Re: [Xen-devel] [PATCH v6 3/5] IOMMU: Make the pcidevs_lock a recursive one

["Xu, Quan" <quan.xu@xxxxxxxxx>]

On March 07, 2016 7:49pm, <JBeulich@xxxxxxxx> wrote:
> >>> On 07.03.16 at 12:42, <quan.xu@xxxxxxxxx> wrote:
> > On March 07, 2016 7:36pm, <JBeulich@xxxxxxxx> wrote:
> >> >>> On 07.03.16 at 12:23, <quan.xu@xxxxxxxxx> wrote:
> >> > On March 07, 2016 7:14pm, <JBeulich@xxxxxxxx> wrote:
> >> >> >>> On 07.03.16 at 08:05, <quan.xu@xxxxxxxxx> wrote:
> >> >
> >> >
> >> >> > A quick question, is it '-ERESTART', instead of '-EBUSY' ?
> >> >>
> >> >> No idea what this question is about in this context.
> >> >>
> >> >
> >> > it is in xen/drivers/passthrough/pci.c, assign_device().
> >> >
> >> > static int assign_device()
> >> > {
> >> >    ....
> >> >     if ( !spin_trylock(&pcidevs_lock) )
> >> >         return -ERESTART;
> >> >    ....
> >> > }
> >>
> >> But I still don't understand what you're trying to find out or point out.
> >
> > Jan, sorry.
> > Now the return error code is '-ERESTART' for ' if (
> > !spin_trylock(&pcidevs_lock) ', in assign_device(), in
> > xen/drivers/passthrough/pci.c.
> > I think it would be '-EBUSY'.
> 
> Oh - definitely not. Just follow the call chain back up, and you should find 
> that
> this gets taken as an indication to create a continuation, whereas -EBUSY 
> would
> bubble back up to the original (user space) caller (which is _not_ what we 
> want
> here).
> 

Got it. thanks.
Quan

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