Re: [PULL}: latest tip/cpus4096 changes

[Ingo Molnar <mingo@elte.hu>]

* Mike Travis <travis@sgi.com> wrote:

> diff --git a/kernel/kgdb.c b/kernel/kgdb.c
> index e4dcfb2..21fde60 100644
> --- a/kernel/kgdb.c
> +++ b/kernel/kgdb.c
> @@ -72,7 +72,7 @@ struct kgdb_state {
>  static struct debuggerinfo_struct {
>  	void			*debuggerinfo;
>  	struct task_struct	*task;
> -} kgdb_info[NR_CPUS];
> +} *kgdb_info;
>  
>  /**
>   * kgdb_connected - Is a host GDB connected to us?
> @@ -1651,6 +1651,13 @@ int kgdb_register_io_module(struct kgdb_io *new_kgdb_io_ops)
>  		return -EBUSY;
>  	}
>  
> +	kgdb_info = kmalloc(nr_cpu_ids * sizeof(*kgdb_info), GFP_KERNEL);
> +	if (unlikely(!kgdb_info)) {
> +		spin_unlock(&kgdb_registration_lock);
> +		printk(KERN_ERR "kgdb: No memory for kgdb_info\n");
> +		return -ENOMEM;
> +	}
> +
>  	if (new_kgdb_io_ops->init) {
>  		err = new_kgdb_io_ops->init();
>  		if (err) {

Look how it continues:

                        spin_unlock(&kgdb_registration_lock);
                        return err;
                }
        }

See the memory leak? This is _trivially_ broken. When you add dynamic 
allocation to any codepath you _need_ to be careul and you need to check 
all interim paths of return.

Also, please submit kgdb patches via the KGDB maintainer:

 KGDB
 P:      Jason Wessel
 M:      jason.wessel@windriver.com
 L:      kgdb-bugreport@lists.sourceforge.net
 S:      Maintained

	Ingo