Documentation/firmware_class/README

   1
   2  request_firmware() hotplug interface:
   3  ------------------------------------
   4         Copyright (C) 2003 Manuel Estrada Sainz
   5
   6  Why:
   7  ---
   8
   9  Today, the most extended way to use firmware in the Linux kernel is linking
  10  it statically in a header file. Which has political and technical issues:
  11
  12   1) Some firmware is not legal to redistribute.
  13   2) The firmware occupies memory permanently, even though it often is just
  14      used once.
  15   3) Some people, like the Debian crowd, don't consider some firmware free
  16      enough and remove entire drivers (e.g.: keyspan).
  17
  18  High level behavior (mixed):
  19  ============================
  20
  21  1), kernel(driver):
  22         - calls request_firmware(&fw_entry, $FIRMWARE, device)
  23         - kernel searchs the fimware image with name $FIRMWARE directly
  24         in the below search path of root filesystem:
  25                 User customized search path by module parameter 'path'[1]
  26                 "/lib/firmware/updates/" UTS_RELEASE,
  27                 "/lib/firmware/updates",
  28                 "/lib/firmware/" UTS_RELEASE,
  29                 "/lib/firmware"
  30         - If found, goto 7), else goto 2)
  31
  32         [1], the 'path' is a string parameter which length should be less
  33         than 256, user should pass 'firmware_class.path=$CUSTOMIZED_PATH'
  34         if firmware_class is built in kernel(the general situation)
  35
  36  2), userspace:
  37         - /sys/class/firmware/xxx/{loading,data} appear.
  38         - hotplug gets called with a firmware identifier in $FIRMWARE
  39           and the usual hotplug environment.
  40                 - hotplug: echo 1 > /sys/class/firmware/xxx/loading
  41
  42  3), kernel: Discard any previous partial load.
  43
  44  4), userspace:
  45                 - hotplug: cat appropriate_firmware_image > \
  46                                         /sys/class/firmware/xxx/data
  47
  48  5), kernel: grows a buffer in PAGE_SIZE increments to hold the image as it
  49          comes in.
  50
  51  6), userspace:
  52                 - hotplug: echo 0 > /sys/class/firmware/xxx/loading
  53
  54  7), kernel: request_firmware() returns and the driver has the firmware
  55          image in fw_entry->{data,size}. If something went wrong
  56          request_firmware() returns non-zero and fw_entry is set to
  57          NULL.
  58
  59  8), kernel(driver): Driver code calls release_firmware(fw_entry) releasing
  60                  the firmware image and any related resource.
  61
  62  High level behavior (driver code):
  63  ==================================
  64
  65          if(request_firmware(&fw_entry, $FIRMWARE, device) == 0)
  66                 copy_fw_to_device(fw_entry->data, fw_entry->size);
  67          release(fw_entry);
  68
  69  Sample/simple hotplug script:
  70  ============================
  71
  72         # Both $DEVPATH and $FIRMWARE are already provided in the environment.
  73
  74         HOTPLUG_FW_DIR=/usr/lib/hotplug/firmware/
  75
  76         echo 1 > /sys/$DEVPATH/loading
  77         cat $HOTPLUG_FW_DIR/$FIRMWARE > /sysfs/$DEVPATH/data
  78         echo 0 > /sys/$DEVPATH/loading
  79
  80  Random notes:
  81  ============
  82
  83  - "echo -1 > /sys/class/firmware/xxx/loading" will cancel the load at
  84    once and make request_firmware() return with error.
  85
  86  - firmware_data_read() and firmware_loading_show() are just provided
  87    for testing and completeness, they are not called in normal use.
  88
  89  - There is also /sys/class/firmware/timeout which holds a timeout in
  90    seconds for the whole load operation.
  91
  92  - request_firmware_nowait() is also provided for convenience in
  93    user contexts to request firmware asynchronously, but can't be called
  94    in atomic contexts.
  95
  96
  97  about in-kernel persistence:
  98  ---------------------------
  99  Under some circumstances, as explained below, it would be interesting to keep
 100  firmware images in non-swappable kernel memory or even in the kernel image
 101  (probably within initramfs).
 102
 103  Note that this functionality has not been implemented.
 104
 105  - Why OPTIONAL in-kernel persistence may be a good idea sometimes:
 106
 107         - If the device that needs the firmware is needed to access the
 108           filesystem. When upon some error the device has to be reset and the
 109           firmware reloaded, it won't be possible to get it from userspace.
 110           e.g.:
 111                 - A diskless client with a network card that needs firmware.
 112                 - The filesystem is stored in a disk behind an scsi device
 113                   that needs firmware.
 114         - Replacing buggy DSDT/SSDT ACPI tables on boot.
 115           Note: this would require the persistent objects to be included
 116           within the kernel image, probably within initramfs.
 117
 118    And the same device can be needed to access the filesystem or not depending
 119    on the setup, so I think that the choice on what firmware to make
 120    persistent should be left to userspace.
 121
 122  about firmware cache:
 123  --------------------
 124  After firmware cache mechanism is introduced during system sleep,
 125  request_firmware can be called safely inside device's suspend and
 126  resume callback, and callers need't cache the firmware by
 127  themselves any more for dealing with firmware loss during system
 128  resume.