Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.

Dismiss

freebsd-sparc-digest V5 #223

0 views

Skip to first unread message

owner-freebsd...@freebsd.org

unread,

Jan 24, 2003, 10:54:20 AM1/24/03

freebsd-sparc-digest Friday, January 24 2003 Volume 05 : Number 223

In this issue:
td_wmesg not cleared leads to panic when doing 'ps ax'
CVSup: nag nag nag
Re: CVSup: nag nag nag
Re: CVSup: nag nag nag
Re: CVSup: nag nag nag
Re: CVSup: nag nag nag
Re: CVSup: nag nag nag
Re: CVSup: nag nag nag
Re: CVSup: nag nag nag
Re: CVSup: nag nag nag
Re: CVSup: nag nag nag
Re: CVSup: nag nag nag
any idea what I'm doing wrong
any idea what I'm doing wrong
any idea what I'm doing wrong
Re: any idea what I'm doing wrong
Sparc64 Fbsd5.0

----------------------------------------------------------------------

Date: Thu, 23 Jan 2003 15:40:52 +0100 (CET)
From: Hartmut Brandt <bra...@fokus.fraunhofer.de>
Subject: td_wmesg not cleared leads to panic when doing 'ps ax'

>Submitter-Id: current-users
>Originator: Hartmut Brandt
>Organization:
>Confidential: no
>Synopsis: td_wmesg not cleared leads to panic when doing 'ps ax'
>Severity: serious
>Priority: high
>Category: kern
>Class: sw-bug
>Release: FreeBSD 5.0-CURRENT sparc64
>Environment:
System: FreeBSD catssrv.fokus.gmd.de 5.0-CURRENT FreeBSD 5.0-CURRENT #17: Thu Jan 23 15:26:01 CET 2003 h...@catssrv.fokus.gmd.de:/opt/obj/usr/src/sys/CATSSRV sparc64

>Description:

The field td_wmesg of struct thread holds a pointer to a constant string
and is set, when a thread calls either msleep or waits on a condition variable.
When the thread continues, this fields is not cleared. When the msleep or
cv_wait was called from a module, that is unloaded, the pointer in td_wmesg
appears to become invalid. When now someone retrieves the process table via
sysctl, then the kernel handler for this sysctl may access the now invalid
pointer. On the SPARC this leads to a panic. i386 probably simply uses an
invalid string (at least I have not seen that panic there).

>How-To-Repeat:

On the sparc:

1. load an interface driver that uses cv_wait or msleep.
2. unload the driver via kldunload
3. ps ax

leads to an immediate crash.

>Fix:

Apply the attached patch. This sets td_wmesg to NULL when an msleep exits
either via the timeout or because of wakeup. It sets also td_wmesg to NULL
when a thread is remove from the wait queue of a condition variable.

Index: sys/kern/kern_condvar.c
===================================================================
RCS file: /home/cvs/freebsd/src/sys/kern/kern_condvar.c,v
retrieving revision 1.35
diff -c -r1.35 kern_condvar.c
*** sys/kern/kern_condvar.c 28 Dec 2002 01:23:07 -0000 1.35
- --- sys/kern/kern_condvar.c 23 Jan 2003 14:31:51 -0000
***************
*** 535,540 ****
- --- 535,541 ----
if ((cvp = td->td_wchan) != NULL && td->td_flags & TDF_CVWAITQ) {
TAILQ_REMOVE(&cvp->cv_waitq, td, td_slpq);
td->td_flags &= ~TDF_CVWAITQ;
+ td->td_wmesg = NULL;
TD_CLR_ON_SLEEPQ(td);
}
}
Index: sys/kern/kern_synch.c
===================================================================
RCS file: /home/cvs/freebsd/src/sys/kern/kern_synch.c,v
retrieving revision 1.209
diff -c -r1.209 kern_synch.c
*** sys/kern/kern_synch.c 28 Dec 2002 01:23:07 -0000 1.209
- --- sys/kern/kern_synch.c 23 Jan 2003 14:31:51 -0000
***************
*** 330,335 ****
- --- 330,336 ----
TAILQ_REMOVE(&slpque[LOOKUP(td->td_wchan)], td, td_slpq);
TD_CLR_ON_SLEEPQ(td);
td->td_flags |= TDF_TIMEOUT;
+ td->td_wmesg = NULL;
} else {
td->td_flags |= TDF_TIMOFAIL;
}
***************
*** 374,379 ****
- --- 375,381 ----
if (TD_ON_SLEEPQ(td)) {
TAILQ_REMOVE(&slpque[LOOKUP(td->td_wchan)], td, td_slpq);
TD_CLR_ON_SLEEPQ(td);
+ td->td_wmesg = NULL;
}
mtx_unlock_spin(&sched_lock);
}

------------------------------

Date: Thu, 23 Jan 2003 16:31:27 -0800 (PST)
From: John Polstra <j...@polstra.com>
Subject: CVSup: nag nag nag

Are any of you testing the new version of CVSup
(cvsup-sparc64-2003-01-19.gz) I put up on January 19? Does it work
reliably now without @M3nopreemption on the command line?

That URL once again: http://people.freebsd.org/~jdp/cvsup-sparc64/

John
- --
John Polstra
John D. Polstra & Co., Inc. Seattle, Washington USA
"Disappointment is a good sign of basic intelligence." -- Chögyam Trungpa

------------------------------

Date: Fri, 24 Jan 2003 02:12:48 +0100
From: Thomas Moestl <tmo...@gmx.net>
Subject: Re: CVSup: nag nag nag

On Thu, 2003/01/23 at 16:31:27 -0800, John Polstra wrote:
> Are any of you testing the new version of CVSup
> (cvsup-sparc64-2003-01-19.gz) I put up on January 19? Does it work
> reliably now without @M3nopreemption on the command line?
>
> That URL once again: http://people.freebsd.org/~jdp/cvsup-sparc64/

I tested it when you first announced the new version by checking out
most of the repo (doc, ports, src, www) from my local mirror, which
worked without a hitch; sorry, should have replied earlier. Thanks
again for making this work!

- Thomas

- --
Thomas Moestl <tmo...@gmx.net> http://www.tu-bs.de/~y0015675/
<t...@FreeBSD.org> http://people.FreeBSD.org/~tmm/
PGP fingerprint: 1C97 A604 2BD0 E492 51D0 9C0F 1FE6 4F1D 419C 776C

------------------------------

Date: Thu, 23 Jan 2003 17:10:10 -0800
From: Will Andrews <wi...@csociety.org>
Subject: Re: CVSup: nag nag nag

On Fri, Jan 24, 2003 at 02:12:48AM +0100, Thomas Moestl wrote:
> I tested it when you first announced the new version by checking out
> most of the repo (doc, ports, src, www) from my local mirror, which
> worked without a hitch; sorry, should have replied earlier. Thanks
> again for making this work!

Works great on my Blade 100 too. With your original version,
John. Haven't tested the newer version. Thanks!

Regards,
- --
wca

------------------------------

Date: Thu, 23 Jan 2003 17:18:44 -0800 (PST)
From: John Polstra <j...@polstra.com>
Subject: Re: CVSup: nag nag nag

In article <20030124011...@procyon.firepipe.net>,
Will Andrews <wi...@csociety.org> wrote:
>
> Works great on my Blade 100 too. With your original version,
> John. Haven't tested the newer version. Thanks!

Ah, but the original version will fail eventually at some random
time, even though it may appear to work OK so far. That's why I'm
nagging folks to get the new version and use it instead. I want to
make sure that my "fix" really fixed it.

John
- --
John Polstra
John D. Polstra & Co., Inc. Seattle, Washington USA
"Disappointment is a good sign of basic intelligence." -- Chögyam Trungpa

------------------------------

Date: Thu, 23 Jan 2003 17:27:53 -0800
From: Will Andrews <wi...@csociety.org>
Subject: Re: CVSup: nag nag nag

On Thu, Jan 23, 2003 at 05:18:44PM -0800, John Polstra wrote:
> Ah, but the original version will fail eventually at some random
> time, even though it may appear to work OK so far. That's why I'm
> nagging folks to get the new version and use it instead. I want to
> make sure that my "fix" really fixed it.

I'll upgrade when you commit the port. ;-)

Regards,
- --
wca

------------------------------

Date: Thu, 23 Jan 2003 17:40:39 -0800 (PST)
From: John Polstra <j...@polstra.com>
Subject: Re: CVSup: nag nag nag

In article <20030124012...@procyon.firepipe.net>,
Will Andrews <wi...@csociety.org> wrote:
>
> I'll upgrade when you commit the port. ;-)

I don't know what that winkie means, but I need it to be tested before
then. The port could be delayed a while due to problems on another
platform (Solaris) that I haven't solved yet.

Please? I put quite a few full weekends into this. I'm just asking
people to test it. I don't have a SPARC machine myself.

John
- --
John Polstra
John D. Polstra & Co., Inc. Seattle, Washington USA
"Disappointment is a good sign of basic intelligence." -- Chögyam Trungpa

------------------------------

Date: Thu, 23 Jan 2003 21:42:06 -0500
From: Garrett Rooney <roo...@electricjellyfish.net>
Subject: Re: CVSup: nag nag nag

On Thursday, January 23, 2003, at 07:31 PM, John Polstra wrote:

> Are any of you testing the new version of CVSup
> (cvsup-sparc64-2003-01-19.gz) I put up on January 19? Does it work
> reliably now without @M3nopreemption on the command line?
>
> That URL once again: http://people.freebsd.org/~jdp/cvsup-sparc64/
>

i was able to cvsup all of src and ports with it on my ultra 10 without
any problems.

- -garrett

------------------------------

Date: Thu, 23 Jan 2003 19:18:25 -0800
From: Will Andrews <wi...@csociety.org>
Subject: Re: CVSup: nag nag nag

On Thu, Jan 23, 2003 at 05:40:39PM -0800, John Polstra wrote:
> I don't know what that winkie means, but I need it to be tested before
> then. The port could be delayed a while due to problems on another
> platform (Solaris) that I haven't solved yet.
>
> Please? I put quite a few full weekends into this. I'm just asking
> people to test it. I don't have a SPARC machine myself.

Since you asked me so nicely, I did your bidding. As expected,
the client works fine.

Regards,
- --
wca

------------------------------

Date: Thu, 23 Jan 2003 23:05:11 -0500
From: Garance A Drosihn <dro...@rpi.edu>
Subject: Re: CVSup: nag nag nag

At 4:31 PM -0800 1/23/03, John Polstra wrote:
>Are any of you testing the new version of CVSup
>(cvsup-sparc64-2003-01-19.gz) I put up on January 19? Does it work
>reliably now without @M3nopreemption on the command line?
>
>That URL once again: http://people.freebsd.org/~jdp/cvsup-sparc64/

I've been using it since the day it came out, seems to be working
fine. And for reasons unrelated to cvsup, I've been doing things
like:
rm -Rf /usr/ports/net
cvsup ~/cvsup/all_ports
and comparing the entire ports directory on my Sparc ultra-10 machine
to the same tree on my dual-Athlon machine. [aside: The dual-athlon
is faster...]. I've just been comparing for the existence of files
and directories though, I haven't been doing an actual diff-type
comparison of the contents of any files. I could do that too.

As long as I'm replying to a question on cvsup, here's a question
that I can't seem to find in the man pages. Is there a way I could
say "do all the work of a normal CVSup run, and tell me what *would*
change, but don't change anything"?

Thanks a lot for working on this. I'd love to see cvsup more widely
used, and on more hardware platforms so it would be more attractive
to the openbsd and netbsd folks. and openafs. and samba. and...

- --
Garance Alistair Drosehn = g...@gilead.netel.rpi.edu
Senior Systems Programmer or g...@freebsd.org
Rensselaer Polytechnic Institute or dro...@rpi.edu

------------------------------

Date: Thu, 23 Jan 2003 22:18:01 -0800 (PST)
From: John Polstra <j...@polstra.com>
Subject: Re: CVSup: nag nag nag

In article <20030124031...@procyon.firepipe.net>,
Will Andrews <wi...@csociety.org> wrote:
>
> Since you asked me so nicely, I did your bidding. As expected,
> the client works fine.

Thanks, much appreciated! If anything goes wrong, please let me
know.

John
- --
John Polstra
John D. Polstra & Co., Inc. Seattle, Washington USA
"Disappointment is a good sign of basic intelligence." -- Chögyam Trungpa

------------------------------

Date: Thu, 23 Jan 2003 22:23:13 -0800 (PST)
From: John Polstra <j...@polstra.com>
Subject: Re: CVSup: nag nag nag

In article <p05200f12ba566a02be1b@[128.113.24.47]>,
Garance A Drosihn <dro...@rpi.edu> wrote:
> I've been using it since the day it came out, seems to be working
> fine. And for reasons unrelated to cvsup, I've been doing things
> like:
> rm -Rf /usr/ports/net
> cvsup ~/cvsup/all_ports
> and comparing the entire ports directory on my Sparc ultra-10 machine
> to the same tree on my dual-Athlon machine. [aside: The dual-athlon
> is faster...]. I've just been comparing for the existence of files
> and directories though, I haven't been doing an actual diff-type
> comparison of the contents of any files. I could do that too.

I'm not too worried about that. I'm mostly interested in hearing
about any unexpected runtime errors or crashes.

> As long as I'm replying to a question on cvsup, here's a question
> that I can't seem to find in the man pages. Is there a way I could
> say "do all the work of a normal CVSup run, and tell me what *would*
> change, but don't change anything"?

Well ... sort of. You can add an extra argument to the command line
which is the name of an empty directory. Whenever CVSup updates a
file, it will put the new version into the appropriate place under
that directory rather than where it really belongs. It's kind of
like DESTDIR with make buildworld. I should warn you, though, that I
have not really been trying to maintain this feature for the past few
years. I'm not absolutely sure it works right any more.

John
- --
John Polstra
John D. Polstra & Co., Inc. Seattle, Washington USA
"Disappointment is a good sign of basic intelligence." -- Chögyam Trungpa

------------------------------

Date: Fri, 24 Jan 2003 09:06:51 -0000
From: "Bri" <br...@ukip.com>
Subject: any idea what I'm doing wrong

lighthouse# make
cc -c -x
assembler-with-cpp -DLOCORE -O -pipe -Wall -Wredundant-decls -Wnested-
externs -Wstrict-prototypes -Wmissing-prototypes -Wpointer-arith -Winline -
Wcas
t-qual -ansi -g -nostdinc -I- -I. -I../../.. -I../../../dev -I../../../con
trib
/dev/acpica -I../../../contrib/ipfilter -D_KERNEL -include
opt_global.h -mcmodel
=medlow -msoft-float -fno-common -ffreestanding -Werror
../../../sparc64/sparc6
4/locore.S
cc -c -x
assembler-with-cpp -DLOCORE -O -pipe -Wall -Wredundant-decls -Wnested-
externs -Wstrict-prototypes -Wmissing-prototypes -Wpointer-arith -Winline -
Wcas
t-qual -ansi -g -nostdinc -I- -I. -I../../.. -I../../../dev -I../../../con
trib
/dev/acpica -I../../../contrib/ipfilter -D_KERNEL -include
opt_global.h -mcmodel
=medlow -msoft-float -fno-common -ffreestanding -Werror
../../../sparc64/sparc6
4/exception.S
../../../sparc64/sparc64/exception.S: Assembler messages:
../../../sparc64/sparc64/exception.S:2181: Error: Architecture mismatch on
"rd".
../../../sparc64/sparc64/exception.S:2181: (Requires v9a|v9b; requested
archite
cture is v9.)
../../../sparc64/sparc64/exception.S:2248: Error: Architecture mismatch on
"wr".
../../../sparc64/sparc64/exception.S:2248: (Requires v9a|v9b; requested
archite
cture is v9.)
../../../sparc64/sparc64/exception.S:2268: Error: Architecture mismatch on
"rd".
../../../sparc64/sparc64/exception.S:2268: (Requires v9a|v9b; requested
archite
cture is v9.)
../../../sparc64/sparc64/exception.S:2448: Error: Architecture mismatch on
"wr".
../../../sparc64/sparc64/exception.S:2448: (Requires v9a|v9b; requested
archite
cture is v9.)
../../../sparc64/sparc64/exception.S:2771: Error: Architecture mismatch on
"wr".
../../../sparc64/sparc64/exception.S:2771: (Requires v9a|v9b; requested
archite
cture is v9.)
*** Error code 1

Stop in /builds/CURRENT/src/sys/sparc64/compile/GENERIC.

the make depend went fine but the make breaks I originally try compiling my
custom kernel but I get the same.

these are also brand new sources.

lighthouse# uname -a
FreeBSD lighthouse.shacknet.nu 5.0-20021031-SNAP FreeBSD 5.0-20021031-SNAP
#0: Fri Nov 1 01:06:53 GMT 2002
ro...@u60.locore.ca:/usr/obj/usr/src/sys/GENERIC sparc64

Bri,

------------------------------

Date: Fri, 24 Jan 2003 09:46:55 -0000
From: "Bri" <br...@ukip.com>
Subject: any idea what I'm doing wrong

Stop in /builds/CURRENT/src/sys/sparc64/compile/GENERIC.

the make depend went fine but the make breaks I originally try compiling my
custom kernel but I get the same.

these are also brand new sources.

lighthouse# uname -a
FreeBSD lighthouse.shacknet.nu 5.0-20021031-SNAP FreeBSD 5.0-20021031-SNAP
#0: Fri Nov 1 01:06:53 GMT 2002
ro...@u60.locore.ca:/usr/obj/usr/src/sys/GENERIC sparc64

Bri,

------------------------------

Date: Fri, 24 Jan 2003 11:09:22 -0000
From: "Bri" <br...@ukip.com>
Subject: any idea what I'm doing wrong

Stop in /builds/CURRENT/src/sys/sparc64/compile/GENERIC.

the make depend went fine but the make breaks I originally try compiling my
custom kernel but I get the same.

these are also brand new sources.

lighthouse# uname -a
FreeBSD lighthouse.shacknet.nu 5.0-20021031-SNAP FreeBSD 5.0-20021031-SNAP
#0: Fri Nov 1 01:06:53 GMT 2002
ro...@u60.locore.ca:/usr/obj/usr/src/sys/GENERIC sparc64

Bri,

------------------------------

Date: Fri, 24 Jan 2003 12:47:15 +0100
From: Thomas Moestl <tmo...@gmx.net>
Subject: Re: any idea what I'm doing wrong

On Fri, 2003/01/24 at 09:06:51 -0000, Bri wrote:
> ./../../sparc64/sparc64/exception.S: Assembler messages:
> ./../../sparc64/sparc64/exception.S:2181: Error: Architecture mismatch on
> "rd".
> ./../../sparc64/sparc64/exception.S:2181: (Requires v9a|v9b; requested
> archite
> cture is v9.)
> ./../../sparc64/sparc64/exception.S:2248: Error: Architecture mismatch on
> "wr".
> ./../../sparc64/sparc64/exception.S:2248: (Requires v9a|v9b; requested
> archite
> cture is v9.)
> ./../../sparc64/sparc64/exception.S:2268: Error: Architecture mismatch on
> "rd".
> ./../../sparc64/sparc64/exception.S:2268: (Requires v9a|v9b; requested
> archite
> cture is v9.)
> ./../../sparc64/sparc64/exception.S:2448: Error: Architecture mismatch on
> "wr".
> ./../../sparc64/sparc64/exception.S:2448: (Requires v9a|v9b; requested
> archite
> cture is v9.)
> ./../../sparc64/sparc64/exception.S:2771: Error: Architecture mismatch on
> "wr".
> ./../../sparc64/sparc64/exception.S:2771: (Requires v9a|v9b; requested
> archite
> cture is v9.)
> *** Error code 1

Looks like you are trying to compile a recent kernel with a quite old
gcc. The gcc build was not-so-recently changed so that the correct
flags will be passed to the assembler. You will need to update your
world.

- Thomas

- --
Thomas Moestl <tmo...@gmx.net> http://www.tu-bs.de/~y0015675/
<t...@FreeBSD.org> http://people.FreeBSD.org/~tmm/
PGP fingerprint: 1C97 A604 2BD0 E492 51D0 9C0F 1FE6 4F1D 419C 776C

------------------------------

Date: Fri, 24 Jan 2003 09:54:10 -0600
From: "Eric Humphries" <e...@bitstream.net>
Subject: Sparc64 Fbsd5.0

I recently purchased a Sun Ultra 30. I would like to run FreeBSD 5.0 on
it, but after I select which terminal type I want and proceed to the
next terminal screen, my keyboard doesn't appear to be functioning
properly.

Keep in mind I've tried all 4 or 5 terminal types, some display just
find, but others severely mangle the install menu. Regardless of which
one I choose, I'm not able to arrow around or select anything other than
install notes and exit for some reason. I've also tried installing via
serial console, but the same happens. Any ideas / suggestions would be
greatly appreciated.

Thank You,
Eric Humphries

------------------------------

End of freebsd-sparc-digest V5 #223
***********************************

To Unsubscribe: send mail to majo...@FreeBSD.org
with unsubscribe freebsd-sparc-digest in the body of the message

0 new messages