From: Eduardo Pereira Habkost <ehabkost@conectiva.com.br>
To: Robert de Bath <robert$@mayday.cix.co.uk>
Cc: Linux-8086 <linux-8086@vger.kernel.org>
Subject: [bcc patch] Fix variable offset macros for #asm (was: Re: ELKS memcpy_fromfs() failing. Wrong variable offsets)
Date: Wed, 2 Jun 2004 13:10:49 -0300 [thread overview]
Message-ID: <20040602161048.GB26330@duckman.distro.conectiva> (raw)
In-Reply-To: <609c173af82cc283@mayday.cix.co.uk>
[-- Attachment #1: Type: text/plain, Size: 4342 bytes --]
On Wed, Jun 02, 2004 at 09:47:25AM +0100, Robert de Bath wrote:
<snip>
>
> You have highlighted a bug though, the definition some of the 'set'
> variables is actually wrong, they should have memcpy_fromfs.off added
> when it might be non-zero.
I am sending a fix to this bug. I hope that I didn't anything wrong,
as I am not familiar with bcc code.
I guess that creating an outoffset(variable) function that uses
<function>.off when needed would be better, but that was just a quick fix.
>
> BTW: You should really be using those variables to access C arguments
> and locals, the '_memcpy_fromfs.ds' is for stack pointer relative the
> '.memcpy_fromfs.ds' is for 'bp' relative.
As we have a fix for bcc >= 0.16.8, I will send a fix for ELKS code,
for using .<function>.<variable>, now. :)
>
> I can fix the bug in one of two ways; either actually add the
> 'memcpy_fromfs.off' variable to the 'set' variables or simply
> have the compiler assume that any '#asm' will use si and di.
> (The asm("...") function will not however.)
>
> Actually I think assuming that #asm uses si/di is best because as this
> example has shown it is quite likely to be true, plus, it's the failsafe
> option.
I prefer the code as it is now: if some #asm will use some register,
then it should save it. Or maybe a better one (IMO): having a way to
tell the compiler which registers will be used.
And even if assuming that all #asm will use SI and DI, it is a little
safer fixing the offsets, too. Just in case someone make more changes
to the code, try to optmize #asm, and break things because there are
implict things (like: dumplocs() only work because when using #asm,
<function>.off is zero) they didn't see.
Which fix do you plan to include on the next releases of bcc?
(BTW, I really missed a BCC_VERSION macro or equivalent. It would be
very useful in cases like this one)
--
Eduardo
--- dev86-0.16.15/bcc/codefrag.c 2002-08-03 13:38:27.000000000 -0300
+++ dev86-current/bcc/codefrag.c 2004-06-02 12:38:28.233060704 -0300
@@ -105,7 +105,6 @@
# define outlswitch() (outload(), outstr(ireg0str), outncregname(DREG))
# define outnc1() outnstr(",*1")
# define outsbc() outop3str("sbb\t")
-# define outset() outstr ("\tset\t")
# define outsl() outop2str("shl\t")
# define outsr() outop2str("sar\t")
# define outtransfer() outload()
@@ -134,6 +133,10 @@
outstr(acclostr);
outncregname(BREG);
}
+PUBLIC void outset()
+{
+ outstr ("\tset\t");
+}
PUBLIC void comment()
{
outstr("! ");
@@ -417,7 +420,6 @@
# define outpil2switch() outnop2str("LDD\tD,X")
# define outrolhi() outnop1str("ROLA");
# define outsbc() outop2str("SBC")
-# define outset() outstr ("\tSET\t")
# define outsl() outop1str("LSL")
# define outtransfer() outop2str("TFR\t")
# define reclaimfactor() outnstr ("++") /* discard factor from stack */
@@ -434,6 +436,10 @@
# define tfrhilo() outnop2str("TFR\tA,B")
# define tfrlohi() outnop2str("TFR\tB,A")
# define usr1() (outnop1str("LSRA"), outnop1str("RORB"))
+PUBLIC void outset()
+{
+ outstr ("\tSET\t");
+}
PUBLIC void clrBreg()
{
outnop1str("CLRB");
--- dev86-0.16.15/bcc/table.c 2002-07-28 04:43:13.000000000 -0300
+++ dev86-current/bcc/table.c 2004-06-02 12:48:12.176287840 -0300
@@ -402,8 +402,38 @@
for (i = 0; i < HASHTABSIZE; ++i)
for (symptr = hashtab[i]; symptr != NULL; symptr = symptr->next)
- if (symptr->storage == LOCAL)
- set(symptr->name.namea, symptr->offset.offi - sp);
+ if (symptr->storage == LOCAL) {
+ /* Variable offset relative to sp */
+ outccname(funcname);
+ outbyte(LOCALSTARTCHAR);
+ outstr(symptr->name.namea);
+ outset();
+ outshex(symptr->offset.offi - sp);
+ outnl();
+
+#ifdef FRAMEPOINTER
+ /* Variable offset relative to bp */
+ if (framep)
+ {
+ outbyte(LOCALSTARTCHAR);
+ outstr(funcname);
+ outbyte(LOCALSTARTCHAR);
+ outstr(symptr->name.namea);
+ outset();
+ outshex(symptr->offset.offi - framep);
+#ifdef I8088
+#ifndef NO_DEL_PUSH
+ if (optimise && !callersaves && symptr->offset.offi < framep) {
+ outbyte('+');
+ outstr(funcname);
+ outstr(".off");
+ }
+#endif
+#endif
+ outnl();
+ }
+#endif
+ }
}
#ifdef HOLDSTRINGS
[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]
next prev parent reply other threads:[~2004-06-02 16:10 UTC|newest]
Thread overview: 5+ messages / expand[flat|nested] mbox.gz Atom feed top
[not found] <20040601222140.GV21172@duckman.distro.conectiva>
2004-06-02 8:47 ` ELKS memcpy_fromfs() failing. Wrong variable offsets Robert de Bath
2004-06-02 12:52 ` Eduardo Pereira Habkost
2004-06-02 16:10 ` Eduardo Pereira Habkost [this message]
2004-06-03 14:20 ` [bcc patch] Fix variable offset macros for #asm (was: Re: ELKS memcpy_fromfs() failing. Wrong variable offsets) Miguel Bolanos
2004-06-03 14:19 ` ELKS memcpy_fromfs() failing. Wrong variable offsets Miguel Bolanos
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20040602161048.GB26330@duckman.distro.conectiva \
--to=ehabkost@conectiva.com.br \
--cc=linux-8086@vger.kernel.org \
--cc=robert$@mayday.cix.co.uk \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox