[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] Question about the XEN platform pci

From: karim.allah.ahmed@xxxxxxxxx [mailto:karim.allah.ahmed@xxxxxxxxx] 
Sent: 2016年4月13日 15:00
To: Konrad Rzeszutek Wilk
Cc: Wu, Bob; xen-devel@xxxxxxxxxxxxxxxxxxxx
Subject: Re: [Xen-devel] Question about the XEN platform pci

On Tue, Apr 12, 2016 at 5:45 PM, Konrad Rzeszutek Wilk <konrad.wilk@xxxxxxxxxx> 
> On Tue, Apr 12, 2016 at 05:33:47PM +0200, karim.allah.ahmed@xxxxxxxxx wrote:
>> The INTx interrupt of this platform device can be used by Xen in HVM 
>> case to notify the guest of pending events in the event channel. 
>> However that's usually not used in favor of vector callbacks support 
>> in Xen where a vector is injected directly to the vCPU bypassing LAPIC.
>> (that said, the platform-pci driver in linux is actually broken when 
>> vector callbacks are not used anyway)
> Oh? Is there an report/bug somewhere?

I'm not sure if it's reported as a bug somewhere or not.

I've always assumed that INTx is deperecated and vector callbacks are the one 
that's supported that's why I never tried to fix it, in addition I never tried 
to reproduce it I just looked at the code and it seemed a little bit off as 
explained below.

Mainly xenbus_init is called during postcore_initcall which will eventually try 
to read a value from XenStore and will get stuck on read_reply at xenbus 
forever since the platform driver is not probed yet and its INTx interrupt 
handler is not registered yet which basically means that the guest can not be 
notified at this moment of any pending event channels and none of the per-event 
handlers will ever be invoked (including the XenStore one) and the reply will 
never be picked up by the kernel.

The exact stack where things get stuck during xenbus_init:


> Thanks!
>> I also think that the grant-table lives on this PCI device MMIO BAR 
>> (?!)
> The area may be usurped for grant-table as the OS won't touch that 
> memory area (it after all belongs to the device).
>> If you looked at hw/i386/xen/xen_platform.c in QEMU source , you will 
>> get a general idea what this device is supposed todo (like logging to 
>> syslog stuff for example).
>> >>
>> >> -----Original Message-----
>> >> From: Konrad Rzeszutek Wilk [mailto:konrad.wilk@xxxxxxxxxx]
>> >> Sent: 2016年4月11日 22:24
>> >> To: Wu, Bob
>> >> Cc: xen-devel@xxxxxxxxxxxxx
>> >> Subject: Re: [Xen-devel] Question about the XEN platform pci
>> >>
>> >> On Fri, Apr 08, 2016 at 08:52:08AM +0000, Wu, Bob wrote:
>> >> >
>> >> > Sorry bother, I read the XEN source code recently, and found the 
>> >> > XEN platform PCI code under drivers/xen/platform-pci.c, and I can't 
>> >> > fully under this driver's affect, can anybody explain a little for me?
>> >> >
>> >> > Is the platform PCI driver for PV-split-PCI-driver-model such as the 
>> >> > pci-frontend/pci-backend? or for PCI pass-through model? Or for other 
>> >> > purpose?
>> >> > I saw the xenbus_pcifront_driver/ xenbus_xen_pcibk_driver are 
>> >> > registered on XENBUS, so I guess the platform-PCI-driver is not for PV 
>> >> > PCI driver.
>> >> >
>> >>
>> >> It is for the QEMU driver. To tell QEMU to stop emulating the IDE/network.
>> >>
>> >> > Really thank you for your replay.
>> >> >
>> >> > Thanks,
>> >> > Bob
>> >> >

Hi, Karim,

I still have one more question, can you share your understanding?
The PCI platform driver just register a pci driver on pci bus, but when the 
real device attaching occur?
I mean who trigger the real device attach behavior? The QEMU or others? I there 
any piece of code can indicate it?

Xen-devel mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.