pathless nvme drive crashes block discovery

Bug #1847771 reported by Dimitri John Ledkov
12
This bug affects 2 people
Affects Status Importance Assigned to Milestone
curtin
Fix Released
Critical
Michael Hudson-Doyle
subiquity
Fix Released
Undecided
Unassigned

Bug Description

pathless nvme drive crashes block discovery

I have an nvme drive, and a regular sda usb stick plugged in.

Somehow multipath declares a path to my nmve drive.... without any map setup.

This seems to crash everything.

Revision history for this message
Dimitri John Ledkov (xnox) wrote :
Revision history for this message
Dimitri John Ledkov (xnox) wrote :
Revision history for this message
Dimitri John Ledkov (xnox) wrote :
Revision history for this message
Dimitri John Ledkov (xnox) wrote :

2019-10-11 14:49:46,529 curtin:1230 Validation error: None is not of type 'string' in
{
 "id": "disk-sda",
 "multipath": null,
 "path": "/dev/sda",
 "ptable": "dos",
 "serial": "SanDisk_Ultra_USB_3.0_4C530001200208100222-0:0",
 "type": "disk"
}
NoneType: None

The null in multipath is weird, but check this out - a mapless multipath:

    "multipath": {
        "paths": [
            {
                "device": "nvme0n1",
                "serial": "27FF079419DE00131387",
                "multipath": "[orphan]",
                "host_wwnn": "[undef]",
                "target_wwnn": "[undef]",
                "host_wwpn": "[undef]",
                "target_wwpn": "[undef]",
                "host_adapter": "[undef]"
            }
        ]
    },

Revision history for this message
Dimitri John Ledkov (xnox) wrote :

  File "/usr/lib/python3/dist-packages/curtin/storage_config.py", line 1232, in extract_storage_config
    raise RuntimeError("Extract storage config does not validate.")
RuntimeError: Extract storage config does not validate.

Changed in curtin (Ubuntu):
importance: Undecided → Critical
tags: added: rls-ee-incoming
Revision history for this message
Launchpad Janitor (janitor) wrote :

Status changed to 'Confirmed' because the bug affects multiple users.

Changed in curtin (Ubuntu):
status: New → Confirmed
Revision history for this message
Michael Hudson-Doyle (mwhudson) wrote :

Yeah, I have a similar config in my laptop, and:

(master)mwhudson@anduril:~/src/probert$ sudo multipathd show paths
hcil dev dev_t pri dm_st chk_st dev_st next_check
0:1:1:1 nvme0n1 259:0 1 undef undef unknown orphan
(master)mwhudson@anduril:~/src/probert$ sudo multipathd show maps
(master)mwhudson@anduril:~/src/probert$

what is going on.

Revision history for this message
Michael Hudson-Doyle (mwhudson) wrote :

Ah so curtin takes DM_MULTIPATH_DEVICE_PATH being set at all as an indication of a multipath device, but per the follow up to https://www.spinics.net/lists/dm-devel/msg35965.html that's wrong: only a value of "1" indicates a multipath device.

Revision history for this message
Michael Hudson-Doyle (mwhudson) wrote :

This is fixed in the stable/ubuntu-19.10 channel now. If you boot an eoan daily (up to and including 20191014, the next one will contain the fix), it will offer to update you to 19.10.2 which should then be able to complete the install.

Changed in subiquity:
status: New → Fix Committed
Revision history for this message
Dan Watkins (oddbloke) wrote :

Dropped the Ubuntu tasks as (IIUC) the code that caused this problem isn't in the Ubuntu package, we're only seeing it because subiquity is/was consuming curtin master.

Changed in curtin:
status: New → Fix Committed
assignee: nobody → Michael Hudson-Doyle (mwhudson)
no longer affects: curtin (Ubuntu)
no longer affects: curtin (Ubuntu Eoan)
Changed in curtin:
importance: Undecided → Critical
Changed in subiquity:
status: Fix Committed → Fix Released
Changed in curtin:
status: Fix Committed → Fix Released
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.