dunfell patch review
cover dunfell patch review
Two additional for reveiw. This include reverting the jsoncpp update that may have introduced an abi change.
The following changes since commit 3add820b373564478591e244226704371c1d34d1:
wireshark: Update to 3.2.5 (2020-07-12 19:20:59 -0700)
are available in the Git repository at:
git://git.openembedded.org/meta-openembedded-contrib stable/dunfell-nut
http://cgit.openembedded.org/meta-openembedded-contrib/log/?h=stable/dunfell-nut
Armin Kuster (1):
Revert "jsoncpp: upgrade 1.9.2 -> 1.9.3"
Ovidiu Panait (1):
nss: Fix CVE-2020-12399
.../{jsoncpp_1.9.3.bb => jsoncpp_1.9.2.bb} | 2 +-
...e-a-fixed-length-for-DSA-exponentiat.patch | 110 ++++++++++++++++++
meta-oe/recipes-support/nss/nss_3.51.1.bb | 1 +
3 files changed, 112 insertions(+), 1 deletion(-)
rename meta-oe/recipes-devtools/jsoncpp/{jsoncpp_1.9.3.bb => jsoncpp_1.9.2.bb} (93%)
create mode 100644 meta-oe/recipes-support/nss/nss/0001-Bug-1631576-Force-a-fixed-length-for-DSA-exponentiat.patchSignature · 1 line
2.17.1
PATCH 1/2 Revert "jsoncpp: upgrade 1.9.2 -> 1.9.3"
This reverts commit 2b384c59733c437027f9b14cc32da19251efd97b.
It appears that there was a change in soname not noted in the changelog.
https://github.com/open-source-parsers/jsoncpp/commit/8b7ea09b8055df01866a5ce4142b12ed8f9f13eb
ABI change appears to have occured.
https://abi-laboratory.pro/index.php?view=timeline&l=jsoncpp
---
.../jsoncpp/{jsoncpp_1.9.3.bb => jsoncpp_1.9.2.bb} | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
rename meta-oe/recipes-devtools/jsoncpp/{jsoncpp_1.9.3.bb => jsoncpp_1.9.2.bb} (93%)meta-oe/recipes-devtools/jsoncpp/jsoncpp_1.9.2.bb renamed from meta-oe/recipes-devtools/jsoncpp/jsoncpp_1.9.3.bb +1 -1
@@ -11,7 +11,7 @@ SECTION = "libs" LICENSE = "MIT" LIC_FILES_CHKSUM = "file://LICENSE;md5=fa2a23dd1dc6c139f35105379d76df2b" -SRCREV = "6aba23f4a8628d599a9ef7fa4811c4ff6e4070e2" +SRCREV = "d2e6a971f4544c55b8e3b25cf96db266971b778f" SRC_URI = "git://github.com/open-source-parsers/jsoncpp" S = "${WORKDIR}/git"
Signature · 1 line
2.17.1
PATCH 2/2 nss: Fix CVE-2020-12399
In reply to akusterFrom: Ovidiu Panait <[email protected]>
Master (nss version 3.54) is not affected by this issue. This is a backport from nss version 3.54.
NSS has shown timing differences when performing DSA signatures, which was exploitable and could eventually leak private keys. This vulnerability affects Thunderbird < 68.9.0, Firefox < 77, and Firefox ESR < 68.9.
Upstream patch:
https://hg.mozilla.org/projects/nss/rev/daa823a4a29bcef0fec33a379ec83857429aea2e
--- ...e-a-fixed-length-for-DSA-exponentiat.patch | 110 ++++++++++++++++++ meta-oe/recipes-support/nss/nss_3.51.1.bb | 1 + 2 files changed, 111 insertions(+) create mode 100644 meta-oe/recipes-support/nss/nss/0001-Bug-1631576-Force-a-fixed-length-for-DSA-exponentiat.patch
meta-oe/recipes-support/nss/nss/0001-Bug-1631576-Force-a-fixed-length-for-DSA-exponentiat.patch new file +110 -0
@@ -0,0 +1,110 @@ +From 5942c26888ba12ad5e0d92fb62f23d7cde6dc159 Mon Sep 17 00:00:00 2001 +From: Ovidiu Panait <[email protected]> +Date: Mon, 13 Jul 2020 06:25:56 +0000 +Subject: [PATCH] Bug 1631576 - Force a fixed length for DSA exponentiation + r=pereida,bbrumley + +Differential Revision: https://phabricator.services.mozilla.com/D72011 + +Upstream-Status: Backport [https://hg.mozilla.org/projects/nss/rev/daa823a4a29bcef0fec33a379ec83857429aea2e] + +Authored-by: Robert Relyea <[email protected]> +Signed-off-by: Ovidiu Panait <[email protected]> +--- + nss/lib/freebl/dsa.c | 45 ++++++++++++++++++++++++++++++++++---------- + 1 file changed, 35 insertions(+), 10 deletions(-) + +diff --git a/nss/lib/freebl/dsa.c b/nss/lib/freebl/dsa.c +index aef3539..389c9de 100644 +--- a/nss/lib/freebl/dsa.c ++++ b/nss/lib/freebl/dsa.c +@@ -313,13 +313,14 @@ DSA_NewKeyFromSeed(const PQGParams *params, + + static SECStatus + dsa_SignDigest(DSAPrivateKey *key, SECItem *signature, const SECItem *digest, +- const unsigned char *kb) ++ const unsigned char *kbytes) + { + mp_int p, q, g; /* PQG parameters */ + mp_int x, k; /* private key & pseudo-random integer */ + mp_int r, s; /* tuple (r, s) is signature) */ + mp_int t; /* holding tmp values */ + mp_int ar; /* holding blinding values */ ++ mp_digit fuzz; /* blinding multiplier for q */ + mp_err err = MP_OKAY; + SECStatus rv = SECSuccess; + unsigned int dsa_subprime_len, dsa_signature_len, offset; +@@ -373,6 +374,7 @@ dsa_SignDigest(DSAPrivateKey *key, SECItem *signature, const SECItem *digest, + CHECK_MPI_OK(mp_init(&s)); + CHECK_MPI_OK(mp_init(&t)); + CHECK_MPI_OK(mp_init(&ar)); ++ + /* + ** Convert stored PQG and private key into MPI integers. + */ +@@ -380,14 +382,28 @@ dsa_SignDigest(DSAPrivateKey *key, SECItem *signature, const SECItem *digest, + SECITEM_TO_MPINT(key->params.subPrime, &q); + SECITEM_TO_MPINT(key->params.base, &g); + SECITEM_TO_MPINT(key->privateValue, &x); +- OCTETS_TO_MPINT(kb, &k, dsa_subprime_len); ++ OCTETS_TO_MPINT(kbytes, &k, dsa_subprime_len); ++ ++ /* k blinding create a single value that has the high bit set in ++ * the mp_digit*/ ++ if (RNG_GenerateGlobalRandomBytes(&fuzz, sizeof(mp_digit)) != SECSuccess) { ++ PORT_SetError(SEC_ERROR_NEED_RANDOM); ++ rv = SECFailure; ++ goto cleanup; ++ } ++ fuzz |= 1ULL << ((sizeof(mp_digit) * PR_BITS_PER_BYTE - 1)); + /* + ** FIPS 186-1, Section 5, Step 1 + ** + ** r = (g**k mod p) mod q + */ +- CHECK_MPI_OK(mp_exptmod(&g, &k, &p, &r)); /* r = g**k mod p */ +- CHECK_MPI_OK(mp_mod(&r, &q, &r)); /* r = r mod q */ ++ CHECK_MPI_OK(mp_mul_d(&q, fuzz, &t)); /* t = q*fuzz */ ++ CHECK_MPI_OK(mp_add(&k, &t, &t)); /* t = k+q*fuzz */ ++ /* length of t is now fixed, bits in k have been blinded */ ++ CHECK_MPI_OK(mp_exptmod(&g, &t, &p, &r)); /* r = g**t mod p */ ++ /* r is now g**(k+q*fuzz) == g**k mod p */ ++ CHECK_MPI_OK(mp_mod(&r, &q, &r)); /* r = r mod q */ ++ + /* + ** FIPS 186-1, Section 5, Step 2 + ** +@@ -411,15 +427,24 @@ dsa_SignDigest(DSAPrivateKey *key, SECItem *signature, const SECItem *digest, + /* Using mp_invmod on k directly would leak bits from k. */ + CHECK_MPI_OK(mp_mul(&k, &ar, &k)); /* k = k * ar */ + CHECK_MPI_OK(mp_mulmod(&k, &t, &q, &k)); /* k = k * t mod q */ +- CHECK_MPI_OK(mp_invmod(&k, &q, &k)); /* k = k**-1 mod q */ ++ /* k is now k*t*ar */ ++ CHECK_MPI_OK(mp_invmod(&k, &q, &k)); /* k = k**-1 mod q */ ++ /* k is now (k*t*ar)**-1 */ + CHECK_MPI_OK(mp_mulmod(&k, &t, &q, &k)); /* k = k * t mod q */ +- SECITEM_TO_MPINT(localDigest, &s); /* s = HASH(M) */ ++ /* k is now (k*ar)**-1 */ ++ SECITEM_TO_MPINT(localDigest, &s); /* s = HASH(M) */ + /* To avoid leaking secret bits here the addition is blinded. */ +- CHECK_MPI_OK(mp_mul(&x, &ar, &x)); /* x = x * ar */ +- CHECK_MPI_OK(mp_mulmod(&x, &r, &q, &x)); /* x = x * r mod q */ ++ CHECK_MPI_OK(mp_mul(&x, &ar, &x)); /* x = x * ar */ ++ /* x is now x*ar */ ++ CHECK_MPI_OK(mp_mulmod(&x, &r, &q, &x)); /* x = x * r mod q */ ++ /* x is now x*r*ar */ + CHECK_MPI_OK(mp_mulmod(&s, &ar, &q, &t)); /* t = s * ar mod q */ +- CHECK_MPI_OK(mp_add(&t, &x, &s)); /* s = t + x */ +- CHECK_MPI_OK(mp_mulmod(&s, &k, &q, &s)); /* s = s * k mod q */ ++ /* t is now hash(M)*ar */ ++ CHECK_MPI_OK(mp_add(&t, &x, &s)); /* s = t + x */ ++ /* s is now (HASH(M)+x*r)*ar */ ++ CHECK_MPI_OK(mp_mulmod(&s, &k, &q, &s)); /* s = s * k mod q */ ++ /* s is now (HASH(M)+x*r)*ar*(k*ar)**-1 = (k**-1)*(HASH(M)+x*r) */ ++ + /* + ** verify r != 0 and s != 0 + ** mentioned as optional in FIPS 186-1. +-- +2.18.1 +
meta-oe/recipes-support/nss/nss_3.51.1.bb +1 -0
@@ -34,6 +34,7 @@ SRC_URI = "http://ftp.mozilla.org/pub/mozilla.org/security/nss/releases/${VERSIO file://0001-freebl-add-a-configure-option-to-disable-ARM-HW-cryp.patch \ file://riscv.patch \ file://0001-Enable-uint128-on-mips64.patch \ + file://0001-Bug-1631576-Force-a-fixed-length-for-DSA-exponentiat.patch \ " SRC_URI[md5sum] = "6acaf1ddff69306ae30a908881c6f233"
Signature · 1 line
2.17.1
Re: [oe] [dunfell 1/2] Revert "jsoncpp: upgrade 1.9.2 -> 1.9.3"
In reply to akustermeta-oe/recipes-devtools/jsoncpp/jsoncpp_1.9.2.bb
Quoted 4 lines · akuster
...
meta-oe/recipes-devtools/jsoncpp/jsoncpp_1.9.2.bb renamed from meta-oe/recipes-devtools/jsoncpp/jsoncpp_1.9.3.bb +0
-0
This would make the version going backwards, please add
PV = "1.9.3+really1.9.2"
cu
Adrian
Re: [oe] [dunfell 1/2] Revert "jsoncpp: upgrade 1.9.2 -> 1.9.3"
Quoted 7 lines · Adrian Bunk
meta-oe/recipes-devtools/jsoncpp/jsoncpp_1.9.2.bb
Quoted 4 lines · akuster
...
meta-oe/recipes-devtools/jsoncpp/jsoncpp_1.9.2.bb renamed from meta-oe/recipes-devtools/jsoncpp/jsoncpp_1.9.3.bb +0
-0This would make the version going backwards, please add
PV = "1.9.3+really1.9.2"
right. Wouldn't EP = "1" do the same?
Quoted 3 lines
cu
Adrian
Re: [oe] [dunfell 1/2] Revert "jsoncpp: upgrade 1.9.2 -> 1.9.3"
Quoted 4 lines · On Fri, Jul 17, 2020 at 12:16:10PM -0700, akuster808 wrote:
Quoted 1 line · Adrian Bunk
On Fri, Jul 17, 2020 at 08:20:54AM -0700, akuster wrote:
Quoted 8 lines
Quoted 6 lines
meta-oe/recipes-devtools/jsoncpp/jsoncpp_1.9.2.bb
Quoted 4 lines
...
meta-oe/recipes-devtools/jsoncpp/jsoncpp_1.9.2.bb renamed from meta-oe/recipes-devtools/jsoncpp/jsoncpp_1.9.3.bb +0
-0This would make the version going backwards, please add
PV = "1.9.3+really1.9.2"right. Wouldn't EP = "1" do the same?
PE (not EP) would also solve this.
The problem with an epoch is that it is permanent while a +really version will go away with 1.9.4.
cu
Adrian
Re: [oe] [dunfell 1/2] Revert "jsoncpp: upgrade 1.9.2 -> 1.9.3"
Quoted 4 lines · Adrian Bunk
Quoted 3 lines · On Fri, Jul 17, 2020 at 12:16:10PM -0700, akuster808 wrote:
Quoted 1 line · Adrian Bunk
On Fri, Jul 17, 2020 at 08:20:54AM -0700, akuster wrote:
Quoted 8 lines
Quoted 7 lines
Quoted 6 lines
meta-oe/recipes-devtools/jsoncpp/jsoncpp_1.9.2.bb
Quoted 4 lines
...
meta-oe/recipes-devtools/jsoncpp/jsoncpp_1.9.2.bb renamed from meta-oe/recipes-devtools/jsoncpp/jsoncpp_1.9.3.bb +0
-0This would make the version going backwards, please add
PV = "1.9.3+really1.9.2"right. Wouldn't EP = "1" do the same?
PE (not EP) would also solve this.
right.
Quoted 3 lines
The problem with an epoch is that it is permanent while a +really version will go away with 1.9.4.
I don't expect to every update to 1.9.4 as .3 may have introduced ABI changes otherwise I would not have reverted this to begin with.
-armin
Quoted 3 lines
cu
Adrian
Re: [oe] [dunfell 1/2] Revert "jsoncpp: upgrade 1.9.2 -> 1.9.3"
Quoted 13 lines · On Fri, Jul 17, 2020 at 01:57:24PM -0700, akuster808 wrote:
Quoted 10 lines · Adrian Bunk
Quoted 9 lines · On Fri, Jul 17, 2020 at 12:16:10PM -0700, akuster808 wrote:
Quoted 7 lines · Adrian Bunk
meta-oe/recipes-devtools/jsoncpp/jsoncpp_1.9.2.bb
Quoted 4 lines · akuster
...
meta-oe/recipes-devtools/jsoncpp/jsoncpp_1.9.2.bb renamed from meta-oe/recipes-devtools/jsoncpp/jsoncpp_1.9.3.bb +0
-0This would make the version going backwards, please add
PV = "1.9.3+really1.9.2"
Quoted 8 lines
Quoted 2 lines
Quoted 1 line
right. Wouldn't EP = "1" do the same?
PE (not EP) would also solve this.
right.
Quoted 3 lines
The problem with an epoch is that it is permanent while a +really version will go away with 1.9.4.
I don't expect to every update to 1.9.4 as .3 may have introduced ABI changes otherwise I would not have reverted this to begin with.
Not in dunfell.
PE would have to stay forever in master.
cu
Adrian
Re: [oe] [dunfell 1/2] Revert "jsoncpp: upgrade 1.9.2 -> 1.9.3"
Quoted 14 lines · Adrian Bunk
Quoted 13 lines · On Fri, Jul 17, 2020 at 01:57:24PM -0700, akuster808 wrote:
Quoted 11 lines · Adrian Bunk
Quoted 9 lines · On Fri, Jul 17, 2020 at 12:16:10PM -0700, akuster808 wrote:
Quoted 7 lines · Adrian Bunk
meta-oe/recipes-devtools/jsoncpp/jsoncpp_1.9.2.bb
Quoted 4 lines · akuster
...
meta-oe/recipes-devtools/jsoncpp/jsoncpp_1.9.2.bb renamed from meta-oe/recipes-devtools/jsoncpp/jsoncpp_1.9.3.bb +0
-0This would make the version going backwards, please add
PV = "1.9.3+really1.9.2"right. Wouldn't EP = "1" do the same?
PE (not EP) would also solve this.
Quoted 8 lines
Quoted 5 lines
right.
Quoted 2 lines
The problem with an epoch is that it is permanent while a +really version will go away with 1.9.4.
I don't expect to every update to 1.9.4 as .3 may have introduced ABI changes otherwise I would not have reverted this to begin with.
Not in dunfell.
PE would have to stay forever in master.
Master did not change. Dunfell changed.
Quoted 3 lines
cu
Adrian
Re: [oe] [dunfell 1/2] Revert "jsoncpp: upgrade 1.9.2 -> 1.9.3"
In reply to Adrian BunkHi,
Quoted 8 lines · Adrian Bunk
meta-oe/recipes-devtools/jsoncpp/jsoncpp_1.9.2.bb
Quoted 4 lines · akuster
...
meta-oe/recipes-devtools/jsoncpp/jsoncpp_1.9.2.bb renamed from meta-oe/recipes-devtools/jsoncpp/jsoncpp_1.9.3.bb +0
-0This would make the version going backwards, please add
PV = "1.9.3+really1.9.2"
This will break CVE version checks unless CVE_VERSION is set to 1.9.2. Please increment PE instead.
Cheers,
-Mikko
Quoted 4 lines
cu
Adrian