mmap_anonymous will exhaust memory on `riscv64gc-unknown-linux-gnu` when memory size is big.

[yuyang-ok]

I didn't know this issue related to this project or qemu-riscv64.

abc.tar.gz

[yuyang-ok]

I test two version of qemu, all of there have this problem.

yuyang@yuyang-Inspiron-7590:~/projects/wasmtime/tests$ qemu-riscv64  --version 
qemu-riscv64 version 7.0.94 (v7.1.0-rc4)
Copyright (c) 2003-2022 Fabrice Bellard and the QEMU Project developers

yuyang@yuyang-Inspiron-7590:~/projects/abc-rust$ qemu-riscv64-static --version 
qemu-riscv64 version 6.2.0 (Debian 1:6.2+dfsg-2ubuntu6.4)
Copyright (c) 2003-2021 Fabrice Bellard and the QEMU Project developers

[bjorn3]

Did you use the version of qemu that wasmtime uses in CI? It has a patch to fix a bug in qemu that I believe is the same as this one.

[yuyang-ok]

No, I am using from official.

[yuyang-ok]

I met qemu-riscv64 problem before.
https://bytecodealliance.zulipchat.com/#narrow/stream/206238-general/topic/.E2.9C.94.20target-riscv64.20.20decommit_table_pages.2E
I download the latest version qemu , this problem solved.
the two problem are related?

[sunfishcode]

This doesn't appear to be a rustix bug; I can reproduce the same behavior with libc:

        let r = libc::mmap(
            ptr::null_mut(),
            6666 << 30,
            libc::PROT_NONE,
            libc::MAP_PRIVATE | libc::MAP_ANONYMOUS,
            -1,
            0,
        );
        assert_ne!(r, libc::MAP_FAILED);

[yuyang-ok]

I did some test using c.

#include <stdio.h>
#include <sys/mman.h>
#include <errno.h>

int main(int argc, char **argv)
{
    void *addr;
    size_t length = 6666 << 30;
    addr = mmap(NULL, length, PROT_NONE, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
    if (addr == -1)
    {
        printf("errno:%d\n", errno);
        return 1;
    }
    printf("ptr:%d", addr);
    getchar();
    return 0;
}

yuyang@yuyang-Inspiron-7590:~/projects/abc-rust/src$ /home/yuyang/projects/qemu/build/qemu-riscv64 -L /usr/riscv64-linux-gnu -E LD_LIBRARY_PATH=/usr/riscv64-linux-gnu/lib ./a.out 
errno:12

@sunfishcode is that rustix should behave just like c did.

[yuyang-ok]

@bjorn3 I patched the code.

I am using qemu.

yuyang@yuyang-Inspiron-7590:~/projects/qemu/build$ git log 
commit 9a99f964b152f8095949bbddca7841744ad418da (HEAD -> master, tag: v7.1.0-rc4, origin/master, origin/HEAD)
Author: Richard Henderson <richard.henderson@linaro.org>
Date:   Wed Aug 24 19:27:56 2022 -0700

    Update version for v7.1.0-rc4 release
    
    Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

I still get the error code 12.

[sunfishcode]

That C program isn't equivalent because it's evaluating 6666 << 30 as type int, which overflows, producing a different value when subsequently converted to size_t.

[yuyang-ok]

@sunfishcode sorry.

#include <stdio.h>
#include <sys/mman.h>
#include <errno.h>

int main(int argc, char **argv)
{
    void *addr;
    int a = 6666;
    size_t length = a << 30;
    addr = mmap(NULL, length, PROT_NONE, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
    if (addr == MAP_FAILED)
    {
        printf("errno:%d\n", errno);
        return 1;
    }
    getchar();
    return 0;
}
yuyang@yuyang-Inspiron-7590:~/projects/abc-rust/src$ /home/yuyang/projects/qemu/build/qemu-riscv64 -L /usr/riscv64-linux-gnu -E LD_LIBRARY_PATH=/usr/riscv64-linux-gnu/lib ./a.out 
errno:12

I still get this error code.

[yuyang-ok]

abc.tar.gz

[sunfishcode]

That C code still computes 6666 << 30 in type int, so it still overflows before converting to size_t.

[yuyang-ok]

@sunfishcode ok, thanks.