Re: [Crash-utility] [PATCH v3 4/4] Add assert check if an element is installed one more time
On Fri, Sep 17, 2021 at 5:09 PM Philipp Rudo <prudo(a)redhat.com> wrote:
Hi Tao,
On Tue, 14 Sep 2021 17:01:54 +0800
Tao Liu <ltao(a)redhat.com> wrote:
> symname_hash_install won't check if spn has been installed before. If
> it does, the second install will corrupt the hash table as well as
> spn->cnt counting. It should never happen in normal cases, but we need
> get notified if it happens.
>
> Signed-off-by: Tao Liu <ltao(a)redhat.com>
> ---
> symbols.c | 1 +
> 1 file changed, 1 insertion(+)
>
> diff --git a/symbols.c b/symbols.c
> index f8b4998..a453a6a 100644
> --- a/symbols.c
> +++ b/symbols.c
> @@ -1167,6 +1167,7 @@ symname_hash_install(struct syment *table[], struct syment
*spn)
> spn->name_hash_next = NULL;
> } else {
> while (sp) {
> + assert(sp != spn);
the change makes totally sense. Although I'm not sure if the 'assert'
isn't too strict. In my opinion printing a warning + ignoring the
duplicate symbol would be better.
In my opinion an 'assert' should only be used, in situations a program
cannot be recovered from. But for the case when a symbol is added twice
we can 'recover' from by simply ignoring the second request.
OK. I will replace it with a function which can check if spn exists in
the hashtable.
Thanks,
Tao Liu
Thanks
Philipp
> if (STREQ(sp->name, spn->name)) {
> sp->cnt++;
> spn->cnt++;