[PATCH v1 1/1] lib/test_scanf: Handle n_bits == 0 in random tests

[PATCH v1 1/1] lib/test_scanf: Handle n_bits == 0 in random tests

[Andy Shevchenko]

UBSAN reported (via LKP)

[   11.021349][    T1] UBSAN: shift-out-of-bounds in lib/test_scanf.c:275:51
[   11.022782][    T1] shift exponent 32 is too large for 32-bit type 'unsigned int'

When n_bits == 0, the shift is out of range. Switch code to use GENMASK
to handle this case.

Fixes: 50f530e176ea ("lib: test_scanf: Add tests for sscanf number conversion")
Reported-by: kernel test robot <oliver.sang@intel.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
---
 lib/test_scanf.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/lib/test_scanf.c b/lib/test_scanf.c
index 84fe09eaf55e..abae88848972 100644
--- a/lib/test_scanf.c
+++ b/lib/test_scanf.c
@@ -271,7 +271,7 @@ static u32 __init next_test_random(u32 max_bits)
 {
 	u32 n_bits = hweight32(prandom_u32_state(&rnd_state)) % (max_bits + 1);
 
-	return prandom_u32_state(&rnd_state) & (UINT_MAX >> (32 - n_bits));
+	return prandom_u32_state(&rnd_state) & GENMASK(n_bits, 0);
 }
 
 static unsigned long long __init next_test_random_ull(void)
@@ -280,7 +280,7 @@ static unsigned long long __init next_test_random_ull(void)
 	u32 n_bits = (hweight32(rand1) * 3) % 64;
 	u64 val = (u64)prandom_u32_state(&rnd_state) * rand1;
 
-	return val & (ULLONG_MAX >> (64 - n_bits));
+	return val & GENMASK_ULL(n_bits, 0);
 }
 
 #define random_for_type(T)				\
-- 
2.30.2

Re: [PATCH v1 1/1] lib/test_scanf: Handle n_bits == 0 in random tests

[Richard Fitzgerald]

On 27/07/2021 16:01, Andy Shevchenko wrote:
> UBSAN reported (via LKP)
> 
> [   11.021349][    T1] UBSAN: shift-out-of-bounds in lib/test_scanf.c:275:51
> [   11.022782][    T1] shift exponent 32 is too large for 32-bit type 'unsigned int'
> 
> When n_bits == 0, the shift is out of range. Switch code to use GENMASK
> to handle this case.
> 
> Fixes: 50f530e176ea ("lib: test_scanf: Add tests for sscanf number conversion")
> Reported-by: kernel test robot <oliver.sang@intel.com>
> Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
> ---

Reviewed-by: Richard Fitzgerald <rf@opensource.cirrus.com>

Sorry about that. Would have been sensible to use GENMASK anyway.

Re: [PATCH v1 1/1] lib/test_scanf: Handle n_bits == 0 in random tests

[Petr Mladek]

On Tue 2021-07-27 18:01:32, Andy Shevchenko wrote:
> UBSAN reported (via LKP)
> 
> [   11.021349][    T1] UBSAN: shift-out-of-bounds in lib/test_scanf.c:275:51
> [   11.022782][    T1] shift exponent 32 is too large for 32-bit type 'unsigned int'
> 
> When n_bits == 0, the shift is out of range. Switch code to use GENMASK
> to handle this case.
> 
> Fixes: 50f530e176ea ("lib: test_scanf: Add tests for sscanf number conversion")
> Reported-by: kernel test robot <oliver.sang@intel.com>
> Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>

The patch is committed in linux/printk.git, branch for-5.15.

I would send it for 4.14 if there was another urgent fix needed.
But this one does not look important enough to hurry it up alone.

Best Regards,
Petr