Re: [BUG] jfs: possible uninitialized-variable access in xtSplitUp()

From: Li Tuo
Date: Mon Aug 02 2021 - 22:46:36 EST

Thanks for your feedback, and any feedback about the further check would be appreciated!

Best wishes,
Tuo Li

On 2021/8/3 3:04, Dave Kleikamp wrote:
On 7/31/21 1:31 AM, Li Tuo wrote:

Our static analysis tool finds a possible uninitialized-variable access in the jfs driver in Linux 5.14.0-rc3:

At the beginning of the function xtSplitUp(), the variable rbn is not initialized.
If sp->header.flag & BT_ROOT is true,
780:    rc = (sp->header.flag & BT_ROOT) ? xtSplitRoot(tid, ip, split, &rmp) : xtSplitPage(tid, ip, split, &rmp, &rbn);

the varialbe rbn will remain uninitialized.
However, it is accessed through:
814:    rcbn = rbn;

I am not quite sure whether this possible uninitialized-variable access is real and how to fix it if it is real.
Any feedback would be appreciated, thanks!

I think the logic that protects us is that in the case where rbn is uninitialized, sp->header.flag & BT_ROOT, which means it SHOULD BE the last entry in btstack, so we shouldn't enter the loop:
    while ((parent = BT_POP(btstack)) != NULL) {

It does seem that some type of sanity check is warranted. I'll take a closer look and see if I can add some kind of error path if things are out of sync.

Reported-by: TOTE Robot <oslab@xxxxxxxxxxxxxxx>

Best wishes,
Tuo Li