OPINIONATED OMARCHY
TROUBLESHOOTING CORPUS
790
RECORDS
790
AUDITED
0
UNCHECKED
2388
SOURCES
LIVE
network high common DISPUTE

Stabilise a Realtek RTL8821CE or RTL8822CE that keeps dropping or crawls

omarchyarchendeavouroscachyosmanjarolaptop
CORRECTED An auditor found the fix wrong or incomplete and rewrote it.
AUDIT NOTE

The fix is exactly right and I verified every module and parameter name against the cited kernel sources rather than from memory. drivers/net/wireless/realtek/rtw88/pci.c lines 20-23 declare module_param_named(disable_msi, ...) and module_param_named(disable_aspm, rtw_pci_disable_aspm, bool, 0644), and main.c line 38 declares module_param_named(disable_lps_deep, rtw_disable_lps_deep_mode, bool, 0644). The rtw88 Makefile confirms main.o builds into rtw88_core and pci.o into rtw88_pci, and that rtw88_8821ce, rtw88_8822ce, rtw88_8822be and rtw88_8723de are real module names, so /etc/modprobe.d/70-rtw88.conf as written is correct, and the sysfs paths in the verify step exist. rtl8821ce-dkms-git and rtw88-dkms-git both exist in the AUR. Nothing here is Omarchy-3 shaped and the yay -S note correctly avoids a bare pacman upgrade. THE CAUSE IS WRONG ON A CHECKABLE SPECIFIC: it says the upstream DMI quirk table "only lists a handful of HP models". rtw_pci_quirks[] in pci.c has exactly TWO entries, "HP Notebook - P3S95EA#ACB" and "ASUS TUF Gaming A15 FA506II", so it is two machines from two vendors, not a handful of HP models. That is precisely the kind of invented specific that reads more authoritative than the text around it. Two smaller points folded into the corrected text: the RTL8723DE is 802.11n, not a Wi-Fi 5 part, and pci.c also carries a narrower runtime workaround (rx_no_aspm for 8821C behind an Intel bridge) worth knowing about. Finally, rtl8821ce-dkms-git was last updated 2023-01 per the AUR RPC, which is a real build risk against a 7.x kernel and belongs in the text rather than only implied by the danger note.

The cause above was rewritten on 2026-09-01 to match this note; the fix was corrected by the audit itself.
RISK

Blacklisting rtw88_8821ce in favour of a DKMS module means that if the DKMS build fails on the next kernel upgrade you boot with no Wi-Fi at all and no way to download a fix. Keep an ethernet cable or a USB Wi-Fi dongle to hand, always install the matching linux-headers before upgrading, and check dkms status after every kernel bump. Do not blacklist the in-tree driver until you have confirmed the DKMS one actually loads.

Symptom

Wi-Fi on an RTL8821CE / RTL8822BE / RTL8822CE (very common in budget HP, Lenovo and Asus laptops) connects but is unusable: a few hundred kbit/s, pings jumping to seconds, the link dropping every couple of minutes and reconnecting, or the card going dead after a resume. lspci -k shows Kernel driver in use: rtw88_8821ce. Users describe it as "my Wi-Fi works on Windows but is unusable on Arch".

Cause

The in-kernel rtw88 driver enables PCIe ASPM and the deep low-power-save mode on hardware whose platform implementation is broken, so the link stalls, crawls or dies after resume. Upstream rtw88/pci.c carries a DMI quirk table (rtw_pci_quirks[]) that force-disables ASPM and deep LPS, but it has only two entries, "HP Notebook - P3S95EA#ACB" and "ASUS TUF Gaming A15 FA506II", so every other affected machine has to set the module parameter by hand. The driver also force-disables ASPM at runtime for an 8821C sitting behind an Intel PCIe bridge (rx_no_aspm), which is why the same chip misbehaves on some boards and not others. This is the rtw88 sibling of the already-documented rtw89 problem, and rtw88 covers the older Realtek PCIe parts (8821CE, 8822BE, 8822CE and the 802.11n 8723DE) while rtw89 covers the Wi-Fi 6 ones.

Fix

Identify the chip and the driver actually bound:


lspci -knn | grep -A3 -i 'network\|wireless'

Then disable ASPM and deep power save. The parameter names come straight from the driver source. rtw88/pci.c declares module_param_named(disable_aspm, rtw_pci_disable_aspm, bool, 0644) in module rtw88_pci, and rtw88/main.c declares disable_lps_deep in rtw88_core:


sudo tee /etc/modprobe.d/70-rtw88.conf >/dev/null <<'EOF'
options rtw88_pci disable_aspm=y
options rtw88_core disable_lps_deep=y
EOF

sudo modprobe -r rtw88_8821ce rtw88_pci rtw88_core 2>/dev/null
sudo modprobe rtw88_8821ce

(Substitute rtw88_8822ce / rtw88_8822be / rtw88_8723de for your part. Doing this over ssh will drop the connection. Run it at the console.) Confirm:


cat /sys/module/rtw88_pci/parameters/disable_aspm
cat /sys/module/rtw88_core/parameters/disable_lps_deep

If disable_msi is also needed on your board (some report MSI interrupt trouble), the same module exposes it: add options rtw88_pci disable_msi=y.

If the in-tree driver is still unusable after that, fall back to an out-of-tree DKMS driver. It needs kernel headers, so install those first:


# Omarchy: use the AUR helper, never a bare pacman -Syu
yay -S linux-headers rtw88-dkms-git

Prefer rtw88-dkms-git: it is a backport of the whole mainline rtw88 series and is still maintained. The older single-chip rtl8821ce-dkms-git also exists, but its last AUR update was January 2023, so expect it to fail to build against a current kernel:


yay -S linux-headers rtl8821ce-dkms-git

Only after dkms status shows the replacement module built for your running kernel should you blacklist the in-tree one:


sudo dkms status
echo 'blacklist rtw88_8821ce' | sudo tee /etc/modprobe.d/blacklist-rtw88.conf
sudo mkinitcpio -P
sudo reboot

After the reboot lspci -k should show the replacement driver in Kernel driver in use:.

Verify

cat /sys/module/rtw88_pci/parameters/disable_aspm prints Y. Then run a sustained transfer and watch for drops: ping -i 0.2 -c 200 <your-router-ip> should show no gaps over one second and no packet loss, and journalctl -kf | grep -i rtw88 should be quiet during it.

Sources

rtw88-rtl8821ce-unstable-disable-aspm