Cacti: Package Import Signature Validation Bypass Allows Self-Signed Packages

Cacti is an open source performance and fault management framework. Versions 1.2.30 and prior have a package import signature validation bypass allows which allows self-signed packages. This issue has been fixed in version 1.2.31.

accel/ethosu: reject NPU_OP_RESIZE commands from userspace

In the Linux kernel, the following vulnerability has been resolved: accel/ethosu: reject NPU_OP_RESIZE commands from userspace NPU_OP_RESIZE is a U85-only command that the driver does not yet implement. The existing WARN_ON(1) placeholder fires unconditionally whenever userspace submits this command via DRM_IOCTL_ETHOSU_GEM_CREATE, causing unbounded kernel log spam. If panic_on_warn is set the kernel panics, giving any unprivileged user with access to the DRM device a trivial denial-of-service primitive. Replace the WARN_ON(1) with an explicit -EINVAL return so the ioctl rejects the command before it reaches hardware.

xfrm: iptfs: fix ABBA deadlock in iptfs_destroy_state()

In the Linux kernel, the following vulnerability has been resolved: xfrm: iptfs: fix ABBA deadlock in iptfs_destroy_state() iptfs_destroy_state() calls hrtimer_cancel() while holding a spinlock that the timer callback also acquires, leading to an ABBA deadlock on SMP systems. For the output timer (iptfs_timer): - iptfs_destroy_state() holds x->lock, calls hrtimer_cancel() - iptfs_delay_timer() callback takes x->lock For the drop timer (drop_timer): - iptfs_destroy_state() holds drop_lock, calls hrtimer_cancel() - iptfs_drop_timer() callback takes drop_lock Both timers use HRTIMER_MODE_REL_SOFT, so their callbacks run in softirq context. When hrtimer_cancel() is called for a soft timer that is currently executing on another CPU, hrtimer_cancel_wait_running() spins on softirq_expiry_lock -- the same lock held by the softirq running the callback. If the callback is blocked waiting for the spinlock held by the caller of hrtimer_cancel(), a circular dependency forms: CPU 0: holds lock_A -> waits for softirq_expiry_lock CPU 1: holds softirq_expiry_lock -> waits for lock_A Fix by calling hrtimer_cancel() before acquiring the respective locks. hrtimer_cancel() is safe to call without holding any lock and will wait for any in-progress callback to complete. For the output timer, the lock is still acquired afterwards to drain the packet queue. For the drop timer, the lock/unlock pair is removed entirely since it only existed to serialize with the timer callback, which hrtimer_cancel() already guarantees. Found by source code audit.

libnfs through 6.0.2 before 935b8db has an xid integer underflow in READ_IOVEC in rpc_read_from_socket in lib/socket.c during a connection to a crafted NFS server, when the expected pdu size exceeds the absolute pdu size from the xid/record-marker.

drm/v3d: Fix vaddr leak when indirect CSD has zeroed workgroups

In the Linux kernel, the following vulnerability has been resolved: drm/v3d: Fix vaddr leak when indirect CSD has zeroed workgroups v3d_rewrite_csd_job_wg_counts_from_indirect() maps both the indirect buffer and the workgroup buffer and is expected to release them before returning. When any of the workgroup counts read from the buffer is zero, the function bailed out early and skipped the cleanup, leaking the vaddr mappings of both BOs. Jump to the cleanup path instead of returning directly, so the mappings are always dropped.

rseq: Fix using an uninitialized stack variable in rseq_exit_user_update()

In the Linux kernel, the following vulnerability has been resolved: rseq: Fix using an uninitialized stack variable in rseq_exit_user_update() There is an bug in which an uninitialized stack variable is used in rseq_exit_user_update() as reported by syzbot: BUG: KMSAN: kernel-infoleak in rseq_set_ids_get_csaddr include/linux/rseq_entry.h:502 [inline] The local variable: struct rseq_ids ids = { .cpu_id = task_cpu(t), .mm_cid = task_mm_cid(t), .node_id = cpu_to_node(ids.cpu_id), }; According to the C standard, the evaluation order of expressions in an initializer list is indeterminately sequenced. The compiler (Clang, in this KMSAN build) evaluates `cpu_to_node(ids.cpu_id)` *before* `ids.cpu_id` is initialized with `task_cpu(t)`. This is fixed by moving the assignment of ids.node_id outside the structure initialization.

netfilter: nft_meta_bridge: fix stale stack leak via IIFHWADDR register

In the Linux kernel, the following vulnerability has been resolved: netfilter: nft_meta_bridge: fix stale stack leak via IIFHWADDR register NFT_META_BRI_IIFHWADDR declares its destination register with len = ETH_ALEN (6 bytes), which the register-init tracking rounds up to two 32-bit registers (8 bytes). nft_meta_bridge_get_eval() then does memcpy(dest, br_dev->dev_addr, ETH_ALEN), writing only 6 bytes and leaving the upper 2 bytes of the second register as uninitialised nft_do_chain() stack. A downstream load of that register span leaks those stale bytes to userspace. Zero the second register before the memcpy so the full declared span is written.

Cacti: Open Redirect via HTTP_REFERER substring check in auth_login_redirect

Cacti is an open source performance and fault management framework. Versions 1.2.30 and prior are vulnerable to Open Redirect through a substring check rather than a host check at str_contains($referer, CACTI_PATH_URL). When the user's login_opts == '1' (redirect to referer after login), the function used $_SERVER['HTTP_REFERER'] directly. An attacker could craft a referer such as https://evil.com/cacti/. Where CACTI_PATH_URL is /cacti/, the substring matches and the user is redirected to evil.com after login. The pre-existing validate_redirect_url() helper at lib/html_utility.php performed proper validation but was not invoked from auth_login_redirect(). This issue has been fixed in version 1.2.31.

Wildcard DNS SAN bypasses CA name-constraint checks

Certificates with wildcard DNS SANs (e.g. *.example.com) bypassed CA name-constraint checks. A certificate with a wildcard DNS SAN that should be rejected by the issuing CA's permitted/excluded DNS name constraints could be accepted.

Vim: Out-of-bounds Write in Spell File Word Count

Vim is an open source, command line text editor. Prior to 9.2.0653, the tree_count_words() function in src/spellfile.c fills in the word-count fields of a spell-file word trie by walking it iteratively with a depth counter. The counter is bounded only by the trie structure itself; it is never checked against the size of the fixed MAXWLEN-element stack arrays it indexes (arridx[], curi[], wordcount[]). A crafted .spl/.sug file pair, loaded when the user invokes spell suggestion, can drive the descent arbitrarily deep, so the function writes past the end of those arrays. This is a stack out-of-bounds write that corrupts the call frame and crashes the editor. This vulnerability is fixed in 9.2.0653.

X.509 trust-chain bypass in wolfSSL_X509_verify_cert() via untrusted intermediate anchoring

X.509 trust-chain bypass in the OpenSSL compatibility certificate verifier (wolfSSL_X509_verify_cert()). This affects only builds with --enable-opensslextra (OPENSSL_EXTRA) and whose application validates certificates by calling X509_verify_cert() with caller-supplied untrusted intermediate certificates; for those users it is critical, otherwise the library is unaffected. In particular, native wolfSSL TLS/DTLS usage is not impacted. wolfSSL's X509_verify_cert() temporarily loads each caller-supplied untrusted intermediate into the certificate manager but failed to drop them before the trusted-store check, so an untrusted intermediate could anchor the path itself. An attacker can present a chain that never reaches a configured trust anchor and have it accepted, resulting in acceptance of an attacker-controlled certificate. This is certificate verification independent of TLS (e.g. S/MIME/CMS, code/firmware signing, JWT/JWS x5c), is not specific to any key type or algorithm, and a single untrusted intermediate suffices. The default wolfSSL TLS handshake (WOLFSSL_VERIFY_PEER) is not affected; only TLS applications doing manual or deferred peer verification through this API are, which also requires --enable-sessioncerts.

Out-of-bounds heap read in SM2/SM3 certificate Subject Key Identifier computation

Out-of-bounds heap read during SM2/SM3 certificate signature verification. When parsing a certificate with an SM3wSM2 signature, the Subject Key Identifier computation reads the trailing 65 bytes of the public key without checking that the key is at least that long. A public key shorter than 65 bytes results in an out-of-bounds heap read, leading to a potential crash (denial of service); there is no out-of-bounds write. Note this only affects builds with SM2 support (--enable-sm2 or --enable-all).

Renesas TSIP TLS 1.3 transcript buffer out-of-bounds write in tsip_StoreMessage

Out-of-bounds write in the Renesas TSIP TLS 1.3 transcript buffer. In tsip_StoreMessage() the capacity check guarding the fixed message bag (MSGBAG_SIZE) sets an error code but fails to return, so execution falls through to an XMEMCPY that writes past the end of the buffer once the accumulated TLS 1.3 handshake transcript exceeds MSGBAG_SIZE (8 KB), corrupting adjacent heap state and potentially causing a remote denial of service crash. The bag is sized to hold a normal handshake, so this is reached only by an unusually large but valid certificate chain, or by a malicious or man-in-the-middle server sending an oversized handshake message to a client that does not strictly verify the chain. This only affects builds using the Renesas TSIP TLS port (WOLFSSL_RENESAS_TSIP_TLS) as a TLS 1.3 client on Renesas MCUs with TSIP hardware enabled, and is rated High within those builds. All other configurations are unaffected.

Un-negotiated Raw Public Key (RFC 7250) accepted in place of X.509, bypassing chain validation

Un-negotiated Raw Public Key (RFC 7250) accepted in place of an X.509 certificate, bypassing chain validation. A raw public key has no chain, so ParseCertRelative() accepts it without performing any trust verification; it must therefore only be accepted when RPK was actually negotiated for that peer. The check now defaults the expected type to X.509 (per RFC 7250/8446) when no type was negotiated, comparing against the received server certificate type on the client and the selected client certificate type on the server, and rejects any mismatch, including an un-negotiated raw public key, with UNSUPPORTED_CERTIFICATE. Only affects builds with Raw Public Key support (HAVE_RPK) enabled - disabled by default in a standalone build, but included in --enable-all.

Chain intermediate CA:TRUE without keyCertSign accepted as a signing CA (temporary CA exemption)

Chain intermediate CA:TRUE without keyCertSign accepted as a signing CA. Intermediate CA certificates are required to have the keyCertSign key usage when a Key Usage extension is present, but chain-supplied temporary CAs (WOLFSSL_TEMP_CA) added while building a certificate path were previously exempted from this check, so an intermediate asserting CA:TRUE but lacking keyCertSign was accepted as a signing CA. The check now applies to chain-supplied temporary CAs as well; only operator-loaded root certificates (WOLFSSL_USER_CA) and self-signed roots remain exempt. Per RFC 5280 an absent Key Usage extension implies all usages, so the requirement is enforced only when the extension is actually present (extKeyUsageSet). Affects the OpenSSL-compatibility certificate-path-building path (X509_verify_cert / X509_STORE, OPENSSL_EXTRA/OPENSSL_ALL), where untrusted chain intermediates are added as temporary CAs; native (non-OpenSSL-compat) certificate verification does not create temporary CAs and is unaffected. Within those builds, the check applies unless ALLOW_INVALID_CERTSIGN is defined.

EVoke Systems EVoke CSMS Improper Restriction of Excessive Authentication Attempts

The WebSocket Application Programming Interface lacks restrictions on the number of authentication requests. This absence of rate limiting may allow an attacker to conduct denial-of-service attacks or brute-force attacks to gain unauthorized access.

File Browser: Cross-user unauthorized share-link deletion via unbounded prefix match in DeleteWithPathPrefix

File Browser is a file managing interface for uploading, deleting, previewing, renaming, and editing files within a specified directory. Prior to 2.63.6, a low-privileged authenticated user of filebrowser (with create + delete permissions in their own isolated scope) can silently destroy share-link records belonging to any other user — including the administrator — by performing a legitimate DELETE on a file in their own directory whose logical path happens to be a byte-prefix of another user's stored share.Link.Path. The file contents of the victim are not exposed, but the victim's share links are irrevocably wiped. This vulnerability is fixed in 2.63.6.

EVoke Systems EVoke CSMS Insufficient Session Expiration

The WebSocket backend uses charging station identifiers to uniquely associate sessions but allows multiple endpoints to connect using the same session identifier. This implementation results in predictable session identifiers. This vulnerability may allow unauthorized users to authenticate as other users or enable a malicious actor to cause a denial-of-service condition by overwhelming the backend with valid session requests.

HMAC zero-length tag forgery in EVP_DigestVerifyFinal

HMAC zero-length tag forgery in EVP_DigestVerifyFinal, where a zero-length tag could be accepted as valid during HMAC verification. In the OpenSSL-compatibility HMAC verify path the supplied signature length was only checked as not exceeding the MAC length, so a zero-length or otherwise truncated tag could pass verification. The fix requires the supplied tag length to exactly equal the MAC length and rejects a zero-length MAC, so a forged short or empty tag is no longer accepted.

X25519 x86_64 assembly final reduction leaves non-canonical field element

The X25519 x86_64 assembly implementation fails to clear the most significant bit during the final modular reduction, so the computed result may not be fully reduced modulo the field prime 2^255 - 19. This can leave the field element in a non-canonical form, producing an incorrect result from the scalar multiplication and potentially a wrong shared secret. The final carry-propagation chains in the x64 and AVX2 reduction routines could overflow into the top bit, and the high limb was not masked afterward, so the 255-bit field element was left non-canonical.