Even for writes to already allocated clusters, an l2meta is allocated,
though it stays effectively unused. After this patch, only allocating
requests still have one. Each l2meta now describes an in-flight request
that writes to clusters that are not yet hooked up in the L2 table.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
There's no real reason to have an l2meta for normal requests that don't
allocate anything. Before we can get rid of it, we must return the host
cluster offset in a different way.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
As soon as delayed COW is introduced, the l2meta struct is needed even
after completion of the request, so it can't live on the stack.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This makes it easier to address the areas for which a COW must be
performed. As a nice side effect, the COW code in
qcow2_alloc_cluster_link_l2 becomes really trivial.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The offset within the cluster is already present as n_start and this is
what the code uses. QCowL2Meta.offset is only needed at a cluster
granularity.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Per the datasheet, the DAY (wday) register is user defined. Implement this.
Signed-off-by: Antoine Mathys <barsamin@gmail.com>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Per the datasheet, the mapping between 12 and 24 hours modes is:
0 <-> 12 PM
1-12 <-> 1-12 AM
13-23 <-> 1-11 PM
Signed-off-by: Antoine Mathys <barsamin@gmail.com>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
The post_load timer was being freed, but not deleted. This could cause
problems when the timer is armed, but the device is hot-unplugged before
the callback is executed.
Signed-off-by: Amit Shah <amit.shah@redhat.com>
This saves us a few bytes in the VirtIOSerial struct. Not a big
savings, but since the entire structure is used only during a short
while after migration, it's helpful to keep the struct cleaner and
smaller.
Signed-off-by: Amit Shah <amit.shah@redhat.com>
The virtio_serial_load() function became too big, split the code that
gets the port info from the source into a separate function.
Signed-off-by: Amit Shah <amit.shah@redhat.com>
Tray statuses should be also reseted. Some guests may lock the tray
and after reset before any kernel is loaded the tray should be unlocked.
Also if you reset the real computer the tray is closed. We should
do the same in qemu.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
To make it easier to move code around without breaking
build at intermedite steps, tweak makefiles
to look in pci/ and hw/ for include files, automatically.
This will be reverted at the end of the reorganization.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
For tap, we currently assume the vnet header size is 10
(the default value) but that might not be the case
if tap is persistent and has been used by qemu previously.
To fix, set vnet header size correctly on open.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Cleanup the q35/ich9 license headers.
Signed-off-by: Jason Baron <jbaron@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Isaku Yamahata <yamahata@valinux.co.jp>
madvise(DONTNEED) will throw away the contents of the whole page at the
given address, even if the given length is less than the page size. One
can argue about whether that's the correct behaviour, but that's what it's
done for a long time in Linux at least.
That means that the madvise() in ram_load(), on a setup where
TARGET_PAGE_SIZE is smaller than the host page size, can throw away data
in guest pages adjacent to the one it's actually processing right now,
leading to guest memory corruption on an incoming migration.
This patch therefore, disables the madvise() if the host page size is
larger than TARGET_PAGE_SIZE. This means we don't get the benefits of that
madvise() in this case, but a more complete fix is more difficult to
accomplish. This at least fixes the guest memory corruption.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Reported-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
The code for migrating (or savevm-ing) memory pages starts off by creating
a dirty bitmap and filling it with 1s. Except, actually, because bit
addresses are 0-based it fills every bit except bit 0 with 1s and puts an
extra 1 beyond the end of the bitmap, potentially corrupting unrelated
memory. Oops. This patch fixes it.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
This adds some first tests for qcow2's dependency handling when two
parallel write requests access the same cluster.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
We want to use these events to suspend requests for testing concurrent
AIO requests. Suspending requests while they are holding the CoMutex is
rather boring for this purpose.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This makes the blkdebug suspend/resume functionality available in
qemu-io. Use it like this:
$ ./qemu-io blkdebug::/tmp/test.qcow2
qemu-io> break write_aio req_a
qemu-io> aio_write 0 4k
qemu-io> blkdebug: Suspended request 'req_a'
qemu-io> resume req_a
blkdebug: Resuming request 'req_a'
qemu-io> wrote 4096/4096 bytes at offset 0
4 KiB, 1 ops; 0:00:30.71 (133.359788 bytes/sec and 0.0326 ops/sec)
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This allows more systematic AIO testing. The patch adds three new
operations to blkdebug:
* Setting a "breakpoint" on a blkdebug event. The next request that
triggers this breakpoint is suspended and is tagged with a name.
The breakpoint is removed after a request has triggered it.
* A suspended request (identified by it's tag) can be resumed
* It's possible to check whether a suspended request with a given
tag exists. This can be used for waiting for an event.
Ideally, we would instead tag requests right when they are created and
set breakpoints for individual requests. However, at this point the
block layer doesn't allow this easily, and breakpoints that trigger for
any request already allow a lot of useful testing.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The cleanup work to remove a rule depends on the type of the rule. It's
easy for the existing rules as there is no data that must be cleaned up
and is specific to a type yet, but the next patch will change this.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
As soon as new rules can be set during runtime, as introduced by the
next patch, blkdebug makes sense even without a config file.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
We will use qemu_opts_create_nofail function, it can make code
more readable.
Signed-off-by: Dong Xu Wang <wdongxu@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
While id is NULL, qemu_opts_create can not fail, so ignore
errors is fine.
Signed-off-by: Dong Xu Wang <wdongxu@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
It will set opt->str in qemu_opt_set_bool, without opt->str, there
will be some potential bugs.
These are uses of opt->str, and what happens when it isn't set:
* qemu_opt_get(): returns NULL, which means "not set". Bug can bite
when value isn't the default value.
* qemu_opt_parse(): passes NULL to parse_option_bool(), which treats it
like "on". Wrong if the value is actually false. Bug can bite when
qemu_opts_validate() runs after qemu_opt_set_bool().
* qemu_opt_del(): passes NULL to g_free(), which is just fine.
* qemu_opt_foreach(): passes NULL to the callback, which is unlikely to
be prepared for it.
* qemu_opts_print(): prints NULL, which crashes on some systems.
* qemu_opts_to_qdict(): passes NULL to qstring_from_str(), which
crashes.
It also makes qemu_opt_set_bool more readable by using find_desc_by_name
and opts_accepts_any.
It is based on Luiz's patch and uses Markus's comments. Discussions can
be found at:
http://lists.nongnu.org/archive/html/qemu-devel/2012-07/msg02716.html
Signed-off-by: Dong Xu Wang <wdongxu@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Use opts_accepts_any() and find_desc_by_name().
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
Signed-off-by: Dong Xu Wang <wdongxu@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The new functions are opts_accepts_any() and find_desc_by_name(), which
are also going to be used by qemu_opts_validate() (see next commit).
This also makes opt_set() slightly more readable.
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
Signed-off-by: Dong Xu Wang <wdongxu@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Check for a 0 "distance" value to avoid infinite loop when the
expired FCR timer was not programed with auto-increment.
With this change the behavior is coherent with the same type
of code in the exynos4210_gfrc_restart() function in the same
file.
Linux seems to mostly use this timer with auto-increment
which explain why it is not a problem most of the time.
However other OS might have a problem with this if they
don't use the auto-increment feature.
Signed-off-by: Jean-Christophe DUBOIS <jcd@tribudubois.net>
Reviewed-by: Evgeny Voevodin <e.voevodin@samsung.com>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Fix a bug on the ARM GIC model where interrupts are not
set pending on the correct target CPUs when they are
triggered by writes to the Interrupt Set Enable or
Set Pending registers.
Signed-off-by: Daniel Sangorrin <dsl@ertl.jp>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
commit 7b482bcf xilinx_zynq: added QSPI controller
Adds one QSPI controller, which has two spi buses, one is for
spi0, and another is for spi1. But when initializing the spi1
bus, "dev" has been overwrited by the ssi_create_slave_no_init() function,
so that qdev_get_child_bus() returns NULL and the last two m25p80 flashes
won't be attached to the spi1 bus, but to main-system-bus.
Here we add one variable to avoid overwriting.
Signed-off-by: Liming Wang <walimisdev@gmail.com>
Reviewed-by: Peter Crosthwaite <peter.crosthwaite@xilinx.com>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
The GIC architecture specification for v1 and v2 GICs (as found
on the Cortex-A9 and newer) states that the GICC_PMR reset value
is zero; this differs from the 0xf0 reset value used on 11MPCore.
The NVIC is different again in not having a CPU interface; since
we share the GIC code we must force the priority mask field to
allow through all interrupts.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Igor Mitsyanko <i.mitsyanko@samsung.com>
The GIC spec states that only interrupts with higher priority
than the value in the GICC_PMR priority mask register are
passed through to the processor. We were incorrectly allowing
through interrupts with a priority equal to the specified
value: correct the comparison operation to match the spec.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Igor Mitsyanko <i.mitsyanko@samsung.com>
Fix the code in the secondary CPU boot stubs so that it correctly
initialises the GIC rather than relying on bugs or implementation
dependent aspects of the QEMU GIC implementation:
* set the GIC_PMR.Priority field to all-ones, so that all
interrupts are passed through. The default of all-zeroes
means all interrupts are masked, and QEMU only booted because
of a bug in the priority masking in our GIC implementation.
* add a barrier after GIC setup and before WFI to ensure that
GIC config is complete before we go into a possible low power
state. This isn't needed with the software GIC model but could
be required when using KVM and executing this code on the
real hardware CPU.
Note that of the three secondary stub implementations, only
the common generic one needs to support both v6 and v7 DSB
encodings; highbank and exynos4210 will always be v7 CPUs.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Igor Mitsyanko <i.mitsyanko@samsung.com>
An error has occurred if the return value is invalid_set_file_pointer
and getlasterror doesn't return no_error.
Signed-off-by: Fabien Chouteau <chouteau@adacore.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
This one fixes a race which qemu had also in iscsi block driver
between cancellation and io completition.
qemu_rbd_aio_cancel was not synchronously waiting for the end of
the command.
To archieve this it introduces a new status flag which uses
-EINPROGRESS.
Signed-off-by: Stefan Priebe <s.priebe@profihost.ag>
Reviewed-by: Stefan Hajnoczi <stefanha@gmail.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This makes it easier to create images with both compressed and
uncompressed clusters for testing.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Corey Bryant <coreyb@linux.vnet.ibm.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
