The consistency check on open is necessary in order to fix inconsistent
table offsets left as a result of a crash mid-operation. Images with a
backing file actually flush before updating table offsets and are
therefore guaranteed to be consistent. Do not mark these images dirty.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The len and is_write arguments to cpu_physical_memory_unmap() were
swapped. This patch changes calls to use the correct argument ordering.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
It is not possible to use virtio-ioeventfd when building without an I/O
thread. We rely on a signal to kick us out of vcpu execution. Timers
and AIO use SIGALRM and SIGUSR2 respectively. Unfortunately eventfd
does not support O_ASYNC (SIGIO) so eventfd cannot be used in a signal
driven manner.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
USB Mass Storage Devices sometimes have the RMB (removable) bit set in
the SCSI INQUIRY response. Thumbdrives tend to have the bit set whereas
hard disks do not.
Operating systems differentiate between removable devices and fixed
devices. Under Linux, the anaconda installer looks for removable
devices. Under Windows, only fixed devices may have more than one
partition and AutoRun is also affected by the removable bit.
For these reasons, allow USB Mass Storage Devices to override the
removable bit:
qemu -usb
-drive if=none,file=test.img,cache=none,id=disk0
-device usb-storage,drive=disk0,removable=on
The default is off.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
scsi-disk devices may wish to override the removable bit. Add support
for a qdev property on SCSI devices. This is will be used by usb-msd.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Provide the "removable" qdev property bit to override the SCSI INQUIRY
removable (RMB) bit for non-CDROM devices. This will be used by USB
Mass Storage Devices, which sometimes have this guest-visible bit set
and sometimes do not. They therefore requires a means for user
configuration.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The backing format should be honored during image creation. For some
reason we currently use the image format to open the backing file. This
fails when the backing file has a different format than the image being
created. Keep the image and backing format drivers completely separate.
Also print the backing filename if there is an error opening the backing
file instead of the image filename.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Acked-by: Jes Sorensen <Jes.Sorensen@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
QED relies on the underlying filesystem to extend the file and maintain
its size. Check that images are not created on a block device.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Virtqueue notify is currently handled synchronously in userspace virtio. This
prevents the vcpu from executing guest code while hardware emulation code
handles the notify.
On systems that support KVM, the ioeventfd mechanism can be used to make
virtqueue notify a lightweight exit by deferring hardware emulation to the
iothread and allowing the VM to continue execution. This model is similar to
how vhost receives virtqueue notifies.
The result of this change is improved performance for userspace virtio devices.
Virtio-blk throughput increases especially for multithreaded scenarios and
virtio-net transmit throughput increases substantially.
Some virtio devices are known to have guest drivers which expect a notify to be
processed synchronously and spin waiting for completion.
For virtio-net, this also seems to interact with the guest stack in strange
ways so that TCP throughput for small message sizes (~200bytes)
is harmed. Only enable ioeventfd for virtio-blk for now.
Care must be taken not to interfere with vhost-net, which uses host
notifiers. If the set_host_notifier() API is used by a device
virtio-pci will disable virtio-ioeventfd and let the device deal with
host notifiers as it wishes.
Finally, there used to be a limit of 6 KVM io bus devices inside the
kernel. On such a kernel, don't use ioeventfd for virtqueue host
notification since the limit is reached too easily. This ensures that
existing vhost-net setups (which always use ioeventfd) have ioeventfds
available so they can continue to work.
After migration and on VM change state (running/paused) virtio-ioeventfd
will enable/disable itself.
* VIRTIO_CONFIG_S_DRIVER_OK -> enable virtio-ioeventfd
* !VIRTIO_CONFIG_S_DRIVER_OK -> disable virtio-ioeventfd
* virtio_pci_set_host_notifier() -> disable virtio-ioeventfd
* vm_change_state(running=0) -> disable virtio-ioeventfd
* vm_change_state(running=1) -> enable virtio-ioeventfd
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
There used to be a limit of 6 KVM io bus devices in the kernel.
On such a kernel, we can't use many ioeventfds for host notification
since the limit is reached too easily.
Add an API to test for this condition.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
The VirtIOPCIProxy bugs field is currently used to enable workarounds
for older guests. Rename it to flags so that other per-device behavior
can be tracked.
A later patch uses the flags field to remember whether ioeventfd should
be used for virtqueue host notification.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
For some reason the carets ('^') in the QED specification disappeared.
This patch puts them back.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This patch adds support for the qemu-img check command. It also
introduces a dirty bit in the qed header to mark modified images as
needing a check. This bit is cleared when the image file is closed
cleanly.
If an image file is opened and it has the dirty bit set, a consistency
check will run and try to fix corrupted table offsets. These
corruptions may occur if there is power loss while an allocating write
is performed. Once the image is fixed it opens as normal again.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This patch implements the read/write state machine. Operations are
fully asynchronous and multiple operations may be active at any time.
Allocating writes lock tables to ensure metadata updates do not
interfere with each other. If two allocating writes need to update the
same L2 table they will run sequentially. If two allocating writes need
to update different L2 tables they will run in parallel.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This patch adds code to look up data cluster offsets in the image via
the L1/L2 tables. The L2 tables are writethrough cached in memory for
performance (each read/write requires a lookup so it is essential to
cache the tables).
With cluster lookup code in place it is possible to implement
bdrv_is_allocated() to query the number of contiguous
allocated/unallocated clusters.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This patch introduces the qed on-disk layout and implements image
creation. Later patches add read/write and other functionality.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
discard [-Cq] off len -- discards a number of bytes at a specified
offset
discards a range of bytes from the given offset
Example:
'discard 512 1k' - discards 1 kilobyte from 512 bytes into the file
Discards a segment of the currently open file.
-C, -- report statistics in a machine parsable format
-q, -- quite mode, do not show I/O statistics
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
We register the vm change state handler in a PCI BAR map() function.
This function can be called multiple times throughout the lifetime of a
PCI IDE device. This results in duplicate vm change state handlers
being register, none of which are ever unregistered.
Instead, register the vm change state handler in the device's init
function once and for all.
piix tested, cmd646 and via not tested.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Backing filenames may contain a protocol. The code currently doesn't
consider this case and produces filenames that embed "<protocol>:".
Don't combine filenames if the backing filename contains a protocol.
Based on an earlier patch by Anthony Liguori <aliguori@us.ibm.com>.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The bdrv_find_protocol() function returns NULL if an unknown protocol
name is given. It returns the "file" protocol when the filename
contains no protocol at all. This makes it difficult to distinguish
between paths which contain a protocol and those which do not.
Factor out a helper function that tests whether or not a filename has a
protocol. The next patch makes use of this function.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The qemu-img create command should check the backing format to ensure
only image files with valid backing formats are created. By checking in
qemu-img.c we can print a useful error message.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Free option parameter lists in the img_create() error return path.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
parse_option_parameters() may need to create a new option parameter list
from a template list. Use append_option_parameters() instead of
duplicating the code.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Filenames may start with "<protocol>:" to explicitly use a protocol like
nbd. Filenames with unknown protocols are rejected in most of QEMU
except for bdrv_create_file(). Even if a file with an invalid filename
can be created, QEMU cannot use it since all the other relevant
functions reject such paths. Make bdrv_create_file() consistent.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Remove the debugging fprintf() slipped in via the following commit:
commit b2e0a138e7
Author: Michael S. Tsirkin <mst@redhat.com>
Date: Mon Nov 22 19:52:34 2010 +0200
migration: stable ram block ordering
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
SCSI read/write requests should not be re-issued before the current
fragment of I/O completes. There are asserts in scsi-disk.c that guard
this constraint but they trigger on SPARC Linux 2.4. It turns out that
the asserts are too early in the code path and don't allow for read
requests to terminate.
Only the read assert needs to be moved but move the write assert too for
consistency.
Reported-by: Nigel Horne <njh@bandsman.co.uk>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Errors should be logged using error_report() so they go to the
appropriate monitor.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
VM state change notifications are invoked from vm_start()/vm_stop().
Trace these state changes so we can reason about the state of the VM
from trace output.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Since commit 4bed983730 an .fd_read()
handler that deletes its IOHandler is exposed to .fd_write() being
called on the deleted IOHandler.
This patch fixes deletion so that .fd_read() and .fd_write() are never
called on an IOHandler that is marked for deletion.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
The L1 table grow operation includes a size calculation that bumps up
the new L1 table size in order to anticipate the size needs of vmstate
data. This helps reduce the number of times that the L1 table has to be
grown when vmstate data is appended.
This size overhead is not necessary during image creation,
bdrv_truncate(), or snapshot goto operations. In fact, existing
qemu-iotests that exercise table growth are no longer able to trigger it
because image creation preallocates an L1 table that is too large after
changes to qcow_create2().
This patch keeps the size calculation but also adds exact growth for
callers that do not want to inflate the L1 table size unnecessarily.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Document the restriction that format strings must begin and end with
double quotes. This is for easy parsing since we don't run cpp over
trace-events.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
The regular expression to parse trace event definitions assumed the
format string would be a simple double-quoted string. However, we now
use PRI?64 for portability which splits string literals. The regular
expression can disregard the format string entirely since simpletrace.py
never needs to use it.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Timestamp files were recently added to reduce make churn on source files
that use tracing. The timestamp files should never be committed and
should not be visible in git status.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Observing block layer aio readv/writev operations is useful for
debugging image formats or understanding guest disk I/O patterns.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
The LTTng UserSpace Tracer formerly used TPPROTO() and TPARGS() instead
of TP_PROTO() and TP_ARGS() like the kernel uses. This has been changed
so QEMU needs to follow.
I am not aware of a graceful way of making the transition but since no
one complained that the UST build is broken, it should be fine to just
switch over without compatibility for old UST headers. The newer UST
headers are shipping in distro packages so it is realistic to make this
change now.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Although comment lines must be skipped, the '#' character can occur in
valid format strings. Be more careful when checking for comments.
Leave comments at the end of the line where they will not interfere with
other processing.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
It is not portable to use "%ld" for int64_t because int64_t may have
type long on 64-bit platforms and long long on 32-bit platforms. Use
the standard library PRId64 macros to keep format strings portable.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
The console_select() function does not check that active_console is
non-NULL before dereferencing it. When invoked with qemu -nodefaults it
is possible to hit this case.
This patch checks that active_console is non-NULL before stashing away
the old console dimensions in console_select().
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Acked-by: Gerd Hoffmann <kraxel@redhat.com>
Acked-by: Stefan Weil <weil@mail.berlios.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
The blkverify block driver makes investigating image format data
corruption much easier. A raw image initialized with the same contents
as the test image (e.g. qcow2 file) must be provided. The raw image
mirrors read/write operations and is used to verify that data read from
the test image is correct.
See docs/blkverify.txt for more information.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The OpenIndiana (Solaris) e1000g driver drops frames that are too long
or too short. It expects to receive frames of at least the Ethernet
minimum size. ARP requests in particular are small and will be dropped
if they are not padded appropriately, preventing a Solaris VM from
becoming visible on the network.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
This patch adds trace events for virtqueue operations including
adding/removing buffers, notifying the guest, and receiving a notify
from the guest.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
It is often useful to instrument memory management functions in order to
find leaks or performance problems. This patch adds trace events for
the memory allocation primitives.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
This patch adds LTTng Userspace Tracer (UST) backend support. The UST
system requires no kernel support but libust and liburcu must be
installed.
$ ./configure --trace-backend ust
$ make
Start the UST daemon:
$ ustd &
List available tracepoints and enable some:
$ ustctl --list-markers $(pgrep qemu)
[...]
{PID: 5458, channel/marker: ust/paio_submit, state: 0, fmt: "acb %p
opaque %p sector_num %lu nb_sectors %lu type %lu" 0x4b32ba}
$ ustctl --enable-marker "ust/paio_submit" $(pgrep qemu)
Run the trace:
$ ustctl --create-trace $(pgrep qemu)
$ ustctl --start-trace $(pgrep qemu)
[...]
$ ustctl --stop-trace $(pgrep qemu)
$ ustctl --destroy-trace $(pgrep qemu)
Trace results can be viewed using lttv-gui.
More information about UST:
http://lttng.org/ust
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
trace: Check for LTTng Userspace Tracer headers
When using the 'ust' backend, check if the relevant headers are
available at host.
Signed-off-by: Prerna Saxena <prerna@linux.vnet.ibm.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
This patch adds the trace-file command:
trace-file [on|off|flush]
Open, close, or flush the trace file. If no argument is given,
the status of the trace file is displayed.
The trace file is turned on by default but is only written out when the
trace buffer becomes full. The flush operation can be used to force
write out at any time.
Turning off the trace file does not change the state of trace events;
tracing will continue to the trace buffer. When the trace file is off,
use "info trace" to display the contents of the trace buffer in memory.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
This commit also contains the trace-file sub-command from the following
commit:
commit 5ce8d1a957afae2c52ad748944ce72848ccf57bd
Author: Prerna Saxena <prerna@linux.vnet.ibm.com>
Date: Wed Aug 4 16:23:54 2010 +0530
trace: Add options to specify trace file name at startup and runtime
This patch adds an optional command line switch '-trace' to specify the
filename to write traces to, when qemu starts.
Eg, If compiled with the 'simple' trace backend,
[temp@system]$ qemu -trace FILENAME IMAGE
Allows the binary traces to be written to FILENAME instead of the option
set at config-time.
Also, this adds monitor sub-command 'set' to trace-file commands to
dynamically change trace log file at runtime.
Eg,
(qemu)trace-file set FILENAME
This allows one to set trace outputs to FILENAME from the default
specified at startup.
Signed-off-by: Prerna Saxena <prerna@linux.vnet.ibm.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Sometimes it is useful to disable a trace event. Removing the event
from trace-events is not enough since source code will call the
trace_*() function for the event.
This patch makes it easy to build without specific trace events by
marking them disabled in trace-events:
disable multiwrite_cb(void *mcb, int ret) "mcb %p ret %d"
This builds without the multiwrite_cb trace event.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
trace: Allow bulk enabling/disabling of trace events at compile time
For 'simple' trace backend, allow bulk enabling/disabling of trace
events at compile time. Trace events that are preceded by 'disable'
keyword are compiled in, but turned off by default. These can
individually be turned on using the monitor. All other trace events are
enabled by default.
TODO :
This could be enhanced when the trace-event namespace is partitioned into a
group and an ID within that group. In such a case, marking a group as enabled
would automatically enable all trace-events listed under it.
Signed-off-by: Prerna Saxena <prerna@linux.vnet.ibm.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
This patch adds a simple tracer which produces binary trace files. To
try out the simple backend:
$ ./configure --trace-backend=simple
$ make
After running QEMU you can pretty-print the trace:
$ ./simpletrace.py trace-events trace.log
The output of simpletrace.py looks like this:
qemu_realloc 0.699 ptr=0x24363f0 size=0x3 newptr=0x24363f0
qemu_free 0.768 ptr=0x24363f0
^ ^---- timestamp delta (us)
|____ trace event name
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
trace: Make trace record fields 64-bit
Explicitly use 64-bit fields in trace records so that timestamps and
magic numbers work for 32-bit host builds.
Includes fixes from Prerna Saxena <prerna@linux.vnet.ibm.com>.
Signed-off-by: Prerna Saxena <prerna@linux.vnet.ibm.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
This patch introduces the trace-events file where trace events can be
declared like so:
qemu_malloc(size_t size) "size %zu"
qemu_free(void *ptr) "ptr %p"
These trace event declarations are processed by a new tool called
tracetool to generate code for the trace events. Trace event
declarations are independent of the backend tracing system (LTTng User
Space Tracing, ftrace markers, DTrace).
The default "nop" backend generates empty trace event functions.
Therefore trace events are disabled by default.
The trace-events file serves two purposes:
1. Adding trace events is easy. It is not necessary to understand the
details of a backend tracing system. The trace-events file is a
single location where trace events can be declared without code
duplication.
2. QEMU is not tightly coupled to one particular backend tracing system.
In order to support tracing across QEMU host platforms and to
anticipate new backend tracing systems that are currently maturing,
it is important to be flexible and not tied to one system.
This commit includes fixes from Prerna Saxena
<prerna@linux.vnet.ibm.com> and Blue Swirl <blauwirbel@gmail.com>.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
There is no indication whether or not the sector is allocated when
nb_sectors=1:
sector allocated at offset 64 KiB
This message is produced whether or not the sector is allocated.
Simply use the same message as the plural case, I don't think the
English is so broken that we need special case output here:
0/1 sectors allocated at offset 64 KiB
This change does not affect qemu-iotests since nb_sectors=1 is not used
there.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The "info blockstats" documentation was copy-pasted as "info block"
instead of "info blockstats". The documentation for "commands", "jit",
"numa", "qdm", and "roms" is missing. This patch resolves these issues
in qemu-monitor.hx.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
This patch expands tabs on a few lines so the code formats nicely and
follows the QEMU coding style.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The VirtIOBlockRequest structure is about 40 KB in size. This patch
avoids zeroing every request by only initializing fields that are read.
The other fields are either written to or may not be used at all.
Oprofile shows about 10% of CPU samples in memset called by
virtio_blk_alloc_request(). The workload is
dd if=/dev/vda of=/dev/null iflag=direct bs=8k running concurrently 4
times. This patch makes memset disappear to the bottom of the profile.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This patch adds the ability to grow qcow2 images in-place using
bdrv_truncate(). This enables qemu-img resize command support for
qcow2.
Snapshots are not supported and bdrv_truncate() will return -ENOTSUP.
The notion of resizing an image with snapshots could lead to confusion:
users may expect snapshots to remain unchanged, but this is not possible
with the current qcow2 on-disk format where the header.size field is
global instead of per-snapshot. Others may expect snapshots to change
size along with the current image data. I think it is safest to not
support snapshots and perhaps add behavior later if there is a
consensus.
Backing images continue to work. If the image is now larger than its
backing image, zeroes are read when accessing beyond the end of the
backing image.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This patch adds a 'resize' command to grow/shrink disk images. This
allows changing the size of disk images without copying to a new image
file. Currently only raw files support resize.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The BlockDriver bdrv_getlength function is called from the I/O code path
when checking that the request falls within the device. Unfortunately
this involves an lseek system call in the raw protocol; every read or
write request will incur this lseek cost.
Jan Kiszka <jan.kiszka@siemens.com> identified this issue and its
latency overhead. This patch caches device length in the existing
total_sectors variable so lseek calls can be avoided for fixed size
devices.
Growable devices fall back to the full bdrv_getlength code path because
I have not added logic to detect extending the size of the device in a
write.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This patch combines the lseek+read/write calls to use pread/pwrite
instead. This will result in fewer system calls and is already used by
AIO.
Thanks to Jan Kiszka <jan.kiszka@siemens.com> for identifying excessive
lseek and Christoph Hellwig <hch@lst.de> for confirming that this
approach should work.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The i loop iterator is shadowed by the next free cluster index. Both
using the variable name 'i' makes the code harder to read.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
It is safer to set backing_hd to NULL after deleting it so that any use
after deletion is obvious during development. Happy segfaulting!
This patch should be applied after Kevin Wolf's "vmdk: Convert to
bdrv_open" so that vmdk does not segfault on close.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
A new iovec array is allocated when creating a merged write request.
This patch ensures that the iovec array is deleted in addition to its
qiov owner.
Reported-by: Leszek Urbanski <tygrys@moo.pl>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Several commands have code to create a BlockDriverState and open a file.
The bdrv_new_open() function can be used to perform these steps. This
patch converts the qemu-img commands to actually use bdrv_new_open().
Replaced the bdrv_new_open() 'readonly' argument with bdrv_open()-style
flags to support generic flags like BDRV_O_NO_BACKING.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The bdrv_first linked list of BlockDriverStates is currently extern so
that block migration can iterate the list. However, since there is
already a bdrv_iterate() function there is no need to expose bdrv_first.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Stefan Hajnoczi