2.6.18-rt6

2.6.18-rt6

[Ingo Molnar]

i've released the 2.6.18-rt6 tree, which can be downloaded from the 
usual place:

  http://redhat.com/~mingo/realtime-preempt/

this is a fixes-mostly release. Changes since -rt4:

 - fix for module loading / symbol table crash (John Stultz)
 - scheduler fix (Mike Galbraith)
 - fix x86_64 NMI watchdog & preempt-rcu interaction
 - fix time-warp false positives
 - jiffies_to_timespec export fix (Steven Rostedt)
 - ll_rw_block.c warning fix (Mike Galbraith)
 - PPC updates (Daniel Walker)
 - MIPS updates (Manish Lachwani)
 - ARM oprofile fix (Kevin Hilman)
 - traditional futexes queued via plists (Séstien Duguése)
 - (various other smaller fixes)

to build a 2.6.18-rt6 tree, the following patches should be applied:

  http://kernel.org/pub/linux/kernel/v2.6/linux-2.6.18.tar.bz2
  http://redhat.com/~mingo/realtime-preempt/patch-2.6.18-rt6

as usual, bugreports, fixes and suggestions are welcome,

	Ingo

Re: 2.6.18-rt6

[Sébastien Dugué]

  Hi Ingo, 

  nice work,

On Wed, 2006-10-18 at 10:39 +0200, Ingo Molnar wrote:
> i've released the 2.6.18-rt6 tree, which can be downloaded from the 
> usual place:
> 
>   http://redhat.com/~mingo/realtime-preempt/
> 
> this is a fixes-mostly release. Changes since -rt4:
> 
>  - fix for module loading / symbol table crash (John Stultz)
>  - scheduler fix (Mike Galbraith)
>  - fix x86_64 NMI watchdog & preempt-rcu interaction
>  - fix time-warp false positives
>  - jiffies_to_timespec export fix (Steven Rostedt)
>  - ll_rw_block.c warning fix (Mike Galbraith)
>  - PPC updates (Daniel Walker)
>  - MIPS updates (Manish Lachwani)
>  - ARM oprofile fix (Kevin Hilman)
>  - traditional futexes queued via plists (Séstien Duguése)

  Thanks for including this in your tree.
 BTW, my name is Sébastien Dugué.

>  - (various other smaller fixes)
> 
> to build a 2.6.18-rt6 tree, the following patches should be applied:
> 
>   http://kernel.org/pub/linux/kernel/v2.6/linux-2.6.18.tar.bz2
>   http://redhat.com/~mingo/realtime-preempt/patch-2.6.18-rt6
> 
> as usual, bugreports, fixes and suggestions are welcome,
> 
> 	Ingo

  Sébastien.

Re: 2.6.18-rt6

[Ingo Molnar]

* Sébastien Dugué <sebastien.dugue@bull.net> wrote:

> >  - traditional futexes queued via plists (Séstien Duguése)
> 
>   Thanks for including this in your tree.
>  BTW, my name is Sébastien Dugué.

doh - weird cut & paste error :-/

	Ingo

Re: 2.6.18-rt6

[Paul E. McKenney]

On Wed, Oct 18, 2006 at 10:39:21AM +0200, Ingo Molnar wrote:
> 
> i've released the 2.6.18-rt6 tree, which can be downloaded from the 
> usual place:
> 
>   http://redhat.com/~mingo/realtime-preempt/
> 
> this is a fixes-mostly release. Changes since -rt4:
> 
>  - fix for module loading / symbol table crash (John Stultz)
>  - scheduler fix (Mike Galbraith)
>  - fix x86_64 NMI watchdog & preempt-rcu interaction
>  - fix time-warp false positives
>  - jiffies_to_timespec export fix (Steven Rostedt)
>  - ll_rw_block.c warning fix (Mike Galbraith)
>  - PPC updates (Daniel Walker)
>  - MIPS updates (Manish Lachwani)
>  - ARM oprofile fix (Kevin Hilman)
>  - traditional futexes queued via plists (Séstien Duguése)
>  - (various other smaller fixes)
> 
> to build a 2.6.18-rt6 tree, the following patches should be applied:
> 
>   http://kernel.org/pub/linux/kernel/v2.6/linux-2.6.18.tar.bz2
>   http://redhat.com/~mingo/realtime-preempt/patch-2.6.18-rt6
> 
> as usual, bugreports, fixes and suggestions are welcome,

A nit from IPv6, since I happened by chance to run this on an IPv6
machine -- there are a couple of smp_processor_id() calls that need
to be raw_smp_processor_id() to suppress warnings.  I believe that this
is the correct change, as it seems to me that the locking protects
things so that preemption is not a problem.  That said, I cannot claim
to be an IPv6 expert.  Tested on x86.

Signed-off-by: Paul E. McKenney <paulmck@us.ibm.com>
---

 ip6_tables.c |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)

diff -urpNa -X dontdiff linux-2.6.18-rt3/net/ipv6/netfilter/ip6_tables.c linux-2.6.18-rt3-ip6t_do_table/net/ipv6/netfilter/ip6_tables.c
--- linux-2.6.18-rt3/net/ipv6/netfilter/ip6_tables.c	2006-09-19 20:42:06.000000000 -0700
+++ linux-2.6.18-rt3-ip6t_do_table/net/ipv6/netfilter/ip6_tables.c	2006-10-17 17:44:55.000000000 -0700
@@ -285,7 +285,7 @@ ip6t_do_table(struct sk_buff **pskb,
 	read_lock_bh(&table->lock);
 	private = table->private;
 	IP_NF_ASSERT(table->valid_hooks & (1 << hook));
-	table_base = (void *)private->entries[smp_processor_id()];
+	table_base = (void *)private->entries[raw_smp_processor_id()];
 	e = get_entry(table_base, private->hook_entry[hook]);
 
 	/* For return from builtin chain */
@@ -1110,7 +1110,7 @@ do_add_counters(void __user *user, unsig
 
 	i = 0;
 	/* Choose the copy that is on our node */
-	loc_cpu_entry = private->entries[smp_processor_id()];
+	loc_cpu_entry = private->entries[raw_smp_processor_id()];
 	IP6T_ENTRY_ITERATE(loc_cpu_entry,
 			  private->size,
 			  add_counter_to_entry,

Re: 2.6.18-rt6

[Esben Nielsen]

[-- Attachment #1: Type: TEXT/PLAIN, Size: 1767 bytes --]



On Wed, 18 Oct 2006, Ingo Molnar wrote:

>
> i've released the 2.6.18-rt6 tree, which can be downloaded from the
> usual place:
>
>  http://redhat.com/~mingo/realtime-preempt/
>
> this is a fixes-mostly release. Changes since -rt4:
>
> - fix for module loading / symbol table crash (John Stultz)
> - scheduler fix (Mike Galbraith)
> - fix x86_64 NMI watchdog & preempt-rcu interaction
> - fix time-warp false positives
> - jiffies_to_timespec export fix (Steven Rostedt)
> - ll_rw_block.c warning fix (Mike Galbraith)
> - PPC updates (Daniel Walker)
> - MIPS updates (Manish Lachwani)
> - ARM oprofile fix (Kevin Hilman)
> - traditional futexes queued via plists (Séstien Duguése)
> - (various other smaller fixes)

I assume this goes under small fixes:

I see that
 	spin_lock(&pi_state->pi_mutex.wait_lock);
is added in futex.c:570 in wake_futex_pi().
I do but I see some missing unlocks at returns.

I have a totally untested (it compiles) patch below.

Esben

  kernel/futex.c |    8 ++++++--
  1 file changed, 6 insertions(+), 2 deletions(-)

Index: linux-2.6.18-rt/kernel/futex.c
===================================================================
--- linux-2.6.18-rt.orig/kernel/futex.c
+++ linux-2.6.18-rt/kernel/futex.c
@@ -590,10 +590,14 @@ static int wake_futex_pi(u32 __user *uad
  		inc_preempt_count();
  		curval = futex_atomic_cmpxchg_inatomic(uaddr, uval, newval);
  		dec_preempt_count();
-		if (curval == -EFAULT)
+		if (curval == -EFAULT) {
+			spin_unlock(&pi_state->pi_mutex.wait_lock);
  			return -EFAULT;
-		if (curval != uval)
+		}
+		if (curval != uval) {
+			spin_unlock(&pi_state->pi_mutex.wait_lock);
  			return -EINVAL;
+		}
  	}

  	spin_lock_irq(&pi_state->owner->pi_lock);

Re: 2.6.18-rt6

[Sergio Monteiro Basto]

[-- Attachment #1: Type: text/plain, Size: 2264 bytes --]

I just test rt6 it in my problematic VIA Board with one PENTIUM D (DUAL)

1. don't apply cleaning to 2.16.18.1 (on sparc arch )
2. I usual boot with notsc (because without it give me many lost time
tickets)
    I try 2.16.18.1-rt6 without notsc and freeze on boot. 
    I try with notsc and says 
        Time: acpi_pm clocksource has been installed.
        Looks stable and don't show any lost timer ticket on dmesg.

But I want also work with Nvidia dri which is a close drive from NVIDIA.
I install that drive and I enable DRI, After a few minutes I got a
spontaneous reboot.
I will keep testing without nvidia close source.

And I like to know if this rt6 patch make this new clocksource (acpi_pm)
or just found it and use it ? 

cat /sys/devices/system/clocksource/clocksource0/available_clocksource
acpi_pm jiffies
cat /sys/devices/system/clocksource/clocksource0/current_clocksource
acpi_pm

Thanks, 



On Wed, 2006-10-18 at 10:39 +0200, Ingo Molnar wrote:
> i've released the 2.6.18-rt6 tree, which can be downloaded from the 
> usual place:
> 
>   http://redhat.com/~mingo/realtime-preempt/
> 
> this is a fixes-mostly release. Changes since -rt4:
> 
>  - fix for module loading / symbol table crash (John Stultz)
>  - scheduler fix (Mike Galbraith)
>  - fix x86_64 NMI watchdog & preempt-rcu interaction
>  - fix time-warp false positives
>  - jiffies_to_timespec export fix (Steven Rostedt)
>  - ll_rw_block.c warning fix (Mike Galbraith)
>  - PPC updates (Daniel Walker)
>  - MIPS updates (Manish Lachwani)
>  - ARM oprofile fix (Kevin Hilman)
>  - traditional futexes queued via plists (Séstien Duguése)
>  - (various other smaller fixes)
> 
> to build a 2.6.18-rt6 tree, the following patches should be applied:
> 
>   http://kernel.org/pub/linux/kernel/v2.6/linux-2.6.18.tar.bz2
>   http://redhat.com/~mingo/realtime-preempt/patch-2.6.18-rt6
> 
> as usual, bugreports, fixes and suggestions are welcome,
> 
> 	Ingo
> -
> 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/
-- 
Sérgio M.B.

[-- Attachment #2: smime.p7s --]
[-- Type: application/x-pkcs7-signature, Size: 2166 bytes --]

Re: 2.6.18-rt6

[john stultz]

On Fri, 2006-10-20 at 01:23 +0100, Sergio Monteiro Basto wrote:
> I just test rt6 it in my problematic VIA Board with one PENTIUM D (DUAL)
> 
> 1. don't apply cleaning to 2.16.18.1 (on sparc arch )
> 2. I usual boot with notsc (because without it give me many lost time
> tickets)
>     I try 2.16.18.1-rt6 without notsc and freeze on boot. 
>     I try with notsc and says 
>         Time: acpi_pm clocksource has been installed.
>         Looks stable and don't show any lost timer ticket on dmesg.

Out of curiosity, does 2.6.18.1 require notsc (or booting w/
clocksource=acpi_pm) to be stable? Or is it just -rt6 that has this
problem?

> But I want also work with Nvidia dri which is a close drive from NVIDIA.
> I install that drive and I enable DRI, After a few minutes I got a
> spontaneous reboot.
> I will keep testing without nvidia close source.
> 
> And I like to know if this rt6 patch make this new clocksource (acpi_pm)
> or just found it and use it ? 

-rt6 includes a patch from Thomas Gleixner that disables the TSC if
dynticks are enabled in your config.

thanks
-john

Re: 2.6.18-rt6 and compile fail with rt7 on x86_64

[Sergio Monteiro Basto]

[-- Attachment #1: Type: text/plain, Size: 2628 bytes --]

On Thu, 2006-10-19 at 17:59 -0700, john stultz wrote:
> On Fri, 2006-10-20 at 01:23 +0100, Sergio Monteiro Basto wrote:
> > I just test rt6 it in my problematic VIA Board with one PENTIUM D (DUAL)
> > 
> > 1. don't apply cleaning to 2.16.18.1 (on sparc arch )
> > 2. I usual boot with notsc (because without it give me many lost time
> > tickets)
> >     I try 2.16.18.1-rt6 without notsc and freeze on boot. 
> >     I try with notsc and says 
> >         Time: acpi_pm clocksource has been installed.
> >         Looks stable and don't show any lost timer ticket on dmesg.
> 
> Out of curiosity, does 2.6.18.1 require notsc (or booting w/
> clocksource=acpi_pm) to be stable? Or is it just -rt6 that has this
> problem?
> 

yes, 2.6.18.1 require notsc to work properly , but without notsc boots
normally but with  many lost timer ticks 
    time.c: Lost 300 timer tick(s)! rip mwait_idle+0x3f/0x54)
    psmouse.c: Wheel Mouse at isa0060/serio1/input0 lost
synchronization, throwing 2 bytes away.
    time.c: Lost 300 timer tick(s)! rip mwait_idle+0x3f/0x54)

booting with clocksource=acpi_pm , don't find/use   acpi_pm
cat /sys/devices/system/clocksource/clocksource0/available_clocksource
jiffies

> > But I want also work with Nvidia dri which is a close drive from NVIDIA.
> > I install that drive and I enable DRI, After a few minutes I got a
> > spontaneous reboot.
> > I will keep testing without nvidia close source.
> > 
> > And I like to know if this rt6 patch make this new clocksource (acpi_pm)
> > or just found it and use it ? 
> 
> -rt6 includes a patch from Thomas Gleixner that disables the TSC if
> dynticks are enabled in your config.

I will try that 

Thanks, 
> 
> thanks
> -john
> 

compiles fails with rt7 
+ KernelImage=arch/x86_64/boot/bzImage
+ make -s ARCH=x86_64 oldconfig
+ make -s ARCH=x86_64 -j2 bzImage
  CHK     include/linux/version.h
  UPD     include/linux/version.h
  CHK     include/linux/utsrelease.h
  SYMLINK include/asm -> include/asm-x86_64
  UPD     include/linux/utsrelease.h
  CHK     include/linux/compile.h
  UPD     include/linux/compile.h
kernel/rtmutex.c:938:48: error: macro "rt_release_bkl" passed 2
arguments, but takes just 1
kernel/rtmutex.c: In function 'rt_mutex_slowlock':
kernel/rtmutex.c:938: error: 'rt_release_bkl' undeclared (first use in
this function)
kernel/rtmutex.c:938: error: (Each undeclared identifier is reported
only once
kernel/rtmutex.c:938: error: for each function it appears in.)
make[1]: *** [kernel/rtmutex.o] Error 1
make: *** [kernel] Error 2


thanks 
-- 
Sérgio M.B.

[-- Attachment #2: smime.p7s --]
[-- Type: application/x-pkcs7-signature, Size: 2166 bytes --]

Re: 2.6.18-rt6

[Lee Revell]

On Wed, 2006-10-18 at 10:39 +0200, Ingo Molnar wrote:
> i've released the 2.6.18-rt6 tree, which can be downloaded from the 
> usual place:
> 
>   http://redhat.com/~mingo/realtime-preempt/

This does not work here.  It boots but then wants to fsck my disks, and
dies with a sig 11 in fsck.ext3.  This is 100% reproducible and booting
2.6.18-rt5 works and does not want to fsck the disks. 

Sorry I don't have more information, the box is headless and in
production so I have limited debugging bandwidth.

Do you need my .config?

Lee

Re: 2.6.18-rt6

[Mark Knecht]

On 10/20/06, Lee Revell <rlrevell@joe-job.com> wrote:
> On Wed, 2006-10-18 at 10:39 +0200, Ingo Molnar wrote:
> > i've released the 2.6.18-rt6 tree, which can be downloaded from the
> > usual place:
> >
> >   http://redhat.com/~mingo/realtime-preempt/
>
> This does not work here.  It boots but then wants to fsck my disks, and
> dies with a sig 11 in fsck.ext3.  This is 100% reproducible and booting
> 2.6.18-rt5 works and does not want to fsck the disks.
>
> Sorry I don't have more information, the box is headless and in
> production so I have limited debugging bandwidth.
>
> Do you need my .config?
>
> Lee
>

I've been running 2.6.18-rt6 for a couple of days now. No problems so far.

mark@lightning ~ $ uname -a
Linux lightning 2.6.18-rt6 #2 PREEMPT Wed Oct 18 10:18:51 PDT 2006
x86_64 AMD Athlon(tm) 64 Processor 3000+ GNU/Linux
mark@lightning ~ $ uptime
 08:52:52 up 21:32,  2 users,  load average: 0.93, 0.73, 0.47
mark@lightning ~ $


I am single processor. HRT is turned on. I am not using the DynTicks
thing as I don't know what it's supposed to do.

Anyway, 2.6.18-rt5 wouldn't boot reliably for me so it seems we are
out of phase with each other. Maybe compare .config's?

- Mark

-rt7 announcement? (was Re: 2.6.18-rt6)

[Lee Revell]

On Fri, 2006-10-20 at 11:00 -0400, Lee Revell wrote:
> On Wed, 2006-10-18 at 10:39 +0200, Ingo Molnar wrote:
> > i've released the 2.6.18-rt6 tree, which can be downloaded from the 
> > usual place:
> > 
> >   http://redhat.com/~mingo/realtime-preempt/
> 
> This does not work here.  It boots but then wants to fsck my disks, and
> dies with a sig 11 in fsck.ext3.  This is 100% reproducible and booting
> 2.6.18-rt5 works and does not want to fsck the disks. 

I see that -rt7 is posted.  The patch is a huge diff from -rt6.  Where
are the release notes?

Lee

Re: -rt7 announcement? (was Re: 2.6.18-rt6)

[Mark Knecht]

On 10/23/06, Lee Revell <rlrevell@joe-job.com> wrote:
> On Fri, 2006-10-20 at 11:00 -0400, Lee Revell wrote:
> > On Wed, 2006-10-18 at 10:39 +0200, Ingo Molnar wrote:
> > > i've released the 2.6.18-rt6 tree, which can be downloaded from the
> > > usual place:
> > >
> > >   http://redhat.com/~mingo/realtime-preempt/
> >
> > This does not work here.  It boots but then wants to fsck my disks, and
> > dies with a sig 11 in fsck.ext3.  This is 100% reproducible and booting
> > 2.6.18-rt5 works and does not want to fsck the disks.
>
> I see that -rt7 is posted.  The patch is a huge diff from -rt6.  Where
> are the release notes?
>
> Lee

I've been running it for a few hours on my AMD64. No problems so far.

- Mark

lightning ~ # uname -a
Linux lightning 2.6.18-rt7 #1 PREEMPT Mon Oct 23 07:28:51 PDT 2006
x86_64 AMD Athlon(tm) 64 Processor 3000+ GNU/Linux
lightning ~ # uptime
 11:28:25 up  3:53,  2 users,  load average: 0.20, 0.10, 0.28
lightning ~ #

Re: -rt7 announcement? (was Re: 2.6.18-rt6)

[Thomas Gleixner]

On Mon, 2006-10-23 at 12:34 -0400, Lee Revell wrote:
> On Fri, 2006-10-20 at 11:00 -0400, Lee Revell wrote:
> > On Wed, 2006-10-18 at 10:39 +0200, Ingo Molnar wrote:
> > > i've released the 2.6.18-rt6 tree, which can be downloaded from the 
> > > usual place:
> > > 
> > >   http://redhat.com/~mingo/realtime-preempt/
> > 
> > This does not work here.  It boots but then wants to fsck my disks, and
> > dies with a sig 11 in fsck.ext3.  This is 100% reproducible and booting
> > 2.6.18-rt5 works and does not want to fsck the disks. 
> 
> I see that -rt7 is posted.  The patch is a huge diff from -rt6.  Where
> are the release notes?

Basically we merged the latest hrt-dyntick queue into -rt.

	tglx

Re: -rt7 announcement? (was Re: 2.6.18-rt6)

[Mark Knecht]

On 10/23/06, Thomas Gleixner <tglx@linutronix.de> wrote:
> On Mon, 2006-10-23 at 12:34 -0400, Lee Revell wrote:
> > On Fri, 2006-10-20 at 11:00 -0400, Lee Revell wrote:
> > > On Wed, 2006-10-18 at 10:39 +0200, Ingo Molnar wrote:
> > > > i've released the 2.6.18-rt6 tree, which can be downloaded from the
> > > > usual place:
> > > >
> > > >   http://redhat.com/~mingo/realtime-preempt/
> > >
> > > This does not work here.  It boots but then wants to fsck my disks, and
> > > dies with a sig 11 in fsck.ext3.  This is 100% reproducible and booting
> > > 2.6.18-rt5 works and does not want to fsck the disks.
> >
> > I see that -rt7 is posted.  The patch is a huge diff from -rt6.  Where
> > are the release notes?
>
> Basically we merged the latest hrt-dyntick queue into -rt.
>
>         tglx

Hi Thomas,
   Some differences/questions between 2.6.18-rt6 and 2.6.18-rt7:

   In 2.6.18-rt6, using make menuconfig, there were options on the
front page for HRT Support. This seems to have moved under Processor
Type with 2.6.18-rt7. Was that on purpose?

   In 2.6.18-rt6 I turned on HRT support, left 1000 nanoseconds for
the timing, but did not enable dynamic ticks since I wasn't sure it
was OK on AMD64. Should I be using DynTicks with an AMD64 single
processor? With a dual-processor?

   On 2.6.18-rt7 it seems there is no time value setting or it's been
moved somewhere I haven't found. does that 1000 nanosecond time
setting still apply?

Thanks,
Mark

Re: -rt7 announcement? (was Re: 2.6.18-rt6)

[Thomas Gleixner]

On Mon, 2006-10-23 at 11:44 -0700, Mark Knecht wrote:
> Hi Thomas,
>    Some differences/questions between 2.6.18-rt6 and 2.6.18-rt7:
> 
>    In 2.6.18-rt6, using make menuconfig, there were options on the
> front page for HRT Support. This seems to have moved under Processor
> Type with 2.6.18-rt7. Was that on purpose?

Yes

>    In 2.6.18-rt6 I turned on HRT support, left 1000 nanoseconds for
> the timing, but did not enable dynamic ticks since I wasn't sure it
> was OK on AMD64. Should I be using DynTicks with an AMD64 single
> processor? With a dual-processor?

Should work

>    On 2.6.18-rt7 it seems there is no time value setting or it's been
> moved somewhere I haven't found. does that 1000 nanosecond time
> setting still apply?

Basically yes. It is handled internally.

	tglx

Re: -rt7 announcement? (was Re: 2.6.18-rt6)

[Mark Knecht]

On 10/23/06, Thomas Gleixner <tglx@linutronix.de> wrote:
> On Mon, 2006-10-23 at 11:44 -0700, Mark Knecht wrote:
<SNIP>
>
> >    In 2.6.18-rt6 I turned on HRT support, left 1000 nanoseconds for
> > the timing, but did not enable dynamic ticks since I wasn't sure it
> > was OK on AMD64. Should I be using DynTicks with an AMD64 single
> > processor? With a dual-processor?
>
> Should work
>
<SNIP>

I turned on DynTicks. When I rebooted with the new kernel I got a lot
of fsck messages about file systems times being in the future and
being fixed. Is this a 1-time deal?

Other than that no problems. dmesg appears to be clean so far.

- Mark

Re: -rt7 announcement? (was Re: 2.6.18-rt6)

[Lee Revell]

On Mon, 2006-10-23 at 20:35 +0200, Thomas Gleixner wrote:
> On Mon, 2006-10-23 at 12:34 -0400, Lee Revell wrote:
> > On Fri, 2006-10-20 at 11:00 -0400, Lee Revell wrote:
> > > On Wed, 2006-10-18 at 10:39 +0200, Ingo Molnar wrote:
> > > > i've released the 2.6.18-rt6 tree, which can be downloaded from the 
> > > > usual place:
> > > > 
> > > >   http://redhat.com/~mingo/realtime-preempt/
> > > 
> > > This does not work here.  It boots but then wants to fsck my disks, and
> > > dies with a sig 11 in fsck.ext3.  This is 100% reproducible and booting
> > > 2.6.18-rt5 works and does not want to fsck the disks. 
> > 
> > I see that -rt7 is posted.  The patch is a huge diff from -rt6.  Where
> > are the release notes?
> 
> Basically we merged the latest hrt-dyntick queue into -rt.

Thanks.  -rt7 boots and seems to work on this machine where -rt6 did
not.

Is the bug where the NMI watchdog caused lockups thought to be fixed?

Lee

Re: -rt7 announcement? (was Re: 2.6.18-rt6)

[Thomas Gleixner]

On Mon, 2006-10-23 at 14:49 -0400, Lee Revell wrote:
> > > I see that -rt7 is posted.  The patch is a huge diff from -rt6.  Where
> > > are the release notes?
> > 
> > Basically we merged the latest hrt-dyntick queue into -rt.
> 
> Thanks.  -rt7 boots and seems to work on this machine where -rt6 did
> not.
> 
> Is the bug where the NMI watchdog caused lockups thought to be fixed?

Yep,

	tglx

Re: -rt7 announcement? (was Re: 2.6.18-rt6)

[Sergio Monteiro Basto]

[-- Attachment #1: Type: text/plain, Size: 887 bytes --]

On Mon, 2006-10-23 at 20:35 +0200, Thomas Gleixner wrote:
> On Mon, 2006-10-23 at 12:34 -0400, Lee Revell wrote:
> > On Fri, 2006-10-20 at 11:00 -0400, Lee Revell wrote:
> > > On Wed, 2006-10-18 at 10:39 +0200, Ingo Molnar wrote:
> > > > i've released the 2.6.18-rt6 tree, which can be downloaded from the 
> > > > usual place:
> > > > 
> > > >   http://redhat.com/~mingo/realtime-preempt/
> > > 
> > > This does not work here.  It boots but then wants to fsck my disks, and
> > > dies with a sig 11 in fsck.ext3.  This is 100% reproducible and booting
> > > 2.6.18-rt5 works and does not want to fsck the disks. 
> > 
> > I see that -rt7 is posted.  The patch is a huge diff from -rt6.  Where
> > are the release notes?
> 
> Basically we merged the latest hrt-dyntick queue into -rt.


rt7 should be to be applied on 2.6.18.1 
still for 2.6.18 

-- 
Sérgio M.B.

[-- Attachment #2: smime.p7s --]
[-- Type: application/x-pkcs7-signature, Size: 2166 bytes --]

Re: -rt7 announcement? (was Re: 2.6.18-rt6)

[Lee Revell]

On Mon, 2006-10-23 at 21:26 +0100, Sergio Monteiro Basto wrote:
> rt7 should be to be applied on 2.6.18.1 
> still for 2.6.18 
> 

The -rt patch has always been against the most recent base kernel.  It
could be rebased against -stable but that would be more work for the
maintainers...

Lee

Re: -rt7 announcement? (was Re: 2.6.18-rt6) and more info about a compile error

[Sergio Monteiro Basto]

[-- Attachment #1: Type: text/plain, Size: 1105 bytes --]

On Mon, 2006-10-23 at 16:40 -0400, Lee Revell wrote:
> On Mon, 2006-10-23 at 21:26 +0100, Sergio Monteiro Basto wrote:
> > rt7 should be to be applied on 2.6.18.1 
> > still for 2.6.18 
> > 
> 
> The -rt patch has always been against the most recent base kernel.  It
> could be rebased against -stable but that would be more work for the
> maintainers...
 
For me the most recent stable kernel is 2.6.18.1. 
Normally change for .1 are very small but in this case I got 1, just 1,
reject which I don't know to fix and prefer don't try it. My luck is the
rej in a sparc arch and I can ignore it.

I got this compile error if I don't use CONFIG_PREEMPT_BKL in .config

kernel/rtmutex.c:938:48: error: macro "rt_release_bkl" passed 2
arguments, but takes just 1
kernel/rtmutex.c: In function 'rt_mutex_slowlock':
kernel/rtmutex.c:938: error: 'rt_release_bkl' undeclared (first use in
this function)
kernel/rtmutex.c:938: error: (Each undeclared identifier is reported
only once
kernel/rtmutex.c:938: error: for each function it appears in.)


Thanks, 
> Lee
> 
-- 
Sérgio M.B.

[-- Attachment #2: smime.p7s --]
[-- Type: application/x-pkcs7-signature, Size: 2166 bytes --]

Re: -rt7 announcement? (was Re: 2.6.18-rt6) and more info about a compile error

[Thomas Gleixner]

On Tue, 2006-10-24 at 02:26 +0100, Sergio Monteiro Basto wrote:
> On Mon, 2006-10-23 at 16:40 -0400, Lee Revell wrote:
> > On Mon, 2006-10-23 at 21:26 +0100, Sergio Monteiro Basto wrote:
> > > rt7 should be to be applied on 2.6.18.1 
> > > still for 2.6.18 
> > > 
> > 
> > The -rt patch has always been against the most recent base kernel.  It
> > could be rebased against -stable but that would be more work for the
> > maintainers...
>  
> For me the most recent stable kernel is 2.6.18.1. 
> Normally change for .1 are very small but in this case I got 1, just 1,
> reject which I don't know to fix and prefer don't try it. My luck is the
> rej in a sparc arch and I can ignore it.

-rtXX is always against 2.6.X, never against the .stable versions.

> I got this compile error if I don't use CONFIG_PREEMPT_BKL in .config
> 
> kernel/rtmutex.c:938:48: error: macro "rt_release_bkl" passed 2
> arguments, but takes just 1
> kernel/rtmutex.c: In function 'rt_mutex_slowlock':
> kernel/rtmutex.c:938: error: 'rt_release_bkl' undeclared (first use in
> this function)
> kernel/rtmutex.c:938: error: (Each undeclared identifier is reported
> only once
> kernel/rtmutex.c:938: error: for each function it appears in.)

Fix below.

	tglx

Index: linux-2.6.18/kernel/rtmutex.c
===================================================================
--- linux-2.6.18.orig/kernel/rtmutex.c	2006-10-24 06:33:02.000000000 +0200
+++ linux-2.6.18/kernel/rtmutex.c	2006-10-24 06:31:55.000000000 +0200
@@ -902,7 +902,10 @@ static inline void rt_reacquire_bkl(int 
 }
 
 #else
-# define rt_release_bkl(x)	(-1)
+static inline int rt_release_bkl(struct rt_mutex *lock, unsigned long flags)
+{
+	return -1;
+}
 # define rt_reacquire_bkl(x)	do { } while (0)
 #endif
 

Re: -rt7 announcement? (was Re: 2.6.18-rt6) and more info about a compile error

[Daniel Walker]

Is this new functionality? I thought -rt7 was just moving to the -mm'ish
hrtimers. What's changing the rtmutex?

Daniel

On Tue, 2006-10-24 at 06:38 +0200, Thomas Gleixner wrote:
> Index: linux-2.6.18/kernel/rtmutex.c
> ===================================================================
> --- linux-2.6.18.orig/kernel/rtmutex.c	2006-10-24 06:33:02.000000000 +0200
> +++ linux-2.6.18/kernel/rtmutex.c	2006-10-24 06:31:55.000000000 +0200
> @@ -902,7 +902,10 @@ static inline void rt_reacquire_bkl(int 
>  }
>  
>  #else
> -# define rt_release_bkl(x)	(-1)
> +static inline int rt_release_bkl(struct rt_mutex *lock, unsigned long flags)
> +{
> +	return -1;
> +}
>  # define rt_reacquire_bkl(x)	do { } while (0)
>  #endif
>  
> 
>