====================================================================== X.Org Security Advisory: October 07, 2026
Issues in X.Org X server prior to xorg-server-21.1.25 and Xwayland prior to xwayland-24.1.14 ====================================================================== Multiple issues have been found in the X server and Xwayland implementations published by X.Org for which we are releasing security fixes for in xorg-server-21.1.25 and xwayland-24.1.14. * CVE-2026-88812: XKB SetGeometry TextDoodad Double Free _CheckSetDoodad() in xkb/xkb.c frees doodad->text.text on error but does not NULL the pointer. When the error path continues, _XkbClearDoodad() is called during cleanup and frees the same pointer again, resulting in a double free. An authenticated X client can trigger this by sending a crafted XkbSetGeometry request with a malformed TextDoodad entry that causes the initial error path to be taken. The double free can lead to heap corruption, potentially enabling arbitrary code execution or denial of service (crash). Fixed in: xorg-server-21.1.25 and xwayland-24.1.14 Fix: https://gitlab.freedesktop.org/xorg/xserver/-/commit/0d1b1b0bad86 Found by: Anonymous working with TrendAI Zero Day Initiative. (ZDI-CAN-31221) * CVE-2026-93515: Present Extension Cross-Window Notify Use-After-Free present_clear_window_notifies() does not unlink notify entries from the per-window list before freeing window_priv. When the window is subsequently destroyed or reused, the stale list entries are traversed, resulting in a use-after-free. An authenticated X client can trigger this by creating cross-window Present notifies and then destroying the target window. Both the Present and SYNC extensions must be enabled (they are by default). The use-after-free can lead to denial of service (crash) or potentially information disclosure. Fixed in: xorg-server-21.1.25 and xwayland-24.1.14 Fix: https://gitlab.freedesktop.org/xorg/xserver/-/commit/1b6955c3108e Found by: Anonymous working with TrendAI Zero Day Initiative. (ZDI-CAN-31830) * CVE-2026-93516: XInput Passive Grab modifierDevice Use-After-Free GrabRec stores a raw DeviceIntPtr for the modifierDevice without any lifetime management. When a master device is removed via XIRemoveMaster, the device is freed but the passive grab still holds a dangling pointer to it. Subsequent operations that dereference the modifierDevice pointer trigger a use-after-free. An authenticated X client with the ability to add and remove master devices (XI2) can trigger this by creating a passive grab, then removing the master device referenced as the modifier device. The use-after-free can lead to denial of service (crash) or potentially arbitrary code execution. Fixed in: xorg-server-21.1.25 and xwayland-24.1.14 Fix: https://gitlab.freedesktop.org/xorg/xserver/-/commit/be5726341588 Found by: Anonymous working with TrendAI Zero Day Initiative. (ZDI-CAN-31832) * CVE-2026-93517: GLX RenderLarge Heap Buffer Overflow In the GLX RenderLarge request handling, the buffer is allocated to the size of cmdlen but the subsequent memcpy copies dataBytes into it. There is no validation that dataBytes is less than or equal to cmdlen, allowing an attacker to write past the end of the allocated buffer. An authenticated X client can trigger this by sending a crafted GLX RenderLarge request where dataBytes exceeds cmdlen, causing a heap buffer overflow. The heap buffer overflow can lead to arbitrary code execution or denial of service. Fixed in: xorg-server-21.1.25 and xwayland-24.1.14 Fix: https://gitlab.freedesktop.org/xorg/xserver/-/commit/2abe4632d793 Found by: Anonymous working with TrendAI Zero Day Initiative. (ZDI-CAN-31833) * CVE-2026-93518: XKB ResizeKeyType Numeric Truncation In XkbResizeKeyType(), the size_syms variable is declared as unsigned short. The expression (nTotal * 15) / 10 can exceed 65535 for large nTotal values, and the result is truncated when assigned to the unsigned short variable. This causes an undersized allocation, and subsequent writes overflow the heap buffer. An authenticated X client can trigger this by sending XKB requests that cause a key type resize with a sufficiently large nTotal value. The heap buffer overflow resulting from the truncated allocation can lead to arbitrary code execution or denial of service. Fixed in: xorg-server-21.1.25 and xwayland-24.1.14 Fix: https://gitlab.freedesktop.org/xorg/xserver/-/commit/89101a6c6618 Found by: Anonymous working with TrendAI Zero Day Initiative. (ZDI-CAN-31834) * CVE-2026-93519: XFixes Pointer Barrier Event List Buffer Overflow input_constrain_cursor() writes barrier events into a fixed-size buffer without checking bounds. When more than 100 pointer barriers are active, the writes overflow the fixed buffer on the stack or heap. An authenticated X client can trigger this by creating more than 100 pointer barriers (using the XFIXES extension) and then generating pointer motion events (e.g. via XTEST). Both extensions are enabled by default. The buffer overflow can lead to arbitrary code execution or denial of service. Fixed in: xorg-server-21.1.25 and xwayland-24.1.14 Fix: https://gitlab.freedesktop.org/xorg/xserver/-/commit/1f42cc1f00d9 Found by: Anonymous working with TrendAI Zero Day Initiative. (ZDI-CAN-31938) * CVE-2026-93520: XKB ChangeKeycodeRange Heap Out-of-Bounds Write This is caused by an incomplete fix in commit a3171732d. XkbAllocNames() still uses max_key_code+1 for the names->keys allocation size, while memset in the ChangeKeycodeRange path uses MAP_LENGTH which can be larger. This causes a heap out-of-bounds write when the keycode range is changed to a value where MAP_LENGTH exceeds max_key_code+1. An authenticated X client can trigger this by sending XKB requests that change the keycode range. The heap out-of-bounds write can lead to arbitrary code execution or denial of service. Fixed in: xorg-server-21.1.25 and xwayland-24.1.14 Fix: https://gitlab.freedesktop.org/xorg/xserver/-/commit/b941a473e06f Found by: Anonymous working with TrendAI Zero Day Initiative. (ZDI-CAN-31941) * CVE-2026-93521: RandR ChangeProviderProperty Heap Buffer Overflow In RRChangeProviderProperty() with PrependMode, new_value.size is set to len instead of total_len. Additionally, the old_data offset calculation uses prop_value->size instead of len for the memcpy of existing data. This causes a heap buffer overflow when prepending property data, as the buffer is undersized and the copy offset is incorrect. This mirrors a bug pattern that was previously fixed in RRChangeOutputProperty but was not applied to the provider property path. An authenticated X client can trigger this by sending RandR ChangeProviderProperty requests with PrependMode on a system with RandR providers (standard on modern GPUs). The heap buffer overflow can lead to arbitrary code execution or denial of service. Fixed in: xorg-server-21.1.25 and xwayland-24.1.14 Fix: https://gitlab.freedesktop.org/xorg/xserver/-/commit/5dc9efd5a199 Found by: Anonymous working with TrendAI Zero Day Initiative. (ZDI-CAN-31944) * CVE-2026-93522: Glamor CopyArea CPU-FBO Heap Buffer Overflow In glamor's CopyArea CPU-to-FBO path, the temporary buffer (tmp_bits) is sized based on the destination height but is indexed using source coordinates. When the source region is taller than the destination, writes overflow the allocated buffer. An authenticated X client can trigger this by performing a CopyArea operation between drawables with mismatched depth (depth-24 to depth-32 or vice versa) where the source is taller than the destination, on a system using glamor/GPU acceleration. The heap buffer overflow can lead to arbitrary code execution or denial of service. This issue does not affect xorg-server-21.1.x Fixed in: xwayland-24.1.14 Fix: https://gitlab.freedesktop.org/xorg/xserver/-/commit/ad26c26bf795 Found by: Anonymous working with TrendAI Zero Day Initiative. (ZDI-CAN-32361) * CVE-2026-93523: XInput2 PassiveUngrabDevice Modifier Out-of-Bounds Write The XI2 passive ungrab path (XIPassiveUngrabDevice) is missing the modifier validation that exists in the grab path and in legacy XI paths. Without checking against AllModifiersMask, a client can supply out-of-range modifier values that lead to an out-of-bounds write when the modifier is used as an index. An authenticated X client can trigger this by sending an XIPassiveUngrabDevice request with an out-of-range modifier value. The out-of-bounds write can lead to arbitrary code execution or denial of service. Fixed in: xorg-server-21.1.25 and xwayland-24.1.14 Fix: https://gitlab.freedesktop.org/xorg/xserver/-/commit/37a1847a60c6 Found by: Anonymous working with TrendAI Zero Day Initiative. (ZDI-CAN-32366) * CVE-2026-93524: XKB SetMap Key Width/Action Count Desync Out-Of-Bounds Read CheckKeySyms() in xkb/xkb.c populates a symsPerKey[] validation array with new key widths for keys in the SetMap request range, then continues filling entries for keys beyond the range. However, the second loop starts at index i = nKeySyms (the first loop's counter) rather than i = firstKeySym + nKeySyms. When firstKeySym is large and nKeySyms is small, the second loop overwrites symsPerKey[target] with the old width from the existing map. CheckKeyActions() then validates the wire action count against the stale old width and accepts it. In _XkbSetMap(), SetKeySyms() widens the key (resizing actions to nGroups * newWidth), but SetKeyActions() then resizes again to the old count. A subsequent XkbGetMap request reads XkbKeyNumActions() entries (derived from the new wider key_sym_map.width) from the shorter action allocation, causing an out-of-bounds heap read. The overread bytes are copied into the GetMap reply and sent back to the requesting client, producing a bounded heap information disclosure. Fixed in: xorg-server-21.1.25 and xwayland-24.1.14 Fix: https://gitlab.freedesktop.org/xorg/xserver/-/commit/cea71d0273e4 Found by: WONJOON HWANG (@joon1337) working with TrendAI Zero Day Initiative. (ZDI-CAN-32408) * CVE-2026-93536: GestureBuildSprite Use-After-Free GestureBuildSprite() copies raw WindowPtr values into the gesture sprite trace via CopySprite() without any reference counting or lifetime management. WindowGone() only repairs touch sprite traces when a window is destroyed -- gesture sprite traces are never checked or cleaned. This leaves stale WindowPtr references in gesture sprites. When DeliverOneGestureEvent() later dereferences the stale WindowPtr via DeepestSpriteWin(&gi->sprite)->drawable.id, it produces a use-after-free. The freed 4 bytes (drawable.id) can be read back, potentially leaking information about the contents of the freed memory region. Fixed in: xorg-server-21.1.25 and xwayland-24.1.14 Fix: https://gitlab.freedesktop.org/xorg/xserver/-/commit/efcfd8acc754 Found by: 4nibhal working with TrendAI Zero Day Initiative. (ZDI-CAN-32753)
