Re: [PATCH v1 08/12] powerpc/pseries: CMM: Implement balloon compaction

[David Hildenbrand <david@redhat.com>]

On 31.10.19 15:29, David Hildenbrand wrote:
> We can now get rid of the cmm_lock and completely rely on the balloon
> compaction internals, which now also manage the page list and the lock.
> 
> Inflated/"loaned" pages are now movable. Memory blocks that contain
> such apges can get offlined. Also, all such pages will be marked
> PageOffline() and can therefore be excluded in memory dumps using recent
> versions of makedumpfile.
> 
> Don't switch to balloon_page_alloc() yet (due to the GFP_NOIO). Will
> do that separately to discuss this change in detail.
> 
> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
> Cc: Paul Mackerras <paulus@samba.org>
> Cc: Michael Ellerman <mpe@ellerman.id.au>
> Cc: Thiago Jung Bauermann <bauerman@linux.ibm.com>
> Cc: Geert Uytterhoeven <geert@linux-m68k.org>
> Cc: Anshuman Khandual <khandual@linux.vnet.ibm.com>
> Cc: "Oliver O'Halloran" <oohall@gmail.com>
> Cc: Alexey Kardashevskiy <aik@ozlabs.ru>
> Cc: "Enrico Weigelt, metux IT consult" <info@metux.net>
> Cc: Thomas Gleixner <tglx@linutronix.de>
> Cc: Allison Randal <allison@lohutok.net>
> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
> Cc: Pavel Tatashin <pasha.tatashin@soleen.com>
> Cc: Arun KS <arunks@codeaurora.org>
> Cc: Todd Kjos <tkjos@google.com>
> Cc: Christian Brauner <christian@brauner.io>
> Cc: Gao Xiang <xiang@kernel.org>
> Cc: Greg Hackmann <ghackmann@google.com>
> Cc: David Howells <dhowells@redhat.com>
> Signed-off-by: David Hildenbrand <david@redhat.com>
> ---
>   arch/powerpc/platforms/pseries/Kconfig |   1 +
>   arch/powerpc/platforms/pseries/cmm.c   | 132 ++++++++++++++++++++++---
>   include/uapi/linux/magic.h             |   1 +
>   3 files changed, 120 insertions(+), 14 deletions(-)
> 
> diff --git a/arch/powerpc/platforms/pseries/Kconfig b/arch/powerpc/platforms/pseries/Kconfig
> index 9e35cddddf73..595e9f8a6539 100644
> --- a/arch/powerpc/platforms/pseries/Kconfig
> +++ b/arch/powerpc/platforms/pseries/Kconfig
> @@ -108,6 +108,7 @@ config PPC_SMLPAR
>   config CMM
>   	tristate "Collaborative memory management"
>   	depends on PPC_SMLPAR
> +	select MEMORY_BALLOON
>   	default y
>   	help
>   	  Select this option, if you want to enable the kernel interface
> diff --git a/arch/powerpc/platforms/pseries/cmm.c b/arch/powerpc/platforms/pseries/cmm.c
> index 3a55dd1fdd39..235fd7fe9df1 100644
> --- a/arch/powerpc/platforms/pseries/cmm.c
> +++ b/arch/powerpc/platforms/pseries/cmm.c
> @@ -19,6 +19,10 @@
>   #include <linux/stringify.h>
>   #include <linux/swap.h>
>   #include <linux/device.h>
> +#include <linux/mount.h>
> +#include <linux/pseudo_fs.h>
> +#include <linux/magic.h>
> +#include <linux/balloon_compaction.h>
>   #include <asm/firmware.h>
>   #include <asm/hvcall.h>
>   #include <asm/mmu.h>
> @@ -77,13 +81,11 @@ static atomic_long_t loaned_pages;
>   static unsigned long loaned_pages_target;
>   static unsigned long oom_freed_pages;
>   
> -static LIST_HEAD(cmm_page_list);
> -static DEFINE_SPINLOCK(cmm_lock);
> -
>   static DEFINE_MUTEX(hotplug_mutex);
>   static int hotplug_occurred; /* protected by the hotplug mutex */
>   
>   static struct task_struct *cmm_thread_ptr;
> +static struct balloon_dev_info b_dev_info;
>   
>   static long plpar_page_set_loaned(struct page *page)
>   {
> @@ -149,19 +151,16 @@ static long cmm_alloc_pages(long nr)
>   				  __GFP_NOMEMALLOC);
>   		if (!page)
>   			break;
> -		spin_lock(&cmm_lock);
>   		rc = plpar_page_set_loaned(page);
>   		if (rc) {
>   			pr_err("%s: Can not set page to loaned. rc=%ld\n", __func__, rc);
> -			spin_unlock(&cmm_lock);
>   			__free_page(page);
>   			break;
>   		}
>   
> -		list_add(&page->lru, &cmm_page_list);
> +		balloon_page_enqueue(&b_dev_info, page);
>   		atomic_long_inc(&loaned_pages);
>   		adjust_managed_page_count(page, -1);
> -		spin_unlock(&cmm_lock);
>   		nr--;
>   	}
>   
> @@ -178,21 +177,19 @@ static long cmm_alloc_pages(long nr)
>    **/
>   static long cmm_free_pages(long nr)
>   {
> -	struct page *page, *tmp;
> +	struct page *page;
>   
>   	cmm_dbg("Begin free of %ld pages.\n", nr);
> -	spin_lock(&cmm_lock);
> -	list_for_each_entry_safe(page, tmp, &cmm_page_list, lru) {
> -		if (!nr)
> +	while (nr) {
> +		page = balloon_page_dequeue(&b_dev_info);
> +		if (!page)
>   			break;
>   		plpar_page_set_active(page);
> -		list_del(&page->lru);
>   		adjust_managed_page_count(page, 1);
>   		__free_page(page);
>   		atomic_long_dec(&loaned_pages);
>   		nr--;
>   	}
> -	spin_unlock(&cmm_lock);
>   	cmm_dbg("End request with %ld pages unfulfilled\n", nr);
>   	return nr;
>   }
> @@ -484,6 +481,105 @@ static struct notifier_block cmm_mem_nb = {
>   	.priority = CMM_MEM_HOTPLUG_PRI
>   };
>   
> +#ifdef CONFIG_BALLOON_COMPACTION
> +static struct vfsmount *balloon_mnt;
> +
> +static int cmm_init_fs_context(struct fs_context *fc)
> +{
> +	return init_pseudo(fc, PPC_CMM_MAGIC) ? 0 : -ENOMEM;
> +}
> +
> +static struct file_system_type balloon_fs = {
> +	.name = "ppc-cmm",
> +	.init_fs_context = cmm_init_fs_context,
> +	.kill_sb = kill_anon_super,
> +};
> +
> +static int cmm_migratepage(struct balloon_dev_info *b_dev_info,
> +			   struct page *newpage, struct page *page,
> +			   enum migrate_mode mode)
> +{
> +	unsigned long flags;
> +
> +	/*
> +	 * loan/"inflate" the newpage first.
> +	 *
> +	 * We might race against the cmm_thread who might discover after our
> +	 * loan request that another page is to be unloaned. However, once
> +	 * the cmm_thread runs again later, this error will automatically
> +	 * be corrected.
> +	 */
> +	if (plpar_page_set_loaned(newpage)) {
> +		/* Unlikely, but possible. Tell the caller not to retry now. */
> +		pr_err_ratelimited("%s: Cannot set page to loaned.", __func__);
> +		return -EBUSY;
> +	}
> +
> +	/* balloon page list reference */
> +	get_page(newpage);
> +
> +	spin_lock_irqsave(&b_dev_info->pages_lock, flags);
> +	balloon_page_insert(b_dev_info, newpage);
> +	balloon_page_delete(page);

I think I am missing a b_dev_info->isolated_pages-- here.

> +	spin_unlock_irqrestore(&b_dev_info->pages_lock, flags);


-- 

Thanks,

David / dhildenb