Skip to contentSkip to conversation

FAILED: patch "[PATCH] ksmbd: validate security descriptor offsets before pointer" failed to apply to 6.1-stable tree

PATCH Applied [email protected] 3 participants 5 messages Nov 3 – Nov 5 1 message not in the archive

Subsystems (2)
lore
Share & tools
lore

[email protected]

author maintainer
#0 lore raw

The patch below does not apply to the 6.1-stable tree.
If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <[email protected]>.

To reproduce the conflict and resubmit, you may use the following commands:

git fetch https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/ linux-6.1.y
git checkout FETCH_HEAD
git cherry-pick -x 3474e1a304fbf9f7c996b24adcd519e031227a90
# <resolve conflicts, build, test, etc.>
git commit -s
git send-email --to '<[email protected]>' --in-reply-to '2026110319-default-yelp-9aa0@gregkh' --subject-prefix 'PATCH 6.1.y' 'HEAD^..'

Possible dependencies:

thanks,

greg k-h

------------------ original commit in Linus's tree ------------------

From 3474e1a304fbf9f7c996b24adcd519e031227a90 Mon Sep 17 00:00:00 2001
From: Namjae Jeon <[email protected]>
Date: Sat, 17 Oct 2026 09:44:48 +0900
Subject: [PATCH] ksmbd: validate security descriptor offsets before pointer
arithmetic

parse_sec_desc() constructs SID and DACL pointers by adding untrusted 32-bit offsets to the security descriptor before checking their bounds. The existing checks reject offsets into the descriptor header, but do not reject offsets beyond the descriptor's length.

On a 32-bit kernel, a large offset can wrap the pointer before the descriptor. The end-only checks in parse_sid() and parse_dacl() can then accept it, allowing an authenticated SMB2 SET_INFO request to access memory outside the security descriptor and fault the kernel.

Check each nonzero offset against acl_len before constructing its pointer. Initialize the optional owner and group SID pointers to NULL so an omitted SID remains absent when parsing the DACL.

Fixes 0626e6641f6b ("cifsd: add server handler for central processing and tranport layers")
fs/smb/server/smbacl.c +11 -9
index 96428df33b43..f3b1efd8e85a 100644
@@ -900,7 +900,7 @@ int parse_sec_desc(const struct mnt_idmap *idmap, struct smb_ntsd *pntsd,
 		   int acl_len, struct smb_fattr *fattr)
 {
 	int rc = 0;
-	struct smb_sid *owner_sid_ptr, *group_sid_ptr;
+	struct smb_sid *owner_sid_ptr = NULL, *group_sid_ptr = NULL;
 	struct smb_acl *dacl_ptr; /* no need for SACL ptr */
 	char *end_of_acl = ((char *)pntsd) + acl_len;
 	__u32 dacloffset;
@@ -912,12 +912,7 @@ int parse_sec_desc(const struct mnt_idmap *idmap, struct smb_ntsd *pntsd,
 	if (acl_len < sizeof(struct smb_ntsd))
 		return -EINVAL;
 
-	owner_sid_ptr = (struct smb_sid *)((char *)pntsd +
-			le32_to_cpu(pntsd->osidoffset));
-	group_sid_ptr = (struct smb_sid *)((char *)pntsd +
-			le32_to_cpu(pntsd->gsidoffset));
 	dacloffset = le32_to_cpu(pntsd->dacloffset);
-	dacl_ptr = (struct smb_acl *)((char *)pntsd + dacloffset);
 	ksmbd_debug(SMB,
 		    "revision %d type 0x%x ooffset 0x%x goffset 0x%x sacloffset 0x%x dacloffset 0x%x\n",
 		    pntsd->revision, pntsd->type, le32_to_cpu(pntsd->osidoffset),
@@ -933,9 +928,12 @@ int parse_sec_desc(const struct mnt_idmap *idmap, struct smb_ntsd *pntsd,
 	pntsd->type = cpu_to_le16(DACL_PRESENT);
 
 	if (pntsd->osidoffset) {
-		if (le32_to_cpu(pntsd->osidoffset) < sizeof(struct smb_ntsd))
+		if (le32_to_cpu(pntsd->osidoffset) < sizeof(struct smb_ntsd) ||
+		    le32_to_cpu(pntsd->osidoffset) >= acl_len)
 			return -EINVAL;
 
+		owner_sid_ptr = (struct smb_sid *)((char *)pntsd +
+				le32_to_cpu(pntsd->osidoffset));
 		rc = parse_sid(owner_sid_ptr, end_of_acl);
 		if (rc) {
 			pr_err("%s: Error %d parsing Owner SID\n", __func__, rc);
@@ -951,9 +949,12 @@ int parse_sec_desc(const struct mnt_idmap *idmap, struct smb_ntsd *pntsd,
 	}
 
 	if (pntsd->gsidoffset) {
-		if (le32_to_cpu(pntsd->gsidoffset) < sizeof(struct smb_ntsd))
+		if (le32_to_cpu(pntsd->gsidoffset) < sizeof(struct smb_ntsd) ||
+		    le32_to_cpu(pntsd->gsidoffset) >= acl_len)
 			return -EINVAL;
 
+		group_sid_ptr = (struct smb_sid *)((char *)pntsd +
+				le32_to_cpu(pntsd->gsidoffset));
 		rc = parse_sid(group_sid_ptr, end_of_acl);
 		if (rc) {
 			pr_err("%s: Error %d mapping Owner SID to gid\n",
@@ -975,9 +976,10 @@ int parse_sec_desc(const struct mnt_idmap *idmap, struct smb_ntsd *pntsd,
 		pntsd->type |= cpu_to_le16(DACL_PROTECTED);
 
 	if (dacloffset) {
-		if (dacloffset < sizeof(struct smb_ntsd))
+		if (dacloffset < sizeof(struct smb_ntsd) || dacloffset >= acl_len)
 			return -EINVAL;
 
+		dacl_ptr = (struct smb_acl *)((char *)pntsd + dacloffset);
 		rc = parse_dacl(idmap, dacl_ptr, end_of_acl,
 				owner_sid_ptr, group_sid_ptr, fattr);
 		if (rc)
#1 lore raw

Re: Backport: FAILED: patch "[PATCH] ksmbd: validate security descriptor offsets before pointer" failed to apply to 6.1-stable tree

From: Namjae Jeon <[email protected]>

[ Upstream commit 3474e1a304fbf9f7c996b24adcd519e031227a90 ]

parse_sec_desc() constructs SID and DACL pointers by adding untrusted 32-bit offsets to the security descriptor before checking their bounds. The existing checks reject offsets into the descriptor header, but do not reject offsets beyond the descriptor's length.

On a 32-bit kernel, a large offset can wrap the pointer before the descriptor. The end-only checks in parse_sid() and parse_dacl() can then accept it, allowing an authenticated SMB2 SET_INFO request to access memory outside the security descriptor and fault the kernel.

Check each nonzero offset against acl_len before constructing its pointer. Initialize the optional owner and group SID pointers to NULL so an omitted SID remains absent when parsing the DACL.

The 6.1 tree uses a void parse_dacl(). Preserve that interface in this backport.

Fixes 0626e6641f6b ("cifsd: add server handler for central processing and tranport layers")
(cherry picked from commit 3474e1a304fbf9f7c996b24adcd519e031227a90)
---
 fs/smb/server/smbacl.c | 20 +++++++++++---------
 1 file changed, 11 insertions(+), 9 deletions(-)
fs/smb/server/smbacl.c +11 -9
index b0bf27cab..4a8812336 100644
@@ -885,7 +885,7 @@ int parse_sec_desc(struct user_namespace *user_ns, struct smb_ntsd *pntsd,
                   int acl_len, struct smb_fattr *fattr)
 {
        int rc = 0;
-       struct smb_sid *owner_sid_ptr, *group_sid_ptr;
+       struct smb_sid *owner_sid_ptr = NULL, *group_sid_ptr = NULL;
        struct smb_acl *dacl_ptr; /* no need for SACL ptr */
        char *end_of_acl = ((char *)pntsd) + acl_len;
        __u32 dacloffset;
@@ -897,12 +897,7 @@ int parse_sec_desc(struct user_namespace *user_ns, struct smb_ntsd *pntsd,
        if (acl_len < sizeof(struct smb_ntsd))
                return -EINVAL;
 
-       owner_sid_ptr = (struct smb_sid *)((char *)pntsd +
-                       le32_to_cpu(pntsd->osidoffset));
-       group_sid_ptr = (struct smb_sid *)((char *)pntsd +
-                       le32_to_cpu(pntsd->gsidoffset));
        dacloffset = le32_to_cpu(pntsd->dacloffset);
-       dacl_ptr = (struct smb_acl *)((char *)pntsd + dacloffset);
        ksmbd_debug(SMB,
                    "revision %d type 0x%x ooffset 0x%x goffset 0x%x sacloffset 0x%x dacloffset 0x%x\n",
                    pntsd->revision, pntsd->type, le32_to_cpu(pntsd->osidoffset),
@@ -918,9 +913,12 @@ int parse_sec_desc(struct user_namespace *user_ns, struct smb_ntsd *pntsd,
        pntsd->type = cpu_to_le16(DACL_PRESENT);
 
        if (pntsd->osidoffset) {
-               if (le32_to_cpu(pntsd->osidoffset) < sizeof(struct smb_ntsd))
+               if (le32_to_cpu(pntsd->osidoffset) < sizeof(struct smb_ntsd) ||
+                   le32_to_cpu(pntsd->osidoffset) >= acl_len)
                        return -EINVAL;
 
+               owner_sid_ptr = (struct smb_sid *)((char *)pntsd +
+                               le32_to_cpu(pntsd->osidoffset));
                rc = parse_sid(owner_sid_ptr, end_of_acl);
                if (rc) {
                        pr_err("%s: Error %d parsing Owner SID\n", __func__, rc);
@@ -936,9 +934,12 @@ int parse_sec_desc(struct user_namespace *user_ns, struct smb_ntsd *pntsd,
        }
 
        if (pntsd->gsidoffset) {
-               if (le32_to_cpu(pntsd->gsidoffset) < sizeof(struct smb_ntsd))
+               if (le32_to_cpu(pntsd->gsidoffset) < sizeof(struct smb_ntsd) ||
+                   le32_to_cpu(pntsd->gsidoffset) >= acl_len)
                        return -EINVAL;
 
+               group_sid_ptr = (struct smb_sid *)((char *)pntsd +
+                               le32_to_cpu(pntsd->gsidoffset));
                rc = parse_sid(group_sid_ptr, end_of_acl);
                if (rc) {
                        pr_err("%s: Error %d mapping Owner SID to gid\n",
@@ -960,9 +961,10 @@ int parse_sec_desc(struct user_namespace *user_ns, struct smb_ntsd *pntsd,
                pntsd->type |= cpu_to_le16(DACL_PROTECTED);
 
        if (dacloffset) {
-               if (dacloffset < sizeof(struct smb_ntsd))
+               if (dacloffset < sizeof(struct smb_ntsd) || dacloffset >= acl_len)
                        return -EINVAL;
 
+               dacl_ptr = (struct smb_acl *)((char *)pntsd + dacloffset);
                parse_dacl(user_ns, dacl_ptr, end_of_acl,
                           owner_sid_ptr, group_sid_ptr, fattr);
        }

--
2.43.0

maintainer
#2 lore raw

Re: Backport: FAILED: patch "[PATCH] ksmbd: validate security descriptor offsets before pointer" failed to apply to 6.1-stable tree

Quoted 1 line · Ratnaker Gautam

The 6.1 tree uses a void parse_dacl(). Preserve that interface in t

Could you fix your mail client and try again please? :)

Signature · 2 lines
Thanks,
Sasha

A message that is not in the archive (sent off-list or not yet received): DB5PR08MB100627622C33C426D3138E1FAFD682@DB5PR08MB10062.eurprd08.prod.outlook.com

#4 lore raw

PATCH 1/1 Fixed ksmbd: validate security descriptor offsets before pointer arithmetic

From: Namjae Jeon <[email protected]>

[ Upstream commit 3474e1a304fbf9f7c996b24adcd519e031227a90 ]

parse_sec_desc() constructs SID and DACL pointers by adding untrusted 32-bit offsets to the security descriptor before checking their bounds. The existing checks reject offsets into the descriptor header, but do not reject offsets beyond the descriptor's length.

On a 32-bit kernel, a large offset can wrap the pointer before the descriptor. The end-only checks in parse_sid() and parse_dacl() can then accept it, allowing an authenticated SMB2 SET_INFO request to access memory outside the security descriptor and fault the kernel.

Check each nonzero offset against acl_len before constructing its pointer. Initialize the optional owner and group SID pointers to NULL so an omitted SID remains absent when parsing the DACL.

The 6.1 tree uses a void parse_dacl(). Preserve that interface in this backport.

Fixes 0626e6641f6b ("cifsd: add server handler for central processing and tranport layers")
(cherry picked from commit 3474e1a304fbf9f7c996b24adcd519e031227a90)

---

My earlier reply in this thread had its tabs expanded by Outlook. Please disregard that copy.

 fs/smb/server/smbacl.c | 20 +++++++++++---------
 1 file changed, 11 insertions(+), 9 deletions(-)
fs/smb/server/smbacl.c +11 -9
index b0bf27cab..4a8812336 100644
@@ -885,7 +885,7 @@ int parse_sec_desc(struct user_namespace *user_ns, struct smb_ntsd *pntsd,
 		   int acl_len, struct smb_fattr *fattr)
 {
 	int rc = 0;
-	struct smb_sid *owner_sid_ptr, *group_sid_ptr;
+	struct smb_sid *owner_sid_ptr = NULL, *group_sid_ptr = NULL;
 	struct smb_acl *dacl_ptr; /* no need for SACL ptr */
 	char *end_of_acl = ((char *)pntsd) + acl_len;
 	__u32 dacloffset;
@@ -897,12 +897,7 @@ int parse_sec_desc(struct user_namespace *user_ns, struct smb_ntsd *pntsd,
 	if (acl_len < sizeof(struct smb_ntsd))
 		return -EINVAL;
 
-	owner_sid_ptr = (struct smb_sid *)((char *)pntsd +
-			le32_to_cpu(pntsd->osidoffset));
-	group_sid_ptr = (struct smb_sid *)((char *)pntsd +
-			le32_to_cpu(pntsd->gsidoffset));
 	dacloffset = le32_to_cpu(pntsd->dacloffset);
-	dacl_ptr = (struct smb_acl *)((char *)pntsd + dacloffset);
 	ksmbd_debug(SMB,
 		    "revision %d type 0x%x ooffset 0x%x goffset 0x%x sacloffset 0x%x dacloffset 0x%x\n",
 		    pntsd->revision, pntsd->type, le32_to_cpu(pntsd->osidoffset),
@@ -918,9 +913,12 @@ int parse_sec_desc(struct user_namespace *user_ns, struct smb_ntsd *pntsd,
 	pntsd->type = cpu_to_le16(DACL_PRESENT);
 
 	if (pntsd->osidoffset) {
-		if (le32_to_cpu(pntsd->osidoffset) < sizeof(struct smb_ntsd))
+		if (le32_to_cpu(pntsd->osidoffset) < sizeof(struct smb_ntsd) ||
+		    le32_to_cpu(pntsd->osidoffset) >= acl_len)
 			return -EINVAL;
 
+		owner_sid_ptr = (struct smb_sid *)((char *)pntsd +
+				le32_to_cpu(pntsd->osidoffset));
 		rc = parse_sid(owner_sid_ptr, end_of_acl);
 		if (rc) {
 			pr_err("%s: Error %d parsing Owner SID\n", __func__, rc);
@@ -936,9 +934,12 @@ int parse_sec_desc(struct user_namespace *user_ns, struct smb_ntsd *pntsd,
 	}
 
 	if (pntsd->gsidoffset) {
-		if (le32_to_cpu(pntsd->gsidoffset) < sizeof(struct smb_ntsd))
+		if (le32_to_cpu(pntsd->gsidoffset) < sizeof(struct smb_ntsd) ||
+		    le32_to_cpu(pntsd->gsidoffset) >= acl_len)
 			return -EINVAL;
 
+		group_sid_ptr = (struct smb_sid *)((char *)pntsd +
+				le32_to_cpu(pntsd->gsidoffset));
 		rc = parse_sid(group_sid_ptr, end_of_acl);
 		if (rc) {
 			pr_err("%s: Error %d mapping Owner SID to gid\n",
@@ -960,9 +961,10 @@ int parse_sec_desc(struct user_namespace *user_ns, struct smb_ntsd *pntsd,
 		pntsd->type |= cpu_to_le16(DACL_PROTECTED);
 
 	if (dacloffset) {
-		if (dacloffset < sizeof(struct smb_ntsd))
+		if (dacloffset < sizeof(struct smb_ntsd) || dacloffset >= acl_len)
 			return -EINVAL;
 
+		dacl_ptr = (struct smb_acl *)((char *)pntsd + dacloffset);
 		parse_dacl(user_ns, dacl_ptr, end_of_acl,
 			   owner_sid_ptr, group_sid_ptr, fattr);
 	}
Signature · 1 line
2.43.0
maintainer Applied
#5 lore raw

Re: [PATCH 6.1.y] Fixed ksmbd: validate security descriptor offsets before pointer arithmetic

Quoted 3 lines

From: Namjae Jeon <[email protected]>

[ Upstream commit 3474e1a304fbf9f7c996b24adcd519e031227a90 ]

Queued for 6.1 and 5.15, thanks.

Going forward please fix your mail client and follow the submission rules: send with git send-email, use a proper [PATCH 6.1.y] / [PATCH 5.15.y] subject, and do not send patches as Outlook replies.

Signature · 2 lines
Thanks,
Sasha