| Message ID | 20170916130034.17706-1-Jason@zx2c4.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
Am Samstag, 16. September 2017, 15:00:34 CEST schrieb Jason A. Donenfeld: Hi Jason, > This started out as just replacing the use of crypto/rng with > get_random_bytes_wait, This change is a challenge. The use of the kernel crypto API's DRNG has been made to allow FIPS 140-2 compliance. Otherwise, the entire key generation logic will not be using the right(TM) DRNG. Thus, I would not suggest to replace that for a stable tree. Note, I am currently working on a pluggable DRNG apporach for /dev/random and /dev/urandom to be able to get rid of the use of the kernel crypto API's DRNG API. It is ready and I will air that solution shortly. Yet, it needs work to be integrated upstream (and approval from Ted Tso). Ciao Stephan
On Mon, Sep 18, 2017 at 10:49:56AM +0200, Stephan Mueller wrote: > Am Samstag, 16. September 2017, 15:00:34 CEST schrieb Jason A. Donenfeld: > > Hi Jason, > > > This started out as just replacing the use of crypto/rng with > > get_random_bytes_wait, > > This change is a challenge. The use of the kernel crypto API's DRNG has been > made to allow FIPS 140-2 compliance. Otherwise, the entire key generation > logic will not be using the right(TM) DRNG. Thus, I would not suggest to > replace that for a stable tree. Why not? What is the issue here, there is only one "DRNG" in the kernel now (and probably for a long time...) > Note, I am currently working on a pluggable DRNG apporach for /dev/random and > /dev/urandom to be able to get rid of the use of the kernel crypto API's DRNG > API. It is ready and I will air that solution shortly. Yet, it needs work to > be integrated upstream (and approval from Ted Tso). We don't postpone work for potential future patches that might or might not ever happen or get merged. That's how NetBSD died... thanks, greg k-h
Am Montag, 18. September 2017, 11:04:55 CEST schrieb Greg KH: Hi Greg, > On Mon, Sep 18, 2017 at 10:49:56AM +0200, Stephan Mueller wrote: > > Am Samstag, 16. September 2017, 15:00:34 CEST schrieb Jason A. Donenfeld: > > > > Hi Jason, > > > > > This started out as just replacing the use of crypto/rng with > > > get_random_bytes_wait, > > > > This change is a challenge. The use of the kernel crypto API's DRNG has > > been made to allow FIPS 140-2 compliance. Otherwise, the entire key > > generation logic will not be using the right(TM) DRNG. Thus, I would not > > suggest to replace that for a stable tree. > > Why not? An SP800-90A-compliant DRNG must be used in those circumstances. > What is the issue here, there is only one "DRNG" in the kernel > now (and probably for a long time...) There are more DRNGs implemented in the kernel crypto API (see crypto/drbg.c or crypto/ansi-cprng.c). > > > Note, I am currently working on a pluggable DRNG apporach for /dev/random > > and /dev/urandom to be able to get rid of the use of the kernel crypto > > API's DRNG API. It is ready and I will air that solution shortly. Yet, it > > needs work to be integrated upstream (and approval from Ted Tso). > > We don't postpone work for potential future patches that might or might > not ever happen or get merged. That's how NetBSD died... Then I would recommend to leave it as is or hear complaints from other users. Ciao Stephan
Hello Stephan, I realize you have your Linux RNG project, which is a very worthwhile goal that I support you in. I hope you're eventually successful, and I apologize for not being more available in the last 2.5 months to chime in on that patchset thread you posted. However, your message to this thread here is a completely inappropriate and nonsensical hijacking, which makes me entirely question your overall judgement. (David -- please don't let such hijacking derail or delay these critical vulnerability fixes from landing.) > This change is a challenge. The use of the kernel crypto API's DRNG has been > made to allow FIPS 140-2 compliance. Otherwise, the entire key generation > logic will not be using the right(TM) DRNG. Thus, I would not suggest to > replace that for a stable tree. Complete and utter nonsense. This patch has a history (this is already v6), and it's pretty obvious from prior discussion and conclusion that the only reason "crypto/rng" was used for this is because the original author just had no idea what he was doing (thus leading to using ECB as well). Besides, from what RNG do you think the seed for that was generated? > Note, I am currently working on a pluggable DRNG apporach for /dev/random and > /dev/urandom to be able to get rid of the use of the kernel crypto API's DRNG > API. It is ready and I will air that solution shortly. Good to hear you're still working on that project. > Yet, it needs work to > be integrated upstream (and approval from Ted Tso). Good luck with getting approval... While Ted and I have our differences like any two kernel developers, I really tend agree with him in his attitude about this FIPS silliness. It's unlikely you're going to be able to shovel this stuff into random.c, and I think doing so will undermine your entire LRNG effort. > An SP800-90A-compliant DRNG must be used in those circumstances. Again, complete and utter unsubstantiated nonsense. > Then I would recommend to leave it as is or hear complaints from other users. What a poor lack of judgement. I get it -- this FIPS compliance backwardness is something you're keen about. But don't start hijacking every thread that involves randomness with a demand to replace calls to get_random_bytes with wrappers in outdated, flawed, government compliance crypto API key expanders. From a cryptographic point of view it's a ridiculous demand. And from an engineering point of view, it's a ridiculous demand too, for if get_random_bytes already returned FIPS-compliant randomness, you wouldn't be writing this email here. Instead, if you want your FIPS garbage in the Linux RNG, take your battle for that over to random.c. I'll oppose you to the end on it, but at the very least it will the the appropriate venue for that discussion. In the meantime, stop hijacking this thread. Thanks, Jason PS: apologies for what's probably perceived as an unfriendly and overly hostile tone of this email. It's not my intention to alienate you, but I do feel necessary in these circumstances to write as directly as possible.
On Mon, Sep 18, 2017 at 01:24:18PM +0200, Jason A. Donenfeld wrote: > Good luck with getting approval... While Ted and I have our > differences like any two kernel developers, I really tend agree with > him in his attitude about this FIPS silliness. It's unlikely you're > going to be able to shovel this stuff into random.c, and I think doing > so will undermine your entire LRNG effort. Let me add one more reason why FIPS compliance for the kernel is just ***stupid***. The way FIPS compliance works, you have to pay hundreds of thousands of dollars to a FIPS certification lab to certify a specific binary, complete with the exact build environment (compiler, binutils, etc.) used to build that kernel binary. The moment you need to make a change --- say, to fix a critical zero-day security bug --- this breaks the FIPS certification, and you then have to go back to the FIPS certification lab, and pay another hundreds of thousands of dollars for another certification. This will take weeks/months, and while you are waiting for the results to come back from the FIPS certification lab, the hackers will be busy extracting another 143 million credit histories, or another 4.1 million SF-86 Security Clearance Forms from the systems involved. :-) You might say that FIPS certification != FIPS compliance. Sure, but the only silly people who care about FIPS compliance also need FIPS certification, for the US Goverment signoff. Realistically, people who need FIPS certification will need to use FIPS certified crypto in hardware. In which case the FIPS certified RNG, as well as the FIPS certified crypto, will all be in a single certified lump of hardware, which doesn't have to change when we need to fix various kernel bugs. Cheers, - Ted
On Tue, Sep 19, 2017 at 9:39 AM, Theodore Ts'o <tytso@mit.edu> wrote: > On Mon, Sep 18, 2017 at 01:24:18PM +0200, Jason A. Donenfeld wrote: >> Good luck with getting approval... While Ted and I have our >> differences like any two kernel developers, I really tend agree with >> him in his attitude about this FIPS silliness. ... > > Let me add one more reason why FIPS compliance for the kernel is just > ***stupid***. The way FIPS compliance works, you have to pay hundreds > of thousands of dollars to a FIPS certification lab to certify a > specific binary, complete with the exact build environment (compiler, > binutils, etc.) used to build that kernel binary. > > The moment you need to make a change --- say, to fix a critical > zero-day security bug --- this breaks the FIPS certification, ... > > You might say that FIPS certification != FIPS compliance. Sure, but > the only silly people who care about FIPS compliance also need FIPS > certification, for the US Goverment signoff. I do not think it is just the US that matters here. If I understand Stefan correctly, one of his concerns is German (or EU?) gov't standards that are somehow related. I'm very hazy on details. I emphatically agree with Ted on some points here. Making FIPS certification a goal for kernel development would be really dumb. Having multiple RNGs available & compile-time options to select among them also looks silly to me; we just need one good one. On the other hand, I do not see why the driver should not use a FIPS-compliant PRNG where it can. This would make things easier for anyone who does seek certification. One of the big distro vendors? A gov't department or contractor that wants to use Linux? A corporation's sys admin or security reviewer? Stefan? I cannot see much downside to this. Is the current PRNG more efficient? More easily maintained? Is it still the case that random(4) must use only hashes, not ciphers, to avoid restrictions under export laws?
On Tue, Sep 19, 2017 at 9:04 PM, Sandy Harris <sandyinchina@gmail.com> wrote:
> On the other hand, I do not see why the driver should not
From my perspective, this discussion is over. I'll be carrying out
David's requested patchset changes and submitting it. If you'd like
different cryptography, you'll have to find somebody else to do that
work for you.
On Tue, Sep 19, 2017 at 03:04:29PM -0400, Sandy Harris wrote: > On the other hand, I do not see why the driver should not > use a FIPS-compliant PRNG where it can. This would make > things easier for anyone who does seek certification. One > of the big distro vendors? A gov't department or contractor > that wants to use Linux? A corporation's sys admin or > security reviewer? Stefan? First, making it easier for a sysadmin to seek certification is creating an attractive nuisance. That means that after said company sinks $100,000+ into getting a certification, they will be hesitant to take the kernel update to fix that zero-day bug, less it causes them to lose that certification. Secondly, I've worked with a defense contractor wanting to use (and did use) Linux. Specifically, on the Zumwalt class destroyer, DD-21, although back when I worked on it was the DD(X) program. I can assure you the fact that /dev/random wasn't FIPS certified wasn't a problem with either Raytheon or the US Navy. Really. If you're really serious about crypto, and you do government work, it will be type 1 ciphers implemented in hardware, courtesy of the NSA. So really. You can use Linux without getting FIPS certification. Lots of copies of Linux are used in the government already, without FIPS certification. > I cannot see much downside to this. Is the current PRNG > more efficient? More easily maintained? Is it still the case > that random(4) must use only hashes, not ciphers, to > avoid restrictions under export laws? Linux is now using a Chacha20 based random number generator, much like OpenBSD. It's stream cipher-based CSPRNG, which is much more efficient than a block cipher or HMAC based DRBG. Unfortunately, it's also not one of the types defined in NIST 800-90A rev 1. - Ted
diff --git a/security/keys/Kconfig b/security/keys/Kconfig index a7a23b5541f8..91eafada3164 100644 --- a/security/keys/Kconfig +++ b/security/keys/Kconfig @@ -45,10 +45,8 @@ config BIG_KEYS bool "Large payload keys" depends on KEYS depends on TMPFS - depends on (CRYPTO_ANSI_CPRNG = y || CRYPTO_DRBG = y) select CRYPTO_AES - select CRYPTO_ECB - select CRYPTO_RNG + select CRYPTO_GCM help This option provides support for holding large keys within the kernel (for example Kerberos ticket caches). The data may be stored out to diff --git a/security/keys/big_key.c b/security/keys/big_key.c index 6acb00f6f22c..36682686f8c2 100644 --- a/security/keys/big_key.c +++ b/security/keys/big_key.c @@ -1,5 +1,6 @@ /* Large capacity key type * + * Copyright (C) 2017 Jason A. Donenfeld <Jason@zx2c4.com>. All Rights Reserved. * Copyright (C) 2013 Red Hat, Inc. All Rights Reserved. * Written by David Howells (dhowells@redhat.com) * @@ -16,10 +17,10 @@ #include <linux/shmem_fs.h> #include <linux/err.h> #include <linux/scatterlist.h> +#include <linux/random.h> #include <keys/user-type.h> #include <keys/big_key-type.h> -#include <crypto/rng.h> -#include <crypto/skcipher.h> +#include <crypto/aead.h> /* * Layout of key payload words. @@ -49,7 +50,12 @@ enum big_key_op { /* * Key size for big_key data encryption */ -#define ENC_KEY_SIZE 16 +#define ENC_KEY_SIZE 32 + +/* + * Authentication tag length + */ +#define ENC_AUTHTAG_SIZE 16 /* * big_key defined keys take an arbitrary string as the description and an @@ -64,57 +70,62 @@ struct key_type key_type_big_key = { .destroy = big_key_destroy, .describe = big_key_describe, .read = big_key_read, + /* no ->update(); don't add it without changing big_key_crypt() nonce */ }; /* - * Crypto names for big_key data encryption + * Crypto names for big_key data authenticated encryption */ -static const char big_key_rng_name[] = "stdrng"; -static const char big_key_alg_name[] = "ecb(aes)"; +static const char big_key_alg_name[] = "gcm(aes)"; /* - * Crypto algorithms for big_key data encryption + * Crypto algorithms for big_key data authenticated encryption */ -static struct crypto_rng *big_key_rng; -static struct crypto_skcipher *big_key_skcipher; +static struct crypto_aead *big_key_aead; /* - * Generate random key to encrypt big_key data + * Since changing the key affects the entire object, we need a mutex. */ -static inline int big_key_gen_enckey(u8 *key) -{ - return crypto_rng_get_bytes(big_key_rng, key, ENC_KEY_SIZE); -} +static DEFINE_MUTEX(big_key_aead_lock); /* * Encrypt/decrypt big_key data */ static int big_key_crypt(enum big_key_op op, u8 *data, size_t datalen, u8 *key) { - int ret = -EINVAL; + int ret; struct scatterlist sgio; - SKCIPHER_REQUEST_ON_STACK(req, big_key_skcipher); - - if (crypto_skcipher_setkey(big_key_skcipher, key, ENC_KEY_SIZE)) { + struct aead_request *aead_req; + /* We always use a zero nonce. The reason we can get away with this is + * because we're using a different randomly generated key for every + * different encryption. Notably, too, key_type_big_key doesn't define + * an .update function, so there's no chance we'll wind up reusing the + * key to encrypt updated data. Simply put: one key, one encryption. + */ + u8 zero_nonce[crypto_aead_ivsize(big_key_aead)]; + + aead_req = aead_request_alloc(big_key_aead, GFP_KERNEL); + if (!aead_req) + return -ENOMEM; + + memset(zero_nonce, 0, sizeof(zero_nonce)); + sg_init_one(&sgio, data, datalen + (op == BIG_KEY_ENC ? ENC_AUTHTAG_SIZE : 0)); + aead_request_set_crypt(aead_req, &sgio, &sgio, datalen, zero_nonce); + aead_request_set_callback(aead_req, CRYPTO_TFM_REQ_MAY_SLEEP, NULL, NULL); + aead_request_set_ad(aead_req, 0); + + mutex_lock(&big_key_aead_lock); + if (crypto_aead_setkey(big_key_aead, key, ENC_KEY_SIZE)) { ret = -EAGAIN; goto error; } - - skcipher_request_set_tfm(req, big_key_skcipher); - skcipher_request_set_callback(req, CRYPTO_TFM_REQ_MAY_SLEEP, - NULL, NULL); - - sg_init_one(&sgio, data, datalen); - skcipher_request_set_crypt(req, &sgio, &sgio, datalen, NULL); - if (op == BIG_KEY_ENC) - ret = crypto_skcipher_encrypt(req); + ret = crypto_aead_encrypt(aead_req); else - ret = crypto_skcipher_decrypt(req); - - skcipher_request_zero(req); - + ret = crypto_aead_decrypt(aead_req); error: + mutex_unlock(&big_key_aead_lock); + aead_request_free(aead_req); return ret; } @@ -146,16 +157,13 @@ int big_key_preparse(struct key_preparsed_payload *prep) * * File content is stored encrypted with randomly generated key. */ - size_t enclen = ALIGN(datalen, crypto_skcipher_blocksize(big_key_skcipher)); + size_t enclen = datalen + ENC_AUTHTAG_SIZE; loff_t pos = 0; - /* prepare aligned data to encrypt */ data = kmalloc(enclen, GFP_KERNEL); if (!data) return -ENOMEM; - memcpy(data, prep->data, datalen); - memset(data + datalen, 0x00, enclen - datalen); /* generate random key */ enckey = kmalloc(ENC_KEY_SIZE, GFP_KERNEL); @@ -163,13 +171,10 @@ int big_key_preparse(struct key_preparsed_payload *prep) ret = -ENOMEM; goto error; } - - ret = big_key_gen_enckey(enckey); - if (ret) - goto err_enckey; + get_random_bytes_wait(enckey, ENC_KEY_SIZE); /* encrypt aligned data */ - ret = big_key_crypt(BIG_KEY_ENC, data, enclen, enckey); + ret = big_key_crypt(BIG_KEY_ENC, data, datalen, enckey); if (ret) goto err_enckey; @@ -195,7 +200,7 @@ int big_key_preparse(struct key_preparsed_payload *prep) *path = file->f_path; path_get(path); fput(file); - kfree(data); + kzfree(data); } else { /* Just store the data in a buffer */ void *data = kmalloc(datalen, GFP_KERNEL); @@ -211,9 +216,9 @@ int big_key_preparse(struct key_preparsed_payload *prep) err_fput: fput(file); err_enckey: - kfree(enckey); + kzfree(enckey); error: - kfree(data); + kzfree(data); return ret; } @@ -227,7 +232,7 @@ void big_key_free_preparse(struct key_preparsed_payload *prep) path_put(path); } - kfree(prep->payload.data[big_key_data]); + kzfree(prep->payload.data[big_key_data]); } /* @@ -259,7 +264,7 @@ void big_key_destroy(struct key *key) path->mnt = NULL; path->dentry = NULL; } - kfree(key->payload.data[big_key_data]); + kzfree(key->payload.data[big_key_data]); key->payload.data[big_key_data] = NULL; } @@ -295,7 +300,7 @@ long big_key_read(const struct key *key, char __user *buffer, size_t buflen) struct file *file; u8 *data; u8 *enckey = (u8 *)key->payload.data[big_key_data]; - size_t enclen = ALIGN(datalen, crypto_skcipher_blocksize(big_key_skcipher)); + size_t enclen = datalen + ENC_AUTHTAG_SIZE; loff_t pos = 0; data = kmalloc(enclen, GFP_KERNEL); @@ -328,7 +333,7 @@ long big_key_read(const struct key *key, char __user *buffer, size_t buflen) err_fput: fput(file); error: - kfree(data); + kzfree(data); } else { ret = datalen; if (copy_to_user(buffer, key->payload.data[big_key_data], @@ -344,47 +349,31 @@ long big_key_read(const struct key *key, char __user *buffer, size_t buflen) */ static int __init big_key_init(void) { - struct crypto_skcipher *cipher; - struct crypto_rng *rng; int ret; - rng = crypto_alloc_rng(big_key_rng_name, 0, 0); - if (IS_ERR(rng)) { - pr_err("Can't alloc rng: %ld\n", PTR_ERR(rng)); - return PTR_ERR(rng); - } - - big_key_rng = rng; - - /* seed RNG */ - ret = crypto_rng_reset(rng, NULL, crypto_rng_seedsize(rng)); - if (ret) { - pr_err("Can't reset rng: %d\n", ret); - goto error_rng; - } - /* init block cipher */ - cipher = crypto_alloc_skcipher(big_key_alg_name, 0, CRYPTO_ALG_ASYNC); - if (IS_ERR(cipher)) { - ret = PTR_ERR(cipher); + big_key_aead = crypto_alloc_aead(big_key_alg_name, 0, CRYPTO_ALG_ASYNC); + if (IS_ERR(big_key_aead)) { + ret = PTR_ERR(big_key_aead); pr_err("Can't alloc crypto: %d\n", ret); - goto error_rng; + return ret; + } + ret = crypto_aead_setauthsize(big_key_aead, ENC_AUTHTAG_SIZE); + if (ret < 0) { + pr_err("Can't set crypto auth tag len: %d\n", ret); + goto free_aead; } - - big_key_skcipher = cipher; ret = register_key_type(&key_type_big_key); if (ret < 0) { pr_err("Can't register type: %d\n", ret); - goto error_cipher; + goto free_aead; } return 0; -error_cipher: - crypto_free_skcipher(big_key_skcipher); -error_rng: - crypto_free_rng(big_key_rng); +free_aead: + crypto_free_aead(big_key_aead); return ret; }
This started out as just replacing the use of crypto/rng with get_random_bytes_wait, so that we wouldn't use bad randomness at boot time. But, upon looking further, it appears that there were even deeper underlying cryptographic problems, and that this seems to have been committed with very little crypto review. So, I rewrote the whole thing, trying to keep to the conventions introduced by the previous author, to fix these cryptographic flaws. It makes no sense to seed crypto/rng at boot time and then keep using it like this, when in fact there's already get_random_bytes_wait, which can ensure there's enough entropy and be a much more standard way of generating keys. Since this sensitive material is being stored untrusted, using ECB and no authentication is simply not okay at all. I find it surprising and a bit horrifying that this code even made it past basic crypto review, which perhaps points to some larger issues. This patch moves from using AES-ECB to using AES-GCM. Since keys are uniquely generated each time, we can set the nonce to zero. There was also a race condition in which the same key would be reused at the same time in different threads. A mutex fixes this issue now. And, some error paths forgot to zero out sensitive material, so this patch changes a kfree into a kzfree. So, to summarize, this commit fixes the following vulnerabilities: * Low entropy key generation, allowing an attacker to potentially guess or predict keys. * Unauthenticated encryption, allowing an attacker to modify the cipher text in particular ways in order to manipulate the plaintext, which is is even more frightening considering the next point. * Use of ECB mode, allowing an attacker to trivially swap blocks or compare identical plaintext blocks. * Key re-use. * Faulty memory zeroing. Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com> Cc: David Howells <dhowells@redhat.com> Cc: Eric Biggers <ebiggers3@gmail.com> Cc: Herbert Xu <herbert@gondor.apana.org.au> Cc: Kirill Marinushkin <k.marinushkin@gmail.com> Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org> Cc: Ilhan Gurel <ilhan.gurel@gmail.com> Cc: security@kernel.org Cc: stable@vger.kernel.org --- security/keys/Kconfig | 4 +- security/keys/big_key.c | 139 ++++++++++++++++++++++-------------------------- 2 files changed, 65 insertions(+), 78 deletions(-) Changes v3->v4: - Patchset resurrected from the dead. Just like the original use of ECB, which this patch fixes, crypto-related things are sometimes neglected, unfortunately. - Rebased on top of 4.13+ changes. - Now that get_random_bytes_wait has been merged, we can actually use it, just like in v1 of this patch.