In the Linux kernel, the following vulnerability has been resolved:
perf: Fix use-after-free when perf mmap() revival races with the last munmap()
perf_mmap_close() drops rb->mmap_count without holding
event->mmap_mutex (the refcount_dec_and_test() right before the
refcount_dec_and_mutex_lock() of event->mmap_count). A concurrent
perf_mmap_rb() can slot its entire "revival" path into that window
(perf_mmap holds event->mmap_mutex for its whole duration, including
rb_alloc):
munmap side (perf_mmap_close) mmap side (perf_mmap_rb)
----------------------------------- --------------------------------
rb->mmap_count 1 -> 0 (no lock) (holds event->mmap_mutex)
inc_not_zero(rb->mmap_count) fails
ring_buffer_attach(event, NULL)
rb_alloc() + attach new rb
refcount_set(&event->mmap_count, 1)
lock; event->mmap_count 1 -> 0
ring_buffer_attach(event, NULL)
ring_buffer_put() -> frees the new rb
The revival's refcount_set(&event->mmap_count, 1) is an invisible
1 -> 1 write: the close frees the just-revived buffer although the
other process still has it mapped -- a page-level use-after-free
allowing local privilege escalation to root by any unprivileged user
(default kernel.perf_event_paranoid=2).
Swap the order of the two counter updates: event->mmap_count is
dropped first via refcount_dec_and_mutex_lock(), so its 1 -> 0
transition and the ring_buffer_attach() stay serialized with
perf_mmap(). rb->mmap_count == 0 then implies every event using the
buffer is detached already, so the result of the rb->mmap_count drop
can gate the remaining teardown directly and detach_rest is no longer
needed.
An earlier fix for this race from Kyle Zeng and David Lee takes
event->mmap_mutex around both counter updates [0]; here the not-last
close stays lockless.
CVSS Vector: CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H
CVSS Score: 7.8
AV:L - The UAF is reached only via local syscalls: perf_event_open(2) then mmap(2)/munmap(2) on the resulting perf fd (perf_fops.mmap → perf_mmap → perf_mmap_rb racing perf_mmap_close). There is no network, adjacent-radio, or physical path into the ring-buffer revival race.
AC:L - The attacker drives both sides from threads or processes sharing the fd: one munmap() runs perf_mmap_close() while a concurrent mmap() holds event->mmap_mutex through rb_alloc() and map_range(). That is an attacker-controlled, freely retryable race, not an uninfluenced layout or victim-state condition.
PR:L - Default kernel.perf_event_paranoid=2 still allows an unprivileged per-task software event with exclude_kernel set; perf_mmap() only calls the optional LSM security_perf_event_read hook and does not require CAP_PERFMON or CAP_SYS_ADMIN. The fix notes local privilege escalation by any unprivileged user.
UI:N - The attacker opens the event and races mmap/munmap on that fd in their own process or a helper that inherited the fd. No victim action such as mounting a filesystem or opening a file is required.
S:U - Impact is host-kernel page use-after-free of the perf ring buffer, i.e. standard local privilege escalation in the same OS security authority. It is not a KVM/Xen guest-to-host escape or IOMMU/DMA boundary bypass.
C:H - perf_mmap_close() frees the revived ring-buffer pages (rb_free) while the concurrent mmap's VM_PFNMAP VMA still maps those PFNs. That page-level UAF lets the attacker reclaim the pages and read arbitrary kernel memory through the still-valid mapping.
I:H - The same still-mapped PFNs remain writable on the user control page (perf_mmap_pfn_mkwrite), so reclaim/spray of the freed pages yields a kernel write primitive. The fix commit describes this as local privilege escalation to root, matching UAF integrity High.
A:H - Use-after-free of the ring buffer while it remains mapped causes kernel oops or panic on reuse, and the reporters documented a panic reproducer. Any kernel UAF is availability High even without a completed privilege-escalation exploit.
| Attack Vector |
Local |
Scope |
Unchanged |
| Attack Complexity |
Low |
Confidentiality Impact |
High |
| Privileges Required |
Low |
Integrity Impact |
High |
| User Interaction |
None |
Availability Impact |
High |
AV:L - The UAF is reached only via local syscalls: perf_event_open(2) then mmap(2)/munmap(2) on the resulting perf fd (perf_fops.mmap → perf_mmap → perf_mmap_rb racing perf_mmap_close). There is no network, adjacent-radio, or physical path into the ring-buffer revival race.
AC:L - The attacker drives both sides from threads or processes sharing the fd: one munmap() runs perf_mmap_close() while a concurrent mmap() holds event->mmap_mutex through rb_alloc() and map_range(). That is an attacker-controlled, freely retryable race, not an uninfluenced layout or victim-state condition.
PR:L - Default kernel.perf_event_paranoid=2 still allows an unprivileged per-task software event with exclude_kernel set; perf_mmap() only calls the optional LSM security_perf_event_read hook and does not require CAP_PERFMON or CAP_SYS_ADMIN. The fix notes local privilege escalation by any unprivileged user.
UI:N - The attacker opens the event and races mmap/munmap on that fd in their own process or a helper that inherited the fd. No victim action such as mounting a filesystem or opening a file is required.
S:U - Impact is host-kernel page use-after-free of the perf ring buffer, i.e. standard local privilege escalation in the same OS security authority. It is not a KVM/Xen guest-to-host escape or IOMMU/DMA boundary bypass.
C:H - perf_mmap_close() frees the revived ring-buffer pages (rb_free) while the concurrent mmap's VM_PFNMAP VMA still maps those PFNs. That page-level UAF lets the attacker reclaim the pages and read arbitrary kernel memory through the still-valid mapping.
I:H - The same still-mapped PFNs remain writable on the user control page (perf_mmap_pfn_mkwrite), so reclaim/spray of the freed pages yields a kernel write primitive. The fix commit describes this as local privilege escalation to root, matching UAF integrity High.
A:H - Use-after-free of the ring buffer while it remains mapped causes kernel oops or panic on reuse, and the reporters documented a panic reproducer. Any kernel UAF is availability High even without a completed privilege-escalation exploit.
CVSS 3.1