Skip to contentSkip to conversation

random: fix 32MB+ reads from /dev/urandom

PATCH In review Alexey Dobriyan 3 participants 3 messages Jul 23, 2014 – Aug 9, 2014 Random number driver

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 random: fix 32MB+ reads from /dev/urandom

https://bugzilla.kernel.org/show_bug.cgi?id=80981

Steps to reproduce:

strace dd if=/dev/urandom of=/dev/null bs=64M count=1

Before, dd did 1 read() syscall and returned full length. Now, dd stops at 32 MB.

---

	STABLE needs 79a8468747c5f95ed3d5ce8376a3e82e0c5857fc
	and this patch.
 drivers/char/random.c |   24 ++++++++++++++++++------
 1 file changed, 18 insertions(+), 6 deletions(-)
drivers/char/random.c +18 -6
@@ -1377,20 +1377,32 @@ random_read(struct file *file, char __user *buf, size_t nbytes, loff_t *ppos)
 }
 
 static ssize_t
-urandom_read(struct file *file, char __user *buf, size_t nbytes, loff_t *ppos)
+urandom_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
 {
-	int ret;
+	ssize_t ret;
 
 	if (unlikely(nonblocking_pool.initialized == 0))
 		printk_once(KERN_NOTICE "random: %s urandom read "
 			    "with %d bits of entropy available\n",
 			    current->comm, nonblocking_pool.entropy_total);
 
-	nbytes = min_t(size_t, nbytes, INT_MAX >> (ENTROPY_SHIFT + 3));
-	ret = extract_entropy_user(&nonblocking_pool, buf, nbytes);
+	ret = 0;
+	while (count > 0) {
+		size_t count1;
+		ssize_t ret1;
+
+		count1 = min_t(size_t, count, INT_MAX >> (ENTROPY_SHIFT + 3));
+		ret1 = extract_entropy_user(&nonblocking_pool, buf, count1);
+		if (ret1 < 0)
+			return ret1;
 
-	trace_urandom_read(8 * nbytes, ENTROPY_BITS(&nonblocking_pool),
-			   ENTROPY_BITS(&input_pool));
+		trace_urandom_read(8 * count1, ENTROPY_BITS(&nonblocking_pool),
+				   ENTROPY_BITS(&input_pool));
+
+		buf += ret1;
+		count -= ret1;
+		ret += ret1;
+	}
 	return ret;
 }
 
maintainer
#1 lore raw

I don't think this patch is worthwhile. There is no legitimate reason to be using such large reads like this, and read(2) is always allowed to return a short read (for example, if a signal comes in while the read is happening). The only difference is that we are now guaranteed to return a short-read if you try to read more than 32MB. Given that there is no valid reason to ask for more than 256 or maybe 512 bytes at a time, I'm not particularly worried about this.

If you really care, you can ask dd keep reading after a short read by using "iflag=fullblock".

- Ted

#2 lore raw
Quoted 5 lines · Theodore Ts'o

I don't think this patch is worthwhile. There is no legitimate reason to be using such large reads like this, and read(2) is always allowed to return a short read (for example, if a signal comes in while the read is happening). The only difference is that we are now guaranteed to return a short-read if you try to read more than 32MB. Given

Regression from previous versions would be the reason. I'm pretty sure someone has that dd in the scripts somewhere. (And signals are not an issue in dd case.)
Pavel

Signature · 2 lines
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html