Re: NGROUPS 2.6.2rc2

[Andrew Morton]

Tim Hockin <thockin@xxxxxxx> wrote:
>
> Attached is a patch to remove the NGROUPS limit (again).

+/* export the group_info to a user-space array */
+static int groups_to_user(gid_t *grouplist, struct group_info __user *info)
+{
+	int i;
+	int count = info->ngroups;
+
+	for (i = 0; i < info->nblocks; i++) {
+		int cp_count = min(NGROUPS_BLOCK, count);
+		int off = i * NGROUPS_BLOCK;
+		int len = cp_count * sizeof(*grouplist);
+
+		if (copy_to_user(grouplist+off, info->blocks[i], len))
+			return -EFAULT;
+

This had me thorougly confused for a while ;) The __user tag here should
apply to grouplist, not to info.


+static int groups16_to_user(old_gid_t __user *grouplist,
+    struct group_info *info)
+{
+	int i;
+	old_gid_t group;
+
+	if (info->ngroups > TASK_SIZE/sizeof(group))
+		return -EFAULT;
+	if (!access_ok(VERIFY_WRITE, grouplist, info->ngroups * sizeof(group)))
+		return -EFAULT;

Why are many functions playing with TASK_SIZE?

--- 1.2/fs/nfsd/auth.c	Tue Jun 17 16:31:29 2003
+++ edited/fs/nfsd/auth.c	Tue Jan 27 12:40:02 2004
@@ -10,12 +10,15 @@
 #include <linux/sunrpc/svcauth.h>
 #include <linux/nfsd/nfsd.h>
 
+extern asmlinkage long sys_setgroups(int gidsetsize, gid_t *grouplist);
+

rant.  We have soooo many syscalls declared in .c files.  We had a bug due
to this a while back.  Problem is, we have no anointed header in which to
place them.  include/linux/syscalls.h would suit.  And unistd.h for
arch-specific syscalls.  But that's not appropriate to this patch.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/