Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 13 May 2002 12:15:59 -0700
From:      Marcel Moolenaar <marcel@cup.hp.com>
To:        Andrew Atrens <andrew_atrens@yahoo.ca>
Cc:        java@freebsd.org
Subject:   Re: linux-ibm-jdk-1.3.1[x] broken for non-root users on -stable
Message-ID:  <20020513191559.GA42806@gauss.cup.hp.com>
In-Reply-To: <20020513181818.16354.qmail@web11305.mail.yahoo.com>
References:  <20020513181818.16354.qmail@web11305.mail.yahoo.com>

next in thread | previous in thread | raw e-mail | index | archive | help
On Mon, May 13, 2002 at 02:18:18PM -0400, Andrew Atrens wrote:
> linux ibm jdk1.3.1_1 run as joe user -
> 
>  26966 java     CALL  linux_modify_ldt(0x11,0xbfbff8ac,0x10)
>  26966 java     RET   linux_modify_ldt 8190/0x1ffe
>  26966 java     CALL  linux_sigaltstack(0xbfbff8ac,0x806ccf0)
>  26966 java     RET   linux_sigaltstack 0
>  26966 java     PSIG  SIGBUS SIG_DFL
>  26966 java     NAMI  "java.core"
> 
> linux ibm jdk1.3.1_1 run as root -
> 
>  27019 java     CALL  linux_modify_ldt(0x11,0xbfbff8ac,0x10)
>  27019 java     RET   linux_modify_ldt 8190/0x1ffe
>  27019 java     CALL  linux_sigaltstack(0xbfbff8ac,0x806ccf0)
>  27019 java     RET   linux_sigaltstack 0
>  27019 java     CALL  mprotect(0xbfbc8000,0x1000,0)
>  27019 java     RET   mprotect 0

Hi Andrew,

The kernel trace does not indicate there's a problem with emulation
per se. The SIGBUS is caused within java and unfortunately, there's
no way a kernel trace can tell us if the SIGBUS is caused by how
emulation works. I suspect it does, but it doesn't tell me where I
need to look.

The best thing to do now is to disassemble the code between the
call to sigaltstack and the call to mprotect and figure out where
exactly the SIGBUS is coming from. You can backtrace from there.

I suspect the root of the problem to be prior to the sigaltstack
though...

Another approach would be to look at high risk syscalls (such as
modify_ldt) and see if they behave differently for joe and root.
I'm not surprised to see silent "adjustments" to parameters and/or
results based on privileges. This is a more random approach, but
if you have 1 or 2 strong suspects then it's the quickest way to
remove any wandering doubts from your brain that may block real
progress..

HTH,

-- 
 Marcel Moolenaar	  USPA: A-39004		 marcel@xcllnt.net

To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-java" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20020513191559.GA42806>