LKML Archive on lore.kernel.org
help / color / mirror / Atom feed
From: Mel Gorman <mel@csn.ul.ie>
To: Andi Kleen <andi@firstfloor.org>
Cc: linux-kernel@vger.kernel.org, pj@sgi.com, linux-mm@kvack.org,
	nickpiggin@yahoo.com.au
Subject: Re: [PATCH] [5/18] Expand the hugetlbfs sysctls to handle arrays for all hstates
Date: Tue, 18 Mar 2008 14:34:38 +0000	[thread overview]
Message-ID: <20080318143438.GE23866@csn.ul.ie> (raw)
In-Reply-To: <20080317015818.E30041B41E0@basil.firstfloor.org>

On (17/03/08 02:58), Andi Kleen didst pronounce:
> - I didn't bother with hugetlb_shm_group and treat_as_movable,
> these are still single global.

I cannot imagine why either of those would be per-pool anyway.
Potentially shm_group could become a per-mount value which is both
outside the scope of this patchset and not per-pool so unsuitable for
hstate. 

> - Also improve error propagation for the sysctl handlers a bit
> 
> 
> Signed-off-by: Andi Kleen <ak@suse.de>
> 
> ---
>  include/linux/hugetlb.h |    5 +++--
>  kernel/sysctl.c         |    2 +-
>  mm/hugetlb.c            |   43 +++++++++++++++++++++++++++++++------------
>  3 files changed, 35 insertions(+), 15 deletions(-)
> 
> Index: linux/include/linux/hugetlb.h
> ===================================================================
> --- linux.orig/include/linux/hugetlb.h
> +++ linux/include/linux/hugetlb.h
> @@ -32,8 +32,6 @@ int hugetlb_fault(struct mm_struct *mm, 
>  int hugetlb_reserve_pages(struct inode *inode, long from, long to);
>  void hugetlb_unreserve_pages(struct inode *inode, long offset, long freed);
>  
> -extern unsigned long max_huge_pages;
> -extern unsigned long sysctl_overcommit_huge_pages;
>  extern unsigned long hugepages_treat_as_movable;
>  extern const unsigned long hugetlb_zero, hugetlb_infinity;
>  extern int sysctl_hugetlb_shm_group;
> @@ -258,6 +256,9 @@ static inline unsigned huge_page_shift(s
>  	return h->order + PAGE_SHIFT;
>  }
>  
> +extern unsigned long max_huge_pages[HUGE_MAX_HSTATE];
> +extern unsigned long sysctl_overcommit_huge_pages[HUGE_MAX_HSTATE];

Any particular reason for moving them?

Also, offhand it's not super-clear why max_huge_pages is not part of
hstate as we only expect one hstate per pagesize anyway.

> +
>  #else
>  struct hstate {};
>  #define hstate_file(f) NULL
> Index: linux/kernel/sysctl.c
> ===================================================================
> --- linux.orig/kernel/sysctl.c
> +++ linux/kernel/sysctl.c
> @@ -935,7 +935,7 @@ static struct ctl_table vm_table[] = {
>  	 {
>  		.procname	= "nr_hugepages",
>  		.data		= &max_huge_pages,
> -		.maxlen		= sizeof(unsigned long),
> +		.maxlen 	= sizeof(max_huge_pages),
>  		.mode		= 0644,
>  		.proc_handler	= &hugetlb_sysctl_handler,
>  		.extra1		= (void *)&hugetlb_zero,
> Index: linux/mm/hugetlb.c
> ===================================================================
> --- linux.orig/mm/hugetlb.c
> +++ linux/mm/hugetlb.c
> @@ -22,8 +22,8 @@
>  #include "internal.h"
>  
>  const unsigned long hugetlb_zero = 0, hugetlb_infinity = ~0UL;
> -unsigned long max_huge_pages;
> -unsigned long sysctl_overcommit_huge_pages;
> +unsigned long max_huge_pages[HUGE_MAX_HSTATE];
> +unsigned long sysctl_overcommit_huge_pages[HUGE_MAX_HSTATE];
>  static gfp_t htlb_alloc_mask = GFP_HIGHUSER;
>  unsigned long hugepages_treat_as_movable;
>  
> @@ -496,11 +496,11 @@ static int __init hugetlb_init_hstate(st
>  
>  	h->hugetlb_next_nid = first_node(node_online_map);
>  
> -	for (i = 0; i < max_huge_pages; ++i) {
> +	for (i = 0; i < max_huge_pages[h - hstates]; ++i) {
>  		if (!alloc_fresh_huge_page(h))
>  			break;
>  	}
> -	max_huge_pages = h->free_huge_pages = h->nr_huge_pages = i;
> +	max_huge_pages[h - hstates] = h->free_huge_pages = h->nr_huge_pages = i;
>  

hmm ok, it looks a little weird to be working out h - hstates multiple times
in a loop when it is invariant but functionally, it's fine.

>  	printk(KERN_INFO "Total HugeTLB memory allocated, %ld %dMB pages\n",
>  			h->free_huge_pages,
> @@ -531,8 +531,9 @@ void __init huge_add_hstate(unsigned ord
>  
>  static int __init hugetlb_setup(char *s)
>  {
> -	if (sscanf(s, "%lu", &max_huge_pages) <= 0)
> -		max_huge_pages = 0;
> +	unsigned long *mhp = &max_huge_pages[parsed_hstate - hstates];

This looks like we are assuming there is only ever one other
parsed_hstate. For the purposes of what you aim to achieve in this set,
it's not important but a comment over parsed_hstate about this
assumption is probably necessary.

> +	if (sscanf(s, "%lu", mhp) <= 0)
> +		*mhp = 0;
>  	return 1;
>  }
>  __setup("hugepages=", hugetlb_setup);
> @@ -584,10 +585,12 @@ static inline void try_to_free_low(unsig
>  #endif
>  
>  #define persistent_huge_pages(h) (h->nr_huge_pages - h->surplus_huge_pages)
> -static unsigned long set_max_huge_pages(unsigned long count)
> +static unsigned long
> +set_max_huge_pages(struct hstate *h, unsigned long count, int *err)
>  {
>  	unsigned long min_count, ret;
> -	struct hstate *h = &global_hstate;
> +
> +	*err = 0;
>  

What is updating err to anything else in set_max_huge_pages()?

>  	/*
>  	 * Increase the pool size
> @@ -659,8 +662,20 @@ int hugetlb_sysctl_handler(struct ctl_ta
>  			   struct file *file, void __user *buffer,
>  			   size_t *length, loff_t *ppos)
>  {
> -	proc_doulongvec_minmax(table, write, file, buffer, length, ppos);
> -	max_huge_pages = set_max_huge_pages(max_huge_pages);
> +	int err = 0;
> +	struct hstate *h;
> +	int i;
> +	err = proc_doulongvec_minmax(table, write, file, buffer, length, ppos);
> +	if (err)
> +		return err;
> +	i = 0;
> +	for_each_hstate (h) {
> +		max_huge_pages[i] = set_max_huge_pages(h, max_huge_pages[i],
> +							&err);

hmm, this is saying when I write 10 to nr_hugepages, I am asking for 10
2MB pages and 10 1GB pages potentially. Is that what you want?

> +		if (err)
> +			return err;

I'm failing to see how the error handling is improved when
set_max_huge_pages() is not updating err. Maybe it happens in another
patch.

> +		i++;
> +	}
>  	return 0;
>  }
>  
> @@ -680,10 +695,14 @@ int hugetlb_overcommit_handler(struct ct
>  			struct file *file, void __user *buffer,
>  			size_t *length, loff_t *ppos)
>  {
> -	struct hstate *h = &global_hstate;
> +	struct hstate *h;
> +	int i = 0;
>  	proc_doulongvec_minmax(table, write, file, buffer, length, ppos);
>  	spin_lock(&hugetlb_lock);
> -	h->nr_overcommit_huge_pages = sysctl_overcommit_huge_pages;
> +	for_each_hstate (h) {
> +		h->nr_overcommit_huge_pages = sysctl_overcommit_huge_pages[i];
> +		i++;
> +	}

Similar to the other sysctl here, the overcommit value is being set for
all the huge page sizes.

>  	spin_unlock(&hugetlb_lock);
>  	return 0;
>  }
> 

-- 
Mel Gorman
Part-time Phd Student                          Linux Technology Center
University of Limerick                         IBM Dublin Software Lab

  reply	other threads:[~2008-03-19 21:58 UTC|newest]

Thread overview: 76+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2008-03-17  1:58 [PATCH] [0/18] GB pages hugetlb support Andi Kleen
2008-03-17  1:58 ` [PATCH] [1/18] Convert hugeltlb.c over to pass global state around in a structure Andi Kleen
2008-03-17 20:15   ` Adam Litke
2008-03-18 12:05   ` Mel Gorman
2008-03-17  1:58 ` [PATCH] [2/18] Add basic support for more than one hstate in hugetlbfs Andi Kleen
2008-03-17 20:22   ` Adam Litke
2008-03-17 20:44     ` Andi Kleen
2008-03-18 12:23   ` Mel Gorman
2008-03-23 10:38   ` KOSAKI Motohiro
2008-03-23 11:28     ` Andi Kleen
2008-03-23 11:30       ` KOSAKI Motohiro
2008-03-17  1:58 ` [PATCH] [3/18] Convert /proc output code over to report multiple hstates Andi Kleen
2008-03-18 12:28   ` Mel Gorman
2008-03-17  1:58 ` [PATCH] [4/18] Add basic support for more than one hstate in hugetlbfs Andi Kleen
2008-03-17  8:09   ` Paul Jackson
2008-03-17  8:15     ` Andi Kleen
2008-03-17 20:28   ` Adam Litke
2008-03-18 14:11   ` Mel Gorman
2008-03-17  1:58 ` [PATCH] [5/18] Expand the hugetlbfs sysctls to handle arrays for all hstates Andi Kleen
2008-03-18 14:34   ` Mel Gorman [this message]
2008-03-18 16:49     ` Andi Kleen
2008-03-18 17:01       ` Mel Gorman
2008-03-17  1:58 ` [PATCH] [6/18] Add support to have individual hstates for each hugetlbfs mount Andi Kleen
2008-03-18 14:10   ` Adam Litke
2008-03-18 15:02   ` Mel Gorman
2008-03-17  1:58 ` [PATCH] [7/18] Abstract out the NUMA node round robin code into a separate function Andi Kleen
2008-03-18 15:42   ` Mel Gorman
2008-03-18 15:47     ` Andi Kleen
2008-03-18 16:04       ` Mel Gorman
2008-03-17  1:58 ` [PATCH] [8/18] Add a __alloc_bootmem_node_nopanic Andi Kleen
2008-03-18 15:54   ` Mel Gorman
2008-03-17  1:58 ` [PATCH] [9/18] Export prep_compound_page to the hugetlb allocator Andi Kleen
2008-03-17  1:58 ` [PATCH] [10/18] Factor out new huge page preparation code into separate function Andi Kleen
2008-03-17 20:31   ` Adam Litke
2008-03-18 16:02   ` Mel Gorman
2008-03-17  1:58 ` [PATCH] [11/18] Fix alignment bug in bootmem allocator Andi Kleen
2008-03-17  2:19   ` Yinghai Lu
2008-03-17  7:02     ` Andi Kleen
2008-03-17  7:17       ` Yinghai Lu
2008-03-17  7:31         ` Yinghai Lu
2008-03-17  7:41           ` Andi Kleen
2008-03-17  7:53             ` Yinghai Lu
2008-03-17  8:10               ` Yinghai Lu
2008-03-17  8:17                 ` Andi Kleen
2008-03-17  8:56               ` Andi Kleen
2008-03-17 18:52                 ` Yinghai Lu
2008-03-17 21:27                   ` Yinghai Lu
2008-03-18  2:06                     ` Yinghai Lu
2008-03-18 16:18   ` Mel Gorman
2008-03-17  1:58 ` [PATCH] [12/18] Add support to allocate hugetlb pages that are larger than MAX_ORDER Andi Kleen
2008-03-18 16:27   ` Mel Gorman
2008-04-09 16:05   ` Andrew Hastings
2008-04-09 17:56     ` Andi Kleen
2008-03-17  1:58 ` [PATCH] [13/18] Add support to allocate hugepages of different size with hugepages= Andi Kleen
2008-03-18 16:32   ` Mel Gorman
2008-03-18 16:45     ` Andi Kleen
2008-03-18 16:46       ` Mel Gorman
2008-03-17  1:58 ` [PATCH] [14/18] Clean up hugetlb boot time printk Andi Kleen
2008-03-18 16:37   ` Mel Gorman
2008-03-17  1:58 ` [PATCH] [15/18] Add support to x86-64 to allocate and lookup GB pages in hugetlb Andi Kleen
2008-03-17  1:58 ` [PATCH] [16/18] Add huge pud support to hugetlbfs Andi Kleen
2008-03-17  1:58 ` [PATCH] [17/18] Add huge pud support to mm/memory.c Andi Kleen
2008-03-17  1:58 ` [PATCH] [18/18] Implement hugepagesz= option for x86-64 Andi Kleen
2008-03-17  9:29   ` Paul Jackson
2008-03-17  9:59     ` Andi Kleen
2008-03-17 10:02       ` Paul Jackson
2008-03-17  3:11 ` [PATCH] [0/18] GB pages hugetlb support Paul Jackson
2008-03-17  7:00   ` Andi Kleen
2008-03-17  7:00     ` Paul Jackson
2008-03-17  7:29       ` Andi Kleen
2008-03-17  5:35 ` Paul Jackson
2008-03-17  6:58   ` Andi Kleen
2008-03-17  9:26 ` Paul Jackson
2008-03-17 15:05 ` Adam Litke
2008-03-17 15:33   ` Andi Kleen
2008-03-17 15:59     ` Adam Litke

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20080318143438.GE23866@csn.ul.ie \
    --to=mel@csn.ul.ie \
    --cc=andi@firstfloor.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-mm@kvack.org \
    --cc=nickpiggin@yahoo.com.au \
    --cc=pj@sgi.com \
    --subject='Re: [PATCH] [5/18] Expand the hugetlbfs sysctls to handle arrays for all hstates' \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).