ipc: shm: Make shmid_kernel timestamps y2038 safe
authorDeepa Dinamani <deepa.kernel@gmail.com>
Thu, 3 Aug 2017 02:51:14 +0000 (19:51 -0700)
committerAl Viro <viro@zeniv.linux.org.uk>
Mon, 4 Sep 2017 00:24:29 +0000 (20:24 -0400)
time_t is not y2038 safe. Replace all uses of
time_t by y2038 safe time64_t.

Similarly, replace the calls to get_seconds() with
y2038 safe ktime_get_real_seconds().
Note that this preserves fast access on 64 bit systems,
but 32 bit systems need sequence counters.

The syscall interfaces themselves are not changed as part of
the patch. They will be part of a different series.

Signed-off-by: Deepa Dinamani <deepa.kernel@gmail.com>
Reviewed-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
include/linux/shm.h
ipc/shm.c

index 04e8818..7bb897b 100644 (file)
@@ -12,9 +12,9 @@ struct shmid_kernel /* private to the kernel */
        struct file             *shm_file;
        unsigned long           shm_nattch;
        unsigned long           shm_segsz;
-       time_t                  shm_atim;
-       time_t                  shm_dtim;
-       time_t                  shm_ctim;
+       time64_t                shm_atim;
+       time64_t                shm_dtim;
+       time64_t                shm_ctim;
        pid_t                   shm_cprid;
        pid_t                   shm_lprid;
        struct user_struct      *mlock_user;
index 342024d..f3d6408 100644 (file)
--- a/ipc/shm.c
+++ b/ipc/shm.c
@@ -200,7 +200,7 @@ static int __shm_open(struct vm_area_struct *vma)
        if (IS_ERR(shp))
                return PTR_ERR(shp);
 
-       shp->shm_atim = get_seconds();
+       shp->shm_atim = ktime_get_real_seconds();
        shp->shm_lprid = task_tgid_vnr(current);
        shp->shm_nattch++;
        shm_unlock(shp);
@@ -287,7 +287,7 @@ static void shm_close(struct vm_area_struct *vma)
                goto done; /* no-op */
 
        shp->shm_lprid = task_tgid_vnr(current);
-       shp->shm_dtim = get_seconds();
+       shp->shm_dtim = ktime_get_real_seconds();
        shp->shm_nattch--;
        if (shm_may_destroy(ns, shp))
                shm_destroy(ns, shp);
@@ -592,7 +592,7 @@ static int newseg(struct ipc_namespace *ns, struct ipc_params *params)
        shp->shm_cprid = task_tgid_vnr(current);
        shp->shm_lprid = 0;
        shp->shm_atim = shp->shm_dtim = 0;
-       shp->shm_ctim = get_seconds();
+       shp->shm_ctim = ktime_get_real_seconds();
        shp->shm_segsz = size;
        shp->shm_nattch = 0;
        shp->shm_file = file;
@@ -846,7 +846,7 @@ static int shmctl_down(struct ipc_namespace *ns, int shmid, int cmd,
                err = ipc_update_perm(&shmid64->shm_perm, ipcp);
                if (err)
                        goto out_unlock0;
-               shp->shm_ctim = get_seconds();
+               shp->shm_ctim = ktime_get_real_seconds();
                break;
        default:
                err = -EINVAL;
@@ -1586,7 +1586,7 @@ static int sysvipc_shm_proc_show(struct seq_file *s, void *it)
 
        seq_printf(s,
                   "%10d %10d  %4o " SIZE_SPEC " %5u %5u  "
-                  "%5lu %5u %5u %5u %5u %10lu %10lu %10lu "
+                  "%5lu %5u %5u %5u %5u %10llu %10llu %10llu "
                   SIZE_SPEC " " SIZE_SPEC "\n",
                   shp->shm_perm.key,
                   shp->shm_perm.id,