Re: [Xen-devel] [PATCH 5 of 9 v2] libxl: wait for devices to initialize upon addition to the domain

[Ian Campbell <Ian.Campbell@xxxxxxxxxx>]

On Fri, 2011-11-18 at 11:59 +0000, Roger Pau Monne wrote:
> # HG changeset patch
> # User Roger Pau Monne <roger.pau@xxxxxxxxxxxxx>
> # Date 1321612154 -3600
> # Node ID 4ad998fddb16a299cb230ea23ba944d84adbd2bf
> # Parent  a77bba3717402fdf24a72d028ae0f3d482e6f427
> libxl: wait for devices to initialize upon addition to the domain.
> 
> Block waiting for devices to initialize (switch to state 2). This is
> necessary because we need to call the hotplug scripts when state is
> 2, otherwise the execution fails. Make use of the newly introduced
> function libxl__wait_for_device_state.

It would be better to use the XenbusStateFoo constants than the numbers.
This also applies to the prviosu patch, I just didn't think of it...

Otherwise this looks good.

I'm a little concerned about the synchronous wait here (since it will
serialise adding all devices). Eventually it should be possible to do
all the adds and then wait for them all but this is dependent on the
event infrastructure.

> Signed-off-by: Roger Pau Monne <roger.pau@xxxxxxxxxxxxx>
> 
> diff -r a77bba371740 -r 4ad998fddb16 tools/libxl/libxl.c
> --- a/tools/libxl/libxl.c     Fri Nov 18 10:49:00 2011 +0100
> +++ b/tools/libxl/libxl.c     Fri Nov 18 11:29:14 2011 +0100
> @@ -971,6 +971,8 @@ int libxl_device_disk_add(libxl_ctx *ctx
>      flexarray_t *front;
>      flexarray_t *back;
>      char *dev;
> +    char *be_path, *state_path, *state;
> +    struct timeval tv;
>      libxl__device device;
>      int major, minor, rc;
>  
> @@ -1075,6 +1077,23 @@ int libxl_device_disk_add(libxl_ctx *ctx
>                               libxl__xs_kvs_of_flexarray(&gc, back, 
> back->count),
>                               libxl__xs_kvs_of_flexarray(&gc, front, 
> front->count));
>  
> +    be_path = libxl__device_backend_path(&gc, &device);
> +    state_path = libxl__sprintf(&gc, "%s/state", be_path);
> +    state = libxl__xs_read(&gc, XBT_NULL, state_path);
> +
> +    if (atoi(state) != 2) {
> +        xs_watch(ctx->xsh, state_path, be_path);
> +        tv.tv_sec = LIBXL_DESTROY_TIMEOUT;
> +        tv.tv_usec = 0;
> +        rc = libxl__wait_for_device_state(&gc, &tv, 2, NULL);
> +        xs_unwatch(ctx->xsh, state_path, be_path);
> +        if (rc < 0) {
> +            LIBXL__LOG(ctx, LIBXL__LOG_ERROR,
> +                       "unable to initialize disk device: %s\n",
> +                       disk->pdev_path);
> +            goto out_free;
> +        }
> +    }
>      rc = 0;
>  
>  out_free:
> @@ -1450,6 +1469,8 @@ int libxl_device_nic_add(libxl_ctx *ctx,
>      flexarray_t *back;
>      libxl__device device;
>      char *dompath, **l;
> +    char *be_path, *state_path, *state;
> +    struct timeval tv;
>      unsigned int nb, rc;
>  
>      front = flexarray_make(16, 1);
> @@ -1518,8 +1539,25 @@ int libxl_device_nic_add(libxl_ctx *ctx,
>                               libxl__xs_kvs_of_flexarray(&gc, back, 
> back->count),
>                               libxl__xs_kvs_of_flexarray(&gc, front, 
> front->count));
>  
> -    /* FIXME: wait for plug */
> +    be_path = libxl__device_backend_path(&gc, &device);
> +    state_path = libxl__sprintf(&gc, "%s/state", be_path);
> +    state = libxl__xs_read(&gc, XBT_NULL, state_path);
> +
> +    if (atoi(state) != 2) {
> +        xs_watch(ctx->xsh, state_path, be_path);
> +        tv.tv_sec = LIBXL_DESTROY_TIMEOUT;
> +        tv.tv_usec = 0;
> +        rc = libxl__wait_for_device_state(&gc, &tv, 2, NULL);
> +        xs_unwatch(ctx->xsh, state_path, be_path);
> +        if (rc < 0) {
> +            LIBXL__LOG(ctx, LIBXL__LOG_ERROR,
> +                       "unable to initialize nic device: %s\n",
> +                       nic->ifname);
> +            goto out_free;
> +        }
> +    }
>      rc = 0;
> +
>  out_free:
>      flexarray_free(back);
>      flexarray_free(front);
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@xxxxxxxxxxxxxxxxxxx
> http://lists.xensource.com/xen-devel



_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel