Convert the ptimer test cases to the transaction-based ptimer API,
by changing to ptimer_init(), dropping the now-unused QEMUBH
variables, and surrounding each set of changes to the ptimer
state in ptimer_transaction_begin/commit calls.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20191008171740.9679-4-peter.maydell@linaro.org
Provide the new transaction-based API. If a ptimer is created
using ptimer_init() rather than ptimer_init_with_bh(), then
instead of providing a QEMUBH, it provides a pointer to the
callback function directly, and has opted into the transaction
API. All calls to functions which modify ptimer state:
- ptimer_set_period()
- ptimer_set_freq()
- ptimer_set_limit()
- ptimer_set_count()
- ptimer_run()
- ptimer_stop()
must be between matched calls to ptimer_transaction_begin()
and ptimer_transaction_commit(). When ptimer_transaction_commit()
is called it will evaluate the state of the timer after all the
changes in the transaction, and call the callback if necessary.
In the old API the individual update functions generally would
call ptimer_trigger() immediately, which would schedule the QEMUBH.
In the new API the update functions will instead defer the
"set s->next_event and call ptimer_reload()" work to
ptimer_transaction_commit().
Because ptimer_trigger() can now immediately call into the
device code which may then call other ptimer functions that
update ptimer_state fields, we must be more careful in
ptimer_reload() not to cache fields from ptimer_state across
the ptimer_trigger() call. (This was harmless with the QEMUBH
mechanism as the BH would not be invoked until much later.)
We use assertions to check that:
* the functions modifying ptimer state are not called outside
a transaction block
* ptimer_transaction_begin() and _commit() calls are paired
* the transaction API is not used with a QEMUBH ptimer
There is some slight repetition of code:
* most of the set functions have similar looking "if s->bh
call ptimer_reload, otherwise set s->need_reload" code
* ptimer_init() and ptimer_init_with_bh() have similar code
We deliberately don't try to avoid this repetition, because
it will all be deleted when the QEMUBH version of the API
is removed.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20191008171740.9679-3-peter.maydell@linaro.org
Currently the ptimer design uses a QEMU bottom-half as its
mechanism for calling back into the device model using the
ptimer when the timer has expired. Unfortunately this design
is fatally flawed, because it means that there is a lag
between the ptimer updating its own state and the device
callback function updating device state, and guest accesses
to device registers between the two can return inconsistent
device state.
We want to replace the bottom-half design with one where
the guest device's callback is called either immediately
(when the ptimer triggers by timeout) or when the device
model code closes a transaction-begin/end section (when the
ptimer triggers because the device model changed the
ptimer's count value or other state). As the first step,
rename ptimer_init() to ptimer_init_with_bh(), to free up
the ptimer_init() name for the new API. We can then convert
all the ptimer users away from ptimer_init_with_bh() before
removing it entirely.
(Commit created with
git grep -l ptimer_init | xargs sed -i -e 's/ptimer_init/ptimer_init_with_bh/'
and three overlong lines folded by hand.)
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20191008171740.9679-2-peter.maydell@linaro.org
Host kernel within [4.18, 5.3] report an erroneous KVM_MAX_VCPUS=512
for ARM. The actual capability to instantiate more than 256 vcpus
was fixed in 5.4 with the upgrade of the KVM_IRQ_LINE ABI to support
vcpu id encoded on 12 bits instead of 8 and a redistributor consuming
a single KVM IO device instead of 2.
So let's check this capability when attempting to use more than 256
vcpus within any ARM kvm accelerated machine.
Signed-off-by: Eric Auger <eric.auger@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Andrew Jones <drjones@redhat.com>
Acked-by: Marc Zyngier <maz@kernel.org>
Message-id: 20191003154640.22451-4-eric.auger@redhat.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Host kernels that expose the KVM_CAP_ARM_IRQ_LINE_LAYOUT_2 capability
allow injection of interrupts along with vcpu ids larger than 255.
Let's encode the vpcu id on 12 bits according to the upgraded KVM_IRQ_LINE
ABI when needed.
Given that we have two callsites that need to assemble
the value for kvm_set_irq(), a new helper routine, kvm_arm_set_irq
is introduced.
Without that patch qemu exits with "kvm_set_irq: Invalid argument"
message.
Signed-off-by: Eric Auger <eric.auger@redhat.com>
Reported-by: Zenghui Yu <yuzenghui@huawei.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Andrew Jones <drjones@redhat.com>
Acked-by: Marc Zyngier <maz@kernel.org>
Message-id: 20191003154640.22451-3-eric.auger@redhat.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Block layer patches:
- block: Fix crash with qcow2 partial cluster COW with small cluster
sizes (misaligned write requests with BDRV_REQ_NO_FALLBACK)
- qcow2: Fix integer overflow potentially causing corruption with huge
requests
- vhdx: Detect truncated image files
- tools: Support help options for --object
- Various block-related replay improvements
- iotests/028: Fix for long $TEST_DIRs
# gpg: Signature made Mon 14 Oct 2019 17:02:54 BST
# gpg: using RSA key 7F09B272C88F2FD6
# gpg: Good signature from "Kevin Wolf <kwolf@redhat.com>" [full]
# Primary key fingerprint: DC3D EB15 9A9A F95D 3D74 56FE 7F09 B272 C88F 2FD6
* remotes/kevin/tags/for-upstream:
iotests: Test large write request to qcow2 file
qcow2: Limit total allocation range to INT_MAX
qemu-nbd: Support help options for --object
qemu-img: Support help options for --object
qemu-io: Support help options for --object
vl: Split off user_creatable_print_help()
iotests/028: Fix for long $TEST_DIRs
block: Reject misaligned write requests with BDRV_REQ_NO_FALLBACK
replay: add BH oneshot event for block layer
replay: finish record/replay before closing the disks
replay: don't drain/flush bdrv queue while RR is working
replay: update docs for record/replay with block devices
replay: disable default snapshot for record/replay
block: implement bdrv_snapshot_goto for blkreplay
block/vhdx: add check for truncated image files
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
The following statement produces a SyntaxWarning with Python 3.8:
if len(format) is 0:
scripts/tracetool/__init__.py:459: SyntaxWarning: "is" with a literal. Did you mean "=="?
Use the conventional len(x) == 0 syntax instead.
Reported-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20191010122154.10553-1-stefanha@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Without HEAD^, the following happens when you attempt a large write
request to a qcow2 file such that the number of bytes covered by all
clusters involved in a single allocation will exceed INT_MAX:
(A) handle_alloc_space() decides to fill the whole area with zeroes and
fails because bdrv_co_pwrite_zeroes() fails (the request is too
large).
(B) If handle_alloc_space() does not do anything, but merge_cow()
decides that the requests can be merged, it will create a too long
IOV that later cannot be written.
(C) Otherwise, all parts will be written separately, so those requests
will work.
In either B or C, though, qcow2_alloc_cluster_link_l2() will have an
overflow: We use an int (i) to iterate over nb_clusters, and then
calculate the L2 entry based on "i << s->cluster_bits" -- which will
overflow if the range covers more than INT_MAX bytes. This then leads
to image corruption because the L2 entry will be wrong (it will be
recognized as a compressed cluster).
Even if that were not the case, the .cow_end area would be empty
(because handle_alloc() will cap avail_bytes and nb_bytes at INT_MAX, so
their difference (which is the .cow_end size) will be 0).
So this test checks that on such large requests, the image will not be
corrupted. Unfortunately, we cannot check whether COW will be handled
correctly, because that data is discarded when it is written to null-co
(but we have to use null-co, because writing 2 GB of data in a test is
not quite reasonable).
Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
When the COW areas are included, the size of an allocation can exceed
INT_MAX. This is kind of limited by handle_alloc() in that it already
caps avail_bytes at INT_MAX, but the number of clusters still reflects
the original length.
This can have all sorts of effects, ranging from the storage layer write
call failing to image corruption. (If there were no image corruption,
then I suppose there would be data loss because the .cow_end area is
forced to be empty, even though there might be something we need to
COW.)
Fix all of it by limiting nb_clusters so the equivalent number of bytes
will not exceed INT_MAX.
Cc: qemu-stable@nongnu.org
Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Instead of parsing help options as normal object properties and
returning an error, provide the same help functionality as the system
emulator in qemu-nbd, too.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Instead of parsing help options as normal object properties and
returning an error, provide the same help functionality as the system
emulator in qemu-img, too.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Instead of parsing help options as normal object properties and
returning an error, provide the same help functionality as the system
emulator in qemu-io, too.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Printing help for --object is something that we not only want in the
system emulator, but also in tools that support --object. Move it into a
separate function in qom/object_interfaces.c to make the code accessible
for tools.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
For long test image paths, the order of the "Formatting" line and the
"(qemu)" prompt after a drive_backup HMP command may be reversed. In
fact, the interaction between the prompt and the line may lead to the
"Formatting" to being greppable at all after "read"-ing it (if the
prompt injects an IFS character into the "Formatting" string).
So just wait until we get a prompt. At that point, the block job must
have been started, so "info block-jobs" will only return "No active
jobs" once it is done.
Reported-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The BDRV_REQ_NO_FALLBACK flag means that an operation should only be
performed if it can be offloaded or otherwise performed efficiently.
However a misaligned write request requires a RMW so we should return
an error and let the caller decide how to proceed.
This hits an assertion since commit c8bb23cbdb if the required
alignment is larger than the cluster size:
qemu-img create -f qcow2 -o cluster_size=2k img.qcow2 4G
qemu-io -c "open -o driver=qcow2,file.align=4k blkdebug::img.qcow2" \
-c 'write 0 512'
qemu-io: block/io.c:1127: bdrv_driver_pwritev: Assertion `!(flags & BDRV_REQ_NO_FALLBACK)' failed.
Aborted
The reason is that when writing to an unallocated cluster we try to
skip the copy-on-write part and zeroize it using BDRV_REQ_NO_FALLBACK
instead, resulting in a write request that is too small (2KB cluster
size vs 4KB required alignment).
Signed-off-by: Alberto Garcia <berto@igalia.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Replay is capable of recording normal BH events, but sometimes
there are single use callbacks scheduled with aio_bh_schedule_oneshot
function. This patch enables recording and replaying such callbacks.
Block layer uses these events for calling the completion function.
Replaying these calls makes the execution deterministic.
Signed-off-by: Pavel Dovgalyuk <Pavel.Dovgaluk@ispras.ru>
Acked-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
After recent updates block devices cannot be closed on qemu exit.
This happens due to the block request polling when replay is not finished.
Therefore now we stop execution recording before closing the block devices.
Signed-off-by: Pavel Dovgalyuk <pavel.dovgaluk@ispras.ru>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
In record/replay mode bdrv queue is controlled by replay mechanism.
It does not allow saving or loading the snapshots
when bdrv queue is not empty. Stopping the VM is not blocked by nonempty
queue, but flushing the queue is still impossible there,
because it may cause deadlocks in replay mode.
This patch disables bdrv_drain_all and bdrv_flush_all in
record/replay mode.
Stopping the machine when the IO requests are not finished is needed
for the debugging. E.g., breakpoint may be set at the specified step,
and forcing the IO requests to finish may break the determinism
of the execution.
Signed-off-by: Pavel Dovgalyuk <pavel.dovgaluk@ispras.ru>
Acked-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This patch updates the description of the command lines for using
record/replay with attached block devices.
Signed-off-by: Pavel Dovgalyuk <Pavel.Dovgaluk@ispras.ru>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This patch disables setting '-snapshot' option on by default
in record/replay mode. This is needed for creating vmstates in record
and replay modes.
Signed-off-by: Pavel Dovgalyuk <pavel.dovgaluk@ispras.ru>
Acked-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Peter Maydell
Eric AugerandPeter Maydell
Stefan Hajnoczi
Max ReitzandKevin Wolf
Alberto GarciaandKevin Wolf
Pavel DovgalyukandKevin Wolf