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, 4 Mar 2019 06:36:56 +0000
From:   Peng Fan <peng.fan@....com>
To:     Dennis Zhou <dennis@...nel.org>
CC:     Tejun Heo <tj@...nel.org>, Christoph Lameter <cl@...ux.com>,
        Vlad Buslov <vladbu@...lanox.com>,
        "kernel-team@...com" <kernel-team@...com>,
        "linux-mm@...ck.org" <linux-mm@...ck.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: RE: [PATCH 11/12] percpu: convert chunk hints to be based on
 pcpu_block_md



> -----Original Message-----
> From: Dennis Zhou [mailto:dennis@...nel.org]
> Sent: 2019年3月4日 4:22
> To: Peng Fan <peng.fan@....com>
> Cc: Dennis Zhou <dennis@...nel.org>; Tejun Heo <tj@...nel.org>; Christoph
> Lameter <cl@...ux.com>; Vlad Buslov <vladbu@...lanox.com>;
> kernel-team@...com; linux-mm@...ck.org; linux-kernel@...r.kernel.org
> Subject: Re: [PATCH 11/12] percpu: convert chunk hints to be based on
> pcpu_block_md
> 
> On Sun, Mar 03, 2019 at 08:18:49AM +0000, Peng Fan wrote:
> >
> >
> > > -----Original Message-----
> > > From: owner-linux-mm@...ck.org [mailto:owner-linux-mm@...ck.org]
> On
> > > Behalf Of Dennis Zhou
> > > Sent: 2019年2月28日 10:19
> > > To: Dennis Zhou <dennis@...nel.org>; Tejun Heo <tj@...nel.org>;
> > > Christoph Lameter <cl@...ux.com>
> > > Cc: Vlad Buslov <vladbu@...lanox.com>; kernel-team@...com;
> > > linux-mm@...ck.org; linux-kernel@...r.kernel.org
> > > Subject: [PATCH 11/12] percpu: convert chunk hints to be based on
> > > pcpu_block_md
> > >
> > > As mentioned in the last patch, a chunk's hints are no different
> > > than a block just responsible for more bits. This converts chunk
> > > level hints to use a pcpu_block_md to maintain them. This lets us
> > > reuse the same hint helper functions as a block. The left_free and
> > > right_free are unused by the chunk's pcpu_block_md.
> > >
> > > Signed-off-by: Dennis Zhou <dennis@...nel.org>
> > > ---
> > >  mm/percpu-internal.h |   5 +-
> > >  mm/percpu-stats.c    |   5 +-
> > >  mm/percpu.c          | 120
> +++++++++++++++++++------------------------
> > >  3 files changed, 57 insertions(+), 73 deletions(-)
> > >
> > > diff --git a/mm/percpu-internal.h b/mm/percpu-internal.h index
> > > 119bd1119aa7..0468ba500bd4 100644
> > > --- a/mm/percpu-internal.h
> > > +++ b/mm/percpu-internal.h
> > > @@ -39,9 +39,7 @@ struct pcpu_chunk {
> > >
> > >  	struct list_head	list;		/* linked to pcpu_slot lists */
> > >  	int			free_bytes;	/* free bytes in the chunk */
> > > -	int			contig_bits;	/* max contiguous size hint */
> > > -	int			contig_bits_start; /* contig_bits starting
> > > -						      offset */
> > > +	struct pcpu_block_md	chunk_md;
> > >  	void			*base_addr;	/* base address of this chunk */
> > >
> > >  	unsigned long		*alloc_map;	/* allocation map */
> > > @@ -49,7 +47,6 @@ struct pcpu_chunk {
> > >  	struct pcpu_block_md	*md_blocks;	/* metadata blocks */
> > >
> > >  	void			*data;		/* chunk data */
> > > -	int			first_bit;	/* no free below this */
> > >  	bool			immutable;	/* no [de]population allowed */
> > >  	int			start_offset;	/* the overlap with the previous
> > >  						   region to have a page aligned diff --git
> > > a/mm/percpu-stats.c b/mm/percpu-stats.c index
> > > b5fdd43b60c9..ef5034a0464e 100644
> > > --- a/mm/percpu-stats.c
> > > +++ b/mm/percpu-stats.c
> > > @@ -53,6 +53,7 @@ static int find_max_nr_alloc(void)  static void
> > > chunk_map_stats(struct seq_file *m, struct pcpu_chunk *chunk,
> > >  			    int *buffer)
> > >  {
> > > +	struct pcpu_block_md *chunk_md = &chunk->chunk_md;
> > >  	int i, last_alloc, as_len, start, end;
> > >  	int *alloc_sizes, *p;
> > >  	/* statistics */
> > > @@ -121,9 +122,9 @@ static void chunk_map_stats(struct seq_file *m,
> > > struct pcpu_chunk *chunk,
> > >  	P("nr_alloc", chunk->nr_alloc);
> > >  	P("max_alloc_size", chunk->max_alloc_size);
> > >  	P("empty_pop_pages", chunk->nr_empty_pop_pages);
> > > -	P("first_bit", chunk->first_bit);
> > > +	P("first_bit", chunk_md->first_free);
> > >  	P("free_bytes", chunk->free_bytes);
> > > -	P("contig_bytes", chunk->contig_bits * PCPU_MIN_ALLOC_SIZE);
> > > +	P("contig_bytes", chunk_md->contig_hint *
> PCPU_MIN_ALLOC_SIZE);
> > >  	P("sum_frag", sum_frag);
> > >  	P("max_frag", max_frag);
> > >  	P("cur_min_alloc", cur_min_alloc); diff --git a/mm/percpu.c
> > > b/mm/percpu.c index 7cdf14c242de..197479f2c489 100644
> > > --- a/mm/percpu.c
> > > +++ b/mm/percpu.c
> > > @@ -233,10 +233,13 @@ static int pcpu_size_to_slot(int size)
> > >
> > >  static int pcpu_chunk_slot(const struct pcpu_chunk *chunk)  {
> > > -	if (chunk->free_bytes < PCPU_MIN_ALLOC_SIZE || chunk->contig_bits
> > > == 0)
> > > +	const struct pcpu_block_md *chunk_md = &chunk->chunk_md;
> > > +
> > > +	if (chunk->free_bytes < PCPU_MIN_ALLOC_SIZE ||
> > > +	    chunk_md->contig_hint == 0)
> > >  		return 0;
> > >
> > > -	return pcpu_size_to_slot(chunk->contig_bits * PCPU_MIN_ALLOC_SIZE);
> > > +	return pcpu_size_to_slot(chunk_md->contig_hint *
> > > PCPU_MIN_ALLOC_SIZE);
> > >  }
> > >
> > >  /* set the pointer to a chunk in a page struct */ @@ -592,54 +595,6
> > > @@ static inline bool pcpu_region_overlap(int a, int b, int x, int y)
> > >  	return false;
> > >  }
> > >
> > > -/**
> > > - * pcpu_chunk_update - updates the chunk metadata given a free area
> > > - * @chunk: chunk of interest
> > > - * @bit_off: chunk offset
> > > - * @bits: size of free area
> > > - *
> > > - * This updates the chunk's contig hint and starting offset given a free
> area.
> > > - * Choose the best starting offset if the contig hint is equal.
> > > - */
> > > -static void pcpu_chunk_update(struct pcpu_chunk *chunk, int
> > > bit_off, int bits) -{
> > > -	if (bits > chunk->contig_bits) {
> > > -		chunk->contig_bits_start = bit_off;
> > > -		chunk->contig_bits = bits;
> > > -	} else if (bits == chunk->contig_bits && chunk->contig_bits_start &&
> > > -		   (!bit_off ||
> > > -		    __ffs(bit_off) > __ffs(chunk->contig_bits_start))) {
> > > -		/* use the start with the best alignment */
> > > -		chunk->contig_bits_start = bit_off;
> > > -	}
> > > -}
> > > -
> > > -/**
> > > - * pcpu_chunk_refresh_hint - updates metadata about a chunk
> > > - * @chunk: chunk of interest
> > > - *
> > > - * Iterates over the metadata blocks to find the largest contig area.
> > > - * It also counts the populated pages and uses the delta to update
> > > the
> > > - * global count.
> > > - *
> > > - * Updates:
> > > - *      chunk->contig_bits
> > > - *      chunk->contig_bits_start
> > > - */
> > > -static void pcpu_chunk_refresh_hint(struct pcpu_chunk *chunk) -{
> > > -	int bit_off, bits;
> > > -
> > > -	/* clear metadata */
> > > -	chunk->contig_bits = 0;
> > > -
> > > -	bit_off = chunk->first_bit;
> > > -	bits = 0;
> > > -	pcpu_for_each_md_free_region(chunk, bit_off, bits) {
> > > -		pcpu_chunk_update(chunk, bit_off, bits);
> > > -	}
> > > -}
> > > -
> > >  /**
> > >   * pcpu_block_update - updates a block given a free area
> > >   * @block: block of interest
> > > @@ -753,6 +708,29 @@ static void pcpu_block_update_scan(struct
> > > pcpu_chunk *chunk, int bit_off,
> > >  	pcpu_block_update(block, s_off, e_off);  }
> > >
> > > +/**
> > > + * pcpu_chunk_refresh_hint - updates metadata about a chunk
> > > + * @chunk: chunk of interest
> > > + *
> > > + * Iterates over the metadata blocks to find the largest contig area.
> > > + * It also counts the populated pages and uses the delta to update
> > > +the
> > > + * global count.
> > > + */
> > > +static void pcpu_chunk_refresh_hint(struct pcpu_chunk *chunk) {
> > > +	struct pcpu_block_md *chunk_md = &chunk->chunk_md;
> > > +	int bit_off, bits;
> > > +
> > > +	/* clear metadata */
> > > +	chunk_md->contig_hint = 0;
> > > +
> > > +	bit_off = chunk_md->first_free;
> > > +	bits = 0;
> > > +	pcpu_for_each_md_free_region(chunk, bit_off, bits) {
> > > +		pcpu_block_update(chunk_md, bit_off, bit_off + bits);
> > > +	}
> > > +}
> > > +
> > >  /**
> > >   * pcpu_block_refresh_hint
> > >   * @chunk: chunk of interest
> > > @@ -800,6 +778,7 @@ static void pcpu_block_refresh_hint(struct
> > > pcpu_chunk *chunk, int index)  static void
> > > pcpu_block_update_hint_alloc(struct pcpu_chunk *chunk, int bit_off,
> > >  					 int bits)
> > >  {
> > > +	struct pcpu_block_md *chunk_md = &chunk->chunk_md;
> > >  	int nr_empty_pages = 0;
> > >  	struct pcpu_block_md *s_block, *e_block, *block;
> > >  	int s_index, e_index;	/* block indexes of the freed allocation */
> > > @@ -910,8 +889,9 @@ static void pcpu_block_update_hint_alloc(struct
> > > pcpu_chunk *chunk, int bit_off,
> > >  	 * contig hint is broken.  Otherwise, it means a smaller space
> > >  	 * was used and therefore the chunk contig hint is still correct.
> > >  	 */
> > > -	if (pcpu_region_overlap(chunk->contig_bits_start,
> > > -				chunk->contig_bits_start + chunk->contig_bits,
> > > +	if (pcpu_region_overlap(chunk_md->contig_hint_start,
> > > +				chunk_md->contig_hint_start +
> > > +				chunk_md->contig_hint,
> > >  				bit_off,
> > >  				bit_off + bits))
> > >  		pcpu_chunk_refresh_hint(chunk);
> > > @@ -930,9 +910,10 @@ static void pcpu_block_update_hint_alloc(struct
> > > pcpu_chunk *chunk, int bit_off,
> > >   *
> > >   * A chunk update is triggered if a page becomes free, a block becomes
> free,
> > >   * or the free spans across blocks.  This tradeoff is to minimize
> > > iterating
> > > - * over the block metadata to update chunk->contig_bits.
> > > chunk->contig_bits
> > > - * may be off by up to a page, but it will never be more than the
> > > available
> > > - * space.  If the contig hint is contained in one block, it will be accurate.
> > > + * over the block metadata to update chunk_md->contig_hint.
> > > + * chunk_md->contig_hint may be off by up to a page, but it will
> > > + never be more
> > > + * than the available space.  If the contig hint is contained in
> > > + one block, it
> > > + * will be accurate.
> > >   */
> > >  static void pcpu_block_update_hint_free(struct pcpu_chunk *chunk,
> > > int bit_off,
> > >  					int bits)
> > > @@ -1026,8 +1007,9 @@ static void pcpu_block_update_hint_free(struct
> > > pcpu_chunk *chunk, int bit_off,
> > >  	if (((end - start) >= PCPU_BITMAP_BLOCK_BITS) || s_index !=
> e_index)
> > >  		pcpu_chunk_refresh_hint(chunk);
> > >  	else
> > > -		pcpu_chunk_update(chunk, pcpu_block_off_to_off(s_index, start),
> > > -				  end - start);
> > > +		pcpu_block_update(&chunk->chunk_md,
> > > +				  pcpu_block_off_to_off(s_index, start),
> > > +				  end);
> > >  }
> > >
> > >  /**
> > > @@ -1082,6 +1064,7 @@ static bool pcpu_is_populated(struct
> > > pcpu_chunk *chunk, int bit_off, int bits,  static int
> > > pcpu_find_block_fit(struct pcpu_chunk *chunk, int alloc_bits,
> > >  			       size_t align, bool pop_only)  {
> > > +	struct pcpu_block_md *chunk_md = &chunk->chunk_md;
> > >  	int bit_off, bits, next_off;
> > >
> > >  	/*
> > > @@ -1090,12 +1073,12 @@ static int pcpu_find_block_fit(struct
> > > pcpu_chunk *chunk, int alloc_bits,
> > >  	 * cannot fit in the global hint, there is memory pressure and
> creating
> > >  	 * a new chunk would happen soon.
> > >  	 */
> > > -	bit_off = ALIGN(chunk->contig_bits_start, align) -
> > > -		  chunk->contig_bits_start;
> > > -	if (bit_off + alloc_bits > chunk->contig_bits)
> > > +	bit_off = ALIGN(chunk_md->contig_hint_start, align) -
> > > +		  chunk_md->contig_hint_start;
> > > +	if (bit_off + alloc_bits > chunk_md->contig_hint)
> > >  		return -1;
> > >
> > > -	bit_off = chunk->first_bit;
> > > +	bit_off = chunk_md->first_free;
> > >  	bits = 0;
> > >  	pcpu_for_each_fit_region(chunk, alloc_bits, align, bit_off, bits) {
> > >  		if (!pop_only || pcpu_is_populated(chunk, bit_off, bits, @@
> > > -1190,6
> > > +1173,7 @@ static unsigned long pcpu_find_zero_area(unsigned long
> > > +*map,
> > > static int pcpu_alloc_area(struct pcpu_chunk *chunk, int alloc_bits,
> > >  			   size_t align, int start)
> > >  {
> > > +	struct pcpu_block_md *chunk_md = &chunk->chunk_md;
> > >  	size_t align_mask = (align) ? (align - 1) : 0;
> > >  	unsigned long area_off = 0, area_bits = 0;
> > >  	int bit_off, end, oslot;
> > > @@ -1222,8 +1206,8 @@ static int pcpu_alloc_area(struct pcpu_chunk
> > > *chunk, int alloc_bits,
> > >  	chunk->free_bytes -= alloc_bits * PCPU_MIN_ALLOC_SIZE;
> > >
> > >  	/* update first free bit */
> > > -	if (bit_off == chunk->first_bit)
> > > -		chunk->first_bit = find_next_zero_bit(
> > > +	if (bit_off == chunk_md->first_free)
> > > +		chunk_md->first_free = find_next_zero_bit(
> > >  					chunk->alloc_map,
> > >  					pcpu_chunk_map_bits(chunk),
> > >  					bit_off + alloc_bits);
> > > @@ -1245,6 +1229,7 @@ static int pcpu_alloc_area(struct pcpu_chunk
> > > *chunk, int alloc_bits,
> > >   */
> > >  static void pcpu_free_area(struct pcpu_chunk *chunk, int off)  {
> > > +	struct pcpu_block_md *chunk_md = &chunk->chunk_md;
> > >  	int bit_off, bits, end, oslot;
> > >
> > >  	lockdep_assert_held(&pcpu_lock);
> > > @@ -1264,7 +1249,7 @@ static void pcpu_free_area(struct pcpu_chunk
> > > *chunk, int off)
> > >  	chunk->free_bytes += bits * PCPU_MIN_ALLOC_SIZE;
> > >
> > >  	/* update first free bit */
> > > -	chunk->first_bit = min(chunk->first_bit, bit_off);
> > > +	chunk_md->first_free = min(chunk_md->first_free, bit_off);
> > >
> > >  	pcpu_block_update_hint_free(chunk, bit_off, bits);
> > >
> > > @@ -1285,6 +1270,9 @@ static void pcpu_init_md_blocks(struct
> > > pcpu_chunk
> > > *chunk)  {
> > >  	struct pcpu_block_md *md_block;
> > >
> > > +	/* init the chunk's block */
> > > +	pcpu_init_md_block(&chunk->chunk_md,
> > > pcpu_chunk_map_bits(chunk));
> > > +
> > >  	for (md_block = chunk->md_blocks;
> > >  	     md_block != chunk->md_blocks +
> pcpu_chunk_nr_blocks(chunk);
> > >  	     md_block++)
> > > @@ -1352,7 +1340,6 @@ static struct pcpu_chunk * __init
> > > pcpu_alloc_first_chunk(unsigned long tmp_addr,
> > >  	chunk->nr_populated = chunk->nr_pages;
> > >  	chunk->nr_empty_pop_pages = chunk->nr_pages;
> > >
> > > -	chunk->contig_bits = map_size / PCPU_MIN_ALLOC_SIZE;
> > >  	chunk->free_bytes = map_size;
> > >
> > >  	if (chunk->start_offset) {
> > > @@ -1362,7 +1349,7 @@ static struct pcpu_chunk * __init
> > > pcpu_alloc_first_chunk(unsigned long tmp_addr,
> > >  		set_bit(0, chunk->bound_map);
> > >  		set_bit(offset_bits, chunk->bound_map);
> > >
> > > -		chunk->first_bit = offset_bits;
> > > +		chunk->chunk_md.first_free = offset_bits;
> > >
> > >  		pcpu_block_update_hint_alloc(chunk, 0, offset_bits);
> > >  	}
> > > @@ -1415,7 +1402,6 @@ static struct pcpu_chunk
> > > *pcpu_alloc_chunk(gfp_t
> > > gfp)
> > >  	pcpu_init_md_blocks(chunk);
> > >
> > >  	/* init metadata */
> > > -	chunk->contig_bits = region_bits;
> > >  	chunk->free_bytes = chunk->nr_pages * PAGE_SIZE;
> > >
> > >  	return chunk;
> >
> > Reviewed-by: Peng Fan <peng.fan@....com>
> >
> > Nitpick, how about name a function __pcpu_md_update, and let
> > pcpu_chunk_update and pcpu_block_update to call __pcpu_md_update. If
> > you prefer, I could submit a patch.
> >
> 
> I don't have a preference. But I do not really want to have 2 functions that just
> wrap the same function. I think overall it'll make sense to rename it, but I
> haven't thought about what to rename it to as it'll probably be influence by
> the possible direction that percpu ends up going.

ok, then I just leave it up to you.

Thanks,
Peng.

> 
> Thanks,
> Dennis

Powered by blists - more mailing lists