Understanding ZFS Channel Programs
klarasystems.com
klarasystems.com
For example, consider what the API must support to make a channel program that lists all the child datasets of a given dataset, and takes a snapshot of each child dataset that has a certain ZFS property.
The looping cannot be done by the user program, as that would introduce edge cases when datasets are created or destroyed between listing and generating the snapshot commands.
As such, picking Lua as the API isn't a bad choice I think.
1. Find the latest three snapshots, and then 2. delete those three snapshots.
With a generic API, that’d take two calls, with a possible TOCTOU in between so it wouldn’t be atomic.
But if you send that same program to the ZFS subsystem, and have it evaluate and execute the script, the subsystem would be able to guarantee atomicity, right?
begin()
loop {}To enable the client program to submit "the whole thing", you really need to represent that sequence of operations as some kind of instruction stream loaded into a buffer — either a text-based one (source code in some language), or bytecode in some ISA.
(Why not capture a sequence of API operations that do nothing when called, but instead build such an instruction stream; where "committing" them actually runs the instruction stream? Y'know, like a Redis pipeline? Because a "pipeline" like that can't branch based on the result of executing part of it, unless branching/looping/etc are also implemented as operations in this hypothetical fluent-API-DSL. At which point, you're just programming with the semantics of another language, with extra steps.)
And if a static analysis pass is going to occur to check the code for runtime complexity, then you're already transforming the input into a different verified output with an intermediate AST-like structure; so at that point you may as well accept a text-based format and expose a "compile" system call. Just as GPU drivers do for compute shaders.
I'm guessing that there's probably an internal ZFS API allowing you to instead pass in some kind of bytecode (LuaJIT bytecode?) and just have it directly static-analyzed and plopped into a kernel handle, rather than parsed, then static-analyzed, then codegen'ed. They'd definitely need something like that for testing.
For example, the way this is handled in foundationdb is that you create a transaction, then within that transaction issue a series of read and write calls. The read calls are executed immediately, and they see the database as if other concurrent changes weren’t happening. The write calls get batched up and at the end of the transaction block they are submitted together - with information about which keys were read. On submission, the transaction block can fail with E_RETRY - which means a conflict happened, the transaction was entirely aborted and the userland program should run the transaction again from the start.
I’d absolutely adore an API like this for filesystem operations. There are all sorts of half baked, dirty workarounds to get around the lack of atomicity in posix filesystems. Eg, that whole “write a file then rename it” thing - which almost nobody implements correctly because it’s error prone and most correct solutions on top of posix have terrible performance.
Picture a database server where once you begin a transaction, the whole DB server goes single-threaded and doesn't serve any other clients until the TX completes. Fundamentally, at least for its more arcane global operations, that's exactly what a filesystem is. Usually this doesn't matter, because even these arcane filesystem operations take on the order of microseconds to complete (just with lots of non-locked pre/post execution overhead that inflates the total syscall time.) But a userspace program can do arbitrary-much stuff in a callback. It can even just get into an infinite loop, and never get back to the kernel. This is why there isn't such a thing as a system call with userspace callbacks! (Or any API equivalent to one—e.g. one with tx_begin + tx_commit calls, where tx_begin acquires a global kernel lock before returning to userspace, under the expectation that userspace call tx_commit to release the kernel lock.)
On the other hand, submitting a "callback program" as a whole to the kernel, allows the kernel to statically analyze the behavior of the "callback program" in advance; and only if it is determined to be "simple" (e.g. if it predictably halts after a short time as a static property), will the kernel issue the program a capability for actually running that "callback program."
This is how compute shaders work. This is how eBPF works. And this is (apparently) how this ZFS Lua thing works. They all do it for similar reasons: to ensure that the program is a complete, soft-realtime practical unit of work to be executing in some bounded-time context.
You do get atomicity from this, but it's not MVCC atomicity. It's more like Redis's atomicity: during the execution of a ZFS Lua program, all other clients making system calls against the filesystem will wait on acquiring the filesystem mutex, and so will block until the program completes. There's never a case where other activity might "interleave" with the Lua program's execution, and cause it to retry. There is no other activity. The only case in which the Lua program can actually fail (and so get rolled back), is if it dies/is killed in the middle of execution, due to e.g. a hardware power cut. That's the kind of atomicity that ZFS is concerned about—guaranteeing that changes that make it into the filesystem journal, are complete valid change units. In this case, one Lua program is one change unit in the journal.
In this theoretical design, you could just block all other administrative modifications. I don't think you need to tie up an entire CPU core, and I'm fairly sure that these zfs operations don't block regular reads and writes.
I think you had it right in your initial comment. There's no good way express branching with an implementation which incrementally submits operations to be committed as a batch. You'd have to take an admin lock on an entire zpool.
EDIT: talked to a zfs dev, said this would take the txg sync lock.
Think of it like trying to deschedule a userspace thread in the middle of it having jumped to kernelspace to handle an interrupt. It just wouldn't work; that's not a pre-emptible state, not one that can be cleanly represented during a context switch with a PUSHA, not one where pre-emption would leave the kernel in a known state, etc.
So the CPU core is tied up because the original thread can't be descheduled, and instead would still be "stuck" in the middle of the system call, doing a busy-wait on the result of the callback. To make the callback actually happen in this hypothetical design, the execution of the callback would need to be scheduled onto another CPU core, using some system-global callback-scheduler like Apple's libdispatch.
Note that this is also why, in Linux, processes stuck in the D state are unkillable. They're stuck "inside" a blocking system call, and so cannot be descheduled, even by the process manager trying to hard-kill them (which, in the end, requires the system call to at least return to the kernel so that the kernel resources involved can reach a known postcondition state.)
And this is why innovations like io_uring make so much sense in Linux — they allow a userspace process to 1. make a long-running blocking syscall, while also 2. spawning a worker subprocess to communicate asynchronously with the logic inside the running syscall, by queuing messages back and forth through the kernel rings. (Picture, say, sendfile(2) messaging your worker to let you observe the progress of the operation, and/or to signal it on a channel to gracefully cancel the operation-in-progress.)
In this imaginary design, the syscalls you make would look something like:
- BeginChannelTx -> return ChannelTxID
- ReadZFSProperties(ChannelTxID, params) -> return data
- DestroySomeDatasets(ChannelTxID, params) -> ok
- CommitChannelTx(ChannelTxID)
Notably, DestroySomeDatasets doesn't actually do any work. It merely records which datasets you want to destroy. There are no callbacks as far as I can see: there's no kernel thread waiting on a user thread to do something. This way also lets you express branching.
The drawback of this approach is you need a lock on all mutating administrative commands when you call BeginChannelTX. I talked to a ZFS dev, and he said that with ZFS' design, that's actually the txg sync lock. This means that while reads will proceed, writes will only proceed for a short period of time, and nothing will make it to disk. The overhead of making all these syscalls was also judged to be problematic.
So at some point I bet you'll be basically uploading BPF like code to peripherals directly in order to accomplish really fast I/O. Don't know how that's going to look at all from a post-POSIX API perspective.
Compare to eBPF, which not only uses a JIT compiler in the kernel, but on popular Linux distributions like Fedora and Ubuntu permits unprivileged users to run eBPF programs.
I would never assume a lack of exploitable bugs in the fork of the Lua interpreter used in FreeBSD, but the fact is that eBPF has seen a steady, if infrequent, stream of actual exploits over the years and in particular unprivileged user exploits.
Call me a pedant but this grinds my gears. The speed my CPU will carry out instructions is not determined by if it's in a kernel or user context. Avoiding making switches between those contexts certainly means less work though.
For starters, it doesn’t run on the channel subsystem, but uses the CPU for it. Also, channel programs (the IBM ones) can do a lot of stuff - IBM’s ISAM uses self-modifying channel programs. These channels can’t be that flexible because they run within the kernel and being too flexible would be a security risk.
Low/mid range mainframes don't really make sense anymore, but channel programs did run on the main CPU on quite a few low/mid range mainframe systems.
> Also, channel programs (the IBM ones) can do a lot of stuff - IBM’s ISAM uses self-modifying channel programs. These channels can’t be that flexible because they run within the kernel and being too flexible would be a security risk.
There's no reason this couldn't do that either as long as it's run in a preemptable context (which I'm pretty sure is the case from looking at the example program). Schemes like eBPF mainly control code flow in pursuit of the goal of running in non preemptable contexts like interrupts.