smp_utils 0.93 beta, adds sgv4 interface

smp_utils 0.93 beta, adds sgv4 interface

[Douglas Gilbert]

smp_utils is a package of command line utilities for invoking
SMP functions to monitor and manage SAS expanders. SMP is the
Serial Attached SCSI (SAS) Management Protocol (SMP). A SAS Host
Bus Adapter (HBA) includes a SMP initiator (along with a SSP and
STP initiator). A SAS expander contains a SMP target.

Previous versions have supported two interfaces: the "mptctl"
pass through used by MPT Fusion SAS HBAs and the "smp_portal"
sysfs attribute pass through used by Luben's aic94xx driver.

This beta adds preliminary support for an interface based on
the scsi generic version 4 ("sgv4") interface. It works, apart
from a few shortcomings, with the MPT Fusion SAS driver found
in lk 2.6.24-rc4 . My aic94xx based hardware refuses to
work with the mainline driver (but works with Luben's
driver) so I'm unable to test it with the "sgv4" interface. I
would be grateful if someone would test it and tell me how it
fairs. Since the sgv4 interface essentially talks to the
linux SAS transport layer, other SAS HBAs may work as well.

For an overview and examples of smp_utils see:
http://www.torque.net/sg/smp_utils.html
A tarball can be found in table 2.

Version 0.93 ChangeLog:
Changelog for smp_utils-0.93 [20071211]
  - sgv4: add new interface
  - smp_phy_control: sync with sas2r10
  - smp_rep_phy_sata: affiliation contexts, added sas2r08
  - smp_discover: add '--list' option for attribute=value output
  - smp_rep_general: sync with sas2r09..12
  - smp_discover, smp_discover_list: sync with sas2r09..12
    - warn if expander's SAS address is different from requested
  - add smp_rep_exp_route_tbl utility
  - rename this file from CHANGELOG to ChangeLog
  - place code in subversion repository
  - move smp_lib.h into include directory
  - move man pages into doc directory


Doug Gilbert

Re: smp_utils 0.93 beta, adds sgv4 interface

[James Bottomley]

On Wed, 2007-12-19 at 23:46 -0500, Douglas Gilbert wrote:
> smp_utils is a package of command line utilities for invoking
> SMP functions to monitor and manage SAS expanders. SMP is the
> Serial Attached SCSI (SAS) Management Protocol (SMP). A SAS Host
> Bus Adapter (HBA) includes a SMP initiator (along with a SSP and
> STP initiator). A SAS expander contains a SMP target.
> 
> Previous versions have supported two interfaces: the "mptctl"
> pass through used by MPT Fusion SAS HBAs and the "smp_portal"
> sysfs attribute pass through used by Luben's aic94xx driver.
> 
> This beta adds preliminary support for an interface based on
> the scsi generic version 4 ("sgv4") interface. It works, apart
> from a few shortcomings, with the MPT Fusion SAS driver found
> in lk 2.6.24-rc4 . My aic94xx based hardware refuses to
> work with the mainline driver (but works with Luben's
> driver) so I'm unable to test it with the "sgv4" interface. I
> would be grateful if someone would test it and tell me how it
> fairs. Since the sgv4 interface essentially talks to the
> linux SAS transport layer, other SAS HBAs may work as well.

I'm pleased to report that it works just fine for me with my aic94xx
setup and the sgv4 interface and a couple of LSI expanders (which,
unfortunately don't take all the commands ... it looks like you're using
the SAS2.0 set).

I'll also plug in my mpt1068 and see if I can get it to function through
the sgv4 interface as well.

One of the things which might be useful would be for SAS hosts to
respond to SMP requests so they could also be managed with these tools.
To do this, we'd have to put a SMP interpreter for them inside the
transport class, probably just for a minimal set of SMP frames ... what
do people think about this?

Thanks for doing this,

James

Re: smp_utils 0.93 beta, adds sgv4 interface

[Douglas Gilbert]

James Bottomley wrote:
> On Wed, 2007-12-19 at 23:46 -0500, Douglas Gilbert wrote:
>> smp_utils is a package of command line utilities for invoking
>> SMP functions to monitor and manage SAS expanders. SMP is the
>> Serial Attached SCSI (SAS) Management Protocol (SMP). A SAS Host
>> Bus Adapter (HBA) includes a SMP initiator (along with a SSP and
>> STP initiator). A SAS expander contains a SMP target.
>>
>> Previous versions have supported two interfaces: the "mptctl"
>> pass through used by MPT Fusion SAS HBAs and the "smp_portal"
>> sysfs attribute pass through used by Luben's aic94xx driver.
>>
>> This beta adds preliminary support for an interface based on
>> the scsi generic version 4 ("sgv4") interface. It works, apart
>> from a few shortcomings, with the MPT Fusion SAS driver found
>> in lk 2.6.24-rc4 . My aic94xx based hardware refuses to
>> work with the mainline driver (but works with Luben's
>> driver) so I'm unable to test it with the "sgv4" interface. I
>> would be grateful if someone would test it and tell me how it
>> fairs. Since the sgv4 interface essentially talks to the
>> linux SAS transport layer, other SAS HBAs may work as well.
> 
> I'm pleased to report that it works just fine for me with my aic94xx
> setup and the sgv4 interface and a couple of LSI expanders (which,
> unfortunately don't take all the commands ... it looks like you're using
> the SAS2.0 set).

Good.
Yes, I am tracking the SAS-2 drafts but they do keep backward
compatibility with SAS-1.1 and SAS-1 . See the COVERAGE file
in the smp_utils tarball.
I'm not aware that LSI have released any SAS-2 expanders.
When manufacturers start talking about zoning in their
expanders they should be supporting at least some of the extra
SAS-2 expander functionality.

> I'll also plug in my mpt1068 and see if I can get it to function through
> the sgv4 interface as well.
> 
> One of the things which might be useful would be for SAS hosts to
> respond to SMP requests so they could also be managed with these tools.
> To do this, we'd have to put a SMP interpreter for them inside the
> transport class, probably just for a minimal set of SMP frames ... what
> do people think about this?

The MPT Fusion SAS HBAs already supports a virtual SMP target
on the HBA itself. Try:
   smp_rep_manufacturer /sys/class/bsg/sas_host4
where sas_host4 is a MPT SAS HBA. This virtual SMP target is not
visible to other devices in the same SAS domain.

I believe this allows access to the sideband signals (SGPIO)
on a wide internal cable. Some enclosure management chips
(e.g. MG9072) offer SES-2 target functionality on a SGPIO
interface. So are you proposing that other SAS HBAs (e.g. aic94xx)
should add code to their drivers so that the SMP READ and WRITE
GPIO REGISTER functions control those SGPIO sideband signals on
a wide internal connector? Or did you have something else in mind ...


Doug Gilbert

Re: smp_utils 0.93 beta, adds sgv4 interface

[James Bottomley]

On Fri, 2007-12-21 at 10:10 -0500, Douglas Gilbert wrote:
> James Bottomley wrote:
> > On Wed, 2007-12-19 at 23:46 -0500, Douglas Gilbert wrote:
> >> smp_utils is a package of command line utilities for invoking
> >> SMP functions to monitor and manage SAS expanders. SMP is the
> >> Serial Attached SCSI (SAS) Management Protocol (SMP). A SAS Host
> >> Bus Adapter (HBA) includes a SMP initiator (along with a SSP and
> >> STP initiator). A SAS expander contains a SMP target.
> >>
> >> Previous versions have supported two interfaces: the "mptctl"
> >> pass through used by MPT Fusion SAS HBAs and the "smp_portal"
> >> sysfs attribute pass through used by Luben's aic94xx driver.
> >>
> >> This beta adds preliminary support for an interface based on
> >> the scsi generic version 4 ("sgv4") interface. It works, apart
> >> from a few shortcomings, with the MPT Fusion SAS driver found
> >> in lk 2.6.24-rc4 . My aic94xx based hardware refuses to
> >> work with the mainline driver (but works with Luben's
> >> driver) so I'm unable to test it with the "sgv4" interface. I
> >> would be grateful if someone would test it and tell me how it
> >> fairs. Since the sgv4 interface essentially talks to the
> >> linux SAS transport layer, other SAS HBAs may work as well.
> > 
> > I'm pleased to report that it works just fine for me with my aic94xx
> > setup and the sgv4 interface and a couple of LSI expanders (which,
> > unfortunately don't take all the commands ... it looks like you're using
> > the SAS2.0 set).
> 
> Good.
> Yes, I am tracking the SAS-2 drafts but they do keep backward
> compatibility with SAS-1.1 and SAS-1 . See the COVERAGE file
> in the smp_utils tarball.
> I'm not aware that LSI have released any SAS-2 expanders.
> When manufacturers start talking about zoning in their
> expanders they should be supporting at least some of the extra
> SAS-2 expander functionality.

Yes ... I've no idea when (if) I'll get SAS-2 expanders.

> > I'll also plug in my mpt1068 and see if I can get it to function through
> > the sgv4 interface as well.
> > 
> > One of the things which might be useful would be for SAS hosts to
> > respond to SMP requests so they could also be managed with these tools.
> > To do this, we'd have to put a SMP interpreter for them inside the
> > transport class, probably just for a minimal set of SMP frames ... what
> > do people think about this?
> 
> The MPT Fusion SAS HBAs already supports a virtual SMP target
> on the HBA itself. Try:
>    smp_rep_manufacturer /sys/class/bsg/sas_host4
> where sas_host4 is a MPT SAS HBA. This virtual SMP target is not
> visible to other devices in the same SAS domain.
> 
> I believe this allows access to the sideband signals (SGPIO)
> on a wide internal cable. Some enclosure management chips
> (e.g. MG9072) offer SES-2 target functionality on a SGPIO
> interface. So are you proposing that other SAS HBAs (e.g. aic94xx)
> should add code to their drivers so that the SMP READ and WRITE
> GPIO REGISTER functions control those SGPIO sideband signals on
> a wide internal connector? Or did you have something else in mind ...

The way I was planning to implement it was to stick an SMP interpreter
in libsas (sort of like the way the fusion does it by using the firmware
page) and make it translate to the existing phy functions.  However, as
best as I can discover (at least according to the docs I have) the
aic94xx has no external SGPIO bus (although perhaps someone with better
documentation knows better?), so support would only be over an expander
link (that also means we have no SGPIO support in libsas itself,
either).

James