Re: [libvirt] [PATCH v2 7/9] hostdev: Update comments

  > diff --git a/src/util/virhostdev.c b/src/util/virhostdev.c > index ab17547..0892dbf 100644 > --- a/src/util/virhostdev.c > +++ b/src/util/virhostdev.c   This comment block starts "We have to use 9 loops here." - that should be adjusted too as there are now 7 steps.  (Is it any coincidence that there are also 7 stages of grief and 7 steps to great health? ;-))

> +    /* Detaching devices from the host involves several steps; each of them > +     * is described at length below */

> +    /* Step 1: perform safety checks, eg. ensure the devices are assignable */ >      for (i = 0; i < virPCIDeviceListCount(pcidevs); i++) { >          virPCIDevicePtr dev = virPCIDeviceListGet(pcidevs, i); >          bool strict_acs_check = !!(flags & VIR_HOSTDEV_STRICT_ACS_CHECK); > @@ -710,7 +709,8 @@ virHostdevPreparePCIDevices(virHostdevManagerPtr hostdev_mgr, >          } >      } >   > -    /* Loop 2: detach managed devices (i.e. bind to appropriate stub driver) */ > +    /* Step 2: detach managed devices (i.e. bind to appropriate stub driver). > +     *         detachPCIDevices() will also mark devices as inactive */   s/detachPCIDevices/virHostdevOnlyDetachPCIDevice   Or whatever it ends up being named.

s/will also mark devices as inactive/will place device on inactive list */

> +    /* Step 5: reattach managed devices to their host drivers; unmanaged > +     *         devices need no further processing. reattachPCIDevices() will   s/reattachPCIDevices/virHostdevOnlyReattachPCIDevice   (or whatever the final name is)

> +     *         remove devices from the inactive list as they are reattached > +     *         to the host */ >      for (i = 0; i < virPCIDeviceListCount(pcidevs); i++) { >          virPCIDevicePtr dev = virPCIDeviceListGet(pcidevs, i); >   >          ignore_value(virHostdevOnlyReattachPCIDevice(hostdev_mgr, dev, true)); >      } >   > +    /* At this point, all devices are either marked as inactive (if they > +     * were unmanaged), or have been reattached to the host driver (if they > +     * were managed) and are no longer tracked by any of our device lists */ > +   Ahh, but if you kept the goto cleanup code during what is step 2 here, then this wouldn't be necessarily true...