No SCSI Ultra 160 with Adaptec Controller

No SCSI Ultra 160 with Adaptec Controller

[Tom Sightler]

Hi All,

While trying to determine why my SCSI Ultra 160 drives don't work on my Dell 
PowerApp.Web 100 I noticed this section of code:

      /*
       * This is needed to work around a sequencer bug for now.  Regardless
       * of the controller in use, if we have a Quantum drive, we need to
       * limit the speed to 80MByte/sec.  As soon as I get a fixed version
       * of the sequencer, this code will get yanked.
       */
      if(!strncmp(buffer + 8, "QUANTUM", 7) &&
         p->transinfo[tindex].goal_options )

Since this machine has Quantum drives I guess this is my problem.  Does anyone 
know if this code is still actually neccessary?  It seems it's been there a 
while.  It's dissapointing to not get full performance out of the hardware you 
have.

Thanks,
Tom
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: No SCSI Ultra 160 with Adaptec Controller

[Matt Domsch]

On Wed, 24 Jan 2001, Tom Sightler wrote:

> I temporarily disabled that code and the
> increase in IO's per second is measurable, though not earth shattering, but
> I was afraid to leave it that way because fast corrupted data is worth much
> less that only slightly slower good data.

I don't believe the problem is data corruption, but that there could be
some CRC data residual from an I/O which causes the driver to issue a
SCSI bus reset.  As bus resets really kill performance, Doug thought it
better to slow the drive to 80 rather than run at 160 and have occasional
bus resets.

-- 
Matt Domsch
Dell Linux Systems Group
Linux OS Development
www.dell.com/linux


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

RE: No SCSI Ultra 160 with Adaptec Controller

[Nathan Black]

What problem are you trying to solve, I just joined the group in hopes of
finding a solution to a problem of mine.

It sounds like they may be related.
I am trying to write to a drive(raw, blank, whatever you want to call it) on
kernel 2.4.1-pre10. I get great throughput at first, but then it slows to
about 25 MB/sec. This isn't a problem except after writing a large amount of
data, the system becomes unusable. I did not say it halts, oops, or anything
like that, just extremely slow. ps takes about 10 minutes or more to
complete.(4 tasks).

I have a compaq smart array controller 2. 

Before I upgraded the kernel( I was at 2.4.0 released). I would get an oops.
and a kernel panic/halt.
I did see that running in uniprocessor mode with NMIs turned off would let
me write without problems, but I was never sure how to fix it.

Any ideas?

-----Original Message-----
From: Tom Sightler [mailto:ttsig@tuxyturvy.com]
Sent: Wednesday, January 24, 2001 3:08 PM
To: mjacob@feral.com; Matt_Domsch@dell.com
Cc: linux-kernel@vger.kernel.org
Subject: Re: No SCSI Ultra 160 with Adaptec Controller


> Actually, aren't a number of newer drives getting upwards of 30MB/s?
>

Well, at 80MB/sec these drives are able to come in at an average of about
34MB/s across the board.

> > Therefore, you only exceed the 80MB/sec bus speed if you
> > have more than 4 disks all doing maximum I/O at the same time.  Since
the
> > PowerApp.web 100 has at most 2 disks internally, you really shouldn't
see
> > any significant performance difference.

I wouldn't dare argue that I'd get some huge performance boost, but I am
running software raid1 on multiple spindles (we have some drives attached
externally) so it should help some.
Also, this unit is doing heavy read/write of small files and the lower
latency, and
burstable transfers do help some.  I temporarily disabled that code and the
increase in IO's per second is measurable, though not earth shattering, but
I
was afraid to leave it that way because fast corrupted data is worth much
less
that only slightly slower good data.

It was really just an issue of if the code still needed to be there.
Workarounds that are put in as temporary have a tendency to never get
revisited
until someone brings them up, and on top of that they make Linux look sloppy
to
me (the same drives work correctly at Ultra160 under that other OS, or at
least appear to).  I know it
won't make a huge difference to me, but if some user has a 12 disk RAID
array
full of Quantum disks it will make a big difference to them.

Later,
Tom



-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: No SCSI Ultra 160 with Adaptec Controller

[Tom Sightler]

> Actually, aren't a number of newer drives getting upwards of 30MB/s?
>

Well, at 80MB/sec these drives are able to come in at an average of about
34MB/s across the board.

> > Therefore, you only exceed the 80MB/sec bus speed if you
> > have more than 4 disks all doing maximum I/O at the same time.  Since
the
> > PowerApp.web 100 has at most 2 disks internally, you really shouldn't
see
> > any significant performance difference.

I wouldn't dare argue that I'd get some huge performance boost, but I am
running software raid1 on multiple spindles (we have some drives attached
externally) so it should help some.
Also, this unit is doing heavy read/write of small files and the lower
latency, and
burstable transfers do help some.  I temporarily disabled that code and the
increase in IO's per second is measurable, though not earth shattering, but
I
was afraid to leave it that way because fast corrupted data is worth much
less
that only slightly slower good data.

It was really just an issue of if the code still needed to be there.
Workarounds that are put in as temporary have a tendency to never get
revisited
until someone brings them up, and on top of that they make Linux look sloppy
to
me (the same drives work correctly at Ultra160 under that other OS, or at
least appear to).  I know it
won't make a huge difference to me, but if some user has a 12 disk RAID
array
full of Quantum disks it will make a big difference to them.

Later,
Tom



-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

RE: No SCSI Ultra 160 with Adaptec Controller

[Matthew Jacob]

Well, this is all what comes from not spending money on this stuff myself- I
guess I just have too many 3 year old drives...


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

RE: No SCSI Ultra 160 with Adaptec Controller

[John Jasen]

On Tue, 23 Jan 2001, Matthew Jacob wrote:

> Actually, aren't a number of newer drives getting upwards of 30MB/s?

It depends .... tests I've done here, with scsi/160 and FC on seagate
drives, the read/write speeds start at ~35MB/s, and peter off to ~22MB/s.

I admit my methodology was crude, but I was more curious than scientificly
precise.

--
-- John E. Jasen (jjasen1@umbc.edu)
-- In theory, theory and practise are the same. In practise, they aren't.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: No SCSI Ultra 160 with Adaptec Controller

[Matt Domsch]

>      /dev/sda:
>       Timing buffered disk reads:  64 MB in  1.59 seconds =40.25 MB/sec
>
> They can do more like 40MB/s, so only two disks could saturate the 80MB/s.

Apparently I was misinformed as to the speed of these disks.  My
apologies for the confusion this caused.
-Matt


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: No SCSI Ultra 160 with Adaptec Controller

[Tim Sullivan]

Matt_Domsch@Dell.com wrote:
> 
> > Justin's 6.0.9beta(latest release) hasn't corrected the problem yet.
> >
> > scsi0 : Adaptec AIC7XXX EISA/VLB/PCI SCSI HBA DRIVER, Rev 6.0.9 BETA
> >         <Adaptec 29160 Ultra160 SCSI adapter>
> >         aic7892: Wide Channel A, SCSI Id=7, 32/255 SCBs
> >
> >   Vendor: QUANTUM   Model: ATLAS10K2-TY367L  Rev: DDD6
> >   Type:   Direct-Access                      ANSI SCSI revision: 03
> > (scsi0:A:0): async, 8bit
> > scsi0:0:0:0: Tagged Queuing enabled.  Depth 8
> > Detected scsi disk sda at scsi0, channel 0, id 0, lun 0
> > (scsi0:A:0): async, 16bit
> > (scsi0:A:0): synchronous at 80.0MHz DT, offset 0x7f, 16bit
> > SCSI device sda: 71721820 512-byte hdwr sectors (36722 MB)
> 
> I could still be wrong, but 80 MHz * 16 bits (2 bytes) = 160 MB/sec.
> We are testing Justin's driver internally.  I'll grab our SCSI disk guys and
> make sure it's up to speed.
> 
> Thanks,
> Matt

My apologies, you're right of course :-)  I was seeing MHz and thinking
MB!

Thanks,
tim
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: No SCSI Ultra 160 with Adaptec Controller

[Martin Josefsson]

On Tue, 23 Jan 2001, Tim Sullivan wrote:

> Matt_Domsch@Dell.com wrote:
> > 
> > Yes, that code is still necessary.  There's a new aic7xxx driver by Justin
> > Gibbs at Adaptec which is now being beta tested which corrects this issue.
> 
> Justin's 6.0.9beta(latest release) hasn't corrected the problem yet.
> 
> scsi0 : Adaptec AIC7XXX EISA/VLB/PCI SCSI HBA DRIVER, Rev 6.0.9 BETA
>         <Adaptec 29160 Ultra160 SCSI adapter>
>         aic7892: Wide Channel A, SCSI Id=7, 32/255 SCBs
> 
>   Vendor: QUANTUM   Model: ATLAS10K2-TY367L  Rev: DDD6
>   Type:   Direct-Access                      ANSI SCSI revision: 03
> (scsi0:A:0): async, 8bit
> scsi0:0:0:0: Tagged Queuing enabled.  Depth 8
> Detected scsi disk sda at scsi0, channel 0, id 0, lun 0
> (scsi0:A:0): async, 16bit
> (scsi0:A:0): synchronous at 80.0MHz DT, offset 0x7f, 16bit
> SCSI device sda: 71721820 512-byte hdwr sectors (36722 MB)

I just had to jump into this thread and say this:

The way I see it:
	80MHz * 16bit = 160MB/s

This is output from kernel 2.4.1-pre10 with Justin's 6.0.9beta driver:

scsi0 : Adaptec AIC7XXX EISA/VLB/PCI SCSI HBA DRIVER, Rev 6.0.9 BETA
        <Adaptec 29160 Ultra160 SCSI adapter>
        aic7892: Wide Channel A, SCSI Id=7, 32/255 SCBs
 
  Vendor: IBM       Model: DDYS-T18350N      Rev: S80D
  Type:   Direct-Access                      ANSI SCSI revision: 03
(scsi0:A:6): async, 8bit
scsi0:0:6:0: Tagged Queuing enabled.  Depth 32
Detected scsi disk sda at scsi0, channel 0, id 6, lun 0
async, 16bit
synchronous at 80.0MHz DT, offset 0x3f, 16bit
SCSI device sda: 35843670 512-byte hdwr sectors (18352 MB)

/Martin

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: No SCSI Ultra 160 with Adaptec Controller

[I Lee Hetherington]

Matt_Domsch@Dell.com wrote:

> Something to note, however: the media transfer rate for those disks is at
> most ~20MB/sec.

Hmm...

     scsi0 : Adaptec AHA274x/284x/294x (EISA/VLB/PCI-Fast SCSI)
     5.1.31/3.2.4
            <Adaptec AIC-7899 Ultra 160/m SCSI host adapter>
     Vendor: QUANTUM   Model: ATLAS10K2-TY092L  Rev: DA40
       Type:   Direct-Access                      ANSI SCSI revision: 03

     /dev/sda:
      Timing buffered disk reads:  64 MB in  1.59 seconds =40.25 MB/sec

They can do more like 40MB/s, so only two disks could saturate the 80MB/s.

This is in a Dell Precision 620.

--Lee Hethernigton


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: No SCSI Ultra 160 with Adaptec Controller

[Tim Sullivan]

Matt_Domsch@Dell.com wrote:
> 
> Yes, that code is still necessary.  There's a new aic7xxx driver by Justin
> Gibbs at Adaptec which is now being beta tested which corrects this issue.

Justin's 6.0.9beta(latest release) hasn't corrected the problem yet.

scsi0 : Adaptec AIC7XXX EISA/VLB/PCI SCSI HBA DRIVER, Rev 6.0.9 BETA
        <Adaptec 29160 Ultra160 SCSI adapter>
        aic7892: Wide Channel A, SCSI Id=7, 32/255 SCBs

  Vendor: QUANTUM   Model: ATLAS10K2-TY367L  Rev: DDD6
  Type:   Direct-Access                      ANSI SCSI revision: 03
(scsi0:A:0): async, 8bit
scsi0:0:0:0: Tagged Queuing enabled.  Depth 8
Detected scsi disk sda at scsi0, channel 0, id 0, lun 0
(scsi0:A:0): async, 16bit
(scsi0:A:0): synchronous at 80.0MHz DT, offset 0x7f, 16bit
SCSI device sda: 71721820 512-byte hdwr sectors (36722 MB)

-tim
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

RE: No SCSI Ultra 160 with Adaptec Controller

[Matthew Jacob]

On Tue, 23 Jan 2001 Matt_Domsch@Dell.com wrote:

> Hi Tom.  Thanks for writing.
> 
> > Since this machine has Quantum drives I guess this is my 
> > problem.  Does anyone 
> > know if this code is still actually necessary?  It seems 
> > it's been there a 
> > while.  It's disappointing to not get full performance out of 
> > the hardware you 
> > have.
> 
> Yes, that code is still necessary.  There's a new aic7xxx driver by Justin
> Gibbs at Adaptec which is now being beta tested which corrects this issue.
> Something to note, however: the media transfer rate for those disks is at
> most ~20MB/sec. 


Actually, aren't a number of newer drives getting upwards of 30MB/s?

> Therefore, you only exceed the 80MB/sec bus speed if you
> have more than 4 disks all doing maximum I/O at the same time.  Since the
> PowerApp.web 100 has at most 2 disks internally, you really shouldn't see
> any significant performance difference.
> 
> Hope this helps.
> Thanks for buying Dell!
> Matt Domsch
> Dell Linux Systems Group
> Linux OS Development
> 
> 
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> Please read the FAQ at http://www.tux.org/lkml/
> 

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

RE: No SCSI Ultra 160 with Adaptec Controller

[Matt_Domsch]

Hi Tom.  Thanks for writing.

> Since this machine has Quantum drives I guess this is my 
> problem.  Does anyone 
> know if this code is still actually necessary?  It seems 
> it's been there a 
> while.  It's disappointing to not get full performance out of 
> the hardware you 
> have.

Yes, that code is still necessary.  There's a new aic7xxx driver by Justin
Gibbs at Adaptec which is now being beta tested which corrects this issue.
Something to note, however: the media transfer rate for those disks is at
most ~20MB/sec.  Therefore, you only exceed the 80MB/sec bus speed if you
have more than 4 disks all doing maximum I/O at the same time.  Since the
PowerApp.web 100 has at most 2 disks internally, you really shouldn't see
any significant performance difference.

Hope this helps.
Thanks for buying Dell!
Matt Domsch
Dell Linux Systems Group
Linux OS Development


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/