This patch changes where the throttling state is stored (used to be the
BlockDriverState, now it is the BlockBackend), but it doesn't actually
make it a BB level feature yet. For example, throttling is still
disabled when the BDS is detached from the BB.
Signed-off-by: NKevin Wolf <kwolf@redhat.com>
Acked-by: NStefan Hajnoczi <stefanha@redhat.com>

Signed-off-by: NKevin Wolf <kwolf@redhat.com>
Reviewed-by: NAlberto Garcia <berto@igalia.com>
Acked-by: NStefan Hajnoczi <stefanha@redhat.com>

Some features, like I/O throttling, are implemented outside
block-backend.c, but still want to keep information in BlockBackend,
e.g. list entries that allow keeping a list of BlockBackends.

In order to avoid exposing the whole struct layout in the public header
file, this patch introduces an embedded public struct where such
information can be added and a pair of functions to convert between
BlockBackend and BlockBackendPublic.
Signed-off-by: NKevin Wolf <kwolf@redhat.com>
Reviewed-by: NAlberto Garcia <berto@igalia.com>
Reviewed-by: NEric Blake <eblake@redhat.com>
Acked-by: NStefan Hajnoczi <stefanha@redhat.com>

It was already true in principle that a throttled BDS always has a BB
attached, except that the order of operations while attaching or
detaching a BDS to/from a BB wasn't careful enough.

This commit breaks graph manipulations while I/O throttling is enabled.
It would have been possible to keep things working with some temporary
hacks, but quite cumbersome, so it's not worth the hassle. We'll fix
things again in a minute.
Signed-off-by: NKevin Wolf <kwolf@redhat.com>
Reviewed-by: NEric Blake <eblake@redhat.com>
Reviewed-by: NAlberto Garcia <berto@igalia.com>
Acked-by: NStefan Hajnoczi <stefanha@redhat.com>

Now that there are no remaining clients, we can drop the
sector-based blk_read(), blk_write(), blk_aio_readv(), and
blk_aio_writev().  Sadly, there are still remaining
sector-based interfaces, such as blk_*discard(), or
blk_write_compressed(); those will have to wait for another
day.
Signed-off-by: NEric Blake <eblake@redhat.com>
Signed-off-by: NKevin Wolf <kwolf@redhat.com>

blk_aio_readv() and blk_aio_writev() are annoying in that they
can't access sub-sector granularity, and cannot pass flags.
Also, they require the caller to pass redundant information
about the size of the I/O (qiov->size in bytes must match
nb_sectors in sectors).

Add new blk_aio_preadv() and blk_aio_pwritev() functions to fix
the flaws. The next few patches will upgrade callers, then
finally delete the old interfaces.
Signed-off-by: NEric Blake <eblake@redhat.com>
Signed-off-by: NKevin Wolf <kwolf@redhat.com>

Sector-based blk_write() should die; convert the one-off
variant blk_write_zeroes() to use an offset/count interface
instead.  Likewise for blk_co_write_zeroes() and
blk_aio_write_zeroes().
Signed-off-by: NEric Blake <eblake@redhat.com>
Signed-off-by: NKevin Wolf <kwolf@redhat.com>

Sector-based blk_read() should die; convert the one-off
variant blk_read_unthrottled().
Signed-off-by: NEric Blake <eblake@redhat.com>
Signed-off-by: NKevin Wolf <kwolf@redhat.com>

We have several block drivers that understand BDRV_REQ_FUA,
and emulate it in the block layer for the rest by a full flush.
But without a way to actually request BDRV_REQ_FUA during a
pass-through blk_pwrite(), FUA-aware block drivers like NBD are
forced to repeat the emulation logic of a full flush regardless
of whether the backend they are writing to could do it more
efficiently.

This patch just wires up a flags argument; followup patches
will actually make use of it in the NBD driver and in qemu-io.
Signed-off-by: NEric Blake <eblake@redhat.com>
Acked-by: NDenis V. Lunev <den@openvz.org>
Signed-off-by: NKevin Wolf <kwolf@redhat.com>

It used to be an internal helper function just for implementing
bdrv_co_do_readv/writev(), but now that it's a public interface, it
deserves a name without "do" in it.
Signed-off-by: NKevin Wolf <kwolf@redhat.com>
Reviewed-by: NEric Blake <eblake@redhat.com>
Reviewed-by: NFam Zheng <famz@redhat.com>

Extract the handling of throttling from bdrv_flush_io_queue.  These
new functions will soon become BdrvChildRole callbacks, as they can
be generalized to "beginning of drain" and "end of drain".
Reviewed-by: NAlberto Garcia <berto@igalia.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>
Acked-by: NStefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: NKevin Wolf <kwolf@redhat.com>

Commit 57d6a428 neglected to pass the given flags to blk_aio_prwv(),
which broke discard by WRITE SAME for scsi-disk (the UNMAP bit would be
ignored).

Commit fc1453cd introduced the same bug for blk_write_zeroes(). This is
used for 'qemu-img convert' without has_zero_init (e.g. on a block
device) and for preallocation=falloc in parallels.

Commit 8896e088 is the version for blk_co_write_zeroes(). This function
is only used in qemu-io.
Reported-by: NMax Reitz <mreitz@redhat.com>
Signed-off-by: NKevin Wolf <kwolf@redhat.com>
Reviewed-by: NEric Blake <eblake@redhat.com>

Commit 57d6a428 broke blk_aio_write_zeroes() because in some write
functions in the call path don't have an explicit length argument but
reuse qiov->size instead. Which is great, except that write_zeroes
doesn't have a qiov, which this commit interprets as 0 bytes.
Consequently, blk_aio_write_zeroes() didn't effectively do anything.

This patch introduces an explicit acb->bytes in BlkAioEmAIOCB and uses
that instead of acb->rwco.size.

The synchronous version of the function is okay because it does pass a
qiov (with the right size and a NULL pointer as its base).
Signed-off-by: NKevin Wolf <kwolf@redhat.com>
Reviewed-by: NMax Reitz <mreitz@redhat.com>

The previous patches have successively made blk->enable_write_cache the
true source for the information whether a writethrough mode must be
implemented. The corresponding BDRV_O_CACHE_WB is only useless baggage
we're carrying around, so now's the time to remove it.

At the same time, we remove the 'cache.writeback' option parsing on the
BDS level as the only effect was setting the BDRV_O_CACHE_WB flag.

This change requires test cases that explicitly enabled the option to
drop it. Other than that and the change of the error message when
writethrough is enabled on the BDS level (from "Can't set writethrough
mode" to "doesn't support the option"), there should be no change in
behaviour.
Signed-off-by: NKevin Wolf <kwolf@redhat.com>
Reviewed-by: NMax Reitz <mreitz@redhat.com>

Whether a write cache is used or not is a decision that concerns the
user (e.g. the guest device) rather than the backend. It was already
logically part of the BB level as bdrv_move_feature_fields() always kept
it on top of the BDS tree; with this patch, the core of it (the actual
flag and the additional flushes) is also implemented there.

Direct callers of bdrv_open() must pass BDRV_O_CACHE_WB now if bs
doesn't have a BlockBackend attached.
Signed-off-by: NKevin Wolf <kwolf@redhat.com>
Reviewed-by: NMax Reitz <mreitz@redhat.com>

All callers of blk_new_open() either don't rely on the WCE bit set after
blk_new_open() because they explicitly set it anyway, or they pass
BDRV_O_CACHE_WB unconditionally.

This patch changes blk_new_open() so that it always enables writeback
mode and asserts that BDRV_O_CACHE_WB is clear. For those callers that
used to pass BDRV_O_CACHE_WB unconditionally, the flag is removed now.
Signed-off-by: NKevin Wolf <kwolf@redhat.com>
Reviewed-by: NMax Reitz <mreitz@redhat.com>

The function is unused since commit f21d96d0 ('block: Use BdrvChild in
BlockBackend').
Signed-off-by: NKevin Wolf <kwolf@redhat.com>
Reviewed-by: NEric Blake <eblake@redhat.com>

Move declarations out of qemu-common.h for functions declared in
utils/ files: e.g. include/qemu/path.h for utils/path.c.
Move inline functions out of qemu-common.h and into new files (e.g.
include/qemu/bcd.h)
Signed-off-by: NVeronia Bahaa <veroniabahaa@gmail.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

Signed-off-by: NKevin Wolf <kwolf@redhat.com>

Signed-off-by: NKevin Wolf <kwolf@redhat.com>

Signed-off-by: NKevin Wolf <kwolf@redhat.com>

Signed-off-by: NKevin Wolf <kwolf@redhat.com>

Use blk_read(), so that it goes through blk_co_preadv() like all read
requests from the BB to the BDS.
Signed-off-by: NKevin Wolf <kwolf@redhat.com>

Signed-off-by: NKevin Wolf <kwolf@redhat.com>

This patch introduces blk_co_preadv() as a central function on the
BlockBackend level that is supposed to handle all read requests from the
BB to its root BDS eventually.
Signed-off-by: NKevin Wolf <kwolf@redhat.com>

Signed-off-by: NKevin Wolf <kwolf@redhat.com>

This function iterates over all BDSs attached to a BB. We are going to
need it when rewriting bdrv_next() so it no longer uses bdrv_states.
Signed-off-by: NMax Reitz <mreitz@redhat.com>
Signed-off-by: NKevin Wolf <kwolf@redhat.com>

Move bdrv_commit_all() and bdrv_flush_all() to the BlockBackend level.
Signed-off-by: NMax Reitz <mreitz@redhat.com>
Signed-off-by: NKevin Wolf <kwolf@redhat.com>

We can basically inline it in hmp_drive_del(); monitor_remove_blk() is
called already, so we just need to call bdrv_make_anon(), too.
Signed-off-by: NMax Reitz <mreitz@redhat.com>
Signed-off-by: NKevin Wolf <kwolf@redhat.com>

Before this patch, blk_new() automatically assigned a name to the new
BlockBackend and considered it referenced by the monitor. This patch
removes the implicit monitor_add_blk() call from blk_new() (and
consequently the monitor_remove_blk() call from blk_delete(), too) and
thus blk_new() (and related functions) no longer take a BB name
argument.

In fact, there is only a single point where blk_new()/blk_new_open() is
called and the new BB is monitor-owned, and that is in blockdev_init().
Besides thus relieving us from having to invent names for all of the BBs
we use in qemu-img, this fixes a bug where qemu cannot create a new
image if there already is a monitor-owned BB named "image".

If a BB and its BDS tree are created in a single operation, as of this
patch the BDS tree will be created before the BB is given a name
(whereas it was the other way around before). This results in minor
change to the output of iotest 087, whose reference output is amended
accordingly.
Signed-off-by: NMax Reitz <mreitz@redhat.com>
Signed-off-by: NKevin Wolf <kwolf@redhat.com>

Introduce separate functions (monitor_add_blk() and
monitor_remove_blk()) which set or unset a BB name. Since the name is
equivalent to the monitor's reference to a BB, adding a name the same as
declaring the BB to be monitor-owned and removing it revokes this
status, hence the function names.
Signed-off-by: NMax Reitz <mreitz@redhat.com>
Signed-off-by: NKevin Wolf <kwolf@redhat.com>

While monitor_block_backends contains nearly all BBs, we sometimes
really need all BBs. To this end, this patch adds the block_backend
list.
Signed-off-by: NMax Reitz <mreitz@redhat.com>
Signed-off-by: NKevin Wolf <kwolf@redhat.com>

The blk_backends list does not contain all BlockBackends but only the
ones which are referenced by the monitor, and that is not necessarily
true for every BlockBackend. Rename the list to monitor_block_backends
to make that fact clear.
Signed-off-by: NMax Reitz <mreitz@redhat.com>
Signed-off-by: NKevin Wolf <kwolf@redhat.com>

Later, we will remove bdrv_commit_all() and move its contents here, and
in order to replace bdrv_commit_all() calls by calls to blk_commit_all()
before doing so, we need to add it as an alias now.
Signed-off-by: NMax Reitz <mreitz@redhat.com>
Reviewed-by: NKevin Wolf <kwolf@redhat.com>
Signed-off-by: NKevin Wolf <kwolf@redhat.com>

Instead of iterating directly through blk_backends, we can use
blk_next() instead. This gives us some abstraction from the list itself
which we can use to rename it, for example.
Signed-off-by: NMax Reitz <mreitz@redhat.com>
Signed-off-by: NKevin Wolf <kwolf@redhat.com>

We check that the guest can't write beyond the end of its disk, but for
other internal users it can make sense to allow growing a file.
Signed-off-by: NKevin Wolf <kwolf@redhat.com>

When bdrv_close_all() is called, instead of force-closing all root
BlockDriverStates, it is better to just drop the reference from all
BlockBackends and let them be closed automatically. This prevents BDS
from getting closed that are still referenced by other BDS, which may
result in loss of cached data.

This patch adds a function for doing that, but does not yet incorporate
it in bdrv_close_all().
Signed-off-by: NMax Reitz <mreitz@redhat.com>
Reviewed-by: NKevin Wolf <kwolf@redhat.com>
Signed-off-by: NKevin Wolf <kwolf@redhat.com>

Signed-off-by: NMax Reitz <mreitz@redhat.com>
Reviewed-by: NKevin Wolf <kwolf@redhat.com>
Reviewed-by: NFam Zheng <famz@redhat.com>
Signed-off-by: NKevin Wolf <kwolf@redhat.com>

It is unused now, so we can remove it.
Signed-off-by: NMax Reitz <mreitz@redhat.com>
Reviewed-by: NFam Zheng <famz@redhat.com>
Reviewed-by: NKevin Wolf <kwolf@redhat.com>
Signed-off-by: NKevin Wolf <kwolf@redhat.com>

bdrv_close() no longer signifies ejection of a medium, this is now done
by removing the BDS from the BB. Therefore, we want to have a notifier
for that in the BB instead of a close notifier in the BDS. The former is
added now, the latter is removed later.

Symmetrically, another notifier list is added that is invoked whenever a
BDS is inserted. We will need that for virtio-blk and virtio-scsi, which
can then remove their op blockers on BDS ejection and set them up on
insertion.
Signed-off-by: NMax Reitz <mreitz@redhat.com>
Reviewed-by: NKevin Wolf <kwolf@redhat.com>
Reviewed-by: NFam Zheng <famz@redhat.com>
Signed-off-by: NKevin Wolf <kwolf@redhat.com>