Re:[PATCH] eth0: Error -16 writing packet header to BAP

Re:[PATCH] eth0: Error -16 writing packet header to BAP

[Jim Houston]

Hi Brian,

I believe the patch which follows will fix your "Error -16..."
problem with the orinoco_cs driver.

I sent the patch to David Gibson a couple weeks of ago but have
not heard from him yet. 

Jim Houston

--

Hi David,

The patch which follows "fixes" a problem I'm seeing with the
orinoco_cs wireless driver.  The problem results in the error
"Error -110 writing packet header to BAP," and the connection locks up.
This happens almost immediately with the link active. I see this problem
with old "Wavelan Bronze" cards which the driver reports as "Firmware ID 1F
vendor 0x1 (Lucent) version 3.06."  I'm assuming that the problem
is a feature of this old hardware/firmware.  

The change is to some existing error handling (avoidance?) code
which retries if it detects an error setting a BAP offset.  My change
also does this retry if the operation times out.  I looked at the
wvlan_cs driver (which just works) to see if it had a similar
workaround.  It doesn't.

--
Jim Houston


diff -urN -X /home/jim/dontdiff linux-2.4.7-orig/drivers/net/wireless/hermes.c linux-2.4.7/drivers/net/wireless/hermes.c
--- linux-2.4.7-orig/drivers/net/wireless/hermes.c      Tue Jul 31 22:55:00 2001
+++ linux-2.4.7/drivers/net/wireless/hermes.c   Tue Jul 31 22:34:09 2001
@@ -53,9 +53,9 @@
 #include <stdarg.h>
 
 #define DMSG(stuff...) do {printk(KERN_DEBUG "hermes @ 0x%x: " , hw->iobase); \
-                       printk(#stuff);} while (0)
+                       printk(##stuff);} while (0)
 
-#define DEBUG(lvl, stuff...) if ( (lvl) <= HERMES_DEBUG) DMSG(#stuff)
+#define DEBUG(lvl, stuff...) if ( (lvl) <= HERMES_DEBUG) DMSG(##stuff)
 
 #else /* ! HERMES_DEBUG */
 
@@ -330,20 +330,17 @@
                reg = hermes_read_reg(hw, oreg);
        }
 
-       if (reg & HERMES_OFFSET_BUSY) {
-               DEBUG(0,"hermes_bap_seek: returning ETIMEDOUT...\n");
-               return -ETIMEDOUT;
-       }
-
        /* For some reason, seeking the BAP seems to randomly fail somewhere
           (firmware bug?). We retry a few times before giving up. */
-       if (reg & HERMES_OFFSET_ERR) {
-               if (l--) {
-                       udelay(1);
-                       goto retry;
-               } else
-                       return -EIO;
+       if ( (reg & (HERMES_OFFSET_BUSY | HERMES_OFFSET_ERR)) && l--) {
+               udelay(1);
+               goto retry;
        }
+       if (reg & HERMES_OFFSET_BUSY)
+               return -ETIMEDOUT;
+
+       if (reg & HERMES_OFFSET_ERR)
+               return -EIO;
 
        return 0;
 }