| CVE |
Vendors |
Products |
Updated |
CVSS v3.1 |
| In the Linux kernel, the following vulnerability has been resolved:
RDMA/mlx5: Drain RCU callbacks during module teardown
devx_free_subscription() can remain queued after the last DevX event file
drops its module reference or an auxiliary driver detaches its devices.
mlx5_ib can then unload before the callback runs.
Registration error unwind has the same risk because driver registration
can attach existing devices before failing. Wait after all drivers have
stopped. |
| In the Linux kernel, the following vulnerability has been resolved:
clk: qcom: camcc-sc8280xp: unregister CAMCC_GDSC_CLK
With the introduction of sync_state support in the clk and pmdomain
subsystems, the following warning happens when the unused clocks are
shutdown in camcc-sc8280xp:
[ 15.408367] titan_top_gdsc status stuck at 'on'
[ 15.408429] WARNING: drivers/clk/qcom/gdsc.c:178 at gdsc_toggle_logic+0x14c/0x160, CPU#2: kworker/u32:1/14
[ 15.408462] Modules linked in: bnep vfat fat ath11k_pci(+) ath11k mac80211 cfg80211 mhi libarc4 snd_soc_wcd938x snd_soc_wcd938x_sdw snd_soc_wcd_classh hci_uart snd_soc_wcd_common
snd_soc_sc8280xp soundwire_qcom snd_soc_wcd_mbhc snd_soc_qcom_sdw slimbus snd_soc_qcom_common regmap_sdw btqca btrtl qcom_camss soundwire_bus btbcm btintel snd_soc_sdca snd_soc_lpass_wsa_macro
bluetooth snd_soc_lpass_tx_macro snd_soc_lpass_va_macro snd_soc_lpass_rx_macro snd_soc_hdmi_codec snd_soc_lpass_macro_common videobuf2_dma_sg ov5675 v4l2_fwnode videobuf2_memops
qcom_spmi_adc5 snd_soc_core qcom_spmi_adc_tm5 videobuf2_v4l2 snd_seq snd_seq_device videobuf2_common v4l2_async qcom_vadc_common qcom_spmi_temp_alarm pm8941_pwrkey industrialio videodev
snd_compress rfkill ac97_bus snd_pcm_dmaengine qcom_tsens mc qcom_edac snd_pcm pci_pwrctrl_pwrseq qcom_cpufreq_hw snd_timer snd qcomtee soundcore tee leds_gpio joydev binfmt_misc zram
lz4hc_compress governor_simpleondemand panel_edp msm xhci_plat_hcd nvme nvme_core dwc3 qcom_pm8008_regulator
[ 15.408688] ucsi_glink nvme_keyring nvme_auth pmic_glink_altmode udc_core typec_ucsi aux_hpd_bridge qcom_battmgr ulpi ubwc_config socinfo ocmem drm_gpuvm qcom_q6v5_pas drm_exec
qcom_pil_info leds_qcom_lpg gpu_sched led_class_multicolor rtc_pm8xxx qcom_pbs qcom_common drm_display_helper qcom_pon qcom_glink_smem qcom_glink ghash_ce pwrseq_qcom_wcn gpio_sbu_mux
qcom_stats phy_qcom_qmp_combo qcom_q6v5 gf128mul cec dispcc_sc8280xp phy_qcom_edp camcc_sc8280xp i2c_qcom_cci qcom_sysmon drm_dp_aux_bus mdt_loader aux_bridge qcom_pm8008 i2c_hid_of_elan
dwc3_qcom_legacy llcc_qcom icc_bwmon gpi typec qcom_refgen_regulator phy_qcom_qmp_usb nvmem_qfprom qcom_ipcc phy_qcom_snps_femto_v2 gpucc_sc8280xp pinctrl_sc8280xp_lpass_lpi qcom_hwspinlock
pinctrl_lpass_lpi lpasscc_sc8280xp qrtr qcom_aoss pmic_glink pdr_interface phy_qcom_qmp_pcie qcom_smd qcom_pdr_msg icc_osm_l3 qcom_wdt qmi_helpers qcom_rng smp2p rpmsg_core gpio_keys pwm_bl
smem hid_multitouch fuse i2c_dev
[ 15.408928] CPU: 2 UID: 0 PID: 14 Comm: kworker/u32:1 Not tainted 7.1.0+ #2 PREEMPT(lazy)
[ 15.408937] Hardware name: LENOVO 21BX0016US/21BX0016US, BIOS N3HET88W (1.60 ) 03/14/2024
[ 15.408942] Workqueue: pm pm_runtime_work
[ 15.408959] pstate: 60400005 (nZCv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
[ 15.408967] pc : gdsc_toggle_logic+0x14c/0x160
[ 15.408978] lr : gdsc_toggle_logic+0x14c/0x160
[ 15.408987] sp : ffff8000800f3b40
[ 15.408991] x29: ffff8000800f3b40 x28: 0000000000000000 x27: 0000000000000000
[ 15.409003] x26: 0000000000000000 x25: 0000000000000000 x24: 0000000000000000
[ 15.409014] x23: 0000000000000000 x22: 0000000000000001 x21: ffffa33f298fca88
[ 15.409024] x20: 0000000000000000 x19: ffffa33f298fc5b0 x18: 00cd15db75dacefd
[ 15.409035] x17: 000000040044ffff x16: ffffa33f3b1a3d88 x15: 726f776b80000002
[ 15.409045] x14: ffffffffffffffff x13: 0000000000000028 x12: 0101010101010101
[ 15.409056] x11: 7f7f7f7f7f7f7f7f x10: fefeff3039313274 x9 : ffffa33f3a5edafc
[ 15.409067] x8 : ffff8000800f3780 x7 : 0000000000000001 x6 : 0000000000000001
[ 15.409078] x5 : ffff000bf3ca1288 x4 : 0000000000000000 x3 : ffff5cccb6a3f000
[ 15.409088] x2 : 0000000000000000 x1 : 0000000000000000 x0 : ffff000080ae0000
[ 15.409098] Call trace:
[ 15.409103] gdsc_toggle_logic+0x14c/0x160 (P)
[ 15.409115] gdsc_disable+0x4c/0x190
[ 15.409126] _genp
---truncated--- |
| In the Linux kernel, the following vulnerability has been resolved:
platform/x86: asus-wmi: fix resource leaks on probe failure
During driver initialization in asus_wmi_add(), various subsystems are
registered sequentially. However, the error path labels are out of order
relative to the registration sequence.
Specifically:
1. If asus_wmi_custom_fan_curve_init() fails, the driver jumps to
fail_custom_fan_curve. Because this label is placed below fail_sysfs,
it bypasses the cleanup calls for the input device and sysfs groups,
which were successfully registered before, leaking those resources.
2. If asus_screenpad_init() fails, the driver jumps to fail_screenpad.
Because fail_screenpad is placed below fail_backlight, it bypasses the
cleanup calls for backlight and rfkill, leaking those resources.
Fix these resource leaks by reordering the error path labels in
asus_wmi_add() to match the exact reverse order of the resource
allocations. |
| In the Linux kernel, the following vulnerability has been resolved:
usb: fix UAF when probe runs concurrent to dyn ID removal
Dynamic IDs are only guaranteed to be valid when usb_dynids_lock is held,
as remove_id_store can free the node. Thus, make a copy in
usb_probe_interface. Clarify the documentation that the id parameter is
only valid during the probe.
USB serial has the same pattern, but it does not need fixing as the IDs
cannot be removed via sysfs. |
| In the Linux kernel, the following vulnerability has been resolved:
platform/x86: asus-wireless: Fail probe when there is no ACPI match
Every platform driver can be forced to match a device that does not match
its list of device IDs because of device_match_driver_override(), so
platform drivers that rely on the existence of a device ACPI companion
object need to verify its presence.
asus_wireless_probe() returns success when acpi_match_acpi_device()
finds no match, leaving behind an input device that never reports
anything because the notify handler is not installed. Worse, when the
driver is force-bound to a device without an ACPI companion, probe
still succeeds and stores a NULL companion pointer, which
asus_wireless_remove() later passes to acpi_dev_remove_notify_handler(),
leading to a NULL pointer dereference on unbind.
Return -ENODEV when the device does not match the ID table. This also
covers the missing-companion case, because acpi_match_acpi_device()
rejects a NULL device. Perform the check before allocating any driver
state, instead of after the input device has already been registered. |
| In the Linux kernel, the following vulnerability has been resolved:
bus: mhi: ep: Fix device refcount leak in the error path of MHI device creation
mhi_ep_create_device() takes one device reference for the UL channel and
another for the DL channel after allocating the transfer device. These
references are normally released by mhi_ep_destroy_device() before the
device itself is removed.
If dev_set_name() or device_add() fails, the error path currently drops
only one reference. The remaining channel references keep the device
from being released and leave the channels associated with a device that
was never registered.
Route both failures through a common unwind path that drops the DL
channel reference, the UL channel reference, and the initial reference
from device_initialize(). |
| In the Linux kernel, the following vulnerability has been resolved:
drm/amd/display: Fix dangling pointer in plane reset function
amdgpu_dm_plane_drm_plane_reset() frees the old state before allocating
a new one. If kzalloc() fails, the function returns without updating
the state pointer, leaving a dangling pointer to already freed memory.
Fix this by allocating the new state first. On allocation failure, the
old state remains untouched and the function safely returns.
Found by Linux Verification Center (linuxtesting.org) with SVACE.
[adjust for movement around current amd-staging-drm-next] |
| In the Linux kernel, the following vulnerability has been resolved:
drm/amdgpu/pm/powerplay: bounds-check voltage index in Vega10 lookup
vddInd, vddciInd and mvddInd from VBIOS-parsed tables index into vddc,
vddci and vddmem lookup tables without bounds checks across nine sites.
Return -EINVAL when any index is out of range. |
| In the Linux kernel, the following vulnerability has been resolved:
drm/lima: call drm_mm_init() with a valid allocation range
lima_vm_create() is currently run before va_start and va_end are set up,
meaning they are both 0. lima_vm_create() runs drm_mm_init() with them
as arguments for the allocator, and if DRM_DEBUG_MM is enabled the
DRM_MM_BUG_ON check in drm_mm_init then fires, as seen here on
exynos4412-odroid-u2:
[ 1.736297] ------------[ cut here ]------------
[ 1.740370] kernel BUG at drivers/gpu/drm/drm_mm.c:931!
[ 1.745574] Internal error: Oops - BUG: 0 [#1] SMP ARM
[ 1.750697] Modules linked in:
[ 1.753734] CPU: 0 UID: 0 PID: 41 Comm: kworker/u16:1 Not tainted 7.0.10-postmarketos-exynos4 #11 PREEMPT
[ 1.763372] Hardware name: Samsung Exynos (Flattened Device Tree)
[ 1.769446] Workqueue: events_unbound deferred_probe_work_func
[ 1.775261] PC is at drm_mm_init+0x9c/0xa4
[ 1.779339] LR is at lima_vm_create+0x144/0x17c
[ ... ]
Fix the issue by moving the lima_vm_create() call after va_start and
va_end are set up. |
| In the Linux kernel, the following vulnerability has been resolved:
ASoC: rt700-sdw: always drain jack work on remove
rt700_sdw_remove() drains jack_detect_work and jack_btn_check_work only
when rt700->hw_init is true. That state bit is cleared by
rt700_update_status() when the SoundWire slave becomes UNATTACHED, but a
jack work item can already have been queued by rt700_interrupt_callback()
or rt700_jack_init() while the device was initialized.
Do not use hw_init as the remove-time guard for draining these work
objects. The delayed works are initialized during rt700_init(), so remove
can cancel them unconditionally and pair the object lifetime with the
codec-private data lifetime instead of a mutable hardware state bit.
This issue was found by our static analysis tool and then confirmed by
manual review of the SoundWire status, interrupt and remove paths. The
remove path should drain work based on whether the work object exists, not
on a runtime hardware state bit that can change after the work was queued.
A QEMU PoC queued jack_detect_work, simulated SDW_SLAVE_UNATTACHED, and
then entered remove. DEBUG_OBJECTS reported an active timer/work object
associated with the rt700 jack work path after remove skipped the cancel.
This is sent as an RFC because the practical trigger depends on SoundWire
core remove ordering after an UNATTACHED status update. If remove cannot
run after hw_init has been cleared while jack work is still pending, this
is a defensive lifecycle cleanup rather than a reachable race on current
systems. |
| In the Linux kernel, the following vulnerability has been resolved:
udf: Mark LVID buffer as uptodate before marking it dirty
When an I/O error occurs while writing the Logical Volume Integrity
Descriptor (LVID) buffer to the block device, the block layer's completion
handler (`end_buffer_write_sync()`) clears the `BH_Uptodate` flag on the
buffer. However, the buffer still contains valid LVID data in memory. If
the filesystem is subsequently remounted read-write or synced,
`udf_open_lvid()` or `udf_sync_fs()` will modify the LVID buffer and call
`mark_buffer_dirty()`. This triggers a spurious
`WARN_ON_ONCE(!buffer_uptodate(bh))` warning in `mark_buffer_dirty()`
because the buffer is not marked uptodate, even though its in-memory
contents are valid and are about to be overwritten.
To prevent this spurious warning, unconditionally set the `BH_Uptodate`
flag before calling `mark_buffer_dirty()` in `udf_open_lvid()` and
`udf_sync_fs()`. This acknowledges that the in-memory buffer is valid and
matches the workaround previously applied to `udf_close_lvid()` in commit
853a0c25baf9 ("udf: Mark LVID buffer as uptodate before marking it dirty").
Extending this workaround ensures consistent behavior across all LVID
updates.
Buffer I/O error on dev loop0, logical block 128, lost sync page write
------------[ cut here ]------------
!buffer_uptodate(bh)
WARNING: fs/buffer.c:1087 at mark_buffer_dirty+0x299/0x410 fs/buffer.c:1087
...
Call Trace:
<TASK>
udf_open_lvid+0x369/0x5b0 fs/udf/super.c:2078
udf_reconfigure+0x336/0x540 fs/udf/super.c:679
reconfigure_super+0x232/0x8f0 fs/super.c:1080
vfs_cmd_reconfigure fs/fsopen.c:268 [inline]
vfs_fsconfig_locked+0x171/0x320 fs/fsopen.c:297
__do_sys_fsconfig fs/fsopen.c:463 [inline]
__se_sys_fsconfig+0x6b9/0x810 fs/fsopen.c:350
do_syscall_64+0x174/0x580 arch/x86/entry/syscall_64.c:94
</TASK> |
| In the Linux kernel, the following vulnerability has been resolved:
time/namespace: Validate nanosecond field in proc_timens_set_offset()
The function validates tv_sec to be within [-KTIME_SEC_MAX, KTIME_SEC_MAX]
but never validates that tv_nsec is within the valid range of
[0, NSEC_PER_SEC-1] before using it in timespec64_add().
timespec64_add() expects both timespec64 structures to have normalized
values with tv_nsec in the range [0, 999999999]. If off->val.tv_nsec
contains invalid values (negative or >= NSEC_PER_SEC), it could lead to
incorrect calculations or unexpected behavior.
Add validation to ensure tv_nsec is within the valid range before
performing the addition. |
| In the Linux kernel, the following vulnerability has been resolved:
wifi: mac80211_hwsim: avoid NULL skb in stop queue drain
mac80211_hwsim_stop() drops any frames left in data->pending. The loop
currently checks skb_queue_empty() and then dequeues separately.
That split is racy with TX status handling, which can remove a pending
frame under the queue lock. If the last entry is removed after the empty
check, skb_dequeue() returns NULL and the stop path passes that NULL skb
to ieee80211_free_txskb().
Use skb_dequeue() as the loop condition instead. The dequeue result is the
object that stop owns and frees, and a concurrent status completion that
empties the queue simply makes the loop terminate. |
| In the Linux kernel, the following vulnerability has been resolved:
cgroup/cpuset: Make nr_deadline_tasks an atomic_t
The nr_deadline_tasks variable in the cpuset structure was introduced by
commit 6c24849f5515 ("sched/cpuset: Keep track of SCHED_DEADLINE task
in cpusets"). It is reported by sashiko [1] that nr_deadline_tasks
can currently be modified by inc_dl_tasks_cs() under rq->lock and
by cpuset_attach() under cpuset_mutex. So if both updates happen
simultaneously, the nr_deadline_tasks variable can be corrupted leading
to incorrect operations down the road.
Fix that by changing its type to atomic_t so that nr_deadline_tasks
are always atomically updated. This fix patch is a low hanging fruit.
It can handle some of the races between a concurrent sched_setscheduler()
and cpuset_can_attach()/cpuset_attach() calls, but not all of them like
the other issue raised by sashiko [2]. This will be handled hopefully
in a future follow up patch.
[1] https://sashiko.dev/#/patchset/20260626181923.133658-1-longman%40redhat.com
[2] https://sashiko.dev/#/patchset/20260630033344.352702-1-longman%40redhat.com |
| In the Linux kernel, the following vulnerability has been resolved:
hwrng: core - fix rng list on registration error
hwrng_register(rng) does the following:
1. Checks if rng has name and read methods set
2. Checks if the name already exists
3. Adds rng to global rng_list
4. May try to set rng to current_rng
If step 4 fails, it returns an error. However, it does not remove the
rng from rng_list, causing a dangling reference which can result in
use-after-free if the caller frees rng, since registration failed.
Add a list_del_init() cleanup step. |
| In the Linux kernel, the following vulnerability has been resolved:
uprobes/x86: Move optimized uprobe from nop5 to nop10
Andrii reported an issue with optimized uprobes [1] that can clobber
redzone area with call instruction storing return address on stack
where user code may keep temporary data without adjusting rsp.
Fixing this by moving the optimized uprobes on top of 10-bytes nop
instruction, so we can squeeze another instruction to escape the
redzone area before doing the call, like:
lea -0x80(%rsp), %rsp
call tramp
Note the lea instruction is used to adjust the rsp register without
changing the flags.
We use nop10 and following transformation to optimized instructions
above and back as suggested by Peterz [2].
Optimize path (int3_update_optimize):
1) Initial state after set_swbp() installed the uprobe:
cc 2e 0f 1f 84 00 00 00 00 00
From offset 0 this is INT3 followed by the tail of the original
10-byte NOP.
After a previous unoptimization bytes 5..9 may still contain the
old call instruction, which remains valid for threads already there.
2) Rewrite the LEA tail and call displacement:
cc [8d 64 24 80 e8 d0 d1 d2 d3]
From offset 0 this traps on the uprobe INT3. Bytes 1..9 are not
executable entry points while byte 0 is trapped.
3) Publish the first LEA byte:
[48] 8d 64 24 80 e8 d0 d1 d2 d3
From offset 0 this is:
lea -0x80(%rsp), %rsp
call <uprobe-trampoline>
Unoptimize path (int3_update_unoptimize):
1) Initial optimized state:
48 8d 64 24 80 e8 d0 d1 d2 d3
Same as 3) above.
2) Trap new entries before restoring the NOP bytes:
[cc] 8d 64 24 80 e8 d0 d1 d2 d3
From offset 0 this traps. A thread that had already executed the
LEA can still reach the intact CALL at offset 5.
3) Restore bytes 1..4 of the original NOP while keeping byte 0 trapped
and byte 5 as CALL.
cc [2e 0f 1f 84] e8 d0 d1 d2 d3
From offset 0 this still traps. Offset 5 is still the CALL for any
thread that was already past the first LEA byte.
4) Publish the first byte of the original NOP:
[66] 2e 0f 1f 84 e8 d0 d1 d2 d3
From offset 0 this is the restored 10-byte NOP; the CALL opcode and
displacement are now only NOP operands. Offset 5 still decodes as
CALL for a thread that was already there.
Tthere is only a single target uprobe-trampoline for the given nop10
instruction address, so the CALL instruction will not be changed across
unoptimization/optimization cycles.
Therefore, any task that is preempted at the CALL instruction is guaranteed
to observe that CALL and not anything else.
Note as explained in [2] we need to use following nop10:
PF1 PF2 ESC NOPL MOD SIB DISP32
NOP10: 0x66, 0x2e, 0x0f, 0x1f, 0x84, 0x00, 0x00, 0x00, 0x00, 0x00 -- cs nopw 0x00000000(%rax,%rax,1)
which means we need to allow 0x2e prefix which maps to INAT_PFX_CS
attribute in is_prefix_bad function.
Also changing the uprobe syscall error when called out of uprobe
trampoline to -EPROTO, so we are able to detect the fixed kernel.
The optimized uprobe performance stays the same:
uprobe-nop : 3.129 ± 0.013M/s
uprobe-push : 3.045 ± 0.006M/s
uprobe-ret : 1.095 ± 0.004M/s
--> uprobe-nop10 : 7.170 ± 0.020M/s
uretprobe-nop : 2.143 ± 0.021M/s
uretprobe-push : 2.090 ± 0.000M/s
uretprobe-ret : 0.942 ± 0.000M/s
--> uretprobe-nop10: 3.381 ± 0.003M/s
usdt-nop : 3.245 ± 0.004M/s
--> usdt-nop10 : 7.256 ± 0.023M/s
[1] https://lore.kernel.org/bpf/20260509003146.976844-1-andrii@kernel.org/
[2] https://lore.kernel.org/bpf/20260518104306.GU3102624@noisy.programming.kicks-ass.net/#t |
| In the Linux kernel, the following vulnerability has been resolved:
dmaengine: zynqmp_dma: fix race between runtime PM and device removal
In zynqmp_dma_remove(), runtime PM was disabled only after checking
state and doing a manual suspend. This can race with runtime PM in the
remove/unbind (rmmod) path.
Disable runtime PM first, then suspend only if the device is not already
suspended. To prevent any further runtime PM transitions. |
| In the Linux kernel, the following vulnerability has been resolved:
drm/rockchip: analogix_dp: Fix OF node reference leak via auto cleanup
Sashiko reported a reference leak in rockchip_dp_drm_encoder_enable(),
the of_get_child_by_name() function does not call of_node_put() in a
symmetrical way [1].
Fix the device node reference leak by using __free(device_node) to
automatically manage of_node_put() for all device nodes. |
| In the Linux kernel, the following vulnerability has been resolved:
nvdimm: virtio_pmem: refcount requests for token lifetime
KASAN reports slab-use-after-free in __wake_up_common():
BUG: KASAN: slab-use-after-free in __wake_up_common+0x114/0x160
Read of size 8 at addr ffff88810fdcb710 by task swapper/0/0
CPU: 0 UID: 0 PID: 0 Comm: swapper/0 Not tainted
6.19.0-next-20260220-00006-g1eae5f204ec3 #4 PREEMPT(full)
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Arch Linux
1.17.0-2-2 04/01/2014
Call Trace:
<IRQ>
dump_stack_lvl+0x6d/0xb0
print_report+0x170/0x4e2
? __pfx__raw_spin_lock_irqsave+0x10/0x10
? __virt_addr_valid+0x1dc/0x380
kasan_report+0xbc/0xf0
? __wake_up_common+0x114/0x160
? __wake_up_common+0x114/0x160
__wake_up_common+0x114/0x160
? __pfx__raw_spin_lock_irqsave+0x10/0x10
__wake_up+0x36/0x60
virtio_pmem_host_ack+0x11d/0x3b0
? sched_balance_domains+0x29f/0xb00
? __pfx_virtio_pmem_host_ack+0x10/0x10
? _raw_spin_lock_irqsave+0x98/0x100
? __pfx__raw_spin_lock_irqsave+0x10/0x10
vring_interrupt+0x1c9/0x5e0
? __pfx_vp_interrupt+0x10/0x10
vp_vring_interrupt+0x87/0x100
? __pfx_vp_interrupt+0x10/0x10
__handle_irq_event_percpu+0x17f/0x550
? __pfx__raw_spin_lock+0x10/0x10
handle_irq_event+0xab/0x1c0
handle_fasteoi_irq+0x276/0xae0
__common_interrupt+0x65/0x130
common_interrupt+0x78/0xa0
</IRQ>
virtio_pmem_host_ack() wakes a request that has already been freed by the
submitter.
This happens when the request token is still reachable via the virtqueue,
but virtio_pmem_flush() returns and frees it.
Fix the token lifetime by refcounting struct virtio_pmem_request.
virtio_pmem_flush() holds a submitter reference, and the virtqueue holds an
extra reference once the request is queued. The completion path drops the
virtqueue reference, and the submitter drops its reference before
returning. |
| In the Linux kernel, the following vulnerability has been resolved:
memcg: move LRU size accounting on reparenting instead of copying it
When a memory cgroup is offlined its LRU folios are reparented to the
parent. lruvec_reparent_lru() splices the child's lists into the
parent's and credits the parent with the child's per-zone
lru_zone_size[], but never clears the child's copy, so the size is
copied rather than moved. lru_gen_reparent_memcg() does the same for
MGLRU.
The parent is left correct, credited with exactly the folios it took
over. The stale value sits on the child and nothing will correct it:
folio->memcg_data now resolves to the parent, so every later
update_lru_size() for those folios goes there.
Dying cgroups are not freed immediately and mem_cgroup_iter() still
walks them, so shrink_lruvec() keeps being called on them.
get_scan_count() reads the phantom counter through lruvec_lru_size() and
the scan loop then grinds through nr[] in SWAP_CLUSTER_MAX steps against
an empty list, for as long as the dead cgroup lives. Under MGLRU the
MGLRU scanner runs instead, but count_shadow_nodes() sums all of
NR_LRU_LISTS through lruvec_lru_size() and over-budgets the shadow node
limit just the same.
On one 251 GiB host a sweep of every mz->lru_zone_size[] found 380
counters describing folios on no list at all: 124777314 pages, 476 GiB,
1.89x the machine's RAM, across 57 cgroups. All were on memcgs with
CSS_DYING set and CSS_ONLINE clear, and parent/child pairs reported
byte-identical sizes.
LRU_UNEVICTABLE needs its size moved too. Its list is deliberately not
spliced because lruvec_init() poisons the head - the unevictable LRU is
imaginary and folios are never threaded on it - but the size is kept by
lruvec_add_folio()/lruvec_del_folio() and those folios account to the
parent from here on.
This depends on commit bf4ade7dbd76 ("memcg: keep folio's objcg same as
its node") and must not be backported ahead of it. Without that
invariant a folio's objcg can belong to another node, so a folio already
spliced onto the parent's list can still resolve to the child's lruvec
until the objcg's node is reparented in a later iteration of
memcg_reparent_objcgs(); clearing the child's counter early then lets
lruvec_del_folio() underflow it and trip the WARN_ONCE()/VM_BUG_ON() in
mem_cgroup_update_lru_size(). |
