netdev - Re: [PATCH net] tls: do not return error when the tls_bigint overflows in tls_advance_record

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <ZPiXYkYewO1Z7WRN@hog>
Date: Wed, 6 Sep 2023 17:14:42 +0200
From: Sabrina Dubroca <sd@...asysnail.net>
To: Jakub Kicinski <kuba@...nel.org>
Cc: Liu Jian <liujian56@...wei.com>, borisp@...dia.com,
	john.fastabend@...il.com, davem@...emloft.net, edumazet@...gle.com,
	pabeni@...hat.com, vfedorenko@...ek.ru, netdev@...r.kernel.org
Subject: Re: [PATCH net] tls: do not return error when the tls_bigint
 overflows in tls_advance_record_sn()

2023-09-06, 08:02:31 -0700, Jakub Kicinski wrote:
> On Wed, 6 Sep 2023 13:02:37 +0200 Sabrina Dubroca wrote:
> > I've been running the selftests with async crypto and have collected a
> > few fixes that I was going to post this week (but not this one, since
> > we don't have a selftest for wrapping rec_seq). One of the patches
> > adds -EBUSY checks for all existing -EINPROGRESS, since the crypto API
> > can return -EBUSY as well if we're going through the backlog queue.
> 
> BTW is it possible to fake async crypto for a test or does one need 
> to have an actual accelerator?

That's what I did for my tests, forcing AESNI to go async. I'm going
to send my changes as RFC to linux-crypto@. I think syzbot would find
a few more bugs if they let it loose with forced async crypto.

Short version (without the debugfs toggles):

diff --git a/crypto/simd.c b/crypto/simd.c
index edaa479a1ec5..e3f3bf31fcca 100644
--- a/crypto/simd.c
+++ b/crypto/simd.c
@@ -317,7 +317,7 @@ static int simd_aead_encrypt(struct aead_request *req)
 	subreq = aead_request_ctx(req);
 	*subreq = *req;
 
-	if (!crypto_simd_usable() ||
+	if (true /* force async */ || !crypto_simd_usable() ||
 	    (in_atomic() && cryptd_aead_queued(ctx->cryptd_tfm)))
 		child = &ctx->cryptd_tfm->base;
 	else
@@ -338,7 +338,7 @@ static int simd_aead_decrypt(struct aead_request *req)
 	subreq = aead_request_ctx(req);
 	*subreq = *req;
 
-	if (!crypto_simd_usable() ||
+	if (true /* force async */ || !crypto_simd_usable() ||
 	    (in_atomic() && cryptd_aead_queued(ctx->cryptd_tfm)))
 		child = &ctx->cryptd_tfm->base;
 	else


-- 
Sabrina