linux/Documentation/DocBook/device-drivers.tmpl

459 lines
15 KiB
Cheetah
Raw Normal View History

<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE book PUBLIC "-//OASIS//DTD DocBook XML V4.1.2//EN"
"http://www.oasis-open.org/docbook/xml/4.1.2/docbookx.dtd" []>
<book id="LinuxDriversAPI">
<bookinfo>
<title>Linux Device Drivers</title>
<legalnotice>
<para>
This documentation is free software; you can redistribute
it and/or modify it under the terms of the GNU General Public
License as published by the Free Software Foundation; either
version 2 of the License, or (at your option) any later
version.
</para>
<para>
This program is distributed in the hope that it will be
useful, but WITHOUT ANY WARRANTY; without even the implied
warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
See the GNU General Public License for more details.
</para>
<para>
You should have received a copy of the GNU General Public
License along with this program; if not, write to the Free
Software Foundation, Inc., 59 Temple Place, Suite 330, Boston,
MA 02111-1307 USA
</para>
<para>
For more details see the file COPYING in the source
distribution of Linux.
</para>
</legalnotice>
</bookinfo>
<toc></toc>
<chapter id="Basics">
<title>Driver Basics</title>
<sect1><title>Driver Entry and Exit points</title>
!Iinclude/linux/init.h
</sect1>
<sect1><title>Atomic and pointer manipulation</title>
!Iarch/x86/include/asm/atomic.h
</sect1>
<sect1><title>Delaying, scheduling, and timer routines</title>
!Iinclude/linux/sched.h
!Ekernel/sched/core.c
!Ikernel/sched/cpupri.c
!Ikernel/sched/fair.c
!Iinclude/linux/completion.h
!Ekernel/time/timer.c
</sect1>
<sect1><title>Wait queues and Wake events</title>
!Iinclude/linux/wait.h
!Ekernel/sched/wait.c
</sect1>
<sect1><title>High-resolution timers</title>
!Iinclude/linux/ktime.h
!Iinclude/linux/hrtimer.h
!Ekernel/time/hrtimer.c
</sect1>
<sect1><title>Workqueues and Kevents</title>
!Ekernel/workqueue.c
</sect1>
<sect1><title>Internal Functions</title>
!Ikernel/exit.c
!Ikernel/signal.c
!Iinclude/linux/kthread.h
!Ekernel/kthread.c
</sect1>
<sect1><title>Kernel objects manipulation</title>
<!--
X!Iinclude/linux/kobject.h
-->
!Elib/kobject.c
</sect1>
<sect1><title>Kernel utility functions</title>
!Iinclude/linux/kernel.h
!Ekernel/printk/printk.c
!Ekernel/panic.c
!Ekernel/sys.c
!Ekernel/rcu/srcu.c
!Ekernel/rcu/tree.c
!Ekernel/rcu/tree_plugin.h
!Ekernel/rcu/update.c
</sect1>
<sect1><title>Device Resource Management</title>
!Edrivers/base/devres.c
</sect1>
</chapter>
<chapter id="devdrivers">
<title>Device drivers infrastructure</title>
<sect1><title>The Basic Device Driver-Model Structures </title>
!Iinclude/linux/device.h
</sect1>
<sect1><title>Device Drivers Base</title>
!Idrivers/base/init.c
!Edrivers/base/driver.c
!Edrivers/base/core.c
!Edrivers/base/syscore.c
!Edrivers/base/class.c
!Idrivers/base/node.c
!Edrivers/base/firmware_class.c
!Edrivers/base/transport_class.c
<!-- Cannot be included, because
attribute_container_add_class_device_adapter
and attribute_container_classdev_to_container
exceed allowed 44 characters maximum
X!Edrivers/base/attribute_container.c
-->
!Edrivers/base/dd.c
<!--
X!Edrivers/base/interface.c
-->
!Iinclude/linux/platform_device.h
!Edrivers/base/platform.c
!Edrivers/base/bus.c
</sect1>
<sect1><title>Device Drivers DMA Management</title>
!Edrivers/dma-buf/dma-buf.c
fence: dma-buf cross-device synchronization (v18) A fence can be attached to a buffer which is being filled or consumed by hw, to allow userspace to pass the buffer without waiting to another device. For example, userspace can call page_flip ioctl to display the next frame of graphics after kicking the GPU but while the GPU is still rendering. The display device sharing the buffer with the GPU would attach a callback to get notified when the GPU's rendering-complete IRQ fires, to update the scan-out address of the display, without having to wake up userspace. A driver must allocate a fence context for each execution ring that can run in parallel. The function for this takes an argument with how many contexts to allocate: + fence_context_alloc() A fence is transient, one-shot deal. It is allocated and attached to one or more dma-buf's. When the one that attached it is done, with the pending operation, it can signal the fence: + fence_signal() To have a rough approximation whether a fence is fired, call: + fence_is_signaled() The dma-buf-mgr handles tracking, and waiting on, the fences associated with a dma-buf. The one pending on the fence can add an async callback: + fence_add_callback() The callback can optionally be cancelled with: + fence_remove_callback() To wait synchronously, optionally with a timeout: + fence_wait() + fence_wait_timeout() When emitting a fence, call: + trace_fence_emit() To annotate that a fence is blocking on another fence, call: + trace_fence_annotate_wait_on(fence, on_fence) A default software-only implementation is provided, which can be used by drivers attaching a fence to a buffer when they have no other means for hw sync. But a memory backed fence is also envisioned, because it is common that GPU's can write to, or poll on some memory location for synchronization. For example: fence = custom_get_fence(...); if ((seqno_fence = to_seqno_fence(fence)) != NULL) { dma_buf *fence_buf = seqno_fence->sync_buf; get_dma_buf(fence_buf); ... tell the hw the memory location to wait ... custom_wait_on(fence_buf, seqno_fence->seqno_ofs, fence->seqno); } else { /* fall-back to sw sync * / fence_add_callback(fence, my_cb); } On SoC platforms, if some other hw mechanism is provided for synchronizing between IP blocks, it could be supported as an alternate implementation with it's own fence ops in a similar way. enable_signaling callback is used to provide sw signaling in case a cpu waiter is requested or no compatible hardware signaling could be used. The intention is to provide a userspace interface (presumably via eventfd) later, to be used in conjunction with dma-buf's mmap support for sw access to buffers (or for userspace apps that would prefer to do their own synchronization). v1: Original v2: After discussion w/ danvet and mlankhorst on #dri-devel, we decided that dma-fence didn't need to care about the sw->hw signaling path (it can be handled same as sw->sw case), and therefore the fence->ops can be simplified and more handled in the core. So remove the signal, add_callback, cancel_callback, and wait ops, and replace with a simple enable_signaling() op which can be used to inform a fence supporting hw->hw signaling that one or more devices which do not support hw signaling are waiting (and therefore it should enable an irq or do whatever is necessary in order that the CPU is notified when the fence is passed). v3: Fix locking fail in attach_fence() and get_fence() v4: Remove tie-in w/ dma-buf.. after discussion w/ danvet and mlankorst we decided that we need to be able to attach one fence to N dma-buf's, so using the list_head in dma-fence struct would be problematic. v5: [ Maarten Lankhorst ] Updated for dma-bikeshed-fence and dma-buf-manager. v6: [ Maarten Lankhorst ] I removed dma_fence_cancel_callback and some comments about checking if fence fired or not. This is broken by design. waitqueue_active during destruction is now fatal, since the signaller should be holding a reference in enable_signalling until it signalled the fence. Pass the original dma_fence_cb along, and call __remove_wait in the dma_fence_callback handler, so that no cleanup needs to be performed. v7: [ Maarten Lankhorst ] Set cb->func and only enable sw signaling if fence wasn't signaled yet, for example for hardware fences that may choose to signal blindly. v8: [ Maarten Lankhorst ] Tons of tiny fixes, moved __dma_fence_init to header and fixed include mess. dma-fence.h now includes dma-buf.h All members are now initialized, so kmalloc can be used for allocating a dma-fence. More documentation added. v9: Change compiler bitfields to flags, change return type of enable_signaling to bool. Rework dma_fence_wait. Added dma_fence_is_signaled and dma_fence_wait_timeout. s/dma// and change exports to non GPL. Added fence_is_signaled and fence_enable_sw_signaling calls, add ability to override default wait operation. v10: remove event_queue, use a custom list, export try_to_wake_up from scheduler. Remove fence lock and use a global spinlock instead, this should hopefully remove all the locking headaches I was having on trying to implement this. enable_signaling is called with this lock held. v11: Use atomic ops for flags, lifting the need for some spin_lock_irqsaves. However I kept the guarantee that after fence_signal returns, it is guaranteed that enable_signaling has either been called to completion, or will not be called any more. Add contexts and seqno to base fence implementation. This allows you to wait for less fences, by testing for seqno + signaled, and then only wait on the later fence. Add FENCE_TRACE, FENCE_WARN, and FENCE_ERR. This makes debugging easier. An CONFIG_DEBUG_FENCE will be added to turn off the FENCE_TRACE spam, and another runtime option can turn it off at runtime. v12: Add CONFIG_FENCE_TRACE. Add missing documentation for the fence->context and fence->seqno members. v13: Fixup CONFIG_FENCE_TRACE kconfig description. Move fence_context_alloc to fence. Simplify fence_later. Kill priv member to fence_cb. v14: Remove priv argument from fence_add_callback, oops! v15: Remove priv from documentation. Explicitly include linux/atomic.h. v16: Add trace events. Import changes required by android syncpoints. v17: Use wake_up_state instead of try_to_wake_up. (Colin Cross) Fix up commit description for seqno_fence. (Rob Clark) v18: Rename release_fence to fence_release. Move to drivers/dma-buf/. Rename __fence_is_signaled and __fence_signal to *_locked. Rename __fence_init to fence_init. Make fence_default_wait return a signed long, and fix wait ops too. Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com> Signed-off-by: Thierry Reding <thierry.reding@gmail.com> #use smp_mb__before_atomic() Acked-by: Sumit Semwal <sumit.semwal@linaro.org> Acked-by: Daniel Vetter <daniel@ffwll.ch> Reviewed-by: Rob Clark <robdclark@gmail.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2014-07-01 18:57:14 +08:00
!Edrivers/dma-buf/fence.c
!Edrivers/dma-buf/seqno-fence.c
fence: dma-buf cross-device synchronization (v18) A fence can be attached to a buffer which is being filled or consumed by hw, to allow userspace to pass the buffer without waiting to another device. For example, userspace can call page_flip ioctl to display the next frame of graphics after kicking the GPU but while the GPU is still rendering. The display device sharing the buffer with the GPU would attach a callback to get notified when the GPU's rendering-complete IRQ fires, to update the scan-out address of the display, without having to wake up userspace. A driver must allocate a fence context for each execution ring that can run in parallel. The function for this takes an argument with how many contexts to allocate: + fence_context_alloc() A fence is transient, one-shot deal. It is allocated and attached to one or more dma-buf's. When the one that attached it is done, with the pending operation, it can signal the fence: + fence_signal() To have a rough approximation whether a fence is fired, call: + fence_is_signaled() The dma-buf-mgr handles tracking, and waiting on, the fences associated with a dma-buf. The one pending on the fence can add an async callback: + fence_add_callback() The callback can optionally be cancelled with: + fence_remove_callback() To wait synchronously, optionally with a timeout: + fence_wait() + fence_wait_timeout() When emitting a fence, call: + trace_fence_emit() To annotate that a fence is blocking on another fence, call: + trace_fence_annotate_wait_on(fence, on_fence) A default software-only implementation is provided, which can be used by drivers attaching a fence to a buffer when they have no other means for hw sync. But a memory backed fence is also envisioned, because it is common that GPU's can write to, or poll on some memory location for synchronization. For example: fence = custom_get_fence(...); if ((seqno_fence = to_seqno_fence(fence)) != NULL) { dma_buf *fence_buf = seqno_fence->sync_buf; get_dma_buf(fence_buf); ... tell the hw the memory location to wait ... custom_wait_on(fence_buf, seqno_fence->seqno_ofs, fence->seqno); } else { /* fall-back to sw sync * / fence_add_callback(fence, my_cb); } On SoC platforms, if some other hw mechanism is provided for synchronizing between IP blocks, it could be supported as an alternate implementation with it's own fence ops in a similar way. enable_signaling callback is used to provide sw signaling in case a cpu waiter is requested or no compatible hardware signaling could be used. The intention is to provide a userspace interface (presumably via eventfd) later, to be used in conjunction with dma-buf's mmap support for sw access to buffers (or for userspace apps that would prefer to do their own synchronization). v1: Original v2: After discussion w/ danvet and mlankhorst on #dri-devel, we decided that dma-fence didn't need to care about the sw->hw signaling path (it can be handled same as sw->sw case), and therefore the fence->ops can be simplified and more handled in the core. So remove the signal, add_callback, cancel_callback, and wait ops, and replace with a simple enable_signaling() op which can be used to inform a fence supporting hw->hw signaling that one or more devices which do not support hw signaling are waiting (and therefore it should enable an irq or do whatever is necessary in order that the CPU is notified when the fence is passed). v3: Fix locking fail in attach_fence() and get_fence() v4: Remove tie-in w/ dma-buf.. after discussion w/ danvet and mlankorst we decided that we need to be able to attach one fence to N dma-buf's, so using the list_head in dma-fence struct would be problematic. v5: [ Maarten Lankhorst ] Updated for dma-bikeshed-fence and dma-buf-manager. v6: [ Maarten Lankhorst ] I removed dma_fence_cancel_callback and some comments about checking if fence fired or not. This is broken by design. waitqueue_active during destruction is now fatal, since the signaller should be holding a reference in enable_signalling until it signalled the fence. Pass the original dma_fence_cb along, and call __remove_wait in the dma_fence_callback handler, so that no cleanup needs to be performed. v7: [ Maarten Lankhorst ] Set cb->func and only enable sw signaling if fence wasn't signaled yet, for example for hardware fences that may choose to signal blindly. v8: [ Maarten Lankhorst ] Tons of tiny fixes, moved __dma_fence_init to header and fixed include mess. dma-fence.h now includes dma-buf.h All members are now initialized, so kmalloc can be used for allocating a dma-fence. More documentation added. v9: Change compiler bitfields to flags, change return type of enable_signaling to bool. Rework dma_fence_wait. Added dma_fence_is_signaled and dma_fence_wait_timeout. s/dma// and change exports to non GPL. Added fence_is_signaled and fence_enable_sw_signaling calls, add ability to override default wait operation. v10: remove event_queue, use a custom list, export try_to_wake_up from scheduler. Remove fence lock and use a global spinlock instead, this should hopefully remove all the locking headaches I was having on trying to implement this. enable_signaling is called with this lock held. v11: Use atomic ops for flags, lifting the need for some spin_lock_irqsaves. However I kept the guarantee that after fence_signal returns, it is guaranteed that enable_signaling has either been called to completion, or will not be called any more. Add contexts and seqno to base fence implementation. This allows you to wait for less fences, by testing for seqno + signaled, and then only wait on the later fence. Add FENCE_TRACE, FENCE_WARN, and FENCE_ERR. This makes debugging easier. An CONFIG_DEBUG_FENCE will be added to turn off the FENCE_TRACE spam, and another runtime option can turn it off at runtime. v12: Add CONFIG_FENCE_TRACE. Add missing documentation for the fence->context and fence->seqno members. v13: Fixup CONFIG_FENCE_TRACE kconfig description. Move fence_context_alloc to fence. Simplify fence_later. Kill priv member to fence_cb. v14: Remove priv argument from fence_add_callback, oops! v15: Remove priv from documentation. Explicitly include linux/atomic.h. v16: Add trace events. Import changes required by android syncpoints. v17: Use wake_up_state instead of try_to_wake_up. (Colin Cross) Fix up commit description for seqno_fence. (Rob Clark) v18: Rename release_fence to fence_release. Move to drivers/dma-buf/. Rename __fence_is_signaled and __fence_signal to *_locked. Rename __fence_init to fence_init. Make fence_default_wait return a signed long, and fix wait ops too. Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com> Signed-off-by: Thierry Reding <thierry.reding@gmail.com> #use smp_mb__before_atomic() Acked-by: Sumit Semwal <sumit.semwal@linaro.org> Acked-by: Daniel Vetter <daniel@ffwll.ch> Reviewed-by: Rob Clark <robdclark@gmail.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2014-07-01 18:57:14 +08:00
!Iinclude/linux/fence.h
!Iinclude/linux/seqno-fence.h
!Edrivers/dma-buf/reservation.c
!Iinclude/linux/reservation.h
!Edrivers/base/dma-coherent.c
!Edrivers/base/dma-mapping.c
</sect1>
<sect1><title>Device Drivers Power Management</title>
!Edrivers/base/power/main.c
</sect1>
<sect1><title>Device Drivers ACPI Support</title>
<!-- Internal functions only
X!Edrivers/acpi/sleep/main.c
X!Edrivers/acpi/sleep/wakeup.c
X!Edrivers/acpi/motherboard.c
X!Edrivers/acpi/bus.c
-->
!Edrivers/acpi/scan.c
!Idrivers/acpi/scan.c
<!-- No correct structured comments
X!Edrivers/acpi/pci_bind.c
-->
</sect1>
<sect1><title>Device drivers PnP support</title>
!Idrivers/pnp/core.c
<!-- No correct structured comments
X!Edrivers/pnp/system.c
-->
!Edrivers/pnp/card.c
!Idrivers/pnp/driver.c
!Edrivers/pnp/manager.c
!Edrivers/pnp/support.c
</sect1>
<sect1><title>Userspace IO devices</title>
!Edrivers/uio/uio.c
!Iinclude/linux/uio_driver.h
</sect1>
</chapter>
<chapter id="parportdev">
<title>Parallel Port Devices</title>
!Iinclude/linux/parport.h
!Edrivers/parport/ieee1284.c
!Edrivers/parport/share.c
!Idrivers/parport/daisy.c
</chapter>
<chapter id="message_devices">
<title>Message-based devices</title>
<sect1><title>Fusion message devices</title>
!Edrivers/message/fusion/mptbase.c
!Idrivers/message/fusion/mptbase.c
!Edrivers/message/fusion/mptscsih.c
!Idrivers/message/fusion/mptscsih.c
!Idrivers/message/fusion/mptctl.c
!Idrivers/message/fusion/mptspi.c
!Idrivers/message/fusion/mptfc.c
!Idrivers/message/fusion/mptlan.c
</sect1>
</chapter>
<chapter id="snddev">
<title>Sound Devices</title>
!Iinclude/sound/core.h
!Esound/sound_core.c
!Iinclude/sound/pcm.h
!Esound/core/pcm.c
!Esound/core/device.c
!Esound/core/info.c
!Esound/core/rawmidi.c
!Esound/core/sound.c
!Esound/core/memory.c
!Esound/core/pcm_memory.c
!Esound/core/init.c
!Esound/core/isadma.c
!Esound/core/control.c
!Esound/core/pcm_lib.c
!Esound/core/hwdep.c
!Esound/core/pcm_native.c
!Esound/core/memalloc.c
<!-- FIXME: Removed for now since no structured comments in source
X!Isound/sound_firmware.c
-->
</chapter>
<chapter id="uart16x50">
<title>16x50 UART Driver</title>
!Edrivers/tty/serial/serial_core.c
!Edrivers/tty/serial/8250/8250_core.c
</chapter>
<chapter id="fbdev">
<title>Frame Buffer Library</title>
<para>
The frame buffer drivers depend heavily on four data structures.
These structures are declared in include/linux/fb.h. They are
fb_info, fb_var_screeninfo, fb_fix_screeninfo and fb_monospecs.
The last three can be made available to and from userland.
</para>
<para>
fb_info defines the current state of a particular video card.
Inside fb_info, there exists a fb_ops structure which is a
collection of needed functions to make fbdev and fbcon work.
fb_info is only visible to the kernel.
</para>
<para>
fb_var_screeninfo is used to describe the features of a video card
that are user defined. With fb_var_screeninfo, things such as
depth and the resolution may be defined.
</para>
<para>
The next structure is fb_fix_screeninfo. This defines the
properties of a card that are created when a mode is set and can't
be changed otherwise. A good example of this is the start of the
frame buffer memory. This "locks" the address of the frame buffer
memory, so that it cannot be changed or moved.
</para>
<para>
The last structure is fb_monospecs. In the old API, there was
little importance for fb_monospecs. This allowed for forbidden things
such as setting a mode of 800x600 on a fix frequency monitor. With
the new API, fb_monospecs prevents such things, and if used
correctly, can prevent a monitor from being cooked. fb_monospecs
will not be useful until kernels 2.5.x.
</para>
<sect1><title>Frame Buffer Memory</title>
!Edrivers/video/fbdev/core/fbmem.c
</sect1>
<!--
<sect1><title>Frame Buffer Console</title>
X!Edrivers/video/console/fbcon.c
</sect1>
-->
<sect1><title>Frame Buffer Colormap</title>
!Edrivers/video/fbdev/core/fbcmap.c
</sect1>
<!-- FIXME:
drivers/video/fbgen.c has no docs, which stuffs up the sgml. Comment
out until somebody adds docs. KAO
<sect1><title>Frame Buffer Generic Functions</title>
X!Idrivers/video/fbgen.c
</sect1>
KAO -->
<sect1><title>Frame Buffer Video Mode Database</title>
!Idrivers/video/fbdev/core/modedb.c
!Edrivers/video/fbdev/core/modedb.c
</sect1>
<sect1><title>Frame Buffer Macintosh Video Mode Database</title>
!Edrivers/video/fbdev/macmodes.c
</sect1>
<sect1><title>Frame Buffer Fonts</title>
<para>
Refer to the file lib/fonts/fonts.c for more information.
</para>
<!-- FIXME: Removed for now since no structured comments in source
X!Ilib/fonts/fonts.c
-->
</sect1>
</chapter>
<chapter id="input_subsystem">
<title>Input Subsystem</title>
<sect1><title>Input core</title>
!Iinclude/linux/input.h
!Edrivers/input/input.c
!Edrivers/input/ff-core.c
!Edrivers/input/ff-memless.c
</sect1>
<sect1><title>Multitouch Library</title>
!Iinclude/linux/input/mt.h
!Edrivers/input/input-mt.c
</sect1>
<sect1><title>Polled input devices</title>
!Iinclude/linux/input-polldev.h
!Edrivers/input/input-polldev.c
</sect1>
<sect1><title>Matrix keyboars/keypads</title>
!Iinclude/linux/input/matrix_keypad.h
</sect1>
<sect1><title>Sparse keymap support</title>
!Iinclude/linux/input/sparse-keymap.h
!Edrivers/input/sparse-keymap.c
</sect1>
</chapter>
<chapter id="spi">
<title>Serial Peripheral Interface (SPI)</title>
<para>
SPI is the "Serial Peripheral Interface", widely used with
embedded systems because it is a simple and efficient
interface: basically a multiplexed shift register.
Its three signal wires hold a clock (SCK, often in the range
of 1-20 MHz), a "Master Out, Slave In" (MOSI) data line, and
a "Master In, Slave Out" (MISO) data line.
SPI is a full duplex protocol; for each bit shifted out the
MOSI line (one per clock) another is shifted in on the MISO line.
Those bits are assembled into words of various sizes on the
way to and from system memory.
An additional chipselect line is usually active-low (nCS);
four signals are normally used for each peripheral, plus
sometimes an interrupt.
</para>
<para>
The SPI bus facilities listed here provide a generalized
interface to declare SPI busses and devices, manage them
according to the standard Linux driver model, and perform
input/output operations.
At this time, only "master" side interfaces are supported,
where Linux talks to SPI peripherals and does not implement
such a peripheral itself.
(Interfaces to support implementing SPI slaves would
necessarily look different.)
</para>
<para>
The programming interface is structured around two kinds of driver,
and two kinds of device.
A "Controller Driver" abstracts the controller hardware, which may
be as simple as a set of GPIO pins or as complex as a pair of FIFOs
connected to dual DMA engines on the other side of the SPI shift
register (maximizing throughput). Such drivers bridge between
whatever bus they sit on (often the platform bus) and SPI, and
expose the SPI side of their device as a
<structname>struct spi_master</structname>.
SPI devices are children of that master, represented as a
<structname>struct spi_device</structname> and manufactured from
<structname>struct spi_board_info</structname> descriptors which
are usually provided by board-specific initialization code.
A <structname>struct spi_driver</structname> is called a
"Protocol Driver", and is bound to a spi_device using normal
driver model calls.
</para>
<para>
The I/O model is a set of queued messages. Protocol drivers
submit one or more <structname>struct spi_message</structname>
objects, which are processed and completed asynchronously.
(There are synchronous wrappers, however.) Messages are
built from one or more <structname>struct spi_transfer</structname>
objects, each of which wraps a full duplex SPI transfer.
A variety of protocol tweaking options are needed, because
different chips adopt very different policies for how they
use the bits transferred with SPI.
</para>
!Iinclude/linux/spi/spi.h
!Fdrivers/spi/spi.c spi_register_board_info
!Edrivers/spi/spi.c
</chapter>
<chapter id="i2c">
<title>I<superscript>2</superscript>C and SMBus Subsystem</title>
<para>
I<superscript>2</superscript>C (or without fancy typography, "I2C")
is an acronym for the "Inter-IC" bus, a simple bus protocol which is
widely used where low data rate communications suffice.
Since it's also a licensed trademark, some vendors use another
name (such as "Two-Wire Interface", TWI) for the same bus.
I2C only needs two signals (SCL for clock, SDA for data), conserving
board real estate and minimizing signal quality issues.
Most I2C devices use seven bit addresses, and bus speeds of up
to 400 kHz; there's a high speed extension (3.4 MHz) that's not yet
found wide use.
I2C is a multi-master bus; open drain signaling is used to
arbitrate between masters, as well as to handshake and to
synchronize clocks from slower clients.
</para>
<para>
The Linux I2C programming interfaces support only the master
side of bus interactions, not the slave side.
The programming interface is structured around two kinds of driver,
and two kinds of device.
An I2C "Adapter Driver" abstracts the controller hardware; it binds
to a physical device (perhaps a PCI device or platform_device) and
exposes a <structname>struct i2c_adapter</structname> representing
each I2C bus segment it manages.
On each I2C bus segment will be I2C devices represented by a
<structname>struct i2c_client</structname>. Those devices will
be bound to a <structname>struct i2c_driver</structname>,
which should follow the standard Linux driver model.
(At this writing, a legacy model is more widely used.)
There are functions to perform various I2C protocol operations; at
this writing all such functions are usable only from task context.
</para>
<para>
The System Management Bus (SMBus) is a sibling protocol. Most SMBus
systems are also I2C conformant. The electrical constraints are
tighter for SMBus, and it standardizes particular protocol messages
and idioms. Controllers that support I2C can also support most
SMBus operations, but SMBus controllers don't support all the protocol
options that an I2C controller will.
There are functions to perform various SMBus protocol operations,
either using I2C primitives or by issuing SMBus commands to
i2c_adapter devices which don't support those I2C operations.
</para>
!Iinclude/linux/i2c.h
!Fdrivers/i2c/i2c-boardinfo.c i2c_register_board_info
!Edrivers/i2c/i2c-core.c
</chapter>
<chapter id="hsi">
<title>High Speed Synchronous Serial Interface (HSI)</title>
<para>
High Speed Synchronous Serial Interface (HSI) is a
serial interface mainly used for connecting application
engines (APE) with cellular modem engines (CMT) in cellular
handsets.
HSI provides multiplexing for up to 16 logical channels,
low-latency and full duplex communication.
</para>
!Iinclude/linux/hsi/hsi.h
!Edrivers/hsi/hsi.c
</chapter>
</book>