I second that ...
Having i unsigned violates the rule of "least surprise".
If you need it unsigned make it clearly visible, also adding
a simple comment may help.
jm2c,
wh
________________________________________
Von: Dan Carpenter <dan.carpenter@xxxxxxxxxx>
Gesendet: Dienstag, 9. Februar 2021 10:19:23
An: Will Deacon
Cc: Colin King; Joerg Roedel; Matthias Brugger; Anan sun; Yong Wu; Chao Hao; Tomasz Figa; iommu@xxxxxxxxxxxxxxxxxxxxxxxxxx; linux-arm-kernel@xxxxxxxxxxxxxxxxxxx; linux-mediatek@xxxxxxxxxxxxxxxxxxx; kernel-janitors@xxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx
Betreff: Re: [PATCH][next] iommu/mediatek: Fix unsigned domid comparison with less than zero
On Thu, Feb 04, 2021 at 09:25:58AM +0000, Will Deacon wrote:
> On Wed, Feb 03, 2021 at 01:59:36PM +0000, Colin King wrote:
> > From: Colin Ian King <colin.king@xxxxxxxxxxxxx>
> >
> > Currently the check for domid < 0 is always false because domid
> > is unsigned. Fix this by making it signed.
> >
> > Addresses-CoverityL ("Unsigned comparison against 0")
>
> Typo here ('L' instead of ':')
>
> > Fixes: ab1d5281a62b ("iommu/mediatek: Add iova reserved function")
> > Signed-off-by: Colin Ian King <colin.king@xxxxxxxxxxxxx>
> > ---
> > drivers/iommu/mtk_iommu.c | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/drivers/iommu/mtk_iommu.c b/drivers/iommu/mtk_iommu.c
> > index 0ad14a7604b1..823d719945b2 100644
> > --- a/drivers/iommu/mtk_iommu.c
> > +++ b/drivers/iommu/mtk_iommu.c
> > @@ -640,7 +640,7 @@ static void mtk_iommu_get_resv_regions(struct device *dev,
> > struct list_head *head)
> > {
> > struct mtk_iommu_data *data = dev_iommu_priv_get(dev);
> > - unsigned int domid = mtk_iommu_get_domain_id(dev, data->plat_data), i;
> > + int domid = mtk_iommu_get_domain_id(dev, data->plat_data), i;
>
> Not sure if it's intentional, but this also makes 'i' signed. It probably
> should remain 'unsigned' to match 'iova_region_nr' in
> 'struct mtk_iommu_plat_data'.
iova_region_nr is either 1 or 5 so unsigned doesn't matter.
I once almost introduced a bug where the iterator was supposed to be
size_t. I fixed a bug by making it signed but I ended up introducing a
new bug. But generally that's pretty rare. The more common case is
that making iterators unsigned introduces bugs.
It's better to default to "int i;" and if more complicated types are
required that should stand out. "size_t pg_idx;" or whatever.
regards,
dan carpenter
