Re: [PATCH] Btrfs: use rcu to protect device->name V2

[Josef Bacik <josef@redhat.com>]

On Mon, Jun 11, 2012 at 02:10:34PM -0700, Zach Brown wrote:
> 
> >-		if (state->print_mask&  BTRFSIC_PRINT_MASK_SUPERBLOCK_WRITE)
> >+		if (state->print_mask&  BTRFSIC_PRINT_MASK_SUPERBLOCK_WRITE) {
> >+			struct rcu_string *name;
> >+
> >+			rcu_read_lock();
> >+			name = rcu_dereference(device->name);
> >  			printk(KERN_INFO "New initial S-block (bdev %p, %s)"
> >  			       " @%llu (%s/%llu/%d)\n",
> >-			       superblock_bdev, device->name,
> >+			       superblock_bdev, name->str,
> >  			       (unsigned long long)dev_bytenr,
> >  			       dev_state->name,
> >  			       (unsigned long long)dev_bytenr,
> >  			       superblock_mirror_num);
> >+			rcu_read_unlock();
> >+		}
> 
> That's a whole lot of noise at every call site.  How about some helpers?
> In sloppy-code..
> 
> #define printk_in_rcu(args) do { rcu; printk; rcu; } while (0)
> 
> #define device_name_rcu(dev) ({
> 	struct rcu_string *name = rcu_dereference(dev->name);
> 	name->str; )}
> 
> 	printk_in_rcu("HELLO FRIENDS %s", device_name_rcu(dev));
> 
> It's kind of annoying to have a wrapper for printk(), pr_debug(), etc,
> but it sure seems better than implementing the rcu consistency rules
> around every output.  (And multiple rcu_string locals for multiple
> device names..)
> 

Yeah I agree I just didn't feel like figuring out how to make a macro work with
printk args ;).  Thanks,

Josef