Re: [Qemu-devel] [PATCH/RFC] block: Ensure that block size constraints are considered

[Paolo Bonzini]

Il 21/11/2012 10:15, Kevin Wolf ha scritto:
>> > +    if ((bs->open_flags & BDRV_O_NOCACHE)) {
>> > +        bs->file->buffer_alignment = align;
>> > +    }
> Any reason to restrict this to BDRV_O_NOCACHE?
> 
> There have been patches to change the BDRV_O_NOCACHE flag from the
> monitor, in which case bdrv_set_buffer_alignment() wouldn't be called
> anew and O_DIRECT requests start to fail again.
> 

bdrv_set_buffer_alignment() is completely broken.  It should set host
alignment, but in fact it is passed the guest alignment.

In practice, we only support logical_block_size matching the host's or
bigger (which is unsafe due to torn writes, but works).

So I suggest that we just look at writes outside the device models, and
"fix" them to always read a multiple of 4k.

Paolo