KDB: SMP locking problem

[Keith Owens]

On Mon, 9 Jun 2003 15:58:04 -0400 , 
"Sanders, Richard" <Richard.Sanders at stratus.com> wrote:
>Here is a patch for  kdb/kdb_io.c which fixes a SMP race on the printf lock.
>Patch is against 4.2.
>--- linux/kdb/kdb_io.c.orig	Mon Jun  9 15:45:54 2003
>+++ linux/kdb/kdb_io.c	Mon Jun  9 15:47:08 2003
>@@ -493,10 +493,11 @@
> 	 * But if any cpu goes recursive in kdb, just print the output,
> 	 * even if it is interleaved with any other text.
> 	 */
>-	if (!KDB_STATE(PRINTF_LOCK)) {
>+	spin_lock(&kdb_printf_lock);
>+	if (!KDB_STATE(PRINTF_LOCK)) 
> 		KDB_STATE_SET(PRINTF_LOCK);
>-		spin_lock(&kdb_printf_lock);
>-	}
>+	else
>+		spin_unlock(&kdb_printf_lock);
> 
> 	diag = kdbgetintenv("LINES", &linecount);
> 	if (diag || linecount <= 1)
>@@ -592,8 +593,8 @@
> 		console_loglevel = saved_loglevel;
> 	}
> 	if (KDB_STATE(PRINTF_LOCK)) {
>-		spin_unlock(&kdb_printf_lock);
> 		KDB_STATE_CLEAR(PRINTF_LOCK);
>+		spin_unlock(&kdb_printf_lock);
> 	}
> 	if (do_longjmp)
> #ifdef KDB_HAVE_LONGJMP

Rejected.  Read the comments at the start of your patch, you just
introduced a recursion deadlock.

BTW, your patch was in quoted printable.  Like Linus, I tend to ignore
encoded patches unless they are so large that they have to be zipped.
Please send small patchs in plain text, not as encoded attachments.