Re: [PATCH net v2 0/3] fix xa_alloc_cyclic() return checks

[Dan Carpenter <dan.carpenter@linaro.org>]

On Fri, Mar 14, 2025 at 01:52:58PM +0100, Przemek Kitszel wrote:
> On 3/14/25 11:23, Dan Carpenter wrote:
> > On Wed, Mar 12, 2025 at 10:52:48AM +0100, Michal Swiatkowski wrote:
> > > Pierre Riteau <pierre@stackhpc.com> found suspicious handling an error
> > > from xa_alloc_cyclic() in scheduler code [1]. The same is done in few
> > > other places.
> > > 
> > > v1 --> v2: [2]
> > >   * add fixes tags
> > >   * fix also the same usage in dpll and phy
> > > 
> > > [1] https://lore.kernel.org/netdev/20250213223610.320278-1-pierre@stackhpc.com/
> > > [2] https://lore.kernel.org/netdev/20250214132453.4108-1-michal.swiatkowski@linux.intel.com/
> > > 
> > > Michal Swiatkowski (3):
> > >    devlink: fix xa_alloc_cyclic() error handling
> > >    dpll: fix xa_alloc_cyclic() error handling
> > >    phy: fix xa_alloc_cyclic() error handling
> > 
> > Maybe there should be a wrapper around xa_alloc_cyclic() for people who
> > don't care about the 1 return?
> 
> What about changing init flags instead, and add a new one for this
> purpose?, say:
> XA_FLAGS_ALLOC_RET0

Right now I have a static checker rule for passing 1 to ERR_PTR().
It's not specific to this function but it catches the bugs here.  If we
added a XA_FLAGS_ALLOC_RET0 then I'd have to silence the checker rule for
xa_alloc_cyclic().

I was also thinking about creating another more specific rule for just this
function to warn about when callers which treat 1 and negative error
codes the same, but that wouldn't be possible.

On the other hand, people who pass XA_FLAGS_ALLOC_RET0 probably will
understand what it means and not introduce bugs so static analysis becomes
less important in that case.

regards,
dan carpenter