A new device driver for a video frame grabber

1
A new device driver for a video frame grabber
Term Project Kernel Device Summer 2006

• Binh Tran Lucie Ngnepieba

2
Outline

• Overview
• Our goals
• Our achievement
• 4. Difficulties and our solutions
• 5. Discovered facts and our learns
• 6. Future works
• 7. Conclusion

3
Overview

ISA greyscale card
PCI color card
4
Goals

• To compile correctly on the Linux distribution
  2.6.16
• To improve the previous Video frame grabber
  device driver to support Video4Linux.
• To integrate the previous driver and make it
  work on both grayscale and color support.
• To work on the low-level operation of the frame
  grabber to support the new interrupt-driven I/O
  feature (if time permits).

5
Previous students works

• Taken from the previous students journal (Atulya
  Mahajan, Huan Keat ("Sean") Toh)
• The device driver ran on Linux distribution
  2.6.10 and 2.6.11
• The greyscale part was running good, but the
  color part was not working correctly
• The cleanup function was inconsistent
• Kernel panic when unload the module

6
Our achievements

• Our HRT Frame grabber driver works well and
  stable with the 2.6.16.14 kernel (goal 1)
• Our driver supports Video4linux (goal 2)
• Our driver had improved the previous driver to
  work with both greyscale and color cards (goal
  3)
• We made our driver work properly, consistently,
  and non-errors for actions load module, unload
  module , auto-detect different devices, no kernel
  panic.

7
Difficulties encountered and our solution

• 1. The HRT module is loaded, but no devices are
  registered, no major number shows up
• Sol we generated a series of debugging messages
  to test, and we did fix the hrt_init_module,
  hrt_cleanup_module, hrt_load, hrt_unload.
• 2. The two cards are detected as grayscale
  cards, and give us a warning message about sysfs
  needed.
• Sol Switched the two cards, reboot, and load
  module again. It works. We diagnose the cards
  slots are sensitive. Moreover, sysfs is not a
  main issue.

8
Difficulties encountered and our solution

• The pci color card was not unregistered when we
  unload the module
• Sol we fixed the hrt_cleanup and
  hrt_pci_cleanup functions
• 4. The color doesnt display well, still get
  fuzzy
• Sol pixel mapping between kernel buf and user
  buf is not correct. We did fix the grab_field
  function

9
Difficulties encountered and our solution (cont)

• 5. Pixels mapping among user space, kernel
  buffer, and I/O devices do not work correctly,
  the color frames are still distorted
• Sol We modified the grab_field functions,
  and changed the resolution in the test files as
  well
• Missing libraries needed for Video4linux
• Sol installed SDL libraries.
• Issues of compiling the hrttest.c that uses the
  Video4linux
• Sol modified the Makefile
• Understand the otherss source code (gt65 pages
  with more than 3700 lines of codes)
• Sol we tried to do mapping, analysis
  statically, and draw links among functions

10
Discovered facts and our learns

• Learn how to read the long long source code
• Learn how to debug subtle errors
• Improve how to use the following techniques
• Char Device driver
• Memory mapped I/O
• Internal mutual exclusion (spinlocks and
  semaphores)
• Interrupt Timer management (tasklets and kernel
  timer API) Learn more about Linux like making a
  make file, etc
• Improve knowledge about Linux

11
Works left

• To work on the low-level operation of the frame
  grabber to support the new interrupt-driven I/O
  feature (if time permits)

12
Conclusion

• We have achieved most of our main goals