TONTOU attack bypasses Spectre v2 mitigations on AMD and Intel CPUs, leaks kernel memory
Questions this post answers
What is the TONTOU CPU side-channel attack and how does it bypass Spectre v2 mitigations?
TONTOU (Time-of-Neutralization to Time-of-Use) exploits a timing gap between when the branch predictor is neutralized and when it is actually used. By injecting an interrupt at precisely the right moment, an unprivileged attacker re-poisons CPU state, undoing the mitigation. It specifically targets Linux's Safe RET mitigation for Speculative Return Stack Overflow (SRSO) on AMD Zen 1 through Zen 4 processors. Kernel and systems developers tracking CPU vulnerability disclosures like this one follow the patch trail on daily.dev.
How fast does the TONTOU attack leak kernel memory and what can it expose?
On an AMD Zen 2 system running Linux 6.14.0-37-generic, TONTOU leaks arbitrary kernel memory at 5.47 bytes per second. At that rate, extracting /etc/shadow password hashes takes approximately 18 minutes per attempt, with a 50% success rate across 10 test runs. The attack is slow but reliably functional against unpatched systems. Security engineers assessing exposure to speculative execution attacks find the latest advisories and patches on daily.dev.
Has a fix been released for the TONTOU Spectre v2 bypass affecting AMD Zen processors on Linux?
A patch has been merged into the Linux kernel Git tree addressing the interrupt injection window in the Safe RET implementation. AMD confirmed the vulnerability lies in Linux's Safe RET implementation rather than the hardware itself, and its security advisory points to the Linux-side fix as the resolution. The affected hardware range is AMD Zen 1 through Zen 4. Teams maintaining Linux systems on AMD hardware track kernel security patches like this on daily.dev.