Move the internal documentation about qemu threading to the knowledge base. The conversion included rstizing of the text document, mainly just fixing of the headline and enclosing function names and code examples into code block sections. Signed-off-by: Peter Krempa <pkrempa(a)redhat.com&gt; --- docs/kbase/index.rst | 3 + docs/kbase/internals/meson.build | 1 + .../kbase/internals/qemu-threads.rst | 136 +++++++++--------- 3 files changed, 69 insertions(+), 71 deletions(-) rename src/qemu/THREADS.txt => docs/kbase/internals/qemu-threads.rst (70%) diff --git a/docs/kbase/index.rst b/docs/kbase/index.rst index 0848467d51..f1cd143fab 100644 --- a/docs/kbase/index.rst +++ b/docs/kbase/index.rst @@ -101,3 +101,6 @@ Internals `RPC protocol & APIs <internals/rpc.html>`__ RPC protocol information and API / dispatch guide + +`QEMU driver threading <internals/qemu-threads.html>`__ + Basics of locking and threaded access to qemu driver primitives. diff --git a/docs/kbase/internals/meson.build b/docs/kbase/internals/meson.build index 8b5daad1f9..3e84b398b2 100644 --- a/docs/kbase/internals/meson.build +++ b/docs/kbase/internals/meson.build @@ -5,6 +5,7 @@ docs_kbase_internals_files = [ 'locking', 'migration', 'overview', + 'qemu-threads', 'rpc', ] diff --git a/src/qemu/THREADS.txt b/docs/kbase/internals/qemu-threads.rst similarity index 70% rename from src/qemu/THREADS.txt rename to docs/kbase/internals/qemu-threads.rst index b5f54f203c..c68512d1b3 100644 --- a/src/qemu/THREADS.txt +++ b/docs/kbase/internals/qemu-threads.rst @@ -1,5 +1,7 @@ - QEMU Driver Threading: The Rules - ================================= +QEMU Driver Threading: The Rules +================================ + +.. contents:: This document describes how thread safety is ensured throughout the QEMU driver. The criteria for this model are: @@ -8,37 +10,36 @@ the QEMU driver. The criteria for this model are: - Code which sleeps must be able to time out after suitable period - Must be safe against dispatch of asynchronous events from monitor - Basic locking primitives ------------------------ There are a number of locks on various objects - * virQEMUDriver * + ``virQEMUDriver`` - The qemu_conf.h file has inline comments describing the locking + The ``qemu_conf.h`` file has inline comments describing the locking needs for each field. Any field marked immutable, self-locking can be accessed without the driver lock. For other fields there - are typically helper APIs in qemu_conf.c that provide serialized - access to the data. No code outside qemu_conf.c should ever + are typically helper APIs in ``qemu_conf.c`` that provide serialized + access to the data. No code outside ``qemu_conf.c`` should ever acquire this lock - * virDomainObj * + ``virDomainObj`` Will be locked and the reference counter will be increased after calling - any of the virDomainObjListFindBy{ID,Name,UUID} methods. The preferred way + any of the ``virDomainObjListFindBy{ID,Name,UUID}`` methods. The preferred way of decrementing the reference counter and unlocking the domain is using the - virDomainObjEndAPI() function. + ``virDomainObjEndAPI()`` function. - Lock must be held when changing/reading any variable in the virDomainObj * + Lock must be held when changing/reading any variable in the ``virDomainObj`` This lock must not be held for anything which sleeps/waits (i.e. monitor commands). - * qemuMonitorPrivatePtr: Job conditions + ``qemuMonitorPrivatePtr`` job conditions - Since virDomainObj *lock must not be held during sleeps, the job + Since ``virDomainObj`` lock must not be held during sleeps, the job conditions provide additional protection for code making updates. QEMU driver uses three kinds of job conditions: asynchronous, agent @@ -61,30 +62,30 @@ There are a number of locks on various objects Agent job condition is then used when thread wishes to talk to qemu agent monitor. It is possible to acquire just agent job - (qemuDomainObjBeginAgentJob), or only normal job (qemuDomainObjBeginJob) + (``qemuDomainObjBeginAgentJob``), or only normal job (``qemuDomainObjBeginJob``) but not both at the same time. Holding an agent job and a normal job would allow an unresponsive or malicious agent to block normal libvirt API and potentially result in a denial of service. Which type of job to grab depends whether caller wishes to communicate only with agent socket, or only with qemu monitor socket. - Immediately after acquiring the virDomainObj *lock, any method + Immediately after acquiring the ``virDomainObj`` lock, any method which intends to update state must acquire asynchronous, normal or - agent job . The virDomainObj *lock is released while blocking on + agent job . The ``virDomainObj`` lock is released while blocking on these condition variables. Once the job condition is acquired, a - method can safely release the virDomainObj *lock whenever it hits + method can safely release the ``virDomainObj`` lock whenever it hits a piece of code which may sleep/wait, and re-acquire it after the sleep/wait. Whenever an asynchronous job wants to talk to the monitor, it needs to acquire nested job (a special kind of normal job) to obtain exclusive access to the monitor. - Since the virDomainObj *lock was dropped while waiting for the + Since the ``virDomainObj`` lock was dropped while waiting for the job condition, it is possible that the domain is no longer active when the condition is finally obtained. The monitor lock is only safe to grab after verifying that the domain is still active. - * qemuMonitor *: Mutex + ``qemuMonitor`` mutex Lock to be used when invoking any monitor command to ensure safety wrt any asynchronous events that may be dispatched from the monitor. @@ -92,118 +93,111 @@ There are a number of locks on various objects The job condition *MUST* be held before acquiring the monitor lock - The virDomainObj *lock *MUST* be held before acquiring the monitor + The ``virDomainObj`` lock *MUST* be held before acquiring the monitor lock. - The virDomainObj *lock *MUST* then be released when invoking the + The ``virDomainObj`` lock *MUST* then be released when invoking the monitor command. Helper methods -------------- -To lock the virDomainObj * - - virObjectLock() - - Acquires the virDomainObj *lock +To lock the ``virDomainObj`` - virObjectUnlock() - - Releases the virDomainObj *lock + ``virObjectLock()`` + - Acquires the ``virDomainObj`` lock + ``virObjectUnlock()`` + - Releases the ``virDomainObj`` lock To acquire the normal job condition - qemuDomainObjBeginJob() + ``qemuDomainObjBeginJob()`` - Waits until the job is compatible with current async job or no async job is running - - Waits for job.cond condition 'job.active != 0' using virDomainObj * + - Waits for ``job.cond`` condition ``job.active != 0`` using ``virDomainObj`` mutex - Rechecks if the job is still compatible and repeats waiting if it isn't - - Sets job.active to the job type + - Sets ``job.active`` to the job type - - qemuDomainObjEndJob() + ``qemuDomainObjEndJob()`` - Sets job.active to 0 - Signals on job.cond condition - To acquire the agent job condition - qemuDomainObjBeginAgentJob() + ``qemuDomainObjBeginAgentJob()`` - Waits until there is no other agent job set - - Sets job.agentActive tp the job type - - qemuDomainObjEndAgentJob() - - Sets job.agentActive to 0 - - Signals on job.cond condition + - Sets ``job.agentActive`` to the job type + ``qemuDomainObjEndAgentJob()`` + - Sets ``job.agentActive`` to 0 + - Signals on ``job.cond`` condition To acquire the asynchronous job condition - qemuDomainObjBeginAsyncJob() + ``qemuDomainObjBeginAsyncJob()`` - Waits until no async job is running - - Waits for job.cond condition 'job.active != 0' using virDomainObj * + - Waits for ``job.cond`` condition ``job.active != 0`` using ``virDomainObj`` mutex - - Rechecks if any async job was started while waiting on job.cond + - Rechecks if any async job was started while waiting on ``job.cond`` and repeats waiting in that case - - Sets job.asyncJob to the asynchronous job type - - - qemuDomainObjEndAsyncJob() - - Sets job.asyncJob to 0 - - Broadcasts on job.asyncCond condition + - Sets ``job.asyncJob`` to the asynchronous job type + ``qemuDomainObjEndAsyncJob()`` + - Sets ``job.asyncJob`` to 0 + - Broadcasts on ``job.asyncCond`` condition To acquire the QEMU monitor lock - qemuDomainObjEnterMonitor() - - Acquires the qemuMonitorObjPtr lock - - Releases the virDomainObj *lock + ``qemuDomainObjEnterMonitor()`` + - Acquires the ``qemuMonitorObj`` lock + - Releases the ``virDomainObj`` lock - qemuDomainObjExitMonitor() - - Releases the qemuMonitorObjPtr lock - - Acquires the virDomainObj *lock + ``qemuDomainObjExitMonitor()`` + - Releases the ``qemuMonitorObj`` lock + - Acquires the ``virDomainObj`` lock These functions must not be used by an asynchronous job. To acquire the QEMU monitor lock as part of an asynchronous job - qemuDomainObjEnterMonitorAsync() + ``qemuDomainObjEnterMonitorAsync()`` - Validates that the right async job is still running - - Acquires the qemuMonitorObjPtr lock - - Releases the virDomainObj *lock + - Acquires the ``qemuMonitorObj`` lock + - Releases the ``virDomainObj`` lock - Validates that the VM is still active qemuDomainObjExitMonitor() - - Releases the qemuMonitorObjPtr lock - - Acquires the virDomainObj *lock + - Releases the ``qemuMonitorObj`` lock + - Acquires the ``virDomainObj`` lock These functions are for use inside an asynchronous job; the caller must check for a return of -1 (VM not running, so nothing to exit). - Helper functions may also call this with VIR_ASYNC_JOB_NONE when + Helper functions may also call this with ``VIR_ASYNC_JOB_NONE`` when used from a sync job (such as when first starting a domain). To keep a domain alive while waiting on a remote command - qemuDomainObjEnterRemote() - - Releases the virDomainObj *lock + ``qemuDomainObjEnterRemote()`` + - Releases the ``virDomainObj`` lock - qemuDomainObjExitRemote() - - Acquires the virDomainObj *lock + ``qemuDomainObjExitRemote()`` + - Acquires the ``virDomainObj`` lock Design patterns --------------- - - * Accessing something directly to do with a virDomainObj * + * Accessing something directly to do with a ``virDomainObj``:: virDomainObj *obj; @@ -214,7 +208,7 @@ Design patterns virDomainObjEndAPI(&obj); - * Updating something directly to do with a virDomainObj * + * Updating something directly to do with a ``virDomainObj``:: virDomainObj *obj; @@ -229,7 +223,7 @@ Design patterns virDomainObjEndAPI(&obj); - * Invoking a monitor command on a virDomainObj * + * Invoking a monitor command on a ``virDomainObj``:: virDomainObj *obj; qemuDomainObjPrivate *priv; @@ -252,7 +246,7 @@ Design patterns virDomainObjEndAPI(&obj); - * Invoking an agent command on a virDomainObj * + * Invoking an agent command on a ``virDomainObj``:: virDomainObj *obj; qemuAgent *agent; @@ -276,7 +270,7 @@ Design patterns virDomainObjEndAPI(&obj); - * Running asynchronous job + * Running asynchronous job:: virDomainObj *obj; qemuDomainObjPrivate *priv; @@ -316,7 +310,7 @@ Design patterns virDomainObjEndAPI(&obj); - * Coordinating with a remote server for migration + * Coordinating with a remote server for migration:: virDomainObj *obj; qemuDomainObjPrivate *priv; -- 2.35.3

Signed-off-by: Peter Krempa <pkrempa(a)redhat.com&gt; --- docs/kbase/index.rst | 3 ++ docs/kbase/internals/meson.build | 1 + .../kbase/internals/qemu-event-handlers.rst | 44 ++++++++++--------- 3 files changed, 28 insertions(+), 20 deletions(-) rename src/qemu/EVENTHANDLERS.txt => docs/kbase/internals/qemu-event-handlers.rst (64%) diff --git a/docs/kbase/index.rst b/docs/kbase/index.rst index d0f2167be8..8b710db85a 100644 --- a/docs/kbase/index.rst +++ b/docs/kbase/index.rst @@ -107,3 +107,6 @@ Internals `QEMU migration internals <internals/qemu-migration.html>`__ Description of migration phases in the ``v2`` and ``v3`` migration protocol. + +`QEMU monitor event handling <internals/qemu-event-handlers.html>`__ + Brief outline how events emitted by qemu on the monitor are handlded. diff --git a/docs/kbase/internals/meson.build b/docs/kbase/internals/meson.build index 4f7b223786..a16d5a290b 100644 --- a/docs/kbase/internals/meson.build +++ b/docs/kbase/internals/meson.build @@ -5,6 +5,7 @@ docs_kbase_internals_files = [ 'locking', 'migration', 'overview', + 'qemu-event-handlers', 'qemu-migration', 'qemu-threads', 'rpc', diff --git a/src/qemu/EVENTHANDLERS.txt b/docs/kbase/internals/qemu-event-handlers.rst similarity index 64% rename from src/qemu/EVENTHANDLERS.txt rename to docs/kbase/internals/qemu-event-handlers.rst index 39094d793e..3589c4c48c 100644 --- a/src/qemu/EVENTHANDLERS.txt +++ b/docs/kbase/internals/qemu-event-handlers.rst @@ -1,10 +1,14 @@ +=================== +QEMU event handlers +=================== + This is a short description of how an example qemu event can be used to trigger handler code that is called from the context of a worker thread, rather than directly from the event thread (which should itself never block, and can't do things like send qemu monitor commands, etc). -In this case (the NIC_RX_FILTER_CHANGED event) the event is handled by +In this case (the ``NIC_RX_FILTER_CHANGED`` event) the event is handled by calling a qemu monitor command to get the current RX filter state, then executing ioctls/sending netlink messages on the host in response to changes in that filter state. This event is *not* propagated to the @@ -14,39 +18,39 @@ to the end of this document, please do!). Hopefully this narration will be helpful when adding handlers for other qemu events in the future. ----------------------------------------------------- +QEMU monitor events +------------------- Any event emitted by qemu is received by -qemu_monitor_json.c:qemuMonitorJSONIOProcessEvent(). It looks up the -event by name in the table eventHandlers (in the same file), which +``qemu_monitor_json.c:qemuMonitorJSONIOProcessEvent()``. It looks up the +event by name in the table ``eventHandlers`` (in the same file), which should have an entry like this for each event that libvirt -understands: +understands:: { "NIC_RX_FILTER_CHANGED", qemuMonitorJSONHandleNicRxFilterChanged, }, - NB: This table is searched with bsearch, so it *must* be - alphabetically sorted. +NB: This table is searched with bsearch, so it *must* be alphabetically sorted. -qemuMonitorJSONIOProcessEvent calls the function listed in -eventHandlers, e.g.: +``qemuMonitorJSONIOProcessEvent`` calls the function listed in +``eventHandlers``, e.g.:: qemu_monitor_json.c:qemuMonitorJSONHandleNicRxFilterChanged() which extracts any required data from the JSON ("name" in this case), -and calls: +and calls:: qemu_monitor.c:qemuMonitorEmitNicRxFilterChanged() -which uses QEMU_MONITOR_CALLBACK() to call -mon->cb->domainNicRxFilterChanged(). domainNicRxFilterChanged is one -in a list of function pointers in qemu_process.c:monitorCallbacks. For -our example, it has been set to: +which uses ``QEMU_MONITOR_CALLBACK()`` to call +``mon->cb->domainNicRxFilterChanged()``. ``domainNicRxFilterChanged`` is one +in a list of function pointers in ``qemu_process.c:monitorCallbacks``. For +our example, it has been set to:: qemuProcessHandleNicRxFilterChanged() -This function allocates a qemuProcessEvent object, and queues an event -named QEMU_PROCESS_EVENT_NIC_RX_FILTER_CHANGED (you'll want to add an -enum to qemu_domain.h:qemuProcessEventType for your event) for a +This function allocates a ``qemuProcessEvent`` object, and queues an event +named ``QEMU_PROCESS_EVENT_NIC_RX_FILTER_CHANGED`` (you'll want to add an +enum to ``qemu_domain.h:qemuProcessEventType`` for your event) for a worker thread to handle. (Everything up to this point has happened in the context of the thread @@ -56,17 +60,17 @@ monitor. Everything after this is handled in the context of a worker thread, so it has more freedom to make qemu monitor calls and blocking system calls on the host.) -When the worker thread gets the event, it calls +When the worker thread gets the event, it calls:: qemuProcessEventHandler() which switches on the eventType (in our example, -QEMU_PROCESS_EVENT_NIC_RX_FILTER_CHANGED) and decides to call: +``QEMU_PROCESS_EVENT_NIC_RX_FILTER_CHANGED``) and decides to call:: processNicRxFilterChangedEvent() and *that* is where the actual work will be done (and any -event-specific memory allocated during qemuProcessHandleXXX() will be +event-specific memory allocated during ``qemuProcessHandleXXX()`` will be freed). Note that this function must do proper refcounting of the domain object, and assure that the domain is still active prior to performing any operations - it is possible that the domain could have -- 2.35.3