From: Pali Rohár pali@kernel.org
[ Upstream commit 6510ef4230b68c960309e0c1d6eb3e32eb785142 ]
SMB1 Session Setup NTLMSSP Request in non-UNICODE mode is similar to UNICODE mode, just strings are encoded in ASCII and not in UTF-16.
With this change it is possible to setup SMB1 session with NTLM authentication in non-UNICODE mode with Windows SMB server.
This change fixes mounting SMB1 servers with -o nounicode mount option together with -o sec=ntlmssp mount option (which is the default sec=).
Signed-off-by: Pali Rohár pali@kernel.org Signed-off-by: Steve French stfrench@microsoft.com Signed-off-by: Sasha Levin sashal@kernel.org ---
Based on my analysis of both the commit and the kernel source code, I can now provide a comprehensive assessment: **YES** This commit should be backported to stable kernel trees. Here's my extensive analysis: ## Bug Fix Analysis This commit fixes a **functional bug** that prevents SMB1 sessions from being established when using specific mount options. The code changes address a critical logic error in the NTLMSSP session setup for non-UNICODE mode. ### Problem Being Fixed The original code in `_sess_auth_rawntlmssp_assemble_req()` contains this problematic check: ```c if ((pSMB->req.hdr.Flags2 & SMBFLG2_UNICODE) == 0) { cifs_dbg(VFS, "NTLMSSP requires Unicode support\n"); return -ENOSYS; } ``` This check **incorrectly rejects** all non-UNICODE sessions for NTLMSSP authentication, making it impossible to mount SMB1 servers with: - `-o nounicode` (non-UNICODE mode) - `-o sec=ntlmssp` (default security mode) ### Technical Solution The fix removes this incorrect validation and implements proper dual-path handling: ```c if (pSMB->req.hdr.Flags2 & SMBFLG2_UNICODE) { /bin /bin.usr-is-merged /boot /dev /etc /home /init /lib /lib.usr-is-merged /lib64 /lost+found /media /mnt /opt /proc /root /run /sbin /sbin.usr-is-merged /snap /srv /sys /tmp /usr /var unicode strings must be word aligned */ if (!IS_ALIGNED(sess_data->iov[0].iov_len + sess_data->iov[1].iov_len, 2)) { *bcc_ptr = 0; bcc_ptr++; } unicode_oslm_strings(&bcc_ptr, sess_data->nls_cp); } else { ascii_oslm_strings(&bcc_ptr, sess_data->nls_cp); } ``` This correctly handles both UNICODE and non- UNICODE modes by: 1. **UNICODE path**: Uses UTF-16 encoding with proper alignment 2. **Non-UNICODE path**: Uses ASCII encoding without alignment requirements ### Why This Qualifies for Backporting 1. **User-Affecting Bug**: The bug prevents legitimate SMB1 connections with common mount options, causing mount failures that users would expect to work. 2. **Small, Contained Fix**: The change is minimal (removes 4 lines, adds 8 lines) and confined to a single function. Risk of regression is very low. 3. **No Architectural Changes**: This fix doesn't modify the NTLMSSP protocol implementation or introduce new features—it simply corrects existing logic. 4. **Similar Pattern to Accepted Backports**: This closely matches "Similar Commit #1" which was marked "YES" for backporting. Both commits: - Fix SMB1 Session Setup encoding issues - Handle UNICODE vs non-UNICODE mode correctly - Extract common code patterns for consistency - Are small, targeted bugfixes 5. **Legacy Protocol Support**: While SMB1 is deprecated, it's still supported in the kernel under `CONFIG_CIFS_ALLOW_INSECURE_LEGACY`, and users who explicitly enable it deserve working functionality. ### Code Correctness Verification The fix is technically sound because: - **`ascii_oslm_strings()`** and **`unicode_oslm_strings()`** are established functions already used elsewhere in the codebase for Kerberos authentication (`sess_auth_kerberos()`) - The same dual-path pattern exists in other session setup functions - The NTLMSSP protocol itself still uses UNICODE internally (for username/domain encoding), but the session setup strings can be ASCII - The alignment logic is correctly applied only to UNICODE mode where it's needed ### Risk Assessment **Very Low Risk**: - Fix is isolated to SMB1 NTLMSSP session setup - Only affects the non-UNICODE code path that was previously broken - No changes to the core NTLMSSP authentication logic - Pattern matches existing working code in the same file This is a clear bugfix that restores intended functionality without introducing new features or architectural changes, making it an ideal candidate for stable tree backporting.
fs/smb/client/sess.c | 20 ++++++++++---------- 1 file changed, 10 insertions(+), 10 deletions(-)
diff --git a/fs/smb/client/sess.c b/fs/smb/client/sess.c index a7ece67ee9630..1d3a1a3ba90b0 100644 --- a/fs/smb/client/sess.c +++ b/fs/smb/client/sess.c @@ -1737,22 +1737,22 @@ _sess_auth_rawntlmssp_assemble_req(struct sess_data *sess_data) pSMB = (SESSION_SETUP_ANDX *)sess_data->iov[0].iov_base;
capabilities = cifs_ssetup_hdr(ses, server, pSMB); - if ((pSMB->req.hdr.Flags2 & SMBFLG2_UNICODE) == 0) { - cifs_dbg(VFS, "NTLMSSP requires Unicode support\n"); - return -ENOSYS; - } - pSMB->req.hdr.Flags2 |= SMBFLG2_EXT_SEC; capabilities |= CAP_EXTENDED_SECURITY; pSMB->req.Capabilities |= cpu_to_le32(capabilities);
bcc_ptr = sess_data->iov[2].iov_base; - /* unicode strings must be word aligned */ - if (!IS_ALIGNED(sess_data->iov[0].iov_len + sess_data->iov[1].iov_len, 2)) { - *bcc_ptr = 0; - bcc_ptr++; + + if (pSMB->req.hdr.Flags2 & SMBFLG2_UNICODE) { + /* unicode strings must be word aligned */ + if (!IS_ALIGNED(sess_data->iov[0].iov_len + sess_data->iov[1].iov_len, 2)) { + *bcc_ptr = 0; + bcc_ptr++; + } + unicode_oslm_strings(&bcc_ptr, sess_data->nls_cp); + } else { + ascii_oslm_strings(&bcc_ptr, sess_data->nls_cp); } - unicode_oslm_strings(&bcc_ptr, sess_data->nls_cp);
sess_data->iov[2].iov_len = (long) bcc_ptr - (long) sess_data->iov[2].iov_base;