d@nny disc@ mc²so then you find the commit that set it off
git show 91db696adea4d76017b1e1f45915a5cbf04e8da3
commit 91db696adea4d76017b1e1f45915a5cbf04e8da3
Author: David Howells <[email protected]>
Date: Mon Jan 26 11:46:58 2026 +0000
pkcs7: Allow authenticatedAttributes for ML-DSA
Allow the rejection of authenticatedAttributes in PKCS#7 (signedAttrs in
CMS) to be waived in the kernel config for ML-DSA when used for module
signing. This reflects the issue that openssl < 4.0 cannot do this and
openssl-4 has not yet been released.
This does not permit RSA, ECDSA or ECRDSA to be so waived (behaviour
unchanged).
(1) when people blame a project for not having functionality that's one thing
(2) when this fuckin guy says yeah backdoor in the kernel sounds like the right way to solve this that's another thing
(3) READ this shit "Allow authenticatedAttributes" => "Allow the rejection of authenticatedAttributes"
(4) WHEN USED FOR MODULE SIGNING!!!!!!!!!!
(5) oh i bet i'm about to see some behaviour that's so unchanged
It's tempting to search for a clever cryptographic solution, but nobody has yet proposed one.
not a cryptographic problem, fuckboy. diffoscope is the problem. 'oh i can calculate granular diffs but can't ignore a specified list of files"
+config PKCS7_WAIVE_AUTHATTRS_REJECTION_FOR_MLDSA
+ bool "Waive rejection of authenticatedAttributes for ML-DSA"
+ depends on PKCS7_MESSAGE_PARSER
+ depends on CRYPTO_MLDSA
+ help
+ Due to use of CMS_NOATTR with ML-DSA not being supported in
+ OpenSSL < 4.0 (and thus any released version), enabling this
+ allows authenticatedAttributes to be used with ML-DSA for
+ module signing. Use of authenticatedAttributes in this
+ context is normally rejected.
why is this considered a normal thing to push
Use of authenticatedAttributes in this context is normally rejected.
do people think when they pull in kernel kconfig they learn about new ways to hack people and then try to run this and get hacked
do you think linus has done the full kconfig cannonball run recently
i have
insane shit
+#ifdef CONFIG_PKCS7_WAIVE_AUTHATTRS_REJECTION_FOR_MLDSA
+ bool authattrs_rej_waivable; /* T if authatts rejection can be waived */
+#endif
"authatts" what is the purpose of this typo besides signaling a typo
this one is low-effort. get a real job
git show 965e9a2cf23b066d8bdeb690dff9cd7089c5f667
commit 965e9a2cf23b066d8bdeb690dff9cd7089c5f667 (tag: keys-next-20260206, dhowells/keys-next)
Author: David Howells <[email protected]>
Date: Thu Feb 5 13:02:41 2026 +0000
pkcs7: Change a pr_warn() to pr_warn_once()
Only display the "PKCS7: Waived invalid module sig (has authattrs)" once.
Suggested-by: Lenny Szubowicz <[email protected]>
Signed-off-by: David Howells <[email protected]>
Tested-by: Lenny Szubowicz <[email protected]>
cc: Lukas Wunner <[email protected]>
cc: Ignat Korchagin <[email protected]>
cc: Jarkko Sakkinen <[email protected]>
cc: Stephan Mueller <[email protected]>
cc: Eric Biggers <[email protected]>
cc: Herbert Xu <[email protected]>
cc: [email protected]
cc: [email protected]
diff --git a/crypto/asymmetric_keys/pkcs7_verify.c b/crypto/asymmetric_keys/pkcs7_verify.c
index 519eecfe6778..474e2c1ae21b 100644
--- a/crypto/asymmetric_keys/pkcs7_verify.c
+++ b/crypto/asymmetric_keys/pkcs7_verify.c
@@ -427,7 +427,7 @@ int pkcs7_verify(struct pkcs7_message *pkcs7,
if (pkcs7->have_authattrs) {
#ifdef CONFIG_PKCS7_WAIVE_AUTHATTRS_REJECTION_FOR_MLDSA
if (pkcs7->authattrs_rej_waivable) {
- pr_warn("Waived invalid module sig (has authattrs)\n");
+ pr_warn_once("Waived invalid module sig (has authattrs)\n");
break;
}
#endif
linux-crypto is not........hence the linus email..........
anyway this is all still up of course
this is the one that had me reading that brain destroying nist spec
git show f3eccecd782dbaf33d5ad0d1fd22ea277300acdb
commit f3eccecd782dbaf33d5ad0d1fd22ea277300acdb
Author: David Howells <[email protected]>
Date: Wed Jan 21 12:41:42 2026 +0000
pkcs7: Allow the signing algo to do whatever digestion it wants itself
this one is SO close to sounding like a real commit message. except for all the specifics
Allow the data to be verified in a PKCS#7 or CMS message to be passed
directly to an asymmetric cipher algorithm (e.g. ML-DSA) if it wants to do
whatever passes for hashing/digestion itself.
so GODDAMN pissed he would create a capability that does not exist like this
this is, btw, one intended result of NIST's document that frames a "signature" as distinct from "hashing"
these people don't deserve to work for the government
The normal digestion of the
data is then skipped as that would be ignored unless another signed info in
the message has some other algorithm that needs it.
The normal digestion of the data is then skipped
nope! nope. not something that you would ever allow in the real world
Note that whilst ML-DSA does allow for an "external mu", CMS doesn't yet
have that standardised.
i didn't actually spend any time looking this up because i assumed it was completely made up
no fucking way am i adding an early return 0 to my real crypto because someone thinks quantum computing is real and can hurt you. pascal's wager
+ if (!sinfo->authattrs && sig->algo_takes_data) {
+ /* There's no intermediate digest and the signature algo
+ * doesn't want the data prehashing.
+ */
+ sig->m = (void *)pkcs7->data;
+ sig->m_size = pkcs7->data_len;
+ sig->m_free = false;
+ return 0;
+ }
+
oh yeah this is where i started screaming and saying this is all completely insane shit to be doing in the kernel
this would be headline on lwn if it was in a gnu project
+ if (!sinfo->sig->m_free) {
+ pr_notice_once("%s: No digest available\n", __func__);
+ return -EINVAL; /* TODO: MLDSA doesn't necessarily calculate an
+ * intermediate digest. */
+ }
none of this shit makes any sense btw. you can just type shit. IETF material
this is i think when i became convinced oh this is for real like this is gonna be released with signed modules upon 7.0
diff --git a/include/crypto/public_key.h b/include/crypto/public_key.h
index bd38ba4d217d..4c5199b20338 100644
--- a/include/crypto/public_key.h
+++ b/include/crypto/public_key.h
@@ -46,6 +46,8 @@ struct public_key_signature {
u8 *m; /* Message data to pass to verifier */
u32 s_size; /* Number of bytes in signature */
u32 m_size; /* Number of bytes in ->m */
+ bool m_free; /* T if ->m needs freeing */
+ bool algo_takes_data; /* T if public key algo operates on data, not a hash */
const char *pkey_algo;
const char *hash_algo;
const char *encoding;
the fuck are you doing in public_key.h asshole?????
i'm going backwards so this comes before the previous
git show f728074f1f577565c97e465652c3d4afb0964013
commit f728074f1f577565c97e465652c3d4afb0964013
Author: David Howells <[email protected]>
Date: Mon Jan 26 09:49:24 2026 +0000
pkcs7, x509: Rename ->digest to ->m
went into a brief fugue state
diff --git a/crypto/asymmetric_keys/asymmetric_type.c b/crypto/asymmetric_keys/asymmetric_type.c
index 348966ea2175..2326743310b1 100644
--- a/crypto/asymmetric_keys/asymmetric_type.c
+++ b/crypto/asymmetric_keys/asymmetric_type.c
@@ -593,10 +593,10 @@ static int asymmetric_key_verify_signature(struct kernel_pkey_params *params,
{
struct public_key_signature sig = {
.s_size = params->in2_len,
- .digest_size = params->in_len,
+ .m_size = params->in_len,
.encoding = params->encoding,
.hash_algo = params->hash_algo,
- .digest = (void *)in,
+ .m = (void *)in,
.s = (void *)in2,
};
this is very pythonic
diff --git a/include/crypto/public_key.h b/include/crypto/public_key.h
index 81098e00c08f..bd38ba4d217d 100644
--- a/include/crypto/public_key.h
+++ b/include/crypto/public_key.h
@@ -43,9 +43,9 @@ extern void public_key_free(struct public_key *key);
struct public_key_signature {
struct asymmetric_key_id *auth_ids[3];
u8 *s; /* Signature */
- u8 *digest;
+ u8 *m; /* Message data to pass to verifier */
u32 s_size; /* Number of bytes in signature */
- u32 digest_size; /* Number of bytes in digest */
+ u32 m_size; /* Number of bytes in ->m */
const char *pkey_algo;
const char *hash_algo;
const char *encoding;
no. no. that's not real. that's not a thing
he keeps getting his boys to rizz him up
Signed-off-by: David Howells <[email protected]>
Reviewed-by: Jarkko Sakkinen <[email protected]>
cc: Lukas Wunner <[email protected]>
cc: Ignat Korchagin <[email protected]>
cc: Stephan Mueller <[email protected]>
cc: Eric Biggers <[email protected]>
cc: Herbert Xu <[email protected]>
cc: [email protected]
cc: [email protected]
this one's where he acts like people think that going back through time in git is more invisible?
git show 2c62068ac86bdd917a12eef49ba82ec8b091208b
commit 2c62068ac86bdd917a12eef49ba82ec8b091208b
Author: David Howells <[email protected]>
Date: Wed Jan 21 13:33:37 2026 +0000
x509: Separately calculate sha256 for blacklist
Calculate the SHA256 hash for blacklisting purposes independently of the
signature hash (which may be something other than SHA256).
This is necessary because when ML-DSA is used, no digest is calculated.
Note that this represents a change of behaviour in that the hash used for
the blacklist check would previously have been whatever digest was used
for, say, RSA-based signatures. It may be that this is inadvisable.
This is necessary because when ML-DSA is used, no digest is calculated.
yeah figured that out after wasting an hour of my life
Note that this represents a change of behaviour
yeah i mean the only thing i don't know is why you would say this out loud. i was literally reading the kbuild docs should i keep going
It may be that this is inadvisable.
i can't understand the theory of this process? like you can act like the IETF in the linux kernel? the place where people do close review?
public_key.h in ring 0 like it's a fucking joke (it has to be a joke for long enough until it's real)noticed another completely insane thing https://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-fs.git/commit/?h=keys-pqc&id=d3b6dd90e23ef1b57143e60668175ecd890948d1
+config CRYPTO_MLDSA
+ tristate "ML-DSA (Module-Lattice-Based Digital Signature Algorithm)"
+ select CRYPTO_SIG
+ select CRYPTO_LIB_MLDSA
+ help
+ ML-DSA (Module-Lattice-Based Digital Signature Algorithm) (FIPS-204).
+
+ Only signature verification is implemented.
+
actually i guess that's the reason for the hackneyed openssl stuff if openssl can sign or something? but anyway only implementing verification is the weirdest way to implement asymmetric crypto
i was gonna say also the code looks wrong bc how does that implement the brain destroying nist standard but ohohoho https://www.phoronix.com/news/Linux-7.0-Crypto-ML-DSA (note: phoronix posted genuine reiser apologia like two years ago. this is not a link because phoronix is good to link to)
This ML-DSA introduction for signing within the kernel comes in the same release of Linux finally removing SHA-1 module signing support.
it is TEXTBOOK fucking IETF to pull this shit
In addition to that initial ML-DSA support is also AES library updates in the now-merged crypto library pull.
yeah that's a normal way to talk about things in a pull request
moth bitch