🚨 CVE-2026-45910
In the Linux kernel, the following vulnerability has been resolved:

RDMA/rxe: Fix race condition in QP timer handlers

I encontered the following warning:
WARNING: drivers/infiniband/sw/rxe/rxe_task.c:249 at rxe_sched_task+0x1c8/0x238 [rdma_rxe], CPU#0: swapper/0/0
...
libsha1 [last unloaded: ip6_udp_tunnel]
CPU: 0 UID: 0 PID: 0 Comm: swapper/0 Tainted: G C 6.19.0-rc5-64k-v8+ #37 PREEMPT
Tainted: [C]=CRAP
Hardware name: Raspberry Pi 4 Model B Rev 1.2
Call trace:
rxe_sched_task+0x1c8/0x238 [rdma_rxe] (P)
retransmit_timer+0x130/0x188 [rdma_rxe]
call_timer_fn+0x68/0x4d0
__run_timers+0x630/0x888
...
WARNING: drivers/infiniband/sw/rxe/rxe_task.c:38 at rxe_sched_task+0x1c0/0x238 [rdma_rxe], CPU#0: swapper/0/0
...
WARNING: drivers/infiniband/sw/rxe/rxe_task.c:111 at do_work+0x488/0x5c8 [rdma_rxe], CPU#3: kworker/u17:4/93400
...
refcount_t: underflow; use-after-free.
WARNING: lib/refcount.c:28 at refcount_warn_saturate+0x138/0x1a0, CPU#3: kworker/u17:4/93400

The issue is caused by a race condition between retransmit_timer() and
rxe_destroy_qp, leading to the Queue Pair's (QP) reference count dropping
to zero during timer handler execution.

It seems this warning is harmless because rxe_qp_do_cleanup() will flush
all pending timers and requests.

Example of flow causing the issue:

CPU0 CPU1
retransmit_timer() {
spin_lock_irqsave
rxe_destroy_qp()
__rxe_cleanup()
__rxe_put() // qp->ref_count decrease to 0
rxe_qp_do_cleanup() {
if (qp->valid) {
rxe_sched_task() {
WARN_ON(rxe_read(task->qp) <= 0);
}
}
spin_unlock_irqrestore
}
spin_lock_irqsave
qp->valid = 0
spin_unlock_irqrestore
}

Ensure the QP's reference count is maintained and its validity is checked
within the timer callbacks by adding calls to rxe_get(qp) and corresponding
rxe_put(qp) after use.

🎖@cveNotify

🚨 CVE-2026-45911
In the Linux kernel, the following vulnerability has been resolved:

usb: cdns3: fix role switching during resume

If the role change while we are suspended, the cdns3 driver switches to the
new mode during resume. However, switching to host mode in this context
causes a NULL pointer dereference.

The host role's start() operation registers a xhci-hcd device, but its
probe is deferred while we are in the resume path. The host role's resume()
operation assumes the xhci-hcd device is already probed, which is not the
case, leading to the dereference. Since the start() operation of the new
role is already called, the resume operation can be skipped.

So skip the resume operation for the new role if a role switch occurs
during resume. Once the resume sequence is complete, the xhci-hcd device
can be probed in case of host mode.

Unable to handle kernel NULL pointer dereference at virtual address 0000000000000208
Mem abort info:
...
Data abort info:
...
[0000000000000208] pgd=0000000000000000, p4d=0000000000000000
Internal error: Oops: 0000000096000004 [#1] SMP
Modules linked in:
CPU: 0 UID: 0 PID: 146 Comm: sh Not tainted
6.19.0-rc7-00013-g6e64f4aabfae-dirty #135 PREEMPT
Hardware name: Texas Instruments J7200 EVM (DT)
pstate: 20000005 (nzCv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--)
pc : usb_hcd_is_primary_hcd+0x0/0x1c
lr : cdns_host_resume+0x24/0x5c
...
Call trace:
usb_hcd_is_primary_hcd+0x0/0x1c (P)
cdns_resume+0x6c/0xbc
cdns3_controller_resume.isra.0+0xe8/0x17c
cdns3_plat_resume+0x18/0x24
platform_pm_resume+0x2c/0x68
dpm_run_callback+0x90/0x248
device_resume+0x100/0x24c
dpm_resume+0x190/0x2ec
dpm_resume_end+0x18/0x34
suspend_devices_and_enter+0x2b0/0xa44
pm_suspend+0x16c/0x5fc
state_store+0x80/0xec
kobj_attr_store+0x18/0x2c
sysfs_kf_write+0x7c/0x94
kernfs_fop_write_iter+0x130/0x1dc
vfs_write+0x240/0x370
ksys_write+0x70/0x108
__arm64_sys_write+0x1c/0x28
invoke_syscall+0x48/0x10c
el0_svc_common.constprop.0+0x40/0xe0
do_el0_svc+0x1c/0x28
el0_svc+0x34/0x108
el0t_64_sync_handler+0xa0/0xe4
el0t_64_sync+0x198/0x19c
Code: 52800003 f9407ca5 d63f00a0 17ffffe4 (f9410401)
---[ end trace 0000000000000000 ]---

🎖@cveNotify

🚨 CVE-2026-45912
In the Linux kernel, the following vulnerability has been resolved:

ext4: don't cache extent during splitting extent

Caching extents during the splitting process is risky, as it may result
in stale extents remaining in the status tree. Moreover, in most cases,
the corresponding extent block entries are likely already cached before
the split happens, making caching here not particularly useful.

Assume we have an unwritten extent, and then DIO writes the first half.

[UUUUUUUUUUUUUUUU] on-disk extent U: unwritten extent
[UUUUUUUUUUUUUUUU] extent status tree
|<- ->| ----> dio write this range

First, when ext4_split_extent_at() splits this extent, it truncates the
existing extent and then inserts a new one. During this process, this
extent status entry may be shrunk, and calls to ext4_find_extent() and
ext4_cache_extents() may occur, which could potentially insert the
truncated range as a hole into the extent status tree. After the split
is completed, this hole is not replaced with the correct status.

[UUUUUUU|UUUUUUUU] on-disk extent U: unwritten extent
[UUUUUUU|HHHHHHHH] extent status tree H: hole

Then, the outer calling functions will not correct this remaining hole
extent either. Finally, if we perform a delayed buffer write on this
latter part, it will re-insert the delayed extent and cause an error in
space accounting.

In adition, if the unwritten extent cache is not shrunk during the
splitting, ext4_cache_extents() also conflicts with existing extents
when caching extents. In the future, we will add checks when caching
extents, which will trigger a warning. Therefore, Do not cache extents
that are being split.

🎖@cveNotify

🚨 CVE-2026-45913
In the Linux kernel, the following vulnerability has been resolved:

net: bridge: mcast: always update mdb_n_entries for vlan contexts

syzbot triggered a warning[1] about the number of mdb entries in a context.
It turned out that there are multiple ways to trigger that warning today
(some got added during the years), the root cause of the problem is that
the increase is done conditionally, and over the years these different
conditions increased so there were new ways to trigger the warning, that is
to do a decrease which wasn't paired with a previous increase.

For example one way to trigger it is with flush:
$ ip l add br0 up type bridge vlan_filtering 1 mcast_snooping 1
$ ip l add dumdum up master br0 type dummy
$ bridge mdb add dev br0 port dumdum grp 239.0.0.1 permanent vid 1
$ ip link set dev br0 down
$ ip link set dev br0 type bridge mcast_vlan_snooping 1
^^^^ this will enable snooping, but will not update mdb_n_entries
because in __br_multicast_enable_port_ctx() we check !netif_running
$ bridge mdb flush dev br0
^^^ this will trigger the warning because it will delete the pg which
we added above, which will try to decrease mdb_n_entries

Fix the problem by removing the conditional increase and always keep the
count up-to-date while the vlan exists. In order to do that we have to
first initialize it on port-vlan context creation, and then always increase
or decrease the value regardless of mcast options. To keep the current
behaviour we have to enforce the mdb limit only if the context is port's or
if the port-vlan's mcast snooping is enabled.

[1]
------------[ cut here ]------------
n == 0
WARNING: net/bridge/br_multicast.c:718 at br_multicast_port_ngroups_dec_one net/bridge/br_multicast.c:718 [inline], CPU#0: syz.4.4607/22043
WARNING: net/bridge/br_multicast.c:718 at br_multicast_port_ngroups_dec net/bridge/br_multicast.c:771 [inline], CPU#0: syz.4.4607/22043
WARNING: net/bridge/br_multicast.c:718 at br_multicast_del_pg+0x1bbe/0x1e20 net/bridge/br_multicast.c:825, CPU#0: syz.4.4607/22043
Modules linked in:
CPU: 0 UID: 0 PID: 22043 Comm: syz.4.4607 Not tainted syzkaller #0 PREEMPT(full)
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/24/2026
RIP: 0010:br_multicast_port_ngroups_dec_one net/bridge/br_multicast.c:718 [inline]
RIP: 0010:br_multicast_port_ngroups_dec net/bridge/br_multicast.c:771 [inline]
RIP: 0010:br_multicast_del_pg+0x1bbe/0x1e20 net/bridge/br_multicast.c:825
Code: 41 5f 5d e9 04 7a 48 f7 e8 3f 73 5c f7 90 0f 0b 90 e9 cf fd ff ff e8 31 73 5c f7 90 0f 0b 90 e9 16 fd ff ff e8 23 73 5c f7 90 <0f> 0b 90 e9 60 fd ff ff e8 15 73 5c f7 eb 05 e8 0e 73 5c f7 48 8b
RSP: 0018:ffffc9000c207220 EFLAGS: 00010293
RAX: ffffffff8a68042d RBX: ffff88807c6f1800 RCX: ffff888066e90000
RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000000000000
RBP: 0000000000000000 R08: ffff888066e90000 R09: 000000000000000c
R10: 000000000000000c R11: 0000000000000000 R12: ffff8880303ef800
R13: dffffc0000000000 R14: ffff888050eb11c4 R15: 1ffff1100a1d6238
FS: 00007fa45921b6c0(0000) GS:ffff8881256f5000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00007fa4591f9ff8 CR3: 0000000081df2000 CR4: 00000000003526f0
Call Trace:
<TASK>
br_mdb_flush_pgs net/bridge/br_mdb.c:1525 [inline]
br_mdb_flush net/bridge/br_mdb.c:1544 [inline]
br_mdb_del_bulk+0x5e2/0xb20 net/bridge/br_mdb.c:1561
rtnl_mdb_del+0x48a/0x640 net/core/rtnetlink.c:-1
rtnetlink_rcv_msg+0x77e/0xbe0 net/core/rtnetlink.c:6967
netlink_rcv_skb+0x232/0x4b0 net/netlink/af_netlink.c:2550
netlink_unicast_kernel net/netlink/af_netlink.c:1318 [inline]
netlink_unicast+0x80f/0x9b0 net/netlink/af_netlink.c:1344
netlink_sendmsg+0x813/0xb40 net/netlink/af_netlink.c:1894
sock_sendmsg_nosec net/socket.c:727 [inline]
__sock_sendmsg net/socket.c:742 [inline]
____sys_sendmsg+0xa68/0xad0 net/socket.c:2592
___sys_sendmsg+0x2a5/0x360 net/socke
---truncated---

🎖@cveNotify

🚨 CVE-2026-45914
In the Linux kernel, the following vulnerability has been resolved:

Revert "hwmon: (ibmpex) fix use-after-free in high/low store"

This reverts commit 6946c726c3f4c36f0f049e6f97e88c510b15f65d.

Jean Delvare points out that the patch does not completely
fix the reported problem, that it in fact introduces a
(new) race condition, and that it may actually not be needed in
the first place.

Various AI reviews agree. Specific and relevant AI feedback:

"
This reordering sets the driver data to NULL before removing the sensor
attributes in the loop below.

ibmpex_show_sensor() retrieves this driver data via dev_get_drvdata() but
does not check if it is NULL before dereferencing it to access
data->sensors[].

If a userspace process reads a sensor file (like temp1_input) while this
delete function is running, could it race with the dev_set_drvdata(...,
NULL) call here and crash in ibmpex_show_sensor()?

Would it be safer to keep the original order where device_remove_file() is
called before clearing the driver data? device_remove_file() should wait
for any active sysfs callbacks to complete, which might already prevent the
use-after-free this patch intends to fix.
"

Revert the offending patch. If it can be shown that the originally reported
alleged race condition does indeed exist, it can always be re-introduced
with a complete fix.

🎖@cveNotify

🚨 CVE-2026-32174
Improper authentication in Azure Bot Service allows an authorized attacker to elevate privileges over a network.

🎖@cveNotify

🚨 CVE-2026-54303
n8n is an open source workflow automation platform. Prior to 2.24.0, an endpoint in the Meta and Microsoft Teams trigger nodes reflects a query parameter into the HTTP response without sanitization or Content-Security-Policy headers, enabling reflected XSS in the n8n origin when a logged-in user visits a crafted URL. This vulnerability is fixed in 2.24.0.

🎖@cveNotify

🚨 CVE-2026-54314
n8n is an open source workflow automation platform. Prior to 2.24.0, the Compression node's Decompress operation expanded attacker-controlled archives into memory without enforcing limits on decompressed output size. An unauthenticated attacker could send a small compressed archive to a public webhook workflow using this node, causing the n8n process to terminate due to memory exhaustion and disrupting all workflows in the same instance. This vulnerability is fixed in 2.24.0.

🎖@cveNotify

🚨 CVE-2026-56692
NanoClaw before 2.1.17 contains a symlink following vulnerability in forwardAttachedFiles that allows container-controlled agents to exfiltrate host-readable files. The host validates attachment filenames using only isSafeAttachmentName before copying with fs.copyFileSync, which follows symlinks without containment checks, allowing malicious agents to disclose arbitrary host files.

🎖@cveNotify

🚨 CVE-2026-56695
OpenHarness ohmo gateway /resume and /summary slash commands default remote_invocable to True, allowing admitted remote senders to enumerate and load arbitrary session snapshots by ID. Attackers can exploit this to access victim snapshots containing private prompts, credentials, tool output, and file paths via shared gateway channels.

🎖@cveNotify

🚨 CVE-2026-44789
n8n is an open source workflow automation platform. Prior to 1.123.43, 2.22.1, and 2.20.7, an authenticated user with permission to create or modify workflows could achieve global prototype pollution via an unvalidated pagination parameter in the HTTP Request node. Combined with other techniques this could lead to RCE on the instance. This vulnerability is fixed in 1.123.43, 2.22.1, and 2.20.7.

🎖@cveNotify

🚨 CVE-2026-48520
Langflow is a tool for building and deploying AI-powered agents and workflows. Prior to 1.10.0, the "Shareable Playground" (or "Public Flows" in code) contains a potential arbitrary file-read vulnerability, depending on the exact flow configuration used. By making a flow public, public execution of the flow is allowed. The execution request can contain a list of files that gets read by Langflow and fed into the LLM. The files path can be any path supported by the storage - it can be either a local file or S3 path if supported by the local configuration This vulnerability is fixed in 1.10.0.

🎖@cveNotify

🚨 CVE-2026-54301
n8n is an open source workflow automation platform. Prior to 1.123.55, 2.25.7, and 2.26.2, an authenticated user with workflow edit access could configure a Respond to Webhook node to serve binary content with an attacker-controlled Content-Type. The binary response path bypassed the central Content-Security-Policy sandbox header, allowing a public webhook to execute JavaScript in the n8n origin when visited by an authenticated user, with access to that user's session. This vulnerability is fixed in 1.123.55, 2.25.7, and 2.26.2.

🎖@cveNotify

🚨 CVE-2026-54308
n8n is an open source workflow automation platform. Prior to 2.25.7 and 2.26.2, the MicrosoftAgent365Trigger and StripeTrigger node did not validate that inbound requests. As a result, an unauthenticated attacker who knows the webhook URL could submit a forged payload and cause the workflow to execute with attacker-controlled data. This vulnerability is fixed in 2.25.7 and 2.26.2.

🎖@cveNotify

🚨 CVE-2026-56114
dhcpcd through 10.3.2, fixed in commit 2f00c7b, contains a one-byte stack out-of-bounds write vulnerability in dhcp6_makemessage() in src/dhcp6.c that allows unauthenticated same-link attackers to write beyond a fixed local buffer by serializing an oversized RFC6603 OPTION_PD_EXCLUDE option body. Attackers can send a crafted DHCPv6 ADVERTISE message containing an IA_PD IAPREFIX /0 with a valid OPTION_PD_EXCLUDE using an exclude prefix length of /121 through /128 to trigger the out-of-bounds write and potentially corrupt adjacent stack memory.

🎖@cveNotify

🚨 CVE-2026-56117
dhcpcd through 10.3.2, fixed in commit 78ea09e, contains a heap use-after-free vulnerability in the control socket handling within src/control.c that allows local unprivileged attackers to trigger memory corruption when privilege separation is disabled. Attackers can connect to the control socket and send a privileged command such as -x, causing control_recvdata() to free the client object while the same READ+HANGUP event subsequently reaches control_hangup() with the stale pointer, resulting in a use-after-free condition exploitable in deployments using --disable-privsep or where privsep initialization has failed with the control socket operating in mode 0666.

🎖@cveNotify

🚨 CVE-2026-49859
Deno is a JavaScript, TypeScript, and WebAssembly runtime. Prior to 2.8.1, when fetch() was called, Deno checked the destination hostname against --deny-net rules but did not re-check the IP addresses that hostname resolved to. An attacker-controlled script could use a specially crafted domain name that passes the hostname check yet resolves to a denied IP, bypassing the network restriction entirely. This vulnerability is fixed in 2.8.1.

🎖@cveNotify

🚨 CVE-2026-54009
Open WebUI is a self-hosted artificial intelligence platform designed to operate entirely offline. Prior to 0.9.6, POST /api/chat/completions accepts an image_url.url value that, when it does NOT start with http://, https://, or data:image/, is interpreted as a file id and resolved against the global file table with no ownership check. an authenticated user can therefore set image_url.url to another user's file id, the server reads that file from disk, base64-encodes it, and injects the data URI into the LLM request. the user then prompts the LLM to describe / OCR the file and reads the content back. This vulnerability is fixed in 0.9.6.

🎖@cveNotify

🚨 CVE-2026-54015
Open WebUI is a self-hosted artificial intelligence platform designed to operate entirely offline. Prior to 0.9.6, Open WebUI's prompt version-history endpoints authorize the prompt_id in the URL but then act on caller-supplied history IDs without verifying that the history row belongs to that prompt (history_entry.prompt_id == prompt.id). This affects /api/v1/prompts/id/{prompt_id}/history/diff, /api/v1/prompts/id/{prompt_id}/update/version, and /api/v1/prompts/id/{prompt_id}/history/{history_id}. An authenticated user with access to any prompt they control, plus a victim prompt_history.id, can read or delete another user's private prompt history. This vulnerability is fixed in 0.9.6.

🎖@cveNotify

🚨 CVE-2026-46551
NocoDB is software for building databases as spreadsheets. Prior to 2026.04.4, the uploadViaURL path in the v1/v2 attachment API did not enforce NC_ATTACHMENT_FIELD_SIZE against the remote content-length or against the response stream. An authenticated user (Editor+) could direct the server to download arbitrarily large files, exhausting disk space and causing denial of service. In packages/nocodb/src/services/attachments.service.ts, the HEAD probe read content-length but never compared it to NC_ATTACHMENT_FIELD_SIZE; the subsequent storageAdapter.fileCreateByUrl() performed the download without maxContentLength. This vulnerability is fixed in 2026.04.4.

🎖@cveNotify

🚨 CVE-2026-47376
NocoDB is software for building databases as spreadsheets. Prior to 2026.04.1, the password-reset page rendered the URL token directly into a JavaScript string literal in a server-rendered EJS template. EJS <%= %> HTML-entity-encodes a fixed set of characters but does not escape single quotes or backslashes, so a crafted token could break out of the JS string context and execute attacker-controlled script in the NocoDB origin. Triggering required only that a victim follow a malicious password-reset link. This vulnerability is fixed in 2026.04.1.

🎖@cveNotify