Vulnerabilities

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> perf/core: Exit early on perf_mmap() fail<br /> <br /> When perf_mmap() fails to allocate a buffer, it still invokes the<br /> event_mapped() callback of the related event. On X86 this might increase<br /> the perf_rdpmc_allowed reference counter. But nothing undoes this as<br /> perf_mmap_close() is never called in this case, which causes another<br /> reference count leak.<br /> <br /> Return early on failure to prevent that.

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> nfsd: avoid ref leak in nfsd_open_local_fh()<br /> <br /> If two calls to nfsd_open_local_fh() race and both successfully call<br /> nfsd_file_acquire_local(), they will both get an extra reference to the<br /> net to accompany the file reference stored in *pnf.<br /> <br /> One of them will fail to store (using xchg()) the file reference in<br /> *pnf and will drop that reference but WON&amp;#39;T drop the accompanying<br /> reference to the net. This leak means that when the nfs server is shut<br /> down it will hang in nfsd_shutdown_net() waiting for<br /> &amp;nn-&gt;nfsd_net_free_done.<br /> <br /> This patch adds the missing nfsd_net_put().

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> benet: fix BUG when creating VFs<br /> <br /> benet crashes as soon as SRIOV VFs are created:<br /> <br /> kernel BUG at mm/vmalloc.c:3457!<br /> Oops: invalid opcode: 0000 [#1] SMP KASAN NOPTI<br /> CPU: 4 UID: 0 PID: 7408 Comm: test.sh Kdump: loaded Not tainted 6.16.0+ #1 PREEMPT(voluntary)<br /> [...]<br /> RIP: 0010:vunmap+0x5f/0x70<br /> [...]<br /> Call Trace:<br /> <br /> __iommu_dma_free+0xe8/0x1c0<br /> be_cmd_set_mac_list+0x3fe/0x640 [be2net]<br /> be_cmd_set_mac+0xaf/0x110 [be2net]<br /> be_vf_eth_addr_config+0x19f/0x330 [be2net]<br /> be_vf_setup+0x4f7/0x990 [be2net]<br /> be_pci_sriov_configure+0x3a1/0x470 [be2net]<br /> sriov_numvfs_store+0x20b/0x380<br /> kernfs_fop_write_iter+0x354/0x530<br /> vfs_write+0x9b9/0xf60<br /> ksys_write+0xf3/0x1d0<br /> do_syscall_64+0x8c/0x3d0<br /> <br /> be_cmd_set_mac_list() calls dma_free_coherent() under a spin_lock_bh.<br /> Fix it by freeing only after the lock has been released.

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> sunrpc: fix handling of server side tls alerts<br /> <br /> Scott Mayhew discovered a security exploit in NFS over TLS in<br /> tls_alert_recv() due to its assumption it can read data from<br /> the msg iterator&amp;#39;s kvec..<br /> <br /> kTLS implementation splits TLS non-data record payload between<br /> the control message buffer (which includes the type such as TLS<br /> aler or TLS cipher change) and the rest of the payload (say TLS<br /> alert&amp;#39;s level/description) which goes into the msg payload buffer.<br /> <br /> This patch proposes to rework how control messages are setup and<br /> used by sock_recvmsg().<br /> <br /> If no control message structure is setup, kTLS layer will read and<br /> process TLS data record types. As soon as it encounters a TLS control<br /> message, it would return an error. At that point, NFS can setup a<br /> kvec backed msg buffer and read in the control message such as a<br /> TLS alert. Msg iterator can advance the kvec pointer as a part of<br /> the copy process thus we need to revert the iterator before calling<br /> into the tls_alert_recv.

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> net/sched: mqprio: fix stack out-of-bounds write in tc entry parsing<br /> <br /> TCA_MQPRIO_TC_ENTRY_INDEX is validated using<br /> NLA_POLICY_MAX(NLA_U32, TC_QOPT_MAX_QUEUE), which allows the value<br /> TC_QOPT_MAX_QUEUE (16). This leads to a 4-byte out-of-bounds stack<br /> write in the fp[] array, which only has room for 16 elements (0–15).<br /> <br /> Fix this by changing the policy to allow only up to TC_QOPT_MAX_QUEUE - 1.

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> eth: fbnic: unlink NAPIs from queues on error to open<br /> <br /> CI hit a UaF in fbnic in the AF_XDP portion of the queues.py test.<br /> The UaF is in the __sk_mark_napi_id_once() call in xsk_bind(),<br /> NAPI has been freed. Looks like the device failed to open earlier,<br /> and we lack clearing the NAPI pointer from the queue.

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> sunrpc: fix client side handling of tls alerts<br /> <br /> A security exploit was discovered in NFS over TLS in tls_alert_recv<br /> due to its assumption that there is valid data in the msghdr&amp;#39;s<br /> iterator&amp;#39;s kvec.<br /> <br /> Instead, this patch proposes the rework how control messages are<br /> setup and used by sock_recvmsg().<br /> <br /> If no control message structure is setup, kTLS layer will read and<br /> process TLS data record types. As soon as it encounters a TLS control<br /> message, it would return an error. At that point, NFS can setup a kvec<br /> backed control buffer and read in the control message such as a TLS<br /> alert. Scott found that a msg iterator can advance the kvec pointer<br /> as a part of the copy process thus we need to revert the iterator<br /> before calling into the tls_alert_recv.

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> usb: gadget: uvc: Initialize frame-based format color matching descriptor<br /> <br /> Fix NULL pointer crash in uvcg_framebased_make due to uninitialized color<br /> matching descriptor for frame-based format which was added in<br /> commit f5e7bdd34aca ("usb: gadget: uvc: Allow creating new color matching<br /> descriptors") that added handling for uncompressed and mjpeg format.<br /> <br /> Crash is seen when userspace configuration (via configfs) does not<br /> explicitly define the color matching descriptor. If color_matching is not<br /> found, config_group_find_item() returns NULL. The code then jumps to<br /> out_put_cm, where it calls config_item_put(color_matching);. If<br /> color_matching is NULL, this will dereference a null pointer, leading to a<br /> crash.<br /> <br /> [ 2.746440] Unable to handle kernel NULL pointer dereference at virtual address 000000000000008c<br /> [ 2.756273] Mem abort info:<br /> [ 2.760080] ESR = 0x0000000096000005<br /> [ 2.764872] EC = 0x25: DABT (current EL), IL = 32 bits<br /> [ 2.771068] SET = 0, FnV = 0<br /> [ 2.771069] EA = 0, S1PTW = 0<br /> [ 2.771070] FSC = 0x05: level 1 translation fault<br /> [ 2.771071] Data abort info:<br /> [ 2.771072] ISV = 0, ISS = 0x00000005, ISS2 = 0x00000000<br /> [ 2.771073] CM = 0, WnR = 0, TnD = 0, TagAccess = 0<br /> [ 2.771074] GCS = 0, Overlay = 0, DirtyBit = 0, Xs = 0<br /> [ 2.771075] user pgtable: 4k pages, 39-bit VAs, pgdp=00000000a3e59000<br /> [ 2.771077] [000000000000008c] pgd=0000000000000000, p4d=0000000000000000, pud=0000000000000000<br /> [ 2.771081] Internal error: Oops: 0000000096000005 [#1] PREEMPT SMP<br /> [ 2.771084] Dumping ftrace buffer:<br /> [ 2.771085] (ftrace buffer empty)<br /> [ 2.771138] CPU: 7 PID: 486 Comm: ln Tainted: G W E 6.6.58-android15<br /> [ 2.771139] Hardware name: Qualcomm Technologies, Inc. SunP QRD HDK (DT)<br /> [ 2.771140] pstate: 61400005 (nZCv daif +PAN -UAO -TCO +DIT -SSBS BTYPE=--)<br /> [ 2.771141] pc : __uvcg_fill_strm+0x198/0x2cc<br /> [ 2.771145] lr : __uvcg_iter_strm_cls+0xc8/0x17c<br /> [ 2.771146] sp : ffffffc08140bbb0<br /> [ 2.771146] x29: ffffffc08140bbb0 x28: ffffff803bc81380 x27: ffffff8023bbd250<br /> [ 2.771147] x26: ffffff8023bbd250 x25: ffffff803c361348 x24: ffffff803d8e6768<br /> [ 2.771148] x23: 0000000000000004 x22: 0000000000000003 x21: ffffffc08140bc48<br /> [ 2.771149] x20: 0000000000000000 x19: ffffffc08140bc48 x18: ffffffe9f8cf4a00<br /> [ 2.771150] x17: 000000001bf64ec3 x16: 000000001bf64ec3 x15: ffffff8023bbd250<br /> [ 2.771151] x14: 000000000000000f x13: 004c4b40000f4240 x12: 000a2c2a00051615<br /> [ 2.771152] x11: 000000000000004f x10: ffffffe9f76b40ec x9 : ffffffe9f7e389d0<br /> [ 2.771153] x8 : ffffff803d0d31ce x7 : 000f4240000a2c2a x6 : 0005161500028b0a<br /> [ 2.771154] x5 : ffffff803d0d31ce x4 : 0000000000000003 x3 : 0000000000000000<br /> [ 2.771155] x2 : ffffffc08140bc50 x1 : ffffffc08140bc48 x0 : 0000000000000000<br /> [ 2.771156] Call trace:<br /> [ 2.771157] __uvcg_fill_strm+0x198/0x2cc<br /> [ 2.771157] __uvcg_iter_strm_cls+0xc8/0x17c<br /> [ 2.771158] uvcg_streaming_class_allow_link+0x240/0x290<br /> [ 2.771159] configfs_symlink+0x1f8/0x630<br /> [ 2.771161] vfs_symlink+0x114/0x1a0<br /> [ 2.771163] do_symlinkat+0x94/0x28c<br /> [ 2.771164] __arm64_sys_symlinkat+0x54/0x70<br /> [ 2.771164] invoke_syscall+0x58/0x114<br /> [ 2.771166] el0_svc_common+0x80/0xe0<br /> [ 2.771168] do_el0_svc+0x1c/0x28<br /> [ 2.771169] el0_svc+0x3c/0x70<br /> [ 2.771172] el0t_64_sync_handler+0x68/0xbc<br /> [ 2.771173] el0t_64_sync+0x1a8/0x1ac<br /> <br /> Initialize color matching descriptor for frame-based format to prevent<br /> NULL pointer crash by mirroring the handling done for uncompressed and<br /> mjpeg formats.

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> platform/x86/intel/pmt: fix a crashlog NULL pointer access<br /> <br /> Usage of the intel_pmt_read() for binary sysfs, requires a pcidev. The<br /> current use of the endpoint value is only valid for telemetry endpoint<br /> usage.<br /> <br /> Without the ep, the crashlog usage causes the following NULL pointer<br /> exception:<br /> <br /> BUG: kernel NULL pointer dereference, address: 0000000000000000<br /> Oops: Oops: 0000 [#1] SMP NOPTI<br /> RIP: 0010:intel_pmt_read+0x3b/0x70 [pmt_class]<br /> Code:<br /> Call Trace:<br /> <br /> ? sysfs_kf_bin_read+0xc0/0xe0<br /> kernfs_fop_read_iter+0xac/0x1a0<br /> vfs_read+0x26d/0x350<br /> ksys_read+0x6b/0xe0<br /> __x64_sys_read+0x1d/0x30<br /> x64_sys_call+0x1bc8/0x1d70<br /> do_syscall_64+0x6d/0x110<br /> <br /> Augment struct intel_pmt_entry with a pointer to the pcidev to avoid<br /> the NULL pointer exception.

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> ksmbd: fix null pointer dereference error in generate_encryptionkey<br /> <br /> If client send two session setups with krb5 authenticate to ksmbd,<br /> null pointer dereference error in generate_encryptionkey could happen.<br /> sess-&gt;Preauth_HashValue is set to NULL if session is valid.<br /> So this patch skip generate encryption key if session is valid.

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> perf/core: Prevent VMA split of buffer mappings<br /> <br /> The perf mmap code is careful about mmap()&amp;#39;ing the user page with the<br /> ringbuffer and additionally the auxiliary buffer, when the event supports<br /> it. Once the first mapping is established, subsequent mapping have to use<br /> the same offset and the same size in both cases. The reference counting for<br /> the ringbuffer and the auxiliary buffer depends on this being correct.<br /> <br /> Though perf does not prevent that a related mapping is split via mmap(2),<br /> munmap(2) or mremap(2). A split of a VMA results in perf_mmap_open() calls,<br /> which take reference counts, but then the subsequent perf_mmap_close()<br /> calls are not longer fulfilling the offset and size checks. This leads to<br /> reference count leaks.<br /> <br /> As perf already has the requirement for subsequent mappings to match the<br /> initial mapping, the obvious consequence is that VMA splits, caused by<br /> resizing of a mapping or partial unmapping, have to be prevented.<br /> <br /> Implement the vm_operations_struct::may_split() callback and return<br /> unconditionally -EINVAL.<br /> <br /> That ensures that the mapping offsets and sizes cannot be changed after the<br /> fact. Remapping to a different fixed address with the same size is still<br /> possible as it takes the references for the new mapping and drops those of<br /> the old mapping.

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> perf/core: Handle buffer mapping fail correctly in perf_mmap()<br /> <br /> After successful allocation of a buffer or a successful attachment to an<br /> existing buffer perf_mmap() tries to map the buffer read only into the page<br /> table. If that fails, the already set up page table entries are zapped, but<br /> the other perf specific side effects of that failure are not handled. The<br /> calling code just cleans up the VMA and does not invoke perf_mmap_close().<br /> <br /> This leaks reference counts, corrupts user-&gt;vm accounting and also results<br /> in an unbalanced invocation of event::event_mapped().<br /> <br /> Cure this by moving the event::event_mapped() invocation before the<br /> map_range() call so that on map_range() failure perf_mmap_close() can be<br /> invoked without causing an unbalanced event::event_unmapped() call.<br /> <br /> perf_mmap_close() undoes the reference counts and eventually frees buffers.