wolfsshd: load PKCS#8 PEM host keys (e.g. ML-DSA)#1118
Conversation
|
Can one of the admins verify this patch? |
|
See ZD 22185 |
ejohnstown
left a comment
ejohnstown
left a comment
There was a problem hiding this comment.
I can't accept that commit from Claude. Can you squash it into the other commit and force-push it again, please?
There was a problem hiding this comment.
Pull request overview
This PR updates wolfsshd host-key loading so it can successfully load PKCS#8 PEM private keys (-----BEGIN PRIVATE KEY-----), including the format produced for ML-DSA keys, while preserving existing DER host-key behavior.
Changes:
- Detect host-key input as DER vs PEM by content and decode PEM using
wc_KeyPemToDer()(supports PKCS#1/SEC1/PKCS#8). - Continue passing DER bytes unchanged to
wolfSSH_CTX_UsePrivateKey_buffer()as ASN.1. - Add cleanup for the temporary DER buffer used during PEM conversion.
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
wolfSSL-Fenrir-bot
left a comment
wolfSSL-Fenrir-bot
left a comment
There was a problem hiding this comment.
Fenrir Automated Review — PR #1118
Scan targets checked: wolfssh-bugs, wolfssh-src
No new issues found in the changed files. ✅
6d6373e to
af83564
Compare
wolfsshd could not load a host key stored as a PKCS#8 "-----BEGIN PRIVATE KEY-----" PEM file (the form emitted for ML-DSA keys, and by `openssl genpkey`). SetupCTX() pre-converted the file with wc_PemToDer(..., PRIVATEKEY_TYPE, ...), which only recognizes the classic "RSA/EC PRIVATE KEY" headers. On a PKCS#8 body that call returns success but produces a malformed DER (leading 0x04 rather than a 0x30 SEQUENCE); the bytes were then passed to wolfSSH_CTX_UsePrivateKey_buffer() as WOLFSSH_FORMAT_ASN1, which rejected them with WS_BAD_FILETYPE_E. Detect PEM vs DER by content and decode PEM with wc_KeyPemToDer(), which handles PKCS#1, SEC1 and PKCS#8 private-key bodies. DER keys are unchanged. Tested (master + wolfSSL master): an ML-DSA-65 PEM host key now loads (wolfSSH_CTX_UsePrivateKey_buffer ret = 0, was WS_BAD_FILETYPE_E); ML-DSA-65 DER and ECDSA SEC1 PEM host keys continue to load (no regression). Also harden the host-key load path per review: reject an empty (0-byte) key file explicitly instead of taking the PEM branch into WMALLOC(0), and zeroize the raw file buffer (not just the decoded DER) before freeing so private-key material does not linger in the heap.
af83564 to
ac50255
Compare
|
Squashed to a single commit, and addressed both Copilot findings in it: empty (0-byte) host-key files now return a file error before the PEM path (no WMALLOC(0)), and the raw key buffer is zeroized before free alongside the decoded DER. Thanks for the patience! |
wolfSSL-Fenrir-bot
left a comment
wolfSSL-Fenrir-bot
left a comment
There was a problem hiding this comment.
Fenrir Automated Review — PR #1118
Scan targets checked: wolfssh-bugs, wolfssh-src
No new issues found in the changed files. ✅
wolfsshdcould not load a host key stored as a PKCS#8-----BEGIN PRIVATE KEY-----PEM file (the form emitted for ML-DSA keys, and byopenssl genpkey).SetupCTX()pre-converted the file withwc_PemToDer(..., PRIVATEKEY_TYPE, ...), which only recognizes the classicRSA/EC PRIVATE KEYheaders. On a PKCS#8 body that call returns success but produces a malformed DER (leading0x04rather than a0x30SEQUENCE); the bytes were then passed towolfSSH_CTX_UsePrivateKey_buffer()asWOLFSSH_FORMAT_ASN1, which rejected them withWS_BAD_FILETYPE_E(-1016).Detect PEM vs DER by content and decode PEM with
wc_KeyPemToDer(), which handles PKCS#1, SEC1 and PKCS#8 private-key bodies. DER keys are unchanged.Testing
Built against wolfSSL master; same
wolfsshdbinary run against three host keys with the patch toggled on and off, reading thewolfSSH_CTX_UsePrivateKey_buffer()return from-D -doutput:ret = -1016, "Failed to use host private key"ret = 0ret = 0ret = 0(no regression)ret = 0ret = 0(no regression)The ML-DSA PEM key was generated with
wolfssl genkey ml-dsa -level 3 -output priv -outform PEM.No
ChangeLog.mdentry is included — the top section is the released v1.5.0; happy to add one wherever you'd like it. Suggested wording: "Fixed loading PKCS#8 PEM host keys in wolfSSHd (e.g. ML-DSA)."