[PATCH v3 0/2] mm/show_mem: Bug fix for print mem alloc info

[PATCH v3 0/2] mm/show_mem: Bug fix for print mem alloc info

[Yueyang Pan]

This patch set fixes two issues we saw in production rollout. 

The first issue is that we saw all zero output of memory allocation 
profiling information from show_mem() if CONFIG_MEM_ALLOC_PROFILING 
is set and sysctl.vm.mem_profiling=0. This cause ambiguity as we 
don't know what 0B actually means in the output. It can mean either 
memory allocation profiling is temporary disabled or the allocation 
at that position is actually 0. Such ambiguity will make further 
parsing harder as we cannot differentiate between two case.

The second issue is that multiple entities can call show_mem() 
which messed up the allocation info in dmesg. We saw outputs like this:  
```
    327 MiB    83635 mm/compaction.c:1880 func:compaction_alloc
   48.4 GiB 12684937 mm/memory.c:1061 func:folio_prealloc
   7.48 GiB    10899 mm/huge_memory.c:1159 func:vma_alloc_anon_folio_pmd
    298 MiB    95216 kernel/fork.c:318 func:alloc_thread_stack_node
    250 MiB    63901 mm/zsmalloc.c:987 func:alloc_zspage
    1.42 GiB   372527 mm/memory.c:1063 func:folio_prealloc
    1.17 GiB    95693 mm/slub.c:2424 func:alloc_slab_page
     651 MiB   166732 mm/readahead.c:270 func:page_cache_ra_unbounded
     419 MiB   107261 net/core/page_pool.c:572 func:__page_pool_alloc_pages_slow
     404 MiB   103425 arch/x86/mm/pgtable.c:25 func:pte_alloc_one
```
The above example is because one kthread invokes show_mem() 
from __alloc_pages_slowpath while kernel itself calls 
oom_kill_process()

Revision History
=================
Changes from v2 [2]
- Merge status dump with "Memory allocations:" starting line
- Move the definition of spinlock within ifdef 

Changes from v1 [1]
- Dump status of memory allocation profiling instead of disabling 
the output following Vishal's advise.
- Move lock from file scope to within __show_mem() and replace mutex 
with spinlock following Andrew, Vlastimil and Shakeel's advice.

[1] https://lore.kernel.org/linux-mm/cover.1756318426.git.pyyjason@gmail.com/
[2] https://lore.kernel.org/linux-mm/cover.1756827906.git.pyyjason@gmail.com/

Yueyang Pan (2):
  mm/show_mem: Dump the status of the mem alloc profiling  before
    printing
  mm/show_mem: Add trylock while printing alloc info

 mm/show_mem.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

-- 
2.47.3

[PATCH v3 1/2] mm/show_mem: Dump the status of the mem alloc profiling before printing

[Yueyang Pan]

This patch prints the status of the memory allocation profiling
before __show_mem actually prints the detailed allocation info.
This way will let us know the `0B` we saw in allocation info is
because the profiling is disabled or the allocation is actually
0B.

Signed-off-by: Yueyang Pan <pyyjason@gmail.com>
---
 mm/show_mem.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/mm/show_mem.c b/mm/show_mem.c
index ecf20a93ea54..fd85a028a926 100644
--- a/mm/show_mem.c
+++ b/mm/show_mem.c
@@ -427,7 +427,8 @@ void __show_mem(unsigned int filter, nodemask_t *nodemask, int max_zone_idx)
 
 		nr = alloc_tag_top_users(tags, ARRAY_SIZE(tags), false);
 		if (nr) {
-			pr_notice("Memory allocations:\n");
+			pr_notice("Memory allocations (profiling is currently turned %s):\n",
+				mem_alloc_profiling_enabled() ? "on" : "off");
 			for (i = 0; i < nr; i++) {
 				struct codetag *ct = tags[i].ct;
 				struct alloc_tag *tag = ct_to_alloc_tag(ct);
-- 
2.47.3

[PATCH v3 2/2] mm/show_mem: Add trylock while printing alloc info

[Yueyang Pan]

In production, show_mem() can be called concurrently from two
different entities, for example one from oom_kill_process()
another from __alloc_pages_slowpath from another kthread. This
patch adds a spinlock and invokes trylock before printing out the
kernel alloc info in show_mem(). This way two alloc info won't
interleave with each other, which then makes parsing easier.

Signed-off-by: Yueyang Pan <pyyjason@gmail.com>
---
 mm/show_mem.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/mm/show_mem.c b/mm/show_mem.c
index fd85a028a926..e9701d07549b 100644
--- a/mm/show_mem.c
+++ b/mm/show_mem.c
@@ -421,7 +421,9 @@ void __show_mem(unsigned int filter, nodemask_t *nodemask, int max_zone_idx)
 	printk("%lu pages hwpoisoned\n", atomic_long_read(&num_poisoned_pages));
 #endif
 #ifdef CONFIG_MEM_ALLOC_PROFILING
-	{
+	static DEFINE_SPINLOCK(mem_alloc_profiling_spinlock);
+
+	if (spin_trylock(&mem_alloc_profiling_spinlock)) {
 		struct codetag_bytes tags[10];
 		size_t i, nr;
 
@@ -448,6 +450,7 @@ void __show_mem(unsigned int filter, nodemask_t *nodemask, int max_zone_idx)
 						  ct->lineno, ct->function);
 			}
 		}
+		spin_unlock(&mem_alloc_profiling_spinlock);
 	}
 #endif
 }
-- 
2.47.3

Re: [PATCH v3 2/2] mm/show_mem: Add trylock while printing alloc info

[Usama Arif]

On 03/09/2025 12:16, Yueyang Pan wrote:
> In production, show_mem() can be called concurrently from two
> different entities, for example one from oom_kill_process()
> another from __alloc_pages_slowpath from another kthread. This
> patch adds a spinlock and invokes trylock before printing out the
> kernel alloc info in show_mem(). This way two alloc info won't
> interleave with each other, which then makes parsing easier.
> 
> Signed-off-by: Yueyang Pan <pyyjason@gmail.com>

Acked-by: Usama Arif <usamaarif642@gmail.com>

> ---
>  mm/show_mem.c | 5 ++++-
>  1 file changed, 4 insertions(+), 1 deletion(-)
> 
> diff --git a/mm/show_mem.c b/mm/show_mem.c
> index fd85a028a926..e9701d07549b 100644
> --- a/mm/show_mem.c
> +++ b/mm/show_mem.c
> @@ -421,7 +421,9 @@ void __show_mem(unsigned int filter, nodemask_t *nodemask, int max_zone_idx)
>  	printk("%lu pages hwpoisoned\n", atomic_long_read(&num_poisoned_pages));
>  #endif
>  #ifdef CONFIG_MEM_ALLOC_PROFILING
> -	{
> +	static DEFINE_SPINLOCK(mem_alloc_profiling_spinlock);
> +
> +	if (spin_trylock(&mem_alloc_profiling_spinlock)) {
>  		struct codetag_bytes tags[10];
>  		size_t i, nr;
>  
> @@ -448,6 +450,7 @@ void __show_mem(unsigned int filter, nodemask_t *nodemask, int max_zone_idx)
>  						  ct->lineno, ct->function);
>  			}
>  		}
> +		spin_unlock(&mem_alloc_profiling_spinlock);
>  	}
>  #endif
>  }

Re: [PATCH v3 1/2] mm/show_mem: Dump the status of the mem alloc profiling before printing

[Usama Arif]

On 03/09/2025 12:16, Yueyang Pan wrote:
> This patch prints the status of the memory allocation profiling
> before __show_mem actually prints the detailed allocation info.
> This way will let us know the `0B` we saw in allocation info is
> because the profiling is disabled or the allocation is actually
> 0B.
> 
> Signed-off-by: Yueyang Pan <pyyjason@gmail.com>

Acked-by: Usama Arif <usamaarif642@gmail.com>

> ---
>  mm/show_mem.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/mm/show_mem.c b/mm/show_mem.c
> index ecf20a93ea54..fd85a028a926 100644
> --- a/mm/show_mem.c
> +++ b/mm/show_mem.c
> @@ -427,7 +427,8 @@ void __show_mem(unsigned int filter, nodemask_t *nodemask, int max_zone_idx)
>  
>  		nr = alloc_tag_top_users(tags, ARRAY_SIZE(tags), false);
>  		if (nr) {
> -			pr_notice("Memory allocations:\n");
> +			pr_notice("Memory allocations (profiling is currently turned %s):\n",
> +				mem_alloc_profiling_enabled() ? "on" : "off");
>  			for (i = 0; i < nr; i++) {
>  				struct codetag *ct = tags[i].ct;
>  				struct alloc_tag *tag = ct_to_alloc_tag(ct);

Re: [PATCH v3 1/2] mm/show_mem: Dump the status of the mem alloc profiling before printing

[Vlastimil Babka]

On 9/3/25 13:16, Yueyang Pan wrote:
> This patch prints the status of the memory allocation profiling
> before __show_mem actually prints the detailed allocation info.
> This way will let us know the `0B` we saw in allocation info is
> because the profiling is disabled or the allocation is actually
> 0B.
> 
> Signed-off-by: Yueyang Pan <pyyjason@gmail.com>

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

> ---
>  mm/show_mem.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/mm/show_mem.c b/mm/show_mem.c
> index ecf20a93ea54..fd85a028a926 100644
> --- a/mm/show_mem.c
> +++ b/mm/show_mem.c
> @@ -427,7 +427,8 @@ void __show_mem(unsigned int filter, nodemask_t *nodemask, int max_zone_idx)
>  
>  		nr = alloc_tag_top_users(tags, ARRAY_SIZE(tags), false);
>  		if (nr) {
> -			pr_notice("Memory allocations:\n");
> +			pr_notice("Memory allocations (profiling is currently turned %s):\n",
> +				mem_alloc_profiling_enabled() ? "on" : "off");
>  			for (i = 0; i < nr; i++) {
>  				struct codetag *ct = tags[i].ct;
>  				struct alloc_tag *tag = ct_to_alloc_tag(ct);

Re: [PATCH v3 2/2] mm/show_mem: Add trylock while printing alloc info

[Vlastimil Babka]

On 9/3/25 13:16, Yueyang Pan wrote:
> In production, show_mem() can be called concurrently from two
> different entities, for example one from oom_kill_process()
> another from __alloc_pages_slowpath from another kthread. This
> patch adds a spinlock and invokes trylock before printing out the
> kernel alloc info in show_mem(). This way two alloc info won't
> interleave with each other, which then makes parsing easier.
> 
> Signed-off-by: Yueyang Pan <pyyjason@gmail.com>

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

> ---
>  mm/show_mem.c | 5 ++++-
>  1 file changed, 4 insertions(+), 1 deletion(-)
> 
> diff --git a/mm/show_mem.c b/mm/show_mem.c
> index fd85a028a926..e9701d07549b 100644
> --- a/mm/show_mem.c
> +++ b/mm/show_mem.c
> @@ -421,7 +421,9 @@ void __show_mem(unsigned int filter, nodemask_t *nodemask, int max_zone_idx)
>  	printk("%lu pages hwpoisoned\n", atomic_long_read(&num_poisoned_pages));
>  #endif
>  #ifdef CONFIG_MEM_ALLOC_PROFILING
> -	{
> +	static DEFINE_SPINLOCK(mem_alloc_profiling_spinlock);
> +
> +	if (spin_trylock(&mem_alloc_profiling_spinlock)) {
>  		struct codetag_bytes tags[10];
>  		size_t i, nr;
>  
> @@ -448,6 +450,7 @@ void __show_mem(unsigned int filter, nodemask_t *nodemask, int max_zone_idx)
>  						  ct->lineno, ct->function);
>  			}
>  		}
> +		spin_unlock(&mem_alloc_profiling_spinlock);
>  	}
>  #endif
>  }

Re: [PATCH v3 1/2] mm/show_mem: Dump the status of the mem alloc profiling before printing

[Zi Yan]

On 3 Sep 2025, at 7:16, Yueyang Pan wrote:

> This patch prints the status of the memory allocation profiling
> before __show_mem actually prints the detailed allocation info.
> This way will let us know the `0B` we saw in allocation info is
> because the profiling is disabled or the allocation is actually
> 0B.
>
> Signed-off-by: Yueyang Pan <pyyjason@gmail.com>
> ---
>  mm/show_mem.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
>
Acked-by: Zi Yan <ziy@nvidia.com>

Best Regards,
Yan, Zi

Re: [PATCH v3 2/2] mm/show_mem: Add trylock while printing alloc info

[Zi Yan]

On 3 Sep 2025, at 7:16, Yueyang Pan wrote:

> In production, show_mem() can be called concurrently from two
> different entities, for example one from oom_kill_process()
> another from __alloc_pages_slowpath from another kthread. This
> patch adds a spinlock and invokes trylock before printing out the
> kernel alloc info in show_mem(). This way two alloc info won't
> interleave with each other, which then makes parsing easier.
>
> Signed-off-by: Yueyang Pan <pyyjason@gmail.com>
> ---
>  mm/show_mem.c | 5 ++++-
>  1 file changed, 4 insertions(+), 1 deletion(-)
>

Acked-by: Zi Yan <ziy@nvidia.com>

Best Regards,
Yan, Zi

Re: [PATCH v3 1/2] mm/show_mem: Dump the status of the mem alloc profiling before printing

[Suren Baghdasaryan]

On Wed, Sep 3, 2025 at 9:29 AM Zi Yan <ziy@nvidia.com> wrote:
>
> On 3 Sep 2025, at 7:16, Yueyang Pan wrote:
>
> > This patch prints the status of the memory allocation profiling
> > before __show_mem actually prints the detailed allocation info.
> > This way will let us know the `0B` we saw in allocation info is
> > because the profiling is disabled or the allocation is actually
> > 0B.
> >
> > Signed-off-by: Yueyang Pan <pyyjason@gmail.com>
> > ---
> >  mm/show_mem.c | 3 ++-
> >  1 file changed, 2 insertions(+), 1 deletion(-)
> >
> Acked-by: Zi Yan <ziy@nvidia.com>

Acked-by: Suren Baghdasaryan <surenb@google.com>

>
> Best Regards,
> Yan, Zi

Re: [PATCH v3 2/2] mm/show_mem: Add trylock while printing alloc info

[Suren Baghdasaryan]

On Wed, Sep 3, 2025 at 9:30 AM Zi Yan <ziy@nvidia.com> wrote:
>
> On 3 Sep 2025, at 7:16, Yueyang Pan wrote:
>
> > In production, show_mem() can be called concurrently from two
> > different entities, for example one from oom_kill_process()
> > another from __alloc_pages_slowpath from another kthread. This
> > patch adds a spinlock and invokes trylock before printing out the
> > kernel alloc info in show_mem(). This way two alloc info won't
> > interleave with each other, which then makes parsing easier.
> >
> > Signed-off-by: Yueyang Pan <pyyjason@gmail.com>
> > ---
> >  mm/show_mem.c | 5 ++++-
> >  1 file changed, 4 insertions(+), 1 deletion(-)
> >
>
> Acked-by: Zi Yan <ziy@nvidia.com>

Acked-by: Suren Baghdasaryan <surenb@google.com>


>
> Best Regards,
> Yan, Zi

Re: [PATCH v3 1/2] mm/show_mem: Dump the status of the mem alloc profiling before printing

[Vishal Moola (Oracle)]

On Wed, Sep 03, 2025 at 04:16:13AM -0700, Yueyang Pan wrote:
> This patch prints the status of the memory allocation profiling
> before __show_mem actually prints the detailed allocation info.
> This way will let us know the `0B` we saw in allocation info is
> because the profiling is disabled or the allocation is actually
> 0B.
> 
> Signed-off-by: Yueyang Pan <pyyjason@gmail.com>

Reviewed-by: Vishal Moola (Oracle) <vishal.moola@gmail.com>