Skip to contentSkip to conversation

iniparser: prevent signed integer underflow in line length calculation

PATCH In review Anton Moryakov 2 participants 2 messages Jan 27 – Apr 13

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 iniparser: prevent signed integer underflow in line length calculation

In iniparser_load_file(), the code computed:

    len = (int)strlen(line) - 1;

If the input line was empty (e.g. started with a null byte), strlen() returned 0, leading to len = -1. This caused:
- A signed integer underflow (detected by static analyzers)
- Potential out-of-bounds access when checking line[len]

Fix by:
- Using size_t for strlen() result
- Checking for zero length before subtraction
- Computing len as (int)(line_len - 1) only when safe

---
 lib/libiniparser.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)
lib/libiniparser.c +7 -1
index 4b21b34..1d6dbac 100644
@@ -697,7 +697,13 @@ dictionary * iniparser_load_file(FILE * in, const char * ininame)
 
     while (fgets(line+last, ASCIILINESZ-last, in)!=NULL) {
         lineno++ ;
-        len = (int)strlen(line)-1;
+        size_t line_len = strlen(line);
+                if (line_len == 0) {
+            memset(line, 0, ASCIILINESZ);
+            last = 0;
+            continue;
+        }
+        len = (int)(line_len - 1);
         if (len<=0)
             continue;
         /* Safety check against buffer overflows */
Signature · 6 lines
2.39.2


______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/
#1 lore raw

Hi,

according to your description, this was found by static analysis? I'm a little confused how it should be possible to trigger this bug:

Quoted 4 lines · Anton Moryakov
  
      while (fgets(line+last, ASCIILINESZ-last, in)!=NULL) {
          lineno++ ;

- len = (int)strlen(line)-1;

The return value of strlen is cast to an int, which AFAIK has higher precedence than the following subtraction of 1 (also, by default, an integer).

The variable len is declared as an int at the beginning of the function, it is signed and *can* be negative, there is no underflow here.

Directly after the code you inserted is this:

Quoted 2 lines
          if (len<=0)
              continue;

which should catch that exact case you are trying to fix, no?

Greetings,

David

______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/