Re: [Qemu-devel] [PATCH v4] virtio-blk: set correct config size for the host driver

[Stefano Garzarella]

On Wed, Feb 13, 2019 at 09:32:27AM +0100, Stefano Garzarella wrote:
> On Wed, Feb 13, 2019 at 04:01:43PM +0800, Stefan Hajnoczi wrote:
> > On Wed, Feb 13, 2019 at 09:48:57AM +0800, Changpeng Liu wrote:
> > > Commit caa1ee43 "vhost-user-blk: add discard/write zeroes features
> > > support" added fields to struct virtio_blk_config. This changes
> > > the size of the config space and breaks migration from QEMU 3.1
> > > and older:
> > > 
> > > qemu-system-ppc64: get_pci_config_device: Bad config data: i=0x10 read: 
> > > 41 device: 1 cmask: ff wmask: 80 w1cmask:0
> > > qemu-system-ppc64: Failed to load PCIDevice:config
> > > qemu-system-ppc64: Failed to load virtio-blk:virtio
> > > qemu-system-ppc64: error while loading state for instance 0x0 of device 
> > > 'address@hidden:01.0/virtio-blk'
> > > qemu-system-ppc64: load of migration failed: Invalid argument
> > > 
> > > Since virtio-blk doesn't support the "discard" and "write zeroes"
> > > features, it shouldn't even expose the associated fields in the
> > > config space actually. Just include all fields up to num_queues to
> > > match QEMU 3.1 and older.
> > > 
> > > Signed-off-by: Changpeng Liu <address@hidden>
> > > ---
> > >  hw/block/virtio-blk.c | 13 +++++++++----
> > >  1 file changed, 9 insertions(+), 4 deletions(-)
> > 
> > Stefano: Please rebase your DISCARD/WRITE_ZEROES series onto this and
> > check that the config space is correctly sized.  Machine types before
> > 4.0 shouldn't have these fields so that the config space size remains
> > unchanged.
> 
> Sure!
> 
> Since I should set a correct config size checking if new features are
> enabled or not at runtime, should be better to add a variable and an array
> of sizes like in virtio-net?

In my series "[PATCH v4 0/6] virtio-blk: add DISCARD and WRITE_ZEROES"
I'm adding the host_features field in VirtIOBlock. Then, I could add
something like the following patch (proof of concept) inspired by the
virtio-net approach, that would be simplest to maintain when we will add
new features.

What do you think?

Thanks,
Stefano


diff --git a/hw/block/virtio-blk.c b/hw/block/virtio-blk.c
index 843bb2bec8..84dcc1406c 100644
--- a/hw/block/virtio-blk.c
+++ b/hw/block/virtio-blk.c
@@ -28,6 +28,51 @@
 #include "hw/virtio/virtio-bus.h"
 #include "hw/virtio/virtio-access.h"
 
+/*
+ * Calculate the number of bytes up to and including the given 'field' of
+ * 'container'.
+ */
+#define endof(container, field) \
+    (offsetof(container, field) + sizeof_field(container, field))
+
+typedef struct VirtIOFeature {
+    uint64_t flags;
+    size_t end;
+} VirtIOFeature;
+
+static VirtIOFeature feature_sizes[] = {
+    {.flags = 1ULL << VIRTIO_NET_F_SIZE_MAX,
+     .end = endof(struct virtio_blk_config, size_max)},
+    {.flags = 1ULL << VIRTIO_NET_F_SEG_MAX,
+     .end = endof(struct virtio_blk_config, seg_max)},
+    {.flags = 1ULL << VIRTIO_NET_F_GEOMETRY,
+     .end = endof(struct virtio_blk_config, geometry)},
+    {.flags = 1ULL << VIRTIO_NET_F_BLK_SIZE,
+     .end = endof(struct virtio_blk_config, blk_size)},
+    {.flags = 1ULL << VIRTIO_NET_F_TOPOLOGY,
+     .end = endof(struct virtio_blk_config, opt_io_size)},
+    {.flags = 1ULL << VIRTIO_NET_F_CONFIG_WCE,
+     .end = endof(struct virtio_blk_config, wce)},
+    {.flags = 1ULL << VIRTIO_NET_F_MQ,
+     .end = endof(struct virtio_blk_config, num_queues)},
+    {}
+};
+
+static void virtio_blk_set_config_size(VirtIOBlock *s, uint64_t host_features)
+{
+    int i, config_size;
+
+    config_size = endof(struct virtio_blk_config, capacity);
+
+    for (i = 0; feature_sizes[i].flags != 0; i++) {
+        if (host_features & feature_sizes[i].flags) {
+            config_size = MAX(feature_sizes[i].end, config_size);
+        }
+    }
+
+    s->config_size = config_size;
+}
+
 static void virtio_blk_init_request(VirtIOBlock *s, VirtQueue *vq,
                                     VirtIOBlockReq *req)
 {
@@ -757,7 +802,7 @@ static void virtio_blk_update_config(VirtIODevice *vdev, 
uint8_t *config)
     blkcfg.alignment_offset = 0;
     blkcfg.wce = blk_enable_write_cache(s->blk);
     virtio_stw_p(vdev, &blkcfg.num_queues, s->conf.num_queues);
-    memcpy(config, &blkcfg, sizeof(struct virtio_blk_config));
+    memcpy(config, &blkcfg, s->config_size);
 }
 
 static void virtio_blk_set_config(VirtIODevice *vdev, const uint8_t *config)
@@ -765,7 +810,7 @@ static void virtio_blk_set_config(VirtIODevice *vdev, const 
uint8_t *config)
     VirtIOBlock *s = VIRTIO_BLK(vdev);
     struct virtio_blk_config blkcfg;
 
-    memcpy(&blkcfg, config, sizeof(blkcfg));
+    memcpy(&blkcfg, config, s->config_size);
 
     aio_context_acquire(blk_get_aio_context(s->blk));
     blk_set_enable_write_cache(s->blk, blkcfg.wce != 0);
@@ -948,8 +993,9 @@ static void virtio_blk_device_realize(DeviceState *dev, 
Error **errp)
         return;
     }
 
-    virtio_init(vdev, "virtio-blk", VIRTIO_ID_BLOCK,
-                sizeof(struct virtio_blk_config));
+    virtio_blk_set_config_size(s, s->host_features);
+
+    virtio_init(vdev, "virtio-blk", VIRTIO_ID_BLOCK, s->config_size);
 
     s->blk = conf->conf.blk;
     s->rq = NULL;