kvm/ia64 TODO List. (Draft).

kvm/ia64 TODO List. (Draft).

[Zhang, Xiantao]

Hi,  folks
In the past months, kvm/ia64 has been merged into linux mainline kernel
with the help from community, and also get a relatively stable status.
Cheers! Thanks for your effort!! You know, kvm/ia64 bits will be
included in Linux-2.6.26 release. 
In the next stage, kvm/ia64 may be integrated into some commercial Linux
OS, and before that, we need to ensure its stability and greats feature
enabled for it. So, now we need your hands to join in the development
for enhancing the features, and make it more and more stable in the next
weeks. 
But compared with x86 side, we still have some good features to catch
up.  In this mail, I will list them and their dependcies in detail. If
you are interest in one or more tasks, please reply the mail directly
and let other guys to track their progress and know the task's holders.
Welcome to join us!! Appreciate any effort from you!

TODO List (draft):

1. Live Migration and Save restore.  (Xiantao is working on it.)  
2. Host-memory swapping.  (Good and urgent feature to catch up.)
3. virto-io enabling. (Fast IO for kvm guests.)
4. Guest reboot and shutdown support. (Basic feature)
5. vt-d enabling.  (Maybe share the code with x86 side, but only to
implement ia64-specific side.  Depend on VT-d platform for ia64)
6. superpage support for guest. 
7. kvm trace enabling. 

If I omit some good features or you have some good proposals, please
feel free to bring forward them! Thanks in advace for your support!  :-)


BTW. Kvm/ia64 mailing list has switched to kernel.org, and the
sourceforge list has been dropped from now on.  Please subscribe the new
one from http://vger.kernel.org/vger-lists.html.  

Xiantao

Re: [kvm-ia64-devel] kvm/ia64 TODO List. (Draft).

[Akio Takebe]

[-- Attachment #1: Mail message body --]
[-- Type: text/plain, Size: 2496 bytes --]

Hi, Xiantao

I want to add supporting windows2008.

Thank you for your good proposal.
I tried to boot windows2008 with supporting more pal calls before,
I couldn't do it. I'm not sure the casuse yet.
FYI, I attached the patch.

Best Regards,

Akio Takebe

>Hi,  folks
>In the past months, kvm/ia64 has been merged into linux mainline kernel
>with the help from community, and also get a relatively stable status.
>Cheers! Thanks for your effort!! You know, kvm/ia64 bits will be
>included in Linux-2.6.26 release. 
>In the next stage, kvm/ia64 may be integrated into some commercial Linux
>OS, and before that, we need to ensure its stability and greats feature
>enabled for it. So, now we need your hands to join in the development
>for enhancing the features, and make it more and more stable in the next
>weeks. 
>But compared with x86 side, we still have some good features to catch
>up.  In this mail, I will list them and their dependcies in detail. If
>you are interest in one or more tasks, please reply the mail directly
>and let other guys to track their progress and know the task's holders.
>Welcome to join us!! Appreciate any effort from you!
>
>TODO List (draft):
>
>1. Live Migration and Save restore.  (Xiantao is working on it.)  
>2. Host-memory swapping.  (Good and urgent feature to catch up.)
>3. virto-io enabling. (Fast IO for kvm guests.)
>4. Guest reboot and shutdown support. (Basic feature)
>5. vt-d enabling.  (Maybe share the code with x86 side, but only to
>implement ia64-specific side.  Depend on VT-d platform for ia64)
>6. superpage support for guest. 
>7. kvm trace enabling. 
>
>If I omit some good features or you have some good proposals, please
>feel free to bring forward them! Thanks in advace for your support!  :-)
>
>
>BTW. Kvm/ia64 mailing list has switched to kernel.org, and the
>sourceforge list has been dropped from now on.  Please subscribe the new
>one from http://vger.kernel.org/vger-lists.html.  
>
>Xiantao
>
>-------------------------------------------------------------------------
>Sponsored by: SourceForge.net Community Choice Awards: VOTE NOW!
>Studies have shown that voting for your favorite open source project,
>along with a healthy diet, reduces your potential for chronic lameness
>and boredom. Vote Now at http://www.sourceforge.net/community/cca08
>_______________________________________________
>kvm-ia64-devel mailing list
>kvm-ia64-devel@lists.sourceforge.net
>https://lists.sourceforge.net/lists/listinfo/kvm-ia64-devel

[-- Attachment #2: support_more_palcall.patch --]
[-- Type: application/octet-stream, Size: 3597 bytes --]

diff --git a/arch/ia64/kvm/kvm_fw.c b/arch/ia64/kvm/kvm_fw.c
index 091f936..c29a3ea 100644
--- a/arch/ia64/kvm/kvm_fw.c
+++ b/arch/ia64/kvm/kvm_fw.c
@@ -112,6 +112,19 @@ static void remote_pal_cache_flush(void *data)
 	args->status = status;
 }
 
+static void
+remote_pal_prefetch_visibility(void *v)
+{
+	s64 trans_type = (s64)v;
+	ia64_pal_prefetch_visibility(trans_type);
+}
+
+static void
+remote_pal_mc_drain(void *v)
+{
+	ia64_pal_mc_drain();
+}
+
 static struct ia64_pal_retval pal_cache_flush(struct kvm_vcpu *vcpu)
 {
 	u64 gr28, gr29, gr30, gr31;
@@ -158,6 +171,38 @@ static struct ia64_pal_retval pal_cache_flush(struct kvm_vcpu *vcpu)
 	return result;
 }
 
+static struct ia64_pal_retval pal_prefetch_visibility(struct kvm_vcpu *vcpu)
+{
+	struct ia64_pal_retval result = {0, 0, 0, 0};
+	unsigned long in0, in1, in2, in3;
+	kvm_get_pal_call_data(vcpu, &in0, &in1, &in2, &in3);
+	result.status = ia64_pal_prefetch_visibility(in1);
+	if (result.status == 0) {
+		/* must be performed on all remote processors 
+		   in the coherence domain. */
+		smp_call_function(remote_pal_prefetch_visibility,
+				  (void *)in1, 1, 1);
+		 /* no more necessary on remote processor */
+		result.status = 1;
+	}
+	return result;
+}
+
+static struct ia64_pal_retval pal_get_brand_info(struct kvm_vcpu *vcpu)
+{
+	struct ia64_pal_retval result = {0, 0, 0, 0};
+	unsigned long in0, in1, in2, in3;
+	kvm_get_pal_call_data(vcpu, &in0, &in1, &in2, &in3);
+	if (in1 == 0) {
+		char brand_info[128];
+		result.status = ia64_pal_get_brand_info(brand_info);
+		if (result.status == PAL_STATUS_SUCCESS)
+			copy_to_user((void *)in2, brand_info, 128);
+	} else {
+		result.status = PAL_STATUS_EINVAL;
+	}
+	return result;
+}
 struct ia64_pal_retval pal_cache_summary(struct kvm_vcpu *vcpu)
 {
 
@@ -241,6 +286,24 @@ static struct ia64_pal_retval pal_cache_info(struct kvm_vcpu *vcpu)
 	return ((struct ia64_pal_retval){status, r9, r10, 0});
 }
 
+static struct ia64_pal_retval pal_perf_mon_info(struct kvm_vcpu *vcpu)
+{
+	long status;
+	unsigned long in0, in1, in2, in3, r9;
+	unsigned long pm_buffer[16];
+
+	kvm_get_pal_call_data(vcpu, &in0, &in1, &in2, &in3);
+	status = ia64_pal_perf_mon_info( pm_buffer, (pal_perf_mon_info_u_t *) &r9);
+	if (status != 0) {
+		printk("PAL_PERF_MON_INFO fails ret=%ld\n", status);
+	} else {
+		if (copy_to_user((void *)in1, pm_buffer, sizeof(pm_buffer))) {
+			status = PAL_STATUS_EINVAL;
+		}
+	}
+	return ((struct ia64_pal_retval){status, r9, 0, 0});
+}
+
 #define GUEST_IMPL_VA_MSB	59
 #define GUEST_RID_BITS		18
 
@@ -334,6 +397,9 @@ int kvm_pal_emul(struct kvm_vcpu *vcpu, struct kvm_run *run)
 	case PAL_PLATFORM_ADDR :
 		result = pal_platform_addr(vcpu);
 		break;
+	case PAL_PERF_MON_INFO:
+		result = pal_perf_mon_info(vcpu);
+		break;
 	case PAL_CACHE_INFO:
 		result = pal_cache_info(vcpu);
 		break;
@@ -366,6 +432,22 @@ int kvm_pal_emul(struct kvm_vcpu *vcpu, struct kvm_run *run)
 		result.status = PAL_STATUS_SUCCESS;
 		result.v0 = vcpu->vcpu_id;
 		break;
+	case PAL_PREFETCH_VISIBILITY:
+		result = pal_prefetch_visibility(vcpu);
+		break;
+	case PAL_MC_DRAIN:
+		result.status = ia64_pal_mc_drain();
+		/* FIXME: All vcpus likely call PAL_MC_DRAIN.
+		   That causes the congestion. */
+		smp_call_function(remote_pal_mc_drain, NULL, 1, 1);
+		break;
+	case PAL_BRAND_INFO:
+		result = pal_get_brand_info(vcpu);
+		break;
+	case PAL_GET_PSTATE:
+	case PAL_CACHE_SHARED_INFO:
+		/* Optional, no need to complain about being unimplemented */
+		break;
 	default:
 		INIT_PAL_STATUS_UNIMPLEMENTED(result);
 		printk(KERN_WARNING"kvm: Unsupported pal call,"

RE: [kvm-ia64-devel] kvm/ia64 TODO List. (Draft).

[Zhang, Xiantao]

Akio Takebe wrote:
> Hi, Xiantao
> 
> I want to add supporting windows2008.
> 
> Thank you for your good proposal.
> I tried to boot windows2008 with supporting more pal calls before,
> I couldn't do it. I'm not sure the casuse yet.
> FYI, I attached the patch.

Hi, Akio 
	Thanks for your work! So, with the extra pal calls support, you
still can't boot windows 2008 ?  :)
Xiantao


> Best Regards,
> 
> Akio Takebe
> 
>> Hi,  folks
>> In the past months, kvm/ia64 has been merged into linux mainline
>> kernel with the help from community, and also get a relatively
>> stable status. Cheers! Thanks for your effort!! You know, kvm/ia64
>> bits will be included in Linux-2.6.26 release.
>> In the next stage, kvm/ia64 may be integrated into some commercial
>> Linux OS, and before that, we need to ensure its stability and
>> greats feature enabled for it. So, now we need your hands to join in
>> the development for enhancing the features, and make it more and
>> more stable in the next weeks. But compared with x86 side, we still
>> have some good features to catch up.  In this mail, I will list them
>> and their dependcies in detail. If you are interest in one or more
>> tasks, please reply the mail directly and let other guys to track
>> their progress and know the task's holders. Welcome to join us!!
>> Appreciate any effort from you! 
>> 
>> TODO List (draft):
>> 
>> 1. Live Migration and Save restore.  (Xiantao is working on it.)
>> 2. Host-memory swapping.  (Good and urgent feature to catch up.)
>> 3. virto-io enabling. (Fast IO for kvm guests.)
>> 4. Guest reboot and shutdown support. (Basic feature)
>> 5. vt-d enabling.  (Maybe share the code with x86 side, but only to
>> implement ia64-specific side.  Depend on VT-d platform for ia64)
>> 6. superpage support for guest.
>> 7. kvm trace enabling.
>> 
>> If I omit some good features or you have some good proposals, please
>> feel free to bring forward them! Thanks in advace for your support! 
>> :-) 
>> 
>> 
>> BTW. Kvm/ia64 mailing list has switched to kernel.org, and the
>> sourceforge list has been dropped from now on.  Please subscribe the
>> new one from http://vger.kernel.org/vger-lists.html.
>> 
>> Xiantao
>> 
>>
------------------------------------------------------------------------
-
>> Sponsored by: SourceForge.net Community Choice Awards: VOTE NOW!
>> Studies have shown that voting for your favorite open source project,
>> along with a healthy diet, reduces your potential for chronic
>> lameness and boredom. Vote Now at
>> http://www.sourceforge.net/community/cca08
>> _______________________________________________ 
>> kvm-ia64-devel mailing list
>> kvm-ia64-devel@lists.sourceforge.net
>> https://lists.sourceforge.net/lists/listinfo/kvm-ia64-devel

RE: [kvm-ia64-devel] kvm/ia64 TODO List. (Draft).

[Zhang, Xiantao]

Akio Takebe wrote:
> Hi, Xiantao
> 
> I want to add supporting windows2008.

So, do you have time to work on it ? :)
Xiantao

 
> Thank you for your good proposal.
> I tried to boot windows2008 with supporting more pal calls before,
> I couldn't do it. I'm not sure the casuse yet.
> FYI, I attached the patch.
> 
> Best Regards,
> 
> Akio Takebe
> 
>> Hi,  folks
>> In the past months, kvm/ia64 has been merged into linux mainline
>> kernel with the help from community, and also get a relatively
>> stable status. Cheers! Thanks for your effort!! You know, kvm/ia64
>> bits will be included in Linux-2.6.26 release.
>> In the next stage, kvm/ia64 may be integrated into some commercial
>> Linux OS, and before that, we need to ensure its stability and
>> greats feature enabled for it. So, now we need your hands to join in
>> the development for enhancing the features, and make it more and
>> more stable in the next weeks. But compared with x86 side, we still
>> have some good features to catch up.  In this mail, I will list them
>> and their dependcies in detail. If you are interest in one or more
>> tasks, please reply the mail directly and let other guys to track
>> their progress and know the task's holders. Welcome to join us!!
>> Appreciate any effort from you! 
>> 
>> TODO List (draft):
>> 
>> 1. Live Migration and Save restore.  (Xiantao is working on it.)
>> 2. Host-memory swapping.  (Good and urgent feature to catch up.)
>> 3. virto-io enabling. (Fast IO for kvm guests.)
>> 4. Guest reboot and shutdown support. (Basic feature)
>> 5. vt-d enabling.  (Maybe share the code with x86 side, but only to
>> implement ia64-specific side.  Depend on VT-d platform for ia64)
>> 6. superpage support for guest.
>> 7. kvm trace enabling.
>> 
>> If I omit some good features or you have some good proposals, please
>> feel free to bring forward them! Thanks in advace for your support! 
>> :-) 
>> 
>> 
>> BTW. Kvm/ia64 mailing list has switched to kernel.org, and the
>> sourceforge list has been dropped from now on.  Please subscribe the
>> new one from http://vger.kernel.org/vger-lists.html.
>> 
>> Xiantao
>> 
>>
------------------------------------------------------------------------
-
>> Sponsored by: SourceForge.net Community Choice Awards: VOTE NOW!
>> Studies have shown that voting for your favorite open source project,
>> along with a healthy diet, reduces your potential for chronic
>> lameness and boredom. Vote Now at
>> http://www.sourceforge.net/community/cca08
>> _______________________________________________ 
>> kvm-ia64-devel mailing list
>> kvm-ia64-devel@lists.sourceforge.net
>> https://lists.sourceforge.net/lists/listinfo/kvm-ia64-devel

RE: [kvm-ia64-devel] kvm/ia64 TODO List. (Draft).

[Akio Takebe]

Hi, Xiantao

>Akio Takebe wrote:
>> Hi, Xiantao
>> 
>> I want to add supporting windows2008.
>> 
>> Thank you for your good proposal.
>> I tried to boot windows2008 with supporting more pal calls before,
>> I couldn't do it. I'm not sure the casuse yet.
>> FYI, I attached the patch.
>
>Hi, Akio 
>	Thanks for your work! So, with the extra pal calls support, you
>still can't boot windows 2008 ?  :)
>Xiantao
>
Yes, I couldn't boot it completely.
Windows 2008 installer hanged up if I applied the patch.

Best Regards,

Akio Takebe 

RE: [kvm-ia64-devel] kvm/ia64 TODO List. (Draft).

[Akio Takebe]

Hi, Xiantao

>Akio Takebe wrote:
>> Hi, Xiantao
>> 
>> I want to add supporting windows2008.
>
>So, do you have time to work on it ? :)
>Xiantao
I don't have enough time now.
But I'm going to try while my rest time. :-)

Best Regards,

Akio Takebe

RE: [kvm-ia64-devel] kvm/ia64 TODO List. (Draft).

[Zhang, Xiantao]

Akio Takebe wrote:
> Hi, Xiantao
> 
>> Akio Takebe wrote:
>>> Hi, Xiantao
>>> 
>>> I want to add supporting windows2008.
>> 
>> So, do you have time to work on it ? :)
>> Xiantao
> I don't have enough time now.
> But I'm going to try while my rest time. :-)

OK, Thank you!   That shouldn't be a big problem.  I will add it to the
TODO list and temporarily assign you as the owner :-)
Xiantao

Re: kvm/ia64 TODO List. (Draft).

[Jes Sorensen]

Zhang, Xiantao wrote:
> TODO List (draft):
> 
> 1. Live Migration and Save restore.  (Xiantao is working on it.)  
> 2. Host-memory swapping.  (Good and urgent feature to catch up.)
> 3. virto-io enabling. (Fast IO for kvm guests.)
> 4. Guest reboot and shutdown support. (Basic feature)
> 5. vt-d enabling.  (Maybe share the code with x86 side, but only to
> implement ia64-specific side.  Depend on VT-d platform for ia64)
> 6. superpage support for guest. 
> 7. kvm trace enabling. 

Hi,

IMHO there's other stuff we need to look at as well - I find KVM/ia64
really slow to boot, I don't know if it is my system, but it seems
frustratingly slow :-(

Otherwise I'd like to focus work on qemu and scalability, something we
really need, but which is also shared with x86.

Cheers,
Jes

RE: kvm/ia64 TODO List. (Draft).

[Zhang, Xiantao]

Jes Sorensen wrote:
> Zhang, Xiantao wrote:
>> TODO List (draft):
>> 
>> 1. Live Migration and Save restore.  (Xiantao is working on it.)
>> 2. Host-memory swapping.  (Good and urgent feature to catch up.)
>> 3. virto-io enabling. (Fast IO for kvm guests.)
>> 4. Guest reboot and shutdown support. (Basic feature)
>> 5. vt-d enabling.  (Maybe share the code with x86 side, but only to
>> implement ia64-specific side.  Depend on VT-d platform for ia64)
>> 6. superpage support for guest.
>> 7. kvm trace enabling.
> 
> Hi,

Hi, Jes
     Thanks for your reply!

> IMHO there's other stuff we need to look at as well - I find KVM/ia64
> really slow to boot, I don't know if it is my system, but it seems
> frustratingly slow :-(

It is versy strange! We didn't meet this issue at our side.  Maybe it is
caused by old PAL.  Which version is the machine's PAL?  

> Otherwise I'd like to focus work on qemu and scalability, something we
> really need, but which is also shared with x86.
Agree. We really need effort to identify such tasks and work on them.
Do you have some proposals on this area? :)

Xiantao

Re: kvm/ia64 TODO List. (Draft).

[Jes Sorensen]

Zhang, Xiantao wrote:
> It is versy strange! We didn't meet this issue at our side.  Maybe it is
> caused by old PAL.  Which version is the machine's PAL?  

Hi Xiantao,

This is the output from the host:

[root@mustafa ~]# cat /proc/pal/cpu0/version_info
PAL_vendor : 0xff (min=0xff)
PAL_A      : 01.08 (min\x01.03)
PAL_B      : 01.14 (min\x01.00)

>> Otherwise I'd like to focus work on qemu and scalability, something we
>> really need, but which is also shared with x86.
> Agree. We really need effort to identify such tasks and work on them.
> Do you have some proposals on this area? :)

There's a lot to do really, first priority is to get qemu and qemu-kvm
in sync. Then we need to look at making it more scalable for when
running on larger systems with many I/O devices presented to the guest.
I should have the hardware to work on that though, at some point :-)

Cheers,
Jes

RE: kvm/ia64 TODO List. (Draft).

[Zhang, Xiantao]

Jes Sorensen wrote:
> Zhang, Xiantao wrote:
>> It is versy strange! We didn't meet this issue at our side.  Maybe
>> it is caused by old PAL.  Which version is the machine's PAL?
> 
> Hi Xiantao,
> 
> This is the output from the host:
> 
> [root@mustafa ~]# cat /proc/pal/cpu0/version_info
> PAL_vendor : 0xff (min=0xff)
> PAL_A      : 01.08 (min\x01.03)
> PAL_B      : 01.14 (min\x01.00)

The same dump info as following from my box. Seems your machine are
using the very old PAL which can't support VT-I friendly.  Maybe you
need to upgrade it and try it again :)
[root@vti-xiantao ~]# cat /proc/pal/cpu0/version_info
PAL_vendor : 0xff (min=0xff)
PAL_A      : 07.46 (min\a.00)
PAL_B      : 09.68 (min\x05.00)

>>> Otherwise I'd like to focus work on qemu and scalability, something
>>> we really need, but which is also shared with x86.
>> Agree. We really need effort to identify such tasks and work on them.
>> Do you have some proposals on this area? :)
> There's a lot to do really, first priority is to get qemu and qemu-kvm
> in sync. Then we need to look at making it more scalable for when
> running on larger systems with many I/O devices presented to the
> guest. I should have the hardware to work on that though, at some
> point :-) 
Yes.  We do need to catch up some new features with x86 side in
userspace. Hope your machine to work well with VT-I after you upgrading
the PAL . You know, we still have a lot of work to do for getting a
production-level's kvm, and so we expect more hands to join the
development and contribute on it.   
Xiantao