Smatch complains that 'str' can be used without being initialized:

    drivers/platform/x86/intel/intel_plr_tpmi.c:178 plr_print_bits()
    error: uninitialized symbol 'str'.

In this loop, we iterate over all the set bits and print the name of the
bit.  The intention is that if there is a bit which is between 0-31 we
look for the name in the first array plr_coarse_reasons[] which has 10
elements.  If the bit is in the 32-63 range we look for it in the
plr_fine_reasons[] array which has 30 elements.  If the bit is in the
invalid ranges, 10-31 or 62-63, then we should print "UNKNOWN(%d)".

The problem is that 'str' needs to be initialized at the start of each
iteration, otherwise if we can't find the string then instead of printing
"UNKNOWN(%d)", we will re-print whatever the previous bit was.

Fixes: 9e9397a41b7b ("platform/x86/intel/tpmi/plr: Add support for the plr mailbox")
Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
---
v2: initialize str at the start of each iteration

 drivers/platform/x86/intel/intel_plr_tpmi.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/platform/x86/intel/intel_plr_tpmi.c b/drivers/platform/x86/intel/intel_plr_tpmi.c
index c1aa52c23d25..69ace6a629bc 100644
--- a/drivers/platform/x86/intel/intel_plr_tpmi.c
+++ b/drivers/platform/x86/intel/intel_plr_tpmi.c
@@ -162,10 +162,11 @@ static int plr_clear_cpu_status(struct tpmi_plr_die *plr_die, int cpu)
 static void plr_print_bits(struct seq_file *s, u64 val, int bits)
 {
 	const unsigned long mask[] = { BITMAP_FROM_U64(val) };
-	const char *str;
 	int bit, index;
 
 	for_each_set_bit(bit, mask, bits) {
+		const char *str = NULL;
+
 		if (bit < PLR_COARSE_REASON_BITS) {
 			if (bit < ARRAY_SIZE(plr_coarse_reasons))
 				str = plr_coarse_reasons[bit];
-- 
2.43.0

On Mon, 15 Jul 2024, Dan Carpenter wrote:

> Smatch complains that 'str' can be used without being initialized:
> 
>     drivers/platform/x86/intel/intel_plr_tpmi.c:178 plr_print_bits()
>     error: uninitialized symbol 'str'.
> 
> In this loop, we iterate over all the set bits and print the name of the
> bit.  The intention is that if there is a bit which is between 0-31 we
> look for the name in the first array plr_coarse_reasons[] which has 10
> elements.  If the bit is in the 32-63 range we look for it in the
> plr_fine_reasons[] array which has 30 elements.  If the bit is in the
> invalid ranges, 10-31 or 62-63, then we should print "UNKNOWN(%d)".
> 
> The problem is that 'str' needs to be initialized at the start of each
> iteration, otherwise if we can't find the string then instead of printing
> "UNKNOWN(%d)", we will re-print whatever the previous bit was.
> 
> Fixes: 9e9397a41b7b ("platform/x86/intel/tpmi/plr: Add support for the plr mailbox")
> Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
> ---
> v2: initialize str at the start of each iteration

Thanks for the update, I've applied this fix now (will go along with the 
for-next PR to Linus).

-- 
 i.

> 
>  drivers/platform/x86/intel/intel_plr_tpmi.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/platform/x86/intel/intel_plr_tpmi.c b/drivers/platform/x86/intel/intel_plr_tpmi.c
> index c1aa52c23d25..69ace6a629bc 100644
> --- a/drivers/platform/x86/intel/intel_plr_tpmi.c
> +++ b/drivers/platform/x86/intel/intel_plr_tpmi.c
> @@ -162,10 +162,11 @@ static int plr_clear_cpu_status(struct tpmi_plr_die *plr_die, int cpu)
>  static void plr_print_bits(struct seq_file *s, u64 val, int bits)
>  {
>  	const unsigned long mask[] = { BITMAP_FROM_U64(val) };
> -	const char *str;
>  	int bit, index;
>  
>  	for_each_set_bit(bit, mask, bits) {
> +		const char *str = NULL;
> +
>  		if (bit < PLR_COARSE_REASON_BITS) {
>  			if (bit < ARRAY_SIZE(plr_coarse_reasons))
>  				str = plr_coarse_reasons[bit];
>