Thread (33 messages) 33 messages, 7 authors, 2013-01-03

[PATCH RESEND 6/6] clk: s5p-g2d: Fix incorrect usage of IS_ERR_OR_NULL

From: Dan Carpenter <hidden>
Date: 2013-01-03 09:05:54
Also in: kernel-janitors, linux-media, lkml

On Wed, Jan 02, 2013 at 06:31:53PM +1300, Tony Prisk wrote:
On Wed, 2013-01-02 at 08:10 +0300, Dan Carpenter wrote:
quoted
clk_get() returns NULL if CONFIG_HAVE_CLK is disabled.

I told Tony about this but everyone has been gone with end of year
holidays so it hasn't been addressed.

Tony, please fix it so people don't apply these patches until
clk_get() is updated to not return NULL.  It sucks to have to revert
patches.

regards,
dan carpenter
I posted the query to Mike Turquette, linux-kernel and linux-arm-kernel
mailing lists, regarding the return of NULL when HAVE_CLK is undefined.

Short Answer: A return value of NULL is valid and not an error therefore
we should be using IS_ERR, not IS_ERR_OR_NULL on clk_get results.

I see the obvious problem this creates, and asked this question:

If the driver can't operate with a NULL clk, it should use a
IS_ERR_OR_NULL test to test for failure, rather than IS_ERR.


And Russell's answer:

Why should a _consumer_ of a clock care?  It is _very_ important that
people get this idea - to a consumer, the struct clk is just an opaque
cookie.  The fact that it appears to be a pointer does _not_ mean that
the driver can do any kind of dereferencing on that pointer - it should
never do so.

Thread can be viewed here:
https://lkml.org/lkml/2012/12/20/105
Ah.  Grand.  Thanks...

Btw. The documentation for clk_get() really should include some of
this information.  I know Russell thinks that the driver authors are
stupid and lazy, and it's probably true.  But if everyone makes the
same mistake over and over, then it probably means we could put a
special note:

"Do not check this with IS_ERR_OR_NULL().  Null values are not an
error.  Drivers should treat the return value as an opaque cookie
and they should not dereference it."

This is probably there in the file somewhere else, but I searched
for "opaque", "cookie", and "dereference" and I didn't find
anything.  I'm not saying the documentation isn't perfect, just that
driver authors are lazy and stupid but we can't kill them so we have
to live with them.

regards,
dan carpenter
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help