lkml.org 
[lkml]   [2002]   [Jan]   [22]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
SubjectRe: missing memset in divas and eicon in 2.2.20
Date
"A month of sundays ago Keith Owens wrote:"
> On Tue, 22 Jan 2002 10:19:15 +0100 (MET),
> Armin Schindler <mac@melware.de> wrote:
> >Did you use plain 2.2.20 ?
> >I cannot reproduce this problem here, can you please send me your
> >kernel config.
> >
> >On Tue, 22 Jan 2002, Peter T. Breuer wrote:
> >> betty:/usr/local/src/linux-2.2.20% sudo depmod -ae -F System.map 2.2.20-SMP
> >> depmod: *** Unresolved symbols in
> >> /lib/modules/2.2.20-SMP/misc/divas.o depmod: memset
> >> depmod: *** Unresolved symbols in
> >> /lib/modules/2.2.20-SMP/misc/eicon.o depmod: memset
>
> This can be a gcc problem. Some versions of gcc generate internal

Possibly. As I recall it self-selected gcc 2.7.2 to do the compilation
on my system, although egcs 2.91 is the installed compiler of choice.
I didn't pay too much attention as I was only compiling 2.2.20 in order
to win the arms race against stephen tweedies ext3 patch for the 2.2
series. So ... uh, ok, let's seee...


/usr/bin/gcc272 -D__KERNEL__ -I/usr/local/src/linux-2.2.20/include
-Wall -Wstrict-prototypes -O2 -fomit-frame-pointer -D__SMP__ -pipe
-fno-strength-reduce -m486 -malign-loops=2 -malign-jumps=2
-malign-functions=2 -DCPU=686 -DUTS_MACHINE='"i386"' -c -o
init/version.o init/version.c

Yep. Wonder how that got there. This is a debian system. Oh, it's the
scripts/kwhich stuff. Clever lad.

CC =$(shell if [ -n "$(CROSS_COMPILE)" ]; then echo $(CROSS_COMPILE)gcc; else \
$(CONFIG_SHELL) scripts/kwhich gcc272 2>/dev/null || $(CONFIG_SHELL) scripts/kwhich kgcc 2>/dev/null || echo cc; fi) \
-D__KERNEL__ -I$(HPATH)


> calls to memset and memcpy when manipulating structures. Because these
> internal calls are created after cpp they end up as phantom calls to
> functions instead of being converted by string.h.


In any case, don't worry about it. It's a question of adding an
#include <linux/strings.h> to every source file of divas.o and eicon.o
that mentions the word "memset". I did, and that cured the problem.

Something like vim -o `grep -w memset *.c | cut -d: -f 1 | uniq`
in the directory.


Now, obviously, all the other files in the kernel don't have this
problem with memset (I compile about 100 modules normally), so the
problem IS with the diva/eicon stuff, I am afraid. It's fixed by the
include I suggest above.

> If it is a gcc problem, you track it down by first identifying the object
>
> nm -A drivers/isdn/eicon/*.o | fgrep memset
>
> then compile the xxx object with -S
>
> make CFLAGS_xxx.o=-S SUBDIRS=drivers/isdn/eicon modules
>
> vi drivers/isdn/eicon/xxx.o looking for calls to memset. Scroll up
> until you find the function that that is generating the call, then
> eyeball the code looking for structure assignments like s = *foo or s =
> 0. Replace the assignments with explicit calls to memcpy or memset.

Well, I'd try that next, if I were desperate. It was just sufficiently
annoying that I fixed it in my own way.


Peter
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2005-03-22 13:18    [W:0.064 / U:8.392 seconds]
©2003-2017 Jasper Spaans. hosted at Digital OceanAdvertise on this site