[patch] fix for urandom read(2) not interruptible

[andrea () E-MIND COM (Andrea Arcangeli)]

After having read phrak54 about Linux /dev/u?random (and this is the reason
I am CCing also to bugtraq ;), I was playing a bit
with the random driver it and I noticed that was difficult to kill `dd
if=/dev/urandom of=/dev/null bs=100000k count=20000' once started ;)). The
machine was eavily loaded and the process was unkillable and I the fastest
thing to restore the system is been a reset...

It's a bug in random.c that doesn' t check for signal pending inside the
read(2) code, so you have no chance to kill the process via signals until
the read(2) syscall is finished, and it could take a lot of time before
return, if the buffer given to the read syscall is very big...

Here the fix against 2.1.132:

Index: linux/drivers/char/random.c
diff -u linux/drivers/char/random.c:1.1.1.1 linux/drivers/char/random.c:1.1.1.1.2.3
--- linux/drivers/char/random.c:1.1.1.1 Fri Nov 20 00:02:25 1998
+++ linux/drivers/char/random.c Sun Dec 27 20:19:16 1998
@@ -232,6 +232,11 @@
  * Eastlake, Steve Crocker, and Jeff Schiller.
  */

+/*
+ * Added a check for signal pending in the extract_entropy() loop to allow
+ * the read(2) syscall to be interrupted. Copyright (C) 1998  Andrea Arcangeli
+ */
+
 #include <linux/utsname.h>
 #include <linux/config.h>
 #include <linux/kernel.h>
@@ -1269,7 +1274,14 @@
                buf += i;
                add_timer_randomness(r, &extract_timer_state, nbytes);
                if (to_user && current->need_resched)
+               {
+                       if (signal_pending(current))
+                       {
+                               ret = -EINTR;
+                               break;
+                       }
                        schedule();
+               }
        }

        /* Wipe data just returned from memory */


And here a fix against 2.0.36:

--- linux/drivers/char/random.c.orig    Sun Dec 27 20:22:53 1998
+++ linux/drivers/char/random.c Sun Dec 27 20:24:17 1998
@@ -226,6 +226,11 @@
  * Eastlake, Steve Crocker, and Jeff Schiller.
  */

+/*
+ * Added a check for signal pending in the extract_entropy() loop to allow
+ * the read(2) syscall to be interrupted. Copyright (C) 1998  Andrea Arcangeli
+ */
+
 #include <linux/config.h> /* CONFIG_RST_COOKIES and CONFIG_SYN_COOKIES */
 #include <linux/utsname.h>
 #include <linux/kernel.h>
@@ -1004,7 +1009,14 @@
                buf += i;
                add_timer_randomness(r, &extract_timer_state, nbytes);
                if (to_user && need_resched)
+               {
+                       if (signal_pending(current))
+                       {
+                               ret = -EINTR;
+                               break;
+                       }
                        schedule();
+               }
        }

        /* Wipe data from memory */

Andrea Arcangeli