linux-stable.git/drivers/soc/apple, branch linux-6.3.y

soc: apple: rtkit: Do not copy the reg state structure to the stack

2023-02-13T19:20:34+00:00

The register state struct is 848 bytes, which ends up bloating the
apple_rtkit_crashlog_dump_regs stack frame beyond 1024 on some
32-bit platforms, triggering compile warnings.

This doesn't matter for 64BIT/ARM64, but there's also no good reason to
copy the structure to the stack in this case. We can use __packed to
avoid alignment issues, there are no double-read hazards, and this is a
fatal error path so performance does not matter.

Fixes: 22991d8d5725 ("soc: apple: rtkit: Add register dump decoding to crashlog")
Signed-off-by: Asahi Lina 
Reviewed-by: Eric Curtin 
Signed-off-by: Arnd Bergmann

soc: apple: rtkit: Add register dump decoding to crashlog

2023-01-31T11:44:47+00:00

When the coprocessor crashes, it's useful to get a proper register dump
so we can find out what the firmware was doing. Add a decoder for this.

Originally this had ESR decoding by reusing the ARM64 arch header for
this, but that introduces some module linking and cross-arch compilation
issues, so let's leave that out for now.

Reviewed-by: Sven Peter 
Reviewed-by: Eric Curtin 
Signed-off-by: Asahi Lina 
Signed-off-by: Hector Martin

soc: apple: rtkit: Export non-devm init/free functions

2023-01-31T11:40:14+00:00

While we normally encourage devm usage by drivers, some consumers (and
in particular the upcoming Rust abstractions) might want to manually
manage memory. Export the raw functions to make this possible.

Signed-off-by: Asahi Lina 
Reviewed-by: Sven Peter 
Reviewed-by: Eric Curtin 
Signed-off-by: Hector Martin

soc: apple: apple-pmgr-pwrstate: Switch to IRQ-safe mode

2023-01-31T11:37:07+00:00

This requires changing the reset path locking primitives to the spinlock
path in genpd, instead of the mutex path.

Reviewed-by: Eric Curtin 
Reviewed-by: Sven Peter 
Signed-off-by: Hector Martin

soc: apple: rtkit: Add apple_rtkit_idle() function

2023-01-31T11:35:47+00:00

This is yet another low power mode, used by DCP.

Reviewed-by: Eric Curtin 
Reviewed-by: Sven Peter 
Signed-off-by: Hector Martin

soc: apple: rtkit: Stop casting function pointer signatures

2022-11-28T11:34:09+00:00

Fixes: 9bd1d9a0d8bb ("soc: apple: Add RTKit IPC library")
Reported-by: kernel test robot 
Signed-off-by: Sven Peter 
Signed-off-by: Hector Martin

soc: apple: sart: Stop casting function pointer signatures

2022-11-28T11:34:09+00:00

Fixes: b170143ae111 ("soc: apple: Add SART driver")
Reported-by: kernel test robot 
Signed-off-by: Sven Peter 
Signed-off-by: Hector Martin

soc: apple: rtkit: Add apple_rtkit_poll

2022-09-17T17:53:29+00:00

This allows a client to receive messages in atomic context, by polling.

Signed-off-by: Hector Martin 
Signed-off-by: Russell King (Oracle) 
Reviewed-by: Sven Peter 
Reviewed-by: Eric Curtin 
Reviewed-by: Linus Walleij 
Signed-off-by: Arnd Bergmann

soc: apple: Add SART driver

2022-05-02T15:24:45+00:00

The NVMe co-processor on the Apple M1 uses a DMA address filter called
SART for some DMA transactions. This adds a simple driver used to
configure the memory regions from which DMA transactions are allowed.

Unlike a real IOMMU, SART does not support any pagetables and can't be
implemented inside the IOMMU subsystem using iommu_ops.

It also can't be implemented using dma_map_ops since not all DMA
transactions of the NVMe controller are filtered by SART.
Instead, most buffers have to be registered using the integrated NVMe
IOMMU and we can't have two separate dma_map_ops implementations for a
single device.

Co-developed-by: Hector Martin 
Reviewed-by: Arnd Bergmann 
Signed-off-by: Hector Martin 
Signed-off-by: Sven Peter

soc: apple: Add RTKit IPC library

2022-05-01T14:55:06+00:00

Apple SoCs such as the M1 come with multiple embedded co-processors
running proprietary firmware. Communication with those is established
over a simple mailbox using the RTKit IPC protocol.

This cannot be implemented inside the mailbox subsystem since on top
of communication over channels we also need support for starting,
hibernating and resetting these co-processors. We also need to
handle shared memory allocations differently depending on the
co-processor and don't want to split that across multiple drivers.

Reviewed-by: Arnd Bergmann 
Signed-off-by: Sven Peter