On Wed, 2009-02-11 at 14:54 -0800, Alok Kataria wrote:
> Hi Tetsuo,
>
> Thanks for reporting this. Yes i tried it here with a slightly
> different .config and i did see the problem. I will try to figure out
> what is happening over here, but before doing that let me ask if this
> bug is only in the stable series ?
> Do we know for which kernel does the "nolapic" option doesn't hit this
> bug ?
>
> Also, I assume its not present in 2.6.29-rc ?
Okay i went through the whole thread, and notice that all of my queries
were previously answered.
And what happened to this patch of checking the stats only if its not
null.
+if (!lb_stats) {
+ printk(KERN_INFO "lb_stats == NULL for CPU %u\n", i);
+ continue;
+
}
>
> Thanks,
> Alok
>
> On Mon, 2009-02-09 at 03:19 -0800, Tetsuo Handa wrote:
> > Hello.
> >
> > > Bug-Entry : http://bugzilla.kernel.org/show_bug.cgi?id=12409
> > > Subject : NULL pointer dereference at get_stats()
> > > Submitter : Tetsuo Handa <[email protected]>
> > > Date : 2008-12-30 12:53 (21 days old)
> > > References : http://marc.info/?l=linux-kernel&m=123064167008695&w=4
> > > Handled-By : Frederik Deweerdt <[email protected]>
> >
> > I made custom initramfs and tested on native environment and KVM and VMware.
> > This problem happened only in VMware.
> >
> > Alok and Daniel (cc: added), can you reproduce this problem?
> >
> > Problem description:
> >
> > Compiling 2.6.28.4 using config at
> > http://I-love.SAKURA.ne.jp/tmp/config-2.6.28-bug
> > and booting with "nolapic" option on x86 machine with 2 CPUs, NULL pointer
> > dereference happens at get_stats() because for_each_possible_cpu() reaches
> > CPU 1 while "nolapic" option should prevent for_each_possible_cpu() from
> > reaching CPU 1.
> >
> > Compiling 2.6.28.4 using config at
> > http://I-love.SAKURA.ne.jp/tmp/config-2.6.28-nobug
> > (all options in "Power management and ACPI options" section disabled)
> > solves this problem.
> >
> > Also, this problem seems to be VMware specific.
> > I couldn't reproduce this problem on native environment and KVM.
> >
> > Native environment (no problem):
> >
> > CentOS 5.2 (x86_64) and Ubuntu 8.04 (i386) on ThinkPad X60
> >
> > KVM environment (no problem):
> >
> > kvm -kernel /boot/vmlinuz-2.6.28.4 -initrd /boot/init -net none -hda
> > /var/tmp/image.img -vnc xx.xx.xx.xx:0 --append "ro nolapic" -smp 2
> > on Ubuntu 8.04 (i386) on ThinkPad X60
> >
> > VMware environment (problematic):
> >
> > Debian Sarge (i386) / CentOS 5.2 (i386) on VMware workstation 6.5.1
> > (x86_64)
> > on CentOS 5.2 (x86_64) on ThinkPad X60
> >
> > Problematic kernel versions:
> >
> > 2.6.27.x and 2.6.28.x .
> >
> > Regards.
> >
> > ----- Custom initramfs -----
> > /*
> > gcc -Wall -O3 -static -o init init.c
> > echo init | cpio -o -H newc | gzip -9 > /boot/init
> > */
> > #include <sys/mount.h>
> > #include <sys/types.h>
> > #include <sys/stat.h>
> > #include <fcntl.h>
> > #include <unistd.h>
> > #include <stdio.h>
> > #include <string.h>
> >
> > int main(int argc, char *argv[]) {
> > FILE *fp;
> > char buffer[1024];
> > memset(buffer, 0, sizeof(buffer));
> > mkdir("/proc", 0755);
> > mount("none", "/proc", "proc", 0, NULL);
> > fp = fopen("/proc/cpuinfo", "r");
> > while (fgets(buffer, sizeof(buffer) - 1, fp))
> > if (strstr(buffer, "processor"))
> > printf("%s", buffer);
> > fclose(fp);
> > fp = fopen("/proc/cmdline", "r");
> > fgets(buffer, sizeof(buffer) - 1, fp);
> > fclose(fp);
> > printf("%s", buffer);
> > fflush(stdout);
> > if (mkdir("/sys", 0755) || mount("/sys/", "/sys/", "sysfs", 0, NULL)) {
> > printf("mount failed\n");
> > while (1)
> > sleep(1);
> > }
> > fp = fopen("/sys/class/net/lo/statistics/rx_packets", "r");
> > if (!fp) {
> > printf("open failed\n");
> > while (1)
> > sleep(1);
> > }
> > while (fgets(buffer, sizeof(buffer) - 1, fp))
> > ;
> > fclose(fp);
> > printf("done\n");
> > while (1)
> > sleep(1);
> > return 0;
> > }
--
To unsubscribe from this list: send the line "unsubscribe kernel-testers" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
