Re: User-space to kernel communication
Re: User-space to kernel communication
- Subject: Re: User-space to kernel communication
- From: Dan Markarian <email@hidden>
- Date: Tue, 9 Jan 2007 11:48:48 -0500
Hey Greg,
IOUserClient does support variable length data.
http://lists.apple.com/archives/darwin-development/2001/May/
msg00760.html
You can find documentation on IOUserClient at:
http://developer.apple.com/documentation/DeviceDrivers/Conceptual/
WritingDeviceDriver/MakingHWAccessible/chapter_5_section_2.html
http://developer.apple.com/documentation/DeviceDrivers/Conceptual/
WritingDeviceDriver/MakingHWAccessible/chapter_5_section_3.html
http://developer.apple.com/documentation/DeviceDrivers/Conceptual/
WritingDeviceDriver/MakingHWAccessible/chapter_5_section_4.html
It seems that the easiest means to communicate to your I/O Kit
driver, for your purposes, is to send a CFData property to it with
IORegistryEntrySetCFProperties() in user space, which is
automatically routed to your IOService::setProperties() method in
your driver in the kernel. See the "Accessing Device Properties"
section for further documentation.
Dan
On 8 Jan 2007, at 9:19 PM, Greg wrote:
Thanks for pointing this out! I've looked into this some more and
found documentation on kernel control sockets here:
http://developer.apple.com/documentation/Darwin/Conceptual/
NKEConceptual/index.html?http://developer.apple.com/documentation/
Darwin/Conceptual/NKEConceptual/control/chapter_3_section_2.html
This looks very interesting and so far it seems to be exactly what
I want as it uses a familiar API (bsd sockets) and allows for
bidirectional communication. With this available I don't see why
anyone would want to use sysctl or the IOUserClient stuff... are
there any disadvantages associated with using kernel control
sockets over the aforementioned sysctl/IOConnectMethod stuff?
- Greg
On Jan 8, 2007, at 8:04 PM, Brian Bergstrand wrote:
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
On Jan 8, 2007, at 6:18 PM, Greg wrote:
From what I've read about IOUserClient and its
IOConnectMethod*'s, it does not seem like they support passing
variable length data. Perhaps I'm missing something, could you
point me to what you're referring to or perhaps a code example?
No, you'd have to combine memory mapping with an IOUserKitClient
to get true variable length support (by passing a ptr in a struct
arg to your method). However, you say you only need around 50
bytes, so just have a static buffer in your struct large enough to
fix the max amount of data:
typedef struct {
unsigned char buffer[256];
unsigned char bufferSize;
} MyIOUserClientArg;
Then just cast 'buffer' as needed.
As for an example, that's about all I can give you for
IOUserClient, but I do have a public project that uses a kernel
control socket to send variable sized data if you are interested
in seeing that:
http://prdownloads.sourceforge.net/peerguardian/
PeerGuardianSrc_1.4.2.zip?download
The interesting bits as far as kernel<->user communication are in:
pplib/ppmsg.h
pplib/pplib.c
kern/PeerProtector.c
kern/pp_session.c
kern/pplog.c
- Greg
On Jan 8, 2007, at 7:07 PM, Brian Bergstrand wrote:
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
There is no "correct" way. There are ways that make more "sense"
given your current setup, but that's about it. If you want bi-
directional communication and you are already an IOKit kext, the
easiest method will be an IOUserClient nub (as Josh already
stated). You could also easily use a kernel control socket or
memory mapping (from personal experience, the former is a tad
easier than the latter because of certain kernel restrictions on
mapping memory from userland into the kernel).
Also from personal experience, sysctls and ioctls do not make
for easy bi-directional communication channels.
On Jan 8, 2007, at 5:41 PM, Greg wrote:
Sorry for the followup, but perhaps somebody could just advice
me on a specific method.
That document does provide details for the sysctl API but skims
over memory-mapping and a few others.
Apple provides an IODataQueue example but that is only for
kernel -> user-client communication, I'm not exactly sure how I
would reverse that direction.
However, it seems that sysctl is a viable option for my
purposes. It does seem to be able to handle variable length
data of moderate size, I guess I'll just use that. The reason
for my confusion is simply the shear number of options and the
rather sparse documentation on them; I'm just trying to find
the "correct" way of doing this...
- Greg
On Jan 8, 2007, at 6:17 PM, Greg Miller wrote:
Take a look at the document at
http://developer.apple.com/documentation/Darwin/Conceptual/
KernelProgramming/boundaries/chapter_14_section_1.html#//
apple_ref/doc/uid/TP30000905-CH217-DontLinkElementID_9
G~
On 1/8/07, Greg <email@hidden> wrote:
Like I said, I'm trying to send the kernel variable length
data of
around 50 bytes. I am using the IOKit but I would also like
to know
how to do this without it. Is there any documentation on
this? I
would really prefer to avoid blindly going through header files.
- Greg
On Jan 8, 2007, at 3:15 PM, Josh Graessley wrote:
>
> This depends on what you are trying to communicate with in the
> kernel. If you are already using IOKit, I would imagine an
> IOUserClient would be the way to go. If you're writing a
non-IOKit
> kext, you may want to look in to a kernel control socket (sys/
> kern_control.h).
>
> -josh
>
> On Jan 7, 2007, at 6:40 PM, Greg wrote:
>
>> Hi, what is the best way(s) to send variable length data
(around
>> 50 bytes or so) to a kernel extension from a user-land
>> application? Are there different ways to do this if the
kext uses
>> the IOUserClient schema to talk to user-land apps? If you
could
>> also point me to documentation or a tutorial on how to do
this I
>> would be much obliged!
>>
>> Thanks,
>>
Brian Bergstrand
<http://www.bergstrand.org/brian/> PGP Key ID: 0xB6C7B6A2
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (Darwin)
iD8DBQFFouoOedHYW7bHtqIRAmtZAJ4hLO5Ip3DX8dAVZHxlyd5chFcxmQCfQX7p
g3XhL63bYRraNiZPGHU4csw=
=Z7ob
-----END PGP SIGNATURE-----
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Darwin-kernel mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
40apple.com
This email sent to email@hidden
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Darwin-kernel mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden