CVE-2026-93164 in Linux
Summary
by MITRE • 09/18/2026
In the Linux kernel, the following vulnerability has been resolved:
uprobes/x86: Move optimized uprobe from nop5 to nop10
Andrii reported an issue with optimized uprobes [1] that can clobber
redzone area with call instruction storing return address on stack where user code may keep temporary data without adjusting rsp.
Fixing this by moving the optimized uprobes on top of 10-bytes nop instruction, so we can squeeze another instruction to escape the redzone area before doing the call, like:
lea -0x80(%rsp), %rsp call tramp
Note the lea instruction is used to adjust the rsp register without changing the flags.
We use nop10 and following transformation to optimized instructions above and back as suggested by Peterz [2].
Optimize path (int3_update_optimize):
1) Initial state after set_swbp() installed the uprobe: cc 2e 0f 1f 84 00 00 00 00 00
From offset 0 this is INT3 followed by the tail of the original 10-byte NOP.
After a previous unoptimization bytes 5..9 may still contain the old call instruction, which remains valid for threads already there.
2) Rewrite the LEA tail and call displacement: cc [8d 64 24 80 e8 d0 d1 d2 d3]
From offset 0 this traps on the uprobe INT3. Bytes 1..9 are not executable entry points while byte 0 is trapped.
3) Publish the first LEA byte: [48] 8d 64 24 80 e8 d0 d1 d2 d3
From offset 0 this is: lea -0x80(%rsp), %rsp call <uprobe-trampoline>
Unoptimize path (int3_update_unoptimize):
1) Initial optimized state: 48 8d 64 24 80 e8 d0 d1 d2 d3 Same as 3) above.
2) Trap new entries before restoring the NOP bytes: [cc] 8d 64 24 80 e8 d0 d1 d2 d3
From offset 0 this traps. A thread that had already executed the LEA can still reach the intact CALL at offset 5.
3) Restore bytes 1..4 of the original NOP while keeping byte 0 trapped and byte 5 as CALL. cc [2e 0f 1f 84] e8 d0 d1 d2 d3
From offset 0 this still traps. Offset 5 is still the CALL for any thread that was already past the first LEA byte.
4) Publish the first byte of the original NOP: [66] 2e 0f 1f 84 e8 d0 d1 d2 d3
From offset 0 this is the restored 10-byte NOP; the CALL opcode and displacement are now only NOP operands. Offset 5 still decodes as CALL for a thread that was already there.
Tthere is only a single target uprobe-trampoline for the given nop10 instruction address, so the CALL instruction will not be changed across unoptimization/optimization cycles. Therefore, any task that is preempted at the CALL instruction is guaranteed to observe that CALL and not anything else.
Note as explained in [2] we need to use following nop10:
PF1 PF2 ESC NOPL MOD SIB DISP32 NOP10: 0x66, 0x2e, 0x0f, 0x1f, 0x84, 0x00, 0x00, 0x00, 0x00, 0x00 -- cs nopw 0x00000000(%rax,%rax,1)
which means we need to allow 0x2e prefix which maps to INAT_PFX_CS attribute in is_prefix_bad function.
Also changing the uprobe syscall error when called out of uprobe trampoline to -EPROTO, so we are able to detect the fixed kernel.
The optimized uprobe performance stays the same:
uprobe-nop : 3.129 ± 0.013M/s uprobe-push : 3.045 ± 0.006M/s uprobe-ret : 1.095 ± 0.004M/s --> uprobe-nop10 : 7.170 ± 0.020M/s uretprobe-nop : 2.143 ± 0.021M/s uretprobe-push : 2.090 ± 0.000M/s uretprobe-ret : 0.942 ± 0.000M/s --> uretprobe-nop10: 3.381 ± 0.003M/s usdt-nop : 3.245 ± 0.004M/s --> usdt-nop10 : 7.256 ± 0.023M/s
[1] https://lore.kernel.org/bpf/[email protected]/
[2] https://lore.kernel.org/bpf/[email protected]/#t
VulDB is the best source for vulnerability data and more expert information about this specific topic.
Analysis
by VulDB Data Team • 09/18/2026
The Linux kernel has addressed a critical memory safety vulnerability within the x86 architecture implementation of optimized uprobes, specifically concerning stack redzone corruption. The flaw originated from the use of five-byte NOP instructions as padding for uprobe trampoline code. When an optimized uprobe is installed, it replaces the original instruction with a breakpoint interrupt (INT3) followed by a jump to a trampoline function that executes the original logic in kernel space. To ensure atomic updates and prevent race conditions during optimization transitions, the kernel uses NOP instructions as placeholders for the remaining bytes of the replaced instruction. The previous implementation utilized nop5 sequences, which provided insufficient padding between the INT3 trap and the subsequent call instruction leading to the trampoline. This lack of spacing caused the stack pointer adjustment logic within the trampoline setup or execution path to overwrite memory in the redzone area—a region reserved by the Application Binary Interface (ABI) for temporary data storage that user-space applications do not explicitly manage via register adjustments like rsp.
This vulnerability allows a local attacker, who has the ability to insert uprobes into target processes, to corrupt the stack redzone of those processes. By clobbering this area with call instruction return addresses or other control flow modifications during the optimized uprobe execution path, an attacker can potentially cause undefined behavior in user-space applications that rely on the integrity of their redzones for temporary data storage without explicit stack pointer adjustments. While direct arbitrary code execution is not immediately guaranteed by this specific flaw alone due to the controlled nature of the trampoline jump, it represents a significant stability risk and could be chained with other vulnerabilities or exploited through timing attacks to cause denial of service via segmentation faults or memory corruption errors in affected applications. The issue falls under CWE-124 Buffer Underwrite if considering stack growth directionality or more accurately CWE-787 Out-of-bounds Write relative to the expected redzone boundaries, and it impacts the integrity and availability aspects of security as defined by standard CIA triads.
The resolution involves transitioning from nop5 to a ten-byte NOP instruction sequence (nop10) for optimized uprobes on x86 platforms. This expansion provides adequate space to insert an additional instruction that explicitly adjusts the stack pointer away from the redzone before executing the call to the trampoline function. Specifically, the fix employs a lea -0x80(%rsp), %rsp instruction followed by the call operation. The use of the LEA (Load Effective Address) instruction is critical because it modifies the stack pointer without affecting CPU flags, ensuring that subsequent operations within the trampoline do not encounter unexpected flag states that could alter control flow unpredictably. This structural change ensures that any data stored in the redzone remains untouched during the uprobe interception and execution cycle, thereby preserving user-space memory integrity.
The implementation of this fix requires careful handling of instruction decoding prefixes to ensure compatibility with existing kernel validation logic. The chosen nop10 sequence utilizes specific byte patterns including 0x66, 0x2e, 0x0f, etc., which necessitates allowing the 0x2e prefix in the is_prefix_bad function mapping it to INAT_PFX_CS attribute. This adjustment ensures that the kernel correctly identifies and processes these extended NOP sequences without rejecting them as invalid instructions during uprobe installation or optimization phases. Furthermore, the patch includes a mechanism to detect this fixed state by changing the error code returned when an uprobe syscall is invoked outside of the expected trampoline context from its previous value to -EPROTO. This diagnostic aid helps in verifying that systems are running with the corrected kernel version and can distinguish between legacy vulnerable states and patched environments during debugging or automated security assessments.
From a performance perspective, this architectural change does not degrade system efficiency; rather, it maintains or slightly improves uprobe throughput metrics compared to previous implementations using different optimization strategies like push/ret sequences. Benchmarking data indicates that the nop10 approach yields higher instructions per second for both uprobes and uretprobes compared to older methods, demonstrating that security hardening can coexist with high-performance tracing capabilities. Security practitioners should ensure their Linux distributions are updated to include this kernel patch to mitigate risks associated with stack redzone corruption in applications utilizing dynamic instrumentation tools such as SystemTap or BPF-based uprobe probes. Mitigation strategies for organizations unable to immediately patch involve monitoring application stability and reviewing the usage of uprobes in production environments, particularly where sensitive data processing occurs without explicit stack management by user-space code.
This vulnerability highlights the importance of adhering strictly to ABI specifications regarding redzone preservation when modifying execution flow at runtime. It serves as a reminder that low-level kernel modifications affecting instruction streams must account for all architectural constraints and side effects on adjacent memory regions. The fix aligns with best practices in secure coding by ensuring that temporary data areas are isolated from control-flow manipulation mechanisms. For further technical details regarding the patch series, references can be made to discussions involving Andrii Nakryiko and Peter Zijlstra on kernel mailing lists, which detail the rationale behind selecting nop10 over other alternatives like push/ret or simple call sequences that might still encroach upon protected stack regions depending on alignment and offset calculations.