Re: [rtl] use of RT Fifo leads to IRQ be lost (RTL 2.2)

[Gérard LASSAILLY <g.lassailly@wanadoo.fr>]

Many thanks for your excellent reply.

In my test apps, ISR are not writing in RT FIFO. It's a periodic thread.
I volontary make "big" write in RT FIFO because I think that under RTL2.2
the use of RT FIFO mask all IRQ.
I have look at RT FIFO code from RTL 3.1, it seem that change have been made
to "spin lock" only access to FIFO and not to mask interrupt during all the
time of write (or read).
I can't change now from RTL2.2 to RTL3.1, so I search somebody who know if
what I say is true (RTL2.2 versus RTL3.1).

I know that if I want to transfert lot of data between RT LINUX and LINUX, I
need to use shared memory and not RT FIFO. But, I prefer RT FIFO for small
transfert (cmd, etc ...), ONLY if that don't change IRQ latency in such
proportion I observed with RTL2.2.
I want to be sure that RTL3.1 solve this pb before migrating from RTL2.2 to
RTL3.1.

Gérard


----- Original Message -----
From: "Dresner, Norman A." <norman_a_dresner@md.northgrum.com>
To: <rtl@fsmlabs.com>
Cc: <g.lassailly@wanadoo.fr>
Sent: Tuesday, November 06, 2001 3:54 PM
Subject: RE: [rtl] use of RT Fifo leads to IRQ be lost RTL 2.2


> While you might not want a real-time FIFO to interfere with IRQ latency,
> consider that the RT-kernel has to copy the a large amount of data from
your
> buffer into its buffer in a way that guarantees the atomicity of the
> operation -- most likely , with interrupts disabled.  Similarly the
> operation of reading it out of the FIFO at the user-task's end requires
the
> full overhead of a system call and then the copy of an equal amount of
data
> from the FIFO's internal buffer into the user's buffer, again with some
> guarantee of atomicity.  While there might be a similar number of copy
> operations when your ISR writes it into shared memory and the user reads
it
> out, the time when the interrupts are disabled can be reduced to the time
to
> acquire and release some sort of mutex which controls the simulated FIFO
in
> shared memory; once the pointers are acquired atomically, memory transfer
> can proceed as a normal operation -- and it's possible that the user-task
> doesn't need to copy the data out of shared memory but can use it in
place,
> thus saving even more CPU bandwidth
>
> Norm
>
> > -----Original Message-----
> > From: Gérard LASSAILLY [SMTP:g.lassailly@wanadoo.fr]
> > Sent: Thursday, November 01, 2001 7:49 AM
> > To: rtl@fsmlabs.com
> > Subject: Re: [rtl] use of RT Fifo leads to IRQ be lost RTL 2.2
> >
> > Thanks for your reply.
> >
> > I also think that using shared memory is THE solution to move large
block
> > of
> > data between user space and RT LINUX.
> >
> > But, I think that the use of RT Fifo should not affect IRQ latency on a
> > Real
> > Time OS (a can see on my oscilloscope IRQ latency > 1ms !!! when i'm
using
> > rtl_put() with 1,5 Mbytes under RTL2.2 / kernel 2.2.14 / Pentium 266
MMX).
> >
> > Somebody know if change have been made on RT FIFO code ( RTL 2.2 -> RTL
> > 3.1)
> > to reduce IRQ latency?
> >
> > Gérard
> >
> > ----- Original Message -----
> > From: "Norm Dresner" <ndrez@att.net>
> > To: <rtl@fsmlabs.com>
> > Sent: Wednesday, October 31, 2001 2:51 AM
> > Subject: Re: [rtl] use of RT Fifo leads to IRQ be lost RTL 2.2
> >
> >
> > > This doesn't seem like a good design to me.  Why can't you
> > > write the data to shared memory (very fast) and just send a
> > > pointer through the FIFO telling the user-task where it can
> > > find the data.
> > >
> > >     Norm
> > >
> > > ----- Original Message -----
> > > From: Gérard LASSAILLY
> > > To: rtl@fsmlabs.com
> > > Sent: Tuesday, October 30, 2001 4:47 PM
> > > Subject: [rtl] use of RT Fifo leads to IRQ be lost RTL 2.2
> > >
> > >
> > > I'm using 4 RT Fifo under RT Linux 2.2.
> > > I write blocs of 1.4 Mbytes in  RT FIFO (1.4 Mbytes by
> > > rtf_put() ).
> > > This write in RT Fifo is repeat 25 times by second.
> > >
> > > During this, IRQ are lost or delayed.
> > >
> > > I don't use RT Fifo in ISR but only to communicate between
> > > RT task and user task under LINUX.
> > >
> > > I think that IRQ are masked during rtf_put() during a long
> > > time.
> > >
> > > Some knowledge please ?
> > > Do you know if RT Linux 3.0 or 3.1 workaround this
> > > probleme?
> > >
> > > Thanks for your reply.