[dm-devel] Re: 2.4 kernel patches: dm-daemon.c & spin_lock_irq

[Brian <brian photoresearchers com>]

I had to make two modifications to the device-mapper source in order to 
get the RHEL 3 ES latest kernel (2.4.21-9.0.1) to compile with the 
latest stable device-mapper release (1.00.08).  The first was changing 
the argument to the spin_[un]lock_irq() functions from 
"&current->sigmask_lock" to "&current->sighand->siglock".  The second 
was to block the redefinition of fls() in dm-io.c, which is now defined 
in the kernel's "bitops.h".  I'll leave it to the folk more closely 
involved with the releases to determine how, if at all, these 
alterations might be incorporated.  My dm-io.c fix is pretty much a 
hack; I'm sure there's a better way to check for definitions and make 
sure that the DM code doesn't conflict with the kernel's bitops 
functions.  I've attached a patch file as well.

-Brian

--- drivers/md/dm-daemon.c.orig	2004-04-01 13:15:46.000000000 -0500
+++ drivers/md/dm-daemon.c	2004-04-05 12:44:40.000000000 -0400
@@ -19,10 +19,10 @@
 	reparent_to_init();
 
 	/* block all signals */
-	spin_lock_irq(&current->sigmask_lock);
+	spin_lock_irq(&current->sighand->siglock);
 	sigfillset(&current->blocked);
 	flush_signals(current);
-	spin_unlock_irq(&current->sigmask_lock);
+	spin_unlock_irq(&current->sighand->siglock);
 
 	strcpy(current->comm, dd->name);
 	atomic_set(&dd->please_die, 0);
--- drivers/md/dm-io.c.orig	2004-04-01 13:15:46.000000000 -0500
+++ drivers/md/dm-io.c	2004-04-05 13:00:33.000000000 -0400
@@ -178,10 +178,12 @@
 /*
  * Primitives for alignment calculations.
  */
+#ifndef fls(x)
 int fls(unsigned n)
 {
 	return generic_fls32(n);
 }
+#endif
 
 static inline int log2_floor(unsigned n)
 {