source: doc/theses/thierry_delisle_PhD/thesis/text/io.tex @ d3b95f1

\chapter{User Level \io}
As mentioned in Section~\ref{prev:io}, User-Level \io requires multiplexing the \io operations of many \glspl{thrd} onto fewer \glspl{proc} using asynchronous \io operations.
Different operating systems offer various forms of asynchronous operations and as mentioned in Chapter~\ref{intro}, this work is exclusively focused on the Linux operating-system.

\section{Kernel Interface}
Since this work fundamentally depends on operating-system support, the first step of any design is to discuss the available interfaces and pick one (or more) as the foundations of the non-blocking \io subsystem.

\subsection{\lstinline{O_NONBLOCK}}
In Linux, files can be opened with the flag @O_NONBLOCK@~\cite{MAN:open} (or @SO_NONBLOCK@~\cite{MAN:accept}, the equivalent for sockets) to use the file descriptors in ``nonblocking mode''.
In this mode, ``Neither the @open()@ nor any subsequent \io operations on the [opened file descriptor] will cause the calling process to wait''~\cite{MAN:open}.
This feature can be used as the foundation for the non-blocking \io subsystem.
However, for the subsystem to know when an \io operation completes, @O_NONBLOCK@ must be use in conjunction with a system call that monitors when a file descriptor becomes ready, \ie, the next \io operation on it does not cause the process to wait
\footnote{In this context, ready means \emph{some} operation can be performed without blocking.
It does not mean an operation returning \lstinline{EAGAIN} succeeds on the next try.
For example, a ready read may only return a subset of bytes and the read must be issues again for the remaining bytes, at which point it may return \lstinline{EAGAIN}.}.
This mechanism is also crucial in determining when all \glspl{thrd} are blocked and the application \glspl{kthrd} can now block.

There are three options to monitor file descriptors in Linux
\footnote{For simplicity, this section omits \lstinline{pselect} and \lstinline{ppoll}.
The difference between these system calls and \lstinline{select} and \lstinline{poll}, respectively, is not relevant for this discussion.},
@select@~\cite{MAN:select}, @poll@~\cite{MAN:poll} and @epoll@~\cite{MAN:epoll}.
All three of these options offer a system call that blocks a \gls{kthrd} until at least one of many file descriptors becomes ready.
The group of file descriptors being waited is called the \newterm{interest set}.

\paragraph{\lstinline{select}} is the oldest of these options, it takes as an input a contiguous array of bits, where each bits represent a file descriptor of interest.
On return, it modifies the set in place to identify which of the file descriptors changed status.
This destructive change means that calling select in a loop requires re-initializing the array each time and the number of file descriptors supported has a hard limit.
Another limit of @select@ is that once the call is started, the interest set can no longer be modified.
Monitoring a new file descriptor generally requires aborting any in progress call to @select@
\footnote{Starting a new call to \lstinline{select} is possible but requires a distinct kernel thread, and as a result is not an acceptable multiplexing solution when the interest set is large and highly dynamic unless the number of parallel calls to \lstinline{select} can be strictly bounded.}.

\paragraph{\lstinline{poll}} is an improvement over select, which removes the hard limit on the number of file descriptors and the need to re-initialize the input on every call.
It works using an array of structures as an input rather than an array of bits, thus allowing a more compact input for small interest sets.
Like @select@, @poll@ suffers from the limitation that the interest set cannot be changed while the call is blocked.

\paragraph{\lstinline{epoll}} further improves these two functions by allowing the interest set to be dynamically added to and removed from while a \gls{kthrd} is blocked on an @epoll@ call.
This dynamic capability is accomplished by creating an \emph{epoll instance} with a persistent interest set, which is used across multiple calls.
This capability significantly reduces synchronization overhead on the part of the caller (in this case the \io subsystem), since the interest set can be modified when adding or removing file descriptors without having to synchronize with other \glspl{kthrd} potentially calling @epoll@.

However, all three of these system calls have limitations.
The @man@ page for @O_NONBLOCK@ mentions that ``[@O_NONBLOCK@] has no effect for regular files and block devices'', which means none of these three system calls are viable multiplexing strategies for these types of \io operations.
Furthermore, @epoll@ has been shown to have problems with pipes and ttys~\cit{Peter's examples in some fashion}.
Finally, none of these are useful solutions for multiplexing \io operations that do not have a corresponding file descriptor and can be awkward for operations using multiple file descriptors.

\subsection{POSIX asynchronous I/O (AIO)}
An alternative to @O_NONBLOCK@ is the AIO interface.
Its interface lets programmers enqueue operations to be performed asynchronously by the kernel.
Completions of these operations can be communicated in various ways: either by spawning a new \gls{kthrd}, sending a Linux signal, or by polling for completion of one or more operation.
For this work, spawning a new \gls{kthrd} is counter-productive but a related solution is discussed in Section~\ref{io:morethreads}.
Using interrupts handlers can also lead to fairly complicated interactions between subsystems and has non-trivial cost.
Leaving polling for completion, which is similar to the previous system calls.
AIO only supports read and write operations to file descriptors, it does not have the same limitation as @O_NONBLOCK@, \ie, the file descriptors can be regular files and blocked devices.
It also supports batching multiple operations in a single system call.

AIO offers two different approach to polling: @aio_error@ can be used as a spinning form of polling, returning @EINPROGRESS@ until the operation is completed, and @aio_suspend@ can be used similarly to @select@, @poll@ or @epoll@, to wait until one or more requests have completed.
For the purpose of \io multiplexing, @aio_suspend@ is the best interface.
However, even if AIO requests can be submitted concurrently, @aio_suspend@ suffers from the same limitation as @select@ and @poll@, \ie, the interest set cannot be dynamically changed while a call to @aio_suspend@ is in progress.
AIO also suffers from the limitation of specifying which requests have completed, \ie programmers have to poll each request in the interest set using @aio_error@ to identify the completed requests.
This limitation means that, like @select@ and @poll@ but not @epoll@, the time needed to examine polling results increases based on the total number of requests monitored, not the number of completed requests.
Finally, AIO does not seem to be a popular interface, which I believe is due in part to this poor polling interface.
Linus Torvalds talks about this interface as follows:

\begin{displayquote}
        AIO is a horrible ad-hoc design, with the main excuse being ``other,
        less gifted people, made that design, and we are implementing it for
        compatibility because database people - who seldom have any shred of
        taste - actually use it''.

        But AIO was always really really ugly.

        \begin{flushright}
                -- Linus Torvalds\cit{https://lwn.net/Articles/671657/}
        \end{flushright}
\end{displayquote}

Interestingly, in this e-mail, Linus goes on to describe
``a true \textit{asynchronous system call} interface''
that does
``[an] arbitrary system call X with arguments A, B, C, D asynchronously using a kernel thread''
in
``some kind of arbitrary \textit{queue up asynchronous system call} model''.
This description is actually quite close to the interface described in the next section.

\subsection{\lstinline{io_uring}}
A very recent addition to Linux, @io_uring@~\cite{MAN:io_uring}, is a framework that aims to solve many of the problems listed in the above interfaces.
Like AIO, it represents \io operations as entries added to a queue.
But like @epoll@, new requests can be submitted while a blocking call waiting for requests to complete is already in progress.
The @io_uring@ interface uses two ring buffers (referred to simply as rings) at its core: a submit ring to which programmers push \io requests and a completion ring from which programmers poll for completion.

One of the big advantages over the prior interfaces is that @io_uring@ also supports a much wider range of operations.
In addition to supporting reads and writes to any file descriptor like AIO, it supports other operations like @open@, @close@, @fsync@, @accept@, @connect@, @send@, @recv@, @splice@, \etc.

On top of these, @io_uring@ adds many extras like avoiding copies between the kernel and user-space using shared memory, allowing different mechanisms to communicate with device drivers, and supporting chains of requests, \ie, requests that automatically trigger followup requests on completion.

\subsection{Extra Kernel Threads}\label{io:morethreads}
Finally, if the operating system does not offer a satisfactory form of asynchronous \io operations, an ad-hoc solution is to create a pool of \glspl{kthrd} and delegate operations to it to avoid blocking \glspl{proc}, which is a compromise for multiplexing.
In the worst case, where all \glspl{thrd} are consistently blocking on \io, it devolves into 1-to-1 threading.
However, regardless of the frequency of \io operations, it achieves the fundamental goal of not blocking \glspl{proc} when \glspl{thrd} are ready to run.
This approach is used by languages like Go\cit{Go} and frameworks like libuv\cit{libuv}, since it has the advantage that it can easily be used across multiple operating systems.
This advantage is especially relevant for languages like Go, which offer a homogeneous \glsxtrshort{api} across all platforms.
As opposed to C, which has a very limited standard api for \io, \eg, the C standard library has no networking.

\subsection{Discussion}
These options effectively fall into two broad camps: waiting for \io to be ready versus waiting for \io to complete.
All operating systems that support asynchronous \io must offer an interface along one of these lines, but the details vary drastically.
For example, Free BSD offers @kqueue@~\cite{MAN:bsd/kqueue}, which behaves similarly to @epoll@, but with some small quality of use improvements, while Windows (Win32)~\cit{https://docs.microsoft.com/en-us/windows/win32/fileio/synchronous-and-asynchronous-i-o} offers ``overlapped I/O'', which handles submissions similarly to @O_NONBLOCK@ with extra flags on the synchronous system call, but waits for completion events, similarly to @io_uring@.

For this project, I selected @io_uring@, in large parts because of its generality.
While @epoll@ has been shown to be a good solution for socket \io (\cite{DBLP:journals/pomacs/KarstenB20}), @io_uring@'s transparent support for files, pipes, and more complex operations, like @splice@ and @tee@, make it a better choice as the foundation for a general \io subsystem.

\section{Event-Engine}
An event engine's responsibility is to use the kernel interface to multiplex many \io operations onto few \glspl{kthrd}.
In concrete terms, this means \glspl{thrd} enter the engine through an interface, the event engines then starts the operation and parks the calling \glspl{thrd}, returning control to the \gls{proc}.
The parked \glspl{thrd} are then rescheduled by the event engine once the desired operation has completed.

\subsection{\lstinline{io_uring} in depth}
Before going into details on the design of my event engine, more details on @io_uring@ usage are presented, each important in the design of the engine.
Figure~\ref{fig:iouring} shows an overview of an @io_uring@ instance.
Two ring buffers are used to communicate with the kernel: one for submissions~(left) and one for completions~(right).
The submission ring contains entries, \newterm{Submit Queue Entries} (SQE), produced (appended) by the application when an operation starts and then consumed by the kernel.
The completion ring contains entries, \newterm{Completion Queue Entries} (CQE), produced (appended) by the kernel when an operation completes and then consumed by the application.
The submission ring contains indexes into the SQE array (denoted \emph{S} in the figure) containing entries describing the I/O operation to start;
the completion ring contains entries for the completed I/O operation.
Multiple @io_uring@ instances can be created, in which case they each have a copy of the data structures in the figure.

\begin{figure}
        \centering
        \input{io_uring.pstex_t}
        \caption[Overview of \lstinline{io_uring}]{Overview of \lstinline{io_uring} \smallskip\newline Two ring buffer are used to communicate with the kernel, one for completions~(right) and one for submissions~(left). The submission ring indexes into a pre-allocated array (denoted \emph{S}) instead.}
        \label{fig:iouring}
\end{figure}

New \io operations are submitted to the kernel following 4 steps, which use the components shown in the figure.
\begin{enumerate}
\item
An SQE is allocated from the pre-allocated array (denoted \emph{S} in Figure~\ref{fig:iouring}).
This array is created at the same time as the @io_uring@ instance, is in kernel-locked memory visible by both the kernel and the application, and has a fixed size determined at creation.
How these entries are allocated is not important for the functioning of @io_uring@, the only requirement is that no entry is reused before the kernel has consumed it.
\item
The SQE is filled according to the desired operation.
This step is straight forward, the only detail worth mentioning is that SQEs have a @user_data@ field that must be filled in order to match submission and completion entries.
\item
The SQE is submitted to the submission ring by appending the index of the SQE to the ring following regular ring buffer steps: \lstinline{buffer[head] = item; head++}.
Since the head is visible to the kernel, some memory barriers may be required to prevent the compiler from reordering these operations.
Since the submission ring is a regular ring buffer, more than one SQE can be added at once and the head is updated only after all entries are updated.
\item
The kernel is notified of the change to the ring using the system call @io_uring_enter@.
The number of elements appended to the submission ring is passed as a parameter and the number of elements consumed is returned.
The @io_uring@ instance can be constructed so this step is not required, but this requires elevated privilege.% and an early version of @io_uring@ had additional restrictions.
\end{enumerate}

\begin{sloppypar}
The completion side is simpler: applications call @io_uring_enter@ with the flag @IORING_ENTER_GETEVENTS@ to wait on a desired number of operations to complete.
The same call can be used to both submit SQEs and wait for operations to complete.
When operations do complete, the kernel appends a CQE to the completion ring and advances the head of the ring.
Each CQE contains the result of the operation as well as a copy of the @user_data@ field of the SQE that triggered the operation.
It is not necessary to call @io_uring_enter@ to get new events because the kernel can directly modify the completion ring.
The system call is only needed if the application wants to block waiting for operations to complete.
\end{sloppypar}

The @io_uring_enter@ system call is protected by a lock inside the kernel.
This protection means that concurrent call to @io_uring_enter@ using the same instance are possible, but there is no performance gained from parallel calls to @io_uring_enter@.
It is possible to do the first three submission steps in parallel, however, doing so requires careful synchronization.

@io_uring@ also introduces constraints on the number of simultaneous operations that can be ``in flight''.
Obviously, SQEs are allocated from a fixed-size array, meaning that there is a hard limit to how many SQEs can be submitted at once.
In addition, the @io_uring_enter@ system call can fail because ``The  kernel [...] ran out of resources to handle [a request]'' or ``The application is attempting to overcommit the number of requests it can  have  pending.''.
This restriction means \io request bursts may have to be subdivided and submitted in chunks at a later time.

\subsection{Multiplexing \io: Submission}
The submission side is the most complicated aspect of @io_uring@ and the completion side effectively follows from the design decisions made in the submission side.
While it is possible to do the first steps of submission in parallel, the duration of the system call scales with number of entries submitted.
The consequence is that the amount of parallelism used to prepare submissions for the next system call is limited.
Beyond this limit, the length of the system call is the throughput limiting factor.
I concluded from early experiments that preparing submissions seems to take about as long as the system call itself, which means that with a single @io_uring@ instance, there is no benefit in terms of \io throughput to having more than two \glspl{hthrd}.
Therefore the design of the submission engine must manage multiple instances of @io_uring@ running in parallel, effectively sharding @io_uring@ instances.
Similarly to scheduling, this sharding can be done privately, \ie, one instance per \glspl{proc}, in decoupled pools, \ie, a pool of \glspl{proc} use a pool of @io_uring@ instances without one-to-one coupling between any given instance and any given \gls{proc}, or some mix of the two.
Since completions are sent to the instance where requests were submitted, all instances with pending operations must be polled continously
\footnote{As will be described in Chapter~\ref{practice}, this does not translate into constant cpu usage.}.

A complicating aspect of submission is @io_uring@'s support for chains of operations, where the completion of an operation triggers the submission of the next operation on the link.
SQEs forming a chain must be allocated from the same instance and must be contiguous in the Submission Ring (see Figure~\ref{fig:iouring}).
The consequence of this feature is that filling SQEs can be arbitrarly complex and therefore users may need to run arbitrary code between allocation and submission.
Supporting chains is a requirement of the \io subsystem, but it is still valuable.
Support for this feature can be fulfilled simply to supporting arbitrary user code between allocation and submission.

\subsubsection{Public Instances}
One approach is to have multiple shared instances.
\Glspl{thrd} attempting \io operations pick one of the available instances and submit operations to that instance.
Since there is no coupling between \glspl{proc} and @io_uring@ instances in this approach, \glspl{thrd} running on more than one \gls{proc} can attempt to submit to the same instance concurrently.
Since @io_uring@ effectively sets the amount of sharding needed to avoid contention on its internal locks, performance in this approach is based on two aspects: the synchronization needed to submit does not induce more contention than @io_uring@ already does and the scheme to route \io requests to specific @io_uring@ instances does not introduce contention.
This second aspect has an oversized importance because it comes into play before the sharding of instances, and as such, all \glspl{hthrd} can contend on the routing algorithm.

Allocation in this scheme can be handled fairly easily.
Free SQEs, \ie, SQEs that aren't currently being used to represent a request, can be written to safely and have a field called @user_data@ which the kernel only reads to copy to @cqe@s.
Allocation also requires no ordering guarantee as all free SQEs are interchangeable.
This requires a simple concurrent bag.
The only added complexity is that the number of SQEs is fixed, which means allocation can fail.

Allocation failures need to be pushed up to the routing algorithm: \glspl{thrd} attempting \io operations must not be directed to @io_uring@ instances without sufficient SQEs available.
Furthermore, the routing algorithm should block operations up-front if none of the instances have available SQEs.

Once an SQE is allocated, \glspl{thrd} can fill them normally, they simply need to keep track of the SQE index and which instance it belongs to.

Once an SQE is filled in, what needs to happen is that the SQE must be added to the submission ring buffer, an operation that is not thread-safe on itself, and the kernel must be notified using the @io_uring_enter@ system call.
The submission ring buffer is the same size as the pre-allocated SQE buffer, therefore pushing to the ring buffer cannot fail
\footnote{This is because it is invalid to have the same \lstinline{sqe} multiple times in the ring buffer.}.
However, as mentioned, the system call itself can fail with the expectation that it will be retried once some of the already submitted operations complete.
Since multiple SQEs can be submitted to the kernel at once, it is important to strike a balance between batching and latency.
Operations that are ready to be submitted should be batched together in few system calls, but at the same time, operations should not be left pending for long period of times before being submitted.
This can be handled by either designating one of the submitting \glspl{thrd} as the being responsible for the system call for the current batch of SQEs or by having some other party regularly submitting all ready SQEs, \eg, the poller \gls{thrd} mentioned later in this section.

In the case of designating a \gls{thrd}, ideally, when multiple \glspl{thrd} attempt to submit operations to the same @io_uring@ instance, all requests would be batched together and one of the \glspl{thrd} would do the system call on behalf of the others, referred to as the \newterm{submitter}.
In practice however, it is important that the \io requests are not left pending indefinitely and as such, it may be required to have a current submitter and a next submitter.
Indeed, as long as there is a ``next'' submitter, \glspl{thrd} submitting new \io requests can move on, knowing that some future system call will include their request.
Once the system call is done, the submitter must also free SQEs so that the allocator can reused them.

Finally, the completion side is much simpler since the @io_uring@ system call enforces a natural synchronization point.
Polling simply needs to regularly do the system call, go through the produced CQEs and communicate the result back to the originating \glspl{thrd}.
Since CQEs only own a signed 32 bit result, in addition to the copy of the @user_data@ field, all that is needed to communicate the result is a simple future~\cite{wiki:future}.
If the submission side does not designate submitters, polling can also submit all SQEs as it is polling events.
A simple approach to polling is to allocate a \gls{thrd} per @io_uring@ instance and simply let the poller \glspl{thrd} poll their respective instances when scheduled.
This design is especially convenient for reasons explained in Chapter~\ref{practice}.

With this pool of instances approach, the big advantage is that it is fairly flexible.
It does not impose restrictions on what \glspl{thrd} submitting \io operations can and cannot do between allocations and submissions.
It also can gracefully handles running out of ressources, SQEs or the kernel returning @EBUSY@.
The down side to this is that many of the steps used for submitting need complex synchronization to work properly.
The routing and allocation algorithm needs to keep track of which ring instances have available SQEs, block incoming requests if no instance is available, prevent barging if \glspl{thrd} are already queued up waiting for SQEs and handle SQEs being freed.
The submission side needs to safely append SQEs to the ring buffer, correctly handle chains, make sure no SQE is dropped or left pending forever, notify the allocation side when SQEs can be reused and handle the kernel returning @EBUSY@.
All this synchronization may have a significant cost and, compare to the next approach presented, this synchronization is entirely overhead.

\subsubsection{Private Instances}
Another approach is to simply create one ring instance per \gls{proc}.
This alleviate the need for synchronization on the submissions, requiring only that \glspl{thrd} are not interrupted in between two submission steps.
This is effectively the same requirement as using @thread_local@ variables.
Since SQEs that are allocated must be submitted to the same ring, on the same \gls{proc}, this effectively forces the application to submit SQEs in allocation order
\footnote{The actual requirement is that \glspl{thrd} cannot context switch between allocation and submission.
This requirement means that from the subsystem's point of view, the allocation and submission are sequential.
To remove this requirement, a \gls{thrd} would need the ability to ``yield to a specific \gls{proc}'', \ie, park with the promise that it will be run next on a specific \gls{proc}, the \gls{proc} attached to the correct ring.}
, greatly simplifying both allocation and submission.
In this design, allocation and submission form a ring partitionned ring buffer as shown in Figure~\ref{fig:pring}.
Once added to the ring buffer, the attached \gls{proc} has a significant amount of flexibility with regards to when to do the system call.
Possible options are: when the \gls{proc} runs out of \glspl{thrd} to run, after running a given number of threads \glspl{thrd}, etc.

\begin{figure}
        \centering
        \input{pivot_ring.pstex_t}
        \caption[Partitioned ring buffer]{Partitioned ring buffer \smallskip\newline Allocated sqes are appending to the first partition.
        When submitting, the partition is advanced.
        The kernel considers the partition as the head of the ring.}
        \label{fig:pring}
\end{figure}

This approach has the advantage that it does not require much of the synchronization needed in the shared approach.
This comes at the cost that \glspl{thrd} submitting \io operations have less flexibility, they cannot park or yield, and several exceptional cases are handled poorly.
Instances running out of SQEs cannot run \glspl{thrd} wanting to do \io operations, in such a case the \gls{thrd} needs to be moved to a different \gls{proc}, the only current way of achieving this would be to @yield()@ hoping to be scheduled on a different \gls{proc}, which is not guaranteed.

A more involved version of this approach can seem to solve most of these problems, using a pattern called \newterm{helping}.
\Glspl{thrd} that wish to submit \io operations but cannot do so
\footnote{either because of an allocation failure or because they were migrate to a different \gls{proc} between allocation and submission}
create an object representing what they wish to achieve and add it to a list somewhere.
For this particular problem, one solution would be to have a list of pending submissions per \gls{proc} and a list of pending allocations, probably per cluster.
The problem with these ``solutions'' is that they are still bound by the strong coupling between \glspl{proc} and @io_uring@ instances.
These data structures would allow moving \glspl{thrd} to a specific \gls{proc} when the current \gls{proc} cannot fulfill the \io request.

Imagine a simple case with two \glspl{thrd} on two \glspl{proc}, one \gls{thrd} submits an \io operation and then sets a flag, the other \gls{thrd} spins until the flag is set.
If the first \gls{thrd} is preempted between allocation and submission and moves to the other \gls{proc}, the original \gls{proc} could start running the spinning \gls{thrd}.
If this happens, the helping ``solution'' is for the \io \gls{thrd}to added append an item to the submission list of the \gls{proc} where the allocation was made.
No other \gls{proc} can help the \gls{thrd} since @io_uring@ instances are strongly coupled to \glspl{proc}.
However, in this case, the \gls{proc} is unable to help because it is executing the spinning \gls{thrd} mentioned when first expression this case
\footnote{This particular example is completely artificial, but in the presence of many more \glspl{thrd}, it is not impossible that this problem would arise ``in the wild''.
Furthermore, this pattern is difficult to reliably detect and avoid.}
resulting in a deadlock.
Once in this situation, the only escape is to interrupted the execution of the \gls{thrd}, either directly or due to regular preemption, only then can the \gls{proc} take the time to handle the pending request to help.
Interrupting \glspl{thrd} for this purpose is far from desireable, the cost is significant and the situation may be hard to detect.
However, a more subtle reason why interrupting the \gls{thrd} is not a satisfying solution is that the \gls{proc} is not actually using the instance it is tied to.
If it were to use it, then helping could be done as part of the usage.
Interrupts are needed here entirely because the \gls{proc} is tied to an instance it is not using.
Therefore a more satisfying solution would be for the \gls{thrd} submitting the operation to simply notice that the instance is unused and simply go ahead and use it.
This is the approach presented next.

\subsubsection{Instance borrowing}
Both of the approaches presented above have undesirable aspects that stem from too loose or too tight coupling between @io_uring@ and \glspl{proc}.
In the first approach, loose coupling meant that all operations have synchronization overhead that a tighter coupling can avoid.
The second approach on the other hand suffers from tight coupling causing problems when the \gls{proc} do not benefit from the coupling.
While \glspl{proc} are continously issuing \io operations tight coupling is valuable since it avoids synchronization costs.
However, in unlikely failure cases or when \glspl{proc} are not making use of their instance, tight coupling is no longer advantageous.
A compromise between these approaches would be to allow tight coupling but have the option to revoke this coupling dynamically when failure cases arise.
I call this approach ``instance borrowing''\footnote{While it looks similar to work-sharing and work-stealing, I think it is different enough from either to warrant a different verb to avoid confusion.}.

In this approach, each cluster owns a pool of @io_uring@ instances managed by an arbiter.
When a \gls{thrd} attempts to issue an \io operation, it ask for an instance from the arbiter and issues requests to that instance.
However, in doing so it ties to the instance to the \gls{proc} it is currently running on.
This coupling is kept until the arbiter decides to revoke it, taking back the instance and reverting the \gls{proc} to its initial state with respect to \io.
This tight coupling means that synchronization can be minimal since only one \gls{proc} can use the instance at any given time, akin to the private instances approach.
However, where it differs is that revocation from the arbiter means this approach does not suffer from the deadlock scenario described above.

Arbitration is needed in the following cases:
\begin{enumerate}
        \item The current \gls{proc} does not currently hold an instance.
        \item The current instance does not have sufficient SQEs to satisfy the request.
        \item The current \gls{proc} has the wrong instance, this happens if the submitting \gls{thrd} context-switched between allocation and submission.
        I will refer to these as \newterm{External Submissions}.
\end{enumerate}
However, even when the arbiter is not directly needed, \glspl{proc} need to make sure that their ownership of the instance is not being revoked.
This can be accomplished by a lock-less handshake\footnote{Note that the handshake is not Lock-\emph{Free} since it lacks the proper progress guarantee.}.
A \gls{proc} raises a local flag before using its borrowed instance and checks if the instance is marked as revoked or if the arbiter has raised its flag.
If not it proceeds, otherwise it delegates the operation to the arbiter.
Once the operation is completed, the \gls{proc} lowers its local flag.

Correspondingly, before revoking an instance the arbiter marks the instance and then waits for the \gls{proc} using it to lower its local flag.
Only then does it reclaim the instance and potentially assign it to an other \gls{proc}.

The arbiter maintains four lists around which it makes its decisions:
\begin{enumerate}
        \item A list of pending submissions.
        \item A list of pending allocations.
        \item A list of instances currently borrowed by \glspl{proc}.
        \item A list of instances currently available.
\end{enumerate}

\paragraph{External Submissions} are handled by the arbiter by revoking the appropriate instance and adding the submission to the submission ring.
There is no need to immediately revoke the instance however.
External submissions must simply be added to the ring before the next system call, \ie, when the submission ring is flushed.
This means that whoever is responsible for the system call first checks if the instance has any external submissions.
If it is the case, it asks the arbiter to revoke the instance and add the external submissions to the ring.

\paragraph{Pending Allocations} can be more complicated to handle.
If the arbiter has available instances, the arbiter can attempt to directly hand over the instance and satisfy the request.
Otherwise


% Verbs of this design

% Allocation: obtaining an sqe from which to fill in the io request, enforces the io instance to use since it must be the one which provided the sqe. Must interact with the arbiter if the instance does not have enough sqe for the allocation. (Typical allocation will ask for only one sqe, but chained sqe must be allocated from the same context so chains of sqe must be allocated in bulks)

% Submition: simply adds the sqe(s) to some data structure to communicate that they are ready to go. This operation can't fail because there are as many spots in the submit buffer than there are sqes. Must interact with the arbiter only if the thread was moved between the allocation and the submission.

% Flushing: Taking all the sqes that were submitted and making them visible to the kernel, also counting them in order to figure out what to_submit should be. Must be thread-safe with submission. Has to interact with the Arbiter if there are external submissions. Can't simply use a protected queue because adding to the array is not safe if the ring is still available for submitters. Flushing must therefore: check if there are external pending requests if so, ask the arbiter to flush otherwise use the fast flush operation.

% Collect: Once the system call is done, it returns how many sqes were consumed by the system. These must be freed for allocation. Must interact with the arbiter to notify that things are now ready.

% Handle: process all the produced cqe. No need to interact with any of the submission operations or the arbiter.




% alloc():
%       proc.io->in_use = true, __ATOMIC_ACQUIRE
%       if cltr.io.flag || !proc.io || proc.io->flag:
%               return alloc_slow(cltr.io, proc.io)

%       a = alloc_fast(proc.io)
%       if a:
%               proc.io->in_use = false, __ATOMIC_RELEASE
%               return a

%       return alloc_slow(cltr.io)

% alloc_fast()
%       left = proc.io->submit_q.free.tail - proc.io->submit_q.free.head
%       if num_entries - left < want:
%               return None

%       a = ready[head]
%       head = head + 1, __ATOMIC_RELEASE

% alloc_slow()
%       cltr.io.flag = true, __ATOMIC_ACQUIRE
%       while(proc.io && proc.io->in_use) pause;



% submit(a):
%       proc.io->in_use = true, __ATOMIC_ACQUIRE
%       if cltr.io.flag || proc.io != alloc.io || proc.io->flag:
%               return submit_slow(cltr.io)

%       submit_fast(proc.io, a)
%       proc.io->in_use = false, __ATOMIC_RELEASE

% polling()
%       loop:
%               yield
%               flush()
%               io_uring_enter
%               collect
%               handle()

\section{Interface}
Finally, the last important part of the \io subsystem is it's interface. There are multiple approaches that can be offered to programmers, each with advantages and disadvantages. The new \io subsystem can replace the C runtime's API or extend it. And in the later case the interface can go from very similar to vastly different. The following sections discuss some useful options using @read@ as an example. The standard Linux interface for C is :

@ssize_t read(int fd, void *buf, size_t count);@.

\subsection{Replacement}
Replacing the C \glsxtrshort{api}

\subsection{Synchronous Extension}

\subsection{Asynchronous Extension}

\subsection{Interface directly to \lstinline{io_uring}}