Re: [PATCH v4 38/38] tests/tcg: Add a test for info proc mappings

[Richard Henderson <richard.henderson@linaro.org>]

On 6/30/23 11:04, Alex Bennée wrote:
> From: Ilya Leoshkevich <iii@linux.ibm.com>
> 
> Add a small test to prevent regressions.
> 
> Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
> Message-Id: <20230621203627.1808446-9-iii@linux.ibm.com>
> Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
...
> +def run_test():
> +    """Run through the tests one by one"""
> +    try:
> +        mappings = gdb.execute("info proc mappings", False, True)
> +    except gdb.error as exc:
> +        exc_str = str(exc)
> +        if "Not supported on this target." in exc_str:
> +            # Detect failures due to an outstanding issue with how GDB handles
> +            # the x86_64 QEMU's target.xml, which does not contain the
> +            # definition of orig_rax. Skip the test in this case.
> +            print("SKIP: {}".format(exc_str))
> +            return
> +        raise
> +    report(isinstance(mappings, str), "Fetched the mappings from the inferior")
> +    report("/sha1" in mappings, "Found the test binary name in the mappings")

This test fails on ppc64 host, or indeed any host with page size != 4k.

When host page size != target page size, and the executable image is small, such as sha1, 
then target_mmap cannot directly map the executable file, but must implement the mmap via 
MAP_ANON + pread.  Which leaves us with no binary name in the host /proc/self/maps for us 
to copy to the artificial guest /proc/self/maps.

One of the very many issues with page size mismatch...

I'm tempted to remove the test, but I suppose we could check host page size in python.


r~