Bug#419903: openafs-client: aklog crashes with LWP: stack overflow in process IO MANAGER!

[Thomas Bushnell BSG]

On Wed, 2007-04-18 at 11:21 -0700, Russ Allbery wrote:
> Can you try applying the following patch to the source package and
> rebuilding?  I don't have a Debian PPC system running AFS to easily test
> with, but we think this may fix the problem.

Sorry I couldn't get to it earlier, but I can confirm that the uploaded
patch does solve the problem on my end, as was predicted.

Thomas



signature.asc
Description: This is a digitally signed message part

Bug#419903: openafs-client: aklog crashes with LWP: stack overflow in process IO MANAGER!

[Russ Allbery]

Christian Guggenberger <[EMAIL PROTECTED]> writes:

>> Can you try applying the following patch to the source package and
>> rebuilding?  I don't have a Debian PPC system running AFS to easily test
>> with, but we think this may fix the problem.

> FYI, I can confirm that the patch in question fixes the aklog problems on
> ppc.

Excellent, thank you.  I'll get it into the next release.

-- 
Russ Allbery ([EMAIL PROTECTED])   


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Bug#419903: openafs-client: aklog crashes with LWP: stack overflow in process IO MANAGER!

[Christian Guggenberger]

>Can you try applying the following patch to the source package and
>rebuilding?  I don't have a Debian PPC system running AFS to easily test
>with, but we think this may fix the problem.

FYI, I can confirm that the patch in question fixes the aklog problems on
ppc.

cheers.
 - Christian
 


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Bug#419903: openafs-client: aklog crashes with LWP: stack overflow in process IO MANAGER!

[Russ Allbery]

Thomas Bushnell BSG <[EMAIL PROTECTED]> writes:
> On Wed, 2007-04-18 at 11:21 -0700, Russ Allbery wrote:

>> Can you try applying the following patch to the source package and
>> rebuilding?  I don't have a Debian PPC system running AFS to easily
>> test with, but we think this may fix the problem.

> I'm in the middle of a crunch moment in the preparation of my doctoral
> thesis.  I'll be able to deal this weekend, or maybe Monday, but not
> before then.

> I'm grateful for the patch, and I hope it does work!

Okay, thanks.  There are some other near-RC fixes that I need to make to
OpenAFS, so I may aim at getting them all into the next upload.

-- 
Russ Allbery ([EMAIL PROTECTED])   


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Bug#419903: openafs-client: aklog crashes with LWP: stack overflow in process IO MANAGER!

[Thomas Bushnell BSG]

On Wed, 2007-04-18 at 11:21 -0700, Russ Allbery wrote:
> Can you try applying the following patch to the source package and
> rebuilding?  I don't have a Debian PPC system running AFS to easily test
> with, but we think this may fix the problem.

I'm in the middle of a crunch moment in the preparation of my doctoral
thesis.  I'll be able to deal this weekend, or maybe Monday, but not
before then.

I'm grateful for the patch, and I hope it does work!

Thomas



signature.asc
Description: This is a digitally signed message part

Bug#419903: openafs-client: aklog crashes with LWP: stack overflow in process IO MANAGER!

[Russ Allbery]

Thomas Bushnell BSG <[EMAIL PROTECTED]> writes:

> Package: openafs-client
> Version: 1.4.4.dfsg1-1
> Severity: grave
> Justification: renders package unusable

> When I run aklog, I get the following helpful error message:

> ~$ aklog -d
> Authenticating to cell becket.net (server aquinas.becket.net).
> We've deduced that we need to authenticate to realm BECKET.NET.
> Getting tickets: afs/[EMAIL PROTECTED]
> Principal not found, trying alternate service name: [EMAIL PROTECTED]
> Using Kerberos V5 ticket natively
> About to resolve name tb to id in cell becket.net.
> stackcheck = 66051: stack = 66051
> topstack = 0x3007b034: stackptr = 0x3004b008: stacksize = 0x3
> Wed Apr 18 08:35:47 2007 LWP: stack overflow in process IO MANAGER!
> Aborted

Can you try applying the following patch to the source package and
rebuilding?  I don't have a Debian PPC system running AFS to easily test
with, but we think this may fix the problem.

--- src/config/param.ppc_linux26.h  (revision 2485)
+++ src/config/param.ppc_linux26.h  (local)
@@ -74,6 +74,11 @@
 #endif
 #endif /* KERNEL */
 
+#ifdef __GLIBC__
+#if (__GLIBC__ > 2) || (__GLIBC__ == 2 && __GLIBC_MINOR__ > 3)
+#define USE_UCONTEXT
+#endif
+#endif
 #endif /* AFS_PARAM_H */
 
 #else /* !defined(UKERNEL) */

-- 
Russ Allbery ([EMAIL PROTECTED])   


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Bug#419903: openafs-client: aklog crashes with LWP: stack overflow in process IO MANAGER!

[Russ Allbery]

Thomas Bushnell BSG <[EMAIL PROTECTED]> writes:

> Package: openafs-client
> Version: 1.4.4.dfsg1-1
> Severity: grave
> Justification: renders package unusable

> When I run aklog, I get the following helpful error message:

> ~$ aklog -d
> Authenticating to cell becket.net (server aquinas.becket.net).
> We've deduced that we need to authenticate to realm BECKET.NET.
> Getting tickets: afs/[EMAIL PROTECTED]
> Principal not found, trying alternate service name: [EMAIL PROTECTED]
> Using Kerberos V5 ticket natively
> About to resolve name tb to id in cell becket.net.
> stackcheck = 66051: stack = 66051
> topstack = 0x3007b034: stackptr = 0x3004b008: stacksize = 0x3
> Wed Apr 18 08:35:47 2007 LWP: stack overflow in process IO MANAGER!
> Aborted

> No tokens are reported by the "tokens" program, and klist shows
> a correct [EMAIL PROTECTED] ticket.

> This makes the package unusuable since the purpose of the package is to
> provide AFS file services to users, and without the ability to get
> tokens, it is similarly not possible to use AFS effectively in almost
> every environment.

> Is this related perhaps to the end of linuxthreads in Debian?

It's probably related to the fact that glibc 2.5 is not fully compatible
with glibc 2.3 on any platform, at least for what OpenAFS is doing in its
LWP threading library (which I believe involves using setjmp/longjmp and
probably violating an abstraction boundary in the process).  I already
found that client binaries built with 2.3 won't work with glibc 2.5.  On
x86 and amd64, rebuilding with glibc 2.5 fixes the problem, but apparently
that isn't the case on PowerPC.

AFS is slowly moving towards using pthreads everywhere, but not every
platform supports them and up until recently the LWP code had been more
stable.

I'll forward this upstream; this is an area in which there's been some
current work and investigation, so I'm hoping that someone else will
already have an idea of how to fix it.

-- 
Russ Allbery ([EMAIL PROTECTED])   


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Bug#419903: openafs-client: aklog crashes with LWP: stack overflow in process IO MANAGER!

[Thomas Bushnell BSG]

Package: openafs-client
Version: 1.4.4.dfsg1-1
Severity: grave
Justification: renders package unusable

When I run aklog, I get the following helpful error message:

~$ aklog -d
Authenticating to cell becket.net (server aquinas.becket.net).
We've deduced that we need to authenticate to realm BECKET.NET.
Getting tickets: afs/[EMAIL PROTECTED]
Principal not found, trying alternate service name: [EMAIL PROTECTED]
Using Kerberos V5 ticket natively
About to resolve name tb to id in cell becket.net.
stackcheck = 66051: stack = 66051
topstack = 0x3007b034: stackptr = 0x3004b008: stacksize = 0x3
Wed Apr 18 08:35:47 2007 LWP: stack overflow in process IO MANAGER!
Aborted

No tokens are reported by the "tokens" program, and klist shows
a correct [EMAIL PROTECTED] ticket.

This makes the package unusuable since the purpose of the package is to
provide AFS file services to users, and without the ability to get
tokens, it is similarly not possible to use AFS effectively in almost
every environment.

Is this related perhaps to the end of linuxthreads in Debian?

-- System Information:
Debian Release: lenny/sid
  APT prefers unstable
  APT policy: (500, 'unstable')
Architecture: powerpc (ppc)

Kernel: Linux 2.6.12kempis
Locale: LANG=en_US, LC_CTYPE=en_US (charmap=ISO-8859-1)
Shell: /bin/sh linked to /bin/bash

Versions of packages openafs-client depends on:
ii  debconf [debconf-2.0] 1.5.13 Debian configuration
management sy
ii  libc6 2.5-2  GNU C Library: Shared
libraries
ii  libncurses5   5.5-5  Shared libraries for
terminal hand

Versions of packages openafs-client recommends:
ii  lsof4.77.dfsg.1-3List open files
ii  openafs-modules-2.6.12  1.4.1-2+kempis.6 AFS distributed filesystem
kernel
ii  openafs-modules-source  1.4.4.dfsg1-1AFS distributed filesystem
kernel

-- debconf information:
* openafs-client/fakestat: true
* openafs-client/afsdb: true
* openafs-client/run-client: false
* openafs-client/dynroot: true
* openafs-client/crypt: true
* openafs-client/thiscell: becket.net
* openafs-client/cachesize: 5
* openafs-client/cell-info: aquinas.becket.net



signature.asc
Description: This is a digitally signed message part