This is the mail archive of the newlib@sourceware.org mailing list for the newlib project.

Index Nav:	[Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav:	[Date Prev] [Date Next]	[Thread Prev] [Thread Next]
Other format:	[Raw text]

Re: Bug in m68k memcpy/memcmp

From: Jeff Johnston <jjohnstn at redhat dot com>
To: Eric Norum <norume at aps dot anl dot gov>
Cc: kazu at codesourcery dot com, newlib at sources dot redhat dot com
Date: Fri, 27 Apr 2007 18:12:55 -0400
Subject: Re: Bug in m68k memcpy/memcmp
References: <48CED6E1-1E52-4EB4-911D-B4E395A4ED76@aps.anl.gov>

Being that it is Friday, I have made a patch to temporarily remove memcpy.S and memset.S from being built until this can be resolved.

-- Jeff J.

Eric Norum wrote:

The proposed patch to the M68k versions of memcpy/memset has a nasty bug:
+.Lcopy:
+#if !defined (__mcoldfire__)
+    dbra    %d0,1b
+#else
+    subq.l    #1,%d0
+    bpl    1b
+#endif
The non-ColdFire version uses dbra -- which uses only the lower 16-bits of the counter register. The above code will fail for transfers greater than 16*65535 bytes.

There needs to be another decrement of the most-significant 16-bits of the count register and another conditional branch and conditional branch added to all the 'dbra' loop tests.
Off the top of my head, I think that there needs to be something like:
+.Lcopy:
+#if !defined (__mcoldfire__)
+    dbra    %d0,1b
+      subl         #0x10000,d0
+     bpl           1b
+#else
+    subq.l    #1,%d0
+    bpl    1b
+#endif
--Eric Norum <norume@aps.anl.gov>
Advanced Photon Source
Argonne National Laboratory
(630) 252-4793

References:
- Bug in m68k memcpy/memcmp
  - From: Eric Norum

Index Nav:	[Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav:	[Date Prev] [Date Next]	[Thread Prev] [Thread Next]