Working primitive, slow, prototype.

Fortune for collect's current commit: Small blessing − 小吉
safe-memfd
Avril 2 years ago
parent 2aa8c832de
commit a5ad02a6de

10
.gitignore vendored

@ -1 +1,9 @@
collect
/target
*~
# Profiling
*.png
*.svg
perf.*
vgcore.*

57
Cargo.lock generated

@ -0,0 +1,57 @@
# This file is automatically @generated by Cargo.
# It is not intended for manual editing.
version = 3
[[package]]
name = "bytes"
version = "1.1.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "c4872d67bab6358e59559027aa3b9157c53d9358c51423c17554809a8858e0f8"
[[package]]
name = "cc"
version = "1.0.73"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "2fff2a6927b3bb87f9595d67196a70493f627687a71d87a0d692242c33f58c11"
[[package]]
name = "collect"
version = "0.1.0"
dependencies = [
"bytes",
"jemallocator",
"libc",
]
[[package]]
name = "fs_extra"
version = "1.2.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "2022715d62ab30faffd124d40b76f4134a550a87792276512b18d63272333394"
[[package]]
name = "jemalloc-sys"
version = "0.3.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "0d3b9f3f5c9b31aa0f5ed3260385ac205db665baa41d49bb8338008ae94ede45"
dependencies = [
"cc",
"fs_extra",
"libc",
]
[[package]]
name = "jemallocator"
version = "0.3.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "43ae63fcfc45e99ab3d1b29a46782ad679e98436c3169d15a167a1108a724b69"
dependencies = [
"jemalloc-sys",
"libc",
]
[[package]]
name = "libc"
version = "0.2.122"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "ec647867e2bf0772e28c8bcde4f0d19a9216916e890543b5a03ed8ef27b8f259"

@ -0,0 +1,33 @@
[package]
name = "collect"
version = "0.1.0"
edition = "2021"
# See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
[features]
default = ["jemalloc"]
# TODO: mmap, memfd_create() ver
# TODO: make `bytes` optional, and use Vec<u8> by default instead? idk...
# Use jemalloc instead of system malloc.
# Seems to reduce overall memory usage at the cost of a very small speed drop.
jemalloc = ["jemallocator"]
[profile.release]
opt-level = 3
lto = "fat"
codegen-units = 1
strip=true
[profile.symbols]
inherits="release"
#incremental=true
strip=false
[dependencies]
bytes = "1.1.0"
jemallocator = { version = "0.3.2", optional = true }
libc = "0.2.122"

@ -1,289 +0,0 @@
// Collect all stdin into memory, then
#define _GNU_SOURCE
#include <stdio.h>
#include <stdlib.h>
#include <stdint.h>
#include <stdbool.h>
#include <unistd.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <sys/mman.h>
#include <errno.h>
#define LIKELY(expr) __builtin_expect(!!(expr), true)
#define UNLIKELY(expr) __builtin_expect(!!(expr), false)
#define _$_if__L true
#define _$_if__U false
#define $if(l, expr) if(__builtin_expect(!!(expr), _$_if__ ## l))
#define F_STDIN 0
#define F_STDOUT 1
#define F_STDERR 2
typedef union arguments {
struct {
off_t pages_per_buffer;
} sized;
struct {
size_t buffsz;
} unsized;
} option_t;
#define DEFAULT_OPTION ((option_t){ .sized = { .pages_per_buffer = 8 } })
static bool has_size(int fd, off_t* restrict size)
{
struct stat st;
if( fstat(fd, &st) < 0 ) { perror("failed to stat stdin"); return false; }
else if (st.st_size > 0) {
// Non-zero size
*size = st.st_size;
return true;
}
fprintf(stderr, "returned sz (fd %d): %ld\n", fd, st.st_size);
return false;
}
int collect_sized(off_t sz, const option_t* opt);
int collect_unsized(const option_t* opt);
int main(void)
{
off_t sz;
option_t args = DEFAULT_OPTION;
if(has_size(F_STDIN, &sz)) {
return collect_sized((size_t)sz, &args);
} else {
return collect_unsized(&args);
}
}
inline static
const void* map_input_buffer(int fd, size_t sz)
{
void* map = mmap(NULL, sz, PROT_READ, MAP_PRIVATE, fd, 0);
if(UNLIKELY(map == MAP_FAILED)) {
perror("input mmap()");
return NULL;
}
return map;
}
inline static
bool unmap_mem(void* mem, size_t len)
{
if(UNLIKELY( munmap(mem, len) != 0 )) {
perror("munmap()");
return false;
}
return true;
}
static int page_size()
{
static int _page_size=0;
if(UNLIKELY(!_page_size)) return _page_size = getpagesize();
return _page_size;
}
inline static
bool alloc_pages(off_t pages, int *restrict _fd, size_t* restrict _size)
{
int fd = memfd_create("collect-sized-buffer", O_RDWR);
$if(U, fd < 0) goto _e_memfd;
$if(U, fallocate(fd, 0, 0, __builtin_constant_p(_size) && !_size
? pages * page_size()
: _size ? (off_t)( *_size = pages * page_size() )
: pages * page_size()) != 0) goto _e_fallocate;
$if(L, _fd) *_fd = fd;
else close(fd);
return true;
// +Unwind+ //
_e_fallocate:
perror("fallocate()");
close(fd);
if(0)
_e_memfd:
perror("memfd_create()");
// -Unwind- //
return false;
}
struct map_fd {
void* map;
size_t len;
int fd;
};
static
bool map_pages(off_t pages, struct map_fd* restrict out)
{
$if(U, !out) return alloc_pages(pages, NULL, NULL);
$if(U, !alloc_pages(pages, &out->fd, &out->len)) goto _e_ap;
$if(U, (out->map = mmap(NULL, out->len, PROT_READ|PROT_WRITE, MAP_PRIVATE, out->fd, 0)) == MAP_FAILED) goto _e_map;
$if(U, madvise(out->map, out->len, MADV_MERGEABLE | MADV_WILLNEED)) goto _e_madv;
return true;
// +Unwind+ //
_e_madv:
perror("madv()");
munmap(out->map, out->len);
if(0)
_e_map:
perror("mmap()");
close(out->fd);
if(0)
_e_ap:
(void)0; // no perror() needed
// -Unwind- //
return false;
}
inline static
void unmap_pages(struct map_fd in, int *restrict keep_fd)
{
$if(U, munmap(in.map, in.len)) perror("munmap()");
if(__builtin_constant_p(keep_fd) && keep_fd) *keep_fd = in.fd;
else {
if(!keep_fd) {
$if(U, close(in.fd)) perror("close()");
} else *keep_fd = in.fd;
}
}
int collect_sized(off_t isz, const option_t* gopt)
{
register int rc=0;
__auto_type opt = gopt->sized;
const off_t real_max_size = page_size() * opt.pages_per_buffer;
// const off_t pages_per_isz = isz % page_size();
// const off_t page_leftover_isz = isz / page_size();
struct map_fd buffer;
if(!map_pages(opt.pages_per_buffer, &buffer)) return 1;
if(isz > real_max_size) {
// Multiple buffers needed
} else $if(U, isz == real_max_size) {
// Exactly one buffer (unlikely, but possible)
ssize_t r = splice(F_STDIN, NULL,
buffer.fd, NULL,
(size_t)isz,
SPLICE_F_MOVE);
switch(r) {
case -1: goto _e_splice;
case 0: /* TODO: splice reported end-of-input, should we ignore this? */
rc = 10;
goto _cleanup_splice;
default: {
fprintf(stderr, "splice()'d %lu bytes into buffer (%ld size @ %d)\n", r, buffer.len, buffer.fd);
}
break;
}
//TODO: splice() all bytes from that buffer into STDOUT
rc = 0;
} else {
// Less than one buffer
ssize_t r = splice(F_STDIN, NULL, // TODO: XXX: WHY does splice() **ALWAYS** fail??? it literally never works???
buffer.fd, NULL,
(size_t)isz,
SPLICE_F_MOVE);
switch(r) {
case -1: goto _e_splice;
case 0: /* TODO: splice reported end-of-input, should we ignore this? */
rc = 10;
goto _cleanup_splice;
default: {
fprintf(stderr, "splice()'d %lu bytes into buffer (%ld size @ %d)\n", r, buffer.len, buffer.fd);
}
break;
}
// TODO: splice() isz bytes from buffer into stdout
rc = 0;
}
// +Cleanup+ //
_cleanup_splice: if(0)
_e_splice: rc = (perror("splice()"), -1);
unmap_pages(buffer, NULL);
// -Cleanup- //
return rc;
}
int collect_unsized(const option_t* opt)
{
return 0;
}
#if 0
int collect_sized(off_t isz, const option_t* opt)
{
const size_t sz = (size_t)isz;
fprintf(stderr, "size of input: %lu, max size of mapping: %lu (buffers %lu / lo %lu)\n", sz, opt->sized.maxsz,
sz % opt->sized.maxsz,
sz / opt->sized.maxsz);
//fcntl(F_STDOUT, ... SOMETHING to make splice() work here...
//TODO :: XXX: : WHY can't we splice() here???? w/e..
#if 1
if( fallocate(F_STDOUT, 0 /* | FALLOC_FL_KEEP_SIZE*/, 0, isz) != 0) {
perror("fallocate(STDOUT)");
// return 1;
}
#endif
if( fcntl(F_STDOUT, F_SETFL, fcntl(F_STDOUT, F_GETFL) & ~O_APPEND) < 0 )
{
perror("fcntl(stdout) + O_APPEND");
return -O_APPEND;
}
ssize_t sprc = splice(F_STDIN, NULL,
F_STDOUT, NULL, //TODO: XXX: Why does this always fail? I've seen splice(1, 2) work before...
sz,
SPLICE_F_MOVE);
switch(sprc) {
case -1: perror("splice() whole buffer failed");
return 1;
case 0:
fprintf(stderr, "splice() reported end-of-input. TODO: continue splicing, or ignore?\n");
return 2;
default:
if((size_t)sprc == sz) return 0;
else if (sprc < sz) {
fprintf(stderr, "splice() moved only %ld / %lu bytes. TODO: move the other %lu bytes\n",
sprc, sz,
sz - (size_t)sprc);
return 3;
} else if(sprc > sz) fprintf(stderr, "splice() somehow moved %ld / %lu (+ %ld bytes more)\n",
sprc, sz,
(size_t)sprc - sz);
return -1;
}
#if 0
// Map stdin
const void* stdin_map = map_input_buffer(F_STDIN, sz);
if(!stdin_map) goto e_map_input;
cleanup:
unmap_mem((void*)stdin_map, sz);
if(0)
e_map_input:
{ fprintf(stderr, "failed to map stdin (%lu)\n", sz); rc = 1; }
return rc;
#endif
}
#endif

@ -0,0 +1,64 @@
#[cfg(feature="jemalloc")]
extern crate jemallocator;
#[cfg(feature="jemalloc")]
const _:() = {
#[global_allocator]
static ALLOC: jemallocator::Jemalloc = jemallocator::Jemalloc;
};
use std::{
io,
mem::MaybeUninit,
os::unix::prelude::*,
num::NonZeroUsize,
};
use bytes::{
BytesMut,
Buf,
BufMut,
};
fn try_get_size<R: ?Sized>(reader: &R) -> Option<NonZeroUsize>
where R: AsRawFd
{
let fd = reader.as_raw_fd();
use libc::{
fstat64,
stat64,
};
if fd < 0 {
return None;
}
let mut st: MaybeUninit<stat64> = MaybeUninit::uninit();
unsafe {
match fstat64(fd, st.as_mut_ptr()) {
0 => {
NonZeroUsize::new(st.assume_init().st_size as usize)
},
_ => None,
}
}
}
fn main() -> io::Result<()> {
let (bytes, read) = {
let stdin = io::stdin();
let mut bytes = match try_get_size(&stdin) {
Some(sz) => BytesMut::with_capacity(sz.into()),
None => BytesMut::new(),
};
let read = io::copy(&mut stdin.lock(), &mut (&mut bytes).writer())?;
(bytes.freeze(), read as usize)
};
let written = io::copy(&mut bytes.slice(..read).reader() , &mut io::stdout().lock())?;
if read != written as usize {
return Err(io::Error::new(io::ErrorKind::BrokenPipe, format!("read {read} bytes, but only wrote {written}")));
}
Ok(())
}
Loading…
Cancel
Save