* remotes/armbru/tags/pull-monitor-2018-07-03-v2: (32 commits)
qapi: Polish command flags documentation in qapi-code-gen.txt
monitor: Improve some comments
qmp: Clean up capability negotiation after commit 02130314d8c
qobject: Let qobject_from_jsonf() fail instead of abort
qmp: Switch timestamp_put() to qdict_from_jsonf_nofail()
qmp: Add some comments around null responses
qmp: Simplify monitor_qmp_respond()
qmp: Replace get_qmp_greeting() by qmp_greeting()
qmp: Replace monitor_json_emitter{,raw}() by qmp_{queue,send}_response()
qmp: Use QDict * instead of QObject * for response objects
qmp: De-duplicate error response building
qobject: New qdict_from_jsonf_nofail()
monitor: Peel off @mon_global wrapper
monitor: Rename use_io_thr to use_io_thread
qmp: Don't let JSON errors jump the queue
qmp: Don't let malformed in-band commands jump the queue
tests/qmp-test: Demonstrate QMP errors jumping the queue
qmp: Simplify code around monitor_qmp_dispatch_one()
qmp: Always free QMPRequest with qmp_request_free()
qmp: Revert change to handle_qmp_command tracepoint
...
Peter Maydell [Thu, 5 Jul 2018 07:21:25 +0000 (08:21 +0100)]
Merge remote-tracking branch 'remotes/mdroth/tags/qga-pull-2018-07-03-tag' into staging
qemu-ga patch queue for soft-freeze
* add systemd suspend support
* add used/total space stats for guest-get-fsinfo
* fixes for guest-get-fsinfo over PCI bridges
* MSI installer and schema doc fixes
* guard against unbounded allocations in guest-file-read
* add some additional qga test cases
* remotes/mdroth/tags/qga-pull-2018-07-03-tag:
qga: removing bios_supports_mode
qga: systemd hibernate/suspend/hybrid-sleep support
qga: removing switch statements, adding run_process_child
qga: guest_suspend: decoupling pm-utils and sys logic
qga: bios_supports_mode: decoupling pm-utils and sys logic
qga: refactoring qmp_guest_suspend_* functions
qemu-ga: make get-fsinfo work over pci bridges
qga-win: Fixing msi upgrade disallow in WiX file
qga/schema: fix documentation for GuestOSInfo
test-qga: add trivial tests for some commands
qga-win: add driver path usage to GuestFilesystemInfo
qga: add mountpoint usage info to GuestFilesystemInfo
qga: check bytes count read by guest-file-read
qga: unset frozen state if no mount points are frozen
qmp: Clean up capability negotiation after commit 02130314d8c
qmp_greeting() offers capabilities to the client, and
qmp_qmp_capabilities() accepts or denies capabilities requested by the
client. The two compute the set of available capabilities
independently. Not nice.
Clean this up as follows. Compute available capabilities just once in
monitor_qmp_caps_reset(), and store them in Monitor member
qmp.capab_offered[]. Have qmp_greeting() and qmp_qmp_capabilities()
use that. Both are now oblivious of capability details.
qobject: Let qobject_from_jsonf() fail instead of abort
qobject_from_jsonf() aborts on error, unlike qobject_from_jsonv(),
which returns null. Since all remaining users of qobject_from_jsonf()
cope fine with null, change it to return null.
monitor_qmp_respond() takes both a response object and an error
object. If an error object is non-null, the response object must be
null, and the response is built from the error object.
Of the two callers, one always passes a null response object, and one
a null error object. Move building the response object from the error
object to the latter, and drop the error object parameter.
qmp: Replace monitor_json_emitter{,raw}() by qmp_{queue,send}_response()
monitor_json_emitter() and monitor_json_emitter_raw() are
unnecessarily general: they can send arbitrary JSON values, even
though we only ever use them for QMP, which may send only JSON
objects.
Specialize the argument from QObject * to QDict *, and rename to
qmp_queue_response(), qmp_send_response().
All callers but one lose an upcast. The lone exception gains a
downcast; the next commit will get rid of it.
All callers of qmp_build_error_object() duplicate the code to wrap it
in a response object. Replace it by qmp_error_response() that
captures the duplicated code, including error_free().
Many uses of qobject_from_jsonf() convert JSON objects. Create new
convenience function qdict_from_jsonf_nofail() that includes the
conversion to QDict. The next few commits will put it to use.
handle_qmp_command() reports JSON syntax errors right away. This is
wrong when OOB is enabled, because the errors can "jump the queue"
then.
The previous commit fixed the same bug for semantic errors, by
delaying the checking until dispatch. We can't delay the checking, so
delay the reporting.
qmp: Don't let malformed in-band commands jump the queue
handle_qmp_command() reports certain errors right away. This is wrong
when OOB is enabled, because the errors can "jump the queue" then, as
the previous commit demonstrates.
To fix, we need to delay errors until dispatch. Do that for semantic
errors, mostly by reverting ill-advised parts of commit cf869d53172
"qmp: support out-of-band (oob) execution". Bonus: doesn't run
qmp_dispatch_check_obj() twice, once in handle_qmp_command(), and
again in do_qmp_dispatch(). That's also due to commit cf869d53172.
The next commit will fix queue jumping for syntax errors.
tests/qmp-test: Demonstrate QMP errors jumping the queue
When OOB is enabled, out-of-band commands are executed right away,
everything else is queued. This lets out-of-band commands "jump the
queue".
However, certain errors are always reported right away, and therefore
can jump the queue even when the erroneous input does not request
out-of-band execution. These errors are pretty unlikely to occur in
production, but it's wrong all the same. Mark FIXME.
qmp: Simplify code around monitor_qmp_dispatch_one()
Change monitor_qmp_dispatch_one() to take its parameters unwrapped,
move monitor_resume() to the one caller that needs it, rename the
function to monitor_qmp_dispatch().
qmp: Always free QMPRequest with qmp_request_free()
monitor_qmp_dispatch_one() frees a QMPRequest manually, because it
needs to keep a reference to ->id. Premature optimization. Take an
additional reference so we can use qmp_request_free().
qmp: Revert change to handle_qmp_command tracepoint
Commit 71da4667db6 "monitor: separate QMP parser and dispatcher" moved
the handle_qmp_command tracepoint from handle_qmp_command() to
monitor_qmp_dispatch_one(). This delays tracing from enqueue time to
dequeue time. Revert that. Dequeue remains adequately visible via
tracepoint monitor_qmp_cmd_in_band.
qmp: Redo how the client requests out-of-band execution
Commit cf869d53172 "qmp: support out-of-band (oob) execution" added a
general mechanism for command-independent arguments just for an
out-of-band flag:
The "control" key is introduced to store this extra flag. "control"
field is used to store arguments that are shared by all the commands,
rather than command specific arguments. Let "run-oob" be the first.
However, it failed to reject unknown members of "control". For
instance, in QMP command
Instead of fixing this, revert the general "control" mechanism
(because YAGNI), and do it the way I initially proposed, with key
"exec-oob". Simpler code, simpler interface.
qmp qemu-ga: Revert change that accidentally made qemu-ga accept "id"
Commit cf869d53172 "qmp: support out-of-band (oob) execution" changed
how we check "id":
Note that in the patch I exported qmp_dispatch_check_obj() to be
used to check the request earlier, and at the same time allowed
"id" field to be there since actually we always allow that.
The part after "and" is ill-advised: it makes qemu-ga accept and
ignore "id". Revert.
qmp: Make "id" optional again even in "oob" monitors
Commit cf869d53172 "qmp: support out-of-band (oob) execution" made
"id" mandatory for all commands when the client accepted capability
"oob". This is rather onerous when you play with QMP by hand, and
unnecessarily so: only out-of-band commands need an ID for reliable
matching of response to command.
Revert that part of commit cf869d53172 for now, but have documentation
advise on the need to use "id" with out-of-band commands.
tests/qmp-test tests an out-of-band command overtaking a slow in-band
command. To do that, it needs:
1. An in-band command that *reliably* takes long enough to be
overtaken.
2. An out-of-band command to do the overtaking.
3. To avoid delays, a way to make the in-band command complete quickly
after it was overtaken.
To satisfy these needs, commit 469638f9cb3 provides the rather
peculiar oob-capable QMP command x-oob-test:
* With "lock": true, it waits for a global semaphore.
* With "lock": false, it signals the global semaphore.
To satisfy 1., the test runs x-oob-test in-band with "lock": true.
To satisfy 2. and 3., it runs x-oob-test out-of-band with "lock": false.
Note that waiting for a semaphore violates the rules for oob-capable
commands. Running x-oob-test with "lock": true hangs the monitor
until you run x-oob-test with "lock": false on another monitor (which
you might not have set up).
Having an externally visible QMP command that may hang the monitor is
not nice. Let's apply a little more ingenuity to the problem. Idea:
have an existing command block on reading a FIFO special file, unblock
it by opening the FIFO for writing.
where ID2 is a different arbitrary string. Since there's no migration
to pause, the command will fail, but that's fine; instant failure is
still a test of out-of-band responses overtaking in-band commands.
Events are broadcast to all monitors. If another monitor's client has
a command with the same ID in flight, the event will incorrectly claim
that command was dropped. This must be fixed before out-of-band
execution can graduate from "experimental".
OOB documentation is spread over qmp-spec.txt sections 2.2.1
Capabilities and 2.3 Issuing Commands. The amount of detail is a bit
distracting there. Move the meat of the matter to new section 2.3.1
Out of band execution.
Throw in a few other improvements while there:
* 2.2 Server Greeting: Drop advice to search entire capabilities
array; should be obvious.
* 3. QMP Examples
- 3.1 Server Greeting: Update greeting to the one we expect for the
release. Now shows capability "oob". Update qmp-intro.txt
likewise.
- 3.2 Capabilities negotiation: Show client accepting capability
"oob".
bios_support_mode verifies if the guest has support for a certain
suspend mode but it doesn't inform back which suspend tool
provides it. The caller, guest_suspend, executes all suspend
strategies in order again.
After adding systemd suspend support, bios_support_mode now will
verify for support for systemd, then pmutils, then Linux sys state
file. In a worst case scenario where both systemd and pmutils isn't
supported but Linux sys state is:
- bios_supports_mode will check for systemd, then pmutils, then
Linux sys state. It will tell guest_suspend that there is support,
but it will not tell who provides it;
- guest_suspend will try to execute (and fail) systemd suspend,
then pmutils suspend, to only then use the Linux sys suspend.
The time spent executing systemd and pmutils suspend was wasted
and could be avoided, but only bios_support_mode knew it but
didn't inform it back.
A quicker approach is to nuke bios_supports_mode and control
whether we found support at all with a bool flag inside
guest_suspend. guest_suspend will search for suspend support
and execute it as soon as possible. If the a given suspend
mechanism fails, continue to the next. If no suspend
support is found, the "not supported" message is still being
sent back to the user.
qga: systemd hibernate/suspend/hybrid-sleep support
pmutils isn't being supported by newer OSes like Fedora 27
or Mint. This means that the only suspend option QGA offers
for these guests are writing directly into the Linux sys state
file. This also means that QGA also loses the ability to do
hybrid suspend in those guests - this suspend mode is only
available when using pmutils.
Newer guests can use systemd facilities to do all the suspend
types QGA supports. The mapping in comparison with pmutils is:
This is a cleanup of the resulting code after detaching
pmutils and Linux sys state file logic:
- remove the SUSPEND_MODE_* macros and use an enumeration
instead. At the same time, drop the switch statements
at the start of each function and use the enumeration
index to get the right binary/argument;
- create a new function called run_process_child(). This
function uses g_spawn_sync() to execute a shell command,
returning the exit code. This is a common operation in the
pmutils functions and will be used in the systemd implementation
as well, so this function will avoid code repetition.
There are more places inside commands-posix.c where this new
run_process_child function can also be used, but one step
at a time.
Signed-off-by: Daniel Henrique Barboza <[email protected]>
*check/propagate local_err before setting errp directly Signed-off-by: Michael Roth <[email protected]>
qga: guest_suspend: decoupling pm-utils and sys logic
Following the same logic of the previous patch, let's also
decouple the suspend logic from guest_suspend into specialized
functions, one for each strategy we support at this moment.
qga: bios_supports_mode: decoupling pm-utils and sys logic
In bios_supports_mode there is a verification to assert if
the chosen suspend mode is supported by the pmutils tools and,
if not, we see if the Linux sys state files supports it.
This verification is done in the same function, one after
the other, and it works for now. But, when adding a new
suspend mechanism that will not necessarily follow the same
return 0 or 1 logic of pmutils, this code will be hard
to deal with.
This patch decouple the two existing logics into their own
functions, pmutils_supports_mode and linux_sys_state_supports_mode,
which in turn are used inside bios_support_mode. The existing
logic is kept but now it's easier to extend it.
To be able to add new suspend mechanisms we need to detach
the existing QMP functions from the current implementation
specifics.
At this moment we have functions such as qmp_guest_suspend_ram
calling bios_suspend_mode and guest_suspend passing the
pmutils command and arguments as parameters. This patch
removes this logic from the QMP functions, moving them to
the respective functions that will have to deal with which
binary to use.
Iterate over the PCI bridges to lookup the PCI device associated with
the block device.
This allows to lookup the driver under the following syspath:
/sys/devices/pci0000:00/0000:00:02.2/0000:03:00.0/virtio2/block/vda/vda3
It also works with an "old-style" Q35 libvirt hierarchy: root complex
-> DMI-PCI bridge -> PCI-PCI bridge -> virtio controller, ex:
/sys/devices/pci0000:00/0000:00:03.0/0000:01:01.0/0000:02:01.0/virtio1/block/vda/vda3
The setup can be reproduced with the following qemu command line
(Thanks Marcel for help):
Issue: When upgrading qemu-ga using the msi from an old version
to a newer one, the upgrade is not allowed by the msi
showing this error message "Another version of this product
is already installed."
Fix: For the upgrade to be allowed by the msi the WiX file must
provide three things:
1. Changing product's Id. (assigning it to "*")
2. Constant product's UpgradeId. (exists)
3. Changing version. (exists)
Before the patch, the commands for !CONFIG_VNC are stubs that fail
like this:
{"error": {"class": "GenericError",
"desc": "The feature 'vnc' is not enabled"}}
Afterwards, they fail like this:
{"error": {"class": "CommandNotFound",
"desc": "The command FOO has not been found"}}
I call that an improvement, because it lets clients distinguish
between command unavailable (class CommandNotFound) and command failed
(class GenericError).
Will return "unknown command: 'info vnc'" when VNC is compiled
out (same as error for spice when --disable-spice)
Occurrences of VNC (case insensitive) in the schema that aren't
covered by this change:
* add_client
Command has other uses, including "socket bases character devices".
These are unconditional as far as I can tell.
* set_password, expire_password
In theory, these commands could be used for managing any service's
password. In practice, they're used for VNC and SPICE services.
They're documented for "remote display session" / "remote display
server".
The service is selected by argument @protocol. The code special-cases
protocol-specific argument checking, then calls a protocol-specific
function to do the work. If it fails, the command fails with "Could
not set password". It does when the service isn't compiled in (it's a
stub then).
We could make these commands conditional on the conjunction of all
services [currently: defined(CONFIG_VNC) || defined(CONFIG_SPICE)],
but I doubt it's worthwhile.
* change
Command has other uses, namely changing media.
This patch inlines a stub; no functional change.
Wrap generated code with #if/#endif using an 'ifcontext' on
QAPIGenCSnippet objects.
This makes a conditional event's qapi_event_send_FOO() compile-time
conditional, but its enum QAPIEvent member remains unconditional for
now. A follow up patch "qapi-event: add 'if' condition to implicit
event enum" will improve this.
qapi-introspect: add preprocessor conditions to generated QLit
This commit adds 'ifcond' conditions to top-level QLit objects.
Future work will add them to object and enum type members, i.e. within
QLit objects.
Extend the QLit generator to_qlit() to accept (@obj, @cond) tuples in
addition to just @obj. The tuple causes the QLit generated for
objects for @obj with #if/#endif conditions for @cond.
See generated tests/test-qmp-introspect.c. Example diff after this
patch:
Add helpers to wrap generated code with #if/#endif lines.
A later patch wants to use QAPIGen for generating C snippets rather
than full C files with copyright headers etc. Splice in class
QAPIGenCCode between QAPIGen and QAPIGenC.
Add a 'with' statement context manager that will be used to wrap
generator visitor methods. The manager will check if code was
generated before adding #if/#endif lines on QAPIGenCSnippet
objects. Used in the following patches.
qapi: leave the ifcond attribute undefined until check()
We commonly initialize attributes to None in .init(), then set their
real value in .check(). Accessing the attribute before .check()
yields None. If we're lucky, the code that accesses the attribute
prematurely chokes on None.
It won't for .ifcond, because None is a legitimate value.
Leave the ifcond attribute undefined until check().
qapi: pass 'if' condition into QAPISchemaEntity objects
Built-in objects remain unconditional. Explicitly defined objects use
the condition specified in the schema. Implicitly defined objects
inherit their condition from their users. For most of them, there is
exactly one user, so the condition to use is obvious. The exception
is wrapped types generated for simple union variants, which can be
shared by any number of simple unions. The tight condition would be
the disjunction of the conditions of these simple unions. For now,
use the wrapped type's condition instead. Much simpler and good
enough for now.
Chen Hanxiao [Thu, 14 Jun 2018 08:06:06 +0000 (16:06 +0800)]
qga: add mountpoint usage info to GuestFilesystemInfo
This patch adds support for getting the usage of mounted
filesystem.
The usage of fs stored as used_bytes and total_bytes.
It's very useful when we try to monitor guest's filesystem.
Accept 'if' key in top-level elements, accepted as string or list of
string type. The following patches will modify the test visitor to
check the value is correctly saved, and generate #if/#endif code (as a
single #if/endif line or a series for a list).
Example of 'if' key:
{ 'struct': 'TestIfStruct', 'data': { 'foo': 'int' },
'if': 'defined(TEST_IF_STRUCT)' }
The generated code is for now *unconditional*. Later patches generate
the conditionals.
* remotes/kraxel/tags/usb-20180703-pull-request:
xhci: fix guest-triggerable assert
ehci: Don't fetch a NULL current qtd but advance the queue instead.
While reading file content via 'guest-file-read' command,
'qmp_guest_file_read' routine allocates buffer of count+1
bytes. It could overflow for large values of 'count'.
Add check to avoid it.
Chen Hanxiao [Thu, 14 Jun 2018 08:10:13 +0000 (16:10 +0800)]
qga: unset frozen state if no mount points are frozen
If we set mountpoints to qmp_guest_fsfreeze_freeze_list,
we may got nothing to freeze as all mountpoints are
not valid.
So call ga_unset_frozen in this senario.
Also, if we return 0 frozen fs, there is no need to call
guest-fsfreeze-thaw.
Peter Maydell [Tue, 3 Jul 2018 15:04:41 +0000 (16:04 +0100)]
Merge remote-tracking branch 'remotes/shorne/tags/pull-or-20180703' into staging
OpenRISC cleanups and Fixes for QEMU 3.0
Mostly patches from Richard Henderson fixing multiple things:
* Fix singlestepping in GDB.
* Use more TB linking.
* Fixes to exit TB after updating SPRs to enable registering of state
changes.
* Significant optimizations and refactors to the TLB
* Split out disassembly from translation.
* Add qemu-or1k to qemu-binfmt-conf.sh.
* Implement signal handling for linux-user.
Then there are a few fixups from me:
* Fix delay slot detections to match hardware, this was masking a bug
in the linus kernel.
* Fix stores to the PIC mask register
# gpg: Signature made Tue 03 Jul 2018 14:44:10 BST
# gpg: using RSA key C3B31C2D5E6627E4
# gpg: Good signature from "Stafford Horne <[email protected]>"
# gpg: WARNING: This key is not certified with a trusted signature!
# gpg: There is no indication that the signature belongs to the owner.
# Primary key fingerprint: D9C4 7354 AEF8 6C10 3A25 EFF1 C3B3 1C2D 5E66 27E4
* remotes/shorne/tags/pull-or-20180703: (25 commits)
target/openrisc: Fix writes to interrupt mask register
target/openrisc: Fix delay slot exception flag to match spec
linux-user: Fix struct sigaltstack for openrisc
linux-user: Implement signals for openrisc
target/openrisc: Add support in scripts/qemu-binfmt-conf.sh
target/openrisc: Reorg tlb lookup
target/openrisc: Increase the TLB size
target/openrisc: Stub out handle_mmu_fault for softmmu
target/openrisc: Use identical sizes for ITLB and DTLB
target/openrisc: Fix cpu_mmu_index
target/openrisc: Fix tlb flushing in mtspr
target/openrisc: Reduce tlb to a single dimension
target/openrisc: Merge mmu_helper.c into mmu.c
target/openrisc: Remove indirect function calls for mmu
target/openrisc: Merge tlb allocation into CPUOpenRISCState
target/openrisc: Form the spr index from tcg
target/openrisc: Exit the TB after l.mtspr
target/openrisc: Split out is_user
target/openrisc: Link more translation blocks
target/openrisc: Fix singlestep_enabled
...
Peter Maydell [Tue, 3 Jul 2018 13:59:27 +0000 (14:59 +0100)]
Merge remote-tracking branch 'remotes/dgibson/tags/ppc-for-3.0-20180703' into staging
ppc patch queue 2018-07-03
Here's a last minue pull request before today's soft freeze. Ideally
I would have sent this earlier, but I was waiting for a couple of
extra fixes I knew were close. And the freeze crept up on me, like
always.
Most of the changes here are bugfixes in any case. There are some
cleanups as well, which have been in my staging tree for a little
while. There are a couple of truly new features (some extensions to
the sam460ex platform), but these are low risk, since they only affect
a new and not really stabilized machine type anyway.
Higlights are:
* Mac platform improvements from Mark Cave-Ayland
* Sam460ex improvements from BALATON Zoltan et al.
* XICS interrupt handler cleanups from Cédric Le Goater
* TCG improvements for atomic loads and stores from Richard
Henderson
* Assorted other bugfixes
* remotes/dgibson/tags/ppc-for-3.0-20180703: (35 commits)
ppc: Include vga cirrus card into the compiling process
target/ppc: Relax reserved bitmask of indexed store instructions
target/ppc: set is_jmp on ppc_tr_breakpoint_check
spapr: compute default value of "hpt-max-page-size" later
target/ppc/kvm: don't pass cpu to kvm_get_smmu_info()
target/ppc/kvm: get rid of kvm_get_fallback_smmu_info()
ppc440_uc: Basic emulation of PPC440 DMA controller
sam460ex: Add RTC device
hw/timer: Add basic M41T80 emulation
ppc4xx_i2c: Rewrite to model hardware more closely
hw/ppc: Give sam46ex its own config option
fpu_helper.c: fix setting FPSCR[FI] bit
target/ppc: Implement the rest of gen_st_atomic
target/ppc: Implement the rest of gen_ld_atomic
target/ppc: Use atomic min/max helpers
target/ppc: Use MO_ALIGN for EXIWX and ECOWX
target/ppc: Split out gen_st_atomic
target/ppc: Split out gen_ld_atomic
target/ppc: Split out gen_load_locked
target/ppc: Tidy gen_conditional_store
...
Signed-off-by: Peter Maydell <[email protected]>
# Conflicts:
# hw/ppc/spapr.c
target/openrisc: Fix writes to interrupt mask register
The interrupt controller mask register (PICMR) allows writing any value
to any of the 32 interrupt mask bits. Writing a 0 masks the interrupt
writing a 1 unmasks (enables) the the interrupt.
For some reason the old code was or'ing the write values to the PICMR
meaning it was not possible to ever mask a interrupt once it was
enabled.
I have tested this by running linux 4.18 and my regular checks, I don't
see any issues.
target/openrisc: Fix delay slot exception flag to match spec
The delay slot exception flag is only set on the SR register during
exception. Previously it was being set on both the ESR and SR this
caused QEMU to differ from the spec. The was apparent as the linux
kernel had a bug where it could boot on QEMU but not on real hardware.
All of the existing code was boilerplate from elsewhere,
and would crash the guest upon the first signal.
Signed-off-by: Richard Henderson <[email protected]> Signed-off-by: Stafford Horne <[email protected]>
---
v2:
Add a comment to the new definition of target_pt_regs.
Install the signal mask into the ucontext.
v3:
Incorporate feedback from Laurent.
While openrisc has a split i/d tlb, qemu does not. Perform a
lookup on both i & d tlbs in parallel and put the composite
rights into qemu's tlb. This avoids ping-ponging the qemu tlb
between EXEC and READ.
The key can be secretly shared between clients and servers. Clients
must specify the directory containing the "keys.psk" file and a
username (defaults to "qemu"). Servers must specify only the
directory.
Example NBD client:
$ qemu-img info \
--object tls-creds-psk,id=tls0,dir=/tmp/keys,username=rjones,endpoint=client \
--image-opts \
file.driver=nbd,file.host=localhost,file.port=10809,file.tls-creds=tls0,file.export=/
* remotes/cody/tags/block-pull-request:
backup: Use copy offloading
block: Honour BDRV_REQ_NO_SERIALISING in copy range
block: Fix parameter checking in bdrv_co_copy_range_internal
Peter Maydell [Tue, 3 Jul 2018 09:47:02 +0000 (10:47 +0100)]
Merge remote-tracking branch 'remotes/ericb/tags/pull-nbd-2018-07-02' into staging
nbd patches for 2018-07-02
Bug fixes and iotest exposure of fleecing via NBD (serving a
read-only point-in-time view via blockdev-backup sync:none,
as well as serving dirty bitmaps over NBD), including a new
x-dirty-bitmap parameter when opening NBD clients as the
counterpart to x-nbd-server-add-bitmap. Also a random fix
for iscsi block_status spotted by Coverity that missed other
miscellaneous trees.
- Eric Blake: nbd/server: Fix dirty bitmap logic regression
- Eric Blake: iscsi: Avoid potential for get_status overflow
- John Snow/Vladimir Sementsov-Ogievskiy: 0/2 block: formalize and test fleecing
- Eric Blake: 0/2 test NBD bitmap export
* remotes/ericb/tags/pull-nbd-2018-07-02:
iotests: New test 223 for exporting dirty bitmap over NBD
nbd/client: Add x-dirty-bitmap to query bitmap from server
iotests: add 222 to test basic fleecing
blockdev: enable non-root nodes for backup source
iscsi: Avoid potential for get_status overflow
nbd/server: Fix dirty bitmap logic regression
vga: disable global_vmstate for 3.0+ machine types
Move global_vmstate from vga_common_init() parameter to VGACommonState
field. Set global_vmstate to true for isa vga devices, so nothing
changes here. virtio-vga and secondary-vga already set global_vmstate
to false so no change here either. All other pci vga devices get a new
global-vmstate property, defaulting to false. A compat property flips
it to true for older machine types.
With this in place you don't get a vmstate section naming conflict any
more when adding multiple pci vga devices to your vm.
Sebastian Bauer [Mon, 25 Jun 2018 22:27:18 +0000 (00:27 +0200)]
ehci: Don't fetch a NULL current qtd but advance the queue instead.
Fetching qtd with the NULL address most likely makes no sense so from now
on, we handle it this case similarly as if the terminate (T) bit is not
set, which is already an exception as according to section 3.6 of the EHCI
spec there is no T bit defined for the current_qtd field.
The spec is a bit vague on how an EHCI driver should initialize these
fields: "The general operational model is that the host controller can
detect whether the overlay area contains a description of an active
transfer" (p. 49). QEMU primarily uses the QTD_TOKEN_ACTIVE bit of the
queue header to infer the activity state but there are other ways
conceivable.
This change allows QEMU to boot further into AmigaOS. The public available
version of the EHCI driver recycles queue heads in some rare conditions but
only clears the current_qtd field but not the status field. This works with
many available EHCI PCI cards but e.g., not with the Freescale USB
controller's found on the P5040. On the emulated EHCI controller of QEMU
the consequence is that some garbage was read in, which resulted in a
reset of the controller. This change fixes the problem.
The implementation is similar to the 'qemu-img convert'. In the
beginning of the job, offloaded copy is attempted. If it fails, further
I/O will go through the existing bounce buffer code path.
Then, as Kevin pointed out, both this and qemu-img convert can benefit
from a local check if one request fails because of, for example, the
offset is beyond EOF, but another may well be accepted by the protocol
layer. This will be implemented separately.
block: Fix parameter checking in bdrv_co_copy_range_internal
src may be NULL if BDRV_REQ_ZERO_WRITE flag is set, in this case only
check dst and dst->bs. This bug was introduced when moving in the
request tracking code from bdrv_co_copy_range, in 37aec7d75eb.
This especially fixes the possible segfault when initializing src_bs
with a NULL src.
Sebastian Bauer [Mon, 2 Jul 2018 19:13:45 +0000 (21:13 +0200)]
ppc: Include vga cirrus card into the compiling process
Drivers for this card exists on PPC-based AmigaOS guests so it is useful to
allow users to emulate the graphics card for PPC machines.
As cirrus vga is currently preferred over std(vga) in absence of any user
choice, this change also sets the default display of spapr machines to
std as otherwise qemu refuses to start these machines. Not specifying an
explicit graphics mode is for instance done by 'make check'.
BALATON Zoltan [Thu, 28 Jun 2018 22:38:33 +0000 (00:38 +0200)]
target/ppc: Relax reserved bitmask of indexed store instructions
The PPC440 User Manual says that if bit 31 is set, the contents of
CR[CR0] are undefined for indexed store instructions but this form is
not invalid. Other PPC variants confirming to recent ISA where this
bit may be reserved should ignore reserved bits and not raise invalid
instruction exception. In particular, MorphOS has an stwx instruction
with bit 31 set and fails to boot currently because of this. With this
patch it gets further.
Eric Blake [Mon, 2 Jul 2018 19:14:58 +0000 (14:14 -0500)]
iotests: New test 223 for exporting dirty bitmap over NBD
Although this test is NOT a full test of image fleecing (as it
intentionally uses just a single block device directly exported
over NBD, rather than trying to set up a blockdev-backup job with
multiple BDS involved), it DOES prove that qemu as a server is
able to properly expose a dirty bitmap over NBD.
When coupled with image fleecing, it is then possible for a
third-party client to do an incremental backup by using
qemu-img map with the x-dirty-bitmap option to learn which parts
of the file are dirty (perhaps confusingly, they are the portions
mapped as "data":false - which is part of the reason this is
still in the x- experimental namespace), along with another
normal client (perhaps 'qemu-nbd -c' to expose the server over
/dev/nbd0 and then just use normal I/O on that block device) to
read the dirty sections.
projects / qemu.git / log