dell-dw5930e¶
Dell DW5930e — 5G modem on PCIe/MHI, a documented gap for parking with the radio switch as the way to quiet it
Vendor: Dell
Status: untested — catalogued from documented sources, not yet proven here.
Not maintainer-verified: nobody on this project has run this hardware yet. The two facts are recorded separately on purpose (D-027) — a correct recipe and a tested device are different claims.
What it is¶
A Dell 5G modem, a PCIe card that talks to the host over MHI and is driven by mhi_pci_generic. The maintainer's bring-up notes (not part of this repository, and not re-read on 2026-10-01: the card is out of the machine) identify it as a Snapdragon X55 module, Foxconn T99W175. It is the 5G card the field laptop is meant to carry; the 4G DW5821e is in its slot at the moment.
What you can do with it¶
Carry the station's data link and, through its diagnostic channel, host cellular tooling (issue #69). To quiet it, use the radio switch; parking through the helper is not available for a PCIe/MHI card.
Setup¶
Nothing to install; ModemManager and NetworkManager drive it. The maintainer's machine carries a udev rule that holds the card out of runtime suspend, because without it the card crash-loops; it is the maintainer's, it is not shipped, and this entry does not change it.
Known problems¶
Cannot be parked: see the gap above. The antenna leads and the runtime-PM behaviour are the physical and kernel problems the maintainer's bring-up records; neither is a catalog matter.
Power control¶
This entry declares pci_runtime, which the helper refuses: it cannot be parked or woken today. The note below says why and what to use instead.
Not implemented, by design: pci_runtime is refused at plan time (D-056) and this card is not a USB node, so nothing here can be parked. The route is the radio switch, nmcli radio wwan off (the helper's radio off wwan verb and the tray's Radios toggle are to wrap it, planned and not yet released), which silences the radio without detaching the card and needs no root. A runtime-suspended PCIe function is not parked in a sense state could read back, and a card that crash-loops on D3cold (this one did, until its own udev rule) is the wrong thing to suspend by hand. Unmeasured: nothing about parking this card has been run.
See device power control for what parking does to a machine, how to inspect it and how to reverse it.
What is not yet known¶
No USB identifier exists to carry: the card is PCIe/MHI. The catalog's identifier model is USB-only, so the PCI identifier (105b:e0b1, recorded by the maintainer's bring-up notes when the card was fitted on 2026-09-21) is prose here, not a field, and was not re-read on 2026-10-01 because the card is not in the machine at the moment; the DW5821e is. Closing this needs the card back in the machine and lspci -nn plus a decision on a PCI identifier model, which is the maintainer's.
Who can close it: the maintainer owns this hardware; the gap closes at the next capture session.