[Qemu-devel] what's on qemu tree for pci related fixes

[Qemu-devel] what's on qemu tree for pci related fixes

[Michael S. Tsirkin]

Here's the list of patches on my pci fixes tree:
git://git.kernel.org/pub/scm/linux/kernel/git/mst/qemu.git pci

Here are the pci fixes I know of that haven't been applied already.
All, any more comments on any of them?
If yes, please reply on the original patch, not here.

eee1bff qemu: clean up target page usage in msix
52a5bce fix comment on cpu_register_physical_memory_offset
9305505 qemu/virtio-pci: remove unnecessary check
c128c88 qemu/pci: reset device registers on bus reset
74b7e8b qemu/pci: refactor code/symbolic constants
5307502 qemu/virtio: fix reset with device removal
c1f2e99 qemu/qdev: type safety in reset handler

Isaku Yamahata, would you like me to put some of the fixes you posted
onto this tree, to reduce the chance if conflict? I can do this, but
you'll have to start including changelog when you repost your patches so
I can figure out that some patch was not changed.

Anthony, I decided not to rebase the tree, I'll merge master in,
instead, on the off chance that you finally get converted to using git
pull.  But if this makes things inconvenient for you, pls let me know:
maybe I can keep both a rebased and non-rebased branches around.

Thanks,

-- 
MST

[Qemu-devel] what's on qemu tree for pci related fixes

[Michael S. Tsirkin]

Here's the list of patches on my pci fixes tree:
git://git.kernel.org/pub/scm/linux/kernel/git/mst/qemu.git pci

Here are the pci fixes I know of that haven't been applied already.
They mostly fix up minor nits.  All, any more comments on any of them?
If yes, please reply on the original patch, not here.

8c205ed pci: move apb specific stuff to apb_pci.c
9d4a672 pci: pci.h cleanup: move out stuff not in pci.c
d9f806f pci: fix pci_info_device().
b19ef75 pci: move pci_data_{read, write}() declaration from pci.h to pci_host.h
33a8716 pci: simplify (pci_/pcie_mmcfg_)data_read()
34df7d5 pci: remove pci_addr_to_config() by open code
95586cc pci: rename (pci_/pcie_mmcfg_)addr_to_dev
996a107 pci: shorten pci_host_{conf, data}_register_xxx function a bit.
dec2505 pci: remove pci_sub_bus() by open coding.
f0940fd pci: s/pci_find_host_bus/pci_find_root_bus/g
89071ce pci_host: remove unnecessary & 0xff.
49ab987 pci: kill unnecessary included in pci.c
4400dda pci: clean up of pci_init_wmask().
f464d9e pci: remove some unnecessary comment in pci.h
dbb55bc pci: move typedef, PCIHostState, PCIExpressHost to qemu-common.h.
e518813 pci: remove unused constants.
91880da pci: clean up of pci_update_mappings()
958610d pci: split up up pci_update mappings
f2e78cc pci: remove magic number, 256 in pci.c
09b5e0d pci: fix pci_config_get_io_base().
270b6cc pci: pci bridge related clean up.
eba4cea pci: convert goto into scope in bridge_filter
6ef641c configure: use correct cflags in compiler checks

(last commit isn't pci specific but I need it to
 build qemu, so I let it in).

Thanks,

-- 
MST

[Qemu-devel] Re: what's on qemu tree for pci related fixes

[Anthony Liguori]

Michael S. Tsirkin wrote:
> Here's the list of patches on my pci fixes tree:
> git://git.kernel.org/pub/scm/linux/kernel/git/mst/qemu.git pci
>
> Here are the pci fixes I know of that haven't been applied already.
> They mostly fix up minor nits.  All, any more comments on any of them?
> If yes, please reply on the original patch, not here.
>   

Most of these are in staging minus Isaku's latest series. The last 
series caused some gnarly regressions so I've got them in a separate 
branch for more thorough testing.

Regards,

Anthony Liguori

[Qemu-devel] Re: what's on qemu tree for pci related fixes

[Michael S. Tsirkin]

On Sun, Nov 22, 2009 at 09:40:52AM -0600, Anthony Liguori wrote:
> Michael S. Tsirkin wrote:
>> Here's the list of patches on my pci fixes tree:
>> git://git.kernel.org/pub/scm/linux/kernel/git/mst/qemu.git pci
>>
>> Here are the pci fixes I know of that haven't been applied already.
>> They mostly fix up minor nits.  All, any more comments on any of them?
>> If yes, please reply on the original patch, not here.
>>   
>
> Most of these are in staging minus Isaku's latest series. The last  
> series caused some gnarly regressions

You mean the regressions caused by series that were already applied?
This series is actually about fixing up nits
from that last one :)

> so I've got them in a separate  
> branch for more thorough testing.

It isn't a public branch, is it?

> Regards,
>
> Anthony Liguori

-- 
MST

[Qemu-devel] Re: what's on qemu tree for pci related fixes

[Michael S. Tsirkin]

On Sun, Nov 22, 2009 at 09:40:52AM -0600, Anthony Liguori wrote:
> Michael S. Tsirkin wrote:
>> Here's the list of patches on my pci fixes tree:
>> git://git.kernel.org/pub/scm/linux/kernel/git/mst/qemu.git pci
>>
>> Here are the pci fixes I know of that haven't been applied already.
>> They mostly fix up minor nits.  All, any more comments on any of them?
>> If yes, please reply on the original patch, not here.
>>   
>
> Most of these are in staging minus Isaku's latest series. The last  
> series caused some gnarly regressions so I've got them in a separate  
> branch for more thorough testing.
>
> Regards,
>
> Anthony Liguori

Hope these are on their way to staging now as well.
I prefer building on top of these cleanups.

-- 
MST

[Qemu-devel] what's on qemu tree for pci related fixes

[Michael S. Tsirkin]

Here's the list of patches on my pci fixes tree:
git://git.kernel.org/pub/scm/linux/kernel/git/mst/qemu.git pci
I have rebased this tree to latest master and rearranged the patches in
the order of their being ready for upstream.  The tree is currently on
top of qemu master: I would base it on staging tree, but that is
constantly rebased, is it not?

I have slightly tweaked commit log comments on a couple of patches,
fixing typos, and adding acks. Not sure it's worth the noise to repost
on list just for this.  Anthony, can you just take patches from the
tree so that comments look nicer?

Here are the pci fixes I know of that haven't been applied on staging
already.  All, any more comments on any of them?
If yes, please reply on the original patch, not here.

        pci: move apb specific stuff to apb_pci.c
        configure: use correct cflags in compiler checks

The above are in staging now.

        pci: fix pci_info_device().
        pci: move pci_data_{read, write}() declaration from pci.h to pci_host.h
        pci: simplify (pci_/pcie_mmcfg_)data_read()
        pci: remove pci_addr_to_config() by open code
        pci: rename (pci_/pcie_mmcfg_)addr_to_dev
        pci: shorten pci_host_{conf, data}_register_xxx function a bit.
        pci: remove pci_sub_bus() by open coding.
        pci: s/pci_find_host_bus/pci_find_root_bus/g
        pci_host: remove unnecessary & 0xff.
        pci: kill unnecessary included in pci.c
        pci: clean up of pci_init_wmask().
        pci: remove some unnecessary comment in pci.h
        pci: move typedef, PCIHostState, PCIExpressHost to qemu-common.h.
        pci: remove unused constants.
        pci: clean up of pci_update_mappings()
        pci: split up up pci_update mappings
        pci: remove magic number, 256 in pci.c
        pci: fix pci_config_get_io_base().
        pci: pci bridge related clean up.
        pci: convert goto into scope in bridge_filter

The above are minor cleanups, they run fine for me, look very safe, have
at least one ack, and were out for more than 2 weeks now. Time to apply,
IMO.

        msix: add helper to mark all msix entries
        virtio: do not reset msix state on soft reset
        msix: fix mask bit state after reset
        msix: fix reset value for enable bit
        msix: clear pending bit of an unused vector

The above all look pretty obvious, safe, and fix crashes reported with
windows guests.  I think they are needed on stable-0.11 as well.
Out only about 1 week though ... more comments?

        msix: macro rename for function mask support
        msix: function mask support
        pci: prepare irq code for interrupt state
        pci: interrupt status bit implementation
        pci: interrupt disable bit support

These have been out less than 1 week, and they aren't trivial. We need
them for 0.12 so that WHQL can pass there, but maybe not for 0.11.
No known issues/unaddressed comments though.

-- 
MST