Pekka Enberg
59d28eac93
core: Switch Completion "is_completed" to use Cell
...
Suggested by Jussi
2025-05-27 14:05:07 +03:00
Pekka Enberg
05df548b10
core/io: Add wait_for_completion() to I/O dispatcher
2025-05-27 13:47:40 +03:00
Pekka Enberg
eca9a5b703
core/io: Switch to Arc<Completion>
2025-05-27 11:28:49 +03:00
pedrocarlo
0c22382f3c
shared lock on file and throw ReadOnly error in transaction
2025-05-02 16:30:48 -03:00
Diego Reis
09d83aadf3
Fix dumb conflict errors
2025-04-13 11:10:06 -03:00
Diego Reis
d9bf383507
core/io: Untie MemoryIO's lifetime of the IO layer
2025-04-13 11:10:06 -03:00
Diego Reis
79f8b83cbe
Fix dumb clippy errors
2025-04-13 11:10:06 -03:00
Diego Reis
b519509349
core/io: Add internal in-memory MemoryIO to all IO layers
...
Honestly I don't have 100% sure if this is a good idea, the reasoning is that in any IO we'll want to do memory only operations like creating tables etc, so may want a common way to access it
2025-04-13 11:10:05 -03:00
Jussi Saurio
60a13c129f
io/linux: make syscallio the default (io_uring is really slow)
2025-04-10 13:32:26 +03:00
Pekka Enberg
e21e2b5449
Merge 'core: Fix syscall VFS on Linux' from Pekka Enberg
...
Fix the syscall VFS on Linux not to use `PlatformIO`, which is just an
alias for `io_uring`.
Reviewed-by: Jussi Saurio <jussi.saurio@gmail.com>
Closes #1285
2025-04-09 19:38:45 +03:00
Pekka Enberg
2d009083ba
core: Fix syscall VFS on Linux
...
Fix the syscall VFS on Linux not to use `PlatformIO`, which is just an
alias for `io_uring`.
2025-04-09 19:27:58 +03:00
Avinash Sajjanshetty
2873c36b31
Implement trait Clock for implemented IOs
...
Replace `chrono::Local::now()` to return `Instant` but containing
same info
2025-04-06 23:34:13 +05:30
Avinash Sajjanshetty
8d6d50d9d5
Update IO trait
...
- Remove existing `get_current_time() -> String`
- Add super trait bound `Clock`
2025-04-06 23:06:57 +05:30
Avinash Sajjanshetty
02ee2cf90e
Add Clock trait
2025-04-06 23:02:41 +05:30
PThorpe92
e9420e7d2b
Fix platform specific ffi c ptr types
2025-03-24 22:48:07 -04:00
Daniel Boll
fc50609491
Fix io_uring WAL write corruption by ensuring buffer lifetime
...
Ensure the Arc<RefCell<Buffer>> in UringFile::pwrite remains alive until
the io_uring write completes by referencing it in the completion callback.
This prevents WAL file corruption where the correct buffer data was
overwritten with stale memory (e.g., 00 18 27 xx instead of 37 7f 06 82).
Validation:
- Tested with limbo -v io_uring and WAL operations.
- Verified with xxd and wal-browser.
Signed-off-by: Daniel Boll <danielboll.academico@gmail.com>
2025-03-19 23:55:43 -03:00
PThorpe92
57d4aa7216
Reorganize ext library and feature gate vfs to more easily prevent wasm build issues
2025-03-19 10:17:11 -04:00
PThorpe92
8e4705d6c6
Fix read completion error in vfs
2025-03-14 18:08:49 -04:00
PThorpe92
44f6054657
Impl copy + clone for io openflags
2025-03-12 21:52:51 -04:00
PThorpe92
8d3c44cf00
Impl IO trait for VfsMod type
2025-03-12 21:52:50 -04:00
Pere Diaz Bou
b63f541150
impl Send Sync for GenericFile
2025-03-05 14:07:48 +01:00
Pere Diaz Bou
d1e7f0826b
impl Send Sync for WindowsFile
2025-03-05 14:07:48 +01:00
Pere Diaz Bou
3d3cdbeb0c
fix generic and windows sync compilation
2025-03-05 14:07:48 +01:00
Pere Diaz Bou
ab2fc3fb13
fix clippy arc warnings for now
2025-03-05 14:07:48 +01:00
Pere Diaz Bou
d1c7d758c4
fix uring Sync,Sync unsafe impl
2025-03-05 14:07:48 +01:00
Pere Diaz Bou
e4a8ee5402
move load extensions to Connection
...
Extensions are loaded per connection and not per database as per SQLite
behaviour. This also helps with removing locks.
2025-03-05 14:07:48 +01:00
Pere Diaz Bou
8daf7666d1
Make database Sync + Send
2025-03-05 14:07:48 +01:00
Pekka Enberg
01671d1d1e
Merge 'Performance improvements for Unix I/O backend' from Preston Thorpe
...
This PR reworks the unix I/O backend, removing runtime reference
counting/borrow checking and optimizing away the hashmap in favor of a
static array, with an unlikely fallback vec.
The only reason the fallback vec is there is because unlike the
`io_uring` module, we cannot simply index into the array with the fd as
the OS could theoretically give us a fd up to I believe 1024 so keeping
an array of that size for a few elements is unnecessary.
Closes #940
2025-02-12 09:29:41 +02:00
PThorpe92
e3ab80ae28
Replace Option with MaybeUninit in unix IO
2025-02-11 09:09:49 -05:00
PThorpe92
9098237a12
Add as_read method to completion enum
2025-02-11 09:09:47 -05:00
PThorpe92
975cf95c37
Reformat unix io with wrappers for unsafecell
2025-02-11 09:06:49 -05:00
PThorpe92
159e2fbd71
Remove rc refcell in unix io module
2025-02-11 09:01:57 -05:00
Doug Anderson444
0aa5de6d9f
rm log, switch all to tracing
2025-02-11 09:03:36 -04:00
Pekka Enberg
0638550be7
Merge 'Remove unnecessary reference counting from completion I/O callbacks' from Preston Thorpe
...
I am on a bit of a mission to revisit a lot of the ref counting, this
was an easy first win.
It seems to be a linear path of function calls or hashmaps which can own
the completions directly, no cloning needed.
Reviewed-by: Jussi Saurio <jussi.saurio@gmail.com>
Closes #912
2025-02-10 12:11:30 +02:00
Jorge López
83b158fb3a
core: silence some unused warnings when building without default features
2025-02-09 01:13:12 +01:00
PThorpe92
75898027a0
Remove unnecessary reference counting from completion io callbacks
2025-02-08 08:20:19 -05:00
PThorpe92
575a524d04
Replace hashmap for io_uring pending ops with static array
2025-02-07 09:05:08 -05:00
PThorpe92
b5a5b34329
Use Cell for size in memoryio in place of unsafecell
2025-02-01 23:04:42 -05:00
PThorpe92
5c00b3dc6e
Improve in-memory IO performance, remove runtime borrow checking
2025-02-01 21:54:39 -05:00
Jorge López
55e06b0c72
core/io: make file locks non-blocking so they fail right away
2025-01-14 11:01:11 +01:00
Pekka Enberg
5c9505e8f7
Revert "core/io/io_uring: replace nix and libc calls with their rustix counterparts."
...
This reverts commit b146f5d4cb
2025-01-14 10:25:23 +02:00
Jorge López
2f90a06533
core/io/unix: replace O_NONBLOCK flag from libc with equivalent from rustix
2025-01-13 21:03:05 +01:00
Jorge López
5e9cb58f04
core/io/io_uring: remove unnecessary path prefix for log macros, and replace one unwrap with ?
2025-01-13 20:21:37 +01:00
Jorge López
b146f5d4cb
core/io/io_uring: replace nix and libc calls with their rustix counterparts.
...
core: remove dependency on nix. We keep depending on libc, though, because crate io_uring requires libc's iovec.
2025-01-13 20:21:37 +01:00
Jorge López
b1e8f2da73
core/io/unix: minor formatting
2025-01-13 20:21:37 +01:00
Jorge López
7b5e5efd14
core/io/unix: replace libc calls and types with their rustix counterparts
2025-01-13 20:21:37 +01:00
Jorge López
7808665c92
core: make MAX_IOVECS u32 instead of usize, to match the type expected by io_uring
2025-01-13 20:21:37 +01:00
Jorge López
5a45df84db
core: add debug line when an IO backend is created. A user can check that the correct one has been selected.
2025-01-11 16:08:41 +01:00
Jorge López
2596e0800e
core: expose UnixIO apart from PlatformIO on any Unix, and expose UringIO on Linux with feature io_uring.
...
cli: add a new argument to select I/O backend (more than one option only for Linux with io_uring feature).
cli: make both Limbo::new() and Limbo::open_db() use get_io(), unifying parsing of database path and eliminating duplicated code.
2025-01-11 15:21:26 +01:00
Jorge López
511c0b495d
Rename LinuxIOError to UringIOError to match the IO backend renames
2025-01-07 15:19:31 +01:00
