Changeset 3653

Timestamp:

07/04/02 01:57:05 (11 years ago)

Author:

mds

Message:

(mds)

So I should check this change in, right?

yep - should do no harm to do it the right way ;)

Simon

DI Simon Vogl wrote:

Mark D. Studebaker wrote:

Dori Eldar of Intel reports:

---------------
Speaking of i2c-algo-bit.c there's a bug in line 147: instead of
setscl(1) you need also to wait for the slave
to raise his CLK line so instead you need to cal sclhi(), which does
just
that.

---------------

Is this a bug?

It seems so, indeed. I have not come across a situation where this has played
a role, but this should be the clean way to do it - for devices that don't hold
down the scl line, it makes no difference.

Simon

Location:

i2c/trunk

Files:

• CHANGES (modified) (1 diff)
• kernel/i2c-algo-bit.c (modified) (1 diff)

i2c/trunk/CHANGES

@@ -21 +21 @@
   i2c.h: Add defines for SMBus 2.0
   i2c-algo-8xx.[ch]: Compile updates, Add GPL
+  i2c-algo-bit.c: Wait for clock low extension by slow devices
   i2c-core.c: Add SMBus 2.0 PEC support; change back to 'old' CONFIG names;
               fix i2c emulated process call

i2c/trunk/kernel/i2c-algo-bit.c

@@ -145 +145 @@
         DEBPROTO(printk(" Sr "));
         setsda(adap,1);
-        setscl(adap,1);
+        sclhi(adap);
         udelay(adap->udelay);