On Fri, 2 May 2025 10:10:33 +0300 Dan Carpenter <dan.carpenter@xxxxxxxxxx> wrote:
> On Fri, May 02, 2025 at 03:46:21PM +0900, Honggyu Kim wrote:
> > Hi Dan,
> >
> > On 4/23/2025 5:24 PM, Dan Carpenter wrote:
> > > Return -EEXIST if the node already exists. Don't return success.
> > >
> > > Fixes: 1bf270ac1b0a ("mm/mempolicy: support memory hotplug in weighted interleave")
> > > Signed-off-by: Dan Carpenter <dan.carpenter@xxxxxxxxxx>
> > > ---
> > > Potentially returning success was intentional? This is from static
> > > analysis and I can't be totally sure.
> > >
> > > mm/mempolicy.c | 3 ++-
> > > 1 file changed, 2 insertions(+), 1 deletion(-)
> > >
> > > diff --git a/mm/mempolicy.c b/mm/mempolicy.c
> > > index f43951668c41..0538a994440a 100644
> > > --- a/mm/mempolicy.c
> > > +++ b/mm/mempolicy.c
> > > @@ -3539,7 +3539,7 @@ static const struct kobj_type wi_ktype = {
> > > static int sysfs_wi_node_add(int nid)
> > > {
> > > - int ret = 0;
> > > + int ret;
> > > char *name;
> > > struct iw_node_attr *new_attr;
> > > @@ -3569,6 +3569,7 @@ static int sysfs_wi_node_add(int nid)
> > > if (wi_group->nattrs[nid]) {
> > > mutex_unlock(&wi_group->kobj_lock);
> > > pr_info("node%d already exists\n", nid);
> > > + ret = -EEXIST;
> >
> > Returning -EEXIST here looks good to me, but could you remove the above pr_info
> > as well? I mean the following change is needed.
> >
> > - pr_info("node%d already exists\n", nid)
> > + ret = -EEXIST;
> >
> > We don't need the above pr_info here because we delegate a warning message to
> > its caller wi_node_notifier().
> >
> > This can close another warning report below.
> > https://lore.kernel.org/all/202505020458.yLHRAaW9-lkp@xxxxxxxxx
> >
> > If you apply my suggestion then please add
> >
> > Reviewed-by: Honggyu Kim <honggyu.kim@xxxxxx>
> >
>
> Rakie Kim was pretty confident that returning 0 was intentional. Btw,
> Smatch considers it intentional if the "ret = 0;" is within 5
> lines of the goto. Or we could add a comment, which wouldn't silence
> the warning but it would help people reading the code.
>
Hi Dan,
Thank you for taking the time to review this code and point out the issue.
I believe there may have been some confusion related to the behavior in
previous versions.
In the latest revision, the `wi_node_notifier()` function that calls
`sysfs_wi_node_add()` has been updated to always return `NOTIFY_OK`,
regardless of the return value from `sysfs_wi_node_add()`. This ensures that
no memory hotplug event will be blocked by our notifier logic.
static int wi_node_notifier(struct notifier_block *nb,
unsigned long action, void *data)
{
...
switch (action) {
case MEM_ONLINE:
err = sysfs_wi_node_add(nid);
if (err)
pr_err("failed to add sysfs for node%d during hotplug: %d\n",
nid, err);
break;
...
return NOTIFY_OK;
}
Given that, it is appropriate for `sysfs_wi_node_add()` to return `-EEXIST`
when the node already exists. Since the error message is already logged by
`wi_node_notifier()`, I agree with Honggyu's suggestion to remove the
redundant `pr_info()` statement as well:
- pr_info("node%d already exists\n", nid);
+ ret = -EEXIST;
Once again, thank you very much for your review and for helping us improve
the code.
Reviewed-by: Rakie Kim <rakie.kim@xxxxxx>
Rakie
> regards,
> dan carpenter
>
>
