Re: [RFC PATCH] mm: reduce noise in show_mem for lowmem allocations

Re: [RFC PATCH] mm: reduce noise in show_mem for lowmem allocations

[Mel Gorman]

On Tue, Aug 23, 2022 at 11:22:30AM +0200, Michal Hocko wrote:
> Hi,
> while discussing early DMA pool pre-allocation failure with Christoph
> [1] I have realized that the allocation failure warning is rather noisy
> for constrained allocations like GFP_DMA{32}. Those zones are usually
> not populated on all nodes very often as their memory ranges are
> constrained.
> 
> This is an attempt to reduce the ballast that doesn't provide any
> relevant information for those allocation failures investigation.
> Please note that I have only compile tested it (in my default config
> setup) and I am throwing it mostly to see what people think about it.
> 
> [1] http://lkml.kernel.org/r/20220817060647.1032426-1-hch@lst.de

No objections.

-- 
Mel Gorman
SUSE Labs

Re: [RFC PATCH] mm: reduce noise in show_mem for lowmem allocations

[Vlastimil Babka]

On 8/23/22 11:22, Michal Hocko wrote:
> All but node0 are really completely irrelevant for this allocation
> because they do not have ZONE_DMA yet it swamps the log and makes it
> harder to visually inspect.
> 
> Address this by providing gfp_maks parameter to show_mem and filter the
> output to only those zones/nodes which are relevant for the allocation.
> That means nodes which have at least one managed zone which is usable
> for the allocation (zone_idx(zone) <= gfp_zone(gfp_mask)).
> The resulting output for the same failure would become:

Looks good to me.

> [...]
> [   14.017605][    T1] Mem-Info:

Maybe print the gfp_mask (or just max zone) here again, to make it more
obvious in case somebody sents a report without the top header?

> [   14.017956][    T1] active_anon:0 inactive_anon:0 isolated_anon:0
> [   14.017956][    T1]  active_file:0 inactive_file:0 isolated_file:0
> [   14.017956][    T1]  unevictable:0 dirty:0 writeback:0
> [   14.017956][    T1]  slab_reclaimable:876 slab_unreclaimable:30300
> [   14.017956][    T1]  mapped:0 shmem:0 pagetables:12 bounce:0
> [   14.017956][    T1]  free:3170151735 free_pcp:6868 free_cma:0
> [   14.017962][    T1] Node 0 active_anon:0kB inactive_anon:0kB active_file:0kB inactive_file:0kB unevictable:0kB isolated(anon):0kB isolated(file):0kB mapped:0kB dirty:0kB writeback:0kB shmem:0kB shmem_thp: 0kB shmem_pmdmapped: 0kB anon_thp: 0kB writeback_tmp:0kB kernel_stack:7200kB pagetables:4kB all_unreclaimable? no
> [   14.018026][    T1] Node 0 DMA free:160kB boost:0kB min:0kB low:0kB high:0kB reserved_highatomic:0KB active_anon:0kB inactive_anon:0kB active_file:0kB inactive_file:0kB unevictable:0kB writepending:0kB present:15996kB managed:15360kB mlocked:0kB bounce:0kB free_pcp:0kB local_pcp:0kB free_cma:0kB
> [   14.018035][    T1] lowmem_reserve[]: 0 0 0 0 0
> [   14.018339][    T1] Node 0 DMA: 0*4kB 0*8kB 0*16kB 1*32kB (U) 0*64kB 1*128kB (U) 0*256kB 0*512kB 0*1024kB 0*2048kB 0*4096kB = 160kB
> [   14.018480][    T1] 0 total pagecache pages
> [   14.018483][    T1] 0 pages in swap cache
> [   14.018484][    T1] Swap cache stats: add 0, delete 0, find 0/0
> [   14.018486][    T1] Free swap  = 0kB
> [   14.018487][    T1] Total swap = 0kB
> [   14.018488][    T1] 3221164600 pages RAM
> [   14.018489][    T1] 0 pages HighMem/MovableOnly
> [   14.018490][    T1] 50531051 pages reserved
> [   14.018491][    T1] 0 pages cma reserved
> [   14.018492][    T1] 0 pages hwpoisoned
> 
> Signed-off-by: Michal Hocko <mhocko@suse.com>

Re: [RFC PATCH] mm: reduce noise in show_mem for lowmem allocations

[Michal Hocko]

On Thu 25-08-22 11:52:09, Vlastimil Babka wrote:
> On 8/23/22 11:22, Michal Hocko wrote:
> > All but node0 are really completely irrelevant for this allocation
> > because they do not have ZONE_DMA yet it swamps the log and makes it
> > harder to visually inspect.
> > 
> > Address this by providing gfp_maks parameter to show_mem and filter the
> > output to only those zones/nodes which are relevant for the allocation.
> > That means nodes which have at least one managed zone which is usable
> > for the allocation (zone_idx(zone) <= gfp_zone(gfp_mask)).
> > The resulting output for the same failure would become:
> 
> Looks good to me.
> 
> > [...]
> > [   14.017605][    T1] Mem-Info:
> 
> Maybe print the gfp_mask (or just max zone) here again, to make it more
> obvious in case somebody sents a report without the top header?

I have tried to not alter the output but rather filter it out. The gfp
mask is the first line of the allocation failure and from my past
experience it is usually included in reports.
> 
> > [   14.017956][    T1] active_anon:0 inactive_anon:0 isolated_anon:0
> > [   14.017956][    T1]  active_file:0 inactive_file:0 isolated_file:0
> > [   14.017956][    T1]  unevictable:0 dirty:0 writeback:0
> > [   14.017956][    T1]  slab_reclaimable:876 slab_unreclaimable:30300
> > [   14.017956][    T1]  mapped:0 shmem:0 pagetables:12 bounce:0
> > [   14.017956][    T1]  free:3170151735 free_pcp:6868 free_cma:0
> > [   14.017962][    T1] Node 0 active_anon:0kB inactive_anon:0kB active_file:0kB inactive_file:0kB unevictable:0kB isolated(anon):0kB isolated(file):0kB mapped:0kB dirty:0kB writeback:0kB shmem:0kB shmem_thp: 0kB shmem_pmdmapped: 0kB anon_thp: 0kB writeback_tmp:0kB kernel_stack:7200kB pagetables:4kB all_unreclaimable? no
> > [   14.018026][    T1] Node 0 DMA free:160kB boost:0kB min:0kB low:0kB high:0kB reserved_highatomic:0KB active_anon:0kB inactive_anon:0kB active_file:0kB inactive_file:0kB unevictable:0kB writepending:0kB present:15996kB managed:15360kB mlocked:0kB bounce:0kB free_pcp:0kB local_pcp:0kB free_cma:0kB
> > [   14.018035][    T1] lowmem_reserve[]: 0 0 0 0 0
> > [   14.018339][    T1] Node 0 DMA: 0*4kB 0*8kB 0*16kB 1*32kB (U) 0*64kB 1*128kB (U) 0*256kB 0*512kB 0*1024kB 0*2048kB 0*4096kB = 160kB
> > [   14.018480][    T1] 0 total pagecache pages
> > [   14.018483][    T1] 0 pages in swap cache
> > [   14.018484][    T1] Swap cache stats: add 0, delete 0, find 0/0
> > [   14.018486][    T1] Free swap  = 0kB
> > [   14.018487][    T1] Total swap = 0kB
> > [   14.018488][    T1] 3221164600 pages RAM
> > [   14.018489][    T1] 0 pages HighMem/MovableOnly
> > [   14.018490][    T1] 50531051 pages reserved
> > [   14.018491][    T1] 0 pages cma reserved
> > [   14.018492][    T1] 0 pages hwpoisoned
> > 
> > Signed-off-by: Michal Hocko <mhocko@suse.com>

-- 
Michal Hocko
SUSE Labs

Re: [RFC PATCH] mm: reduce noise in show_mem for lowmem allocations

[Johannes Weiner]

On Tue, Aug 23, 2022 at 11:22:30AM +0200, Michal Hocko wrote:
> Signed-off-by: Michal Hocko <mhocko@suse.com>

Acked-by: Johannes Weiner <hannes@cmpxchg.org>

Re: [RFC PATCH] mm: reduce noise in show_mem for lowmem allocations

[Vlastimil Babka]

On 8/25/22 11:59, Michal Hocko wrote:
> On Thu 25-08-22 11:52:09, Vlastimil Babka wrote:
>> On 8/23/22 11:22, Michal Hocko wrote:
>> > All but node0 are really completely irrelevant for this allocation
>> > because they do not have ZONE_DMA yet it swamps the log and makes it
>> > harder to visually inspect.
>> > 
>> > Address this by providing gfp_maks parameter to show_mem and filter the
>> > output to only those zones/nodes which are relevant for the allocation.
>> > That means nodes which have at least one managed zone which is usable
>> > for the allocation (zone_idx(zone) <= gfp_zone(gfp_mask)).
>> > The resulting output for the same failure would become:
>> 
>> Looks good to me.
>> 
>> > [...]
>> > [   14.017605][    T1] Mem-Info:
>> 
>> Maybe print the gfp_mask (or just max zone) here again, to make it more
>> obvious in case somebody sents a report without the top header?
> 
> I have tried to not alter the output but rather filter it out. The gfp
> mask is the first line of the allocation failure and from my past
> experience it is usually included in reports.

OK

>> > [   14.017956][    T1] active_anon:0 inactive_anon:0 isolated_anon:0
>> > [   14.017956][    T1]  active_file:0 inactive_file:0 isolated_file:0
>> > [   14.017956][    T1]  unevictable:0 dirty:0 writeback:0
>> > [   14.017956][    T1]  slab_reclaimable:876 slab_unreclaimable:30300
>> > [   14.017956][    T1]  mapped:0 shmem:0 pagetables:12 bounce:0
>> > [   14.017956][    T1]  free:3170151735 free_pcp:6868 free_cma:0
>> > [   14.017962][    T1] Node 0 active_anon:0kB inactive_anon:0kB active_file:0kB inactive_file:0kB unevictable:0kB isolated(anon):0kB isolated(file):0kB mapped:0kB dirty:0kB writeback:0kB shmem:0kB shmem_thp: 0kB shmem_pmdmapped: 0kB anon_thp: 0kB writeback_tmp:0kB kernel_stack:7200kB pagetables:4kB all_unreclaimable? no
>> > [   14.018026][    T1] Node 0 DMA free:160kB boost:0kB min:0kB low:0kB high:0kB reserved_highatomic:0KB active_anon:0kB inactive_anon:0kB active_file:0kB inactive_file:0kB unevictable:0kB writepending:0kB present:15996kB managed:15360kB mlocked:0kB bounce:0kB free_pcp:0kB local_pcp:0kB free_cma:0kB
>> > [   14.018035][    T1] lowmem_reserve[]: 0 0 0 0 0
>> > [   14.018339][    T1] Node 0 DMA: 0*4kB 0*8kB 0*16kB 1*32kB (U) 0*64kB 1*128kB (U) 0*256kB 0*512kB 0*1024kB 0*2048kB 0*4096kB = 160kB
>> > [   14.018480][    T1] 0 total pagecache pages
>> > [   14.018483][    T1] 0 pages in swap cache
>> > [   14.018484][    T1] Swap cache stats: add 0, delete 0, find 0/0
>> > [   14.018486][    T1] Free swap  = 0kB
>> > [   14.018487][    T1] Total swap = 0kB
>> > [   14.018488][    T1] 3221164600 pages RAM
>> > [   14.018489][    T1] 0 pages HighMem/MovableOnly
>> > [   14.018490][    T1] 50531051 pages reserved
>> > [   14.018491][    T1] 0 pages cma reserved
>> > [   14.018492][    T1] 0 pages hwpoisoned
>> > 
>> > Signed-off-by: Michal Hocko <mhocko@suse.com>

Acked-by: Vlastimil Babka <vbabka@suse.cz>

Re: [RFC PATCH] mm: reduce noise in show_mem for lowmem allocations

[Michal Hocko]

[Cc Dan]
Dan has brought up[1] that the use of gfp mask has confused his static
analyzer which assumes that GFP_HIGHUSER_MOVABLE implies a sleeping
allocation and that wouldn't be a great idea from the panic path. I
would add that most callers of this function would be really bad to
allocate.

The report itself is a false positive but it made me think a bit about
this. Even if the check is too simplistic I guess it resembles how many
developers are thinking (including me). If I see GFP_HIGHUSER_MOVABLE or
GF_KERNEL I automatically assume a sleeping allocation down the road.
And who know somebody might add one in the future even into show_mem
because the gfp parameter would be too tempting to not (ab)use.

My original intention was to use a natural allocation speak but this can
backfire so maybe it would be better to give the argument its real
meaning and that is the high_zone_idx. This is cryptic for code outside
of MM but that is not all that many callers and we can hide this fact
from them. In other words does the thing below looks better (incremental
for illustration, I will make it a proper patch if yes)?

[1] https://lore.kernel.org/all/Yw2ugrlZ8bwE5/hh@kili/?q=http%3A%2F%2Flkml.kernel.org%2Fr%2FYw2ugrlZ8bwE5%2Fhh%40kili

--- 
diff --git a/arch/powerpc/xmon/xmon.c b/arch/powerpc/xmon/xmon.c
index 19b1a94b0c00..3d9782ea3fa7 100644
--- a/arch/powerpc/xmon/xmon.c
+++ b/arch/powerpc/xmon/xmon.c
@@ -1086,7 +1086,7 @@ cmds(struct pt_regs *excp)
 				memzcan();
 				break;
 			case 'i':
-				show_mem(0, NULL, GFP_HIGHUSER_MOVABLE);
+				show_mem(0, NULL);
 				break;
 			default:
 				termch = cmd;
diff --git a/arch/sparc/kernel/setup_32.c b/arch/sparc/kernel/setup_32.c
index 2a1e5641a7cc..c8e0dd99f370 100644
--- a/arch/sparc/kernel/setup_32.c
+++ b/arch/sparc/kernel/setup_32.c
@@ -83,7 +83,7 @@ static void prom_sync_me(void)
 			     "nop\n\t" : : "r" (&trapbase));
 
 	prom_printf("PROM SYNC COMMAND...\n");
-	show_free_areas(0, NULL, GFP_HIGHUSER_MOVABLE);
+	show_free_areas(0, NULL);
 	if (!is_idle_task(current)) {
 		local_irq_enable();
 		ksys_sync();
diff --git a/drivers/tty/sysrq.c b/drivers/tty/sysrq.c
index c6b2b42d2367..d2b2720db6ca 100644
--- a/drivers/tty/sysrq.c
+++ b/drivers/tty/sysrq.c
@@ -342,7 +342,7 @@ static const struct sysrq_key_op sysrq_ftrace_dump_op = {
 
 static void sysrq_handle_showmem(int key)
 {
-	show_mem(0, NULL, GFP_HIGHUSER_MOVABLE);
+	show_mem(0, NULL);
 }
 static const struct sysrq_key_op sysrq_showmem_op = {
 	.handler	= sysrq_handle_showmem,
diff --git a/drivers/tty/vt/keyboard.c b/drivers/tty/vt/keyboard.c
index 9ceeea531cf6..be8313cdbac3 100644
--- a/drivers/tty/vt/keyboard.c
+++ b/drivers/tty/vt/keyboard.c
@@ -606,7 +606,7 @@ static void fn_scroll_back(struct vc_data *vc)
 
 static void fn_show_mem(struct vc_data *vc)
 {
-	show_mem(0, NULL, GFP_HIGHUSER_MOVABLE);
+	show_mem(0, NULL);
 }
 
 static void fn_show_state(struct vc_data *vc)
diff --git a/include/linux/mm.h b/include/linux/mm.h
index e0d0ac52770c..0756798feaff 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -1827,7 +1827,11 @@ extern void pagefault_out_of_memory(void);
  */
 #define SHOW_MEM_FILTER_NODES		(0x0001u)	/* disallowed nodes */
 
-extern void show_free_areas(unsigned int flags, nodemask_t *nodemask, gfp_t gfp_mask);
+extern void __show_free_areas(unsigned int flags, nodemask_t *nodemask, int max_zone_idx);
+static void __maybe_unused show_free_areas(unsigned int flags, nodemask_t *nodemask)
+{
+	__show_free_areas(flags, nodemask, MAX_NR_ZONES - 1);
+}
 
 #ifdef CONFIG_MMU
 extern bool can_do_mlock(void);
@@ -2563,7 +2567,12 @@ extern void calculate_min_free_kbytes(void);
 extern int __meminit init_per_zone_wmark_min(void);
 extern void mem_init(void);
 extern void __init mmap_init(void);
-extern void show_mem(unsigned int flags, nodemask_t *nodemask, gfp_t gfp_mask);
+
+extern void __show_mem(unsigned int flags, nodemask_t *nodemask, int max_zone_idx);
+static inline void show_mem(unsigned int flags, nodemask_t *nodemask)
+{
+	__show_mem(flags, nodemask, MAX_NR_ZONES - 1);
+}
 extern long si_mem_available(void);
 extern void si_meminfo(struct sysinfo * val);
 extern void si_meminfo_node(struct sysinfo *val, int nid);
diff --git a/init/initramfs.c b/init/initramfs.c
index ec966cb78363..18229cfe8906 100644
--- a/init/initramfs.c
+++ b/init/initramfs.c
@@ -63,7 +63,7 @@ static void panic_show_mem(const char *fmt, ...)
 {
 	va_list args;
 
-	show_mem(0, NULL, GFP_HIGHUSER_MOVABLE);
+	show_mem(0, NULL);
 	va_start(args, fmt);
 	panic(fmt, args);
 	va_end(args);
diff --git a/kernel/panic.c b/kernel/panic.c
index ba5f7691d15c..a3308af28a21 100644
--- a/kernel/panic.c
+++ b/kernel/panic.c
@@ -187,7 +187,7 @@ static void panic_print_sys_info(bool console_flush)
 		show_state();
 
 	if (panic_print & PANIC_PRINT_MEM_INFO)
-		show_mem(0, NULL, GFP_HIGHUSER_MOVABLE);
+		show_mem(0, NULL);
 
 	if (panic_print & PANIC_PRINT_TIMER_INFO)
 		sysrq_timer_list_show();
diff --git a/lib/show_mem.c b/lib/show_mem.c
index b97461a6c4bb..ade71096e241 100644
--- a/lib/show_mem.c
+++ b/lib/show_mem.c
@@ -8,13 +8,13 @@
 #include <linux/mm.h>
 #include <linux/cma.h>
 
-void show_mem(unsigned int filter, nodemask_t *nodemask, gfp_t gfp_mask)
+void __show_mem(unsigned int filter, nodemask_t *nodemask, int max_zone_idx)
 {
 	pg_data_t *pgdat;
 	unsigned long total = 0, reserved = 0, highmem = 0;
 
 	printk("Mem-Info:\n");
-	show_free_areas(filter, nodemask, gfp_mask);
+	show_free_areas(filter, nodemask, max_zone_idx);
 
 	for_each_online_pgdat(pgdat) {
 		int zoneid;
diff --git a/mm/nommu.c b/mm/nommu.c
index 46cff4a51fa0..9d7afc2d959e 100644
--- a/mm/nommu.c
+++ b/mm/nommu.c
@@ -1047,7 +1047,7 @@ static int do_mmap_private(struct vm_area_struct *vma,
 enomem:
 	pr_err("Allocation of length %lu from process %d (%s) failed\n",
 	       len, current->pid, current->comm);
-	show_free_areas(0, NULL, GFP_KERNEL);
+	show_free_areas(0, NULL);
 	return -ENOMEM;
 }
 
@@ -1270,13 +1270,13 @@ unsigned long do_mmap(struct file *file,
 	kmem_cache_free(vm_region_jar, region);
 	pr_warn("Allocation of vma for %lu byte allocation from process %d failed\n",
 			len, current->pid);
-	show_free_areas(0, NULL, GFP_KERNEL);
+	show_free_areas(0, NULL);
 	return -ENOMEM;
 
 error_getting_region:
 	pr_warn("Allocation of vm region for %lu byte allocation from process %d failed\n",
 			len, current->pid);
-	show_free_areas(0, NULL, GFP_KERNEL);
+	show_free_areas(0, NULL);
 	return -ENOMEM;
 }
 
diff --git a/mm/oom_kill.c b/mm/oom_kill.c
index ffeaad3b348d..94804504be9c 100644
--- a/mm/oom_kill.c
+++ b/mm/oom_kill.c
@@ -461,7 +461,7 @@ static void dump_header(struct oom_control *oc, struct task_struct *p)
 	if (is_memcg_oom(oc))
 		mem_cgroup_print_oom_meminfo(oc->memcg);
 	else {
-		show_mem(SHOW_MEM_FILTER_NODES, oc->nodemask, oc->gfp_mask);
+		__show_mem(SHOW_MEM_FILTER_NODES, oc->nodemask, gfp_zone(oc->gfp_mask));
 		if (should_dump_unreclaim_slab())
 			dump_unreclaimable_slab();
 	}
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 208e3f8b38f8..110a16ea848a 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -4248,7 +4248,7 @@ static void warn_alloc_show_mem(gfp_t gfp_mask, nodemask_t *nodemask)
 	if (!in_task() || !(gfp_mask & __GFP_DIRECT_RECLAIM))
 		filter &= ~SHOW_MEM_FILTER_NODES;
 
-	show_mem(filter, nodemask, gfp_mask);
+	__show_mem(filter, nodemask, gfp_zone(gfp_mask));
 }
 
 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...)
@@ -5944,10 +5944,9 @@ static bool node_has_managed_zones(pg_data_t *pgdat, int max_zone_idx)
  * SHOW_MEM_FILTER_NODES: suppress nodes that are not allowed by current's
  *   cpuset.
  */
-void show_free_areas(unsigned int filter, nodemask_t *nodemask, gfp_t gfp_mask)
+void __show_free_areas(unsigned int filter, nodemask_t *nodemask, int max_zone_idx)
 {
 	unsigned long free_pcp = 0;
-	int max_zone_idx = gfp_zone(gfp_mask);
 	int cpu;
 	struct zone *zone;
 	pg_data_t *pgdat;
-- 
Michal Hocko
SUSE Labs

Re: [RFC PATCH] mm: reduce noise in show_mem for lowmem allocations

[Vlastimil Babka]

On 8/30/22 09:34, Michal Hocko wrote:
> [Cc Dan]
> Dan has brought up[1] that the use of gfp mask has confused his static
> analyzer which assumes that GFP_HIGHUSER_MOVABLE implies a sleeping
> allocation and that wouldn't be a great idea from the panic path. I
> would add that most callers of this function would be really bad to
> allocate.
> 
> The report itself is a false positive but it made me think a bit about
> this. Even if the check is too simplistic I guess it resembles how many
> developers are thinking (including me). If I see GFP_HIGHUSER_MOVABLE or
> GF_KERNEL I automatically assume a sleeping allocation down the road.
> And who know somebody might add one in the future even into show_mem
> because the gfp parameter would be too tempting to not (ab)use.
> 
> My original intention was to use a natural allocation speak but this can
> backfire so maybe it would be better to give the argument its real
> meaning and that is the high_zone_idx. This is cryptic for code outside
> of MM but that is not all that many callers and we can hide this fact
> from them. In other words does the thing below looks better (incremental
> for illustration, I will make it a proper patch if yes)?

Yeah, looks better to me this way. Thanks!

> 
> [1] https://lore.kernel.org/all/Yw2ugrlZ8bwE5/hh@kili/?q=http%3A%2F%2Flkml.kernel.org%2Fr%2FYw2ugrlZ8bwE5%2Fhh%40kili
> 
> --- 
> diff --git a/arch/powerpc/xmon/xmon.c b/arch/powerpc/xmon/xmon.c
> index 19b1a94b0c00..3d9782ea3fa7 100644
> --- a/arch/powerpc/xmon/xmon.c
> +++ b/arch/powerpc/xmon/xmon.c
> @@ -1086,7 +1086,7 @@ cmds(struct pt_regs *excp)
>  				memzcan();
>  				break;
>  			case 'i':
> -				show_mem(0, NULL, GFP_HIGHUSER_MOVABLE);
> +				show_mem(0, NULL);
>  				break;
>  			default:
>  				termch = cmd;
> diff --git a/arch/sparc/kernel/setup_32.c b/arch/sparc/kernel/setup_32.c
> index 2a1e5641a7cc..c8e0dd99f370 100644
> --- a/arch/sparc/kernel/setup_32.c
> +++ b/arch/sparc/kernel/setup_32.c
> @@ -83,7 +83,7 @@ static void prom_sync_me(void)
>  			     "nop\n\t" : : "r" (&trapbase));
>  
>  	prom_printf("PROM SYNC COMMAND...\n");
> -	show_free_areas(0, NULL, GFP_HIGHUSER_MOVABLE);
> +	show_free_areas(0, NULL);
>  	if (!is_idle_task(current)) {
>  		local_irq_enable();
>  		ksys_sync();
> diff --git a/drivers/tty/sysrq.c b/drivers/tty/sysrq.c
> index c6b2b42d2367..d2b2720db6ca 100644
> --- a/drivers/tty/sysrq.c
> +++ b/drivers/tty/sysrq.c
> @@ -342,7 +342,7 @@ static const struct sysrq_key_op sysrq_ftrace_dump_op = {
>  
>  static void sysrq_handle_showmem(int key)
>  {
> -	show_mem(0, NULL, GFP_HIGHUSER_MOVABLE);
> +	show_mem(0, NULL);
>  }
>  static const struct sysrq_key_op sysrq_showmem_op = {
>  	.handler	= sysrq_handle_showmem,
> diff --git a/drivers/tty/vt/keyboard.c b/drivers/tty/vt/keyboard.c
> index 9ceeea531cf6..be8313cdbac3 100644
> --- a/drivers/tty/vt/keyboard.c
> +++ b/drivers/tty/vt/keyboard.c
> @@ -606,7 +606,7 @@ static void fn_scroll_back(struct vc_data *vc)
>  
>  static void fn_show_mem(struct vc_data *vc)
>  {
> -	show_mem(0, NULL, GFP_HIGHUSER_MOVABLE);
> +	show_mem(0, NULL);
>  }
>  
>  static void fn_show_state(struct vc_data *vc)
> diff --git a/include/linux/mm.h b/include/linux/mm.h
> index e0d0ac52770c..0756798feaff 100644
> --- a/include/linux/mm.h
> +++ b/include/linux/mm.h
> @@ -1827,7 +1827,11 @@ extern void pagefault_out_of_memory(void);
>   */
>  #define SHOW_MEM_FILTER_NODES		(0x0001u)	/* disallowed nodes */
>  
> -extern void show_free_areas(unsigned int flags, nodemask_t *nodemask, gfp_t gfp_mask);
> +extern void __show_free_areas(unsigned int flags, nodemask_t *nodemask, int max_zone_idx);
> +static void __maybe_unused show_free_areas(unsigned int flags, nodemask_t *nodemask)
> +{
> +	__show_free_areas(flags, nodemask, MAX_NR_ZONES - 1);
> +}
>  
>  #ifdef CONFIG_MMU
>  extern bool can_do_mlock(void);
> @@ -2563,7 +2567,12 @@ extern void calculate_min_free_kbytes(void);
>  extern int __meminit init_per_zone_wmark_min(void);
>  extern void mem_init(void);
>  extern void __init mmap_init(void);
> -extern void show_mem(unsigned int flags, nodemask_t *nodemask, gfp_t gfp_mask);
> +
> +extern void __show_mem(unsigned int flags, nodemask_t *nodemask, int max_zone_idx);
> +static inline void show_mem(unsigned int flags, nodemask_t *nodemask)
> +{
> +	__show_mem(flags, nodemask, MAX_NR_ZONES - 1);
> +}
>  extern long si_mem_available(void);
>  extern void si_meminfo(struct sysinfo * val);
>  extern void si_meminfo_node(struct sysinfo *val, int nid);
> diff --git a/init/initramfs.c b/init/initramfs.c
> index ec966cb78363..18229cfe8906 100644
> --- a/init/initramfs.c
> +++ b/init/initramfs.c
> @@ -63,7 +63,7 @@ static void panic_show_mem(const char *fmt, ...)
>  {
>  	va_list args;
>  
> -	show_mem(0, NULL, GFP_HIGHUSER_MOVABLE);
> +	show_mem(0, NULL);
>  	va_start(args, fmt);
>  	panic(fmt, args);
>  	va_end(args);
> diff --git a/kernel/panic.c b/kernel/panic.c
> index ba5f7691d15c..a3308af28a21 100644
> --- a/kernel/panic.c
> +++ b/kernel/panic.c
> @@ -187,7 +187,7 @@ static void panic_print_sys_info(bool console_flush)
>  		show_state();
>  
>  	if (panic_print & PANIC_PRINT_MEM_INFO)
> -		show_mem(0, NULL, GFP_HIGHUSER_MOVABLE);
> +		show_mem(0, NULL);
>  
>  	if (panic_print & PANIC_PRINT_TIMER_INFO)
>  		sysrq_timer_list_show();
> diff --git a/lib/show_mem.c b/lib/show_mem.c
> index b97461a6c4bb..ade71096e241 100644
> --- a/lib/show_mem.c
> +++ b/lib/show_mem.c
> @@ -8,13 +8,13 @@
>  #include <linux/mm.h>
>  #include <linux/cma.h>
>  
> -void show_mem(unsigned int filter, nodemask_t *nodemask, gfp_t gfp_mask)
> +void __show_mem(unsigned int filter, nodemask_t *nodemask, int max_zone_idx)
>  {
>  	pg_data_t *pgdat;
>  	unsigned long total = 0, reserved = 0, highmem = 0;
>  
>  	printk("Mem-Info:\n");
> -	show_free_areas(filter, nodemask, gfp_mask);
> +	show_free_areas(filter, nodemask, max_zone_idx);
>  
>  	for_each_online_pgdat(pgdat) {
>  		int zoneid;
> diff --git a/mm/nommu.c b/mm/nommu.c
> index 46cff4a51fa0..9d7afc2d959e 100644
> --- a/mm/nommu.c
> +++ b/mm/nommu.c
> @@ -1047,7 +1047,7 @@ static int do_mmap_private(struct vm_area_struct *vma,
>  enomem:
>  	pr_err("Allocation of length %lu from process %d (%s) failed\n",
>  	       len, current->pid, current->comm);
> -	show_free_areas(0, NULL, GFP_KERNEL);
> +	show_free_areas(0, NULL);
>  	return -ENOMEM;
>  }
>  
> @@ -1270,13 +1270,13 @@ unsigned long do_mmap(struct file *file,
>  	kmem_cache_free(vm_region_jar, region);
>  	pr_warn("Allocation of vma for %lu byte allocation from process %d failed\n",
>  			len, current->pid);
> -	show_free_areas(0, NULL, GFP_KERNEL);
> +	show_free_areas(0, NULL);
>  	return -ENOMEM;
>  
>  error_getting_region:
>  	pr_warn("Allocation of vm region for %lu byte allocation from process %d failed\n",
>  			len, current->pid);
> -	show_free_areas(0, NULL, GFP_KERNEL);
> +	show_free_areas(0, NULL);
>  	return -ENOMEM;
>  }
>  
> diff --git a/mm/oom_kill.c b/mm/oom_kill.c
> index ffeaad3b348d..94804504be9c 100644
> --- a/mm/oom_kill.c
> +++ b/mm/oom_kill.c
> @@ -461,7 +461,7 @@ static void dump_header(struct oom_control *oc, struct task_struct *p)
>  	if (is_memcg_oom(oc))
>  		mem_cgroup_print_oom_meminfo(oc->memcg);
>  	else {
> -		show_mem(SHOW_MEM_FILTER_NODES, oc->nodemask, oc->gfp_mask);
> +		__show_mem(SHOW_MEM_FILTER_NODES, oc->nodemask, gfp_zone(oc->gfp_mask));
>  		if (should_dump_unreclaim_slab())
>  			dump_unreclaimable_slab();
>  	}
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> index 208e3f8b38f8..110a16ea848a 100644
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -4248,7 +4248,7 @@ static void warn_alloc_show_mem(gfp_t gfp_mask, nodemask_t *nodemask)
>  	if (!in_task() || !(gfp_mask & __GFP_DIRECT_RECLAIM))
>  		filter &= ~SHOW_MEM_FILTER_NODES;
>  
> -	show_mem(filter, nodemask, gfp_mask);
> +	__show_mem(filter, nodemask, gfp_zone(gfp_mask));
>  }
>  
>  void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...)
> @@ -5944,10 +5944,9 @@ static bool node_has_managed_zones(pg_data_t *pgdat, int max_zone_idx)
>   * SHOW_MEM_FILTER_NODES: suppress nodes that are not allowed by current's
>   *   cpuset.
>   */
> -void show_free_areas(unsigned int filter, nodemask_t *nodemask, gfp_t gfp_mask)
> +void __show_free_areas(unsigned int filter, nodemask_t *nodemask, int max_zone_idx)
>  {
>  	unsigned long free_pcp = 0;
> -	int max_zone_idx = gfp_zone(gfp_mask);
>  	int cpu;
>  	struct zone *zone;
>  	pg_data_t *pgdat;

Re: [RFC PATCH] mm: reduce noise in show_mem for lowmem allocations

[Michal Hocko]

On Thu 01-09-22 16:14:39, Vlastimil Babka wrote:
> On 8/30/22 09:34, Michal Hocko wrote:
> > [Cc Dan]
> > Dan has brought up[1] that the use of gfp mask has confused his static
> > analyzer which assumes that GFP_HIGHUSER_MOVABLE implies a sleeping
> > allocation and that wouldn't be a great idea from the panic path. I
> > would add that most callers of this function would be really bad to
> > allocate.
> > 
> > The report itself is a false positive but it made me think a bit about
> > this. Even if the check is too simplistic I guess it resembles how many
> > developers are thinking (including me). If I see GFP_HIGHUSER_MOVABLE or
> > GF_KERNEL I automatically assume a sleeping allocation down the road.
> > And who know somebody might add one in the future even into show_mem
> > because the gfp parameter would be too tempting to not (ab)use.
> > 
> > My original intention was to use a natural allocation speak but this can
> > backfire so maybe it would be better to give the argument its real
> > meaning and that is the high_zone_idx. This is cryptic for code outside
> > of MM but that is not all that many callers and we can hide this fact
> > from them. In other words does the thing below looks better (incremental
> > for illustration, I will make it a proper patch if yes)?
> 
> Yeah, looks better to me this way. Thanks!

Thanks for looking into this. Unless somebody objects I will send a
consolidated patch early next week.
-- 
Michal Hocko
SUSE Labs