[Qemu-devel] [PATCH for-2.5 0/4] block: replace IOV_MAX with BlockLimits.max_iov

[Stefan Hajnoczi <stefanha@redhat.com>]

This series depends on "[PATCH] block/mirror: limit qiov to IOV_MAX elements".

IOV_MAX has been hardcoded in several places since preadv()/pwritev()/etc
refuse to take more iovecs per call.  That's actually an implementation detail
of raw-posix and we shouldn't spread that across the codebase.

This series adds BlockLimits.max_iov (similar to BlockLimits.max_transfer_len
and other fields).  By default it is INT_MAX but raw-posix sets it to IOV_MAX.

Current IOV_MAX users are converted to blk_get_max_iov().

By the way, the IOV_MAX vs BlockLimits.max_iov naming is slightly confusing but
follows the BlockLimits field naming convention.  Once IOV_MAX is banished no
one will be bothered by the reverse naming anymore.

Suggested-by: Kevin Wolf <kwolf@redhat.com>

Stefan Hajnoczi (4):
  block: add BlockLimits.max_iov field
  block-backend: add blk_get_max_iov()
  block: replace IOV_MAX with BlockLimits.max_iov
  block/mirror: replace IOV_MAX with blk_get_max_iov()

 block/block-backend.c          | 5 +++++
 block/io.c                     | 6 +++++-
 block/mirror.c                 | 6 ++++--
 block/raw-posix.c              | 1 +
 hw/block/virtio-blk.c          | 2 +-
 include/block/block_int.h      | 3 +++
 include/sysemu/block-backend.h | 1 +
 7 files changed, 20 insertions(+), 4 deletions(-)

-- 
2.4.3