ipc/sem.c: remove sem_base, embed struct sem
authorManfred Spraul <manfred@colorfullife.com>
Wed, 12 Jul 2017 21:34:38 +0000 (14:34 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Wed, 12 Jul 2017 23:26:01 +0000 (16:26 -0700)
sma->sem_base is initialized with

sma->sem_base = (struct sem *) &sma[1];

The current code has four problems:
 - There is an unnecessary pointer dereference - sem_base is not needed.
 - Alignment for struct sem only works by chance.
 - The current code causes false positive for static code analysis.
 - This is a cast between different non-void types, which the future
   randstruct GCC plugin warns on.

And, as bonus, the code size gets smaller:

  Before:
    0 .text         00003770
  After:
    0 .text         0000374e

[manfred@colorfullife.com: s/[0]/[]/, per hch]
Link: http://lkml.kernel.org/r/20170525185107.12869-2-manfred@colorfullife.com
Link: http://lkml.kernel.org/r/20170515171912.6298-2-manfred@colorfullife.com
Signed-off-by: Manfred Spraul <manfred@colorfullife.com>
Acked-by: Kees Cook <keescook@chromium.org>
Cc: Kees Cook <keescook@chromium.org>
Cc: <1vier1@web.de>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Fabian Frederick <fabf@skynet.be>
Cc: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
include/linux/sem.h
ipc/sem.c

index 9edec92..9db1409 100644 (file)
@@ -8,11 +8,29 @@
 
 struct task_struct;
 
+/* One semaphore structure for each semaphore in the system. */
+struct sem {
+       int     semval;         /* current value */
+       /*
+        * PID of the process that last modified the semaphore. For
+        * Linux, specifically these are:
+        *  - semop
+        *  - semctl, via SETVAL and SETALL.
+        *  - at task exit when performing undo adjustments (see exit_sem).
+        */
+       int     sempid;
+       spinlock_t      lock;   /* spinlock for fine-grained semtimedop */
+       struct list_head pending_alter; /* pending single-sop operations */
+                                       /* that alter the semaphore */
+       struct list_head pending_const; /* pending single-sop operations */
+                                       /* that do not alter the semaphore*/
+       time_t  sem_otime;      /* candidate for sem_otime */
+} ____cacheline_aligned_in_smp;
+
 /* One sem_array data structure for each set of semaphores in the system. */
 struct sem_array {
        struct kern_ipc_perm    sem_perm;       /* permissions .. see ipc.h */
        time_t                  sem_ctime;      /* last change time */
-       struct sem              *sem_base;      /* ptr to first semaphore in array */
        struct list_head        pending_alter;  /* pending operations */
                                                /* that alter the array */
        struct list_head        pending_const;  /* pending complex operations */
@@ -21,6 +39,8 @@ struct sem_array {
        int                     sem_nsems;      /* no. of semaphores in array */
        int                     complex_count;  /* pending complex operations */
        unsigned int            use_global_lock;/* >0: global lock required */
+
+       struct sem              sems[];
 };
 
 #ifdef CONFIG_SYSVIPC
index 947dc23..fff8337 100644 (file)
--- a/ipc/sem.c
+++ b/ipc/sem.c
 #include <linux/uaccess.h>
 #include "util.h"
 
-/* One semaphore structure for each semaphore in the system. */
-struct sem {
-       int     semval;         /* current value */
-       /*
-        * PID of the process that last modified the semaphore. For
-        * Linux, specifically these are:
-        *  - semop
-        *  - semctl, via SETVAL and SETALL.
-        *  - at task exit when performing undo adjustments (see exit_sem).
-        */
-       int     sempid;
-       spinlock_t      lock;   /* spinlock for fine-grained semtimedop */
-       struct list_head pending_alter; /* pending single-sop operations */
-                                       /* that alter the semaphore */
-       struct list_head pending_const; /* pending single-sop operations */
-                                       /* that do not alter the semaphore*/
-       time_t  sem_otime;      /* candidate for sem_otime */
-} ____cacheline_aligned_in_smp;
 
 /* One queue for each sleeping process in the system. */
 struct sem_queue {
@@ -175,7 +157,7 @@ static int sysvipc_sem_proc_show(struct seq_file *s, void *it);
  *     sem_array.sem_undo
  *
  * b) global or semaphore sem_lock() for read/write:
- *     sem_array.sem_base[i].pending_{const,alter}:
+ *     sem_array.sems[i].pending_{const,alter}:
  *
  * c) special:
  *     sem_undo_list.list_proc:
@@ -250,7 +232,7 @@ static void unmerge_queues(struct sem_array *sma)
         */
        list_for_each_entry_safe(q, tq, &sma->pending_alter, list) {
                struct sem *curr;
-               curr = &sma->sem_base[q->sops[0].sem_num];
+               curr = &sma->sems[q->sops[0].sem_num];
 
                list_add_tail(&q->list, &curr->pending_alter);
        }
@@ -270,7 +252,7 @@ static void merge_queues(struct sem_array *sma)
 {
        int i;
        for (i = 0; i < sma->sem_nsems; i++) {
-               struct sem *sem = sma->sem_base + i;
+               struct sem *sem = &sma->sems[i];
 
                list_splice_init(&sem->pending_alter, &sma->pending_alter);
        }
@@ -306,7 +288,7 @@ static void complexmode_enter(struct sem_array *sma)
        sma->use_global_lock = USE_GLOBAL_LOCK_HYSTERESIS;
 
        for (i = 0; i < sma->sem_nsems; i++) {
-               sem = sma->sem_base + i;
+               sem = &sma->sems[i];
                spin_lock(&sem->lock);
                spin_unlock(&sem->lock);
        }
@@ -366,7 +348,7 @@ static inline int sem_lock(struct sem_array *sma, struct sembuf *sops,
         *
         * Both facts are tracked by use_global_mode.
         */
-       sem = sma->sem_base + sops->sem_num;
+       sem = &sma->sems[sops->sem_num];
 
        /*
         * Initial check for use_global_lock. Just an optimization,
@@ -421,7 +403,7 @@ static inline void sem_unlock(struct sem_array *sma, int locknum)
                complexmode_tryleave(sma);
                ipc_unlock_object(&sma->sem_perm);
        } else {
-               struct sem *sem = sma->sem_base + locknum;
+               struct sem *sem = &sma->sems[locknum];
                spin_unlock(&sem->lock);
        }
 }
@@ -487,7 +469,7 @@ static int newary(struct ipc_namespace *ns, struct ipc_params *params)
        if (ns->used_sems + nsems > ns->sc_semmns)
                return -ENOSPC;
 
-       size = sizeof(*sma) + nsems * sizeof(struct sem);
+       size = sizeof(*sma) + nsems * sizeof(sma->sems[0]);
        sma = ipc_rcu_alloc(size);
        if (!sma)
                return -ENOMEM;
@@ -504,12 +486,10 @@ static int newary(struct ipc_namespace *ns, struct ipc_params *params)
                return retval;
        }
 
-       sma->sem_base = (struct sem *) &sma[1];
-
        for (i = 0; i < nsems; i++) {
-               INIT_LIST_HEAD(&sma->sem_base[i].pending_alter);
-               INIT_LIST_HEAD(&sma->sem_base[i].pending_const);
-               spin_lock_init(&sma->sem_base[i].lock);
+               INIT_LIST_HEAD(&sma->sems[i].pending_alter);
+               INIT_LIST_HEAD(&sma->sems[i].pending_const);
+               spin_lock_init(&sma->sems[i].lock);
        }
 
        sma->complex_count = 0;
@@ -612,7 +592,7 @@ static int perform_atomic_semop_slow(struct sem_array *sma, struct sem_queue *q)
        un = q->undo;
 
        for (sop = sops; sop < sops + nsops; sop++) {
-               curr = sma->sem_base + sop->sem_num;
+               curr = &sma->sems[sop->sem_num];
                sem_op = sop->sem_op;
                result = curr->semval;
 
@@ -639,7 +619,7 @@ static int perform_atomic_semop_slow(struct sem_array *sma, struct sem_queue *q)
        sop--;
        pid = q->pid;
        while (sop >= sops) {
-               sma->sem_base[sop->sem_num].sempid = pid;
+               sma->sems[sop->sem_num].sempid = pid;
                sop--;
        }
 
@@ -661,7 +641,7 @@ undo:
        sop--;
        while (sop >= sops) {
                sem_op = sop->sem_op;
-               sma->sem_base[sop->sem_num].semval -= sem_op;
+               sma->sems[sop->sem_num].semval -= sem_op;
                if (sop->sem_flg & SEM_UNDO)
                        un->semadj[sop->sem_num] += sem_op;
                sop--;
@@ -692,7 +672,7 @@ static int perform_atomic_semop(struct sem_array *sma, struct sem_queue *q)
         * until the operations can go through.
         */
        for (sop = sops; sop < sops + nsops; sop++) {
-               curr = sma->sem_base + sop->sem_num;
+               curr = &sma->sems[sop->sem_num];
                sem_op = sop->sem_op;
                result = curr->semval;
 
@@ -716,7 +696,7 @@ static int perform_atomic_semop(struct sem_array *sma, struct sem_queue *q)
        }
 
        for (sop = sops; sop < sops + nsops; sop++) {
-               curr = sma->sem_base + sop->sem_num;
+               curr = &sma->sems[sop->sem_num];
                sem_op = sop->sem_op;
                result = curr->semval;
 
@@ -815,7 +795,7 @@ static int wake_const_ops(struct sem_array *sma, int semnum,
        if (semnum == -1)
                pending_list = &sma->pending_const;
        else
-               pending_list = &sma->sem_base[semnum].pending_const;
+               pending_list = &sma->sems[semnum].pending_const;
 
        list_for_each_entry_safe(q, tmp, pending_list, list) {
                int error = perform_atomic_semop(sma, q);
@@ -856,7 +836,7 @@ static int do_smart_wakeup_zero(struct sem_array *sma, struct sembuf *sops,
                for (i = 0; i < nsops; i++) {
                        int num = sops[i].sem_num;
 
-                       if (sma->sem_base[num].semval == 0) {
+                       if (sma->sems[num].semval == 0) {
                                got_zero = 1;
                                semop_completed |= wake_const_ops(sma, num, wake_q);
                        }
@@ -867,7 +847,7 @@ static int do_smart_wakeup_zero(struct sem_array *sma, struct sembuf *sops,
                 * Assume all were changed.
                 */
                for (i = 0; i < sma->sem_nsems; i++) {
-                       if (sma->sem_base[i].semval == 0) {
+                       if (sma->sems[i].semval == 0) {
                                got_zero = 1;
                                semop_completed |= wake_const_ops(sma, i, wake_q);
                        }
@@ -909,7 +889,7 @@ static int update_queue(struct sem_array *sma, int semnum, struct wake_q_head *w
        if (semnum == -1)
                pending_list = &sma->pending_alter;
        else
-               pending_list = &sma->sem_base[semnum].pending_alter;
+               pending_list = &sma->sems[semnum].pending_alter;
 
 again:
        list_for_each_entry_safe(q, tmp, pending_list, list) {
@@ -922,7 +902,7 @@ again:
                 * be in the  per semaphore pending queue, and decrements
                 * cannot be successful if the value is already 0.
                 */
-               if (semnum != -1 && sma->sem_base[semnum].semval == 0)
+               if (semnum != -1 && sma->sems[semnum].semval == 0)
                        break;
 
                error = perform_atomic_semop(sma, q);
@@ -959,9 +939,9 @@ again:
 static void set_semotime(struct sem_array *sma, struct sembuf *sops)
 {
        if (sops == NULL) {
-               sma->sem_base[0].sem_otime = get_seconds();
+               sma->sems[0].sem_otime = get_seconds();
        } else {
-               sma->sem_base[sops[0].sem_num].sem_otime =
+               sma->sems[sops[0].sem_num].sem_otime =
                                                        get_seconds();
        }
 }
@@ -1067,9 +1047,9 @@ static int count_semcnt(struct sem_array *sma, ushort semnum,
        semcnt = 0;
        /* First: check the simple operations. They are easy to evaluate */
        if (count_zero)
-               l = &sma->sem_base[semnum].pending_const;
+               l = &sma->sems[semnum].pending_const;
        else
-               l = &sma->sem_base[semnum].pending_alter;
+               l = &sma->sems[semnum].pending_alter;
 
        list_for_each_entry(q, l, list) {
                /* all task on a per-semaphore list sleep on exactly
@@ -1124,7 +1104,7 @@ static void freeary(struct ipc_namespace *ns, struct kern_ipc_perm *ipcp)
                wake_up_sem_queue_prepare(q, -EIDRM, &wake_q);
        }
        for (i = 0; i < sma->sem_nsems; i++) {
-               struct sem *sem = sma->sem_base + i;
+               struct sem *sem = &sma->sems[i];
                list_for_each_entry_safe(q, tq, &sem->pending_const, list) {
                        unlink_queue(sma, q);
                        wake_up_sem_queue_prepare(q, -EIDRM, &wake_q);
@@ -1174,9 +1154,9 @@ static time_t get_semotime(struct sem_array *sma)
        int i;
        time_t res;
 
-       res = sma->sem_base[0].sem_otime;
+       res = sma->sems[0].sem_otime;
        for (i = 1; i < sma->sem_nsems; i++) {
-               time_t to = sma->sem_base[i].sem_otime;
+               time_t to = sma->sems[i].sem_otime;
 
                if (to > res)
                        res = to;
@@ -1325,7 +1305,7 @@ static int semctl_setval(struct ipc_namespace *ns, int semid, int semnum,
                return -EIDRM;
        }
 
-       curr = &sma->sem_base[semnum];
+       curr = &sma->sems[semnum];
 
        ipc_assert_locked_object(&sma->sem_perm);
        list_for_each_entry(un, &sma->list_id, list_id)
@@ -1402,7 +1382,7 @@ static int semctl_main(struct ipc_namespace *ns, int semid, int semnum,
                        }
                }
                for (i = 0; i < sma->sem_nsems; i++)
-                       sem_io[i] = sma->sem_base[i].semval;
+                       sem_io[i] = sma->sems[i].semval;
                sem_unlock(sma, -1);
                rcu_read_unlock();
                err = 0;
@@ -1450,8 +1430,8 @@ static int semctl_main(struct ipc_namespace *ns, int semid, int semnum,
                }
 
                for (i = 0; i < nsems; i++) {
-                       sma->sem_base[i].semval = sem_io[i];
-                       sma->sem_base[i].sempid = task_tgid_vnr(current);
+                       sma->sems[i].semval = sem_io[i];
+                       sma->sems[i].sempid = task_tgid_vnr(current);
                }
 
                ipc_assert_locked_object(&sma->sem_perm);
@@ -1476,7 +1456,7 @@ static int semctl_main(struct ipc_namespace *ns, int semid, int semnum,
                err = -EIDRM;
                goto out_unlock;
        }
-       curr = &sma->sem_base[semnum];
+       curr = &sma->sems[semnum];
 
        switch (cmd) {
        case GETVAL:
@@ -1932,7 +1912,7 @@ SYSCALL_DEFINE4(semtimedop, int, semid, struct sembuf __user *, tsops,
         */
        if (nsops == 1) {
                struct sem *curr;
-               curr = &sma->sem_base[sops->sem_num];
+               curr = &sma->sems[sops->sem_num];
 
                if (alter) {
                        if (sma->complex_count) {
@@ -2146,7 +2126,7 @@ void exit_sem(struct task_struct *tsk)
 
                /* perform adjustments registered in un */
                for (i = 0; i < sma->sem_nsems; i++) {
-                       struct sem *semaphore = &sma->sem_base[i];
+                       struct sem *semaphore = &sma->sems[i];
                        if (un->semadj[i]) {
                                semaphore->semval += un->semadj[i];
                                /*