The Manageability Engine Interface (aka HECI) allows applications to
communicate with the Intel(R) Manageability Engine (ME) firmware.
It is meant to be used by user-space manageability applications to
access ME features such as Intel(R) Active Management Technology,
Intel(R) Quiet System Technology and ASF.
The patch is large so I'm placing the diff on the web for download
rather than attaching it here. Download:
http://download.openamt.org/intel-MEI.diff
Anas
Hi,
Le lundi 22 octobre 2007 ? 13:22 -0400, Anas Nashif a ?crit :
> The Manageability Engine Interface (aka HECI) allows applications to
> communicate with the Intel(R) Manageability Engine (ME) firmware.
>
> It is meant to be used by user-space manageability applications to
> access ME features such as Intel(R) Active Management Technology,
> Intel(R) Quiet System Technology and ASF.
Could you briefly explain all these terms ?
Thanks,
Xav
On Mon, 22 Oct 2007 13:22:38 -0400 Anas Nashif <[email protected]> wrote:
> The Manageability Engine Interface (aka HECI) allows applications to
> communicate with the Intel(R) Manageability Engine (ME) firmware.
>
> It is meant to be used by user-space manageability applications to
> access ME features such as Intel(R) Active Management Technology,
> Intel(R) Quiet System Technology and ASF.
>
> The patch is large so I'm placing the diff on the web for download
> rather than attaching it here. Download:
>
> http://download.openamt.org/intel-MEI.diff
Please get the patches mailed out somehow. Consider splitting the patch
up. I think you could send it as-is (200k will fit, I believe) but it's
rather too large to review effectively.
The code looks good from a quick scan. Immediate impressions from a
quick scan, mainly trivia:
- Consider using open-coded "0" in place of ESUCCESS
- Remove #pragme pack(1), use __attribute__((packed))
- Why does a new driver have "additional char device for legacy mode"?
- Remove the INFO(), ERR() and WARN() macros, just open-code the printk calls
- Remove the commented-out DEF_PARM, HECI_DEBUG, any others
- You can't have a global variable called "debug"!
- Remove private definitions of TRUE and FALSE. Use standard "true" and
"false" or just open-code 0 and 1.
- This:
+ /**
+ * virtual void GetParam(const char* UserParam);
+ * read write messages to/from heci fw
+ */
appears to be dead code, should be called get_param, should use
tabstops not spaces
- These:
+#define SHUTDOWN_METHOD(method) .shutdown = method,
+#define HECI_REBOOT_NOTIFIER(reboot_notifier, driver, reboot_function)
+#define REGISTER_REBOOT_NOTIFIER(reboot_notifier)
+#define UNREGISTER_REBOOT_NOTIFIER(reboot_notifier)
+#define heci_reboot_notifier
will be unpopular. See if they can be removed via suitable means?
- heci_init.c has these:
+/**
+ * heci init function prototypes
+ */
+int host_start_message(struct iamt_heci_device *dev);
+int host_enum_clients_message(struct iamt_heci_device *dev);
+int allocate_me_clents_storage(struct iamt_heci_device *dev);
+void heci_disable(struct iamt_heci_device *dev);
+void host_init_wd(struct iamt_heci_device *dev);
+void host_init_legacy(struct iamt_heci_device *dev);
+int heci_wait_event_int_timeout(struct iamt_heci_device *dev, long timeout);
which are extern-declarations-in-C which checkpatch failed to detect.
They should be placed in a header file or just removed.
- please review all global symbols, check that they really do need to be
global, and that they have suitable names (ie: not like "debug")
- mysterious msleep(100) in host_start_message()? Needs a comment.
- Please prefer to put a blank line between the declaration of locals and
the start of code in each function.
- Here:
+ if (0 == memcmp(&heci_wd_guid,
we boringly prefer "if (foo == 0)" rather than "if (0 == foo)". (lots
of places).
- The changelog could do with some expansion. What is "Intel(R)
Manageability Engine (ME) firmware"? Why do we want to include this code
in Linux? What value has it to our users, etc? Bascially: tell us more
stuff.
On Monday 22 October 2007 19:22:38 Anas Nashif wrote:
> The Manageability Engine Interface (aka HECI) allows applications to
> communicate with the Intel(R) Manageability Engine (ME) firmware.
>
> It is meant to be used by user-space manageability applications to
> access ME features such as Intel(R) Active Management Technology,
> Intel(R) Quiet System Technology and ASF.
>
> The patch is large so I'm placing the diff on the web for download
> rather than attaching it here. Download:
>
> http://download.openamt.org/intel-MEI.diff
>
>
> Anas
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.tux.org/lkml/
>
Will integrated sensors on DG965 motherboard be supported?
They can be accessed only through the
HECI interface, but the board has no AMT, and even if it had one, I don't
think AMT includes HW sensors.
Will it be supported?
I really need it (I have to reboot to bios each time to see the sensors, which is
both inaccurate, and annoying)
Best regards,
Maxim Levitsky
Andrew,
Thanks for the feedback.
Andrew Morton wrote:
> On Mon, 22 Oct 2007 13:22:38 -0400 Anas Nashif <[email protected]> wrote:
>>
>> The patch is large so I'm placing the diff on the web for download
>> rather than attaching it here. Download:
>>
>> http://download.openamt.org/intel-MEI.diff
>
> Please get the patches mailed out somehow. Consider splitting the patch
> up. I think you could send it as-is (200k will fit, I believe) but it's
> rather too large to review effectively.
>
> The code looks good from a quick scan. Immediate impressions from a
> quick scan, mainly trivia:
>
We will fix the issues below and send the revised patch to the list.
>
>
> - Why does a new driver have "additional char device for legacy mode"?
>
It is not quite new. What is currently considered legacy was supported
on Linux with a driver that was never submitted upstream (although it is
open-source and available from e1000.sf.net).
Some applications still use the legacy interface (KCS style) and have to
be supported with the new driver as well.
(...)
> - The changelog could do with some expansion. What is "Intel(R)
> Manageability Engine (ME) firmware"? Why do we want to include this code
> in Linux? What value has it to our users, etc? Bascially: tell us more
> stuff.
The core hardware architecture of Intel Active Management Technology
(Intel AMT) is resident in firmware. The micro-controller within the
chipset's graphics and memory controller (MCH) hub houses the Management
Engine (ME) firmware, which implements various services on behalf of
management applications. Additionally, flash memory houses system BIOS,
code used by the management engine, and a third-party data store (3PDS)
that enables applications to store information as needed in non-volatile
memory.
Communication between the host OS and the ME is accomplished by means of
the Intel Management Engine Interface (aka HECI: Host Embedded
Controller Interface ). MEI is bi-directional, and either the host or
Intel AMT firmware can initiate transactions.
Some of the ME subsystems that can be access via MEI driver:
- Intel(R) Quiet System Technology (QST) is implemented as a firmware
subsystem that runs in the ME. Programs that wish to expose the
health monitoring and fan speed control capabilities of Intel(R) QST
will need to use the MEI driver to communicate with the ME sub-system.
- ASF is the "Alert Standard Format" which is an DMTF manageability
standard. It is implemented in the PC's hardware and firmware, and is
managed from a remote console.
Most recent Intel desktop chipsets have one or more of the above ME
services. The MEI driver will make it possible to support the above
features on Linux and provides applications access to the ME and it's
features. The MEI drivers will also help bridge a current gap related to
lm_sensors support on recent desktop chipsets.
Anas
Hi,
Maxim Levitsky wrote:
>
> Will integrated sensors on DG965 motherboard be supported?
> They can be accessed only through the
> HECI interface, but the board has no AMT, and even if it had one, I don't
> think AMT includes HW sensors.
>
True, this is supported using QST.
> Will it be supported?
The MEI driver (HECI) is the first step toward supporting this in Linux.
We are working on a QST release for Linux in the form of an SDK that can
be used for example by lm_sensors to expose the sensors.
> I really need it (I have to reboot to bios each time to see the sensors, which is
> both inaccurate, and annoying)
>
Yes, we know how annoying it is and we are working on it.
Anas
>
> Best regards,
> Maxim Levitsky
On Tuesday 23 October 2007 15:31:07 Anas Nashif wrote:
> Andrew,
>
> Thanks for the feedback.
>
> Andrew Morton wrote:
> > On Mon, 22 Oct 2007 13:22:38 -0400 Anas Nashif <[email protected]> wrote:
> >>
> >> The patch is large so I'm placing the diff on the web for download
> >> rather than attaching it here. Download:
> >>
> >> http://download.openamt.org/intel-MEI.diff
> >
> > Please get the patches mailed out somehow. Consider splitting the patch
> > up. I think you could send it as-is (200k will fit, I believe) but it's
> > rather too large to review effectively.
> >
> > The code looks good from a quick scan. Immediate impressions from a
> > quick scan, mainly trivia:
> >
>
> We will fix the issues below and send the revised patch to the list.
>
> >
> >
> > - Why does a new driver have "additional char device for legacy mode"?
> >
>
> It is not quite new. What is currently considered legacy was supported
> on Linux with a driver that was never submitted upstream (although it is
> open-source and available from e1000.sf.net).
> Some applications still use the legacy interface (KCS style) and have to
> be supported with the new driver as well.
>
> (...)
>
> > - The changelog could do with some expansion. What is "Intel(R)
> > Manageability Engine (ME) firmware"? Why do we want to include this code
> > in Linux? What value has it to our users, etc? Bascially: tell us more
> > stuff.
>
> The core hardware architecture of Intel Active Management Technology
> (Intel AMT) is resident in firmware. The micro-controller within the
> chipset's graphics and memory controller (MCH) hub houses the Management
> Engine (ME) firmware, which implements various services on behalf of
> management applications. Additionally, flash memory houses system BIOS,
> code used by the management engine, and a third-party data store (3PDS)
> that enables applications to store information as needed in non-volatile
> memory.
>
> Communication between the host OS and the ME is accomplished by means of
> the Intel Management Engine Interface (aka HECI: Host Embedded
> Controller Interface ). MEI is bi-directional, and either the host or
> Intel AMT firmware can initiate transactions.
>
> Some of the ME subsystems that can be access via MEI driver:
>
> - Intel(R) Quiet System Technology (QST) is implemented as a firmware
> subsystem that runs in the ME. Programs that wish to expose the
> health monitoring and fan speed control capabilities of Intel(R) QST
> will need to use the MEI driver to communicate with the ME sub-system.
> - ASF is the "Alert Standard Format" which is an DMTF manageability
> standard. It is implemented in the PC's hardware and firmware, and is
> managed from a remote console.
>
> Most recent Intel desktop chipsets have one or more of the above ME
> services. The MEI driver will make it possible to support the above
> features on Linux and provides applications access to the ME and it's
> features. The MEI drivers will also help bridge a current gap related to
> lm_sensors support on recent desktop chipsets.
Care to tell me, and other users of recent intel motherboards,
when QST tool to access HW sensors will be released, or at least
is it developed?
>
>
> Anas
>
>
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.tux.org/lkml/
>
Best regards,
Maxim Levitsky
On Tuesday 23 October 2007 15:35:37 Anas Nashif wrote:
> Hi,
>
> Maxim Levitsky wrote:
> >
> > Will integrated sensors on DG965 motherboard be supported?
> > They can be accessed only through the
> > HECI interface, but the board has no AMT, and even if it had one, I don't
> > think AMT includes HW sensors.
> >
> True, this is supported using QST.
>
> > Will it be supported?
>
> The MEI driver (HECI) is the first step toward supporting this in Linux.
> We are working on a QST release for Linux in the form of an SDK that can
> be used for example by lm_sensors to expose the sensors.\
This would be great, any dates?, at least approximate...
>
> > I really need it (I have to reboot to bios each time to see the sensors, which is
> > both inaccurate, and annoying)
> >
>
> Yes, we know how annoying it is and we are working on it.
>
> Anas
>
> >
> > Best regards,
> > Maxim Levitsky
>
Best regards,
Maxim Levitsky
On Mon, Oct 22, 2007 at 11:40:01AM -0700, Andrew Morton wrote:
> - Here:
>
> + if (0 == memcmp(&heci_wd_guid,
>
> we boringly prefer "if (foo == 0)" rather than "if (0 == foo)". (lots
> of places).
But 0 == blah is safer. If you accidentally do 0 = blah the compiler
will tell you. Just because people have always done it the other way
around doesn't make it the right way to do it. I have noticed many
people have started to realize this in the last few years.
It is also much clearer that you are comparing against a constant and
not doing an assignment when the constant comes before the variable.
I think to encourage people doing it the less safe way is just silly.
Some places in the kernel that already uses the constant first are:
./sound/oss/btaudio.c
./sound/usb/usx2y/
./net/sctp/
./arch/x86_64/mm/init.c
./arch/cris/arch-v32/drivers/cryptocop.c
./drivers/i2c/algos/i2c-algo-pcf.c
./drivers/mtd/devices/doc2000.c
./drivers/net/iseries_veth.c
./drivers/net/hp100.c
./drivers/net/phy/phy.c
./drivers/net/xen-netfront.c
./drivers/scsi/scsi_error.c
./drivers/scsi/sr_ioctl.c
./drivers/scsi/ch.c
./drivers/scsi/sg.c
./drivers/scsi/sr_vendor.c
./drivers/scsi/scsi_debug.c
./drivers/media/video/tea6415c.c
./drivers/media/video/tvaudio.c
./drivers/media/video/bt8xx/
./drivers/char/ip2/
./drivers/isdn/hisax/
./drivers/sbus/char/
./fs/nfsd/
and I am sure there a quite a few more. It ought to be encouraged for
all new code to avoid stupid typo bugs that are hard to see.
--
Len Sorensen
On Tue, 23 Oct 2007 09:31:07 -0400 Anas Nashif <[email protected]> wrote:
> Andrew,
>
> Thanks for the feedback.
>
> Andrew Morton wrote:
> > On Mon, 22 Oct 2007 13:22:38 -0400 Anas Nashif <[email protected]> wrote:
> >>
> >> The patch is large so I'm placing the diff on the web for download
> >> rather than attaching it here. Download:
> >>
> >> http://download.openamt.org/intel-MEI.diff
> >
> > Please get the patches mailed out somehow. Consider splitting the patch
> > up. I think you could send it as-is (200k will fit, I believe) but it's
> > rather too large to review effectively.
> >
> > The code looks good from a quick scan. Immediate impressions from a
> > quick scan, mainly trivia:
> >
>
> We will fix the issues below and send the revised patch to the list.
>
> >
> >
> > - Why does a new driver have "additional char device for legacy mode"?
> >
>
> It is not quite new. What is currently considered legacy was supported
> on Linux with a driver that was never submitted upstream (although it is
> open-source and available from e1000.sf.net).
> Some applications still use the legacy interface (KCS style) and have to
> be supported with the new driver as well.
>
It would be better to remove the lecacy mode support from the new driver
and to continue to ship a patch for those people who use the old interface.
They've been patching in the whole driver thus far so I assume all the
processes are already in place for this.
>
> > - The changelog could do with some expansion. What is "Intel(R)
> > Manageability Engine (ME) firmware"? Why do we want to include this code
> > in Linux? What value has it to our users, etc? Bascially: tell us more
> > stuff.
>
> The core hardware architecture of Intel Active Management Technology
> (Intel AMT) is resident in firmware. The micro-controller within the
> chipset's graphics and memory controller (MCH) hub houses the Management
> Engine (ME) firmware, which implements various services on behalf of
> management applications. Additionally, flash memory houses system BIOS,
> code used by the management engine, and a third-party data store (3PDS)
> that enables applications to store information as needed in non-volatile
> memory.
>
> Communication between the host OS and the ME is accomplished by means of
> the Intel Management Engine Interface (aka HECI: Host Embedded
> Controller Interface ). MEI is bi-directional, and either the host or
> Intel AMT firmware can initiate transactions.
>
> Some of the ME subsystems that can be access via MEI driver:
>
> - Intel(R) Quiet System Technology (QST) is implemented as a firmware
> subsystem that runs in the ME. Programs that wish to expose the
> health monitoring and fan speed control capabilities of Intel(R) QST
> will need to use the MEI driver to communicate with the ME sub-system.
> - ASF is the "Alert Standard Format" which is an DMTF manageability
> standard. It is implemented in the PC's hardware and firmware, and is
> managed from a remote console.
>
> Most recent Intel desktop chipsets have one or more of the above ME
> services. The MEI driver will make it possible to support the above
> features on Linux and provides applications access to the ME and it's
> features. The MEI drivers will also help bridge a current gap related to
> lm_sensors support on recent desktop chipsets.
>
I see, thanks. That would be a fine addition to the patch's changelog,
please.
On Tue, 23 Oct 2007 11:04:29 -0400
[email protected] (Lennart Sorensen) wrote:
> On Mon, Oct 22, 2007 at 11:40:01AM -0700, Andrew Morton wrote:
> > - Here:
> >
> > + if (0 == memcmp(&heci_wd_guid,
> >
> > we boringly prefer "if (foo == 0)" rather than "if (0 == foo)".
> > (lots of places).
>
> But 0 == blah is safer. If you accidentally do 0 = blah the compiler
> will tell you.
gcc will tell you in the other direction just as well.
and people read from left to right (at least in english) so coding in
that direction is generally preferred in the Linux kernel as well.
On Tue, 23 Oct 2007 11:04:29 -0400 [email protected] (Lennart Sorensen) wrote:
> On Mon, Oct 22, 2007 at 11:40:01AM -0700, Andrew Morton wrote:
> > - Here:
> >
> > + if (0 == memcmp(&heci_wd_guid,
> >
> > we boringly prefer "if (foo == 0)" rather than "if (0 == foo)". (lots
> > of places).
>
> But 0 == blah is safer. If you accidentally do 0 = blah the compiler
> will tell you.
If you do 'if (blah = 0)' then compiler will tell you too. To all intents
and purposes this invalidates the reasons for doing `if (0 == blah)'.
> Just because people have always done it the other way
> around doesn't make it the right way to do it. I have noticed many
> people have started to realize this in the last few years.
>
> It is also much clearer that you are comparing against a constant and
> not doing an assignment when the constant comes before the variable.
>
> I think to encourage people doing it the less safe way is just silly.
It isn't less safe.
> Some places in the kernel that already uses the constant first are:
It impacts readability. All the aio code was implemented that way for a
few years and it drove everyone so batty that we undid it.
On Tue, Oct 23, 2007 at 09:23:33AM -0700, Arjan van de Ven wrote:
> gcc will tell you in the other direction just as well.
>
> and people read from left to right (at least in english) so coding in
> that direction is generally preferred in the Linux kernel as well.
What does gcc have to say about if (foo = 0){ rather than if (foo == 0){
Both are legal C so it shouldn't say anything. Of course the first is
usually a bug (or very bad style).
On the other hand if (0 = foo) will give an error.
It isn't about how you read in english, it is about not making mistakes.
And why can't you say if 0 is equal to the variable foo rather than if
the variable foo is equal to 0? Both are valid english, so that is just
a crappy excuse for sticking with a bad idea.
--
Len Sorensen
> What does gcc have to say about if (foo = 0){ rather than if (foo == 0){
It's not a hard experiment to do.
The answer is:
warning: suggest parentheses around assignment used as truth value
- R.
On Tue, Oct 23, 2007 at 11:22:50AM -0700, Roland Dreier wrote:
> It's not a hard experiment to do.
>
> The answer is:
>
> warning: suggest parentheses around assignment used as truth value
A warning is not an error. It won't abort the compile.
The warning (which I don't remember gcc doing in the past) is a nice
idea though.
--
Len Sorensen
On Die, 2007-10-23 at 15:35 -0400, Lennart Sorensen wrote:
> On Tue, Oct 23, 2007 at 11:22:50AM -0700, Roland Dreier wrote:
> > It's not a hard experiment to do.
> >
> > The answer is:
> >
> > warning: suggest parentheses around assignment used as truth value
>
> A warning is not an error. It won't abort the compile.
Add "-Werror".
> The warning (which I don't remember gcc doing in the past) is a nice
> idea though.
That's the case since many years - I don't remember how long.
Bernd
--
Firmix Software GmbH http://www.firmix.at/
mobil: +43 664 4416156 fax: +43 1 7890849-55
Embedded Linux Development and Services
Maxim Levitsky wrote:
> Will integrated sensors on DG965 motherboard be supported?
> They can be accessed only through the
> HECI interface, but the board has no AMT, and even if it had one, I don't
> think AMT includes HW sensors.
While talking about AMT ...
<ad>
While talking about AMT: One quite useful thing is AMT provides is
*remote* management (most useful for kernel hackers: serial-over-lan
console, remote reset). I've hacked up some tools to use this stuff.
Package is is called amtterm.
Source code: http://dl.bytesex.org/releases/amtterm/
Fedora 7+8 have ready-to-go packages in the repositories, so you can
just "yum install amtterm".
SuSE packages are in the buildservice:
http://download.opensuse.org/repositories/home:/kraxel/
</ad>
HTH,
Gerd