[opencryptoki-users] pkcsslotd crashes on Linux 2.6.10 uclibc

[King G.G. <kin...@gm...>]

Hi all
 here is the description of problems I have:
 System configuration:
Linux 2.6.10 uclibc, BusyBox v1.00-rc3,
openCryptoki 2.1.5-6 built with i386 uclibc gcc 3.3.4

Any application which calls C_Initialize and stays in the memory for a few
seconds causes
pkcsslotd to 'disappear' from memory without freeing shared memory. Most
likely pkcsslotd
crashes and shared memory stays until next reboot since despite pkcsslotd i=
s
no longer
running any call to C_XXX functions succeed and applications work fine.
After spending days
I found that even there is no need to call C_XXX functions from
PKCS11_API.so. To make pkcsslotd
crash it's enough to attach to shared memory and write into 2 fields:
Slot_Mgr_Proc_t::inuse and
Slot_Mgr_Proc_t::proc_id, and stay in memory for a few seconds:

key_t tok;
int shmid =3D 0;
void* shmp =3D NULL;
Slot_Mgr_Shr_t *shm;
Slot_Mgr_Proc_t *procp;

tok =3D ftok("/usr/sbin/pkcsslotd",'b');
shmid =3D shmget(tok, sizeof(Slot_Mgr_Shr_t),S_IWUSR|S_IWGRP|S_IRGRP|S_IRUS=
R);
printf("shmid =3D %d\n",shmid);
shmp =3D shmat(shmid,NULL,0);
shm =3D shmp;
procp =3D shm->proc_table;
procp->inuse =3D TRUE;
procp->proc_id =3D getpid();
while(1);

Couple of days later I found the following code in garbage_linux.c, line
846:

memset(&p->cmd, 0, sizeof(p->cmd));
sscanf(buf, "%d (%15c", &p->pid, p->cmd); // comm[16] in kernel

where cmd is an unallocated char*. After changing definition of cmd in
proc_t structure
from char* to char[16] this issue was solved.
To me it seems this is the only place where cmd is used. But since pkcsslot=
d
is very
important part I'm afraid of breaking something in it's functionality by
this change. Could you
please tell me is this a bug or this is how it supposed to be? What is the
meaning of that memset?
Is it supposed to zeroise string pointed by cmd or make cmd NULL-pointer? I=
f
zeroise, how many bytes?

Are my changes correct or not? May they affect some other parts of
openCryptoki or somehow cause
malfunctioning?
Or if that cmd is not used at all then may be it's better to skip it?
 Gev