On Wednesday 06 July 2005 08:34 am, Arjan van de Ven wrote:
> On Tue, 2005-07-05 at 14:12 -0700, Andrew Morton wrote:
> > Bongani Hlope <[EMAIL PROTECTED]> wrote:
> > >
> > > I haven't tested 2.6.12.2 but the problem was introduced around
> > > 2.6.11-mm1 and
> > > found its way to 2.6.12-rcX. First try to run the following command
> > > (this works for me)
> > > echo 0 > /proc/sys/kernel/randomize_va_space
> > > I got an email from Juan Gallego (cc'd), he says that command does not
> > > work for him though.
> > >
> > > Andrew,
> > > Should I log this on the kernel's bugzilla?
> >
> > Yes please. This is a tough one, and having one place to go to for the
> > info would be useful.
>
> key for this one is to make sure we separate the cases carefully, and
> not end up with one big bucket of "something broke" that has a gazilion
> different and unrelated causes.
> For the cases where a vm layout thing is suspected of causing the
> breakage we also really need a /proc/<pid>/maps *at the time of the
> breakage* realistically for doing any kind of diagnostics.
>
The problem is, it is hard to get the /proc/<pid>/maps file because the crashes
are just random.
If I set /proc/sys/kernel/randomize_va_space back to 1, and do a make -j4 on
the kernel source.
gcc will start to segfault or cause protection errors. I added a printk in the
randomize_stack_top function
just to print the pid and the random_variable (yes it is silly but...), the
problem went away.
I tried to capture the /proc/<pid>/maps for the broken apps using the code
below (it captures some, but it is brain dead),
but the problem goes away. I'll try to write a bash script that does the same
and test.
#include <sys/types.h>
#include <sys/wait.h>
#include <unistd.h>
#include <cstdio>
#include <iostream>
#include <fstream>
using namespace std;
int main(int argc, char** argv) {
pid_t child;
child = fork();
if(child == 0) {
execvp("/usr/bin/g++", argv);
}
else if (child > 0) {
char* name = new char[256];
sprintf(name,"/proc/%ld/maps", child);
ifstream in(name);
delete[] name;
ofstream out("/home/bongani/log/procs.txt", ofstream::app);
name = new char[4096];
out<<"--------------------------------------------"<<"\n";
out<<"g++("<<child<<")"<<"\n";
while(!in.eof()) {
in.getline(name, 4096);
out<<child<<" "<<name<<"\n";
}
out<<"----------------------------------------------"<<endl;
do {
}
while(child != wait(NULL));
delete[] name;
}
else {
printf("fork failed");
}
return 0;
}
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
