Re: [Scst-devel] iscsi_scst: Unknown symbol backport_dependency_symbol

[Carsten A. <car...@ae...>]

Hi Bart (et al.),

On 07/17/18 18:46, Bart Van Assche wrote:
> I think a MOFED support contract has to be bought separately and that it
> is not included in the adapter warranty. I'm not sure that this is the
> right document but I think the following document explains the details:
> http://www.mellanox.com/pdf/products/M1Global_Brochure.pdf.
> 

I see, thus if we wanted to go with MOFED we should go with at least
their silver plan - understood and filed ;)


> The SCST trunk should work fine. The only feature in the trunk of which
> we know that it does not work reliably in combination with recent kernel
> versions is SCSI pass-through. This refers to the dev_* handlers, e.g.
> dev_disk. There have been significant changes in the Linux kernel v4.11
> block layer API and we need to review the integration of SCST and
> kernels >= 4.11. The target drivers in the trunk and the scst_vdisk and
> scst_fileio handlers should work fine however.

OK, first of, please bear with me as I'm still learning (a lot).

My understanding is to hook up our Solaris initiators via iSER, I want
our Linux SCST targets to run mlx4_core (for the connect X3 card),
ib_umad to enable opensm on the Linux end as both machines are currently
connected back to back via a single port.

Then for the initial handshake ib_ipoib plus IP configuration. Right now
in datagram mode but I have been able to also run in connected mode with
maximum possible mtu (only tested with simple ping -Mdo -s<payload>).

As suggested to possibly avoid the problems seen with Debian's 4.9
kernel I am now using Debian's stretch-backports kernel

4.16.0-0.bpo.2-amd64 #1 SMP Debian 4.16.16-2~bpo9+1 (2018-06-26) x86_64
GNU/Linux

which means I must use svn trunk as even the 3.3 branch will not compile
the scst module anymore.

No problem, but on its first iSER connection this kernel/module bug is
triggered:

Jul 18 12:13:17 fc1 kernel: scst: Using autogenerated relative target id
1 for target iqn.2008-05.local.atlas:sam1
Jul 18 12:13:17 fc1 scst[6604]: Loading and configuring SCST.
Jul 18 12:13:17 fc1 systemd[1]: Started LSB: SCST - A Generic SCSI
Target Subsystem.
Jul 18 12:13:46 fc1 kernel: iscsi-scst: isert_cm_evt:CONNECT_REQUEST(4)
status:0 portal:0000000026a69e0a cm_id:000000006256209c
Jul 18 12:13:46 fc1 kernel: iscsi-scst: iser created device:00000000feba73ad
Jul 18 12:13:46 fc1 kernel: iscsi-scst: iser created
cm_id:000000006256209c qp:0x20B
Jul 18 12:13:46 fc1 kernel: iscsi-scst: iser accepted connection
cm_id:000000006256209c 192.168.110.40->192.168.110.1
Jul 18 12:13:46 fc1 kernel: iscsi-scst: iser rx pdu before conn
established, pdu saved
Jul 18 12:13:46 fc1 kernel: iscsi-scst: isert_cm_evt:ESTABLISHED(9)
status:0 portal:0000000026a69e0a cm_id:000000006256209c
Jul 18 12:13:46 fc1 iscsi-scstd[6684]: iSER Connect to 192.168.110.1:3260
Jul 18 12:13:46 fc1 kernel: iscsi-scst: iser push saved rx pdu
Jul 18 12:13:46 fc1 kernel: usercopy: Kernel memory exposure attempt
detected from SLUB object 'scst_cmd' (offset 200, size 48)!
Jul 18 12:13:46 fc1 kernel: ------------[ cut here ]------------
Jul 18 12:13:46 fc1 kernel: kernel BUG at
/build/linux-Sci2oS/linux-4.16.16/mm/usercopy.c:100!
Jul 18 12:13:46 fc1 kernel: invalid opcode: 0000 [#1] SMP PTI

full log in attachment.

And again, I'm hitting a point where I don't know how to proceed (sorry
again), do you?

Cheers
Carsten

-- 
Dr. Carsten Aulbert, Max Planck Institute for Gravitational Physics,
Callinstraße 38, 30167 Hannover, Germany
Phone: +49 511 762 17185