user/sven/linux.git/include/linux, branch v6.16.1

mm: fix a UAF when vma->mm is freed after vma->vm_refcnt got dropped

2025-08-15T14:39:36Z

commit 9bbffee67ffd16360179327b57f3b1245579ef08 upstream. By inducing delays in the right places, Jann Horn created a reproducer for a hard to hit UAF issue that became possible after VMAs were allowed to be recycled by adding SLAB_TYPESAFE_BY_RCU to their cache. Race description is borrowed from Jann's discovery report: lock_vma_under_rcu() looks up a VMA locklessly with mas_walk() under rcu_read_lock(). At that point, the VMA may be concurrently freed, and it can be recycled by another process. vma_start_read() then increments the vma->vm_refcnt (if it is in an acceptable range), and if this succeeds, vma_start_read() can return a recycled VMA. In this scenario where the VMA has been recycled, lock_vma_under_rcu() will then detect the mismatching ->vm_mm pointer and drop the VMA through vma_end_read(), which calls vma_refcount_put(). vma_refcount_put() drops the refcount and then calls rcuwait_wake_up() using a copy of vma->vm_mm. This is wrong: It implicitly assumes that the caller is keeping the VMA's mm alive, but in this scenario the caller has no relation to the VMA's mm, so the rcuwait_wake_up() can cause UAF. The diagram depicting the race: T1 T2 T3 == == == lock_vma_under_rcu mas_walk mmap vma_start_read __refcount_inc_not_zero_limited_acquire munmap __vma_enter_locked refcount_add_not_zero vma_end_read vma_refcount_put __refcount_dec_and_test rcuwait_wait_event rcuwait_wake_up [UAF] Note that rcuwait_wait_event() in T3 does not block because refcount was already dropped by T1. At this point T3 can exit and free the mm causing UAF in T1. To avoid this we move vma->vm_mm verification into vma_start_read() and grab vma->vm_mm to stabilize it before vma_refcount_put() operation. [surenb@google.com: v3] Link: https://lkml.kernel.org/r/20250729145709.2731370-1-surenb@google.com Link: https://lkml.kernel.org/r/20250728175355.2282375-1-surenb@google.com Fixes: 3104138517fc ("mm: make vma cache SLAB_TYPESAFE_BY_RCU") Signed-off-by: Suren Baghdasaryan Reported-by: Jann Horn Closes: https://lore.kernel.org/all/CAG48ez0-deFbVH=E3jbkWx=X3uVbd8nWeo6kbJPQ0KoUD+m2tA@mail.gmail.com/ Reviewed-by: Vlastimil Babka Acked-by: Lorenzo Stoakes Cc: Jann Horn Cc: Liam Howlett Cc: Signed-off-by: Andrew Morton Signed-off-by: Suren Baghdasaryan Signed-off-by: Greg Kroah-Hartman

net: usbnet: Avoid potential RCU stall on LINK_CHANGE event

2025-08-15T14:39:32Z

commit 0d9cfc9b8cb17dbc29a98792d36ec39a1cf1395f upstream. The Gemalto Cinterion PLS83-W modem (cdc_ether) is emitting confusing link up and down events when the WWAN interface is activated on the modem-side. Interrupt URBs will in consecutive polls grab: * Link Connected * Link Disconnected * Link Connected Where the last Connected is then a stable link state. When the system is under load this may cause the unlink_urbs() work in __handle_link_change() to not complete before the next usbnet_link_change() call turns the carrier on again, allowing rx_submit() to queue new SKBs. In that event the URB queue is filled faster than it can drain, ending up in a RCU stall: rcu: INFO: rcu_sched detected expedited stalls on CPUs/tasks: { 0-.... } 33108 jiffies s: 201 root: 0x1/. rcu: blocking rcu_node structures (internal RCU debug): Sending NMI from CPU 1 to CPUs 0: NMI backtrace for cpu 0 Call trace: arch_local_irq_enable+0x4/0x8 local_bh_enable+0x18/0x20 __netdev_alloc_skb+0x18c/0x1cc rx_submit+0x68/0x1f8 [usbnet] rx_alloc_submit+0x4c/0x74 [usbnet] usbnet_bh+0x1d8/0x218 [usbnet] usbnet_bh_tasklet+0x10/0x18 [usbnet] tasklet_action_common+0xa8/0x110 tasklet_action+0x2c/0x34 handle_softirqs+0x2cc/0x3a0 __do_softirq+0x10/0x18 ____do_softirq+0xc/0x14 call_on_irq_stack+0x24/0x34 do_softirq_own_stack+0x18/0x20 __irq_exit_rcu+0xa8/0xb8 irq_exit_rcu+0xc/0x30 el1_interrupt+0x34/0x48 el1h_64_irq_handler+0x14/0x1c el1h_64_irq+0x68/0x6c _raw_spin_unlock_irqrestore+0x38/0x48 xhci_urb_dequeue+0x1ac/0x45c [xhci_hcd] unlink1+0xd4/0xdc [usbcore] usb_hcd_unlink_urb+0x70/0xb0 [usbcore] usb_unlink_urb+0x24/0x44 [usbcore] unlink_urbs.constprop.0.isra.0+0x64/0xa8 [usbnet] __handle_link_change+0x34/0x70 [usbnet] usbnet_deferred_kevent+0x1c0/0x320 [usbnet] process_scheduled_works+0x2d0/0x48c worker_thread+0x150/0x1dc kthread+0xd8/0xe8 ret_from_fork+0x10/0x20 Get around the problem by delaying the carrier on to the scheduled work. This needs a new flag to keep track of the necessary action. The carrier ok check cannot be removed as it remains required for the LINK_RESET event flow. Fixes: 4b49f58fff00 ("usbnet: handle link change") Cc: stable@vger.kernel.org Signed-off-by: John Ernberg Link: https://patch.msgid.link/20250723102526.1305339-1-john.ernberg@actia.se Signed-off-by: Jakub Kicinski Signed-off-by: Greg Kroah-Hartman

ipv6: reject malicious packets in ipv6_gso_segment()

2025-08-15T14:39:25Z

[ Upstream commit d45cf1e7d7180256e17c9ce88e32e8061a7887fe ] syzbot was able to craft a packet with very long IPv6 extension headers leading to an overflow of skb->transport_header. This 16bit field has a limited range. Add skb_reset_transport_header_careful() helper and use it from ipv6_gso_segment() WARNING: CPU: 0 PID: 5871 at ./include/linux/skbuff.h:3032 skb_reset_transport_header include/linux/skbuff.h:3032 [inline] WARNING: CPU: 0 PID: 5871 at ./include/linux/skbuff.h:3032 ipv6_gso_segment+0x15e2/0x21e0 net/ipv6/ip6_offload.c:151 Modules linked in: CPU: 0 UID: 0 PID: 5871 Comm: syz-executor211 Not tainted 6.16.0-rc6-syzkaller-g7abc678e3084 #0 PREEMPT(full) Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 07/12/2025 RIP: 0010:skb_reset_transport_header include/linux/skbuff.h:3032 [inline] RIP: 0010:ipv6_gso_segment+0x15e2/0x21e0 net/ipv6/ip6_offload.c:151 Call Trace: skb_mac_gso_segment+0x31c/0x640 net/core/gso.c:53 nsh_gso_segment+0x54a/0xe10 net/nsh/nsh.c:110 skb_mac_gso_segment+0x31c/0x640 net/core/gso.c:53 __skb_gso_segment+0x342/0x510 net/core/gso.c:124 skb_gso_segment include/net/gso.h:83 [inline] validate_xmit_skb+0x857/0x11b0 net/core/dev.c:3950 validate_xmit_skb_list+0x84/0x120 net/core/dev.c:4000 sch_direct_xmit+0xd3/0x4b0 net/sched/sch_generic.c:329 __dev_xmit_skb net/core/dev.c:4102 [inline] __dev_queue_xmit+0x17b6/0x3a70 net/core/dev.c:4679 Fixes: d1da932ed4ec ("ipv6: Separate ipv6 offload support") Reported-by: syzbot+af43e647fd835acc02df@syzkaller.appspotmail.com Closes: https://lore.kernel.org/netdev/688a1a05.050a0220.5d226.0008.GAE@google.com/T/#u Signed-off-by: Eric Dumazet Reviewed-by: Dawid Osuchowski Reviewed-by: Willem de Bruijn Link: https://patch.msgid.link/20250730131738.3385939-1-edumazet@google.com Signed-off-by: Jakub Kicinski Signed-off-by: Sasha Levin

block: Fix default IO priority if there is no IO context

2025-08-15T14:39:24Z

[ Upstream commit e2ba58ccc9099514380c3300cbc0750b5055fc1c ] Upstream commit 53889bcaf536 ("block: make __get_task_ioprio() easier to read") changes the IO priority returned to the caller if no IO context is defined for the task. Prior to this commit, the returned IO priority was determined by task_nice_ioclass() and task_nice_ioprio(). Now it is always IOPRIO_DEFAULT, which translates to IOPRIO_CLASS_NONE with priority 0. However, task_nice_ioclass() returns IOPRIO_CLASS_IDLE, IOPRIO_CLASS_RT, or IOPRIO_CLASS_BE depending on the task scheduling policy, and task_nice_ioprio() returns a value determined by task_nice(). This causes regressions in test code checking the IO priority and class of IO operations on tasks with no IO context. Fix the problem by returning the IO priority calculated from task_nice_ioclass() and task_nice_ioprio() if no IO context is defined to match earlier behavior. Fixes: 53889bcaf536 ("block: make __get_task_ioprio() easier to read") Cc: Jens Axboe Cc: Bart Van Assche Signed-off-by: Guenter Roeck Reviewed-by: Yu Kuai Reviewed-by: Damien Le Moal Link: https://lore.kernel.org/r/20250731044953.1852690-1-linux@roeck-us.net Signed-off-by: Jens Axboe Signed-off-by: Sasha Levin

padata: Remove comment for reorder_work

2025-08-15T14:39:21Z

[ Upstream commit 82a0302e7167d0b7c6cde56613db3748f8dd806d ] Remove comment for reorder_work which no longer exists. Reported-by: Stephen Rothwell Fixes: 71203f68c774 ("padata: Fix pd UAF once and for all") Signed-off-by: Herbert Xu Signed-off-by: Sasha Levin

vfio/pci: Do vf_token checks for VFIO_DEVICE_BIND_IOMMUFD

2025-08-15T14:39:21Z

[ Upstream commit 86624ba3b522b6512def25534341da93356c8da4 ] This was missed during the initial implementation. The VFIO PCI encodes the vf_token inside the device name when opening the device from the group FD, something like: "0000:04:10.0 vf_token=bd8d9d2b-5a5f-4f5a-a211-f591514ba1f3" This is used to control access to a VF unless there is co-ordination with the owner of the PF. Since we no longer have a device name in the cdev path, pass the token directly through VFIO_DEVICE_BIND_IOMMUFD using an optional field indicated by VFIO_DEVICE_BIND_FLAG_TOKEN. Fixes: 5fcc26969a16 ("vfio: Add VFIO_DEVICE_BIND_IOMMUFD") Tested-by: Shameer Kolothum Reviewed-by: Yi Liu Signed-off-by: Jason Gunthorpe Reviewed-by: Kevin Tian Link: https://lore.kernel.org/r/0-v3-bdd8716e85fe+3978a-vfio_token_jgg@nvidia.com Signed-off-by: Alex Williamson Signed-off-by: Sasha Levin

module: Restore the moduleparam prefix length check

2025-08-15T14:39:16Z

[ Upstream commit bdc877ba6b7ff1b6d2ebeff11e63da4a50a54854 ] The moduleparam code allows modules to provide their own definition of MODULE_PARAM_PREFIX, instead of using the default KBUILD_MODNAME ".". Commit 730b69d22525 ("module: check kernel param length at compile time, not runtime") added a check to ensure the prefix doesn't exceed MODULE_NAME_LEN, as this is what param_sysfs_builtin() expects. Later, commit 58f86cc89c33 ("VERIFY_OCTAL_PERMISSIONS: stricter checking for sysfs perms.") removed this check, but there is no indication this was intentional. Since the check is still useful for param_sysfs_builtin() to function properly, reintroduce it in __module_param_call(), but in a modernized form using static_assert(). While here, clean up the __module_param_call() comments. In particular, remove the comment "Default value instead of permissions?", which comes from commit 9774a1f54f17 ("[PATCH] Compile-time check re world-writeable module params"). This comment was related to the test variable __param_perm_check_##name, which was removed in the previously mentioned commit 58f86cc89c33. Fixes: 58f86cc89c33 ("VERIFY_OCTAL_PERMISSIONS: stricter checking for sysfs perms.") Signed-off-by: Petr Pavlu Reviewed-by: Daniel Gomez Link: https://lore.kernel.org/r/20250630143535.267745-4-petr.pavlu@suse.com Signed-off-by: Daniel Gomez Signed-off-by: Sasha Levin

i3c: fix module_i3c_i2c_driver() with I3C=n

2025-08-15T14:39:16Z

[ Upstream commit 5523a466e905b6287b94654ddb364536f2f948cf ] When CONFIG_I3C is disabled and the i3c_i2c_driver_register() happens to not be inlined, any driver calling it still references the i3c_driver instance, which then causes a link failure: x86_64-linux-ld: drivers/hwmon/lm75.o: in function `lm75_i3c_reg_read': lm75.c:(.text+0xc61): undefined reference to `i3cdev_to_dev' x86_64-linux-ld: lm75.c:(.text+0xd25): undefined reference to `i3c_device_do_priv_xfers' x86_64-linux-ld: lm75.c:(.text+0xdd8): undefined reference to `i3c_device_do_priv_xfers' This issue was part of the original i3c code, but only now caused problems when i3c support got added to lm75. Change the 'inline' annotations in the header to '__always_inline' to ensure that the dead-code-elimination pass in the compiler can optimize it out as intended. Fixes: 6071d10413ff ("hwmon: (lm75) add I3C support for P3T1755") Fixes: 3a379bbcea0a ("i3c: Add core I3C infrastructure") Signed-off-by: Arnd Bergmann Reviewed-by: Randy Dunlap Tested-by: Randy Dunlap Reviewed-by: Guenter Roeck Reviewed-by: Frank Li Link: https://lore.kernel.org/r/20250725090609.2456262-1-arnd@kernel.org Signed-off-by: Alexandre Belloni Signed-off-by: Sasha Levin

bpf: Fix oob access in cgroup local storage

2025-08-15T14:39:14Z

[ Upstream commit abad3d0bad72a52137e0c350c59542d75ae4f513 ] Lonial reported that an out-of-bounds access in cgroup local storage can be crafted via tail calls. Given two programs each utilizing a cgroup local storage with a different value size, and one program doing a tail call into the other. The verifier will validate each of the indivial programs just fine. However, in the runtime context the bpf_cg_run_ctx holds an bpf_prog_array_item which contains the BPF program as well as any cgroup local storage flavor the program uses. Helpers such as bpf_get_local_storage() pick this up from the runtime context: ctx = container_of(current->bpf_ctx, struct bpf_cg_run_ctx, run_ctx); storage = ctx->prog_item->cgroup_storage[stype]; if (stype == BPF_CGROUP_STORAGE_SHARED) ptr = &READ_ONCE(storage->buf)->data[0]; else ptr = this_cpu_ptr(storage->percpu_buf); For the second program which was called from the originally attached one, this means bpf_get_local_storage() will pick up the former program's map, not its own. With mismatching sizes, this can result in an unintended out-of-bounds access. To fix this issue, we need to extend bpf_map_owner with an array of storage_cookie[] to match on i) the exact maps from the original program if the second program was using bpf_get_local_storage(), or ii) allow the tail call combination if the second program was not using any of the cgroup local storage maps. Fixes: 7d9c3427894f ("bpf: Make cgroup storages shared between programs on the same cgroup") Reported-by: Lonial Con Signed-off-by: Daniel Borkmann Link: https://lore.kernel.org/r/20250730234733.530041-4-daniel@iogearbox.net Signed-off-by: Alexei Starovoitov Signed-off-by: Sasha Levin

bpf: Move cgroup iterator helpers to bpf.h

2025-08-15T14:39:14Z

[ Upstream commit 9621e60f59eae87eb9ffe88d90f24f391a1ef0f0 ] Move them into bpf.h given we also need them in core code. Signed-off-by: Daniel Borkmann Link: https://lore.kernel.org/r/20250730234733.530041-3-daniel@iogearbox.net Signed-off-by: Alexei Starovoitov Stable-dep-of: abad3d0bad72 ("bpf: Fix oob access in cgroup local storage") Signed-off-by: Sasha Levin