memory: tegra: Fix integer overflow on tick value calculation
authorDmitry Osipenko <digetx@gmail.com>
Thu, 11 Apr 2019 22:12:48 +0000 (01:12 +0300)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Sat, 25 May 2019 16:23:32 +0000 (18:23 +0200)
commit b906c056b6023c390f18347169071193fda57dde upstream.

Multiplying the Memory Controller clock rate by the tick count results
in an integer overflow and in result the truncated tick value is being
programmed into hardware, such that the GR3D memory client performance is
reduced by two times.

Cc: stable <stable@vger.kernel.org>
Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/memory/tegra/mc.c

index bd25faf..c8f1666 100644 (file)
@@ -280,7 +280,7 @@ static int tegra_mc_setup_latency_allowance(struct tegra_mc *mc)
        u32 value;
 
        /* compute the number of MC clock cycles per tick */
-       tick = mc->tick * clk_get_rate(mc->clk);
+       tick = (unsigned long long)mc->tick * clk_get_rate(mc->clk);
        do_div(tick, NSEC_PER_SEC);
 
        value = readl(mc->regs + MC_EMEM_ARB_CFG);