home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.os.linux
- Path: sparky!uunet!zaphod.mps.ohio-state.edu!news.acns.nwu.edu!nucsrl!hpa
- From: hpa@eecs.nwu.edu (H. Peter Anvin N9ITP)
- Subject: Dumb question: new executable format?
- Message-ID: <1993Jan12.065656.15070@eecs.nwu.edu>
- Reply-To: hpa@nwu.edu (H. Peter Anvin)
- Organization: You must be kidding!
- Date: Tue, 12 Jan 1993 06:56:56 GMT
- Lines: 32
-
- Since the new extfs now is in alpha, and will (sometime) support block
- sizes other than 1K, and such a change is eventually slated for the
- Linux kernel as far as I have understood, I would like to ask the
- question: doesn't that cause a problem with the current Minix-style
- executable format?
-
- With a 1K block size, everything is dandy. But with a 4K block size,
- which has been discussed (or really anything > 1K)
- the 1K header drives the pages in the executable out of alignment:
-
- 1K blocks | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | A | B | ...
- 4K blocks | 0 | 1 | 2 | ...
- Exec file |Hdr| Page 0 | Page 1 | Page 2 | ...
-
- Either we could adopt BSD-style executables with a 4K header and lots
- of wasted space, or we could do as someone suggested on comp.os.linux
- (sorry can't remember your name, but for the record it was not my
- idea) and put the header in the first page of the text segment,
- supposedly including a header version number and a pointer (or even
- better, a jump instruction) to the beginning of the actual code.
-
- This would clearly require modifications of both the kernel and
- linkers, possibly even compilers. Maybe Linus and HJ could tell how
- much of a hassle it would be to implement such a change.
-
- /hpa
-
- --
- INTERNET: hpa@nwu.edu TALK: hpa@delta.eecs.nwu.edu
- BITNET: HPA@NUACC IBMNET: 16331@IBMX400
- HAM RADIO: N9ITP NeXTMAIL: hpa@lenny.acns.nwu.edu
- while ( 1 ) ; cp /dev/zero /dev/null & end
-