This protocol extension reuses the same set of grant pages for all
transactions between the front/back drivers, avoiding expensive tlb
flushes, grant table lock contention and switches between userspace
and kernel space. The full description of the protocol can be found in
the public blkif.h header.
http://xenbits.xen.org/gitweb/?p=xen.git;a=blob_plain;f=xen/include/public/io/blkif.h
Speed improvement with 15 guests performing I/O is ~450%.
Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
On ioreq_release the full ioreq was memset to 0, loosing all the data
and memory allocations inside the QEMUIOVector, which leads to a
memory leak. Create a new function to specifically reset ioreq.
Reported-by: Maik Wessler <maik.wessler@yahoo.com>
Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Touching char/char.h basically causes the whole of QEMU to
be rebuilt. Avoid this, it is usually unnecessary.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Because xs.h will be remove in future release of Xen, this patch removes the
extra includes of this headers.
Also, it removes the extra includes of xenctrl.h and xen/io/xenbus.h as there
already are in xen_common.h.
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
First offender is xen_config_dev_blk()'s use of disk->bdrv->filename.
Get the filename from disk->opts instead. Same result, except for
snapshots: there, we now get the filename specified by the user
instead of the name of the temporary image created by bdrv_open().
Should be an improvement.
Second offender is blk_init()'s use of blkdev->bs->drv->format_name.
Simply use the appropriate interface to get the format name.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Legacy (non-pvops) gntdev drivers may require this to be done when the
number of grants intended to be used simultaneously exceeds a certain
driver specific default limit.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
While for the "normal" case (called from blk_send_response_all())
decrementing requests_finished is correct, doing so in the parse error
case is wrong; requests_inflight needs to be decremented instead.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Use bdrv_aio_flush instead of bdrv_flush.
Make sure to call bdrv_aio_writev/readv after the presync bdrv_aio_flush is fully
completed and make sure to call the postsync bdrv_aio_flush after
bdrv_aio_writev/readv is fully completed.
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
This patch removes a dead option.
The same can be achieved removing BDRV_O_NOCACHE and BDRV_O_CACHE_WB
from the flags passed to bdrv_open.
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
* sstabellini/disk_io:
xen_disk: when using AIO flush after the operation is completed
xen_disk: open disk with BDRV_O_NOCACHE | BDRV_O_CACHE_WB | BDRV_O_NATIVE_AIO
We need to detach the blkdev from the BlockDriverState before calling
bdrv_delete.
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
All files under GPLv2 will get GPLv2+ changes starting tomorrow.
event_notifier.c and exec-obsolete.h were only ever touched by Red Hat
employees and can be relicensed now.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Xen_disk.c has support for using synchronous I/O instead of asynchronous,
but it is compiled out by default. Remove it.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The synchronous .bdrv_flush callback doesn't exist any more and a device really
shouldn't poke into the block layer internals anyway. All drivers are supposed
to have a correctly working bdrv_flush, so let's just hard-code this.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Rename the existing xendev 'connect' op to 'initialised' and introduce
a new 'connected' op. This new op, if defined, is called when the
backend is connected. Note that since there is no state transition this
may be called more than once.
Signed-off-by: John Haxby <john.haxby@oracle.com>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
For now, this just protects against programming errors like having the
same drive back multiple non-qdev devices, or untimely bdrv_delete().
Later commits will add other interesting uses.
While there, rename BlockDriverState member peer to dev, bdrv_attach()
to bdrv_attach_dev(), bdrv_detach() to bdrv_detach_dev(), and
bdrv_get_attached() to bdrv_get_attached_dev().
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Decouple the I/O accounting from bdrv_aio_readv/writev/flush and
make the hardware models call directly into the accounting helpers.
This means:
- we do not count internal requests from image formats in addition
to guest originating I/O
- we do not double count I/O ops if the device model handles it
chunk wise
- we only account I/O once it actuall is done
- can extent I/O accounting to synchronous or coroutine I/O easily
- implement I/O latency tracking easily (see the next patch)
I've conveted the existing device model callers to the new model,
device models that are using synchronous I/O and weren't accounted
before haven't been updated yet. Also scsi hasn't been converted
to the end-to-end accounting as I want to defer that after the pending
scsi layer overhaul.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Sometimes the toolstack uses "aio" without an additional format
identifier, in such cases use "raw".
Updated in v2:
- fix code style.
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
When disk is a cdrom and the drive is empty the "params" node in
xenstore might be missing completely: cope with it instead of
segfaulting.
Updated in v2:
- actually removed the strchr(blkdev->params, ':') that caused the
segfault;
- free all the allocated strings from xenstore before returning;
Updated in v3:
- set blkdev fields to NULL after free'ing them.
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
cppcheck report:
hw/xen_disk.c:309: style:
Variable 'len' is assigned a value that is never used
Cc: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefan Weil <weil@mail.berlios.de>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This patch updates the libxenctrl calls in Qemu to use the new interface,
otherwise Qemu wouldn't be able to build against new versions of the
library.
We check libxenctrl version in configure, from Xen 3.3.0 to Xen
unstable.
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Acked-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Alexander Graf <agraf@suse.de>
And put braces for blocks with a single statement.
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Acked-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Alexander Graf <agraf@suse.de>
Bug fix: routines 'ioreq_runio_qemu_sync' and 'ioreq_runio_qemu_aio'
won't call 'ioreq_unmap' or 'ioreq_finish' on errors, leaving ioreq in
the blkdev->inflight list and a leak.
Signed-off-by: Feiran Zheng <famcool@gmail.com>
Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
In hw/xen_disk.c, async writing ioreq is leaked when
ioreq->req.nr_segments==0, because `aio_inflight` flag is not released
properly (skipped by misplaced "break").
Signed-off-by: Feiran Zheng <famcool@gmail.com>
Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This patch can be applied to both qemu-xen and qemu and adds support
for empty write barriers to xen_disk.
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Acked-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Changing block.h or blockdev.h resulted in recompiling most objects.
Move DriveInfo typedef and BlockInterfaceType enum definitions
to qemu-common.h and rearrange blockdev.h use to decrease churn.
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
What is known today as bdrv_open2 becomes the new bdrv_open. All remaining
callers of the old function are converted to the new one. In some places they
even know the right format, so they should have used bdrv_open2 from the
beginning.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Instead of using the field 'readonly' of the BlockDriverState struct for passing the request,
pass the request in the flags parameter to the function.
Signed-off-by: Naphtali Sprei <nsprei@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
We have code for a quite a few block formats. While I trust that all
of these formats are useful at least for some people in some
circumstances, some of them are of a kind that friends don't let
friends use in production.
This patch provides an optional block format whitelist, default off.
If a whitelist is configured with --block-drv-whitelist, QEMU proper
can use only whitelisted formats. Other programs, like qemu-img, are
not affected.
Drivers for formats off the whitelist still participate in format
probing, to ensure all programs probe exactly the same. Without that,
QEMU proper would be prone to treat images with a format off the
whitelist as raw when the image's format is probed.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Problem: Our file sys-queue.h is a copy of the BSD file, but there are
some additions and it's not entirely compatible. Because of that, there have
been conflicts with system headers on BSD systems. Some hacks have been
introduced in the commits 15cc923584,
f40d753718,
96555a96d7 and
3990d09adf but the fixes were fragile.
Solution: Avoid the conflict entirely by renaming the functions and the
file. Revert the previous hacks.
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
First step cleaning up the drives handling. This one does nothing but
removing drives_table[], still it became seriously big.
drive_get_index() is gone and is replaced by drives_get() which hands
out DriveInfo pointers instead of a table index. This needs adaption in
*tons* of places all over.
The drives are now maintained as linked list.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
This patch adds a block device backend driver to qemu. It is a pure
userspace implemention using the gntdev interface. It uses "qdisk" as
backend name in xenstore so it doesn't interfere with the other existing
backends (blkback aka "vbd" and tapdisk aka "tap").
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@7223 c046a42c-6fe2-441c-8c8c-71466251a162
