linux-kernel - Re: [PATCH] x86/lib: Fix num_digits() signed overflow for INT

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <6d28349e-9912-474d-a750-71488e2fc976@formalgen.com>
Date: Tue, 20 Jan 2026 18:48:54 +0100
From: David Desobry <david.desobry@...malgen.com>
To: "H. Peter Anvin" <hpa@...or.com>, tglx@...nel.org, mingo@...hat.com,
 bp@...en8.de, dave.hansen@...ux.intel.com
Cc: x86@...nel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] x86/lib: Fix num_digits() signed overflow for INT_MIN

Fair point! I've sent a v2 that replaces the loop with a switch 
statement (using GCC ranges). It's faster, handles INT_MIN properly via 
an unsigned cast, and I've cleaned up that mobile-submission comment 
while I was at it.

Le 20/01/2026 à 17:23, H. Peter Anvin a écrit :
> On January 20, 2026 1:42:58 AM PST, David Desobry <david.desobry@...malgen.com> wrote:
>> In num_digits(), the negation of the input value "val = -val"
>> causes undefined behavior when val is INT_MIN, as its absolute
>> value cannot be represented as a signed 32-bit integer.
>>
>> This leads to incorrect results (returning 2 instead of 11).
>> By promoting the value to long long before negation, we ensure
>> the absolute value is correctly handled.
>>
>> Signed-off-by: David Desobry <david.desobry@...malgen.com>
>> ---
>> arch/x86/lib/misc.c | 7 ++++---
>> 1 file changed, 4 insertions(+), 3 deletions(-)
>>
>> diff --git a/arch/x86/lib/misc.c b/arch/x86/lib/misc.c
>> index 40b81c338ae5..c975db6ccb9f 100644
>> --- a/arch/x86/lib/misc.c
>> +++ b/arch/x86/lib/misc.c
>> @@ -8,15 +8,16 @@
>>   */
>> int num_digits(int val)
>> {
>> +	long long v = val;
>> 	long long m = 10;
>> 	int d = 1;
>>
>> -	if (val < 0) {
>> +	if (v < 0) {
>> 		d++;
>> -		val = -val;
>> +		v = -v;
>> 	}
>>
>> -	while (val >= m) {
>> +	while (v >= m) {
>> 		m *= 10;
>> 		d++;
>> 	}
> That has got to be the dumbest possible implementation of that task, bug or no bug.
>
> A switch statement would be simpler and faster.