diff mbox series

[v2] xen-block: handle resize callback

Message ID 20190130161948.15888-1-paul.durrant@citrix.com (mailing list archive)
State New, archived
Headers show
Series [v2] xen-block: handle resize callback | expand

Commit Message

Paul Durrant Jan. 30, 2019, 4:19 p.m. UTC
Some frontend drivers will handle dynamic resizing of PV disks, so set up
the BlockDevOps resize_cb() method during xen_block_realize() to allow
this to be done.

Signed-off-by: Paul Durrant <paul.durrant@citrix.com>
---
Cc: Stefan Hajnoczi <stefanha@redhat.com>
Cc: Stefano Stabellini <sstabellini@kernel.org>
Cc: Anthony Perard <anthony.perard@citrix.com>
Cc: Kevin Wolf <kwolf@redhat.com>
Cc: Max Reitz <mreitz@redhat.com>

v2:
 - re-write backend state after updating sectors
---
 hw/block/dataplane/xen-block.c |  4 +---
 hw/block/trace-events          |  1 +
 hw/block/xen-block.c           | 40 ++++++++++++++++++++++++++++++----
 3 files changed, 38 insertions(+), 7 deletions(-)

Comments

Anthony PERARD Jan. 31, 2019, 3:20 p.m. UTC | #1
On Wed, Jan 30, 2019 at 04:19:48PM +0000, Paul Durrant wrote:
> Some frontend drivers will handle dynamic resizing of PV disks, so set up
> the BlockDevOps resize_cb() method during xen_block_realize() to allow
> this to be done.
> 
> Signed-off-by: Paul Durrant <paul.durrant@citrix.com>
> ---


> +    /*
> +     * Mimic the behaviour of Linux xen-blkback and re-write the state
> +     * to trigger the frontend watch.
> +     */
> +    xen_device_backend_set_state(xendev, backend_state);

:(, that function doesn't write the state again if it hasn't changed.
So in my testing, Linux never did anything.
Paul Durrant Jan. 31, 2019, 3:22 p.m. UTC | #2
> -----Original Message-----
> From: Anthony PERARD [mailto:anthony.perard@citrix.com]
> Sent: 31 January 2019 15:21
> To: Paul Durrant <Paul.Durrant@citrix.com>
> Cc: qemu-devel@nongnu.org; qemu-block@nongnu.org; xen-
> devel@lists.xenproject.org; Stefan Hajnoczi <stefanha@redhat.com>; Stefano
> Stabellini <sstabellini@kernel.org>; Kevin Wolf <kwolf@redhat.com>; Max
> Reitz <mreitz@redhat.com>
> Subject: Re: [PATCH v2] xen-block: handle resize callback
> 
> On Wed, Jan 30, 2019 at 04:19:48PM +0000, Paul Durrant wrote:
> > Some frontend drivers will handle dynamic resizing of PV disks, so set
> up
> > the BlockDevOps resize_cb() method during xen_block_realize() to allow
> > this to be done.
> >
> > Signed-off-by: Paul Durrant <paul.durrant@citrix.com>
> > ---
> 
> 
> > +    /*
> > +     * Mimic the behaviour of Linux xen-blkback and re-write the state
> > +     * to trigger the frontend watch.
> > +     */
> > +    xen_device_backend_set_state(xendev, backend_state);
> 
> :(, that function doesn't write the state again if it hasn't changed.
> So in my testing, Linux never did anything.

Gah! I forgot about that. Alright, it's going to have to be a bit more crude.

  Paul

> 
> --
> Anthony PERARD
Anthony PERARD Jan. 31, 2019, 3:25 p.m. UTC | #3
On Thu, Jan 31, 2019 at 03:22:18PM +0000, Paul Durrant wrote:
> > -----Original Message-----
> > From: Anthony PERARD [mailto:anthony.perard@citrix.com]
> > Sent: 31 January 2019 15:21
> > To: Paul Durrant <Paul.Durrant@citrix.com>
> > Cc: qemu-devel@nongnu.org; qemu-block@nongnu.org; xen-
> > devel@lists.xenproject.org; Stefan Hajnoczi <stefanha@redhat.com>; Stefano
> > Stabellini <sstabellini@kernel.org>; Kevin Wolf <kwolf@redhat.com>; Max
> > Reitz <mreitz@redhat.com>
> > Subject: Re: [PATCH v2] xen-block: handle resize callback
> > 
> > On Wed, Jan 30, 2019 at 04:19:48PM +0000, Paul Durrant wrote:
> > > Some frontend drivers will handle dynamic resizing of PV disks, so set
> > up
> > > the BlockDevOps resize_cb() method during xen_block_realize() to allow
> > > this to be done.
> > >
> > > Signed-off-by: Paul Durrant <paul.durrant@citrix.com>
> > > ---
> > 
> > 
> > > +    /*
> > > +     * Mimic the behaviour of Linux xen-blkback and re-write the state
> > > +     * to trigger the frontend watch.
> > > +     */
> > > +    xen_device_backend_set_state(xendev, backend_state);
> > 
> > :(, that function doesn't write the state again if it hasn't changed.
> > So in my testing, Linux never did anything.
> 
> Gah! I forgot about that. Alright, it's going to have to be a bit more crude.

more crude > Yes, I tried to ignore the check in _set_state and end-up
with an infinit loop.
Paul Durrant Jan. 31, 2019, 3:26 p.m. UTC | #4
> -----Original Message-----
> From: Anthony PERARD [mailto:anthony.perard@citrix.com]
> Sent: 31 January 2019 15:25
> To: Paul Durrant <Paul.Durrant@citrix.com>
> Cc: qemu-devel@nongnu.org; qemu-block@nongnu.org; xen-
> devel@lists.xenproject.org; Stefan Hajnoczi <stefanha@redhat.com>; Stefano
> Stabellini <sstabellini@kernel.org>; Kevin Wolf <kwolf@redhat.com>; Max
> Reitz <mreitz@redhat.com>
> Subject: Re: [PATCH v2] xen-block: handle resize callback
> 
> On Thu, Jan 31, 2019 at 03:22:18PM +0000, Paul Durrant wrote:
> > > -----Original Message-----
> > > From: Anthony PERARD [mailto:anthony.perard@citrix.com]
> > > Sent: 31 January 2019 15:21
> > > To: Paul Durrant <Paul.Durrant@citrix.com>
> > > Cc: qemu-devel@nongnu.org; qemu-block@nongnu.org; xen-
> > > devel@lists.xenproject.org; Stefan Hajnoczi <stefanha@redhat.com>;
> Stefano
> > > Stabellini <sstabellini@kernel.org>; Kevin Wolf <kwolf@redhat.com>;
> Max
> > > Reitz <mreitz@redhat.com>
> > > Subject: Re: [PATCH v2] xen-block: handle resize callback
> > >
> > > On Wed, Jan 30, 2019 at 04:19:48PM +0000, Paul Durrant wrote:
> > > > Some frontend drivers will handle dynamic resizing of PV disks, so
> set
> > > up
> > > > the BlockDevOps resize_cb() method during xen_block_realize() to
> allow
> > > > this to be done.
> > > >
> > > > Signed-off-by: Paul Durrant <paul.durrant@citrix.com>
> > > > ---
> > >
> > >
> > > > +    /*
> > > > +     * Mimic the behaviour of Linux xen-blkback and re-write the
> state
> > > > +     * to trigger the frontend watch.
> > > > +     */
> > > > +    xen_device_backend_set_state(xendev, backend_state);
> > >
> > > :(, that function doesn't write the state again if it hasn't changed.
> > > So in my testing, Linux never did anything.
> >
> > Gah! I forgot about that. Alright, it's going to have to be a bit more
> crude.
> 
> more crude > Yes, I tried to ignore the check in _set_state and end-up
> with an infinit loop.

Yes, you would. That check is there for a reason :-) I'll replace the call in xen-block.c with a call to xen_device_backend_printf().

  Paul

> 
> --
> Anthony PERARD
diff mbox series

Patch

diff --git a/hw/block/dataplane/xen-block.c b/hw/block/dataplane/xen-block.c
index d0d8905a33..c6a15da024 100644
--- a/hw/block/dataplane/xen-block.c
+++ b/hw/block/dataplane/xen-block.c
@@ -50,7 +50,6 @@  struct XenBlockDataPlane {
     unsigned int nr_ring_ref;
     void *sring;
     int64_t file_blk;
-    int64_t file_size;
     int protocol;
     blkif_back_rings_t rings;
     int more_work;
@@ -189,7 +188,7 @@  static int xen_block_parse_request(XenBlockRequest *request)
                request->req.seg[i].first_sect + 1) * dataplane->file_blk;
         request->size += len;
     }
-    if (request->start + request->size > dataplane->file_size) {
+    if (request->start + request->size > blk_getlength(dataplane->blk)) {
         error_report("error: access beyond end of file");
         goto err;
     }
@@ -638,7 +637,6 @@  XenBlockDataPlane *xen_block_dataplane_create(XenDevice *xendev,
     dataplane->xendev = xendev;
     dataplane->file_blk = conf->logical_block_size;
     dataplane->blk = conf->blk;
-    dataplane->file_size = blk_getlength(dataplane->blk);
 
     QLIST_INIT(&dataplane->inflight);
     QLIST_INIT(&dataplane->freelist);
diff --git a/hw/block/trace-events b/hw/block/trace-events
index d0851953c5..8020f9226a 100644
--- a/hw/block/trace-events
+++ b/hw/block/trace-events
@@ -126,6 +126,7 @@  xen_block_realize(const char *type, uint32_t disk, uint32_t partition) "%s d%up%
 xen_block_connect(const char *type, uint32_t disk, uint32_t partition) "%s d%up%u"
 xen_block_disconnect(const char *type, uint32_t disk, uint32_t partition) "%s d%up%u"
 xen_block_unrealize(const char *type, uint32_t disk, uint32_t partition) "%s d%up%u"
+xen_block_size(const char *type, uint32_t disk, uint32_t partition, int64_t sectors) "%s d%up%u %"PRIi64
 xen_disk_realize(void) ""
 xen_disk_unrealize(void) ""
 xen_cdrom_realize(void) ""
diff --git a/hw/block/xen-block.c b/hw/block/xen-block.c
index a636487b3e..0f0799e861 100644
--- a/hw/block/xen-block.c
+++ b/hw/block/xen-block.c
@@ -144,6 +144,38 @@  static void xen_block_unrealize(XenDevice *xendev, Error **errp)
     }
 }
 
+static void xen_block_set_size(XenBlockDevice *blockdev)
+{
+    const char *type = object_get_typename(OBJECT(blockdev));
+    XenBlockVdev *vdev = &blockdev->props.vdev;
+    BlockConf *conf = &blockdev->props.conf;
+    int64_t sectors = blk_getlength(conf->blk) / conf->logical_block_size;
+    XenDevice *xendev = XEN_DEVICE(blockdev);
+
+    trace_xen_block_size(type, vdev->disk, vdev->partition, sectors);
+
+    xen_device_backend_printf(xendev, "sectors", "%"PRIi64, sectors);
+}
+
+static void xen_block_resize_cb(void *opaque)
+{
+    XenBlockDevice *blockdev = opaque;
+    XenDevice *xendev = XEN_DEVICE(blockdev);
+    enum xenbus_state backend_state = xen_device_backend_get_state(xendev);
+
+    xen_block_set_size(blockdev);
+
+    /*
+     * Mimic the behaviour of Linux xen-blkback and re-write the state
+     * to trigger the frontend watch.
+     */
+    xen_device_backend_set_state(xendev, backend_state);
+}
+
+static const BlockDevOps xen_block_dev_ops = {
+    .resize_cb = xen_block_resize_cb,
+};
+
 static void xen_block_realize(XenDevice *xendev, Error **errp)
 {
     XenBlockDevice *blockdev = XEN_BLOCK_DEVICE(xendev);
@@ -180,7 +212,7 @@  static void xen_block_realize(XenDevice *xendev, Error **errp)
     }
 
     if (!blkconf_apply_backend_options(conf, blockdev->info & VDISK_READONLY,
-                                       false, errp)) {
+                                       true, errp)) {
         return;
     }
 
@@ -197,6 +229,7 @@  static void xen_block_realize(XenDevice *xendev, Error **errp)
         return;
     }
 
+    blk_set_dev_ops(conf->blk, &xen_block_dev_ops, blockdev);
     blk_set_guest_block_size(conf->blk, conf->logical_block_size);
 
     if (conf->discard_granularity > 0) {
@@ -215,9 +248,8 @@  static void xen_block_realize(XenDevice *xendev, Error **errp)
 
     xen_device_backend_printf(xendev, "sector-size", "%u",
                               conf->logical_block_size);
-    xen_device_backend_printf(xendev, "sectors", "%"PRIi64,
-                              blk_getlength(conf->blk) /
-                              conf->logical_block_size);
+
+    xen_block_set_size(blockdev);
 
     blockdev->dataplane =
         xen_block_dataplane_create(xendev, conf, blockdev->props.iothread);