[Bug 219857] panic in scsi_cd code

[bugzilla-noreply at freebsd.org]

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=219857

--- Comment #14 from Dexuan Cui <decui at microsoft.com> ---
(In reply to Kenneth D. Merry from comment #13)
Hi Ken, we found a regression caused by the patch: when we tested the patch
with recent 12-CURRENT (July 11) in a *UP* FreeBSD Virtual Machine running on
Windows Server 2012 R2, the VM hangs after printing 

Trying to mount root from ufs:/dev/da0s1a [rw]...

If I use a SMP VM (e.g. 2 CPUs or more), the issue will go away.
If I use a UP VM, and insert a .iso file into the virtual DVD drive, the issue
will go away as well.

However, the same UP VM image can run fine on a Windows Server 2016 host with 1
CPU, and no .iso file in the virtual DVD drive.

So I can't say this patch has a bug. My gut feeling is that: maybe the patch
somehow triggers a host issue, e.g. maybe the host's emulation of DVD drive
returns some invalid value to the VM and the VM hangs? When the issue happens,
the VM's CPU utilization is 0%, so I'm sure the VM is not in a busy loop.
Probably the VM is waiting for some condition?

I'm attaching the good log (I insert a .iso file into the virtual DVD) and the
bad log (the virtual DVD is empty). I got the log with the same VM running on
Windows Server 2012 R2. 


The good one is:
...
cd0 at ata1 bus 0 scbus1 target 0 lun 0
cd0: <Msft Virtual CD/ROM 1.0> Removable CD-ROM SPC-3 SCSI device
cd0: 16.700MB/s transfers (WDMA2, ATAPI 12bytes, PIO 65534bytes)
cd0: 669MB (342562 2048 byte sectors)
...
Trying to mount root from ufs:/dev/da0s1a [rw]...
GEOM: new disk da0
GEOM: new disk cd0
...
start_init: trying /sbin/init
...
login:


The bad one is:
cd0 at ata1 bus 0 scbus1 target 0 lun 0
cd0: <Msft Virtual CD/ROM 1.0> Removable CD-ROM SPC-3 SCSI device
cd0: 16.700MB/s transfers (WDMA2, ATAPI 12bytes, PIO 65534bytes)
cd0: Attempt to query device size failed: NOT READY, Medium not present
...
Trying to mount root from ufs:/dev/da0s1a [rw]...
GEOM: new disk cd0
GEOM: new disk da0
(Dexuan: it hangs here for ever)


How can I know where it hangs? Any insight or suggestion?  Thanks!

-- 
You are receiving this mail because:
You are on the CC list for the bug.