[lvc-project] [PATCH] mmc: dw_mmc: Fix potential null pointer risk

[Ulf Hansson]

On Thu, 7 Mar 2024 at 20:07, Aleksandr Mishin <amishin at t-argos.ru> wrote:
>
>
>
> 07.03.2024 13:57, Ulf Hansson wrote:
> > On Thu, 7 Mar 2024 at 09:53, Aleksandr Mishin <amishin at t-argos.ru> wrote:
> >>
> >> In dw_mci_runtime_resume() 'host->slot' could be null, but check is not cover all corresponding code.
> >> Fix this bug by changing check place.
> >
> > In fact host->slot can never be NULL in dw_mci_runtime_resume() or in
> > dw_mci_runtime_suspend().
> >
> > A better fix would thus be to remove the redundant checks.
> >
> > Kind regards
> > Uffe
> >
> >>
> >> Found by Linux Verification Center (linuxtesting.org) with SVACE.
> >>
> >> Fixes: 4a835afd808a (mmc: dw_mmc: Fix potential null pointer risk)
> >> Signed-off-by: Aleksandr Mishin <amishin at t-argos.ru>
> >> ---
> >>   drivers/mmc/host/dw_mmc.c | 4 +++-
> >>   1 file changed, 3 insertions(+), 1 deletion(-)
> >>
> >> diff --git a/drivers/mmc/host/dw_mmc.c b/drivers/mmc/host/dw_mmc.c
> >> index 829af2c98a44..a4f124452abc 100644
> >> --- a/drivers/mmc/host/dw_mmc.c
> >> +++ b/drivers/mmc/host/dw_mmc.c
> >> @@ -3570,8 +3570,10 @@ int dw_mci_runtime_resume(struct device *dev)
> >>                     DW_MCI_ERROR_FLAGS);
> >>          mci_writel(host, CTRL, SDMMC_CTRL_INT_ENABLE);
> >>
> >> +       if (!host->slot)
> >> +               goto err;
> >>
> >> -       if (host->slot && host->slot->mmc->pm_flags & MMC_PM_KEEP_POWER)
> >> +       if (host->slot->mmc->pm_flags & MMC_PM_KEEP_POWER)
> >>                  dw_mci_set_ios(host->slot->mmc, &host->slot->mmc->ios);
> >>
> >>          /* Force setup bus to guarantee available clock output */
> >> --
> >> 2.30.2
> >>
> >>
> >
>
> At the same time there are few checks such as "if (host->slot)" in
> dw_mci_runtime_resume() and commit
> 4a835afd808a3dbbac44bb399a902b822dc7445c message contains: "we
> previously assumed 'host->slot' could be null, null pointer judgment
> should be added" and replaces "if (host->slot->mmc->pm_flags &
> MMC_PM_KEEP_POWER)" with "if (host->slot && host->slot->mmc->pm_flags &
> MMC_PM_KEEP_POWER)"
> So where is the truth?

It looks to me that the runtime PM callbacks are prevented from being
called, unless we have a host->slot assigned.

Just adding checks because it looks like the code could need it, isn't
always the correct thing to do. I would rather try to remove the
checks altogether and give it some tests to see how it plays.

Kind regards
Uffe