[PATCH 3/3] Increase lockdep limits: MAX_LOCKDEP_CHAINS_BITS

[PATCH 3/3] Increase lockdep limits: MAX_LOCKDEP_CHAINS_BITS

[Joao Correia]

(Applies to current Linus tree, as of 2.6.31-rc2)

A third limit becomes apparent as being too low after raising
MAX_STACK_TRACE_ENTRIES and MAX_LOCK_DEPTH, although this one is more
elusive to trigger.


Signed-off-by: Joao Correia <joaomiguelcorreia@gmail.com>

---
 kernel/lockdep_internals.h | 2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/kernel/lockdep_internals.h b/kernel/lockdep_internals.h
index 699a2ac..93af1f1 100644
--- a/kernel/lockdep_internals.h
+++ b/kernel/lockdep_internals.h
@@ -56,7 +56,7 @@ enum {
  */
 #define MAX_LOCKDEP_ENTRIES    16384UL

-#define MAX_LOCKDEP_CHAINS_BITS        15
+#define MAX_LOCKDEP_CHAINS_BITS        16
 #define MAX_LOCKDEP_CHAINS     (1UL << MAX_LOCKDEP_CHAINS_BITS)

 #define MAX_LOCKDEP_CHAIN_HLOCKS (MAX_LOCKDEP_CHAINS*5)
---

Re: [PATCH 3/3] Increase lockdep limits: MAX_LOCKDEP_CHAINS_BITS

[Peter Zijlstra]

On Tue, 2009-07-07 at 16:25 +0100, Joao Correia wrote:
> (Applies to current Linus tree, as of 2.6.31-rc2)
> 
> A third limit becomes apparent as being too low after raising
> MAX_STACK_TRACE_ENTRIES and MAX_LOCK_DEPTH, although this one is more
> elusive to trigger.

Hrmm, I really wonder what takes all these chains, I never run into this
stuff. We recently fixed the dma-debug code to generate less classes,
could there be more such bugs?

> Signed-off-by: Joao Correia <joaomiguelcorreia@gmail.com>
> 
> ---
>  kernel/lockdep_internals.h | 2 +-
>  1 files changed, 1 insertions(+), 1 deletions(-)
> 
> diff --git a/kernel/lockdep_internals.h b/kernel/lockdep_internals.h
> index 699a2ac..93af1f1 100644
> --- a/kernel/lockdep_internals.h
> +++ b/kernel/lockdep_internals.h
> @@ -56,7 +56,7 @@ enum {
>   */
>  #define MAX_LOCKDEP_ENTRIES    16384UL
> 
> -#define MAX_LOCKDEP_CHAINS_BITS        15
> +#define MAX_LOCKDEP_CHAINS_BITS        16
>  #define MAX_LOCKDEP_CHAINS     (1UL << MAX_LOCKDEP_CHAINS_BITS)
> 
>  #define MAX_LOCKDEP_CHAIN_HLOCKS (MAX_LOCKDEP_CHAINS*5)
> ---

Re: [PATCH 3/3] Increase lockdep limits: MAX_LOCKDEP_CHAINS_BITS

[Peter Zijlstra]

On Tue, 2009-07-07 at 16:25 +0100, Joao Correia wrote:
> (Applies to current Linus tree, as of 2.6.31-rc2)
> 
> A third limit becomes apparent as being too low after raising
> MAX_STACK_TRACE_ENTRIES and MAX_LOCK_DEPTH, although this one is more
> elusive to trigger.

Would this involve reloading modules a lot?

Re: [PATCH 3/3] Increase lockdep limits: MAX_LOCKDEP_CHAINS_BITS

[Joao Correia]

On Wed, Jul 8, 2009 at 7:34 PM, Peter Zijlstra<a.p.zijlstra@chello.nl> wrote:
> On Tue, 2009-07-07 at 16:25 +0100, Joao Correia wrote:
>> (Applies to current Linus tree, as of 2.6.31-rc2)
>>
>> A third limit becomes apparent as being too low after raising
>> MAX_STACK_TRACE_ENTRIES and MAX_LOCK_DEPTH, although this one is more
>> elusive to trigger.
>
> Would this involve reloading modules a lot?
>
>

All the other limits were triggered immediatly upon boot. This one
happens during regular system usage, after a couple of hours. Not
loading more modules than needed, the system stays at around 45
modules loaded, give or take. I do have qemu running, if that helps
anything.

Joao Correia

Re: [PATCH 3/3] Increase lockdep limits: MAX_LOCKDEP_CHAINS_BITS

[Joao Correia]

On Wed, Jul 8, 2009 at 8:32 PM, Joao Correia<joaomiguelcorreia@gmail.com> wrote:
> On Wed, Jul 8, 2009 at 7:34 PM, Peter Zijlstra<a.p.zijlstra@chello.nl> wrote:
>> On Tue, 2009-07-07 at 16:25 +0100, Joao Correia wrote:
>>> (Applies to current Linus tree, as of 2.6.31-rc2)
>>>
>>> A third limit becomes apparent as being too low after raising
>>> MAX_STACK_TRACE_ENTRIES and MAX_LOCK_DEPTH, although this one is more
>>> elusive to trigger.
>>
>> Would this involve reloading modules a lot?
>>
>>
>
> All the other limits were triggered immediatly upon boot. This one
> happens during regular system usage, after a couple of hours. Not
> loading more modules than needed, the system stays at around 45
> modules loaded, give or take. I do have qemu running, if that helps
> anything.
>
> Joao Correia
>

qemu running as a -host-, not the system running inside it. Just to clear it up.

Joao Correia

Re: [PATCH 3/3] Increase lockdep limits: MAX_LOCKDEP_CHAINS_BITS

[Dave Jones]

On Wed, Jul 08, 2009 at 08:33:37PM +0100, Joao Correia wrote:
 > On Wed, Jul 8, 2009 at 8:32 PM, Joao Correia<joaomiguelcorreia@gmail.com> wrote:
 > > On Wed, Jul 8, 2009 at 7:34 PM, Peter Zijlstra<a.p.zijlstra@chello.nl> wrote:
 > >> On Tue, 2009-07-07 at 16:25 +0100, Joao Correia wrote:
 > >>> (Applies to current Linus tree, as of 2.6.31-rc2)
 > >>>
 > >>> A third limit becomes apparent as being too low after raising
 > >>> MAX_STACK_TRACE_ENTRIES and MAX_LOCK_DEPTH, although this one is more
 > >>> elusive to trigger.
 > >>
 > >> Would this involve reloading modules a lot?
 > >>
 > >>
 > >
 > > All the other limits were triggered immediatly upon boot. This one
 > > happens during regular system usage, after a couple of hours. Not
 > > loading more modules than needed, the system stays at around 45
 > > modules loaded, give or take. I do have qemu running, if that helps
 > > anything.
 > >
 > > Joao Correia
 > >
 > 
 > qemu running as a -host-, not the system running inside it. Just to clear it up.

Could you upload your /proc/lockdep someplace ? Maybe there's some obvious
clues in there like the ones we saw in mine.

	Dave