Re: [Xen-devel] [PATCH 1/2] IOMMU/spinlock: Fix a bug found in AMD IOMMU initialization.

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 08.03.16 at 12:09, <quan.xu@xxxxxxxxx> wrote:
> Doing what we do serves as a fix for a bug found in AMD IOMMU initialization.
> 
> The current code is using spin_lock{_irqsave(), _irqrestore()} to
> protect pci_get_dev() in the set_iommu_interrupt_handler(). However,
> this can only be called during AMD IOMMU initialization, with interrupt
> enabled, so at least it is not necessary to disable interrupts, or
> save/restore interrupt flag.

On top of what Dario said: This description isn't very accurate either:
If the code in question runs with interrupts enabled (which I'm not
sure it does; would take me following back up the call chain to see
whether this happens before or after interrupts get enabled for the
first time), then it may very well be necessary to disable interrupts
for a particular purpose - from an abstract pov. That's not the case
here, but the description of such a change shouldn't make incorrect
claims or statements.

Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel