Re: A question of writing file in rtlinux, thanks

  Such are our discussion about kernel side writing
file, I repost it to mailing list. Hope everybody can
profit from it, :)  Any suggestion else will be
welcome too.

  Thanks to Alejandro Lucero. I've got what u mean.
That is acquiring and writing asynchronously. 

 --- [email protected] 的正文：
> On Sunday 23 January 2005 16:10, you wrote:
> > First thanks very very much for your kind help~ :)
> 
> You're welcome.
> 
> >  --- [email protected] 的正文：
> >
> > > On Saturday 22 January 2005 17:32, WuXin wrote:
> > > > Dear Everybody:
> > > >         Nice to meet u in the web!
> > > >         Now I'm using rtlinux/free v3.1 and I
> want
> > >
> > > to
> > >
> > > > write some data to a file once per 5 ms. I'm
> > >
> > > little
> > >
> > > > confused whether I can only use rtlinux's
> > >
> > > fifo/mbuff
> > >
> > > > or  I can use some normal kernel function such
> as
> > > > "filp_open" or kernel side system call such as
> > > > "open/write"?
> > >
> > > You can use these functions but at your own
> risk,
> > > since you enter in the Linux
> > > kernel behaviour "no real time" world. You can
> use
> > > two threads, one acquiring
> > > the data and writing it to a share buffer, and
> other
> > > writting data from the
> > > shared buffer to disk using the linux kernel
> > > functions. The last should be
> > > less priority. If the system load (linux) is not
> > > very high, this approach
> > > will work, but you must decide the size of the
> > > shared buffer to avoid losing
> > > data.
> >
> > What do u mean here? The thread acquiring data is
> > RTLinux thread, but what's the other thread? In my
> > opinion, it's also RTLinux thread. Then we'll meet
> the
> > same question, how can use normal linux kernel
> > functions in RTLinux thread?
> 
> Both are rtlinux threads. The division is due to
> when linux kernel functions 
> are used the delay can be so high that your 5ms
> requirements could be 
> overpassed. With this approach you makes sure one
> thread will always be ready 
> to acquire data. The other can be blocked by Linux
> kernel, but this is not a 
> problem because:
> 
> 1) you use a shared buffer, so writer (in the
> buffer) thread can still works
> 2) the reader (from the buffer, but writer to disk)
> thread can make writes 
> calls of large sizes.
> 
> In this way, meanwhile one thread writes B bytes to
> the shared buffer 200 
> times by second, the other can read from the buffer
> B*10 bytes in each loop, 
> then saving 10 write system calls.
> 
> About how you can use this Linux kernel functions,
> you must do some tricky 
> things, since you are using system calls from kernel
> inside. This is not 
> hard, and you can learn it fast. I worked with some
> similar in a old project, 
> and I learned from khttp, the http server which runs
> inside of the kernel. 
> You can see that in older linux kernel sources .
> 
> >
> > Btw: My method now is
> > RTLinux thread acquires data and writes them into
> a
> > size-precounted mbuff. When the acquiring job is
> over,
> > i execute a user side program, read data from
> mbuff
> > and write to file with user-side functions such as
> > "fopen" "fprintf".
> >
> > It works, but not direct and not perfect i think.
> :)
> 
> It depends of how many bytes you want to acquire. If
> you have enough memory 
> this is a god approach and simple.
> 
> > > By other hand, there are two things you can do:
> > >
> > > 	- First, Hofrat did what you want to do and he
> can
> > > give some detailed
> > > explanations. You can see the ArvidStaub paper
> in
> > > the last Real Time Linux
> > > Workshop too.
> >
> > Plz forgive my basic of How can I contact Mr.
> Hofrat?
> 
> Hofrat is the rtlinux mailing list maintainer, so I
> guess he will answer you 
> soon.
> 
> > Arvid Staub's "RAT -- Real-Time Audio Tools" is
> > downloading and also your paper "IDE Driver and
> RTLFS
> > file system for RTLinux ".
> >
> > Keep learning from you! :)
> > Really help me a lot, deeply moved.
> 
> I hope you can go forward. If you decide to use my
> code, please inform me 
> about your problems/goals.
> 
> > Another question: I'm the first time using mailing
> > list, so should I still post this reply to "rtl
> list"
> > rather than to you only?
> 
> well, I've seen that I answered you directly. I did
> wrongly. It's better to 
> use the mailing list. Others can profit from it.
> 
> > > 	- I worked in using IDE disks from RT tasks. In
> > > this way you avoid the Linux
> > > impact (at least it is minimized). You can work
> > > writting directly into the
> > > disk, or using the rtlfs file system (designed
> to
> > > benefit real time tasks).
> > > The drawback with this approach is you need a
> > > specific disk partition to work
> > > with. See www.lutherfs.com
> > >
> > > >          Very appreciate for your reply.
> Thanks a
> > >
> > > lot.
> > >
> > > >                                       Yours
> > > >                                       Ardon
> >
> >
> 
> _______________________________________________
> > > > Rtl mailing list
> > > > [email protected]
> > > > http://www2.fsmlabs.com/mailman/listinfo/rtl
> > >
> > > --
> > > Alejandro Lucero
> > > OS3, OS Serveis i Solucions
> > > www.os3sl.com
> > > Ingeniería Informática
> > > +34 665687168
> > > Av.Benjamin Franklin
> > > CEEI. Parque Tecnológico de Paterna
> > > Valencia(Spain)
> >
> >
> 
> 

_________________________________________________________
Do You Yahoo!?
150万曲MP3疯狂搜，带您闯入音乐殿堂
http://music.yisou.com/
美女明星应有尽有，搜遍美图、艳图和酷图
http://image.yisou.com
1G就是1000兆，雅虎电邮自助扩容！
http://cn.rd.yahoo.com/mail_cn/tag/1g/*http://cn.mail.yahoo.com/event/mail_1g/
_______________________________________________
Rtl mailing list
[email protected]
http://www2.fsmlabs.com/mailman/listinfo/rtl