Message-ID: <20150729175656.GE23178@dtor-ws>
Date:	Wed, 29 Jul 2015 10:56:56 -0700
From:	Dmitry Torokhov <dmitry.torokhov@...il.com>
To:	"Dr. H. Nikolaus Schaller" <hns@...delico.com>
Cc:	Rob Herring <robherring2@...il.com>,
	Belisko Marek <marek.belisko@...il.com>,
	"linux-input@...r.kernel.org" <linux-input@...r.kernel.org>,
	LKML <linux-kernel@...r.kernel.org>,
	Rob Herring <robh+dt@...nel.org>,
	Grant Likely <grant.likely@...aro.org>
Subject: Re: [PATCH] input: twl4030-vibra: Fix ERROR: Bad of_node_put()
 warning

On Wed, Jul 29, 2015 at 07:50:24PM +0200, Dr. H. Nikolaus Schaller wrote:
> 
> Am 29.07.2015 um 19:26 schrieb Dmitry Torokhov <dmitry.torokhov@...il.com>:
> 
> > On Tue, Jul 28, 2015 at 10:13:54PM -0500, Rob Herring wrote:
> >> On Tue, Jul 28, 2015 at 3:23 PM, Belisko Marek <marek.belisko@...il.com> wrote:
> >>> Hi Dmitry,
> >>> 
> >>> On Thu, Jul 23, 2015 at 10:53 PM, Dmitry Torokhov
> >>> <dmitry.torokhov@...il.com> wrote:
> >>>> On Thu, Jul 23, 2015 at 10:38:34PM +0200, Marek Belisko wrote:
> >>>>> Fix following:
> >>>>> [    8.862274] ERROR: Bad of_node_put() on /ocp/i2c@...70000/twl@...audio
> >>>>> [    8.869293] CPU: 0 PID: 1003 Comm: modprobe Not tainted 4.2.0-rc2-letux+ #1175
> >>>>> [    8.876922] Hardware name: Generic OMAP36xx (Flattened Device Tree)
> >>>>> [    8.883514] [<c00159e0>] (unwind_backtrace) from [<c0012488>] (show_stack+0x10/0x14)
> >>>>> [    8.891693] [<c0012488>] (show_stack) from [<c05cb810>] (dump_stack+0x78/0x94)
> >>>>> [    8.899322] [<c05cb810>] (dump_stack) from [<c02cfd5c>] (kobject_release+0x68/0x7c)
> >>>>> [    8.907409] [<c02cfd5c>] (kobject_release) from [<bf0040c4>] (twl4030_vibra_probe+0x74/0x188 [twl4030_vibra])
> >>>>> [    8.917877] [<bf0040c4>] (twl4030_vibra_probe [twl4030_vibra]) from [<c03816ac>] (platform_drv_probe+0x48/0x90)
> >>>>> [    8.928497] [<c03816ac>] (platform_drv_probe) from [<c037feb4>] (really_probe+0xd4/0x238)
> >>>>> [    8.937103] [<c037feb4>] (really_probe) from [<c0380160>] (driver_probe_device+0x30/0x48)
> >>>>> [    8.945678] [<c0380160>] (driver_probe_device) from [<c03801e0>] (__driver_attach+0x68/0x8c)
> >>>>> [    8.954589] [<c03801e0>] (__driver_attach) from [<c037ea60>] (bus_for_each_dev+0x50/0x84)
> >>>>> [    8.963226] [<c037ea60>] (bus_for_each_dev) from [<c037f828>] (bus_add_driver+0xcc/0x1e4)
> >>>>> [    8.971832] [<c037f828>] (bus_add_driver) from [<c0380b60>] (driver_register+0x9c/0xe0)
> >>>>> [    8.980255] [<c0380b60>] (driver_register) from [<c00097e0>] (do_one_initcall+0x100/0x1b8)
> >>>>> [    8.988983] [<c00097e0>] (do_one_initcall) from [<c00b8008>] (do_init_module+0x58/0x1c0)
> >>>>> [    8.997497] [<c00b8008>] (do_init_module) from [<c00b8cac>] (SyS_init_module+0x54/0x64)
> >>>>> [    9.005950] [<c00b8cac>] (SyS_init_module) from [<c000ed20>] (ret_fast_syscall+0x0/0x54)
> >>>>> [    9.015838] input: twl4030:vibrator as /devices/platform/68000000.ocp/48070000.i2c/i2c-0/0-0048/48070000.i2c:twl@48:audio/input/input2
> >>>>> 
> >>>>> node passed to of_find_node_by_name is put inside that function and new node
> >>>>> is returned if found. Free returned node not already freed node.
> >>>> 
> >>>> Hmm, if of_find_node_by_name() "puts" passed in node should we not "get"
> >>>> it before calling of_find_node_by_name()? The node pointer in question
> >>>> is simply copied from parent device.
> >>> I'm not sure. what I can say is that I cannot see such error in 4.1
> >>> but only in 4.2-rcx.
> >>> Adding Grant and Rob to CC, maybe they know what should be done and
> >>> why I see such error in 4.2-rcx.
> >> 
> >> The problem was that node passed into of_find_node_by_name is the the
> >> starting point to search, but you should be doing the put on the
> >> returned node. So the patch below is correct.
> >> 
> >> As far as why in 4.2, it seems you have OF_DYNAMIC enabled in your
> >> config either because you have DT unit test or overlays enabled.
> >> Overlays are now user enable-able in 4.2.
> > 
> > Right, but the question was whether we should also "get" the node that
> > we are passing into of_find_node_by_name(), or, maybe better, stop
> > of_find_node_by_name() from "putting" the node that is passed in? It is
> > really surprising behavior.
> 
> I agree that it is quite unexpected and would be much easier to understand
> if it would not put the node.
> 
> But I guess it is intended to be a convenience to make it easier to walk down
> the tree, i.e. that you can simply write
> 
> np = of_find_node_by_name(NULL, “level11”);
> np = of_find_node_by_name(np, “level2”);
> np = of_find_node_by_name(np, “level3”);

Right, also for_each_node_by_name() relies on this behavior (but we can
fix it to use something like __of_find_node_by_name_and_put_from()).

> 
> Otherwise it would need some temporary variable and explicit calls to put the
> parent level after finding a child node.
> 
> On the other hand greping for of_find_node_by_name returns many more
> calls with parent = NULL than others. So the put is ignored anyways.
> 
> But it is a major change in semantics of a very low level function so it is
> easy to introduce regressions (especially in out-of-the tree drivers).

I am not sure how much we should care about out of tree drivers; but I
worry that there are several callers that simply do:

	np = of_find_node_by_bame(dev->of_node, "blah");

in our tree...

-- 
Dmitry
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists