mm: coalesce split strings
authorJoe Perches <joe@perches.com>
Thu, 17 Mar 2016 21:19:47 +0000 (14:19 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Thu, 17 Mar 2016 22:09:34 +0000 (15:09 -0700)
Kernel style prefers a single string over split strings when the string is
'user-visible'.

Miscellanea:

 - Add a missing newline
 - Realign arguments

Signed-off-by: Joe Perches <joe@perches.com>
Acked-by: Tejun Heo <tj@kernel.org> [percpu]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
20 files changed:
mm/dmapool.c
mm/huge_memory.c
mm/kasan/report.c
mm/kmemcheck.c
mm/kmemleak.c
mm/memblock.c
mm/memory_hotplug.c
mm/mempolicy.c
mm/mmap.c
mm/oom_kill.c
mm/page_alloc.c
mm/page_owner.c
mm/percpu.c
mm/slab.c
mm/slab_common.c
mm/slub.c
mm/sparse-vmemmap.c
mm/sparse.c
mm/swapfile.c
mm/vmalloc.c

index 57312b5d6e12aaf2da3f6acd38a19bbe28959077..2821500e81230112305f5948170a7c7367ab0b86 100644 (file)
@@ -452,13 +452,11 @@ void dma_pool_free(struct dma_pool *pool, void *vaddr, dma_addr_t dma)
                        }
                        spin_unlock_irqrestore(&pool->lock, flags);
                        if (pool->dev)
-                               dev_err(pool->dev, "dma_pool_free %s, dma %Lx "
-                                       "already free\n", pool->name,
-                                       (unsigned long long)dma);
+                               dev_err(pool->dev, "dma_pool_free %s, dma %Lx already free\n",
+                                       pool->name, (unsigned long long)dma);
                        else
-                               printk(KERN_ERR "dma_pool_free %s, dma %Lx "
-                                       "already free\n", pool->name,
-                                       (unsigned long long)dma);
+                               printk(KERN_ERR "dma_pool_free %s, dma %Lx already free\n",
+                                       pool->name, (unsigned long long)dma);
                        return;
                }
        }
index bb944c771c82c7042dd72911f5ee3c82a2e950b6..e1a177c2079170deeefc62fb66c724584ce42452 100644 (file)
@@ -168,8 +168,7 @@ static void set_recommended_min_free_kbytes(void)
 
        if (recommended_min > min_free_kbytes) {
                if (user_min_free_kbytes >= 0)
-                       pr_info("raising min_free_kbytes from %d to %lu "
-                               "to help transparent hugepage allocations\n",
+                       pr_info("raising min_free_kbytes from %d to %lu to help transparent hugepage allocations\n",
                                min_free_kbytes, recommended_min);
 
                min_free_kbytes = recommended_min;
index 12f222d0224b93120ed7131a80172666e78276d4..745aa8f3602861edbfd4de9a606e04373fa28ef6 100644 (file)
@@ -214,8 +214,7 @@ static void kasan_report_error(struct kasan_access_info *info)
         */
        kasan_disable_current();
        spin_lock_irqsave(&report_lock, flags);
-       pr_err("================================="
-               "=================================\n");
+       pr_err("==================================================================\n");
        if (info->access_addr <
                        kasan_shadow_to_mem((void *)KASAN_SHADOW_START)) {
                if ((unsigned long)info->access_addr < PAGE_SIZE)
@@ -236,8 +235,7 @@ static void kasan_report_error(struct kasan_access_info *info)
                print_address_description(info);
                print_shadow_for_address(info->first_bad_addr);
        }
-       pr_err("================================="
-               "=================================\n");
+       pr_err("==================================================================\n");
        add_taint(TAINT_BAD_PAGE, LOCKDEP_NOW_UNRELIABLE);
        spin_unlock_irqrestore(&report_lock, flags);
        kasan_enable_current();
index 6f4f424037c0dafd4b034fb33d90d9f038733810..e5f83333066e1187aa42540c7cb284e054195885 100644 (file)
@@ -20,8 +20,7 @@ void kmemcheck_alloc_shadow(struct page *page, int order, gfp_t flags, int node)
        shadow = alloc_pages_node(node, flags | __GFP_NOTRACK, order);
        if (!shadow) {
                if (printk_ratelimit())
-                       printk(KERN_ERR "kmemcheck: failed to allocate "
-                               "shadow bitmap\n");
+                       printk(KERN_ERR "kmemcheck: failed to allocate shadow bitmap\n");
                return;
        }
 
index a81cd76ea2827f9408468b00290e2296ec00852f..e6429926e95741055fe1aa87a3082e6ed073f909 100644 (file)
@@ -596,8 +596,7 @@ static struct kmemleak_object *create_object(unsigned long ptr, size_t size,
                else if (parent->pointer + parent->size <= ptr)
                        link = &parent->rb_node.rb_right;
                else {
-                       kmemleak_stop("Cannot insert 0x%lx into the object "
-                                     "search tree (overlaps existing)\n",
+                       kmemleak_stop("Cannot insert 0x%lx into the object search tree (overlaps existing)\n",
                                      ptr);
                        /*
                         * No need for parent->lock here since "parent" cannot
@@ -670,8 +669,8 @@ static void delete_object_part(unsigned long ptr, size_t size)
        object = find_and_remove_object(ptr, 1);
        if (!object) {
 #ifdef DEBUG
-               kmemleak_warn("Partially freeing unknown object at 0x%08lx "
-                             "(size %zu)\n", ptr, size);
+               kmemleak_warn("Partially freeing unknown object at 0x%08lx (size %zu)\n",
+                             ptr, size);
 #endif
                return;
        }
@@ -717,8 +716,8 @@ static void paint_ptr(unsigned long ptr, int color)
 
        object = find_and_get_object(ptr, 0);
        if (!object) {
-               kmemleak_warn("Trying to color unknown object "
-                             "at 0x%08lx as %s\n", ptr,
+               kmemleak_warn("Trying to color unknown object at 0x%08lx as %s\n",
+                             ptr,
                              (color == KMEMLEAK_GREY) ? "Grey" :
                              (color == KMEMLEAK_BLACK) ? "Black" : "Unknown");
                return;
@@ -1463,8 +1462,8 @@ static void kmemleak_scan(void)
        if (new_leaks) {
                kmemleak_found_leaks = true;
 
-               pr_info("%d new suspected memory leaks (see "
-                       "/sys/kernel/debug/kmemleak)\n", new_leaks);
+               pr_info("%d new suspected memory leaks (see /sys/kernel/debug/kmemleak)\n",
+                       new_leaks);
        }
 
 }
@@ -1795,8 +1794,7 @@ static void kmemleak_do_cleanup(struct work_struct *work)
        if (!kmemleak_found_leaks)
                __kmemleak_do_cleanup();
        else
-               pr_info("Kmemleak disabled without freeing internal data. "
-                       "Reclaim the memory with \"echo clear > /sys/kernel/debug/kmemleak\"\n");
+               pr_info("Kmemleak disabled without freeing internal data. Reclaim the memory with \"echo clear > /sys/kernel/debug/kmemleak\".\n");
 }
 
 static DECLARE_WORK(cleanup_work, kmemleak_do_cleanup);
index fc7824fa1b422f044c37dfa3d899a0df7eaeb26a..b570dddb4cb9930d6beda24473bc8a7ff6fc97d8 100644 (file)
@@ -238,8 +238,7 @@ phys_addr_t __init_memblock memblock_find_in_range_node(phys_addr_t size,
                 * so we use WARN_ONCE() here to see the stack trace if
                 * fail happens.
                 */
-               WARN_ONCE(1, "memblock: bottom-up allocation failed, "
-                            "memory hotunplug may be affected\n");
+               WARN_ONCE(1, "memblock: bottom-up allocation failed, memory hotunplug may be affected\n");
        }
 
        return __memblock_find_range_top_down(start, end, size, align, nid,
index f5758b678608b2516271e8d7bf954be46236c251..aa34431c3f31769b05a43820aca0a75c6975e8a3 100644 (file)
@@ -1970,8 +1970,7 @@ static int check_memblock_offlined_cb(struct memory_block *mem, void *arg)
 
                beginpa = PFN_PHYS(section_nr_to_pfn(mem->start_section_nr));
                endpa = PFN_PHYS(section_nr_to_pfn(mem->end_section_nr + 1))-1;
-               pr_warn("removing memory fails, because memory "
-                       "[%pa-%pa] is onlined\n",
+               pr_warn("removing memory fails, because memory [%pa-%pa] is onlined\n",
                        &beginpa, &endpa);
        }
 
index 8cbc74387df38c5572001bd2e06bd8da1af5ab70..b25de27b83d02087a135ee2e9b66ffa283e9fe46 100644 (file)
@@ -2559,9 +2559,7 @@ static void __init check_numabalancing_enable(void)
                set_numabalancing_state(numabalancing_override == 1);
 
        if (num_online_nodes() > 1 && !numabalancing_override) {
-               pr_info("%s automatic NUMA balancing. "
-                       "Configure with numa_balancing= or the "
-                       "kernel.numa_balancing sysctl",
+               pr_info("%s automatic NUMA balancing. Configure with numa_balancing= or the kernel.numa_balancing sysctl\n",
                        numabalancing_default ? "Enabling" : "Disabling");
                set_numabalancing_state(numabalancing_default);
        }
index 14641926c97f9f8514239cfc6602bb9bf1175373..e06345aafa036fcb48cba1dd1771e03d4e22724d 100644 (file)
--- a/mm/mmap.c
+++ b/mm/mmap.c
@@ -2517,9 +2517,8 @@ SYSCALL_DEFINE5(remap_file_pages, unsigned long, start, unsigned long, size,
        unsigned long ret = -EINVAL;
        struct file *file;
 
-       pr_warn_once("%s (%d) uses deprecated remap_file_pages() syscall. "
-                       "See Documentation/vm/remap_file_pages.txt.\n",
-                       current->comm, current->pid);
+       pr_warn_once("%s (%d) uses deprecated remap_file_pages() syscall. See Documentation/vm/remap_file_pages.txt.\n",
+                    current->comm, current->pid);
 
        if (prot)
                return ret;
@@ -2885,8 +2884,7 @@ bool may_expand_vm(struct mm_struct *mm, vm_flags_t flags, unsigned long npages)
        if (is_data_mapping(flags) &&
            mm->data_vm + npages > rlimit(RLIMIT_DATA) >> PAGE_SHIFT) {
                if (ignore_rlimit_data)
-                       pr_warn_once("%s (%d): VmData %lu exceed data ulimit "
-                                    "%lu. Will be forbidden soon.\n",
+                       pr_warn_once("%s (%d): VmData %lu exceed data ulimit %lu. Will be forbidden soon.\n",
                                     current->comm, current->pid,
                                     (mm->data_vm + npages) << PAGE_SHIFT,
                                     rlimit(RLIMIT_DATA));
index 63ced708eafd1686feaae600fc5a13f7adb3b7c9..fde3d374c0afa79d229d28bc7a0fec56ad99be34 100644 (file)
@@ -383,8 +383,7 @@ static void dump_tasks(struct mem_cgroup *memcg, const nodemask_t *nodemask)
 static void dump_header(struct oom_control *oc, struct task_struct *p,
                        struct mem_cgroup *memcg)
 {
-       pr_warn("%s invoked oom-killer: gfp_mask=%#x(%pGg), order=%d, "
-                       "oom_score_adj=%hd\n",
+       pr_warn("%s invoked oom-killer: gfp_mask=%#x(%pGg), order=%d, oom_score_adj=%hd\n",
                current->comm, oc->gfp_mask, &oc->gfp_mask, oc->order,
                current->signal->oom_score_adj);
 
index 30f01c6f6b888a7beed20fe4da325a60fb570ba6..42cf199652a530aec9f7c508cc91ba1890514342 100644 (file)
@@ -4074,8 +4074,7 @@ static int __parse_numa_zonelist_order(char *s)
                user_zonelist_order = ZONELIST_ORDER_ZONE;
        } else {
                printk(KERN_WARNING
-                       "Ignoring invalid numa_zonelist_order value:  "
-                       "%s\n", s);
+                      "Ignoring invalid numa_zonelist_order value:  %s\n", s);
                return -EINVAL;
        }
        return 0;
@@ -4539,12 +4538,11 @@ void __ref build_all_zonelists(pg_data_t *pgdat, struct zone *zone)
        else
                page_group_by_mobility_disabled = 0;
 
-       pr_info("Built %i zonelists in %s order, mobility grouping %s.  "
-               "Total pages: %ld\n",
-                       nr_online_nodes,
-                       zonelist_order_name[current_zonelist_order],
-                       page_group_by_mobility_disabled ? "off" : "on",
-                       vm_total_pages);
+       pr_info("Built %i zonelists in %s order, mobility grouping %s.  Total pages: %ld\n",
+               nr_online_nodes,
+               zonelist_order_name[current_zonelist_order],
+               page_group_by_mobility_disabled ? "off" : "on",
+               vm_total_pages);
 #ifdef CONFIG_NUMA
        pr_info("Policy zone: %s\n", zone_names[policy_zone]);
 #endif
@@ -6142,22 +6140,21 @@ void __init mem_init_print_info(const char *str)
 
 #undef adj_init_size
 
-       pr_info("Memory: %luK/%luK available "
-              "(%luK kernel code, %luK rwdata, %luK rodata, "
-              "%luK init, %luK bss, %luK reserved, %luK cma-reserved"
+       pr_info("Memory: %luK/%luK available (%luK kernel code, %luK rwdata, %luK rodata, %luK init, %luK bss, %luK reserved, %luK cma-reserved"
 #ifdef CONFIG_HIGHMEM
-              ", %luK highmem"
+               ", %luK highmem"
 #endif
-              "%s%s)\n",
-              nr_free_pages() << (PAGE_SHIFT-10), physpages << (PAGE_SHIFT-10),
-              codesize >> 10, datasize >> 10, rosize >> 10,
-              (init_data_size + init_code_size) >> 10, bss_size >> 10,
-              (physpages - totalram_pages - totalcma_pages) << (PAGE_SHIFT-10),
-              totalcma_pages << (PAGE_SHIFT-10),
+               "%s%s)\n",
+               nr_free_pages() << (PAGE_SHIFT - 10),
+               physpages << (PAGE_SHIFT - 10),
+               codesize >> 10, datasize >> 10, rosize >> 10,
+               (init_data_size + init_code_size) >> 10, bss_size >> 10,
+               (physpages - totalram_pages - totalcma_pages) << (PAGE_SHIFT - 10),
+               totalcma_pages << (PAGE_SHIFT - 10),
 #ifdef CONFIG_HIGHMEM
-              totalhigh_pages << (PAGE_SHIFT-10),
+               totalhigh_pages << (PAGE_SHIFT - 10),
 #endif
-              str ? ", " : "", str ? str : "");
+               str ? ", " : "", str ? str : "");
 }
 
 /**
index 44ad1f00c4e1ce8e356c94bbc751fc544f7c7229..ac3d8d129974398cb98bf0bbf2b56207fdc18642 100644 (file)
@@ -198,9 +198,8 @@ void __dump_page_owner(struct page *page)
                return;
        }
 
-       pr_alert("page allocated via order %u, migratetype %s, "
-                       "gfp_mask %#x(%pGg)\n", page_ext->order,
-                       migratetype_names[mt], gfp_mask, &gfp_mask);
+       pr_alert("page allocated via order %u, migratetype %s, gfp_mask %#x(%pGg)\n",
+                page_ext->order, migratetype_names[mt], gfp_mask, &gfp_mask);
        print_stack_trace(&trace, 0);
 
        if (page_ext->last_migrate_reason != -1)
index 847814b15233bbf16b981b3bdc6cab7447d51b62..1571547e7b013300f29d4ac849b8c265beddf086 100644 (file)
@@ -888,8 +888,8 @@ static void __percpu *pcpu_alloc(size_t size, size_t align, bool reserved,
        size = ALIGN(size, 2);
 
        if (unlikely(!size || size > PCPU_MIN_UNIT_SIZE || align > PAGE_SIZE)) {
-               WARN(true, "illegal size (%zu) or align (%zu) for "
-                    "percpu allocation\n", size, align);
+               WARN(true, "illegal size (%zu) or align (%zu) for percpu allocation\n",
+                    size, align);
                return NULL;
        }
 
index e1f6c27c3db5c028ed4ffc963756f3005805400e..e558f8593a22529a0abb1954e03060817a4ce0a6 100644 (file)
--- a/mm/slab.c
+++ b/mm/slab.c
@@ -1566,11 +1566,9 @@ static void dump_line(char *data, int offset, int limit)
        if (bad_count == 1) {
                error ^= POISON_FREE;
                if (!(error & (error - 1))) {
-                       printk(KERN_ERR "Single bit error detected. Probably "
-                                       "bad RAM.\n");
+                       printk(KERN_ERR "Single bit error detected. Probably bad RAM.\n");
 #ifdef CONFIG_X86
-                       printk(KERN_ERR "Run memtest86+ or a similar memory "
-                                       "test tool.\n");
+                       printk(KERN_ERR "Run memtest86+ or a similar memory test tool.\n");
 #else
                        printk(KERN_ERR "Run a memory test tool.\n");
 #endif
@@ -1693,11 +1691,9 @@ static void slab_destroy_debugcheck(struct kmem_cache *cachep,
                }
                if (cachep->flags & SLAB_RED_ZONE) {
                        if (*dbg_redzone1(cachep, objp) != RED_INACTIVE)
-                               slab_error(cachep, "start of a freed object "
-                                          "was overwritten");
+                               slab_error(cachep, "start of a freed object was overwritten");
                        if (*dbg_redzone2(cachep, objp) != RED_INACTIVE)
-                               slab_error(cachep, "end of a freed object "
-                                          "was overwritten");
+                               slab_error(cachep, "end of a freed object was overwritten");
                }
        }
 }
@@ -2398,11 +2394,9 @@ static void cache_init_objs_debug(struct kmem_cache *cachep, struct page *page)
 
                if (cachep->flags & SLAB_RED_ZONE) {
                        if (*dbg_redzone2(cachep, objp) != RED_INACTIVE)
-                               slab_error(cachep, "constructor overwrote the"
-                                          " end of an object");
+                               slab_error(cachep, "constructor overwrote the end of an object");
                        if (*dbg_redzone1(cachep, objp) != RED_INACTIVE)
-                               slab_error(cachep, "constructor overwrote the"
-                                          " start of an object");
+                               slab_error(cachep, "constructor overwrote the start of an object");
                }
                /* need to poison the objs? */
                if (cachep->flags & SLAB_POISON) {
@@ -2469,8 +2463,8 @@ static void slab_put_obj(struct kmem_cache *cachep,
        /* Verify double free bug */
        for (i = page->active; i < cachep->num; i++) {
                if (get_free_obj(page, i) == objnr) {
-                       printk(KERN_ERR "slab: double free detected in cache "
-                                       "'%s', objp %p\n", cachep->name, objp);
+                       printk(KERN_ERR "slab: double free detected in cache '%s', objp %p\n",
+                              cachep->name, objp);
                        BUG();
                }
        }
@@ -2901,8 +2895,7 @@ static void *cache_alloc_debugcheck_after(struct kmem_cache *cachep,
        if (cachep->flags & SLAB_RED_ZONE) {
                if (*dbg_redzone1(cachep, objp) != RED_INACTIVE ||
                                *dbg_redzone2(cachep, objp) != RED_INACTIVE) {
-                       slab_error(cachep, "double free, or memory outside"
-                                               " object was overwritten");
+                       slab_error(cachep, "double free, or memory outside object was overwritten");
                        printk(KERN_ERR
                                "%p: redzone 1:0x%llx, redzone 2:0x%llx\n",
                                objp, *dbg_redzone1(cachep, objp),
@@ -4028,8 +4021,7 @@ void slabinfo_show_stats(struct seq_file *m, struct kmem_cache *cachep)
                unsigned long node_frees = cachep->node_frees;
                unsigned long overflows = cachep->node_overflow;
 
-               seq_printf(m, " : globalstat %7lu %6lu %5lu %4lu "
-                          "%4lu %4lu %4lu %4lu %4lu",
+               seq_printf(m, " : globalstat %7lu %6lu %5lu %4lu %4lu %4lu %4lu %4lu %4lu",
                           allocs, high, grown,
                           reaped, errors, max_freeable, node_allocs,
                           node_frees, overflows);
index 8addc3c4df37ecc8923098f793215724559386e9..e885e11a316f0f43663d46bc879200c3eb725559 100644 (file)
@@ -726,8 +726,8 @@ void kmem_cache_destroy(struct kmem_cache *s)
                err = shutdown_cache(s, &release, &need_rcu_barrier);
 
        if (err) {
-               pr_err("kmem_cache_destroy %s: "
-                      "Slab cache still has objects\n", s->name);
+               pr_err("kmem_cache_destroy %s: Slab cache still has objects\n",
+                      s->name);
                dump_stack();
        }
 out_unlock:
@@ -1047,13 +1047,11 @@ static void print_slabinfo_header(struct seq_file *m)
 #else
        seq_puts(m, "slabinfo - version: 2.1\n");
 #endif
-       seq_puts(m, "# name            <active_objs> <num_objs> <objsize> "
-                "<objperslab> <pagesperslab>");
+       seq_puts(m, "# name            <active_objs> <num_objs> <objsize> <objperslab> <pagesperslab>");
        seq_puts(m, " : tunables <limit> <batchcount> <sharedfactor>");
        seq_puts(m, " : slabdata <active_slabs> <num_slabs> <sharedavail>");
 #ifdef CONFIG_DEBUG_SLAB
-       seq_puts(m, " : globalstat <listallocs> <maxobjs> <grown> <reaped> "
-                "<error> <maxfreeable> <nodeallocs> <remotefrees> <alienoverflow>");
+       seq_puts(m, " : globalstat <listallocs> <maxobjs> <grown> <reaped> <error> <maxfreeable> <nodeallocs> <remotefrees> <alienoverflow>");
        seq_puts(m, " : cpustat <allochit> <allocmiss> <freehit> <freemiss>");
 #endif
        seq_putc(m, '\n');
index 64ed5f3a3046222caa6bc426303788ce4a40cb0f..7277413ebc8b9a75806fba91103f57557b248cc3 100644 (file)
--- a/mm/slub.c
+++ b/mm/slub.c
@@ -950,14 +950,14 @@ static int on_freelist(struct kmem_cache *s, struct page *page, void *search)
                max_objects = MAX_OBJS_PER_PAGE;
 
        if (page->objects != max_objects) {
-               slab_err(s, page, "Wrong number of objects. Found %d but "
-                       "should be %d", page->objects, max_objects);
+               slab_err(s, page, "Wrong number of objects. Found %d but should be %d",
+                        page->objects, max_objects);
                page->objects = max_objects;
                slab_fix(s, "Number of objects adjusted.");
        }
        if (page->inuse != page->objects - nr) {
-               slab_err(s, page, "Wrong object count. Counter is %d but "
-                       "counted were %d", page->inuse, page->objects - nr);
+               slab_err(s, page, "Wrong object count. Counter is %d but counted were %d",
+                        page->inuse, page->objects - nr);
                page->inuse = page->objects - nr;
                slab_fix(s, "Object count adjusted.");
        }
@@ -1117,8 +1117,8 @@ static inline int free_consistency_checks(struct kmem_cache *s,
 
        if (unlikely(s != page->slab_cache)) {
                if (!PageSlab(page)) {
-                       slab_err(s, page, "Attempt to free object(0x%p) "
-                               "outside of slab", object);
+                       slab_err(s, page, "Attempt to free object(0x%p) outside of slab",
+                                object);
                } else if (!page->slab_cache) {
                        pr_err("SLUB <none>: no slab for object 0x%p.\n",
                               object);
@@ -3439,10 +3439,9 @@ static int kmem_cache_open(struct kmem_cache *s, unsigned long flags)
        free_kmem_cache_nodes(s);
 error:
        if (flags & SLAB_PANIC)
-               panic("Cannot create slab %s size=%lu realsize=%u "
-                       "order=%u offset=%u flags=%lx\n",
-                       s->name, (unsigned long)s->size, s->size,
-                       oo_order(s->oo), s->offset, flags);
+               panic("Cannot create slab %s size=%lu realsize=%u order=%u offset=%u flags=%lx\n",
+                     s->name, (unsigned long)s->size, s->size,
+                     oo_order(s->oo), s->offset, flags);
        return -EINVAL;
 }
 
index b60802b3e5ead91c346ffd1b47d90b77155bdf79..d3511f9ad0f90112ade77f9eddbae8cb3297028c 100644 (file)
@@ -166,8 +166,8 @@ void __meminit vmemmap_verify(pte_t *pte, int node,
        int actual_node = early_pfn_to_nid(pfn);
 
        if (node_distance(actual_node, node) > LOCAL_DISTANCE)
-               printk(KERN_WARNING "[%lx-%lx] potential offnode "
-                       "page_structs\n", start, end - 1);
+               printk(KERN_WARNING "[%lx-%lx] potential offnode page_structs\n",
+                      start, end - 1);
 }
 
 pte_t * __meminit vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node)
@@ -292,8 +292,8 @@ void __init sparse_mem_maps_populate_node(struct page **map_map,
                if (map_map[pnum])
                        continue;
                ms = __nr_to_section(pnum);
-               printk(KERN_ERR "%s: sparsemem memory map backing failed "
-                       "some memory will not be available.\n", __func__);
+               printk(KERN_ERR "%s: sparsemem memory map backing failed some memory will not be available.\n",
+                      __func__);
                ms->section_mem_map = 0;
        }
 
index 3717ceed4177c9183626673134116b6e0f87adda..7cdb27d9f01f582911e6a86b4605bb3ba0759652 100644 (file)
@@ -428,8 +428,8 @@ void __init sparse_mem_maps_populate_node(struct page **map_map,
                if (map_map[pnum])
                        continue;
                ms = __nr_to_section(pnum);
-               printk(KERN_ERR "%s: sparsemem memory map backing failed "
-                       "some memory will not be available.\n", __func__);
+               printk(KERN_ERR "%s: sparsemem memory map backing failed some memory will not be available.\n",
+                      __func__);
                ms->section_mem_map = 0;
        }
 }
@@ -456,8 +456,8 @@ static struct page __init *sparse_early_mem_map_alloc(unsigned long pnum)
        if (map)
                return map;
 
-       printk(KERN_ERR "%s: sparsemem memory map backing failed "
-                       "some memory will not be available.\n", __func__);
+       printk(KERN_ERR "%s: sparsemem memory map backing failed some memory will not be available.\n",
+              __func__);
        ms->section_mem_map = 0;
        return NULL;
 }
index d2c37365e2d6e429cf0f7cdbb62308f5e76aea93..b86cf26a586b6a6e71b6dfb79e0d1c2442c5e77a 100644 (file)
@@ -2526,8 +2526,7 @@ SYSCALL_DEFINE2(swapon, const char __user *, specialfile, int, swap_flags)
                  (swap_flags & SWAP_FLAG_PRIO_MASK) >> SWAP_FLAG_PRIO_SHIFT;
        enable_swap_info(p, prio, swap_map, cluster_info, frontswap_map);
 
-       pr_info("Adding %uk swap on %s.  "
-                       "Priority:%d extents:%d across:%lluk %s%s%s%s%s\n",
+       pr_info("Adding %uk swap on %s.  Priority:%d extents:%d across:%lluk %s%s%s%s%s\n",
                p->pages<<(PAGE_SHIFT-10), name->name, p->prio,
                nr_extents, (unsigned long long)span<<(PAGE_SHIFT-10),
                (p->flags & SWP_SOLIDSTATE) ? "SS" : "",
index d4b2e34adae003214ef923b04bdd39867a19bf18..e86c24ee94453df67c8d73b073e80ae404a2cfe6 100644 (file)
@@ -469,8 +469,8 @@ overflow:
                goto retry;
        }
        if (printk_ratelimit())
-               pr_warn("vmap allocation for size %lu failed: "
-                       "use vmalloc=<size> to increase size.\n", size);
+               pr_warn("vmap allocation for size %lu failed: use vmalloc=<size> to increase size\n",
+                       size);
        kfree(va);
        return ERR_PTR(-EBUSY);
 }