On Tue, 26 Jan 2016, Krzysztof Kozlowski wrote:

On 26.01.2016 15:24, Julia Lawall wrote:

quoted

On Tue, 26 Jan 2016, Krzysztof Kozlowski wrote:

quoted

On 26.01.2016 00:21, Amitoj Kaur Chawla wrote:

quoted

for_each_available_child_of_node and for_each_child_of_node perform an
of_node_get on each iteration, so to break out of the loop an of_node_put is
required.

Found using Coccinelle. The simplified version of the semantic patch
that is used for this is as follows:

// <smpl>
@@
local idexpression n;
expression e,r;
@@

 for_each_available_child_of_node(r,n) {
   ...
(
   of_node_put(n);
|
   e = n
|
+  of_node_put(n);
?  break;
)
   ...
 }
... when != n
// </smpl>

Patch iselft looks correct but why are you pasting coccinelle script
into the message?

The script is already present in Linux kernel:
scripts/coccinelle/iterators/device_node_continue.cocci

I don't think so.  The continue one takes care of the case where there is 
an extraneous of_node_put before a continue, not a missing one before a 
break.  But OK to drop it if it doesn't seem useful.

julia

You are right - this is not covered by that cocci patch... but I think
is covered by scripts/coccinelle/iterators/fen.cocci, isn't it?

Not quite.  That is for of_node_puts after normal loop completion (not 
sure that this problem comes up any more, but at one point there were a 
number of them). There are indeed a lot of ways in which the management of 
reference counts can go wrong...

Anyway, the rule that Amitoj used seems to be pretty reliable, so I'll try 
to get it into the kernel source tree some day soon.

julia

BR,
Krzysztof

quoted

quoted

This just extends the commit message without any meaningful data so with
removal of coccinelle script above:
Reviewed-by: Krzysztof Kozlowski <redacted>

Best regards,
Krzysztof

quoted

Signed-off-by: Amitoj Kaur Chawla <redacted>
---
 drivers/media/platform/exynos4-is/media-dev.c | 12 +++++++++---
 1 file changed, 9 insertions(+), 3 deletions(-)