amd64/92889: xdr double buffer overflow
Menshutin Anton
may at chg.ru
Mon Feb 6 05:00:16 PST 2006
>Number: 92889
>Category: amd64
>Synopsis: xdr double buffer overflow
>Confidential: no
>Severity: serious
>Priority: low
>Responsible: freebsd-amd64
>State: open
>Quarter:
>Keywords:
>Date-Required:
>Class: sw-bug
>Submitter-Id: current-users
>Arrival-Date: Mon Feb 06 13:00:14 GMT 2006
>Closed-Date:
>Last-Modified:
>Originator: Menshutin Anton
>Release: 5.4 amd64
>Organization:
ITP
>Environment:
FreeBSD ***.***.ru 5.4-RELEASE FreeBSD 5.4-RELEASE #0: Thu Nov 24 13:33:07 UTC 2005 ***@***.***.ru:/usr/src/sys/amd64/compile/SMP amd64
>Description:
Problem in xdr functions from libc on amd64. Sending doubles or floats with pvm library causes out of range access in receiving program. First 4 bytes of double are being received correctly (when sending from i386 machine) but the last 4 bytes come to the wrong place, owerwriting 4 bytes memory AFTER the double. Seems to be a problem with libc xdr packing\unpacking functions.
>How-To-Repeat:
Write a simple test program sending double from one host to another.
//test_pvm.c
#include <stdio.h>
#include "pvm3.h"
char *hostname="localhost"; //hostname where to run slave
char buf[1024];
int main()
{
int tid;
pvm_catchout(stdout);
pvm_spawn("test_pvm_slave",NULL,PvmTaskHost,
hostname,1,&tid);
pvm_initsend(PvmDataDefault);
double t=0.123;
double *p;
int i;
unsigned int k;
p=buf;
*p=t;
printf("Data before sending:");
for (i=0;i<16;i++)
{
printf("%hhu ",buf[i]);
}
printf("\n");
pvm_pkdouble(p,1,1);
pvm_send(tid,0);
pvm_exit();
return(0);
}
//test_pvm_slave.c
#include "pvm3.h"
#include <stdio.h>
char buf[1024];
int main()
{
int tid;
double t=0.123;
double *p;
int i;
fprintf(stderr,"I am slave\n");
pvm_recv(pvm_parent(),0);
p=buf;
pvm_upkdouble(p,1,1);
t=*p;
for (i=0;i<16;i++)
{
printf("%hhu ",buf[i]);
}
printf("\n");
pvm_exit();
return(0);
}
>Fix:
>Release-Note:
>Audit-Trail:
>Unformatted:
More information about the freebsd-amd64
mailing list