Re: [Scst-devel] QLogic Target driver 0.9.3.5

SourceForge Headquarters 225 Broadway Suite 1600 San Diego, CA 92101 +1 (858) 454-5900

I'll try testing these changes this week. Since my test system is 2.6.14 an=
d
it would take me quite a bit to update to 2.6.15 I'd have to test it again
on my production system which I can't take down tonight. I would persue the
upgrade of my test system if we were focusing on that version.

- Mark.

On 2/21/06, Nathaniel Clark <na...@mi...> wrote:
>
> Mark,
>
> I'll put in a check for what I think is causing this in my drop today,
> but if you could rerun the test with that code, I'd be very grateful.
>
> - nate
>
> Mark Buechler wrote:
>
> > Look in /sys/class/scsi_host/host<x>/target_mode_enable or something
> > like that.
> >
> > I compiled and installed kernel 2.6.15 tonight with this version of
> > qla2x00tgt and have the following:
> >
> > -------------------------------------------------------
> > ansrv1 login: QLogic Fibre Channel HBA Driver
> > qla2200 0000:02:04.0: Found an ISP2200, irq 177, iobase 0xf8bec000
> > qla2200 0000:02:04.0: Configuring PCI space...
> > qla2200 0000:02:04.0: Configure NVRAM parameters...
> > qla2200 0000:02:04.0: Verifying loaded RISC code...
> > qla2200 0000:02:04.0: Waiting for LIP to complete...
> > qla2200 0000:02:04.0: LOOP UP detected (1 Gbps).
> > qla2200 0000:02:04.0: Topology - (F_Port), Host Loop address 0xffff
> > scsi5 : qla2xxx
> > qla2200 0000:02:04.0:
> >  QLogic Fibre Channel HBA Driver: 8.01.03-k
> >   QLogic QLA22xx -
> >   ISP2200: PCI (33 MHz) @ 0000:02:04.0 hdma-, host#=3D5, fw=3D2.02.08 T=
P
> > qla2200 0000:02:05.0: Found an ISP2200, irq 169, iobase 0xf8c0e000
> > qla2200 0000:02:05.0: Configuring PCI space...
> > qla2200 0000:02:05.0: Configure NVRAM parameters...
> > qla2200 0000:02:05.0: Verifying loaded RISC code...
> > qla2200 0000:02:05.0: Waiting for LIP to complete...
> > qla2200 0000:02: 05.0: LOOP UP detected (1 Gbps).
> > qla2200 0000:02:05.0: Topology - (F_Port), Host Loop address 0xffff
> > scsi6 : qla2xxx
> > qla2200 0000:02:05.0:
> >  QLogic Fibre Channel HBA Driver: 8.01.03-k
> >   QLogic QLA22xx -
> >   ISP2200: PCI (33 MHz) @ 0000:02: 05.0 hdma-, host#=3D6, fw=3D2.02.08 =
TP
> > qla2x00tgt: no version for "scst_unregister" found: kernel tainted.
> > q2t_host_action:qla2x00tgt(0): Using 32 Bit PCI Addressing
> > qla2200 0000:02:04.0: Performing ISP error recovery - ha=3D f5088284.
> > qla2200 0000:02:04.0: LOOP UP detected (1 Gbps).
> > q2t_handle_imm_notify:Port config changed, Global TPRLO or Global LOGO
> > (46)
> > q2t_make_name:***ERROR*** qla2x00tgt(0): Unable to find wwn login for
> > loop id 0, using loop id instead
> > [10734]: scst_init_session:scst: Name 0 not found, using default group
> > [10734]: scst_alloc_add_tgt_dev:Virtual device SCST lun=3D0
> > [10734]: scst_alloc_add_tgt_dev:Virtual device SCST lun=3D1
> > q2t_build_ctio_pkt:handle(scst_cmd) -> 00008001, timeout 30 L 0x0 -> I
> > 0x0 E 0xbef0
> > Unable to handle kernel NULL pointer dereference at virtual address
> > 00000009
> >  printing eip:
> > f8c30a14
> > *pde =3D 00000000
> > Oops: 0000 [#1]
> > Modules linked in: qla2x00tgt qla2200 qla2xxx firmware_class
> > scst_fileio scsi_tgt ipv6 eeprom lm92 i2c_isa i2c_dev ipmi_devi
> > ntf ipmi_msghandler ide_disk ide_cd cdrom serio_raw sg floppy evdev
> > pcspkr rtc ext3 mbcache raid6 raid5 xor raid10 raid0 lin
> > ear ide_generic generic shpchp pci_hotplug ohci_hcd usbcore
> > serverworks ide_core i2c_piix4 i2c_core e100 mii jbd raid1 md_mo
> > d lpfc scsi_transport_fc sym53c8xx aic7xxx aic79xx scsi_transport_spi
> > sd_mod scsi_mod dm_mod
> > CPU:    0
> > EIP:    0060:[<f8c30a14>]    Tainted: GF     VLI
> > EFLAGS: 00010046   (2.6.15-2-686-scst)
> > EIP is at q2t_do_ctio_completion+0x804/0x820 [qla2x00tgt]
> > eax: 00002200   ebx: f5088284   ecx: 00008040   edx: 00000000
> > esi: f5088284   edi: 00008001   ebp: 00008001   esp: c0313e48
> > ds: 007b   es: 007b   ss: 0068
> > Process swapper (pid: 0, threadinfo=3Dc0312000 task=3Dc02d0b00)
> > Stack: 00000010 00000001 00000000 00000020 f503964c f57f1da0 f60e16b4
> > f50355c0
> >        f8c315ba 00000000 00000001 f5088284 f5088284 f5088284 00008001
> > 00000001
> >        f8c30b71 f5088284 00008001 00000001 00000000 f5088284 f5088284
> > 00000000
> > Call Trace:
> >  [<f8c315ba>] q2t_send_cmd_to_scst+0x22a/0x6d0 [qla2x00tgt]
> >  [<f8c30b71>] q2t_ctio_completion+0x141/0x170 [qla2x00tgt]
> >  [<f8cd6c59>] qla2x00_process_completed_request+0x39/0x100 [qla2xxx]
> >  [<f8cd6e30>] qla2x00_process_response_queue+0x110/0x1c0 [qla2xxx]
> >  [<f8cd6aa5>] qla2x00_async_event+0x685/0x800 [qla2xxx]
> >  [<c01143ca>] scheduler_tick+0x1a/0x2f0
> >  [<f8cd5fdd>] qla2100_intr_handler+0x10d/0x1e0 [qla2xxx]
> >  [<c0136f30>] handle_IRQ_event+0x30/0x70
> >  [<c0136fc2>] __do_IRQ+0x52/0xb0
> >  [<c011c66b>] __do_softirq+0x7b/0x90
> >  [<c0105229>] do_IRQ+0x19/0x30
> >  [<c01036be>] common_interrupt+0x1a/0x20
> >  [<c0100cf1>] default_idle+0x41/0x70
> >  [<c0100d8f>] cpu_idle+0x4f/0x60
> >  [<c0314882>] start_kernel+0x182/0x1c0
> >  [<c03143c0>] unknown_bootoption+0x0/0x1e0
> > Code: 24 08 e8 40 73 4e c7 e9 b3 f8 ff ff c7 04 24 28 4e c3 f8 b8 f6
> > 49 c3 f8 89 44 24 04 e8 26 73 4e c7 e9 89 f8 ff ff 90 8
> > b 54 24 24 <0f> b6 52 09 89 54 24 20 e9 3e f8 ff ff eb 0d 90 90 90 90
> > 90 90
> >  <0>Kernel panic - not syncing: Fatal exception in interrupt
> > ------------------------------------------------------
> >
> > It seems as soon as I enable target mode an initiator is seen. Since
> > all my initiators are powered OFF, I can only assume it sees itself.
> >
> > - Mark.
> >
> > On 2/20/06, * Ming Zhang* <mi...@el...
> > <mailto:mi...@el...>> wrote:
> >
> >     sorry a dumb question.
> >
> >     how to enable the target mode on this new qla driver? :P
> >
> >     ming
> >
> >
> >     On Mon, 2006-02-20 at 17:30 -0500, Ming Zhang wrote:
> >     > could u consider merging this trivial patch to remove some compil=
e
> >     > warning on 64bit? thanks!
> >     >
> >     > ming
> >     >
> >     > On Mon, 2006-02-20 at 15:37 -0500, Nathaniel Clark wrote:
> >     > > On Monday 20 February 2006 1:18 pm, you wrote:
> >     > > > Nate,
> >     > > >
> >     > > > The code looks OK. Reviewing it I have some questions:
> >     > > >
> >     > > > 1. You somewhere replaced memset() on MEMSET(). Why? I
> >     haven't found its
> >     > > > definition anywhere
> >     > >
> >     > > ooo...  damn! yea I moved the definitions of those around in
> >     my internal
> >     > > codebase and the condition definition didn't make it in. it
> >     should look like
> >     > > this: (patch attached, I'll re-release tomorrow, I'm not at
> >     work today)
> >     > >
> >     > > #ifndef MEMCPY
> >     > > /* Debugging to see how many memcopys/memsets happen */
> >     > > #define MEMCPY(d, s, l) do
> {                                    \
> >     > >                 TRACE_MEM("memcpy to %p size %d", d,
> l);        \
> >     > >                 memcpy(d, s,
> l);                                \
> >     > >         } while (0)
> >     > > #endif
> >     > >
> >     > > #ifndef MEMSET
> >     > > #define MEMSET(d, c, l) do
> {                                    \
> >     > >                 TRACE_MEM("memset of %p size %d", d,
> l);        \
> >     > >                 memset(d, c,
> l);                                \
> >     > >         } while (0)
> >     > > #endif
> >     > >
> >     > > >
> >     > > > 2. If you replaced tgt->sess array by sess_list to save
> >     memory, but I
> >     > > > would consider to implement q2t_find_sess_by_lid() function
> >     using hash
> >     > > > search instead of linear one, because for 512 devices (old
> >     > > > MAX_FIBRE_DEVICES value) linear search is quite ineffective.
> >     I can't
> >     > > > imagine so many devices, though :)
> >     > >
> >     > > No, I removed tgt->sess because you can have up to 2k devices
> >     with the 16bit
> >     > > Loop ID's.  And the MAX_FIBRE_DEVICES is not standard across
> >     all chips.
> >     > >
> >     > > >
> >     > > > 3. Do you think I should publish this version or left in CVS?
> >     > >
> >     > > Lemme respin a release tomorrow.  This one is okay for CVS
> >     though.
> >     > >
> >     > > >
> >     > > > Vlad
> >     > > >
> >     > > > Nathaniel Clark wrote:
> >     > > > > Changes from 0.9.3.4 <http://0.9.3.4> to 0.9.3.5
> >     <http://0.9.3.5>
> >     > > > > -----------------------------------------
> >     > > > > * qLogic's 2K loop ID's are now supported fully (including
> >     reading from
> >     > > > > the IOCB's)  This will fix a problem when Initiators did
> >     Port Logouts
> >     > > > > and the target would get VERY confused, and stop
> >     functioning.  This only
> >     > > > > happened w/ multiple initiators (and possibly only in
> >     Fabric mode).
> >     > > > >
> >     > > > > * Target Mode is now enabled/disabled differently.  The
> >     chip will now
> >     > > > > not report being a target until target mode is actually
> >     enabled (it will
> >     > > > > then do a chip reset).  This will fix the extra devices
> >     report by
> >     > > > > various OS's (including Linux).
> >     > > > >
> >     > > > > * Hopefully even more stable.
> >     > > > >
> >     > > > > * Extra files in qla2xxx sysfs scsi_host directory (good
> >     for debugging).
> >     > >
> >
> >
> >
> >     -------------------------------------------------------
> >     This SF.net email is sponsored by: Splunk Inc. Do you grep through
> >     log files
> >     for problems?  Stop!  Download the new AJAX search engine that make=
s
> >     searching your log files as easy as surfing the  web.  DOWNLOAD
> >     SPLUNK!
> >
> http://sel.as-us.falkag.net/sel?cmd=3Dlnk&kid=3D103432&bid=3D230486&dat=
=3D121642
> >     <
> http://sel.as-us.falkag.net/sel?cmd=3Dlnk&kid=3D103432&bid=3D230486&dat=
=3D121642>
> >     _______________________________________________
> >     Scst-devel mailing list
> >     Scs...@li...
> >     <mailto:Scs...@li...>
> >     https://lists.sourceforge.net/lists/listinfo/scst-devel
> >     <https://lists.sourceforge.net/lists/listinfo/scst-devel>
> >
> >
>
>