Skip to content
This repository


Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP
100644 73 lines (58 sloc) 3.091 kb
6fdf2496 »
2010-05-12 initial import from old git tree
1 Here's a summary of the current state of the Linux Driver core and sysfs
2 subsystems.
4 If the information in here is incorrect, or anyone knows of any
5 outstanding issues not listed here, please let me know.
7 List of outstanding regressions from 2.6.15:
8 - none known.
10 List of outstanding regressions from older kernel versions:
11 - none known.
13 There are no outstanding bugs for these subsystems in the
14 system that I know of. If this is not true, please
15 point me at the bugs you feel I should know about.
17 Future stuff:
18 All of the patches that will be sent in for 2.6.17 can be found in my
19 quilt tree at
20 and
21 are included automagically in the -mm releases.
23 Here's a summary of the changes that can be found there:
24 - There's a big pending AOE driver update that missed the 2.6.16
25 cut-off timeframe. If you use this driver, I would recommend
26 getting these patches.
27 - pollable sysfs attributes will be coming soon. I think Neil
28 is going to rework the existing patches, due to some issues
29 found in -mm reviews, but the functionality should remain the
30 same (adds a new sysfs function you can call to let userspace
31 know that a sysfs file has changed contents. Use this instead
32 of the old uevent interface to notify userspace in an easier
33 manner.)
34 - change semaphores to mutexes.
35 - module sysfs file refcount fixes
37 - relayfs as a stand-alone filesystem is gone, to be replaced
38 with the ability to use relayfs files from within sysfs.
39 - other minor bugfixes.
41 I am currently working on allowing struct device to be bound to a struct
42 class as part of the migration away from struct class_device and toward
43 allowing everything to be a struct device in the sysfs device tree.
44 That work is being tested out on the "usbfs2" work which adds a dev file
45 to every USB endpoint in the system, allowing async-io to be used from
46 userspace instead of the ioctl mess we have in usbfs today. If anyone
47 is interested in this work, look in my patch queue for the initial cut
48 of it. It will be rapidly changing this week as I get the time to work
49 more on it.
51 The above mentioned changes, do not change any existing sysfs or driver
52 core apis, but only add a few more functions. So there are no API
53 changes coming in the near future that I am aware of.
56 It is also noted that a lot of people still don't know how to use a
57 kobject properly, and that more sysfs files are containing more than one
58 value (hey, look a table of numbers in the cpufreq file...) Please be
59 aware that people will be walking around with big sticks and whacking
60 the fingers of any programmers who do this. When in doubt, _please_ ask
61 how to do this kind of low-level programming as it is very easy to get
62 wrong...
64 Also the kernel/userspace API documentation patch is pending me getting
65 the chance to work on it today or tomorrow, don't worry, I haven't
66 forgotten about it :)
69 Was this summary useful for people? Anything that I should add to it?
71 thanks,
73 greg k-h
Something went wrong with that request. Please try again.