The calculated ideal rate can easily overflow an unsigned long, thus making the best div selection buggy as soon as no ideal match is found before the overflow occurs.
Fixes: 4731a72df273 ("drm/sun4i: request exact rates to our parents") Cc: stable@vger.kernel.org Signed-off-by: Boris Brezillon boris.brezillon@bootlin.com Acked-by: Maxime Ripard maxime.ripard@bootlin.com --- Changes in v2: - Add a comment to explain why we bail out after an overflow - Add Maxime ack - Use a goto instead of a break --- drivers/gpu/drm/sun4i/sun4i_dotclock.c | 12 +++++++++++- 1 file changed, 11 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/sun4i/sun4i_dotclock.c b/drivers/gpu/drm/sun4i/sun4i_dotclock.c index e36004fbe453..2a15f2f9271e 100644 --- a/drivers/gpu/drm/sun4i/sun4i_dotclock.c +++ b/drivers/gpu/drm/sun4i/sun4i_dotclock.c @@ -81,9 +81,19 @@ static long sun4i_dclk_round_rate(struct clk_hw *hw, unsigned long rate, int i;
for (i = tcon->dclk_min_div; i <= tcon->dclk_max_div; i++) { - unsigned long ideal = rate * i; + u64 ideal = (u64)rate * i; unsigned long rounded;
+ /* + * ideal has overflowed the max value that can be stored in an + * unsigned long, and every clk operation we might do on a + * truncated u64 value will give us incorrect results. + * Let's just stop there since bigger dividers will result in + * the same overflow issue. + */ + if (ideal > ULONG_MAX) + goto out; + rounded = clk_hw_round_rate(clk_hw_get_parent(hw), ideal);