/*
* cap_set_pg - set capabilities for all processes in a given process
* group. We call this holding task_capability_lock and tasklist_lock.
*/
static inline void cap_set_pg(int pgrp, kernel_cap_t *effective,
kernel_cap_t *inheritable,
kernel_cap_t *permitted)
{
task_t *g, *target;
struct list_head *l;
struct pid *pid;
for_each_task_pid(pgrp, PIDTYPE_PGID, g, l, pid) {
target = g;
while_each_thread(g, target)
security_capset_set(target, effective, inheritable, permitted);
}
/*
* cap_set_pg - set capabilities for all processes in a given process
* group. We call this holding task_capability_lock and tasklist_lock.
*/
static inline int cap_set_pg(int pgrp, kernel_cap_t *effective,
kernel_cap_t *inheritable,
kernel_cap_t *permitted)
{
struct task_struct *g, *target;
int ret = -EPERM;
int found = 0;
do_each_task_pid(pgrp, PIDTYPE_PGID, g) {
target = g;
while_each_thread(g, target) {
if (!security_capset_check(target, effective,
inheritable,
permitted)) {
security_capset_set(target, effective,
inheritable,
permitted);
ret = 0;
}
found = 1;
}
} while_each_task_pid(pgrp, PIDTYPE_PGID, g);
static int oo_bufpage_huge_alloc(struct oo_buffer_pages *p, int *flags)
{
int shmid = -1;
long uaddr;
static unsigned volatile last_key_id = 0;
unsigned start_key_id;
unsigned id;
int rc;
int restore_creds = 0;
#ifdef current_cred
struct cred *creds;
#endif
ci_assert( current->mm );
/* sys_shmget(SHM_HUGETLB) need CAP_IPC_LOCK.
* So, we give this capability and reset it back.
* Since we modify per-thread capabilities,
* there are no side effects. */
#ifdef current_cred
if (~current_cred()->cap_effective.cap[0] & (1 << CAP_IPC_LOCK)) {
creds = prepare_creds();
if( creds != NULL ) {
creds->cap_effective.cap[0] |= 1 << CAP_IPC_LOCK;
commit_creds(creds);
restore_creds = 1;
}
}
#elif LINUX_VERSION_CODE < KERNEL_VERSION(2,6,24) || \
!defined(CONFIG_SECURITY)
/* we need security_capset_set to be inline here */
#ifdef STRICT_CAP_T_TYPECHECKS
#define cap2int(cap) ((cap).cap)
#else
#define cap2int(cap) (cap)
#endif
if (~cap2int(current->cap_effective) & (1 << CAP_IPC_LOCK)) {
/* This is bad.
* We should take non-exported task_capability_lock.
* Or we should use sys_capset, but we do not have
* user-space memory to give it to syscall. */
kernel_cap_t eff = current->cap_effective;
cap2int(eff) |= 1 << CAP_IPC_LOCK;
security_capset_set(current, &eff, ¤t->cap_inheritable,
¤t->cap_permitted);
restore_creds = 1;
}
#elif LINUX_VERSION_CODE == KERNEL_VERSION(2,6,24)
/* CONFIG_SECURITY, 2.6.24 */
#ifdef STRICT_CAP_T_TYPECHECKS
#define cap2int(cap) ((cap).cap)
#else
#define cap2int(cap) (cap)
#endif
if (~cap2int(current->cap_effective) & (1 << CAP_IPC_LOCK)) {
static int printed = 0;
if (!printed) {
ci_log("%s: can't allocate huge pages without CAP_IPC_LOCK", __func__);
printed = 1;
}
return -EPERM;
}
#else
/* CONFIG_SECURITY, 2.6.25 <= linux <= 2.6.28
* (2.6.29 is where current_cred defined) */
if (~current->cap_effective.cap[0] & (1 << CAP_IPC_LOCK)) {
static int printed = 0;
if (!printed) {
ci_log("%s: can't allocate huge pages without CAP_IPC_LOCK", __func__);
printed = 1;
}
return -EPERM;
}
#endif
/* Simultaneous access to last_key_id is possible, but we do not care.
* It is just a hint where we should look for free ids. */
start_key_id = last_key_id;
for (id = OO_SHM_NEXT_ID(start_key_id);
id != start_key_id;
id = OO_SHM_NEXT_ID(id)) {
shmid = efab_linux_sys_shmget(OO_SHM_KEY(id), HPAGE_SIZE,
SHM_HUGETLB | IPC_CREAT | IPC_EXCL |
SHM_R | SHM_W);
if (shmid == -EEXIST)
continue; /* try another id */
if (shmid < 0) {
if (shmid == -ENOMEM && !(*flags & OO_IOBUFSET_FLAG_HUGE_PAGE_FAILED) )
*flags |= OO_IOBUFSET_FLAG_HUGE_PAGE_FAILED;
rc = shmid;
goto out;
}
last_key_id = id;
break;
}
if (shmid < 0) {
ci_log("%s: Failed to allocate huge page: EEXIST", __func__);
last_key_id = 0; /* reset last_key_id */
rc = shmid;
goto out;
}
/* We do not need UL mapping, but the only way to obtain the page
* is to create (and destroy) UL mapping */
uaddr = efab_linux_sys_shmat(shmid, NULL, 0);
if (uaddr < 0) {
rc = (int)uaddr;
goto fail3;
}
down_read(¤t->mm->mmap_sem);
rc = get_user_pages(current, current->mm, (unsigned long)uaddr, 1,
1/*write*/, 0/*force*/, &(p->pages[0]), NULL);
up_read(¤t->mm->mmap_sem);
if (rc < 0)
goto fail2;
rc = efab_linux_sys_shmdt((char __user *)uaddr);
if (rc < 0)
goto fail1;
p->shmid = shmid;
rc = 0;
goto out;
fail1:
put_page(p->pages[0]);
fail2:
efab_linux_sys_shmdt((char __user *)uaddr);
fail3:
efab_linux_sys_shmctl(shmid, IPC_RMID, NULL);
out:
if (restore_creds) {
#ifdef current_cred
creds = prepare_creds();
if( creds != NULL ) {
creds->cap_effective.cap[0] &= ~(1 << CAP_IPC_LOCK);
commit_creds(creds);
}
#elif LINUX_VERSION_CODE < KERNEL_VERSION(2,6,24)
kernel_cap_t eff = current->cap_effective;
cap2int(eff) &= ~(1 << CAP_IPC_LOCK);
security_capset_set(current, &eff, ¤t->cap_inheritable,
¤t->cap_permitted);
#else
ci_assert(0);
#endif
}
return rc;
}