Adaptec aic79xx performance

Adaptec aic79xx performance

[Denny Page]

[Ok, once more with the correct subject :-]

Hello,

I have a question regarding the support condition of the aic79xx driver.

I am running kernel version 2.6.14 (gentoo) on a dual processor system
PCI-X system.  I am using an Adaptec scsi adapter to run an HP Ultrium
tape drive.  Nothing else is attached to the controller.

I was previously using an Adaptec 29160N (aic7892), which is a 32 bit,
66Mhz controller.  With this controller, the tape was not able to
maintain full streaming, or achieve full speed.  Average rate was
approximately 27MB per second.  In order to address the performance
problem, HP recommended that I change to a 64 bit U160 or U320 controller.

I've put in an Adaptec 29320ALP-R (aic7901), which is a 32 bit, 133Mhz
controller.  With this controller, performance has dropped to an abysmal
2.7MB per second!

I am running the controller at 133Mhz, but I have also tried locking it
to 66Mhz with no effect.  I am using the same cable with both
controllers, and the terminator is built into the cable.  Nothing stands
out in /var/log/messages.

So, can anyone offer ideas as to why the 79xx controller performs at one
tenth the rate of the 78xx controller and how to fix it?  Any help
appreciated.

Thanks,

Denny

Re: Adaptec aic79xx performance

[Denny Page]

Further info, indicating a negotiation problem:

/proc/scsi/aic7xxx:
Target 3 Negotiation Settings
     User: 160.000MB/s transfers (80.000MHz DT, offset 127, 16bit)
     Goal: 160.000MB/s transfers (80.000MHz DT, offset 127, 16bit)
     Curr: 160.000MB/s transfers (80.000MHz DT, offset 127, 16bit)
     Channel A Target 3 Lun 0 Settings
             Commands Queued 24
             Commands Active 0
             Command Openings 1
             Max Tagged Openings 0
             Device Queue Frozen Count 0

/proc/scsi/aic79xx:
Target 3 Negotiation Settings
     User: 320.000MB/s transfers (160.000MHz RDSTRM|DT|IU|RTI|QAS, 16bit)
     Goal: 3.300MB/s transfers
     Curr: 3.300MB/s transfers
     Channel A Target 3 Lun 0 Settings
             Commands Queued 19
             Commands Active 0
             Command Openings 1
             Max Tagged Openings 0
             Device Queue Frozen Count 0

Kinda explains the abysmal data rate.  I am guessing that this is a bug. 
  Anyone looked at this?

Thanks,

Denny

Re: Adaptec aic79xx performance

[Denny Page]

Boot verbose output for the aic79xx controller follows.

Denny

-----

Dec  3 02:46:07 marvin ahd_pci:6:1:0: Reading VPD from 
SEEPROM...ahd_pci:6:1:0: VPD parsing successful
Dec  3 02:46:07 marvin ahd_pci:6:1:0: Reading SEEPROM...done.
Dec  3 02:46:07 marvin ahd_pci:6:1:0: STPWLEVEL is on
Dec  3 02:46:07 marvin ahd_pci:6:1:0: Manual Primary Termination
Dec  3 02:46:07 marvin ahd_pci:6:1:0: Manual Secondary Termination
Dec  3 02:46:07 marvin ahd_pci:6:1:0: Primary High byte termination Disabled
Dec  3 02:46:07 marvin ahd_pci:6:1:0: Primary Low byte termination Enabled
Dec  3 02:46:07 marvin ahd_pci:6:1:0: Secondary High byte termination 
Disabled
Dec  3 02:46:07 marvin ahd_pci:6:1:0: Secondary Low byte termination 
Disabled
Dec  3 02:46:07 marvin ahd_pci:6:1:0: Downloading Sequencer Program... 
650 instructions downloaded
Dec  3 02:46:07 marvin ahd_pci:6:1:0: Features 0x3c001, Bugs 0x500000, 
Flags 0x3d1
Dec  3 02:46:07 marvin scsi0 : Adaptec AIC79XX PCI-X SCSI HBA DRIVER, 
Rev 1.3.11
Dec  3 02:46:07 marvin <Adaptec 29320ALP Ultra320 SCSI adapter>
Dec  3 02:46:07 marvin aic7901: Ultra320 Wide Channel A, SCSI Id=7, 
PCI-X 101-133Mhz, 512 SCBs
Dec  3 02:46:07 marvin
Dec  3 02:46:07 marvin scsi0: Slave Alloc 0
Dec  3 02:46:07 marvin scsi0: Slave Destroy 0
Dec  3 02:46:07 marvin scsi0: Slave Alloc 1
Dec  3 02:46:07 marvin scsi0: Slave Destroy 1
Dec  3 02:46:07 marvin scsi0: Slave Alloc 2
Dec  3 02:46:07 marvin scsi0: Slave Destroy 2
Dec  3 02:46:07 marvin scsi0: Slave Alloc 3
Dec  3 02:46:07 marvin Vendor: HP        Model: Ultrium 2-SCSI    Rev: F5CD
Dec  3 02:46:07 marvin Type:   Sequential-Access                  ANSI 
SCSI revision: 03
Dec  3 02:46:07 marvin scsi0: Slave Configure 3
Dec  3 02:46:07 marvin target0:0:3: asynchronous.
Dec  3 02:46:07 marvin target0:0:3: Beginning Domain Validation
Dec  3 02:46:07 marvin (scsi0:A:3:0): Sending WDTR 1
Dec  3 02:46:07 marvin (scsi0:A:3:0): Received WDTR 1 filtered to 1
Dec  3 02:46:07 marvin target0:0:3: FAST-5 WIDE SCSI 2.0 MB/s ST (1020 
ns, offset 255)
Dec  3 02:46:07 marvin scsi0: target 3 using 16bit transfers
Dec  3 02:46:07 marvin (scsi0:A:3:0): Sending SDTR period 44, offset 0
Dec  3 02:46:07 marvin (scsi0:A:3:0): Received SDTR period 44, offset 0
Dec  3 02:46:07 marvin 	Filtered to period 0, offset 0
Dec  3 02:46:07 marvin target0:0:3: wide asynchronous.
Dec  3 02:46:07 marvin scsi0: target 3 using asynchronous transfers
Dec  3 02:46:07 marvin (scsi0:A:3:0): Sending WDTR 1
Dec  3 02:46:07 marvin (scsi0:A:3:0): Received WDTR 1 filtered to 1
Dec  3 02:46:07 marvin target0:0:3: FAST-5 WIDE SCSI 2.0 MB/s ST (1020 
ns, offset 255)
Dec  3 02:46:07 marvin scsi0: target 3 using 16bit transfers
Dec  3 02:46:07 marvin (scsi0:A:3:0): Sending SDTR period 44, offset 0
Dec  3 02:46:07 marvin (scsi0:A:3:0): Received SDTR period 44, offset 0
Dec  3 02:46:07 marvin 	Filtered to period 0, offset 0
Dec  3 02:46:07 marvin target0:0:3: wide asynchronous.
Dec  3 02:46:07 marvin scsi0: target 3 using asynchronous transfers
Dec  3 02:46:07 marvin (scsi0:A:3:0): Sending WDTR 1
Dec  3 02:46:07 marvin (scsi0:A:3:0): Received WDTR 1 filtered to 1
Dec  3 02:46:07 marvin target0:0:3: FAST-5 WIDE SCSI 2.0 MB/s ST (1020 
ns, offset 255)
Dec  3 02:46:07 marvin scsi0: target 3 using 16bit transfers
Dec  3 02:46:07 marvin (scsi0:A:3:0): Sending SDTR period 44, offset 0
Dec  3 02:46:07 marvin (scsi0:A:3:0): Received SDTR period 44, offset 0
Dec  3 02:46:07 marvin 	Filtered to period 0, offset 0
Dec  3 02:46:07 marvin target0:0:3: wide asynchronous.
Dec  3 02:46:07 marvin scsi0: target 3 using asynchronous transfers
Dec  3 02:46:07 marvin (scsi0:A:3:0): Sending WDTR 1
Dec  3 02:46:07 marvin (scsi0:A:3:0): Received WDTR 1 filtered to 1
Dec  3 02:46:07 marvin target0:0:3: FAST-5 WIDE SCSI 2.0 MB/s ST (1020 
ns, offset 255)
Dec  3 02:46:07 marvin scsi0: target 3 using 16bit transfers
Dec  3 02:46:07 marvin (scsi0:A:3:0): Sending SDTR period 44, offset 0
Dec  3 02:46:07 marvin (scsi0:A:3:0): Received SDTR period 44, offset 0
Dec  3 02:46:07 marvin 	Filtered to period 0, offset 0
Dec  3 02:46:07 marvin target0:0:3: wide asynchronous.
Dec  3 02:46:07 marvin scsi0: target 3 using asynchronous transfers
Dec  3 02:46:07 marvin (scsi0:A:3:0): Sending WDTR 1
Dec  3 02:46:07 marvin (scsi0:A:3:0): Received WDTR 1 filtered to 1
Dec  3 02:46:07 marvin target0:0:3: FAST-5 WIDE SCSI 2.0 MB/s ST (1020 
ns, offset 255)
Dec  3 02:46:07 marvin scsi0: target 3 using 16bit transfers
Dec  3 02:46:07 marvin (scsi0:A:3:0): Sending SDTR period 44, offset 0
Dec  3 02:46:07 marvin (scsi0:A:3:0): Received SDTR period 44, offset 0
Dec  3 02:46:07 marvin 	Filtered to period 0, offset 0
Dec  3 02:46:07 marvin target0:0:3: wide asynchronous.
Dec  3 02:46:07 marvin scsi0: target 3 using asynchronous transfers
Dec  3 02:46:07 marvin target0:0:3: Domain Validation skipping write tests
Dec  3 02:46:07 marvin (scsi0:A:3:0): Sending PPR bus_width 1, period 8, 
offset fe, ppr_options f3
Dec  3 02:46:07 marvin (scsi0:A:3:0): Received PPR width 1, period 9, 
offset fe,options 2
Dec  3 02:46:07 marvin 	Filtered to width 0, period 0, offset 0, options 0
Dec  3 02:46:07 marvin target0:0:3: asynchronous.
Dec  3 02:46:07 marvin scsi0: target 3 using 8bit transfers
Dec  3 02:46:07 marvin target0:0:3: Ending Domain Validation
Dec  3 02:46:07 marvin scsi0: Slave Alloc 4
Dec  3 02:46:07 marvin scsi0: Slave Destroy 4
Dec  3 02:46:07 marvin scsi0: Slave Alloc 5
Dec  3 02:46:07 marvin scsi0: Slave Destroy 5
Dec  3 02:46:07 marvin scsi0: Slave Alloc 6
Dec  3 02:46:07 marvin scsi0: Slave Destroy 6
Dec  3 02:46:07 marvin scsi0: Slave Alloc 8
Dec  3 02:46:07 marvin scsi0: Slave Destroy 8
Dec  3 02:46:07 marvin scsi0: Slave Alloc 9
Dec  3 02:46:07 marvin scsi0: Slave Destroy 9
Dec  3 02:46:07 marvin scsi0: Slave Alloc 10
Dec  3 02:46:07 marvin scsi0: Slave Destroy 10
Dec  3 02:46:07 marvin scsi0: Slave Alloc 11
Dec  3 02:46:07 marvin scsi0: Slave Destroy 11
Dec  3 02:46:07 marvin scsi0: Slave Alloc 12
Dec  3 02:46:07 marvin scsi0: Slave Destroy 12
Dec  3 02:46:07 marvin scsi0: Slave Alloc 13
Dec  3 02:46:07 marvin scsi0: Slave Destroy 13
Dec  3 02:46:07 marvin scsi0: Slave Alloc 14
Dec  3 02:46:07 marvin scsi0: Slave Destroy 14
Dec  3 02:46:07 marvin scsi0: Slave Alloc 15
Dec  3 02:46:07 marvin scsi0: Slave Destroy 15

Re: Adaptec aic79xx performance

[Denny Page]

Further research shows that this defect is new to kernel version 2.6.14. 
  I've tested 2.6.13 and a couple of previous kernels, and they 
correctly negotiate U160.

With 2.6.13, I am now receiving the expected 60+ MB per second with 
uncompressed data.

The defect appears to still be present in 2.6.15-rc4.

Hints and suggestions welcomed.

Thanks,

Denny

Re: Adaptec aic79xx performance

[Denny Page]

Is there anyone that would offer any suggestions on progressing the 
diagnosis?

Thanks,

Denny


Denny Page wrote:

> Further research shows that this defect is new to kernel version 
> 2.6.14.  I've tested 2.6.13 and a couple of previous kernels, and they 
> correctly negotiate U160.
>
> With 2.6.13, I am now receiving the expected 60+ MB per second with 
> uncompressed data.
>
> The defect appears to still be present in 2.6.15-rc4.
>
> Hints and suggestions welcomed.
>
> Thanks,
>
> Denny
> -
> To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>