freebsd stable - Feb 2014 - What is difference between loading module with loader and loading module wtih kldload?

Hello, Hackers.

  I've upgraded my fileserver to 10-STABLE and got very strange (but
 very-very painful) problem, which I could not reproduce on virtual machine
 (VirtualBox).

 I'm using geom_raid5 module (and I'm its maintainer, yes) and module,
built
for 10-STABLE (after world & kernel build & install & reboot), and
loaded
via /boot/loader.conf is reason to almost instant crash after boot.
Sometimes system mounts filesystems before crash and sometimes not. Most of
time it is "page write: page not present", in different places. PS/2
keyboard is always blocked after that, I could not drop to debugger. No
memory dump performed. Several times it turend off video output (!) right
after crash.

 But if I boot without this module, drop to single-user mode, load module
with kldload and continue booting with "exit" everything work smoothly
for
hours!

 I understand, that it it some incompatibility between module new kernel,
but I could not reproduce it on VirtualBox instance, and I'm puzzled, that
this crash does not occur if module loaded by kldload! Maybe, here is some
hint in this?

-- 
// Black Lion AKA Lev Serebryakov <lev at FreeBSD.org>

> Hello, Hackers.
>
>   I've upgraded my fileserver to 10-STABLE and got very strange (but
>  very-very painful) problem, which I could not reproduce on virtual machine
>  (VirtualBox).
>
>  I'm using geom_raid5 module (and I'm its maintainer, yes) and
module, built
> for 10-STABLE (after world & kernel build & install & reboot),
and loaded
> via /boot/loader.conf is reason to almost instant crash after boot.
> Sometimes system mounts filesystems before crash and sometimes not. Most of
> time it is "page write: page not present", in different places.
PS/2
> keyboard is always blocked after that, I could not drop to debugger. No
> memory dump performed. Several times it turend off video output (!) right
> after crash.
>
>  But if I boot without this module, drop to single-user mode, load module
> with kldload and continue booting with "exit" everything work
smoothly for
> hours!
>
>  I understand, that it it some incompatibility between module new kernel,
> but I could not reproduce it on VirtualBox instance, and I'm puzzled,
that
> this crash does not occur if module loaded by kldload! Maybe, here is some
> hint in this?
Warning, I'm no expert on 10.x. I see too many issues by ppl on the lists.
So it
seems too shakey for me. But, I have a couple of ideas, hoping they might help.
I had a similar problem when using a very large drive, or mirror set (geom), a
few years ago. FreeBSD didn't see/allocate all the memory available on the
system. So it barfed, because it didn't have enough to allocate to handle
the
drive size. I'm sorry, but I have forgotten how to TELL FreeBSD how much
memory it can use during boot tho. But if you know, then this might help you
too.

Hope this helps.

--Chris
>
> --
> // Black Lion AKA Lev Serebryakov <lev at FreeBSD.org>
>
> _______________________________________________
> freebsd-stable at freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-stable
> To unsubscribe, send any mail to "freebsd-stable-unsubscribe at
freebsd.org"
>

Lev Serebryakov wrote this message on Wed, Feb 26, 2014 at 01:31
+0400:
>   I've upgraded my fileserver to 10-STABLE and got very strange (but
>  very-very painful) problem, which I could not reproduce on virtual machine
>  (VirtualBox).
> 
>  I'm using geom_raid5 module (and I'm its maintainer, yes) and
module, built
> for 10-STABLE (after world & kernel build & install & reboot),
and loaded
> via /boot/loader.conf is reason to almost instant crash after boot.
> Sometimes system mounts filesystems before crash and sometimes not. Most of
> time it is "page write: page not present", in different places.
PS/2
> keyboard is always blocked after that, I could not drop to debugger. No
> memory dump performed. Several times it turend off video output (!) right
> after crash.
Can you give us an exact error message?  I am not finding the string:
"page write: page not present" anywhere in the tree, on close thing is
in trap, where it could be user/supervisor write/read instruction/data
page not present, where words seperated by slashes could be one or the
other...

This sounds like it could be a buffer overflow...  Could you try turning
on INVARIANTS and other related debugging on 10-STABLE since these were
turned off for the RELEASE?

Also uname -a would be helpful to know which arch you are on...
>  But if I boot without this module, drop to single-user mode, load module
> with kldload and continue booting with "exit" everything work
smoothly for
> hours!
> 
>  I understand, that it it some incompatibility between module new kernel,
> but I could not reproduce it on VirtualBox instance, and I'm puzzled,
that
> this crash does not occur if module loaded by kldload! Maybe, here is some
> hint in this?
There are a few differences between a boot time loaded module and a
runtime loaded module...  The linker runs earlier at boot time to link
up the module before things start, the module may be partly run w/ cold
set (which informs us the interrupts and other things may not be fully
working).. Sysinits are run in a slightly different order (i.e. kernel
SYSINITs that appear after your modules will be run first)...

-- 
  John-Mark Gurney				Voice: +1 415 225 5579

     "All that I will do, has been done, All that I have, has not."

Hello, Lev.
You wrote 26 февраля 2014 г., 1:31:07:

LS>  But if I boot without this module, drop to single-user mode, load module
LS> with kldload and continue booting with "exit" everything work
smoothly for
LS> hours!
  But then crashes too :(

-- 
// Black Lion AKA Lev Serebryakov <lev@FreeBSD.org>

_______________________________________________
freebsd-stable@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to
"freebsd-stable-unsubscribe@freebsd.org"

Hello, Lev.
You wrote 26 ??????? 2014 ?., 1:31:07:

LS>  But if I boot without this module, drop to single-user mode, load module
LS> with kldload and continue booting with "exit" everything work
smoothly for
LS> hours!
  But then crashes too :(

-- 
// Black Lion AKA Lev Serebryakov <lev at FreeBSD.org>

On 2/25/14, Lev Serebryakov <lev at freebsd.org>
wrote:
> Hello, Hackers.
>
>   I've upgraded my fileserver to 10-STABLE and got very strange (but
>  very-very painful) problem, which I could not reproduce on virtual machine
>  (VirtualBox).
>
>  I'm using geom_raid5 module (and I'm its maintainer, yes) and
module,
> built
> for 10-STABLE (after world & kernel build & install & reboot),
and loaded
> via /boot/loader.conf is reason to almost instant crash after boot.
> Sometimes system mounts filesystems before crash and sometimes not. Most of
> time it is "page write: page not present", in different places.
PS/2
> keyboard is always blocked after that, I could not drop to debugger. No
> memory dump performed. Several times it turend off video output (!) right
> after crash.
I have a similar crash with geom_sched. This introduced after fine
graded geom locks.
>
>  But if I boot without this module, drop to single-user mode, load module
> with kldload and continue booting with "exit" everything work
smoothly for
> hours!
>
>  I understand, that it it some incompatibility between module new kernel,
> but I could not reproduce it on VirtualBox instance, and I'm puzzled,
that
> this crash does not occur if module loaded by kldload! Maybe, here is some
> hint in this?
>
> --
> // Black Lion AKA Lev Serebryakov <lev at FreeBSD.org>
>
> _______________________________________________
> freebsd-stable at freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-stable
> To unsubscribe, send any mail to "freebsd-stable-unsubscribe at
freebsd.org"
>

On 2/25/14, Lev Serebryakov <lev@freebsd.org>
wrote:
> Hello, Hackers.
>
>   I've upgraded my fileserver to 10-STABLE and got very strange (but
>  very-very painful) problem, which I could not reproduce on virtual machine
>  (VirtualBox).
>
>  I'm using geom_raid5 module (and I'm its maintainer, yes) and
module,
> built
> for 10-STABLE (after world & kernel build & install & reboot),
and loaded
> via /boot/loader.conf is reason to almost instant crash after boot.
> Sometimes system mounts filesystems before crash and sometimes not. Most of
> time it is "page write: page not present", in different places.
PS/2
> keyboard is always blocked after that, I could not drop to debugger. No
> memory dump performed. Several times it turend off video output (!) right
> after crash.
I have a similar crash with geom_sched. This introduced after fine
graded geom locks.
>
>  But if I boot without this module, drop to single-user mode, load module
> with kldload and continue booting with "exit" everything work
smoothly for
> hours!
>
>  I understand, that it it some incompatibility between module new kernel,
> but I could not reproduce it on VirtualBox instance, and I'm puzzled,
that
> this crash does not occur if module loaded by kldload! Maybe, here is some
> hint in this?
>
> --
> // Black Lion AKA Lev Serebryakov <lev@FreeBSD.org>
>
> _______________________________________________
> freebsd-stable@freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-stable
> To unsubscribe, send any mail to
"freebsd-stable-unsubscribe@freebsd.org"
>
_______________________________________________
freebsd-stable@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to
"freebsd-stable-unsubscribe@freebsd.org"

Am 25.02.2014 22:31, schrieb Lev Serebryakov:
> Hello, Hackers.
>
>    I've upgraded my fileserver to 10-STABLE and got very strange (but
>   very-very painful) problem, which I could not reproduce on virtual
machine
>   (VirtualBox).
>
>   I'm using geom_raid5 module (and I'm its maintainer, yes) and
module, built
> for 10-STABLE (after world & kernel build & install & reboot),
and loaded
> via /boot/loader.conf is reason to almost instant crash after boot.
> Sometimes system mounts filesystems before crash and sometimes not. Most of
> time it is "page write: page not present", in different places.
PS/2
> keyboard is always blocked after that, I could not drop to debugger. No
> memory dump performed. Several times it turend off video output (!) right
> after crash.
Can you find some hint?: 
https://www.google.de/search?q=page+write%3A+page+not+present
>   But if I boot without this module, drop to single-user mode, load module
> with kldload and continue booting with "exit" everything work
smoothly for
> hours!
>
>   I understand, that it it some incompatibility between module new kernel,
> but I could not reproduce it on VirtualBox instance, and I'm puzzled,
that
> this crash does not occur if module loaded by kldload! Maybe, here is some
> hint in this?
>
Greetings