CVE-2026-72107

Summary (CVE-2026-72107) In the Linux kernel, the vulnerability is in the dm-era code path where writes are tracked in target-relative blocks. The root cause is that era_map() calculates the writeset block before applying the target offset, allowing a non-zero start sector to pass an absolute map...

CVE-2026-72105

The CVE-2026-72105 entry concerns the Linux kernel dm-log component, where a bitset_size overflow on 32‑bit architectures could occur during region_count handling. The root cause is a miscalculation of bitset_size (size_t) when computing the bitmap memory based on region_count, which could wrap t...

CVE-2026-72106

In the Linux kernel, the dm-ioctl path in list_version_get_info has a potential buffer overflow: sizeof(tt->version) is 12 bytes, but the code writes 16 bytes into the output buffer (info->vers->version[0..2] and info->vers->next). The fix replaces sizeof(tt->version) with sizeo...

CVE-2026-72104

The CVE-2026-72104 entry concerns the Linux kernel component dm-pcache (pcache target). The vulnerability arises when optional arguments are parsed as name/value pairs and a table advertises an option (e.g., cache_mode) but provides only the option name; after consuming the name, the parser decre...

CVE-2026-72103

CVE-2026-72103 relates to the Linux kernel’s dm-crypt flow. A refactoring bug caused the caller’s thread keyring to be kept alive beyond the caller’s lifetime, making the LUKS volume key linger in memory. As a result, cryptsetup luksSuspend could fail to wipe the key, even though luksOpen passes ...

CVE-2026-72102

The CVE-2026-72102 entry concerns the Linux kernel. Affected area: dm_early_create logic related to dynamic device-mapper table handling. Root cause: when dm_resume fails, a table is freed via dm_table_destroy even though it was already instantiated with dm_swap_table, leading to a potential use-...

CVE-2026-72101

The CVE-2026-72101 involves the Linux kernel dm‑integrity module. A bug caused zeroing of the remaining space to fail when the hash size was smaller than the device tuple size, leading to leaking uninitialized kernel memory. The issue has been fixed by a commit that corrects the zeroing behavior....

CVE-2026-72099

Summary (CVE-2026-72099) In the Linux kernel, the dm-integrity component had an issue where hash_offset could be incremented twice; the loop already increments hash_offset, so a second increment was unnecessary. The vulnerability was resolved by correcting this behavior in the relevant code path ...

CVE-2026-72100

CVE-2026-72100 concerns the Linux kernel function dm-integrity. A bug in dm_integrity_check_limits could exit with DM_MAPIO_KILL while the affected range was still locked, causing a potential deadlock and leaked resources. The fix moves the limit check up so that on exit no resources are leaked. ...

CVE-2026-72098

The CVE-2026-72098 issue affects the Linux kernel’s dm-verity FEC calculation. A buffer overflow occurs in dm-verity-fec when computing erasures: if (neras && *neras fec->roots) fio->erasures[(*neras)++] = i;. The post-increment can push *neras to roots+1, which is passed as no_eras to deco...

CVE-2026-72096

CVE-2026-72096 affects the Linux kernel’s dm-verity component, where the error counter v->corrupted_errs was not atomic, potentially causing race conditions. This could allow the call to dm_audit_log_target("max-corrupted-errors") to be skipped due to races. The issue has been resolved in the ...

CVE-2026-72097

CVE-2026-72097 involves the Linux kernel, fixing a possible NULL pointer dereference in dm-verity, occurring in dm_verity_loadpin_is_bdev_trusted when a device has no table. The document set does not provide exploitation details; it simply notes the fix. Patches are referenced in kernel.org commi...

CVE-2026-72095

The CVE-2026-72095 issue affects the Linux kernel dma_fence_dedup_array() function, which incorrectly returned 1 for a 0-length input, contradicting the documented contract. The bug could allow a zero local count path to reach a dedup call and dereference an uninitialized fence slot (notably in a...

CVE-2026-72094

The CVE-2026-72094 entry describes a Linux kernel vulnerability in dma-buf: dma-fence where a race between the signaled-bit and the dma_fence_ops pointer could allow a NULL dereference on weakly ordered platforms. The fix decouples fence producers and consumers by using the signaled-bit as the de...

CVE-2026-72092

The CVE-2026-72092 issue affects the Linux kernel’s accel/amdxdna path: aie2_dev_ops defines cmd_submit while aie4_vf_ops does not, so the AMDXDNA_EXEC_CMD ioctl on AIE4 devices may call a NULL function pointer and oops. The kernel now rejects submission early with -EOPNOTSUPP when no cmd_submit ...

CVE-2026-72093

CVE-2026-72093 (Linux kernel) : The vulnerability concerns the accel/amdxdna path for amdxdna_gem_dmabuf_mmap. When vm_insert_pages() fails, the error path invokes vma->vm_ops->close(vma), which calls drm_gem_vm_close() → drm_gem_object_put() to release the GEM object. However, the close_vm...

CVE-2026-72091

The CVE-2026-72091 in the Linux kernel accelerator amdxdna driver describes a NULL dereference when a user submits an AMDXDNA_EXEC_CMD with cmd handle 0 (AMDXDNA_INVALID_BO_HANDLE). The sequence: accel/amdxdna: submit_execbuf passes a user BO handle to amdxdna_cmd_submit with drv_cmd == NULL; sin...

CVE-2026-72090

CVE-2026-72090 affects the Linux kernel component accel/amdxdna. The root cause is in the ioctl flow for SYNC_DIRECT_FROM_DEVICE: amdxdna_drm_sync_bo_ioctl() looked up args->handle in the caller’s drm_file and then invoked amdxdna_hwctx_sync_debug_bo() with abo->client. The function uses th...

CVE-2026-72088

The CVE-2026-72088 entry concerns the Linux kernel SCSI HPSA driver IOACCEL2 reset path. The root cause was a DMA mapping leak when phys_disk->in_reset is set, causing an early return without cleaning up resources allocated for the command. The fix adds proper error cleanup: unmapping the IOAC...

CVE-2026-72089

The CVE-2026-72089 entry describes a Linux kernel issue in accel/ivpu firmware log handling. fw_log_from_bo() validates header_size but not log->size, so if a log’s size is smaller than header (including 0), fw_log_print_buffer() computes data_size = log->size - log->header_size and unde...