(I had posted this message earlier when probably everyone was accidentally
removed from the list. As this list is active now I post this again)
Hi,
This is response to the feedback to a mail I had posted some time back. I
would like to thank you all for the responses.
I had some questions regarding the feedback; I am building an
implementation of a wireless protocol stack for which I need realtime
guarantees at the physical layer. I am using RTL version 1.1 with Linux
2.0.36.
>Subject: Re: [Re: [rtl] Communicating with the linux kernel]
>From: [EMAIL PROTECTED]
>On Mon, May 24, 1999 at 10:27:16AM +0000, Paolo Mantegazza wrote:
>> Tomasz Motylewski wrote:
>> >
>> > On Thu, 20 May 1999, Sugat Jain wrote:
>> >
>> > > My problem right now is an implementation one: When a received packet
is to
>> > > be handed over to the linux part for processing by higher layers, when
a
>> >
>> > Yes, this is the problem to which I knew no satisfactory answer. What is
>> > needed is a "bottom half" for RT interrupt. So, when your RT-ISR sees a
>> > normal packet to be passed through netif_rx it can set some flag and
exit.
>> >
>
> The problem of bottom hakves can be solved if RTL implemented a feature,
> available in RTAI, i.e system requests (srq). This a a mechanism which
>The problem of bottom halves was solved in RTL with the mechanism used by
>fifo handlers -- easily adaptable to other operations. Since V1 we have
>also had the ability for rt irq handlers to pend interrupts for Linux
>handlers. In V2 we will have special "soft interrupts" for RTL to send
>to linux.
I was going through the various methods for having the ISR in RT space
schedule a bottom half to run in Linux space.
In an earlier discussion (in the archives) on Bottom Halves it was
mentioned that mark_ bh would cause race conditions if called from RT space
and so queue_task_rt() could be used. In my implementation I need to receive a
packet, process it and send out a packet appropriately within 10ms. As the
system timer runs at 100hz I will not be able to use the scheduler queue, rt
queue or the timer queue. I could use the Immediate bh queue but I will have
to call mark_bh with it. (I could set HZ to 500 or 1000 & have 2 or 1ms timer
ticks instead of 10ms - but I was wondering if there could be a better
solution than this)
Another solution mentioned was that since V1 it is possible for rt irq
handlers to pend interrupts for Linux Handlers. I would be grateful if someone
could explain me the necessary steps to do this.
Tomasz Motylewski had suggested trying :
SFREQ |= MASK(WAVELAN_IRQ); /* analog of "mark_bh" */
This didn't seem to work though.
I do not fully understand the intricacies of interrupts and related things
in Linux. I would appreciate it if someone could explain me the problem in
using mark_bh.
Thanks in advance,
Sugat Jain
____________________________________________________________________
Get your own FREE, personal Netscape WebMail account today at
http://webmail.netscape.com.
--- [rtl] ---
To unsubscribe:
echo "unsubscribe rtl" | mail [EMAIL PROTECTED] OR
echo "unsubscribe rtl <Your_email>" | mail [EMAIL PROTECTED]
----
For more information on Real-Time Linux see:
http://www.rtlinux.org/~rtlinux/
