Linux cryptographic layer development
 help / color / mirror / Atom feed
From: Dominik Brodowski <linux@dominikbrodowski.net>
To: "Jason A. Donenfeld" <Jason@zx2c4.com>
Cc: Theodore Ts'o <tytso@mit.edu>,
	LKML <linux-kernel@vger.kernel.org>,
	Linux Crypto Mailing List <linux-crypto@vger.kernel.org>
Subject: Re: [PATCH 2/2] random: only call crng_finalize_init() for primary_crng
Date: Mon, 31 Jan 2022 17:55:52 +0100	[thread overview]
Message-ID: <YfgUmAYoaxmlWe54@owl.dominikbrodowski.net> (raw)
In-Reply-To: <CAHmME9rqoweMb5zO95B1QsAHRFkSmcumbJKZTA+tqqRHN3mzGw@mail.gmail.com>

Am Sun, Jan 30, 2022 at 11:11:22PM +0100 schrieb Jason A. Donenfeld:
> Thanks, I'll apply this. I do wonder, though, do we have locking
> concerns around crng_init transitioning from 1 to 2, or with calls to
> crng_need_final_init? For example, can crng_reseed be called at the
> same time as rand_initialize? Or are we still single core at this
> point in the boot sequence? I don't think that this patch changes
> anything from that perspective, which is why it seems reasonable to
> apply, but I do wonder.

Well, the comment

	 * crng_init is protected by primary_crng->lock

is currently not adhered to. It's unproblematic to set it at
rand_initialize() time (by calling crng_finalize_init()), as the system
is still running with IRQs disabled and only the boot CPU active (but
not yet in PID 1). So its call to crng_finalize_init() will not race
with crng_reseed() calling crng_finalize_init().

However, I think the other sites setting crng_init
	- crng_reseed() calling crng_finalize_init()
	- crng_fast_load()
might race, in particular two parallel calls to crng_reseed(). So let's
try to keep the promise to increase[*] crng_init only while holding
primary_crng->lock. UNTESTED, not even compile-tested patch below.

What do you think?

Thanks,
	Dominik

[*] The read sites still need to be checked, but at a first glance, I did
not notice any obvious problematic code.

diff --git a/drivers/char/random.c b/drivers/char/random.c
index 7ed910c23858..e21c73cadcc2 100644
--- a/drivers/char/random.c
+++ b/drivers/char/random.c
@@ -465,7 +465,7 @@ static struct crng_state primary_crng = {
  * its value (from 0->1->2).
  */
 static int crng_init = 0;
-static bool crng_need_final_init = false;
+static bool crng_needs_numa_init = false;
 #define crng_ready() (likely(crng_init > 1))
 static int crng_init_cnt = 0;
 static unsigned long crng_global_init_time = 0;
@@ -788,31 +788,29 @@ static void crng_initialize_secondary(struct crng_state *crng)
 	crng->init_time = jiffies - CRNG_RESEED_INTERVAL - 1;
 }
 
+static void crng_finalize_init(void)
+{
+	invalidate_batched_entropy();
+	/* We can't call numa_crng_init() until we have workqueues,
+	 * but we will pick this up in rand_initialize() */
+	if (system_wq)
+		numa_crng_init();
+	else
+		crng_needs_numa_init = true;
+	crng_init = 2;
+}
+
 static void __init crng_initialize_primary(void)
 {
 	_extract_entropy(&primary_crng.state[4], sizeof(u32) * 12);
 	if (crng_init_try_arch_early() && trust_cpu && crng_init < 2) {
-		invalidate_batched_entropy();
-		numa_crng_init();
-		crng_init = 2;
+		crng_finalize_init();
 		pr_notice("crng init done (trusting CPU's manufacturer)\n");
 	}
 	primary_crng.init_time = jiffies - CRNG_RESEED_INTERVAL - 1;
 }
 
-static void crng_finalize_init(void)
-{
-	if (!system_wq) {
-		/* We can't call numa_crng_init until we have workqueues,
-		 * so mark this for processing later. */
-		crng_need_final_init = true;
-		return;
-	}
-
-	invalidate_batched_entropy();
-	numa_crng_init();
-	crng_init = 2;
-	crng_need_final_init = false;
+static void crng_late_init(void) {
 	process_random_ready_list();
 	wake_up_interruptible(&crng_init_wait);
 	kill_fasync(&fasync, SIGIO, POLL_IN);
@@ -896,12 +894,13 @@ static size_t crng_fast_load(const u8 *cp, size_t len)
 		p[crng_init_cnt % CHACHA_KEY_SIZE] ^= *cp;
 		cp++; crng_init_cnt++; len--; ret++;
 	}
-	spin_unlock_irqrestore(&primary_crng.lock, flags);
 	if (crng_init_cnt >= CRNG_INIT_CNT_THRESH) {
 		invalidate_batched_entropy();
 		crng_init = 1;
-		pr_notice("fast init done\n");
 	}
+	spin_unlock_irqrestore(&primary_crng.lock, flags);
+	if (crng_init == 1)
+		pr_notice("fast init done\n");
 	return ret;
 }
 
@@ -954,6 +953,7 @@ static void crng_reseed(struct crng_state *crng, bool use_input_pool)
 {
 	unsigned long flags;
 	int i, num;
+	bool needs_late_init = false;
 	union {
 		u8 block[CHACHA_BLOCK_SIZE];
 		u32 key[8];
@@ -978,9 +978,17 @@ static void crng_reseed(struct crng_state *crng, bool use_input_pool)
 	}
 	memzero_explicit(&buf, sizeof(buf));
 	WRITE_ONCE(crng->init_time, jiffies);
-	spin_unlock_irqrestore(&crng->lock, flags);
-	if (crng == &primary_crng && crng_init < 2)
+	if (crng == &primary_crng && crng_init < 2) {
 		crng_finalize_init();
+		/* crng_late_init() is only needed if crng_init progresses to 2
+		 * after rand_initialize(). Note that while userspace may reset
+		 * crng_global_init_time to 0, it cannot reset crng_init to 2 */
+		if (crng_global_init_time > 0)
+			needs_late_init = true;
+	}
+	spin_unlock_irqrestore(&crng->lock, flags);
+	if (needs_late_init)
+		crng_late_init();
 }
 
 static void _extract_crng(struct crng_state *crng, u8 out[CHACHA_BLOCK_SIZE])
@@ -1696,8 +1704,8 @@ static void __init init_std_data(void)
 int __init rand_initialize(void)
 {
 	init_std_data();
-	if (crng_need_final_init)
-		crng_finalize_init();
+	if (crng_needs_numa_init)
+		numa_crng_init();
 	crng_initialize_primary();
 	crng_global_init_time = jiffies;
 	if (ratelimit_disable) {



  reply	other threads:[~2022-01-31 16:56 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-01-30 21:03 [PATCH 1/2] random: access primary_pool directly rather than through pointer Dominik Brodowski
2022-01-30 21:03 ` [PATCH 2/2] random: only call crng_finalize_init() for primary_crng Dominik Brodowski
2022-01-30 22:11   ` Jason A. Donenfeld
2022-01-31 16:55     ` Dominik Brodowski [this message]
2022-01-30 22:04 ` [PATCH 1/2] random: access primary_pool directly rather than through pointer Jason A. Donenfeld

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=YfgUmAYoaxmlWe54@owl.dominikbrodowski.net \
    --to=linux@dominikbrodowski.net \
    --cc=Jason@zx2c4.com \
    --cc=linux-crypto@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=tytso@mit.edu \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox