Xen devel - Feb 2004 - Kernel panic while compiling kernel

I know you Xen developers are beginning to hate me ;) but...

While trying to compile 2.4.24 under DOM0, quickly after issuing ''make
dep'' I
got:

Kernel panic: Failed mmu update: c015bf80, 4

I suspect this address probably isn''t to helpful but this is all I
have. I am
currently booted in 2.4.21-SuSE and compiling the 2.4.24 regarding another 
thread here.

Regarding server hardware: I have this server (HP tc2120) for a few months now 
and I have never experienced such stability problems with it under different 
kernels (2.4.34, 2.4.24, 2.6.0, 2.6.1, 2.6.2). So I think this is Xen 
related.


-------------------------------------------------------
The SF.Net email is sponsored by EclipseCon 2004
Premiere Conference on Open Tools Development and Integration
See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
http://www.eclipsecon.org/osdn
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

Hmmm... well we know there are bugs in Xen v1.3, but v1.2 has been
stable for a while now (it passes our regression tests).

If you can, turn on debugging in Xen (edit arch/i386/Rules.mk and
remove -DNDEBUG from CFLAGS), re-compile, and send us the full serial
output for a Xen/Xenolinxu session which includes this crash.

 -- Keir
> 
> I know you Xen developers are beginning to hate me ;) but...
> 
> While trying to compile 2.4.24 under DOM0, quickly after issuing
''make dep'' I
> got:
> 
> Kernel panic: Failed mmu update: c015bf80, 4
> 
> I suspect this address probably isn''t to helpful but this is all I
have. I am
> currently booted in 2.4.21-SuSE and compiling the 2.4.24 regarding another 
> thread here.
> 
> Regarding server hardware: I have this server (HP tc2120) for a few months
now
> and I have never experienced such stability problems with it under
different
> kernels (2.4.34, 2.4.24, 2.6.0, 2.6.1, 2.6.2). So I think this is Xen 
> related.
> 
> 
> -------------------------------------------------------
> The SF.Net email is sponsored by EclipseCon 2004
> Premiere Conference on Open Tools Development and Integration
> See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
> http://www.eclipsecon.org/osdn
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/xen-devel


-------------------------------------------------------
The SF.Net email is sponsored by EclipseCon 2004
Premiere Conference on Open Tools Development and Integration
See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
http://www.eclipsecon.org/osdn
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

On Wednesday 11 February 2004 12:22, Keir Fraser wrote:
> Hmmm... well we know there are bugs in Xen v1.3, but v1.2 has been
> stable for a while now (it passes our regression tests).
>
> If you can, turn on debugging in Xen (edit arch/i386/Rules.mk and
> remove -DNDEBUG from CFLAGS), re-compile, and send us the full serial
> output for a Xen/Xenolinxu session which includes this crash.
Something is really bad, after sending you the report of working tg3 witd Xen 
DEBUG I planned to put some load on the DOM0 to see if I can reproduce the 
problem from this thread. While I was trying to ssh to DOM0 I got the same 
kernel panic. Fortunately I had an minicom session active :)

The complete log is the same as one I have sent you in tg3 thread. Just append 
this:

DOM0: Kernel panic: Failed mmu update: c015bf80, 4
Domain 0 killed: rebooting machine!



-------------------------------------------------------
The SF.Net email is sponsored by EclipseCon 2004
Premiere Conference on Open Tools Development and Integration
See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
http://www.eclipsecon.org/osdn
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

On Wednesday 11 February 2004 12:22, Keir Fraser wrote:
> Hmmm... well we know there are bugs in Xen v1.3, but v1.2 has been
> stable for a while now (it passes our regression tests).
>
> If you can, turn on debugging in Xen (edit arch/i386/Rules.mk and
> remove -DNDEBUG from CFLAGS), re-compile, and send us the full serial
> output for a Xen/Xenolinxu session which includes this crash.
Here come the latest update:

- I have compiled Xen + XenoLinux with GCC 2.95.3
- tg3 works (although debug is still on so this one is not definitive)
- after starting DOM1, it starts booting and then panics with mmu error I was 
getting yesterday in DOM0 (although different address)
- it restarts and restarts until it manages to survive the boot process
- domain-1.log is the complete log until domain is successfully booted 
- then I torture it a little, some from DOM0 some from DOMx
- after few disk i/o and cpu tests everything is working stable ( find / -exec 
ls -l {} \; , bzip2-c /dev/zero >/dev/null )
- then I started yast2 runlevel editor (gave me 100% panic probability 
yesterday) - the same, panic
- domain-2.log is the log starting with panic caused by runlevel editor 
followed by one unsuccessful boot (again)
- then I tried to run runlevel editor in DOM0:
DOM0: Kernel panic: Failed mmu update: c013eea8, 4
Domain 0 killed: rebooting machine!

So, I was unable to find the exact load pattern which causes the crash. But if 
there is any test you want me to do I will be glad to do it.

> getting yesterday in DOM0 (although different address)
> - it restarts and restarts until it manages to survive the boot process
> - domain-1.log is the complete log until domain is successfully booted 
> - then I torture it a little, some from DOM0 some from DOMx
> - after few disk i/o and cpu tests everything is working stable ( find /
-exec
> ls -l {} \; , bzip2-c /dev/zero >/dev/null )
> - then I started yast2 runlevel editor (gave me 100% panic probability 
> yesterday) - the same, panic
> - domain-2.log is the log starting with panic caused by runlevel editor 
> followed by one unsuccessful boot (again)
> - then I tried to run runlevel editor in DOM0:
> DOM0: Kernel panic: Failed mmu update: c013eea8, 4
> Domain 0 killed: rebooting machine!
> 
> So, I was unable to find the exact load pattern which causes the crash. But
if
> there is any test you want me to do I will be glad to do it.
Whoops. I forgot that in Xen 1.2, to get useful memory debugging you
will need to edit common/memory.c and enable MEM_LOG() (change ''#if
0''
to ''#if 1'').

I''ll try to find some time to do a bit of testing myself today.

 -- Keir


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

On Thursday 12 February 2004 10:44, Keir Fraser wrote:
> Whoops. I forgot that in Xen 1.2, to get useful memory debugging you
> will need to edit common/memory.c and enable MEM_LOG() (change
''#if 0''
> to ''#if 1'').
There were two compilation issues in that file after enablig that, I 
quick-fixed id by adding two dummy parameteres (and formats) to the two 
offending lines. Those were ones in form of MSG_LOG("text and no
parms");
> I''ll try to find some time to do a bit of testing myself today.
If it saves you some testing time here is the log with MSG_LOG turned on, 
runlevel editor did the trick in DOM0 (didn''t start other domain this
time).

> On Thursday 12 February 2004 10:44, Keir Fraser wrote:
> > Whoops. I forgot that in Xen 1.2, to get useful memory debugging you
> > will need to edit common/memory.c and enable MEM_LOG() (change
''#if 0''
> > to ''#if 1'').
> 
> There were two compilation issues in that file after enablig that, I 
> quick-fixed id by adding two dummy parameteres (and formats) to the two 
> offending lines. Those were ones in form of MSG_LOG("text and no
parms");
> 
> > I''ll try to find some time to do a bit of testing myself
today.
> 
> If it saves you some testing time here is the log with MSG_LOG turned on, 
> runlevel editor did the trick in DOM0 (didn''t start other domain
this time).
Okay, that could be useful.

Perhaps you could add the follwoing (I assume you knwo a bit of C):
 1. Extend the "Pfn already pinned" message to include every field in
 the page structure (e.g., page->flags, page->type_count, and so on).

 2. At the start of the switch case MMUEXT_PIN_L2_TABLE, add
    printk(" ********* L2 pin pfn == %08lx\n", pfn);

 3. At teh start o fthe switch case MMUEXT_UNPIN_TABLE, add
    if ( (page->flags & PG_type_mask) == PGT_l2_page_table )
      printk(" ********* L2 unpin pfn == %08lx\n", pfn);

This should let us see whether this is a Xenolinux or a Xen problem.

 -- Keir


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

On Thursday 12 February 2004 11:26, Keir Fraser wrote:
> Perhaps you could add the follwoing (I assume you knwo a bit of C):
No problem at all. Yes, I am pretty fluent in C (did some kernel patches and 
wrote some drivers for custom hw).
> This should let us see whether this is a Xenolinux or a Xen problem.
Here is is. All after this line:
DOM0: Skipped services in runlevel 3: [80C[12D[1;33macpid[m
Is caused by starting ''yast2 runlevel''.

This part of the trace indicates that it''s a Xenolinux problem --
it''s
trying to pin the same page multiple times (a bad thing). Hmmm...

An alternative is that Xen has misparsed the comamnds coming down from
Xenolinux. 

If you use our pre-built Xen/Xenolinux (a link near the bottom of
http://www.cl.cam.ac.uk/Research/SRG/netos/xen/downloads.html), can
you still reproduce the error?

 -- Keir
>  ********* L2 pin pfn == 00007baf
>  ********* L2 pin pfn == 00007b6a
>  ********* L2 pin pfn == 00006f27
>  ********* L2 pin pfn == 00006f27
> DOM0: (file=memory.c, line=769) Pfn 00006f27 already pinned (L2);
flags:22000000 totcnt:3 typecnt:2
> DOM0: Kernel panic: Failed mmu update: c013eea8, 4
>  ********* L2 pin pfn == 00006f27
> DOM0: (file=memory.c, line=769) Pfn 00006f27 already pinned (L2);
flags:22000000 totcnt:3 typecnt:2


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

On Thursday 12 February 2004 12:06, Keir Fraser wrote:
> This part of the trace indicates that it''s a Xenolinux problem --
it''s
> trying to pin the same page multiple times (a bad thing). Hmmm...
>
> An alternative is that Xen has misparsed the comamnds coming down from
> Xenolinux.
>
> If you use our pre-built Xen/Xenolinux (a link near the bottom of
> http://www.cl.cam.ac.uk/Research/SRG/netos/xen/downloads.html), can
> you still reproduce the error?
I will try but it is one big download for my company''s 256kbit link
which I
must not hog. :( Can I download just xen.gz and xenolinux.gz from that 
package somewhere?


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

On Thursday 12 February 2004 12:06, Keir Fraser wrote:
> If you use our pre-built Xen/Xenolinux (a link near the bottom of
> http://www.cl.cam.ac.uk/Research/SRG/netos/xen/downloads.html), can
> you still reproduce the error?
I''m afraid I can. No debug of course, but again:

DOM0: Kernel panic: Failed mmu update: c01ae080, 4
Domain 0 killed: rebooting machine!


Versions:

 Xen version 1.2 (root@cl.cam.ac.uk) (gcc version 3.2.2 20030222 (Red Hat 
Linux 3.2.2-5)) Mon Feb 9 19:34:26 GMT 2004

DOM0: Linux version 2.4.24-xeno (root@labyrinth.cl.cam.ac.uk) (gcc version 
3.2.2 20030222 (Red Hat Linux 3.2.2-5)) #27 Mon Feb 9 19:29:51 GMT 2004


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

> On Thursday 12 February 2004 12:06, Keir Fraser wrote:
> 
> > If you use our pre-built Xen/Xenolinux (a link near the bottom of
> > http://www.cl.cam.ac.uk/Research/SRG/netos/xen/downloads.html), can
> > you still reproduce the error?
> 
> I''m afraid I can. No debug of course, but again:
> 
> DOM0: Kernel panic: Failed mmu update: c01ae080, 4
> Domain 0 killed: rebooting machine!
> 
> 
> Versions:
> 
>  Xen version 1.2 (root@cl.cam.ac.uk) (gcc version 3.2.2 20030222 (Red Hat 
> Linux 3.2.2-5)) Mon Feb 9 19:34:26 GMT 2004
> 
> DOM0: Linux version 2.4.24-xeno (root@labyrinth.cl.cam.ac.uk) (gcc version 
> 3.2.2 20030222 (Red Hat Linux 3.2.2-5)) #27 Mon Feb 9 19:29:51 GMT 2004
Okay, then out of interest can you trigger the same problem with
Xen/Xenolinux v1.3?

 http://www.cl.cam.ac.uk/~kaf24/xen.gz
 http://www.cl.cam.ac.uk/~kaf24/xenolinux.gz

 -- Keir


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

On Thursday 12 February 2004 12:44, Keir Fraser wrote:
> Okay, then out of interest can you trigger the same problem with
> Xen/Xenolinux v1.3?
Yep! :-(

> On Thursday 12 February 2004 12:44, Keir Fraser wrote:
> 
> > Okay, then out of interest can you trigger the same problem with
> > Xen/Xenolinux v1.3?
> 
> Yep! :-(
Okay, I think I have a fix. Please try:

 http://www.cl.cam.ac.uk/~kaf24/xen.gz
 http://www.cl.cam.ac.uk/~kaf24/xenolinux.gz

 -- Keir


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

On Thursday 12 February 2004 15:07, Keir Fraser wrote:
> Okay, I think I have a fix. Please try:
The same, log attached.

> On Thursday 12 February 2004 15:07, Keir Fraser wrote:
> 
> > Okay, I think I have a fix. Please try:
> 
> The same, log attached.
Okay, try this Xen (use the same Xenolinux as before):

 http://www.cl.cam.ac.uk/~kaf24/xen.gz

It will crash just teh same, but should print a big stack dump from
Xenolinux. Hopefully I can then reconstruct a backtrace.

I think I''m close to the bug. :-)

 -- Keir


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

On Thursday 12 February 2004 15:42, Keir Fraser wrote:
> I think I''m close to the bug. :-)
Great, I hope you really are! When finished, can it be backported to 1.2 or is 
it safe to use unstable (aka 1.3-devel) for production?

Log attached:

> On Thursday 12 February 2004 15:42, Keir Fraser wrote:
> 
> > I think I''m close to the bug. :-)
> 
> Great, I hope you really are! When finished, can it be backported to 1.2 or
is
> it safe to use unstable (aka 1.3-devel) for production?
Okay, please try again with:
 http://www.cl.cam.ac.uk/~kaf24/xen.gz
 http://www.cl.cam.ac.uk/~kaf24/xenolinux.gz

This should do the trick. :-)

 -- Keir


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

On Thursday 12 February 2004 18:32, Keir Fraser wrote:
> Okay, please try again with:
>  http://www.cl.cam.ac.uk/~kaf24/xen.gz
>  http://www.cl.cam.ac.uk/~kaf24/xenolinux.gz
>
> This should do the trick. :-)
Indeed it does! There are no ''interesting'' logs to report this
time. :)



-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

On Thursday 12 February 2004 18:32, Keir Fraser wrote:
> This should do the trick. :-)
I see 1.2 tools are not working with Xen-1.3, could you send me the patch for 
1.2 or just let me know when can i bk pull-it? I appreciate it!

From pure curiosity, what do you suspect is different on my server that 
triggered that bug? Chipset (ServerWorks GCNB-LE)?
 


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id56&alloc_id438&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

> From pure curiosity, what do you suspect is different on my server that 
> triggered that bug? Chipset (ServerWorks GCNB-LE)?
Nope, libc/libpthread version.

Please can you tell us a bit about the origins of your root file
system, and send us the output of:

 ldconfig -p | egrep ''pthread|libc.so''
 cat /proc/cpuinfo 
 

Thanks,
Ian


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

On Friday 13 February 2004 09:50, Ian Pratt wrote:
> > From pure curiosity, what do you suspect is different on my server
that
> > triggered that bug? Chipset (ServerWorks GCNB-LE)?
>
> Nope, libc/libpthread version.
>
> Please can you tell us a bit about the origins of your root file
> system, and send us the output of:
Of course, I am glad I can help!  

Currently it is a minimal SuSE 9.0 installation, so nothing fancy about that.


xen:~ # ldconfig -p | egrep ''pthread|libc.so''
        libpthread.so.0 (libc6, hwcap: 0x8000000000000, OS ABI: Linux 2.4.1) 
=> /lib/i686/libpthread.so.0
        libpthread.so.0 (libc6, OS ABI: Linux 2.2.5) => /lib/libpthread.so.0
        libc.so.6 (libc6, hwcap: 0x8000000000000, OS ABI: Linux 2.4.1) 
=> /lib/i686/libc.so.6
        libc.so.6 (libc6, OS ABI: Linux 2.2.5) => /lib/libc.so.6

xen:~ # cat /proc/cpuinfo
processor       : 0
vendor_id       : GenuineIntel
cpu family      : 15
model           : 2
model name      : Intel(R) Pentium(R) 4 CPU 2.66GHz
stepping        : 9
cpu MHz         : 2666.051
cache size      : 512 KB
fdiv_bug        : no
hlt_bug         : no
f00f_bug        : no
coma_bug        : no
fpu             : yes
fpu_exception   : yes
cpuid level     : 2
wp              : yes
flags           : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca 
cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe cid
bogomips        : 26633.83

xen:~ # rpm -qa | grep libc
glibc-2.3.2-88


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

> On Thursday 12 February 2004 18:32, Keir Fraser wrote:
> > This should do the trick. :-)
> 
> I see 1.2 tools are not working with Xen-1.3, could you send me the patch
for
> 1.2 or just let me know when can i bk pull-it? I appreciate it!
The fix is now checked in to both v1.2 and v1.3 repositories on
bkbits.net. 

 -- Keir


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

On Friday 13 February 2004 10:08, Keir Fraser wrote:
> The fix is now checked in to both v1.2 and v1.3 repositories on
> bkbits.net.
More bad news :(

''yast2 runlevel'' test case is fixed, but there is still that
other one which
panics just after boot (somewhere in init scripts I presume).

It results in:

DOM1: [m[?25hMaster Resource Control: runlevel 3 has been [80C[10D[1mreached[m
DOM1: INIT: no more processes left in this runlevel
DOM1: Kernel panic: Failed mmu update: c0140ea0, 1
Killing domain 1
Releasing task 1

It completely booted after 11 tries:

DOM11: INIT: no more processes left in this runlevel

Then I tried to SSH to it:

DOM11: Kernel panic: Failed mmu update: c0140ea0, 1
Killing domain 11
Releasing task 11

What debugging options should I try this time? :)
If possible for 1.2 because I suspect those session with 1.3-devel somehow 
trashed my VD root partition. If that makes any sense...




-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

> What debugging options should I try this time? :)
> If possible for 1.2 because I suspect those session with 1.3-devel somehow 
> trashed my VD root partition. If that makes any sense...
Okay, try:
 http://www.cl.cam.ac.uk/~kaf24/xen.gz
 http://www.cl.cam.ac.uk/~kaf24/xenolinux.gz

This should crash again, but will get us some useful debugging out of
Xen (including a dump of the Xenolinux stack). Please send me the
complete output trace for a crash with these image files.

 -- Keir


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

On Friday 13 February 2004 12:10, Keir Fraser wrote:
> This should crash again, but will get us some useful debugging out of
> Xen (including a dump of the Xenolinux stack). Please send me the
> complete output trace for a crash with these image files.
Here it goes:

> On Friday 13 February 2004 12:10, Keir Fraser wrote:
> 
> > This should crash again, but will get us some useful debugging out of
> > Xen (including a dump of the Xenolinux stack). Please send me the
> > complete output trace for a crash with these image files.
> 
> Here it goes:
You used a different kernel image for DOM1. Be careful to use the
correct image when setting up guest OSes from within DOM0!

If you use the correct image for DOM1 (i.e., one of the first lines
from DOM1 should have a timestamp saying 11:04) then can you still
cause a crash?

 -- Keir


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

On Friday 13 February 2004 12:36, Keir Fraser wrote:
> You used a different kernel image for DOM1. Be careful to use the
> correct image when setting up guest OSes from within DOM0!
Damn, of course! :) It is not recommended to work on several things at the 
same time or something like this is bound to happen. I apologize for this.
> If you use the correct image for DOM1 (i.e., one of the first lines
> from DOM1 should have a timestamp saying 11:04) then can you still
> cause a crash?
Not yet :) Just kidding, everything seems to be fine now.


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

On Friday 13 February 2004 12:42, Tvrtko A. Uršulin wrote:
> > If you use the correct image for DOM1 (i.e., one of the first lines
> > from DOM1 should have a timestamp saying 11:04) then can you still
> > cause a crash?
>
> Not yet :) Just kidding, everything seems to be fine now.
To reply to my own post, with your version (11:04) everything is fine. But my 
1.2 (bk pulled this morning) compiled with gcc 2.95.3 is still causing kernel 
panic.

I must leave work now and don''t think will be back until Monday. 

Thank you once more for this debugging sessions!

Best regards,
Tvrtko A. Ursulin


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id56&alloc_id438&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

> On Friday 13 February 2004 12:42, Tvrtko A. Ur¹ulin wrote:
> 
> > > If you use the correct image for DOM1 (i.e., one of the first
lines
> > > from DOM1 should have a timestamp saying 11:04) then can you
still
> > > cause a crash?
> >
> > Not yet :) Just kidding, everything seems to be fine now.
> 
> To reply to my own post, with your version (11:04) everything is fine. But
my
> 1.2 (bk pulled this morning) compiled with gcc 2.95.3 is still causing
kernel
> panic.
Pull the very latest 1.2 (Changeset 1.720). Build Xen/Xenolinux from
scratch with 2.95.3. Make sure you use _your_ image files for booting
Xen, DOM0, and any other domains that you create. Also make sure that
you use the latest xc tools (reinstall them from the latest
repository) -- it shouldn''t make a difference (interfaces
haven''t
changed in v1.2 for quite a while), but better safe than sorry. 

If you can still cause a crash then let me know. :-)

 -- Keir


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

On Friday 13 February 2004 15:01, Keir Fraser wrote:
> If you can still cause a crash then let me know. :-)
After solving the xc_dom_create.py issue in another thread, additional domain 
was created. But now we return to the old problem. The panic after init 
scripts is still there. 

Logs are attached:

P.S. Ignore those Killing/Releasing 1...2...3 lines, those are from buggy Xc.c

> On Friday 13 February 2004 15:01, Keir Fraser wrote:
> 
> > If you can still cause a crash then let me know. :-)
> 
> After solving the xc_dom_create.py issue in another thread, additional
domain
> was created. But now we return to the old problem. The panic after init 
> scripts is still there. 
> 
> Logs are attached:
> 
> P.S. Ignore those Killing/Releasing 1...2...3 lines, those are from buggy
Xc.c
It seems that SuSE is the kiss of death for Xen at the moment. :-(

Okay, try:
 http://www.cl.cam.ac.uk/~kaf24/xen.gz
 http://www.cl.cam.ac.uk/~kaf24/xenolinux.gz

If it crashes then we should get a stack dump.

 -- Keir


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

> It seems that SuSE is the kiss of death for Xen at the moment. :-(
>
> Okay, try:
>  http://www.cl.cam.ac.uk/~kaf24/xen.gz
>  http://www.cl.cam.ac.uk/~kaf24/xenolinux.gz
>
> If it crashes then we should get a stack dump.
Hmm.. I cannot reproduce it with your image, strange... 

I just recompiled Xen and XenoLinux with gcc 3.3.1 and now even DOM0 panics in 
init:
DOM0: INIT: version 2.82 booting
DOM0: Kernel panic: Failed mmu update: c7c7ded8, 1
Domain 0 killed: rebooting machine!

So gcc 2.95.3 yields somehow better results than 3.3.1.

Do you have some ideas what is happening here?


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

> 
> > It seems that SuSE is the kiss of death for Xen at the moment. :-(
> >
> > Okay, try:
> >  http://www.cl.cam.ac.uk/~kaf24/xen.gz
> >  http://www.cl.cam.ac.uk/~kaf24/xenolinux.gz
> >
> > If it crashes then we should get a stack dump.
> 
> Hmm.. I cannot reproduce it with your image, strange... 
> 
> I just recompiled Xen and XenoLinux with gcc 3.3.1 and now even DOM0 panics
in
> init:
> DOM0: INIT: version 2.82 booting
> DOM0: Kernel panic: Failed mmu update: c7c7ded8, 1
> Domain 0 killed: rebooting machine!
> 
> So gcc 2.95.3 yields somehow better results than 3.3.1.
> 
> Do you have some ideas what is happening here?
Not really. I''ll try out a few GCC versions here and see if I can
reproduce the behaviour that you''re seeing.

Something pretty odd is going on here -- there are many people using
Xen successfully without the large number of problems you''ve been
experiencing! 

 -- Keir


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

> 
> > It seems that SuSE is the kiss of death for Xen at the moment. :-(
> >
> > Okay, try:
> >  http://www.cl.cam.ac.uk/~kaf24/xen.gz
> >  http://www.cl.cam.ac.uk/~kaf24/xenolinux.gz
> >
> > If it crashes then we should get a stack dump.
> 
> Hmm.. I cannot reproduce it with your image, strange... 
> 
> I just recompiled Xen and XenoLinux with gcc 3.3.1 and now even DOM0 panics
in
> init:
> DOM0: INIT: version 2.82 booting
> DOM0: Kernel panic: Failed mmu update: c7c7ded8, 1
> Domain 0 killed: rebooting machine!
> 
> So gcc 2.95.3 yields somehow better results than 3.3.1.
> 
> Do you have some ideas what is happening here?
It does seem that your builds have a much higher failure rate than
usual. You are doing a completely clean build in all cases, with no
modifications from teh default configurations?

 -- Keir


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

On Monday 16 February 2004 14:59, Keir Fraser wrote:
> It does seem that your builds have a much higher failure rate than
> usual. You are doing a completely clean build in all cases, with no
> modifications from teh default configurations?
Erm O:-) not quite, but I didn''t realize it is not safe to turn some
options
on (like modular netfilter, QoS, ...)
I must say that when I recompiled with make defconfig (that is oldconfig 
without .config, only added reiserfs), mmu update panic is gone.

I have attached my .config in case you are interested.

> On Monday 16 February 2004 14:59, Keir Fraser wrote:
> 
> > It does seem that your builds have a much higher failure rate than
> > usual. You are doing a completely clean build in all cases, with no
> > modifications from teh default configurations?
> 
> Erm O:-) not quite, but I didn''t realize it is not safe to turn
some options
> on (like modular netfilter, QoS, ...)
> I must say that when I recompiled with make defconfig (that is oldconfig 
> without .config, only added reiserfs), mmu update panic is gone.
> CONFIG_MPENTIUM4=y
> CONFIG_X86_L1_CACHE_SHIFT=7
You''ve probably been bitten by one of these CPU config options. All
your other config changes should be okay (shouldn''t break anything).
In particular, I expect that we haven''t done testing with any other
cache-line size than 32 bytes.

I''ll look into this... it''s something I hadn''t
considered. :-)

*BUT* take it as a reminder to list everything non-standard about your
build and environment when reporting a bug.

 -- Keir

PS. Can you try building your own Xenolinux with all your config
changes _except_ the ones listed above? i.e. you should have
CONFIG_M686=y and CONFIG_X86_L1_CACHE_SHIFT=5. I expect that the
resulting kernel should not crash.


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

> *BUT* take it as a reminder to list everything non-standard about your
> build and environment when reporting a bug.
Ok, I understand.
> PS. Can you try building your own Xenolinux with all your config
> changes _except_ the ones listed above? i.e. you should have
> CONFIG_M686=y and CONFIG_X86_L1_CACHE_SHIFT=5. I expect that the
> resulting kernel should not crash.
It does crash. :)


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

> 
> > *BUT* take it as a reminder to list everything non-standard about your
> > build and environment when reporting a bug.
> 
> Ok, I understand.
> 
> > PS. Can you try building your own Xenolinux with all your config
> > changes _except_ the ones listed above? i.e. you should have
> > CONFIG_M686=y and CONFIG_X86_L1_CACHE_SHIFT=5. I expect that the
> > resulting kernel should not crash.
> 
> It does crash. :)
If you''re sure that the CPU config options were set to the default,
and that you booted the correct kernel, then a useful thing to do is
the following:
Incrementally remove your config options (i.e. move towards the
default Xenolinux configuration) until your kernel image no longer
crashes. 
Or alternatively:
Starting from teh default Xenolinux configuration, incrementally add
your desired extra config options until the resulting kernel image
starts to crash.

In this way we can pinpoint which kernel config option is causing
Xenolinux to fail.

 -- Keir


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

> 
> > *BUT* take it as a reminder to list everything non-standard about your
> > build and environment when reporting a bug.
> 
> Ok, I understand.
> 
> > PS. Can you try building your own Xenolinux with all your config
> > changes _except_ the ones listed above? i.e. you should have
> > CONFIG_M686=y and CONFIG_X86_L1_CACHE_SHIFT=5. I expect that the
> > resulting kernel should not crash.
> 
> It does crash. :)
I''ve built a kernel with your exact configuration and it ran just
fine. So I''m rather stuck as to what your problem might be...

I have checked in some fixes to prevent GCC optimising code
surrounding calls from Xenolinux into Xen. This might help.

 -- Keir



-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

Just an observation - I''ve noticed that several of the include files 
include linux/config.h and some include xeno/config.h - also for some 
other included files this is the case.  Also,  -nostdinc only seems to 
be applied to the arch/i386 branch - which means the drivers may be 
getting a different set of include paths.  Perhaps some of your config 
settings are different in different .c files as a result with 
non-standard config options.  Although I kind of doubt the problem is 
this simple, it seems reasonable to clean up the build tree so that it 
has a well isolated include structure.  I think I just volunteered 
myself unless anyone else is already working on it.

Zachary Amsden
zamsden@cisco.com

Keir Fraser wrote:
>>>*BUT* take it as a reminder to list everything non-standard about
your
>>>build and environment when reporting a bug.
>>>      
>>>
>>Ok, I understand.
>>
>>    
>>
>>>PS. Can you try building your own Xenolinux with all your config
>>>changes _except_ the ones listed above? i.e. you should have
>>>CONFIG_M686=y and CONFIG_X86_L1_CACHE_SHIFT=5. I expect that the
>>>resulting kernel should not crash.
>>>      
>>>
>>It does crash. :)
>>    
>>
>
>If you''re sure that the CPU config options were set to the default,
>and that you booted the correct kernel, then a useful thing to do is
>the following:
>Incrementally remove your config options (i.e. move towards the
>default Xenolinux configuration) until your kernel image no longer
>crashes. 
>Or alternatively:
>Starting from teh default Xenolinux configuration, incrementally add
>your desired extra config options until the resulting kernel image
>starts to crash.
>
>In this way we can pinpoint which kernel config option is causing
>Xenolinux to fail.
>
> -- Keir
>
>
>-------------------------------------------------------
>SF.Net is sponsored by: Speed Start Your Linux Apps Now.
>Build and deploy apps & Web services for Linux with
>a free DVD software kit from IBM. Click Now!
>http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
>_______________________________________________
>Xen-devel mailing list
>Xen-devel@lists.sourceforge.net
>https://lists.sourceforge.net/lists/listinfo/xen-devel
>  
>


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

Zachary Amsden wrote:
> Just an observation - I''ve noticed that several of the include
files
> include linux/config.h and some include xeno/config.h - also for some 
> other included files this is the case.  Also,  -nostdinc only seems to 
Now that you mention it, debian sid (and sarge, I think) now include 
linux''s 2.6 kernel headers. I know that it caused some compile trouble 
in user-mode-linux. This is just a wild guess and maybe unrelated to 
xen... :)

Regards,
Nuno Silva



-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

> Just an observation - I''ve noticed that several of the include
files
> include linux/config.h and some include xeno/config.h - also for some 
> other included files this is the case.  Also,  -nostdinc only seems to 
> be applied to the arch/i386 branch - which means the drivers may be 
> getting a different set of include paths.  Perhaps some of your config 
> settings are different in different .c files as a result with 
> non-standard config options.  Although I kind of doubt the problem is 
> this simple, it seems reasonable to clean up the build tree so that it 
> has a well isolated include structure.  I think I just volunteered 
> myself unless anyone else is already working on it.
> 
> Zachary Amsden
> zamsden@cisco.com
If you can work out what''s going on then we''d be glad to
accept a
patch. It''s quite possible that the arch/xeno master Makefile has got
out of sync with arch/i386 as we''ve ported to more recent 2.4 kernels.

 -- Keir


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

On Wednesday 18 February 2004 02:49, Zachary Amsden
wrote:
> Just an observation - I''ve noticed that several of the include
files
> include linux/config.h and some include xeno/config.h - also for some
Hmmm... which ones? I have just grepped the whole tree and didn''t find
any...
> other included files this is the case.  Also,  -nostdinc only seems to
> be applied to the arch/i386 branch - which means the drivers may be
> getting a different set of include paths.  Perhaps some of your config
...also true for this one. Where did you notice this no -nostdinc situation? 


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

Tvrtko A. Uršulin wrote:
>On Wednesday 18 February 2004 02:49, Zachary Amsden wrote:
>  
>
>>Just an observation - I''ve noticed that several of the include
files
>>include linux/config.h and some include xeno/config.h - also for some
>>    
>>
>
>Hmmm... which ones? I have just grepped the whole tree and didn''t
find any...
>
>  
>
xen/include/asm-i386/apic.h for one... and there are more dangerous 
looking things:

xen/include/asm-i386/dma.h includes both "linux/config.h" and 
"linux/spinklock.h".  This is very likely to be built into the
drivers,
and likely to cause major havoc if you change the cacheline size.
>>other included files this is the case.  Also,  -nostdinc only seems to
>>be applied to the arch/i386 branch - which means the drivers may be
>>getting a different set of include paths.  Perhaps some of your config
>>    
>>
>
>...also true for this one. Where did you notice this no -nostdinc situation?
>  
>
Only arch/i386/Rules.mk specifies -nostdinc.  This is really dangerous 
if you are not building on a 2.4.24 or very close kernel, and makes 
cross-compiling out of the question.

I should have a patch for a slightly cleaner build today that you could 
try out.  It''s likely to be largish, since I may need to pull in some 
additional headers from 2.4.24.

Zachary Amsden
zamsden@cisco.com


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

> xen/include/asm-i386/apic.h for one... and there are more dangerous 
> looking things:
> 
> xen/include/asm-i386/dma.h includes both "linux/config.h" and 
> "linux/spinklock.h".  This is very likely to be built into the
drivers,
> and likely to cause major havoc if you change the cacheline size.
>
> Only arch/i386/Rules.mk specifies -nostdinc.  This is really dangerous 
> if you are not building on a 2.4.24 or very close kernel, and makes 
> cross-compiling out of the question.
> 
> I should have a patch for a slightly cleaner build today that you could 
> try out.  It''s likely to be largish, since I may need to pull in
some
> additional headers from 2.4.24.
I''m not sure that I understand what you are trying to do. 
#include <linux/foo.h> and #include <xeno/foo.h> should mean the
same
thing when building Xen -- all kernel object files should be built
with -nostdinc, and incude/linux is a symlink to include/xeno.

When building Linux for Xen, it does not use any Xen headers except
those in include/hypervisor-ifs -- all other headers are private to
Xen!

It sounds like you may be spending time on something that
isn''''t
broken. :-(

 -- Keir


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

Keir Fraser wrote:
>>xen/include/asm-i386/apic.h for one... and there are more dangerous 
>>looking things:
>>
>>xen/include/asm-i386/dma.h includes both "linux/config.h" and 
>>"linux/spinklock.h".  This is very likely to be built into the
drivers,
>>and likely to cause major havoc if you change the cacheline size.
>>
>>Only arch/i386/Rules.mk specifies -nostdinc.  This is really dangerous 
>>if you are not building on a 2.4.24 or very close kernel, and makes 
>>cross-compiling out of the question.
>>
>>I should have a patch for a slightly cleaner build today that you could 
>>try out.  It''s likely to be largish, since I may need to pull
in some
>>additional headers from 2.4.24.
>>    
>>
>
>I''m not sure that I understand what you are trying to do. 
>#include <linux/foo.h> and #include <xeno/foo.h> should mean the
same
>thing when building Xen -- all kernel object files should be built
>with -nostdinc, and incude/linux is a symlink to include/xeno.
>
>When building Linux for Xen, it does not use any Xen headers except
>those in include/hypervisor-ifs -- all other headers are private to
>Xen!
>  
>
I''m not talking about Linux -- I''m talking about Xen itself. 
Xen
depends on the local system includes, which is not good.  Xen is 
definitely not built with -nostdinc, which is what I am trying to fix.

Zachary Amsden
zamsden@cisco.com


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

> I''m not talking about Linux -- I''m talking about Xen
itself.  Xen
> depends on the local system includes, which is not good.  Xen is 
> definitely not built with -nostdinc, which is what I am trying to fix.
Really? From within the Xen bukd tree, try:
 make clean
 make | grep "^gcc" | grep -v nostdinc

No kernel object files will be printed.

Note that all directories pull in arch/i386/Rules.mk, and that file
adds -nostdinc to CFLAGS.

 -- Keir


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

Keir Fraser wrote:
>>I''m not talking about Linux -- I''m talking about Xen
itself.  Xen
>>depends on the local system includes, which is not good.  Xen is 
>>definitely not built with -nostdinc, which is what I am trying to fix.
>>    
>>
>
>Really? From within the Xen bukd tree, try:
> make clean
> make | grep "^gcc" | grep -v nostdinc
>
>No kernel object files will be printed.
>
>Note that all directories pull in arch/i386/Rules.mk, and that file
>adds -nostdinc to CFLAGS.
>
> -- Keir
>  
>
Ahh, missed that.  It seems that -nostdinc shouldn''t really be arch 
dependent, but I don''t see much point in reworking the build now.

Zachary Amsden
zamsden@cisco.com


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel