Re: kern/55699: Panic certainly in iSCSI initiator

[BERTRAND Joël <joel.bertrand%systella.fr@localhost>]

Michael van Elst a écrit :
> The following reply was made to PR kern/55699; it has been noted by GNATS.
> 
> From: mlelstv%serpens.de@localhost (Michael van Elst)
> To: gnats-bugs%netbsd.org@localhost
> Cc: 
> Subject: Re: kern/55699: Panic certainly in iSCSI initiator
> Date: Wed, 7 Oct 2020 13:55:15 -0000 (UTC)
> 
>  joel.bertrand%systella.fr@localhost writes:
>  
>  >	Randomly, when bacula lauches a job, system panics with :
>  >[  3409.124192] uvm_fault(0xffffffff8151f760, 0xffffffff819ca000, 4) -> e
>  >[  3409.124192] fatal page fault in supervisor mode
>  >[  3409.124192] trap type 6 code 0x10 rip 0xffffffff819ca061 cs 0x8 rflags
>  >0x10293 cr2 0xffffffff819ca061 ilevel 0 rsp 0xffffae013d79abb0
>  >[  3409.124192] curlwp 0xfffffa749e1ef060 pid 160.1 lowest kstack
>  >0xffffae013d7982c0
>  >[  3409.124192] panic: trap
>  >[  3409.124192] cpu5: Begin traceback...
>  >[  3409.133355] vpanic() at netbsd:vpanic+0x160
>  >[  3409.133355] snprintf() at netbsd:snprintf
>  >[  3409.133355] startlwp() at netbsd:startlwp
>  >[  3409.133355] alltraps() at netbsd:alltraps+0xbb
>  >[  3409.133355] usbd_devinfo_vp() at netbsd:usbd_devinfo_vp+0xa9
>  
>  That crash is unrelated to backup, network or iSCSI.

	I have seen that this panic was related to usb. Unfortunately, this
server only uses keyboard and mouse :

legendre# usbdevs
addr 0: xHCI root hub, NetBSD
addr 0: xHCI root hub, NetBSD
 addr 1: Microsoft 3-Button Mouse with IntelliEye(TM), Microsoft
 addr 2: USB Keyboard, Logitech
addr 1: EHCI root hub, NetBSD
 addr 2: Rate Matching Hub, Intel
addr 1: EHCI root hub, NetBSD
 addr 2: Rate Matching Hub, Intel
legendre#

>  Something is running usbdevs or calls an equivalent function and
>  the USB driver crashes. Maybe bacula has support of USB devices
>  and probes for them.

	I've never seen this kind of devices in bacula.

>  The rip is the same as the fault adress (cr2), which means some
>  code should be executed that is no longer in memory. My guess is
>  that the usbverbose module is unloaded while in use (which would
>  be a bug). A workaround would then be to load that module manually.

	JKB