[U-Boot-Users] What if eth_init() fails?

[Wolfgang Denk]

Dear Upakul,

in message <bb58ac4d0711152213y4e4520ay44ae0b9d5302b9c2 at mail.gmail.com> you wrote:
> 
> Thanks for the replies. I am attaching herewith, the patch which I suppose
> should fix the issue in NetLoop().

I'm sorry, but I think this is actually not a good idea.

Tradition is that a function returns <0 (typical -1) in case of
problems, and a return code >=0 indicates success (eventually
including a useful return value).

It seems that the original call was based on that expectation, too:

--- u-boot-1.2.0_orig/net/net.c	2007-01-07 04:43:11.000000000 +0530
+++ u-boot-1.2.0/net/net.c	2007-11-14 18:03:03.000000000 +0530
@@ -305,7 +305,7 @@
 #ifdef CONFIG_NET_MULTI
 	eth_set_current();
 #endif
-	if (eth_init(bd) < 0) {

The test for  "< 0" reads to me: "if there was an error"...

+	if (eth_init(bd) > 0) {

Now this is completely misleading.


Assuming that eth_init() can only result in sucess or failure, the
test should be written either as

	if (eth_init(bd) != 0) ... /* error handling */

or be left as is:

	if (eth_init(bd) < 0) ... /* error handling */

Actually I strongly prefer the second form which perfectly matches my
internal C parser :-)

So the real fix for this problem is to change the code of eth_init()
instead and to make it return -1 in case of errors (instead of 1).

Note that all places where eth_init() is called should be checked.


And BTW: please stop posting HTML!

Best regards,

Wolfgang Denk

-- 
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
There is an order of things in this universe.
	-- Apollo, "Who Mourns for Adonais?" stardate 3468.1