lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Mon, 26 Feb 2024 12:06:29 -0500
From: Steven Rostedt <rostedt@...dmis.org>
To: Richard Chang <richardycc@...gle.com>
Cc: mhiramat@...nel.org, mathieu.desnoyers@...icios.com,
 akpm@...ux-foundation.org, liumartin@...gle.com, surenb@...gle.com,
 minchan@...nel.org, linux-kernel@...r.kernel.org,
 linux-trace-kernel@...r.kernel.org, linux-mm@...ck.org
Subject: Re: [PATCH] mm: add alloc_contig_migrate_range allocation
 statistics

On Mon, 26 Feb 2024 10:00:15 +0000
Richard Chang <richardycc@...gle.com> wrote:

> alloc_contig_migrate_range has every information to be able to
> understand big contiguous allocation latency. For example, how many
> pages are migrated, how many times they were needed to unmap from
> page tables.
> 
> This patch adds the trace event to collect the allocation statistics.
> In the field, it was quite useful to understand CMA allocation
> latency.
> 
> Signed-off-by: Richard Chang <richardycc@...gle.com>
> ---
>  include/trace/events/kmem.h | 39 +++++++++++++++++++++++++++++++++++++
>  mm/internal.h               |  3 ++-
>  mm/page_alloc.c             | 30 +++++++++++++++++++++++-----
>  mm/page_isolation.c         |  2 +-
>  4 files changed, 67 insertions(+), 7 deletions(-)
> 
> diff --git a/include/trace/events/kmem.h b/include/trace/events/kmem.h
> index 58688768ef0f..964704d76f9f 100644
> --- a/include/trace/events/kmem.h
> +++ b/include/trace/events/kmem.h
> @@ -304,6 +304,45 @@ TRACE_EVENT(mm_page_alloc_extfrag,
>  		__entry->change_ownership)
>  );
>  
> +TRACE_EVENT(mm_alloc_contig_migrate_range_info,
> +
> +	TP_PROTO(unsigned long start,
> +		 unsigned long end,
> +		 int migratetype,
> +		 unsigned long nr_migrated,
> +		 unsigned long nr_reclaimed,
> +		 unsigned long nr_mapped),
> +
> +	TP_ARGS(start, end, migratetype,
> +		nr_migrated, nr_reclaimed, nr_mapped),
> +
> +	TP_STRUCT__entry(
> +		__field(unsigned long, start)
> +		__field(unsigned long, end)
> +		__field(int, migratetype)


Please move the int to the end of the longs, as it will cause a 4 byte hole
in 64 bit machines otherwise.


> +		__field(unsigned long, nr_migrated)
> +		__field(unsigned long, nr_reclaimed)
> +		__field(unsigned long, nr_mapped)
> +	),
> +
> +	TP_fast_assign(
> +		__entry->start = start;
> +		__entry->end = end;
> +		__entry->migratetype = migratetype;
> +		__entry->nr_migrated = nr_migrated;
> +		__entry->nr_reclaimed = nr_reclaimed;
> +		__entry->nr_mapped = nr_mapped;
> +	),
> +
> +	TP_printk("start=0x%lx end=0x%lx migratetype=%d nr_migrated=%lu nr_reclaimed=%lu nr_mapped=%lu",
> +		  __entry->start,
> +		  __entry->end,
> +		  __entry->migratetype,
> +		  __entry->nr_migrated,
> +		  __entry->nr_reclaimed,
> +		  __entry->nr_mapped)
> +);
> +
>  /*
>   * Required for uniquely and securely identifying mm in rss_stat tracepoint.
>   */
> diff --git a/mm/internal.h b/mm/internal.h
> index f309a010d50f..e114c647e278 100644
> --- a/mm/internal.h
> +++ b/mm/internal.h
> @@ -537,7 +537,8 @@ isolate_migratepages_range(struct compact_control *cc,
>  			   unsigned long low_pfn, unsigned long end_pfn);
>  
>  int __alloc_contig_migrate_range(struct compact_control *cc,
> -					unsigned long start, unsigned long end);
> +					unsigned long start, unsigned long end,
> +					int migratetype);
>  
>  /* Free whole pageblock and set its migration type to MIGRATE_CMA. */
>  void init_cma_reserved_pageblock(struct page *page);
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> index 150d4f23b010..f840bc785afa 100644
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -6219,9 +6219,14 @@ static void alloc_contig_dump_pages(struct list_head *page_list)
>  	}
>  }
>  
> -/* [start, end) must belong to a single zone. */
> +/*
> + * [start, end) must belong to a single zone.
> + * @migratetype: using migratetype to filter the type of migration in
> + *		trace_mm_alloc_contig_migrate_range_info.
> + */
>  int __alloc_contig_migrate_range(struct compact_control *cc,
> -					unsigned long start, unsigned long end)
> +					unsigned long start, unsigned long end,
> +					int migratetype)
>  {
>  	/* This function is based on compact_zone() from compaction.c. */
>  	unsigned int nr_reclaimed;
> @@ -6232,6 +6237,10 @@ int __alloc_contig_migrate_range(struct compact_control *cc,
>  		.nid = zone_to_nid(cc->zone),
>  		.gfp_mask = GFP_USER | __GFP_MOVABLE | __GFP_RETRY_MAYFAIL,
>  	};
> +	struct page *page;
> +	unsigned long total_mapped = 0;
> +	unsigned long total_migrated = 0;
> +	unsigned long total_reclaimed = 0;
>  
>  	lru_cache_disable();
>  
> @@ -6257,9 +6266,16 @@ int __alloc_contig_migrate_range(struct compact_control *cc,
>  							&cc->migratepages);
>  		cc->nr_migratepages -= nr_reclaimed;
>  
> +		total_reclaimed += nr_reclaimed;
> +		list_for_each_entry(page, &cc->migratepages, lru)
> +			total_mapped += page_mapcount(page);

You're doing this calculation regardless of if tracing is enabled or not
and it's only used for tracing? Please add:

		if (trace_mm_alloc_contig_migrate_range_info_enabled()) {
			total_reclaimed += nr_reclaimed;
			list_for_each_entry(page, &cc->migratepages, lru)
				total_mapped += page_mapcount(page);
		}

-- Steve

> +
>  		ret = migrate_pages(&cc->migratepages, alloc_migration_target,
>  			NULL, (unsigned long)&mtc, cc->mode, MR_CONTIG_RANGE, NULL);
>  
> +		if (!ret)

May want the above to be:

		if (trace_mm_alloc_config_migrate_range_info_enabled() && !ret)

too.

-- Steve

> +			total_migrated += cc->nr_migratepages;
> +
>  		/*
>  		 * On -ENOMEM, migrate_pages() bails out right away. It is pointless
>  		 * to retry again over this error, so do the same here.
> @@ -6273,9 +6289,13 @@ int __alloc_contig_migrate_range(struct compact_control *cc,
>  		if (!(cc->gfp_mask & __GFP_NOWARN) && ret == -EBUSY)
>  			alloc_contig_dump_pages(&cc->migratepages);
>  		putback_movable_pages(&cc->migratepages);
> -		return ret;
>  	}
> -	return 0;
> +
> +	trace_mm_alloc_contig_migrate_range_info(start, end, migratetype,
> +						 total_migrated,
> +						 total_reclaimed,
> +						 total_mapped);
> +	return (ret < 0) ? ret : 0;
>  }
>  
>  /**
> @@ -6355,7 +6375,7 @@ int alloc_contig_range(unsigned long start, unsigned long end,
>  	 * allocated.  So, if we fall through be sure to clear ret so that
>  	 * -EBUSY is not accidentally used or returned to caller.
>  	 */
> -	ret = __alloc_contig_migrate_range(&cc, start, end);
> +	ret = __alloc_contig_migrate_range(&cc, start, end, migratetype);
>  	if (ret && ret != -EBUSY)
>  		goto done;
>  	ret = 0;
> diff --git a/mm/page_isolation.c b/mm/page_isolation.c
> index cd0ea3668253..a5c8fa4c2a75 100644
> --- a/mm/page_isolation.c
> +++ b/mm/page_isolation.c
> @@ -434,7 +434,7 @@ static int isolate_single_pageblock(unsigned long boundary_pfn, int flags,
>  				}
>  
>  				ret = __alloc_contig_migrate_range(&cc, head_pfn,
> -							head_pfn + nr_pages);
> +							head_pfn + nr_pages, page_mt);
>  
>  				/*
>  				 * restore the page's migratetype so that it can


Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ