CVE-2026-89984: perf/x86/intel: Fix kernel address leakages in LBR stack
In the Linux kernel, the following vulnerability has been resolved:
perf/x86/intel: Fix kernel address leakages in LBR stack
Before Arch LBR gained CPL filtering support, a user-only branch stack could still contain kernel addresses. As a result, kernel branch records may be exposed to user space even when PERFSAMPLEBRANCHUSER is requested.
For example, on Intel Tiger Lake, the following command can still report SYSRET/ERET entries with kernel-space from addresses:
$ ./perf record -e cycles:p -o - --branch-filter any,savetype,u -- \ ./perf bench syscall basic --loop 1000 | \ ./perf script -i - --fields brstack|tr ' ' '\n'| \ grep -E '0x[89a-f][0-9a-f]{15}'
Total time: 0.000 [sec]
0.219000 usecs/op 4,566,210 ops/sec [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.551 MB - ] 0xffffffff93c001c8/0x7f12a2b1d647/P/-/-/16959/SYSRET/- 0xffffffff93c001c8/0x7f12a2b1d5c2/P/-/-/17535/SYSRET/- 0xffffffff93c01928/0x7f12a2861000/P/-/-/6719/ERET/- 0xffffffff93c01928/0x7f12a297a000/P/-/-/8575/ERET/-
The problem is that intelpmulbrfilter() does not fully validate the privilege level of sampled entries. It filters some mismatches based on the branch type and the to address, but it does not reject entries whose from address violates the requested branch privilege filter.
Fix this by extending software filtering to validate both from and to addresses against brsel. Any LBR entry contains kernel address does not match the requested user filter is dropped. This prevents kernel addresses from appearing in user-only branch stacks.
Affected Software
Event History
Frequently Asked Questions
Which systems are most likely to be affected?
Systems using Intel architectural Last Branch Records (LBR) before CPL filtering support are affected. Intel Tiger Lake is specifically identified as exhibiting the issue.
How can I check whether kernel branch addresses are being exposed?
Collect a user-only branch stack with perf using a branch filter such as "any,save_type,u" and inspect the branch-stack output for kernel-space addresses. SYSRET or ERET entries whose from addresses begin with values such as 0xffffffff indicate exposure.
What is required to trigger the exposure?
The issue occurs when perf requests user-only branch samples through PERF_SAMPLE_BRANCH_USER or an equivalent user-only branch filter. The LBR filtering can then retain kernel branch records because sampled entry privilege levels are not fully validated.