On Thu, Apr 07, 2011 at 07:10:00AM +0800, Lyu Mitnick wrote: > Would you mind to share your solution about this feature?? My idea is > replace > all bdrv_pread(), bdrv_pwrite_sync(), bdrv_pwrite() in block/vpc.c to > bdrv_pread_split(), bdrv_pwrite_sync_split(), bdrv_pwrite_split() > corresponding. > bdrv_pread_split(), bdrv_pwrite_sync_split(), bdrv_pwrite_split() should > deal with > data splitted at different files via bdrv_pread(), bdrv_pwrite_sync(), > bdrv_pwrite(), > give an interface of sequential byte accessed. Is it suitable to solve this > issue??
For the synchronous bdrv_read/bdrv_write used in vmdk the problem actually is rather trivial, as it already split it's reads/write on a cluster boundary internally, and the higher level split is always alignment to a cluster boundary. So basically no new work there.