Bug 204769

On Wed, 2019-09-04 at 15:18 +0000, bugzilla-daemon@bugzilla.kernel.org
wrote:
> https://bugzilla.kernel.org/show_bug.cgi?id=204769
> 
>             Bug ID: 204769
>            Summary: SCSI devices missing for disks attached to
> controller
>            Product: SCSI Drivers
>            Version: 2.5
>     Kernel Version: 5.12.11
>           Hardware: Intel
>                 OS: Linux
>               Tree: Mainline
>             Status: NEW
>           Severity: normal
>           Priority: P1
>          Component: AACRAID
>           Assignee: scsi_drivers-aacraid@kernel-bugs.osdl.org
>           Reporter: linux@lanrules.de
>         Regression: No
> 
> We use an Adaptec ASR8405 RAID controller with LSI backplane. The
> RAID
> controller is configured to expose attached disks to the system
> ("Expose RAW").
> 
> Up to Linux 5.1.12, devices would show as expected. lsscsi -v
> reports:
> > [0:0:0:0]    disk    ASR8405  storage          V1.0  /dev/sda
> >   dir: /sys/bus/scsi/devices/0:0:0:0 
> >  
> > [/sys/devices/pci0000:60/0000:60:03.1/0000:61:00.0/host0/target0:0:
> > 0/0:0:0:0]
> > [0:1:4:0]    disk    ATA      HGST HUH721212AL T3D0  -
> >   dir: /sys/bus/scsi/devices/0:1:4:0 
> >  
> > [/sys/devices/pci0000:60/0000:60:03.1/0000:61:00.0/host0/target0:1:
> > 4/0:1:4:0]
> > [0:1:5:0]    disk    ATA      HGST HUH721212AL T3D0  -
> >   dir: /sys/bus/scsi/devices/0:1:5:0 
> >  
> > [/sys/devices/pci0000:60/0000:60:03.1/0000:61:00.0/host0/target0:1:
> > 5/0:1:5:0]
> > [0:1:6:0]    disk    ATA      HGST HUH721212AL T3D0  -
> >   dir: /sys/bus/scsi/devices/0:1:6:0 
> >  
> > [/sys/devices/pci0000:60/0000:60:03.1/0000:61:00.0/host0/target0:1:
> > 6/0:1:6:0]
> > [0:1:7:0]    disk    ATA      HGST HUH721212AL T3D0  -
> >   dir: /sys/bus/scsi/devices/0:1:7:0 
> >  
> > [/sys/devices/pci0000:60/0000:60:03.1/0000:61:00.0/host0/target0:1:
> > 7/0:1:7:0]
> > [0:1:8:0]    disk    ATA      HGST HUH721212AL T3D0  -
> >   dir: /sys/bus/scsi/devices/0:1:8:0 
> >  
> > [/sys/devices/pci0000:60/0000:60:03.1/0000:61:00.0/host0/target0:1:
> > 8/0:1:8:0]
> > [0:1:9:0]    disk    ATA      HGST HUH721212AL T3D0  -
> >   dir: /sys/bus/scsi/devices/0:1:9:0 
> >  
> > [/sys/devices/pci0000:60/0000:60:03.1/0000:61:00.0/host0/target0:1:
> > 9/0:1:9:0]
> > [0:1:10:0]   disk    ATA      HGST HUH721212AL T3D0  -
> >   dir: /sys/bus/scsi/devices/0:1:10:0 
> >  
> > [/sys/devices/pci0000:60/0000:60:03.1/0000:61:00.0/host0/target0:1:
> > 10/0:1:10:0]
> > [0:1:11:0]   disk    ATA      HGST HUH721212AL T3D0  -
> >   dir: /sys/bus/scsi/devices/0:1:11:0 
> >  
> > [/sys/devices/pci0000:60/0000:60:03.1/0000:61:00.0/host0/target0:1:
> > 11/0:1:11:0]
> > [0:1:12:0]   disk    ATA      HGST HUH721212AL T3D0  -
> >   dir: /sys/bus/scsi/devices/0:1:12:0 
> >  
> > [/sys/devices/pci0000:60/0000:60:03.1/0000:61:00.0/host0/target0:1:
> > 12/0:1:12:0]
> > [0:1:13:0]   disk    ATA      HGST HUH721212AL T3D0  -
> >   dir: /sys/bus/scsi/devices/0:1:13:0 
> >  
> > [/sys/devices/pci0000:60/0000:60:03.1/0000:61:00.0/host0/target0:1:
> > 13/0:1:13:0]
> > [0:3:0:0]    enclosu LSI      SAS3x36          0601  -
> >   dir: /sys/bus/scsi/devices/0:3:0:0 
> >  
> > [/sys/devices/pci0000:60/0000:60:03.1/0000:61:00.0/host0/target0:3:
> > 0/0:3:0:0]
> > [N:0:4:1]    dsk/nvm Samsung SSD 970 PRO
> > 1TB__1                 /dev/nvme0n1
> >   dir: /sys/class/nvme/nvme0/nvme0n1 
> >  
> > [/sys/devices/pci0000:10/0000:10:01.3/0000:13:00.0/nvme/nvme0/nvme0
> > n1]
> > [N:1:4:1]    dsk/nvm Samsung SSD 970 PRO
> > 1TB__1                 /dev/nvme1n1
> >   dir: /sys/class/nvme/nvme1/nvme1n1 
> >  
> > [/sys/devices/pci0000:10/0000:10:01.4/0000:14:00.0/nvme/nvme1/nvme1
> > n1]
> 
> On Linux 5.12.11, the devices are missing. lsscsi -v reports:
> > [0:0:0:0]    disk    ASR8405  storage          V1.0  /dev/sda
> >   dir: /sys/bus/scsi/devices/0:0:0:0 
> >  
> > [/sys/devices/pci0000:60/0000:60:03.1/0000:61:00.0/host0/target0:0:
> > 0/0:0:0:0]
> > [0:3:0:0]    enclosu LSI      SAS3x36          0601  -
> >   dir: /sys/bus/scsi/devices/0:3:0:0 
> >  
> > [/sys/devices/pci0000:60/0000:60:03.1/0000:61:00.0/host0/target0:3:
> > 0/0:3:0:0]
> > [N:1:4:1]    dsk/nvm Samsung SSD 970 PRO
> > 1TB__1                 /dev/nvme1n1
> >   dir: /sys/class/nvme/nvme0/nvme1n1 
> >  
> > [/sys/devices/pci0000:10/0000:10:01.3/0000:13:00.0/nvme/nvme0/nvme1
> > n1]
> > [N:0:4:1]    dsk/nvm Samsung SSD 970 PRO
> > 1TB__1                 /dev/nvme0n1
> >   dir: /sys/class/nvme/nvme1/nvme0n1 
> >  
> > [/sys/devices/pci0000:10/0000:10:01.4/0000:14:00.0/nvme/nvme1/nvme0
> > n1]
> 
> Output in dmesg is basically the same, except that the following
> lines are
> missing with the newer kernel:
> 
> > [   17.016661] scsi 0:1:4:0: Direct-Access     ATA      HGST
> > HUH721212AL T3D0
> > PQ: 1 ANSI: 6
> > [   17.018313] scsi 0:1:5:0: Direct-Access     ATA      HGST
> > HUH721212AL T3D0
> > PQ: 1 ANSI: 6
> > [   17.020129] scsi 0:1:6:0: Direct-Access     ATA      HGST
> > HUH721212AL T3D0
> > PQ: 1 ANSI: 6
> > [   17.022530] scsi 0:1:7:0: Direct-Access     ATA      HGST
> > HUH721212AL T3D0
> > PQ: 1 ANSI: 6
> > [   17.024291] scsi 0:1:8:0: Direct-Access     ATA      HGST
> > HUH721212AL T3D0
> > PQ: 1 ANSI: 6
> > [   17.026041] scsi 0:1:9:0: Direct-Access     ATA      HGST
> > HUH721212AL T3D0
> > PQ: 1 ANSI: 6
> > [   17.027725] scsi 0:1:10:0: Direct-Access     ATA      HGST
> > HUH721212AL
> > T3D0 PQ: 1 ANSI: 6
> > [   17.029395] scsi 0:1:11:0: Direct-Access     ATA      HGST
> > HUH721212AL
> > T3D0 PQ: 1 ANSI: 6
> > [   17.031145] scsi 0:1:12:0: Direct-Access     ATA      HGST
> > HUH721212AL
> > T3D0 PQ: 1 ANSI: 6
> > [   17.032672] scsi 0:1:13:0: Direct-Access     ATA      HGST
> > HUH721212AL
> > T3D0 PQ: 1 ANSI: 6
> 
> The only change on the system to obtain the different behavior is
> upgrade/downgrade of the kernel via kexec.

So the driver is loading as /dev/sda is being seen above.
Seems likely to be the enclosure then that is not being probed
[0:0:0:0]    disk    ASR8405  storage          V1.0  /dev/sda 
The enclosure looks like its on another port correct ?
0:1:x:x

I am looking at what may have changed in the latest commits
>

> The enclosure looks like its on another port correct ?
> 0:1:x:x
Funny enough, the enclosure is 0:3:0:0, and is properly exposed by both kernel versions.

Btw., it seems like this is a duplicate of https://bugzilla.kernel.org/show_bug.cgi?id=204173. Sorry I didn't spot it when submitting, got confused by how Bugzilla orders search results.

I am seeing the exact same issue.
No hardware changes, but updating beyond 5.2 breaks this.
I'll note that if you force the expose_physicals Parm to 1, the disks do still become exposed as /dev/ah* and /dev/sd* even behind an array.
In protect mode( -1 ), sg* should be RO but is not exposed at all.

I believe I've run into the same issue with my Debian Distro kernel (5.10.x) and reported it in Debian's bug tracker at:

https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1027921.

This Kernel.org bug was linked in Debian's Tracker and this comment is intended to provide the backlink.

https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1027921#47 also suggests that https://github.com/torvalds/linux/commit/948e922fc44611ee2de0c89583ca958cb5307d36 might be the cause of this bug. I'm not personally qualified to tell.

Personally, I'd say that 948e922fc446 ("scsi: core: map PQ=1, PDT=other values to SCSI_SCAN_TARGET_PRESENT") is indeed broken.

But this has been discussed previously, and the SCSI maintainer disagrees:
https://www.mail-archive.com/linux-scsi@vger.kernel.org/msg84659.html

Back then, the proposed solution was to "either have a SCSI host flag to override the behavior or consider masking PQ in the driver", but apparently neither idea has been implemented.