Re: [PATCH 15/22] Implement mlock(2), munlock(2), mlockall(2), munlockall(2), madvise(2), minherit(2)

[Richard Henderson]

On 8/19/23 21:37, Warner Losh wrote:
>     +/* madvise(2) */
>     +static inline abi_long do_bsd_madvise(abi_long arg1, abi_long arg2,
>     +        abi_long arg3)
>     +{
>     +    /*
>     +     * A straight passthrough may not be safe because qemu sometimes
>     +     * turns private file-backed mapping into anonymous mappings. This
>     +     * will break MADV_DONTNEED.  This is a hint, so ignoring and returing
>     +     * success is ok.
>     +     */
>     +    return get_errno(0);
>
>
> This looks like it was copied from an early linux-user implementation, and
> that seems to have been fixed to no longer cause problems. Can someone
> that knows about the linux-user history here comment?

We now track pages that are "passthrough" and ok for DONTNEED etc.


    case MADV_DONTNEED:
        if (page_check_range(start, len, PAGE_PASSTHROUGH)) {
            ret = get_errno(madvise(g2h_untagged(start), len, advice));
            if ((advice == MADV_DONTNEED) && (ret == 0)) {
                page_reset_target_data(start, start + len - 1);
            }
        }


It's still not ideal, but it's something.


r~