Re: [PATCH 1/2] virtio: use virtio accessor to access packed descriptor flags

2021-11-14 Thread Jason Wang
On Fri, Nov 12, 2021 at 6:04 PM Michael S. Tsirkin  wrote:
>
> On Fri, Nov 12, 2021 at 10:23:12AM +0800, Jason Wang wrote:
> > On Thu, Nov 11, 2021 at 4:27 PM Michael S. Tsirkin  wrote:
> > >
> > > On Thu, Nov 11, 2021 at 02:38:53PM +0800, Jason Wang wrote:
> > > > We used to access packed descriptor flags via
> > > > address_space_{write|read}_cached(). When we hit the cache, memcpy()
> > > > is used which is not an atomic operation which may lead a wrong value
> > > > is read or wrote.
> > >
> > > Could you clarify where's the memcpy that you see?
> > > Thanks!
> >
> > In the address_space_{write|read}_cached it self:
> >
> > static inline MemTxResult
> > =>dress_space_write_cached(MemoryRegionCache *cache, hwaddr addr,
> >const void *buf, hwaddr len)
> > {
> > assert(addr < cache->len && len <= cache->len - addr);
> > if (likely(cache->ptr)) {
> > memcpy(cache->ptr + addr, buf, len);
> > return MEMTX_OK;
> > } else {
> > return address_space_write_cached_slow(cache, addr, buf, len);
> > }
> > }
> >
> > Thanks
>
> But that's a copy from the cache, not from guest memory.

I may miss something but it looks to me the cache is the cache of
address translation not the cache from guest memory:

See address_space_cache_init():

cache->ptr = qemu_ram_ptr_length(mr->ram_block, cache->xlat, , true);

> I don't see how it can change so I don't see why it needs
> to be atomic. Was there an actual issue you observed or
> is this theoretical?

During code review.

Thanks

>
>
> > >
> > > > So this patch switches to use virito_{stw|lduw}_phys_cached() to make
> > > > sure the aceess is atomic.
> > > >
> > > > Fixes: 86044b24e865f ("virtio: basic packed virtqueue support")
> > > > Cc: qemu-sta...@nongnu.org
> > > > Signed-off-by: Jason Wang 
> > > > ---
> > > >  hw/virtio/virtio.c | 11 ---
> > > >  1 file changed, 4 insertions(+), 7 deletions(-)
> > > >
> > > > diff --git a/hw/virtio/virtio.c b/hw/virtio/virtio.c
> > > > index cc69a9b881..939bcbfeb9 100644
> > > > --- a/hw/virtio/virtio.c
> > > > +++ b/hw/virtio/virtio.c
> > > > @@ -507,11 +507,9 @@ static void 
> > > > vring_packed_desc_read_flags(VirtIODevice *vdev,
> > > >   MemoryRegionCache *cache,
> > > >   int i)
> > > >  {
> > > > -address_space_read_cached(cache,
> > > > -  i * sizeof(VRingPackedDesc) +
> > > > -  offsetof(VRingPackedDesc, flags),
> > > > -  flags, sizeof(*flags));
> > > > -virtio_tswap16s(vdev, flags);
> > > > +hwaddr off = i * sizeof(VRingPackedDesc) + 
> > > > offsetof(VRingPackedDesc, flags);
> > > > +
> > > > +*flags = virtio_lduw_phys_cached(vdev, cache, off);
> > > >  }
> > > >
> > > >  static void vring_packed_desc_read(VirtIODevice *vdev,
> > > > @@ -564,8 +562,7 @@ static void 
> > > > vring_packed_desc_write_flags(VirtIODevice *vdev,
> > > >  {
> > > >  hwaddr off = i * sizeof(VRingPackedDesc) + 
> > > > offsetof(VRingPackedDesc, flags);
> > > >
> > > > -virtio_tswap16s(vdev, >flags);
> > > > -address_space_write_cached(cache, off, >flags, 
> > > > sizeof(desc->flags));
> > > > +virtio_stw_phys_cached(vdev, cache, off, desc->flags);
> > > >  address_space_cache_invalidate(cache, off, sizeof(desc->flags));
> > > >  }
> > > >
> > > > --
> > > > 2.25.1
> > >
>




Re: [PATCH 1/2] virtio: use virtio accessor to access packed descriptor flags

2021-11-12 Thread Michael S. Tsirkin
On Fri, Nov 12, 2021 at 10:23:12AM +0800, Jason Wang wrote:
> On Thu, Nov 11, 2021 at 4:27 PM Michael S. Tsirkin  wrote:
> >
> > On Thu, Nov 11, 2021 at 02:38:53PM +0800, Jason Wang wrote:
> > > We used to access packed descriptor flags via
> > > address_space_{write|read}_cached(). When we hit the cache, memcpy()
> > > is used which is not an atomic operation which may lead a wrong value
> > > is read or wrote.
> >
> > Could you clarify where's the memcpy that you see?
> > Thanks!
> 
> In the address_space_{write|read}_cached it self:
> 
> static inline MemTxResult
> =>dress_space_write_cached(MemoryRegionCache *cache, hwaddr addr,
>const void *buf, hwaddr len)
> {
> assert(addr < cache->len && len <= cache->len - addr);
> if (likely(cache->ptr)) {
> memcpy(cache->ptr + addr, buf, len);
> return MEMTX_OK;
> } else {
> return address_space_write_cached_slow(cache, addr, buf, len);
> }
> }
> 
> Thanks

But that's a copy from the cache, not from guest memory.
I don't see how it can change so I don't see why it needs
to be atomic. Was there an actual issue you observed or
is this theoretical?


> >
> > > So this patch switches to use virito_{stw|lduw}_phys_cached() to make
> > > sure the aceess is atomic.
> > >
> > > Fixes: 86044b24e865f ("virtio: basic packed virtqueue support")
> > > Cc: qemu-sta...@nongnu.org
> > > Signed-off-by: Jason Wang 
> > > ---
> > >  hw/virtio/virtio.c | 11 ---
> > >  1 file changed, 4 insertions(+), 7 deletions(-)
> > >
> > > diff --git a/hw/virtio/virtio.c b/hw/virtio/virtio.c
> > > index cc69a9b881..939bcbfeb9 100644
> > > --- a/hw/virtio/virtio.c
> > > +++ b/hw/virtio/virtio.c
> > > @@ -507,11 +507,9 @@ static void 
> > > vring_packed_desc_read_flags(VirtIODevice *vdev,
> > >   MemoryRegionCache *cache,
> > >   int i)
> > >  {
> > > -address_space_read_cached(cache,
> > > -  i * sizeof(VRingPackedDesc) +
> > > -  offsetof(VRingPackedDesc, flags),
> > > -  flags, sizeof(*flags));
> > > -virtio_tswap16s(vdev, flags);
> > > +hwaddr off = i * sizeof(VRingPackedDesc) + offsetof(VRingPackedDesc, 
> > > flags);
> > > +
> > > +*flags = virtio_lduw_phys_cached(vdev, cache, off);
> > >  }
> > >
> > >  static void vring_packed_desc_read(VirtIODevice *vdev,
> > > @@ -564,8 +562,7 @@ static void 
> > > vring_packed_desc_write_flags(VirtIODevice *vdev,
> > >  {
> > >  hwaddr off = i * sizeof(VRingPackedDesc) + offsetof(VRingPackedDesc, 
> > > flags);
> > >
> > > -virtio_tswap16s(vdev, >flags);
> > > -address_space_write_cached(cache, off, >flags, 
> > > sizeof(desc->flags));
> > > +virtio_stw_phys_cached(vdev, cache, off, desc->flags);
> > >  address_space_cache_invalidate(cache, off, sizeof(desc->flags));
> > >  }
> > >
> > > --
> > > 2.25.1
> >




Re: [PATCH 1/2] virtio: use virtio accessor to access packed descriptor flags

2021-11-11 Thread Jason Wang
On Thu, Nov 11, 2021 at 4:27 PM Michael S. Tsirkin  wrote:
>
> On Thu, Nov 11, 2021 at 02:38:53PM +0800, Jason Wang wrote:
> > We used to access packed descriptor flags via
> > address_space_{write|read}_cached(). When we hit the cache, memcpy()
> > is used which is not an atomic operation which may lead a wrong value
> > is read or wrote.
>
> Could you clarify where's the memcpy that you see?
> Thanks!

In the address_space_{write|read}_cached it self:

static inline MemTxResult
=>dress_space_write_cached(MemoryRegionCache *cache, hwaddr addr,
   const void *buf, hwaddr len)
{
assert(addr < cache->len && len <= cache->len - addr);
if (likely(cache->ptr)) {
memcpy(cache->ptr + addr, buf, len);
return MEMTX_OK;
} else {
return address_space_write_cached_slow(cache, addr, buf, len);
}
}

Thanks

>
> > So this patch switches to use virito_{stw|lduw}_phys_cached() to make
> > sure the aceess is atomic.
> >
> > Fixes: 86044b24e865f ("virtio: basic packed virtqueue support")
> > Cc: qemu-sta...@nongnu.org
> > Signed-off-by: Jason Wang 
> > ---
> >  hw/virtio/virtio.c | 11 ---
> >  1 file changed, 4 insertions(+), 7 deletions(-)
> >
> > diff --git a/hw/virtio/virtio.c b/hw/virtio/virtio.c
> > index cc69a9b881..939bcbfeb9 100644
> > --- a/hw/virtio/virtio.c
> > +++ b/hw/virtio/virtio.c
> > @@ -507,11 +507,9 @@ static void vring_packed_desc_read_flags(VirtIODevice 
> > *vdev,
> >   MemoryRegionCache *cache,
> >   int i)
> >  {
> > -address_space_read_cached(cache,
> > -  i * sizeof(VRingPackedDesc) +
> > -  offsetof(VRingPackedDesc, flags),
> > -  flags, sizeof(*flags));
> > -virtio_tswap16s(vdev, flags);
> > +hwaddr off = i * sizeof(VRingPackedDesc) + offsetof(VRingPackedDesc, 
> > flags);
> > +
> > +*flags = virtio_lduw_phys_cached(vdev, cache, off);
> >  }
> >
> >  static void vring_packed_desc_read(VirtIODevice *vdev,
> > @@ -564,8 +562,7 @@ static void vring_packed_desc_write_flags(VirtIODevice 
> > *vdev,
> >  {
> >  hwaddr off = i * sizeof(VRingPackedDesc) + offsetof(VRingPackedDesc, 
> > flags);
> >
> > -virtio_tswap16s(vdev, >flags);
> > -address_space_write_cached(cache, off, >flags, 
> > sizeof(desc->flags));
> > +virtio_stw_phys_cached(vdev, cache, off, desc->flags);
> >  address_space_cache_invalidate(cache, off, sizeof(desc->flags));
> >  }
> >
> > --
> > 2.25.1
>




Re: [PATCH 1/2] virtio: use virtio accessor to access packed descriptor flags

2021-11-11 Thread Michael S. Tsirkin
On Thu, Nov 11, 2021 at 02:38:53PM +0800, Jason Wang wrote:
> We used to access packed descriptor flags via
> address_space_{write|read}_cached(). When we hit the cache, memcpy()
> is used which is not an atomic operation which may lead a wrong value
> is read or wrote.

Could you clarify where's the memcpy that you see?
Thanks!

> So this patch switches to use virito_{stw|lduw}_phys_cached() to make
> sure the aceess is atomic.
> 
> Fixes: 86044b24e865f ("virtio: basic packed virtqueue support")
> Cc: qemu-sta...@nongnu.org
> Signed-off-by: Jason Wang 
> ---
>  hw/virtio/virtio.c | 11 ---
>  1 file changed, 4 insertions(+), 7 deletions(-)
> 
> diff --git a/hw/virtio/virtio.c b/hw/virtio/virtio.c
> index cc69a9b881..939bcbfeb9 100644
> --- a/hw/virtio/virtio.c
> +++ b/hw/virtio/virtio.c
> @@ -507,11 +507,9 @@ static void vring_packed_desc_read_flags(VirtIODevice 
> *vdev,
>   MemoryRegionCache *cache,
>   int i)
>  {
> -address_space_read_cached(cache,
> -  i * sizeof(VRingPackedDesc) +
> -  offsetof(VRingPackedDesc, flags),
> -  flags, sizeof(*flags));
> -virtio_tswap16s(vdev, flags);
> +hwaddr off = i * sizeof(VRingPackedDesc) + offsetof(VRingPackedDesc, 
> flags);
> +
> +*flags = virtio_lduw_phys_cached(vdev, cache, off);
>  }
>  
>  static void vring_packed_desc_read(VirtIODevice *vdev,
> @@ -564,8 +562,7 @@ static void vring_packed_desc_write_flags(VirtIODevice 
> *vdev,
>  {
>  hwaddr off = i * sizeof(VRingPackedDesc) + offsetof(VRingPackedDesc, 
> flags);
>  
> -virtio_tswap16s(vdev, >flags);
> -address_space_write_cached(cache, off, >flags, 
> sizeof(desc->flags));
> +virtio_stw_phys_cached(vdev, cache, off, desc->flags);
>  address_space_cache_invalidate(cache, off, sizeof(desc->flags));
>  }
>  
> -- 
> 2.25.1




Re: [PATCH 1/2] virtio: use virtio accessor to access packed descriptor flags

2021-11-10 Thread Philippe Mathieu-Daudé
On 11/11/21 07:38, Jason Wang wrote:
> We used to access packed descriptor flags via
> address_space_{write|read}_cached(). When we hit the cache, memcpy()
> is used which is not an atomic operation which may lead a wrong value
> is read or wrote.
> 
> So this patch switches to use virito_{stw|lduw}_phys_cached() to make
> sure the aceess is atomic.
> 
> Fixes: 86044b24e865f ("virtio: basic packed virtqueue support")
> Cc: qemu-sta...@nongnu.org
> Signed-off-by: Jason Wang 
> ---
>  hw/virtio/virtio.c | 11 ---
>  1 file changed, 4 insertions(+), 7 deletions(-)

Reviewed-by: Philippe Mathieu-Daudé