# Linux 7.2 Protects Against Crafted Perf Data From Going Rogue > Source: > Published: 2026-06-24 10:26:28+00:00 # Linux 7.2 Protects Against Crafted Perf Data From Going Rogue With the help of Claude Opus 4.6, the Linux 7.2 kernel added protections to fend off specially crafted or corrupted perf data for the perf tool that could cause a number of issues for the running system. The Arnaldo Carvalho de Melo of Red Hat worked on a set of 29 patches with Claude Opus 4.6 to better protect the perf data handling to avoid any accidental or nefarious issues around the perf data handling. The improved validation and handling prevents against out-of-bounds reads/writes, infinite loops, heap overflows, and segmentation faults. There is now defense-in-depth validation for file parsing and a number of other improvements made. Plus a number of bugs were also uncovered and fixed during this big rework. Besides Claude, the Linux kernel's Sashiko also helped on the AI side. Also as part of the perf tooling work for Linux 7.2 is a set of patches that went through twenty revisions for a new The The *perf.data*output and consumed by Linux's*perf*performance analyzing tool is a binary format containing performance counter data, CPU events, and other useful data. But to now there hasn't been a good job at hardening it to ensure that nothing goes awry if it's corrupted or manipulated for malicious intentArnaldo Carvalho de Melo of Red Hat worked on a set of 29 patches with Claude Opus 4.6 to better protect the perf data handling to avoid any accidental or nefarious issues around the perf data handling. The improved validation and handling prevents against out-of-bounds reads/writes, infinite loops, heap overflows, and segmentation faults. There is now defense-in-depth validation for file parsing and a number of other improvements made. Plus a number of bugs were also uncovered and fixed during this big rework. Besides Claude, the Linux kernel's Sashiko also helped on the AI side. Also as part of the perf tooling work for Linux 7.2 is a set of patches that went through twenty revisions for a new *perf inject --aslr*feature to better code with Address Space Layout Randomization (ASLR) differences between systems or reboots. This new option will remap virtual memory addresses or drop physical memory event leaks when profile record data is shared between systems.The