First, number of CPUs can't be negative number.
Second, different signnnedness leads to suboptimal code in the following
cases:
1)
	kmalloc(nr_cpu_ids * sizeof(X));
"int" has to be sign extended to size_t.
2)
	while (loff_t *pos < nr_cpu_ids)
MOVSXD is 1 byte longed than the same MOV.
Other cases exist as well. Basically compiler is told that nr_cpu_ids
can't be negative which can't be deduced if it is "int".
Code savings on allyesconfig kernel: -3KB
	add/remove: 0/0 grow/shrink: 25/264 up/down: 261/-3631 (-3370)
	function                                     old     new   delta
	coretemp_cpu_online                          450     512     +62
	rcu_init_one                                1234    1272     +38
	pci_device_probe                             374     399     +25
				...
	pgdat_reclaimable_pages                      628     556     -72
	select_fallback_rq                           446     369     -77
	task_numa_find_cpu                          1923    1807    -116
Link: http://lkml.kernel.org/r/20170819114959.GA30580@avx2
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
                                      acpi_parse_gic_cpu_interface, 0);
 
        if (cpu_count > nr_cpu_ids)
-               pr_warn("Number of cores (%d) exceeds configured maximum of %d - clipping\n",
+               pr_warn("Number of cores (%d) exceeds configured maximum of %u - clipping\n",
                        cpu_count, nr_cpu_ids);
 
        if (!bootcpu_valid) {
 
        paca = __va(memblock_alloc_base(paca_size, PAGE_SIZE, limit));
        memset(paca, 0, paca_size);
 
-       printk(KERN_DEBUG "Allocated %u bytes for %d pacas at %p\n",
+       printk(KERN_DEBUG "Allocated %u bytes for %u pacas at %p\n",
                paca_size, nr_cpu_ids, paca);
 
        allocate_lppacas(nr_cpu_ids, limit);
 
                if (maxcpus > nr_cpu_ids) {
                        printk(KERN_WARNING
                               "Partition configured for %d cpus, "
-                              "operating system maximum is %d.\n",
+                              "operating system maximum is %u.\n",
                               maxcpus, nr_cpu_ids);
                        maxcpus = nr_cpu_ids;
                } else
 
 static void xive_native_setup_pools(void)
 {
        /* Allocate a pool big enough */
-       pr_debug("XIVE: Allocating VP block for pool size %d\n", nr_cpu_ids);
+       pr_debug("XIVE: Allocating VP block for pool size %u\n", nr_cpu_ids);
 
        xive_pool_vps = xive_native_alloc_vp_block(nr_cpu_ids);
        if (WARN_ON(xive_pool_vps == XIVE_INVALID_VP))
                pr_err("XIVE: Failed to allocate pool VP, KVM might not function\n");
 
-       pr_debug("XIVE: Pool VPs allocated at 0x%x for %d max CPUs\n",
+       pr_debug("XIVE: Pool VPs allocated at 0x%x for %u max CPUs\n",
                 xive_pool_vps, nr_cpu_ids);
 }
 
 
         * We use a struct cpumask for this, so it must be big enough.
         */
        if ((smp_height * smp_width) > nr_cpu_ids)
-               early_panic("Hypervisor %d x %d grid too big for Linux NR_CPUS %d\n",
+               early_panic("Hypervisor %d x %d grid too big for Linux NR_CPUS %u\n",
                            smp_height, smp_width, nr_cpu_ids);
 #endif
 
 
 
        /* Allocate a new cpuid. */
        if (nr_logical_cpuids >= nr_cpu_ids) {
-               WARN_ONCE(1, "APIC: NR_CPUS/possible_cpus limit of %i reached. "
+               WARN_ONCE(1, "APIC: NR_CPUS/possible_cpus limit of %u reached. "
                             "Processor %d/0x%x and the rest are ignored.\n",
                             nr_cpu_ids, nr_logical_cpuids, apicid);
                return -EINVAL;
 
        unsigned long delta;
        int rc;
 
-       pr_info("NR_CPUS:%d nr_cpumask_bits:%d nr_cpu_ids:%d nr_node_ids:%d\n",
+       pr_info("NR_CPUS:%d nr_cpumask_bits:%d nr_cpu_ids:%u nr_node_ids:%d\n",
                NR_CPUS, nr_cpumask_bits, nr_cpu_ids, nr_node_ids);
 
        /*
 
 
        /* nr_cpu_ids could be reduced via nr_cpus= */
        if (possible > nr_cpu_ids) {
-               pr_warn("%d Processors exceeds NR_CPUS limit of %d\n",
+               pr_warn("%d Processors exceeds NR_CPUS limit of %u\n",
                        possible, nr_cpu_ids);
                possible = nr_cpu_ids;
        }
 
                        buf[n++] = ',';
 
                if (nr_cpu_ids == total_cpus-1)
-                       n += snprintf(&buf[n], len - n, "%d", nr_cpu_ids);
+                       n += snprintf(&buf[n], len - n, "%u", nr_cpu_ids);
                else
-                       n += snprintf(&buf[n], len - n, "%d-%d",
+                       n += snprintf(&buf[n], len - n, "%u-%d",
                                                      nr_cpu_ids, total_cpus-1);
        }
 
 
                return error;
        }
        if (submit_queues > nr_cpu_ids) {
-               pr_warn("%s: trim submit_queues (was %d) to nr_cpu_ids=%d\n",
+               pr_warn("%s: trim submit_queues (was %d) to nr_cpu_ids=%u\n",
                        my_name, submit_queues, nr_cpu_ids);
                submit_queues = nr_cpu_ids;
        }
 
 #define cpumask_pr_args(maskp)         nr_cpu_ids, cpumask_bits(maskp)
 
 #if NR_CPUS == 1
-#define nr_cpu_ids             1
+#define nr_cpu_ids             1U
 #else
-extern int nr_cpu_ids;
+extern unsigned int nr_cpu_ids;
 #endif
 
 #ifdef CONFIG_CPUMASK_OFFSTACK
 /* Assuming NR_CPUS is huge, a runtime limit is more efficient.  Also,
  * not all bits may be allocated. */
-#define nr_cpumask_bits        ((unsigned int)nr_cpu_ids)
+#define nr_cpumask_bits        nr_cpu_ids
 #else
 #define nr_cpumask_bits        ((unsigned int)NR_CPUS)
 #endif
 
        if (rcu_fanout_leaf == RCU_FANOUT_LEAF &&
            nr_cpu_ids == NR_CPUS)
                return;
-       pr_info("RCU: Adjusting geometry for rcu_fanout_leaf=%d, nr_cpu_ids=%d\n",
+       pr_info("RCU: Adjusting geometry for rcu_fanout_leaf=%d, nr_cpu_ids=%u\n",
                rcu_fanout_leaf, nr_cpu_ids);
 
        /*
 
        if (rcu_fanout_leaf != RCU_FANOUT_LEAF)
                pr_info("\tBoot-time adjustment of leaf fanout to %d.\n", rcu_fanout_leaf);
        if (nr_cpu_ids != NR_CPUS)
-               pr_info("\tRCU restricting CPUs from NR_CPUS=%d to nr_cpu_ids=%d.\n", NR_CPUS, nr_cpu_ids);
+               pr_info("\tRCU restricting CPUs from NR_CPUS=%d to nr_cpu_ids=%u.\n", NR_CPUS, nr_cpu_ids);
 #ifdef CONFIG_RCU_BOOST
        pr_info("\tRCU priority boosting: priority %d delay %d ms.\n", kthread_prio, CONFIG_RCU_BOOST_DELAY);
 #endif
 
        alloc_bootmem_cpumask_var(&cpu_isolated_map);
        ret = cpulist_parse(str, cpu_isolated_map);
        if (ret) {
-               pr_err("sched: Error, all isolcpus= values must be between 0 and %d\n", nr_cpu_ids);
+               pr_err("sched: Error, all isolcpus= values must be between 0 and %u\n", nr_cpu_ids);
                return 0;
        }
        return 1;
 
 early_param("maxcpus", maxcpus);
 
 /* Setup number of possible processor ids */
-int nr_cpu_ids __read_mostly = NR_CPUS;
+unsigned int nr_cpu_ids __read_mostly = NR_CPUS;
 EXPORT_SYMBOL(nr_cpu_ids);
 
 /* An arch may set nr_cpu_ids earlier if needed, so this would be redundant */
 
 
 static __init int init_graph_trace(void)
 {
-       max_bytes_for_cpu = snprintf(NULL, 0, "%d", nr_cpu_ids - 1);
+       max_bytes_for_cpu = snprintf(NULL, 0, "%u", nr_cpu_ids - 1);
 
        if (!register_trace_event(&graph_trace_entry_event)) {
                pr_warn("Warning: could not register graph trace events\n");
 
        cpuhp_setup_state_nocalls(CPUHP_SLUB_DEAD, "slub:dead", NULL,
                                  slub_cpu_dead);
 
-       pr_info("SLUB: HWalign=%d, Order=%d-%d, MinObjects=%d, CPUs=%d, Nodes=%d\n",
+       pr_info("SLUB: HWalign=%d, Order=%d-%d, MinObjects=%d, CPUs=%u, Nodes=%d\n",
                cache_line_size(),
                slub_min_order, slub_max_order, slub_min_objects,
                nr_cpu_ids, nr_node_ids);