Linux 3.0-rc1 - what Xen components went in.

Linux 3.0-rc1 - what Xen components went in.

[Konrad Rzeszutek Wilk]

3.0-rc1 (2.6.40-rc1) is out and while in the past it took us all the way up to
rc7 to have it working under Xen, this is not case with this release!

And my VP summarized this event in this blog:
http://blogs.oracle.com/wim/entry/linux_mainline_contains_all_the

But if you want to get more technical:

Features that went in:
 - cleancache shim to Xen Transcendent Memory. In other words, you can use tmem
   (not yet in, but as an outside tree), or zcache (in the staging directory) to
   speed up your I/Os. Details on how to use it:
   http://oss.oracle.com/projects/tmem/dist/documentation/internals/linuxpatch

 - The usage of VM_IO to distinguish between pages that are "System RAM" or
   "PCI" is slowly going away. While it worked in the past it also brought us
   lots of bugs that took time to get right. The new mechanism, based on the P2M
   1-1 mapping bypasses the need for the VM_IO flag. Still have some more work..

 - Updates to the backend code so that Xen PCI backend can easily run against
   the kernel (Note: xen pciback has not yet been merged)

 - Lots of cleanups done in the general code and in the balloon code. The latter preps
   the code base for using the more generic memory hotplugging.

 - IRQ rework. In the past our IRQ code was bit of dance. We had to distinguish between
   the different types of interrupts (think edge vs level) coming from the Xen hypervisor.
   Except that we weren't doing it cleanly and were overloading the chip functions that dealt
   with this and causing us to do do some extra hack^H^H^H fancy unmasking/masking. Gone!

 - Xen block backend. Yes, it is in!

Bugs:
 - Only four of them and they are also in the 2.6.39.1 stable tree.

That is it... rc1 is starting and please report any bugs to the mailing list.

The shortlog is as follow:

Chris Lalancette (1):
      xen/blkback: Flush blkback data when connecting.

Dan Magenheimer (1):
      xen: cleancache shim to Xen Transcendent Memory

Daniel De Graaf (1):
      xen/gntdev,gntalloc: Remove unneeded VM flags

Daniel Kiper (13):
      xen/balloon: Use PageHighMem() for high memory page detection
      xen/balloon: Simplify HVM integration
      xen/balloon: Clarify credit calculation
      xen/balloon: Move dec_totalhigh_pages() from __balloon_append() to balloon_append()
      xen/setup: Fix for incorrect xen_extra_mem_start initialization under 32-bit
      arch/x86/xen/irq: Cleanup code/data sections definitions
      arch/x86/xen/enlighten: Cleanup code/data sections definitions
      arch/x86/xen/setup: Cleanup code/data sections definitions
      arch/x86/xen/mmu: Cleanup code/data sections definitions
      arch/x86/xen/xen-ops: Cleanup code/data sections definitions
      arch/x86/xen/time: Cleanup code/data sections definitions
      arch/x86/xen/smp: Cleanup code/data sections definitions
      drivers/xen/sys-hypervisor: Cleanup code/data sections definitions

Eric Dumazet (1):
      xen: netfront: fix declaration order


Ian Campbell (6):
      xen: netfront: assume all hw features are available until backend connection setup
      Signed bit field; int have_hotplug_status_watch:1
      xen: rename blkbk module xen-blkback.
      xen: tidy up whitespace in drivers/xen/Makefile
      xen: netback: use __CONST_RING_SIZE not __RING_SIZE
      xen: netfront: hold RTNL when updating features.

Jan Beulich (2):
      xen/blkback: simplify address translations
      xen/blkback: don't fail empty barrier requests

Jeremy Fitzhardinge (14):
      xen-blkback-porting
      xen/blkback: don't include xen/evtchn.h
      xen/blkback: deal with hardsect_size to logical_block_size rename
      xen/blkback: remove spurious debug output noise
      xen/blkback: little cleanups
      xen/blkback: use drv_get/set_drvdata rather than directly accessing driver_data.
      xen/blkback: add accessor for xenbus backend device
      xen: drop all the special iomap pte paths.
      xen: use mmu_update for xen_set_pte_at()
      xen: condense everything onto xen_set_pte
      xen: make a pile of mmu pvop functions static
      xen: use normal virt_to_machine for ptes
      xen/mmu: remove all ad-hoc stats stuff
      xen: fix compile without CONFIG_XEN_DEBUG_FS

John Stultz (1):
      x86: Convert remaining x86 clocksources to clocksource_register_hz/khz

K. Y. Srinivasan (3):
      xen/blkback: Propagate changed size of VBDs
      xen/blkback: Print additional information when a vbd is resized.


Keir Fraser (1):
      blkback: Fix CVE-2010-3699

Konrad Rzeszutek Wilk (68):
      xen/pci: Add xen_[find|register|unregister]_device_domain_owner functions.
      xen/irq: Check if the PCI device is owned by a domain different than DOMID_SELF.
      xen/irq: Add support to check if IRQ line is shared with other domains.
      xen/irq: Export 'xen_pirq_from_irq' function.
      xen/irq: The Xen hypervisor cleans up the PIRQs if the other domain forgot.
      xen: add blkback support
      Fix compile warnings: ignoring return value of 'xenbus_register_backend' ..
      xen/blkback: Replace WRITE_BARRIER with (REQ_FLUSH | REQ_FUA)
      xen/blkback: Update to use blkdev_get_by_dev instead of open_by_devnum.
      xen/blkback: Move global/static variables into struct xen_blkbk.
      xen/blkback: Union the blkif_request request specific fields
      xen/blkback: Use 'vzalloc' for page arrays and pre-allocate pages.
      xen/blkback: Utilize the M2P override mechanism for GNTMAP_host_map
      xen/blkback: Use kzalloc's, and GFP_KERNEL for data structures.
      xen/blkback: Add some comments.
      xen/blkback: blkif->struct blkif_st
      xen/blkback: Fix checkpatch warnings in vbd.c
      xen/blkback: Fix interface.c checkpatch warnings .. except
      xen/blkback: Fix checkpatch warnings of xenbus.c
      xen/blkback: Fix checkpatch warnings in blkback.c
      xen/blkback: remove per-queue plugging
      xen/blkback: Seperate the bio allocation and the bio submission.
      xen/blkback: Cleanup move the code a bit around.
      xen/blkback: Shuffle code around (vbd_translate moved higher).
      xen/blkback: Move the check for misaligned I/O higher.
      xen/blkback: Change fast_flush_area to xen_blkbk_unmap, and tweak xen_blk_map_seg.
      xen/p2m/m2p/gnttab: Support GNTMAP_host_map in the M2P override.
      xen/blkback: Move the check for misaligned I/O once more.
      xen/blkback: Get the 'requeust_queue' properly.
      block, xen/blkback: remove blk_[get|put]_queue calls.
      xen/blkback: Move it from drivers/xen to drivers/block
      xen/p2m: Add EXPORT_SYMBOL_GPL to the M2P override functions.
      xen/blkback: Squash vbd.c,interface.c in blkback.c and xenbus.c respectivly.
      xen-blkback: Remove from the copyright notice the address.
      xen-blkback: Inline some of the functions that were moved from vbd/interface.c
      xen/blkback: Prefix exposed functions with xen_
      xen/blkback: Move the plugging/unplugging to a higher level.
      xen/blkback: Stick REQ_SYNC on WRITEs to deal with CFQ I/O scheduler.
      Revert "xen/blkback: Move the plugging/unplugging to a higher level."
      xen-blkfront: Provide for 'feature-flush-cache' the BLKIF_OP_WRITE_FLUSH_CACHE operation.
      xen/blkback: Add support for BLKIF_OP_FLUSH_DISKCACHE and drop BLKIF_OP_WRITE_BARRIER.
      xen/blkback: Squash the checking for operation into dispatch_rw_block_io
      xen/blkback: Fix up some of the comments.
      xen/blkback: Fixed up comments and converted spaces to tabs.
      xen-blkfront: Provide for 'feature-flush-cache' the BLKIF_OP_WRITE_FLUSH_CACHE operation.
      xen-blkfront: Introduce BLKIF_OP_FLUSH_DISKCACHE support.
      xen/setup: Ignore E820_UNUSABLE when setting 1-1 mappings.
      xen/p2m: Create entries in the P2M_MFN trees's to track 1-1 mappings
      xen/blkback: Change printk/DPRINTK to pr_.. type variant.
      xen/blkback: Make the DPRINTK uniform.
      xen/blkback: Use the DRV_PFX in the pr_.. macros.
      xen/blkback: Removing the debug_lvl option.
      xen/blkback: Move blkif_get_x86_[32|64]_req to common.h in block/xen-blkback dir.
      xen/blkback: Fix spelling mistakes.
      xen/blkback: Flesh out the description in the Kconfig.
      xen/blkback: Fix checkpatch.pl warnings about more than 80 lines.
      xen/blkback: Checkpatch.pl recommend against multiple assigments.
      xen/blkback: Fixing some more of the cleanpatch.pl warnings.
      xen/blkback: Move include/xen/blkif.h into drivers/block/xen-blkback/common.h
      xen/blkback: Remove the unused typedefs.
      xen/blkback: Change structure name blkif_st to xen_blkif.
      xen/blkback: Prefix 'vbd' with 'xen' in structs and functions.
      xen/blkback: Add the prefix XEN in the common.h.
      xen/blkback: if log_stats is enabled print out the data.
      xen/blkback: Align the tabs on the structure.
      xen/pci: Fix compiler error when CONFIG_XEN_PRIVILEGED_GUEST is not set.
      xen/p2m/m2p/gnttab: Support GNTMAP_host_map in the M2P override.
      xen/p2m: Add EXPORT_SYMBOL_GPL to the M2P override functions.

Laszlo Ersek (1):
      xen/blkback: fix xenbus_transaction_start() hang caused by double xenbus_transaction_end()

Marek Marczykowski (1):
      xen-blkfront: fix data size for xenbus_gather in blkfront_connect

Michał Mirosław (3):
      net: convert xen-netfront to hw_features
      net: netxen: convert to hw_features
      net: xen-netback: convert to hw_features

Stefano Stabellini (1):
      xen: do not clear and mask evtchns in __xen_evtchn_do_upcall

Tian, Kevin (3):
      xen mmu: fix a race window causing leave_mm BUG()
      x86: Skip migrating IRQF_PER_CPU irqs in fixup_irqs()
      x86: Don't unmask disabled irqs when migrating them

Tom Goetz (1):
      xen/blkback: Fix the WRITE_BARRIER

Re: [Xen-devel] Linux 3.0-rc1 - what Xen components went in.

[Joseph Glanville]

Wow, excellent work!

Congrats to the Xen devel team. :)

Joseph.

On 1 June 2011 02:12, Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> wrote:
> 3.0-rc1 (2.6.40-rc1) is out and while in the past it took us all the way up to
> rc7 to have it working under Xen, this is not case with this release!
>
> And my VP summarized this event in this blog:
> http://blogs.oracle.com/wim/entry/linux_mainline_contains_all_the
>
> But if you want to get more technical:
>
> Features that went in:
>  - cleancache shim to Xen Transcendent Memory. In other words, you can use tmem
>   (not yet in, but as an outside tree), or zcache (in the staging directory) to
>   speed up your I/Os. Details on how to use it:
>   http://oss.oracle.com/projects/tmem/dist/documentation/internals/linuxpatch
>
>  - The usage of VM_IO to distinguish between pages that are "System RAM" or
>   "PCI" is slowly going away. While it worked in the past it also brought us
>   lots of bugs that took time to get right. The new mechanism, based on the P2M
>   1-1 mapping bypasses the need for the VM_IO flag. Still have some more work..
>
>  - Updates to the backend code so that Xen PCI backend can easily run against
>   the kernel (Note: xen pciback has not yet been merged)
>
>  - Lots of cleanups done in the general code and in the balloon code. The latter preps
>   the code base for using the more generic memory hotplugging.
>
>  - IRQ rework. In the past our IRQ code was bit of dance. We had to distinguish between
>   the different types of interrupts (think edge vs level) coming from the Xen hypervisor.
>   Except that we weren't doing it cleanly and were overloading the chip functions that dealt
>   with this and causing us to do do some extra hack^H^H^H fancy unmasking/masking. Gone!
>
>  - Xen block backend. Yes, it is in!
>
> Bugs:
>  - Only four of them and they are also in the 2.6.39.1 stable tree.
>
> That is it... rc1 is starting and please report any bugs to the mailing list.
>
> The shortlog is as follow:
>
> Chris Lalancette (1):
>      xen/blkback: Flush blkback data when connecting.
>
> Dan Magenheimer (1):
>      xen: cleancache shim to Xen Transcendent Memory
>
> Daniel De Graaf (1):
>      xen/gntdev,gntalloc: Remove unneeded VM flags
>
> Daniel Kiper (13):
>      xen/balloon: Use PageHighMem() for high memory page detection
>      xen/balloon: Simplify HVM integration
>      xen/balloon: Clarify credit calculation
>      xen/balloon: Move dec_totalhigh_pages() from __balloon_append() to balloon_append()
>      xen/setup: Fix for incorrect xen_extra_mem_start initialization under 32-bit
>      arch/x86/xen/irq: Cleanup code/data sections definitions
>      arch/x86/xen/enlighten: Cleanup code/data sections definitions
>      arch/x86/xen/setup: Cleanup code/data sections definitions
>      arch/x86/xen/mmu: Cleanup code/data sections definitions
>      arch/x86/xen/xen-ops: Cleanup code/data sections definitions
>      arch/x86/xen/time: Cleanup code/data sections definitions
>      arch/x86/xen/smp: Cleanup code/data sections definitions
>      drivers/xen/sys-hypervisor: Cleanup code/data sections definitions
>
> Eric Dumazet (1):
>      xen: netfront: fix declaration order
>
>
> Ian Campbell (6):
>      xen: netfront: assume all hw features are available until backend connection setup
>      Signed bit field; int have_hotplug_status_watch:1
>      xen: rename blkbk module xen-blkback.
>      xen: tidy up whitespace in drivers/xen/Makefile
>      xen: netback: use __CONST_RING_SIZE not __RING_SIZE
>      xen: netfront: hold RTNL when updating features.
>
> Jan Beulich (2):
>      xen/blkback: simplify address translations
>      xen/blkback: don't fail empty barrier requests
>
> Jeremy Fitzhardinge (14):
>      xen-blkback-porting
>      xen/blkback: don't include xen/evtchn.h
>      xen/blkback: deal with hardsect_size to logical_block_size rename
>      xen/blkback: remove spurious debug output noise
>      xen/blkback: little cleanups
>      xen/blkback: use drv_get/set_drvdata rather than directly accessing driver_data.
>      xen/blkback: add accessor for xenbus backend device
>      xen: drop all the special iomap pte paths.
>      xen: use mmu_update for xen_set_pte_at()
>      xen: condense everything onto xen_set_pte
>      xen: make a pile of mmu pvop functions static
>      xen: use normal virt_to_machine for ptes
>      xen/mmu: remove all ad-hoc stats stuff
>      xen: fix compile without CONFIG_XEN_DEBUG_FS
>
> John Stultz (1):
>      x86: Convert remaining x86 clocksources to clocksource_register_hz/khz
>
> K. Y. Srinivasan (3):
>      xen/blkback: Propagate changed size of VBDs
>      xen/blkback: Print additional information when a vbd is resized.
>
>
> Keir Fraser (1):
>      blkback: Fix CVE-2010-3699
>
> Konrad Rzeszutek Wilk (68):
>      xen/pci: Add xen_[find|register|unregister]_device_domain_owner functions.
>      xen/irq: Check if the PCI device is owned by a domain different than DOMID_SELF.
>      xen/irq: Add support to check if IRQ line is shared with other domains.
>      xen/irq: Export 'xen_pirq_from_irq' function.
>      xen/irq: The Xen hypervisor cleans up the PIRQs if the other domain forgot.
>      xen: add blkback support
>      Fix compile warnings: ignoring return value of 'xenbus_register_backend' ..
>      xen/blkback: Replace WRITE_BARRIER with (REQ_FLUSH | REQ_FUA)
>      xen/blkback: Update to use blkdev_get_by_dev instead of open_by_devnum.
>      xen/blkback: Move global/static variables into struct xen_blkbk.
>      xen/blkback: Union the blkif_request request specific fields
>      xen/blkback: Use 'vzalloc' for page arrays and pre-allocate pages.
>      xen/blkback: Utilize the M2P override mechanism for GNTMAP_host_map
>      xen/blkback: Use kzalloc's, and GFP_KERNEL for data structures.
>      xen/blkback: Add some comments.
>      xen/blkback: blkif->struct blkif_st
>      xen/blkback: Fix checkpatch warnings in vbd.c
>      xen/blkback: Fix interface.c checkpatch warnings .. except
>      xen/blkback: Fix checkpatch warnings of xenbus.c
>      xen/blkback: Fix checkpatch warnings in blkback.c
>      xen/blkback: remove per-queue plugging
>      xen/blkback: Seperate the bio allocation and the bio submission.
>      xen/blkback: Cleanup move the code a bit around.
>      xen/blkback: Shuffle code around (vbd_translate moved higher).
>      xen/blkback: Move the check for misaligned I/O higher.
>      xen/blkback: Change fast_flush_area to xen_blkbk_unmap, and tweak xen_blk_map_seg.
>      xen/p2m/m2p/gnttab: Support GNTMAP_host_map in the M2P override.
>      xen/blkback: Move the check for misaligned I/O once more.
>      xen/blkback: Get the 'requeust_queue' properly.
>      block, xen/blkback: remove blk_[get|put]_queue calls.
>      xen/blkback: Move it from drivers/xen to drivers/block
>      xen/p2m: Add EXPORT_SYMBOL_GPL to the M2P override functions.
>      xen/blkback: Squash vbd.c,interface.c in blkback.c and xenbus.c respectivly.
>      xen-blkback: Remove from the copyright notice the address.
>      xen-blkback: Inline some of the functions that were moved from vbd/interface.c
>      xen/blkback: Prefix exposed functions with xen_
>      xen/blkback: Move the plugging/unplugging to a higher level.
>      xen/blkback: Stick REQ_SYNC on WRITEs to deal with CFQ I/O scheduler.
>      Revert "xen/blkback: Move the plugging/unplugging to a higher level."
>      xen-blkfront: Provide for 'feature-flush-cache' the BLKIF_OP_WRITE_FLUSH_CACHE operation.
>      xen/blkback: Add support for BLKIF_OP_FLUSH_DISKCACHE and drop BLKIF_OP_WRITE_BARRIER.
>      xen/blkback: Squash the checking for operation into dispatch_rw_block_io
>      xen/blkback: Fix up some of the comments.
>      xen/blkback: Fixed up comments and converted spaces to tabs.
>      xen-blkfront: Provide for 'feature-flush-cache' the BLKIF_OP_WRITE_FLUSH_CACHE operation.
>      xen-blkfront: Introduce BLKIF_OP_FLUSH_DISKCACHE support.
>      xen/setup: Ignore E820_UNUSABLE when setting 1-1 mappings.
>      xen/p2m: Create entries in the P2M_MFN trees's to track 1-1 mappings
>      xen/blkback: Change printk/DPRINTK to pr_.. type variant.
>      xen/blkback: Make the DPRINTK uniform.
>      xen/blkback: Use the DRV_PFX in the pr_.. macros.
>      xen/blkback: Removing the debug_lvl option.
>      xen/blkback: Move blkif_get_x86_[32|64]_req to common.h in block/xen-blkback dir.
>      xen/blkback: Fix spelling mistakes.
>      xen/blkback: Flesh out the description in the Kconfig.
>      xen/blkback: Fix checkpatch.pl warnings about more than 80 lines.
>      xen/blkback: Checkpatch.pl recommend against multiple assigments.
>      xen/blkback: Fixing some more of the cleanpatch.pl warnings.
>      xen/blkback: Move include/xen/blkif.h into drivers/block/xen-blkback/common.h
>      xen/blkback: Remove the unused typedefs.
>      xen/blkback: Change structure name blkif_st to xen_blkif.
>      xen/blkback: Prefix 'vbd' with 'xen' in structs and functions.
>      xen/blkback: Add the prefix XEN in the common.h.
>      xen/blkback: if log_stats is enabled print out the data.
>      xen/blkback: Align the tabs on the structure.
>      xen/pci: Fix compiler error when CONFIG_XEN_PRIVILEGED_GUEST is not set.
>      xen/p2m/m2p/gnttab: Support GNTMAP_host_map in the M2P override.
>      xen/p2m: Add EXPORT_SYMBOL_GPL to the M2P override functions.
>
> Laszlo Ersek (1):
>      xen/blkback: fix xenbus_transaction_start() hang caused by double xenbus_transaction_end()
>
> Marek Marczykowski (1):
>      xen-blkfront: fix data size for xenbus_gather in blkfront_connect
>
> Michał Mirosław (3):
>      net: convert xen-netfront to hw_features
>      net: netxen: convert to hw_features
>      net: xen-netback: convert to hw_features
>
> Stefano Stabellini (1):
>      xen: do not clear and mask evtchns in __xen_evtchn_do_upcall
>
> Tian, Kevin (3):
>      xen mmu: fix a race window causing leave_mm BUG()
>      x86: Skip migrating IRQF_PER_CPU irqs in fixup_irqs()
>      x86: Don't unmask disabled irqs when migrating them
>
> Tom Goetz (1):
>      xen/blkback: Fix the WRITE_BARRIER
>
>
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel
>



-- 
Kind regards,
Joseph.
Founder | Director
Orion Virtualisation Solutions | www.orionvm.com.au | Phone: 1300 56
99 52 | Mobile: 0428 754 846

Re: Linux 3.0-rc1 - what Xen components went in.

[Stephan Seitz]

[-- Attachment #1.1: Type: text/plain, Size: 1434 bytes --]

On Tue, May 31, 2011 at 12:12:24PM -0400, Konrad Rzeszutek Wilk wrote:
> - Xen block backend. Yes, it is in!

I tried 3.0-rc2 as a Dom0 kernel, but I have problems with the block 
backend support.
Both Dom0 and DomU are Debian/Testing systems. XEN version is 4.1.0.

The 3.0-rc2 compiled and started without problems, even the DomU started.  
But when I looked at the DomU console I noticed that the system had 
problems with the filesystems. ext3 and ext4 partitions produced journal 
errors and got mounted read only.

Going back to my 2.6.34.8 with xen patches the DomU problems vanished.

Here is the XEN configuration part for the DomU block devices:

disk = [ 'phy:mapper/mroot,xvda1,w', 'phy:mapper/mswap,xvda2,w', 
'phy:mapper/mvar,xvda5,w', 'phy:mapper/msquid,xvda6,w', 
'phy:mapper/musr,xvda7,w', 'phy:mapper/mhome,xvda8,w' ]

Do I have to change something? The devices are LUKS encrypted LVM 
devices.

The DomU kernel is a 2.6.36.4 vanilla kernel. I noticed that I can’t use 
a newer kernel. Every kernel > 2.6.36 crashes the DomU immediatley (I 
don’t even get a console). The log said that DomU was crashing too fast, 
so it was disabled to avoid loops. This happens with 2.6.34.8 as Dom0 
kernel and with 3.0-rc2.

Shade and sweet water!

	Stephan

-- 
| Stephan Seitz             E-Mail: stse@fsing.rootsland.net |
| PGP Public Keys: http://fsing.rootsland.net/~stse/pgp.html |

[-- Attachment #1.2: Digital signature --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

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

Re: Re: Linux 3.0-rc1 - what Xen components went in.

[Konrad Rzeszutek Wilk]

On Thu, Jun 16, 2011 at 11:48:20AM +0200, Stephan Seitz wrote:
> On Tue, May 31, 2011 at 12:12:24PM -0400, Konrad Rzeszutek Wilk wrote:
> >- Xen block backend. Yes, it is in!
> 
> I tried 3.0-rc2 as a Dom0 kernel, but I have problems with the block
> backend support.
> Both Dom0 and DomU are Debian/Testing systems. XEN version is 4.1.0.
> 
> The 3.0-rc2 compiled and started without problems, even the DomU
> started.  But when I looked at the DomU console I noticed that the
> system had problems with the filesystems. ext3 and ext4 partitions
> produced journal errors and got mounted read only.

Hmm, are barriers somehow being turned on? (They shouldn't). It would
be helpfull if you included the error logs.

> 
> Going back to my 2.6.34.8 with xen patches the DomU problems vanished.

So using the older Dom0. Are there any errors/warnings reported by
the frontend? Is there a difference in the dmesg output from the DomU
booted on 2.6.34 vs 3.0?

> 
> Here is the XEN configuration part for the DomU block devices:
> 
> disk = [ 'phy:mapper/mroot,xvda1,w', 'phy:mapper/mswap,xvda2,w',
> 'phy:mapper/mvar,xvda5,w', 'phy:mapper/msquid,xvda6,w',
> 'phy:mapper/musr,xvda7,w', 'phy:mapper/mhome,xvda8,w' ]
> 

that is pretty much how I've it too. Well:

disk = [ 'phy:/dev/vg_guest_1/data-xfs,xvda,w', 'phy:/dev/vg_guest_1/data-ext4,xvdb,w','phy:/dev/vg_guest_1/data-ext3,xvdc,w', 'phy:/dev/vg_guest_1/data-swap ,xvdd,w']

I don't try to put disks as partitions. Let me try that. Are the sizes
that xne-blkfront report the same?

> Do I have to change something? The devices are LUKS encrypted LVM
> devices.

Ah, hadn't tried LUKS. What happens if they are _not_ on LUKS?
Let me set up LUKs on my box in the meantime.

> 
> The DomU kernel is a 2.6.36.4 vanilla kernel. I noticed that I can’t
> use a newer kernel. Every kernel > 2.6.36 crashes the DomU
> immediatley (I don’t even get a console). The log said that DomU was
> crashing too fast, so it was disabled to avoid loops. This happens
> with 2.6.34.8 as Dom0 kernel and with 3.0-rc2.

Can you boot it with 'earlyprintk=xenboot debug loglevel=10' please?
Are you using a special version of .config? Is it available somewhere?

Re: Linux 3.0-rc1 - what Xen components went in.

[Stephan Seitz]

[-- Attachment #1.1: Type: text/plain, Size: 2540 bytes --]

On Thu, Jun 16, 2011 at 11:07:06AM -0400, Konrad Rzeszutek Wilk wrote:
>On Thu, Jun 16, 2011 at 11:48:20AM +0200, Stephan Seitz wrote:
>> The 3.0-rc2 compiled and started without problems, even the DomU
>> started.  But when I looked at the DomU console I noticed that the
>> system had problems with the filesystems. ext3 and ext4 partitions
>> produced journal errors and got mounted read only.
>Hmm, are barriers somehow being turned on? (They shouldn't). It would
>be helpfull if you included the error logs.

At least with 2.6.36.4-DomU barriers seems to be enabled:
blkfront: xvda1: barriers enabled (tag)
blkfront: xvda2: barriers enabled (tag)
blkfront: xvda5: barriers enabled (tag)
blkfront: xvda6: barriers enabled (tag)
blkfront: xvda7: barriers enabled (tag)
blkfront: xvda8: barriers enabled (tag)

xvda1 is ext3 (root fs) with the mount option barrier=1, but this doesn’t 
seem to work (EXT3-fs: barriers not enabled). The other filesystems are 
ext4 without explicit barrier options and don’t have log output 
concerning barriers.

For now I can’t make tests since I’m not at home for the next days. Then 
I will try to get logs as well. Since the filesystems are changed to read 
only, the logs are not written to disk, so I have to use other methods.

>I don't try to put disks as partitions. Let me try that. Are the sizes
>that xne-blkfront report the same?

xen-blkfront? Is this a program or kernel output? Dom0 or DomU?

>Ah, hadn't tried LUKS. What happens if they are _not_ on LUKS?

For now I can’t try it. Besides /boot in Dom0 all my partitions are LUKS 
encrypted.

>> The DomU kernel is a 2.6.36.4 vanilla kernel. I noticed that I can’t
>> use a newer kernel. Every kernel > 2.6.36 crashes the DomU
>> immediatley (I don’t even get a console). The log said that DomU was
>> crashing too fast, so it was disabled to avoid loops. This happens
>> with 2.6.34.8 as Dom0 kernel and with 3.0-rc2.
>Can you boot it with 'earlyprintk=xenboot debug loglevel=10' please?

Is this a boot option I pass to the DomU kernel? I will do the test next 
week.

>Are you using a special version of .config? Is it available somewhere?

http://fsing.rootsland.net/~stse/config-2.6.38.3-DomU (not working)
http://fsing.rootsland.net/~stse/config-2.6.36.4-DomU (working)

Thanks for your help.

Shade and sweet water!

	Stephan

-- 
| Stephan Seitz             E-Mail: stse@fsing.rootsland.net |
| PGP Public Keys: http://fsing.rootsland.net/~stse/pgp.html |

[-- Attachment #1.2: Digital signature --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

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

Re: Re: Linux 3.0-rc1 - what Xen components went in.

[Konrad Rzeszutek Wilk]

On Thu, Jun 16, 2011 at 07:31:43PM +0200, Stephan Seitz wrote:
> On Thu, Jun 16, 2011 at 11:07:06AM -0400, Konrad Rzeszutek Wilk wrote:
> >On Thu, Jun 16, 2011 at 11:48:20AM +0200, Stephan Seitz wrote:
> >>The 3.0-rc2 compiled and started without problems, even the DomU
> >>started.  But when I looked at the DomU console I noticed that the
> >>system had problems with the filesystems. ext3 and ext4 partitions
> >>produced journal errors and got mounted read only.
> >Hmm, are barriers somehow being turned on? (They shouldn't). It would
> >be helpfull if you included the error logs.
> 
> At least with 2.6.36.4-DomU barriers seems to be enabled:
> blkfront: xvda1: barriers enabled (tag)
> blkfront: xvda2: barriers enabled (tag)
> blkfront: xvda5: barriers enabled (tag)
> blkfront: xvda6: barriers enabled (tag)
> blkfront: xvda7: barriers enabled (tag)
> blkfront: xvda8: barriers enabled (tag)

That is wrong. It should not have turned them on. The 3.0 version
does not export the feature-barrier at all. Looks like a bug
in xen-blkfront in 2.6.36 ..
> 
> xvda1 is ext3 (root fs) with the mount option barrier=1, but this
> doesn’t seem to work (EXT3-fs: barriers not enabled). The other
> filesystems are ext4 without explicit barrier options and don’t have
> log output concerning barriers.

Uh, what? Do you see the errors if the barriers are turned off.  As
in -o barrier=0 or -o writeback=1
> 
> For now I can’t make tests since I’m not at home for the next days.
> Then I will try to get logs as well. Since the filesystems are
> changed to read only, the logs are not written to disk, so I have to
> use other methods.

Uh, you can log the output from the kernel using the 'xl crreate -c'.
You might need to edit the Linux command line options..
> 
> >I don't try to put disks as partitions. Let me try that. Are the sizes
> >that xne-blkfront report the same?
> 
> xen-blkfront? Is this a program or kernel output? Dom0 or DomU?

kernel output.
> 
> >Ah, hadn't tried LUKS. What happens if they are _not_ on LUKS?
> 
> For now I can’t try it. Besides /boot in Dom0 all my partitions are
> LUKS encrypted.

OK? Can you try in a week or so? I am going to try to reproduce LUKS here.
You could just create new partition.

> 
> >>The DomU kernel is a 2.6.36.4 vanilla kernel. I noticed that I can’t
> >>use a newer kernel. Every kernel > 2.6.36 crashes the DomU
> >>immediatley (I don’t even get a console). The log said that DomU was
> >>crashing too fast, so it was disabled to avoid loops. This happens
> >>with 2.6.34.8 as Dom0 kernel and with 3.0-rc2.
> >Can you boot it with 'earlyprintk=xenboot debug loglevel=10' please?
> 
> Is this a boot option I pass to the DomU kernel? I will do the test
> next week.

Yes.
> 
> >Are you using a special version of .config? Is it available somewhere?
> 
> http://fsing.rootsland.net/~stse/config-2.6.38.3-DomU (not working)
> http://fsing.rootsland.net/~stse/config-2.6.36.4-DomU (working)
> 
> Thanks for your help.
> 
> Shade and sweet water!
> 
> 	Stephan
> 
> -- 
> | Stephan Seitz             E-Mail: stse@fsing.rootsland.net |
> | PGP Public Keys: http://fsing.rootsland.net/~stse/pgp.html |



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

Re: Linux 3.0-rc1 - what Xen components went in.

[Stephan Seitz]

[-- Attachment #1.1: Type: text/plain, Size: 1231 bytes --]

On Thu, Jun 16, 2011 at 11:48:20AM +0200, Stephan Seitz wrote:
>I tried 3.0-rc2 as a Dom0 kernel, but I have problems with the block 
>backend support.
>Both Dom0 and DomU are Debian/Testing systems. XEN version is 4.1.0.
>
>The 3.0-rc2 compiled and started without problems, even the DomU 
>started.  But when I looked at the DomU console I noticed that the 
>system had problems with the filesystems. ext3 and ext4 partitions 
>produced journal errors and got mounted read only.

I could solve this problem by using the mount option „barrier=0” for all 
ext3 and ext4 filesystems in /etc/fstab.

>The DomU kernel is a 2.6.36.4 vanilla kernel. I noticed that I can’t use 
>a newer kernel. Every kernel > 2.6.36 crashes the DomU immediatley (I 
>don’t even get a console). The log said that DomU was crashing too fast, 

This seems to be fixed in 3.0 as well. At least I could use 3.0-rc3 as 
DomU kernel.

I have now a running Dom0 and DomU with 3.0-rc5, and everything is 
working fine.

Thanks for your help and your work!

Shade and sweet water!

	Stephan

-- 
| Stephan Seitz             E-Mail: stse@fsing.rootsland.net |
| PGP Public Keys: http://fsing.rootsland.net/~stse/pgp.html |

[-- Attachment #1.2: Digital signature --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

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

Problem with vfb at Xen 4.1 Dom0 ( 3.0.0-5-generic ) Ubuntu 11.10 (alpha 2)

[Boris Derzhavets]

[-- Attachment #1.1: Type: text/plain, Size: 767 bytes --]

Profile f15.cfg ( to install)  :-

name="VF15PV"
memory=2048
disk = ['phy:/dev/sda6,xvda,w' ]
vif = [ 'bridge=br0' ]
# vfb = [ 'type=vnc,vncunused=1']
kernel = "/root/vms/f15/vmlinuz"
ramdisk = "/root/vms/f15/initrd.img"
vcpus=2
on_reboot = 'restart'
on_crash = 'restart'

works fine for

# xl create f15.cfg -c
or
# xm create f15.cfg -c

As soon as i add "vfb"  line .
1. xl fails when parsing
2. xm reports name "VF15PV" doesn't exist.

Runtime profile :-

name="VF15PV"
memory=2048
disk = ['phy:/dev/sda6,xvda,w' ]
vif = [ 'bridge=br0,mac=00:16:3E:66:61:B1' ]
bootloader = "/usr/lib/xen-4.1/bin/pygrub"
vcpus=2
on_reboot = 'restart'
on_crash = 'restart'

behaves same way .  No vfb line, then xl and xm works fine.


Boris.


[-- Attachment #1.2: Type: text/html, Size: 975 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

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

Re: [Xen-devel] Problem with vfb at Xen 4.1 Dom0 ( 3.0.0-5-generic ) Ubuntu 11.10 (alpha 2)

[Stefano Stabellini]

[-- Attachment #1: Type: text/plain, Size: 884 bytes --]

On Sun, 17 Jul 2011, Boris Derzhavets wrote:
> Profile f15.cfg ( to install)  :-
> 
> name="VF15PV"
> memory=2048
> disk = ['phy:/dev/sda6,xvda,w' ]
> vif = [ 'bridge=br0' ]
> # vfb = [ 'type=vnc,vncunused=1']
> kernel = "/root/vms/f15/vmlinuz"
> ramdisk = "/root/vms/f15/initrd.img"
> vcpus=2
> on_reboot = 'restart'
> on_crash = 'restart'
> 
> works fine for
> 
> # xl create f15.cfg -c
> or
> # xm create f15.cfg -c
> 
> As soon as i add "vfb"  line .
> 1. xl fails when parsing
> 2. xm reports name "VF15PV" doesn't exist.
> 
> Runtime profile :-
> 
> name="VF15PV"
> memory=2048
> disk = ['phy:/dev/sda6,xvda,w' ]
> vif = [ 'bridge=br0,mac=00:16:3E:66:61:B1' ]
> bootloader = "/usr/lib/xen-4.1/bin/pygrub"
> vcpus=2
> on_reboot = 'restart'
> on_crash = 'restart'
> 
> behaves same way .  No vfb line, then xl and xm works fine.
> 
 
it should be:

vfb = [ 'vnc=1,vncunused=1']

[-- Attachment #2: Type: text/plain, Size: 137 bytes --]

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

Re: [Xen-devel] Problem with vfb at Xen 4.1 Dom0 ( 3.0.0-5-generic ) Ubuntu 11.10 (alpha 2)

[Boris Derzhavets]

[-- Attachment #1.1: Type: text/plain, Size: 2185 bytes --]

root@boris-P5Q-E:~/vms/f15# cat f15.install
name="VF15S"
memory=2048
disk = ['phy:/dev/sda7,xvda,w' ]
vif = [ 'bridge=br0' ]
vfb = [ 'vnc=1,vncunused=1']
kernel = "/root/vms/f15/vmlinuz"
ramdisk = "/root/vms/f15/initrd.img"
vcpus=2
on_reboot = 'restart'
on_crash = 'restart'

root@boris-P5Q-E:~/vms/f15# xl create f15.install
Parsing config file f15.install
libxl: error: libxl_device.c:476:libxl__wait_for_device_model Device Model not ready
xl: fatal error: libxl_create.c:535, rc=-1: libxl__confirm_device_model_startup

Boris

--- On Mon, 7/18/11, Stefano Stabellini <stefano.stabellini@eu.citrix.com> wrote:

From: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Subject: Re: [Xen-devel] Problem with vfb at Xen 4.1 Dom0 ( 3.0.0-5-generic ) Ubuntu 11.10 (alpha 2)
To: "Boris Derzhavets" <bderzhavets@yahoo.com>
Cc: "xen-devel@lists.xensource.com" <xen-devel@lists.xensource.com>, "xen-users@lists.xensource.com" <xen-users@lists.xensource.com>, "Rzeszutek Wilk" <konrad.wilk@oracle.com>, Konrad@yahoo.com
Date: Monday, July 18, 2011, 7:35 AM

On Sun, 17 Jul 2011, Boris Derzhavets wrote:
> Profile f15.cfg ( to install)  :-
> 
> name="VF15PV"
> memory=2048
> disk = ['phy:/dev/sda6,xvda,w' ]
> vif = [ 'bridge=br0' ]
> # vfb = [ 'type=vnc,vncunused=1']
> kernel = "/root/vms/f15/vmlinuz"
> ramdisk = "/root/vms/f15/initrd.img"
> vcpus=2
> on_reboot = 'restart'
> on_crash = 'restart'
> 
> works fine for
> 
> # xl create f15.cfg -c
> or
> # xm create f15.cfg -c
> 
> As soon as i add "vfb"  line .
> 1. xl fails when parsing
> 2. xm reports name "VF15PV" doesn't exist.
> 
> Runtime profile :-
> 
> name="VF15PV"
> memory=2048
> disk = ['phy:/dev/sda6,xvda,w' ]
> vif = [ 'bridge=br0,mac=00:16:3E:66:61:B1' ]
> bootloader = "/usr/lib/xen-4.1/bin/pygrub"
> vcpus=2
> on_reboot = 'restart'
> on_crash = 'restart'
> 
> behaves same way .  No vfb line, then xl and xm works fine.
> 
 
it should be:

vfb = [ 'vnc=1,vncunused=1']
-----Inline Attachment Follows-----

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

[-- Attachment #1.2: Type: text/html, Size: 2967 bytes --]

[-- Attachment #2: Type: text/plain, Size: 137 bytes --]

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

Re: [Xen-devel] Problem with vfb at Xen 4.1 Dom0 ( 3.0.0-5-generic ) Ubuntu 11.10 (alpha 2)

[Stefano Stabellini]

On Mon, 18 Jul 2011, Boris Derzhavets wrote:
> root@boris-P5Q-E:~/vms/f15# cat f15.install
> name="VF15S"
> memory=2048
> disk = ['phy:/dev/sda7,xvda,w' ]
> vif = [ 'bridge=br0' ]
> vfb = [ 'vnc=1,vncunused=1']
> kernel = "/root/vms/f15/vmlinuz"
> ramdisk = "/root/vms/f15/initrd.img"
> vcpus=2
> on_reboot = 'restart'
> on_crash = 'restart'
> 
> root@boris-P5Q-E:~/vms/f15# xl create f15.install
> Parsing config file f15.install
> libxl: error: libxl_device.c:476:libxl__wait_for_device_model Device Model not ready
> xl: fatal error: libxl_create.c:535, rc=-1: libxl__confirm_device_model_startup
> 

this is another failure to start qemu, probably due to the missing
keymap files, like the other one.

Re: Problem with vfb at Xen 4.1 Dom0 ( 3.0.0-5-generic ) Ubuntu 11.10 (alpha 2)

[Boris Derzhavets]

[-- Attachment #1.1: Type: text/plain, Size: 1571 bytes --]

Yes , install keymaps brings VFB

But, attempt to install F15 PV domU gives just a part of required screen VFB window 
with alive mouse pointer,  for  F14 PV install - VFB  window is good. But mouse pointer is dead.

Boris.


--- On Mon, 7/18/11, Stefano Stabellini <stefano.stabellini@eu.citrix.com> wrote:

From: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Subject: Re: [Xen-devel] Problem with vfb at Xen 4.1 Dom0 ( 3.0.0-5-generic ) Ubuntu 11.10 (alpha 2)
To: "Boris Derzhavets" <bderzhavets@yahoo.com>
Cc: "Stefano Stabellini" <Stefano.Stabellini@eu.citrix.com>, "xen-devel@lists.xensource.com" <xen-devel@lists.xensource.com>, "xen-users@lists.xensource.com" <xen-users@lists.xensource.com>, "Rzeszutek Wilk" <konrad.wilk@oracle.com>, "Konrad@yahoo.com" <Konrad@yahoo.com>
Date: Monday, July 18, 2011, 8:40 AM

On Mon, 18 Jul 2011, Boris Derzhavets wrote:
> root@boris-P5Q-E:~/vms/f15# cat f15.install
> name="VF15S"
> memory=2048
> disk = ['phy:/dev/sda7,xvda,w' ]
> vif = [ 'bridge=br0' ]
> vfb = [ 'vnc=1,vncunused=1']
> kernel = "/root/vms/f15/vmlinuz"
> ramdisk = "/root/vms/f15/initrd.img"
> vcpus=2
> on_reboot = 'restart'
> on_crash = 'restart'
> 
> root@boris-P5Q-E:~/vms/f15# xl create f15.install
> Parsing config file f15.install
> libxl: error: libxl_device.c:476:libxl__wait_for_device_model Device Model not ready
> xl: fatal error: libxl_create.c:535, rc=-1: libxl__confirm_device_model_startup
> 

this is another failure to start qemu, probably due to the missing
keymap files, like the other one.


[-- Attachment #1.2: Type: text/html, Size: 2032 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

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

Re: [Xen-devel] Problem with vfb at Xen 4.1 Dom0 ( 3.0.0-5-generic ) Ubuntu 11.10 (alpha 2)

[Stefano Stabellini]

[-- Attachment #1: Type: text/plain, Size: 505 bytes --]

CC'ing Konrad and Olaf

On Mon, 18 Jul 2011, Boris Derzhavets wrote:
> Yes , install keymaps brings VFB
> 
> But, attempt to install F15 PV domU gives just a part of required screen VFB window
> with alive mouse pointer,  for  F14 PV install - VFB  window is good. But mouse pointer is dead.

This bug sounds familiar. Maybe the guest kernel is missing:

Input: xen-kbdfront - advertise either absolute or relative coordinates

If that is the case we might have to file a bug in the Fedora bug
tracker?

[-- Attachment #2: Type: text/plain, Size: 137 bytes --]

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

Re: Problem with vfb at Xen 4.1 Dom0 ( 3.0.0-5-generic ) Ubuntu 11.10 (alpha 2)

[Olaf Hering]

On Mon, Jul 18, Stefano Stabellini wrote:

> CC'ing Konrad and Olaf
> 
> On Mon, 18 Jul 2011, Boris Derzhavets wrote:
> > Yes , install keymaps brings VFB
> > 
> > But, attempt to install F15 PV domU gives just a part of required screen VFB window
> > with alive mouse pointer,  for  F14 PV install - VFB  window is good. But mouse pointer is dead.
> 
> This bug sounds familiar. Maybe the guest kernel is missing:
> 
> Input: xen-kbdfront - advertise either absolute or relative coordinates
> 
> If that is the case we might have to file a bug in the Fedora bug
> tracker?

Its nice to have the two patches related to the abs/rel coordinate
handling. But the dead mouse pointer happend only during save/restore
and migration. So I guess the F14 issue is something else.


Olaf