2.6.30 dom0 Xen patches

[Andrew Lyon]

I have rebased the xen-patches from opensuse
kernel-source-2.6.30-50.1.src.rpm so that they can be applied to
vanilla 2.6.30, the usual patches tarball and gentoo ebuild are
available at http://code.google.com/p/gentoo-xen-kernel/downloads/list
, if you are not using Gentoo simply apply the patches in numeric
order.

Some notes:

both dom0 and domU require CONFIG_SPARSE_IRQ=y and CONFIG_HOTPLUG=y
64 bit dom0/domU requires CONFIG_IA32_EMULATION=y

Andy

_______________________________________________
Xen-devel mailing list
Xen-devel@...
http://lists.xensource.com/xen-devel

[Andrew Lyon]

On Wed, Jul 1, 2009 at 6:15 PM, Andrew Lyon<andrew.lyon@...> wrote:

> I have rebased the xen-patches from opensuse
> kernel-source-2.6.30-50.1.src.rpm so that they can be applied to
> vanilla 2.6.30, the usual patches tarball and gentoo ebuild are
> available at http://code.google.com/p/gentoo-xen-kernel/downloads/list
> , if you are not using Gentoo simply apply the patches in numeric
> order.
>
> Some notes:
>
> both dom0 and domU require CONFIG_SPARSE_IRQ=y and CONFIG_HOTPLUG=y
> 64 bit dom0/domU requires CONFIG_IA32_EMULATION=y
>
> Andy
>

While this kernel seems to work ok as dom0 or domU it seems to be
incompatible with James Harpers gplpv drivers, probably something has
been changed in the networking code and the drivers need to be
updated, I have contacted James about this and I hope we will be able
to work together to figure it out and fix the problem, but if you use
gplpv please hold off upgrading for now!

Andy

_______________________________________________
Xen-devel mailing list
Xen-devel@...
http://lists.xensource.com/xen-devel

[Daniel Schroeder]

Andrew Lyon wrote:

> I have rebased the xen-patches from opensuse
> kernel-source-2.6.30-50.1.src.rpm so that they can be applied to
> vanilla 2.6.30, the usual patches tarball and gentoo ebuild are
> available at http://code.google.com/p/gentoo-xen-kernel/downloads/list
> , if you are not using Gentoo simply apply the patches in numeric
> order.
>
> Some notes:
>
> both dom0 and domU require CONFIG_SPARSE_IRQ=y and CONFIG_HOTPLUG=y
> 64 bit dom0/domU requires CONFIG_IA32_EMULATION=y
>
> Andy

hello Andy,

many thanks for your work.
I have taken your patch set and applied it on top of vanilla
2.6.30.1-rc1 and applied this patch for gcc4.3 bug:

http://xenbits.xen.org/gitweb?p=xenclient/linux-2.6.27-pq.git;a=blob_plain;f=master/clobber-induction-variable.patch;hb=HEAD

for the lazy people out there, here is the distclean tarball:

http://x17.eu/linux-2.6.30.1-rc1-xen.aka.suse-xenified-2.6.30-50.1-rebased.patches.by.andrew.lyon.tar.gz
http://x17.eu/linux-2.6.30.1-rc1-xen.aka.suse-xenified-2.6.30-50.1-rebased.patches.by.andrew.lyon.tar.gz.md5


--
daniel

_______________________________________________
Xen-devel mailing list
Xen-devel@...
http://lists.xensource.com/xen-devel

[Jan Beulich]

>>> Andrew Lyon <andrew.lyon@...> 01.07.09 19:15 >>>
>I have rebased the xen-patches from opensuse
>kernel-source-2.6.30-50.1.src.rpm so that they can be applied to
>vanilla 2.6.30, the usual patches tarball and gentoo ebuild are
>available at http://code.google.com/p/gentoo-xen-kernel/downloads/list 
>, if you are not using Gentoo simply apply the patches in numeric
>order.
>
>Some notes:
>
>both dom0 and domU require CONFIG_SPARSE_IRQ=y and CONFIG_HOTPLUG=y
>64 bit dom0/domU requires CONFIG_IA32_EMULATION=y

May I ask why you found the first and third ones to be necessary? Unless
it's because of the conditional in the loop right after the commen "Zap
stats across IRQ changes of use." in unbind_from_irq(), which I fixed a
few days ago, but didn't push out to our internal repo so far. (The fix
is to change the dependency from CONFIG_SPARSE_IRQ to
CONFIG_GENERIC_HARDIRQS in the 2.6.30 patch.)

The second one seems less meaningful, as I wouldn't think anyone would
build a general use (Xen) kernel with EMBEDDED=y, and HOTPLUG is not
user selectable without EMBEDDED.

May I also ask for you to let us know about fixes to the patches that you
find necessary which aren't just due to you applying them to plain 2.6.30?

Thanks, Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@...
http://lists.xensource.com/xen-devel

[Daniel Schroeder]

<where>
http://x17.eu/linux-2.6.30.1-rc2-xen.aka.suse-xenified-2.6.30-50.1-rebased.patches.by.andrew.lyon.tar.gz
http://x17.eu/linux-2.6.30.1-rc2-xen.aka.suse-xenified-2.6.30-50.1-rebased.patches.by.andrew.lyon.tar.gz.md5
</where>

<what>
Andrew Lyon wrote:

> I have rebased the xen-patches from opensuse
> kernel-source-2.6.30-50.1.src.rpm so that they can be applied to
> vanilla 2.6.30, the usual patches tarball and gentoo ebuild are
> available at http://code.google.com/p/gentoo-xen-kernel/downloads/list
> , if you are not using Gentoo simply apply the patches in numeric
> order.
>
> Some notes:
>
> both dom0 and domU require CONFIG_SPARSE_IRQ=y and CONFIG_HOTPLUG=y
> 64 bit dom0/domU requires CONFIG_IA32_EMULATION=y
>
> Andy

plus

http://xenbits.xen.org/gitweb?p=xenclient/linux-2.6.27-pq.git;a=blob_plain;f=master/clobber-induction-variable.patch;hb=HEAD
</what>


--
daniel

_______________________________________________
Xen-devel mailing list
Xen-devel@...
http://lists.xensource.com/xen-devel

[dbaxps]

Not sure i did right thing, i've edited 60036_xen3-patch-2.6.30.patch1 as follows:-

@@ -3957,7 +3982,29 @@ void __init probe_nr_irqs_gsi(void)
 
     printk(KERN_DEBUG "nr_irqs_gsi: %d\n", nr_irqs_gsi);
 }
+
+#ifdef CONFIG_GENERIC_HARDIRQS
+int __init arch_probe_nr_irqs(void)
+{
+    int nr;
+
+    if (nr_irqs > (NR_VECTORS * nr_cpu_ids))
+        nr_irqs = NR_VECTORS * nr_cpu_ids;
+
+    nr = nr_irqs_gsi + 8 * nr_cpu_ids;
+#if defined(CONFIG_PCI_MSI) || defined(CONFIG_HT_IRQ)
+    /*
+     * for MSI and HT dyn irq
+     */
+    nr += nr_irqs_gsi * 16;
 #endif

allows to compile unbind_from_irq() in ./linux-2drivers/xen/core/evtchn.c with no errors and xenified kernel been built seems to work OK
Make menuconfig didn't require any changes to enable Xen Dom0 support .

In .config :-

Before build CONFIG_SPARSE_IRQ=y
After build # CONFIG_SPARSE_IRQ is not set

Boris

--- On Thu, 7/2/09, Jan Beulich <JBeulich@...> wrote:

From: Jan Beulich <JBeulich@...>
Subject: Re: [Xen-devel] 2.6.30 dom0 Xen patches
To: "Andrew Lyon" <andrew.lyon@...>
Cc: "Xen-devel" <xen-devel@...>
Date: Thursday, July 2, 2009, 2:53 AM

>>> Andrew Lyon <andrew.lyon@...> 01.07.09 19:15 >>>

>I have rebased the xen-patches from opensuse
>kernel-source-2.6.30-50.1.src.rpm so that they can be applied to
>vanilla 2.6.30, the usual patches tarball and gentoo ebuild are
>available at http://code.google.com/p/gentoo-xen-kernel/downloads/list
>, if you are not using Gentoo simply apply the patches in numeric
>order.
>
>Some notes:
>
>both dom0 and domU require CONFIG_SPARSE_IRQ=y and CONFIG_HOTPLUG=y
>64 bit dom0/domU requires CONFIG_IA32_EMULATION=y

May I ask why you found the first and third ones to be necessary? Unless
it's because of the conditional in the loop right after the commen "Zap
stats across IRQ changes of use." in unbind_from_irq(), which I fixed a
few days ago, but didn't push out to our internal repo so far. (The fix
is to change the dependency from CONFIG_SPARSE_IRQ to
CONFIG_GENERIC_HARDIRQS in the 2.6.30 patch.)

The second one seems less meaningful, as I wouldn't think anyone would
build a general use (Xen) kernel with EMBEDDED=y, and HOTPLUG is not
user selectable without EMBEDDED.

May I also ask for you to let us know about fixes to the patches that you
find necessary which aren't just due to you applying them to plain 2.6.30?

Thanks, Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@...
http://lists.xensource.com/xen-devel

_______________________________________________
Xen-devel mailing list
Xen-devel@...
http://lists.xensource.com/xen-devel

[Daniel Schroeder]

i have created a directory, so you can check this http://x17.eu/xen/ url
after a new vanilla stable release for updates...

<where>
http://x17.eu/xen/
http://x17.eu/xen/linux-2.6.29.6-xen.aka.suse-xenified-2.6.29-62.1-rebased.patches.by.andrew.lyon.tar.gz
http://x17.eu/xen/linux-2.6.29.6-xen.aka.suse-xenified-2.6.29-62.1-rebased.patches.by.andrew.lyon.tar.gz.md5
http://x17.eu/xen/linux-2.6.30.1-xen.aka.suse-xenified-2.6.30-50.1-rebased.patches.by.andrew.lyon.tar.gz
http://x17.eu/xen/linux-2.6.30.1-xen.aka.suse-xenified-2.6.30-50.1-rebased.patches.by.andrew.lyon.tar.gz.md5
</where>


<what>
Andrew Lyon wrote:

> I have rebased the xen-patches from opensuse
> kernel-source-2.6.30-50.1.src.rpm so that they can be applied to
> vanilla 2.6.30, the usual patches tarball and gentoo ebuild are
> available at http://code.google.com/p/gentoo-xen-kernel/downloads/list
> , if you are not using Gentoo simply apply the patches in numeric
> order.
>
> Some notes:
>
> both dom0 and domU require CONFIG_SPARSE_IRQ=y and CONFIG_HOTPLUG=y
> 64 bit dom0/domU requires CONFIG_IA32_EMULATION=y
>
> Andy

plus

http://xenbits.xen.org/gitweb?p=xenclient/linux-2.6.27-pq.git;a=blob_plain;f=master/clobber-induction-variable.patch;hb=HEAD
</what>


--
daniel


_______________________________________________
Xen-devel mailing list
Xen-devel@...
http://lists.xensource.com/xen-devel

[Pasi Kärkkäinen]

On Thu, Jul 02, 2009 at 07:53:09AM +0100, Jan Beulich wrote:

> >>> Andrew Lyon <andrew.lyon@...> 01.07.09 19:15 >>>
> >I have rebased the xen-patches from opensuse
> >kernel-source-2.6.30-50.1.src.rpm so that they can be applied to
> >vanilla 2.6.30, the usual patches tarball and gentoo ebuild are
> >available at http://code.google.com/p/gentoo-xen-kernel/downloads/list 
> >, if you are not using Gentoo simply apply the patches in numeric
> >order.
> >
> >Some notes:
> >
> >both dom0 and domU require CONFIG_SPARSE_IRQ=y and CONFIG_HOTPLUG=y
> >64 bit dom0/domU requires CONFIG_IA32_EMULATION=y
>
> May I ask why you found the first and third ones to be necessary? Unless
> it's because of the conditional in the loop right after the commen "Zap
> stats across IRQ changes of use." in unbind_from_irq(), which I fixed a
> few days ago, but didn't push out to our internal repo so far. (The fix
> is to change the dependency from CONFIG_SPARSE_IRQ to
> CONFIG_GENERIC_HARDIRQS in the 2.6.30 patch.)
>
> The second one seems less meaningful, as I wouldn't think anyone would
> build a general use (Xen) kernel with EMBEDDED=y, and HOTPLUG is not
> user selectable without EMBEDDED.
>
> May I also ask for you to let us know about fixes to the patches that you
> find necessary which aren't just due to you applying them to plain 2.6.30?
>

Hello Jan,

I assume Novell did a lot of testing/patching/bugfixing for SLES11 2.6.27
Xen kernel. Do these OpenSUSE Xen patches contain the fixes from SLES11?

-- Pasi

_______________________________________________
Xen-devel mailing list
Xen-devel@...
http://lists.xensource.com/xen-devel

[Jan Beulich]

>>> Pasi Kärkkäinen<pasik@...> 06.07.09 13:02 >>>
>I assume Novell did a lot of testing/patching/bugfixing for SLES11 2.6.27
>Xen kernel. Do these OpenSUSE Xen patches contain the fixes from SLES11?

Yes, OpenSuSE and SLE11 have been sharing the kernel approximately since
the release of the first kernel maintenance update for SLE11.

Jan

_______________________________________________
Xen-devel mailing list
Xen-devel@...
http://lists.xensource.com/xen-devel

[Tim Post-3]

Hi,

On Mon, 2009-07-06 at 14:04 +0100, Jan Beulich wrote:
> >>> Pasi Kärkkäinen<pasik@...> 06.07.09 13:02 >>>
> >I assume Novell did a lot of testing/patching/bugfixing for SLES11 2.6.27
> >Xen kernel. Do these OpenSUSE Xen patches contain the fixes from SLES11?
>
> Yes, OpenSuSE and SLE11 have been sharing the kernel approximately since
> the release of the first kernel maintenance update for SLE11.
>
> Jan

Is there a consensus now on where the 2.6.30-xen tree is going to live?
I saw a mention of space on xenbits, but did not see a consensus. Did I
miss one in the thread?

No rush or anything, I'm just catching up to the thread and feel as if I
missed a reply or two.

Cheers,
--Tim


_______________________________________________
Xen-devel mailing list
Xen-devel@...
http://lists.xensource.com/xen-devel

[dbaxps]

To build xenified 2.6.30.1 via Andy's 2.6.30 pacth set in it's current state :-

1. Created log1 via script
rm -f log1
touch -f log1
for V in `ls 6*`
do
echo $V >> log1
cat $V|grep CONFIG_SPARSE_IRQ >> log1
done

2. Analyzed log1

60035_xen3-patch-2.6.29.patch1
+# if !defined(CONFIG_SPARSE_IRQ) && NR_CPUS < MAX_IO_APICS
+# elif defined(CONFIG_SPARSE_IRQ) && 8 * NR_CPUS > 32 * MAX_IO_APICS
+#ifdef CONFIG_SPARSE_IRQ
+#ifdef CONFIG_SPARSE_IRQ
+#ifdef CONFIG_SPARSE_IRQ
+#ifdef CONFIG_SPARSE_IRQ
+#ifdef CONFIG_SPARSE_IRQ
+#ifdef CONFIG_SPARSE_IRQ
+#ifndef CONFIG_SPARSE_IRQ

60036_xen3-patch-2.6.30.patch1
-# if !defined(CONFIG_SPARSE_IRQ) && NR_CPUS < MAX_IO_APICS
-# elif defined(CONFIG_SPARSE_IRQ) && 8 * NR_CPUS > 32 * MAX_IO_APICS
+# elif defined(CONFIG_SPARSE_IRQ)


3. Made global replacement CONFIG_SPARSE_IRQ by CONFIG_GENERIC_HARDIRQS in files
60035_xen3-patch-2.6.29.patch1,60036_xen3-patch-2.6.30.patch1

4.Untared kernel-2.6.30.1 tar.bz2 and applied patches. Built 2.6.30.1 xenified kernel with no errors

Loaded Xen 3.4.1-rc6 Dom0 with xenified 2.6.30.1 (been built) on top of Ubuntu 9.04 Server
Tested OSOL 2009.06 loaded with PV Drivers in stub domain OK , also tested OSOL 2009.06 PV DomU and F11 PV DomU.

Boris


--- On Thu, 7/2/09, Jan Beulich <JBeulich@...> wrote:

From: Jan Beulich <JBeulich@...>
Subject: Re: [Xen-devel] 2.6.30 dom0 Xen patches
To: "Andrew Lyon" <andrew.lyon@...>
Cc: "Xen-devel" <xen-devel@...>
Date: Thursday, July 2, 2009, 2:53 AM

>>> Andrew Lyon <andrew.lyon@...> 01.07.09 19:15 >>>

>I have rebased the xen-patches from opensuse
>kernel-source-2.6.30-50.1.src.rpm so that they can be applied to
>vanilla 2.6.30, the usual patches tarball and gentoo ebuild are
>available at http://code.google.com/p/gentoo-xen-kernel/downloads/list
>, if you are not using Gentoo simply apply the patches in numeric
>order.
>
>Some notes:
>
>both dom0 and domU require CONFIG_SPARSE_IRQ=y and CONFIG_HOTPLUG=y
>64 bit dom0/domU requires CONFIG_IA32_EMULATION=y

May I ask why you found the first and third ones to be necessary? Unless
it's because of the conditional in the loop right after the commen "Zap
stats across IRQ changes of use." in unbind_from_irq(), which I fixed a
few days ago, but didn't push out to our internal repo so far. (The fix
is to change the dependency from CONFIG_SPARSE_IRQ to
CONFIG_GENERIC_HARDIRQS in the 2.6.30 patch.)

The second one seems less meaningful, as I wouldn't think anyone would
build a general use (Xen) kernel with EMBEDDED=y, and HOTPLUG is not
user selectable without EMBEDDED.

May I also ask for you to let us know about fixes to the patches that you
find necessary which aren't just due to you applying them to plain 2.6.30?

Thanks, Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@...
http://lists.xensource.com/xen-devel

_______________________________________________
Xen-devel mailing list
Xen-devel@...
http://lists.xensource.com/xen-devel

[dbaxps]

The schema bellow has been tested positive on Core2Duo E8400,
E6600 boxes and caused dropping into stack trace
when loading under Xen 3.4.1-rc6 Hypervisor on Q9550 box.
What reminds an issue had been raised up for PVOPS kernel
in march or april 2009 and finally, fixed by Jeremy Fitzhradinge.
However , building kernel without changing dependency
with CONFIG_SPARSE_IRQ=y in .config file allows to load
kernel on Q9550 box under Xen 3.4.1 as well as on Core2Duo
boxes.

Boris.

--- On Tue, 7/7/09, Boris Derzhavets <bderzhavets@...> wrote:

From: Boris Derzhavets <bderzhavets@...>
Subject: Re: [Xen-devel] 2.6.30 dom0 Xen patches
To: "Andrew Lyon" <andrew.lyon@...>, "Jan Beulich" <JBeulich@...>
Cc: "Xen-devel" <xen-devel@...>
Date: Tuesday, July 7, 2009, 1:12 PM

To build xenified 2.6.30.1 via Andy's 2.6.30 pacth set in it's current state :-

1. Created log1 via script
rm -f log1
touch -f log1
for V in `ls 6*`
do
echo $V >> log1
cat $V|grep CONFIG_SPARSE_IRQ >> log1
done

2. Analyzed log1

60035_xen3-patch-2.6.29.patch1
+# if !defined(CONFIG_SPARSE_IRQ) && NR_CPUS < MAX_IO_APICS
+# elif defined(CONFIG_SPARSE_IRQ) && 8 * NR_CPUS > 32 * MAX_IO_APICS
+#ifdef CONFIG_SPARSE_IRQ
+#ifdef CONFIG_SPARSE_IRQ
+#ifdef CONFIG_SPARSE_IRQ
+#ifdef CONFIG_SPARSE_IRQ
+#ifdef CONFIG_SPARSE_IRQ
+#ifdef CONFIG_SPARSE_IRQ
+#ifndef CONFIG_SPARSE_IRQ

60036_xen3-patch-2.6.30.patch1
-# if !defined(CONFIG_SPARSE_IRQ) && NR_CPUS < MAX_IO_APICS
-# elif defined(CONFIG_SPARSE_IRQ) && 8 * NR_CPUS > 32 * MAX_IO_APICS
+# elif defined(CONFIG_SPARSE_IRQ)


3. Made global replacement CONFIG_SPARSE_IRQ by CONFIG_GENERIC_HARDIRQS in files
60035_xen3-patch-2.6.29.patch1,60036_xen3-patch-2.6.30.patch1

4.Untared kernel-2.6.30.1 tar.bz2 and applied patches. Built 2.6.30.1 xenified kernel with no errors

Loaded Xen 3.4.1-rc6 Dom0 with xenified 2.6.30.1 (been built) on top of Ubuntu 9.04 Server
Tested OSOL 2009.06 loaded with PV Drivers in stub domain OK , also tested OSOL 2009.06 PV DomU and F11 PV DomU.

Boris


--- On Thu, 7/2/09, Jan Beulich <JBeulich@...> wrote:

From: Jan Beulich <JBeulich@...>
Subject: Re: [Xen-devel] 2.6.30 dom0 Xen patches
To: "Andrew Lyon" <andrew.lyon@...>
Cc: "Xen-devel" <xen-devel@...>
Date: Thursday, July 2, 2009, 2:53 AM

>>> Andrew Lyon <andrew.lyon@...> 01.07.09 19:15 >>>

>I have rebased the xen-patches from opensuse
>kernel-source-2.6.30-50.1.src.rpm so that they can be applied to
>vanilla 2.6.30, the usual patches tarball and gentoo ebuild are
>available at http://code.google.com/p/gentoo-xen-kernel/downloads/list
>, if you are not using Gentoo simply apply the patches in numeric
>order.
>
>Some notes:
>
>both dom0 and domU require CONFIG_SPARSE_IRQ=y and CONFIG_HOTPLUG=y
>64 bit dom0/domU requires CONFIG_IA32_EMULATION=y

May I ask why you found the first and third ones to be necessary? Unless
it's because of the conditional in the loop right after the commen "Zap
stats across IRQ changes of use." in unbind_from_irq(), which I fixed a
few days ago, but didn't push out to our internal repo so far. (The fix
is to change the dependency from CONFIG_SPARSE_IRQ to
CONFIG_GENERIC_HARDIRQS in the 2.6.30 patch.)

The second one seems less meaningful, as I wouldn't think anyone would
build a general use (Xen) kernel with EMBEDDED=y, and HOTPLUG is not
user selectable without EMBEDDED.

May I also ask for you to let us know about fixes to the patches that you
find necessary which aren't just due to you applying them to plain 2.6.30?

Thanks, Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@...
http://lists.xensource.com/xen-devel

-----Inline Attachment Follows-----

_______________________________________________
Xen-devel mailing list
Xen-devel@...
http://lists.xensource.com/xen-devel

_______________________________________________
Xen-devel mailing list
Xen-devel@...
http://lists.xensource.com/xen-devel

[Jan Beulich]

>>> Boris Derzhavets <bderzhavets@...> 07.07.09 19:12 >>>
>3. Made global replacement CONFIG_SPARSE_IRQ by CONFIG_GENERIC_HARDIRQS in files
>60035_xen3-patch-2.6.29.patch1,60036_xen3-patch-2.6.30.patch1

This certainly doesn't sound right.

Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@...
http://lists.xensource.com/xen-devel

[dbaxps]

OK. What's the right thing to do ?
Please, advise.

Boris.

--- On Fri, 7/10/09, Jan Beulich <JBeulich@...> wrote:

From: Jan Beulich <JBeulich@...>
Subject: Re: [Xen-devel] 2.6.30 dom0 Xen patches
To: "Andrew Lyon" <andrew.lyon@...>, "Boris Derzhavets" <bderzhavets@...>
Cc: "Xen-devel" <xen-devel@...>
Date: Friday, July 10, 2009, 4:00 AM

>>> Boris Derzhavets <bderzhavets@...> 07.07.09 19:12 >>>
>3. Made global replacement CONFIG_SPARSE_IRQ by CONFIG_GENERIC_HARDIRQS in files
>60035_xen3-patch-2.6.29.patch1,60036_xen3-patch-2.6.30.patch1

This certainly doesn't sound right.

Jan

_______________________________________________
Xen-devel mailing list
Xen-devel@...
http://lists.xensource.com/xen-devel

[Jan Beulich]

I outlined the solution in the mail you responded to here. It's a single place that needs changing, afaict. Jan

>>> Boris Derzhavets <bderzhavets@...> 10.07.09 10:09 >>>
OK. What's the right thing to do ?
Please, advise.

Boris.

--- On Fri, 7/10/09, Jan Beulich <JBeulich@...> wrote:

From: Jan Beulich <JBeulich@...>
Subject: Re: [Xen-devel] 2.6.30 dom0 Xen patches
To: "Andrew Lyon" <andrew.lyon@...>, "Boris Derzhavets" <bderzhavets@...>
Cc: "Xen-devel" <xen-devel@...>
Date: Friday, July 10, 2009, 4:00 AM

>>> Boris Derzhavets <bderzhavets@...> 07.07.09 19:12 >>>
>3. Made global replacement CONFIG_SPARSE_IRQ by CONFIG_GENERIC_HARDIRQS in files
>60035_xen3-patch-2.6.29.patch1,60036_xen3-patch-2.6.30.patch1

This certainly doesn't sound right.

Jan




     


_______________________________________________
Xen-devel mailing list
Xen-devel@...
http://lists.xensource.com/xen-devel

[dbaxps]

Actually, manually setting  CONFIG_SPARSE_IRQ=y in .config also allows
to build working xenified 2.6.30.1 kernel for C2D & Q9550 CPUs .
But, you've also criticize this action. If it's not a secret ,please, explain how to change dependency in 2.6.30 patch properly ?

Boris.

--- On Fri, 7/10/09, Boris Derzhavets <bderzhavets@...> wrote:

From: Boris Derzhavets <bderzhavets@...>
Subject: Re: [Xen-devel] 2.6.30 dom0 Xen patches
To: "Andrew Lyon" <andrew.lyon@...>, "Jan Beulich" <JBeulich@...>
Cc: "Xen-devel" <xen-devel@...>
Date: Friday, July 10, 2009, 4:09 AM

OK. What's the right thing to do ?
Please, advise.

Boris.

--- On Fri, 7/10/09, Jan Beulich <JBeulich@...> wrote:

From: Jan Beulich <JBeulich@...>
Subject: Re: [Xen-devel] 2.6.30 dom0 Xen patches
To: "Andrew Lyon" <andrew.lyon@...>, "Boris Derzhavets" <bderzhavets@...>
Cc: "Xen-devel" <xen-devel@...>
Date: Friday, July 10, 2009, 4:00 AM

>>> Boris Derzhavets <bderzhavets@...> 07.07.09 19:12 >>>
>3. Made global replacement CONFIG_SPARSE_IRQ by CONFIG_GENERIC_HARDIRQS in files
>60035_xen3-patch-2.6.29.patch1,60036_xen3-patch-2.6.30.patch1

This certainly doesn't sound right.

Jan

-----Inline Attachment Follows-----

_______________________________________________
Xen-devel mailing list
Xen-devel@...
http://lists.xensource.com/xen-devel

_______________________________________________
Xen-devel mailing list
Xen-devel@...
http://lists.xensource.com/xen-devel

[Jan Beulich]

>>> Boris Derzhavets <bderzhavets@...> 10.07.09 10:26 >>>
>Actually, manually setting  CONFIG_SPARSE_IRQ=y in .config also allows
>to build working xenified 2.6.30.1 kernel for C2D & Q9550 CPUs .
>But, you've also criticize this action.

I didn't. All I said is that the build succeeding shouldn't depend on what value
you select here.

>If it's not a secret ,please, explain how to change dependency in 2.6.30
>patch properly ?

Quoting my earlier mail:

>> May I ask why you found the first and third ones to be necessary? Unless
>> it's because of the conditional in the loop right after the commen "Zap
>> stats across IRQ changes of use." in unbind_from_irq(), which I fixed a
>> few days ago, but didn't push out to our internal repo so far. (The fix
>> is to change the dependency from CONFIG_SPARSE_IRQ to
>> CONFIG_GENERIC_HARDIRQS in the 2.6.30 patch.)

And there I meant *only* this one place, not a general replacement
everywhere in the patches (and in particular not in the .29 one).

If that isn't enough, you'll have to let me know what build issue you still
encounter.

Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@...
http://lists.xensource.com/xen-devel

[dbaxps]

The first time and  it's in this tread (about week ago), captured entries and selected the only one 2.6.30 patch file follow you instruction,
as i understood it - 60036_xen3-patch-2.6.30.patch1

All entries found :-

-# if !defined(CONFIG_SPARSE_IRQ) && NR_CPUS < MAX_IO_APICS
-# elif defined(CONFIG_SPARSE_IRQ) && 8 * NR_CPUS > 32 * MAX_IO_APICS
+# elif defined(CONFIG_SPARSE_IRQ)
+#ifdef CONFIG_SPARSE_IRQ

I've just replaced in 60036_xen3-patch-2.6.30.patch1

+#ifdef CONFIG_SPARSE_IRQ
   by
+#ifdef CONFIG_GENERIC_HARDIRQS

due to my understanding term dependency ( it might be wrong )

Then applied Andy's patch set and built 2.6.30 xenified kernel with no errors.
This kernel at runtime has serious bugs with loading PV drivers in HVM
and STUB domains. So, is pretty much useless

In case of 2.6.30.1 vanilla kernel this step doesn't help anymore.
Compilation fails in same place ( same function).

Even been compiled with CONFIG_SPARSE_IRQ=y in .config 2.6.30.1
xenified kernel doesn't have bug mentioned above and it's very important.

Thank you.
Boris

--- On Fri, 7/10/09, Jan Beulich <JBeulich@...> wrote:

From: Jan Beulich <JBeulich@...>
Subject: Re: [Xen-devel] 2.6.30 dom0 Xen patches
To: "Andrew Lyon" <andrew.lyon@...>, "Boris Derzhavets" <bderzhavets@...>
Cc: "Xen-devel" <xen-devel@...>
Date: Friday, July 10, 2009, 4:55 AM

>>> Boris Derzhavets <bderzhavets@...> 10.07.09 10:26 >>>
>Actually, manually setting  CONFIG_SPARSE_IRQ=y in .config also allows
>to build working xenified 2.6.30.1 kernel for C2D & Q9550 CPUs .
>But, you've also criticize this action.

I didn't. All I said is that the build succeeding shouldn't depend on what value
you select here.

>If it's not a secret ,please, explain how to change dependency in 2.6.30
>patch properly ?

Quoting my earlier mail:

>> May I ask why you found the first and third ones to be necessary? Unless
>> it's because of the conditional in the loop right after the commen "Zap
>> stats across IRQ changes of use." in unbind_from_irq(), which I fixed a
>> few days ago, but didn't push out to our internal repo so far. (The fix
>> is to change the dependency from CONFIG_SPARSE_IRQ to
>> CONFIG_GENERIC_HARDIRQS in the 2.6.30 patch.)

And there I meant *only* this one place, not a general replacement
everywhere in the patches (and in particular not in the .29 one).

If that isn't enough, you'll have to let me know what build issue you still
encounter.

Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@...
http://lists.xensource.com/xen-devel

_______________________________________________
Xen-devel mailing list
Xen-devel@...
http://lists.xensource.com/xen-devel

[dbaxps]

I still believe in 60035_xen3-patch-2.6.29.patch1 line 11568 

+#ifdef CONFIG_SPARSE_IRQ
+                       irq_to_desc(irq)->kstat_irqs[cpu] = 0;
+#else
                        kstat_cpu(cpu).irqs[irq] = 0;

causes a problem.
Working xenified 2.6.30.1 kernel may be built changing one more dependency

Boris.

--- On Fri, 7/10/09, Boris Derzhavets <bderzhavets@...> wrote:

From: Boris Derzhavets <bderzhavets@...>
Subject: Re: [Xen-devel] 2.6.30 dom0 Xen patches
To: "Andrew Lyon" <andrew.lyon@...>, "Jan Beulich" <JBeulich@...>
Cc: "Xen-devel" <xen-devel@...>
Date: Friday, July 10, 2009, 5:31 AM

The first time and  it's in this tread (about week ago), captured entries and selected the only one 2.6.30 patch file follow you instruction,
as i understood it - 60036_xen3-patch-2.6.30.patch1

All entries found :-

-# if !defined(CONFIG_SPARSE_IRQ) && NR_CPUS < MAX_IO_APICS
-# elif defined(CONFIG_SPARSE_IRQ) && 8 * NR_CPUS > 32 * MAX_IO_APICS
+# elif defined(CONFIG_SPARSE_IRQ)
+#ifdef CONFIG_SPARSE_IRQ

I've just replaced in 60036_xen3-patch-2.6.30.patch1

+#ifdef CONFIG_SPARSE_IRQ
   by
+#ifdef CONFIG_GENERIC_HARDIRQS

due to my understanding term dependency ( it might be wrong )

Then applied Andy's patch set and built 2.6.30 xenified kernel with no errors.
This kernel at runtime has serious bugs with loading PV drivers in HVM
and STUB domains. So, is pretty much useless

In case of 2.6.30.1 vanilla kernel this step doesn't help anymore.
Compilation fails in same place ( same function).

Even been compiled with CONFIG_SPARSE_IRQ=y in .config 2.6.30.1
xenified kernel doesn't have bug mentioned above and it's very important.

Thank you.
Boris

--- On Fri, 7/10/09, Jan Beulich <JBeulich@...> wrote:

From: Jan Beulich <JBeulich@...>
Subject: Re: [Xen-devel] 2.6.30 dom0 Xen patches
To: "Andrew Lyon" <andrew.lyon@...>, "Boris Derzhavets" <bderzhavets@...>
Cc: "Xen-devel" <xen-devel@...>
Date: Friday, July 10, 2009, 4:55 AM

>>> Boris Derzhavets <bderzhavets@...> 10.07.09 10:26 >>>
>Actually, manually setting  CONFIG_SPARSE_IRQ=y in .config also allows
>to build working xenified 2.6.30.1 kernel for C2D & Q9550 CPUs .
>But, you've also criticize this action.

I didn't. All I said is that the build succeeding shouldn't depend on what value
you select here.

>If it's not a secret ,please, explain how to change dependency in 2.6.30
>patch properly ?

Quoting my earlier mail:

>> May I ask why you found the first and third ones to be necessary? Unless
>> it's because of the conditional in the loop right after the commen "Zap
>> stats across IRQ changes of use." in unbind_from_irq(), which I fixed a
>> few days ago, but didn't push out to our internal repo so far. (The fix
>> is to change the dependency from CONFIG_SPARSE_IRQ to
>> CONFIG_GENERIC_HARDIRQS in the 2.6.30 patch.)

And there I meant *only* this one place, not a general replacement
everywhere in the patches (and in particular not in the .29 one).

If that isn't enough, you'll have to let me know what build issue you still
encounter.

Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@...
http://lists.xensource.com/xen-devel

-----Inline Attachment Follows-----

_______________________________________________
Xen-devel mailing list
Xen-devel@...
http://lists.xensource.com/xen-devel

_______________________________________________
Xen-devel mailing list
Xen-devel@...
http://lists.xensource.com/xen-devel

[Jan Beulich]

>>> Boris Derzhavets <bderzhavets@...> 10.07.09 14:40 >>>
>I still believe in 60035_xen3-patch-2.6.29.patch1 line 11568  
>
>+#ifdef CONFIG_SPARSE_IRQ
>+                       irq_to_desc(irq)->kstat_irqs[cpu] = 0;
>+#else
>                        kstat_cpu(cpu).irqs[irq] = 0;
>
>causes a problem.

Sure - that's what I keep telling you: This needs to be replaced by
CONFIG_GENERIC_HARDIRQS in the .30 patch.

Oh, I think I now understand what you first did: You apparently replaced
the *other* +#ifdef CONFIG_SPARSE_IRQ in the .30 patch - that's wrong.
You need to make the .30 patch *additionally* make the change indicated
above. (Of course, you could do that in the .29 patch right away, but then
your .29 patch will no longer properly reflect what's needed for that version
to build and/or work right.)

Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@...
http://lists.xensource.com/xen-devel