Freebsd: community verdicts

The KTLS receive path decrypted each record in place, assuming that the mbufs holding received data were anonymous and safe to modify. This assumption does not hold for data placed on a socket by sendfile(2), which can reference file-backed memory directly through non-anonymous M_EXTPG pages or EXT_SFBUF mbufs. When the sender transmits such data over a loopback connection without enabling KTLS on the transmit side, the file-backed mbufs reach the receiver's decryption path unchanged. Decrypting a record in place then overwrites the backing file's page cache instead of a private copy of the data. An unprivileged local user who can read a file can overwrite its contents with data of their choosing by sending the file over a loopback connection on which they have enabled KTLS receive. The write modifies the page cache directly, so it bypasses file flags such as schg and is written back to disk. By overwriting a setuid binary or other trusted file, a local user can escalate privileges, potentially gaining full control of the affected system.

dsp_mmap_single() validated the requested mapping by checking the sum of the user-supplied offset and length against the buffer size. This addition could overflow, so that a large offset and length wrapped around and passed the check. The offset was then narrowed from 64 to 32 bits when converted to a buffer address, yielding a mapping that extended past the audio buffer into unrelated kernel memory. The /dev/dsp device nodes are world-accessible by default. On a system with an audio device, either issue allows an unprivileged local user to read and write kernel memory, which can be used to escalate privileges, potentially gaining full control of the affected system. At a minimum, an attacker can crash the kernel, resulting in a Denial of Service (DoS).

Second, the audio buffer backing a mapping could be freed when the device was closed even though the mapping remained valid. The freed memory could then be reused elsewhere while still accessible through the stale mapping. The /dev/dsp device nodes are world-accessible by default. On a system with an audio device, either issue allows an unprivileged local user to read and write kernel memory, which can be used to escalate privileges, potentially gaining full control of the affected system. At a minimum, an attacker can crash the kernel, resulting in a Denial of Service (DoS).

The CONS_HISTORY ioctl handler did not adequately validate the requested history size. A large value caused an integer overflow in the buffer size calculation, resulting in a heap allocation smaller than expected. Subsequent initialization of the buffer wrote beyond the end of the allocation. An unprivileged local user with access to a vt(4) device can trigger an out-of-bounds write in the kernel, potentially escalating privileges.

The ELF image activator cleared per-process ASLR preference flags for setuid binaries after the code that computes the PIE base address, rather than before. As a result, a user-requested ASLR disable was still in effect at the point where the base address was chosen. An unprivileged local user can disable ASLR for a setuid PIE binary by calling procctl(2) before execve(2). This makes exploitation of any separate memory corruption vulnerability in that binary significantly easier.

The kernel handler for IPV6_MSFILTER dropped a serializing lock in order to copy the source-filter list from userspace, then reacquired the lock. During this window another thread could free the multicast filter structure, leaving the handler with a stale pointer to freed memory. An unprivileged local user can exploit this use-after-free to escalate privileges.

The Linuxulator determined whether a binary was set-user-ID or set-group-ID by checking the P_SUGID process flag. During execve(2), this flag is not yet set at the point where the auxiliary vector is constructed, so AT_SECURE was incorrectly set to zero for set-user-ID and set-group-ID executables. An unprivileged local user can inject a shared library via LD_PRELOAD into a set-user-ID or set-group-ID Linux binary, gaining the privileges of that binary.

sigqueue(2) was marked as permitted in capability mode with the introduction of Capsicum in 2011, but the implementation of kern_sigqueue did not include a capability mode check restricting signal delivery to the calling process's own PID. A process in capability mode can use sigqueue(2) to send signals to any process it could signal following standard Unix permissions, bypassing the Capsicum sandbox restriction. A compromised sandboxed process could interfere with other processes, for example by sending SIGKILL or SIGSTOP. This could be any process running as the same user, or any process, for a superuser sandboxed process.

When used to deliver a signal to a specific thread, thr_kill2(2) called p_cansignal() to determine whether the operation was permitted but did not check the result before delivering the signal. The signal was sent even when the permission check failed. The system call returned the resulting error to the caller, but by then the signal had already been delivered. The missing check allows an unprivileged local user who knows or can guess a target's process and thread IDs to send any signal to a process they would not normally be permitted to signal, including processes owned by other users or by root. The same check enforces jail boundaries, so a jailed process can signal processes on the host or in other jails. Thread IDs are allocated globally and sequentially, and so can be discovered by brute force with no visibility into the target. An attacker can stop or terminate arbitrary processes, including critical system daemons, resulting in a Denial of Service (DoS).