Messages in this thread | | | Date | Tue, 19 May 2020 11:33:24 +0200 | From | Robert Richter <> | Subject | Re: [PATCH v2 02/10] EDAC/mc: Use int type for parameters of edac_mc_alloc() |
| |
On 23.04.20 19:49:34, Borislav Petkov wrote: > On Wed, Apr 22, 2020 at 01:58:06PM +0200, Robert Richter wrote: > > Most iterators use int type as index. mci->mc_idx is also type int. So > > use int type for parameters of edac_mc_alloc(). Extend the range check > > to check for negative values. There is a type cast now when assigning > > variable n_layers to mci->n_layer, it is safe due to the range check. > > > > While at it, rename the users of edac_mc_alloc() to mc_idx as this > > fits better here. > > > > Signed-off-by: Robert Richter <rrichter@marvell.com> > > --- > > drivers/edac/edac_mc.c | 7 +++---- > > drivers/edac/edac_mc.h | 6 +++--- > > 2 files changed, 6 insertions(+), 7 deletions(-) > > > > diff --git a/drivers/edac/edac_mc.c b/drivers/edac/edac_mc.c > > index 107d7c4de933..57d1d356d69c 100644 > > --- a/drivers/edac/edac_mc.c > > +++ b/drivers/edac/edac_mc.c > > @@ -444,8 +444,7 @@ static int edac_mc_alloc_dimms(struct mem_ctl_info *mci) > > return 0; > > } > > > > -struct mem_ctl_info *edac_mc_alloc(unsigned int mc_num, > > - unsigned int n_layers, > > +struct mem_ctl_info *edac_mc_alloc(int mc_idx, int n_layers, > > struct edac_mc_layer *layers, > > unsigned int sz_pvt) > > { > > @@ -456,7 +455,7 @@ struct mem_ctl_info *edac_mc_alloc(unsigned int mc_num, > > void *pvt, *ptr = NULL; > > bool per_rank = false; > > > > - if (WARN_ON(n_layers > EDAC_MAX_LAYERS || n_layers == 0)) > > + if (WARN_ON(mc_idx < 0 || n_layers < 1 || n_layers > EDAC_MAX_LAYERS)) > > return NULL; > > Yeah, no, this doesn't make sense to me. The memory controller number > and the number of layers can never ever be negative and thus signed. > > And some drivers supply unsigned types and some signed. So if anything, > this should be fixing all the callers to supply unsigned quantities.
mci->mc_idx is of type int and there is a cast here that should be fixed. IMO that should be a signed int as some interfaces (esp. if you search for an index) that require a negative value to report errors or something could not be found.
So let's take this patch out of this series if you want have it different.
Thanks,
-Robert
| |