2.6 kgdb without serial port

2.6 kgdb without serial port

[Pavel Machek]

Hi!

2.6 kgdb patches in -mm tree seem to contain kgdb-over-ethernet stuff,
but still require me to fill in serial port interrupt/address. Is
there easy way to make it work without serial port? [This notebook has
none :-(].
								Pavel
-- 
When do you have heart between your knees?

Re: 2.6 kgdb without serial port

[Andrew Morton]

Pavel Machek <pavel@ucw.cz> wrote:
>
> Hi!
> 
> 2.6 kgdb patches in -mm tree seem to contain kgdb-over-ethernet stuff,
> but still require me to fill in serial port interrupt/address. Is
> there easy way to make it work without serial port? [This notebook has
> none :-(].

That's a bit ugly, but things should still work OK?  Give it some random
UART address but specify an ethernet connection at boot time - the kgdb
stub should never touch the UART.

crypto-loop + highmen -> random crashes in -test11

[Soeren Sonnenburg]

Hi.

I get random crashes/corruption/ init kills when I use cryptoloop on
this highmem enabled ppc machine.

Applying the loop-* patches at
http://www.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.0-test10/2.6.0-test10-mm1/broken-out

seem to fix this problem (at least I did not get any crash due to
cryptoloop in the last 2 days)

Steps to reproduce:

dd if=/dev/zero of=/file bs=1M count=100
losetup -e blowfish /dev/loop0 /file
Password:
mkfs -t ext3 /dev/loop0
mount /dev/loop0 /mnt
cd /mnt  
dd if=/dev/zero of=bla

gives you nice unpredictable crashes.


Soeren

Re: crypto-loop + highmen -> random crashes in -test11

[Andrew Morton]

Soeren Sonnenburg <kernel@nn7.de> wrote:
>
> Hi.
> 
> I get random crashes/corruption/ init kills when I use cryptoloop on
> this highmem enabled ppc machine.

People have reported cryptoloop+highmem crashes on ia32 as well.  I'm not
sure if that was with -mm though.

> Applying the loop-* patches at
> http://www.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.0-test10/2.6.0-test10-mm1/broken-out
> 
> seem to fix this problem (at least I did not get any crash due to
> cryptoloop in the last 2 days)
>
> Steps to reproduce:
> 
> dd if=/dev/zero of=/file bs=1M count=100
> losetup -e blowfish /dev/loop0 /file
> Password:
> mkfs -t ext3 /dev/loop0
> mount /dev/loop0 /mnt
> cd /mnt  
> dd if=/dev/zero of=bla
> 
> gives you nice unpredictable crashes.
>

Nice, but I don't know why those patches fixed it. hrm.

Re: crypto-loop + highmen -> random crashes in -test11

[bill davidsen]

In article <1071570648.3528.50.camel@localhost>,
Soeren Sonnenburg  <kernel@nn7.de> wrote:

| I get random crashes/corruption/ init kills when I use cryptoloop on
| this highmem enabled ppc machine.
| 
| Applying the loop-* patches at
| http://www.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.0-test10/2.6.0-test10-mm1/broken-out
| 
| seem to fix this problem (at least I did not get any crash due to
| cryptoloop in the last 2 days)

Thank you for identifying this. I have been having perfect results iwth
the systems I have running cryptoloop, but they are both small. Now I
not only know why I don't have problems, I have a way to prevent
problems.

Hopefully this will get into 2.6.0-final, since it fixes a hard fail case.
| 
| Steps to reproduce:
| 
| dd if=/dev/zero of=/file bs=1M count=100
| losetup -e blowfish /dev/loop0 /file
| Password:
| mkfs -t ext3 /dev/loop0
| mount /dev/loop0 /mnt
| cd /mnt  
| dd if=/dev/zero of=bla
| 
| gives you nice unpredictable crashes.
| 
| 
| Soeren
| 
| -
| To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
| the body of a message to majordomo@vger.kernel.org
| More majordomo info at  http://vger.kernel.org/majordomo-info.html
| Please read the FAQ at  http://www.tux.org/lkml/
| 


-- 
bill davidsen <davidsen@tmr.com>
  CTO, TMR Associates, Inc
Doing interesting things with little computers since 1979.

Re: crypto-loop + highmen -> random crashes in -test11

[Soeren Sonnenburg]

On Tue, 2003-12-16 at 12:40, Andrew Morton wrote:
> Soeren Sonnenburg <kernel@nn7.de> wrote:
> >
> > Hi.
> > 
> > I get random crashes/corruption/ init kills when I use cryptoloop on
> > this highmem enabled ppc machine.
> 
> People have reported cryptoloop+highmem crashes on ia32 as well.  I'm not
> sure if that was with -mm though.
> 
> > Applying the loop-* patches at
> > http://www.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.0-test10/2.6.0-test10-mm1/broken-out
> > 
> > seem to fix this problem (at least I did not get any crash due to
> > cryptoloop in the last 2 days)
[...]
> Nice, but I don't know why those patches fixed it. hrm.
> 

no idea. all I can say is that it cryptoloop pretends to work reliably since 3 days...

Soeren

Re: 2.6 kgdb without serial port

[Pavel Machek]

Hi!

> > 2.6 kgdb patches in -mm tree seem to contain kgdb-over-ethernet stuff,
> > but still require me to fill in serial port interrupt/address. Is
> > there easy way to make it work without serial port? [This notebook has
> > none :-(].
> 
> That's a bit ugly, but things should still work OK?  Give it some random
> UART address but specify an ethernet connection at boot time - the kgdb
> stub should never touch the UART.

I found out what was biting me: using 2.95 with kgdb is bad idea. 2.95
with kgdb means reboot just after uncompressing kernel -- pretty nasty
to debug. Please apply,
									Pavel
PS: kgdb could use some code-style changes. Will you accept such
patches?

--- tmp/linux/arch/i386/kernel/kgdb_stub.c	2003-12-26 18:50:45.000000000 +0100
+++ linux/arch/i386/kernel/kgdb_stub.c	2003-12-26 18:41:58.000000000 +0100
@@ -121,6 +121,10 @@
 #include <linux/inet.h>
 #include <linux/kallsyms.h>
 
+#if __GNUC__ < 3
+#error Sorry, your GCC is too old to work with kgdb. (It works with 3.3.2)
+#endif
+
 /************************************************************************
  *
  * external low-level support routines

-- 
When do you have a heart between your knees?
[Johanka's followup: and *two* hearts?]

Re: 2.6 kgdb without serial port

[Andrew Morton]

Pavel Machek <pavel@ucw.cz> wrote:
>
> Hi!
> 
> > > 2.6 kgdb patches in -mm tree seem to contain kgdb-over-ethernet stuff,
> > > but still require me to fill in serial port interrupt/address. Is
> > > there easy way to make it work without serial port? [This notebook has
> > > none :-(].
> > 
> > That's a bit ugly, but things should still work OK?  Give it some random
> > UART address but specify an ethernet connection at boot time - the kgdb
> > stub should never touch the UART.
> 
> I found out what was biting me: using 2.95 with kgdb is bad idea. 2.95
> with kgdb means reboot just after uncompressing kernel -- pretty nasty
> to debug. Please apply,

I've been using 2.95.3 on and off for ages, no problems?

> PS: kgdb could use some code-style changes. Will you accept such
> patches?

Spose so.  The kgdb stub really needs a serious reorganisation so that
non-ia32 architectures can share generic things.  And general reduction of
the patch footprint, maybe some feature removal too.   A fairly large job.

Re: 2.6 kgdb without serial port

[Martin J. Bligh]

>> > > 2.6 kgdb patches in -mm tree seem to contain kgdb-over-ethernet stuff,
>> > > but still require me to fill in serial port interrupt/address. Is
>> > > there easy way to make it work without serial port? [This notebook has
>> > > none :-(].
>> > 
>> > That's a bit ugly, but things should still work OK?  Give it some random
>> > UART address but specify an ethernet connection at boot time - the kgdb
>> > stub should never touch the UART.
>> 
>> I found out what was biting me: using 2.95 with kgdb is bad idea. 2.95
>> with kgdb means reboot just after uncompressing kernel -- pretty nasty
>> to debug. Please apply,
> 
> I've been using 2.95.3 on and off for ages, no problems?

2.95.4 seems to work as well, or did last time I tried it. 
Pavel, what version exactly were you using?

early_printk might help debugging, if you're interested.
ftp://ftp.kernel.org/pub/linux/kernel/people/mbligh/patches/2.6.0/2.6.0-mjb1/200-early_printk

M.

Re: 2.6 kgdb without serial port

[Pavel Machek]

Hi!

> > > > 2.6 kgdb patches in -mm tree seem to contain kgdb-over-ethernet stuff,
> > > > but still require me to fill in serial port interrupt/address. Is
> > > > there easy way to make it work without serial port? [This notebook has
> > > > none :-(].
> > > 
> > > That's a bit ugly, but things should still work OK?  Give it some random
> > > UART address but specify an ethernet connection at boot time - the kgdb
> > > stub should never touch the UART.
> > 
> > I found out what was biting me: using 2.95 with kgdb is bad idea. 2.95
> > with kgdb means reboot just after uncompressing kernel -- pretty nasty
> > to debug. Please apply,
> 
> I've been using 2.95.3 on and off for ages, no problems?

Strange... Okay, some details. If I use gcc-2.95.4 (from debian) by
setting GCC = gcc-2.95, it crashes very early during boot. So perhaps
it dislikes "crosscompiling" instead of specific gcc version? [I was
doing that to speed compiles up].
								Pavel		

> Spose so.  The kgdb stub really needs a serious reorganisation so that
> non-ia32 architectures can share generic things.  And general reduction of
> the patch footprint, maybe some feature removal too.   A fairly large job.

Yes, and as I need kgdb on x86-64...

-- 
When do you have a heart between your knees?
[Johanka's followup: and *two* hearts?]