qla2300 at only 1 GBit on kernel 2.6.5

qla2300 at only 1 GBit on kernel 2.6.5

[Jan-Frode Myklebust]

I gave up on the LSI Logic FC HBA, since we couldn't get it to work
with the SMP-kernel, so now I've gotten a qlogic QLA2300 as
replacement.

This seems to be working fine with the RedHat 2.4.21-9.0.1.ELsmp
kernel, but when I try running on the vanilla 2.6.5 kernel, it only
operates in 1 GB mode. The HBA is connected to a Infortrend SATA/RAID
box. If I try forcing the connection to 2 GBit from the Infortrend, I
get an error saying 'Cable unplugged' when loading the qla2300 module
on 2.6.5.

So, are there any module parameters I can give the qla2300, to force
it to 2 Gbit? Or anything else I can try to get full speed on my
channels?

Kernel logs while loading the qla2300 module on 2.4.21-9.0.1.ELsmp:

-------------------------------------------------------------------
qla2x00_set_info starts at address = f8a94060
qla2x00: Found  VID=1077 DID=2300 SSVID=1077 SSDID=9
scsi(1): Found a QLA2300  @ bus 1, device 0x5, irq 20, iobase 0xf8a8e000
scsi(1): Allocated 4096 SRB(s).
scsi(1): Configure NVRAM parameters...
scsi(1): 64 Bit PCI Addressing Enabled.
scsi(1): Scatter/Gather entries= 3584 qla2x00_nvram_config ZIO enabled:intr_timer_delay=3
scsi(1): Verifying loaded RISC code...
scsi(1): Verifying chip...
scsi(1): Waiting for LIP to complete...
scsi(1): LOOP UP detected.
scsi(1): Port database changed.
scsi(1) qla2x00_isr MBA_PORT_UPDATE ignored
scsi(1): Topology - (N_Port-to-N_Port), Host Loop address 0x1
qla2x00_find_all_fabric_devs GNN_FT Failed-Try issuing GAN
scsi1 : QLogic QLA2300 PCI to Fibre Channel Host Adapter: bus 1 device 5 irq 20
        Firmware version:  3.02.13, Driver version 6.06.00b11
 
scsi(1): Waiting for LIP to complete...
scsi(1): Topology - (N_Port-to-N_Port), Host Loop address 0x1
initializing plug timer for queue f6915e18
Starting timer : 1 1
blk: queue f6915e18, I/O limit 4294967295Mb (mask 0xffffffffffffffff)
  Vendor: IFT       Model: A16F-G1A2         Rev: 334A
  Type:   Direct-Access                      ANSI SCSI revision: 03
initializing plug timer for queue f6915c18
Starting timer : 1 1
blk: queue f6915c18, I/O limit 4294967295Mb (mask 0xffffffffffffffff)
scsi(1:0:0:0): Enabled tagged queuing, queue depth 64.
Attached scsi disk sdb at scsi1, channel 0, id 0, lun 0
SCSI device sdb: 1048576000 512-byte hdwr sectors (536871 MB)
 sdb: sdb1
-------------------------------------------------------------------

Kernel logs while loading the qla2300 module on vanila 2.6.5:

-------------------------------------------------------------------
QLogic Fibre Channel HBA Driver (f8b4c000)
qla2300 0000:01:05.0: Found an ISP2300, irq 145, iobase 0xf89b3000
qla2300 0000:01:05.0: Configuring PCI space...
qla2300 0000:01:05.0: Configure NVRAM parameters...
qla2300 0000:01:05.0: Verifying loaded RISC code...
qla2300 0000:01:05.0: Waiting for LIP to complete...
qla2300 0000:01:05.0: LOOP UP detected (1 Gbps).
qla2300 0000:01:05.0: Topology - (N_Port-to-N_Port), Host Loop address 0x1
qla2300 0000:01:05.0: Failed SNS login: loop_id=80 mb[0]=4005 mb[1]=5 mb[2]=0 mb[6]=423 mb[7]=922e
scsi1 : qla2xxx
qla2300 0000:01:05.0:
 QLogic Fibre Channel HBA Driver: 8.00.00b10
  QLogic QLA2300 -
  ISP2300: PCI (33 MHz) @ 0000:01:05.0 hdma-, host#=1, fw=3.02.21 IPX
  Vendor: IFT       Model: A16F-G1A2         Rev: 334A
  Type:   Direct-Access                      ANSI SCSI revision: 03
qla2300 0000:01:05.0: scsi(1:0:0:0): Enabled tagged queuing, queue depth 32.
SCSI device sdb: 1048576000 512-byte hdwr sectors (536871 MB)
SCSI device sdb: drive cache: write through
 sdb:<4>  Warning: Disk has a valid GPT signature but invalid PMBR.
  Assuming this disk is *not* a GPT disk anymore.
  Use gpt kernel option to override.  Use GNU Parted to correct disk.
 sdb1
Attached scsi disk sdb at scsi1, channel 0, id 0, lun 0
Attached scsi generic sg2 at scsi1, channel 0, id 0, lun 0,  type 0
  Vendor: IFT       Model: A16F-G1A2         Rev: 334A
  Type:   Direct-Access                      ANSI SCSI revision: 03
qla2300 0000:01:05.0: scsi(1:0:0:1): Enabled tagged queuing, queue
depth 32.
SCSI device sdc: 1048576000 512-byte hdwr sectors (536871 MB)
SCSI device sdc: drive cache: write through
 sdc: sdc1
Attached scsi disk sdc at scsi1, channel 0, id 0, lun 1
Attached scsi generic sg3 at scsi1, channel 0, id 0, lun 1,  type 0
  Vendor: IFT       Model: A16F-G1A2         Rev: 334A
  Type:   Direct-Access                      ANSI SCSI revision: 03
qla2300 0000:01:05.0: scsi(1:0:0:2): Enabled tagged queuing, queue
depth 32.
SCSI device sdd: 1317883904 512-byte hdwr sectors (674757 MB)
SCSI device sdd: drive cache: write through
 sdd: sdd1
Attached scsi disk sdd at scsi1, channel 0, id 0, lun 2
Attached scsi generic sg4 at scsi1, channel 0, id 0, lun 2,  type 0
-------------------------------------------------------------------

Any advice?


  -jf

RE: qla2300 at only 1 GBit on kernel 2.6.5

[Andrew Vasquez]

On , linux-scsi-owner@vger.kernel.org wrote:

> This seems to be working fine with the RedHat 2.4.21-9.0.1.ELsmp
> kernel,
>

How are you verifying that the 2.4 driver is coming-up in 2Gig rather
than 1gig?  Do you have an analyzer running between the HBA and
storage device?  Or, as you mentioned later in the email when testing
the 8.x driver, did you force 2gig at the RAID box and reload the
2.4 driver?

> but when I try running on the vanilla 2.6.5 kernel, it only
> operates in 1 GB mode.
>

Yes, the 8.x series driver started to display the connection speed
during a loop_up event.

	qla2300 0000:01:05.0: LOOP UP detected (1 Gbps).

> The HBA is connected to a Infortrend
> SATA/RAID box.
>

Which model?  Actually, looking ahead, I can see you are running an

	Vendor: IFT       Model: A16F-G1A2         Rev: 334A

Of the three FC-SATA RAID boxes that are advertised, only the
A16F-J1210-G1 model mentions support for 'full-duplex 2Gb FC-AL'.

> If I try forcing the connection to 2 GBit from the
> Infortrend, I get an error saying 'Cable unplugged' when loading the
> qla2300 module on 2.6.5. 
> 

Hmm, could you go into the BIOS utility (ctrl-q during boot) and check
the 'Data Rate' settings for the HBA?  What is the value set to --
auto/1gb/2gb?  If it is set to auto, could you set it to 2gb and retry
the test.

Regards,
Andrew Vasquez
QLogic Corporation

Re: qla2300 at only 1 GBit on kernel 2.6.5

[Jan-Frode Myklebust]

On Wed, May 05, 2004 at 09:27:10AM -0700, Andrew Vasquez wrote:
> 
> How are you verifying that the 2.4 driver is coming-up in 2Gig rather
> than 1gig?  Do you have an analyzer running between the HBA and
> storage device?  

The storage device has a gui telling me current speed. CurSynClk=2 GHz
for the redhat kernel, 1 GHz for the 2.6.5 and 2.6.6-rc3 kernels.

> Or, as you mentioned later in the email when testing
> the 8.x driver, did you force 2gig at the RAID box and reload the
> 2.4 driver?

Initially I was running kernel 2.6.5  had the connection forced at
2Gbit from the storage device side, then I struggled for a while
trying to understand why I didn't get link.  The qla2300 module said
"Cable unplugged". 

Then I set the storage device to autodetect speed, rebooted, and then
it came up at 1 Gbit.

Then (with the setting still at autodetect) I rebooted to the RedHat
kernel, and it got autodetected to 2 Gbit according to the storage
device.

> Which model?  Actually, looking ahead, I can see you are running an
> 
> 	Vendor: IFT       Model: A16F-G1A2         Rev: 334A
> 
> Of the three FC-SATA RAID boxes that are advertised, only the
> A16F-J1210-G1 model mentions support for 'full-duplex 2Gb FC-AL'.

The data-sheet says "Two 2Gbps Fibre Host Channels; Transfer rate up to
200MB/sec for each":

	http://www.infortrend.com/document/pdf/DS_ESA16FU_G_0311.pdf

> 
> Hmm, could you go into the BIOS utility (ctrl-q during boot) and check
> the 'Data Rate' settings for the HBA?  What is the value set to --
> auto/1gb/2gb?  If it is set to auto, could you set it to 2gb and retry
> the test.

During boot it says it has no 'ROM BIOS'. Is this something I can
install?


BTW: here's the log I get with 2.6.6-rc3 when loading the qla2300 with
speed=2GHz set on the infortrend:

QLogic Fibre Channel HBA Driver (f8aec000)
qla2300 0000:01:05.0: Found an ISP2300, irq 145, iobase 0xf89b4000
qla2300 0000:01:05.0: Configuring PCI space...
qla2300 0000:01:05.0: Configure NVRAM parameters...
qla2300 0000:01:05.0: Verifying loaded RISC code...
qla2300 0000:01:05.0: Waiting for LIP to complete...
qla2300 0000:01:05.0: Cable is unplugged...
scsi2 : qla2xxx
qla2300 0000:01:05.0:
 QLogic Fibre Channel HBA Driver: 8.00.00b11-k
  QLogic QLA2300 -
  ISP2300: PCI (33 MHz) @ 0000:01:05.0 hdma-, host#=2, fw=3.02.26 IPX


  -jf

Re: qla2300 at only 1 GBit on kernel 2.6.5

[Håvard Lygre]

Jan-Frode Myklebust <janfrode@parallab.uib.no> writes:

> During boot it says it has no 'ROM BIOS'. Is this something I can
> install?

It probably says something about the BIOS not being enabled - you only
need the BIOS if you're going to boot from the FC-device.


> BTW: here's the log I get with 2.6.6-rc3 when loading the qla2300 with
> speed=2GHz set on the infortrend:
>
> QLogic Fibre Channel HBA Driver (f8aec000)
> qla2300 0000:01:05.0: Found an ISP2300, irq 145, iobase 0xf89b4000
> qla2300 0000:01:05.0: Configuring PCI space...
> qla2300 0000:01:05.0: Configure NVRAM parameters...
> qla2300 0000:01:05.0: Verifying loaded RISC code...
> qla2300 0000:01:05.0: Waiting for LIP to complete...
> qla2300 0000:01:05.0: Cable is unplugged...
> scsi2 : qla2xxx
> qla2300 0000:01:05.0:
>  QLogic Fibre Channel HBA Driver: 8.00.00b11-k
>   QLogic QLA2300 -
>   ISP2300: PCI (33 MHz) @ 0000:01:05.0 hdma-, host#=2, fw=3.02.26 IPX


Just as a datapoint: Vanilla Linux 2.6.6 gives me 2 Gbps with both
sides set to auto.  This is with a QLogic 2312 (as sold by IBM). I
haven't tried an earlier 2.6-series kernel on this computer.

On a side note: Your driver reports 33 MHz PCI - if this is in a
32-bit slot, isn't 2Gbps more than the bus can handle?





Also notice the discrepancy in the sizes reported for the same device:


2.6.6:
SCSI device sdc: 4294434816 512-byte hdwr sectors (1099239 MB)

and 2.4.26: 
SCSI device sdc: 4294434816 512-byte hdwr sectors (2198751 MB)

- Obviously just a simple miscalculation in sd.c



--- 2.6.6 ---

QLogic Fibre Channel HBA Driver (f895c000)
qla2300 0000:02:06.0: Found an ISP2312, irq 24, iobase 0xf8918000
qla2300 0000:02:06.0: Configuring PCI space...
qla2300 0000:02:06.0: Configure NVRAM parameters...
qla2300 0000:02:06.0: Verifying loaded RISC code...
qla2300 0000:02:06.0: LIP reset occured (f7f7).
qla2300 0000:02:06.0: Waiting for LIP to complete...
qla2300 0000:02:06.0: LIP occured (f7f7).
qla2300 0000:02:06.0: LOOP UP detected (2 Gbps).
qla2300 0000:02:06.0: Topology - (Loop), Host Loop address 0x7d
scsi2 : qla2xxx
qla2300 0000:02:06.0:
 QLogic Fibre Channel HBA Driver: 8.00.00b11-k
  QLogic QLA2340 - 133MHz PCI-X to 2Gb FC, Single Channel
  ISP2312: PCI-X (133 MHz) @ 0000:02:06.0 hdma-, host#=2, fw=3.02.26 IPX
  Vendor: nStor     Model: NexStorWahooSATA  Rev:
  Type:   Direct-Access                      ANSI SCSI revision: 03
qla2300 0000:02:06.0: scsi(2:0:0:0): Enabled tagged queuing, queue depth 32.
SCSI device sdc: 4294434816 512-byte hdwr sectors (1099239 MB)
SCSI device sdc: drive cache: write back
 sdc: sdc1
Attached scsi disk sdc at scsi2, channel 0, id 0, lun 0
  Vendor: nStor     Model: NexStorWahooSATA  Rev:
  Type:   Direct-Access                      ANSI SCSI revision: 03
qla2300 0000:02:06.0: scsi(2:0:0:1): Enabled tagged queuing, queue depth 32.
SCSI device sdd: 605675520 512-byte hdwr sectors (310106 MB)
SCSI device sdd: drive cache: write back
 sdd: sdd1
Attached scsi disk sdd at scsi2, channel 0, id 0, lun 1
  Vendor: nStor     Model: NexStorWahooSATA  Rev:
  Type:   Processor                          ANSI SCSI revision: 03
qla2300 0000:02:06.0: scsi(2:0:0:2): Enabled tagged queuing, queue depth



For reference, here's from 2.4.26 with the qlogic-driver (6.06.10, as
released in debian's qla2x00-source package)

This is when I've just created the sdd LUN, so it hasn't been
partitioned yet.


qla2x00_set_info starts at address = f8973060
qla2x00: Found  VID=1077 DID=2312 SSVID=1077 SSDID=100
scsi(2): Found a QLA2312  @ bus 2, device 0x6, irq 24, iobase 0x0000dc00
scsi(2): Allocated 4096 SRB(s).
scsi(2): Configure NVRAM parameters...
scsi(2): 32 Bit PCI Addressing Enabled.
scsi(2): Scatter/Gather entries= 32
scsi(2): Verifying loaded RISC code...
scsi(2): Verifying chip...
scsi(2): Waiting for LIP to complete...
scsi(2): LIP reset occurred.
scsi(2): LIP occurred.
scsi(2): LOOP UP detected.
scsi(2): Port database changed.
scsi(2): Topology - (Loop), Host Loop address 0x7d
scsi-qla0-adapter-node=200000e08b12bcb6\;
scsi-qla0-adapter-port=210000e08b12bcb6\;
scsi-qla0-tgt-0-di-0-port=2100000bb52088eb\;
scsi2 : QLogic QLA2312 PCI to Fibre Channel Host Adapter: bus 2 device 6 irq 24
        Firmware version:  3.02.16, Driver version 6.06.10
May 13 17:30:46 guinness kernel:
  Vendor: nStor     Model: NexStorWahooSATA  Rev:
  Type:   Direct-Access                      ANSI SCSI revision: 03
  Vendor: nStor     Model: NexStorWahooSATA  Rev:
  Type:   Direct-Access                      ANSI SCSI revision: 03
  Vendor: nStor     Model: NexStorWahooSATA  Rev:
  Type:   Processor                          ANSI SCSI revision: 03
scsi(2:0:0:0): Enabled tagged queuing, queue depth 32.
scsi(2:0:0:1): Enabled tagged queuing, queue depth 32.
scsi(2:0:0:2): Enabled tagged queuing, queue depth 32.
Attached scsi disk sdc at scsi2, channel 0, id 0, lun 0
Attached scsi disk sdd at scsi2, channel 0, id 0, lun 1
Attached scsi generic sg5 at scsi2, channel 0, id 0, lun 2,  type 3
SCSI device sdc: 4294434816 512-byte hdwr sectors (2198751 MB)
 sdc: sdc1
SCSI device sdd: 605675520 512-byte hdwr sectors (310106 MB)
 sdd: unknown partition table




-- 
Håvard Lygre, hklygre@online.no
BLUG: http://blug.linux.no/   RFC1149: http://blug.linux.no/rfc1149/

Re: qla2300 at only 1 GBit on kernel 2.6.5

[Jan-Frode Myklebust]

On Sat, May 22, 2004 at 03:43:28AM +0200, Håvard Lygre wrote:
> 
> Just as a datapoint: Vanilla Linux 2.6.6 gives me 2 Gbps with both
> sides set to auto.  This is with a QLogic 2312 (as sold by IBM). I
> haven't tried an earlier 2.6-series kernel on this computer.

Andrew Vasquez <andrew.vasquez@qlogic.com> found the problem. The 2.4
driver was forcing the data rate to auto negotiate, while the 2.6 (or
8.x series) driver was honouring the NVRAM setting which was at the
default 1 Gbps. Changing to auto negotiate in the qlogic BIOS fixed the
problem.

> 
> On a side note: Your driver reports 33 MHz PCI - if this is in a
> 32-bit slot, isn't 2Gbps more than the bus can handle?

It's a 64 bit slot..


   -jf