Message-ID: <b03bc9be-e555-b282-7dfd-2c94fc56fb34@redhat.com>
Date:   Fri, 1 Nov 2019 22:51:57 +0100
From:   David Hildenbrand <david@...hat.com>
To:     linux-kernel@...r.kernel.org
Cc:     linux-mm@...ck.org, Tang Chen <tangchen@...fujitsu.com>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        "Rafael J. Wysocki" <rafael@...nel.org>,
        Andrew Morton <akpm@...ux-foundation.org>,
        Keith Busch <keith.busch@...el.com>,
        Jiri Olsa <jolsa@...nel.org>,
        "Peter Zijlstra (Intel)" <peterz@...radead.org>,
        Jani Nikula <jani.nikula@...el.com>,
        Nayna Jain <nayna@...ux.ibm.com>,
        Michal Hocko <mhocko@...e.com>,
        Oscar Salvador <osalvador@...e.de>,
        Stephen Rothwell <sfr@...b.auug.org.au>,
        Dan Williams <dan.j.williams@...el.com>,
        Pavel Tatashin <pasha.tatashin@...een.com>
Subject: Re: [PATCH v1] mm/memory_hotplug: Fix try_offline_node()

On 28.10.19 11:54, David Hildenbrand wrote:
> try_offline_node() is pretty much broken right now:
> - The node span is updated when onlining memory, not when adding it. We
>    ignore memory that was mever onlined. Bad.
> - We touch possible garbage memmaps. The pfn_to_nid(pfn) can easily
>    trigger a kernel panic. Bad for memory that is offline but also bad
>    for subsection hotadd with ZONE_DEVICE, whereby the memmap of the first
>    PFN of a section might contain garbage.
> - Sections belonging to mixed nodes are not properly considered.
> 
> As memory blocks might belong to multiple nodes, we would have to walk all
> pageblocks (or at least subsections) within present sections. However,
> we don't have a way to identify whether a memmap that is not online was
> initialized (relevant for ZONE_DEVICE). This makes things more complicated.
> 
> Luckily, we can piggy pack on the node span and the sysfs links between
> memory blocks and the node. Currently, the node span is grown when calling
> move_pfn_range_to_zone() - e.g., when onlining memory, and shrunk when
> removing memory, before calling try_offline_node(). Sysfs links are
> created via link_mem_sections(), e.g., during boot or when adding memory.
> 
> If the node still spans memory or if any memory block is linked to the
> node in sysfs, we don't set the node offline. Without CONFIG_NUMA, or
> without CONFIG_SYSFS, we will simply always detect the node as being
> linked to the memory block and not set the node offline.
> 
> Add a way to test if a sysfs link exists.
> 
> Note: We will soon stop shrinking the ZONE_DEVICE zone and the node span
> when removing ZONE_DEVICE memory to fix similar issues (acess of garbage
> memmaps) - until we have a reliable way to identify whether these memmaps
> were properly initialized. This implies later, that once a node had
> ZONE_DEVICE memory, we won't be able to set a node offline -
> which should be acceptable.
> 
> Since commit f1dd2cd13c4b ("mm, memory_hotplug: do not associate hotadded
> memory to zones until online") memory that is added is not assoziated
> with a zone/node (memmap not initialized). The introducing commit
> 60a5a19e7419 ("memory-hotplug: remove sysfs file of node") already
> missed that we could have multiple nodes for a section and that the
> zone/node span is updated when onlining pages, not when adding them.
> 
> I tested this by hotplugging two DIMMs to a memory-less and cpu-less NUMA
> node. The node is properly onlined when adding the DIMMs. When removing
> the DIMMs, the node is properly offlined.
> 
> Fixes: 60a5a19e7419 ("memory-hotplug: remove sysfs file of node")
> Fixes: f1dd2cd13c4b ("mm, memory_hotplug: do not associate hotadded memory to zones until online") # visiable after d0dc12e86b319
> Cc: Tang Chen <tangchen@...fujitsu.com>
> Cc: Greg Kroah-Hartman <gregkh@...uxfoundation.org>
> Cc: "Rafael J. Wysocki" <rafael@...nel.org>
> Cc: Andrew Morton <akpm@...ux-foundation.org>
> Cc: Keith Busch <keith.busch@...el.com>
> Cc: Jiri Olsa <jolsa@...nel.org>
> Cc: "Peter Zijlstra (Intel)" <peterz@...radead.org>
> Cc: Jani Nikula <jani.nikula@...el.com>
> Cc: Nayna Jain <nayna@...ux.ibm.com>
> Cc: Michal Hocko <mhocko@...e.com>
> Cc: Oscar Salvador <osalvador@...e.de>
> Cc: Stephen Rothwell <sfr@...b.auug.org.au>
> Cc: Dan Williams <dan.j.williams@...el.com>
> Cc: Pavel Tatashin <pasha.tatashin@...een.com>
> Signed-off-by: David Hildenbrand <david@...hat.com>
> ---
> 
> I just realized that we are are missing a try_online_node() call in the
> memremap path. Bad when adding device memory to a memory-less and cpu-less
> node - the pgdat is not initialized. Will also have to be fixed.
> 
> We stop shrinking the ZONE_DEVICE zone after the following patch:
> 	[PATCH v6 04/10] mm/memory_hotplug: Don't access uninitialized memmaps
> 	in shrink_zone_span()
> This implies, the above note regarding ZONE_DEVICE on a node blocking a
> node from getting offlined until we sorted out how to properly shrink
> the ZONE_DEVICE zone.
> 
> This patch is especially important for:
> 	[PATCH v6 05/10] mm/memory_hotplug: Shrink zones when offlining
> 	memory
> As the BUG fixed with this patch becomes now easier to observe when memory
> is offlined (in contrast to when memory would never have been onlined
> before).
> 
> As both patches are stable fixes and in next/master for a long time, we
> should probably pull this patch in front of both and also backport this
> patch at least to
> 	Cc: stable@...r.kernel.org # v4.13+
> I have not checked yet if there are real blockers to do that. I guess not.
> 
> ---
>   drivers/base/node.c   |  9 +++++++++
>   fs/sysfs/symlink.c    | 21 +++++++++++++++++++++
>   include/linux/node.h  |  7 +++++++
>   include/linux/sysfs.h |  6 ++++++
>   mm/memory_hotplug.c   | 42 ++++++++++++++++++++++++++----------------
>   5 files changed, 69 insertions(+), 16 deletions(-)
> 
> diff --git a/drivers/base/node.c b/drivers/base/node.c
> index 98a31bafc8a2..32aeb85f1d4a 100644
> --- a/drivers/base/node.c
> +++ b/drivers/base/node.c
> @@ -833,6 +833,15 @@ int link_mem_sections(int nid, unsigned long start_pfn, unsigned long end_pfn)
>   				  register_mem_sect_under_node);
>   }
>   
> +bool memory_block_registered_under_node(struct memory_block *mem, int nid)
> +{
> +	if (mem->nid == nid)
> +		return true;
> +	/* memory blocks can span multiple nodes. Check against the link. */
> +	return sysfs_link_exists(&mem->dev.kobj,
> +				 kobject_name(&node_devices[nid]->dev.kobj));
> +}
> +
>   #ifdef CONFIG_HUGETLBFS
>   /*
>    * Handle per node hstate attribute [un]registration on transistions
> diff --git a/fs/sysfs/symlink.c b/fs/sysfs/symlink.c
> index c4deecc80f67..b99697a9dae6 100644
> --- a/fs/sysfs/symlink.c
> +++ b/fs/sysfs/symlink.c
> @@ -153,6 +153,27 @@ void sysfs_remove_link(struct kobject *kobj, const char *name)
>   }
>   EXPORT_SYMBOL_GPL(sysfs_remove_link);
>   
> +/**
> + *	sysfs_link_exists - test if a symlink exists in object's directory.
> + *	@kobj:	object we're acting for.
> + *	@name:	name of the symlink to test.
> + */
> +bool sysfs_link_exists(struct kobject *kobj, const char *name)
> +{
> +	struct kernfs_node *parent = NULL, *kn;
> +
> +	if (!kobj)
> +		parent = sysfs_root_kn;
> +	else
> +		parent = kobj->sd;
> +
> +	kn = kernfs_find_and_get(parent, name);
> +	kernfs_put(kn);
> +
> +	return kn != NULL;
> +}
> +EXPORT_SYMBOL_GPL(sysfs_link_exists);
> +
>   /**
>    *	sysfs_rename_link_ns - rename symlink in object's directory.
>    *	@kobj:	object we're acting for.
> diff --git a/include/linux/node.h b/include/linux/node.h
> index 4866f32a02d8..353853018689 100644
> --- a/include/linux/node.h
> +++ b/include/linux/node.h
> @@ -138,6 +138,8 @@ extern void unregister_one_node(int nid);
>   extern int register_cpu_under_node(unsigned int cpu, unsigned int nid);
>   extern int unregister_cpu_under_node(unsigned int cpu, unsigned int nid);
>   extern void unregister_memory_block_under_nodes(struct memory_block *mem_blk);
> +extern bool memory_block_registered_under_node(struct memory_block *mem,
> +					       int nid);
>   
>   extern int register_memory_node_under_compute_node(unsigned int mem_nid,
>   						   unsigned int cpu_nid,
> @@ -171,6 +173,11 @@ static inline int unregister_cpu_under_node(unsigned int cpu, unsigned int nid)
>   static inline void unregister_memory_block_under_nodes(struct memory_block *mem_blk)
>   {
>   }
> +static inline bool memory_block_registered_under_node(struct memory_block *mem,
> +						      int nid)
> +{
> +	return true;
> +}
>   
>   static inline void register_hugetlbfs_with_node(node_registration_func_t reg,
>   						node_registration_func_t unreg)
> diff --git a/include/linux/sysfs.h b/include/linux/sysfs.h
> index fa7ee503fb76..3ae9a69a0786 100644
> --- a/include/linux/sysfs.h
> +++ b/include/linux/sysfs.h
> @@ -265,6 +265,7 @@ int __must_check sysfs_create_link_nowarn(struct kobject *kobj,
>   					  struct kobject *target,
>   					  const char *name);
>   void sysfs_remove_link(struct kobject *kobj, const char *name);
> +bool sysfs_link_exists(struct kobject *kobj, const char *name);
>   
>   int sysfs_rename_link_ns(struct kobject *kobj, struct kobject *target,
>   			 const char *old_name, const char *new_name,
> @@ -420,6 +421,11 @@ static inline void sysfs_remove_link(struct kobject *kobj, const char *name)
>   {
>   }
>   
> +static inline bool sysfs_link_exists(struct kobject *kobj, const char *name)
> +{
> +	return true;
> +}
> +
>   static inline int sysfs_rename_link_ns(struct kobject *k, struct kobject *t,
>   				       const char *old_name,
>   				       const char *new_name, const void *ns)
> diff --git a/mm/memory_hotplug.c b/mm/memory_hotplug.c
> index 0140c20837b6..c75e99b292cd 100644
> --- a/mm/memory_hotplug.c
> +++ b/mm/memory_hotplug.c
> @@ -1634,6 +1634,15 @@ static int check_cpu_on_node(pg_data_t *pgdat)
>   	return 0;
>   }
>   
> +static int check_no_memblock_registered_under_node_cb(struct memory_block *mem,
> +						      void *arg)
> +{
> +
> +	const int nid = *(int *)arg;
> +
> +	return memory_block_registered_under_node(mem, nid) ? -EEXIST : 0;
> +}
> +
>   /**
>    * try_offline_node
>    * @nid: the node ID
> @@ -1645,26 +1654,27 @@ static int check_cpu_on_node(pg_data_t *pgdat)
>    */
>   void try_offline_node(int nid)
>   {
> +	unsigned long end_pfn = section_nr_to_pfn(__highest_present_section_nr);
>   	pg_data_t *pgdat = NODE_DATA(nid);
> -	unsigned long start_pfn = pgdat->node_start_pfn;
> -	unsigned long end_pfn = start_pfn + pgdat->node_spanned_pages;
> -	unsigned long pfn;
> -
> -	for (pfn = start_pfn; pfn < end_pfn; pfn += PAGES_PER_SECTION) {
> -		unsigned long section_nr = pfn_to_section_nr(pfn);
> -
> -		if (!present_section_nr(section_nr))
> -			continue;
> +	int rc;
>   
> -		if (pfn_to_nid(pfn) != nid)
> -			continue;
> +	/*
> +	 * If the node still spans pages (especially ZONE_DEVICE), don't
> +	 * offline it. A node spans memory after move_pfn_range_to_zone(),
> +	 * e.g., after the memory block was onlined.
> +	 */
> +	if (pgdat->node_spanned_pages)
> +		return;
>   
> -		/*
> -		 * some memory sections of this node are not removed, and we
> -		 * can't offline node now.
> -		 */
> +	/*
> +	 * Especially offline memory blocks might not be spanned by the
> +	 * node. They will get spanned by the node once they get onlined.
> +	 * However, they link to the node in sysfs and can get onlined later.
> +	 */
> +	rc = walk_memory_blocks(0, PFN_PHYS(end_pfn), &nid,
> +				check_no_memblock_registered_under_node_cb);
> +	if (rc)
>   		return;
> -	}
>   
>   	if (check_cpu_on_node(pgdat))
>   		return;
> 

I just realized that we don't need the complicated sysfs handling. 
Memory blocks that belong to multiple nodes can't get offlined and are 
therefore always spanned by the node. For offline memory blocks, we can 
rely on the stored nid. Will do a quick test and send the simpler variant :)

-- 

Thanks,

David / dhildenb

Powered by blists - more mailing lists