 55.141464] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 [ 55.141465] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 [ 55.141466] PKRU: 55555554 [ 55.141467] Call Trace: [ 55.141469] <TASK> [ 55.141472] ? pcie_wait_cmd+0xdf/0x220 [ 55.141478] ? mptcp_seq_show+0xe0/0x180 [ 55.141484] __cancel_work_timer+0x124/0x1b0 [ 55.141487] cancel_delayed_work_sync+0x17/0x20 [ 55.141490] drm_kms_helper_poll_disable+0x26/0x40 [drm_kms_helper] [ 55.141516] drm_mode_config_helper_suspend+0x25/0x90 [drm_kms_helper] [ 55.141531] ? __pm_runtime_resume+0x64/0x90 [ 55.141536] bochs_pm_suspend+0x16/0x20 [bochs] [ 55.141540] pci_pm_suspend+0x8b/0x1b0 [ 55.141545] ? __pfx_pci_pm_suspend+0x10/0x10 [ 55.141547] dpm_run_callback+0x4c/0x160 [ 55.141550] __device_suspend+0x14c/0x4c0 [ 55.141553] async_suspend+0x24/0xa0 [ 55.141555] async_run_entry_fn+0x34/0x120 [ 55.141557] process_one_work+0x21a/0x3f0 [ 55.141560] worker_thread+0x4e/0x3c0 [ 55.141563] ? __pfx_worker_thread+0x10/0x10 [ 55.141565] kthread+0xf2/0x120 [ 55.141568] ? __pfx_kthread+0x10/0x10 [ 55.141570] ret_from_fork+0x29/0x50 [ 55.141575] </TASK> [ 55.141575] ---[ end trace 0000000000000000 ]--- Fixes: a4e7717 ("drm/probe_helper: sort out poll_running vs poll_enabled") Signed-off-by: Zongmin Zhou<[email protected]>

Should result in drm apple to be registered as first DRM device replacing simpledrm. Should resolve problems with userspace assuming that card0 is the main displays device. Signed-off-by: Janne Grunau <[email protected]>

This prevents the PD from being shut down on suspend, which we need until we support runtime PM properly again. Signed-off-by: Hector Martin <[email protected]>

Signed-off-by: Janne Grunau <[email protected]>

An atomic_commit for brightness changes will consume a DCP swap without frame buffer updates and will result in a lost frame. After updating the next brightness values wait for 1 frame duration (at 23.976 fps). Check if the brightness update still needs to be send to DVCP or if a swap did that in the meintime. Signed-off-by: Janne Grunau <[email protected]>

It's only needed to configure the display contoller's iommu to share buffers between the DCP co-processor and the display controller. Possible concern is runtime PM for it and its iommu. If we don't set it up the power domain might never go to lower power states even if it could. Signed-off-by: Janne Grunau <[email protected]>

The current use of of dma_get_sgtable/dma_map_sgtable is deemed unsafe. Replace it with an unmanaged iommu domain for the piodma iommu to map the buffers. Signed-off-by: Janne Grunau <[email protected]>

The iommu scatter table/list mapping can only map full iommu page size extents. Just align the actual the allocation to the iommu page size. This could be handled differently using DARTs subpage protection but there's no easy way to integrate that. Signed-off-by: Janne Grunau <[email protected]>

Used on M2 Ultra During startup. Units are unclear. Signed-off-by: Janne Grunau <[email protected]>

Removes support for all firmware versions which report as compatible to 13.3 except 13.5. This will be removed after m1n1 reports firmware 13.5 as "apple,firmware-compat" for a while. The files with "v13_3" will be renamed at a later point to avoid conflicts with development trees. Signed-off-by: Janne Grunau <[email protected]>

Avoids ugly "__maybe_unused". Signed-off-by: Janne Grunau <[email protected]>

It does not do anything for dcp and its iommu only child node. Signed-off-by: Janne Grunau <[email protected]>

To match macOS behavior and in the hope to fix dcpext crashes on t8112. Crashes still occur but let's keep this. Shouldn;t make a difference since we're on the swaps to finish. Signed-off-by: Janne Grunau <[email protected]>

DCP itself uses with the 13.5 firmware a timeout of 8 seconds for modesets. Using a longer timeout prevents overlapping calls to dcp and might improve reliabilty with slower displays. Signed-off-by: Janne Grunau <[email protected]>

This is no longer necessary after introducing "apple,dma-range" for the dart driver. Signed-off-by: Janne Grunau <[email protected]>

This drivers adds support for the display crossbar used to route display controller streams to the three different modes (DP AltMode, USB4 Tunnel #0/Rust-for-Linux#1) of the Type-C ports. Signed-off-by: Sven Peter <[email protected]>

Signed-off-by: Janne Grunau <[email protected]>

Makes traces easier to compare with macOS. Signed-off-by: Janne Grunau <[email protected]>

This is a simplified version and probably should live in a separate file. Even the shared registers are quite different. Signed-off-by: Janne Grunau <[email protected]>

Signed-off-by: Martin Povišer <[email protected]>

Signed-off-by: Sven Peter <[email protected]> Co-developed-by: Martin Povišer <[email protected]> Signed-off-by: Martin Povišer <[email protected]> Co-developed-by: Janne Grunau <[email protected]> Signed-off-by: Janne Grunau <[email protected]>

"Channel numbers" as received by AFK/EPIC are constantly increasing over restarts of the endpoint. Use a linear array of services and match based on the channel number. The number of services per endpoint is too small to make a difference. Signed-off-by: Janne Grunau <[email protected]>

This is required for DP Altmode, DP Thunderbolt tunneling and HDMI output on 14/16-inch Macbook Pros and M2* desktop devices. M2* desktops and 14 and 16 inch Macbook Pros expose a DisplayPort to HDMI converter which is driven by the DP output of one of the DCP/DCPext display coprocessor/controller blocks. Two gpio pins are used for power control. Another gpio pin acts as HDMI hpd. Do not use the hpd as direct drm_connector interrupt since that is already wired to DCPs hotplug notification. Instead use it to trigger link setup via the dptx endpoint. Signed-off-by: Sven Peter <[email protected]> Co-developed-by: Janne Grunau <[email protected]> Signed-off-by: Janne Grunau <[email protected]>

The offsets differ for every DCP instance. Instead of hardcoding offsets for each SoC family offsets and calculate the instance offset move everything to the device tree. This helps multi die SoCs since there is and unexpected offset between both dies. On multi die SoCs device tree changes were necessary to avoid translating the PMGR reg via the seconds die "ranges" property. Signed-off-by: Janne Grunau <[email protected]>

Signed-off-by: Janne Grunau <[email protected]>

The 13.3 firmware has an additional get_max_lane_count call inserted with ID 10. Signed-off-by: Janne Grunau <[email protected]>

Signed-off-by: Janne Grunau <[email protected]>

Configures the phy to the correct dcp(ext) source by abusing submode in the phy_set_mode_ext() call. Signed-off-by: Janne Grunau <[email protected]>

Adapt to behavior seen on j474s with dcp0 driving lpdptx-phy and dp2hdmi using the macOS 13.5 firmware. Signed-off-by: Janne Grunau <[email protected]>

Signed-off-by: Hector Martin <[email protected]>

Just in case, for consistency with macOS. Signed-off-by: Hector Martin <[email protected]>

Signed-off-by: Asahi Lina <[email protected]>

of_platform_device_destroy() can trigger several NULL pointer dereference which have been elusive so far. Comment this for now since the oopses causes the shutdown to hang. Since dcp can not be reloaded this leaks the platform device on shutdown and reboot. Signed-off-by: Janne Grunau <[email protected]>

Avoids out of order messages and already unmapped buffers while tracing with hv/trace_dcp.py. Signed-off-by: Janne Grunau <[email protected]>

This callback is occasionally seen around (failed) modesets. There seems to be no need to handle it so just trace it. Signed-off-by: Janne Grunau <[email protected]>

Signed-off-by: Janne Grunau <[email protected]>

This reverts commit ab69434.

This mirrors what macOS does and should make reconnections more reliable. Signed-off-by: Janne Grunau <[email protected]>

Seems to make disconnect / reconnect more reliable and almost fixes suspend/resume. The drm device tries to modeset too early on resume which leaves the screen blank. This should reduce power consumption after disconnecting the HDMI port. Signed-off-by: Janne Grunau <[email protected]>

Makes connects more reliable. Signed-off-by: Janne Grunau <[email protected]>

DCP notices the disconnects on its own and the parallel handling just results in confusion (both on DRM and developer side). Signed-off-by: Janne Grunau <[email protected]>

Signed-off-by: Janne Grunau <[email protected]>

Triggering modesets from drm_connector_helper_funcs.atomic_check is more in line with DRM/KMS' design and allows returning errors from failed modesets. Ignore hotplug callbacks from DCP during modeset. DCP always does disconnected -> connected on (at least the initial) modeset. Shield drm helpers from this. This improves reliability with externel (dptx based) displays. Signed-off-by: Janne Grunau <[email protected]>

Signed-off-by: Janne Grunau <[email protected]>

squash! drm: apple: Extract modeset crtc's atomic_flush() Fixes: 99d7bb8 ("drm: apple: Extract modeset crtc's atomic_flush()") Signed-off-by: Janne Grunau <[email protected]>

- Do not trigger an hotplug event from disconnect. DCP/iomfb notices that itself. - Check HPD status before disconnecting DPTX in the crtc disable path. - disconnect on suspend to allow an orderly re-connect on resume Signed-off-by: Janne Grunau <[email protected]>

Avoid device wide hotplugs as DCP knowns the affected connector. Signed-off-by: Janne Grunau <[email protected]>

Only observed with dcp/dptx in linux after initialisation and reset in m1n1. On the initial startup dcp sends two D576 (hotPlug_notify_gated) presumendly due to state confusion due to the multiple dptx connections. Signed-off-by: Janne Grunau <[email protected]>

Under unknown but slightly broken conditions dcp sends timing modes without linked color modes. Log a warning when this happens and log the number of valid modes before emitting HPD events. Signed-off-by: Janne Grunau <[email protected]>

DPTX setup from an initialized connection and display with sleeping and reset dcp is unfortunately quite fragile. The display connection has to be stopped and reestablished. Goodbye flicker free boot. If the IOMFB endpoint is started too early dcp might provide incomplete timing modes which prevent modesets. On display standby a HPD is triggered should result in a fully initialized dcp. If not a display cable unplug and plug should help. MacOS doesn't handle this at all and just gives up. Signed-off-by: Janne Grunau <[email protected]>

Signed-off-by: Janne Grunau <[email protected]>

Should make connect more reliable by avoiding hardcoded waits which are either to long or too short. In the second case the display can't be brought up since dcp fails to report any modes during start. Signed-off-by: Janne Grunau <[email protected]>

DCP color mode scoring seems to prefer high bit depth color modes even when it it would require DSC. For example 12-bit 4k 60 Hz YCbCr 4:4:4 over a 600 MHz HDMI 2.0 link. Prefer 8-/10-bit RGB or YCbCr 4:4:4 modes if available. Signed-off-by: Janne Grunau <[email protected]>

Fixes missing physical display dimensions for HDMI display on Macbook Pros. Signed-off-by: Janne Grunau <[email protected]>

Signed-off-by: Janne Grunau <[email protected]>

The payload is not necessarily epic_std_service_ap_call. The powerlog service on the system endpoint passes serialized dictionaries as payload. Signed-off-by: Janne Grunau <[email protected]>

The 13.5 firmware has stopped updating the NITS property on backlight brightness changes. Parse system log events instead which report backlight's brightness in millinits. Fixes the backlight device's "actual_brightness" property used by the systemd backlight service to save and restore brightness. Signed-off-by: Janne Grunau <[email protected]>

With linux-6.8, the kernel warns about functions that have no extern declaration, so mark both of these static. Fixes: 2d782b0 ("gpu: drm: apple: Add sound mode parsing") Signed-off-by: Arnd Bergmann <[email protected]>

Signed-off-by: Alyssa Ross <[email protected]>

Signed-off-by: Jonathan Gray <[email protected]>

If the DCP firmware indicates that it didn't restore the brightness, schedule an update. Wait for 1 frame duration and check if the brightness update has been taken care of by a swap that happened in the meantime. Fixes restoring the brightness after resume when running on a dumb framebuffer where swaps may not happen for a very long time. Signed-off-by: Mark Kettenis <[email protected]>

Add missing training '\n' and remove leftover dev_dbg() statements. Signed-off-by: Janne Grunau <[email protected]>

Not necessarily only a debounce but 500ms sleep in the HPD interrupt handler seems to make the modeset more reliable on M2* desktop devices. Signed-off-by: Janne Grunau <[email protected]>

Signed-off-by: Janne Grunau <[email protected]>

Signed-off-by: Martin Povišer <[email protected]> Co-developed-by: Janne Grunau <[email protected]> Signed-off-by: Janne Grunau <[email protected]>

Signed-off-by: Martin Povišer <[email protected]>

Signed-off-by: Janne Grunau <[email protected]>

Use a completion to do it from avep_init() instead. Signed-off-by: Janne Grunau <[email protected]>

Signed-off-by: Martin Povišer <[email protected]>

This was only necessary for dcp0 on M2* devices presumably because the reset in m1n1 doesn't work as intended. Signed-off-by: Janne Grunau <[email protected]>

This seems to get rid of initialization timeouts / failures. Signed-off-by: Janne Grunau <[email protected]>

Functionally a revert of "drm: apple: av: Do not open AV service from afk receive handler" with more workqueues. Signed-off-by: Janne Grunau <[email protected]>

Those two drivers are closely linked and should always exists together. Signed-off-by: Janne Grunau <[email protected]>

The main advantage is that it allows runtime PM which would have been manually implemented with the ad-hoc instantiated platform driver. This also probes the devices as component of the DRM driver which allows to simplify the the interface between the av endpoint and the audio driver. Signed-off-by: Janne Grunau <[email protected]>

Now that the DP audio driver is a component of the display sub-system probe errors will bring down the whole display initialization. To prevent that the audio driver must not fail. Allow delayed sound card initialization if the DMA controller is not ready, for example because the apple-sio module is missing (at all or just in the initeramfs). In the case apple-sio is available later provide as sysfs file "probe_snd_card" to trigger initialization. Signed-off-by: Janne Grunau <[email protected]>

Signed-off-by: Jonathan Gray <[email protected]>

Signed-off-by: Caspar Schutijser <[email protected]>

The custom commit_tail implementation stopped making after "drm/apple: Disable fake vblank IRQ machinery" which stopped calling drm_vblank_init(). Revert back to the standard helper implementation. Avoids or at least significantly reduces page flips taking approximately one frame time in kwin_wayland 6. Fixes: ("drm/apple: Switch to nonblocking commit handling") Signed-off-by: Janne Grunau <[email protected]>

Two callbacks for IOMFB::MemDescRelay seems to be dropped between 12.3 and 13.5 DCP firmware. This results in the renumbering of MemDescRelay::release_descriptor from D456 to D454. Noticed while when switching the display refresh rate to 50 Hz with a 14.5 system firmware on a M1 Max Macbook Pro. Signed-off-by: Janne Grunau <[email protected]>

The most likely cause for this is an unexpected callback form which the current driver doesn't recover. Warn only once about it. Signed-off-by: Janne Grunau <[email protected]>

Reduce log spam while errors are still likely due missing state checks.

Can be still enabled by adding `apple_dcp.hdmi_audio` the kernel command line. Signed-off-by: Janne Grunau <[email protected]>

Since we don't init/uses drm's vblank support our page flip timestamps are CLOCK_MONOTONIC timestamps during the event generation. Since compositors use the timestamp to schedule their next kms commit this is timing sensitive sop move it under the drivers control. Take the timestamp directly in the swap_complete callback. Framebuffer swaps are unfortunately not fast with DCP. Measured time from swap_submit to swap_complete is ~1.5 ms for dcp and ~2.3 ms for dcpext. This warrants further investigation. Presentation timestamps might help if delay on dcp firmware side occurs after the actual swap. In the meantime doctor the time stamps and move the page flip completion up to 1 ms earler. This fixes half rate refresh on external displays displays using dcpext. Signed-off-by: Janne Grunau <[email protected]>

Signed-off-by: Asahi Lina <[email protected]>

This works even before the DMA device probes. Might help deal with runtime-pm ordering, though it doesn't solve the deferred ordering problem (since we're creating the link while already probing)... Signed-off-by: Asahi Lina <[email protected]>

Allow the DMA device driver to probe late, and still create the sound device upfront. Instead try to request the DMA channel on first PCM open. This should be safe as long as we bail early and don't allow the process to continue to configuring buffers (since that requires the DMA to be configured). Signed-off-by: Asahi Lina <[email protected]>

Signed-off-by: Asahi Lina <[email protected]>

Signed-off-by: Sven Peter <[email protected]>

This appears to be lane count as "2" is observed for USB-C DP alt mode in shared DP/USB3 mode. Signed-off-by: Janne Grunau <[email protected]>

This unfortunately doesn't work relieably with typec-altmode-displayport since the oob hotplug notification arrives before atc-phy is configured to the appropiate DP mode. Signed-off-by: Janne Grunau <[email protected]>

Signed-off-by: Asahi Lina <[email protected]>

Fixes failure to unmap buffers in dcpep_cb_unmap_piodma() due to the unaligned size. Further along this causes kernel log splat when DCP tries to map the buffers again since thye IOVA is still in use. This causes no apparent issue although map_piodma callback signals an errror and returns 0 (unmapped as DVA). It's not clear why this presents only randomly. Possibly some build or uninitialized memory triggers this unmap/free and immediate allocate/map cycle in the DCP firmware. I never notices this with a clang-built kernel on j314c. It showed with gcc build with the Fedora config at least on 6.8.8 based kernels. This did not reproduce on j375d. Signed-off-by: Janne Grunau <[email protected]>

This reverts commit fa86f31. Signed-off-by: Janne Grunau <[email protected]>

The dpavserv endpoint uses various EPICProviderClass depending on the connected display. Observed values: - "AppleDCPAgileCDIDPDisplay" (j134c, dcp, panel) - "AppleDCPMCDP29XX" (j274, dcp, hdmi) - "AppleDCPPS190" (j474s, dcpext0, hdmi) - "DCPDPService" (j474s, dcpext1, typec) So match against against EPICName which is consistent in all cases. This also allows the distinction between 'dcpav-service-epic' and 'dcpdp-service-epic'. Not sure what the second EPIC service is used for. Signed-off-by: Janne Grunau <[email protected]>

Known uses EDID retrieval and raw I2C access. Signed-off-by: Janne Grunau <[email protected]>

External display only since the EDID provided by integrated panels holds no useful / correct information. Signed-off-by: Janne Grunau <[email protected]>

This driver implements support for the SMC (System Management Controller) in Apple Macs. In contrast to the existing applesmc driver, it uses pluggable backends that allow it to support different SMC implementations, and uses the MFD subsystem to expose the core SMC functionality so that specific features (gpio, hwmon, battery, etc.) can be implemented by separate drivers in their respective downstream subsystems. The initial RTKit backend adds support for Apple Silicon Macs (M1 et al). We hope a backend for T2 Macs will be written in the future (since those are not supported by applesmc), and eventually an x86 backend would allow us to fully deprecate applesmc in favor of this driver. Signed-off-by: Hector Martin <[email protected]>

This driver implements the GPIO service on top of the SMC framework on Apple Mac machines. In particular, these are the GPIOs present in the PMU IC which are used to control power to certain on-board devices. Although the underlying hardware supports various pin config settings (input/output, open drain, etc.), this driver does not implement that functionality and leaves it up to the firmware to configure things properly. We also don't yet support interrupts/events. This is sufficient for device power control, which is the only thing we need to support at this point. More features will be implemented when needed. To our knowledge, only Apple Silicon Macs implement this SMC feature. Signed-off-by: Hector Martin <[email protected]>

This driver implements support for battery stats on top of the macsmc framework, to support Apple M1 Mac machines. Co-authored-by: Joey Gouly <[email protected]> Signed-off-by: Hector Martin <[email protected]>

Signed-off-by: Hector Martin <[email protected]>

This driver implements the reboot/shutdown support exposed by the SMC on Apple Silicon machines, such as Apple M1 Macs. Signed-off-by: Hector Martin <[email protected]>

Apple Silicon Macs (M1, etc.) have an RTC that is part of the PMU IC, but most of the PMU functionality is abstracted out by the SMC. On T600x machines, the RTC counter must be accessed via the SMC to get full functionality, and it seems likely that future machines will move towards making SMC handle all RTC functionality. The SMC RTC counter access is implemented on all current machines as of the time of this writing, on firmware 12.x. However, the RTC offset (needed to set the time) is still only accessible via direct PMU access. To handle this, we expose the RTC offset as an NVMEM cell from the SPMI PMU device node, and this driver consumes that cell and uses it to compute/set the current time. Alarm functionality is not yet implemented. This would also go via the PMU today, but could change in the future. Signed-off-by: Hector Martin <[email protected]>

This driver implements power button and lid switch support for Apple Mac devices using SMC controllers driven by the macsmc driver. In addition to basic input support, this also responds to the final shutdown warning (when the power button is held down long enough) by doing an emergency kernel poweroff. This allows the NVMe controller to be cleanly shut down, which prevents data loss for in-cache data. Signed-off-by: Hector Martin <[email protected]>

Signed-off-by: Hector Martin <[email protected]>

The core enables notifications (NTAP) before returning from the probe function, before the backend has a chance to set smc->core. This leads to a race if a notification arrives early. We already rely on the core setting the drvdata to itself, so move that ahead of NTAP=1 and drop smc->core entirely. That way it's safe for the backend notification callback to fire before the core probe function returns. Signed-off-by: Hector Martin <[email protected]>

This helps catch s2idle power stats, since we get early data when the system resumes due to a power button press. Signed-off-by: Hector Martin <[email protected]>

This is a hardcoded charge threshold feature present in firmware 13.0 or newer. Userspace settings are rounded to one of the two possible behaviors. Signed-off-by: Hector Martin <[email protected]>

The generic handling if charge_behaviours in the power_supply core requires power_supply_desc.charge_behaviours to be set. Signed-off-by: Thomas Weißschuh <[email protected]>

Report more voltages from the battery, and also fudge energy numbers from charge numbers. This way userspace doesn't try to convert on its own (and gets it very wrong). Signed-off-by: Hector Martin <[email protected]>

"ioft" is a 48.16 fixed point type. Signed-off-by: Janne Grunau <[email protected]>

Apple Silicon devices, particularly Macs, continue Apple's tradition of integrating a ridiculous number of sensors on their products. Most of these report to the System Management Controller, and their data is exposed by the SMC firmware as simple key-value pairs. This commit adds an hwmon drive for retrieving the data reported by the temperature, voltage, current and power sensors, as well as fan data. Devices each expose their own unique set of sensors and fans, even with the SMC being baked into the SoC. This driver walks a devicetree node in order to discover the sensors present on the current device at runtime. Co-developed-by: Janne Grunau <[email protected]> Signed-off-by: Janne Grunau <[email protected]> Signed-off-by: James Calligeros <[email protected]>

Should be squashed into "hwmon: add macsmc-hwmon driver" Signed-off-by: Janne Grunau <[email protected]>

Since macOS Sequoia firmware, CHLS replaced CHWA and now allows an arbitrary end charge threshold to be configured. Prefer CHWA over CHLS since the SMC firmware from iBoot-10151.1.1 (macOS 14.0) is not compatible with our CHGLS usage. It was working with the SMC firmware from iBoot-10151.121.1 (macOS 14.5). Signed-off-by: Hector Martin <[email protected]> Signed-off-by: Janne Grunau <[email protected]>

Gone in Sequoia firmware. Signed-off-by: Hector Martin <[email protected]>

If a CHLS charge threshold is configured and the current SoC is above the start threshold report a busy BMS as not charging. Signed-off-by: Janne Grunau <[email protected]>

Signed-off-by: Janne Grunau <[email protected]>

The SMC provides an interface for manually controlling the speeds of any fans attached to it. Expose this via the standard hwmon interface. Once a fan is in manual control, the SMC makes no attempts to save users from themselves. It is possible to write arbitrary values outside of the SMC's reported safe range. The driver therefore does its own sanity checking. Since we are unsure whether or not leaving the fans in manual mode can cause damage to Apple Silicon devices, this functionality is gated behind a very explicit and scary-sounding unsafe module parameter. Signed-off-by: James Calligeros <[email protected]>

Many entries in the USB-audio quirk tables have relatively complex expressions. For improving the readability, introduce a few macros. Those are applied in the following patch. Link: https://patch.msgid.link/[email protected] Signed-off-by: Takashi Iwai <[email protected]>

Apply the newly introduced macros for reduce the complex expressions and cast in the quirk table definitions. It results in a significant code reduction, too. There should be no functional changes. Link: https://patch.msgid.link/[email protected] Signed-off-by: Takashi Iwai <[email protected]>

Add trivial support for audio streaming on the RME Digiface USB. Binds only to the first interface to allow userspace to directly drive the complex I/O and matrix mixer controls. Signed-off-by: Cyan Nyan <[email protected]> [Lina: Added 2x/4x sample rate support & boot/format quirks] Co-developed-by: Asahi Lina <[email protected]> Signed-off-by: Asahi Lina <[email protected]> Link: https://patch.msgid.link/[email protected] Signed-off-by: Takashi Iwai <[email protected]>

Implement sync, output format, and input status mixer controls, to allow the interface to be used as a straight ADAT/SPDIF (+ Headphones) I/O interface. This does not implement the matrix mixer, output gain controls, or input level meter feedback. The full mixer interface is only really usable using a dedicated userspace control app (there are too many mixer nodes for alsamixer to be usable), so for now we leave it up to userspace to directly control these features using raw USB control messages. This is similar to how it's done with some FireWire interfaces (ffado-mixer). Signed-off-by: Asahi Lina <[email protected]> Link: https://patch.msgid.link/[email protected] Signed-off-by: Takashi Iwai <[email protected]>

…able() power_supply_property_is_writeable() gets called from the is_visible() callback for the sysfs attributes of power_supply class devices and for the sysfs attributes of power_supply core instantiated hwmon class devices. These sysfs attributes get registered by the device_add() respectively power_supply_add_hwmon_sysfs() calls in power_supply_register(). use_cnt gets initialized to 0 and is incremented only after these calls. So when power_supply_property_is_writeable() gets called it always return -ENODEV because of use_cnt == 0. This causes all the attributes to have permissions of 444 even those which should be writable. This used to be a problem only for hwmon sysfs attributes but since commit be6299c ("power: supply: sysfs: use power_supply_property_is_writeable()") this now also impacts power_supply class sysfs attributes. Fixes: be6299c ("power: supply: sysfs: use power_supply_property_is_writeable()") Fixes: e67d4df ("power: supply: Add HWMON compatibility layer") Cc: [email protected] Cc: Thomas Weißschuh <[email protected]> Cc: Andrey Smirnov <[email protected]> Signed-off-by: Hans de Goede <[email protected]>

Fixes a race condition reported here: AsahiLinux#309 (comment) The whole premise of lockless access to a single-producer-single- consumer queue is that there is just a single producer and single consumer. That means we can't call drm_sched_can_queue() (which is about queueing more work to the hw, not to the spsc queue) from anywhere other than the consumer (wq). This call in the producer is just an optimization to avoid scheduling the consuming worker if it cannot yet queue more work to the hw. It is safe to drop this optimization to avoid the race condition. Suggested-by: Asahi Lina <[email protected]> Fixes: a78422e ("drm/sched: implement dynamic job-flow control") Closes: AsahiLinux#309 Cc: [email protected] Signed-off-by: Rob Clark <[email protected]>

Signed-off-by: Janne Grunau <[email protected]>

Signed-off-by: Asahi Lina <[email protected]>

fixup! rust: Add `name` argument to Module::init() Signed-off-by: Janne Grunau <[email protected]>

Signed-off-by: Asahi Lina <[email protected]>

Expected to be dropped once upstream removes `remove_new(). Signed-off-by: Janne Grunau <[email protected]>

Signed-off-by: Janne Grunau <[email protected]>

Signed-off-by: Asahi Lina <[email protected]>

Makes debugging a bit easier. Signed-off-by: Asahi Lina <[email protected]>

Signed-off-by: Asahi Lina <[email protected]>

This eliminates a potential deadlock under load and improves the fence signaling situation (for when we have a shrinker). Signed-off-by: Asahi Lina <[email protected]>

This only allows binding one VM context at once, which serializes GPU usage between VMs and therefore prevents one faulting VM from affecting others. Signed-off-by: Asahi Lina <[email protected]>

Signed-off-by: Martin Povišer <[email protected]>

*** possible v6.11 conflict: _snd_dmaengine_pcm_close Signed-off-by: Martin Povišer <[email protected]>

Enables j313, j293, j493, j314, j414, j274, j375, j473, j474, j475 *** WARNING FOR DISTRO PACKAGERS WANTING TO APPLY THIS: *** *** YOU ABSOLUTELY NEED THIS PATCH IN YOUR LSP-PLUGINS PACKAGE *** lsp-plugins/lsp-dsp-lib#20 Do NOT enable speakers without that patch, on any model. It can/will result in nasty noise that could damage them. Signed-off-by: Hector Martin <[email protected]>

Enables j316, j413, j415, j416 *** WARNING FOR DISTRO PACKAGERS WANTING TO APPLY THIS: *** *** YOU ABSOLUTELY NEED THIS PATCH IN YOUR LSP-PLUGINS PACKAGE *** lsp-plugins/lsp-dsp-lib#20 Do NOT enable speakers without that patch, on any model. It can/will result in nasty noise that could damage them. Signed-off-by: Hector Martin <[email protected]>

Possibly because we don't have support in the helper program, this is broken and causes channel errors. Hack in high priority for now, which works around it. Use debug_flags 0x1000000000000 to re-enable for testing. Signed-off-by: Asahi Lina <[email protected]>

After commit c96b547: rust: Add `name` argument to Module::init() init needs a name argument. Since both rust samples were made before that, they fail to compile. Signed-off-by: Bruno Oliveira <[email protected]>

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

rust: samples: fix init argument #335

rust: samples: fix init argument #335

Commits on Sep 16, 2024

Commits on Sep 18, 2024

Commits on Sep 21, 2024

Commits on Sep 23, 2024

Commits on Sep 24, 2024

Commits on Oct 18, 2024

rust: samples: fix init argument #335

Are you sure you want to change the base?

rust: samples: fix init argument #335

Commits on Sep 16, 2024

Commits on Sep 18, 2024

Commits on Sep 21, 2024

Commits on Sep 23, 2024

Commits on Sep 24, 2024

Commits on Oct 18, 2024