JDK-4947404 : AS7UR2 ER build crashed in Solaris x86 10 b45
  • Type: Bug
  • Component: hotspot
  • Sub-Component: compiler
  • Affected Version:
    solaris_10,3.5,7.1,1.4.1_05,1.4.2_02 solaris_10,3.5,7.1,1.4.1_05,1.4.2_02
  • Priority: P1
  • Status: Resolved
  • Resolution: Fixed
  • OS: solaris_7,solaris_9,solaris_10
  • CPU: x86,sparc
  • Submitted: 2003-11-01
  • Updated: 2005-02-03
  • Resolved: 2003-12-15
The Version table provides details related to the release that this issue/RFE will be addressed.

Unresolved : Release in which this issue/RFE will be addressed.
Resolved: Release in which this issue/RFE has been resolved.
Fixed : Release in which this issue/RFE has been fixed. The release containing this fix may be available for download as an Early Access Release or a General Availability Release.

To download the current JDK release, click here.
Other Other
1.4.2_03,solaris_10Fixed 1.4.2_04Fixed
Related Reports
Duplicate :  
Duplicate :  
Relates :  
Description
AS7 UR2 ER build:
When we install bundled build for Sol 10 x86 b45, we swa following error:

Created Domain domain1 successfully
 CORE1116: Sun ONE Application Server 7.0.0_02.1_548819
 INFO: CORE3016: daemon is running as super-user
 INFO: CORE5076: Using [Java HotSpot(TM) Server VM,

Version  1.4.2_02] from [Sun Microsystems Inc.]
 INFO: ADM0002:System MBean
initialized:[ias:type=controller]

 INFO: ADM0002:System MBean

initialized:[ias:type=configurator]

 INFO: ADM0001:MBeanServer initialized successfully
 INFO: ADM0005:Timestamp files for configuration

created for:[admin-server]INFO: ADM0102: Starting a thread for tracking manual changes

 INFO: IOP5053: Received a locate request on a disabled connection. 

 Locate requests are permitted.
 INFO: JTS5014: Recoverable JTS instance, serverId

 [100]

 INFO: RAR5060: Install JDBC Datasources ...
 INFO: JMS5015: Install JMS resources ...
 FATAL: CORE3260: Server crash detected (signal SIGFPE)
 Failed to start admin server.
 

We had removed two NSS/SPR pkg name from  bundled_pkg_sol_5.9.txt.

>cat bundled_pkg_sol_5.9.txt
SUNWjhrt
SUNWasac
SUNWasdev
SUNWasdem
SUNWasman
SUNWasr
SUNWasu
SUNWascmn
SUNWaclg
SUNWjaf
SUNWjaxp
SUNWjmail
SUNWxsrt
SUNWxrgrt
SUNWxrpcrt

Sonia also reproduced same problem in another Sol 10 x86 b45 machine.

Following are the emails:
Hi Becky,

I just got the x86 machine and tried to install the
bundled bits. I got the exactly same erro as Yingbin
got.  I think we should report a P1 bug for it. What
do you think?

Thanks
Sonia

----- Original Message -----
From: becky schmitt <###@###.###>
Date: Friday, October 31, 2003 10:12 am
Subject: [Fwd: Re: [Fwd: S1AS7 SE UR2  ER Promoted.]]


> Hi Guys
> 
> Have any of you seen this crash with the bundled

distribution

> for X86 on the exception release?  
> 
> 
> -------- Original Message --------
> Subject: Re: [Fwd: S1AS7 SE UR2  ER Promoted.]
> Date: Thu, 30 Oct 2003 20:44:30 -0800
> From: yingbin <###@###.###>
> Reply-To: ###@###.###
> To: ###@###.###
> CC: pocs <###@###.###>, ###@###.###
> References: <###@###.###>
> 
> Hi
> Does anybody installed x86 bundled build? When I

ran the 

> installation scripts with modified
> pkg list(15 nessesary pkg), I saw a server crash.

See bellow:

> 
> Created Domain domain1 successfully
> CORE1116: Sun ONE Application Server 7.0.0_02.1_548819
> INFO: CORE3016: daemon is running as super-user
> INFO: CORE5076: Using [Java HotSpot(TM) Server VM,

Version 

> 1.4.2_02] from [Sun M
> icrosystems Inc.]
> INFO: ADM0002:System MBean

initialized:[ias:type=controller]

> INFO: ADM0002:System MBean

initialized:[ias:type=configurator]

> INFO: ADM0001:MBeanServer initialized successfully
> INFO: ADM0005:Timestamp files for configuration

created for:[admin-

> server]INFO: ADM0102: Starting a thread for

tracking manual changes

> INFO: IOP5053: Received a locate request on a

disabled connection. 

> Locate reques
> ts are permitted.
> INFO: JTS5014: Recoverable JTS instance, serverId

= [100]

> INFO: RAR5060: Install JDBC Datasources ...
> INFO: JMS5015: Install JMS resources ...
> FATAL: CORE3260: Server crash detected (signal SIGFPE)
> Failed to start admin server.
> 
> 
> Yingbin
> 
> 
> 
 
 
 
###@###.### 2003-10-31
###@###.### 2003-10-31

================================================================================


###@###.### 2003-11-12

I was able to reproduce this bug on S10 build 46 as foillows

Solaris 10 s10_46 X86

jan@brion>which asadmin
/usr/sbin/asadmin
jan@brion>asadmin version
Sun ONE Application Server 7.0.0_01


/usr/j2se/bin/java -version
java version "1.4.2_02"
Java(TM) 2 Runtime Environment, Standard Edition (build 1.4.2_02-b03)
Java HotSpot(TM) Client VM (build 1.4.2_02-b03, mixed mode)

As root

asadmin create-domain --adminport 4849 --adminuser admin --adminpassword admin123 domain1
Created Domain domain1 successfully

cd /var/appserver/domains/domain1/admin-server/bin

startserv
CORE1116: Sun ONE Application Server 7.0.0_01
INFO: CORE3016: daemon is running as super-user
INFO: CORE5076: Using [Java HotSpot(TM) Server VM, Version 1.4.2_02] from [Sun Microsystems Inc.]
INFO: ADM0002:System MBean initialized:[ias:type=controller]
INFO: ADM0002:System MBean initialized:[ias:type=configurator]
INFO: ADM0001:MBeanServer initialized successfully
INFO: ADM0005:Timestamp files for configuration created for:[admin-server]
INFO: ADM0102: Starting a thread for tracking manual changes
INFO: IOP5053: Received a locate request on a disabled connection. Locate requests are permitted.
INFO: JTS5014: Recoverable JTS instance, serverId = [100]
INFO: RAR5060: Install JDBC Datasources ...
INFO: JMS5015: Install JMS resources ...
WARNING: CORE3283: stderr: caught SIGFPE, info 0x6.
FATAL: CORE3260: Server crash detected (signal SIGFPE)

Comments
CONVERTED DATA BugTraq+ Release Management Values COMMIT TO FIX: 1.4.2_03 1.4.2_04 generic tiger-beta2 FIXED IN: 1.4.2_03 1.4.2_04 tiger-beta2 INTEGRATED IN: 1.4.2_03 1.4.2_04 s10_48 tiger-beta2
14-06-2004

WORK AROUND ======================================================================= ###@###.### 2003-11-07 Add JVM flag -XX:UseSSE=0 to Application Server flags. It prevents SSE/SSE2 instructions generation in compiled java methods. ======================================================================= One workaround is to completely disable SSE support on the platform; put set x86_archext & ~0x6000 in /etc/system and reboot. This is an unfortunate solution to the problem because it disables -all- use of the SSE instructions; including the kernel. A better workaround is needed.
11-06-2004

PUBLIC COMMENTS ###@###.### 2003-11-10 MANAGEMENT SUMMARY ================== 1) this is a bug in the JVM, not Solaris or AppServer 2) the effect is that the AppServer will not start after installation 3) reproducing this bug requires the SSE support in Solaris 10 (that was introduced in build 45) and JDK 1.4.2_02.
10-06-2004

EVALUATION I tested it on Solaris 9 x86. It is working fine. I am using QA's machine and I do see the problem. ###@###.### 2003-11-03 This bug is reproducable only if you use Solaris x86 10 Build 45. During admin server startup you get a SIGFPE exception. Thanks Vani ###@###.### 2003-11-06 The problem seems to be that the state of the %mxcsr register (the control/status register for the SSE instructions) is not properly initialized when the JVM is initialized from the appservd program. I'm not exactly sure why this is happening. It's possible that either appservd is doing something to cause this or there is another bug which causes lwp's to not get their fp.p state initialized properly in some cases. I plan to fix this bug by changing the VM to initialize the %mxcsr register during its initialization and for all threads the VM creates. It much safer for the VM to initialize any global state it relies on rather than assuming the system has already done it. ###@###.### 2003-11-12 After implementing the above fix, the appserverd was able to start up and run successfully. It is likely that there is still a bug in Solaris which causes the incorrect initialization of the %mxcsr. In the situation reported, this occurs before the JVM begins execution. Since this has not been seen in normal java execution, the JVM does not seem to trigger the bug. So I am fairly confident that my fix will prevent the JVM from being affected by the bug. ###@###.### 2003-11-17 This has been fixed in tiger under bugid 5003738. ###@###.### 2004-03-23
17-11-2003