[9fans] libc/strcmp bug?

[9fans] libc/strcmp bug?

[Skip Tavakkolian]

I believe there is a bug in /sys/src/libc/port/strcmp.c, because it
does not check for null pointers getting passed in before
dereferencing them.  I ran into it when attempting to Post some mail;
'marshal' dies with this message:

marshal 2014: suicide: sys: trap: fault read addr=0x0 pc=0x00006feb

Inspection points to the strcmp:

term% acid 2014
/proc/2014/text:386 plan 9 executable

/sys/lib/acid/port
/sys/lib/acid/386
acid: stk()
At pc:0x00006feb:strcmp+0xe /sys/src/libc/port/strcmp.c:10
strcmp(s1=0x00000000,s2=0x00013c5a) /sys/src/libc/port/strcmp.c:5
	called from islikeatty+0x32 /sys/src/cmd/upas/common/libsys.c:709
islikeatty(fd=0x00000000) /sys/src/cmd/upas/common/libsys.c:701
	called from holdon+0xf /sys/src/cmd/upas/common/libsys.c:719
holdon() /sys/src/cmd/upas/common/libsys.c:715
	called from main+0x856 /sys/src/cmd/upas/marshal/marshal.c:271
main(argv=0x7fffefec,argc=0x00000000) /sys/src/cmd/upas/marshal/marshal.c:162
	called from _main+0x31 /sys/src/libc/386/main9.s:16
acid:

I'm not sure why the first arg to strcmp was null.  I'm still investigating.

Re: [9fans] libc/strcmp bug?

[Alexander Viro]

On Fri, 9 Aug 2002, Skip Tavakkolian wrote:

> I believe there is a bug in /sys/src/libc/port/strcmp.c, because it
> does not check for null pointers getting passed in before
> dereferencing them.

Not a libc bug.  Making sure that arguments of strcmp() are non-NULL is
responsibility of caller.

Re: [9fans] libc/strcmp bug?

[arisawa]

Hello,

Viro said:
>On Fri, 9 Aug 2002, Skip Tavakkolian wrote:
>
>> I believe there is a bug in /sys/src/libc/port/strcmp.c, because
it
>> does not check for null pointers getting passed in before
>> dereferencing them.
>
>Not a libc bug.  Making sure that arguments of strcmp() are
non-NULL >is responsibility of caller.

I agree with Viro.
If null pointer is allowed, strcmp() should be something like
bellow:

int
xstrcmp(char *s, char *t)
{
        if(s == nil){
                if(t)
                        return -1;
                return 0;
        }
        if(t)
                return strcmp(s,t);
        return 1;
}

I used this code somewhere because it was useful.

Kenji Arisawa

Re: [9fans] libc/strcmp bug?

[david presotto]

That's a bug in marshal.
----- Original Message -----
From: "Skip Tavakkolian" <fst@centurytel.net>
To: <9fans@cse.psu.edu>
Sent: Friday, August 09, 2002 8:35 PM
Subject: [9fans] libc/strcmp bug?


> I believe there is a bug in /sys/src/libc/port/strcmp.c, because it
> does not check for null pointers getting passed in before
> dereferencing them.  I ran into it when attempting to Post some mail;
> 'marshal' dies with this message:
>
> marshal 2014: suicide: sys: trap: fault read addr=0x0 pc=0x00006feb
>
> Inspection points to the strcmp:
>
> term% acid 2014
> /proc/2014/text:386 plan 9 executable
>
> /sys/lib/acid/port
> /sys/lib/acid/386
> acid: stk()
> At pc:0x00006feb:strcmp+0xe /sys/src/libc/port/strcmp.c:10
> strcmp(s1=0x00000000,s2=0x00013c5a) /sys/src/libc/port/strcmp.c:5
> called from islikeatty+0x32 /sys/src/cmd/upas/common/libsys.c:709
> islikeatty(fd=0x00000000) /sys/src/cmd/upas/common/libsys.c:701
> called from holdon+0xf /sys/src/cmd/upas/common/libsys.c:719
> holdon() /sys/src/cmd/upas/common/libsys.c:715
> called from main+0x856 /sys/src/cmd/upas/marshal/marshal.c:271
> main(argv=0x7fffefec,argc=0x00000000)
/sys/src/cmd/upas/marshal/marshal.c:162
> called from _main+0x31 /sys/src/libc/386/main9.s:16
> acid:
>
> I'm not sure why the first arg to strcmp was null.  I'm still
investigating.
>
>

Re: [9fans] libc/strcmp bug?

[Skip Tavakkolian]

Of course. My mistake.

> Not a libc bug.  Making sure that arguments of strcmp() are non-NULL is
> responsibility of caller.