contrib/FAQ/programs/fp-emu/fpemul/bde_trapinfo.mail

From bde@kralizec.zeta.org.au  Sun Jun 27 01:18:32 1993
Received: from ultima.socs.uts.EDU.AU by bsd.coe.montana.edu (5.67/KAOS-1)
        id AA11952; Sun, 27 Jun 93 01:18:32 -0600
Received: by ultima.socs.uts.EDU.AU (5.65+/SMI-3.3)
        id AA03033; Sun, 27 Jun 93 17:10:22 +1000
Received: by kralizec.zeta.org.au (4.0/SMI-4.0)
        id AA15074; Sat, 26 Jun 93 02:32:58 EST
Date: Sat, 26 Jun 93 02:32:58 EST
From: bde@kralizec.zeta.org.au (Bruce Evans)
Message-Id: <9306251632.AA15074@kralizec.zeta.org.au>
To: nate@bsd.coe.montana.edu
Subject: Re:  Trapframe information
Status: OR

tf_isp          original esp (probably spare - popal ignores it)
tf_trapno       s/w trap no (may be spare - trap.c has already looked at it)
tf_err          h/w error code (probably spare - gets discarded before iret)

___fs           not stored in 386BSD pcb.  Constant anyway unless user has
                screwed with it (?).
___gs           ditto
___orig_eip     in linux, this is on the stack just before the call to the
                emulator.  The reason that it's not a local variable is to
                avoid passing around pointers to it - current->frame (or
                whatever) points to everything in the stack frame.  The
                macros hide a lot of slow memory references
                current->frame->var.

>(And I need to see if I can map orig_eip to one of the three that I'm unsure of
>in the BSD sources)

tf_isp is the least evil.

Bruce