The Linux Kernel versions 2.6.38 through 4.14 have a problematic use of pmd_mkdirty() in the touch_pmd() function inside the THP implementation. touch_pmd() can be reached by get_user_pages(). In such case, the pmd will become dirty. This scenario breaks the new can_follow_write_pmd()'s logic - pmd can become dirty without going through a COW cycle. This bug is not as severe as the original "Dirty cow" because an ext4 file (or any other regular file) cannot be mapped using THP. Nevertheless, it does allow us to overwrite read-only huge pages. For example, the zero huge page and sealed shmem files can be overwritten (since their mapping can be populated using THP). Note that after the first write page-fault to the zero page, it will be replaced with a new fresh (and zeroed) thp.

In the Linux kernel, pick_next_rt_entity() may return a type confused entry, not detected by the BUG_ON condition, as the confused entry will not be NULL, but list_head.The buggy error condition would lead to a type confused entry with the list head,which would then be used as a type confused sched_rt_entity,causing memory corruption.

kernel/events/core.c in the performance subsystem in the Linux kernel before 4.0 mismanages locks during certain migrations, which allows local users to gain privileges via a crafted application, aka Android internal bug 31095224.

kernel/events/core.c in the performance subsystem in the Linux kernel before 4.0 mismanages locks during certain migrations, which allows local users to gain privileges via a crafted application, aka Android internal bug 30955111.

A vulnerability, which was classified as critical, has been found in Linux Kernel. Affected by this issue is the function tst_timer of the file drivers/atm/idt77252.c of the component IPsec. The manipulation leads to use after free. It is recommended to apply a patch to fix this issue. VDB-211934 is the identifier assigned to this vulnerability.

The BPF subsystem in the Linux kernel before 4.5.5 mishandles reference counts, which allows local users to cause a denial of service (use-after-free) or possibly have unspecified other impact via a crafted application on (1) a system with more than 32 Gb of memory, related to the program reference count or (2) a 1 Tb system, related to the map reference count.

The msm_ipc_router_bind_control_port function in net/ipc_router/ipc_router_core.c in the IPC router kernel module for the Linux kernel 3.x, as used in Qualcomm Innovation Center (QuIC) Android contributions for MSM devices and other products, does not verify that a port is a client port, which allows attackers to gain privileges or cause a denial of service (race condition and list corruption) by making many BIND_CONTROL_PORT ioctl calls.

A use-after-free exists in drivers/tee/tee_shm.c in the TEE subsystem in the Linux kernel through 5.15.11. This occurs because of a race condition in tee_shm_get_from_id during an attempt to free a shared memory object.

The networking implementation in the Linux kernel through 4.3.3, as used in Android and other products, does not validate protocol identifiers for certain protocol families, which allows local users to cause a denial of service (NULL function pointer dereference and system crash) or possibly gain privileges by leveraging CLONE_NEWUSER support to execute a crafted SOCK_RAW application.

An out-of-bounds memory write flaw was found in the Linux kernel’s Transport Layer Security functionality in how a user calls a function splice with a ktls socket as the destination. This flaw allows a local user to crash or potentially escalate their privileges on the system.

A heap out-of-bounds write vulnerability in the Linux kernel's Performance Events system component can be exploited to achieve local privilege escalation. A perf_event's read_size can overflow, leading to an heap out-of-bounds increment or write in perf_read_group(). We recommend upgrading past commit 382c27f4ed28f803b1f1473ac2d8db0afc795a1b.

A use-after-free flaw was found in the Linux Kernel due to a race problem in the unix garbage collector's deletion of SKB races with unix_stream_read_generic() on the socket that the SKB is queued on.

.A flaw was found in the CAN BCM networking protocol in the Linux kernel, where a local attacker can abuse a flaw in the CAN subsystem to corrupt memory, crash the system or escalate privileges. This race condition in net/can/bcm.c in the Linux kernel allows for local privilege escalation to root.

A null pointer dereference flaw was found in the nft_inner.c functionality of netfilter in the Linux kernel. This issue could allow a local user to crash the system or escalate their privileges on the system.

In the Linux kernel, the following vulnerability has been resolved: spi: sun6i: fix race between DMA RX transfer completion and RX FIFO drain Previously the transfer complete IRQ immediately drained to RX FIFO to read any data remaining in FIFO to the RX buffer. This behaviour is correct when dealing with SPI in interrupt mode. However in DMA mode the transfer complete interrupt still fires as soon as all bytes to be transferred have been stored in the FIFO. At that point data in the FIFO still needs to be picked up by the DMA engine. Thus the drain procedure and DMA engine end up racing to read from RX FIFO, corrupting any data read. Additionally the RX buffer pointer is never adjusted according to DMA progress in DMA mode, thus calling the RX FIFO drain procedure in DMA mode is a bug. Fix corruptions in DMA RX mode by draining RX FIFO only in interrupt mode. Also wait for completion of RX DMA when in DMA mode before returning to ensure all data has been copied to the supplied memory buffer.

In the Linux kernel, the following vulnerability has been resolved: drm/msm/dpu: Add mutex lock in control vblank irq Add a mutex lock to control vblank irq to synchronize vblank enable/disable operations happening from different threads to prevent race conditions while registering/unregistering the vblank irq callback. v4: -Removed vblank_ctl_lock from dpu_encoder_virt, so it is only a parameter of dpu_encoder_phys. -Switch from atomic refcnt to a simple int counter as mutex has now been added v3: Mistakenly did not change wording in last version. It is done now. v2: Slightly changed wording of commit message Patchwork: https://patchwork.freedesktop.org/patch/571854/

In the Linux kernel, the following vulnerability has been resolved: pipe: wakeup wr_wait after setting max_usage Commit c73be61cede5 ("pipe: Add general notification queue support") a regression was introduced that would lock up resized pipes under certain conditions. See the reproducer in [1]. The commit resizing the pipe ring size was moved to a different function, doing that moved the wakeup for pipe->wr_wait before actually raising pipe->max_usage. If a pipe was full before the resize occured it would result in the wakeup never actually triggering pipe_write. Set @max_usage and @nr_accounted before waking writers if this isn't a watch queue. [Christian Brauner <brauner@kernel.org>: rewrite to account for watch queues]

In the Linux kernel, the following vulnerability has been resolved: ksmbd: fix race condition between session lookup and expire Thread A + Thread B ksmbd_session_lookup | smb2_sess_setup sess = xa_load | | | xa_erase(&conn->sessions, sess->id); | | ksmbd_session_destroy(sess) --> kfree(sess) | // UAF! | sess->last_active = jiffies | + This patch add rwsem to fix race condition between ksmbd_session_lookup and ksmbd_expire_session.

In the Linux kernel, the following vulnerability has been resolved: net: bridge: use DEV_STATS_INC() syzbot/KCSAN reported data-races in br_handle_frame_finish() [1] This function can run from multiple cpus without mutual exclusion. Adopt SMP safe DEV_STATS_INC() to update dev->stats fields. Handles updates to dev->stats.tx_dropped while we are at it. [1] BUG: KCSAN: data-race in br_handle_frame_finish / br_handle_frame_finish read-write to 0xffff8881374b2178 of 8 bytes by interrupt on cpu 1: br_handle_frame_finish+0xd4f/0xef0 net/bridge/br_input.c:189 br_nf_hook_thresh+0x1ed/0x220 br_nf_pre_routing_finish_ipv6+0x50f/0x540 NF_HOOK include/linux/netfilter.h:304 [inline] br_nf_pre_routing_ipv6+0x1e3/0x2a0 net/bridge/br_netfilter_ipv6.c:178 br_nf_pre_routing+0x526/0xba0 net/bridge/br_netfilter_hooks.c:508 nf_hook_entry_hookfn include/linux/netfilter.h:144 [inline] nf_hook_bridge_pre net/bridge/br_input.c:272 [inline] br_handle_frame+0x4c9/0x940 net/bridge/br_input.c:417 __netif_receive_skb_core+0xa8a/0x21e0 net/core/dev.c:5417 __netif_receive_skb_one_core net/core/dev.c:5521 [inline] __netif_receive_skb+0x57/0x1b0 net/core/dev.c:5637 process_backlog+0x21f/0x380 net/core/dev.c:5965 __napi_poll+0x60/0x3b0 net/core/dev.c:6527 napi_poll net/core/dev.c:6594 [inline] net_rx_action+0x32b/0x750 net/core/dev.c:6727 __do_softirq+0xc1/0x265 kernel/softirq.c:553 run_ksoftirqd+0x17/0x20 kernel/softirq.c:921 smpboot_thread_fn+0x30a/0x4a0 kernel/smpboot.c:164 kthread+0x1d7/0x210 kernel/kthread.c:388 ret_from_fork+0x48/0x60 arch/x86/kernel/process.c:147 ret_from_fork_asm+0x11/0x20 arch/x86/entry/entry_64.S:304 read-write to 0xffff8881374b2178 of 8 bytes by interrupt on cpu 0: br_handle_frame_finish+0xd4f/0xef0 net/bridge/br_input.c:189 br_nf_hook_thresh+0x1ed/0x220 br_nf_pre_routing_finish_ipv6+0x50f/0x540 NF_HOOK include/linux/netfilter.h:304 [inline] br_nf_pre_routing_ipv6+0x1e3/0x2a0 net/bridge/br_netfilter_ipv6.c:178 br_nf_pre_routing+0x526/0xba0 net/bridge/br_netfilter_hooks.c:508 nf_hook_entry_hookfn include/linux/netfilter.h:144 [inline] nf_hook_bridge_pre net/bridge/br_input.c:272 [inline] br_handle_frame+0x4c9/0x940 net/bridge/br_input.c:417 __netif_receive_skb_core+0xa8a/0x21e0 net/core/dev.c:5417 __netif_receive_skb_one_core net/core/dev.c:5521 [inline] __netif_receive_skb+0x57/0x1b0 net/core/dev.c:5637 process_backlog+0x21f/0x380 net/core/dev.c:5965 __napi_poll+0x60/0x3b0 net/core/dev.c:6527 napi_poll net/core/dev.c:6594 [inline] net_rx_action+0x32b/0x750 net/core/dev.c:6727 __do_softirq+0xc1/0x265 kernel/softirq.c:553 do_softirq+0x5e/0x90 kernel/softirq.c:454 __local_bh_enable_ip+0x64/0x70 kernel/softirq.c:381 __raw_spin_unlock_bh include/linux/spinlock_api_smp.h:167 [inline] _raw_spin_unlock_bh+0x36/0x40 kernel/locking/spinlock.c:210 spin_unlock_bh include/linux/spinlock.h:396 [inline] batadv_tt_local_purge+0x1a8/0x1f0 net/batman-adv/translation-table.c:1356 batadv_tt_purge+0x2b/0x630 net/batman-adv/translation-table.c:3560 process_one_work kernel/workqueue.c:2630 [inline] process_scheduled_works+0x5b8/0xa30 kernel/workqueue.c:2703 worker_thread+0x525/0x730 kernel/workqueue.c:2784 kthread+0x1d7/0x210 kernel/kthread.c:388 ret_from_fork+0x48/0x60 arch/x86/kernel/process.c:147 ret_from_fork_asm+0x11/0x20 arch/x86/entry/entry_64.S:304 value changed: 0x00000000000d7190 -> 0x00000000000d7191 Reported by Kernel Concurrency Sanitizer on: CPU: 0 PID: 14848 Comm: kworker/u4:11 Not tainted 6.6.0-rc1-syzkaller-00236-gad8a69f361b9 #0

Multiple race conditions in drivers/char/adsprpc.c and drivers/char/adsprpc_compat.c in the ADSPRPC driver for the Linux kernel 3.x, as used in Qualcomm Innovation Center (QuIC) Android contributions for MSM devices and other products, allow attackers to cause a denial of service (zero-value write) or possibly have unspecified other impact via a COMPAT_FASTRPC_IOCTL_INVOKE_FD ioctl call.

A race condition was found in the Linux kernel's watch queue due to a missing lock in pipe_resize_ring(). The specific flaw exists within the handling of pipe buffers. The issue results from the lack of proper locking when performing operations on an object. This flaw allows a local user to crash the system or escalate their privileges on the system.

In the Linux kernel before 5.17.3, fs/io_uring.c has a use-after-free due to a race condition in io_uring timeouts. This can be triggered by a local user who has no access to any user namespace; however, the race condition perhaps can only be exploited infrequently.

A vulnerability in the shared library loading mechanism of Cisco AnyConnect Secure Mobility Client for Linux and Mac OS could allow an authenticated, local attacker to perform a shared library hijacking attack on an affected device if the VPN Posture (HostScan) Module is installed on the AnyConnect client. This vulnerability is due to a race condition in the signature verification process for shared library files that are loaded on an affected device. An attacker could exploit this vulnerability by sending a series of crafted interprocess communication (IPC) messages to the AnyConnect process. A successful exploit could allow the attacker to execute arbitrary code on the affected device with root privileges. To exploit this vulnerability, the attacker must have a valid account on the system.

A local privilege escalation was discovered in the Linux kernel before 5.10.13. Multiple race conditions in the AF_VSOCK implementation are caused by wrong locking in net/vmw_vsock/af_vsock.c. The race conditions were implicitly introduced in the commits that added VSOCK multi-transport support.

In the Linux kernel 4.19 through 5.6.7 on the s390 platform, code execution may occur because of a race condition, as demonstrated by code in enable_sacf_uaccess in arch/s390/lib/uaccess.c that fails to protect against a concurrent page table upgrade, aka CID-3f777e19d171. A crash could also occur.

An issue was discovered in the Linux kernel before 6.5.9, exploitable by local users with userspace access to MMIO registers. Incorrect access checking in the #VC handler and instruction emulation of the SEV-ES emulation of MMIO accesses could lead to arbitrary write access to kernel memory (and thus privilege escalation). This depends on a race condition through which userspace can replace an instruction before the #VC handler reads it.

Race condition in mm/gup.c in the Linux kernel 2.x through 4.x before 4.8.3 allows local users to gain privileges by leveraging incorrect handling of a copy-on-write (COW) feature to write to a read-only memory mapping, as exploited in the wild in October 2016, aka "Dirty COW."

A race condition in Linux kernel SCTP sockets (net/sctp/socket.c) before 5.12-rc8 can lead to kernel privilege escalation from the context of a network service or an unprivileged process. If sctp_destroy_sock is called without sock_net(sk)->sctp.addr_wq_lock then an element is removed from the auto_asconf_splist list without any proper locking. This can be exploited by an attacker with network service privileges to escalate to root or from the context of an unprivileged user directly if a BPF_CGROUP_INET_SOCK_CREATE is attached which denies creation of some SCTP socket.

A double free bug in packet_set_ring() in net/packet/af_packet.c can be exploited by a local user through crafted syscalls to escalate privileges or deny service. We recommend upgrading kernel past the effected versions or rebuilding past ec6af094ea28f0f2dda1a6a33b14cd57e36a9755

A flaw was found in btrfs_get_root_ref in fs/btrfs/disk-io.c in the btrfs filesystem in the Linux Kernel due to a double decrement of the reference count. This issue may allow a local attacker with user privilege to crash the system or may lead to leaked internal kernel information.

Multiple stack-based buffer overflows in net/netfilter/ipvs/ip_vs_ctl.c in the Linux kernel before 2.6.33, when CONFIG_IP_VS is used, allow local users to gain privileges by leveraging the CAP_NET_ADMIN capability for (1) a getsockopt system call, related to the do_ip_vs_get_ctl function, or (2) a setsockopt system call, related to the do_ip_vs_set_ctl function.

An array indexing vulnerability was found in the netfilter subsystem of the Linux kernel. A missing macro could lead to a miscalculation of the `h->nets` array offset, providing attackers with the primitive to arbitrarily increment/decrement a memory buffer out-of-bound. This issue may allow a local user to crash the system or potentially escalate their privileges on the system.

A race condition was found in the way the Linux kernel's memory subsystem handled the copy-on-write (COW) breakage of private read-only shared memory mappings. This flaw allows an unprivileged, local user to gain write access to read-only memory mappings, increasing their privileges on the system.

Heap-based buffer overflow in drivers/net/macsec.c in the MACsec module in the Linux kernel through 4.10.12 allows attackers to cause a denial of service or possibly have unspecified other impact by leveraging the use of a MAX_SKB_FRAGS+1 size in conjunction with the NETIF_F_FRAGLIST feature, leading to an error in the skb_to_sgvec function.

An issue was discovered in __split_huge_pmd in mm/huge_memory.c in the Linux kernel before 5.7.5. The copy-on-write implementation can grant unintended write access because of a race condition in a THP mapcount check, aka CID-c444eb564fb1.

An issue was discovered in kmem_cache_alloc_bulk in mm/slub.c in the Linux kernel before 5.5.11. The slowpath lacks the required TID increment, aka CID-fd4d9c7d0c71.

An issue was discovered in mm/mmap.c in the Linux kernel before 5.7.11. There is a race condition between certain expand functions (expand_downwards and expand_upwards) and page-table free operations from an munmap call, aka CID-246c320a8cfe.

A TOCTOU mismatch in the NFS client code in the Linux kernel before 5.8.3 could be used by local attackers to corrupt memory or possibly have unspecified other impact because a size check is in fs/nfs/nfs4proc.c instead of fs/nfs/nfs4xdr.c, aka CID-b4487b935452.

A flaw was found in Linux Kernel because access to the global variable fg_console is not properly synchronized leading to a use after free in con_font_op.

A possible unauthorized memory access flaw was found in the Linux kernel's cpu_entry_area mapping of X86 CPU data to memory, where a user may guess the location of exception stacks or other important data. Based on the previous CVE-2023-0597, the 'Randomize per-cpu entry area' feature was implemented in /arch/x86/mm/cpu_entry_area.c, which works through the init_cea_offsets() function when KASLR is enabled. However, despite this feature, there is still a risk of per-cpu entry area leaks. This issue could allow a local user to gain access to some important data with memory in an expected location and potentially escalate their privileges on the system.

A race condition was found in the Linux kernel's IP framework for transforming packets (XFRM subsystem) when multiple calls to xfrm_probe_algs occurred simultaneously. This flaw could allow a local attacker to potentially trigger an out-of-bounds write or leak kernel heap memory by performing an out-of-bounds read and copying it into a socket.

The fix for CVE-2019-11599, affecting the Linux kernel before 5.0.10 was not complete. A local user could use this flaw to obtain sensitive information, cause a denial of service, or possibly have other unspecified impacts by triggering a race condition with mmget_not_zero or get_task_mm calls.

jbd2_journal_wait_updates in fs/jbd2/transaction.c in the Linux kernel before 5.17.1 has a use-after-free caused by a transaction_t race condition.

arch/powerpc/mm/mmu_context_book3s64.c in the Linux kernel before 5.1.15 for powerpc has a bug where unrelated processes may be able to read/write to one another's virtual memory under certain conditions via an mmap above 512 TB. Only a subset of powerpc systems are affected.

In Docker Desktop before v4.29.0, an attacker who has gained access to the Docker Desktop VM through a container breakout can further escape to the host by passing extensions and dashboard related IPC messages. Docker Desktop v4.29.0 https://docs.docker.com/desktop/release-notes/#4290 fixes the issue on MacOS, Linux and Windows with Hyper-V backend. As exploitation requires "Allow only extensions distributed through the Docker Marketplace" to be disabled, Docker Desktop  v4.31.0 https://docs.docker.com/desktop/release-notes/#4310  additionally changes the default configuration to enable this setting by default.

In the Linux kernel, the following vulnerability has been resolved: wifi: iwlegacy: Clear stale interrupts before resuming device iwl4965 fails upon resume from hibernation on my laptop. The reason seems to be a stale interrupt which isn't being cleared out before interrupts are enabled. We end up with a race beween the resume trying to bring things back up, and the restart work (queued form the interrupt handler) trying to bring things down. Eventually the whole thing blows up. Fix the problem by clearing out any stale interrupts before interrupts get enabled during resume. Here's a debug log of the indicent: [ 12.042589] ieee80211 phy0: il_isr ISR inta 0x00000080, enabled 0xaa00008b, fh 0x00000000 [ 12.042625] ieee80211 phy0: il4965_irq_tasklet inta 0x00000080, enabled 0x00000000, fh 0x00000000 [ 12.042651] iwl4965 0000:10:00.0: RF_KILL bit toggled to enable radio. [ 12.042653] iwl4965 0000:10:00.0: On demand firmware reload [ 12.042690] ieee80211 phy0: il4965_irq_tasklet End inta 0x00000000, enabled 0xaa00008b, fh 0x00000000, flags 0x00000282 [ 12.052207] ieee80211 phy0: il4965_mac_start enter [ 12.052212] ieee80211 phy0: il_prep_station Add STA to driver ID 31: ff:ff:ff:ff:ff:ff [ 12.052244] ieee80211 phy0: il4965_set_hw_ready hardware ready [ 12.052324] ieee80211 phy0: il_apm_init Init card's basic functions [ 12.052348] ieee80211 phy0: il_apm_init L1 Enabled; Disabling L0S [ 12.055727] ieee80211 phy0: il4965_load_bsm Begin load bsm [ 12.056140] ieee80211 phy0: il4965_verify_bsm Begin verify bsm [ 12.058642] ieee80211 phy0: il4965_verify_bsm BSM bootstrap uCode image OK [ 12.058721] ieee80211 phy0: il4965_load_bsm BSM write complete, poll 1 iterations [ 12.058734] ieee80211 phy0: __il4965_up iwl4965 is coming up [ 12.058737] ieee80211 phy0: il4965_mac_start Start UP work done. [ 12.058757] ieee80211 phy0: __il4965_down iwl4965 is going down [ 12.058761] ieee80211 phy0: il_scan_cancel_timeout Scan cancel timeout [ 12.058762] ieee80211 phy0: il_do_scan_abort Not performing scan to abort [ 12.058765] ieee80211 phy0: il_clear_ucode_stations Clearing ucode stations in driver [ 12.058767] ieee80211 phy0: il_clear_ucode_stations No active stations found to be cleared [ 12.058819] ieee80211 phy0: _il_apm_stop Stop card, put in low power state [ 12.058827] ieee80211 phy0: _il_apm_stop_master stop master [ 12.058864] ieee80211 phy0: il4965_clear_free_frames 0 frames on pre-allocated heap on clear. [ 12.058869] ieee80211 phy0: Hardware restart was requested [ 16.132299] iwl4965 0000:10:00.0: START_ALIVE timeout after 4000ms. [ 16.132303] ------------[ cut here ]------------ [ 16.132304] Hardware became unavailable upon resume. This could be a software issue prior to suspend or a hardware issue. [ 16.132338] WARNING: CPU: 0 PID: 181 at net/mac80211/util.c:1826 ieee80211_reconfig+0x8f/0x14b0 [mac80211] [ 16.132390] Modules linked in: ctr ccm sch_fq_codel xt_tcpudp xt_multiport xt_state iptable_filter iptable_nat nf_nat nf_conntrack nf_defrag_ipv4 ip_tables x_tables binfmt_misc joydev mousedev btusb btrtl btintel btbcm bluetooth ecdh_generic ecc iTCO_wdt i2c_dev iwl4965 iwlegacy coretemp snd_hda_codec_analog pcspkr psmouse mac80211 snd_hda_codec_generic libarc4 sdhci_pci cqhci sha256_generic sdhci libsha256 firewire_ohci snd_hda_intel snd_intel_dspcfg mmc_core snd_hda_codec snd_hwdep firewire_core led_class iosf_mbi snd_hda_core uhci_hcd lpc_ich crc_itu_t cfg80211 ehci_pci ehci_hcd snd_pcm usbcore mfd_core rfkill snd_timer snd usb_common soundcore video parport_pc parport intel_agp wmi intel_gtt backlight e1000e agpgart evdev [ 16.132456] CPU: 0 UID: 0 PID: 181 Comm: kworker/u8:6 Not tainted 6.11.0-cl+ #143 [ 16.132460] Hardware name: Hewlett-Packard HP Compaq 6910p/30BE, BIOS 68MCU Ver. F.19 07/06/2010 [ 16.132463] Workqueue: async async_run_entry_fn [ 16.132469] RIP: 0010:ieee80211_reconfig+0x8f/0x14b0 [mac80211] [ 16.132501] Code: da 02 00 0 ---truncated---

The coredump implementation in the Linux kernel before 5.0.10 does not use locking or other mechanisms to prevent vma layout or vma flags changes while it runs, which allows local users to obtain sensitive information, cause a denial of service, or possibly have unspecified other impact by triggering a race condition with mmget_not_zero or get_task_mm calls. This is related to fs/userfaultfd.c, mm/mmap.c, fs/proc/task_mmu.c, and drivers/infiniband/core/uverbs_main.c.

The Siemens R3964 line discipline driver in drivers/tty/n_r3964.c in the Linux kernel before 5.0.8 has multiple race conditions.

In the Linux kernel, the following vulnerability has been resolved: mm/mremap: fix move_normal_pmd/retract_page_tables race In mremap(), move_page_tables() looks at the type of the PMD entry and the specified address range to figure out by which method the next chunk of page table entries should be moved. At that point, the mmap_lock is held in write mode, but no rmap locks are held yet. For PMD entries that point to page tables and are fully covered by the source address range, move_pgt_entry(NORMAL_PMD, ...) is called, which first takes rmap locks, then does move_normal_pmd(). move_normal_pmd() takes the necessary page table locks at source and destination, then moves an entire page table from the source to the destination. The problem is: The rmap locks, which protect against concurrent page table removal by retract_page_tables() in the THP code, are only taken after the PMD entry has been read and it has been decided how to move it. So we can race as follows (with two processes that have mappings of the same tmpfs file that is stored on a tmpfs mount with huge=advise); note that process A accesses page tables through the MM while process B does it through the file rmap: process A process B ========= ========= mremap mremap_to move_vma move_page_tables get_old_pmd alloc_new_pmd *** PREEMPT *** madvise(MADV_COLLAPSE) do_madvise madvise_walk_vmas madvise_vma_behavior madvise_collapse hpage_collapse_scan_file collapse_file retract_page_tables i_mmap_lock_read(mapping) pmdp_collapse_flush i_mmap_unlock_read(mapping) move_pgt_entry(NORMAL_PMD, ...) take_rmap_locks move_normal_pmd drop_rmap_locks When this happens, move_normal_pmd() can end up creating bogus PMD entries in the line `pmd_populate(mm, new_pmd, pmd_pgtable(pmd))`. The effect depends on arch-specific and machine-specific details; on x86, you can end up with physical page 0 mapped as a page table, which is likely exploitable for user->kernel privilege escalation. Fix the race by letting process B recheck that the PMD still points to a page table after the rmap locks have been taken. Otherwise, we bail and let the caller fall back to the PTE-level copying path, which will then bail immediately at the pmd_none() check. Bug reachability: Reaching this bug requires that you can create shmem/file THP mappings - anonymous THP uses different code that doesn't zap stuff under rmap locks. File THP is gated on an experimental config flag (CONFIG_READ_ONLY_THP_FOR_FS), so on normal distro kernels you need shmem THP to hit this bug. As far as I know, getting shmem THP normally requires that you can mount your own tmpfs with the right mount flags, which would require creating your own user+mount namespace; though I don't know if some distros maybe enable shmem THP by default or something like that. Bug impact: This issue can likely be used for user->kernel privilege escalation when it is reachable.