Re: [target:for-next 51/51] ERROR: "__cmpxchg_called_with_bad_pointer" undefined!

Re: [target:for-next 51/51] ERROR: "__cmpxchg_called_with_bad_pointer" undefined!

[Nicholas A. Bellinger]

On Wed, 2014-01-29 at 14:51 +0800, kbuild test robot wrote:
> tree:   git://git.kernel.org/pub/scm/linux/kernel/git/nab/target-pending.git for-next
> head:   7769401d351d54d5cbcb6400ec60c0b916e87a7e
> commit: 7769401d351d54d5cbcb6400ec60c0b916e87a7e [51/51] target: Fix percpu_ref_put race in transport_lun_remove_cmd
> config: make ARCH=mips allmodconfig
> 
> All error/warnings:
> 
> >> ERROR: "__cmpxchg_called_with_bad_pointer" undefined!
> 

So MIPS doesn't like typedef bool as 1-byte char being used for cmpxchg
-> ll/sc instructions..

Fixing this up now by making se_cmd->lun_ref_active use a single word
instead.

Thanks Fengguang!

--nab

diff --git a/include/target/target_core_base.h b/include/target/target_core_base.h
index d284186..909dacb 100644
--- a/include/target/target_core_base.h
+++ b/include/target/target_core_base.h
@@ -552,7 +552,7 @@ struct se_cmd {
        void                    *priv;
 
        /* Used for lun->lun_ref counting */
-       bool                    lun_ref_active;
+       int                     lun_ref_active;
 
        /* DIF related members */
        enum target_prot_op     prot_op;

Re: [target:for-next 51/51] ERROR: "__cmpxchg_called_with_bad_pointer" undefined!

[Ralf Baechle]

On Wed, Jan 29, 2014 at 02:01:38AM -0800, Nicholas A. Bellinger wrote:

> On Wed, 2014-01-29 at 14:51 +0800, kbuild test robot wrote:
> > tree:   git://git.kernel.org/pub/scm/linux/kernel/git/nab/target-pending.git for-next
> > head:   7769401d351d54d5cbcb6400ec60c0b916e87a7e
> > commit: 7769401d351d54d5cbcb6400ec60c0b916e87a7e [51/51] target: Fix percpu_ref_put race in transport_lun_remove_cmd
> > config: make ARCH=mips allmodconfig
> > 
> > All error/warnings:
> > 
> > >> ERROR: "__cmpxchg_called_with_bad_pointer" undefined!
> > 
> 
> So MIPS doesn't like typedef bool as 1-byte char being used for cmpxchg
> -> ll/sc instructions..
> 
> Fixing this up now by making se_cmd->lun_ref_active use a single word
> instead.

Thanks, looking good.

Note that this is a hardware restriction on LL/SC rsp. LLD/SCD which only
operate on natually aligned four rsp. eight byte operands.  Could fixed
but would slow down and inflate every invocation of cmpxchg() which is
currently an inline function and I feel a bit uneasy about hardware
behaviour when mixing LL/SC rsp. LLD/SCD with conventional loads and
stores.

  Ralf