[PATCH v6 13/15] selftests/nolibc: add mmap_bad test case

Zhangjin Wu falcon at tinylab.org
Sun Jul 9 11:33:22 PDT 2023


Hi, Willy

> On Fri, Jul 07, 2023 at 11:05:49PM +0800, Zhangjin Wu wrote:
> > The length argument of mmap() must be greater than 0, passing a zero
> > length argument expects failure with -EINVAL.
> 
> This one doesn't work for me on x86_64 kernel 5.15.112, qemu userland:
> 
>   46 mmap_bad = <0x0> EEXIST  != (<0xffffffffffffffff> EINVAL)    [FAIL]
>

Just rerun 'run-user' on x86_64 with kernel 5.11.0-41-generic, it is ok.

The failure is very interesting, and also rechecked the kernel mmap code:

    unsigned long do_mmap(struct file *file, unsigned long addr,
                        unsigned long len, unsigned long prot,
                        unsigned long flags, unsigned long pgoff,
                        unsigned long *populate, struct list_head *uf)
    {
        ...
        if (!len)
                return -EINVAL; 
        ...
    }

    $ git blame -L 1202,1202 mm/mmap.c
    e37609bb36f70 (Piotr Kwapulinski 2015-06-24 16:58:39 -0700 1202)        if (!len)

    $ git show e37609bb36f706c954e82e580e2e790e9d5caef8:Makefile 
    VERSION = 4
    PATCHLEVEL = 1
    SUBLEVEL = 0
    EXTRAVERSION =

So, the kernel side should be ok from v4.1?

For qemu-user, I have rechecked the following version:

    $ qemu-x86_64 --version
    qemu-x86_64 version 4.2.1 (Debian 1:4.2-3ubuntu6.18)

    $ qemu-x86_64 --version
    qemu-x86_64 version 7.0.0 (Debian 1:7.0+dfsg-7ubuntu2.6~backport20.04.202306190332~ubuntu20.04.1)

    $ build/x86_64/pc/qemu/v8.0.2/qemu-x86_64 --version
    qemu-x86_64 version 8.0.2 (v8.0.2-dirty)

all of them work well, as a comparison, what's your qemu-user version?

> This EEXIST actually is the errno from the previous test. If I run
> the test natively it's OK:
> 
>   $ ./nolibc-test syscall:46
>   Running test 'syscall'
>   46 mmap_bad = <0xffffffffffffffff> EINVAL                        [OK]
>   Errors during this test: 0
> 
> I'll queue it anyway for now but it would be nice that we figure what's
> happening (even if we need to adjust or drop the test if it's a false
> positive) so that we don't get used to "ah this is a normal error".
>

Yes, if there is a failure, we should figure out why. It is ok for me to remove
this one or let's find another errno condition before we find the root cause of
the reported failure. 

Thanks,
Zhangjin
 
> Willy



More information about the linux-riscv mailing list