Skip to contentSkip to conversation

fprobe: Add exit_handler support

bug report Dan Carpenter 2 participants 2 messages Mar 22, 2022 Kernel janitors

lore
Share & tools
lore
#0 lore raw

Hello Masami Hiramatsu,

The patch 5b0ab78998e3: "fprobe: Add exit_handler support" from Mar 15, 2022, leads to the following Smatch static checker warning:

kernel/trace/fprobe.c:155 fprobe_init_rethook()
warn: struct type mismatch 'rethook_node vs fprobe_rethook_node'

kernel/trace/fprobe.c

    134 static int fprobe_init_rethook(struct fprobe *fp, int num)
    135 {
    136         int i, size;
    137 
    138         if (num < 0)
    139                 return -EINVAL;
    140 
    141         if (!fp->exit_handler) {
    142                 fp->rethook = NULL;
    143                 return 0;
    144         }
    145 
    146         /* Initialize rethook if needed */
    147         size = num * num_possible_cpus() * 2;
    148         if (size < 0)
    149                 return -E2BIG;
    150 
    151         fp->rethook = rethook_alloc((void *)fp, fprobe_exit_handler);
    152         for (i = 0; i < size; i++) {
    153                 struct rethook_node *node;
    154 

--> 155 node = kzalloc(sizeof(struct fprobe_rethook_node), GFP_KERNEL);

It's hard to tell if we're deliberately allocating a larger struct or whether it's accidental.

    156                 if (!node) {
    157                         rethook_free(fp->rethook);
    158                         fp->rethook = NULL;
    159                         return -ENOMEM;
    160                 }
    161                 rethook_add_node(fp->rethook, node);
    162         }
    163         return 0;
    164 }

regards,
dan carpenter

#1 lore raw

Hi,

Quoted 30 lines · On Tue, 22 Mar 2022 09:46:20 +0300 Dan Carpenter <[email protected]>…

Hello Masami Hiramatsu,

The patch 5b0ab78998e3: "fprobe: Add exit_handler support" from Mar 15, 2022, leads to the following Smatch static checker warning:

kernel/trace/fprobe.c:155 fprobe_init_rethook()
warn: struct type mismatch 'rethook_node vs fprobe_rethook_node'

kernel/trace/fprobe.c

    134 static int fprobe_init_rethook(struct fprobe *fp, int num)
    135 {
    136         int i, size;
    137 
    138         if (num < 0)
    139                 return -EINVAL;
    140 
    141         if (!fp->exit_handler) {
    142                 fp->rethook = NULL;
    143                 return 0;
    144         }
    145 
    146         /* Initialize rethook if needed */
    147         size = num * num_possible_cpus() * 2;
    148         if (size < 0)
    149                 return -E2BIG;
    150 
    151         fp->rethook = rethook_alloc((void *)fp, fprobe_exit_handler);
    152         for (i = 0; i < size; i++) {
    153                 struct rethook_node *node;
    154 
Quoted 4 lines

--> 155 node = kzalloc(sizeof(struct fprobe_rethook_node), GFP_KERNEL);

It's hard to tell if we're deliberately allocating a larger struct or whether it's accidental.

Good catch! This does not cause any error, but it should be;

struct fprobe_rethook_node *node;

node = kzalloc(sizeof(*node), GFP_KERNEL);
...
rethook_add_node(fp->rethook, &node->node);

Thank you!

Quoted 13 lines
    156                 if (!node) {
    157                         rethook_free(fp->rethook);
    158                         fp->rethook = NULL;
    159                         return -ENOMEM;
    160                 }
    161                 rethook_add_node(fp->rethook, node);
    162         }
    163         return 0;
    164 }

regards,
dan carpenter

Signature · 1 line
Masami Hiramatsu <[email protected]>