2019-01-27 05:11:27

by Yizhuo Zhai

Subject: [PATCH] clk: gemini: Variable "val" in function gemini_clk_probe() could be uninitialized

In function gemini_clk_probe(), local variable "val" could
be uninitialized if function regmap_read() returns -EINVAL.
However, it will be used as index in the later context, which
could potentially be unsafe.

Signed-off-by: Yizhuo <[email protected]>
---
drivers/clk/clk-gemini.c | 10 ++++++++--
1 file changed, 8 insertions(+), 2 deletions(-)

diff --git a/drivers/clk/clk-gemini.c b/drivers/clk/clk-gemini.c
index 5e66e6c0205e..07e1b551d1d9 100644
--- a/drivers/clk/clk-gemini.c
+++ b/drivers/clk/clk-gemini.c
@@ -314,7 +314,10 @@ static int gemini_clk_probe(struct platform_device *pdev)
gemini_clk_data->hws[GEMINI_CLK_RTC] = hw;

/* CPU clock derived as a fixed ratio from the AHB clock */
- regmap_read(map, GEMINI_GLOBAL_STATUS, &val);
+ ret = regmap_read(map, GEMINI_GLOBAL_STATUS, &val);
+ if (ret)
+ return ret;
+
val >>= CPU_AHB_RATIO_SHIFT;
val &= CPU_AHB_RATIO_MASK;
hw = clk_hw_register_fixed_factor(NULL, "cpu", "ahb", 0,
@@ -323,7 +326,10 @@ static int gemini_clk_probe(struct platform_device *pdev)
gemini_clk_data->hws[GEMINI_CLK_CPU] = hw;

/* Security clock is 1:1 or 0.75 of APB */
- regmap_read(map, GEMINI_GLOBAL_CLOCK_CONTROL, &val);
+ ret = regmap_read(map, GEMINI_GLOBAL_CLOCK_CONTROL, &val);
+ if (ret)
+ return ret;
+
if (val & SECURITY_CLK_SEL) {
mult = 1;
div = 1;
--
2.17.1

2019-02-06 19:24:42

by Stephen Boyd

Subject: Re: [PATCH] clk: gemini: Variable "val" in function gemini_clk_probe() could be uninitialized

Quoting Yizhuo (2019-01-26 21:10:12)
> In function gemini_clk_probe(), local variable "val" could
> be uninitialized if function regmap_read() returns -EINVAL.
> However, it will be used as index in the later context, which
> could potentially be unsafe.

Ok. How did you find this? Any pointers?

>
> Signed-off-by: Yizhuo <[email protected]>
> ---
> drivers/clk/clk-gemini.c | 10 ++++++++--
> 1 file changed, 8 insertions(+), 2 deletions(-)

There are other locations in this file where 'val' is used even if
regmap_read() fails. For example gemini_pci_recalc_rate() does this. Can
you fix all the callers in this file? Presumably nobody cares that this
API could fail in this driver because it's a thin wrapper around mmio
read that never fails. Maybe we could have a comment instead that this
is the case and then ignore this patch entirely.