Bug 748902

Description of problem:
qemu returns numerous read errors when run with a file-backed disk which is stored on a disk with 4k physical sectors. The underlying cause of these read errors is pread() returning EINVAL in handle_aiocb_rw_linear(). Configuring the appliance to not use cache=off makes the problem go away.

By default, qemu seems to be aligning reads to 512 bytes. This fails for a file opened with O_DIRECT when the sector size is 4096 bytes. Disks with 4k sectors are only just starting to come to the market.

# fdisk -l /dev/vg_dellper30001/lv_home
Note: sector size is 4096 (not 512)

Disk /dev/vg_dellper30001/lv_home: 5368 MB, 5368709120 bytes
255 heads, 63 sectors/track, 81 cylinders, total 1310720 sectors
Units = sectors of 1 * 4096 = 4096 bytes
Sector size (logical/physical): 4096 bytes / 4096 bytes
I/O size (minimum/optimal): 4096 bytes / 4096 bytes
Disk identifier: 0x00000000

Version-Release number of selected component (if applicable):
qemu-system-x86-0.14.0-8.fc15.x86_64

How reproducible:
100%

Steps to Reproduce:
I hit the problem with libguestfs. If you have a libguestfs source tree, and a physical disk with 4k sectors, after configure and make, run:

make -C images check
  
Actual results:
This will fail almost immediately in pvcreate.

Expected results:
qemu should check the sector size with the appropriate ioctl rather than using a hard-coded value of 512 bytes. It should Just Work(TM).