kern/182139: lock order reversal

John D. Hendrickson and Sara Darnell johnandsara2 at cox.net
Mon Sep 16 20:25:57 UTC 2013


i don't get that message using linux.  however i had to solve a bad 
IDE drive problem once caused by microsoft software the IDE company 
installed on it.

what is your question?

#1) wrong lock order is very serious: it's the whole damn disk hw and 
driver is built for that: if that fails you have dick.  lock order is 
all that is between your data and being trashed by old data before 
getting on disk (if it gets on disk, with wrong locks)

#2) if mount throws messages it could be a hw error, driver problem, 
fat table mutation, etc.  it really depends you have to find out.  i'm 
saying: do you know where and when the problem is thrown?  because 
most locking is in driver code or memory caching code.

for example, maybe a wrong fat table entry causes a glitch which 
causes the kernel to panic while sorting lock order - and you don't 
see or haven't read the source code to know that's what's going on.

i haven't tried bsd yet because it seems so many messages point out 
serious problems recently developing (caused after bsd stable public 
release by hackers making "improvements" to bsd).  i was only going to 
try it becuase it used to be "rock solid" and "fully compiled".  bsd 
channel scares me as to how many problems and why: from any gov agent 
or company all over the world hacks into drivers? ughh.

is there any management or overview of who is hacking into what parts 
of bsd these days?  or why?  i mean, does the univ. of cal. have 
professors that have keys?  or do foreigners have keys and are calling 
it "bsd" when it isn't?

i know in my area as soon as gov workers invested in microsoft stock 
all unix classes closed and classes started requiring ms use (the area 
had few compared to CA to begin with).  ms moved into area and got gov 
contracts.  recently i think a few unix classes have arrived back. 
but nothing with funding or bite, ie, for e-commerce and e-gov. 
though apple does (some) of that.  apple makes progress there.  but 
most classes still require ms boxes.  all a damn scam if you ask me.


well good luck.  hope some release for bsd is called stable and i get 
to try it.


J B wrote:
> The following reply was made to PR kern/182139; it has been noted by GNATS.
> 
> From: J B <jb.1234abcd at gmail.com>
> To: bug-followup at FreeBSD.org, jb.1234abcd at gmail.com
> Cc:  
> Subject: Re: kern/182139: lock order reversal
> Date: Mon, 16 Sep 2013 06:50:08 +0200
> 
>  --f46d0438eb6d4a888e04e678f0c5
>  Content-Type: text/plain; charset=ISO-8859-1
>  
>  I got similar errors, I assume after umount.
>  # mount -t ext2fs /dev/ada0s6 /mnt/
>  # umount /mnt
>  # cat /var/log/messages
>  ...
>  Sep 16 06:35:41 localhost kernel: lock order reversal:
>  Sep 16 06:35:41 localhost kernel: 1st 0xc8077d84 ufs (ufs) @
>  /usr/src/sys/kern/vfs_mount.c:1237
>  Sep 16 06:35:41 localhost kernel: 2nd 0xc8077a30 syncer (syncer) @
>  /usr/src/sys/kern/vfs_subr.c:2210
>  Sep 16 06:35:41 localhost kernel: KDB: stack backtrace:
>  Sep 16 06:35:41 localhost kernel:


More information about the freebsd-bugs mailing list