Make Problem with FS flash

Joel Sherrill joel.sherrill at OARcorp.com
Mon Dec 11 16:25:25 UTC 2000


Lots of questions .. :)

"Fernando RUIZ CASAS (E-mail)" wrote:
> 
> Hi rtems members,
> 
>   I try build a FS flash file system.
>   The flash driver works ok with a sector erase cache memory.
>   It is not a harddisk but it works very well.

Be warned .. the IMFS is not as block oriented as it needs to
be to store all data in Flash.  I posted on this a while back.
See

http://www.oarcorp.com/rtems/maillistArchives/rtems-users/2000/november/msg00034.html


>   To build a new fs I think that the rule is reply imfs directory.
>   Rename all the files and the routines and
>   build a new libfs.a adding the new fs created.
> 
>   Well,
>   after reply the 'imfs' directory into 'fmfs' (flash memory file system) in
> libfs repository
>   and change the Makefile.am in all the directories and run bootstrap in
> every directory in the libfs tree
> 
>   my problem is the wrapup 'Makefile.am'
> 
>   after rebuild rtems with ./bit_rtems the make says that it can't find
> rules for the fmfs
> 
>   Please, is it possible a bref example of Makefile.am for the wrapup
> directory?

The things involved in adding a directory should be:

  + Add that directory to the list of Makefiles in the 
    appropriate configure.in (c/src/libfs/configure.in)
  + Make sure it is descended into by the parent directory
    (c/src/libfs/src/Makefile.am)
  + Added it to any appropriate wrapup/Makefile.am's 
    following the style in that file.
  + Run bootstrap in the directory with the configure.in

>   Which is the recipe to add a new filesystem in the libfs repository?

Other than the above guidelines on Makefiles and configure.in, that
is about it.

>   Thanks in advance.
> 
> P.S.:
>   Is Tanembaum the father of theory for the imfs file system or a newest
> theory is used?

The only (implicit) reference is the original UNIX article in the
Bell Labs Tech Journal which describes the direct, indirect, and
triply indirect block pointers.  I designed a simpler alternative
that achieved the same end result for the IMFS.  
 
>   Do I need read a new theory for a unix filesystem?

No.  It is pretty straightforward is you have read articles
on the design of classic unix filesystems.

NOTE: The parsing of filenames/paths is VERY unique to the 
RTEMS filesystem infrastructure.

>   Linux uses a new ext2 filesystem. Not a Minix filesystem.

The IMFS is its own beast.

-- 
Joel Sherrill, Ph.D.             Director of Research & Development
joel at OARcorp.com                 On-Line Applications Research
Ask me about RTEMS: a free RTOS  Huntsville AL 35805
   Support Available             (256) 722-9985



More information about the users mailing list