class Server

Initializes the CQHCI driver, providing it with all necessary resources.

This call is expected to be made once. The server should reject any additional calls

until the peer end of `virtual_interrupt_lifeline` closes.

# Interrupt Delegation

While the CQHCI driver is running, it takes over the physical IRQ object from its underlying

driver, and directly handles any CQHCI-specific interrupts. Interrupts which it cannot

service (e.g. SDHCI interrupts) are forwarded to the underlying driver via

`virtual_interrupt`. The server should service interrupts exactly as it normally would.

The CQHCI driver will wait for the virtual IRQ to be acked before acking the physical IRQ.

# Arguments

* `virtual_interrupt` - A virtual interrupt to monitor instead of the physical IRQ.

The CQHCI driver will use this to forward interrupts which it cannot directly service.

* `virtual_interrupt_lifeline` - A lifeline for the virtual interrupt. The CQHCI

driver will destroy the peer end upon shutdown, which should be handled by the server as a

signal to stop monitoring the virtual interrupt and return to monitoring the physical

interrupt.

# Returns

* `cqhci_mmio` and `cqhci_mmio_offset` - An MMIO for the CQHCI register block.

MMIO size minus offset must be at least 512 bytes as per the CQHCI specification.

* `sdhci_mmio` and `sdhci_mmio_offset` - An MMIO for the SDHCI register block.

MMIO size minus offset must be at least 512 bytes as per the SDHCI specification.

* `bti` - The BTI to use for pinning pages for DMA.

* `interrupt` - The physical interrupt to monitor.

Enables the Command Queueing Engine. The CQHCI driver is expected to already have called

[`InitializeCommandQueueing`] to handle interrupts.

While CQHCI is enabled, it is the caller's responsibility to only submit regular requests

(via [`Request`]) via DCMD, or after disabling CQHCI.

Disables the Command Queueing Engine. `EnableCqhci` must have previously been called. From

this point onwards, regular requests can be submitted (until `EnableCqhci` is called again).

Get host info.

Set signal voltage.

Set bus width.

Set bus frequency, zero means disable the clock to the card.

Set mmc timing.

Issue a hw reset.

Perform tuning.

Register the given callback to be called when an in-band interrupt is received from the

card. Before calling the callback the protocol implementation will disable the in-band

interrupt in the controller. In-band interrupts can be re-enabled by calling

`AckInBandInterrupt()` once the previous interrupt has been handled by the client.

In the methods below, vmo_id is used to uniquely identify a VMO that will be passed to

Request in an SdmmcBufferRegion. VMO IDs are chosen by the caller, and may be any uint32

value.

Registers a VMO and transfers ownership to the protocol implementation. vmo_rights is a bit

field containing SdmmcVmoRight values, and determines the read/write permissions used by

the implementation when pinning or mapping the VMO. The implementation may pin vmo during

this call or any time it is used in a request, and may keep it pinned until the VMO is

unregistered. client_id may be in [0, SDMMC_MAX_CLIENT_ID] and identifies the ID space for

this VMO to be registered in (that is, two different VMOs may use the same ID if they are

registered for different clients).

The callee unmaps/unpins the VMO and returns ownership to the caller.

Perform the requests in order, atomically (i.e., no intervening requests). Furthermore,

calls are replied to in the order they are received. If an error occurs, abort any remaining

requests. Else, return the response of the last request.

The protocol implementation chooses whether or not to use DMA

depending on the properties of the request and the capabilities of the controller.

Clients are responsible for performing the following cache operations:

After read requests:

- Call zx_cache_flush with ZX_CACHE_FLUSH_DATA | ZX_CACHE_FLUSH_INVALIDATE on buffers that

have been mapped by the client.

- Call zx_vmo_op_range with ZX_VMO_OP_CACHE_CLEAN_INVALIDATE on all other buffers.

Note that writing to any portion of a buffer before Request has returned can corrupt the

received data.

Before write requests:

- Call zx_cache_flush with ZX_CACHE_FLUSH_DATA on buffers that have been mapped by the

client.

- Call zx_vmo_op_range with ZX_VMO_OP_CACHE_CLEAN on all other buffers.

|bind_handler| returns a handler that binds incoming connections to this

server implementation.

The returned handler borrows the server instance.

The server must outlive the provided |dispatcher|. Only after

the dispatcher is shutdown will it be safe to destroy the servers.

The server should not be moved.